* [Xen-devel][PATCH 0/1] cameraif: Add ABI for para-virtualized
@ 2018-07-31 9:31 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-07-31 9:31 UTC (permalink / raw)
To: xen-devel, konrad.wilk, jgross, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
Hello!
At the moment Xen [1] already supports some virtual multimedia
features [2] such as virtual display, sound. It supports keyboards,
pointers and multi-touch devices all allowing Xen to be used in
automotive appliances, In-Vehicle Infotainment (IVI) systems
and many more.
This work adds a new Xen para-virtualized protocol for a virtual
camera device which extends multimedia capabilities of Xen even
farther: video conferencing, IVI, high definition maps etc.
The initial goal is to support most needed functionality with the
final idea to make it possible to extend the protocol if need be:
1. Provide means for base virtual device configuration:
- pixel formats
- resolutions
- frame rates
2. Support basic camera controls:
- contrast
- brightness
- hue
- saturation
3. Support streaming control
4. Support zero-copying use-cases
I hope that Xen and V4L and other communities could give their
valuable feedback on this work, so I can update the protocol
to better fit any additional requirements I might have missed.
Thank you,
Oleksandr Andrushchenko
[1] https://www.xenproject.org/
[2] https://xenbits.xen.org/gitweb/?p=xen.git;a=tree;f=xen/include/public/io
Oleksandr Andrushchenko (1):
cameraif: add ABI for para-virtual camera
xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
1 file changed, 981 insertions(+)
create mode 100644 xen/include/public/io/cameraif.h
--
2.18.0
^ permalink raw reply [flat|nested] 56+ messages in thread
* [PATCH 0/1] cameraif: Add ABI for para-virtualized
@ 2018-07-31 9:31 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-07-31 9:31 UTC (permalink / raw)
To: xen-devel, konrad.wilk, jgross, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
Hello!
At the moment Xen [1] already supports some virtual multimedia
features [2] such as virtual display, sound. It supports keyboards,
pointers and multi-touch devices all allowing Xen to be used in
automotive appliances, In-Vehicle Infotainment (IVI) systems
and many more.
This work adds a new Xen para-virtualized protocol for a virtual
camera device which extends multimedia capabilities of Xen even
farther: video conferencing, IVI, high definition maps etc.
The initial goal is to support most needed functionality with the
final idea to make it possible to extend the protocol if need be:
1. Provide means for base virtual device configuration:
- pixel formats
- resolutions
- frame rates
2. Support basic camera controls:
- contrast
- brightness
- hue
- saturation
3. Support streaming control
4. Support zero-copying use-cases
I hope that Xen and V4L and other communities could give their
valuable feedback on this work, so I can update the protocol
to better fit any additional requirements I might have missed.
Thank you,
Oleksandr Andrushchenko
[1] https://www.xenproject.org/
[2] https://xenbits.xen.org/gitweb/?p=xen.git;a=tree;f=xen/include/public/io
Oleksandr Andrushchenko (1):
cameraif: add ABI for para-virtual camera
xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
1 file changed, 981 insertions(+)
create mode 100644 xen/include/public/io/cameraif.h
--
2.18.0
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-07-31 9:31 ` [PATCH " Oleksandr Andrushchenko
@ 2018-07-31 9:31 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-07-31 9:31 UTC (permalink / raw)
To: xen-devel, konrad.wilk, jgross, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
This is the ABI for the two halves of a para-virtualized
camera driver which extends Xen's reach multimedia capabilities even
farther enabling it for video conferencing, In-Vehicle Infotainment,
high definition maps etc.
The initial goal is to support most needed functionality with the
final idea to make it possible to extend the protocol if need be:
1. Provide means for base virtual device configuration:
- pixel formats
- resolutions
- frame rates
2. Support basic camera controls:
- contrast
- brightness
- hue
- saturation
3. Support streaming control
4. Support zero-copying use-cases
Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
---
xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
1 file changed, 981 insertions(+)
create mode 100644 xen/include/public/io/cameraif.h
diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
new file mode 100644
index 000000000000..bdc6a1262fcf
--- /dev/null
+++ b/xen/include/public/io/cameraif.h
@@ -0,0 +1,981 @@
+/******************************************************************************
+ * cameraif.h
+ *
+ * Unified camera device I/O interface for Xen guest OSes.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Copyright (C) 2018 EPAM Systems Inc.
+ *
+ * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
+ */
+
+#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
+#define __XEN_PUBLIC_IO_CAMERAIF_H__
+
+#include "ring.h"
+#include "../grant_table.h"
+
+/*
+ ******************************************************************************
+ * Protocol version
+ ******************************************************************************
+ */
+#define XENCAMERA_PROTOCOL_VERSION "1"
+
+/*
+ ******************************************************************************
+ * Feature and Parameter Negotiation
+ ******************************************************************************
+ *
+ * Front->back notifications: when enqueuing a new request, sending a
+ * notification can be made conditional on xencamera_req (i.e., the generic
+ * hold-off mechanism provided by the ring macros). Backends must set
+ * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
+ *
+ * Back->front notifications: when enqueuing a new response, sending a
+ * notification can be made conditional on xencamera_resp (i.e., the generic
+ * hold-off mechanism provided by the ring macros). Frontends must set
+ * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
+ *
+ * The two halves of a para-virtual camera driver utilize nodes within
+ * XenStore to communicate capabilities and to negotiate operating parameters.
+ * This section enumerates these nodes which reside in the respective front and
+ * backend portions of XenStore, following the XenBus convention.
+ *
+ * All data in XenStore is stored as strings. Nodes specifying numeric
+ * values are encoded in decimal. Integer value ranges listed below are
+ * expressed as fixed sized integer types capable of storing the conversion
+ * of a properly formatted node string, without loss of information.
+ *
+ ******************************************************************************
+ * Example configuration
+ ******************************************************************************
+ *
+ * This is an example of backend and frontend configuration:
+ *
+ *--------------------------------- Backend -----------------------------------
+ *
+ * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
+ * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
+ * /local/domain/0/backend/vcamera/1/0/state = "4"
+ * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
+ *
+ *--------------------------------- Frontend ----------------------------------
+ *
+ * /local/domain/1/device/vcamera/0/backend-id = "0"
+ * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
+ * /local/domain/1/device/vcamera/0/state = "4"
+ * /local/domain/1/device/vcamera/0/version = "1"
+ * /local/domain/1/device/vcamera/0/be-alloc = "1"
+ *
+ *---------------------------- Device 0 configuration -------------------------
+ *
+ * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
+ * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
+ * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
+ * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
+ * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
+ * /local/domain/1/device/vcamera/0/unique-id = "0"
+ * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
+ * /local/domain/1/device/vcamera/0/req-event-channel = "15"
+ * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
+ * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
+ *
+ *---------------------------- Device 1 configuration -------------------------
+ *
+ * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
+ * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
+ * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
+ * /local/domain/1/device/vcamera/1/unique-id = "1"
+ * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
+ * /local/domain/1/device/vcamera/1/req-event-channel = "17"
+ * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
+ * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
+ *
+ ******************************************************************************
+ * Backend XenBus Nodes
+ ******************************************************************************
+ *
+ *----------------------------- Protocol version ------------------------------
+ *
+ * versions
+ * Values: <string>
+ *
+ * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
+ * by the backend. For example "1,2,3".
+ *
+ ******************************************************************************
+ * Frontend XenBus Nodes
+ ******************************************************************************
+ *
+ *-------------------------------- Addressing ---------------------------------
+ *
+ * dom-id
+ * Values: <uint16_t>
+ *
+ * Domain identifier.
+ *
+ * dev-id
+ * Values: <uint16_t>
+ *
+ * Device identifier.
+ *
+ * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
+ *
+ *----------------------------- Protocol version ------------------------------
+ *
+ * version
+ * Values: <string>
+ *
+ * Protocol version, chosen among the ones supported by the backend.
+ *
+ *------------------------- Backend buffer allocation -------------------------
+ *
+ * be-alloc
+ * Values: "0", "1"
+ *
+ * If value is set to "1", then backend will be the buffer
+ * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
+ * operation.
+ * If value is not "1" or omitted frontend must allocate buffers itself.
+ *
+ *------------------------------- Camera settings -----------------------------
+ *
+ * unique-id
+ * Values: <string>
+ *
+ * After device instance initialization each camera is assigned a
+ * unique ID, so it can be identified by the backend by this ID.
+ * This can be UUID or such.
+ *
+ * controls
+ * Values: <list of string>
+ *
+ * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
+ * Camera controls are expressed as a list of string values w/o any
+ * ordering requirement.
+ *
+ * formats
+ * Values: <format, char[4]>
+ *
+ * Formats are organized as a set of directories one per each
+ * supported pixel format. The name of the directory is an upper case
+ * string of the corresponding FOURCC string label. The next level of
+ * the directory under <formats> represents supported resolutions.
+ *
+ * resolution
+ * Values: <width, uint32_t>x<height, uint32_t>
+ *
+ * Resolutions are organized as a set of directories one per each
+ * supported resolution under corresponding <formats> directory.
+ * The name of the directory is the supported width and height
+ * of the camera resolution in pixels.
+ *
+ * frame-rates
+ * Values: <numerator, uint32_t>/<denominator, uint32_t>
+ *
+ * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
+ * of the camera expressed as numerator and denominator of the
+ * corresponding frame rate.
+ *
+ * The format of the <formats> directory tree with resolutions and frame rates
+ * must be structured in the following format:
+ *
+ * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
+ *
+ * where
+ * i - i-th supported pixel format
+ * j - j-th supported resolution for i-th pixel format
+ * k - k-th supported frame rate for i-th pixel format and j-th
+ * resolution
+ *
+ *------------------- Camera Request Transport Parameters ---------------------
+ *
+ * This communication path is used to deliver requests from frontend to backend
+ * and get the corresponding responses from backend to frontend,
+ * set up per virtual camera device.
+ *
+ * req-event-channel
+ * Values: <uint32_t>
+ *
+ * The identifier of the Xen camera's control event channel
+ * used to signal activity in the ring buffer.
+ *
+ * req-ring-ref
+ * Values: <uint32_t>
+ *
+ * The Xen grant reference granting permission for the backend to map
+ * a sole page of camera's control ring buffer.
+ *
+ *-------------------- Camera Event Transport Parameters ----------------------
+ *
+ * This communication path is used to deliver asynchronous events from backend
+ * to frontend, set up per virtual camera device.
+ *
+ * evt-event-channel
+ * Values: <uint32_t>
+ *
+ * The identifier of the Xen camera's event channel
+ * used to signal activity in the ring buffer.
+ *
+ * evt-ring-ref
+ * Values: <uint32_t>
+ *
+ * The Xen grant reference granting permission for the backend to map
+ * a sole page of camera's event ring buffer.
+ */
+
+/*
+ ******************************************************************************
+ * STATE DIAGRAMS
+ ******************************************************************************
+ *
+ * Tool stack creates front and back state nodes with initial state
+ * XenbusStateInitialising.
+ * Tool stack creates and sets up frontend camera configuration
+ * nodes per domain.
+ *
+ *-------------------------------- Normal flow --------------------------------
+ *
+ * Front Back
+ * ================================= =====================================
+ * XenbusStateInitialising XenbusStateInitialising
+ * o Query backend device identification
+ * data.
+ * o Open and validate backend device.
+ * |
+ * |
+ * V
+ * XenbusStateInitWait
+ *
+ * o Query frontend configuration
+ * o Allocate and initialize
+ * event channels per configured
+ * camera.
+ * o Publish transport parameters
+ * that will be in effect during
+ * this connection.
+ * |
+ * |
+ * V
+ * XenbusStateInitialised
+ *
+ * o Query frontend transport parameters.
+ * o Connect to the event channels.
+ * |
+ * |
+ * V
+ * XenbusStateConnected
+ *
+ * o Create and initialize OS
+ * virtual camera as per
+ * configuration.
+ * |
+ * |
+ * V
+ * XenbusStateConnected
+ *
+ * XenbusStateUnknown
+ * XenbusStateClosed
+ * XenbusStateClosing
+ * o Remove virtual camera device
+ * o Remove event channels
+ * |
+ * |
+ * V
+ * XenbusStateClosed
+ *
+ *------------------------------- Recovery flow -------------------------------
+ *
+ * In case of frontend unrecoverable errors backend handles that as
+ * if frontend goes into the XenbusStateClosed state.
+ *
+ * In case of backend unrecoverable errors frontend tries removing
+ * the virtualized device. If this is possible at the moment of error,
+ * then frontend goes into the XenbusStateInitialising state and is ready for
+ * new connection with backend. If the virtualized device is still in use and
+ * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
+ * until either the virtualized device is removed or backend initiates a new
+ * connection. On the virtualized device removal frontend goes into the
+ * XenbusStateInitialising state.
+ *
+ * Note on XenbusStateReconfiguring state of the frontend: if backend has
+ * unrecoverable errors then frontend cannot send requests to the backend
+ * and thus cannot provide functionality of the virtualized device anymore.
+ * After backend is back to normal the virtualized device may still hold some
+ * state: configuration in use, allocated buffers, client application state etc.
+ * In most cases, this will require frontend to implement complex recovery
+ * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
+ * frontend will make sure no new clients of the virtualized device are
+ * accepted, allow existing client(s) to exit gracefully by signaling error
+ * state etc.
+ * Once all the clients are gone frontend can reinitialize the virtualized
+ * device and get into XenbusStateInitialising state again signaling the
+ * backend that a new connection can be made.
+ *
+ * There are multiple conditions possible under which frontend will go from
+ * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
+ * specific. For example:
+ * 1. The underlying OS framework may provide callbacks to signal that the last
+ * client of the virtualized device has gone and the device can be removed
+ * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
+ * to periodically check if this is the right time to re-try removal of
+ * the virtualized device.
+ * 3. By any other means.
+ *
+ ******************************************************************************
+ * REQUEST CODES
+ ******************************************************************************
+ */
+#define XENCAMERA_OP_SET_CONFIG 0x00
+#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
+#define XENCAMERA_OP_BUF_CREATE 0x02
+#define XENCAMERA_OP_BUF_DESTROY 0x03
+#define XENCAMERA_OP_STREAM_START 0x04
+#define XENCAMERA_OP_STREAM_STOP 0x05
+#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
+#define XENCAMERA_OP_SET_CTRL 0x07
+
+#define XENCAMERA_CTRL_BRIGHTNESS 0x00
+#define XENCAMERA_CTRL_CONTRAST 0x01
+#define XENCAMERA_CTRL_SATURATION 0x02
+#define XENCAMERA_CTRL_HUE 0x03
+
+/*
+ ******************************************************************************
+ * EVENT CODES
+ ******************************************************************************
+ */
+#define XENCAMERA_EVT_FRAME_AVAIL 0x00
+
+/*
+ ******************************************************************************
+ * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
+ ******************************************************************************
+ */
+#define XENCAMERA_DRIVER_NAME "vcamera"
+
+#define XENCAMERA_LIST_SEPARATOR ","
+#define XENCAMERA_RESOLUTION_SEPARATOR "x"
+#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
+
+#define XENCAMERA_FIELD_BE_VERSIONS "versions"
+#define XENCAMERA_FIELD_FE_VERSION "version"
+#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
+#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
+#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
+#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
+#define XENCAMERA_FIELD_CONTROLS "controls"
+#define XENCAMERA_FIELD_FORMATS "formats"
+#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
+#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
+
+#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
+#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
+#define XENCAMERA_CTRL_SATURATION_STR "saturation"
+#define XENCAMERA_CTRL_HUE_STR "hue"
+
+/* Maximum number of buffer planes supported. */
+#define XENCAMERA_MAX_PLANE 4
+
+/*
+ ******************************************************************************
+ * STATUS RETURN CODES
+ ******************************************************************************
+ *
+ * Status return code is zero on success and -XEN_EXX on failure.
+ *
+ ******************************************************************************
+ * Assumptions
+ ******************************************************************************
+ *
+ * - usage of grant reference 0 as invalid grant reference:
+ * grant reference 0 is valid, but never exposed to a PV driver,
+ * because of the fact it is already in use/reserved by the PV console.
+ * - all references in this document to page sizes must be treated
+ * as pages of size XEN_PAGE_SIZE unless otherwise noted.
+ *
+ ******************************************************************************
+ * Description of the protocol between frontend and backend driver
+ ******************************************************************************
+ *
+ * The two halves of a Para-virtual camera driver communicate with
+ * each other using shared pages and event channels.
+ * Shared page contains a ring with request/response packets.
+ *
+ * All reserved fields in the structures below must be 0.
+ *
+ * For all request/response/event packets:
+ * - frame rate parameter is represented as a pair of 4 octet long
+ * numerator and denominator:
+ * - frame_rate_numer - uint32_t, numerator of the frame rate
+ * - frame_rate_denom - uint32_t, denominator of the frame rate
+ * The corresponding frame rate (Hz) is calculated as:
+ * frame_rate = frame_rate_numer / frame_rate_denom
+ * - buffer index is a zero based index of the buffer. Must be less than
+ * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
+ * - index - uint8_t, index of the buffer.
+ *
+ *
+ *---------------------------------- Requests ---------------------------------
+ *
+ * All request packets have the same length (64 octets).
+ * All request packets have common header:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | operation | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * id - uint16_t, private guest value, echoed in response.
+ * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
+ *
+ *
+ * Request configuration set/reset - request to set or reset.
+ * the configuration/mode of the camera:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_SET_CONFIG | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | pixel format | 12
+ * +----------------+----------------+----------------+----------------+
+ * | width | 16
+ * +----------------+----------------+----------------+----------------+
+ * | height | 20
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_numer | 24
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_denom | 28
+ * +----------------+----------------+----------------+----------------+
+ * | num_bufs | reserved | 32
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 36
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * Pass all zeros to reset, otherwise command is treated as configuration set.
+ *
+ * pixel_format - uint32_t, pixel format to be used, FOURCC code.
+ * width - uint32_t, width in pixels.
+ * height - uint32_t, height in pixels.
+ * frame_rate_numer - uint32_t, numerator of the frame rate.
+ * frame_rate_denom - uint32_t, denominator of the frame rate.
+ * num_bufs - uint8_t, desired number of buffers to be used.
+ *
+ * See response format for this request.
+ *
+ * Notes:
+ * - frontend must check the corresponding response in order to see
+ * if the values reported back by the backend do match the desired ones
+ * and can be accepted.
+ * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
+ * sending XENCAMERA_OP_STREAM_START request to update or tune the
+ * configuration.
+ */
+struct xencamera_config {
+ uint32_t pixel_format;
+ uint32_t width;
+ uint32_t height;
+ uint32_t frame_rate_nom;
+ uint32_t frame_rate_denom;
+ uint8_t num_bufs;
+};
+
+/*
+ * Request buffer details - request camera buffer's memory layout.
+ * detailed description:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_GET_BUF_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * See response format for this request.
+ *
+ *
+ * Request camera buffer creation:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_BUF_CREATE | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | gref_directory | 16
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 20
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * An attempt to create multiple buffers with the same index is an error.
+ * index can be re-used after destroying the corresponding camera buffer.
+ *
+ * index - uint8_t, index of the buffer to be created.
+ * gref_directory - grant_ref_t, a reference to the first shared page
+ * describing shared buffer references. The size of the buffer is equal to
+ * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
+ * shared buffer size exceeds what can be addressed by this single page,
+ * then reference to the next shared page must be supplied (see
+ * gref_dir_next_page below).
+ *
+ * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
+ * allocate the buffer with the parameters provided in this request and page
+ * directory is handled as follows:
+ * Frontend on request:
+ * - allocates pages for the directory (gref_directory,
+ * gref_dir_next_page(s)
+ * - grants permissions for the pages of the directory to the backend
+ * - sets gref_dir_next_page fields
+ * Backend on response:
+ * - grants permissions for the pages of the buffer allocated to
+ * the frontend
+ * - fills in page directory with grant references
+ * (gref[] in struct xencamera_page_directory)
+ */
+struct xencamera_buf_create_req {
+ uint8_t index;
+ uint8_t reserved[3];
+ grant_ref_t gref_directory;
+};
+
+/*
+ * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
+ * the request) employs a list of pages, describing all pages of the shared
+ * data buffer:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | gref_dir_next_page | 4
+ * +----------------+----------------+----------------+----------------+
+ * | gref[0] | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | gref[i] | i*4+8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | gref[N - 1] | N*4+8
+ * +----------------+----------------+----------------+----------------+
+ *
+ * gref_dir_next_page - grant_ref_t, reference to the next page describing
+ * page directory. Must be 0 if there are no more pages in the list.
+ * gref[i] - grant_ref_t, reference to a shared page of the buffer
+ * allocated at XENCAMERA_OP_BUF_CREATE.
+ *
+ * Number of grant_ref_t entries in the whole page directory is not
+ * passed, but instead can be calculated as:
+ * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
+ * XEN_PAGE_SIZE
+ */
+struct xencamera_page_directory {
+ grant_ref_t gref_dir_next_page;
+ grant_ref_t gref[1]; /* Variable length */
+};
+
+/*
+ * Request buffer destruction - destroy a previously allocated camera buffer:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_BUF_DESTROY| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 16
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * index - uint8_t, index of the buffer to be destroyed.
+ */
+
+struct xencamera_buf_destroy_req {
+ uint8_t index;
+};
+
+/*
+ * Request camera capture stream start:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_OP_STREAM_START| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ *
+ * Request camera capture stream stop:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_OP_STREAM_STOP | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ *
+ * Request camera control details:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |GET_CTRL_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 16
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * See response format for this request.
+ *
+ * index - uint8_t, index of the control to be queried.
+ */
+struct xencamera_get_ctrl_details_req {
+ uint8_t index;
+};
+
+/*
+ *
+ * Request camera control change:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_SET_CTRL | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | value | 16
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 20
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * See response format for this request.
+ *
+ * index - uint8_t, index of the control.
+ * value - int32_t, new value of the control.
+ */
+struct xencamera_set_ctrl_req {
+ uint8_t index;
+ uint8_t reserved[3];
+ int32_t value;
+};
+
+/*
+ *---------------------------------- Responses --------------------------------
+ *
+ * All response packets have the same length (64 octets).
+ *
+ * All response packets have common header:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | operation | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ *
+ * id - uint16_t, copied from the request.
+ * operation - uint8_t, XENCAMERA_OP_* - copied from request.
+ * status - int32_t, response status, zero on success and -XEN_EXX on failure.
+ *
+ *
+ * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_SET_CONFIG | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ * | pixel format | 12
+ * +----------------+----------------+----------------+----------------+
+ * | width | 16
+ * +----------------+----------------+----------------+----------------+
+ * | height | 20
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_numer | 24
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_denom | 28
+ * +----------------+----------------+----------------+----------------+
+ * | num_bufs | reserved | 32
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 36
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * Meaning of the corresponding values in this response is the same as for
+ * XENCAMERA_OP_SET_CONFIG request.
+ *
+ *
+ * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
+ * request:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_GET_BUF_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ * | size | 12
+ * +----------------+----------------+----------------+----------------+
+ * | num_planes | reserved | 16
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[0] | 20
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[1] | 24
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[2] | 28
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[3] | 32
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[0] | 36
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[1] | 40
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[2] | 44
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[3] | 48
+ * +----------------+----------------+----------------+----------------+
+ * | plane_stride[0] | plane_stride[1] | 52
+ * +----------------+----------------+----------------+----------------+
+ * | plane_stride[2] | plane_stride[3] | 56
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 60
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * size - uint32_t, overall size of the buffer including sizes of the
+ * individual planes and padding if applicable.
+ * num_planes - uint8_t, number of planes for this buffer.
+ * plane_offset - array of uint32_t, offset of the corresponding plane
+ * in octets from the buffer start.
+ * plane_size - array of uint32_t, size in octets of the corresponding plane
+ * including padding.
+ * plane_stride - array of uint32_t, size in octets occupied by the
+ * corresponding single image line including padding if applicable.
+ */
+struct xencamera_buf_details_resp {
+ uint32_t size;
+ uint8_t num_planes;
+ uint8_t reserved[3];
+ uint32_t plane_offset[XENCAMERA_MAX_PLANE];
+ uint32_t plane_size[XENCAMERA_MAX_PLANE];
+ uint16_t plane_stride[XENCAMERA_MAX_PLANE];
+};
+
+/*
+ * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |GET_CTRL_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | type | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | min | 16
+ * +----------------+----------------+----------------+----------------+
+ * | max | 20
+ * +----------------+----------------+----------------+----------------+
+ * | step | 24
+ * +----------------+----------------+----------------+----------------+
+ * | def_val | 28
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 36
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * index - uint8_t, index of the camera control in response.
+ * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
+ * min - int32_t, minimum value of the control.
+ * max - int32_t, maximum value of the control.
+ * step - int32_t, minimum size in which control value can be changed.
+ * def_val - int32_t, default value of the control.
+ */
+struct xencamera_get_ctrl_details_resp {
+ uint8_t index;
+ uint8_t type;
+ uint8_t reserved[2];
+ int32_t min;
+ int32_t max;
+ int32_t step;
+ int32_t def_val;
+};
+
+/*
+ *----------------------------------- Events ----------------------------------
+ *
+ * Events are sent via a shared page allocated by the front and propagated by
+ * evt-event-channel/evt-ring-ref XenStore entries.
+ *
+ * All event packets have the same length (64 octets).
+ * All event packets have common header:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | type | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ *
+ * id - uint16_t, event id, may be used by front.
+ * type - uint8_t, type of the event.
+ *
+ *
+ * Frame captured event - event from back to front when a new captured
+ * frame is available:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_EVT_FRAME_AVAIL| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | used_sz | 16
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 20
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * index - uint8_t, index of the buffer that contains new captured frame.
+ * used_sz - uint32_t, number of octets this frame has. This can be less
+ * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
+ */
+struct xencamera_frame_avail_evt {
+ uint8_t index;
+ uint8_t reserved[3];
+ uint32_t used_sz;
+};
+
+struct xencamera_req {
+ uint16_t id;
+ uint8_t operation;
+ uint8_t reserved[5];
+ union {
+ struct xencamera_config config;
+ struct xencamera_buf_create_req buf_create;
+ struct xencamera_buf_destroy_req buf_destroy;
+ struct xencamera_set_ctrl_req set_ctrl;
+ uint8_t reserved[56];
+ } req;
+};
+
+struct xencamera_resp {
+ uint16_t id;
+ uint8_t operation;
+ uint8_t reserved;
+ int32_t status;
+ union {
+ struct xencamera_config config;
+ struct xencamera_buf_details_resp buf_details;
+ struct xencamera_get_ctrl_details_resp ctrl_details;
+ uint8_t reserved1[56];
+ } resp;
+};
+
+struct xencamera_evt {
+ uint16_t id;
+ uint8_t type;
+ uint8_t reserved[5];
+ union {
+ struct xencamera_frame_avail_evt frame_avail;
+ uint8_t reserved[56];
+ } evt;
+};
+
+DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
+
+/*
+ ******************************************************************************
+ * Back to front events delivery
+ ******************************************************************************
+ * In order to deliver asynchronous events from back to front a shared page is
+ * allocated by front and its granted reference propagated to back via
+ * XenStore entries (evt-ring-ref/evt-event-channel).
+ * This page has a common header used by both front and back to synchronize
+ * access and control event's ring buffer, while back being a producer of the
+ * events and front being a consumer. The rest of the page after the header
+ * is used for event packets.
+ *
+ * Upon reception of an event(s) front may confirm its reception
+ * for either each event, group of events or none.
+ */
+
+struct xencamera_event_page {
+ uint32_t in_cons;
+ uint32_t in_prod;
+ uint8_t reserved[56];
+};
+
+#define XENCAMERA_EVENT_PAGE_SIZE 4096
+#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
+#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
+#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
+#define XENCAMERA_IN_RING(page) \
+ ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
+#define XENCAMERA_IN_RING_REF(page, idx) \
+ (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
+
+#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
2.18.0
^ permalink raw reply related [flat|nested] 56+ messages in thread
* [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-07-31 9:31 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-07-31 9:31 UTC (permalink / raw)
To: xen-devel, konrad.wilk, jgross, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
This is the ABI for the two halves of a para-virtualized
camera driver which extends Xen's reach multimedia capabilities even
farther enabling it for video conferencing, In-Vehicle Infotainment,
high definition maps etc.
The initial goal is to support most needed functionality with the
final idea to make it possible to extend the protocol if need be:
1. Provide means for base virtual device configuration:
- pixel formats
- resolutions
- frame rates
2. Support basic camera controls:
- contrast
- brightness
- hue
- saturation
3. Support streaming control
4. Support zero-copying use-cases
Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
---
xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
1 file changed, 981 insertions(+)
create mode 100644 xen/include/public/io/cameraif.h
diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
new file mode 100644
index 000000000000..bdc6a1262fcf
--- /dev/null
+++ b/xen/include/public/io/cameraif.h
@@ -0,0 +1,981 @@
+/******************************************************************************
+ * cameraif.h
+ *
+ * Unified camera device I/O interface for Xen guest OSes.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Copyright (C) 2018 EPAM Systems Inc.
+ *
+ * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
+ */
+
+#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
+#define __XEN_PUBLIC_IO_CAMERAIF_H__
+
+#include "ring.h"
+#include "../grant_table.h"
+
+/*
+ ******************************************************************************
+ * Protocol version
+ ******************************************************************************
+ */
+#define XENCAMERA_PROTOCOL_VERSION "1"
+
+/*
+ ******************************************************************************
+ * Feature and Parameter Negotiation
+ ******************************************************************************
+ *
+ * Front->back notifications: when enqueuing a new request, sending a
+ * notification can be made conditional on xencamera_req (i.e., the generic
+ * hold-off mechanism provided by the ring macros). Backends must set
+ * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
+ *
+ * Back->front notifications: when enqueuing a new response, sending a
+ * notification can be made conditional on xencamera_resp (i.e., the generic
+ * hold-off mechanism provided by the ring macros). Frontends must set
+ * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
+ *
+ * The two halves of a para-virtual camera driver utilize nodes within
+ * XenStore to communicate capabilities and to negotiate operating parameters.
+ * This section enumerates these nodes which reside in the respective front and
+ * backend portions of XenStore, following the XenBus convention.
+ *
+ * All data in XenStore is stored as strings. Nodes specifying numeric
+ * values are encoded in decimal. Integer value ranges listed below are
+ * expressed as fixed sized integer types capable of storing the conversion
+ * of a properly formatted node string, without loss of information.
+ *
+ ******************************************************************************
+ * Example configuration
+ ******************************************************************************
+ *
+ * This is an example of backend and frontend configuration:
+ *
+ *--------------------------------- Backend -----------------------------------
+ *
+ * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
+ * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
+ * /local/domain/0/backend/vcamera/1/0/state = "4"
+ * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
+ *
+ *--------------------------------- Frontend ----------------------------------
+ *
+ * /local/domain/1/device/vcamera/0/backend-id = "0"
+ * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
+ * /local/domain/1/device/vcamera/0/state = "4"
+ * /local/domain/1/device/vcamera/0/version = "1"
+ * /local/domain/1/device/vcamera/0/be-alloc = "1"
+ *
+ *---------------------------- Device 0 configuration -------------------------
+ *
+ * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
+ * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
+ * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
+ * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
+ * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
+ * /local/domain/1/device/vcamera/0/unique-id = "0"
+ * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
+ * /local/domain/1/device/vcamera/0/req-event-channel = "15"
+ * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
+ * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
+ *
+ *---------------------------- Device 1 configuration -------------------------
+ *
+ * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
+ * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
+ * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
+ * /local/domain/1/device/vcamera/1/unique-id = "1"
+ * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
+ * /local/domain/1/device/vcamera/1/req-event-channel = "17"
+ * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
+ * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
+ *
+ ******************************************************************************
+ * Backend XenBus Nodes
+ ******************************************************************************
+ *
+ *----------------------------- Protocol version ------------------------------
+ *
+ * versions
+ * Values: <string>
+ *
+ * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
+ * by the backend. For example "1,2,3".
+ *
+ ******************************************************************************
+ * Frontend XenBus Nodes
+ ******************************************************************************
+ *
+ *-------------------------------- Addressing ---------------------------------
+ *
+ * dom-id
+ * Values: <uint16_t>
+ *
+ * Domain identifier.
+ *
+ * dev-id
+ * Values: <uint16_t>
+ *
+ * Device identifier.
+ *
+ * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
+ *
+ *----------------------------- Protocol version ------------------------------
+ *
+ * version
+ * Values: <string>
+ *
+ * Protocol version, chosen among the ones supported by the backend.
+ *
+ *------------------------- Backend buffer allocation -------------------------
+ *
+ * be-alloc
+ * Values: "0", "1"
+ *
+ * If value is set to "1", then backend will be the buffer
+ * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
+ * operation.
+ * If value is not "1" or omitted frontend must allocate buffers itself.
+ *
+ *------------------------------- Camera settings -----------------------------
+ *
+ * unique-id
+ * Values: <string>
+ *
+ * After device instance initialization each camera is assigned a
+ * unique ID, so it can be identified by the backend by this ID.
+ * This can be UUID or such.
+ *
+ * controls
+ * Values: <list of string>
+ *
+ * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
+ * Camera controls are expressed as a list of string values w/o any
+ * ordering requirement.
+ *
+ * formats
+ * Values: <format, char[4]>
+ *
+ * Formats are organized as a set of directories one per each
+ * supported pixel format. The name of the directory is an upper case
+ * string of the corresponding FOURCC string label. The next level of
+ * the directory under <formats> represents supported resolutions.
+ *
+ * resolution
+ * Values: <width, uint32_t>x<height, uint32_t>
+ *
+ * Resolutions are organized as a set of directories one per each
+ * supported resolution under corresponding <formats> directory.
+ * The name of the directory is the supported width and height
+ * of the camera resolution in pixels.
+ *
+ * frame-rates
+ * Values: <numerator, uint32_t>/<denominator, uint32_t>
+ *
+ * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
+ * of the camera expressed as numerator and denominator of the
+ * corresponding frame rate.
+ *
+ * The format of the <formats> directory tree with resolutions and frame rates
+ * must be structured in the following format:
+ *
+ * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
+ *
+ * where
+ * i - i-th supported pixel format
+ * j - j-th supported resolution for i-th pixel format
+ * k - k-th supported frame rate for i-th pixel format and j-th
+ * resolution
+ *
+ *------------------- Camera Request Transport Parameters ---------------------
+ *
+ * This communication path is used to deliver requests from frontend to backend
+ * and get the corresponding responses from backend to frontend,
+ * set up per virtual camera device.
+ *
+ * req-event-channel
+ * Values: <uint32_t>
+ *
+ * The identifier of the Xen camera's control event channel
+ * used to signal activity in the ring buffer.
+ *
+ * req-ring-ref
+ * Values: <uint32_t>
+ *
+ * The Xen grant reference granting permission for the backend to map
+ * a sole page of camera's control ring buffer.
+ *
+ *-------------------- Camera Event Transport Parameters ----------------------
+ *
+ * This communication path is used to deliver asynchronous events from backend
+ * to frontend, set up per virtual camera device.
+ *
+ * evt-event-channel
+ * Values: <uint32_t>
+ *
+ * The identifier of the Xen camera's event channel
+ * used to signal activity in the ring buffer.
+ *
+ * evt-ring-ref
+ * Values: <uint32_t>
+ *
+ * The Xen grant reference granting permission for the backend to map
+ * a sole page of camera's event ring buffer.
+ */
+
+/*
+ ******************************************************************************
+ * STATE DIAGRAMS
+ ******************************************************************************
+ *
+ * Tool stack creates front and back state nodes with initial state
+ * XenbusStateInitialising.
+ * Tool stack creates and sets up frontend camera configuration
+ * nodes per domain.
+ *
+ *-------------------------------- Normal flow --------------------------------
+ *
+ * Front Back
+ * ================================= =====================================
+ * XenbusStateInitialising XenbusStateInitialising
+ * o Query backend device identification
+ * data.
+ * o Open and validate backend device.
+ * |
+ * |
+ * V
+ * XenbusStateInitWait
+ *
+ * o Query frontend configuration
+ * o Allocate and initialize
+ * event channels per configured
+ * camera.
+ * o Publish transport parameters
+ * that will be in effect during
+ * this connection.
+ * |
+ * |
+ * V
+ * XenbusStateInitialised
+ *
+ * o Query frontend transport parameters.
+ * o Connect to the event channels.
+ * |
+ * |
+ * V
+ * XenbusStateConnected
+ *
+ * o Create and initialize OS
+ * virtual camera as per
+ * configuration.
+ * |
+ * |
+ * V
+ * XenbusStateConnected
+ *
+ * XenbusStateUnknown
+ * XenbusStateClosed
+ * XenbusStateClosing
+ * o Remove virtual camera device
+ * o Remove event channels
+ * |
+ * |
+ * V
+ * XenbusStateClosed
+ *
+ *------------------------------- Recovery flow -------------------------------
+ *
+ * In case of frontend unrecoverable errors backend handles that as
+ * if frontend goes into the XenbusStateClosed state.
+ *
+ * In case of backend unrecoverable errors frontend tries removing
+ * the virtualized device. If this is possible at the moment of error,
+ * then frontend goes into the XenbusStateInitialising state and is ready for
+ * new connection with backend. If the virtualized device is still in use and
+ * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
+ * until either the virtualized device is removed or backend initiates a new
+ * connection. On the virtualized device removal frontend goes into the
+ * XenbusStateInitialising state.
+ *
+ * Note on XenbusStateReconfiguring state of the frontend: if backend has
+ * unrecoverable errors then frontend cannot send requests to the backend
+ * and thus cannot provide functionality of the virtualized device anymore.
+ * After backend is back to normal the virtualized device may still hold some
+ * state: configuration in use, allocated buffers, client application state etc.
+ * In most cases, this will require frontend to implement complex recovery
+ * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
+ * frontend will make sure no new clients of the virtualized device are
+ * accepted, allow existing client(s) to exit gracefully by signaling error
+ * state etc.
+ * Once all the clients are gone frontend can reinitialize the virtualized
+ * device and get into XenbusStateInitialising state again signaling the
+ * backend that a new connection can be made.
+ *
+ * There are multiple conditions possible under which frontend will go from
+ * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
+ * specific. For example:
+ * 1. The underlying OS framework may provide callbacks to signal that the last
+ * client of the virtualized device has gone and the device can be removed
+ * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
+ * to periodically check if this is the right time to re-try removal of
+ * the virtualized device.
+ * 3. By any other means.
+ *
+ ******************************************************************************
+ * REQUEST CODES
+ ******************************************************************************
+ */
+#define XENCAMERA_OP_SET_CONFIG 0x00
+#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
+#define XENCAMERA_OP_BUF_CREATE 0x02
+#define XENCAMERA_OP_BUF_DESTROY 0x03
+#define XENCAMERA_OP_STREAM_START 0x04
+#define XENCAMERA_OP_STREAM_STOP 0x05
+#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
+#define XENCAMERA_OP_SET_CTRL 0x07
+
+#define XENCAMERA_CTRL_BRIGHTNESS 0x00
+#define XENCAMERA_CTRL_CONTRAST 0x01
+#define XENCAMERA_CTRL_SATURATION 0x02
+#define XENCAMERA_CTRL_HUE 0x03
+
+/*
+ ******************************************************************************
+ * EVENT CODES
+ ******************************************************************************
+ */
+#define XENCAMERA_EVT_FRAME_AVAIL 0x00
+
+/*
+ ******************************************************************************
+ * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
+ ******************************************************************************
+ */
+#define XENCAMERA_DRIVER_NAME "vcamera"
+
+#define XENCAMERA_LIST_SEPARATOR ","
+#define XENCAMERA_RESOLUTION_SEPARATOR "x"
+#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
+
+#define XENCAMERA_FIELD_BE_VERSIONS "versions"
+#define XENCAMERA_FIELD_FE_VERSION "version"
+#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
+#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
+#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
+#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
+#define XENCAMERA_FIELD_CONTROLS "controls"
+#define XENCAMERA_FIELD_FORMATS "formats"
+#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
+#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
+
+#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
+#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
+#define XENCAMERA_CTRL_SATURATION_STR "saturation"
+#define XENCAMERA_CTRL_HUE_STR "hue"
+
+/* Maximum number of buffer planes supported. */
+#define XENCAMERA_MAX_PLANE 4
+
+/*
+ ******************************************************************************
+ * STATUS RETURN CODES
+ ******************************************************************************
+ *
+ * Status return code is zero on success and -XEN_EXX on failure.
+ *
+ ******************************************************************************
+ * Assumptions
+ ******************************************************************************
+ *
+ * - usage of grant reference 0 as invalid grant reference:
+ * grant reference 0 is valid, but never exposed to a PV driver,
+ * because of the fact it is already in use/reserved by the PV console.
+ * - all references in this document to page sizes must be treated
+ * as pages of size XEN_PAGE_SIZE unless otherwise noted.
+ *
+ ******************************************************************************
+ * Description of the protocol between frontend and backend driver
+ ******************************************************************************
+ *
+ * The two halves of a Para-virtual camera driver communicate with
+ * each other using shared pages and event channels.
+ * Shared page contains a ring with request/response packets.
+ *
+ * All reserved fields in the structures below must be 0.
+ *
+ * For all request/response/event packets:
+ * - frame rate parameter is represented as a pair of 4 octet long
+ * numerator and denominator:
+ * - frame_rate_numer - uint32_t, numerator of the frame rate
+ * - frame_rate_denom - uint32_t, denominator of the frame rate
+ * The corresponding frame rate (Hz) is calculated as:
+ * frame_rate = frame_rate_numer / frame_rate_denom
+ * - buffer index is a zero based index of the buffer. Must be less than
+ * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
+ * - index - uint8_t, index of the buffer.
+ *
+ *
+ *---------------------------------- Requests ---------------------------------
+ *
+ * All request packets have the same length (64 octets).
+ * All request packets have common header:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | operation | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * id - uint16_t, private guest value, echoed in response.
+ * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
+ *
+ *
+ * Request configuration set/reset - request to set or reset.
+ * the configuration/mode of the camera:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_SET_CONFIG | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | pixel format | 12
+ * +----------------+----------------+----------------+----------------+
+ * | width | 16
+ * +----------------+----------------+----------------+----------------+
+ * | height | 20
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_numer | 24
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_denom | 28
+ * +----------------+----------------+----------------+----------------+
+ * | num_bufs | reserved | 32
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 36
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * Pass all zeros to reset, otherwise command is treated as configuration set.
+ *
+ * pixel_format - uint32_t, pixel format to be used, FOURCC code.
+ * width - uint32_t, width in pixels.
+ * height - uint32_t, height in pixels.
+ * frame_rate_numer - uint32_t, numerator of the frame rate.
+ * frame_rate_denom - uint32_t, denominator of the frame rate.
+ * num_bufs - uint8_t, desired number of buffers to be used.
+ *
+ * See response format for this request.
+ *
+ * Notes:
+ * - frontend must check the corresponding response in order to see
+ * if the values reported back by the backend do match the desired ones
+ * and can be accepted.
+ * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
+ * sending XENCAMERA_OP_STREAM_START request to update or tune the
+ * configuration.
+ */
+struct xencamera_config {
+ uint32_t pixel_format;
+ uint32_t width;
+ uint32_t height;
+ uint32_t frame_rate_nom;
+ uint32_t frame_rate_denom;
+ uint8_t num_bufs;
+};
+
+/*
+ * Request buffer details - request camera buffer's memory layout.
+ * detailed description:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_GET_BUF_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * See response format for this request.
+ *
+ *
+ * Request camera buffer creation:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_BUF_CREATE | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | gref_directory | 16
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 20
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * An attempt to create multiple buffers with the same index is an error.
+ * index can be re-used after destroying the corresponding camera buffer.
+ *
+ * index - uint8_t, index of the buffer to be created.
+ * gref_directory - grant_ref_t, a reference to the first shared page
+ * describing shared buffer references. The size of the buffer is equal to
+ * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
+ * shared buffer size exceeds what can be addressed by this single page,
+ * then reference to the next shared page must be supplied (see
+ * gref_dir_next_page below).
+ *
+ * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
+ * allocate the buffer with the parameters provided in this request and page
+ * directory is handled as follows:
+ * Frontend on request:
+ * - allocates pages for the directory (gref_directory,
+ * gref_dir_next_page(s)
+ * - grants permissions for the pages of the directory to the backend
+ * - sets gref_dir_next_page fields
+ * Backend on response:
+ * - grants permissions for the pages of the buffer allocated to
+ * the frontend
+ * - fills in page directory with grant references
+ * (gref[] in struct xencamera_page_directory)
+ */
+struct xencamera_buf_create_req {
+ uint8_t index;
+ uint8_t reserved[3];
+ grant_ref_t gref_directory;
+};
+
+/*
+ * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
+ * the request) employs a list of pages, describing all pages of the shared
+ * data buffer:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | gref_dir_next_page | 4
+ * +----------------+----------------+----------------+----------------+
+ * | gref[0] | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | gref[i] | i*4+8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | gref[N - 1] | N*4+8
+ * +----------------+----------------+----------------+----------------+
+ *
+ * gref_dir_next_page - grant_ref_t, reference to the next page describing
+ * page directory. Must be 0 if there are no more pages in the list.
+ * gref[i] - grant_ref_t, reference to a shared page of the buffer
+ * allocated at XENCAMERA_OP_BUF_CREATE.
+ *
+ * Number of grant_ref_t entries in the whole page directory is not
+ * passed, but instead can be calculated as:
+ * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
+ * XEN_PAGE_SIZE
+ */
+struct xencamera_page_directory {
+ grant_ref_t gref_dir_next_page;
+ grant_ref_t gref[1]; /* Variable length */
+};
+
+/*
+ * Request buffer destruction - destroy a previously allocated camera buffer:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_BUF_DESTROY| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 16
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * index - uint8_t, index of the buffer to be destroyed.
+ */
+
+struct xencamera_buf_destroy_req {
+ uint8_t index;
+};
+
+/*
+ * Request camera capture stream start:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_OP_STREAM_START| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ *
+ * Request camera capture stream stop:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_OP_STREAM_STOP | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ *
+ * Request camera control details:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |GET_CTRL_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 16
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * See response format for this request.
+ *
+ * index - uint8_t, index of the control to be queried.
+ */
+struct xencamera_get_ctrl_details_req {
+ uint8_t index;
+};
+
+/*
+ *
+ * Request camera control change:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_SET_CTRL | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | value | 16
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 20
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * See response format for this request.
+ *
+ * index - uint8_t, index of the control.
+ * value - int32_t, new value of the control.
+ */
+struct xencamera_set_ctrl_req {
+ uint8_t index;
+ uint8_t reserved[3];
+ int32_t value;
+};
+
+/*
+ *---------------------------------- Responses --------------------------------
+ *
+ * All response packets have the same length (64 octets).
+ *
+ * All response packets have common header:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | operation | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ *
+ * id - uint16_t, copied from the request.
+ * operation - uint8_t, XENCAMERA_OP_* - copied from request.
+ * status - int32_t, response status, zero on success and -XEN_EXX on failure.
+ *
+ *
+ * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | _OP_SET_CONFIG | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ * | pixel format | 12
+ * +----------------+----------------+----------------+----------------+
+ * | width | 16
+ * +----------------+----------------+----------------+----------------+
+ * | height | 20
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_numer | 24
+ * +----------------+----------------+----------------+----------------+
+ * | frame_rate_denom | 28
+ * +----------------+----------------+----------------+----------------+
+ * | num_bufs | reserved | 32
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 36
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * Meaning of the corresponding values in this response is the same as for
+ * XENCAMERA_OP_SET_CONFIG request.
+ *
+ *
+ * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
+ * request:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_GET_BUF_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ * | size | 12
+ * +----------------+----------------+----------------+----------------+
+ * | num_planes | reserved | 16
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[0] | 20
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[1] | 24
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[2] | 28
+ * +----------------+----------------+----------------+----------------+
+ * | plane_offset[3] | 32
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[0] | 36
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[1] | 40
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[2] | 44
+ * +----------------+----------------+----------------+----------------+
+ * | plane_size[3] | 48
+ * +----------------+----------------+----------------+----------------+
+ * | plane_stride[0] | plane_stride[1] | 52
+ * +----------------+----------------+----------------+----------------+
+ * | plane_stride[2] | plane_stride[3] | 56
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 60
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * size - uint32_t, overall size of the buffer including sizes of the
+ * individual planes and padding if applicable.
+ * num_planes - uint8_t, number of planes for this buffer.
+ * plane_offset - array of uint32_t, offset of the corresponding plane
+ * in octets from the buffer start.
+ * plane_size - array of uint32_t, size in octets of the corresponding plane
+ * including padding.
+ * plane_stride - array of uint32_t, size in octets occupied by the
+ * corresponding single image line including padding if applicable.
+ */
+struct xencamera_buf_details_resp {
+ uint32_t size;
+ uint8_t num_planes;
+ uint8_t reserved[3];
+ uint32_t plane_offset[XENCAMERA_MAX_PLANE];
+ uint32_t plane_size[XENCAMERA_MAX_PLANE];
+ uint16_t plane_stride[XENCAMERA_MAX_PLANE];
+};
+
+/*
+ * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |GET_CTRL_DETAILS| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | status | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | type | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | min | 16
+ * +----------------+----------------+----------------+----------------+
+ * | max | 20
+ * +----------------+----------------+----------------+----------------+
+ * | step | 24
+ * +----------------+----------------+----------------+----------------+
+ * | def_val | 28
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 36
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * index - uint8_t, index of the camera control in response.
+ * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
+ * min - int32_t, minimum value of the control.
+ * max - int32_t, maximum value of the control.
+ * step - int32_t, minimum size in which control value can be changed.
+ * def_val - int32_t, default value of the control.
+ */
+struct xencamera_get_ctrl_details_resp {
+ uint8_t index;
+ uint8_t type;
+ uint8_t reserved[2];
+ int32_t min;
+ int32_t max;
+ int32_t step;
+ int32_t def_val;
+};
+
+/*
+ *----------------------------------- Events ----------------------------------
+ *
+ * Events are sent via a shared page allocated by the front and propagated by
+ * evt-event-channel/evt-ring-ref XenStore entries.
+ *
+ * All event packets have the same length (64 octets).
+ * All event packets have common header:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id | type | reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ *
+ * id - uint16_t, event id, may be used by front.
+ * type - uint8_t, type of the event.
+ *
+ *
+ * Frame captured event - event from back to front when a new captured
+ * frame is available:
+ * 0 1 2 3 octet
+ * +----------------+----------------+----------------+----------------+
+ * | id |_EVT_FRAME_AVAIL| reserved | 4
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 8
+ * +----------------+----------------+----------------+----------------+
+ * | index | reserved | 12
+ * +----------------+----------------+----------------+----------------+
+ * | used_sz | 16
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 20
+ * +----------------+----------------+----------------+----------------+
+ * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
+ * +----------------+----------------+----------------+----------------+
+ * | reserved | 64
+ * +----------------+----------------+----------------+----------------+
+ *
+ * index - uint8_t, index of the buffer that contains new captured frame.
+ * used_sz - uint32_t, number of octets this frame has. This can be less
+ * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
+ */
+struct xencamera_frame_avail_evt {
+ uint8_t index;
+ uint8_t reserved[3];
+ uint32_t used_sz;
+};
+
+struct xencamera_req {
+ uint16_t id;
+ uint8_t operation;
+ uint8_t reserved[5];
+ union {
+ struct xencamera_config config;
+ struct xencamera_buf_create_req buf_create;
+ struct xencamera_buf_destroy_req buf_destroy;
+ struct xencamera_set_ctrl_req set_ctrl;
+ uint8_t reserved[56];
+ } req;
+};
+
+struct xencamera_resp {
+ uint16_t id;
+ uint8_t operation;
+ uint8_t reserved;
+ int32_t status;
+ union {
+ struct xencamera_config config;
+ struct xencamera_buf_details_resp buf_details;
+ struct xencamera_get_ctrl_details_resp ctrl_details;
+ uint8_t reserved1[56];
+ } resp;
+};
+
+struct xencamera_evt {
+ uint16_t id;
+ uint8_t type;
+ uint8_t reserved[5];
+ union {
+ struct xencamera_frame_avail_evt frame_avail;
+ uint8_t reserved[56];
+ } evt;
+};
+
+DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
+
+/*
+ ******************************************************************************
+ * Back to front events delivery
+ ******************************************************************************
+ * In order to deliver asynchronous events from back to front a shared page is
+ * allocated by front and its granted reference propagated to back via
+ * XenStore entries (evt-ring-ref/evt-event-channel).
+ * This page has a common header used by both front and back to synchronize
+ * access and control event's ring buffer, while back being a producer of the
+ * events and front being a consumer. The rest of the page after the header
+ * is used for event packets.
+ *
+ * Upon reception of an event(s) front may confirm its reception
+ * for either each event, group of events or none.
+ */
+
+struct xencamera_event_page {
+ uint32_t in_cons;
+ uint32_t in_prod;
+ uint8_t reserved[56];
+};
+
+#define XENCAMERA_EVENT_PAGE_SIZE 4096
+#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
+#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
+#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
+#define XENCAMERA_IN_RING(page) \
+ ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
+#define XENCAMERA_IN_RING_REF(page, idx) \
+ (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
+
+#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
2.18.0
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply related [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-07-31 9:31 ` [PATCH " Oleksandr Andrushchenko
@ 2018-08-14 8:30 ` Juergen Gross
-1 siblings, 0 replies; 56+ messages in thread
From: Juergen Gross @ 2018-08-14 8:30 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, boris.ostrovsky,
mchehab, linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>
> This is the ABI for the two halves of a para-virtualized
> camera driver which extends Xen's reach multimedia capabilities even
> farther enabling it for video conferencing, In-Vehicle Infotainment,
> high definition maps etc.
>
> The initial goal is to support most needed functionality with the
> final idea to make it possible to extend the protocol if need be:
>
> 1. Provide means for base virtual device configuration:
> - pixel formats
> - resolutions
> - frame rates
> 2. Support basic camera controls:
> - contrast
> - brightness
> - hue
> - saturation
> 3. Support streaming control
> 4. Support zero-copying use-cases
>
> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
Some style issues below...
> ---
> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
> 1 file changed, 981 insertions(+)
> create mode 100644 xen/include/public/io/cameraif.h
>
> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
> new file mode 100644
> index 000000000000..bdc6a1262fcf
> --- /dev/null
> +++ b/xen/include/public/io/cameraif.h
> +struct xencamera_config {
> + uint32_t pixel_format;
> + uint32_t width;
> + uint32_t height;
> + uint32_t frame_rate_nom;
> + uint32_t frame_rate_denom;
> + uint8_t num_bufs;
Add explicit padding?
> +};
> +struct xencamera_req {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved[5];
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_create_req buf_create;
> + struct xencamera_buf_destroy_req buf_destroy;
> + struct xencamera_set_ctrl_req set_ctrl;
No tabs, please.
> + uint8_t reserved[56];
> + } req;
> +};
> +
> +struct xencamera_resp {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved;
> + int32_t status;
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_details_resp buf_details;
> + struct xencamera_get_ctrl_details_resp ctrl_details;
Tab again.
> + uint8_t reserved1[56];
> + } resp;
> +};
Juergen
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-08-14 8:30 ` Juergen Gross
0 siblings, 0 replies; 56+ messages in thread
From: Juergen Gross @ 2018-08-14 8:30 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, boris.ostrovsky,
mchehab, linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>
> This is the ABI for the two halves of a para-virtualized
> camera driver which extends Xen's reach multimedia capabilities even
> farther enabling it for video conferencing, In-Vehicle Infotainment,
> high definition maps etc.
>
> The initial goal is to support most needed functionality with the
> final idea to make it possible to extend the protocol if need be:
>
> 1. Provide means for base virtual device configuration:
> - pixel formats
> - resolutions
> - frame rates
> 2. Support basic camera controls:
> - contrast
> - brightness
> - hue
> - saturation
> 3. Support streaming control
> 4. Support zero-copying use-cases
>
> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
Some style issues below...
> ---
> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
> 1 file changed, 981 insertions(+)
> create mode 100644 xen/include/public/io/cameraif.h
>
> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
> new file mode 100644
> index 000000000000..bdc6a1262fcf
> --- /dev/null
> +++ b/xen/include/public/io/cameraif.h
> +struct xencamera_config {
> + uint32_t pixel_format;
> + uint32_t width;
> + uint32_t height;
> + uint32_t frame_rate_nom;
> + uint32_t frame_rate_denom;
> + uint8_t num_bufs;
Add explicit padding?
> +};
> +struct xencamera_req {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved[5];
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_create_req buf_create;
> + struct xencamera_buf_destroy_req buf_destroy;
> + struct xencamera_set_ctrl_req set_ctrl;
No tabs, please.
> + uint8_t reserved[56];
> + } req;
> +};
> +
> +struct xencamera_resp {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved;
> + int32_t status;
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_details_resp buf_details;
> + struct xencamera_get_ctrl_details_resp ctrl_details;
Tab again.
> + uint8_t reserved1[56];
> + } resp;
> +};
Juergen
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-08-14 8:30 ` [PATCH " Juergen Gross
@ 2018-08-21 5:54 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-08-21 5:54 UTC (permalink / raw)
To: Juergen Gross, xen-devel, konrad.wilk, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 08/14/2018 11:30 AM, Juergen Gross wrote:
> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>
>> This is the ABI for the two halves of a para-virtualized
>> camera driver which extends Xen's reach multimedia capabilities even
>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>> high definition maps etc.
>>
>> The initial goal is to support most needed functionality with the
>> final idea to make it possible to extend the protocol if need be:
>>
>> 1. Provide means for base virtual device configuration:
>> - pixel formats
>> - resolutions
>> - frame rates
>> 2. Support basic camera controls:
>> - contrast
>> - brightness
>> - hue
>> - saturation
>> 3. Support streaming control
>> 4. Support zero-copying use-cases
>>
>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
> Some style issues below...
Will fix all the below, thank you!
I would like to draw some attention of the Linux/V4L community to this
protocol as the plan is that once it is accepted for Xen we plan to
upstream a Linux camera front-end kernel driver which will be based
on this work and will be a V4L2 device driver (this is why I have sent
this patch not only to Xen, but to the corresponding Linux mailing list
as well)
>> ---
>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>> 1 file changed, 981 insertions(+)
>> create mode 100644 xen/include/public/io/cameraif.h
>>
>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>> new file mode 100644
>> index 000000000000..bdc6a1262fcf
>> --- /dev/null
>> +++ b/xen/include/public/io/cameraif.h
>> +struct xencamera_config {
>> + uint32_t pixel_format;
>> + uint32_t width;
>> + uint32_t height;
>> + uint32_t frame_rate_nom;
>> + uint32_t frame_rate_denom;
>> + uint8_t num_bufs;
> Add explicit padding?
>
>> +};
>> +struct xencamera_req {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved[5];
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_create_req buf_create;
>> + struct xencamera_buf_destroy_req buf_destroy;
>> + struct xencamera_set_ctrl_req set_ctrl;
> No tabs, please.
>
>> + uint8_t reserved[56];
>> + } req;
>> +};
>> +
>> +struct xencamera_resp {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved;
>> + int32_t status;
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_details_resp buf_details;
>> + struct xencamera_get_ctrl_details_resp ctrl_details;
> Tab again.
>
>> + uint8_t reserved1[56];
>> + } resp;
>> +};
>
> Juergen
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-08-21 5:54 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-08-21 5:54 UTC (permalink / raw)
To: Juergen Gross, xen-devel, konrad.wilk, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 08/14/2018 11:30 AM, Juergen Gross wrote:
> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>
>> This is the ABI for the two halves of a para-virtualized
>> camera driver which extends Xen's reach multimedia capabilities even
>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>> high definition maps etc.
>>
>> The initial goal is to support most needed functionality with the
>> final idea to make it possible to extend the protocol if need be:
>>
>> 1. Provide means for base virtual device configuration:
>> - pixel formats
>> - resolutions
>> - frame rates
>> 2. Support basic camera controls:
>> - contrast
>> - brightness
>> - hue
>> - saturation
>> 3. Support streaming control
>> 4. Support zero-copying use-cases
>>
>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
> Some style issues below...
Will fix all the below, thank you!
I would like to draw some attention of the Linux/V4L community to this
protocol as the plan is that once it is accepted for Xen we plan to
upstream a Linux camera front-end kernel driver which will be based
on this work and will be a V4L2 device driver (this is why I have sent
this patch not only to Xen, but to the corresponding Linux mailing list
as well)
>> ---
>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>> 1 file changed, 981 insertions(+)
>> create mode 100644 xen/include/public/io/cameraif.h
>>
>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>> new file mode 100644
>> index 000000000000..bdc6a1262fcf
>> --- /dev/null
>> +++ b/xen/include/public/io/cameraif.h
>> +struct xencamera_config {
>> + uint32_t pixel_format;
>> + uint32_t width;
>> + uint32_t height;
>> + uint32_t frame_rate_nom;
>> + uint32_t frame_rate_denom;
>> + uint8_t num_bufs;
> Add explicit padding?
>
>> +};
>> +struct xencamera_req {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved[5];
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_create_req buf_create;
>> + struct xencamera_buf_destroy_req buf_destroy;
>> + struct xencamera_set_ctrl_req set_ctrl;
> No tabs, please.
>
>> + uint8_t reserved[56];
>> + } req;
>> +};
>> +
>> +struct xencamera_resp {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved;
>> + int32_t status;
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_details_resp buf_details;
>> + struct xencamera_get_ctrl_details_resp ctrl_details;
> Tab again.
>
>> + uint8_t reserved1[56];
>> + } resp;
>> +};
>
> Juergen
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-08-21 5:54 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-03 10:16 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-03 10:16 UTC (permalink / raw)
To: Juergen Gross, xen-devel, konrad.wilk, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>
>>> This is the ABI for the two halves of a para-virtualized
>>> camera driver which extends Xen's reach multimedia capabilities even
>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>> high definition maps etc.
>>>
>>> The initial goal is to support most needed functionality with the
>>> final idea to make it possible to extend the protocol if need be:
>>>
>>> 1. Provide means for base virtual device configuration:
>>> - pixel formats
>>> - resolutions
>>> - frame rates
>>> 2. Support basic camera controls:
>>> - contrast
>>> - brightness
>>> - hue
>>> - saturation
>>> 3. Support streaming control
>>> 4. Support zero-copying use-cases
>>>
>>> Signed-off-by: Oleksandr Andrushchenko
>>> <oleksandr_andrushchenko@epam.com>
>> Some style issues below...
> Will fix all the below, thank you!
>
> I would like to draw some attention of the Linux/V4L community to this
> protocol as the plan is that once it is accepted for Xen we plan to
> upstream a Linux camera front-end kernel driver which will be based
> on this work and will be a V4L2 device driver (this is why I have sent
> this patch not only to Xen, but to the corresponding Linux mailing list
> as well)
ping
>
>>> ---
>>> xen/include/public/io/cameraif.h | 981
>>> +++++++++++++++++++++++++++++++
>>> 1 file changed, 981 insertions(+)
>>> create mode 100644 xen/include/public/io/cameraif.h
>>>
>>> diff --git a/xen/include/public/io/cameraif.h
>>> b/xen/include/public/io/cameraif.h
>>> new file mode 100644
>>> index 000000000000..bdc6a1262fcf
>>> --- /dev/null
>>> +++ b/xen/include/public/io/cameraif.h
>>> +struct xencamera_config {
>>> + uint32_t pixel_format;
>>> + uint32_t width;
>>> + uint32_t height;
>>> + uint32_t frame_rate_nom;
>>> + uint32_t frame_rate_denom;
>>> + uint8_t num_bufs;
>> Add explicit padding?
>>
>>> +};
>>> +struct xencamera_req {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved[5];
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_create_req buf_create;
>>> + struct xencamera_buf_destroy_req buf_destroy;
>>> + struct xencamera_set_ctrl_req set_ctrl;
>> No tabs, please.
>>
>>> + uint8_t reserved[56];
>>> + } req;
>>> +};
>>> +
>>> +struct xencamera_resp {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved;
>>> + int32_t status;
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_details_resp buf_details;
>>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>> Tab again.
>>
>>> + uint8_t reserved1[56];
>>> + } resp;
>>> +};
>>
>> Juergen
> Thank you,
> Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-03 10:16 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-03 10:16 UTC (permalink / raw)
To: Juergen Gross, xen-devel, konrad.wilk, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>
>>> This is the ABI for the two halves of a para-virtualized
>>> camera driver which extends Xen's reach multimedia capabilities even
>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>> high definition maps etc.
>>>
>>> The initial goal is to support most needed functionality with the
>>> final idea to make it possible to extend the protocol if need be:
>>>
>>> 1. Provide means for base virtual device configuration:
>>> - pixel formats
>>> - resolutions
>>> - frame rates
>>> 2. Support basic camera controls:
>>> - contrast
>>> - brightness
>>> - hue
>>> - saturation
>>> 3. Support streaming control
>>> 4. Support zero-copying use-cases
>>>
>>> Signed-off-by: Oleksandr Andrushchenko
>>> <oleksandr_andrushchenko@epam.com>
>> Some style issues below...
> Will fix all the below, thank you!
>
> I would like to draw some attention of the Linux/V4L community to this
> protocol as the plan is that once it is accepted for Xen we plan to
> upstream a Linux camera front-end kernel driver which will be based
> on this work and will be a V4L2 device driver (this is why I have sent
> this patch not only to Xen, but to the corresponding Linux mailing list
> as well)
ping
>
>>> ---
>>> xen/include/public/io/cameraif.h | 981
>>> +++++++++++++++++++++++++++++++
>>> 1 file changed, 981 insertions(+)
>>> create mode 100644 xen/include/public/io/cameraif.h
>>>
>>> diff --git a/xen/include/public/io/cameraif.h
>>> b/xen/include/public/io/cameraif.h
>>> new file mode 100644
>>> index 000000000000..bdc6a1262fcf
>>> --- /dev/null
>>> +++ b/xen/include/public/io/cameraif.h
>>> +struct xencamera_config {
>>> + uint32_t pixel_format;
>>> + uint32_t width;
>>> + uint32_t height;
>>> + uint32_t frame_rate_nom;
>>> + uint32_t frame_rate_denom;
>>> + uint8_t num_bufs;
>> Add explicit padding?
>>
>>> +};
>>> +struct xencamera_req {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved[5];
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_create_req buf_create;
>>> + struct xencamera_buf_destroy_req buf_destroy;
>>> + struct xencamera_set_ctrl_req set_ctrl;
>> No tabs, please.
>>
>>> + uint8_t reserved[56];
>>> + } req;
>>> +};
>>> +
>>> +struct xencamera_resp {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved;
>>> + int32_t status;
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_details_resp buf_details;
>>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>> Tab again.
>>
>>> + uint8_t reserved1[56];
>>> + } resp;
>>> +};
>>
>> Juergen
> Thank you,
> Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-03 10:16 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-03 15:25 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-03 15:25 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi Oleksandr,
On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>
>>>> This is the ABI for the two halves of a para-virtualized
>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>> high definition maps etc.
>>>>
>>>> The initial goal is to support most needed functionality with the
>>>> final idea to make it possible to extend the protocol if need be:
>>>>
>>>> 1. Provide means for base virtual device configuration:
>>>> - pixel formats
>>>> - resolutions
>>>> - frame rates
>>>> 2. Support basic camera controls:
>>>> - contrast
>>>> - brightness
>>>> - hue
>>>> - saturation
>>>> 3. Support streaming control
>>>> 4. Support zero-copying use-cases
>>>>
>>>> Signed-off-by: Oleksandr Andrushchenko
>>>> <oleksandr_andrushchenko@epam.com>
>>> Some style issues below...
>> Will fix all the below, thank you!
>>
>> I would like to draw some attention of the Linux/V4L community to this
>> protocol as the plan is that once it is accepted for Xen we plan to
>> upstream a Linux camera front-end kernel driver which will be based
>> on this work and will be a V4L2 device driver (this is why I have sent
>> this patch not only to Xen, but to the corresponding Linux mailing list
>> as well)
> ping
Sorry, this got buried in my mailbox, I only came across it today. I'll try
to review this this week, if not, just ping me again.
I had one high-level question, though:
What types of hardware do you intend to target? This initial version targets
(very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
codecs? Or complex embedded video pipelines?
In other words, where are you planning to draw the line?
Even with just simple cameras there is a difference between regular UVC
webcams and cameras used with embedded systems: for the latter you often
need to provide more control w.r.t. white-balancing etc., things that a
UVC webcam will generally do for you in the webcam's firmware.
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-03 15:25 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-03 15:25 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi Oleksandr,
On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>
>>>> This is the ABI for the two halves of a para-virtualized
>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>> high definition maps etc.
>>>>
>>>> The initial goal is to support most needed functionality with the
>>>> final idea to make it possible to extend the protocol if need be:
>>>>
>>>> 1. Provide means for base virtual device configuration:
>>>> - pixel formats
>>>> - resolutions
>>>> - frame rates
>>>> 2. Support basic camera controls:
>>>> - contrast
>>>> - brightness
>>>> - hue
>>>> - saturation
>>>> 3. Support streaming control
>>>> 4. Support zero-copying use-cases
>>>>
>>>> Signed-off-by: Oleksandr Andrushchenko
>>>> <oleksandr_andrushchenko@epam.com>
>>> Some style issues below...
>> Will fix all the below, thank you!
>>
>> I would like to draw some attention of the Linux/V4L community to this
>> protocol as the plan is that once it is accepted for Xen we plan to
>> upstream a Linux camera front-end kernel driver which will be based
>> on this work and will be a V4L2 device driver (this is why I have sent
>> this patch not only to Xen, but to the corresponding Linux mailing list
>> as well)
> ping
Sorry, this got buried in my mailbox, I only came across it today. I'll try
to review this this week, if not, just ping me again.
I had one high-level question, though:
What types of hardware do you intend to target? This initial version targets
(very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
codecs? Or complex embedded video pipelines?
In other words, where are you planning to draw the line?
Even with just simple cameras there is a difference between regular UVC
webcams and cameras used with embedded systems: for the latter you often
need to provide more control w.r.t. white-balancing etc., things that a
UVC webcam will generally do for you in the webcam's firmware.
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-03 15:25 ` [PATCH " Hans Verkuil
@ 2018-09-04 6:56 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-04 6:56 UTC (permalink / raw)
To: Hans Verkuil, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko, Artem Mygaiev
On 09/03/2018 06:25 PM, Hans Verkuil wrote:
> Hi Oleksandr,
>
> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>
>>>>> This is the ABI for the two halves of a para-virtualized
>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>> high definition maps etc.
>>>>>
>>>>> The initial goal is to support most needed functionality with the
>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>
>>>>> 1. Provide means for base virtual device configuration:
>>>>> - pixel formats
>>>>> - resolutions
>>>>> - frame rates
>>>>> 2. Support basic camera controls:
>>>>> - contrast
>>>>> - brightness
>>>>> - hue
>>>>> - saturation
>>>>> 3. Support streaming control
>>>>> 4. Support zero-copying use-cases
>>>>>
>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>> <oleksandr_andrushchenko@epam.com>
>>>> Some style issues below...
>>> Will fix all the below, thank you!
>>>
>>> I would like to draw some attention of the Linux/V4L community to this
>>> protocol as the plan is that once it is accepted for Xen we plan to
>>> upstream a Linux camera front-end kernel driver which will be based
>>> on this work and will be a V4L2 device driver (this is why I have sent
>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>> as well)
>> ping
> Sorry, this got buried in my mailbox, I only came across it today. I'll try
> to review this this week, if not, just ping me again.
Thank you for your time
>
> I had one high-level question, though:
>
> What types of hardware do you intend to target? This initial version targets
> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
> codecs? Or complex embedded video pipelines?
>
> In other words, where are you planning to draw the line?
>
> Even with just simple cameras there is a difference between regular UVC
> webcams and cameras used with embedded systems: for the latter you often
> need to provide more control w.r.t. white-balancing etc., things that a
> UVC webcam will generally do for you in the webcam's firmware.
The use-cases we want to implement are mostly in automotive/embedded domain,
so there are many performance restrictions apply.
We are not targeting virtualizing very complex hardware and have no
intention
to make a 1:1 mapping of the real hardware: for that one can pass-through
a real HW device to a virtual machine (VM). The goal is to share a single
camera device to multiple virtual machines, no codecs, receivers etc.
Controlling the same HW device from different VMs doesn't look feasible:
what if the same control is set to different values from different VMs?
Of course, this can be achieved if the corresponding backend can
post-process
original camera image with GPU, for example, thus applying different filters
for different VMs effectively emulating camera controls.
But this requires additional CPU/GPU power which we try to avoid.
System partitioning (camera and controls assignment) is done at
configuration
time (remember we are in automotive/embedded world, so most of the time
the set
of VMs requiring cameras is known at this stage and the configuration
remains
static at run-time). So, when para-virtualized (PV) approach is used then we
only implement very basic controls (those found in the protocol), so one can
assign set of controls (all or some) to one of the VMs (main or mission
critical
VM or whatever) allowing that VM to adjusts those for all VMs at once.
For other
VMs think of it as firmware implemented adjustment. And the backend still
controls the rest of the controls of the real HW camera you mention.
Just an example of automotive use-case (we can imagine many more):
1. Driver Domain - owns real camera HW and runs the camera backend.
Uses camera output for mission critical tasks, e.g. parking assistance.
2. In-Vehicle Infotainment domain - uses PV camera for infotainment
purposes,
e.g. taking pictures while in motion.
3. Navigation domain - uses PV camera for high definition maps
Hope, this helps understanding the possible uses of the proposed
protocol, its
intention and restrictions.
> Regards,
>
> Hans
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-04 6:56 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-04 6:56 UTC (permalink / raw)
To: Hans Verkuil, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Artem Mygaiev, Oleksandr Andrushchenko
On 09/03/2018 06:25 PM, Hans Verkuil wrote:
> Hi Oleksandr,
>
> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>
>>>>> This is the ABI for the two halves of a para-virtualized
>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>> high definition maps etc.
>>>>>
>>>>> The initial goal is to support most needed functionality with the
>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>
>>>>> 1. Provide means for base virtual device configuration:
>>>>> - pixel formats
>>>>> - resolutions
>>>>> - frame rates
>>>>> 2. Support basic camera controls:
>>>>> - contrast
>>>>> - brightness
>>>>> - hue
>>>>> - saturation
>>>>> 3. Support streaming control
>>>>> 4. Support zero-copying use-cases
>>>>>
>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>> <oleksandr_andrushchenko@epam.com>
>>>> Some style issues below...
>>> Will fix all the below, thank you!
>>>
>>> I would like to draw some attention of the Linux/V4L community to this
>>> protocol as the plan is that once it is accepted for Xen we plan to
>>> upstream a Linux camera front-end kernel driver which will be based
>>> on this work and will be a V4L2 device driver (this is why I have sent
>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>> as well)
>> ping
> Sorry, this got buried in my mailbox, I only came across it today. I'll try
> to review this this week, if not, just ping me again.
Thank you for your time
>
> I had one high-level question, though:
>
> What types of hardware do you intend to target? This initial version targets
> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
> codecs? Or complex embedded video pipelines?
>
> In other words, where are you planning to draw the line?
>
> Even with just simple cameras there is a difference between regular UVC
> webcams and cameras used with embedded systems: for the latter you often
> need to provide more control w.r.t. white-balancing etc., things that a
> UVC webcam will generally do for you in the webcam's firmware.
The use-cases we want to implement are mostly in automotive/embedded domain,
so there are many performance restrictions apply.
We are not targeting virtualizing very complex hardware and have no
intention
to make a 1:1 mapping of the real hardware: for that one can pass-through
a real HW device to a virtual machine (VM). The goal is to share a single
camera device to multiple virtual machines, no codecs, receivers etc.
Controlling the same HW device from different VMs doesn't look feasible:
what if the same control is set to different values from different VMs?
Of course, this can be achieved if the corresponding backend can
post-process
original camera image with GPU, for example, thus applying different filters
for different VMs effectively emulating camera controls.
But this requires additional CPU/GPU power which we try to avoid.
System partitioning (camera and controls assignment) is done at
configuration
time (remember we are in automotive/embedded world, so most of the time
the set
of VMs requiring cameras is known at this stage and the configuration
remains
static at run-time). So, when para-virtualized (PV) approach is used then we
only implement very basic controls (those found in the protocol), so one can
assign set of controls (all or some) to one of the VMs (main or mission
critical
VM or whatever) allowing that VM to adjusts those for all VMs at once.
For other
VMs think of it as firmware implemented adjustment. And the backend still
controls the rest of the controls of the real HW camera you mention.
Just an example of automotive use-case (we can imagine many more):
1. Driver Domain - owns real camera HW and runs the camera backend.
Uses camera output for mission critical tasks, e.g. parking assistance.
2. In-Vehicle Infotainment domain - uses PV camera for infotainment
purposes,
e.g. taking pictures while in motion.
3. Navigation domain - uses PV camera for high definition maps
Hope, this helps understanding the possible uses of the proposed
protocol, its
intention and restrictions.
> Regards,
>
> Hans
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-07-31 9:31 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-09 10:31 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-09 10:31 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, jgross,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi Oleksandr,
Sorry for the delay in reviewing, I missed this patch until you pinged me, and
I was very busy after that as well.
On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>
> This is the ABI for the two halves of a para-virtualized
> camera driver which extends Xen's reach multimedia capabilities even
> farther enabling it for video conferencing, In-Vehicle Infotainment,
> high definition maps etc.
>
> The initial goal is to support most needed functionality with the
> final idea to make it possible to extend the protocol if need be:
>
> 1. Provide means for base virtual device configuration:
> - pixel formats
> - resolutions
> - frame rates
> 2. Support basic camera controls:
> - contrast
> - brightness
> - hue
> - saturation
> 3. Support streaming control
> 4. Support zero-copying use-cases
>
> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
> ---
> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
> 1 file changed, 981 insertions(+)
> create mode 100644 xen/include/public/io/cameraif.h
>
> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
> new file mode 100644
> index 000000000000..bdc6a1262fcf
> --- /dev/null
> +++ b/xen/include/public/io/cameraif.h
> @@ -0,0 +1,981 @@
> +/******************************************************************************
> + * cameraif.h
> + *
> + * Unified camera device I/O interface for Xen guest OSes.
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a copy
> + * of this software and associated documentation files (the "Software"), to
> + * deal in the Software without restriction, including without limitation the
> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
> + * sell copies of the Software, and to permit persons to whom the Software is
> + * furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> + * DEALINGS IN THE SOFTWARE.
> + *
> + * Copyright (C) 2018 EPAM Systems Inc.
> + *
> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
> + */
Use SPDX tag instead of copying the license text.
> +
> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
> +
> +#include "ring.h"
> +#include "../grant_table.h"
> +
> +/*
> + ******************************************************************************
> + * Protocol version
> + ******************************************************************************
> + */
> +#define XENCAMERA_PROTOCOL_VERSION "1"
> +
> +/*
> + ******************************************************************************
> + * Feature and Parameter Negotiation
> + ******************************************************************************
> + *
> + * Front->back notifications: when enqueuing a new request, sending a
> + * notification can be made conditional on xencamera_req (i.e., the generic
> + * hold-off mechanism provided by the ring macros). Backends must set
> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
> + *
> + * Back->front notifications: when enqueuing a new response, sending a
> + * notification can be made conditional on xencamera_resp (i.e., the generic
> + * hold-off mechanism provided by the ring macros). Frontends must set
> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
> + *
> + * The two halves of a para-virtual camera driver utilize nodes within
> + * XenStore to communicate capabilities and to negotiate operating parameters.
> + * This section enumerates these nodes which reside in the respective front and
> + * backend portions of XenStore, following the XenBus convention.
> + *
> + * All data in XenStore is stored as strings. Nodes specifying numeric
> + * values are encoded in decimal. Integer value ranges listed below are
> + * expressed as fixed sized integer types capable of storing the conversion
> + * of a properly formatted node string, without loss of information.
> + *
> + ******************************************************************************
> + * Example configuration
> + ******************************************************************************
> + *
> + * This is an example of backend and frontend configuration:
> + *
> + *--------------------------------- Backend -----------------------------------
> + *
> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
> + * /local/domain/0/backend/vcamera/1/0/state = "4"
> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
superfluous to me.
> + *
> + *--------------------------------- Frontend ----------------------------------
> + *
> + * /local/domain/1/device/vcamera/0/backend-id = "0"
> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
> + * /local/domain/1/device/vcamera/0/state = "4"
> + * /local/domain/1/device/vcamera/0/version = "1"
> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
> + *
> + *---------------------------- Device 0 configuration -------------------------
> + *
> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
> + * /local/domain/1/device/vcamera/0/unique-id = "0"
> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
> + *
> + *---------------------------- Device 1 configuration -------------------------
> + *
> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
> + * /local/domain/1/device/vcamera/1/unique-id = "1"
> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
> + *
> + ******************************************************************************
> + * Backend XenBus Nodes
> + ******************************************************************************
> + *
> + *----------------------------- Protocol version ------------------------------
> + *
> + * versions
> + * Values: <string>
> + *
> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
> + * by the backend. For example "1,2,3".
> + *
> + ******************************************************************************
> + * Frontend XenBus Nodes
> + ******************************************************************************
> + *
> + *-------------------------------- Addressing ---------------------------------
> + *
> + * dom-id
> + * Values: <uint16_t>
> + *
> + * Domain identifier.
> + *
> + * dev-id
> + * Values: <uint16_t>
> + *
> + * Device identifier.
> + *
> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
> + *
> + *----------------------------- Protocol version ------------------------------
> + *
> + * version
> + * Values: <string>
> + *
> + * Protocol version, chosen among the ones supported by the backend.
> + *
> + *------------------------- Backend buffer allocation -------------------------
> + *
> + * be-alloc
I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
is better to just write 'backend-alloc'.
> + * Values: "0", "1"
> + *
> + * If value is set to "1", then backend will be the buffer
> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
> + * operation.
> + * If value is not "1" or omitted frontend must allocate buffers itself.
> + *
> + *------------------------------- Camera settings -----------------------------
> + *
> + * unique-id
> + * Values: <string>
> + *
> + * After device instance initialization each camera is assigned a
> + * unique ID, so it can be identified by the backend by this ID.
> + * This can be UUID or such.
> + *
> + * controls
> + * Values: <list of string>
> + *
> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
> + * Camera controls are expressed as a list of string values w/o any
> + * ordering requirement.
> + *
> + * formats
> + * Values: <format, char[4]>
> + *
> + * Formats are organized as a set of directories one per each
> + * supported pixel format. The name of the directory is an upper case
> + * string of the corresponding FOURCC string label. The next level of
> + * the directory under <formats> represents supported resolutions.
Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
In addition it is common to set bit 31 of the fourcc to 1 if the format is
big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
keep that convention.
> + *
> + * resolution
> + * Values: <width, uint32_t>x<height, uint32_t>
> + *
> + * Resolutions are organized as a set of directories one per each
> + * supported resolution under corresponding <formats> directory.
> + * The name of the directory is the supported width and height
> + * of the camera resolution in pixels.
What if you are dealing with an HDMI input? Not unreasonable for media
systems. There can be a lot of resolutions/framerates, and the resolution
can change on the fly, or of course disappear.
What is also missing here is a way to report pixel aspect ratio: PAL and
NTSC-based video material doesn't have square pixels.
It's important to decide whether or not you want to support video sources
like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
or disconnect webcams) or if you stick to fixed camera pipelines.
The big difference is that you don't control what someone can connect as
external sources, so you will have to be a lot more careful and robust.
I suspect that you likely will want to support such sources eventually, so
it pays to design this with that in mind.
> + *
> + * frame-rates
> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
> + *
> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
> + * of the camera expressed as numerator and denominator of the
> + * corresponding frame rate.
> + *
> + * The format of the <formats> directory tree with resolutions and frame rates
> + * must be structured in the following format:
> + *
> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
> + *
> + * where
> + * i - i-th supported pixel format
> + * j - j-th supported resolution for i-th pixel format
> + * k - k-th supported frame rate for i-th pixel format and j-th
> + * resolution> + *
> + *------------------- Camera Request Transport Parameters ---------------------
> + *
> + * This communication path is used to deliver requests from frontend to backend
> + * and get the corresponding responses from backend to frontend,
> + * set up per virtual camera device.
> + *
> + * req-event-channel
> + * Values: <uint32_t>
> + *
> + * The identifier of the Xen camera's control event channel
> + * used to signal activity in the ring buffer.
> + *
> + * req-ring-ref
> + * Values: <uint32_t>
> + *
> + * The Xen grant reference granting permission for the backend to map
> + * a sole page of camera's control ring buffer.
> + *
> + *-------------------- Camera Event Transport Parameters ----------------------
> + *
> + * This communication path is used to deliver asynchronous events from backend
> + * to frontend, set up per virtual camera device.
> + *
> + * evt-event-channel
> + * Values: <uint32_t>
> + *
> + * The identifier of the Xen camera's event channel
> + * used to signal activity in the ring buffer.
> + *
> + * evt-ring-ref
> + * Values: <uint32_t>
> + *
> + * The Xen grant reference granting permission for the backend to map
> + * a sole page of camera's event ring buffer.
> + */
> +
> +/*
> + ******************************************************************************
> + * STATE DIAGRAMS
> + ******************************************************************************
> + *
> + * Tool stack creates front and back state nodes with initial state
> + * XenbusStateInitialising.
> + * Tool stack creates and sets up frontend camera configuration
> + * nodes per domain.
> + *
> + *-------------------------------- Normal flow --------------------------------
> + *
> + * Front Back
> + * ================================= =====================================
> + * XenbusStateInitialising XenbusStateInitialising
> + * o Query backend device identification
> + * data.
> + * o Open and validate backend device.
> + * |
> + * |
> + * V
> + * XenbusStateInitWait
> + *
> + * o Query frontend configuration
> + * o Allocate and initialize
> + * event channels per configured
> + * camera.
> + * o Publish transport parameters
> + * that will be in effect during
> + * this connection.
> + * |
> + * |
> + * V
> + * XenbusStateInitialised
> + *
> + * o Query frontend transport parameters.
> + * o Connect to the event channels.
> + * |
> + * |
> + * V
> + * XenbusStateConnected
> + *
> + * o Create and initialize OS
> + * virtual camera as per
> + * configuration.
> + * |
> + * |
> + * V
> + * XenbusStateConnected
> + *
> + * XenbusStateUnknown
> + * XenbusStateClosed
> + * XenbusStateClosing
> + * o Remove virtual camera device
> + * o Remove event channels
> + * |
> + * |
> + * V
> + * XenbusStateClosed
> + *
> + *------------------------------- Recovery flow -------------------------------
> + *
> + * In case of frontend unrecoverable errors backend handles that as
> + * if frontend goes into the XenbusStateClosed state.
> + *
> + * In case of backend unrecoverable errors frontend tries removing
> + * the virtualized device. If this is possible at the moment of error,
> + * then frontend goes into the XenbusStateInitialising state and is ready for
> + * new connection with backend. If the virtualized device is still in use and
> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
> + * until either the virtualized device is removed or backend initiates a new
> + * connection. On the virtualized device removal frontend goes into the
> + * XenbusStateInitialising state.
> + *
> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
> + * unrecoverable errors then frontend cannot send requests to the backend
> + * and thus cannot provide functionality of the virtualized device anymore.
> + * After backend is back to normal the virtualized device may still hold some
> + * state: configuration in use, allocated buffers, client application state etc.
> + * In most cases, this will require frontend to implement complex recovery
> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
> + * frontend will make sure no new clients of the virtualized device are
> + * accepted, allow existing client(s) to exit gracefully by signaling error
> + * state etc.
> + * Once all the clients are gone frontend can reinitialize the virtualized
> + * device and get into XenbusStateInitialising state again signaling the
> + * backend that a new connection can be made.
> + *
> + * There are multiple conditions possible under which frontend will go from
> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
> + * specific. For example:
> + * 1. The underlying OS framework may provide callbacks to signal that the last
> + * client of the virtualized device has gone and the device can be removed
> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
> + * to periodically check if this is the right time to re-try removal of
> + * the virtualized device.
> + * 3. By any other means.
> + *
> + ******************************************************************************
> + * REQUEST CODES
> + ******************************************************************************
> + */
> +#define XENCAMERA_OP_SET_CONFIG 0x00
> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
> +#define XENCAMERA_OP_BUF_CREATE 0x02
> +#define XENCAMERA_OP_BUF_DESTROY 0x03
> +#define XENCAMERA_OP_STREAM_START 0x04
> +#define XENCAMERA_OP_STREAM_STOP 0x05
> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
> +#define XENCAMERA_OP_SET_CTRL 0x07
> +
> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
> +#define XENCAMERA_CTRL_CONTRAST 0x01
> +#define XENCAMERA_CTRL_SATURATION 0x02
> +#define XENCAMERA_CTRL_HUE 0x03
> +
> +/*
> + ******************************************************************************
> + * EVENT CODES
> + ******************************************************************************
> + */
> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
> +
> +/*
> + ******************************************************************************
> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
> + ******************************************************************************
> + */
> +#define XENCAMERA_DRIVER_NAME "vcamera"
Ah, that's where vcamera comes from. How about calling this xen-camera or
virt-camera? With a preference for xen-camera, since that's what you use for the
defines as well.
Or perhaps pv-camera?
Is this driver going to be xen-specific, or more a general approach that everyone
can use? Obviously, the latter would be preferable.
BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
radio and RDS hardware. Contact me if this is of interest to Xen to support this as
well given the automotive use-case.
> +
> +#define XENCAMERA_LIST_SEPARATOR ","
> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
> +
> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
> +#define XENCAMERA_FIELD_FE_VERSION "version"
> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
> +#define XENCAMERA_FIELD_CONTROLS "controls"
> +#define XENCAMERA_FIELD_FORMATS "formats"
> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
> +
> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
> +#define XENCAMERA_CTRL_HUE_STR "hue"
> +
> +/* Maximum number of buffer planes supported. */
> +#define XENCAMERA_MAX_PLANE 4
> +
> +/*
> + ******************************************************************************
> + * STATUS RETURN CODES
> + ******************************************************************************
> + *
> + * Status return code is zero on success and -XEN_EXX on failure.
> + *
> + ******************************************************************************
> + * Assumptions
> + ******************************************************************************
> + *
> + * - usage of grant reference 0 as invalid grant reference:
> + * grant reference 0 is valid, but never exposed to a PV driver,
> + * because of the fact it is already in use/reserved by the PV console.
> + * - all references in this document to page sizes must be treated
> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
> + *
> + ******************************************************************************
> + * Description of the protocol between frontend and backend driver
> + ******************************************************************************
> + *
> + * The two halves of a Para-virtual camera driver communicate with
> + * each other using shared pages and event channels.
> + * Shared page contains a ring with request/response packets.
> + *
> + * All reserved fields in the structures below must be 0.
> + *
> + * For all request/response/event packets:
> + * - frame rate parameter is represented as a pair of 4 octet long
> + * numerator and denominator:
> + * - frame_rate_numer - uint32_t, numerator of the frame rate
> + * - frame_rate_denom - uint32_t, denominator of the frame rate
> + * The corresponding frame rate (Hz) is calculated as:
> + * frame_rate = frame_rate_numer / frame_rate_denom
> + * - buffer index is a zero based index of the buffer. Must be less than
> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
> + * - index - uint8_t, index of the buffer.
> + *
> + *
> + *---------------------------------- Requests ---------------------------------
> + *
> + * All request packets have the same length (64 octets).
> + * All request packets have common header:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | operation | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * id - uint16_t, private guest value, echoed in response.
> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
> + *
> + *
> + * Request configuration set/reset - request to set or reset.
> + * the configuration/mode of the camera:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_SET_CONFIG | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | pixel format | 12
> + * +----------------+----------------+----------------+----------------+
> + * | width | 16
> + * +----------------+----------------+----------------+----------------+
> + * | height | 20
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_numer | 24
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_denom | 28
> + * +----------------+----------------+----------------+----------------+
> + * | num_bufs | reserved | 32
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 36
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * Pass all zeros to reset, otherwise command is treated as configuration set.
> + *
> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
> + * width - uint32_t, width in pixels.
> + * height - uint32_t, height in pixels.
> + * frame_rate_numer - uint32_t, numerator of the frame rate.
> + * frame_rate_denom - uint32_t, denominator of the frame rate.
If you have to support HDMI/SDTV inputs as well, then you also need to know
the interlaced format, unless you have no plans to support that.
> + * num_bufs - uint8_t, desired number of buffers to be used.
Huh? What has that to do with the format? Why would you need this here?
> + *
> + * See response format for this request.
> + *
> + * Notes:
> + * - frontend must check the corresponding response in order to see
> + * if the values reported back by the backend do match the desired ones
> + * and can be accepted.
> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
> + * configuration.
> + */
> +struct xencamera_config {
> + uint32_t pixel_format;
> + uint32_t width;
> + uint32_t height;
> + uint32_t frame_rate_nom;
> + uint32_t frame_rate_denom;
> + uint8_t num_bufs;
> +};
> +
> +/*
> + * Request buffer details - request camera buffer's memory layout.
> + * detailed description:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_GET_BUF_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * See response format for this request.
> + *
> + *
> + * Request camera buffer creation:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_BUF_CREATE | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | gref_directory | 16
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 20
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * An attempt to create multiple buffers with the same index is an error.
> + * index can be re-used after destroying the corresponding camera buffer.
> + *
> + * index - uint8_t, index of the buffer to be created.
> + * gref_directory - grant_ref_t, a reference to the first shared page
> + * describing shared buffer references. The size of the buffer is equal to
> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
> + * shared buffer size exceeds what can be addressed by this single page,
> + * then reference to the next shared page must be supplied (see
> + * gref_dir_next_page below).
It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
does.
> + *
> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
> + * allocate the buffer with the parameters provided in this request and page
> + * directory is handled as follows:
> + * Frontend on request:
> + * - allocates pages for the directory (gref_directory,
> + * gref_dir_next_page(s)
> + * - grants permissions for the pages of the directory to the backend
> + * - sets gref_dir_next_page fields
> + * Backend on response:
> + * - grants permissions for the pages of the buffer allocated to
> + * the frontend
> + * - fills in page directory with grant references
> + * (gref[] in struct xencamera_page_directory)
> + */
> +struct xencamera_buf_create_req {
> + uint8_t index;
> + uint8_t reserved[3];
> + grant_ref_t gref_directory;
> +};
> +
> +/*
> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
> + * the request) employs a list of pages, describing all pages of the shared
> + * data buffer:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | gref_dir_next_page | 4
> + * +----------------+----------------+----------------+----------------+
> + * | gref[0] | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | gref[i] | i*4+8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | gref[N - 1] | N*4+8
> + * +----------------+----------------+----------------+----------------+
> + *
> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
> + * page directory. Must be 0 if there are no more pages in the list.
> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
> + * allocated at XENCAMERA_OP_BUF_CREATE.
> + *
> + * Number of grant_ref_t entries in the whole page directory is not
> + * passed, but instead can be calculated as:
> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
> + * XEN_PAGE_SIZE
> + */
> +struct xencamera_page_directory {
> + grant_ref_t gref_dir_next_page;
> + grant_ref_t gref[1]; /* Variable length */
> +};
> +
> +/*
> + * Request buffer destruction - destroy a previously allocated camera buffer:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_BUF_DESTROY| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 16
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * index - uint8_t, index of the buffer to be destroyed.
> + */
There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
of them.
> +
> +struct xencamera_buf_destroy_req {
> + uint8_t index;
> +};
> +
> +/*
> + * Request camera capture stream start:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_OP_STREAM_START| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + *
> + * Request camera capture stream stop:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_OP_STREAM_STOP | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + *
> + * Request camera control details:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |GET_CTRL_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 16
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * See response format for this request.
> + *
> + * index - uint8_t, index of the control to be queried.
> + */
> +struct xencamera_get_ctrl_details_req {
> + uint8_t index;
> +};
> +
> +/*
> + *
> + * Request camera control change:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_SET_CTRL | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | value | 16
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 20
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * See response format for this request.
> + *
> + * index - uint8_t, index of the control.
> + * value - int32_t, new value of the control.
I would recommend using a int64_t as the control value.
Note that there are also controls with a payload (e.g. string controls).
If there is ever interest in adding radio/RDS support, then that will become
an issue.
> + */
> +struct xencamera_set_ctrl_req {
> + uint8_t index;
> + uint8_t reserved[3];
> + int32_t value;
> +};
> +
> +/*
> + *---------------------------------- Responses --------------------------------
> + *
> + * All response packets have the same length (64 octets).
> + *
> + * All response packets have common header:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | operation | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + *
> + * id - uint16_t, copied from the request.
> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
> + *
> + *
> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_SET_CONFIG | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + * | pixel format | 12
> + * +----------------+----------------+----------------+----------------+
> + * | width | 16
> + * +----------------+----------------+----------------+----------------+
> + * | height | 20
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_numer | 24
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_denom | 28
> + * +----------------+----------------+----------------+----------------+
> + * | num_bufs | reserved | 32
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 36
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * Meaning of the corresponding values in this response is the same as for
> + * XENCAMERA_OP_SET_CONFIG request.
> + *
> + *
> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
> + * request:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_GET_BUF_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + * | size | 12
> + * +----------------+----------------+----------------+----------------+
> + * | num_planes | reserved | 16
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[0] | 20
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[1] | 24
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[2] | 28
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[3] | 32
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[0] | 36
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[1] | 40
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[2] | 44
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[3] | 48
> + * +----------------+----------------+----------------+----------------+
> + * | plane_stride[0] | plane_stride[1] | 52
> + * +----------------+----------------+----------------+----------------+
> + * | plane_stride[2] | plane_stride[3] | 56
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 60
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * size - uint32_t, overall size of the buffer including sizes of the
> + * individual planes and padding if applicable.
> + * num_planes - uint8_t, number of planes for this buffer.
> + * plane_offset - array of uint32_t, offset of the corresponding plane
> + * in octets from the buffer start.
> + * plane_size - array of uint32_t, size in octets of the corresponding plane
> + * including padding.
> + * plane_stride - array of uint32_t, size in octets occupied by the
> + * corresponding single image line including padding if applicable.
Nice!
> + */
> +struct xencamera_buf_details_resp {
> + uint32_t size;
> + uint8_t num_planes;
> + uint8_t reserved[3];
> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
> +};
> +
> +/*
> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |GET_CTRL_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | type | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | min | 16
> + * +----------------+----------------+----------------+----------------+
> + * | max | 20
> + * +----------------+----------------+----------------+----------------+
> + * | step | 24
> + * +----------------+----------------+----------------+----------------+
> + * | def_val | 28
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 36
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * index - uint8_t, index of the camera control in response.
> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
> + * min - int32_t, minimum value of the control.
> + * max - int32_t, maximum value of the control.
> + * step - int32_t, minimum size in which control value can be changed.
> + * def_val - int32_t, default value of the control.
I'd go with 64 bit values for min/max/step/def_val.
I would also add a flags field. Some controls are read-only, write-only
or volatile, things userspace needs to know.
If you want to support menu controls, then you need a way to get the menu
names as well (VIDIOC_QUERYMENU).
None of this is needed for this initial use-case, but you need to think
about this up-front.
> + */
> +struct xencamera_get_ctrl_details_resp {
> + uint8_t index;
> + uint8_t type;
> + uint8_t reserved[2];
> + int32_t min;
> + int32_t max;
> + int32_t step;
> + int32_t def_val;
> +};
> +
> +/*
> + *----------------------------------- Events ----------------------------------
> + *
> + * Events are sent via a shared page allocated by the front and propagated by
> + * evt-event-channel/evt-ring-ref XenStore entries.
> + *
> + * All event packets have the same length (64 octets).
> + * All event packets have common header:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | type | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + *
> + * id - uint16_t, event id, may be used by front.
> + * type - uint8_t, type of the event.
> + *
> + *
> + * Frame captured event - event from back to front when a new captured
> + * frame is available:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_EVT_FRAME_AVAIL| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | used_sz | 16
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 20
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * index - uint8_t, index of the buffer that contains new captured frame.
> + * used_sz - uint32_t, number of octets this frame has. This can be less
> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
> + */
> +struct xencamera_frame_avail_evt {
> + uint8_t index;
> + uint8_t reserved[3];
> + uint32_t used_sz;
> +};
> +
> +struct xencamera_req {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved[5];
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_create_req buf_create;
> + struct xencamera_buf_destroy_req buf_destroy;
> + struct xencamera_set_ctrl_req set_ctrl;
> + uint8_t reserved[56];
> + } req;
> +};
> +
> +struct xencamera_resp {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved;
> + int32_t status;
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_details_resp buf_details;
> + struct xencamera_get_ctrl_details_resp ctrl_details;
> + uint8_t reserved1[56];
> + } resp;
> +};
> +
> +struct xencamera_evt {
> + uint16_t id;
> + uint8_t type;
> + uint8_t reserved[5];
> + union {
> + struct xencamera_frame_avail_evt frame_avail;
> + uint8_t reserved[56];
> + } evt;
> +};
> +
> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
> +
> +/*
> + ******************************************************************************
> + * Back to front events delivery
> + ******************************************************************************
> + * In order to deliver asynchronous events from back to front a shared page is
> + * allocated by front and its granted reference propagated to back via
> + * XenStore entries (evt-ring-ref/evt-event-channel).
> + * This page has a common header used by both front and back to synchronize
> + * access and control event's ring buffer, while back being a producer of the
> + * events and front being a consumer. The rest of the page after the header
> + * is used for event packets.
> + *
> + * Upon reception of an event(s) front may confirm its reception
> + * for either each event, group of events or none.
> + */
> +
> +struct xencamera_event_page {
> + uint32_t in_cons;
> + uint32_t in_prod;
> + uint8_t reserved[56];
> +};
> +
> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
> +#define XENCAMERA_IN_RING(page) \
> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
> +#define XENCAMERA_IN_RING_REF(page, idx) \
> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
> +
> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * tab-width: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
>
I think the most important decision to make here is whether or not you want to support
hotpluggable sources like HDMI. And an additional complication with that is HDCP.
While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
a patch series adding this is expected later this year/early next year. It might not
be an issue in practice if these are all closed systems, but nevertheless, it is
something to think about.
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-09 10:31 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-09 10:31 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, jgross,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi Oleksandr,
Sorry for the delay in reviewing, I missed this patch until you pinged me, and
I was very busy after that as well.
On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>
> This is the ABI for the two halves of a para-virtualized
> camera driver which extends Xen's reach multimedia capabilities even
> farther enabling it for video conferencing, In-Vehicle Infotainment,
> high definition maps etc.
>
> The initial goal is to support most needed functionality with the
> final idea to make it possible to extend the protocol if need be:
>
> 1. Provide means for base virtual device configuration:
> - pixel formats
> - resolutions
> - frame rates
> 2. Support basic camera controls:
> - contrast
> - brightness
> - hue
> - saturation
> 3. Support streaming control
> 4. Support zero-copying use-cases
>
> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
> ---
> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
> 1 file changed, 981 insertions(+)
> create mode 100644 xen/include/public/io/cameraif.h
>
> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
> new file mode 100644
> index 000000000000..bdc6a1262fcf
> --- /dev/null
> +++ b/xen/include/public/io/cameraif.h
> @@ -0,0 +1,981 @@
> +/******************************************************************************
> + * cameraif.h
> + *
> + * Unified camera device I/O interface for Xen guest OSes.
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a copy
> + * of this software and associated documentation files (the "Software"), to
> + * deal in the Software without restriction, including without limitation the
> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
> + * sell copies of the Software, and to permit persons to whom the Software is
> + * furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> + * DEALINGS IN THE SOFTWARE.
> + *
> + * Copyright (C) 2018 EPAM Systems Inc.
> + *
> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
> + */
Use SPDX tag instead of copying the license text.
> +
> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
> +
> +#include "ring.h"
> +#include "../grant_table.h"
> +
> +/*
> + ******************************************************************************
> + * Protocol version
> + ******************************************************************************
> + */
> +#define XENCAMERA_PROTOCOL_VERSION "1"
> +
> +/*
> + ******************************************************************************
> + * Feature and Parameter Negotiation
> + ******************************************************************************
> + *
> + * Front->back notifications: when enqueuing a new request, sending a
> + * notification can be made conditional on xencamera_req (i.e., the generic
> + * hold-off mechanism provided by the ring macros). Backends must set
> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
> + *
> + * Back->front notifications: when enqueuing a new response, sending a
> + * notification can be made conditional on xencamera_resp (i.e., the generic
> + * hold-off mechanism provided by the ring macros). Frontends must set
> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
> + *
> + * The two halves of a para-virtual camera driver utilize nodes within
> + * XenStore to communicate capabilities and to negotiate operating parameters.
> + * This section enumerates these nodes which reside in the respective front and
> + * backend portions of XenStore, following the XenBus convention.
> + *
> + * All data in XenStore is stored as strings. Nodes specifying numeric
> + * values are encoded in decimal. Integer value ranges listed below are
> + * expressed as fixed sized integer types capable of storing the conversion
> + * of a properly formatted node string, without loss of information.
> + *
> + ******************************************************************************
> + * Example configuration
> + ******************************************************************************
> + *
> + * This is an example of backend and frontend configuration:
> + *
> + *--------------------------------- Backend -----------------------------------
> + *
> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
> + * /local/domain/0/backend/vcamera/1/0/state = "4"
> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
superfluous to me.
> + *
> + *--------------------------------- Frontend ----------------------------------
> + *
> + * /local/domain/1/device/vcamera/0/backend-id = "0"
> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
> + * /local/domain/1/device/vcamera/0/state = "4"
> + * /local/domain/1/device/vcamera/0/version = "1"
> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
> + *
> + *---------------------------- Device 0 configuration -------------------------
> + *
> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
> + * /local/domain/1/device/vcamera/0/unique-id = "0"
> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
> + *
> + *---------------------------- Device 1 configuration -------------------------
> + *
> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
> + * /local/domain/1/device/vcamera/1/unique-id = "1"
> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
> + *
> + ******************************************************************************
> + * Backend XenBus Nodes
> + ******************************************************************************
> + *
> + *----------------------------- Protocol version ------------------------------
> + *
> + * versions
> + * Values: <string>
> + *
> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
> + * by the backend. For example "1,2,3".
> + *
> + ******************************************************************************
> + * Frontend XenBus Nodes
> + ******************************************************************************
> + *
> + *-------------------------------- Addressing ---------------------------------
> + *
> + * dom-id
> + * Values: <uint16_t>
> + *
> + * Domain identifier.
> + *
> + * dev-id
> + * Values: <uint16_t>
> + *
> + * Device identifier.
> + *
> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
> + *
> + *----------------------------- Protocol version ------------------------------
> + *
> + * version
> + * Values: <string>
> + *
> + * Protocol version, chosen among the ones supported by the backend.
> + *
> + *------------------------- Backend buffer allocation -------------------------
> + *
> + * be-alloc
I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
is better to just write 'backend-alloc'.
> + * Values: "0", "1"
> + *
> + * If value is set to "1", then backend will be the buffer
> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
> + * operation.
> + * If value is not "1" or omitted frontend must allocate buffers itself.
> + *
> + *------------------------------- Camera settings -----------------------------
> + *
> + * unique-id
> + * Values: <string>
> + *
> + * After device instance initialization each camera is assigned a
> + * unique ID, so it can be identified by the backend by this ID.
> + * This can be UUID or such.
> + *
> + * controls
> + * Values: <list of string>
> + *
> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
> + * Camera controls are expressed as a list of string values w/o any
> + * ordering requirement.
> + *
> + * formats
> + * Values: <format, char[4]>
> + *
> + * Formats are organized as a set of directories one per each
> + * supported pixel format. The name of the directory is an upper case
> + * string of the corresponding FOURCC string label. The next level of
> + * the directory under <formats> represents supported resolutions.
Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
In addition it is common to set bit 31 of the fourcc to 1 if the format is
big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
keep that convention.
> + *
> + * resolution
> + * Values: <width, uint32_t>x<height, uint32_t>
> + *
> + * Resolutions are organized as a set of directories one per each
> + * supported resolution under corresponding <formats> directory.
> + * The name of the directory is the supported width and height
> + * of the camera resolution in pixels.
What if you are dealing with an HDMI input? Not unreasonable for media
systems. There can be a lot of resolutions/framerates, and the resolution
can change on the fly, or of course disappear.
What is also missing here is a way to report pixel aspect ratio: PAL and
NTSC-based video material doesn't have square pixels.
It's important to decide whether or not you want to support video sources
like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
or disconnect webcams) or if you stick to fixed camera pipelines.
The big difference is that you don't control what someone can connect as
external sources, so you will have to be a lot more careful and robust.
I suspect that you likely will want to support such sources eventually, so
it pays to design this with that in mind.
> + *
> + * frame-rates
> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
> + *
> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
> + * of the camera expressed as numerator and denominator of the
> + * corresponding frame rate.
> + *
> + * The format of the <formats> directory tree with resolutions and frame rates
> + * must be structured in the following format:
> + *
> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
> + *
> + * where
> + * i - i-th supported pixel format
> + * j - j-th supported resolution for i-th pixel format
> + * k - k-th supported frame rate for i-th pixel format and j-th
> + * resolution> + *
> + *------------------- Camera Request Transport Parameters ---------------------
> + *
> + * This communication path is used to deliver requests from frontend to backend
> + * and get the corresponding responses from backend to frontend,
> + * set up per virtual camera device.
> + *
> + * req-event-channel
> + * Values: <uint32_t>
> + *
> + * The identifier of the Xen camera's control event channel
> + * used to signal activity in the ring buffer.
> + *
> + * req-ring-ref
> + * Values: <uint32_t>
> + *
> + * The Xen grant reference granting permission for the backend to map
> + * a sole page of camera's control ring buffer.
> + *
> + *-------------------- Camera Event Transport Parameters ----------------------
> + *
> + * This communication path is used to deliver asynchronous events from backend
> + * to frontend, set up per virtual camera device.
> + *
> + * evt-event-channel
> + * Values: <uint32_t>
> + *
> + * The identifier of the Xen camera's event channel
> + * used to signal activity in the ring buffer.
> + *
> + * evt-ring-ref
> + * Values: <uint32_t>
> + *
> + * The Xen grant reference granting permission for the backend to map
> + * a sole page of camera's event ring buffer.
> + */
> +
> +/*
> + ******************************************************************************
> + * STATE DIAGRAMS
> + ******************************************************************************
> + *
> + * Tool stack creates front and back state nodes with initial state
> + * XenbusStateInitialising.
> + * Tool stack creates and sets up frontend camera configuration
> + * nodes per domain.
> + *
> + *-------------------------------- Normal flow --------------------------------
> + *
> + * Front Back
> + * ================================= =====================================
> + * XenbusStateInitialising XenbusStateInitialising
> + * o Query backend device identification
> + * data.
> + * o Open and validate backend device.
> + * |
> + * |
> + * V
> + * XenbusStateInitWait
> + *
> + * o Query frontend configuration
> + * o Allocate and initialize
> + * event channels per configured
> + * camera.
> + * o Publish transport parameters
> + * that will be in effect during
> + * this connection.
> + * |
> + * |
> + * V
> + * XenbusStateInitialised
> + *
> + * o Query frontend transport parameters.
> + * o Connect to the event channels.
> + * |
> + * |
> + * V
> + * XenbusStateConnected
> + *
> + * o Create and initialize OS
> + * virtual camera as per
> + * configuration.
> + * |
> + * |
> + * V
> + * XenbusStateConnected
> + *
> + * XenbusStateUnknown
> + * XenbusStateClosed
> + * XenbusStateClosing
> + * o Remove virtual camera device
> + * o Remove event channels
> + * |
> + * |
> + * V
> + * XenbusStateClosed
> + *
> + *------------------------------- Recovery flow -------------------------------
> + *
> + * In case of frontend unrecoverable errors backend handles that as
> + * if frontend goes into the XenbusStateClosed state.
> + *
> + * In case of backend unrecoverable errors frontend tries removing
> + * the virtualized device. If this is possible at the moment of error,
> + * then frontend goes into the XenbusStateInitialising state and is ready for
> + * new connection with backend. If the virtualized device is still in use and
> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
> + * until either the virtualized device is removed or backend initiates a new
> + * connection. On the virtualized device removal frontend goes into the
> + * XenbusStateInitialising state.
> + *
> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
> + * unrecoverable errors then frontend cannot send requests to the backend
> + * and thus cannot provide functionality of the virtualized device anymore.
> + * After backend is back to normal the virtualized device may still hold some
> + * state: configuration in use, allocated buffers, client application state etc.
> + * In most cases, this will require frontend to implement complex recovery
> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
> + * frontend will make sure no new clients of the virtualized device are
> + * accepted, allow existing client(s) to exit gracefully by signaling error
> + * state etc.
> + * Once all the clients are gone frontend can reinitialize the virtualized
> + * device and get into XenbusStateInitialising state again signaling the
> + * backend that a new connection can be made.
> + *
> + * There are multiple conditions possible under which frontend will go from
> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
> + * specific. For example:
> + * 1. The underlying OS framework may provide callbacks to signal that the last
> + * client of the virtualized device has gone and the device can be removed
> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
> + * to periodically check if this is the right time to re-try removal of
> + * the virtualized device.
> + * 3. By any other means.
> + *
> + ******************************************************************************
> + * REQUEST CODES
> + ******************************************************************************
> + */
> +#define XENCAMERA_OP_SET_CONFIG 0x00
> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
> +#define XENCAMERA_OP_BUF_CREATE 0x02
> +#define XENCAMERA_OP_BUF_DESTROY 0x03
> +#define XENCAMERA_OP_STREAM_START 0x04
> +#define XENCAMERA_OP_STREAM_STOP 0x05
> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
> +#define XENCAMERA_OP_SET_CTRL 0x07
> +
> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
> +#define XENCAMERA_CTRL_CONTRAST 0x01
> +#define XENCAMERA_CTRL_SATURATION 0x02
> +#define XENCAMERA_CTRL_HUE 0x03
> +
> +/*
> + ******************************************************************************
> + * EVENT CODES
> + ******************************************************************************
> + */
> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
> +
> +/*
> + ******************************************************************************
> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
> + ******************************************************************************
> + */
> +#define XENCAMERA_DRIVER_NAME "vcamera"
Ah, that's where vcamera comes from. How about calling this xen-camera or
virt-camera? With a preference for xen-camera, since that's what you use for the
defines as well.
Or perhaps pv-camera?
Is this driver going to be xen-specific, or more a general approach that everyone
can use? Obviously, the latter would be preferable.
BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
radio and RDS hardware. Contact me if this is of interest to Xen to support this as
well given the automotive use-case.
> +
> +#define XENCAMERA_LIST_SEPARATOR ","
> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
> +
> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
> +#define XENCAMERA_FIELD_FE_VERSION "version"
> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
> +#define XENCAMERA_FIELD_CONTROLS "controls"
> +#define XENCAMERA_FIELD_FORMATS "formats"
> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
> +
> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
> +#define XENCAMERA_CTRL_HUE_STR "hue"
> +
> +/* Maximum number of buffer planes supported. */
> +#define XENCAMERA_MAX_PLANE 4
> +
> +/*
> + ******************************************************************************
> + * STATUS RETURN CODES
> + ******************************************************************************
> + *
> + * Status return code is zero on success and -XEN_EXX on failure.
> + *
> + ******************************************************************************
> + * Assumptions
> + ******************************************************************************
> + *
> + * - usage of grant reference 0 as invalid grant reference:
> + * grant reference 0 is valid, but never exposed to a PV driver,
> + * because of the fact it is already in use/reserved by the PV console.
> + * - all references in this document to page sizes must be treated
> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
> + *
> + ******************************************************************************
> + * Description of the protocol between frontend and backend driver
> + ******************************************************************************
> + *
> + * The two halves of a Para-virtual camera driver communicate with
> + * each other using shared pages and event channels.
> + * Shared page contains a ring with request/response packets.
> + *
> + * All reserved fields in the structures below must be 0.
> + *
> + * For all request/response/event packets:
> + * - frame rate parameter is represented as a pair of 4 octet long
> + * numerator and denominator:
> + * - frame_rate_numer - uint32_t, numerator of the frame rate
> + * - frame_rate_denom - uint32_t, denominator of the frame rate
> + * The corresponding frame rate (Hz) is calculated as:
> + * frame_rate = frame_rate_numer / frame_rate_denom
> + * - buffer index is a zero based index of the buffer. Must be less than
> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
> + * - index - uint8_t, index of the buffer.
> + *
> + *
> + *---------------------------------- Requests ---------------------------------
> + *
> + * All request packets have the same length (64 octets).
> + * All request packets have common header:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | operation | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * id - uint16_t, private guest value, echoed in response.
> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
> + *
> + *
> + * Request configuration set/reset - request to set or reset.
> + * the configuration/mode of the camera:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_SET_CONFIG | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | pixel format | 12
> + * +----------------+----------------+----------------+----------------+
> + * | width | 16
> + * +----------------+----------------+----------------+----------------+
> + * | height | 20
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_numer | 24
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_denom | 28
> + * +----------------+----------------+----------------+----------------+
> + * | num_bufs | reserved | 32
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 36
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * Pass all zeros to reset, otherwise command is treated as configuration set.
> + *
> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
> + * width - uint32_t, width in pixels.
> + * height - uint32_t, height in pixels.
> + * frame_rate_numer - uint32_t, numerator of the frame rate.
> + * frame_rate_denom - uint32_t, denominator of the frame rate.
If you have to support HDMI/SDTV inputs as well, then you also need to know
the interlaced format, unless you have no plans to support that.
> + * num_bufs - uint8_t, desired number of buffers to be used.
Huh? What has that to do with the format? Why would you need this here?
> + *
> + * See response format for this request.
> + *
> + * Notes:
> + * - frontend must check the corresponding response in order to see
> + * if the values reported back by the backend do match the desired ones
> + * and can be accepted.
> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
> + * configuration.
> + */
> +struct xencamera_config {
> + uint32_t pixel_format;
> + uint32_t width;
> + uint32_t height;
> + uint32_t frame_rate_nom;
> + uint32_t frame_rate_denom;
> + uint8_t num_bufs;
> +};
> +
> +/*
> + * Request buffer details - request camera buffer's memory layout.
> + * detailed description:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_GET_BUF_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * See response format for this request.
> + *
> + *
> + * Request camera buffer creation:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_BUF_CREATE | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | gref_directory | 16
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 20
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * An attempt to create multiple buffers with the same index is an error.
> + * index can be re-used after destroying the corresponding camera buffer.
> + *
> + * index - uint8_t, index of the buffer to be created.
> + * gref_directory - grant_ref_t, a reference to the first shared page
> + * describing shared buffer references. The size of the buffer is equal to
> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
> + * shared buffer size exceeds what can be addressed by this single page,
> + * then reference to the next shared page must be supplied (see
> + * gref_dir_next_page below).
It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
does.
> + *
> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
> + * allocate the buffer with the parameters provided in this request and page
> + * directory is handled as follows:
> + * Frontend on request:
> + * - allocates pages for the directory (gref_directory,
> + * gref_dir_next_page(s)
> + * - grants permissions for the pages of the directory to the backend
> + * - sets gref_dir_next_page fields
> + * Backend on response:
> + * - grants permissions for the pages of the buffer allocated to
> + * the frontend
> + * - fills in page directory with grant references
> + * (gref[] in struct xencamera_page_directory)
> + */
> +struct xencamera_buf_create_req {
> + uint8_t index;
> + uint8_t reserved[3];
> + grant_ref_t gref_directory;
> +};
> +
> +/*
> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
> + * the request) employs a list of pages, describing all pages of the shared
> + * data buffer:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | gref_dir_next_page | 4
> + * +----------------+----------------+----------------+----------------+
> + * | gref[0] | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | gref[i] | i*4+8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | gref[N - 1] | N*4+8
> + * +----------------+----------------+----------------+----------------+
> + *
> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
> + * page directory. Must be 0 if there are no more pages in the list.
> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
> + * allocated at XENCAMERA_OP_BUF_CREATE.
> + *
> + * Number of grant_ref_t entries in the whole page directory is not
> + * passed, but instead can be calculated as:
> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
> + * XEN_PAGE_SIZE
> + */
> +struct xencamera_page_directory {
> + grant_ref_t gref_dir_next_page;
> + grant_ref_t gref[1]; /* Variable length */
> +};
> +
> +/*
> + * Request buffer destruction - destroy a previously allocated camera buffer:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_BUF_DESTROY| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 16
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * index - uint8_t, index of the buffer to be destroyed.
> + */
There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
of them.
> +
> +struct xencamera_buf_destroy_req {
> + uint8_t index;
> +};
> +
> +/*
> + * Request camera capture stream start:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_OP_STREAM_START| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + *
> + * Request camera capture stream stop:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_OP_STREAM_STOP | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + *
> + * Request camera control details:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |GET_CTRL_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 16
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * See response format for this request.
> + *
> + * index - uint8_t, index of the control to be queried.
> + */
> +struct xencamera_get_ctrl_details_req {
> + uint8_t index;
> +};
> +
> +/*
> + *
> + * Request camera control change:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_SET_CTRL | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | value | 16
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 20
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * See response format for this request.
> + *
> + * index - uint8_t, index of the control.
> + * value - int32_t, new value of the control.
I would recommend using a int64_t as the control value.
Note that there are also controls with a payload (e.g. string controls).
If there is ever interest in adding radio/RDS support, then that will become
an issue.
> + */
> +struct xencamera_set_ctrl_req {
> + uint8_t index;
> + uint8_t reserved[3];
> + int32_t value;
> +};
> +
> +/*
> + *---------------------------------- Responses --------------------------------
> + *
> + * All response packets have the same length (64 octets).
> + *
> + * All response packets have common header:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | operation | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + *
> + * id - uint16_t, copied from the request.
> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
> + *
> + *
> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | _OP_SET_CONFIG | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + * | pixel format | 12
> + * +----------------+----------------+----------------+----------------+
> + * | width | 16
> + * +----------------+----------------+----------------+----------------+
> + * | height | 20
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_numer | 24
> + * +----------------+----------------+----------------+----------------+
> + * | frame_rate_denom | 28
> + * +----------------+----------------+----------------+----------------+
> + * | num_bufs | reserved | 32
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 36
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * Meaning of the corresponding values in this response is the same as for
> + * XENCAMERA_OP_SET_CONFIG request.
> + *
> + *
> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
> + * request:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_GET_BUF_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + * | size | 12
> + * +----------------+----------------+----------------+----------------+
> + * | num_planes | reserved | 16
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[0] | 20
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[1] | 24
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[2] | 28
> + * +----------------+----------------+----------------+----------------+
> + * | plane_offset[3] | 32
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[0] | 36
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[1] | 40
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[2] | 44
> + * +----------------+----------------+----------------+----------------+
> + * | plane_size[3] | 48
> + * +----------------+----------------+----------------+----------------+
> + * | plane_stride[0] | plane_stride[1] | 52
> + * +----------------+----------------+----------------+----------------+
> + * | plane_stride[2] | plane_stride[3] | 56
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 60
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * size - uint32_t, overall size of the buffer including sizes of the
> + * individual planes and padding if applicable.
> + * num_planes - uint8_t, number of planes for this buffer.
> + * plane_offset - array of uint32_t, offset of the corresponding plane
> + * in octets from the buffer start.
> + * plane_size - array of uint32_t, size in octets of the corresponding plane
> + * including padding.
> + * plane_stride - array of uint32_t, size in octets occupied by the
> + * corresponding single image line including padding if applicable.
Nice!
> + */
> +struct xencamera_buf_details_resp {
> + uint32_t size;
> + uint8_t num_planes;
> + uint8_t reserved[3];
> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
> +};
> +
> +/*
> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |GET_CTRL_DETAILS| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | status | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | type | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | min | 16
> + * +----------------+----------------+----------------+----------------+
> + * | max | 20
> + * +----------------+----------------+----------------+----------------+
> + * | step | 24
> + * +----------------+----------------+----------------+----------------+
> + * | def_val | 28
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 36
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * index - uint8_t, index of the camera control in response.
> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
> + * min - int32_t, minimum value of the control.
> + * max - int32_t, maximum value of the control.
> + * step - int32_t, minimum size in which control value can be changed.
> + * def_val - int32_t, default value of the control.
I'd go with 64 bit values for min/max/step/def_val.
I would also add a flags field. Some controls are read-only, write-only
or volatile, things userspace needs to know.
If you want to support menu controls, then you need a way to get the menu
names as well (VIDIOC_QUERYMENU).
None of this is needed for this initial use-case, but you need to think
about this up-front.
> + */
> +struct xencamera_get_ctrl_details_resp {
> + uint8_t index;
> + uint8_t type;
> + uint8_t reserved[2];
> + int32_t min;
> + int32_t max;
> + int32_t step;
> + int32_t def_val;
> +};
> +
> +/*
> + *----------------------------------- Events ----------------------------------
> + *
> + * Events are sent via a shared page allocated by the front and propagated by
> + * evt-event-channel/evt-ring-ref XenStore entries.
> + *
> + * All event packets have the same length (64 octets).
> + * All event packets have common header:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id | type | reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + *
> + * id - uint16_t, event id, may be used by front.
> + * type - uint8_t, type of the event.
> + *
> + *
> + * Frame captured event - event from back to front when a new captured
> + * frame is available:
> + * 0 1 2 3 octet
> + * +----------------+----------------+----------------+----------------+
> + * | id |_EVT_FRAME_AVAIL| reserved | 4
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 8
> + * +----------------+----------------+----------------+----------------+
> + * | index | reserved | 12
> + * +----------------+----------------+----------------+----------------+
> + * | used_sz | 16
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 20
> + * +----------------+----------------+----------------+----------------+
> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
> + * +----------------+----------------+----------------+----------------+
> + * | reserved | 64
> + * +----------------+----------------+----------------+----------------+
> + *
> + * index - uint8_t, index of the buffer that contains new captured frame.
> + * used_sz - uint32_t, number of octets this frame has. This can be less
> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
> + */
> +struct xencamera_frame_avail_evt {
> + uint8_t index;
> + uint8_t reserved[3];
> + uint32_t used_sz;
> +};
> +
> +struct xencamera_req {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved[5];
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_create_req buf_create;
> + struct xencamera_buf_destroy_req buf_destroy;
> + struct xencamera_set_ctrl_req set_ctrl;
> + uint8_t reserved[56];
> + } req;
> +};
> +
> +struct xencamera_resp {
> + uint16_t id;
> + uint8_t operation;
> + uint8_t reserved;
> + int32_t status;
> + union {
> + struct xencamera_config config;
> + struct xencamera_buf_details_resp buf_details;
> + struct xencamera_get_ctrl_details_resp ctrl_details;
> + uint8_t reserved1[56];
> + } resp;
> +};
> +
> +struct xencamera_evt {
> + uint16_t id;
> + uint8_t type;
> + uint8_t reserved[5];
> + union {
> + struct xencamera_frame_avail_evt frame_avail;
> + uint8_t reserved[56];
> + } evt;
> +};
> +
> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
> +
> +/*
> + ******************************************************************************
> + * Back to front events delivery
> + ******************************************************************************
> + * In order to deliver asynchronous events from back to front a shared page is
> + * allocated by front and its granted reference propagated to back via
> + * XenStore entries (evt-ring-ref/evt-event-channel).
> + * This page has a common header used by both front and back to synchronize
> + * access and control event's ring buffer, while back being a producer of the
> + * events and front being a consumer. The rest of the page after the header
> + * is used for event packets.
> + *
> + * Upon reception of an event(s) front may confirm its reception
> + * for either each event, group of events or none.
> + */
> +
> +struct xencamera_event_page {
> + uint32_t in_cons;
> + uint32_t in_prod;
> + uint8_t reserved[56];
> +};
> +
> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
> +#define XENCAMERA_IN_RING(page) \
> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
> +#define XENCAMERA_IN_RING_REF(page, idx) \
> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
> +
> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * tab-width: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
>
I think the most important decision to make here is whether or not you want to support
hotpluggable sources like HDMI. And an additional complication with that is HDCP.
While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
a patch series adding this is expected later this year/early next year. It might not
be an issue in practice if these are all closed systems, but nevertheless, it is
something to think about.
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-04 6:56 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-09 10:42 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-09 10:42 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko, Artem Mygaiev
On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>> Hi Oleksandr,
>>
>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>
>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>> high definition maps etc.
>>>>>>
>>>>>> The initial goal is to support most needed functionality with the
>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>
>>>>>> 1. Provide means for base virtual device configuration:
>>>>>> - pixel formats
>>>>>> - resolutions
>>>>>> - frame rates
>>>>>> 2. Support basic camera controls:
>>>>>> - contrast
>>>>>> - brightness
>>>>>> - hue
>>>>>> - saturation
>>>>>> 3. Support streaming control
>>>>>> 4. Support zero-copying use-cases
>>>>>>
>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>> Some style issues below...
>>>> Will fix all the below, thank you!
>>>>
>>>> I would like to draw some attention of the Linux/V4L community to this
>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>> upstream a Linux camera front-end kernel driver which will be based
>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>> as well)
>>> ping
>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>> to review this this week, if not, just ping me again.
> Thank you for your time
>>
>> I had one high-level question, though:
>>
>> What types of hardware do you intend to target? This initial version targets
>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>> codecs? Or complex embedded video pipelines?
>>
>> In other words, where are you planning to draw the line?
>>
>> Even with just simple cameras there is a difference between regular UVC
>> webcams and cameras used with embedded systems: for the latter you often
>> need to provide more control w.r.t. white-balancing etc., things that a
>> UVC webcam will generally do for you in the webcam's firmware.
> The use-cases we want to implement are mostly in automotive/embedded domain,
> so there are many performance restrictions apply.
> We are not targeting virtualizing very complex hardware and have no
> intention
> to make a 1:1 mapping of the real hardware: for that one can pass-through
> a real HW device to a virtual machine (VM). The goal is to share a single
> camera device to multiple virtual machines, no codecs, receivers etc.
>
> Controlling the same HW device from different VMs doesn't look feasible:
> what if the same control is set to different values from different VMs?
You can do this, actually: in V4L2 you can get an event when another process
changes a control, and update your own GUI/internal state accordingly.
So in this case if one VM changes a control, an event is sent to all others
that the control has changed value.
> Of course, this can be achieved if the corresponding backend can
> post-process
> original camera image with GPU, for example, thus applying different filters
> for different VMs effectively emulating camera controls.
> But this requires additional CPU/GPU power which we try to avoid.
>
> System partitioning (camera and controls assignment) is done at
> configuration
> time (remember we are in automotive/embedded world, so most of the time
> the set
> of VMs requiring cameras is known at this stage and the configuration
> remains
> static at run-time). So, when para-virtualized (PV) approach is used then we
> only implement very basic controls (those found in the protocol), so one can
> assign set of controls (all or some) to one of the VMs (main or mission
> critical
> VM or whatever) allowing that VM to adjusts those for all VMs at once.
> For other
> VMs think of it as firmware implemented adjustment. And the backend still
> controls the rest of the controls of the real HW camera you mention.
>
> Just an example of automotive use-case (we can imagine many more):
> 1. Driver Domain - owns real camera HW and runs the camera backend.
> Uses camera output for mission critical tasks, e.g. parking assistance.
> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
> purposes,
> e.g. taking pictures while in motion.
> 3. Navigation domain - uses PV camera for high definition maps
>
> Hope, this helps understanding the possible uses of the proposed
> protocol, its
> intention and restrictions.
Right, so in this scenario you probably do not want hotpluggable
sources in the Driver Domain. So support for fixed camera's only.
If this is indeed the case, then this should be made very clear in
the API specification.
One additional thing to consider: cameras can break. So what should be
done if that happens? We as media developers have ideas about that, but
nothing has been implemented (yet).
If the HW is simple (one camera is driven by a single driver instance),
then if it breaks, there simply won't be a video device. But if you have
multiple cameras all controlled through the same driver instance, then today
if a single camera breaks, all are gone.
We have ideas on how to address that, but as I said, nothing is implemented
yet. Basically we need to allow for partial bring-up and inform userspace
what is and what is not running.
But this is likely something you also need to consider in this API, given
the use-case you are looking at.
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-09 10:42 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-09 10:42 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Artem Mygaiev, Oleksandr Andrushchenko
On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>> Hi Oleksandr,
>>
>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>
>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>> high definition maps etc.
>>>>>>
>>>>>> The initial goal is to support most needed functionality with the
>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>
>>>>>> 1. Provide means for base virtual device configuration:
>>>>>> - pixel formats
>>>>>> - resolutions
>>>>>> - frame rates
>>>>>> 2. Support basic camera controls:
>>>>>> - contrast
>>>>>> - brightness
>>>>>> - hue
>>>>>> - saturation
>>>>>> 3. Support streaming control
>>>>>> 4. Support zero-copying use-cases
>>>>>>
>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>> Some style issues below...
>>>> Will fix all the below, thank you!
>>>>
>>>> I would like to draw some attention of the Linux/V4L community to this
>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>> upstream a Linux camera front-end kernel driver which will be based
>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>> as well)
>>> ping
>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>> to review this this week, if not, just ping me again.
> Thank you for your time
>>
>> I had one high-level question, though:
>>
>> What types of hardware do you intend to target? This initial version targets
>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>> codecs? Or complex embedded video pipelines?
>>
>> In other words, where are you planning to draw the line?
>>
>> Even with just simple cameras there is a difference between regular UVC
>> webcams and cameras used with embedded systems: for the latter you often
>> need to provide more control w.r.t. white-balancing etc., things that a
>> UVC webcam will generally do for you in the webcam's firmware.
> The use-cases we want to implement are mostly in automotive/embedded domain,
> so there are many performance restrictions apply.
> We are not targeting virtualizing very complex hardware and have no
> intention
> to make a 1:1 mapping of the real hardware: for that one can pass-through
> a real HW device to a virtual machine (VM). The goal is to share a single
> camera device to multiple virtual machines, no codecs, receivers etc.
>
> Controlling the same HW device from different VMs doesn't look feasible:
> what if the same control is set to different values from different VMs?
You can do this, actually: in V4L2 you can get an event when another process
changes a control, and update your own GUI/internal state accordingly.
So in this case if one VM changes a control, an event is sent to all others
that the control has changed value.
> Of course, this can be achieved if the corresponding backend can
> post-process
> original camera image with GPU, for example, thus applying different filters
> for different VMs effectively emulating camera controls.
> But this requires additional CPU/GPU power which we try to avoid.
>
> System partitioning (camera and controls assignment) is done at
> configuration
> time (remember we are in automotive/embedded world, so most of the time
> the set
> of VMs requiring cameras is known at this stage and the configuration
> remains
> static at run-time). So, when para-virtualized (PV) approach is used then we
> only implement very basic controls (those found in the protocol), so one can
> assign set of controls (all or some) to one of the VMs (main or mission
> critical
> VM or whatever) allowing that VM to adjusts those for all VMs at once.
> For other
> VMs think of it as firmware implemented adjustment. And the backend still
> controls the rest of the controls of the real HW camera you mention.
>
> Just an example of automotive use-case (we can imagine many more):
> 1. Driver Domain - owns real camera HW and runs the camera backend.
> Uses camera output for mission critical tasks, e.g. parking assistance.
> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
> purposes,
> e.g. taking pictures while in motion.
> 3. Navigation domain - uses PV camera for high definition maps
>
> Hope, this helps understanding the possible uses of the proposed
> protocol, its
> intention and restrictions.
Right, so in this scenario you probably do not want hotpluggable
sources in the Driver Domain. So support for fixed camera's only.
If this is indeed the case, then this should be made very clear in
the API specification.
One additional thing to consider: cameras can break. So what should be
done if that happens? We as media developers have ideas about that, but
nothing has been implemented (yet).
If the HW is simple (one camera is driven by a single driver instance),
then if it breaks, there simply won't be a video device. But if you have
multiple cameras all controlled through the same driver instance, then today
if a single camera breaks, all are gone.
We have ideas on how to address that, but as I said, nothing is implemented
yet. Basically we need to allow for partial bring-up and inform userspace
what is and what is not running.
But this is likely something you also need to consider in this API, given
the use-case you are looking at.
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-09 10:42 ` [PATCH " Hans Verkuil
@ 2018-09-10 5:59 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 5:59 UTC (permalink / raw)
To: Hans Verkuil, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko, Artem Mygaiev
Hi, Hans!
On 09/09/2018 01:42 PM, Hans Verkuil wrote:
> On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
>> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>>> Hi Oleksandr,
>>>
>>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>>
>>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>>> high definition maps etc.
>>>>>>>
>>>>>>> The initial goal is to support most needed functionality with the
>>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>>
>>>>>>> 1. Provide means for base virtual device configuration:
>>>>>>> - pixel formats
>>>>>>> - resolutions
>>>>>>> - frame rates
>>>>>>> 2. Support basic camera controls:
>>>>>>> - contrast
>>>>>>> - brightness
>>>>>>> - hue
>>>>>>> - saturation
>>>>>>> 3. Support streaming control
>>>>>>> 4. Support zero-copying use-cases
>>>>>>>
>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>>> Some style issues below...
>>>>> Will fix all the below, thank you!
>>>>>
>>>>> I would like to draw some attention of the Linux/V4L community to this
>>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>>> upstream a Linux camera front-end kernel driver which will be based
>>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>>> as well)
>>>> ping
>>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>>> to review this this week, if not, just ping me again.
>> Thank you for your time
>>> I had one high-level question, though:
>>>
>>> What types of hardware do you intend to target? This initial version targets
>>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>>> codecs? Or complex embedded video pipelines?
>>>
>>> In other words, where are you planning to draw the line?
>>>
>>> Even with just simple cameras there is a difference between regular UVC
>>> webcams and cameras used with embedded systems: for the latter you often
>>> need to provide more control w.r.t. white-balancing etc., things that a
>>> UVC webcam will generally do for you in the webcam's firmware.
>> The use-cases we want to implement are mostly in automotive/embedded domain,
>> so there are many performance restrictions apply.
>> We are not targeting virtualizing very complex hardware and have no
>> intention
>> to make a 1:1 mapping of the real hardware: for that one can pass-through
>> a real HW device to a virtual machine (VM). The goal is to share a single
>> camera device to multiple virtual machines, no codecs, receivers etc.
>>
>> Controlling the same HW device from different VMs doesn't look feasible:
>> what if the same control is set to different values from different VMs?
> You can do this, actually: in V4L2 you can get an event when another process
> changes a control, and update your own GUI/internal state accordingly.
>
> So in this case if one VM changes a control, an event is sent to all others
> that the control has changed value.
Well, technically this can be done by introducing one more
event for such a notification. But, from system partitioning
POV, I am still not convinced this should be done: I would prefer
that a single VM owns such a control and even which control and which
VM is decided while configuring the whole system.
So, I would like to keep it as is.
>
>> Of course, this can be achieved if the corresponding backend can
>> post-process
>> original camera image with GPU, for example, thus applying different filters
>> for different VMs effectively emulating camera controls.
>> But this requires additional CPU/GPU power which we try to avoid.
>>
>> System partitioning (camera and controls assignment) is done at
>> configuration
>> time (remember we are in automotive/embedded world, so most of the time
>> the set
>> of VMs requiring cameras is known at this stage and the configuration
>> remains
>> static at run-time). So, when para-virtualized (PV) approach is used then we
>> only implement very basic controls (those found in the protocol), so one can
>> assign set of controls (all or some) to one of the VMs (main or mission
>> critical
>> VM or whatever) allowing that VM to adjusts those for all VMs at once.
>> For other
>> VMs think of it as firmware implemented adjustment. And the backend still
>> controls the rest of the controls of the real HW camera you mention.
>>
>> Just an example of automotive use-case (we can imagine many more):
>> 1. Driver Domain - owns real camera HW and runs the camera backend.
>> Uses camera output for mission critical tasks, e.g. parking assistance.
>> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
>> purposes,
>> e.g. taking pictures while in motion.
>> 3. Navigation domain - uses PV camera for high definition maps
>>
>> Hope, this helps understanding the possible uses of the proposed
>> protocol, its
>> intention and restrictions.
> Right, so in this scenario you probably do not want hotpluggable
> sources in the Driver Domain. So support for fixed camera's only.
Well, some sort of hotplug can already be implemented, please
see [1], [2] as it is done for virtual display: this is
achieved as a response to the backend's state change,
e.g. whenever backend decides to unplug the virtual device
it changes its state accordingly.
>
> If this is indeed the case, then this should be made very clear in
> the API specification.
As I described above this is already assumed by the state
machine of a xenbus_driver
> One additional thing to consider: cameras can break. So what should be
> done if that happens? We as media developers have ideas about that, but
> nothing has been implemented (yet).
>
> If the HW is simple (one camera is driven by a single driver instance),
> then if it breaks, there simply won't be a video device. But if you have
> multiple cameras all controlled through the same driver instance, then today
> if a single camera breaks, all are gone.
Please see above
> We have ideas on how to address that, but as I said, nothing is implemented
> yet. Basically we need to allow for partial bring-up and inform userspace
> what is and what is not running.
>
> But this is likely something you also need to consider in this API, given
> the use-case you are looking at.
>
> Regards,
>
> Hans
Thank you for your valuable comments,
Oleksandr
[1]
https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L721
[2]
https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L582
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 5:59 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 5:59 UTC (permalink / raw)
To: Hans Verkuil, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Artem Mygaiev, Oleksandr Andrushchenko
Hi, Hans!
On 09/09/2018 01:42 PM, Hans Verkuil wrote:
> On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
>> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>>> Hi Oleksandr,
>>>
>>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>>
>>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>>> high definition maps etc.
>>>>>>>
>>>>>>> The initial goal is to support most needed functionality with the
>>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>>
>>>>>>> 1. Provide means for base virtual device configuration:
>>>>>>> - pixel formats
>>>>>>> - resolutions
>>>>>>> - frame rates
>>>>>>> 2. Support basic camera controls:
>>>>>>> - contrast
>>>>>>> - brightness
>>>>>>> - hue
>>>>>>> - saturation
>>>>>>> 3. Support streaming control
>>>>>>> 4. Support zero-copying use-cases
>>>>>>>
>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>>> Some style issues below...
>>>>> Will fix all the below, thank you!
>>>>>
>>>>> I would like to draw some attention of the Linux/V4L community to this
>>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>>> upstream a Linux camera front-end kernel driver which will be based
>>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>>> as well)
>>>> ping
>>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>>> to review this this week, if not, just ping me again.
>> Thank you for your time
>>> I had one high-level question, though:
>>>
>>> What types of hardware do you intend to target? This initial version targets
>>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>>> codecs? Or complex embedded video pipelines?
>>>
>>> In other words, where are you planning to draw the line?
>>>
>>> Even with just simple cameras there is a difference between regular UVC
>>> webcams and cameras used with embedded systems: for the latter you often
>>> need to provide more control w.r.t. white-balancing etc., things that a
>>> UVC webcam will generally do for you in the webcam's firmware.
>> The use-cases we want to implement are mostly in automotive/embedded domain,
>> so there are many performance restrictions apply.
>> We are not targeting virtualizing very complex hardware and have no
>> intention
>> to make a 1:1 mapping of the real hardware: for that one can pass-through
>> a real HW device to a virtual machine (VM). The goal is to share a single
>> camera device to multiple virtual machines, no codecs, receivers etc.
>>
>> Controlling the same HW device from different VMs doesn't look feasible:
>> what if the same control is set to different values from different VMs?
> You can do this, actually: in V4L2 you can get an event when another process
> changes a control, and update your own GUI/internal state accordingly.
>
> So in this case if one VM changes a control, an event is sent to all others
> that the control has changed value.
Well, technically this can be done by introducing one more
event for such a notification. But, from system partitioning
POV, I am still not convinced this should be done: I would prefer
that a single VM owns such a control and even which control and which
VM is decided while configuring the whole system.
So, I would like to keep it as is.
>
>> Of course, this can be achieved if the corresponding backend can
>> post-process
>> original camera image with GPU, for example, thus applying different filters
>> for different VMs effectively emulating camera controls.
>> But this requires additional CPU/GPU power which we try to avoid.
>>
>> System partitioning (camera and controls assignment) is done at
>> configuration
>> time (remember we are in automotive/embedded world, so most of the time
>> the set
>> of VMs requiring cameras is known at this stage and the configuration
>> remains
>> static at run-time). So, when para-virtualized (PV) approach is used then we
>> only implement very basic controls (those found in the protocol), so one can
>> assign set of controls (all or some) to one of the VMs (main or mission
>> critical
>> VM or whatever) allowing that VM to adjusts those for all VMs at once.
>> For other
>> VMs think of it as firmware implemented adjustment. And the backend still
>> controls the rest of the controls of the real HW camera you mention.
>>
>> Just an example of automotive use-case (we can imagine many more):
>> 1. Driver Domain - owns real camera HW and runs the camera backend.
>> Uses camera output for mission critical tasks, e.g. parking assistance.
>> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
>> purposes,
>> e.g. taking pictures while in motion.
>> 3. Navigation domain - uses PV camera for high definition maps
>>
>> Hope, this helps understanding the possible uses of the proposed
>> protocol, its
>> intention and restrictions.
> Right, so in this scenario you probably do not want hotpluggable
> sources in the Driver Domain. So support for fixed camera's only.
Well, some sort of hotplug can already be implemented, please
see [1], [2] as it is done for virtual display: this is
achieved as a response to the backend's state change,
e.g. whenever backend decides to unplug the virtual device
it changes its state accordingly.
>
> If this is indeed the case, then this should be made very clear in
> the API specification.
As I described above this is already assumed by the state
machine of a xenbus_driver
> One additional thing to consider: cameras can break. So what should be
> done if that happens? We as media developers have ideas about that, but
> nothing has been implemented (yet).
>
> If the HW is simple (one camera is driven by a single driver instance),
> then if it breaks, there simply won't be a video device. But if you have
> multiple cameras all controlled through the same driver instance, then today
> if a single camera breaks, all are gone.
Please see above
> We have ideas on how to address that, but as I said, nothing is implemented
> yet. Basically we need to allow for partial bring-up and inform userspace
> what is and what is not running.
>
> But this is likely something you also need to consider in this API, given
> the use-case you are looking at.
>
> Regards,
>
> Hans
Thank you for your valuable comments,
Oleksandr
[1]
https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L721
[2]
https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L582
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-09 10:31 ` [PATCH " Hans Verkuil
@ 2018-09-10 7:16 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 7:16 UTC (permalink / raw)
To: Hans Verkuil, xen-devel, konrad.wilk, jgross, boris.ostrovsky,
mchehab, linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi, Hans!
On 09/09/2018 01:31 PM, Hans Verkuil wrote:
> Hi Oleksandr,
>
> Sorry for the delay in reviewing, I missed this patch until you pinged me, and
> I was very busy after that as well.
I do appreciate you spending time on this!
>
> On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>
>> This is the ABI for the two halves of a para-virtualized
>> camera driver which extends Xen's reach multimedia capabilities even
>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>> high definition maps etc.
>>
>> The initial goal is to support most needed functionality with the
>> final idea to make it possible to extend the protocol if need be:
>>
>> 1. Provide means for base virtual device configuration:
>> - pixel formats
>> - resolutions
>> - frame rates
>> 2. Support basic camera controls:
>> - contrast
>> - brightness
>> - hue
>> - saturation
>> 3. Support streaming control
>> 4. Support zero-copying use-cases
>>
>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>> ---
>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>> 1 file changed, 981 insertions(+)
>> create mode 100644 xen/include/public/io/cameraif.h
>>
>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>> new file mode 100644
>> index 000000000000..bdc6a1262fcf
>> --- /dev/null
>> +++ b/xen/include/public/io/cameraif.h
>> @@ -0,0 +1,981 @@
>> +/******************************************************************************
>> + * cameraif.h
>> + *
>> + * Unified camera device I/O interface for Xen guest OSes.
>> + *
>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>> + * of this software and associated documentation files (the "Software"), to
>> + * deal in the Software without restriction, including without limitation the
>> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
>> + * sell copies of the Software, and to permit persons to whom the Software is
>> + * furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>> + * DEALINGS IN THE SOFTWARE.
>> + *
>> + * Copyright (C) 2018 EPAM Systems Inc.
>> + *
>> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>> + */
> Use SPDX tag instead of copying the license text.
This is yet a Xen header which belongs to Xen project and
all the rest of the protocols have the same license header.
If Xen community decides to use SPDX then I'll definitely follow.
Konrad, do you think this is the right time for such a move?
>
>> +
>> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
>> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
>> +
>> +#include "ring.h"
>> +#include "../grant_table.h"
>> +
>> +/*
>> + ******************************************************************************
>> + * Protocol version
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_PROTOCOL_VERSION "1"
>> +
>> +/*
>> + ******************************************************************************
>> + * Feature and Parameter Negotiation
>> + ******************************************************************************
>> + *
>> + * Front->back notifications: when enqueuing a new request, sending a
>> + * notification can be made conditional on xencamera_req (i.e., the generic
>> + * hold-off mechanism provided by the ring macros). Backends must set
>> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
>> + *
>> + * Back->front notifications: when enqueuing a new response, sending a
>> + * notification can be made conditional on xencamera_resp (i.e., the generic
>> + * hold-off mechanism provided by the ring macros). Frontends must set
>> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
>> + *
>> + * The two halves of a para-virtual camera driver utilize nodes within
>> + * XenStore to communicate capabilities and to negotiate operating parameters.
>> + * This section enumerates these nodes which reside in the respective front and
>> + * backend portions of XenStore, following the XenBus convention.
>> + *
>> + * All data in XenStore is stored as strings. Nodes specifying numeric
>> + * values are encoded in decimal. Integer value ranges listed below are
>> + * expressed as fixed sized integer types capable of storing the conversion
>> + * of a properly formatted node string, without loss of information.
>> + *
>> + ******************************************************************************
>> + * Example configuration
>> + ******************************************************************************
>> + *
>> + * This is an example of backend and frontend configuration:
>> + *
>> + *--------------------------------- Backend -----------------------------------
>> + *
>> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
>> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
>> + * /local/domain/0/backend/vcamera/1/0/state = "4"
>> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
> Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
> superfluous to me.
'v' stands for 'virtual'. I am following Xen convention used
for all other virtual device protocols here.
>> + *
>> + *--------------------------------- Frontend ----------------------------------
>> + *
>> + * /local/domain/1/device/vcamera/0/backend-id = "0"
>> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
>> + * /local/domain/1/device/vcamera/0/state = "4"
>> + * /local/domain/1/device/vcamera/0/version = "1"
>> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
>> + *
>> + *---------------------------- Device 0 configuration -------------------------
>> + *
>> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
>> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
>> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
>> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
>> + * /local/domain/1/device/vcamera/0/unique-id = "0"
>> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
>> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
>> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
>> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
>> + *
>> + *---------------------------- Device 1 configuration -------------------------
>> + *
>> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
>> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
>> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
>> + * /local/domain/1/device/vcamera/1/unique-id = "1"
>> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
>> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
>> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
>> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
>> + *
>> + ******************************************************************************
>> + * Backend XenBus Nodes
>> + ******************************************************************************
>> + *
>> + *----------------------------- Protocol version ------------------------------
>> + *
>> + * versions
>> + * Values: <string>
>> + *
>> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
>> + * by the backend. For example "1,2,3".
>> + *
>> + ******************************************************************************
>> + * Frontend XenBus Nodes
>> + ******************************************************************************
>> + *
>> + *-------------------------------- Addressing ---------------------------------
>> + *
>> + * dom-id
>> + * Values: <uint16_t>
>> + *
>> + * Domain identifier.
>> + *
>> + * dev-id
>> + * Values: <uint16_t>
>> + *
>> + * Device identifier.
>> + *
>> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
>> + *
>> + *----------------------------- Protocol version ------------------------------
>> + *
>> + * version
>> + * Values: <string>
>> + *
>> + * Protocol version, chosen among the ones supported by the backend.
>> + *
>> + *------------------------- Backend buffer allocation -------------------------
>> + *
>> + * be-alloc
> I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
> is better to just write 'backend-alloc'.
Well, 'be' and 'fe' are commonly used in Xen for backend
and frontend, so I'll probably stick to that convention for now.
>
>> + * Values: "0", "1"
>> + *
>> + * If value is set to "1", then backend will be the buffer
>> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
>> + * operation.
>> + * If value is not "1" or omitted frontend must allocate buffers itself.
>> + *
>> + *------------------------------- Camera settings -----------------------------
>> + *
>> + * unique-id
>> + * Values: <string>
>> + *
>> + * After device instance initialization each camera is assigned a
>> + * unique ID, so it can be identified by the backend by this ID.
>> + * This can be UUID or such.
>> + *
>> + * controls
>> + * Values: <list of string>
>> + *
>> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
>> + * Camera controls are expressed as a list of string values w/o any
>> + * ordering requirement.
>> + *
>> + * formats
>> + * Values: <format, char[4]>
>> + *
>> + * Formats are organized as a set of directories one per each
>> + * supported pixel format. The name of the directory is an upper case
>> + * string of the corresponding FOURCC string label. The next level of
>> + * the directory under <formats> represents supported resolutions.
> Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
Ok, no problem - will remove the 'upper case' from the definition
>
> In addition it is common to set bit 31 of the fourcc to 1 if the format is
> big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
> add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
> keep that convention.
I'll think about it, thank you
>
>> + *
>> + * resolution
>> + * Values: <width, uint32_t>x<height, uint32_t>
>> + *
>> + * Resolutions are organized as a set of directories one per each
>> + * supported resolution under corresponding <formats> directory.
>> + * The name of the directory is the supported width and height
>> + * of the camera resolution in pixels.
> What if you are dealing with an HDMI input? Not unreasonable for media
> systems. There can be a lot of resolutions/framerates, and the resolution
> can change on the fly, or of course disappear.
Well, this is a part of the system configuration done
before we actually run the VMs, e.g. at system design
and configuration time. Most of the time you do know which
resolutions, frame rates etc. you want to assign and these
settings remain static for the whole lifetime of that VM.
If you are designing a system which needs these resolutions
to change at run-time then this can be done:
1. Backend changes the state of the frontend to XenbusStateClosed state
2. Xen (xl/libxl) or backend change the configuration in XenStore
3. Backend re-initializes the frontend which reads new configuration
>
> What is also missing here is a way to report pixel aspect ratio: PAL and
> NTSC-based video material doesn't have square pixels.
Hm, indeed, thank you. I'll put this as a fraction under
the corresponding 'resolution':
Now:
/local/domain/2/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
/local/domain/2/device/vcamera/0/formats/YUYV/1280x1024 = "015/2"
Will change to:
/local/domain/2/device/vcamera/0/formats/YUYV/640x480/framerates/ =
"30/1,15/1,15/2"
/local/domain/2/device/vcamera/0/formats/YUYV/640x480/aspectratio = "1/1"
/local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/framerates = "015/2"
/local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/aspectratio =
"59/58"
>
> It's important to decide whether or not you want to support video sources
> like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
> or disconnect webcams) or if you stick to fixed camera pipelines.
I believe that this is all hidden from the frontend by the
backend, so I se nothing we have to put in the protocol
with this respect.
>
> The big difference is that you don't control what someone can connect as
> external sources, so you will have to be a lot more careful and robust.
>
> I suspect that you likely will want to support such sources eventually, so
> it pays to design this with that in mind.
Again, I think that this is the backend to hide these
use-cases from the frontend.
>
>> + *
>> + * frame-rates
>> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
>> + *
>> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
>> + * of the camera expressed as numerator and denominator of the
>> + * corresponding frame rate.
>> + *
>> + * The format of the <formats> directory tree with resolutions and frame rates
>> + * must be structured in the following format:
>> + *
>> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
>> + *
>> + * where
>> + * i - i-th supported pixel format
>> + * j - j-th supported resolution for i-th pixel format
>> + * k - k-th supported frame rate for i-th pixel format and j-th
>> + * resolution> + *
>> + *------------------- Camera Request Transport Parameters ---------------------
>> + *
>> + * This communication path is used to deliver requests from frontend to backend
>> + * and get the corresponding responses from backend to frontend,
>> + * set up per virtual camera device.
>> + *
>> + * req-event-channel
>> + * Values: <uint32_t>
>> + *
>> + * The identifier of the Xen camera's control event channel
>> + * used to signal activity in the ring buffer.
>> + *
>> + * req-ring-ref
>> + * Values: <uint32_t>
>> + *
>> + * The Xen grant reference granting permission for the backend to map
>> + * a sole page of camera's control ring buffer.
>> + *
>> + *-------------------- Camera Event Transport Parameters ----------------------
>> + *
>> + * This communication path is used to deliver asynchronous events from backend
>> + * to frontend, set up per virtual camera device.
>> + *
>> + * evt-event-channel
>> + * Values: <uint32_t>
>> + *
>> + * The identifier of the Xen camera's event channel
>> + * used to signal activity in the ring buffer.
>> + *
>> + * evt-ring-ref
>> + * Values: <uint32_t>
>> + *
>> + * The Xen grant reference granting permission for the backend to map
>> + * a sole page of camera's event ring buffer.
>> + */
>> +
>> +/*
>> + ******************************************************************************
>> + * STATE DIAGRAMS
>> + ******************************************************************************
>> + *
>> + * Tool stack creates front and back state nodes with initial state
>> + * XenbusStateInitialising.
>> + * Tool stack creates and sets up frontend camera configuration
>> + * nodes per domain.
>> + *
>> + *-------------------------------- Normal flow --------------------------------
>> + *
>> + * Front Back
>> + * ================================= =====================================
>> + * XenbusStateInitialising XenbusStateInitialising
>> + * o Query backend device identification
>> + * data.
>> + * o Open and validate backend device.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateInitWait
>> + *
>> + * o Query frontend configuration
>> + * o Allocate and initialize
>> + * event channels per configured
>> + * camera.
>> + * o Publish transport parameters
>> + * that will be in effect during
>> + * this connection.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateInitialised
>> + *
>> + * o Query frontend transport parameters.
>> + * o Connect to the event channels.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateConnected
>> + *
>> + * o Create and initialize OS
>> + * virtual camera as per
>> + * configuration.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateConnected
>> + *
>> + * XenbusStateUnknown
>> + * XenbusStateClosed
>> + * XenbusStateClosing
>> + * o Remove virtual camera device
>> + * o Remove event channels
>> + * |
>> + * |
>> + * V
>> + * XenbusStateClosed
>> + *
>> + *------------------------------- Recovery flow -------------------------------
>> + *
>> + * In case of frontend unrecoverable errors backend handles that as
>> + * if frontend goes into the XenbusStateClosed state.
>> + *
>> + * In case of backend unrecoverable errors frontend tries removing
>> + * the virtualized device. If this is possible at the moment of error,
>> + * then frontend goes into the XenbusStateInitialising state and is ready for
>> + * new connection with backend. If the virtualized device is still in use and
>> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
>> + * until either the virtualized device is removed or backend initiates a new
>> + * connection. On the virtualized device removal frontend goes into the
>> + * XenbusStateInitialising state.
>> + *
>> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
>> + * unrecoverable errors then frontend cannot send requests to the backend
>> + * and thus cannot provide functionality of the virtualized device anymore.
>> + * After backend is back to normal the virtualized device may still hold some
>> + * state: configuration in use, allocated buffers, client application state etc.
>> + * In most cases, this will require frontend to implement complex recovery
>> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
>> + * frontend will make sure no new clients of the virtualized device are
>> + * accepted, allow existing client(s) to exit gracefully by signaling error
>> + * state etc.
>> + * Once all the clients are gone frontend can reinitialize the virtualized
>> + * device and get into XenbusStateInitialising state again signaling the
>> + * backend that a new connection can be made.
>> + *
>> + * There are multiple conditions possible under which frontend will go from
>> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
>> + * specific. For example:
>> + * 1. The underlying OS framework may provide callbacks to signal that the last
>> + * client of the virtualized device has gone and the device can be removed
>> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
>> + * to periodically check if this is the right time to re-try removal of
>> + * the virtualized device.
>> + * 3. By any other means.
>> + *
>> + ******************************************************************************
>> + * REQUEST CODES
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_OP_SET_CONFIG 0x00
>> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
>> +#define XENCAMERA_OP_BUF_CREATE 0x02
>> +#define XENCAMERA_OP_BUF_DESTROY 0x03
>> +#define XENCAMERA_OP_STREAM_START 0x04
>> +#define XENCAMERA_OP_STREAM_STOP 0x05
>> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
>> +#define XENCAMERA_OP_SET_CTRL 0x07
I am thinking about extending the command set a bit as it already
has some flaws, e.g. there is no way for a VM to tell the backend
that the buffer is not in use anymore and can be given back
to the real HW driver, e.g. queue/dequeue in V4L2 terms:
#define XENCAMERA_OP_SET_FORMAT 0x00
- will be used to set format: pixel format, resolution
#define XENCAMERA_OP_SET_FRAME_RATE 0x01
- used to set the frame rate
#define XENCAMERA_OP_BUF_REQUEST 0x02
- asks backend to allocate the given number of buffers,
backend replies with real number of those to be used
#define XENCAMERA_OP_BUF_CREATE 0x03
- create a shared buffer
#define XENCAMERA_OP_BUF_DESTROY 0x04
- destroy a shared buffer
#define XENCAMERA_OP_BUF_QUEUE 0x05
- VM tells the backend that it has access to the shared buffer
and the buffer cannot be sent back to real HW driver
#define XENCAMERA_OP_BUF_DEQUEUE 0x06
- VM tells the backend that the shared buffer is not in use and
can be sent to real HW driver
#define XENCAMERA_OP_CTRL_ENUM 0x07
- get i-th control ranges and settings
#define XENCAMERA_OP_CTRL_GET 0x08
- get control value
#define XENCAMERA_OP_CTRL_SET 0x09
- set control value
#define XENCAMERA_OP_STREAM_START 0x0a
- start streaming
#define XENCAMERA_OP_STREAM_STOP 0x0b
- stop ctreaming
>> +
>> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
>> +#define XENCAMERA_CTRL_CONTRAST 0x01
>> +#define XENCAMERA_CTRL_SATURATION 0x02
>> +#define XENCAMERA_CTRL_HUE 0x03
>> +
>> +/*
>> + ******************************************************************************
>> + * EVENT CODES
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
>> +
>> +/*
>> + ******************************************************************************
>> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_DRIVER_NAME "vcamera"
> Ah, that's where vcamera comes from. How about calling this xen-camera or
> virt-camera? With a preference for xen-camera, since that's what you use for the
> defines as well.
>
> Or perhaps pv-camera?
>
> Is this driver going to be xen-specific, or more a general approach that everyone
> can use? Obviously, the latter would be preferable.
As I have already replied to the cover letter with explanations:
'v' stands for 'virtual' and there is a convention to name the
Xen virtual devices starting with 'v': vif, vkbd etc.
>
> BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
> radio and RDS hardware. Contact me if this is of interest to Xen to support this as
> well given the automotive use-case.
Yes, thank you, but at this stage we are targeting camera only.
Radio can be another topic if time allows ;) And most probably
it will be a dedicated 'vradio' protocol then...
>
>> +
>> +#define XENCAMERA_LIST_SEPARATOR ","
>> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
>> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
>> +
>> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
>> +#define XENCAMERA_FIELD_FE_VERSION "version"
>> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
>> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
>> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
>> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
>> +#define XENCAMERA_FIELD_CONTROLS "controls"
>> +#define XENCAMERA_FIELD_FORMATS "formats"
>> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
>> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
>> +
>> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
>> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
>> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
>> +#define XENCAMERA_CTRL_HUE_STR "hue"
>> +
>> +/* Maximum number of buffer planes supported. */
>> +#define XENCAMERA_MAX_PLANE 4
>> +
>> +/*
>> + ******************************************************************************
>> + * STATUS RETURN CODES
>> + ******************************************************************************
>> + *
>> + * Status return code is zero on success and -XEN_EXX on failure.
>> + *
>> + ******************************************************************************
>> + * Assumptions
>> + ******************************************************************************
>> + *
>> + * - usage of grant reference 0 as invalid grant reference:
>> + * grant reference 0 is valid, but never exposed to a PV driver,
>> + * because of the fact it is already in use/reserved by the PV console.
>> + * - all references in this document to page sizes must be treated
>> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
>> + *
>> + ******************************************************************************
>> + * Description of the protocol between frontend and backend driver
>> + ******************************************************************************
>> + *
>> + * The two halves of a Para-virtual camera driver communicate with
>> + * each other using shared pages and event channels.
>> + * Shared page contains a ring with request/response packets.
>> + *
>> + * All reserved fields in the structures below must be 0.
>> + *
>> + * For all request/response/event packets:
>> + * - frame rate parameter is represented as a pair of 4 octet long
>> + * numerator and denominator:
>> + * - frame_rate_numer - uint32_t, numerator of the frame rate
>> + * - frame_rate_denom - uint32_t, denominator of the frame rate
>> + * The corresponding frame rate (Hz) is calculated as:
>> + * frame_rate = frame_rate_numer / frame_rate_denom
>> + * - buffer index is a zero based index of the buffer. Must be less than
>> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
>> + * - index - uint8_t, index of the buffer.
>> + *
>> + *
>> + *---------------------------------- Requests ---------------------------------
>> + *
>> + * All request packets have the same length (64 octets).
>> + * All request packets have common header:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | operation | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * id - uint16_t, private guest value, echoed in response.
>> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
>> + *
>> + *
>> + * Request configuration set/reset - request to set or reset.
>> + * the configuration/mode of the camera:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_SET_CONFIG | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | pixel format | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | width | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | height | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_numer | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_denom | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | num_bufs | reserved | 32
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * Pass all zeros to reset, otherwise command is treated as configuration set.
>> + *
>> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>> + * width - uint32_t, width in pixels.
>> + * height - uint32_t, height in pixels.
>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
> If you have to support HDMI/SDTV inputs as well, then you also need to know
> the interlaced format, unless you have no plans to support that.
>
>> + * num_bufs - uint8_t, desired number of buffers to be used.
> Huh? What has that to do with the format? Why would you need this here?
Well, the operation name is 'set_config', not 'set_format',
so I thought we can have such a cumulative command assembling
all the parameters of the configuration. But now I am looking at
turning this single 'set_config' command to 3 different commands,
which is more practical and aligned with V4L2 in particular (please
see above in the command set):
1. set format command:
* pixel_format - uint32_t, pixel format to be used, FOURCC code.
* width - uint32_t, width in pixels.
* height - uint32_t, height in pixels.
2. Set frame rate command:
+ * frame_rate_numer - uint32_t, numerator of the frame rate.
+ * frame_rate_denom - uint32_t, denominator of the frame rate.
3. Set/request num bufs:
* num_bufs - uint8_t, desired number of buffers to be used.
>
>> + *
>> + * See response format for this request.
>> + *
>> + * Notes:
>> + * - frontend must check the corresponding response in order to see
>> + * if the values reported back by the backend do match the desired ones
>> + * and can be accepted.
>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>> + * configuration.
>> + */
>> +struct xencamera_config {
>> + uint32_t pixel_format;
>> + uint32_t width;
>> + uint32_t height;
>> + uint32_t frame_rate_nom;
>> + uint32_t frame_rate_denom;
>> + uint8_t num_bufs;
>> +};
>> +
>> +/*
>> + * Request buffer details - request camera buffer's memory layout.
>> + * detailed description:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_GET_BUF_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * See response format for this request.
>> + *
>> + *
>> + * Request camera buffer creation:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_BUF_CREATE | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref_directory | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * An attempt to create multiple buffers with the same index is an error.
>> + * index can be re-used after destroying the corresponding camera buffer.
>> + *
>> + * index - uint8_t, index of the buffer to be created.
>> + * gref_directory - grant_ref_t, a reference to the first shared page
>> + * describing shared buffer references. The size of the buffer is equal to
>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>> + * shared buffer size exceeds what can be addressed by this single page,
>> + * then reference to the next shared page must be supplied (see
>> + * gref_dir_next_page below).
> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
> does.
Well, I still think it is better to have a per buffer interface
in the protocol as it is done for other Xen virtual devices.
So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
what it does internally in the frontend driver
>
>> + *
>> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
>> + * allocate the buffer with the parameters provided in this request and page
>> + * directory is handled as follows:
>> + * Frontend on request:
>> + * - allocates pages for the directory (gref_directory,
>> + * gref_dir_next_page(s)
>> + * - grants permissions for the pages of the directory to the backend
>> + * - sets gref_dir_next_page fields
>> + * Backend on response:
>> + * - grants permissions for the pages of the buffer allocated to
>> + * the frontend
>> + * - fills in page directory with grant references
>> + * (gref[] in struct xencamera_page_directory)
>> + */
>> +struct xencamera_buf_create_req {
>> + uint8_t index;
>> + uint8_t reserved[3];
>> + grant_ref_t gref_directory;
>> +};
>> +
>> +/*
>> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
>> + * the request) employs a list of pages, describing all pages of the shared
>> + * data buffer:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref_dir_next_page | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref[0] | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref[i] | i*4+8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref[N - 1] | N*4+8
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
>> + * page directory. Must be 0 if there are no more pages in the list.
>> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
>> + * allocated at XENCAMERA_OP_BUF_CREATE.
>> + *
>> + * Number of grant_ref_t entries in the whole page directory is not
>> + * passed, but instead can be calculated as:
>> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
>> + * XEN_PAGE_SIZE
>> + */
>> +struct xencamera_page_directory {
>> + grant_ref_t gref_dir_next_page;
>> + grant_ref_t gref[1]; /* Variable length */
>> +};
>> +
>> +/*
>> + * Request buffer destruction - destroy a previously allocated camera buffer:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_BUF_DESTROY| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * index - uint8_t, index of the buffer to be destroyed.
>> + */
> There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
> of them.
This is not specifically related to V4L2, but can be issued
in response to backend's state change etc.
So, we have a pair of commands to create and destroy buffers.
Even more, frontend can be a some-os-based-driver, not V4L2
based. Or even a user-space application if your will.
>
>> +
>> +struct xencamera_buf_destroy_req {
>> + uint8_t index;
>> +};
>> +
>> +/*
>> + * Request camera capture stream start:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_OP_STREAM_START| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + *
>> + * Request camera capture stream stop:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_OP_STREAM_STOP | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + *
>> + * Request camera control details:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |GET_CTRL_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * See response format for this request.
>> + *
>> + * index - uint8_t, index of the control to be queried.
>> + */
>> +struct xencamera_get_ctrl_details_req {
>> + uint8_t index;
>> +};
>> +
>> +/*
>> + *
>> + * Request camera control change:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_SET_CTRL | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | value | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * See response format for this request.
>> + *
>> + * index - uint8_t, index of the control.
>> + * value - int32_t, new value of the control.
> I would recommend using a int64_t as the control value.
Good point, thank you
>
> Note that there are also controls with a payload (e.g. string controls).
Could you please give me an example of such a control?
Do you think such controls can be of use in a VM?
Can we avoid such controls if we target a simple virtual
camera device? If this is for radio use-case, then we'll
have such support in 'vradio' protocol if need be
>
> If there is ever interest in adding radio/RDS support, then that will become
> an issue.
You mean something like station names, ads etc?
>
>> + */
>> +struct xencamera_set_ctrl_req {
>> + uint8_t index;
>> + uint8_t reserved[3];
>> + int32_t value;
>> +};
>> +
>> +/*
>> + *---------------------------------- Responses --------------------------------
>> + *
>> + * All response packets have the same length (64 octets).
>> + *
>> + * All response packets have common header:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | operation | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * id - uint16_t, copied from the request.
>> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
>> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
>> + *
>> + *
>> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_SET_CONFIG | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | pixel format | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | width | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | height | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_numer | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_denom | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | num_bufs | reserved | 32
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * Meaning of the corresponding values in this response is the same as for
>> + * XENCAMERA_OP_SET_CONFIG request.
>> + *
>> + *
>> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
>> + * request:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_GET_BUF_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | size | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | num_planes | reserved | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[0] | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[1] | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[2] | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[3] | 32
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[0] | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[1] | 40
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[2] | 44
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[3] | 48
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_stride[0] | plane_stride[1] | 52
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_stride[2] | plane_stride[3] | 56
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 60
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * size - uint32_t, overall size of the buffer including sizes of the
>> + * individual planes and padding if applicable.
>> + * num_planes - uint8_t, number of planes for this buffer.
>> + * plane_offset - array of uint32_t, offset of the corresponding plane
>> + * in octets from the buffer start.
>> + * plane_size - array of uint32_t, size in octets of the corresponding plane
>> + * including padding.
>> + * plane_stride - array of uint32_t, size in octets occupied by the
>> + * corresponding single image line including padding if applicable.
> Nice!
Thank you
>> + */
>> +struct xencamera_buf_details_resp {
>> + uint32_t size;
>> + uint8_t num_planes;
>> + uint8_t reserved[3];
>> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
>> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
>> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
>> +};
>> +
>> +/*
>> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |GET_CTRL_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | type | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | min | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | max | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | step | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | def_val | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * index - uint8_t, index of the camera control in response.
>> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
>> + * min - int32_t, minimum value of the control.
>> + * max - int32_t, maximum value of the control.
>> + * step - int32_t, minimum size in which control value can be changed.
>> + * def_val - int32_t, default value of the control.
> I'd go with 64 bit values for min/max/step/def_val.
Sure, good idea, thank you
>
> I would also add a flags field. Some controls are read-only, write-only
> or volatile, things userspace needs to know.
Then I'll also add numerical constants for such
>
> If you want to support menu controls, then you need a way to get the menu
> names as well (VIDIOC_QUERYMENU).
>
> None of this is needed for this initial use-case, but you need to think
> about this up-front.
Yes, thank you
>> + */
>> +struct xencamera_get_ctrl_details_resp {
>> + uint8_t index;
>> + uint8_t type;
>> + uint8_t reserved[2];
>> + int32_t min;
>> + int32_t max;
>> + int32_t step;
>> + int32_t def_val;
>> +};
>> +
>> +/*
>> + *----------------------------------- Events ----------------------------------
>> + *
>> + * Events are sent via a shared page allocated by the front and propagated by
>> + * evt-event-channel/evt-ring-ref XenStore entries.
>> + *
>> + * All event packets have the same length (64 octets).
>> + * All event packets have common header:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | type | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * id - uint16_t, event id, may be used by front.
>> + * type - uint8_t, type of the event.
>> + *
>> + *
>> + * Frame captured event - event from back to front when a new captured
>> + * frame is available:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_EVT_FRAME_AVAIL| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | used_sz | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * index - uint8_t, index of the buffer that contains new captured frame.
>> + * used_sz - uint32_t, number of octets this frame has. This can be less
>> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
>> + */
>> +struct xencamera_frame_avail_evt {
>> + uint8_t index;
>> + uint8_t reserved[3];
>> + uint32_t used_sz;
>> +};
>> +
>> +struct xencamera_req {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved[5];
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_create_req buf_create;
>> + struct xencamera_buf_destroy_req buf_destroy;
>> + struct xencamera_set_ctrl_req set_ctrl;
>> + uint8_t reserved[56];
>> + } req;
>> +};
>> +
>> +struct xencamera_resp {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved;
>> + int32_t status;
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_details_resp buf_details;
>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>> + uint8_t reserved1[56];
>> + } resp;
>> +};
>> +
>> +struct xencamera_evt {
>> + uint16_t id;
>> + uint8_t type;
>> + uint8_t reserved[5];
>> + union {
>> + struct xencamera_frame_avail_evt frame_avail;
>> + uint8_t reserved[56];
>> + } evt;
>> +};
>> +
>> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
>> +
>> +/*
>> + ******************************************************************************
>> + * Back to front events delivery
>> + ******************************************************************************
>> + * In order to deliver asynchronous events from back to front a shared page is
>> + * allocated by front and its granted reference propagated to back via
>> + * XenStore entries (evt-ring-ref/evt-event-channel).
>> + * This page has a common header used by both front and back to synchronize
>> + * access and control event's ring buffer, while back being a producer of the
>> + * events and front being a consumer. The rest of the page after the header
>> + * is used for event packets.
>> + *
>> + * Upon reception of an event(s) front may confirm its reception
>> + * for either each event, group of events or none.
>> + */
>> +
>> +struct xencamera_event_page {
>> + uint32_t in_cons;
>> + uint32_t in_prod;
>> + uint8_t reserved[56];
>> +};
>> +
>> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
>> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
>> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
>> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
>> +#define XENCAMERA_IN_RING(page) \
>> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
>> +#define XENCAMERA_IN_RING_REF(page, idx) \
>> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
>> +
>> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
>> +
>> +/*
>> + * Local variables:
>> + * mode: C
>> + * c-file-style: "BSD"
>> + * c-basic-offset: 4
>> + * tab-width: 4
>> + * indent-tabs-mode: nil
>> + * End:
>> + */
>>
> I think the most important decision to make here is whether or not you want to support
> hotpluggable sources like HDMI. And an additional complication with that is HDCP.
> While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
> a patch series adding this is expected later this year/early next year. It might not
> be an issue in practice if these are all closed systems, but nevertheless, it is
> something to think about.
Yes, thank you for raising these questions, it is worth thinking
about such use-cases.
>
> Regards,
>
> Hans
Thank you so much for the comments,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 7:16 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 7:16 UTC (permalink / raw)
To: Hans Verkuil, xen-devel, konrad.wilk, jgross, boris.ostrovsky,
mchehab, linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi, Hans!
On 09/09/2018 01:31 PM, Hans Verkuil wrote:
> Hi Oleksandr,
>
> Sorry for the delay in reviewing, I missed this patch until you pinged me, and
> I was very busy after that as well.
I do appreciate you spending time on this!
>
> On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>
>> This is the ABI for the two halves of a para-virtualized
>> camera driver which extends Xen's reach multimedia capabilities even
>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>> high definition maps etc.
>>
>> The initial goal is to support most needed functionality with the
>> final idea to make it possible to extend the protocol if need be:
>>
>> 1. Provide means for base virtual device configuration:
>> - pixel formats
>> - resolutions
>> - frame rates
>> 2. Support basic camera controls:
>> - contrast
>> - brightness
>> - hue
>> - saturation
>> 3. Support streaming control
>> 4. Support zero-copying use-cases
>>
>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>> ---
>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>> 1 file changed, 981 insertions(+)
>> create mode 100644 xen/include/public/io/cameraif.h
>>
>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>> new file mode 100644
>> index 000000000000..bdc6a1262fcf
>> --- /dev/null
>> +++ b/xen/include/public/io/cameraif.h
>> @@ -0,0 +1,981 @@
>> +/******************************************************************************
>> + * cameraif.h
>> + *
>> + * Unified camera device I/O interface for Xen guest OSes.
>> + *
>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>> + * of this software and associated documentation files (the "Software"), to
>> + * deal in the Software without restriction, including without limitation the
>> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
>> + * sell copies of the Software, and to permit persons to whom the Software is
>> + * furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>> + * DEALINGS IN THE SOFTWARE.
>> + *
>> + * Copyright (C) 2018 EPAM Systems Inc.
>> + *
>> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>> + */
> Use SPDX tag instead of copying the license text.
This is yet a Xen header which belongs to Xen project and
all the rest of the protocols have the same license header.
If Xen community decides to use SPDX then I'll definitely follow.
Konrad, do you think this is the right time for such a move?
>
>> +
>> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
>> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
>> +
>> +#include "ring.h"
>> +#include "../grant_table.h"
>> +
>> +/*
>> + ******************************************************************************
>> + * Protocol version
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_PROTOCOL_VERSION "1"
>> +
>> +/*
>> + ******************************************************************************
>> + * Feature and Parameter Negotiation
>> + ******************************************************************************
>> + *
>> + * Front->back notifications: when enqueuing a new request, sending a
>> + * notification can be made conditional on xencamera_req (i.e., the generic
>> + * hold-off mechanism provided by the ring macros). Backends must set
>> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
>> + *
>> + * Back->front notifications: when enqueuing a new response, sending a
>> + * notification can be made conditional on xencamera_resp (i.e., the generic
>> + * hold-off mechanism provided by the ring macros). Frontends must set
>> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
>> + *
>> + * The two halves of a para-virtual camera driver utilize nodes within
>> + * XenStore to communicate capabilities and to negotiate operating parameters.
>> + * This section enumerates these nodes which reside in the respective front and
>> + * backend portions of XenStore, following the XenBus convention.
>> + *
>> + * All data in XenStore is stored as strings. Nodes specifying numeric
>> + * values are encoded in decimal. Integer value ranges listed below are
>> + * expressed as fixed sized integer types capable of storing the conversion
>> + * of a properly formatted node string, without loss of information.
>> + *
>> + ******************************************************************************
>> + * Example configuration
>> + ******************************************************************************
>> + *
>> + * This is an example of backend and frontend configuration:
>> + *
>> + *--------------------------------- Backend -----------------------------------
>> + *
>> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
>> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
>> + * /local/domain/0/backend/vcamera/1/0/state = "4"
>> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
> Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
> superfluous to me.
'v' stands for 'virtual'. I am following Xen convention used
for all other virtual device protocols here.
>> + *
>> + *--------------------------------- Frontend ----------------------------------
>> + *
>> + * /local/domain/1/device/vcamera/0/backend-id = "0"
>> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
>> + * /local/domain/1/device/vcamera/0/state = "4"
>> + * /local/domain/1/device/vcamera/0/version = "1"
>> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
>> + *
>> + *---------------------------- Device 0 configuration -------------------------
>> + *
>> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
>> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
>> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
>> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
>> + * /local/domain/1/device/vcamera/0/unique-id = "0"
>> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
>> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
>> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
>> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
>> + *
>> + *---------------------------- Device 1 configuration -------------------------
>> + *
>> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
>> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
>> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
>> + * /local/domain/1/device/vcamera/1/unique-id = "1"
>> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
>> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
>> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
>> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
>> + *
>> + ******************************************************************************
>> + * Backend XenBus Nodes
>> + ******************************************************************************
>> + *
>> + *----------------------------- Protocol version ------------------------------
>> + *
>> + * versions
>> + * Values: <string>
>> + *
>> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
>> + * by the backend. For example "1,2,3".
>> + *
>> + ******************************************************************************
>> + * Frontend XenBus Nodes
>> + ******************************************************************************
>> + *
>> + *-------------------------------- Addressing ---------------------------------
>> + *
>> + * dom-id
>> + * Values: <uint16_t>
>> + *
>> + * Domain identifier.
>> + *
>> + * dev-id
>> + * Values: <uint16_t>
>> + *
>> + * Device identifier.
>> + *
>> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
>> + *
>> + *----------------------------- Protocol version ------------------------------
>> + *
>> + * version
>> + * Values: <string>
>> + *
>> + * Protocol version, chosen among the ones supported by the backend.
>> + *
>> + *------------------------- Backend buffer allocation -------------------------
>> + *
>> + * be-alloc
> I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
> is better to just write 'backend-alloc'.
Well, 'be' and 'fe' are commonly used in Xen for backend
and frontend, so I'll probably stick to that convention for now.
>
>> + * Values: "0", "1"
>> + *
>> + * If value is set to "1", then backend will be the buffer
>> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
>> + * operation.
>> + * If value is not "1" or omitted frontend must allocate buffers itself.
>> + *
>> + *------------------------------- Camera settings -----------------------------
>> + *
>> + * unique-id
>> + * Values: <string>
>> + *
>> + * After device instance initialization each camera is assigned a
>> + * unique ID, so it can be identified by the backend by this ID.
>> + * This can be UUID or such.
>> + *
>> + * controls
>> + * Values: <list of string>
>> + *
>> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
>> + * Camera controls are expressed as a list of string values w/o any
>> + * ordering requirement.
>> + *
>> + * formats
>> + * Values: <format, char[4]>
>> + *
>> + * Formats are organized as a set of directories one per each
>> + * supported pixel format. The name of the directory is an upper case
>> + * string of the corresponding FOURCC string label. The next level of
>> + * the directory under <formats> represents supported resolutions.
> Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
Ok, no problem - will remove the 'upper case' from the definition
>
> In addition it is common to set bit 31 of the fourcc to 1 if the format is
> big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
> add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
> keep that convention.
I'll think about it, thank you
>
>> + *
>> + * resolution
>> + * Values: <width, uint32_t>x<height, uint32_t>
>> + *
>> + * Resolutions are organized as a set of directories one per each
>> + * supported resolution under corresponding <formats> directory.
>> + * The name of the directory is the supported width and height
>> + * of the camera resolution in pixels.
> What if you are dealing with an HDMI input? Not unreasonable for media
> systems. There can be a lot of resolutions/framerates, and the resolution
> can change on the fly, or of course disappear.
Well, this is a part of the system configuration done
before we actually run the VMs, e.g. at system design
and configuration time. Most of the time you do know which
resolutions, frame rates etc. you want to assign and these
settings remain static for the whole lifetime of that VM.
If you are designing a system which needs these resolutions
to change at run-time then this can be done:
1. Backend changes the state of the frontend to XenbusStateClosed state
2. Xen (xl/libxl) or backend change the configuration in XenStore
3. Backend re-initializes the frontend which reads new configuration
>
> What is also missing here is a way to report pixel aspect ratio: PAL and
> NTSC-based video material doesn't have square pixels.
Hm, indeed, thank you. I'll put this as a fraction under
the corresponding 'resolution':
Now:
/local/domain/2/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
/local/domain/2/device/vcamera/0/formats/YUYV/1280x1024 = "015/2"
Will change to:
/local/domain/2/device/vcamera/0/formats/YUYV/640x480/framerates/ =
"30/1,15/1,15/2"
/local/domain/2/device/vcamera/0/formats/YUYV/640x480/aspectratio = "1/1"
/local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/framerates = "015/2"
/local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/aspectratio =
"59/58"
>
> It's important to decide whether or not you want to support video sources
> like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
> or disconnect webcams) or if you stick to fixed camera pipelines.
I believe that this is all hidden from the frontend by the
backend, so I se nothing we have to put in the protocol
with this respect.
>
> The big difference is that you don't control what someone can connect as
> external sources, so you will have to be a lot more careful and robust.
>
> I suspect that you likely will want to support such sources eventually, so
> it pays to design this with that in mind.
Again, I think that this is the backend to hide these
use-cases from the frontend.
>
>> + *
>> + * frame-rates
>> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
>> + *
>> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
>> + * of the camera expressed as numerator and denominator of the
>> + * corresponding frame rate.
>> + *
>> + * The format of the <formats> directory tree with resolutions and frame rates
>> + * must be structured in the following format:
>> + *
>> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
>> + *
>> + * where
>> + * i - i-th supported pixel format
>> + * j - j-th supported resolution for i-th pixel format
>> + * k - k-th supported frame rate for i-th pixel format and j-th
>> + * resolution> + *
>> + *------------------- Camera Request Transport Parameters ---------------------
>> + *
>> + * This communication path is used to deliver requests from frontend to backend
>> + * and get the corresponding responses from backend to frontend,
>> + * set up per virtual camera device.
>> + *
>> + * req-event-channel
>> + * Values: <uint32_t>
>> + *
>> + * The identifier of the Xen camera's control event channel
>> + * used to signal activity in the ring buffer.
>> + *
>> + * req-ring-ref
>> + * Values: <uint32_t>
>> + *
>> + * The Xen grant reference granting permission for the backend to map
>> + * a sole page of camera's control ring buffer.
>> + *
>> + *-------------------- Camera Event Transport Parameters ----------------------
>> + *
>> + * This communication path is used to deliver asynchronous events from backend
>> + * to frontend, set up per virtual camera device.
>> + *
>> + * evt-event-channel
>> + * Values: <uint32_t>
>> + *
>> + * The identifier of the Xen camera's event channel
>> + * used to signal activity in the ring buffer.
>> + *
>> + * evt-ring-ref
>> + * Values: <uint32_t>
>> + *
>> + * The Xen grant reference granting permission for the backend to map
>> + * a sole page of camera's event ring buffer.
>> + */
>> +
>> +/*
>> + ******************************************************************************
>> + * STATE DIAGRAMS
>> + ******************************************************************************
>> + *
>> + * Tool stack creates front and back state nodes with initial state
>> + * XenbusStateInitialising.
>> + * Tool stack creates and sets up frontend camera configuration
>> + * nodes per domain.
>> + *
>> + *-------------------------------- Normal flow --------------------------------
>> + *
>> + * Front Back
>> + * ================================= =====================================
>> + * XenbusStateInitialising XenbusStateInitialising
>> + * o Query backend device identification
>> + * data.
>> + * o Open and validate backend device.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateInitWait
>> + *
>> + * o Query frontend configuration
>> + * o Allocate and initialize
>> + * event channels per configured
>> + * camera.
>> + * o Publish transport parameters
>> + * that will be in effect during
>> + * this connection.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateInitialised
>> + *
>> + * o Query frontend transport parameters.
>> + * o Connect to the event channels.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateConnected
>> + *
>> + * o Create and initialize OS
>> + * virtual camera as per
>> + * configuration.
>> + * |
>> + * |
>> + * V
>> + * XenbusStateConnected
>> + *
>> + * XenbusStateUnknown
>> + * XenbusStateClosed
>> + * XenbusStateClosing
>> + * o Remove virtual camera device
>> + * o Remove event channels
>> + * |
>> + * |
>> + * V
>> + * XenbusStateClosed
>> + *
>> + *------------------------------- Recovery flow -------------------------------
>> + *
>> + * In case of frontend unrecoverable errors backend handles that as
>> + * if frontend goes into the XenbusStateClosed state.
>> + *
>> + * In case of backend unrecoverable errors frontend tries removing
>> + * the virtualized device. If this is possible at the moment of error,
>> + * then frontend goes into the XenbusStateInitialising state and is ready for
>> + * new connection with backend. If the virtualized device is still in use and
>> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
>> + * until either the virtualized device is removed or backend initiates a new
>> + * connection. On the virtualized device removal frontend goes into the
>> + * XenbusStateInitialising state.
>> + *
>> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
>> + * unrecoverable errors then frontend cannot send requests to the backend
>> + * and thus cannot provide functionality of the virtualized device anymore.
>> + * After backend is back to normal the virtualized device may still hold some
>> + * state: configuration in use, allocated buffers, client application state etc.
>> + * In most cases, this will require frontend to implement complex recovery
>> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
>> + * frontend will make sure no new clients of the virtualized device are
>> + * accepted, allow existing client(s) to exit gracefully by signaling error
>> + * state etc.
>> + * Once all the clients are gone frontend can reinitialize the virtualized
>> + * device and get into XenbusStateInitialising state again signaling the
>> + * backend that a new connection can be made.
>> + *
>> + * There are multiple conditions possible under which frontend will go from
>> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
>> + * specific. For example:
>> + * 1. The underlying OS framework may provide callbacks to signal that the last
>> + * client of the virtualized device has gone and the device can be removed
>> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
>> + * to periodically check if this is the right time to re-try removal of
>> + * the virtualized device.
>> + * 3. By any other means.
>> + *
>> + ******************************************************************************
>> + * REQUEST CODES
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_OP_SET_CONFIG 0x00
>> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
>> +#define XENCAMERA_OP_BUF_CREATE 0x02
>> +#define XENCAMERA_OP_BUF_DESTROY 0x03
>> +#define XENCAMERA_OP_STREAM_START 0x04
>> +#define XENCAMERA_OP_STREAM_STOP 0x05
>> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
>> +#define XENCAMERA_OP_SET_CTRL 0x07
I am thinking about extending the command set a bit as it already
has some flaws, e.g. there is no way for a VM to tell the backend
that the buffer is not in use anymore and can be given back
to the real HW driver, e.g. queue/dequeue in V4L2 terms:
#define XENCAMERA_OP_SET_FORMAT 0x00
- will be used to set format: pixel format, resolution
#define XENCAMERA_OP_SET_FRAME_RATE 0x01
- used to set the frame rate
#define XENCAMERA_OP_BUF_REQUEST 0x02
- asks backend to allocate the given number of buffers,
backend replies with real number of those to be used
#define XENCAMERA_OP_BUF_CREATE 0x03
- create a shared buffer
#define XENCAMERA_OP_BUF_DESTROY 0x04
- destroy a shared buffer
#define XENCAMERA_OP_BUF_QUEUE 0x05
- VM tells the backend that it has access to the shared buffer
and the buffer cannot be sent back to real HW driver
#define XENCAMERA_OP_BUF_DEQUEUE 0x06
- VM tells the backend that the shared buffer is not in use and
can be sent to real HW driver
#define XENCAMERA_OP_CTRL_ENUM 0x07
- get i-th control ranges and settings
#define XENCAMERA_OP_CTRL_GET 0x08
- get control value
#define XENCAMERA_OP_CTRL_SET 0x09
- set control value
#define XENCAMERA_OP_STREAM_START 0x0a
- start streaming
#define XENCAMERA_OP_STREAM_STOP 0x0b
- stop ctreaming
>> +
>> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
>> +#define XENCAMERA_CTRL_CONTRAST 0x01
>> +#define XENCAMERA_CTRL_SATURATION 0x02
>> +#define XENCAMERA_CTRL_HUE 0x03
>> +
>> +/*
>> + ******************************************************************************
>> + * EVENT CODES
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
>> +
>> +/*
>> + ******************************************************************************
>> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
>> + ******************************************************************************
>> + */
>> +#define XENCAMERA_DRIVER_NAME "vcamera"
> Ah, that's where vcamera comes from. How about calling this xen-camera or
> virt-camera? With a preference for xen-camera, since that's what you use for the
> defines as well.
>
> Or perhaps pv-camera?
>
> Is this driver going to be xen-specific, or more a general approach that everyone
> can use? Obviously, the latter would be preferable.
As I have already replied to the cover letter with explanations:
'v' stands for 'virtual' and there is a convention to name the
Xen virtual devices starting with 'v': vif, vkbd etc.
>
> BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
> radio and RDS hardware. Contact me if this is of interest to Xen to support this as
> well given the automotive use-case.
Yes, thank you, but at this stage we are targeting camera only.
Radio can be another topic if time allows ;) And most probably
it will be a dedicated 'vradio' protocol then...
>
>> +
>> +#define XENCAMERA_LIST_SEPARATOR ","
>> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
>> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
>> +
>> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
>> +#define XENCAMERA_FIELD_FE_VERSION "version"
>> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
>> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
>> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
>> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
>> +#define XENCAMERA_FIELD_CONTROLS "controls"
>> +#define XENCAMERA_FIELD_FORMATS "formats"
>> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
>> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
>> +
>> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
>> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
>> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
>> +#define XENCAMERA_CTRL_HUE_STR "hue"
>> +
>> +/* Maximum number of buffer planes supported. */
>> +#define XENCAMERA_MAX_PLANE 4
>> +
>> +/*
>> + ******************************************************************************
>> + * STATUS RETURN CODES
>> + ******************************************************************************
>> + *
>> + * Status return code is zero on success and -XEN_EXX on failure.
>> + *
>> + ******************************************************************************
>> + * Assumptions
>> + ******************************************************************************
>> + *
>> + * - usage of grant reference 0 as invalid grant reference:
>> + * grant reference 0 is valid, but never exposed to a PV driver,
>> + * because of the fact it is already in use/reserved by the PV console.
>> + * - all references in this document to page sizes must be treated
>> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
>> + *
>> + ******************************************************************************
>> + * Description of the protocol between frontend and backend driver
>> + ******************************************************************************
>> + *
>> + * The two halves of a Para-virtual camera driver communicate with
>> + * each other using shared pages and event channels.
>> + * Shared page contains a ring with request/response packets.
>> + *
>> + * All reserved fields in the structures below must be 0.
>> + *
>> + * For all request/response/event packets:
>> + * - frame rate parameter is represented as a pair of 4 octet long
>> + * numerator and denominator:
>> + * - frame_rate_numer - uint32_t, numerator of the frame rate
>> + * - frame_rate_denom - uint32_t, denominator of the frame rate
>> + * The corresponding frame rate (Hz) is calculated as:
>> + * frame_rate = frame_rate_numer / frame_rate_denom
>> + * - buffer index is a zero based index of the buffer. Must be less than
>> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
>> + * - index - uint8_t, index of the buffer.
>> + *
>> + *
>> + *---------------------------------- Requests ---------------------------------
>> + *
>> + * All request packets have the same length (64 octets).
>> + * All request packets have common header:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | operation | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * id - uint16_t, private guest value, echoed in response.
>> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
>> + *
>> + *
>> + * Request configuration set/reset - request to set or reset.
>> + * the configuration/mode of the camera:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_SET_CONFIG | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | pixel format | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | width | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | height | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_numer | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_denom | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | num_bufs | reserved | 32
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * Pass all zeros to reset, otherwise command is treated as configuration set.
>> + *
>> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>> + * width - uint32_t, width in pixels.
>> + * height - uint32_t, height in pixels.
>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
> If you have to support HDMI/SDTV inputs as well, then you also need to know
> the interlaced format, unless you have no plans to support that.
>
>> + * num_bufs - uint8_t, desired number of buffers to be used.
> Huh? What has that to do with the format? Why would you need this here?
Well, the operation name is 'set_config', not 'set_format',
so I thought we can have such a cumulative command assembling
all the parameters of the configuration. But now I am looking at
turning this single 'set_config' command to 3 different commands,
which is more practical and aligned with V4L2 in particular (please
see above in the command set):
1. set format command:
* pixel_format - uint32_t, pixel format to be used, FOURCC code.
* width - uint32_t, width in pixels.
* height - uint32_t, height in pixels.
2. Set frame rate command:
+ * frame_rate_numer - uint32_t, numerator of the frame rate.
+ * frame_rate_denom - uint32_t, denominator of the frame rate.
3. Set/request num bufs:
* num_bufs - uint8_t, desired number of buffers to be used.
>
>> + *
>> + * See response format for this request.
>> + *
>> + * Notes:
>> + * - frontend must check the corresponding response in order to see
>> + * if the values reported back by the backend do match the desired ones
>> + * and can be accepted.
>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>> + * configuration.
>> + */
>> +struct xencamera_config {
>> + uint32_t pixel_format;
>> + uint32_t width;
>> + uint32_t height;
>> + uint32_t frame_rate_nom;
>> + uint32_t frame_rate_denom;
>> + uint8_t num_bufs;
>> +};
>> +
>> +/*
>> + * Request buffer details - request camera buffer's memory layout.
>> + * detailed description:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_GET_BUF_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * See response format for this request.
>> + *
>> + *
>> + * Request camera buffer creation:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_BUF_CREATE | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref_directory | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * An attempt to create multiple buffers with the same index is an error.
>> + * index can be re-used after destroying the corresponding camera buffer.
>> + *
>> + * index - uint8_t, index of the buffer to be created.
>> + * gref_directory - grant_ref_t, a reference to the first shared page
>> + * describing shared buffer references. The size of the buffer is equal to
>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>> + * shared buffer size exceeds what can be addressed by this single page,
>> + * then reference to the next shared page must be supplied (see
>> + * gref_dir_next_page below).
> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
> does.
Well, I still think it is better to have a per buffer interface
in the protocol as it is done for other Xen virtual devices.
So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
what it does internally in the frontend driver
>
>> + *
>> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
>> + * allocate the buffer with the parameters provided in this request and page
>> + * directory is handled as follows:
>> + * Frontend on request:
>> + * - allocates pages for the directory (gref_directory,
>> + * gref_dir_next_page(s)
>> + * - grants permissions for the pages of the directory to the backend
>> + * - sets gref_dir_next_page fields
>> + * Backend on response:
>> + * - grants permissions for the pages of the buffer allocated to
>> + * the frontend
>> + * - fills in page directory with grant references
>> + * (gref[] in struct xencamera_page_directory)
>> + */
>> +struct xencamera_buf_create_req {
>> + uint8_t index;
>> + uint8_t reserved[3];
>> + grant_ref_t gref_directory;
>> +};
>> +
>> +/*
>> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
>> + * the request) employs a list of pages, describing all pages of the shared
>> + * data buffer:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref_dir_next_page | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref[0] | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref[i] | i*4+8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | gref[N - 1] | N*4+8
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
>> + * page directory. Must be 0 if there are no more pages in the list.
>> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
>> + * allocated at XENCAMERA_OP_BUF_CREATE.
>> + *
>> + * Number of grant_ref_t entries in the whole page directory is not
>> + * passed, but instead can be calculated as:
>> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
>> + * XEN_PAGE_SIZE
>> + */
>> +struct xencamera_page_directory {
>> + grant_ref_t gref_dir_next_page;
>> + grant_ref_t gref[1]; /* Variable length */
>> +};
>> +
>> +/*
>> + * Request buffer destruction - destroy a previously allocated camera buffer:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_BUF_DESTROY| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * index - uint8_t, index of the buffer to be destroyed.
>> + */
> There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
> of them.
This is not specifically related to V4L2, but can be issued
in response to backend's state change etc.
So, we have a pair of commands to create and destroy buffers.
Even more, frontend can be a some-os-based-driver, not V4L2
based. Or even a user-space application if your will.
>
>> +
>> +struct xencamera_buf_destroy_req {
>> + uint8_t index;
>> +};
>> +
>> +/*
>> + * Request camera capture stream start:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_OP_STREAM_START| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + *
>> + * Request camera capture stream stop:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_OP_STREAM_STOP | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + *
>> + * Request camera control details:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |GET_CTRL_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * See response format for this request.
>> + *
>> + * index - uint8_t, index of the control to be queried.
>> + */
>> +struct xencamera_get_ctrl_details_req {
>> + uint8_t index;
>> +};
>> +
>> +/*
>> + *
>> + * Request camera control change:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_SET_CTRL | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | value | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * See response format for this request.
>> + *
>> + * index - uint8_t, index of the control.
>> + * value - int32_t, new value of the control.
> I would recommend using a int64_t as the control value.
Good point, thank you
>
> Note that there are also controls with a payload (e.g. string controls).
Could you please give me an example of such a control?
Do you think such controls can be of use in a VM?
Can we avoid such controls if we target a simple virtual
camera device? If this is for radio use-case, then we'll
have such support in 'vradio' protocol if need be
>
> If there is ever interest in adding radio/RDS support, then that will become
> an issue.
You mean something like station names, ads etc?
>
>> + */
>> +struct xencamera_set_ctrl_req {
>> + uint8_t index;
>> + uint8_t reserved[3];
>> + int32_t value;
>> +};
>> +
>> +/*
>> + *---------------------------------- Responses --------------------------------
>> + *
>> + * All response packets have the same length (64 octets).
>> + *
>> + * All response packets have common header:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | operation | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * id - uint16_t, copied from the request.
>> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
>> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
>> + *
>> + *
>> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | _OP_SET_CONFIG | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | pixel format | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | width | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | height | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_numer | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | frame_rate_denom | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | num_bufs | reserved | 32
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * Meaning of the corresponding values in this response is the same as for
>> + * XENCAMERA_OP_SET_CONFIG request.
>> + *
>> + *
>> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
>> + * request:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_GET_BUF_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | size | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | num_planes | reserved | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[0] | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[1] | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[2] | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_offset[3] | 32
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[0] | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[1] | 40
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[2] | 44
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_size[3] | 48
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_stride[0] | plane_stride[1] | 52
>> + * +----------------+----------------+----------------+----------------+
>> + * | plane_stride[2] | plane_stride[3] | 56
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 60
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * size - uint32_t, overall size of the buffer including sizes of the
>> + * individual planes and padding if applicable.
>> + * num_planes - uint8_t, number of planes for this buffer.
>> + * plane_offset - array of uint32_t, offset of the corresponding plane
>> + * in octets from the buffer start.
>> + * plane_size - array of uint32_t, size in octets of the corresponding plane
>> + * including padding.
>> + * plane_stride - array of uint32_t, size in octets occupied by the
>> + * corresponding single image line including padding if applicable.
> Nice!
Thank you
>> + */
>> +struct xencamera_buf_details_resp {
>> + uint32_t size;
>> + uint8_t num_planes;
>> + uint8_t reserved[3];
>> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
>> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
>> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
>> +};
>> +
>> +/*
>> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |GET_CTRL_DETAILS| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | status | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | type | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | min | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | max | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * | step | 24
>> + * +----------------+----------------+----------------+----------------+
>> + * | def_val | 28
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 36
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * index - uint8_t, index of the camera control in response.
>> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
>> + * min - int32_t, minimum value of the control.
>> + * max - int32_t, maximum value of the control.
>> + * step - int32_t, minimum size in which control value can be changed.
>> + * def_val - int32_t, default value of the control.
> I'd go with 64 bit values for min/max/step/def_val.
Sure, good idea, thank you
>
> I would also add a flags field. Some controls are read-only, write-only
> or volatile, things userspace needs to know.
Then I'll also add numerical constants for such
>
> If you want to support menu controls, then you need a way to get the menu
> names as well (VIDIOC_QUERYMENU).
>
> None of this is needed for this initial use-case, but you need to think
> about this up-front.
Yes, thank you
>> + */
>> +struct xencamera_get_ctrl_details_resp {
>> + uint8_t index;
>> + uint8_t type;
>> + uint8_t reserved[2];
>> + int32_t min;
>> + int32_t max;
>> + int32_t step;
>> + int32_t def_val;
>> +};
>> +
>> +/*
>> + *----------------------------------- Events ----------------------------------
>> + *
>> + * Events are sent via a shared page allocated by the front and propagated by
>> + * evt-event-channel/evt-ring-ref XenStore entries.
>> + *
>> + * All event packets have the same length (64 octets).
>> + * All event packets have common header:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id | type | reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * id - uint16_t, event id, may be used by front.
>> + * type - uint8_t, type of the event.
>> + *
>> + *
>> + * Frame captured event - event from back to front when a new captured
>> + * frame is available:
>> + * 0 1 2 3 octet
>> + * +----------------+----------------+----------------+----------------+
>> + * | id |_EVT_FRAME_AVAIL| reserved | 4
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 8
>> + * +----------------+----------------+----------------+----------------+
>> + * | index | reserved | 12
>> + * +----------------+----------------+----------------+----------------+
>> + * | used_sz | 16
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 20
>> + * +----------------+----------------+----------------+----------------+
>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>> + * +----------------+----------------+----------------+----------------+
>> + * | reserved | 64
>> + * +----------------+----------------+----------------+----------------+
>> + *
>> + * index - uint8_t, index of the buffer that contains new captured frame.
>> + * used_sz - uint32_t, number of octets this frame has. This can be less
>> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
>> + */
>> +struct xencamera_frame_avail_evt {
>> + uint8_t index;
>> + uint8_t reserved[3];
>> + uint32_t used_sz;
>> +};
>> +
>> +struct xencamera_req {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved[5];
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_create_req buf_create;
>> + struct xencamera_buf_destroy_req buf_destroy;
>> + struct xencamera_set_ctrl_req set_ctrl;
>> + uint8_t reserved[56];
>> + } req;
>> +};
>> +
>> +struct xencamera_resp {
>> + uint16_t id;
>> + uint8_t operation;
>> + uint8_t reserved;
>> + int32_t status;
>> + union {
>> + struct xencamera_config config;
>> + struct xencamera_buf_details_resp buf_details;
>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>> + uint8_t reserved1[56];
>> + } resp;
>> +};
>> +
>> +struct xencamera_evt {
>> + uint16_t id;
>> + uint8_t type;
>> + uint8_t reserved[5];
>> + union {
>> + struct xencamera_frame_avail_evt frame_avail;
>> + uint8_t reserved[56];
>> + } evt;
>> +};
>> +
>> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
>> +
>> +/*
>> + ******************************************************************************
>> + * Back to front events delivery
>> + ******************************************************************************
>> + * In order to deliver asynchronous events from back to front a shared page is
>> + * allocated by front and its granted reference propagated to back via
>> + * XenStore entries (evt-ring-ref/evt-event-channel).
>> + * This page has a common header used by both front and back to synchronize
>> + * access and control event's ring buffer, while back being a producer of the
>> + * events and front being a consumer. The rest of the page after the header
>> + * is used for event packets.
>> + *
>> + * Upon reception of an event(s) front may confirm its reception
>> + * for either each event, group of events or none.
>> + */
>> +
>> +struct xencamera_event_page {
>> + uint32_t in_cons;
>> + uint32_t in_prod;
>> + uint8_t reserved[56];
>> +};
>> +
>> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
>> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
>> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
>> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
>> +#define XENCAMERA_IN_RING(page) \
>> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
>> +#define XENCAMERA_IN_RING_REF(page, idx) \
>> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
>> +
>> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
>> +
>> +/*
>> + * Local variables:
>> + * mode: C
>> + * c-file-style: "BSD"
>> + * c-basic-offset: 4
>> + * tab-width: 4
>> + * indent-tabs-mode: nil
>> + * End:
>> + */
>>
> I think the most important decision to make here is whether or not you want to support
> hotpluggable sources like HDMI. And an additional complication with that is HDCP.
> While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
> a patch series adding this is expected later this year/early next year. It might not
> be an issue in practice if these are all closed systems, but nevertheless, it is
> something to think about.
Yes, thank you for raising these questions, it is worth thinking
about such use-cases.
>
> Regards,
>
> Hans
Thank you so much for the comments,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 7:16 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-10 7:53 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 7:53 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, jgross,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi Oleksandr,
On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
> Hi, Hans!
>
> On 09/09/2018 01:31 PM, Hans Verkuil wrote:
>> Hi Oleksandr,
>>
>> Sorry for the delay in reviewing, I missed this patch until you pinged me, and
>> I was very busy after that as well.
> I do appreciate you spending time on this!
>>
>> On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>
>>> This is the ABI for the two halves of a para-virtualized
>>> camera driver which extends Xen's reach multimedia capabilities even
>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>> high definition maps etc.
>>>
>>> The initial goal is to support most needed functionality with the
>>> final idea to make it possible to extend the protocol if need be:
>>>
>>> 1. Provide means for base virtual device configuration:
>>> - pixel formats
>>> - resolutions
>>> - frame rates
>>> 2. Support basic camera controls:
>>> - contrast
>>> - brightness
>>> - hue
>>> - saturation
>>> 3. Support streaming control
>>> 4. Support zero-copying use-cases
>>>
>>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>> ---
>>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>>> 1 file changed, 981 insertions(+)
>>> create mode 100644 xen/include/public/io/cameraif.h
>>>
>>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>>> new file mode 100644
>>> index 000000000000..bdc6a1262fcf
>>> --- /dev/null
>>> +++ b/xen/include/public/io/cameraif.h
>>> @@ -0,0 +1,981 @@
>>> +/******************************************************************************
>>> + * cameraif.h
>>> + *
>>> + * Unified camera device I/O interface for Xen guest OSes.
>>> + *
>>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>>> + * of this software and associated documentation files (the "Software"), to
>>> + * deal in the Software without restriction, including without limitation the
>>> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
>>> + * sell copies of the Software, and to permit persons to whom the Software is
>>> + * furnished to do so, subject to the following conditions:
>>> + *
>>> + * The above copyright notice and this permission notice shall be included in
>>> + * all copies or substantial portions of the Software.
>>> + *
>>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>>> + * DEALINGS IN THE SOFTWARE.
>>> + *
>>> + * Copyright (C) 2018 EPAM Systems Inc.
>>> + *
>>> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>> + */
>> Use SPDX tag instead of copying the license text.
> This is yet a Xen header which belongs to Xen project and
> all the rest of the protocols have the same license header.
> If Xen community decides to use SPDX then I'll definitely follow.
Ah, yes, I was reviewing this as a kernel header, I hadn't realized
that this isn't a kernel header. Since it isn't a kernel header, you
can disregard my comments about style and naming conventions, since
you have your own.
>
> Konrad, do you think this is the right time for such a move?
>>
>>> +
>>> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
>>> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
>>> +
>>> +#include "ring.h"
>>> +#include "../grant_table.h"
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * Protocol version
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_PROTOCOL_VERSION "1"
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * Feature and Parameter Negotiation
>>> + ******************************************************************************
>>> + *
>>> + * Front->back notifications: when enqueuing a new request, sending a
>>> + * notification can be made conditional on xencamera_req (i.e., the generic
>>> + * hold-off mechanism provided by the ring macros). Backends must set
>>> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
>>> + *
>>> + * Back->front notifications: when enqueuing a new response, sending a
>>> + * notification can be made conditional on xencamera_resp (i.e., the generic
>>> + * hold-off mechanism provided by the ring macros). Frontends must set
>>> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
>>> + *
>>> + * The two halves of a para-virtual camera driver utilize nodes within
>>> + * XenStore to communicate capabilities and to negotiate operating parameters.
>>> + * This section enumerates these nodes which reside in the respective front and
>>> + * backend portions of XenStore, following the XenBus convention.
>>> + *
>>> + * All data in XenStore is stored as strings. Nodes specifying numeric
>>> + * values are encoded in decimal. Integer value ranges listed below are
>>> + * expressed as fixed sized integer types capable of storing the conversion
>>> + * of a properly formatted node string, without loss of information.
>>> + *
>>> + ******************************************************************************
>>> + * Example configuration
>>> + ******************************************************************************
>>> + *
>>> + * This is an example of backend and frontend configuration:
>>> + *
>>> + *--------------------------------- Backend -----------------------------------
>>> + *
>>> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
>>> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
>>> + * /local/domain/0/backend/vcamera/1/0/state = "4"
>>> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
>> Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
>> superfluous to me.
> 'v' stands for 'virtual'. I am following Xen convention used
> for all other virtual device protocols here.
>>> + *
>>> + *--------------------------------- Frontend ----------------------------------
>>> + *
>>> + * /local/domain/1/device/vcamera/0/backend-id = "0"
>>> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
>>> + * /local/domain/1/device/vcamera/0/state = "4"
>>> + * /local/domain/1/device/vcamera/0/version = "1"
>>> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
>>> + *
>>> + *---------------------------- Device 0 configuration -------------------------
>>> + *
>>> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
>>> + * /local/domain/1/device/vcamera/0/unique-id = "0"
>>> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
>>> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
>>> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
>>> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
>>> + *
>>> + *---------------------------- Device 1 configuration -------------------------
>>> + *
>>> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
>>> + * /local/domain/1/device/vcamera/1/unique-id = "1"
>>> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
>>> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
>>> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
>>> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
>>> + *
>>> + ******************************************************************************
>>> + * Backend XenBus Nodes
>>> + ******************************************************************************
>>> + *
>>> + *----------------------------- Protocol version ------------------------------
>>> + *
>>> + * versions
>>> + * Values: <string>
>>> + *
>>> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
>>> + * by the backend. For example "1,2,3".
>>> + *
>>> + ******************************************************************************
>>> + * Frontend XenBus Nodes
>>> + ******************************************************************************
>>> + *
>>> + *-------------------------------- Addressing ---------------------------------
>>> + *
>>> + * dom-id
>>> + * Values: <uint16_t>
>>> + *
>>> + * Domain identifier.
>>> + *
>>> + * dev-id
>>> + * Values: <uint16_t>
>>> + *
>>> + * Device identifier.
>>> + *
>>> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
>>> + *
>>> + *----------------------------- Protocol version ------------------------------
>>> + *
>>> + * version
>>> + * Values: <string>
>>> + *
>>> + * Protocol version, chosen among the ones supported by the backend.
>>> + *
>>> + *------------------------- Backend buffer allocation -------------------------
>>> + *
>>> + * be-alloc
>> I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
>> is better to just write 'backend-alloc'.
> Well, 'be' and 'fe' are commonly used in Xen for backend
> and frontend, so I'll probably stick to that convention for now.
>>
>>> + * Values: "0", "1"
>>> + *
>>> + * If value is set to "1", then backend will be the buffer
>>> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
>>> + * operation.
>>> + * If value is not "1" or omitted frontend must allocate buffers itself.
>>> + *
>>> + *------------------------------- Camera settings -----------------------------
>>> + *
>>> + * unique-id
>>> + * Values: <string>
>>> + *
>>> + * After device instance initialization each camera is assigned a
>>> + * unique ID, so it can be identified by the backend by this ID.
>>> + * This can be UUID or such.
>>> + *
>>> + * controls
>>> + * Values: <list of string>
>>> + *
>>> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
>>> + * Camera controls are expressed as a list of string values w/o any
>>> + * ordering requirement.
>>> + *
>>> + * formats
>>> + * Values: <format, char[4]>
>>> + *
>>> + * Formats are organized as a set of directories one per each
>>> + * supported pixel format. The name of the directory is an upper case
>>> + * string of the corresponding FOURCC string label. The next level of
>>> + * the directory under <formats> represents supported resolutions.
>> Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
> Ok, no problem - will remove the 'upper case' from the definition
>>
>> In addition it is common to set bit 31 of the fourcc to 1 if the format is
>> big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
>> add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
>> keep that convention.
> I'll think about it, thank you
>>
>>> + *
>>> + * resolution
>>> + * Values: <width, uint32_t>x<height, uint32_t>
>>> + *
>>> + * Resolutions are organized as a set of directories one per each
>>> + * supported resolution under corresponding <formats> directory.
>>> + * The name of the directory is the supported width and height
>>> + * of the camera resolution in pixels.
>> What if you are dealing with an HDMI input? Not unreasonable for media
>> systems. There can be a lot of resolutions/framerates, and the resolution
>> can change on the fly, or of course disappear.
> Well, this is a part of the system configuration done
> before we actually run the VMs, e.g. at system design
> and configuration time. Most of the time you do know which
> resolutions, frame rates etc. you want to assign and these
> settings remain static for the whole lifetime of that VM.
> If you are designing a system which needs these resolutions
> to change at run-time then this can be done:
> 1. Backend changes the state of the frontend to XenbusStateClosed state
> 2. Xen (xl/libxl) or backend change the configuration in XenStore
> 3. Backend re-initializes the frontend which reads new configuration
>>
>> What is also missing here is a way to report pixel aspect ratio: PAL and
>> NTSC-based video material doesn't have square pixels.
> Hm, indeed, thank you. I'll put this as a fraction under
> the corresponding 'resolution':
>
> Now:
> /local/domain/2/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024 = "015/2"
>
> Will change to:
> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/framerates/ =
> "30/1,15/1,15/2"
> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/aspectratio = "1/1"
You need to be more precise: aspectratio can refer to the display aspect
ratio or the pixel aspect ratio, and you can calculate one from the other.
So this is either pixelaspectratio or displayaspectratio.
Even though V4L2 provides a pixelaspectratio, I would recommend that you
choose to export it as a displayaspectratio. It's easier for applications
to handle.
Please note that in the case of DVD/BluRay source material the same
resolution can have different display aspect ratios: e.g. PAL can be
4:3 or 16:9, depending on whether it is widescreen or not.
>
> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/framerates = "015/2"
> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/aspectratio =
> "59/58"
>
>>
>> It's important to decide whether or not you want to support video sources
>> like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
>> or disconnect webcams) or if you stick to fixed camera pipelines.
> I believe that this is all hidden from the frontend by the
> backend, so I se nothing we have to put in the protocol
> with this respect.
>>
>> The big difference is that you don't control what someone can connect as
>> external sources, so you will have to be a lot more careful and robust.
>>
>> I suspect that you likely will want to support such sources eventually, so
>> it pays to design this with that in mind.
> Again, I think that this is the backend to hide these
> use-cases from the frontend.
I'm not sure you can: say you are playing a bluray connected to the system
with HDMI, then if there is a resolution change, what do you do? You can tear
everything down and build it up again, or you can just tell frontends that
something changed and that they have to look at the new vcamera configuration.
The latter seems to be more sensible to me. It is really not much that you
need to do: all you really need is an event signalling that something changed.
In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>
>>> + *
>>> + * frame-rates
>>> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
>>> + *
>>> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
>>> + * of the camera expressed as numerator and denominator of the
>>> + * corresponding frame rate.
>>> + *
>>> + * The format of the <formats> directory tree with resolutions and frame rates
>>> + * must be structured in the following format:
>>> + *
>>> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
>>> + *
>>> + * where
>>> + * i - i-th supported pixel format
>>> + * j - j-th supported resolution for i-th pixel format
>>> + * k - k-th supported frame rate for i-th pixel format and j-th
>>> + * resolution> + *
>>> + *------------------- Camera Request Transport Parameters ---------------------
>>> + *
>>> + * This communication path is used to deliver requests from frontend to backend
>>> + * and get the corresponding responses from backend to frontend,
>>> + * set up per virtual camera device.
>>> + *
>>> + * req-event-channel
>>> + * Values: <uint32_t>
>>> + *
>>> + * The identifier of the Xen camera's control event channel
>>> + * used to signal activity in the ring buffer.
>>> + *
>>> + * req-ring-ref
>>> + * Values: <uint32_t>
>>> + *
>>> + * The Xen grant reference granting permission for the backend to map
>>> + * a sole page of camera's control ring buffer.
>>> + *
>>> + *-------------------- Camera Event Transport Parameters ----------------------
>>> + *
>>> + * This communication path is used to deliver asynchronous events from backend
>>> + * to frontend, set up per virtual camera device.
>>> + *
>>> + * evt-event-channel
>>> + * Values: <uint32_t>
>>> + *
>>> + * The identifier of the Xen camera's event channel
>>> + * used to signal activity in the ring buffer.
>>> + *
>>> + * evt-ring-ref
>>> + * Values: <uint32_t>
>>> + *
>>> + * The Xen grant reference granting permission for the backend to map
>>> + * a sole page of camera's event ring buffer.
>>> + */
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * STATE DIAGRAMS
>>> + ******************************************************************************
>>> + *
>>> + * Tool stack creates front and back state nodes with initial state
>>> + * XenbusStateInitialising.
>>> + * Tool stack creates and sets up frontend camera configuration
>>> + * nodes per domain.
>>> + *
>>> + *-------------------------------- Normal flow --------------------------------
>>> + *
>>> + * Front Back
>>> + * ================================= =====================================
>>> + * XenbusStateInitialising XenbusStateInitialising
>>> + * o Query backend device identification
>>> + * data.
>>> + * o Open and validate backend device.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateInitWait
>>> + *
>>> + * o Query frontend configuration
>>> + * o Allocate and initialize
>>> + * event channels per configured
>>> + * camera.
>>> + * o Publish transport parameters
>>> + * that will be in effect during
>>> + * this connection.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateInitialised
>>> + *
>>> + * o Query frontend transport parameters.
>>> + * o Connect to the event channels.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateConnected
>>> + *
>>> + * o Create and initialize OS
>>> + * virtual camera as per
>>> + * configuration.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateConnected
>>> + *
>>> + * XenbusStateUnknown
>>> + * XenbusStateClosed
>>> + * XenbusStateClosing
>>> + * o Remove virtual camera device
>>> + * o Remove event channels
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateClosed
>>> + *
>>> + *------------------------------- Recovery flow -------------------------------
>>> + *
>>> + * In case of frontend unrecoverable errors backend handles that as
>>> + * if frontend goes into the XenbusStateClosed state.
>>> + *
>>> + * In case of backend unrecoverable errors frontend tries removing
>>> + * the virtualized device. If this is possible at the moment of error,
>>> + * then frontend goes into the XenbusStateInitialising state and is ready for
>>> + * new connection with backend. If the virtualized device is still in use and
>>> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
>>> + * until either the virtualized device is removed or backend initiates a new
>>> + * connection. On the virtualized device removal frontend goes into the
>>> + * XenbusStateInitialising state.
>>> + *
>>> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
>>> + * unrecoverable errors then frontend cannot send requests to the backend
>>> + * and thus cannot provide functionality of the virtualized device anymore.
>>> + * After backend is back to normal the virtualized device may still hold some
>>> + * state: configuration in use, allocated buffers, client application state etc.
>>> + * In most cases, this will require frontend to implement complex recovery
>>> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
>>> + * frontend will make sure no new clients of the virtualized device are
>>> + * accepted, allow existing client(s) to exit gracefully by signaling error
>>> + * state etc.
>>> + * Once all the clients are gone frontend can reinitialize the virtualized
>>> + * device and get into XenbusStateInitialising state again signaling the
>>> + * backend that a new connection can be made.
>>> + *
>>> + * There are multiple conditions possible under which frontend will go from
>>> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
>>> + * specific. For example:
>>> + * 1. The underlying OS framework may provide callbacks to signal that the last
>>> + * client of the virtualized device has gone and the device can be removed
>>> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
>>> + * to periodically check if this is the right time to re-try removal of
>>> + * the virtualized device.
>>> + * 3. By any other means.
>>> + *
>>> + ******************************************************************************
>>> + * REQUEST CODES
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_OP_SET_CONFIG 0x00
>>> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
>>> +#define XENCAMERA_OP_BUF_CREATE 0x02
>>> +#define XENCAMERA_OP_BUF_DESTROY 0x03
>>> +#define XENCAMERA_OP_STREAM_START 0x04
>>> +#define XENCAMERA_OP_STREAM_STOP 0x05
>>> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
>>> +#define XENCAMERA_OP_SET_CTRL 0x07
> I am thinking about extending the command set a bit as it already
> has some flaws, e.g. there is no way for a VM to tell the backend
> that the buffer is not in use anymore and can be given back
> to the real HW driver, e.g. queue/dequeue in V4L2 terms:
>
> #define XENCAMERA_OP_SET_FORMAT 0x00
> - will be used to set format: pixel format, resolution
>
> #define XENCAMERA_OP_SET_FRAME_RATE 0x01
> - used to set the frame rate
>
> #define XENCAMERA_OP_BUF_REQUEST 0x02
> - asks backend to allocate the given number of buffers,
> backend replies with real number of those to be used
>
> #define XENCAMERA_OP_BUF_CREATE 0x03
> - create a shared buffer
>
> #define XENCAMERA_OP_BUF_DESTROY 0x04
> - destroy a shared buffer
>
> #define XENCAMERA_OP_BUF_QUEUE 0x05
> - VM tells the backend that it has access to the shared buffer
> and the buffer cannot be sent back to real HW driver
>
> #define XENCAMERA_OP_BUF_DEQUEUE 0x06
> - VM tells the backend that the shared buffer is not in use and
> can be sent to real HW driver
This is the wrong way around: QUEUE would queue the shared buffer to
the backend for use with the real HW driver, DEQUEUE would dequeue it
for use in the VM.
>
> #define XENCAMERA_OP_CTRL_ENUM 0x07
> - get i-th control ranges and settings
>
> #define XENCAMERA_OP_CTRL_GET 0x08
> - get control value
>
> #define XENCAMERA_OP_CTRL_SET 0x09
> - set control value
>
> #define XENCAMERA_OP_STREAM_START 0x0a
> - start streaming
>
> #define XENCAMERA_OP_STREAM_STOP 0x0b
> - stop ctreaming
>
>
>>> +
>>> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
>>> +#define XENCAMERA_CTRL_CONTRAST 0x01
>>> +#define XENCAMERA_CTRL_SATURATION 0x02
>>> +#define XENCAMERA_CTRL_HUE 0x03
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * EVENT CODES
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_DRIVER_NAME "vcamera"
>> Ah, that's where vcamera comes from. How about calling this xen-camera or
>> virt-camera? With a preference for xen-camera, since that's what you use for the
>> defines as well.
>>
>> Or perhaps pv-camera?
>>
>> Is this driver going to be xen-specific, or more a general approach that everyone
>> can use? Obviously, the latter would be preferable.
> As I have already replied to the cover letter with explanations:
> 'v' stands for 'virtual' and there is a convention to name the
> Xen virtual devices starting with 'v': vif, vkbd etc.
Yeah, ignore my comment.
>>
>> BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
>> radio and RDS hardware. Contact me if this is of interest to Xen to support this as
>> well given the automotive use-case.
> Yes, thank you, but at this stage we are targeting camera only.
> Radio can be another topic if time allows ;) And most probably
> it will be a dedicated 'vradio' protocol then...
>>
>>> +
>>> +#define XENCAMERA_LIST_SEPARATOR ","
>>> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
>>> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
>>> +
>>> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
>>> +#define XENCAMERA_FIELD_FE_VERSION "version"
>>> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
>>> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
>>> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
>>> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
>>> +#define XENCAMERA_FIELD_CONTROLS "controls"
>>> +#define XENCAMERA_FIELD_FORMATS "formats"
>>> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
>>> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
>>> +
>>> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
>>> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
>>> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
>>> +#define XENCAMERA_CTRL_HUE_STR "hue"
>>> +
>>> +/* Maximum number of buffer planes supported. */
>>> +#define XENCAMERA_MAX_PLANE 4
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * STATUS RETURN CODES
>>> + ******************************************************************************
>>> + *
>>> + * Status return code is zero on success and -XEN_EXX on failure.
>>> + *
>>> + ******************************************************************************
>>> + * Assumptions
>>> + ******************************************************************************
>>> + *
>>> + * - usage of grant reference 0 as invalid grant reference:
>>> + * grant reference 0 is valid, but never exposed to a PV driver,
>>> + * because of the fact it is already in use/reserved by the PV console.
>>> + * - all references in this document to page sizes must be treated
>>> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
>>> + *
>>> + ******************************************************************************
>>> + * Description of the protocol between frontend and backend driver
>>> + ******************************************************************************
>>> + *
>>> + * The two halves of a Para-virtual camera driver communicate with
>>> + * each other using shared pages and event channels.
>>> + * Shared page contains a ring with request/response packets.
>>> + *
>>> + * All reserved fields in the structures below must be 0.
>>> + *
>>> + * For all request/response/event packets:
>>> + * - frame rate parameter is represented as a pair of 4 octet long
>>> + * numerator and denominator:
>>> + * - frame_rate_numer - uint32_t, numerator of the frame rate
>>> + * - frame_rate_denom - uint32_t, denominator of the frame rate
>>> + * The corresponding frame rate (Hz) is calculated as:
>>> + * frame_rate = frame_rate_numer / frame_rate_denom
>>> + * - buffer index is a zero based index of the buffer. Must be less than
>>> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
>>> + * - index - uint8_t, index of the buffer.
>>> + *
>>> + *
>>> + *---------------------------------- Requests ---------------------------------
>>> + *
>>> + * All request packets have the same length (64 octets).
>>> + * All request packets have common header:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | operation | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * id - uint16_t, private guest value, echoed in response.
>>> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
>>> + *
>>> + *
>>> + * Request configuration set/reset - request to set or reset.
>>> + * the configuration/mode of the camera:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | pixel format | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | width | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | height | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_numer | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_denom | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | num_bufs | reserved | 32
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * Pass all zeros to reset, otherwise command is treated as configuration set.
>>> + *
>>> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>> + * width - uint32_t, width in pixels.
>>> + * height - uint32_t, height in pixels.
>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>> If you have to support HDMI/SDTV inputs as well, then you also need to know
>> the interlaced format, unless you have no plans to support that.
>>
>>> + * num_bufs - uint8_t, desired number of buffers to be used.
>> Huh? What has that to do with the format? Why would you need this here?
> Well, the operation name is 'set_config', not 'set_format',
> so I thought we can have such a cumulative command assembling
> all the parameters of the configuration. But now I am looking at
> turning this single 'set_config' command to 3 different commands,
> which is more practical and aligned with V4L2 in particular (please
> see above in the command set):
> 1. set format command:
> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
> * width - uint32_t, width in pixels.
> * height - uint32_t, height in pixels.
>
> 2. Set frame rate command:
> + * frame_rate_numer - uint32_t, numerator of the frame rate.
> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>
> 3. Set/request num bufs:
> * num_bufs - uint8_t, desired number of buffers to be used.
I like this much better. 1+2 could be combined, but 3 should definitely remain
separate.
>>
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + * Notes:
>>> + * - frontend must check the corresponding response in order to see
>>> + * if the values reported back by the backend do match the desired ones
>>> + * and can be accepted.
>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>> + * configuration.
>>> + */
>>> +struct xencamera_config {
>>> + uint32_t pixel_format;
>>> + uint32_t width;
>>> + uint32_t height;
>>> + uint32_t frame_rate_nom;
>>> + uint32_t frame_rate_denom;
>>> + uint8_t num_bufs;
>>> +};
>>> +
>>> +/*
>>> + * Request buffer details - request camera buffer's memory layout.
>>> + * detailed description:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + *
>>> + * Request camera buffer creation:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref_directory | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * An attempt to create multiple buffers with the same index is an error.
>>> + * index can be re-used after destroying the corresponding camera buffer.
>>> + *
>>> + * index - uint8_t, index of the buffer to be created.
>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>> + * describing shared buffer references. The size of the buffer is equal to
>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>> + * shared buffer size exceeds what can be addressed by this single page,
>>> + * then reference to the next shared page must be supplied (see
>>> + * gref_dir_next_page below).
>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>> does.
> Well, I still think it is better to have a per buffer interface
> in the protocol as it is done for other Xen virtual devices.
> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
> what it does internally in the frontend driver
I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
sense. I'm not sure about the naming.
You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
on the backend, and so can fail. Also, the actual number of allocated buffers in
case of success can be more or less than what was requested.
>>
>>> + *
>>> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
>>> + * allocate the buffer with the parameters provided in this request and page
>>> + * directory is handled as follows:
>>> + * Frontend on request:
>>> + * - allocates pages for the directory (gref_directory,
>>> + * gref_dir_next_page(s)
>>> + * - grants permissions for the pages of the directory to the backend
>>> + * - sets gref_dir_next_page fields
>>> + * Backend on response:
>>> + * - grants permissions for the pages of the buffer allocated to
>>> + * the frontend
>>> + * - fills in page directory with grant references
>>> + * (gref[] in struct xencamera_page_directory)
>>> + */
>>> +struct xencamera_buf_create_req {
>>> + uint8_t index;
>>> + uint8_t reserved[3];
>>> + grant_ref_t gref_directory;
>>> +};
>>> +
>>> +/*
>>> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
>>> + * the request) employs a list of pages, describing all pages of the shared
>>> + * data buffer:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref_dir_next_page | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref[0] | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref[i] | i*4+8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref[N - 1] | N*4+8
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
>>> + * page directory. Must be 0 if there are no more pages in the list.
>>> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
>>> + * allocated at XENCAMERA_OP_BUF_CREATE.
>>> + *
>>> + * Number of grant_ref_t entries in the whole page directory is not
>>> + * passed, but instead can be calculated as:
>>> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
>>> + * XEN_PAGE_SIZE
>>> + */
>>> +struct xencamera_page_directory {
>>> + grant_ref_t gref_dir_next_page;
>>> + grant_ref_t gref[1]; /* Variable length */
>>> +};
>>> +
>>> +/*
>>> + * Request buffer destruction - destroy a previously allocated camera buffer:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_BUF_DESTROY| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * index - uint8_t, index of the buffer to be destroyed.
>>> + */
>> There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
>> of them.
> This is not specifically related to V4L2, but can be issued
> in response to backend's state change etc.
> So, we have a pair of commands to create and destroy buffers.
> Even more, frontend can be a some-os-based-driver, not V4L2
> based. Or even a user-space application if your will.
>>
>>> +
>>> +struct xencamera_buf_destroy_req {
>>> + uint8_t index;
>>> +};
>>> +
>>> +/*
>>> + * Request camera capture stream start:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_OP_STREAM_START| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + *
>>> + * Request camera capture stream stop:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_OP_STREAM_STOP | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + *
>>> + * Request camera control details:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + * index - uint8_t, index of the control to be queried.
>>> + */
>>> +struct xencamera_get_ctrl_details_req {
>>> + uint8_t index;
>>> +};
>>> +
>>> +/*
>>> + *
>>> + * Request camera control change:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_SET_CTRL | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | value | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + * index - uint8_t, index of the control.
>>> + * value - int32_t, new value of the control.
>> I would recommend using a int64_t as the control value.
> Good point, thank you
>>
>> Note that there are also controls with a payload (e.g. string controls).
> Could you please give me an example of such a control?
> Do you think such controls can be of use in a VM?
> Can we avoid such controls if we target a simple virtual
> camera device? If this is for radio use-case, then we'll
> have such support in 'vradio' protocol if need be
Right now all string controls are related to RDS receivers/transmitters. If you
ever decide on a vradio protocol, then you need these (Programme Service name
and Radio Text info). And there is an array of Alternate Frequencies, also RDS
specific.
There are some array controls in V4L2, those are used to control motion detection
for surveillance cameras. And 'compound controls' (think of this as C structs) are
appearing for HW codecs.
I don't think any of these are likely to appear for cameras, at least not in a
way that is relevant for Xen.
>>
>> If there is ever interest in adding radio/RDS support, then that will become
>> an issue.
> You mean something like station names, ads etc?
Yup.
>>
>>> + */
>>> +struct xencamera_set_ctrl_req {
>>> + uint8_t index;
>>> + uint8_t reserved[3];
>>> + int32_t value;
>>> +};
>>> +
>>> +/*
>>> + *---------------------------------- Responses --------------------------------
>>> + *
>>> + * All response packets have the same length (64 octets).
>>> + *
>>> + * All response packets have common header:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | operation | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * id - uint16_t, copied from the request.
>>> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
>>> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
>>> + *
>>> + *
>>> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | pixel format | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | width | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | height | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_numer | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_denom | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | num_bufs | reserved | 32
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * Meaning of the corresponding values in this response is the same as for
>>> + * XENCAMERA_OP_SET_CONFIG request.
>>> + *
>>> + *
>>> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
>>> + * request:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | size | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | num_planes | reserved | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[0] | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[1] | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[2] | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[3] | 32
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[0] | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[1] | 40
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[2] | 44
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[3] | 48
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_stride[0] | plane_stride[1] | 52
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_stride[2] | plane_stride[3] | 56
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 60
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * size - uint32_t, overall size of the buffer including sizes of the
>>> + * individual planes and padding if applicable.
>>> + * num_planes - uint8_t, number of planes for this buffer.
>>> + * plane_offset - array of uint32_t, offset of the corresponding plane
>>> + * in octets from the buffer start.
>>> + * plane_size - array of uint32_t, size in octets of the corresponding plane
>>> + * including padding.
>>> + * plane_stride - array of uint32_t, size in octets occupied by the
>>> + * corresponding single image line including padding if applicable.
>> Nice!
> Thank you
>>> + */
>>> +struct xencamera_buf_details_resp {
>>> + uint32_t size;
>>> + uint8_t num_planes;
>>> + uint8_t reserved[3];
>>> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
>>> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
>>> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
>>> +};
>>> +
>>> +/*
>>> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | type | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | min | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | max | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | step | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | def_val | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * index - uint8_t, index of the camera control in response.
>>> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
>>> + * min - int32_t, minimum value of the control.
>>> + * max - int32_t, maximum value of the control.
>>> + * step - int32_t, minimum size in which control value can be changed.
>>> + * def_val - int32_t, default value of the control.
>> I'd go with 64 bit values for min/max/step/def_val.
> Sure, good idea, thank you
>>
>> I would also add a flags field. Some controls are read-only, write-only
>> or volatile, things userspace needs to know.
> Then I'll also add numerical constants for such
>>
>> If you want to support menu controls, then you need a way to get the menu
>> names as well (VIDIOC_QUERYMENU).
>>
>> None of this is needed for this initial use-case, but you need to think
>> about this up-front.
> Yes, thank you
>>> + */
>>> +struct xencamera_get_ctrl_details_resp {
>>> + uint8_t index;
>>> + uint8_t type;
>>> + uint8_t reserved[2];
>>> + int32_t min;
>>> + int32_t max;
>>> + int32_t step;
>>> + int32_t def_val;
>>> +};
>>> +
>>> +/*
>>> + *----------------------------------- Events ----------------------------------
>>> + *
>>> + * Events are sent via a shared page allocated by the front and propagated by
>>> + * evt-event-channel/evt-ring-ref XenStore entries.
>>> + *
>>> + * All event packets have the same length (64 octets).
>>> + * All event packets have common header:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | type | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * id - uint16_t, event id, may be used by front.
>>> + * type - uint8_t, type of the event.
>>> + *
>>> + *
>>> + * Frame captured event - event from back to front when a new captured
>>> + * frame is available:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_EVT_FRAME_AVAIL| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | used_sz | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * index - uint8_t, index of the buffer that contains new captured frame.
>>> + * used_sz - uint32_t, number of octets this frame has. This can be less
>>> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
>>> + */
>>> +struct xencamera_frame_avail_evt {
>>> + uint8_t index;
>>> + uint8_t reserved[3];
>>> + uint32_t used_sz;
>>> +};
>>> +
>>> +struct xencamera_req {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved[5];
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_create_req buf_create;
>>> + struct xencamera_buf_destroy_req buf_destroy;
>>> + struct xencamera_set_ctrl_req set_ctrl;
>>> + uint8_t reserved[56];
>>> + } req;
>>> +};
>>> +
>>> +struct xencamera_resp {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved;
>>> + int32_t status;
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_details_resp buf_details;
>>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>>> + uint8_t reserved1[56];
>>> + } resp;
>>> +};
>>> +
>>> +struct xencamera_evt {
>>> + uint16_t id;
>>> + uint8_t type;
>>> + uint8_t reserved[5];
>>> + union {
>>> + struct xencamera_frame_avail_evt frame_avail;
>>> + uint8_t reserved[56];
>>> + } evt;
>>> +};
>>> +
>>> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * Back to front events delivery
>>> + ******************************************************************************
>>> + * In order to deliver asynchronous events from back to front a shared page is
>>> + * allocated by front and its granted reference propagated to back via
>>> + * XenStore entries (evt-ring-ref/evt-event-channel).
>>> + * This page has a common header used by both front and back to synchronize
>>> + * access and control event's ring buffer, while back being a producer of the
>>> + * events and front being a consumer. The rest of the page after the header
>>> + * is used for event packets.
>>> + *
>>> + * Upon reception of an event(s) front may confirm its reception
>>> + * for either each event, group of events or none.
>>> + */
>>> +
>>> +struct xencamera_event_page {
>>> + uint32_t in_cons;
>>> + uint32_t in_prod;
>>> + uint8_t reserved[56];
>>> +};
>>> +
>>> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
>>> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
>>> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
>>> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
>>> +#define XENCAMERA_IN_RING(page) \
>>> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
>>> +#define XENCAMERA_IN_RING_REF(page, idx) \
>>> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
>>> +
>>> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
>>> +
>>> +/*
>>> + * Local variables:
>>> + * mode: C
>>> + * c-file-style: "BSD"
>>> + * c-basic-offset: 4
>>> + * tab-width: 4
>>> + * indent-tabs-mode: nil
>>> + * End:
>>> + */
>>>
>> I think the most important decision to make here is whether or not you want to support
>> hotpluggable sources like HDMI. And an additional complication with that is HDCP.
>> While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
>> a patch series adding this is expected later this year/early next year. It might not
>> be an issue in practice if these are all closed systems, but nevertheless, it is
>> something to think about.
> Yes, thank you for raising these questions, it is worth thinking
> about such use-cases.
>>
>> Regards,
>>
>> Hans
> Thank you so much for the comments,
> Oleksandr
>
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 7:53 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 7:53 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, jgross,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
Hi Oleksandr,
On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
> Hi, Hans!
>
> On 09/09/2018 01:31 PM, Hans Verkuil wrote:
>> Hi Oleksandr,
>>
>> Sorry for the delay in reviewing, I missed this patch until you pinged me, and
>> I was very busy after that as well.
> I do appreciate you spending time on this!
>>
>> On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>
>>> This is the ABI for the two halves of a para-virtualized
>>> camera driver which extends Xen's reach multimedia capabilities even
>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>> high definition maps etc.
>>>
>>> The initial goal is to support most needed functionality with the
>>> final idea to make it possible to extend the protocol if need be:
>>>
>>> 1. Provide means for base virtual device configuration:
>>> - pixel formats
>>> - resolutions
>>> - frame rates
>>> 2. Support basic camera controls:
>>> - contrast
>>> - brightness
>>> - hue
>>> - saturation
>>> 3. Support streaming control
>>> 4. Support zero-copying use-cases
>>>
>>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>> ---
>>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>>> 1 file changed, 981 insertions(+)
>>> create mode 100644 xen/include/public/io/cameraif.h
>>>
>>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>>> new file mode 100644
>>> index 000000000000..bdc6a1262fcf
>>> --- /dev/null
>>> +++ b/xen/include/public/io/cameraif.h
>>> @@ -0,0 +1,981 @@
>>> +/******************************************************************************
>>> + * cameraif.h
>>> + *
>>> + * Unified camera device I/O interface for Xen guest OSes.
>>> + *
>>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>>> + * of this software and associated documentation files (the "Software"), to
>>> + * deal in the Software without restriction, including without limitation the
>>> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
>>> + * sell copies of the Software, and to permit persons to whom the Software is
>>> + * furnished to do so, subject to the following conditions:
>>> + *
>>> + * The above copyright notice and this permission notice shall be included in
>>> + * all copies or substantial portions of the Software.
>>> + *
>>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>>> + * DEALINGS IN THE SOFTWARE.
>>> + *
>>> + * Copyright (C) 2018 EPAM Systems Inc.
>>> + *
>>> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>> + */
>> Use SPDX tag instead of copying the license text.
> This is yet a Xen header which belongs to Xen project and
> all the rest of the protocols have the same license header.
> If Xen community decides to use SPDX then I'll definitely follow.
Ah, yes, I was reviewing this as a kernel header, I hadn't realized
that this isn't a kernel header. Since it isn't a kernel header, you
can disregard my comments about style and naming conventions, since
you have your own.
>
> Konrad, do you think this is the right time for such a move?
>>
>>> +
>>> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
>>> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
>>> +
>>> +#include "ring.h"
>>> +#include "../grant_table.h"
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * Protocol version
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_PROTOCOL_VERSION "1"
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * Feature and Parameter Negotiation
>>> + ******************************************************************************
>>> + *
>>> + * Front->back notifications: when enqueuing a new request, sending a
>>> + * notification can be made conditional on xencamera_req (i.e., the generic
>>> + * hold-off mechanism provided by the ring macros). Backends must set
>>> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
>>> + *
>>> + * Back->front notifications: when enqueuing a new response, sending a
>>> + * notification can be made conditional on xencamera_resp (i.e., the generic
>>> + * hold-off mechanism provided by the ring macros). Frontends must set
>>> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
>>> + *
>>> + * The two halves of a para-virtual camera driver utilize nodes within
>>> + * XenStore to communicate capabilities and to negotiate operating parameters.
>>> + * This section enumerates these nodes which reside in the respective front and
>>> + * backend portions of XenStore, following the XenBus convention.
>>> + *
>>> + * All data in XenStore is stored as strings. Nodes specifying numeric
>>> + * values are encoded in decimal. Integer value ranges listed below are
>>> + * expressed as fixed sized integer types capable of storing the conversion
>>> + * of a properly formatted node string, without loss of information.
>>> + *
>>> + ******************************************************************************
>>> + * Example configuration
>>> + ******************************************************************************
>>> + *
>>> + * This is an example of backend and frontend configuration:
>>> + *
>>> + *--------------------------------- Backend -----------------------------------
>>> + *
>>> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
>>> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
>>> + * /local/domain/0/backend/vcamera/1/0/state = "4"
>>> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
>> Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
>> superfluous to me.
> 'v' stands for 'virtual'. I am following Xen convention used
> for all other virtual device protocols here.
>>> + *
>>> + *--------------------------------- Frontend ----------------------------------
>>> + *
>>> + * /local/domain/1/device/vcamera/0/backend-id = "0"
>>> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
>>> + * /local/domain/1/device/vcamera/0/state = "4"
>>> + * /local/domain/1/device/vcamera/0/version = "1"
>>> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
>>> + *
>>> + *---------------------------- Device 0 configuration -------------------------
>>> + *
>>> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
>>> + * /local/domain/1/device/vcamera/0/unique-id = "0"
>>> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
>>> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
>>> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
>>> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
>>> + *
>>> + *---------------------------- Device 1 configuration -------------------------
>>> + *
>>> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
>>> + * /local/domain/1/device/vcamera/1/unique-id = "1"
>>> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
>>> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
>>> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
>>> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
>>> + *
>>> + ******************************************************************************
>>> + * Backend XenBus Nodes
>>> + ******************************************************************************
>>> + *
>>> + *----------------------------- Protocol version ------------------------------
>>> + *
>>> + * versions
>>> + * Values: <string>
>>> + *
>>> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
>>> + * by the backend. For example "1,2,3".
>>> + *
>>> + ******************************************************************************
>>> + * Frontend XenBus Nodes
>>> + ******************************************************************************
>>> + *
>>> + *-------------------------------- Addressing ---------------------------------
>>> + *
>>> + * dom-id
>>> + * Values: <uint16_t>
>>> + *
>>> + * Domain identifier.
>>> + *
>>> + * dev-id
>>> + * Values: <uint16_t>
>>> + *
>>> + * Device identifier.
>>> + *
>>> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
>>> + *
>>> + *----------------------------- Protocol version ------------------------------
>>> + *
>>> + * version
>>> + * Values: <string>
>>> + *
>>> + * Protocol version, chosen among the ones supported by the backend.
>>> + *
>>> + *------------------------- Backend buffer allocation -------------------------
>>> + *
>>> + * be-alloc
>> I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
>> is better to just write 'backend-alloc'.
> Well, 'be' and 'fe' are commonly used in Xen for backend
> and frontend, so I'll probably stick to that convention for now.
>>
>>> + * Values: "0", "1"
>>> + *
>>> + * If value is set to "1", then backend will be the buffer
>>> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
>>> + * operation.
>>> + * If value is not "1" or omitted frontend must allocate buffers itself.
>>> + *
>>> + *------------------------------- Camera settings -----------------------------
>>> + *
>>> + * unique-id
>>> + * Values: <string>
>>> + *
>>> + * After device instance initialization each camera is assigned a
>>> + * unique ID, so it can be identified by the backend by this ID.
>>> + * This can be UUID or such.
>>> + *
>>> + * controls
>>> + * Values: <list of string>
>>> + *
>>> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
>>> + * Camera controls are expressed as a list of string values w/o any
>>> + * ordering requirement.
>>> + *
>>> + * formats
>>> + * Values: <format, char[4]>
>>> + *
>>> + * Formats are organized as a set of directories one per each
>>> + * supported pixel format. The name of the directory is an upper case
>>> + * string of the corresponding FOURCC string label. The next level of
>>> + * the directory under <formats> represents supported resolutions.
>> Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
> Ok, no problem - will remove the 'upper case' from the definition
>>
>> In addition it is common to set bit 31 of the fourcc to 1 if the format is
>> big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
>> add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
>> keep that convention.
> I'll think about it, thank you
>>
>>> + *
>>> + * resolution
>>> + * Values: <width, uint32_t>x<height, uint32_t>
>>> + *
>>> + * Resolutions are organized as a set of directories one per each
>>> + * supported resolution under corresponding <formats> directory.
>>> + * The name of the directory is the supported width and height
>>> + * of the camera resolution in pixels.
>> What if you are dealing with an HDMI input? Not unreasonable for media
>> systems. There can be a lot of resolutions/framerates, and the resolution
>> can change on the fly, or of course disappear.
> Well, this is a part of the system configuration done
> before we actually run the VMs, e.g. at system design
> and configuration time. Most of the time you do know which
> resolutions, frame rates etc. you want to assign and these
> settings remain static for the whole lifetime of that VM.
> If you are designing a system which needs these resolutions
> to change at run-time then this can be done:
> 1. Backend changes the state of the frontend to XenbusStateClosed state
> 2. Xen (xl/libxl) or backend change the configuration in XenStore
> 3. Backend re-initializes the frontend which reads new configuration
>>
>> What is also missing here is a way to report pixel aspect ratio: PAL and
>> NTSC-based video material doesn't have square pixels.
> Hm, indeed, thank you. I'll put this as a fraction under
> the corresponding 'resolution':
>
> Now:
> /local/domain/2/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024 = "015/2"
>
> Will change to:
> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/framerates/ =
> "30/1,15/1,15/2"
> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/aspectratio = "1/1"
You need to be more precise: aspectratio can refer to the display aspect
ratio or the pixel aspect ratio, and you can calculate one from the other.
So this is either pixelaspectratio or displayaspectratio.
Even though V4L2 provides a pixelaspectratio, I would recommend that you
choose to export it as a displayaspectratio. It's easier for applications
to handle.
Please note that in the case of DVD/BluRay source material the same
resolution can have different display aspect ratios: e.g. PAL can be
4:3 or 16:9, depending on whether it is widescreen or not.
>
> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/framerates = "015/2"
> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/aspectratio =
> "59/58"
>
>>
>> It's important to decide whether or not you want to support video sources
>> like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
>> or disconnect webcams) or if you stick to fixed camera pipelines.
> I believe that this is all hidden from the frontend by the
> backend, so I se nothing we have to put in the protocol
> with this respect.
>>
>> The big difference is that you don't control what someone can connect as
>> external sources, so you will have to be a lot more careful and robust.
>>
>> I suspect that you likely will want to support such sources eventually, so
>> it pays to design this with that in mind.
> Again, I think that this is the backend to hide these
> use-cases from the frontend.
I'm not sure you can: say you are playing a bluray connected to the system
with HDMI, then if there is a resolution change, what do you do? You can tear
everything down and build it up again, or you can just tell frontends that
something changed and that they have to look at the new vcamera configuration.
The latter seems to be more sensible to me. It is really not much that you
need to do: all you really need is an event signalling that something changed.
In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>
>>> + *
>>> + * frame-rates
>>> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
>>> + *
>>> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
>>> + * of the camera expressed as numerator and denominator of the
>>> + * corresponding frame rate.
>>> + *
>>> + * The format of the <formats> directory tree with resolutions and frame rates
>>> + * must be structured in the following format:
>>> + *
>>> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
>>> + *
>>> + * where
>>> + * i - i-th supported pixel format
>>> + * j - j-th supported resolution for i-th pixel format
>>> + * k - k-th supported frame rate for i-th pixel format and j-th
>>> + * resolution> + *
>>> + *------------------- Camera Request Transport Parameters ---------------------
>>> + *
>>> + * This communication path is used to deliver requests from frontend to backend
>>> + * and get the corresponding responses from backend to frontend,
>>> + * set up per virtual camera device.
>>> + *
>>> + * req-event-channel
>>> + * Values: <uint32_t>
>>> + *
>>> + * The identifier of the Xen camera's control event channel
>>> + * used to signal activity in the ring buffer.
>>> + *
>>> + * req-ring-ref
>>> + * Values: <uint32_t>
>>> + *
>>> + * The Xen grant reference granting permission for the backend to map
>>> + * a sole page of camera's control ring buffer.
>>> + *
>>> + *-------------------- Camera Event Transport Parameters ----------------------
>>> + *
>>> + * This communication path is used to deliver asynchronous events from backend
>>> + * to frontend, set up per virtual camera device.
>>> + *
>>> + * evt-event-channel
>>> + * Values: <uint32_t>
>>> + *
>>> + * The identifier of the Xen camera's event channel
>>> + * used to signal activity in the ring buffer.
>>> + *
>>> + * evt-ring-ref
>>> + * Values: <uint32_t>
>>> + *
>>> + * The Xen grant reference granting permission for the backend to map
>>> + * a sole page of camera's event ring buffer.
>>> + */
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * STATE DIAGRAMS
>>> + ******************************************************************************
>>> + *
>>> + * Tool stack creates front and back state nodes with initial state
>>> + * XenbusStateInitialising.
>>> + * Tool stack creates and sets up frontend camera configuration
>>> + * nodes per domain.
>>> + *
>>> + *-------------------------------- Normal flow --------------------------------
>>> + *
>>> + * Front Back
>>> + * ================================= =====================================
>>> + * XenbusStateInitialising XenbusStateInitialising
>>> + * o Query backend device identification
>>> + * data.
>>> + * o Open and validate backend device.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateInitWait
>>> + *
>>> + * o Query frontend configuration
>>> + * o Allocate and initialize
>>> + * event channels per configured
>>> + * camera.
>>> + * o Publish transport parameters
>>> + * that will be in effect during
>>> + * this connection.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateInitialised
>>> + *
>>> + * o Query frontend transport parameters.
>>> + * o Connect to the event channels.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateConnected
>>> + *
>>> + * o Create and initialize OS
>>> + * virtual camera as per
>>> + * configuration.
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateConnected
>>> + *
>>> + * XenbusStateUnknown
>>> + * XenbusStateClosed
>>> + * XenbusStateClosing
>>> + * o Remove virtual camera device
>>> + * o Remove event channels
>>> + * |
>>> + * |
>>> + * V
>>> + * XenbusStateClosed
>>> + *
>>> + *------------------------------- Recovery flow -------------------------------
>>> + *
>>> + * In case of frontend unrecoverable errors backend handles that as
>>> + * if frontend goes into the XenbusStateClosed state.
>>> + *
>>> + * In case of backend unrecoverable errors frontend tries removing
>>> + * the virtualized device. If this is possible at the moment of error,
>>> + * then frontend goes into the XenbusStateInitialising state and is ready for
>>> + * new connection with backend. If the virtualized device is still in use and
>>> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
>>> + * until either the virtualized device is removed or backend initiates a new
>>> + * connection. On the virtualized device removal frontend goes into the
>>> + * XenbusStateInitialising state.
>>> + *
>>> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
>>> + * unrecoverable errors then frontend cannot send requests to the backend
>>> + * and thus cannot provide functionality of the virtualized device anymore.
>>> + * After backend is back to normal the virtualized device may still hold some
>>> + * state: configuration in use, allocated buffers, client application state etc.
>>> + * In most cases, this will require frontend to implement complex recovery
>>> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
>>> + * frontend will make sure no new clients of the virtualized device are
>>> + * accepted, allow existing client(s) to exit gracefully by signaling error
>>> + * state etc.
>>> + * Once all the clients are gone frontend can reinitialize the virtualized
>>> + * device and get into XenbusStateInitialising state again signaling the
>>> + * backend that a new connection can be made.
>>> + *
>>> + * There are multiple conditions possible under which frontend will go from
>>> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
>>> + * specific. For example:
>>> + * 1. The underlying OS framework may provide callbacks to signal that the last
>>> + * client of the virtualized device has gone and the device can be removed
>>> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
>>> + * to periodically check if this is the right time to re-try removal of
>>> + * the virtualized device.
>>> + * 3. By any other means.
>>> + *
>>> + ******************************************************************************
>>> + * REQUEST CODES
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_OP_SET_CONFIG 0x00
>>> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
>>> +#define XENCAMERA_OP_BUF_CREATE 0x02
>>> +#define XENCAMERA_OP_BUF_DESTROY 0x03
>>> +#define XENCAMERA_OP_STREAM_START 0x04
>>> +#define XENCAMERA_OP_STREAM_STOP 0x05
>>> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
>>> +#define XENCAMERA_OP_SET_CTRL 0x07
> I am thinking about extending the command set a bit as it already
> has some flaws, e.g. there is no way for a VM to tell the backend
> that the buffer is not in use anymore and can be given back
> to the real HW driver, e.g. queue/dequeue in V4L2 terms:
>
> #define XENCAMERA_OP_SET_FORMAT 0x00
> - will be used to set format: pixel format, resolution
>
> #define XENCAMERA_OP_SET_FRAME_RATE 0x01
> - used to set the frame rate
>
> #define XENCAMERA_OP_BUF_REQUEST 0x02
> - asks backend to allocate the given number of buffers,
> backend replies with real number of those to be used
>
> #define XENCAMERA_OP_BUF_CREATE 0x03
> - create a shared buffer
>
> #define XENCAMERA_OP_BUF_DESTROY 0x04
> - destroy a shared buffer
>
> #define XENCAMERA_OP_BUF_QUEUE 0x05
> - VM tells the backend that it has access to the shared buffer
> and the buffer cannot be sent back to real HW driver
>
> #define XENCAMERA_OP_BUF_DEQUEUE 0x06
> - VM tells the backend that the shared buffer is not in use and
> can be sent to real HW driver
This is the wrong way around: QUEUE would queue the shared buffer to
the backend for use with the real HW driver, DEQUEUE would dequeue it
for use in the VM.
>
> #define XENCAMERA_OP_CTRL_ENUM 0x07
> - get i-th control ranges and settings
>
> #define XENCAMERA_OP_CTRL_GET 0x08
> - get control value
>
> #define XENCAMERA_OP_CTRL_SET 0x09
> - set control value
>
> #define XENCAMERA_OP_STREAM_START 0x0a
> - start streaming
>
> #define XENCAMERA_OP_STREAM_STOP 0x0b
> - stop ctreaming
>
>
>>> +
>>> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
>>> +#define XENCAMERA_CTRL_CONTRAST 0x01
>>> +#define XENCAMERA_CTRL_SATURATION 0x02
>>> +#define XENCAMERA_CTRL_HUE 0x03
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * EVENT CODES
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
>>> + ******************************************************************************
>>> + */
>>> +#define XENCAMERA_DRIVER_NAME "vcamera"
>> Ah, that's where vcamera comes from. How about calling this xen-camera or
>> virt-camera? With a preference for xen-camera, since that's what you use for the
>> defines as well.
>>
>> Or perhaps pv-camera?
>>
>> Is this driver going to be xen-specific, or more a general approach that everyone
>> can use? Obviously, the latter would be preferable.
> As I have already replied to the cover letter with explanations:
> 'v' stands for 'virtual' and there is a convention to name the
> Xen virtual devices starting with 'v': vif, vkbd etc.
Yeah, ignore my comment.
>>
>> BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
>> radio and RDS hardware. Contact me if this is of interest to Xen to support this as
>> well given the automotive use-case.
> Yes, thank you, but at this stage we are targeting camera only.
> Radio can be another topic if time allows ;) And most probably
> it will be a dedicated 'vradio' protocol then...
>>
>>> +
>>> +#define XENCAMERA_LIST_SEPARATOR ","
>>> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
>>> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
>>> +
>>> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
>>> +#define XENCAMERA_FIELD_FE_VERSION "version"
>>> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
>>> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
>>> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
>>> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
>>> +#define XENCAMERA_FIELD_CONTROLS "controls"
>>> +#define XENCAMERA_FIELD_FORMATS "formats"
>>> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
>>> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
>>> +
>>> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
>>> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
>>> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
>>> +#define XENCAMERA_CTRL_HUE_STR "hue"
>>> +
>>> +/* Maximum number of buffer planes supported. */
>>> +#define XENCAMERA_MAX_PLANE 4
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * STATUS RETURN CODES
>>> + ******************************************************************************
>>> + *
>>> + * Status return code is zero on success and -XEN_EXX on failure.
>>> + *
>>> + ******************************************************************************
>>> + * Assumptions
>>> + ******************************************************************************
>>> + *
>>> + * - usage of grant reference 0 as invalid grant reference:
>>> + * grant reference 0 is valid, but never exposed to a PV driver,
>>> + * because of the fact it is already in use/reserved by the PV console.
>>> + * - all references in this document to page sizes must be treated
>>> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
>>> + *
>>> + ******************************************************************************
>>> + * Description of the protocol between frontend and backend driver
>>> + ******************************************************************************
>>> + *
>>> + * The two halves of a Para-virtual camera driver communicate with
>>> + * each other using shared pages and event channels.
>>> + * Shared page contains a ring with request/response packets.
>>> + *
>>> + * All reserved fields in the structures below must be 0.
>>> + *
>>> + * For all request/response/event packets:
>>> + * - frame rate parameter is represented as a pair of 4 octet long
>>> + * numerator and denominator:
>>> + * - frame_rate_numer - uint32_t, numerator of the frame rate
>>> + * - frame_rate_denom - uint32_t, denominator of the frame rate
>>> + * The corresponding frame rate (Hz) is calculated as:
>>> + * frame_rate = frame_rate_numer / frame_rate_denom
>>> + * - buffer index is a zero based index of the buffer. Must be less than
>>> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
>>> + * - index - uint8_t, index of the buffer.
>>> + *
>>> + *
>>> + *---------------------------------- Requests ---------------------------------
>>> + *
>>> + * All request packets have the same length (64 octets).
>>> + * All request packets have common header:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | operation | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * id - uint16_t, private guest value, echoed in response.
>>> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
>>> + *
>>> + *
>>> + * Request configuration set/reset - request to set or reset.
>>> + * the configuration/mode of the camera:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | pixel format | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | width | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | height | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_numer | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_denom | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | num_bufs | reserved | 32
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * Pass all zeros to reset, otherwise command is treated as configuration set.
>>> + *
>>> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>> + * width - uint32_t, width in pixels.
>>> + * height - uint32_t, height in pixels.
>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>> If you have to support HDMI/SDTV inputs as well, then you also need to know
>> the interlaced format, unless you have no plans to support that.
>>
>>> + * num_bufs - uint8_t, desired number of buffers to be used.
>> Huh? What has that to do with the format? Why would you need this here?
> Well, the operation name is 'set_config', not 'set_format',
> so I thought we can have such a cumulative command assembling
> all the parameters of the configuration. But now I am looking at
> turning this single 'set_config' command to 3 different commands,
> which is more practical and aligned with V4L2 in particular (please
> see above in the command set):
> 1. set format command:
> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
> * width - uint32_t, width in pixels.
> * height - uint32_t, height in pixels.
>
> 2. Set frame rate command:
> + * frame_rate_numer - uint32_t, numerator of the frame rate.
> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>
> 3. Set/request num bufs:
> * num_bufs - uint8_t, desired number of buffers to be used.
I like this much better. 1+2 could be combined, but 3 should definitely remain
separate.
>>
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + * Notes:
>>> + * - frontend must check the corresponding response in order to see
>>> + * if the values reported back by the backend do match the desired ones
>>> + * and can be accepted.
>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>> + * configuration.
>>> + */
>>> +struct xencamera_config {
>>> + uint32_t pixel_format;
>>> + uint32_t width;
>>> + uint32_t height;
>>> + uint32_t frame_rate_nom;
>>> + uint32_t frame_rate_denom;
>>> + uint8_t num_bufs;
>>> +};
>>> +
>>> +/*
>>> + * Request buffer details - request camera buffer's memory layout.
>>> + * detailed description:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + *
>>> + * Request camera buffer creation:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref_directory | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * An attempt to create multiple buffers with the same index is an error.
>>> + * index can be re-used after destroying the corresponding camera buffer.
>>> + *
>>> + * index - uint8_t, index of the buffer to be created.
>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>> + * describing shared buffer references. The size of the buffer is equal to
>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>> + * shared buffer size exceeds what can be addressed by this single page,
>>> + * then reference to the next shared page must be supplied (see
>>> + * gref_dir_next_page below).
>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>> does.
> Well, I still think it is better to have a per buffer interface
> in the protocol as it is done for other Xen virtual devices.
> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
> what it does internally in the frontend driver
I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
sense. I'm not sure about the naming.
You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
on the backend, and so can fail. Also, the actual number of allocated buffers in
case of success can be more or less than what was requested.
>>
>>> + *
>>> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
>>> + * allocate the buffer with the parameters provided in this request and page
>>> + * directory is handled as follows:
>>> + * Frontend on request:
>>> + * - allocates pages for the directory (gref_directory,
>>> + * gref_dir_next_page(s)
>>> + * - grants permissions for the pages of the directory to the backend
>>> + * - sets gref_dir_next_page fields
>>> + * Backend on response:
>>> + * - grants permissions for the pages of the buffer allocated to
>>> + * the frontend
>>> + * - fills in page directory with grant references
>>> + * (gref[] in struct xencamera_page_directory)
>>> + */
>>> +struct xencamera_buf_create_req {
>>> + uint8_t index;
>>> + uint8_t reserved[3];
>>> + grant_ref_t gref_directory;
>>> +};
>>> +
>>> +/*
>>> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
>>> + * the request) employs a list of pages, describing all pages of the shared
>>> + * data buffer:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref_dir_next_page | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref[0] | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref[i] | i*4+8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | gref[N - 1] | N*4+8
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
>>> + * page directory. Must be 0 if there are no more pages in the list.
>>> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
>>> + * allocated at XENCAMERA_OP_BUF_CREATE.
>>> + *
>>> + * Number of grant_ref_t entries in the whole page directory is not
>>> + * passed, but instead can be calculated as:
>>> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
>>> + * XEN_PAGE_SIZE
>>> + */
>>> +struct xencamera_page_directory {
>>> + grant_ref_t gref_dir_next_page;
>>> + grant_ref_t gref[1]; /* Variable length */
>>> +};
>>> +
>>> +/*
>>> + * Request buffer destruction - destroy a previously allocated camera buffer:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_BUF_DESTROY| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * index - uint8_t, index of the buffer to be destroyed.
>>> + */
>> There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
>> of them.
> This is not specifically related to V4L2, but can be issued
> in response to backend's state change etc.
> So, we have a pair of commands to create and destroy buffers.
> Even more, frontend can be a some-os-based-driver, not V4L2
> based. Or even a user-space application if your will.
>>
>>> +
>>> +struct xencamera_buf_destroy_req {
>>> + uint8_t index;
>>> +};
>>> +
>>> +/*
>>> + * Request camera capture stream start:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_OP_STREAM_START| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + *
>>> + * Request camera capture stream stop:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_OP_STREAM_STOP | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + *
>>> + * Request camera control details:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + * index - uint8_t, index of the control to be queried.
>>> + */
>>> +struct xencamera_get_ctrl_details_req {
>>> + uint8_t index;
>>> +};
>>> +
>>> +/*
>>> + *
>>> + * Request camera control change:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_SET_CTRL | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | value | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * See response format for this request.
>>> + *
>>> + * index - uint8_t, index of the control.
>>> + * value - int32_t, new value of the control.
>> I would recommend using a int64_t as the control value.
> Good point, thank you
>>
>> Note that there are also controls with a payload (e.g. string controls).
> Could you please give me an example of such a control?
> Do you think such controls can be of use in a VM?
> Can we avoid such controls if we target a simple virtual
> camera device? If this is for radio use-case, then we'll
> have such support in 'vradio' protocol if need be
Right now all string controls are related to RDS receivers/transmitters. If you
ever decide on a vradio protocol, then you need these (Programme Service name
and Radio Text info). And there is an array of Alternate Frequencies, also RDS
specific.
There are some array controls in V4L2, those are used to control motion detection
for surveillance cameras. And 'compound controls' (think of this as C structs) are
appearing for HW codecs.
I don't think any of these are likely to appear for cameras, at least not in a
way that is relevant for Xen.
>>
>> If there is ever interest in adding radio/RDS support, then that will become
>> an issue.
> You mean something like station names, ads etc?
Yup.
>>
>>> + */
>>> +struct xencamera_set_ctrl_req {
>>> + uint8_t index;
>>> + uint8_t reserved[3];
>>> + int32_t value;
>>> +};
>>> +
>>> +/*
>>> + *---------------------------------- Responses --------------------------------
>>> + *
>>> + * All response packets have the same length (64 octets).
>>> + *
>>> + * All response packets have common header:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | operation | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * id - uint16_t, copied from the request.
>>> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
>>> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
>>> + *
>>> + *
>>> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | pixel format | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | width | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | height | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_numer | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | frame_rate_denom | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | num_bufs | reserved | 32
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * Meaning of the corresponding values in this response is the same as for
>>> + * XENCAMERA_OP_SET_CONFIG request.
>>> + *
>>> + *
>>> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
>>> + * request:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | size | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | num_planes | reserved | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[0] | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[1] | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[2] | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_offset[3] | 32
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[0] | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[1] | 40
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[2] | 44
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_size[3] | 48
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_stride[0] | plane_stride[1] | 52
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | plane_stride[2] | plane_stride[3] | 56
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 60
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * size - uint32_t, overall size of the buffer including sizes of the
>>> + * individual planes and padding if applicable.
>>> + * num_planes - uint8_t, number of planes for this buffer.
>>> + * plane_offset - array of uint32_t, offset of the corresponding plane
>>> + * in octets from the buffer start.
>>> + * plane_size - array of uint32_t, size in octets of the corresponding plane
>>> + * including padding.
>>> + * plane_stride - array of uint32_t, size in octets occupied by the
>>> + * corresponding single image line including padding if applicable.
>> Nice!
> Thank you
>>> + */
>>> +struct xencamera_buf_details_resp {
>>> + uint32_t size;
>>> + uint8_t num_planes;
>>> + uint8_t reserved[3];
>>> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
>>> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
>>> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
>>> +};
>>> +
>>> +/*
>>> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | status | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | type | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | min | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | max | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | step | 24
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | def_val | 28
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 36
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * index - uint8_t, index of the camera control in response.
>>> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
>>> + * min - int32_t, minimum value of the control.
>>> + * max - int32_t, maximum value of the control.
>>> + * step - int32_t, minimum size in which control value can be changed.
>>> + * def_val - int32_t, default value of the control.
>> I'd go with 64 bit values for min/max/step/def_val.
> Sure, good idea, thank you
>>
>> I would also add a flags field. Some controls are read-only, write-only
>> or volatile, things userspace needs to know.
> Then I'll also add numerical constants for such
>>
>> If you want to support menu controls, then you need a way to get the menu
>> names as well (VIDIOC_QUERYMENU).
>>
>> None of this is needed for this initial use-case, but you need to think
>> about this up-front.
> Yes, thank you
>>> + */
>>> +struct xencamera_get_ctrl_details_resp {
>>> + uint8_t index;
>>> + uint8_t type;
>>> + uint8_t reserved[2];
>>> + int32_t min;
>>> + int32_t max;
>>> + int32_t step;
>>> + int32_t def_val;
>>> +};
>>> +
>>> +/*
>>> + *----------------------------------- Events ----------------------------------
>>> + *
>>> + * Events are sent via a shared page allocated by the front and propagated by
>>> + * evt-event-channel/evt-ring-ref XenStore entries.
>>> + *
>>> + * All event packets have the same length (64 octets).
>>> + * All event packets have common header:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id | type | reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * id - uint16_t, event id, may be used by front.
>>> + * type - uint8_t, type of the event.
>>> + *
>>> + *
>>> + * Frame captured event - event from back to front when a new captured
>>> + * frame is available:
>>> + * 0 1 2 3 octet
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | id |_EVT_FRAME_AVAIL| reserved | 4
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 8
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | index | reserved | 12
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | used_sz | 16
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 20
>>> + * +----------------+----------------+----------------+----------------+
>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>> + * +----------------+----------------+----------------+----------------+
>>> + * | reserved | 64
>>> + * +----------------+----------------+----------------+----------------+
>>> + *
>>> + * index - uint8_t, index of the buffer that contains new captured frame.
>>> + * used_sz - uint32_t, number of octets this frame has. This can be less
>>> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
>>> + */
>>> +struct xencamera_frame_avail_evt {
>>> + uint8_t index;
>>> + uint8_t reserved[3];
>>> + uint32_t used_sz;
>>> +};
>>> +
>>> +struct xencamera_req {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved[5];
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_create_req buf_create;
>>> + struct xencamera_buf_destroy_req buf_destroy;
>>> + struct xencamera_set_ctrl_req set_ctrl;
>>> + uint8_t reserved[56];
>>> + } req;
>>> +};
>>> +
>>> +struct xencamera_resp {
>>> + uint16_t id;
>>> + uint8_t operation;
>>> + uint8_t reserved;
>>> + int32_t status;
>>> + union {
>>> + struct xencamera_config config;
>>> + struct xencamera_buf_details_resp buf_details;
>>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>>> + uint8_t reserved1[56];
>>> + } resp;
>>> +};
>>> +
>>> +struct xencamera_evt {
>>> + uint16_t id;
>>> + uint8_t type;
>>> + uint8_t reserved[5];
>>> + union {
>>> + struct xencamera_frame_avail_evt frame_avail;
>>> + uint8_t reserved[56];
>>> + } evt;
>>> +};
>>> +
>>> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
>>> +
>>> +/*
>>> + ******************************************************************************
>>> + * Back to front events delivery
>>> + ******************************************************************************
>>> + * In order to deliver asynchronous events from back to front a shared page is
>>> + * allocated by front and its granted reference propagated to back via
>>> + * XenStore entries (evt-ring-ref/evt-event-channel).
>>> + * This page has a common header used by both front and back to synchronize
>>> + * access and control event's ring buffer, while back being a producer of the
>>> + * events and front being a consumer. The rest of the page after the header
>>> + * is used for event packets.
>>> + *
>>> + * Upon reception of an event(s) front may confirm its reception
>>> + * for either each event, group of events or none.
>>> + */
>>> +
>>> +struct xencamera_event_page {
>>> + uint32_t in_cons;
>>> + uint32_t in_prod;
>>> + uint8_t reserved[56];
>>> +};
>>> +
>>> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
>>> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
>>> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
>>> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
>>> +#define XENCAMERA_IN_RING(page) \
>>> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
>>> +#define XENCAMERA_IN_RING_REF(page, idx) \
>>> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
>>> +
>>> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
>>> +
>>> +/*
>>> + * Local variables:
>>> + * mode: C
>>> + * c-file-style: "BSD"
>>> + * c-basic-offset: 4
>>> + * tab-width: 4
>>> + * indent-tabs-mode: nil
>>> + * End:
>>> + */
>>>
>> I think the most important decision to make here is whether or not you want to support
>> hotpluggable sources like HDMI. And an additional complication with that is HDCP.
>> While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
>> a patch series adding this is expected later this year/early next year. It might not
>> be an issue in practice if these are all closed systems, but nevertheless, it is
>> something to think about.
> Yes, thank you for raising these questions, it is worth thinking
> about such use-cases.
>>
>> Regards,
>>
>> Hans
> Thank you so much for the comments,
> Oleksandr
>
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 5:59 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-10 8:14 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 8:14 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko, Artem Mygaiev
On 09/10/2018 07:59 AM, Oleksandr Andrushchenko wrote:
> Hi, Hans!
>
> On 09/09/2018 01:42 PM, Hans Verkuil wrote:
>> On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
>>> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>>>> Hi Oleksandr,
>>>>
>>>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>>>
>>>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>>>> high definition maps etc.
>>>>>>>>
>>>>>>>> The initial goal is to support most needed functionality with the
>>>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>>>
>>>>>>>> 1. Provide means for base virtual device configuration:
>>>>>>>> - pixel formats
>>>>>>>> - resolutions
>>>>>>>> - frame rates
>>>>>>>> 2. Support basic camera controls:
>>>>>>>> - contrast
>>>>>>>> - brightness
>>>>>>>> - hue
>>>>>>>> - saturation
>>>>>>>> 3. Support streaming control
>>>>>>>> 4. Support zero-copying use-cases
>>>>>>>>
>>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>>>> Some style issues below...
>>>>>> Will fix all the below, thank you!
>>>>>>
>>>>>> I would like to draw some attention of the Linux/V4L community to this
>>>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>>>> upstream a Linux camera front-end kernel driver which will be based
>>>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>>>> as well)
>>>>> ping
>>>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>>>> to review this this week, if not, just ping me again.
>>> Thank you for your time
>>>> I had one high-level question, though:
>>>>
>>>> What types of hardware do you intend to target? This initial version targets
>>>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>>>> codecs? Or complex embedded video pipelines?
>>>>
>>>> In other words, where are you planning to draw the line?
>>>>
>>>> Even with just simple cameras there is a difference between regular UVC
>>>> webcams and cameras used with embedded systems: for the latter you often
>>>> need to provide more control w.r.t. white-balancing etc., things that a
>>>> UVC webcam will generally do for you in the webcam's firmware.
>>> The use-cases we want to implement are mostly in automotive/embedded domain,
>>> so there are many performance restrictions apply.
>>> We are not targeting virtualizing very complex hardware and have no
>>> intention
>>> to make a 1:1 mapping of the real hardware: for that one can pass-through
>>> a real HW device to a virtual machine (VM). The goal is to share a single
>>> camera device to multiple virtual machines, no codecs, receivers etc.
>>>
>>> Controlling the same HW device from different VMs doesn't look feasible:
>>> what if the same control is set to different values from different VMs?
>> You can do this, actually: in V4L2 you can get an event when another process
>> changes a control, and update your own GUI/internal state accordingly.
>>
>> So in this case if one VM changes a control, an event is sent to all others
>> that the control has changed value.
> Well, technically this can be done by introducing one more
> event for such a notification. But, from system partitioning
> POV, I am still not convinced this should be done: I would prefer
> that a single VM owns such a control and even which control and which
> VM is decided while configuring the whole system.
> So, I would like to keep it as is.
Well, I am not convinced you can avoid this: some controls are set by drivers
when something happens and so applications will subscribe to the control and
wait for changes in their values. While this is for more advanced use-cases
(certainly more than what you propose today), I would suggest that it is wise
to at least think on how this can be added in the future.
Controls and control events are a key part of V4L2.
>>
>>> Of course, this can be achieved if the corresponding backend can
>>> post-process
>>> original camera image with GPU, for example, thus applying different filters
>>> for different VMs effectively emulating camera controls.
>>> But this requires additional CPU/GPU power which we try to avoid.
>>>
>>> System partitioning (camera and controls assignment) is done at
>>> configuration
>>> time (remember we are in automotive/embedded world, so most of the time
>>> the set
>>> of VMs requiring cameras is known at this stage and the configuration
>>> remains
>>> static at run-time). So, when para-virtualized (PV) approach is used then we
>>> only implement very basic controls (those found in the protocol), so one can
>>> assign set of controls (all or some) to one of the VMs (main or mission
>>> critical
>>> VM or whatever) allowing that VM to adjusts those for all VMs at once.
>>> For other
>>> VMs think of it as firmware implemented adjustment. And the backend still
>>> controls the rest of the controls of the real HW camera you mention.
>>>
>>> Just an example of automotive use-case (we can imagine many more):
>>> 1. Driver Domain - owns real camera HW and runs the camera backend.
>>> Uses camera output for mission critical tasks, e.g. parking assistance.
>>> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
>>> purposes,
>>> e.g. taking pictures while in motion.
>>> 3. Navigation domain - uses PV camera for high definition maps
>>>
>>> Hope, this helps understanding the possible uses of the proposed
>>> protocol, its
>>> intention and restrictions.
>> Right, so in this scenario you probably do not want hotpluggable
>> sources in the Driver Domain. So support for fixed camera's only.
> Well, some sort of hotplug can already be implemented, please
> see [1], [2] as it is done for virtual display: this is
> achieved as a response to the backend's state change,
> e.g. whenever backend decides to unplug the virtual device
> it changes its state accordingly.
>>
>> If this is indeed the case, then this should be made very clear in
>> the API specification.
> As I described above this is already assumed by the state
> machine of a xenbus_driver
>> One additional thing to consider: cameras can break. So what should be
>> done if that happens? We as media developers have ideas about that, but
>> nothing has been implemented (yet).
>>
>> If the HW is simple (one camera is driven by a single driver instance),
>> then if it breaks, there simply won't be a video device. But if you have
>> multiple cameras all controlled through the same driver instance, then today
>> if a single camera breaks, all are gone.
> Please see above
I'm not convinced this is sufficient. But it really would have to be tested
with an actual HDMI source.
A key difference between supporting a video input connector such as HDMI
compared to a fixed camera or video output is that you do not control what
you receive. I've been working with this for many years, and it is amazing
what problems you can get due to this.
So received frames can be corrupt (make sure you can handle the
V4L2_BUF_FLAG_ERROR flag as result of a VIDIOC_DQBUF!) or the DMA engine can get
into a bad state, requiring you to stop streaming and reconfigure (EIO returned
by VIDIOC_(D)QBUF).
A good driver to test such errors with is vivid, which can emulate webcams, SDTV or
an HDMI receiver and has error injection to test these scenarios. See here for
extensive documentation of this driver:
https://hverkuil.home.xs4all.nl/spec/v4l-drivers/vivid.html
I would recommend that you use this driver as a test HW driver since it can
be configured in many different ways and is representative of actual HW.
Exception is for HDMI where connect/disconnect is not properly emulated yet.
If you need that, let me know.
Regards,
Hans
>> We have ideas on how to address that, but as I said, nothing is implemented
>> yet. Basically we need to allow for partial bring-up and inform userspace
>> what is and what is not running.
>>
>> But this is likely something you also need to consider in this API, given
>> the use-case you are looking at.
>>
>> Regards,
>>
>> Hans
> Thank you for your valuable comments,
> Oleksandr
>
> [1]
> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L721
> [2]
> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L582
>
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 8:14 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 8:14 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Juergen Gross, xen-devel, konrad.wilk,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Artem Mygaiev, Oleksandr Andrushchenko
On 09/10/2018 07:59 AM, Oleksandr Andrushchenko wrote:
> Hi, Hans!
>
> On 09/09/2018 01:42 PM, Hans Verkuil wrote:
>> On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
>>> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>>>> Hi Oleksandr,
>>>>
>>>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>>>
>>>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>>>> high definition maps etc.
>>>>>>>>
>>>>>>>> The initial goal is to support most needed functionality with the
>>>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>>>
>>>>>>>> 1. Provide means for base virtual device configuration:
>>>>>>>> - pixel formats
>>>>>>>> - resolutions
>>>>>>>> - frame rates
>>>>>>>> 2. Support basic camera controls:
>>>>>>>> - contrast
>>>>>>>> - brightness
>>>>>>>> - hue
>>>>>>>> - saturation
>>>>>>>> 3. Support streaming control
>>>>>>>> 4. Support zero-copying use-cases
>>>>>>>>
>>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>>>> Some style issues below...
>>>>>> Will fix all the below, thank you!
>>>>>>
>>>>>> I would like to draw some attention of the Linux/V4L community to this
>>>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>>>> upstream a Linux camera front-end kernel driver which will be based
>>>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>>>> as well)
>>>>> ping
>>>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>>>> to review this this week, if not, just ping me again.
>>> Thank you for your time
>>>> I had one high-level question, though:
>>>>
>>>> What types of hardware do you intend to target? This initial version targets
>>>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>>>> codecs? Or complex embedded video pipelines?
>>>>
>>>> In other words, where are you planning to draw the line?
>>>>
>>>> Even with just simple cameras there is a difference between regular UVC
>>>> webcams and cameras used with embedded systems: for the latter you often
>>>> need to provide more control w.r.t. white-balancing etc., things that a
>>>> UVC webcam will generally do for you in the webcam's firmware.
>>> The use-cases we want to implement are mostly in automotive/embedded domain,
>>> so there are many performance restrictions apply.
>>> We are not targeting virtualizing very complex hardware and have no
>>> intention
>>> to make a 1:1 mapping of the real hardware: for that one can pass-through
>>> a real HW device to a virtual machine (VM). The goal is to share a single
>>> camera device to multiple virtual machines, no codecs, receivers etc.
>>>
>>> Controlling the same HW device from different VMs doesn't look feasible:
>>> what if the same control is set to different values from different VMs?
>> You can do this, actually: in V4L2 you can get an event when another process
>> changes a control, and update your own GUI/internal state accordingly.
>>
>> So in this case if one VM changes a control, an event is sent to all others
>> that the control has changed value.
> Well, technically this can be done by introducing one more
> event for such a notification. But, from system partitioning
> POV, I am still not convinced this should be done: I would prefer
> that a single VM owns such a control and even which control and which
> VM is decided while configuring the whole system.
> So, I would like to keep it as is.
Well, I am not convinced you can avoid this: some controls are set by drivers
when something happens and so applications will subscribe to the control and
wait for changes in their values. While this is for more advanced use-cases
(certainly more than what you propose today), I would suggest that it is wise
to at least think on how this can be added in the future.
Controls and control events are a key part of V4L2.
>>
>>> Of course, this can be achieved if the corresponding backend can
>>> post-process
>>> original camera image with GPU, for example, thus applying different filters
>>> for different VMs effectively emulating camera controls.
>>> But this requires additional CPU/GPU power which we try to avoid.
>>>
>>> System partitioning (camera and controls assignment) is done at
>>> configuration
>>> time (remember we are in automotive/embedded world, so most of the time
>>> the set
>>> of VMs requiring cameras is known at this stage and the configuration
>>> remains
>>> static at run-time). So, when para-virtualized (PV) approach is used then we
>>> only implement very basic controls (those found in the protocol), so one can
>>> assign set of controls (all or some) to one of the VMs (main or mission
>>> critical
>>> VM or whatever) allowing that VM to adjusts those for all VMs at once.
>>> For other
>>> VMs think of it as firmware implemented adjustment. And the backend still
>>> controls the rest of the controls of the real HW camera you mention.
>>>
>>> Just an example of automotive use-case (we can imagine many more):
>>> 1. Driver Domain - owns real camera HW and runs the camera backend.
>>> Uses camera output for mission critical tasks, e.g. parking assistance.
>>> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
>>> purposes,
>>> e.g. taking pictures while in motion.
>>> 3. Navigation domain - uses PV camera for high definition maps
>>>
>>> Hope, this helps understanding the possible uses of the proposed
>>> protocol, its
>>> intention and restrictions.
>> Right, so in this scenario you probably do not want hotpluggable
>> sources in the Driver Domain. So support for fixed camera's only.
> Well, some sort of hotplug can already be implemented, please
> see [1], [2] as it is done for virtual display: this is
> achieved as a response to the backend's state change,
> e.g. whenever backend decides to unplug the virtual device
> it changes its state accordingly.
>>
>> If this is indeed the case, then this should be made very clear in
>> the API specification.
> As I described above this is already assumed by the state
> machine of a xenbus_driver
>> One additional thing to consider: cameras can break. So what should be
>> done if that happens? We as media developers have ideas about that, but
>> nothing has been implemented (yet).
>>
>> If the HW is simple (one camera is driven by a single driver instance),
>> then if it breaks, there simply won't be a video device. But if you have
>> multiple cameras all controlled through the same driver instance, then today
>> if a single camera breaks, all are gone.
> Please see above
I'm not convinced this is sufficient. But it really would have to be tested
with an actual HDMI source.
A key difference between supporting a video input connector such as HDMI
compared to a fixed camera or video output is that you do not control what
you receive. I've been working with this for many years, and it is amazing
what problems you can get due to this.
So received frames can be corrupt (make sure you can handle the
V4L2_BUF_FLAG_ERROR flag as result of a VIDIOC_DQBUF!) or the DMA engine can get
into a bad state, requiring you to stop streaming and reconfigure (EIO returned
by VIDIOC_(D)QBUF).
A good driver to test such errors with is vivid, which can emulate webcams, SDTV or
an HDMI receiver and has error injection to test these scenarios. See here for
extensive documentation of this driver:
https://hverkuil.home.xs4all.nl/spec/v4l-drivers/vivid.html
I would recommend that you use this driver as a test HW driver since it can
be configured in many different ways and is representative of actual HW.
Exception is for HDMI where connect/disconnect is not properly emulated yet.
If you need that, let me know.
Regards,
Hans
>> We have ideas on how to address that, but as I said, nothing is implemented
>> yet. Basically we need to allow for partial bring-up and inform userspace
>> what is and what is not running.
>>
>> But this is likely something you also need to consider in this API, given
>> the use-case you are looking at.
>>
>> Regards,
>>
>> Hans
> Thank you for your valuable comments,
> Oleksandr
>
> [1]
> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L721
> [2]
> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L582
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 7:53 ` [PATCH " Hans Verkuil
@ 2018-09-10 8:24 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 8:24 UTC (permalink / raw)
To: Hans Verkuil, xen-devel, konrad.wilk, jgross, boris.ostrovsky,
mchehab, linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 09/10/2018 10:53 AM, Hans Verkuil wrote:
> Hi Oleksandr,
>
> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>> Hi, Hans!
>>
>> On 09/09/2018 01:31 PM, Hans Verkuil wrote:
>>> Hi Oleksandr,
>>>
>>> Sorry for the delay in reviewing, I missed this patch until you pinged me, and
>>> I was very busy after that as well.
>> I do appreciate you spending time on this!
>>> On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>
>>>> This is the ABI for the two halves of a para-virtualized
>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>> high definition maps etc.
>>>>
>>>> The initial goal is to support most needed functionality with the
>>>> final idea to make it possible to extend the protocol if need be:
>>>>
>>>> 1. Provide means for base virtual device configuration:
>>>> - pixel formats
>>>> - resolutions
>>>> - frame rates
>>>> 2. Support basic camera controls:
>>>> - contrast
>>>> - brightness
>>>> - hue
>>>> - saturation
>>>> 3. Support streaming control
>>>> 4. Support zero-copying use-cases
>>>>
>>>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>> ---
>>>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>>>> 1 file changed, 981 insertions(+)
>>>> create mode 100644 xen/include/public/io/cameraif.h
>>>>
>>>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>>>> new file mode 100644
>>>> index 000000000000..bdc6a1262fcf
>>>> --- /dev/null
>>>> +++ b/xen/include/public/io/cameraif.h
>>>> @@ -0,0 +1,981 @@
>>>> +/******************************************************************************
>>>> + * cameraif.h
>>>> + *
>>>> + * Unified camera device I/O interface for Xen guest OSes.
>>>> + *
>>>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>>>> + * of this software and associated documentation files (the "Software"), to
>>>> + * deal in the Software without restriction, including without limitation the
>>>> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
>>>> + * sell copies of the Software, and to permit persons to whom the Software is
>>>> + * furnished to do so, subject to the following conditions:
>>>> + *
>>>> + * The above copyright notice and this permission notice shall be included in
>>>> + * all copies or substantial portions of the Software.
>>>> + *
>>>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>>>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>>>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>>>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>>>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>>>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>>>> + * DEALINGS IN THE SOFTWARE.
>>>> + *
>>>> + * Copyright (C) 2018 EPAM Systems Inc.
>>>> + *
>>>> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>> + */
>>> Use SPDX tag instead of copying the license text.
>> This is yet a Xen header which belongs to Xen project and
>> all the rest of the protocols have the same license header.
>> If Xen community decides to use SPDX then I'll definitely follow.
> Ah, yes, I was reviewing this as a kernel header, I hadn't realized
> that this isn't a kernel header. Since it isn't a kernel header, you
> can disregard my comments about style and naming conventions, since
> you have your own.
>
>> Konrad, do you think this is the right time for such a move?
>>>> +
>>>> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
>>>> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
>>>> +
>>>> +#include "ring.h"
>>>> +#include "../grant_table.h"
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * Protocol version
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_PROTOCOL_VERSION "1"
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * Feature and Parameter Negotiation
>>>> + ******************************************************************************
>>>> + *
>>>> + * Front->back notifications: when enqueuing a new request, sending a
>>>> + * notification can be made conditional on xencamera_req (i.e., the generic
>>>> + * hold-off mechanism provided by the ring macros). Backends must set
>>>> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
>>>> + *
>>>> + * Back->front notifications: when enqueuing a new response, sending a
>>>> + * notification can be made conditional on xencamera_resp (i.e., the generic
>>>> + * hold-off mechanism provided by the ring macros). Frontends must set
>>>> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
>>>> + *
>>>> + * The two halves of a para-virtual camera driver utilize nodes within
>>>> + * XenStore to communicate capabilities and to negotiate operating parameters.
>>>> + * This section enumerates these nodes which reside in the respective front and
>>>> + * backend portions of XenStore, following the XenBus convention.
>>>> + *
>>>> + * All data in XenStore is stored as strings. Nodes specifying numeric
>>>> + * values are encoded in decimal. Integer value ranges listed below are
>>>> + * expressed as fixed sized integer types capable of storing the conversion
>>>> + * of a properly formatted node string, without loss of information.
>>>> + *
>>>> + ******************************************************************************
>>>> + * Example configuration
>>>> + ******************************************************************************
>>>> + *
>>>> + * This is an example of backend and frontend configuration:
>>>> + *
>>>> + *--------------------------------- Backend -----------------------------------
>>>> + *
>>>> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
>>>> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
>>>> + * /local/domain/0/backend/vcamera/1/0/state = "4"
>>>> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
>>> Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
>>> superfluous to me.
>> 'v' stands for 'virtual'. I am following Xen convention used
>> for all other virtual device protocols here.
>>>> + *
>>>> + *--------------------------------- Frontend ----------------------------------
>>>> + *
>>>> + * /local/domain/1/device/vcamera/0/backend-id = "0"
>>>> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
>>>> + * /local/domain/1/device/vcamera/0/state = "4"
>>>> + * /local/domain/1/device/vcamera/0/version = "1"
>>>> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
>>>> + *
>>>> + *---------------------------- Device 0 configuration -------------------------
>>>> + *
>>>> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
>>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
>>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
>>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
>>>> + * /local/domain/1/device/vcamera/0/unique-id = "0"
>>>> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
>>>> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
>>>> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
>>>> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
>>>> + *
>>>> + *---------------------------- Device 1 configuration -------------------------
>>>> + *
>>>> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
>>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
>>>> + * /local/domain/1/device/vcamera/1/unique-id = "1"
>>>> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
>>>> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
>>>> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
>>>> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
>>>> + *
>>>> + ******************************************************************************
>>>> + * Backend XenBus Nodes
>>>> + ******************************************************************************
>>>> + *
>>>> + *----------------------------- Protocol version ------------------------------
>>>> + *
>>>> + * versions
>>>> + * Values: <string>
>>>> + *
>>>> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
>>>> + * by the backend. For example "1,2,3".
>>>> + *
>>>> + ******************************************************************************
>>>> + * Frontend XenBus Nodes
>>>> + ******************************************************************************
>>>> + *
>>>> + *-------------------------------- Addressing ---------------------------------
>>>> + *
>>>> + * dom-id
>>>> + * Values: <uint16_t>
>>>> + *
>>>> + * Domain identifier.
>>>> + *
>>>> + * dev-id
>>>> + * Values: <uint16_t>
>>>> + *
>>>> + * Device identifier.
>>>> + *
>>>> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
>>>> + *
>>>> + *----------------------------- Protocol version ------------------------------
>>>> + *
>>>> + * version
>>>> + * Values: <string>
>>>> + *
>>>> + * Protocol version, chosen among the ones supported by the backend.
>>>> + *
>>>> + *------------------------- Backend buffer allocation -------------------------
>>>> + *
>>>> + * be-alloc
>>> I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
>>> is better to just write 'backend-alloc'.
>> Well, 'be' and 'fe' are commonly used in Xen for backend
>> and frontend, so I'll probably stick to that convention for now.
>>>> + * Values: "0", "1"
>>>> + *
>>>> + * If value is set to "1", then backend will be the buffer
>>>> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
>>>> + * operation.
>>>> + * If value is not "1" or omitted frontend must allocate buffers itself.
>>>> + *
>>>> + *------------------------------- Camera settings -----------------------------
>>>> + *
>>>> + * unique-id
>>>> + * Values: <string>
>>>> + *
>>>> + * After device instance initialization each camera is assigned a
>>>> + * unique ID, so it can be identified by the backend by this ID.
>>>> + * This can be UUID or such.
>>>> + *
>>>> + * controls
>>>> + * Values: <list of string>
>>>> + *
>>>> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
>>>> + * Camera controls are expressed as a list of string values w/o any
>>>> + * ordering requirement.
>>>> + *
>>>> + * formats
>>>> + * Values: <format, char[4]>
>>>> + *
>>>> + * Formats are organized as a set of directories one per each
>>>> + * supported pixel format. The name of the directory is an upper case
>>>> + * string of the corresponding FOURCC string label. The next level of
>>>> + * the directory under <formats> represents supported resolutions.
>>> Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
>> Ok, no problem - will remove the 'upper case' from the definition
>>> In addition it is common to set bit 31 of the fourcc to 1 if the format is
>>> big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
>>> add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
>>> keep that convention.
>> I'll think about it, thank you
>>>> + *
>>>> + * resolution
>>>> + * Values: <width, uint32_t>x<height, uint32_t>
>>>> + *
>>>> + * Resolutions are organized as a set of directories one per each
>>>> + * supported resolution under corresponding <formats> directory.
>>>> + * The name of the directory is the supported width and height
>>>> + * of the camera resolution in pixels.
>>> What if you are dealing with an HDMI input? Not unreasonable for media
>>> systems. There can be a lot of resolutions/framerates, and the resolution
>>> can change on the fly, or of course disappear.
>> Well, this is a part of the system configuration done
>> before we actually run the VMs, e.g. at system design
>> and configuration time. Most of the time you do know which
>> resolutions, frame rates etc. you want to assign and these
>> settings remain static for the whole lifetime of that VM.
>> If you are designing a system which needs these resolutions
>> to change at run-time then this can be done:
>> 1. Backend changes the state of the frontend to XenbusStateClosed state
>> 2. Xen (xl/libxl) or backend change the configuration in XenStore
>> 3. Backend re-initializes the frontend which reads new configuration
>>> What is also missing here is a way to report pixel aspect ratio: PAL and
>>> NTSC-based video material doesn't have square pixels.
>> Hm, indeed, thank you. I'll put this as a fraction under
>> the corresponding 'resolution':
>>
>> Now:
>> /local/domain/2/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024 = "015/2"
>>
>> Will change to:
>> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/framerates/ =
>> "30/1,15/1,15/2"
>> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/aspectratio = "1/1"
> You need to be more precise: aspectratio can refer to the display aspect
> ratio or the pixel aspect ratio, and you can calculate one from the other.
>
> So this is either pixelaspectratio or displayaspectratio.
>
> Even though V4L2 provides a pixelaspectratio, I would recommend that you
> choose to export it as a displayaspectratio. It's easier for applications
> to handle.
ok, then "display-aspect-ratio"
>
> Please note that in the case of DVD/BluRay source material the same
> resolution can have different display aspect ratios: e.g. PAL can be
> 4:3 or 16:9, depending on whether it is widescreen or not.
>
ok
>> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/framerates = "015/2"
>> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/aspectratio =
>> "59/58"
>>
>>> It's important to decide whether or not you want to support video sources
>>> like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
>>> or disconnect webcams) or if you stick to fixed camera pipelines.
>> I believe that this is all hidden from the frontend by the
>> backend, so I se nothing we have to put in the protocol
>> with this respect.
>>> The big difference is that you don't control what someone can connect as
>>> external sources, so you will have to be a lot more careful and robust.
>>>
>>> I suspect that you likely will want to support such sources eventually, so
>>> it pays to design this with that in mind.
>> Again, I think that this is the backend to hide these
>> use-cases from the frontend.
> I'm not sure you can: say you are playing a bluray connected to the system
> with HDMI, then if there is a resolution change, what do you do? You can tear
> everything down and build it up again, or you can just tell frontends that
> something changed and that they have to look at the new vcamera configuration.
>
> The latter seems to be more sensible to me. It is really not much that you
> need to do: all you really need is an event signalling that something changed.
> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
well, this complicates things a lot as I'll have to
re-allocate buffers - right?
But anyways, I can add
#define XENCAMERA_EVT_CFG_CHANGE 0x01
in the protocol, so we can address this use-case
>
>>>> + *
>>>> + * frame-rates
>>>> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
>>>> + *
>>>> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
>>>> + * of the camera expressed as numerator and denominator of the
>>>> + * corresponding frame rate.
>>>> + *
>>>> + * The format of the <formats> directory tree with resolutions and frame rates
>>>> + * must be structured in the following format:
>>>> + *
>>>> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
>>>> + *
>>>> + * where
>>>> + * i - i-th supported pixel format
>>>> + * j - j-th supported resolution for i-th pixel format
>>>> + * k - k-th supported frame rate for i-th pixel format and j-th
>>>> + * resolution> + *
>>>> + *------------------- Camera Request Transport Parameters ---------------------
>>>> + *
>>>> + * This communication path is used to deliver requests from frontend to backend
>>>> + * and get the corresponding responses from backend to frontend,
>>>> + * set up per virtual camera device.
>>>> + *
>>>> + * req-event-channel
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The identifier of the Xen camera's control event channel
>>>> + * used to signal activity in the ring buffer.
>>>> + *
>>>> + * req-ring-ref
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The Xen grant reference granting permission for the backend to map
>>>> + * a sole page of camera's control ring buffer.
>>>> + *
>>>> + *-------------------- Camera Event Transport Parameters ----------------------
>>>> + *
>>>> + * This communication path is used to deliver asynchronous events from backend
>>>> + * to frontend, set up per virtual camera device.
>>>> + *
>>>> + * evt-event-channel
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The identifier of the Xen camera's event channel
>>>> + * used to signal activity in the ring buffer.
>>>> + *
>>>> + * evt-ring-ref
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The Xen grant reference granting permission for the backend to map
>>>> + * a sole page of camera's event ring buffer.
>>>> + */
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * STATE DIAGRAMS
>>>> + ******************************************************************************
>>>> + *
>>>> + * Tool stack creates front and back state nodes with initial state
>>>> + * XenbusStateInitialising.
>>>> + * Tool stack creates and sets up frontend camera configuration
>>>> + * nodes per domain.
>>>> + *
>>>> + *-------------------------------- Normal flow --------------------------------
>>>> + *
>>>> + * Front Back
>>>> + * ================================= =====================================
>>>> + * XenbusStateInitialising XenbusStateInitialising
>>>> + * o Query backend device identification
>>>> + * data.
>>>> + * o Open and validate backend device.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateInitWait
>>>> + *
>>>> + * o Query frontend configuration
>>>> + * o Allocate and initialize
>>>> + * event channels per configured
>>>> + * camera.
>>>> + * o Publish transport parameters
>>>> + * that will be in effect during
>>>> + * this connection.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateInitialised
>>>> + *
>>>> + * o Query frontend transport parameters.
>>>> + * o Connect to the event channels.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateConnected
>>>> + *
>>>> + * o Create and initialize OS
>>>> + * virtual camera as per
>>>> + * configuration.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateConnected
>>>> + *
>>>> + * XenbusStateUnknown
>>>> + * XenbusStateClosed
>>>> + * XenbusStateClosing
>>>> + * o Remove virtual camera device
>>>> + * o Remove event channels
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateClosed
>>>> + *
>>>> + *------------------------------- Recovery flow -------------------------------
>>>> + *
>>>> + * In case of frontend unrecoverable errors backend handles that as
>>>> + * if frontend goes into the XenbusStateClosed state.
>>>> + *
>>>> + * In case of backend unrecoverable errors frontend tries removing
>>>> + * the virtualized device. If this is possible at the moment of error,
>>>> + * then frontend goes into the XenbusStateInitialising state and is ready for
>>>> + * new connection with backend. If the virtualized device is still in use and
>>>> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
>>>> + * until either the virtualized device is removed or backend initiates a new
>>>> + * connection. On the virtualized device removal frontend goes into the
>>>> + * XenbusStateInitialising state.
>>>> + *
>>>> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
>>>> + * unrecoverable errors then frontend cannot send requests to the backend
>>>> + * and thus cannot provide functionality of the virtualized device anymore.
>>>> + * After backend is back to normal the virtualized device may still hold some
>>>> + * state: configuration in use, allocated buffers, client application state etc.
>>>> + * In most cases, this will require frontend to implement complex recovery
>>>> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
>>>> + * frontend will make sure no new clients of the virtualized device are
>>>> + * accepted, allow existing client(s) to exit gracefully by signaling error
>>>> + * state etc.
>>>> + * Once all the clients are gone frontend can reinitialize the virtualized
>>>> + * device and get into XenbusStateInitialising state again signaling the
>>>> + * backend that a new connection can be made.
>>>> + *
>>>> + * There are multiple conditions possible under which frontend will go from
>>>> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
>>>> + * specific. For example:
>>>> + * 1. The underlying OS framework may provide callbacks to signal that the last
>>>> + * client of the virtualized device has gone and the device can be removed
>>>> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
>>>> + * to periodically check if this is the right time to re-try removal of
>>>> + * the virtualized device.
>>>> + * 3. By any other means.
>>>> + *
>>>> + ******************************************************************************
>>>> + * REQUEST CODES
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_OP_SET_CONFIG 0x00
>>>> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
>>>> +#define XENCAMERA_OP_BUF_CREATE 0x02
>>>> +#define XENCAMERA_OP_BUF_DESTROY 0x03
>>>> +#define XENCAMERA_OP_STREAM_START 0x04
>>>> +#define XENCAMERA_OP_STREAM_STOP 0x05
>>>> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
>>>> +#define XENCAMERA_OP_SET_CTRL 0x07
>> I am thinking about extending the command set a bit as it already
>> has some flaws, e.g. there is no way for a VM to tell the backend
>> that the buffer is not in use anymore and can be given back
>> to the real HW driver, e.g. queue/dequeue in V4L2 terms:
>>
>> #define XENCAMERA_OP_SET_FORMAT 0x00
>> - will be used to set format: pixel format, resolution
>>
>> #define XENCAMERA_OP_SET_FRAME_RATE 0x01
>> - used to set the frame rate
>>
>> #define XENCAMERA_OP_BUF_REQUEST 0x02
>> - asks backend to allocate the given number of buffers,
>> backend replies with real number of those to be used
>>
>> #define XENCAMERA_OP_BUF_CREATE 0x03
>> - create a shared buffer
>>
>> #define XENCAMERA_OP_BUF_DESTROY 0x04
>> - destroy a shared buffer
>>
>> #define XENCAMERA_OP_BUF_QUEUE 0x05
>> - VM tells the backend that it has access to the shared buffer
>> and the buffer cannot be sent back to real HW driver
>>
>> #define XENCAMERA_OP_BUF_DEQUEUE 0x06
>> - VM tells the backend that the shared buffer is not in use and
>> can be sent to real HW driver
> This is the wrong way around: QUEUE would queue the shared buffer to
> the backend for use with the real HW driver, DEQUEUE would dequeue it
> for use in the VM.
Indeed, thank you for spotting this
>
>> #define XENCAMERA_OP_CTRL_ENUM 0x07
>> - get i-th control ranges and settings
>>
>> #define XENCAMERA_OP_CTRL_GET 0x08
>> - get control value
>>
>> #define XENCAMERA_OP_CTRL_SET 0x09
>> - set control value
>>
>> #define XENCAMERA_OP_STREAM_START 0x0a
>> - start streaming
>>
>> #define XENCAMERA_OP_STREAM_STOP 0x0b
>> - stop ctreaming
>>
>>
>>>> +
>>>> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
>>>> +#define XENCAMERA_CTRL_CONTRAST 0x01
>>>> +#define XENCAMERA_CTRL_SATURATION 0x02
>>>> +#define XENCAMERA_CTRL_HUE 0x03
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * EVENT CODES
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_DRIVER_NAME "vcamera"
>>> Ah, that's where vcamera comes from. How about calling this xen-camera or
>>> virt-camera? With a preference for xen-camera, since that's what you use for the
>>> defines as well.
>>>
>>> Or perhaps pv-camera?
>>>
>>> Is this driver going to be xen-specific, or more a general approach that everyone
>>> can use? Obviously, the latter would be preferable.
>> As I have already replied to the cover letter with explanations:
>> 'v' stands for 'virtual' and there is a convention to name the
>> Xen virtual devices starting with 'v': vif, vkbd etc.
> Yeah, ignore my comment.
>
>>> BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
>>> radio and RDS hardware. Contact me if this is of interest to Xen to support this as
>>> well given the automotive use-case.
>> Yes, thank you, but at this stage we are targeting camera only.
>> Radio can be another topic if time allows ;) And most probably
>> it will be a dedicated 'vradio' protocol then...
>>>> +
>>>> +#define XENCAMERA_LIST_SEPARATOR ","
>>>> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
>>>> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
>>>> +
>>>> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
>>>> +#define XENCAMERA_FIELD_FE_VERSION "version"
>>>> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
>>>> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
>>>> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
>>>> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
>>>> +#define XENCAMERA_FIELD_CONTROLS "controls"
>>>> +#define XENCAMERA_FIELD_FORMATS "formats"
>>>> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
>>>> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
>>>> +
>>>> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
>>>> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
>>>> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
>>>> +#define XENCAMERA_CTRL_HUE_STR "hue"
>>>> +
>>>> +/* Maximum number of buffer planes supported. */
>>>> +#define XENCAMERA_MAX_PLANE 4
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * STATUS RETURN CODES
>>>> + ******************************************************************************
>>>> + *
>>>> + * Status return code is zero on success and -XEN_EXX on failure.
>>>> + *
>>>> + ******************************************************************************
>>>> + * Assumptions
>>>> + ******************************************************************************
>>>> + *
>>>> + * - usage of grant reference 0 as invalid grant reference:
>>>> + * grant reference 0 is valid, but never exposed to a PV driver,
>>>> + * because of the fact it is already in use/reserved by the PV console.
>>>> + * - all references in this document to page sizes must be treated
>>>> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
>>>> + *
>>>> + ******************************************************************************
>>>> + * Description of the protocol between frontend and backend driver
>>>> + ******************************************************************************
>>>> + *
>>>> + * The two halves of a Para-virtual camera driver communicate with
>>>> + * each other using shared pages and event channels.
>>>> + * Shared page contains a ring with request/response packets.
>>>> + *
>>>> + * All reserved fields in the structures below must be 0.
>>>> + *
>>>> + * For all request/response/event packets:
>>>> + * - frame rate parameter is represented as a pair of 4 octet long
>>>> + * numerator and denominator:
>>>> + * - frame_rate_numer - uint32_t, numerator of the frame rate
>>>> + * - frame_rate_denom - uint32_t, denominator of the frame rate
>>>> + * The corresponding frame rate (Hz) is calculated as:
>>>> + * frame_rate = frame_rate_numer / frame_rate_denom
>>>> + * - buffer index is a zero based index of the buffer. Must be less than
>>>> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
>>>> + * - index - uint8_t, index of the buffer.
>>>> + *
>>>> + *
>>>> + *---------------------------------- Requests ---------------------------------
>>>> + *
>>>> + * All request packets have the same length (64 octets).
>>>> + * All request packets have common header:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | operation | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * id - uint16_t, private guest value, echoed in response.
>>>> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
>>>> + *
>>>> + *
>>>> + * Request configuration set/reset - request to set or reset.
>>>> + * the configuration/mode of the camera:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | pixel format | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | width | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | height | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_numer | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_denom | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | num_bufs | reserved | 32
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * Pass all zeros to reset, otherwise command is treated as configuration set.
>>>> + *
>>>> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>> + * width - uint32_t, width in pixels.
>>>> + * height - uint32_t, height in pixels.
>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>> If you have to support HDMI/SDTV inputs as well, then you also need to know
>>> the interlaced format, unless you have no plans to support that.
>>>
>>>> + * num_bufs - uint8_t, desired number of buffers to be used.
>>> Huh? What has that to do with the format? Why would you need this here?
>> Well, the operation name is 'set_config', not 'set_format',
>> so I thought we can have such a cumulative command assembling
>> all the parameters of the configuration. But now I am looking at
>> turning this single 'set_config' command to 3 different commands,
>> which is more practical and aligned with V4L2 in particular (please
>> see above in the command set):
>> 1. set format command:
>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>> * width - uint32_t, width in pixels.
>> * height - uint32_t, height in pixels.
>>
>> 2. Set frame rate command:
>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>
>> 3. Set/request num bufs:
>> * num_bufs - uint8_t, desired number of buffers to be used.
> I like this much better. 1+2 could be combined, but 3 should definitely remain
> separate.
ok, then 1+2 combined + 3 separate.
Do you think we can still name 1+2 as "set_format" or "set_config"
will fit better?
>
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + * Notes:
>>>> + * - frontend must check the corresponding response in order to see
>>>> + * if the values reported back by the backend do match the desired ones
>>>> + * and can be accepted.
>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>> + * configuration.
>>>> + */
>>>> +struct xencamera_config {
>>>> + uint32_t pixel_format;
>>>> + uint32_t width;
>>>> + uint32_t height;
>>>> + uint32_t frame_rate_nom;
>>>> + uint32_t frame_rate_denom;
>>>> + uint8_t num_bufs;
>>>> +};
>>>> +
>>>> +/*
>>>> + * Request buffer details - request camera buffer's memory layout.
>>>> + * detailed description:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + *
>>>> + * Request camera buffer creation:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref_directory | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>> + *
>>>> + * index - uint8_t, index of the buffer to be created.
>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>> + * then reference to the next shared page must be supplied (see
>>>> + * gref_dir_next_page below).
>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>> does.
>> Well, I still think it is better to have a per buffer interface
>> in the protocol as it is done for other Xen virtual devices.
>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>> what it does internally in the frontend driver
> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
> sense. I'm not sure about the naming.
>
> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
> on the backend, and so can fail. Also, the actual number of allocated buffers in
> case of success can be more or less than what was requested.
The buffers can be allocated and shared by either backend or frontend: see
"be-alloc" configuration option telling which domain (VM) shares
the Xen grant references to the pages of the buffer: either frontend
or backend.
So, I was more thinking that in case of V4L2 based frontend driver:
1. Frontend serves REQBUFS ioctl and asks the backend with
XENCAMERA_OP_BUF_REQUEST
if it can handle that many buffers and gets number of buffers to be used
and buffer structure (number of planes, sizes, offsets etc.) as the reply
to that request
2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
3. Frontend returns from REQBUFS ioctl with actual number of buffers
allocated
>>>> + *
>>>> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
>>>> + * allocate the buffer with the parameters provided in this request and page
>>>> + * directory is handled as follows:
>>>> + * Frontend on request:
>>>> + * - allocates pages for the directory (gref_directory,
>>>> + * gref_dir_next_page(s)
>>>> + * - grants permissions for the pages of the directory to the backend
>>>> + * - sets gref_dir_next_page fields
>>>> + * Backend on response:
>>>> + * - grants permissions for the pages of the buffer allocated to
>>>> + * the frontend
>>>> + * - fills in page directory with grant references
>>>> + * (gref[] in struct xencamera_page_directory)
>>>> + */
>>>> +struct xencamera_buf_create_req {
>>>> + uint8_t index;
>>>> + uint8_t reserved[3];
>>>> + grant_ref_t gref_directory;
>>>> +};
>>>> +
>>>> +/*
>>>> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
>>>> + * the request) employs a list of pages, describing all pages of the shared
>>>> + * data buffer:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref_dir_next_page | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref[0] | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref[i] | i*4+8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref[N - 1] | N*4+8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
>>>> + * page directory. Must be 0 if there are no more pages in the list.
>>>> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
>>>> + * allocated at XENCAMERA_OP_BUF_CREATE.
>>>> + *
>>>> + * Number of grant_ref_t entries in the whole page directory is not
>>>> + * passed, but instead can be calculated as:
>>>> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
>>>> + * XEN_PAGE_SIZE
>>>> + */
>>>> +struct xencamera_page_directory {
>>>> + grant_ref_t gref_dir_next_page;
>>>> + grant_ref_t gref[1]; /* Variable length */
>>>> +};
>>>> +
>>>> +/*
>>>> + * Request buffer destruction - destroy a previously allocated camera buffer:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_BUF_DESTROY| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * index - uint8_t, index of the buffer to be destroyed.
>>>> + */
>>> There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
>>> of them.
>> This is not specifically related to V4L2, but can be issued
>> in response to backend's state change etc.
>> So, we have a pair of commands to create and destroy buffers.
>> Even more, frontend can be a some-os-based-driver, not V4L2
>> based. Or even a user-space application if your will.
>>>> +
>>>> +struct xencamera_buf_destroy_req {
>>>> + uint8_t index;
>>>> +};
>>>> +
>>>> +/*
>>>> + * Request camera capture stream start:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_OP_STREAM_START| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + *
>>>> + * Request camera capture stream stop:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_OP_STREAM_STOP | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + *
>>>> + * Request camera control details:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + * index - uint8_t, index of the control to be queried.
>>>> + */
>>>> +struct xencamera_get_ctrl_details_req {
>>>> + uint8_t index;
>>>> +};
>>>> +
>>>> +/*
>>>> + *
>>>> + * Request camera control change:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_SET_CTRL | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | value | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + * index - uint8_t, index of the control.
>>>> + * value - int32_t, new value of the control.
>>> I would recommend using a int64_t as the control value.
>> Good point, thank you
>>> Note that there are also controls with a payload (e.g. string controls).
>> Could you please give me an example of such a control?
>> Do you think such controls can be of use in a VM?
>> Can we avoid such controls if we target a simple virtual
>> camera device? If this is for radio use-case, then we'll
>> have such support in 'vradio' protocol if need be
> Right now all string controls are related to RDS receivers/transmitters. If you
> ever decide on a vradio protocol, then you need these (Programme Service name
> and Radio Text info). And there is an array of Alternate Frequencies, also RDS
> specific.
>
> There are some array controls in V4L2, those are used to control motion detection
> for surveillance cameras. And 'compound controls' (think of this as C structs) are
> appearing for HW codecs.
>
> I don't think any of these are likely to appear for cameras, at least not in a
> way that is relevant for Xen.
ok
>>> If there is ever interest in adding radio/RDS support, then that will become
>>> an issue.
>> You mean something like station names, ads etc?
> Yup.
>
>>>> + */
>>>> +struct xencamera_set_ctrl_req {
>>>> + uint8_t index;
>>>> + uint8_t reserved[3];
>>>> + int32_t value;
>>>> +};
>>>> +
>>>> +/*
>>>> + *---------------------------------- Responses --------------------------------
>>>> + *
>>>> + * All response packets have the same length (64 octets).
>>>> + *
>>>> + * All response packets have common header:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | operation | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * id - uint16_t, copied from the request.
>>>> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
>>>> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
>>>> + *
>>>> + *
>>>> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | pixel format | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | width | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | height | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_numer | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_denom | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | num_bufs | reserved | 32
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * Meaning of the corresponding values in this response is the same as for
>>>> + * XENCAMERA_OP_SET_CONFIG request.
>>>> + *
>>>> + *
>>>> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
>>>> + * request:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | size | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | num_planes | reserved | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[0] | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[1] | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[2] | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[3] | 32
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[0] | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[1] | 40
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[2] | 44
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[3] | 48
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_stride[0] | plane_stride[1] | 52
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_stride[2] | plane_stride[3] | 56
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 60
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * size - uint32_t, overall size of the buffer including sizes of the
>>>> + * individual planes and padding if applicable.
>>>> + * num_planes - uint8_t, number of planes for this buffer.
>>>> + * plane_offset - array of uint32_t, offset of the corresponding plane
>>>> + * in octets from the buffer start.
>>>> + * plane_size - array of uint32_t, size in octets of the corresponding plane
>>>> + * including padding.
>>>> + * plane_stride - array of uint32_t, size in octets occupied by the
>>>> + * corresponding single image line including padding if applicable.
>>> Nice!
>> Thank you
>>>> + */
>>>> +struct xencamera_buf_details_resp {
>>>> + uint32_t size;
>>>> + uint8_t num_planes;
>>>> + uint8_t reserved[3];
>>>> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
>>>> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
>>>> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
>>>> +};
>>>> +
>>>> +/*
>>>> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | type | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | min | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | max | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | step | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | def_val | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * index - uint8_t, index of the camera control in response.
>>>> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
>>>> + * min - int32_t, minimum value of the control.
>>>> + * max - int32_t, maximum value of the control.
>>>> + * step - int32_t, minimum size in which control value can be changed.
>>>> + * def_val - int32_t, default value of the control.
>>> I'd go with 64 bit values for min/max/step/def_val.
>> Sure, good idea, thank you
>>> I would also add a flags field. Some controls are read-only, write-only
>>> or volatile, things userspace needs to know.
>> Then I'll also add numerical constants for such
>>> If you want to support menu controls, then you need a way to get the menu
>>> names as well (VIDIOC_QUERYMENU).
>>>
>>> None of this is needed for this initial use-case, but you need to think
>>> about this up-front.
>> Yes, thank you
>>>> + */
>>>> +struct xencamera_get_ctrl_details_resp {
>>>> + uint8_t index;
>>>> + uint8_t type;
>>>> + uint8_t reserved[2];
>>>> + int32_t min;
>>>> + int32_t max;
>>>> + int32_t step;
>>>> + int32_t def_val;
>>>> +};
>>>> +
>>>> +/*
>>>> + *----------------------------------- Events ----------------------------------
>>>> + *
>>>> + * Events are sent via a shared page allocated by the front and propagated by
>>>> + * evt-event-channel/evt-ring-ref XenStore entries.
>>>> + *
>>>> + * All event packets have the same length (64 octets).
>>>> + * All event packets have common header:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | type | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * id - uint16_t, event id, may be used by front.
>>>> + * type - uint8_t, type of the event.
>>>> + *
>>>> + *
>>>> + * Frame captured event - event from back to front when a new captured
>>>> + * frame is available:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_EVT_FRAME_AVAIL| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | used_sz | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * index - uint8_t, index of the buffer that contains new captured frame.
>>>> + * used_sz - uint32_t, number of octets this frame has. This can be less
>>>> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
>>>> + */
>>>> +struct xencamera_frame_avail_evt {
>>>> + uint8_t index;
>>>> + uint8_t reserved[3];
>>>> + uint32_t used_sz;
>>>> +};
>>>> +
>>>> +struct xencamera_req {
>>>> + uint16_t id;
>>>> + uint8_t operation;
>>>> + uint8_t reserved[5];
>>>> + union {
>>>> + struct xencamera_config config;
>>>> + struct xencamera_buf_create_req buf_create;
>>>> + struct xencamera_buf_destroy_req buf_destroy;
>>>> + struct xencamera_set_ctrl_req set_ctrl;
>>>> + uint8_t reserved[56];
>>>> + } req;
>>>> +};
>>>> +
>>>> +struct xencamera_resp {
>>>> + uint16_t id;
>>>> + uint8_t operation;
>>>> + uint8_t reserved;
>>>> + int32_t status;
>>>> + union {
>>>> + struct xencamera_config config;
>>>> + struct xencamera_buf_details_resp buf_details;
>>>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>>>> + uint8_t reserved1[56];
>>>> + } resp;
>>>> +};
>>>> +
>>>> +struct xencamera_evt {
>>>> + uint16_t id;
>>>> + uint8_t type;
>>>> + uint8_t reserved[5];
>>>> + union {
>>>> + struct xencamera_frame_avail_evt frame_avail;
>>>> + uint8_t reserved[56];
>>>> + } evt;
>>>> +};
>>>> +
>>>> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * Back to front events delivery
>>>> + ******************************************************************************
>>>> + * In order to deliver asynchronous events from back to front a shared page is
>>>> + * allocated by front and its granted reference propagated to back via
>>>> + * XenStore entries (evt-ring-ref/evt-event-channel).
>>>> + * This page has a common header used by both front and back to synchronize
>>>> + * access and control event's ring buffer, while back being a producer of the
>>>> + * events and front being a consumer. The rest of the page after the header
>>>> + * is used for event packets.
>>>> + *
>>>> + * Upon reception of an event(s) front may confirm its reception
>>>> + * for either each event, group of events or none.
>>>> + */
>>>> +
>>>> +struct xencamera_event_page {
>>>> + uint32_t in_cons;
>>>> + uint32_t in_prod;
>>>> + uint8_t reserved[56];
>>>> +};
>>>> +
>>>> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
>>>> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
>>>> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
>>>> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
>>>> +#define XENCAMERA_IN_RING(page) \
>>>> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
>>>> +#define XENCAMERA_IN_RING_REF(page, idx) \
>>>> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
>>>> +
>>>> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
>>>> +
>>>> +/*
>>>> + * Local variables:
>>>> + * mode: C
>>>> + * c-file-style: "BSD"
>>>> + * c-basic-offset: 4
>>>> + * tab-width: 4
>>>> + * indent-tabs-mode: nil
>>>> + * End:
>>>> + */
>>>>
>>> I think the most important decision to make here is whether or not you want to support
>>> hotpluggable sources like HDMI. And an additional complication with that is HDCP.
>>> While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
>>> a patch series adding this is expected later this year/early next year. It might not
>>> be an issue in practice if these are all closed systems, but nevertheless, it is
>>> something to think about.
>> Yes, thank you for raising these questions, it is worth thinking
>> about such use-cases.
>>> Regards,
>>>
>>> Hans
>> Thank you so much for the comments,
>> Oleksandr
>>
> Regards,
>
> Hans
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 8:24 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 8:24 UTC (permalink / raw)
To: Hans Verkuil, xen-devel, konrad.wilk, jgross, boris.ostrovsky,
mchehab, linux-media, sakari.ailus, koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 09/10/2018 10:53 AM, Hans Verkuil wrote:
> Hi Oleksandr,
>
> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>> Hi, Hans!
>>
>> On 09/09/2018 01:31 PM, Hans Verkuil wrote:
>>> Hi Oleksandr,
>>>
>>> Sorry for the delay in reviewing, I missed this patch until you pinged me, and
>>> I was very busy after that as well.
>> I do appreciate you spending time on this!
>>> On 07/31/2018 11:31 AM, Oleksandr Andrushchenko wrote:
>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>
>>>> This is the ABI for the two halves of a para-virtualized
>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>> high definition maps etc.
>>>>
>>>> The initial goal is to support most needed functionality with the
>>>> final idea to make it possible to extend the protocol if need be:
>>>>
>>>> 1. Provide means for base virtual device configuration:
>>>> - pixel formats
>>>> - resolutions
>>>> - frame rates
>>>> 2. Support basic camera controls:
>>>> - contrast
>>>> - brightness
>>>> - hue
>>>> - saturation
>>>> 3. Support streaming control
>>>> 4. Support zero-copying use-cases
>>>>
>>>> Signed-off-by: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>> ---
>>>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>>>> 1 file changed, 981 insertions(+)
>>>> create mode 100644 xen/include/public/io/cameraif.h
>>>>
>>>> diff --git a/xen/include/public/io/cameraif.h b/xen/include/public/io/cameraif.h
>>>> new file mode 100644
>>>> index 000000000000..bdc6a1262fcf
>>>> --- /dev/null
>>>> +++ b/xen/include/public/io/cameraif.h
>>>> @@ -0,0 +1,981 @@
>>>> +/******************************************************************************
>>>> + * cameraif.h
>>>> + *
>>>> + * Unified camera device I/O interface for Xen guest OSes.
>>>> + *
>>>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>>>> + * of this software and associated documentation files (the "Software"), to
>>>> + * deal in the Software without restriction, including without limitation the
>>>> + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
>>>> + * sell copies of the Software, and to permit persons to whom the Software is
>>>> + * furnished to do so, subject to the following conditions:
>>>> + *
>>>> + * The above copyright notice and this permission notice shall be included in
>>>> + * all copies or substantial portions of the Software.
>>>> + *
>>>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>>>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>>>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>>>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>>>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>>>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>>>> + * DEALINGS IN THE SOFTWARE.
>>>> + *
>>>> + * Copyright (C) 2018 EPAM Systems Inc.
>>>> + *
>>>> + * Author: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>> + */
>>> Use SPDX tag instead of copying the license text.
>> This is yet a Xen header which belongs to Xen project and
>> all the rest of the protocols have the same license header.
>> If Xen community decides to use SPDX then I'll definitely follow.
> Ah, yes, I was reviewing this as a kernel header, I hadn't realized
> that this isn't a kernel header. Since it isn't a kernel header, you
> can disregard my comments about style and naming conventions, since
> you have your own.
>
>> Konrad, do you think this is the right time for such a move?
>>>> +
>>>> +#ifndef __XEN_PUBLIC_IO_CAMERAIF_H__
>>>> +#define __XEN_PUBLIC_IO_CAMERAIF_H__
>>>> +
>>>> +#include "ring.h"
>>>> +#include "../grant_table.h"
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * Protocol version
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_PROTOCOL_VERSION "1"
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * Feature and Parameter Negotiation
>>>> + ******************************************************************************
>>>> + *
>>>> + * Front->back notifications: when enqueuing a new request, sending a
>>>> + * notification can be made conditional on xencamera_req (i.e., the generic
>>>> + * hold-off mechanism provided by the ring macros). Backends must set
>>>> + * xencamera_req appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
>>>> + *
>>>> + * Back->front notifications: when enqueuing a new response, sending a
>>>> + * notification can be made conditional on xencamera_resp (i.e., the generic
>>>> + * hold-off mechanism provided by the ring macros). Frontends must set
>>>> + * xencamera_resp appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
>>>> + *
>>>> + * The two halves of a para-virtual camera driver utilize nodes within
>>>> + * XenStore to communicate capabilities and to negotiate operating parameters.
>>>> + * This section enumerates these nodes which reside in the respective front and
>>>> + * backend portions of XenStore, following the XenBus convention.
>>>> + *
>>>> + * All data in XenStore is stored as strings. Nodes specifying numeric
>>>> + * values are encoded in decimal. Integer value ranges listed below are
>>>> + * expressed as fixed sized integer types capable of storing the conversion
>>>> + * of a properly formatted node string, without loss of information.
>>>> + *
>>>> + ******************************************************************************
>>>> + * Example configuration
>>>> + ******************************************************************************
>>>> + *
>>>> + * This is an example of backend and frontend configuration:
>>>> + *
>>>> + *--------------------------------- Backend -----------------------------------
>>>> + *
>>>> + * /local/domain/0/backend/vcamera/1/0/frontend-id = "1"
>>>> + * /local/domain/0/backend/vcamera/1/0/frontend = "/local/domain/1/device/vcamera/0"
>>>> + * /local/domain/0/backend/vcamera/1/0/state = "4"
>>>> + * /local/domain/0/backend/vcamera/1/0/versions = "1,2"
>>> Why vcamera instead of just camera? If 'v' stands for 'video', then that seems
>>> superfluous to me.
>> 'v' stands for 'virtual'. I am following Xen convention used
>> for all other virtual device protocols here.
>>>> + *
>>>> + *--------------------------------- Frontend ----------------------------------
>>>> + *
>>>> + * /local/domain/1/device/vcamera/0/backend-id = "0"
>>>> + * /local/domain/1/device/vcamera/0/backend = "/local/domain/0/backend/vcamera/1"
>>>> + * /local/domain/1/device/vcamera/0/state = "4"
>>>> + * /local/domain/1/device/vcamera/0/version = "1"
>>>> + * /local/domain/1/device/vcamera/0/be-alloc = "1"
>>>> + *
>>>> + *---------------------------- Device 0 configuration -------------------------
>>>> + *
>>>> + * /local/domain/1/device/vcamera/0/controls = "contrast,hue"
>>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>>> + * /local/domain/1/device/vcamera/0/formats/YUYV/1920x1080 = "15/2"
>>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/640x480 = "15/1,15/2"
>>>> + * /local/domain/1/device/vcamera/0/formats/BGRA/1200x720 = "15/2"
>>>> + * /local/domain/1/device/vcamera/0/unique-id = "0"
>>>> + * /local/domain/1/device/vcamera/0/req-ring-ref = "2832"
>>>> + * /local/domain/1/device/vcamera/0/req-event-channel = "15"
>>>> + * /local/domain/1/device/vcamera/0/evt-ring-ref = "387"
>>>> + * /local/domain/1/device/vcamera/0/evt-event-channel = "16"
>>>> + *
>>>> + *---------------------------- Device 1 configuration -------------------------
>>>> + *
>>>> + * /local/domain/1/device/vcamera/1/controls = "brightness,saturation,hue"
>>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/640x480 = "30/1,15/1,15/2"
>>>> + * /local/domain/1/device/vcamera/1/formats/YUYV/1920x1080 = "15/2"
>>>> + * /local/domain/1/device/vcamera/1/unique-id = "1"
>>>> + * /local/domain/1/device/vcamera/1/req-ring-ref = "2833"
>>>> + * /local/domain/1/device/vcamera/1/req-event-channel = "17"
>>>> + * /local/domain/1/device/vcamera/1/evt-ring-ref = "388"
>>>> + * /local/domain/1/device/vcamera/1/evt-event-channel = "18"
>>>> + *
>>>> + ******************************************************************************
>>>> + * Backend XenBus Nodes
>>>> + ******************************************************************************
>>>> + *
>>>> + *----------------------------- Protocol version ------------------------------
>>>> + *
>>>> + * versions
>>>> + * Values: <string>
>>>> + *
>>>> + * List of XENCAMERA_LIST_SEPARATOR separated protocol versions supported
>>>> + * by the backend. For example "1,2,3".
>>>> + *
>>>> + ******************************************************************************
>>>> + * Frontend XenBus Nodes
>>>> + ******************************************************************************
>>>> + *
>>>> + *-------------------------------- Addressing ---------------------------------
>>>> + *
>>>> + * dom-id
>>>> + * Values: <uint16_t>
>>>> + *
>>>> + * Domain identifier.
>>>> + *
>>>> + * dev-id
>>>> + * Values: <uint16_t>
>>>> + *
>>>> + * Device identifier.
>>>> + *
>>>> + * /local/domain/<dom-id>/device/vcamera/<dev-id>/...
>>>> + *
>>>> + *----------------------------- Protocol version ------------------------------
>>>> + *
>>>> + * version
>>>> + * Values: <string>
>>>> + *
>>>> + * Protocol version, chosen among the ones supported by the backend.
>>>> + *
>>>> + *------------------------- Backend buffer allocation -------------------------
>>>> + *
>>>> + * be-alloc
>>> I thought that 'be' referred to 'big-endian', but apparently not. Perhaps it
>>> is better to just write 'backend-alloc'.
>> Well, 'be' and 'fe' are commonly used in Xen for backend
>> and frontend, so I'll probably stick to that convention for now.
>>>> + * Values: "0", "1"
>>>> + *
>>>> + * If value is set to "1", then backend will be the buffer
>>>> + * provider/allocator for this domain during XENCAMERA_OP_BUF_CREATE
>>>> + * operation.
>>>> + * If value is not "1" or omitted frontend must allocate buffers itself.
>>>> + *
>>>> + *------------------------------- Camera settings -----------------------------
>>>> + *
>>>> + * unique-id
>>>> + * Values: <string>
>>>> + *
>>>> + * After device instance initialization each camera is assigned a
>>>> + * unique ID, so it can be identified by the backend by this ID.
>>>> + * This can be UUID or such.
>>>> + *
>>>> + * controls
>>>> + * Values: <list of string>
>>>> + *
>>>> + * List of supported camera controls separated by XENCAMERA_LIST_SEPARATOR.
>>>> + * Camera controls are expressed as a list of string values w/o any
>>>> + * ordering requirement.
>>>> + *
>>>> + * formats
>>>> + * Values: <format, char[4]>
>>>> + *
>>>> + * Formats are organized as a set of directories one per each
>>>> + * supported pixel format. The name of the directory is an upper case
>>>> + * string of the corresponding FOURCC string label. The next level of
>>>> + * the directory under <formats> represents supported resolutions.
>>> Lower-case characters are also use in pixelformats, so I'd just keep this as-is.
>> Ok, no problem - will remove the 'upper case' from the definition
>>> In addition it is common to set bit 31 of the fourcc to 1 if the format is
>>> big-endian (see v4l2_fourcc_be macro). When v4l utilities print this format we
>>> add a -BE suffix, so V4L2_PIX_FMT_ARGB555X becomes "AR15-BE". You might want to
>>> keep that convention.
>> I'll think about it, thank you
>>>> + *
>>>> + * resolution
>>>> + * Values: <width, uint32_t>x<height, uint32_t>
>>>> + *
>>>> + * Resolutions are organized as a set of directories one per each
>>>> + * supported resolution under corresponding <formats> directory.
>>>> + * The name of the directory is the supported width and height
>>>> + * of the camera resolution in pixels.
>>> What if you are dealing with an HDMI input? Not unreasonable for media
>>> systems. There can be a lot of resolutions/framerates, and the resolution
>>> can change on the fly, or of course disappear.
>> Well, this is a part of the system configuration done
>> before we actually run the VMs, e.g. at system design
>> and configuration time. Most of the time you do know which
>> resolutions, frame rates etc. you want to assign and these
>> settings remain static for the whole lifetime of that VM.
>> If you are designing a system which needs these resolutions
>> to change at run-time then this can be done:
>> 1. Backend changes the state of the frontend to XenbusStateClosed state
>> 2. Xen (xl/libxl) or backend change the configuration in XenStore
>> 3. Backend re-initializes the frontend which reads new configuration
>>> What is also missing here is a way to report pixel aspect ratio: PAL and
>>> NTSC-based video material doesn't have square pixels.
>> Hm, indeed, thank you. I'll put this as a fraction under
>> the corresponding 'resolution':
>>
>> Now:
>> /local/domain/2/device/vcamera/0/formats/YUYV/640x480 = "30/1,15/1,15/2"
>> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024 = "015/2"
>>
>> Will change to:
>> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/framerates/ =
>> "30/1,15/1,15/2"
>> /local/domain/2/device/vcamera/0/formats/YUYV/640x480/aspectratio = "1/1"
> You need to be more precise: aspectratio can refer to the display aspect
> ratio or the pixel aspect ratio, and you can calculate one from the other.
>
> So this is either pixelaspectratio or displayaspectratio.
>
> Even though V4L2 provides a pixelaspectratio, I would recommend that you
> choose to export it as a displayaspectratio. It's easier for applications
> to handle.
ok, then "display-aspect-ratio"
>
> Please note that in the case of DVD/BluRay source material the same
> resolution can have different display aspect ratios: e.g. PAL can be
> 4:3 or 16:9, depending on whether it is widescreen or not.
>
ok
>> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/framerates = "015/2"
>> /local/domain/2/device/vcamera/0/formats/YUYV/1280x1024/aspectratio =
>> "59/58"
>>
>>> It's important to decide whether or not you want to support video sources
>>> like that (HDMI, Composite/S-Video inputs, USB ports where users can connect
>>> or disconnect webcams) or if you stick to fixed camera pipelines.
>> I believe that this is all hidden from the frontend by the
>> backend, so I se nothing we have to put in the protocol
>> with this respect.
>>> The big difference is that you don't control what someone can connect as
>>> external sources, so you will have to be a lot more careful and robust.
>>>
>>> I suspect that you likely will want to support such sources eventually, so
>>> it pays to design this with that in mind.
>> Again, I think that this is the backend to hide these
>> use-cases from the frontend.
> I'm not sure you can: say you are playing a bluray connected to the system
> with HDMI, then if there is a resolution change, what do you do? You can tear
> everything down and build it up again, or you can just tell frontends that
> something changed and that they have to look at the new vcamera configuration.
>
> The latter seems to be more sensible to me. It is really not much that you
> need to do: all you really need is an event signalling that something changed.
> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
well, this complicates things a lot as I'll have to
re-allocate buffers - right?
But anyways, I can add
#define XENCAMERA_EVT_CFG_CHANGE 0x01
in the protocol, so we can address this use-case
>
>>>> + *
>>>> + * frame-rates
>>>> + * Values: <numerator, uint32_t>/<denominator, uint32_t>
>>>> + *
>>>> + * List of XENCAMERA_FRAME_RATE_SEPARATOR separated supported frame rates
>>>> + * of the camera expressed as numerator and denominator of the
>>>> + * corresponding frame rate.
>>>> + *
>>>> + * The format of the <formats> directory tree with resolutions and frame rates
>>>> + * must be structured in the following format:
>>>> + *
>>>> + * .../vcamera/<dev-id>/<format[i]>/<resolution[j]>/<frame-rates[k]>
>>>> + *
>>>> + * where
>>>> + * i - i-th supported pixel format
>>>> + * j - j-th supported resolution for i-th pixel format
>>>> + * k - k-th supported frame rate for i-th pixel format and j-th
>>>> + * resolution> + *
>>>> + *------------------- Camera Request Transport Parameters ---------------------
>>>> + *
>>>> + * This communication path is used to deliver requests from frontend to backend
>>>> + * and get the corresponding responses from backend to frontend,
>>>> + * set up per virtual camera device.
>>>> + *
>>>> + * req-event-channel
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The identifier of the Xen camera's control event channel
>>>> + * used to signal activity in the ring buffer.
>>>> + *
>>>> + * req-ring-ref
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The Xen grant reference granting permission for the backend to map
>>>> + * a sole page of camera's control ring buffer.
>>>> + *
>>>> + *-------------------- Camera Event Transport Parameters ----------------------
>>>> + *
>>>> + * This communication path is used to deliver asynchronous events from backend
>>>> + * to frontend, set up per virtual camera device.
>>>> + *
>>>> + * evt-event-channel
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The identifier of the Xen camera's event channel
>>>> + * used to signal activity in the ring buffer.
>>>> + *
>>>> + * evt-ring-ref
>>>> + * Values: <uint32_t>
>>>> + *
>>>> + * The Xen grant reference granting permission for the backend to map
>>>> + * a sole page of camera's event ring buffer.
>>>> + */
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * STATE DIAGRAMS
>>>> + ******************************************************************************
>>>> + *
>>>> + * Tool stack creates front and back state nodes with initial state
>>>> + * XenbusStateInitialising.
>>>> + * Tool stack creates and sets up frontend camera configuration
>>>> + * nodes per domain.
>>>> + *
>>>> + *-------------------------------- Normal flow --------------------------------
>>>> + *
>>>> + * Front Back
>>>> + * ================================= =====================================
>>>> + * XenbusStateInitialising XenbusStateInitialising
>>>> + * o Query backend device identification
>>>> + * data.
>>>> + * o Open and validate backend device.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateInitWait
>>>> + *
>>>> + * o Query frontend configuration
>>>> + * o Allocate and initialize
>>>> + * event channels per configured
>>>> + * camera.
>>>> + * o Publish transport parameters
>>>> + * that will be in effect during
>>>> + * this connection.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateInitialised
>>>> + *
>>>> + * o Query frontend transport parameters.
>>>> + * o Connect to the event channels.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateConnected
>>>> + *
>>>> + * o Create and initialize OS
>>>> + * virtual camera as per
>>>> + * configuration.
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateConnected
>>>> + *
>>>> + * XenbusStateUnknown
>>>> + * XenbusStateClosed
>>>> + * XenbusStateClosing
>>>> + * o Remove virtual camera device
>>>> + * o Remove event channels
>>>> + * |
>>>> + * |
>>>> + * V
>>>> + * XenbusStateClosed
>>>> + *
>>>> + *------------------------------- Recovery flow -------------------------------
>>>> + *
>>>> + * In case of frontend unrecoverable errors backend handles that as
>>>> + * if frontend goes into the XenbusStateClosed state.
>>>> + *
>>>> + * In case of backend unrecoverable errors frontend tries removing
>>>> + * the virtualized device. If this is possible at the moment of error,
>>>> + * then frontend goes into the XenbusStateInitialising state and is ready for
>>>> + * new connection with backend. If the virtualized device is still in use and
>>>> + * cannot be removed, then frontend goes into the XenbusStateReconfiguring state
>>>> + * until either the virtualized device is removed or backend initiates a new
>>>> + * connection. On the virtualized device removal frontend goes into the
>>>> + * XenbusStateInitialising state.
>>>> + *
>>>> + * Note on XenbusStateReconfiguring state of the frontend: if backend has
>>>> + * unrecoverable errors then frontend cannot send requests to the backend
>>>> + * and thus cannot provide functionality of the virtualized device anymore.
>>>> + * After backend is back to normal the virtualized device may still hold some
>>>> + * state: configuration in use, allocated buffers, client application state etc.
>>>> + * In most cases, this will require frontend to implement complex recovery
>>>> + * reconnect logic. Instead, by going into XenbusStateReconfiguring state,
>>>> + * frontend will make sure no new clients of the virtualized device are
>>>> + * accepted, allow existing client(s) to exit gracefully by signaling error
>>>> + * state etc.
>>>> + * Once all the clients are gone frontend can reinitialize the virtualized
>>>> + * device and get into XenbusStateInitialising state again signaling the
>>>> + * backend that a new connection can be made.
>>>> + *
>>>> + * There are multiple conditions possible under which frontend will go from
>>>> + * XenbusStateReconfiguring into XenbusStateInitialising, some of them are OS
>>>> + * specific. For example:
>>>> + * 1. The underlying OS framework may provide callbacks to signal that the last
>>>> + * client of the virtualized device has gone and the device can be removed
>>>> + * 2. Frontend can schedule a deferred work (timer/tasklet/workqueue)
>>>> + * to periodically check if this is the right time to re-try removal of
>>>> + * the virtualized device.
>>>> + * 3. By any other means.
>>>> + *
>>>> + ******************************************************************************
>>>> + * REQUEST CODES
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_OP_SET_CONFIG 0x00
>>>> +#define XENCAMERA_OP_GET_BUF_DETAILS 0x01
>>>> +#define XENCAMERA_OP_BUF_CREATE 0x02
>>>> +#define XENCAMERA_OP_BUF_DESTROY 0x03
>>>> +#define XENCAMERA_OP_STREAM_START 0x04
>>>> +#define XENCAMERA_OP_STREAM_STOP 0x05
>>>> +#define XENCAMERA_OP_GET_CTRL_DETAILS 0x06
>>>> +#define XENCAMERA_OP_SET_CTRL 0x07
>> I am thinking about extending the command set a bit as it already
>> has some flaws, e.g. there is no way for a VM to tell the backend
>> that the buffer is not in use anymore and can be given back
>> to the real HW driver, e.g. queue/dequeue in V4L2 terms:
>>
>> #define XENCAMERA_OP_SET_FORMAT 0x00
>> - will be used to set format: pixel format, resolution
>>
>> #define XENCAMERA_OP_SET_FRAME_RATE 0x01
>> - used to set the frame rate
>>
>> #define XENCAMERA_OP_BUF_REQUEST 0x02
>> - asks backend to allocate the given number of buffers,
>> backend replies with real number of those to be used
>>
>> #define XENCAMERA_OP_BUF_CREATE 0x03
>> - create a shared buffer
>>
>> #define XENCAMERA_OP_BUF_DESTROY 0x04
>> - destroy a shared buffer
>>
>> #define XENCAMERA_OP_BUF_QUEUE 0x05
>> - VM tells the backend that it has access to the shared buffer
>> and the buffer cannot be sent back to real HW driver
>>
>> #define XENCAMERA_OP_BUF_DEQUEUE 0x06
>> - VM tells the backend that the shared buffer is not in use and
>> can be sent to real HW driver
> This is the wrong way around: QUEUE would queue the shared buffer to
> the backend for use with the real HW driver, DEQUEUE would dequeue it
> for use in the VM.
Indeed, thank you for spotting this
>
>> #define XENCAMERA_OP_CTRL_ENUM 0x07
>> - get i-th control ranges and settings
>>
>> #define XENCAMERA_OP_CTRL_GET 0x08
>> - get control value
>>
>> #define XENCAMERA_OP_CTRL_SET 0x09
>> - set control value
>>
>> #define XENCAMERA_OP_STREAM_START 0x0a
>> - start streaming
>>
>> #define XENCAMERA_OP_STREAM_STOP 0x0b
>> - stop ctreaming
>>
>>
>>>> +
>>>> +#define XENCAMERA_CTRL_BRIGHTNESS 0x00
>>>> +#define XENCAMERA_CTRL_CONTRAST 0x01
>>>> +#define XENCAMERA_CTRL_SATURATION 0x02
>>>> +#define XENCAMERA_CTRL_HUE 0x03
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * EVENT CODES
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_EVT_FRAME_AVAIL 0x00
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * XENSTORE FIELD AND PATH NAME STRINGS, HELPERS
>>>> + ******************************************************************************
>>>> + */
>>>> +#define XENCAMERA_DRIVER_NAME "vcamera"
>>> Ah, that's where vcamera comes from. How about calling this xen-camera or
>>> virt-camera? With a preference for xen-camera, since that's what you use for the
>>> defines as well.
>>>
>>> Or perhaps pv-camera?
>>>
>>> Is this driver going to be xen-specific, or more a general approach that everyone
>>> can use? Obviously, the latter would be preferable.
>> As I have already replied to the cover letter with explanations:
>> 'v' stands for 'virtual' and there is a convention to name the
>> Xen virtual devices starting with 'v': vif, vkbd etc.
> Yeah, ignore my comment.
>
>>> BTW, I am not sure if you are aware of this, but the V4L2 API also has support for
>>> radio and RDS hardware. Contact me if this is of interest to Xen to support this as
>>> well given the automotive use-case.
>> Yes, thank you, but at this stage we are targeting camera only.
>> Radio can be another topic if time allows ;) And most probably
>> it will be a dedicated 'vradio' protocol then...
>>>> +
>>>> +#define XENCAMERA_LIST_SEPARATOR ","
>>>> +#define XENCAMERA_RESOLUTION_SEPARATOR "x"
>>>> +#define XENCAMERA_FRAME_RATE_SEPARATOR "/"
>>>> +
>>>> +#define XENCAMERA_FIELD_BE_VERSIONS "versions"
>>>> +#define XENCAMERA_FIELD_FE_VERSION "version"
>>>> +#define XENCAMERA_FIELD_REQ_RING_REF "req-ring-ref"
>>>> +#define XENCAMERA_FIELD_REQ_CHANNEL "req-event-channel"
>>>> +#define XENCAMERA_FIELD_EVT_RING_REF "evt-ring-ref"
>>>> +#define XENCAMERA_FIELD_EVT_CHANNEL "evt-event-channel"
>>>> +#define XENCAMERA_FIELD_CONTROLS "controls"
>>>> +#define XENCAMERA_FIELD_FORMATS "formats"
>>>> +#define XENCAMERA_FIELD_BE_ALLOC "be-alloc"
>>>> +#define XENCAMERA_FIELD_UNIQUE_ID "unique-id"
>>>> +
>>>> +#define XENCAMERA_CTRL_BRIGHTNESS_STR "brightness"
>>>> +#define XENCAMERA_CTRL_CONTRAST_STR "contrast"
>>>> +#define XENCAMERA_CTRL_SATURATION_STR "saturation"
>>>> +#define XENCAMERA_CTRL_HUE_STR "hue"
>>>> +
>>>> +/* Maximum number of buffer planes supported. */
>>>> +#define XENCAMERA_MAX_PLANE 4
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * STATUS RETURN CODES
>>>> + ******************************************************************************
>>>> + *
>>>> + * Status return code is zero on success and -XEN_EXX on failure.
>>>> + *
>>>> + ******************************************************************************
>>>> + * Assumptions
>>>> + ******************************************************************************
>>>> + *
>>>> + * - usage of grant reference 0 as invalid grant reference:
>>>> + * grant reference 0 is valid, but never exposed to a PV driver,
>>>> + * because of the fact it is already in use/reserved by the PV console.
>>>> + * - all references in this document to page sizes must be treated
>>>> + * as pages of size XEN_PAGE_SIZE unless otherwise noted.
>>>> + *
>>>> + ******************************************************************************
>>>> + * Description of the protocol between frontend and backend driver
>>>> + ******************************************************************************
>>>> + *
>>>> + * The two halves of a Para-virtual camera driver communicate with
>>>> + * each other using shared pages and event channels.
>>>> + * Shared page contains a ring with request/response packets.
>>>> + *
>>>> + * All reserved fields in the structures below must be 0.
>>>> + *
>>>> + * For all request/response/event packets:
>>>> + * - frame rate parameter is represented as a pair of 4 octet long
>>>> + * numerator and denominator:
>>>> + * - frame_rate_numer - uint32_t, numerator of the frame rate
>>>> + * - frame_rate_denom - uint32_t, denominator of the frame rate
>>>> + * The corresponding frame rate (Hz) is calculated as:
>>>> + * frame_rate = frame_rate_numer / frame_rate_denom
>>>> + * - buffer index is a zero based index of the buffer. Must be less than
>>>> + * the value of XENCAMERA_OP_SET_CONFIG.num_bufs response:
>>>> + * - index - uint8_t, index of the buffer.
>>>> + *
>>>> + *
>>>> + *---------------------------------- Requests ---------------------------------
>>>> + *
>>>> + * All request packets have the same length (64 octets).
>>>> + * All request packets have common header:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | operation | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * id - uint16_t, private guest value, echoed in response.
>>>> + * operation - uint8_t, operation code, XENCAMERA_OP_XXX.
>>>> + *
>>>> + *
>>>> + * Request configuration set/reset - request to set or reset.
>>>> + * the configuration/mode of the camera:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | pixel format | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | width | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | height | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_numer | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_denom | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | num_bufs | reserved | 32
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * Pass all zeros to reset, otherwise command is treated as configuration set.
>>>> + *
>>>> + * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>> + * width - uint32_t, width in pixels.
>>>> + * height - uint32_t, height in pixels.
>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>> If you have to support HDMI/SDTV inputs as well, then you also need to know
>>> the interlaced format, unless you have no plans to support that.
>>>
>>>> + * num_bufs - uint8_t, desired number of buffers to be used.
>>> Huh? What has that to do with the format? Why would you need this here?
>> Well, the operation name is 'set_config', not 'set_format',
>> so I thought we can have such a cumulative command assembling
>> all the parameters of the configuration. But now I am looking at
>> turning this single 'set_config' command to 3 different commands,
>> which is more practical and aligned with V4L2 in particular (please
>> see above in the command set):
>> 1. set format command:
>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>> * width - uint32_t, width in pixels.
>> * height - uint32_t, height in pixels.
>>
>> 2. Set frame rate command:
>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>
>> 3. Set/request num bufs:
>> * num_bufs - uint8_t, desired number of buffers to be used.
> I like this much better. 1+2 could be combined, but 3 should definitely remain
> separate.
ok, then 1+2 combined + 3 separate.
Do you think we can still name 1+2 as "set_format" or "set_config"
will fit better?
>
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + * Notes:
>>>> + * - frontend must check the corresponding response in order to see
>>>> + * if the values reported back by the backend do match the desired ones
>>>> + * and can be accepted.
>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>> + * configuration.
>>>> + */
>>>> +struct xencamera_config {
>>>> + uint32_t pixel_format;
>>>> + uint32_t width;
>>>> + uint32_t height;
>>>> + uint32_t frame_rate_nom;
>>>> + uint32_t frame_rate_denom;
>>>> + uint8_t num_bufs;
>>>> +};
>>>> +
>>>> +/*
>>>> + * Request buffer details - request camera buffer's memory layout.
>>>> + * detailed description:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + *
>>>> + * Request camera buffer creation:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref_directory | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>> + *
>>>> + * index - uint8_t, index of the buffer to be created.
>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>> + * then reference to the next shared page must be supplied (see
>>>> + * gref_dir_next_page below).
>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>> does.
>> Well, I still think it is better to have a per buffer interface
>> in the protocol as it is done for other Xen virtual devices.
>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>> what it does internally in the frontend driver
> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
> sense. I'm not sure about the naming.
>
> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
> on the backend, and so can fail. Also, the actual number of allocated buffers in
> case of success can be more or less than what was requested.
The buffers can be allocated and shared by either backend or frontend: see
"be-alloc" configuration option telling which domain (VM) shares
the Xen grant references to the pages of the buffer: either frontend
or backend.
So, I was more thinking that in case of V4L2 based frontend driver:
1. Frontend serves REQBUFS ioctl and asks the backend with
XENCAMERA_OP_BUF_REQUEST
if it can handle that many buffers and gets number of buffers to be used
and buffer structure (number of planes, sizes, offsets etc.) as the reply
to that request
2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
3. Frontend returns from REQBUFS ioctl with actual number of buffers
allocated
>>>> + *
>>>> + * If XENCAMERA_FIELD_BE_ALLOC configuration entry is set, then backend will
>>>> + * allocate the buffer with the parameters provided in this request and page
>>>> + * directory is handled as follows:
>>>> + * Frontend on request:
>>>> + * - allocates pages for the directory (gref_directory,
>>>> + * gref_dir_next_page(s)
>>>> + * - grants permissions for the pages of the directory to the backend
>>>> + * - sets gref_dir_next_page fields
>>>> + * Backend on response:
>>>> + * - grants permissions for the pages of the buffer allocated to
>>>> + * the frontend
>>>> + * - fills in page directory with grant references
>>>> + * (gref[] in struct xencamera_page_directory)
>>>> + */
>>>> +struct xencamera_buf_create_req {
>>>> + uint8_t index;
>>>> + uint8_t reserved[3];
>>>> + grant_ref_t gref_directory;
>>>> +};
>>>> +
>>>> +/*
>>>> + * Shared page for XENCAMERA_OP_BUF_CREATE buffer descriptor (gref_directory in
>>>> + * the request) employs a list of pages, describing all pages of the shared
>>>> + * data buffer:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref_dir_next_page | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref[0] | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref[i] | i*4+8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | gref[N - 1] | N*4+8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * gref_dir_next_page - grant_ref_t, reference to the next page describing
>>>> + * page directory. Must be 0 if there are no more pages in the list.
>>>> + * gref[i] - grant_ref_t, reference to a shared page of the buffer
>>>> + * allocated at XENCAMERA_OP_BUF_CREATE.
>>>> + *
>>>> + * Number of grant_ref_t entries in the whole page directory is not
>>>> + * passed, but instead can be calculated as:
>>>> + * num_grefs_total = (XENCAMERA_OP_GET_BUF_DETAILS.size + XEN_PAGE_SIZE - 1) /
>>>> + * XEN_PAGE_SIZE
>>>> + */
>>>> +struct xencamera_page_directory {
>>>> + grant_ref_t gref_dir_next_page;
>>>> + grant_ref_t gref[1]; /* Variable length */
>>>> +};
>>>> +
>>>> +/*
>>>> + * Request buffer destruction - destroy a previously allocated camera buffer:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_BUF_DESTROY| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * index - uint8_t, index of the buffer to be destroyed.
>>>> + */
>>> There is no V4L2 ioctl to destroy specific buffers. You can only destroy all
>>> of them.
>> This is not specifically related to V4L2, but can be issued
>> in response to backend's state change etc.
>> So, we have a pair of commands to create and destroy buffers.
>> Even more, frontend can be a some-os-based-driver, not V4L2
>> based. Or even a user-space application if your will.
>>>> +
>>>> +struct xencamera_buf_destroy_req {
>>>> + uint8_t index;
>>>> +};
>>>> +
>>>> +/*
>>>> + * Request camera capture stream start:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_OP_STREAM_START| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + *
>>>> + * Request camera capture stream stop:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_OP_STREAM_STOP | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + *
>>>> + * Request camera control details:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + * index - uint8_t, index of the control to be queried.
>>>> + */
>>>> +struct xencamera_get_ctrl_details_req {
>>>> + uint8_t index;
>>>> +};
>>>> +
>>>> +/*
>>>> + *
>>>> + * Request camera control change:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_SET_CTRL | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | value | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * See response format for this request.
>>>> + *
>>>> + * index - uint8_t, index of the control.
>>>> + * value - int32_t, new value of the control.
>>> I would recommend using a int64_t as the control value.
>> Good point, thank you
>>> Note that there are also controls with a payload (e.g. string controls).
>> Could you please give me an example of such a control?
>> Do you think such controls can be of use in a VM?
>> Can we avoid such controls if we target a simple virtual
>> camera device? If this is for radio use-case, then we'll
>> have such support in 'vradio' protocol if need be
> Right now all string controls are related to RDS receivers/transmitters. If you
> ever decide on a vradio protocol, then you need these (Programme Service name
> and Radio Text info). And there is an array of Alternate Frequencies, also RDS
> specific.
>
> There are some array controls in V4L2, those are used to control motion detection
> for surveillance cameras. And 'compound controls' (think of this as C structs) are
> appearing for HW codecs.
>
> I don't think any of these are likely to appear for cameras, at least not in a
> way that is relevant for Xen.
ok
>>> If there is ever interest in adding radio/RDS support, then that will become
>>> an issue.
>> You mean something like station names, ads etc?
> Yup.
>
>>>> + */
>>>> +struct xencamera_set_ctrl_req {
>>>> + uint8_t index;
>>>> + uint8_t reserved[3];
>>>> + int32_t value;
>>>> +};
>>>> +
>>>> +/*
>>>> + *---------------------------------- Responses --------------------------------
>>>> + *
>>>> + * All response packets have the same length (64 octets).
>>>> + *
>>>> + * All response packets have common header:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | operation | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * id - uint16_t, copied from the request.
>>>> + * operation - uint8_t, XENCAMERA_OP_* - copied from request.
>>>> + * status - int32_t, response status, zero on success and -XEN_EXX on failure.
>>>> + *
>>>> + *
>>>> + * Set configuration response - response for XENCAMERA_OP_SET_CONFIG:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | _OP_SET_CONFIG | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | pixel format | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | width | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | height | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_numer | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | frame_rate_denom | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | num_bufs | reserved | 32
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * Meaning of the corresponding values in this response is the same as for
>>>> + * XENCAMERA_OP_SET_CONFIG request.
>>>> + *
>>>> + *
>>>> + * Request buffer details response - response for XENCAMERA_OP_GET_BUF_DETAILS
>>>> + * request:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | size | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | num_planes | reserved | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[0] | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[1] | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[2] | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_offset[3] | 32
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[0] | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[1] | 40
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[2] | 44
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_size[3] | 48
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_stride[0] | plane_stride[1] | 52
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | plane_stride[2] | plane_stride[3] | 56
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 60
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * size - uint32_t, overall size of the buffer including sizes of the
>>>> + * individual planes and padding if applicable.
>>>> + * num_planes - uint8_t, number of planes for this buffer.
>>>> + * plane_offset - array of uint32_t, offset of the corresponding plane
>>>> + * in octets from the buffer start.
>>>> + * plane_size - array of uint32_t, size in octets of the corresponding plane
>>>> + * including padding.
>>>> + * plane_stride - array of uint32_t, size in octets occupied by the
>>>> + * corresponding single image line including padding if applicable.
>>> Nice!
>> Thank you
>>>> + */
>>>> +struct xencamera_buf_details_resp {
>>>> + uint32_t size;
>>>> + uint8_t num_planes;
>>>> + uint8_t reserved[3];
>>>> + uint32_t plane_offset[XENCAMERA_MAX_PLANE];
>>>> + uint32_t plane_size[XENCAMERA_MAX_PLANE];
>>>> + uint16_t plane_stride[XENCAMERA_MAX_PLANE];
>>>> +};
>>>> +
>>>> +/*
>>>> + * Get control details response - response for XENCAMERA_OP_GET_CTRL_DETAILS:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |GET_CTRL_DETAILS| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | status | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | type | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | min | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | max | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | step | 24
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | def_val | 28
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 36
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * index - uint8_t, index of the camera control in response.
>>>> + * type - uint8_t, type of the control, one of the XENCAMERA_CTRL_XXX.
>>>> + * min - int32_t, minimum value of the control.
>>>> + * max - int32_t, maximum value of the control.
>>>> + * step - int32_t, minimum size in which control value can be changed.
>>>> + * def_val - int32_t, default value of the control.
>>> I'd go with 64 bit values for min/max/step/def_val.
>> Sure, good idea, thank you
>>> I would also add a flags field. Some controls are read-only, write-only
>>> or volatile, things userspace needs to know.
>> Then I'll also add numerical constants for such
>>> If you want to support menu controls, then you need a way to get the menu
>>> names as well (VIDIOC_QUERYMENU).
>>>
>>> None of this is needed for this initial use-case, but you need to think
>>> about this up-front.
>> Yes, thank you
>>>> + */
>>>> +struct xencamera_get_ctrl_details_resp {
>>>> + uint8_t index;
>>>> + uint8_t type;
>>>> + uint8_t reserved[2];
>>>> + int32_t min;
>>>> + int32_t max;
>>>> + int32_t step;
>>>> + int32_t def_val;
>>>> +};
>>>> +
>>>> +/*
>>>> + *----------------------------------- Events ----------------------------------
>>>> + *
>>>> + * Events are sent via a shared page allocated by the front and propagated by
>>>> + * evt-event-channel/evt-ring-ref XenStore entries.
>>>> + *
>>>> + * All event packets have the same length (64 octets).
>>>> + * All event packets have common header:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id | type | reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * id - uint16_t, event id, may be used by front.
>>>> + * type - uint8_t, type of the event.
>>>> + *
>>>> + *
>>>> + * Frame captured event - event from back to front when a new captured
>>>> + * frame is available:
>>>> + * 0 1 2 3 octet
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | id |_EVT_FRAME_AVAIL| reserved | 4
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 8
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | index | reserved | 12
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | used_sz | 16
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 20
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + * | reserved | 64
>>>> + * +----------------+----------------+----------------+----------------+
>>>> + *
>>>> + * index - uint8_t, index of the buffer that contains new captured frame.
>>>> + * used_sz - uint32_t, number of octets this frame has. This can be less
>>>> + * than the XENCAMERA_OP_GET_BUF_DETAILS.size for compressed formats.
>>>> + */
>>>> +struct xencamera_frame_avail_evt {
>>>> + uint8_t index;
>>>> + uint8_t reserved[3];
>>>> + uint32_t used_sz;
>>>> +};
>>>> +
>>>> +struct xencamera_req {
>>>> + uint16_t id;
>>>> + uint8_t operation;
>>>> + uint8_t reserved[5];
>>>> + union {
>>>> + struct xencamera_config config;
>>>> + struct xencamera_buf_create_req buf_create;
>>>> + struct xencamera_buf_destroy_req buf_destroy;
>>>> + struct xencamera_set_ctrl_req set_ctrl;
>>>> + uint8_t reserved[56];
>>>> + } req;
>>>> +};
>>>> +
>>>> +struct xencamera_resp {
>>>> + uint16_t id;
>>>> + uint8_t operation;
>>>> + uint8_t reserved;
>>>> + int32_t status;
>>>> + union {
>>>> + struct xencamera_config config;
>>>> + struct xencamera_buf_details_resp buf_details;
>>>> + struct xencamera_get_ctrl_details_resp ctrl_details;
>>>> + uint8_t reserved1[56];
>>>> + } resp;
>>>> +};
>>>> +
>>>> +struct xencamera_evt {
>>>> + uint16_t id;
>>>> + uint8_t type;
>>>> + uint8_t reserved[5];
>>>> + union {
>>>> + struct xencamera_frame_avail_evt frame_avail;
>>>> + uint8_t reserved[56];
>>>> + } evt;
>>>> +};
>>>> +
>>>> +DEFINE_RING_TYPES(xen_cameraif, struct xencamera_req, struct xencamera_resp);
>>>> +
>>>> +/*
>>>> + ******************************************************************************
>>>> + * Back to front events delivery
>>>> + ******************************************************************************
>>>> + * In order to deliver asynchronous events from back to front a shared page is
>>>> + * allocated by front and its granted reference propagated to back via
>>>> + * XenStore entries (evt-ring-ref/evt-event-channel).
>>>> + * This page has a common header used by both front and back to synchronize
>>>> + * access and control event's ring buffer, while back being a producer of the
>>>> + * events and front being a consumer. The rest of the page after the header
>>>> + * is used for event packets.
>>>> + *
>>>> + * Upon reception of an event(s) front may confirm its reception
>>>> + * for either each event, group of events or none.
>>>> + */
>>>> +
>>>> +struct xencamera_event_page {
>>>> + uint32_t in_cons;
>>>> + uint32_t in_prod;
>>>> + uint8_t reserved[56];
>>>> +};
>>>> +
>>>> +#define XENCAMERA_EVENT_PAGE_SIZE 4096
>>>> +#define XENCAMERA_IN_RING_OFFS (sizeof(struct xencamera_event_page))
>>>> +#define XENCAMERA_IN_RING_SIZE (XENCAMERA_EVENT_PAGE_SIZE - XENCAMERA_IN_RING_OFFS)
>>>> +#define XENCAMERA_IN_RING_LEN (XENCAMERA_IN_RING_SIZE / sizeof(struct xencamera_evt))
>>>> +#define XENCAMERA_IN_RING(page) \
>>>> + ((struct xencamera_evt *)((char *)(page) + XENCAMERA_IN_RING_OFFS))
>>>> +#define XENCAMERA_IN_RING_REF(page, idx) \
>>>> + (XENCAMERA_IN_RING((page))[(idx) % XENCAMERA_IN_RING_LEN])
>>>> +
>>>> +#endif /* __XEN_PUBLIC_IO_CAMERAIF_H__ */
>>>> +
>>>> +/*
>>>> + * Local variables:
>>>> + * mode: C
>>>> + * c-file-style: "BSD"
>>>> + * c-basic-offset: 4
>>>> + * tab-width: 4
>>>> + * indent-tabs-mode: nil
>>>> + * End:
>>>> + */
>>>>
>>> I think the most important decision to make here is whether or not you want to support
>>> hotpluggable sources like HDMI. And an additional complication with that is HDCP.
>>> While V4L2 doesn't have an API for HDCP at the moment, Cisco is working on this and
>>> a patch series adding this is expected later this year/early next year. It might not
>>> be an issue in practice if these are all closed systems, but nevertheless, it is
>>> something to think about.
>> Yes, thank you for raising these questions, it is worth thinking
>> about such use-cases.
>>> Regards,
>>>
>>> Hans
>> Thank you so much for the comments,
>> Oleksandr
>>
> Regards,
>
> Hans
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 8:14 ` [PATCH " Hans Verkuil
@ 2018-09-10 8:34 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 8:34 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr Andrushchenko, Juergen Gross, xen-devel,
konrad.wilk, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Artem Mygaiev
On 09/10/2018 11:14 AM, Hans Verkuil wrote:
> On 09/10/2018 07:59 AM, Oleksandr Andrushchenko wrote:
>> Hi, Hans!
>>
>> On 09/09/2018 01:42 PM, Hans Verkuil wrote:
>>> On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>>>>> Hi Oleksandr,
>>>>>
>>>>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>>>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>>>>
>>>>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>>>>> high definition maps etc.
>>>>>>>>>
>>>>>>>>> The initial goal is to support most needed functionality with the
>>>>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>>>>
>>>>>>>>> 1. Provide means for base virtual device configuration:
>>>>>>>>> - pixel formats
>>>>>>>>> - resolutions
>>>>>>>>> - frame rates
>>>>>>>>> 2. Support basic camera controls:
>>>>>>>>> - contrast
>>>>>>>>> - brightness
>>>>>>>>> - hue
>>>>>>>>> - saturation
>>>>>>>>> 3. Support streaming control
>>>>>>>>> 4. Support zero-copying use-cases
>>>>>>>>>
>>>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>>>>> Some style issues below...
>>>>>>> Will fix all the below, thank you!
>>>>>>>
>>>>>>> I would like to draw some attention of the Linux/V4L community to this
>>>>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>>>>> upstream a Linux camera front-end kernel driver which will be based
>>>>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>>>>> as well)
>>>>>> ping
>>>>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>>>>> to review this this week, if not, just ping me again.
>>>> Thank you for your time
>>>>> I had one high-level question, though:
>>>>>
>>>>> What types of hardware do you intend to target? This initial version targets
>>>>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>>>>> codecs? Or complex embedded video pipelines?
>>>>>
>>>>> In other words, where are you planning to draw the line?
>>>>>
>>>>> Even with just simple cameras there is a difference between regular UVC
>>>>> webcams and cameras used with embedded systems: for the latter you often
>>>>> need to provide more control w.r.t. white-balancing etc., things that a
>>>>> UVC webcam will generally do for you in the webcam's firmware.
>>>> The use-cases we want to implement are mostly in automotive/embedded domain,
>>>> so there are many performance restrictions apply.
>>>> We are not targeting virtualizing very complex hardware and have no
>>>> intention
>>>> to make a 1:1 mapping of the real hardware: for that one can pass-through
>>>> a real HW device to a virtual machine (VM). The goal is to share a single
>>>> camera device to multiple virtual machines, no codecs, receivers etc.
>>>>
>>>> Controlling the same HW device from different VMs doesn't look feasible:
>>>> what if the same control is set to different values from different VMs?
>>> You can do this, actually: in V4L2 you can get an event when another process
>>> changes a control, and update your own GUI/internal state accordingly.
>>>
>>> So in this case if one VM changes a control, an event is sent to all others
>>> that the control has changed value.
>> Well, technically this can be done by introducing one more
>> event for such a notification. But, from system partitioning
>> POV, I am still not convinced this should be done: I would prefer
>> that a single VM owns such a control and even which control and which
>> VM is decided while configuring the whole system.
>> So, I would like to keep it as is.
> Well, I am not convinced you can avoid this: some controls are set by drivers
> when something happens and so applications will subscribe to the control and
> wait for changes in their values. While this is for more advanced use-cases
> (certainly more than what you propose today), I would suggest that it is wise
> to at least think on how this can be added in the future.
>
> Controls and control events are a key part of V4L2.
>
ok, then I'll add
#define XENCAMERA_EVT_CTRL_CHANGE 0x02
>>>> Of course, this can be achieved if the corresponding backend can
>>>> post-process
>>>> original camera image with GPU, for example, thus applying different filters
>>>> for different VMs effectively emulating camera controls.
>>>> But this requires additional CPU/GPU power which we try to avoid.
>>>>
>>>> System partitioning (camera and controls assignment) is done at
>>>> configuration
>>>> time (remember we are in automotive/embedded world, so most of the time
>>>> the set
>>>> of VMs requiring cameras is known at this stage and the configuration
>>>> remains
>>>> static at run-time). So, when para-virtualized (PV) approach is used then we
>>>> only implement very basic controls (those found in the protocol), so one can
>>>> assign set of controls (all or some) to one of the VMs (main or mission
>>>> critical
>>>> VM or whatever) allowing that VM to adjusts those for all VMs at once.
>>>> For other
>>>> VMs think of it as firmware implemented adjustment. And the backend still
>>>> controls the rest of the controls of the real HW camera you mention.
>>>>
>>>> Just an example of automotive use-case (we can imagine many more):
>>>> 1. Driver Domain - owns real camera HW and runs the camera backend.
>>>> Uses camera output for mission critical tasks, e.g. parking assistance.
>>>> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
>>>> purposes,
>>>> e.g. taking pictures while in motion.
>>>> 3. Navigation domain - uses PV camera for high definition maps
>>>>
>>>> Hope, this helps understanding the possible uses of the proposed
>>>> protocol, its
>>>> intention and restrictions.
>>> Right, so in this scenario you probably do not want hotpluggable
>>> sources in the Driver Domain. So support for fixed camera's only.
>> Well, some sort of hotplug can already be implemented, please
>> see [1], [2] as it is done for virtual display: this is
>> achieved as a response to the backend's state change,
>> e.g. whenever backend decides to unplug the virtual device
>> it changes its state accordingly.
>>> If this is indeed the case, then this should be made very clear in
>>> the API specification.
>> As I described above this is already assumed by the state
>> machine of a xenbus_driver
>>> One additional thing to consider: cameras can break. So what should be
>>> done if that happens? We as media developers have ideas about that, but
>>> nothing has been implemented (yet).
>>>
>>> If the HW is simple (one camera is driven by a single driver instance),
>>> then if it breaks, there simply won't be a video device. But if you have
>>> multiple cameras all controlled through the same driver instance, then today
>>> if a single camera breaks, all are gone.
>> Please see above
> I'm not convinced this is sufficient. But it really would have to be tested
> with an actual HDMI source.
>
> A key difference between supporting a video input connector such as HDMI
> compared to a fixed camera or video output is that you do not control what
> you receive. I've been working with this for many years, and it is amazing
> what problems you can get due to this.
>
> So received frames can be corrupt (make sure you can handle the
> V4L2_BUF_FLAG_ERROR flag as result of a VIDIOC_DQBUF!) or the DMA engine can get
> into a bad state, requiring you to stop streaming and reconfigure (EIO returned
> by VIDIOC_(D)QBUF).
This all seems to be a backend problem: if a frame is corrupt
then backend can simply drop this frame and don't even report
it to the frontends, e.g. we assume that all this is handled
by the backend and this is its responsibility to always deliver
"good" frames.
Another issue is with "requiring you to stop streaming and reconfigure
(EIO returned
by VIDIOC_(D)QBUF)". If the error is not recoverable then
backend can go into XenbusStateClosed and destroy the virtual
camera, otherwise it can try recovering making it all transparent
for the frontend.
> A good driver to test such errors with is vivid, which can emulate webcams, SDTV or
> an HDMI receiver and has error injection to test these scenarios. See here for
> extensive documentation of this driver:
>
> https://hverkuil.home.xs4all.nl/spec/v4l-drivers/vivid.html
>
> I would recommend that you use this driver as a test HW driver since it can
> be configured in many different ways and is representative of actual HW.
>
> Exception is for HDMI where connect/disconnect is not properly emulated yet.
> If you need that, let me know.
>
> Regards,
>
> Hans
>
>>> We have ideas on how to address that, but as I said, nothing is implemented
>>> yet. Basically we need to allow for partial bring-up and inform userspace
>>> what is and what is not running.
>>>
>>> But this is likely something you also need to consider in this API, given
>>> the use-case you are looking at.
>>>
>>> Regards,
>>>
>>> Hans
>> Thank you for your valuable comments,
>> Oleksandr
>>
>> [1]
>> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L721
>> [2]
>> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L582
>>
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 8:34 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 8:34 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr Andrushchenko, Juergen Gross, xen-devel,
konrad.wilk, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Artem Mygaiev
On 09/10/2018 11:14 AM, Hans Verkuil wrote:
> On 09/10/2018 07:59 AM, Oleksandr Andrushchenko wrote:
>> Hi, Hans!
>>
>> On 09/09/2018 01:42 PM, Hans Verkuil wrote:
>>> On 09/04/2018 08:56 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/03/2018 06:25 PM, Hans Verkuil wrote:
>>>>> Hi Oleksandr,
>>>>>
>>>>> On 09/03/2018 12:16 PM, Oleksandr Andrushchenko wrote:
>>>>>> On 08/21/2018 08:54 AM, Oleksandr Andrushchenko wrote:
>>>>>>> On 08/14/2018 11:30 AM, Juergen Gross wrote:
>>>>>>>> On 31/07/18 11:31, Oleksandr Andrushchenko wrote:
>>>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>>>>>>>>
>>>>>>>>> This is the ABI for the two halves of a para-virtualized
>>>>>>>>> camera driver which extends Xen's reach multimedia capabilities even
>>>>>>>>> farther enabling it for video conferencing, In-Vehicle Infotainment,
>>>>>>>>> high definition maps etc.
>>>>>>>>>
>>>>>>>>> The initial goal is to support most needed functionality with the
>>>>>>>>> final idea to make it possible to extend the protocol if need be:
>>>>>>>>>
>>>>>>>>> 1. Provide means for base virtual device configuration:
>>>>>>>>> - pixel formats
>>>>>>>>> - resolutions
>>>>>>>>> - frame rates
>>>>>>>>> 2. Support basic camera controls:
>>>>>>>>> - contrast
>>>>>>>>> - brightness
>>>>>>>>> - hue
>>>>>>>>> - saturation
>>>>>>>>> 3. Support streaming control
>>>>>>>>> 4. Support zero-copying use-cases
>>>>>>>>>
>>>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>>>> <oleksandr_andrushchenko@epam.com>
>>>>>>>> Some style issues below...
>>>>>>> Will fix all the below, thank you!
>>>>>>>
>>>>>>> I would like to draw some attention of the Linux/V4L community to this
>>>>>>> protocol as the plan is that once it is accepted for Xen we plan to
>>>>>>> upstream a Linux camera front-end kernel driver which will be based
>>>>>>> on this work and will be a V4L2 device driver (this is why I have sent
>>>>>>> this patch not only to Xen, but to the corresponding Linux mailing list
>>>>>>> as well)
>>>>>> ping
>>>>> Sorry, this got buried in my mailbox, I only came across it today. I'll try
>>>>> to review this this week, if not, just ping me again.
>>>> Thank you for your time
>>>>> I had one high-level question, though:
>>>>>
>>>>> What types of hardware do you intend to target? This initial version targets
>>>>> (very) simple webcams, but what about HDMI or SDTV receivers? Or hardware
>>>>> codecs? Or complex embedded video pipelines?
>>>>>
>>>>> In other words, where are you planning to draw the line?
>>>>>
>>>>> Even with just simple cameras there is a difference between regular UVC
>>>>> webcams and cameras used with embedded systems: for the latter you often
>>>>> need to provide more control w.r.t. white-balancing etc., things that a
>>>>> UVC webcam will generally do for you in the webcam's firmware.
>>>> The use-cases we want to implement are mostly in automotive/embedded domain,
>>>> so there are many performance restrictions apply.
>>>> We are not targeting virtualizing very complex hardware and have no
>>>> intention
>>>> to make a 1:1 mapping of the real hardware: for that one can pass-through
>>>> a real HW device to a virtual machine (VM). The goal is to share a single
>>>> camera device to multiple virtual machines, no codecs, receivers etc.
>>>>
>>>> Controlling the same HW device from different VMs doesn't look feasible:
>>>> what if the same control is set to different values from different VMs?
>>> You can do this, actually: in V4L2 you can get an event when another process
>>> changes a control, and update your own GUI/internal state accordingly.
>>>
>>> So in this case if one VM changes a control, an event is sent to all others
>>> that the control has changed value.
>> Well, technically this can be done by introducing one more
>> event for such a notification. But, from system partitioning
>> POV, I am still not convinced this should be done: I would prefer
>> that a single VM owns such a control and even which control and which
>> VM is decided while configuring the whole system.
>> So, I would like to keep it as is.
> Well, I am not convinced you can avoid this: some controls are set by drivers
> when something happens and so applications will subscribe to the control and
> wait for changes in their values. While this is for more advanced use-cases
> (certainly more than what you propose today), I would suggest that it is wise
> to at least think on how this can be added in the future.
>
> Controls and control events are a key part of V4L2.
>
ok, then I'll add
#define XENCAMERA_EVT_CTRL_CHANGE 0x02
>>>> Of course, this can be achieved if the corresponding backend can
>>>> post-process
>>>> original camera image with GPU, for example, thus applying different filters
>>>> for different VMs effectively emulating camera controls.
>>>> But this requires additional CPU/GPU power which we try to avoid.
>>>>
>>>> System partitioning (camera and controls assignment) is done at
>>>> configuration
>>>> time (remember we are in automotive/embedded world, so most of the time
>>>> the set
>>>> of VMs requiring cameras is known at this stage and the configuration
>>>> remains
>>>> static at run-time). So, when para-virtualized (PV) approach is used then we
>>>> only implement very basic controls (those found in the protocol), so one can
>>>> assign set of controls (all or some) to one of the VMs (main or mission
>>>> critical
>>>> VM or whatever) allowing that VM to adjusts those for all VMs at once.
>>>> For other
>>>> VMs think of it as firmware implemented adjustment. And the backend still
>>>> controls the rest of the controls of the real HW camera you mention.
>>>>
>>>> Just an example of automotive use-case (we can imagine many more):
>>>> 1. Driver Domain - owns real camera HW and runs the camera backend.
>>>> Uses camera output for mission critical tasks, e.g. parking assistance.
>>>> 2. In-Vehicle Infotainment domain - uses PV camera for infotainment
>>>> purposes,
>>>> e.g. taking pictures while in motion.
>>>> 3. Navigation domain - uses PV camera for high definition maps
>>>>
>>>> Hope, this helps understanding the possible uses of the proposed
>>>> protocol, its
>>>> intention and restrictions.
>>> Right, so in this scenario you probably do not want hotpluggable
>>> sources in the Driver Domain. So support for fixed camera's only.
>> Well, some sort of hotplug can already be implemented, please
>> see [1], [2] as it is done for virtual display: this is
>> achieved as a response to the backend's state change,
>> e.g. whenever backend decides to unplug the virtual device
>> it changes its state accordingly.
>>> If this is indeed the case, then this should be made very clear in
>>> the API specification.
>> As I described above this is already assumed by the state
>> machine of a xenbus_driver
>>> One additional thing to consider: cameras can break. So what should be
>>> done if that happens? We as media developers have ideas about that, but
>>> nothing has been implemented (yet).
>>>
>>> If the HW is simple (one camera is driven by a single driver instance),
>>> then if it breaks, there simply won't be a video device. But if you have
>>> multiple cameras all controlled through the same driver instance, then today
>>> if a single camera breaks, all are gone.
>> Please see above
> I'm not convinced this is sufficient. But it really would have to be tested
> with an actual HDMI source.
>
> A key difference between supporting a video input connector such as HDMI
> compared to a fixed camera or video output is that you do not control what
> you receive. I've been working with this for many years, and it is amazing
> what problems you can get due to this.
>
> So received frames can be corrupt (make sure you can handle the
> V4L2_BUF_FLAG_ERROR flag as result of a VIDIOC_DQBUF!) or the DMA engine can get
> into a bad state, requiring you to stop streaming and reconfigure (EIO returned
> by VIDIOC_(D)QBUF).
This all seems to be a backend problem: if a frame is corrupt
then backend can simply drop this frame and don't even report
it to the frontends, e.g. we assume that all this is handled
by the backend and this is its responsibility to always deliver
"good" frames.
Another issue is with "requiring you to stop streaming and reconfigure
(EIO returned
by VIDIOC_(D)QBUF)". If the error is not recoverable then
backend can go into XenbusStateClosed and destroy the virtual
camera, otherwise it can try recovering making it all transparent
for the frontend.
> A good driver to test such errors with is vivid, which can emulate webcams, SDTV or
> an HDMI receiver and has error injection to test these scenarios. See here for
> extensive documentation of this driver:
>
> https://hverkuil.home.xs4all.nl/spec/v4l-drivers/vivid.html
>
> I would recommend that you use this driver as a test HW driver since it can
> be configured in many different ways and is representative of actual HW.
>
> Exception is for HDMI where connect/disconnect is not properly emulated yet.
> If you need that, let me know.
>
> Regards,
>
> Hans
>
>>> We have ideas on how to address that, but as I said, nothing is implemented
>>> yet. Basically we need to allow for partial bring-up and inform userspace
>>> what is and what is not running.
>>>
>>> But this is likely something you also need to consider in this API, given
>>> the use-case you are looking at.
>>>
>>> Regards,
>>>
>>> Hans
>> Thank you for your valuable comments,
>> Oleksandr
>>
>> [1]
>> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L721
>> [2]
>> https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen/xen_drm_front.c#L582
>>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 8:24 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-10 9:04 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 9:04 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, jgross,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>> Hi Oleksandr,
>>
>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
<snip>
>>>> I suspect that you likely will want to support such sources eventually, so
>>>> it pays to design this with that in mind.
>>> Again, I think that this is the backend to hide these
>>> use-cases from the frontend.
>> I'm not sure you can: say you are playing a bluray connected to the system
>> with HDMI, then if there is a resolution change, what do you do? You can tear
>> everything down and build it up again, or you can just tell frontends that
>> something changed and that they have to look at the new vcamera configuration.
>>
>> The latter seems to be more sensible to me. It is really not much that you
>> need to do: all you really need is an event signalling that something changed.
>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
> well, this complicates things a lot as I'll have to
> re-allocate buffers - right?
Right. Different resolutions means different sized buffers and usually lots of
changes throughout the whole video pipeline, which in this case can even
go into multiple VMs.
One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
has a flags field that tells userspace what changed. Right now that is just the
resolution, but in the future you can expect flags for cases where just the
colorspace information changes, but not the resolution.
Which reminds me of two important missing pieces of information in your protocol:
1) You need to communicate the colorspace data:
- colorspace
- xfer_func
- ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
think you can ignore hsv_enc)
- quantization
See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
and the links to the colorspace sections in the V4L2 spec for details).
This information is part of the format, it is reported by the driver.
2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
each buffer contains a single field), then you need to be able to tell
userspace whether the dequeued buffer contains a top or bottom field.
Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
timestamp that can help detecting that. You probably need something similar.
> But anyways, I can add
> #define XENCAMERA_EVT_CFG_CHANGE 0x01
> in the protocol, so we can address this use-case
>>
<snip>
>>> 1. set format command:
>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>> * width - uint32_t, width in pixels.
>>> * height - uint32_t, height in pixels.
>>>
>>> 2. Set frame rate command:
>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>
>>> 3. Set/request num bufs:
>>> * num_bufs - uint8_t, desired number of buffers to be used.
>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>> separate.
> ok, then 1+2 combined + 3 separate.
> Do you think we can still name 1+2 as "set_format" or "set_config"
> will fit better?
set_format is closer to S_FMT as used in V4L2, so I have a slight preference
for that, but it is really up to you.
>>
>>>>> + *
>>>>> + * See response format for this request.
>>>>> + *
>>>>> + * Notes:
>>>>> + * - frontend must check the corresponding response in order to see
>>>>> + * if the values reported back by the backend do match the desired ones
>>>>> + * and can be accepted.
>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>> + * configuration.
>>>>> + */
>>>>> +struct xencamera_config {
>>>>> + uint32_t pixel_format;
>>>>> + uint32_t width;
>>>>> + uint32_t height;
>>>>> + uint32_t frame_rate_nom;
>>>>> + uint32_t frame_rate_denom;
>>>>> + uint8_t num_bufs;
>>>>> +};
>>>>> +
>>>>> +/*
>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>> + * detailed description:
>>>>> + * 0 1 2 3 octet
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 8
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 64
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + *
>>>>> + * See response format for this request.
>>>>> + *
>>>>> + *
>>>>> + * Request camera buffer creation:
>>>>> + * 0 1 2 3 octet
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 8
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | index | reserved | 12
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | gref_directory | 16
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 20
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 64
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + *
>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>> + *
>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>> + * then reference to the next shared page must be supplied (see
>>>>> + * gref_dir_next_page below).
>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>> does.
>>> Well, I still think it is better to have a per buffer interface
>>> in the protocol as it is done for other Xen virtual devices.
>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>> what it does internally in the frontend driver
>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>> sense. I'm not sure about the naming.
>>
>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>> case of success can be more or less than what was requested.
> The buffers can be allocated and shared by either backend or frontend: see
> "be-alloc" configuration option telling which domain (VM) shares
> the Xen grant references to the pages of the buffer: either frontend
> or backend.
If you want to do zero-copy video capture, then you need to know which
device in your video pipeline (which now covers both actual hardware and
multiple VMs) has the strictest memory layout requirements. Often the
video HW requires contiguous physical memory for the buffers, which means
you can't just give it a piece of non-contig memory allocated elsewhere.
In practice you have two possible memory models you can use with V4L2 drivers:
MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
elsewhere and imported to V4L2, which may fail if it doesn't match the HW
requirements.
>
> So, I was more thinking that in case of V4L2 based frontend driver:
> 1. Frontend serves REQBUFS ioctl and asks the backend with
> XENCAMERA_OP_BUF_REQUEST
> if it can handle that many buffers and gets number of buffers to be used
> and buffer structure (number of planes, sizes, offsets etc.) as the reply
> to that request
> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
> allocated
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 9:04 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 9:04 UTC (permalink / raw)
To: Oleksandr Andrushchenko, xen-devel, konrad.wilk, jgross,
boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Cc: Oleksandr Andrushchenko
On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>> Hi Oleksandr,
>>
>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
<snip>
>>>> I suspect that you likely will want to support such sources eventually, so
>>>> it pays to design this with that in mind.
>>> Again, I think that this is the backend to hide these
>>> use-cases from the frontend.
>> I'm not sure you can: say you are playing a bluray connected to the system
>> with HDMI, then if there is a resolution change, what do you do? You can tear
>> everything down and build it up again, or you can just tell frontends that
>> something changed and that they have to look at the new vcamera configuration.
>>
>> The latter seems to be more sensible to me. It is really not much that you
>> need to do: all you really need is an event signalling that something changed.
>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
> well, this complicates things a lot as I'll have to
> re-allocate buffers - right?
Right. Different resolutions means different sized buffers and usually lots of
changes throughout the whole video pipeline, which in this case can even
go into multiple VMs.
One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
has a flags field that tells userspace what changed. Right now that is just the
resolution, but in the future you can expect flags for cases where just the
colorspace information changes, but not the resolution.
Which reminds me of two important missing pieces of information in your protocol:
1) You need to communicate the colorspace data:
- colorspace
- xfer_func
- ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
think you can ignore hsv_enc)
- quantization
See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
and the links to the colorspace sections in the V4L2 spec for details).
This information is part of the format, it is reported by the driver.
2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
each buffer contains a single field), then you need to be able to tell
userspace whether the dequeued buffer contains a top or bottom field.
Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
timestamp that can help detecting that. You probably need something similar.
> But anyways, I can add
> #define XENCAMERA_EVT_CFG_CHANGE 0x01
> in the protocol, so we can address this use-case
>>
<snip>
>>> 1. set format command:
>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>> * width - uint32_t, width in pixels.
>>> * height - uint32_t, height in pixels.
>>>
>>> 2. Set frame rate command:
>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>
>>> 3. Set/request num bufs:
>>> * num_bufs - uint8_t, desired number of buffers to be used.
>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>> separate.
> ok, then 1+2 combined + 3 separate.
> Do you think we can still name 1+2 as "set_format" or "set_config"
> will fit better?
set_format is closer to S_FMT as used in V4L2, so I have a slight preference
for that, but it is really up to you.
>>
>>>>> + *
>>>>> + * See response format for this request.
>>>>> + *
>>>>> + * Notes:
>>>>> + * - frontend must check the corresponding response in order to see
>>>>> + * if the values reported back by the backend do match the desired ones
>>>>> + * and can be accepted.
>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>> + * configuration.
>>>>> + */
>>>>> +struct xencamera_config {
>>>>> + uint32_t pixel_format;
>>>>> + uint32_t width;
>>>>> + uint32_t height;
>>>>> + uint32_t frame_rate_nom;
>>>>> + uint32_t frame_rate_denom;
>>>>> + uint8_t num_bufs;
>>>>> +};
>>>>> +
>>>>> +/*
>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>> + * detailed description:
>>>>> + * 0 1 2 3 octet
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 8
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 64
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + *
>>>>> + * See response format for this request.
>>>>> + *
>>>>> + *
>>>>> + * Request camera buffer creation:
>>>>> + * 0 1 2 3 octet
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 8
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | index | reserved | 12
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | gref_directory | 16
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 20
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + * | reserved | 64
>>>>> + * +----------------+----------------+----------------+----------------+
>>>>> + *
>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>> + *
>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>> + * then reference to the next shared page must be supplied (see
>>>>> + * gref_dir_next_page below).
>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>> does.
>>> Well, I still think it is better to have a per buffer interface
>>> in the protocol as it is done for other Xen virtual devices.
>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>> what it does internally in the frontend driver
>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>> sense. I'm not sure about the naming.
>>
>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>> case of success can be more or less than what was requested.
> The buffers can be allocated and shared by either backend or frontend: see
> "be-alloc" configuration option telling which domain (VM) shares
> the Xen grant references to the pages of the buffer: either frontend
> or backend.
If you want to do zero-copy video capture, then you need to know which
device in your video pipeline (which now covers both actual hardware and
multiple VMs) has the strictest memory layout requirements. Often the
video HW requires contiguous physical memory for the buffers, which means
you can't just give it a piece of non-contig memory allocated elsewhere.
In practice you have two possible memory models you can use with V4L2 drivers:
MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
elsewhere and imported to V4L2, which may fail if it doesn't match the HW
requirements.
>
> So, I was more thinking that in case of V4L2 based frontend driver:
> 1. Frontend serves REQBUFS ioctl and asks the backend with
> XENCAMERA_OP_BUF_REQUEST
> if it can handle that many buffers and gets number of buffers to be used
> and buffer structure (number of planes, sizes, offsets etc.) as the reply
> to that request
> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
> allocated
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 9:04 ` [PATCH " Hans Verkuil
@ 2018-09-10 9:52 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 9:52 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/10/2018 12:04 PM, Hans Verkuil wrote:
> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>> Hi Oleksandr,
>>>
>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
> <snip>
>
>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>> it pays to design this with that in mind.
>>>> Again, I think that this is the backend to hide these
>>>> use-cases from the frontend.
>>> I'm not sure you can: say you are playing a bluray connected to the system
>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>> everything down and build it up again, or you can just tell frontends that
>>> something changed and that they have to look at the new vcamera configuration.
>>>
>>> The latter seems to be more sensible to me. It is really not much that you
>>> need to do: all you really need is an event signalling that something changed.
>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>> well, this complicates things a lot as I'll have to
>> re-allocate buffers - right?
> Right. Different resolutions means different sized buffers and usually lots of
> changes throughout the whole video pipeline, which in this case can even
> go into multiple VMs.
>
> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
> has a flags field that tells userspace what changed. Right now that is just the
> resolution, but in the future you can expect flags for cases where just the
> colorspace information changes, but not the resolution.
>
> Which reminds me of two important missing pieces of information in your protocol:
>
> 1) You need to communicate the colorspace data:
>
> - colorspace
> - xfer_func
> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
> think you can ignore hsv_enc)
> - quantization
>
> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
> and the links to the colorspace sections in the V4L2 spec for details).
>
> This information is part of the format, it is reported by the driver.
I'll take a look and think what can be put and how into the protocol,
do you think I'll have to implement all the above for
this stage?
>
> 2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
> each buffer contains a single field), then you need to be able to tell
> userspace whether the dequeued buffer contains a top or bottom field.
I think at the first stage we can assume that interlaced
formats are not supported and add such support later if need be.
>
> Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
> timestamp that can help detecting that. You probably need something similar.
Ok, this can be reported as part of XENCAMERA_EVT_FRAME_AVAIL event
>
>> But anyways, I can add
>> #define XENCAMERA_EVT_CFG_CHANGE 0x01
>> in the protocol, so we can address this use-case
> <snip>
>
>>>> 1. set format command:
>>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>> * width - uint32_t, width in pixels.
>>>> * height - uint32_t, height in pixels.
>>>>
>>>> 2. Set frame rate command:
>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>>
>>>> 3. Set/request num bufs:
>>>> * num_bufs - uint8_t, desired number of buffers to be used.
>>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>>> separate.
>> ok, then 1+2 combined + 3 separate.
>> Do you think we can still name 1+2 as "set_format" or "set_config"
>> will fit better?
> set_format is closer to S_FMT as used in V4L2, so I have a slight preference
> for that, but it is really up to you.
I'll probably stick to SET_CONFIG here
>
>>>>>> + *
>>>>>> + * See response format for this request.
>>>>>> + *
>>>>>> + * Notes:
>>>>>> + * - frontend must check the corresponding response in order to see
>>>>>> + * if the values reported back by the backend do match the desired ones
>>>>>> + * and can be accepted.
>>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>>> + * configuration.
>>>>>> + */
>>>>>> +struct xencamera_config {
>>>>>> + uint32_t pixel_format;
>>>>>> + uint32_t width;
>>>>>> + uint32_t height;
>>>>>> + uint32_t frame_rate_nom;
>>>>>> + uint32_t frame_rate_denom;
>>>>>> + uint8_t num_bufs;
>>>>>> +};
>>>>>> +
>>>>>> +/*
>>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>>> + * detailed description:
>>>>>> + * 0 1 2 3 octet
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 8
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 64
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + *
>>>>>> + * See response format for this request.
>>>>>> + *
>>>>>> + *
>>>>>> + * Request camera buffer creation:
>>>>>> + * 0 1 2 3 octet
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 8
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | index | reserved | 12
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | gref_directory | 16
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 20
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 64
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + *
>>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>>> + *
>>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>>> + * then reference to the next shared page must be supplied (see
>>>>>> + * gref_dir_next_page below).
>>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>>> does.
>>>> Well, I still think it is better to have a per buffer interface
>>>> in the protocol as it is done for other Xen virtual devices.
>>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>>> what it does internally in the frontend driver
>>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>>> sense. I'm not sure about the naming.
>>>
>>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>>> case of success can be more or less than what was requested.
>> The buffers can be allocated and shared by either backend or frontend: see
>> "be-alloc" configuration option telling which domain (VM) shares
>> the Xen grant references to the pages of the buffer: either frontend
>> or backend.
> If you want to do zero-copy video capture,
this is the goal
> then you need to know which
> device in your video pipeline (which now covers both actual hardware and
> multiple VMs) has the strictest memory layout requirements. Often the
> video HW requires contiguous physical memory for the buffers, which means
> you can't just give it a piece of non-contig memory allocated elsewhere.
We have already implemented zero copying use-cases for
virtual display, please see [1] and [2] which are dma-buf
based which can cope with real HW restrictions you mention.
And in that case we can implement zero-copying both ways,
e.g. when the Xen grant references are shared by either
backend or frontend. This is different from camera use-cases:
a single buffer needs to be shared with multiple frontends,
so zero-copying is only possible when backend allocates the references
and shares those with frontends. The way when frontend allocates
the buffers and still we can implement zero-copying is when
there is a single frontend in the system, otherwise we
need to copy the images from backend's buffers into frontend's
ones.
> In practice you have two possible memory models you can use with V4L2 drivers:
> MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
> as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
> elsewhere and imported to V4L2, which may fail if it doesn't match the HW
> requirements.
For the frontend it is possible to work with both MMAP/DMABUF
and the rest is on the backend's side - this was proven by
virtual display implementation, so I see no problem here
for virtual camera.
>
>> So, I was more thinking that in case of V4L2 based frontend driver:
>> 1. Frontend serves REQBUFS ioctl and asks the backend with
>> XENCAMERA_OP_BUF_REQUEST
>> if it can handle that many buffers and gets number of buffers to be used
>> and buffer structure (number of planes, sizes, offsets etc.) as the reply
>> to that request
>> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
>> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
>> allocated
> Regards,
>
> Hans
Thank you,
Oleksandr
[1] https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen
[2]
https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/xen/gntdev-dmabuf.c
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 9:52 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 9:52 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/10/2018 12:04 PM, Hans Verkuil wrote:
> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>> Hi Oleksandr,
>>>
>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
> <snip>
>
>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>> it pays to design this with that in mind.
>>>> Again, I think that this is the backend to hide these
>>>> use-cases from the frontend.
>>> I'm not sure you can: say you are playing a bluray connected to the system
>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>> everything down and build it up again, or you can just tell frontends that
>>> something changed and that they have to look at the new vcamera configuration.
>>>
>>> The latter seems to be more sensible to me. It is really not much that you
>>> need to do: all you really need is an event signalling that something changed.
>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>> well, this complicates things a lot as I'll have to
>> re-allocate buffers - right?
> Right. Different resolutions means different sized buffers and usually lots of
> changes throughout the whole video pipeline, which in this case can even
> go into multiple VMs.
>
> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
> has a flags field that tells userspace what changed. Right now that is just the
> resolution, but in the future you can expect flags for cases where just the
> colorspace information changes, but not the resolution.
>
> Which reminds me of two important missing pieces of information in your protocol:
>
> 1) You need to communicate the colorspace data:
>
> - colorspace
> - xfer_func
> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
> think you can ignore hsv_enc)
> - quantization
>
> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
> and the links to the colorspace sections in the V4L2 spec for details).
>
> This information is part of the format, it is reported by the driver.
I'll take a look and think what can be put and how into the protocol,
do you think I'll have to implement all the above for
this stage?
>
> 2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
> each buffer contains a single field), then you need to be able to tell
> userspace whether the dequeued buffer contains a top or bottom field.
I think at the first stage we can assume that interlaced
formats are not supported and add such support later if need be.
>
> Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
> timestamp that can help detecting that. You probably need something similar.
Ok, this can be reported as part of XENCAMERA_EVT_FRAME_AVAIL event
>
>> But anyways, I can add
>> #define XENCAMERA_EVT_CFG_CHANGE 0x01
>> in the protocol, so we can address this use-case
> <snip>
>
>>>> 1. set format command:
>>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>> * width - uint32_t, width in pixels.
>>>> * height - uint32_t, height in pixels.
>>>>
>>>> 2. Set frame rate command:
>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>>
>>>> 3. Set/request num bufs:
>>>> * num_bufs - uint8_t, desired number of buffers to be used.
>>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>>> separate.
>> ok, then 1+2 combined + 3 separate.
>> Do you think we can still name 1+2 as "set_format" or "set_config"
>> will fit better?
> set_format is closer to S_FMT as used in V4L2, so I have a slight preference
> for that, but it is really up to you.
I'll probably stick to SET_CONFIG here
>
>>>>>> + *
>>>>>> + * See response format for this request.
>>>>>> + *
>>>>>> + * Notes:
>>>>>> + * - frontend must check the corresponding response in order to see
>>>>>> + * if the values reported back by the backend do match the desired ones
>>>>>> + * and can be accepted.
>>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>>> + * configuration.
>>>>>> + */
>>>>>> +struct xencamera_config {
>>>>>> + uint32_t pixel_format;
>>>>>> + uint32_t width;
>>>>>> + uint32_t height;
>>>>>> + uint32_t frame_rate_nom;
>>>>>> + uint32_t frame_rate_denom;
>>>>>> + uint8_t num_bufs;
>>>>>> +};
>>>>>> +
>>>>>> +/*
>>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>>> + * detailed description:
>>>>>> + * 0 1 2 3 octet
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 8
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 64
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + *
>>>>>> + * See response format for this request.
>>>>>> + *
>>>>>> + *
>>>>>> + * Request camera buffer creation:
>>>>>> + * 0 1 2 3 octet
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 8
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | index | reserved | 12
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | gref_directory | 16
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 20
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + * | reserved | 64
>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>> + *
>>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>>> + *
>>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>>> + * then reference to the next shared page must be supplied (see
>>>>>> + * gref_dir_next_page below).
>>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>>> does.
>>>> Well, I still think it is better to have a per buffer interface
>>>> in the protocol as it is done for other Xen virtual devices.
>>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>>> what it does internally in the frontend driver
>>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>>> sense. I'm not sure about the naming.
>>>
>>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>>> case of success can be more or less than what was requested.
>> The buffers can be allocated and shared by either backend or frontend: see
>> "be-alloc" configuration option telling which domain (VM) shares
>> the Xen grant references to the pages of the buffer: either frontend
>> or backend.
> If you want to do zero-copy video capture,
this is the goal
> then you need to know which
> device in your video pipeline (which now covers both actual hardware and
> multiple VMs) has the strictest memory layout requirements. Often the
> video HW requires contiguous physical memory for the buffers, which means
> you can't just give it a piece of non-contig memory allocated elsewhere.
We have already implemented zero copying use-cases for
virtual display, please see [1] and [2] which are dma-buf
based which can cope with real HW restrictions you mention.
And in that case we can implement zero-copying both ways,
e.g. when the Xen grant references are shared by either
backend or frontend. This is different from camera use-cases:
a single buffer needs to be shared with multiple frontends,
so zero-copying is only possible when backend allocates the references
and shares those with frontends. The way when frontend allocates
the buffers and still we can implement zero-copying is when
there is a single frontend in the system, otherwise we
need to copy the images from backend's buffers into frontend's
ones.
> In practice you have two possible memory models you can use with V4L2 drivers:
> MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
> as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
> elsewhere and imported to V4L2, which may fail if it doesn't match the HW
> requirements.
For the frontend it is possible to work with both MMAP/DMABUF
and the rest is on the backend's side - this was proven by
virtual display implementation, so I see no problem here
for virtual camera.
>
>> So, I was more thinking that in case of V4L2 based frontend driver:
>> 1. Frontend serves REQBUFS ioctl and asks the backend with
>> XENCAMERA_OP_BUF_REQUEST
>> if it can handle that many buffers and gets number of buffers to be used
>> and buffer structure (number of planes, sizes, offsets etc.) as the reply
>> to that request
>> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
>> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
>> allocated
> Regards,
>
> Hans
Thank you,
Oleksandr
[1] https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/gpu/drm/xen
[2]
https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/xen/gntdev-dmabuf.c
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 9:52 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-10 11:09 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 11:09 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>> Hi Oleksandr,
>>>>
>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>> <snip>
>>
>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>> it pays to design this with that in mind.
>>>>> Again, I think that this is the backend to hide these
>>>>> use-cases from the frontend.
>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>> everything down and build it up again, or you can just tell frontends that
>>>> something changed and that they have to look at the new vcamera configuration.
>>>>
>>>> The latter seems to be more sensible to me. It is really not much that you
>>>> need to do: all you really need is an event signalling that something changed.
>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>> well, this complicates things a lot as I'll have to
>>> re-allocate buffers - right?
>> Right. Different resolutions means different sized buffers and usually lots of
>> changes throughout the whole video pipeline, which in this case can even
>> go into multiple VMs.
>>
>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>> has a flags field that tells userspace what changed. Right now that is just the
>> resolution, but in the future you can expect flags for cases where just the
>> colorspace information changes, but not the resolution.
>>
>> Which reminds me of two important missing pieces of information in your protocol:
>>
>> 1) You need to communicate the colorspace data:
>>
>> - colorspace
>> - xfer_func
>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>> think you can ignore hsv_enc)
>> - quantization
>>
>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>> and the links to the colorspace sections in the V4L2 spec for details).
>>
>> This information is part of the format, it is reported by the driver.
> I'll take a look and think what can be put and how into the protocol,
> do you think I'll have to implement all the above for
> this stage?
Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
They don't *have* to use this information, but it should be there. For cameras
this isn't all that important, for SDTV/HDTV sources this becomes more relevant
(esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
formats this is critical.
The vivid driver can actually reproduce all combinations, so that's a good driver
to test this with.
>
>>
>> 2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
>> each buffer contains a single field), then you need to be able to tell
>> userspace whether the dequeued buffer contains a top or bottom field.
> I think at the first stage we can assume that interlaced
> formats are not supported and add such support later if need be.
Frankly I consider that a smart move :-) Interlaced formats are awful...
You just have to keep this in mind if you ever have to add support for this.
>>
>> Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
>> timestamp that can help detecting that. You probably need something similar.
> Ok, this can be reported as part of XENCAMERA_EVT_FRAME_AVAIL event
>>
>>> But anyways, I can add
>>> #define XENCAMERA_EVT_CFG_CHANGE 0x01
>>> in the protocol, so we can address this use-case
>> <snip>
>>
>>>>> 1. set format command:
>>>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>>> * width - uint32_t, width in pixels.
>>>>> * height - uint32_t, height in pixels.
>>>>>
>>>>> 2. Set frame rate command:
>>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>>>
>>>>> 3. Set/request num bufs:
>>>>> * num_bufs - uint8_t, desired number of buffers to be used.
>>>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>>>> separate.
>>> ok, then 1+2 combined + 3 separate.
>>> Do you think we can still name 1+2 as "set_format" or "set_config"
>>> will fit better?
>> set_format is closer to S_FMT as used in V4L2, so I have a slight preference
>> for that, but it is really up to you.
> I'll probably stick to SET_CONFIG here
>>
>>>>>>> + *
>>>>>>> + * See response format for this request.
>>>>>>> + *
>>>>>>> + * Notes:
>>>>>>> + * - frontend must check the corresponding response in order to see
>>>>>>> + * if the values reported back by the backend do match the desired ones
>>>>>>> + * and can be accepted.
>>>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>>>> + * configuration.
>>>>>>> + */
>>>>>>> +struct xencamera_config {
>>>>>>> + uint32_t pixel_format;
>>>>>>> + uint32_t width;
>>>>>>> + uint32_t height;
>>>>>>> + uint32_t frame_rate_nom;
>>>>>>> + uint32_t frame_rate_denom;
>>>>>>> + uint8_t num_bufs;
>>>>>>> +};
>>>>>>> +
>>>>>>> +/*
>>>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>>>> + * detailed description:
>>>>>>> + * 0 1 2 3 octet
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 8
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 64
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + *
>>>>>>> + * See response format for this request.
>>>>>>> + *
>>>>>>> + *
>>>>>>> + * Request camera buffer creation:
>>>>>>> + * 0 1 2 3 octet
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 8
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | index | reserved | 12
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | gref_directory | 16
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 20
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 64
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + *
>>>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>>>> + *
>>>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>>>> + * then reference to the next shared page must be supplied (see
>>>>>>> + * gref_dir_next_page below).
>>>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>>>> does.
>>>>> Well, I still think it is better to have a per buffer interface
>>>>> in the protocol as it is done for other Xen virtual devices.
>>>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>>>> what it does internally in the frontend driver
>>>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>>>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>>>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>>>> sense. I'm not sure about the naming.
>>>>
>>>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>>>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>>>> case of success can be more or less than what was requested.
>>> The buffers can be allocated and shared by either backend or frontend: see
>>> "be-alloc" configuration option telling which domain (VM) shares
>>> the Xen grant references to the pages of the buffer: either frontend
>>> or backend.
>> If you want to do zero-copy video capture,
> this is the goal
>> then you need to know which
>> device in your video pipeline (which now covers both actual hardware and
>> multiple VMs) has the strictest memory layout requirements. Often the
>> video HW requires contiguous physical memory for the buffers, which means
>> you can't just give it a piece of non-contig memory allocated elsewhere.
> We have already implemented zero copying use-cases for
> virtual display, please see [1] and [2] which are dma-buf
> based which can cope with real HW restrictions you mention.
> And in that case we can implement zero-copying both ways,
> e.g. when the Xen grant references are shared by either
> backend or frontend. This is different from camera use-cases:
> a single buffer needs to be shared with multiple frontends,
> so zero-copying is only possible when backend allocates the references
> and shares those with frontends. The way when frontend allocates
> the buffers and still we can implement zero-copying is when
> there is a single frontend in the system, otherwise we
> need to copy the images from backend's buffers into frontend's
> ones.
OK. The important thing is that you thought about this :-)
>> In practice you have two possible memory models you can use with V4L2 drivers:
>> MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
>> as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
>> elsewhere and imported to V4L2, which may fail if it doesn't match the HW
>> requirements.
> For the frontend it is possible to work with both MMAP/DMABUF
> and the rest is on the backend's side - this was proven by
> virtual display implementation, so I see no problem here
> for virtual camera.
>>
>>> So, I was more thinking that in case of V4L2 based frontend driver:
>>> 1. Frontend serves REQBUFS ioctl and asks the backend with
>>> XENCAMERA_OP_BUF_REQUEST
>>> if it can handle that many buffers and gets number of buffers to be used
>>> and buffer structure (number of planes, sizes, offsets etc.) as the reply
>>> to that request
>>> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
>>> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
>>> allocated
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 11:09 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 11:09 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>> Hi Oleksandr,
>>>>
>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>> <snip>
>>
>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>> it pays to design this with that in mind.
>>>>> Again, I think that this is the backend to hide these
>>>>> use-cases from the frontend.
>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>> everything down and build it up again, or you can just tell frontends that
>>>> something changed and that they have to look at the new vcamera configuration.
>>>>
>>>> The latter seems to be more sensible to me. It is really not much that you
>>>> need to do: all you really need is an event signalling that something changed.
>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>> well, this complicates things a lot as I'll have to
>>> re-allocate buffers - right?
>> Right. Different resolutions means different sized buffers and usually lots of
>> changes throughout the whole video pipeline, which in this case can even
>> go into multiple VMs.
>>
>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>> has a flags field that tells userspace what changed. Right now that is just the
>> resolution, but in the future you can expect flags for cases where just the
>> colorspace information changes, but not the resolution.
>>
>> Which reminds me of two important missing pieces of information in your protocol:
>>
>> 1) You need to communicate the colorspace data:
>>
>> - colorspace
>> - xfer_func
>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>> think you can ignore hsv_enc)
>> - quantization
>>
>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>> and the links to the colorspace sections in the V4L2 spec for details).
>>
>> This information is part of the format, it is reported by the driver.
> I'll take a look and think what can be put and how into the protocol,
> do you think I'll have to implement all the above for
> this stage?
Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
They don't *have* to use this information, but it should be there. For cameras
this isn't all that important, for SDTV/HDTV sources this becomes more relevant
(esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
formats this is critical.
The vivid driver can actually reproduce all combinations, so that's a good driver
to test this with.
>
>>
>> 2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
>> each buffer contains a single field), then you need to be able to tell
>> userspace whether the dequeued buffer contains a top or bottom field.
> I think at the first stage we can assume that interlaced
> formats are not supported and add such support later if need be.
Frankly I consider that a smart move :-) Interlaced formats are awful...
You just have to keep this in mind if you ever have to add support for this.
>>
>> Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
>> timestamp that can help detecting that. You probably need something similar.
> Ok, this can be reported as part of XENCAMERA_EVT_FRAME_AVAIL event
>>
>>> But anyways, I can add
>>> #define XENCAMERA_EVT_CFG_CHANGE 0x01
>>> in the protocol, so we can address this use-case
>> <snip>
>>
>>>>> 1. set format command:
>>>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>>> * width - uint32_t, width in pixels.
>>>>> * height - uint32_t, height in pixels.
>>>>>
>>>>> 2. Set frame rate command:
>>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>>>
>>>>> 3. Set/request num bufs:
>>>>> * num_bufs - uint8_t, desired number of buffers to be used.
>>>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>>>> separate.
>>> ok, then 1+2 combined + 3 separate.
>>> Do you think we can still name 1+2 as "set_format" or "set_config"
>>> will fit better?
>> set_format is closer to S_FMT as used in V4L2, so I have a slight preference
>> for that, but it is really up to you.
> I'll probably stick to SET_CONFIG here
>>
>>>>>>> + *
>>>>>>> + * See response format for this request.
>>>>>>> + *
>>>>>>> + * Notes:
>>>>>>> + * - frontend must check the corresponding response in order to see
>>>>>>> + * if the values reported back by the backend do match the desired ones
>>>>>>> + * and can be accepted.
>>>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>>>> + * configuration.
>>>>>>> + */
>>>>>>> +struct xencamera_config {
>>>>>>> + uint32_t pixel_format;
>>>>>>> + uint32_t width;
>>>>>>> + uint32_t height;
>>>>>>> + uint32_t frame_rate_nom;
>>>>>>> + uint32_t frame_rate_denom;
>>>>>>> + uint8_t num_bufs;
>>>>>>> +};
>>>>>>> +
>>>>>>> +/*
>>>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>>>> + * detailed description:
>>>>>>> + * 0 1 2 3 octet
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 8
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 64
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + *
>>>>>>> + * See response format for this request.
>>>>>>> + *
>>>>>>> + *
>>>>>>> + * Request camera buffer creation:
>>>>>>> + * 0 1 2 3 octet
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 8
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | index | reserved | 12
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | gref_directory | 16
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 20
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + * | reserved | 64
>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>> + *
>>>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>>>> + *
>>>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>>>> + * then reference to the next shared page must be supplied (see
>>>>>>> + * gref_dir_next_page below).
>>>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>>>> does.
>>>>> Well, I still think it is better to have a per buffer interface
>>>>> in the protocol as it is done for other Xen virtual devices.
>>>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>>>> what it does internally in the frontend driver
>>>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>>>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>>>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>>>> sense. I'm not sure about the naming.
>>>>
>>>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>>>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>>>> case of success can be more or less than what was requested.
>>> The buffers can be allocated and shared by either backend or frontend: see
>>> "be-alloc" configuration option telling which domain (VM) shares
>>> the Xen grant references to the pages of the buffer: either frontend
>>> or backend.
>> If you want to do zero-copy video capture,
> this is the goal
>> then you need to know which
>> device in your video pipeline (which now covers both actual hardware and
>> multiple VMs) has the strictest memory layout requirements. Often the
>> video HW requires contiguous physical memory for the buffers, which means
>> you can't just give it a piece of non-contig memory allocated elsewhere.
> We have already implemented zero copying use-cases for
> virtual display, please see [1] and [2] which are dma-buf
> based which can cope with real HW restrictions you mention.
> And in that case we can implement zero-copying both ways,
> e.g. when the Xen grant references are shared by either
> backend or frontend. This is different from camera use-cases:
> a single buffer needs to be shared with multiple frontends,
> so zero-copying is only possible when backend allocates the references
> and shares those with frontends. The way when frontend allocates
> the buffers and still we can implement zero-copying is when
> there is a single frontend in the system, otherwise we
> need to copy the images from backend's buffers into frontend's
> ones.
OK. The important thing is that you thought about this :-)
>> In practice you have two possible memory models you can use with V4L2 drivers:
>> MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
>> as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
>> elsewhere and imported to V4L2, which may fail if it doesn't match the HW
>> requirements.
> For the frontend it is possible to work with both MMAP/DMABUF
> and the rest is on the backend's side - this was proven by
> virtual display implementation, so I see no problem here
> for virtual camera.
>>
>>> So, I was more thinking that in case of V4L2 based frontend driver:
>>> 1. Frontend serves REQBUFS ioctl and asks the backend with
>>> XENCAMERA_OP_BUF_REQUEST
>>> if it can handle that many buffers and gets number of buffers to be used
>>> and buffer structure (number of planes, sizes, offsets etc.) as the reply
>>> to that request
>>> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
>>> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
>>> allocated
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 11:09 ` [PATCH " Hans Verkuil
@ 2018-09-10 11:49 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 11:49 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/10/2018 02:09 PM, Hans Verkuil wrote:
> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>> Hi Oleksandr,
>>>>>
>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>> <snip>
>>>
>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>> it pays to design this with that in mind.
>>>>>> Again, I think that this is the backend to hide these
>>>>>> use-cases from the frontend.
>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>> everything down and build it up again, or you can just tell frontends that
>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>
>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>> need to do: all you really need is an event signalling that something changed.
>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>> well, this complicates things a lot as I'll have to
>>>> re-allocate buffers - right?
>>> Right. Different resolutions means different sized buffers and usually lots of
>>> changes throughout the whole video pipeline, which in this case can even
>>> go into multiple VMs.
>>>
>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>> has a flags field that tells userspace what changed. Right now that is just the
>>> resolution, but in the future you can expect flags for cases where just the
>>> colorspace information changes, but not the resolution.
>>>
>>> Which reminds me of two important missing pieces of information in your protocol:
>>>
>>> 1) You need to communicate the colorspace data:
>>>
>>> - colorspace
>>> - xfer_func
>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>> think you can ignore hsv_enc)
>>> - quantization
>>>
>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>
>>> This information is part of the format, it is reported by the driver.
>> I'll take a look and think what can be put and how into the protocol,
>> do you think I'll have to implement all the above for
>> this stage?
> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
> They don't *have* to use this information, but it should be there. For cameras
> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
> formats this is critical.
ok, then I'll add the following to the set_config request/response:
uint32_t colorspace;
uint32_t xfer_func;
uint32_t ycbcr_enc;
uint32_t quantization;
With this respect, I will need to put some OS agnostic constants
into the protocol, so if backend and frontend are not Linux/V4L2
based they can still talk to each other.
I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
Do you think I can define the same replacing V4L2_ prefix
with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
Do I need to define all those or there can be some subset of the
above for my simpler use-case?
> The vivid driver can actually reproduce all combinations, so that's a good driver
> to test this with.
You mean I can use it on backend side instead of real HW camera and
test all the configurations possible/those of interest?
>>> 2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
>>> each buffer contains a single field), then you need to be able to tell
>>> userspace whether the dequeued buffer contains a top or bottom field.
>> I think at the first stage we can assume that interlaced
>> formats are not supported and add such support later if need be.
> Frankly I consider that a smart move :-) Interlaced formats are awful...
>
> You just have to keep this in mind if you ever have to add support for this.
Agreed
>
>>> Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
>>> timestamp that can help detecting that. You probably need something similar.
>> Ok, this can be reported as part of XENCAMERA_EVT_FRAME_AVAIL event
>>>> But anyways, I can add
>>>> #define XENCAMERA_EVT_CFG_CHANGE 0x01
>>>> in the protocol, so we can address this use-case
>>> <snip>
>>>
>>>>>> 1. set format command:
>>>>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>>>> * width - uint32_t, width in pixels.
>>>>>> * height - uint32_t, height in pixels.
>>>>>>
>>>>>> 2. Set frame rate command:
>>>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>>>>
>>>>>> 3. Set/request num bufs:
>>>>>> * num_bufs - uint8_t, desired number of buffers to be used.
>>>>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>>>>> separate.
>>>> ok, then 1+2 combined + 3 separate.
>>>> Do you think we can still name 1+2 as "set_format" or "set_config"
>>>> will fit better?
>>> set_format is closer to S_FMT as used in V4L2, so I have a slight preference
>>> for that, but it is really up to you.
>> I'll probably stick to SET_CONFIG here
>>>>>>>> + *
>>>>>>>> + * See response format for this request.
>>>>>>>> + *
>>>>>>>> + * Notes:
>>>>>>>> + * - frontend must check the corresponding response in order to see
>>>>>>>> + * if the values reported back by the backend do match the desired ones
>>>>>>>> + * and can be accepted.
>>>>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>>>>> + * configuration.
>>>>>>>> + */
>>>>>>>> +struct xencamera_config {
>>>>>>>> + uint32_t pixel_format;
>>>>>>>> + uint32_t width;
>>>>>>>> + uint32_t height;
>>>>>>>> + uint32_t frame_rate_nom;
>>>>>>>> + uint32_t frame_rate_denom;
>>>>>>>> + uint8_t num_bufs;
>>>>>>>> +};
>>>>>>>> +
>>>>>>>> +/*
>>>>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>>>>> + * detailed description:
>>>>>>>> + * 0 1 2 3 octet
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 8
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 64
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + *
>>>>>>>> + * See response format for this request.
>>>>>>>> + *
>>>>>>>> + *
>>>>>>>> + * Request camera buffer creation:
>>>>>>>> + * 0 1 2 3 octet
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 8
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | index | reserved | 12
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | gref_directory | 16
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 20
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 64
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + *
>>>>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>>>>> + *
>>>>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>>>>> + * then reference to the next shared page must be supplied (see
>>>>>>>> + * gref_dir_next_page below).
>>>>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>>>>> does.
>>>>>> Well, I still think it is better to have a per buffer interface
>>>>>> in the protocol as it is done for other Xen virtual devices.
>>>>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>>>>> what it does internally in the frontend driver
>>>>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>>>>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>>>>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>>>>> sense. I'm not sure about the naming.
>>>>>
>>>>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>>>>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>>>>> case of success can be more or less than what was requested.
>>>> The buffers can be allocated and shared by either backend or frontend: see
>>>> "be-alloc" configuration option telling which domain (VM) shares
>>>> the Xen grant references to the pages of the buffer: either frontend
>>>> or backend.
>>> If you want to do zero-copy video capture,
>> this is the goal
>>> then you need to know which
>>> device in your video pipeline (which now covers both actual hardware and
>>> multiple VMs) has the strictest memory layout requirements. Often the
>>> video HW requires contiguous physical memory for the buffers, which means
>>> you can't just give it a piece of non-contig memory allocated elsewhere.
>> We have already implemented zero copying use-cases for
>> virtual display, please see [1] and [2] which are dma-buf
>> based which can cope with real HW restrictions you mention.
>> And in that case we can implement zero-copying both ways,
>> e.g. when the Xen grant references are shared by either
>> backend or frontend. This is different from camera use-cases:
>> a single buffer needs to be shared with multiple frontends,
>> so zero-copying is only possible when backend allocates the references
>> and shares those with frontends. The way when frontend allocates
>> the buffers and still we can implement zero-copying is when
>> there is a single frontend in the system, otherwise we
>> need to copy the images from backend's buffers into frontend's
>> ones.
> OK. The important thing is that you thought about this :-)
Sure, copying kills performance...
>
>>> In practice you have two possible memory models you can use with V4L2 drivers:
>>> MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
>>> as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
>>> elsewhere and imported to V4L2, which may fail if it doesn't match the HW
>>> requirements.
>> For the frontend it is possible to work with both MMAP/DMABUF
>> and the rest is on the backend's side - this was proven by
>> virtual display implementation, so I see no problem here
>> for virtual camera.
>>>> So, I was more thinking that in case of V4L2 based frontend driver:
>>>> 1. Frontend serves REQBUFS ioctl and asks the backend with
>>>> XENCAMERA_OP_BUF_REQUEST
>>>> if it can handle that many buffers and gets number of buffers to be used
>>>> and buffer structure (number of planes, sizes, offsets etc.) as the reply
>>>> to that request
>>>> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
>>>> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
>>>> allocated
> Regards,
>
> Hans
Thank you,
Oleksandr
[1]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_colorspace
[2]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_ycbcr_encoding
[3]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_quantization
[4]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_xfer_func
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 11:49 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 11:49 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/10/2018 02:09 PM, Hans Verkuil wrote:
> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>> Hi Oleksandr,
>>>>>
>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>> <snip>
>>>
>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>> it pays to design this with that in mind.
>>>>>> Again, I think that this is the backend to hide these
>>>>>> use-cases from the frontend.
>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>> everything down and build it up again, or you can just tell frontends that
>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>
>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>> need to do: all you really need is an event signalling that something changed.
>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>> well, this complicates things a lot as I'll have to
>>>> re-allocate buffers - right?
>>> Right. Different resolutions means different sized buffers and usually lots of
>>> changes throughout the whole video pipeline, which in this case can even
>>> go into multiple VMs.
>>>
>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>> has a flags field that tells userspace what changed. Right now that is just the
>>> resolution, but in the future you can expect flags for cases where just the
>>> colorspace information changes, but not the resolution.
>>>
>>> Which reminds me of two important missing pieces of information in your protocol:
>>>
>>> 1) You need to communicate the colorspace data:
>>>
>>> - colorspace
>>> - xfer_func
>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>> think you can ignore hsv_enc)
>>> - quantization
>>>
>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>
>>> This information is part of the format, it is reported by the driver.
>> I'll take a look and think what can be put and how into the protocol,
>> do you think I'll have to implement all the above for
>> this stage?
> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
> They don't *have* to use this information, but it should be there. For cameras
> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
> formats this is critical.
ok, then I'll add the following to the set_config request/response:
uint32_t colorspace;
uint32_t xfer_func;
uint32_t ycbcr_enc;
uint32_t quantization;
With this respect, I will need to put some OS agnostic constants
into the protocol, so if backend and frontend are not Linux/V4L2
based they can still talk to each other.
I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
Do you think I can define the same replacing V4L2_ prefix
with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
Do I need to define all those or there can be some subset of the
above for my simpler use-case?
> The vivid driver can actually reproduce all combinations, so that's a good driver
> to test this with.
You mean I can use it on backend side instead of real HW camera and
test all the configurations possible/those of interest?
>>> 2) If you support interlaced formats and V4L2_FIELD_ALTERNATE (i.e.
>>> each buffer contains a single field), then you need to be able to tell
>>> userspace whether the dequeued buffer contains a top or bottom field.
>> I think at the first stage we can assume that interlaced
>> formats are not supported and add such support later if need be.
> Frankly I consider that a smart move :-) Interlaced formats are awful...
>
> You just have to keep this in mind if you ever have to add support for this.
Agreed
>
>>> Also, what to do with dropped frames/fields: V4L2 has a sequence counter and
>>> timestamp that can help detecting that. You probably need something similar.
>> Ok, this can be reported as part of XENCAMERA_EVT_FRAME_AVAIL event
>>>> But anyways, I can add
>>>> #define XENCAMERA_EVT_CFG_CHANGE 0x01
>>>> in the protocol, so we can address this use-case
>>> <snip>
>>>
>>>>>> 1. set format command:
>>>>>> * pixel_format - uint32_t, pixel format to be used, FOURCC code.
>>>>>> * width - uint32_t, width in pixels.
>>>>>> * height - uint32_t, height in pixels.
>>>>>>
>>>>>> 2. Set frame rate command:
>>>>>> + * frame_rate_numer - uint32_t, numerator of the frame rate.
>>>>>> + * frame_rate_denom - uint32_t, denominator of the frame rate.
>>>>>>
>>>>>> 3. Set/request num bufs:
>>>>>> * num_bufs - uint8_t, desired number of buffers to be used.
>>>>> I like this much better. 1+2 could be combined, but 3 should definitely remain
>>>>> separate.
>>>> ok, then 1+2 combined + 3 separate.
>>>> Do you think we can still name 1+2 as "set_format" or "set_config"
>>>> will fit better?
>>> set_format is closer to S_FMT as used in V4L2, so I have a slight preference
>>> for that, but it is really up to you.
>> I'll probably stick to SET_CONFIG here
>>>>>>>> + *
>>>>>>>> + * See response format for this request.
>>>>>>>> + *
>>>>>>>> + * Notes:
>>>>>>>> + * - frontend must check the corresponding response in order to see
>>>>>>>> + * if the values reported back by the backend do match the desired ones
>>>>>>>> + * and can be accepted.
>>>>>>>> + * - frontend may send multiple XENCAMERA_OP_SET_CONFIG requests before
>>>>>>>> + * sending XENCAMERA_OP_STREAM_START request to update or tune the
>>>>>>>> + * configuration.
>>>>>>>> + */
>>>>>>>> +struct xencamera_config {
>>>>>>>> + uint32_t pixel_format;
>>>>>>>> + uint32_t width;
>>>>>>>> + uint32_t height;
>>>>>>>> + uint32_t frame_rate_nom;
>>>>>>>> + uint32_t frame_rate_denom;
>>>>>>>> + uint8_t num_bufs;
>>>>>>>> +};
>>>>>>>> +
>>>>>>>> +/*
>>>>>>>> + * Request buffer details - request camera buffer's memory layout.
>>>>>>>> + * detailed description:
>>>>>>>> + * 0 1 2 3 octet
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | id |_GET_BUF_DETAILS| reserved | 4
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 8
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 64
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + *
>>>>>>>> + * See response format for this request.
>>>>>>>> + *
>>>>>>>> + *
>>>>>>>> + * Request camera buffer creation:
>>>>>>>> + * 0 1 2 3 octet
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | id | _OP_BUF_CREATE | reserved | 4
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 8
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | index | reserved | 12
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | gref_directory | 16
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 20
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * |/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/|
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + * | reserved | 64
>>>>>>>> + * +----------------+----------------+----------------+----------------+
>>>>>>>> + *
>>>>>>>> + * An attempt to create multiple buffers with the same index is an error.
>>>>>>>> + * index can be re-used after destroying the corresponding camera buffer.
>>>>>>>> + *
>>>>>>>> + * index - uint8_t, index of the buffer to be created.
>>>>>>>> + * gref_directory - grant_ref_t, a reference to the first shared page
>>>>>>>> + * describing shared buffer references. The size of the buffer is equal to
>>>>>>>> + * XENCAMERA_OP_GET_BUF_DETAILS.size response. At least one page exists. If
>>>>>>>> + * shared buffer size exceeds what can be addressed by this single page,
>>>>>>>> + * then reference to the next shared page must be supplied (see
>>>>>>>> + * gref_dir_next_page below).
>>>>>>> It might be better to allocate all buffers in one go, i.e. what VIDIOC_REQBUFS
>>>>>>> does.
>>>>>> Well, I still think it is better to have a per buffer interface
>>>>>> in the protocol as it is done for other Xen virtual devices.
>>>>>> So, I'll keep this as is for now: VIDIOC_REQBUFS can still do
>>>>>> what it does internally in the frontend driver
>>>>> I may have misunderstood the original API. The newly proposed XENCAMERA_OP_BUF_REQUEST
>>>>> maps to REQBUFS, right? And then BUF_CREATE/DESTROY just set up the shared buffer
>>>>> mappings for the buffers created by REQBUFS. If that's the sequence, then it makes
>>>>> sense. I'm not sure about the naming.
>>>>>
>>>>> You might want to make it clear that XENCAMERA_OP_BUF_REQUEST allocates the buffers
>>>>> on the backend, and so can fail. Also, the actual number of allocated buffers in
>>>>> case of success can be more or less than what was requested.
>>>> The buffers can be allocated and shared by either backend or frontend: see
>>>> "be-alloc" configuration option telling which domain (VM) shares
>>>> the Xen grant references to the pages of the buffer: either frontend
>>>> or backend.
>>> If you want to do zero-copy video capture,
>> this is the goal
>>> then you need to know which
>>> device in your video pipeline (which now covers both actual hardware and
>>> multiple VMs) has the strictest memory layout requirements. Often the
>>> video HW requires contiguous physical memory for the buffers, which means
>>> you can't just give it a piece of non-contig memory allocated elsewhere.
>> We have already implemented zero copying use-cases for
>> virtual display, please see [1] and [2] which are dma-buf
>> based which can cope with real HW restrictions you mention.
>> And in that case we can implement zero-copying both ways,
>> e.g. when the Xen grant references are shared by either
>> backend or frontend. This is different from camera use-cases:
>> a single buffer needs to be shared with multiple frontends,
>> so zero-copying is only possible when backend allocates the references
>> and shares those with frontends. The way when frontend allocates
>> the buffers and still we can implement zero-copying is when
>> there is a single frontend in the system, otherwise we
>> need to copy the images from backend's buffers into frontend's
>> ones.
> OK. The important thing is that you thought about this :-)
Sure, copying kills performance...
>
>>> In practice you have two possible memory models you can use with V4L2 drivers:
>>> MMAP (i.e. allocated by the driver and the buffers can, if needed, be exported
>>> as dmabuf handles with VIDIOC_EXPBUF), or DMABUF where buffers are allocated
>>> elsewhere and imported to V4L2, which may fail if it doesn't match the HW
>>> requirements.
>> For the frontend it is possible to work with both MMAP/DMABUF
>> and the rest is on the backend's side - this was proven by
>> virtual display implementation, so I see no problem here
>> for virtual camera.
>>>> So, I was more thinking that in case of V4L2 based frontend driver:
>>>> 1. Frontend serves REQBUFS ioctl and asks the backend with
>>>> XENCAMERA_OP_BUF_REQUEST
>>>> if it can handle that many buffers and gets number of buffers to be used
>>>> and buffer structure (number of planes, sizes, offsets etc.) as the reply
>>>> to that request
>>>> 2. Frontend creates n buffers with XENCAMERA_OP_BUF_CREATE
>>>> 3. Frontend returns from REQBUFS ioctl with actual number of buffers
>>>> allocated
> Regards,
>
> Hans
Thank you,
Oleksandr
[1]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_colorspace
[2]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_ycbcr_encoding
[3]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_quantization
[4]
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-defs.html#c.v4l2_xfer_func
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 11:49 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-10 12:26 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 12:26 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr_Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>> Hi Oleksandr,
>>>>>>
>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>> <snip>
>>>>
>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>> it pays to design this with that in mind.
>>>>>>> Again, I think that this is the backend to hide these
>>>>>>> use-cases from the frontend.
>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>
>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>> well, this complicates things a lot as I'll have to
>>>>> re-allocate buffers - right?
>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>> changes throughout the whole video pipeline, which in this case can even
>>>> go into multiple VMs.
>>>>
>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>> resolution, but in the future you can expect flags for cases where just the
>>>> colorspace information changes, but not the resolution.
>>>>
>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>
>>>> 1) You need to communicate the colorspace data:
>>>>
>>>> - colorspace
>>>> - xfer_func
>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>> think you can ignore hsv_enc)
>>>> - quantization
>>>>
>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>
>>>> This information is part of the format, it is reported by the driver.
>>> I'll take a look and think what can be put and how into the protocol,
>>> do you think I'll have to implement all the above for
>>> this stage?
>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>> They don't *have* to use this information, but it should be there. For cameras
>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>> formats this is critical.
> ok, then I'll add the following to the set_config request/response:
>
> uint32_t colorspace;
> uint32_t xfer_func;
> uint32_t ycbcr_enc;
> uint32_t quantization;
>
> With this respect, I will need to put some OS agnostic constants
> into the protocol, so if backend and frontend are not Linux/V4L2
> based they can still talk to each other.
> I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
>
> Do you think I can define the same replacing V4L2_ prefix
> with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
Yes.
>
> Do I need to define all those or there can be some subset of the
> above for my simpler use-case?
Most of these defines directly map to standards. I would skip the following
defines:
V4L2_COLORSPACE_DEFAULT (not applicable)
V4L2_COLORSPACE_470_SYSTEM_* (rarely used, if received by the HW the Xen backend
should map this to V4L2_COLORSPACE_SMPTE170M)
V4L2_COLORSPACE_JPEG (historical V4L2 artifact, see here how to map:
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-details.html#col-jpeg)
V4L2_COLORSPACE_SMPTE240M (rarely used, map to V4L2_COLORSPACE_SMPTE170M if seen in backend)
V4L2_XFER_FUNC_SMPTE240M (rarely used, map to V4L2_XFER_FUNC_709)
V4L2_YCBCR_ENC_SMPTE240M (rarely used, map to V4L2_YCBCR_ENC_709)
While V4L2 allows 0 (DEFAULT) values for xfer_func, ycbcr_enc and quantization, and
provides macros to map default values to the actual values (for legacy reasons),
the Xen backend should always fill this in explicitly, using those same mapping
macros (see e.g. V4L2_MAP_XFER_FUNC_DEFAULT).
The V4L2 spec has extensive information on colorspaces (sections 2.14-2.17).
>
>> The vivid driver can actually reproduce all combinations, so that's a good driver
>> to test this with.
> You mean I can use it on backend side instead of real HW camera and
> test all the configurations possible/those of interest?
Right.
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 12:26 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-10 12:26 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr_Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>> Hi Oleksandr,
>>>>>>
>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>> <snip>
>>>>
>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>> it pays to design this with that in mind.
>>>>>>> Again, I think that this is the backend to hide these
>>>>>>> use-cases from the frontend.
>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>
>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>> well, this complicates things a lot as I'll have to
>>>>> re-allocate buffers - right?
>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>> changes throughout the whole video pipeline, which in this case can even
>>>> go into multiple VMs.
>>>>
>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>> resolution, but in the future you can expect flags for cases where just the
>>>> colorspace information changes, but not the resolution.
>>>>
>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>
>>>> 1) You need to communicate the colorspace data:
>>>>
>>>> - colorspace
>>>> - xfer_func
>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>> think you can ignore hsv_enc)
>>>> - quantization
>>>>
>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>
>>>> This information is part of the format, it is reported by the driver.
>>> I'll take a look and think what can be put and how into the protocol,
>>> do you think I'll have to implement all the above for
>>> this stage?
>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>> They don't *have* to use this information, but it should be there. For cameras
>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>> formats this is critical.
> ok, then I'll add the following to the set_config request/response:
>
> uint32_t colorspace;
> uint32_t xfer_func;
> uint32_t ycbcr_enc;
> uint32_t quantization;
>
> With this respect, I will need to put some OS agnostic constants
> into the protocol, so if backend and frontend are not Linux/V4L2
> based they can still talk to each other.
> I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
>
> Do you think I can define the same replacing V4L2_ prefix
> with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
Yes.
>
> Do I need to define all those or there can be some subset of the
> above for my simpler use-case?
Most of these defines directly map to standards. I would skip the following
defines:
V4L2_COLORSPACE_DEFAULT (not applicable)
V4L2_COLORSPACE_470_SYSTEM_* (rarely used, if received by the HW the Xen backend
should map this to V4L2_COLORSPACE_SMPTE170M)
V4L2_COLORSPACE_JPEG (historical V4L2 artifact, see here how to map:
https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-details.html#col-jpeg)
V4L2_COLORSPACE_SMPTE240M (rarely used, map to V4L2_COLORSPACE_SMPTE170M if seen in backend)
V4L2_XFER_FUNC_SMPTE240M (rarely used, map to V4L2_XFER_FUNC_709)
V4L2_YCBCR_ENC_SMPTE240M (rarely used, map to V4L2_YCBCR_ENC_709)
While V4L2 allows 0 (DEFAULT) values for xfer_func, ycbcr_enc and quantization, and
provides macros to map default values to the actual values (for legacy reasons),
the Xen backend should always fill this in explicitly, using those same mapping
macros (see e.g. V4L2_MAP_XFER_FUNC_DEFAULT).
The V4L2 spec has extensive information on colorspaces (sections 2.14-2.17).
>
>> The vivid driver can actually reproduce all combinations, so that's a good driver
>> to test this with.
> You mean I can use it on backend side instead of real HW camera and
> test all the configurations possible/those of interest?
Right.
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 0/1] cameraif: Add ABI for para-virtualized
2018-07-31 9:31 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-10 12:48 ` Laurent Pinchart
-1 siblings, 0 replies; 56+ messages in thread
From: Laurent Pinchart @ 2018-09-10 12:48 UTC (permalink / raw)
To: Oleksandr Andrushchenko
Cc: xen-devel, konrad.wilk, jgross, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm,
Oleksandr Andrushchenko
Hi Oleksandr,
Thank you for the patch.
On Tuesday, 31 July 2018 12:31:41 EEST Oleksandr Andrushchenko wrote:
> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>
> Hello!
>
> At the moment Xen [1] already supports some virtual multimedia
> features [2] such as virtual display, sound. It supports keyboards,
> pointers and multi-touch devices all allowing Xen to be used in
> automotive appliances, In-Vehicle Infotainment (IVI) systems
> and many more.
>
> This work adds a new Xen para-virtualized protocol for a virtual
> camera device which extends multimedia capabilities of Xen even
> farther: video conferencing, IVI, high definition maps etc.
>
> The initial goal is to support most needed functionality with the
> final idea to make it possible to extend the protocol if need be:
>
> 1. Provide means for base virtual device configuration:
> - pixel formats
> - resolutions
> - frame rates
> 2. Support basic camera controls:
> - contrast
> - brightness
> - hue
> - saturation
> 3. Support streaming control
> 4. Support zero-copying use-cases
>
> I hope that Xen and V4L and other communities could give their
> valuable feedback on this work, so I can update the protocol
> to better fit any additional requirements I might have missed.
I'll start with a question : what are the expected use cases ? The ones listed
above sound like they would better be solved by passing the corresponding
device(s) to the guest.
> [1] https://www.xenproject.org/
> [2] https://xenbits.xen.org/gitweb/?p=xen.git;a=tree;f=xen/include/public/io
>
> Oleksandr Andrushchenko (1):
> cameraif: add ABI for para-virtual camera
>
> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
> 1 file changed, 981 insertions(+)
> create mode 100644 xen/include/public/io/cameraif.h
--
Regards,
Laurent Pinchart
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 0/1] cameraif: Add ABI for para-virtualized
@ 2018-09-10 12:48 ` Laurent Pinchart
0 siblings, 0 replies; 56+ messages in thread
From: Laurent Pinchart @ 2018-09-10 12:48 UTC (permalink / raw)
To: Oleksandr Andrushchenko
Cc: jgross, konrad.wilk, Oleksandr Andrushchenko, koji.matsuoka.xm,
sakari.ailus, xen-devel, boris.ostrovsky, mchehab, linux-media
Hi Oleksandr,
Thank you for the patch.
On Tuesday, 31 July 2018 12:31:41 EEST Oleksandr Andrushchenko wrote:
> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>
> Hello!
>
> At the moment Xen [1] already supports some virtual multimedia
> features [2] such as virtual display, sound. It supports keyboards,
> pointers and multi-touch devices all allowing Xen to be used in
> automotive appliances, In-Vehicle Infotainment (IVI) systems
> and many more.
>
> This work adds a new Xen para-virtualized protocol for a virtual
> camera device which extends multimedia capabilities of Xen even
> farther: video conferencing, IVI, high definition maps etc.
>
> The initial goal is to support most needed functionality with the
> final idea to make it possible to extend the protocol if need be:
>
> 1. Provide means for base virtual device configuration:
> - pixel formats
> - resolutions
> - frame rates
> 2. Support basic camera controls:
> - contrast
> - brightness
> - hue
> - saturation
> 3. Support streaming control
> 4. Support zero-copying use-cases
>
> I hope that Xen and V4L and other communities could give their
> valuable feedback on this work, so I can update the protocol
> to better fit any additional requirements I might have missed.
I'll start with a question : what are the expected use cases ? The ones listed
above sound like they would better be solved by passing the corresponding
device(s) to the guest.
> [1] https://www.xenproject.org/
> [2] https://xenbits.xen.org/gitweb/?p=xen.git;a=tree;f=xen/include/public/io
>
> Oleksandr Andrushchenko (1):
> cameraif: add ABI for para-virtual camera
>
> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
> 1 file changed, 981 insertions(+)
> create mode 100644 xen/include/public/io/cameraif.h
--
Regards,
Laurent Pinchart
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 0/1] cameraif: Add ABI for para-virtualized
2018-09-10 12:48 ` [PATCH " Laurent Pinchart
@ 2018-09-10 13:02 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 13:02 UTC (permalink / raw)
To: Laurent Pinchart, Oleksandr_Andrushchenko
Cc: xen-devel, konrad.wilk, jgross, boris.ostrovsky, mchehab,
linux-media, sakari.ailus, koji.matsuoka.xm
Hi, Laurent!
On 09/10/2018 03:48 PM, Laurent Pinchart wrote:
> Hi Oleksandr,
>
> Thank you for the patch.
>
> On Tuesday, 31 July 2018 12:31:41 EEST Oleksandr Andrushchenko wrote:
>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>
>> Hello!
>>
>> At the moment Xen [1] already supports some virtual multimedia
>> features [2] such as virtual display, sound. It supports keyboards,
>> pointers and multi-touch devices all allowing Xen to be used in
>> automotive appliances, In-Vehicle Infotainment (IVI) systems
>> and many more.
>>
>> This work adds a new Xen para-virtualized protocol for a virtual
>> camera device which extends multimedia capabilities of Xen even
>> farther: video conferencing, IVI, high definition maps etc.
>>
>> The initial goal is to support most needed functionality with the
>> final idea to make it possible to extend the protocol if need be:
>>
>> 1. Provide means for base virtual device configuration:
>> - pixel formats
>> - resolutions
>> - frame rates
>> 2. Support basic camera controls:
>> - contrast
>> - brightness
>> - hue
>> - saturation
>> 3. Support streaming control
>> 4. Support zero-copying use-cases
>>
>> I hope that Xen and V4L and other communities could give their
>> valuable feedback on this work, so I can update the protocol
>> to better fit any additional requirements I might have missed.
> I'll start with a question : what are the expected use cases ?
The very basic use-case is to share a capture stream produced
by a single HW camera to multiple VMs for different
purposes: In-Vehicle Infotainment, high definition maps etc.
all running in different (dedicated) VMs at the same time
> The ones listed
> above sound like they would better be solved by passing the corresponding
> device(s) to the guest.
With the above use-case I cannot tell how passing the
corresponding *single* device can serve *multiple* VMs.
Could you please elaborate more on the solution you see?
>
>> [1] https://www.xenproject.org/
>> [2] https://xenbits.xen.org/gitweb/?p=xen.git;a=tree;f=xen/include/public/io
>>
>> Oleksandr Andrushchenko (1):
>> cameraif: add ABI for para-virtual camera
>>
>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>> 1 file changed, 981 insertions(+)
>> create mode 100644 xen/include/public/io/cameraif.h
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 0/1] cameraif: Add ABI for para-virtualized
@ 2018-09-10 13:02 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 13:02 UTC (permalink / raw)
To: Laurent Pinchart, Oleksandr_Andrushchenko
Cc: jgross, konrad.wilk, koji.matsuoka.xm, sakari.ailus, xen-devel,
boris.ostrovsky, mchehab, linux-media
Hi, Laurent!
On 09/10/2018 03:48 PM, Laurent Pinchart wrote:
> Hi Oleksandr,
>
> Thank you for the patch.
>
> On Tuesday, 31 July 2018 12:31:41 EEST Oleksandr Andrushchenko wrote:
>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>
>>
>> Hello!
>>
>> At the moment Xen [1] already supports some virtual multimedia
>> features [2] such as virtual display, sound. It supports keyboards,
>> pointers and multi-touch devices all allowing Xen to be used in
>> automotive appliances, In-Vehicle Infotainment (IVI) systems
>> and many more.
>>
>> This work adds a new Xen para-virtualized protocol for a virtual
>> camera device which extends multimedia capabilities of Xen even
>> farther: video conferencing, IVI, high definition maps etc.
>>
>> The initial goal is to support most needed functionality with the
>> final idea to make it possible to extend the protocol if need be:
>>
>> 1. Provide means for base virtual device configuration:
>> - pixel formats
>> - resolutions
>> - frame rates
>> 2. Support basic camera controls:
>> - contrast
>> - brightness
>> - hue
>> - saturation
>> 3. Support streaming control
>> 4. Support zero-copying use-cases
>>
>> I hope that Xen and V4L and other communities could give their
>> valuable feedback on this work, so I can update the protocol
>> to better fit any additional requirements I might have missed.
> I'll start with a question : what are the expected use cases ?
The very basic use-case is to share a capture stream produced
by a single HW camera to multiple VMs for different
purposes: In-Vehicle Infotainment, high definition maps etc.
all running in different (dedicated) VMs at the same time
> The ones listed
> above sound like they would better be solved by passing the corresponding
> device(s) to the guest.
With the above use-case I cannot tell how passing the
corresponding *single* device can serve *multiple* VMs.
Could you please elaborate more on the solution you see?
>
>> [1] https://www.xenproject.org/
>> [2] https://xenbits.xen.org/gitweb/?p=xen.git;a=tree;f=xen/include/public/io
>>
>> Oleksandr Andrushchenko (1):
>> cameraif: add ABI for para-virtual camera
>>
>> xen/include/public/io/cameraif.h | 981 +++++++++++++++++++++++++++++++
>> 1 file changed, 981 insertions(+)
>> create mode 100644 xen/include/public/io/cameraif.h
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 12:26 ` [PATCH " Hans Verkuil
@ 2018-09-10 13:16 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 13:16 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/10/2018 03:26 PM, Hans Verkuil wrote:
> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>> Hi Oleksandr,
>>>>>>>
>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>> <snip>
>>>>>
>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>> it pays to design this with that in mind.
>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>> use-cases from the frontend.
>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>
>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>> well, this complicates things a lot as I'll have to
>>>>>> re-allocate buffers - right?
>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>> go into multiple VMs.
>>>>>
>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>> colorspace information changes, but not the resolution.
>>>>>
>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>
>>>>> 1) You need to communicate the colorspace data:
>>>>>
>>>>> - colorspace
>>>>> - xfer_func
>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>> think you can ignore hsv_enc)
>>>>> - quantization
>>>>>
>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>
>>>>> This information is part of the format, it is reported by the driver.
>>>> I'll take a look and think what can be put and how into the protocol,
>>>> do you think I'll have to implement all the above for
>>>> this stage?
>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>> They don't *have* to use this information, but it should be there. For cameras
>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>> formats this is critical.
>> ok, then I'll add the following to the set_config request/response:
>>
>> uint32_t colorspace;
>> uint32_t xfer_func;
>> uint32_t ycbcr_enc;
>> uint32_t quantization;
>>
>> With this respect, I will need to put some OS agnostic constants
>> into the protocol, so if backend and frontend are not Linux/V4L2
>> based they can still talk to each other.
>> I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
>>
>> Do you think I can define the same replacing V4L2_ prefix
>> with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
> Yes.
>
>> Do I need to define all those or there can be some subset of the
>> above for my simpler use-case?
> Most of these defines directly map to standards. I would skip the following
> defines:
>
> V4L2_COLORSPACE_DEFAULT (not applicable)
> V4L2_COLORSPACE_470_SYSTEM_* (rarely used, if received by the HW the Xen backend
> should map this to V4L2_COLORSPACE_SMPTE170M)
> V4L2_COLORSPACE_JPEG (historical V4L2 artifact, see here how to map:
> https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-details.html#col-jpeg)
>
> V4L2_COLORSPACE_SMPTE240M (rarely used, map to V4L2_COLORSPACE_SMPTE170M if seen in backend)
>
> V4L2_XFER_FUNC_SMPTE240M (rarely used, map to V4L2_XFER_FUNC_709)
>
> V4L2_YCBCR_ENC_SMPTE240M (rarely used, map to V4L2_YCBCR_ENC_709)
>
> While V4L2 allows 0 (DEFAULT) values for xfer_func, ycbcr_enc and quantization, and
> provides macros to map default values to the actual values (for legacy reasons),
> the Xen backend should always fill this in explicitly, using those same mapping
> macros (see e.g. V4L2_MAP_XFER_FUNC_DEFAULT).
>
> The V4L2 spec has extensive information on colorspaces (sections 2.14-2.17).
>
Thank you for such a detailed explanation!
I'll define the constants as agreed above.
>>> The vivid driver can actually reproduce all combinations, so that's a good driver
>>> to test this with.
>> You mean I can use it on backend side instead of real HW camera and
>> test all the configurations possible/those of interest?
> Right.
>
> Regards,
>
> Hans
It seems that the number of changes discussed are begging
for the v2 of the protocol to be published ;)
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-10 13:16 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-10 13:16 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/10/2018 03:26 PM, Hans Verkuil wrote:
> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>> Hi Oleksandr,
>>>>>>>
>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>> <snip>
>>>>>
>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>> it pays to design this with that in mind.
>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>> use-cases from the frontend.
>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>
>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>> well, this complicates things a lot as I'll have to
>>>>>> re-allocate buffers - right?
>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>> go into multiple VMs.
>>>>>
>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>> colorspace information changes, but not the resolution.
>>>>>
>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>
>>>>> 1) You need to communicate the colorspace data:
>>>>>
>>>>> - colorspace
>>>>> - xfer_func
>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>> think you can ignore hsv_enc)
>>>>> - quantization
>>>>>
>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>
>>>>> This information is part of the format, it is reported by the driver.
>>>> I'll take a look and think what can be put and how into the protocol,
>>>> do you think I'll have to implement all the above for
>>>> this stage?
>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>> They don't *have* to use this information, but it should be there. For cameras
>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>> formats this is critical.
>> ok, then I'll add the following to the set_config request/response:
>>
>> uint32_t colorspace;
>> uint32_t xfer_func;
>> uint32_t ycbcr_enc;
>> uint32_t quantization;
>>
>> With this respect, I will need to put some OS agnostic constants
>> into the protocol, so if backend and frontend are not Linux/V4L2
>> based they can still talk to each other.
>> I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
>>
>> Do you think I can define the same replacing V4L2_ prefix
>> with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
> Yes.
>
>> Do I need to define all those or there can be some subset of the
>> above for my simpler use-case?
> Most of these defines directly map to standards. I would skip the following
> defines:
>
> V4L2_COLORSPACE_DEFAULT (not applicable)
> V4L2_COLORSPACE_470_SYSTEM_* (rarely used, if received by the HW the Xen backend
> should map this to V4L2_COLORSPACE_SMPTE170M)
> V4L2_COLORSPACE_JPEG (historical V4L2 artifact, see here how to map:
> https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-details.html#col-jpeg)
>
> V4L2_COLORSPACE_SMPTE240M (rarely used, map to V4L2_COLORSPACE_SMPTE170M if seen in backend)
>
> V4L2_XFER_FUNC_SMPTE240M (rarely used, map to V4L2_XFER_FUNC_709)
>
> V4L2_YCBCR_ENC_SMPTE240M (rarely used, map to V4L2_YCBCR_ENC_709)
>
> While V4L2 allows 0 (DEFAULT) values for xfer_func, ycbcr_enc and quantization, and
> provides macros to map default values to the actual values (for legacy reasons),
> the Xen backend should always fill this in explicitly, using those same mapping
> macros (see e.g. V4L2_MAP_XFER_FUNC_DEFAULT).
>
> The V4L2 spec has extensive information on colorspaces (sections 2.14-2.17).
>
Thank you for such a detailed explanation!
I'll define the constants as agreed above.
>>> The vivid driver can actually reproduce all combinations, so that's a good driver
>>> to test this with.
>> You mean I can use it on backend side instead of real HW camera and
>> test all the configurations possible/those of interest?
> Right.
>
> Regards,
>
> Hans
It seems that the number of changes discussed are begging
for the v2 of the protocol to be published ;)
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-10 12:26 ` [PATCH " Hans Verkuil
@ 2018-09-11 6:52 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-11 6:52 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Hi, Hans!
On 09/10/2018 03:26 PM, Hans Verkuil wrote:
> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>> Hi Oleksandr,
>>>>>>>
>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>> <snip>
>>>>>
>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>> it pays to design this with that in mind.
>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>> use-cases from the frontend.
>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>
>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>> well, this complicates things a lot as I'll have to
>>>>>> re-allocate buffers - right?
>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>> go into multiple VMs.
>>>>>
>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>> colorspace information changes, but not the resolution.
>>>>>
>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>
>>>>> 1) You need to communicate the colorspace data:
>>>>>
>>>>> - colorspace
>>>>> - xfer_func
>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>> think you can ignore hsv_enc)
>>>>> - quantization
>>>>>
>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>
>>>>> This information is part of the format, it is reported by the driver.
>>>> I'll take a look and think what can be put and how into the protocol,
>>>> do you think I'll have to implement all the above for
>>>> this stage?
>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>> They don't *have* to use this information, but it should be there. For cameras
>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>> formats this is critical.
>> ok, then I'll add the following to the set_config request/response:
>>
>> uint32_t colorspace;
>> uint32_t xfer_func;
>> uint32_t ycbcr_enc;
>> uint32_t quantization;
Yet another question here: are the above (color space, xfer etc.) and
display aspect ratio defined per pixel_format or per pixel_format +
resolution?
If per pixel_format then
.../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
or if per resolution
.../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
>>
>> With this respect, I will need to put some OS agnostic constants
>> into the protocol, so if backend and frontend are not Linux/V4L2
>> based they can still talk to each other.
>> I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
>>
>> Do you think I can define the same replacing V4L2_ prefix
>> with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
> Yes.
>
>> Do I need to define all those or there can be some subset of the
>> above for my simpler use-case?
> Most of these defines directly map to standards. I would skip the following
> defines:
>
> V4L2_COLORSPACE_DEFAULT (not applicable)
> V4L2_COLORSPACE_470_SYSTEM_* (rarely used, if received by the HW the Xen backend
> should map this to V4L2_COLORSPACE_SMPTE170M)
> V4L2_COLORSPACE_JPEG (historical V4L2 artifact, see here how to map:
> https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-details.html#col-jpeg)
>
> V4L2_COLORSPACE_SMPTE240M (rarely used, map to V4L2_COLORSPACE_SMPTE170M if seen in backend)
>
> V4L2_XFER_FUNC_SMPTE240M (rarely used, map to V4L2_XFER_FUNC_709)
>
> V4L2_YCBCR_ENC_SMPTE240M (rarely used, map to V4L2_YCBCR_ENC_709)
>
> While V4L2 allows 0 (DEFAULT) values for xfer_func, ycbcr_enc and quantization, and
> provides macros to map default values to the actual values (for legacy reasons),
> the Xen backend should always fill this in explicitly, using those same mapping
> macros (see e.g. V4L2_MAP_XFER_FUNC_DEFAULT).
>
> The V4L2 spec has extensive information on colorspaces (sections 2.14-2.17).
>
>>> The vivid driver can actually reproduce all combinations, so that's a good driver
>>> to test this with.
>> You mean I can use it on backend side instead of real HW camera and
>> test all the configurations possible/those of interest?
> Right.
>
> Regards,
>
> Hans
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-11 6:52 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-11 6:52 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
Hi, Hans!
On 09/10/2018 03:26 PM, Hans Verkuil wrote:
> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>> Hi Oleksandr,
>>>>>>>
>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>> <snip>
>>>>>
>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>> it pays to design this with that in mind.
>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>> use-cases from the frontend.
>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>
>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>> well, this complicates things a lot as I'll have to
>>>>>> re-allocate buffers - right?
>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>> go into multiple VMs.
>>>>>
>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>> colorspace information changes, but not the resolution.
>>>>>
>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>
>>>>> 1) You need to communicate the colorspace data:
>>>>>
>>>>> - colorspace
>>>>> - xfer_func
>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>> think you can ignore hsv_enc)
>>>>> - quantization
>>>>>
>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>
>>>>> This information is part of the format, it is reported by the driver.
>>>> I'll take a look and think what can be put and how into the protocol,
>>>> do you think I'll have to implement all the above for
>>>> this stage?
>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>> They don't *have* to use this information, but it should be there. For cameras
>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>> formats this is critical.
>> ok, then I'll add the following to the set_config request/response:
>>
>> uint32_t colorspace;
>> uint32_t xfer_func;
>> uint32_t ycbcr_enc;
>> uint32_t quantization;
Yet another question here: are the above (color space, xfer etc.) and
display aspect ratio defined per pixel_format or per pixel_format +
resolution?
If per pixel_format then
.../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
or if per resolution
.../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
>>
>> With this respect, I will need to put some OS agnostic constants
>> into the protocol, so if backend and frontend are not Linux/V4L2
>> based they can still talk to each other.
>> I see that V4L2 already defines constants for the above: [1], [2], [3], [4].
>>
>> Do you think I can define the same replacing V4L2_ prefix
>> with XENCAMERA_, e.g. V4L2_XFER_FUNC_SRGB -> XENCAMERA_XFER_FUNC_SRGB?
> Yes.
>
>> Do I need to define all those or there can be some subset of the
>> above for my simpler use-case?
> Most of these defines directly map to standards. I would skip the following
> defines:
>
> V4L2_COLORSPACE_DEFAULT (not applicable)
> V4L2_COLORSPACE_470_SYSTEM_* (rarely used, if received by the HW the Xen backend
> should map this to V4L2_COLORSPACE_SMPTE170M)
> V4L2_COLORSPACE_JPEG (historical V4L2 artifact, see here how to map:
> https://hverkuil.home.xs4all.nl/spec/uapi/v4l/colorspaces-details.html#col-jpeg)
>
> V4L2_COLORSPACE_SMPTE240M (rarely used, map to V4L2_COLORSPACE_SMPTE170M if seen in backend)
>
> V4L2_XFER_FUNC_SMPTE240M (rarely used, map to V4L2_XFER_FUNC_709)
>
> V4L2_YCBCR_ENC_SMPTE240M (rarely used, map to V4L2_YCBCR_ENC_709)
>
> While V4L2 allows 0 (DEFAULT) values for xfer_func, ycbcr_enc and quantization, and
> provides macros to map default values to the actual values (for legacy reasons),
> the Xen backend should always fill this in explicitly, using those same mapping
> macros (see e.g. V4L2_MAP_XFER_FUNC_DEFAULT).
>
> The V4L2 spec has extensive information on colorspaces (sections 2.14-2.17).
>
>>> The vivid driver can actually reproduce all combinations, so that's a good driver
>>> to test this with.
>> You mean I can use it on backend side instead of real HW camera and
>> test all the configurations possible/those of interest?
> Right.
>
> Regards,
>
> Hans
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-11 6:52 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-11 7:04 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-11 7:04 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr_Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
> Hi, Hans!
>
> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>> Hi Oleksandr,
>>>>>>>>
>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>> <snip>
>>>>>>
>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>> use-cases from the frontend.
>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>
>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>> re-allocate buffers - right?
>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>> go into multiple VMs.
>>>>>>
>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>> colorspace information changes, but not the resolution.
>>>>>>
>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>
>>>>>> 1) You need to communicate the colorspace data:
>>>>>>
>>>>>> - colorspace
>>>>>> - xfer_func
>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>> think you can ignore hsv_enc)
>>>>>> - quantization
>>>>>>
>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>
>>>>>> This information is part of the format, it is reported by the driver.
>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>> do you think I'll have to implement all the above for
>>>>> this stage?
>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>> They don't *have* to use this information, but it should be there. For cameras
>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>> formats this is critical.
>>> ok, then I'll add the following to the set_config request/response:
>>>
>>> uint32_t colorspace;
>>> uint32_t xfer_func;
>>> uint32_t ycbcr_enc;
>>> uint32_t quantization;
> Yet another question here: are the above (color space, xfer etc.) and
> display aspect ratio defined per pixel_format or per pixel_format +
> resolution?
>
> If per pixel_format then
>
> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>
> or if per resolution
>
> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
They are totally independent of resolution or pixelformat, with the
exception of ycbcr_enc which is of course ignored for RGB pixelformats.
They are set by the driver, never by the application.
For HDMI sources these values can change depending on what source is
connected, so they are not fixed and you need to query them whenever
a new source is connected. In fact, then can change midstream, but we
do not have good support for that at the moment.
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-11 7:04 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-11 7:04 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr_Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
> Hi, Hans!
>
> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>> Hi Oleksandr,
>>>>>>>>
>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>> <snip>
>>>>>>
>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>> use-cases from the frontend.
>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>
>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>> re-allocate buffers - right?
>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>> go into multiple VMs.
>>>>>>
>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>> colorspace information changes, but not the resolution.
>>>>>>
>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>
>>>>>> 1) You need to communicate the colorspace data:
>>>>>>
>>>>>> - colorspace
>>>>>> - xfer_func
>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>> think you can ignore hsv_enc)
>>>>>> - quantization
>>>>>>
>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>
>>>>>> This information is part of the format, it is reported by the driver.
>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>> do you think I'll have to implement all the above for
>>>>> this stage?
>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>> They don't *have* to use this information, but it should be there. For cameras
>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>> formats this is critical.
>>> ok, then I'll add the following to the set_config request/response:
>>>
>>> uint32_t colorspace;
>>> uint32_t xfer_func;
>>> uint32_t ycbcr_enc;
>>> uint32_t quantization;
> Yet another question here: are the above (color space, xfer etc.) and
> display aspect ratio defined per pixel_format or per pixel_format +
> resolution?
>
> If per pixel_format then
>
> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>
> or if per resolution
>
> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
They are totally independent of resolution or pixelformat, with the
exception of ycbcr_enc which is of course ignored for RGB pixelformats.
They are set by the driver, never by the application.
For HDMI sources these values can change depending on what source is
connected, so they are not fixed and you need to query them whenever
a new source is connected. In fact, then can change midstream, but we
do not have good support for that at the moment.
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-11 7:04 ` [PATCH " Hans Verkuil
@ 2018-09-11 7:14 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-11 7:14 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/11/2018 10:04 AM, Hans Verkuil wrote:
> On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
>> Hi, Hans!
>>
>> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>>> Hi Oleksandr,
>>>>>>>>>
>>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>>> <snip>
>>>>>>>
>>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>>> use-cases from the frontend.
>>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>>
>>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>>> re-allocate buffers - right?
>>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>>> go into multiple VMs.
>>>>>>>
>>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>>> colorspace information changes, but not the resolution.
>>>>>>>
>>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>>
>>>>>>> 1) You need to communicate the colorspace data:
>>>>>>>
>>>>>>> - colorspace
>>>>>>> - xfer_func
>>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>>> think you can ignore hsv_enc)
>>>>>>> - quantization
>>>>>>>
>>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>>
>>>>>>> This information is part of the format, it is reported by the driver.
>>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>>> do you think I'll have to implement all the above for
>>>>>> this stage?
>>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>>> They don't *have* to use this information, but it should be there. For cameras
>>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>>> formats this is critical.
>>>> ok, then I'll add the following to the set_config request/response:
>>>>
>>>> uint32_t colorspace;
>>>> uint32_t xfer_func;
>>>> uint32_t ycbcr_enc;
>>>> uint32_t quantization;
>> Yet another question here: are the above (color space, xfer etc.) and
>> display aspect ratio defined per pixel_format or per pixel_format +
>> resolution?
>>
>> If per pixel_format then
>>
>> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>>
>> or if per resolution
>>
>> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
> They are totally independent of resolution or pixelformat, with the
> exception of ycbcr_enc which is of course ignored for RGB pixelformats.
>
> They are set by the driver, never by the application.
>
> For HDMI sources these values can change depending on what source is
> connected, so they are not fixed and you need to query them whenever
> a new source is connected. In fact, then can change midstream, but we
> do not have good support for that at the moment.
Ah, great, then I'll define colorspace, xfer_func, quantization
and display aspect ratio as part of virtual camera device configuration
(as vcamera represents a single source) and ycbcr_enc as a part
of pixel format configuration (one ycbcr_enc per each
pixel format)
Does this sound ok?
>
> Regards,
>
> Hans
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-11 7:14 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-11 7:14 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/11/2018 10:04 AM, Hans Verkuil wrote:
> On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
>> Hi, Hans!
>>
>> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>>> Hi Oleksandr,
>>>>>>>>>
>>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>>> <snip>
>>>>>>>
>>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>>> use-cases from the frontend.
>>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>>
>>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>>> re-allocate buffers - right?
>>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>>> go into multiple VMs.
>>>>>>>
>>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>>> colorspace information changes, but not the resolution.
>>>>>>>
>>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>>
>>>>>>> 1) You need to communicate the colorspace data:
>>>>>>>
>>>>>>> - colorspace
>>>>>>> - xfer_func
>>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>>> think you can ignore hsv_enc)
>>>>>>> - quantization
>>>>>>>
>>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>>
>>>>>>> This information is part of the format, it is reported by the driver.
>>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>>> do you think I'll have to implement all the above for
>>>>>> this stage?
>>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>>> They don't *have* to use this information, but it should be there. For cameras
>>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>>> formats this is critical.
>>>> ok, then I'll add the following to the set_config request/response:
>>>>
>>>> uint32_t colorspace;
>>>> uint32_t xfer_func;
>>>> uint32_t ycbcr_enc;
>>>> uint32_t quantization;
>> Yet another question here: are the above (color space, xfer etc.) and
>> display aspect ratio defined per pixel_format or per pixel_format +
>> resolution?
>>
>> If per pixel_format then
>>
>> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>>
>> or if per resolution
>>
>> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
> They are totally independent of resolution or pixelformat, with the
> exception of ycbcr_enc which is of course ignored for RGB pixelformats.
>
> They are set by the driver, never by the application.
>
> For HDMI sources these values can change depending on what source is
> connected, so they are not fixed and you need to query them whenever
> a new source is connected. In fact, then can change midstream, but we
> do not have good support for that at the moment.
Ah, great, then I'll define colorspace, xfer_func, quantization
and display aspect ratio as part of virtual camera device configuration
(as vcamera represents a single source) and ycbcr_enc as a part
of pixel format configuration (one ycbcr_enc per each
pixel format)
Does this sound ok?
>
> Regards,
>
> Hans
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-11 7:14 ` [PATCH " Oleksandr Andrushchenko
@ 2018-09-11 7:52 ` Hans Verkuil
-1 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-11 7:52 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr_Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/11/18 09:14, Oleksandr Andrushchenko wrote:
> On 09/11/2018 10:04 AM, Hans Verkuil wrote:
>> On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
>>> Hi, Hans!
>>>
>>> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>>>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>>>> Hi Oleksandr,
>>>>>>>>>>
>>>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>>>> <snip>
>>>>>>>>
>>>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>>>> use-cases from the frontend.
>>>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>>>
>>>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>>>> re-allocate buffers - right?
>>>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>>>> go into multiple VMs.
>>>>>>>>
>>>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>>>> colorspace information changes, but not the resolution.
>>>>>>>>
>>>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>>>
>>>>>>>> 1) You need to communicate the colorspace data:
>>>>>>>>
>>>>>>>> - colorspace
>>>>>>>> - xfer_func
>>>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>>>> think you can ignore hsv_enc)
>>>>>>>> - quantization
>>>>>>>>
>>>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>>>
>>>>>>>> This information is part of the format, it is reported by the driver.
>>>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>>>> do you think I'll have to implement all the above for
>>>>>>> this stage?
>>>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>>>> They don't *have* to use this information, but it should be there. For cameras
>>>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>>>> formats this is critical.
>>>>> ok, then I'll add the following to the set_config request/response:
>>>>>
>>>>> uint32_t colorspace;
>>>>> uint32_t xfer_func;
>>>>> uint32_t ycbcr_enc;
>>>>> uint32_t quantization;
>>> Yet another question here: are the above (color space, xfer etc.) and
>>> display aspect ratio defined per pixel_format or per pixel_format +
>>> resolution?
>>>
>>> If per pixel_format then
>>>
>>> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>>>
>>> or if per resolution
>>>
>>> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
>> They are totally independent of resolution or pixelformat, with the
>> exception of ycbcr_enc which is of course ignored for RGB pixelformats.
>>
>> They are set by the driver, never by the application.
>>
>> For HDMI sources these values can change depending on what source is
>> connected, so they are not fixed and you need to query them whenever
>> a new source is connected. In fact, then can change midstream, but we
>> do not have good support for that at the moment.
> Ah, great, then I'll define colorspace, xfer_func, quantization
> and display aspect ratio as part of virtual camera device configuration
> (as vcamera represents a single source) and ycbcr_enc as a part
> of pixel format configuration (one ycbcr_enc per each
> pixel format)
>
> Does this sound ok?
Uh, no :-)
ycbcr_enc is not tied to specific pixel formats. The Y'CbCr encoding tells
you how the Y'CbCr values were derived from the R'G'B' values. So this only
makes sense if you are in fact receiving Y'CbCr pixels, otherwise you just
ignore it.
It's up to you what value to assign to ycbcr_enc in that case: V4L2 doesn't
have any hard requirements for that AFAIK, although it will most likely be
set to 0 (V4L2_YCBCR_ENC_DEFAULT).
Regards,
Hans
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-11 7:52 ` Hans Verkuil
0 siblings, 0 replies; 56+ messages in thread
From: Hans Verkuil @ 2018-09-11 7:52 UTC (permalink / raw)
To: Oleksandr Andrushchenko, Oleksandr_Andrushchenko, xen-devel,
konrad.wilk, jgross, boris.ostrovsky, mchehab, linux-media,
sakari.ailus, koji.matsuoka.xm
On 09/11/18 09:14, Oleksandr Andrushchenko wrote:
> On 09/11/2018 10:04 AM, Hans Verkuil wrote:
>> On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
>>> Hi, Hans!
>>>
>>> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>>>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>>>> Hi Oleksandr,
>>>>>>>>>>
>>>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>>>> <snip>
>>>>>>>>
>>>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>>>> use-cases from the frontend.
>>>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>>>
>>>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>>>> re-allocate buffers - right?
>>>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>>>> go into multiple VMs.
>>>>>>>>
>>>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>>>> colorspace information changes, but not the resolution.
>>>>>>>>
>>>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>>>
>>>>>>>> 1) You need to communicate the colorspace data:
>>>>>>>>
>>>>>>>> - colorspace
>>>>>>>> - xfer_func
>>>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>>>> think you can ignore hsv_enc)
>>>>>>>> - quantization
>>>>>>>>
>>>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>>>
>>>>>>>> This information is part of the format, it is reported by the driver.
>>>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>>>> do you think I'll have to implement all the above for
>>>>>>> this stage?
>>>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>>>> They don't *have* to use this information, but it should be there. For cameras
>>>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>>>> formats this is critical.
>>>>> ok, then I'll add the following to the set_config request/response:
>>>>>
>>>>> uint32_t colorspace;
>>>>> uint32_t xfer_func;
>>>>> uint32_t ycbcr_enc;
>>>>> uint32_t quantization;
>>> Yet another question here: are the above (color space, xfer etc.) and
>>> display aspect ratio defined per pixel_format or per pixel_format +
>>> resolution?
>>>
>>> If per pixel_format then
>>>
>>> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>>>
>>> or if per resolution
>>>
>>> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
>> They are totally independent of resolution or pixelformat, with the
>> exception of ycbcr_enc which is of course ignored for RGB pixelformats.
>>
>> They are set by the driver, never by the application.
>>
>> For HDMI sources these values can change depending on what source is
>> connected, so they are not fixed and you need to query them whenever
>> a new source is connected. In fact, then can change midstream, but we
>> do not have good support for that at the moment.
> Ah, great, then I'll define colorspace, xfer_func, quantization
> and display aspect ratio as part of virtual camera device configuration
> (as vcamera represents a single source) and ycbcr_enc as a part
> of pixel format configuration (one ycbcr_enc per each
> pixel format)
>
> Does this sound ok?
Uh, no :-)
ycbcr_enc is not tied to specific pixel formats. The Y'CbCr encoding tells
you how the Y'CbCr values were derived from the R'G'B' values. So this only
makes sense if you are in fact receiving Y'CbCr pixels, otherwise you just
ignore it.
It's up to you what value to assign to ycbcr_enc in that case: V4L2 doesn't
have any hard requirements for that AFAIK, although it will most likely be
set to 0 (V4L2_YCBCR_ENC_DEFAULT).
Regards,
Hans
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera
2018-09-11 7:52 ` [PATCH " Hans Verkuil
@ 2018-09-11 8:09 ` Oleksandr Andrushchenko
-1 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-11 8:09 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/11/2018 10:52 AM, Hans Verkuil wrote:
> On 09/11/18 09:14, Oleksandr Andrushchenko wrote:
>> On 09/11/2018 10:04 AM, Hans Verkuil wrote:
>>> On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
>>>> Hi, Hans!
>>>>
>>>> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>>>>> Hi Oleksandr,
>>>>>>>>>>>
>>>>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>> <snip>
>>>>>>>>>
>>>>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>>>>> use-cases from the frontend.
>>>>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>>>>
>>>>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>>>>> re-allocate buffers - right?
>>>>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>>>>> go into multiple VMs.
>>>>>>>>>
>>>>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>>>>> colorspace information changes, but not the resolution.
>>>>>>>>>
>>>>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>>>>
>>>>>>>>> 1) You need to communicate the colorspace data:
>>>>>>>>>
>>>>>>>>> - colorspace
>>>>>>>>> - xfer_func
>>>>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>>>>> think you can ignore hsv_enc)
>>>>>>>>> - quantization
>>>>>>>>>
>>>>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>>>>
>>>>>>>>> This information is part of the format, it is reported by the driver.
>>>>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>>>>> do you think I'll have to implement all the above for
>>>>>>>> this stage?
>>>>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>>>>> They don't *have* to use this information, but it should be there. For cameras
>>>>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>>>>> formats this is critical.
>>>>>> ok, then I'll add the following to the set_config request/response:
>>>>>>
>>>>>> uint32_t colorspace;
>>>>>> uint32_t xfer_func;
>>>>>> uint32_t ycbcr_enc;
>>>>>> uint32_t quantization;
>>>> Yet another question here: are the above (color space, xfer etc.) and
>>>> display aspect ratio defined per pixel_format or per pixel_format +
>>>> resolution?
>>>>
>>>> If per pixel_format then
>>>>
>>>> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>>>>
>>>> or if per resolution
>>>>
>>>> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
>>> They are totally independent of resolution or pixelformat, with the
>>> exception of ycbcr_enc which is of course ignored for RGB pixelformats.
>>>
>>> They are set by the driver, never by the application.
>>>
>>> For HDMI sources these values can change depending on what source is
>>> connected, so they are not fixed and you need to query them whenever
>>> a new source is connected. In fact, then can change midstream, but we
>>> do not have good support for that at the moment.
>> Ah, great, then I'll define colorspace, xfer_func, quantization
>> and display aspect ratio as part of virtual camera device configuration
>> (as vcamera represents a single source) and ycbcr_enc as a part
>> of pixel format configuration (one ycbcr_enc per each
>> pixel format)
>>
>> Does this sound ok?
> Uh, no :-)
>
> ycbcr_enc is not tied to specific pixel formats. The Y'CbCr encoding tells
> you how the Y'CbCr values were derived from the R'G'B' values. So this only
> makes sense if you are in fact receiving Y'CbCr pixels, otherwise you just
> ignore it.
>
> It's up to you what value to assign to ycbcr_enc in that case: V4L2 doesn't
> have any hard requirements for that AFAIK, although it will most likely be
> set to 0 (V4L2_YCBCR_ENC_DEFAULT).
Thank you for the explanation
> Regards,
>
> Hans
Thank you,
Oleksandr
^ permalink raw reply [flat|nested] 56+ messages in thread
* Re: [PATCH 1/1] cameraif: add ABI for para-virtual camera
@ 2018-09-11 8:09 ` Oleksandr Andrushchenko
0 siblings, 0 replies; 56+ messages in thread
From: Oleksandr Andrushchenko @ 2018-09-11 8:09 UTC (permalink / raw)
To: Hans Verkuil, Oleksandr_Andrushchenko, xen-devel, konrad.wilk,
jgross, boris.ostrovsky, mchehab, linux-media, sakari.ailus,
koji.matsuoka.xm
On 09/11/2018 10:52 AM, Hans Verkuil wrote:
> On 09/11/18 09:14, Oleksandr Andrushchenko wrote:
>> On 09/11/2018 10:04 AM, Hans Verkuil wrote:
>>> On 09/11/2018 08:52 AM, Oleksandr Andrushchenko wrote:
>>>> Hi, Hans!
>>>>
>>>> On 09/10/2018 03:26 PM, Hans Verkuil wrote:
>>>>> On 09/10/2018 01:49 PM, Oleksandr Andrushchenko wrote:
>>>>>> On 09/10/2018 02:09 PM, Hans Verkuil wrote:
>>>>>>> On 09/10/2018 11:52 AM, Oleksandr Andrushchenko wrote:
>>>>>>>> On 09/10/2018 12:04 PM, Hans Verkuil wrote:
>>>>>>>>> On 09/10/2018 10:24 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>>> On 09/10/2018 10:53 AM, Hans Verkuil wrote:
>>>>>>>>>>> Hi Oleksandr,
>>>>>>>>>>>
>>>>>>>>>>> On 09/10/2018 09:16 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>> <snip>
>>>>>>>>>
>>>>>>>>>>>>> I suspect that you likely will want to support such sources eventually, so
>>>>>>>>>>>>> it pays to design this with that in mind.
>>>>>>>>>>>> Again, I think that this is the backend to hide these
>>>>>>>>>>>> use-cases from the frontend.
>>>>>>>>>>> I'm not sure you can: say you are playing a bluray connected to the system
>>>>>>>>>>> with HDMI, then if there is a resolution change, what do you do? You can tear
>>>>>>>>>>> everything down and build it up again, or you can just tell frontends that
>>>>>>>>>>> something changed and that they have to look at the new vcamera configuration.
>>>>>>>>>>>
>>>>>>>>>>> The latter seems to be more sensible to me. It is really not much that you
>>>>>>>>>>> need to do: all you really need is an event signalling that something changed.
>>>>>>>>>>> In V4L2 that's the V4L2_EVENT_SOURCE_CHANGE.
>>>>>>>>>> well, this complicates things a lot as I'll have to
>>>>>>>>>> re-allocate buffers - right?
>>>>>>>>> Right. Different resolutions means different sized buffers and usually lots of
>>>>>>>>> changes throughout the whole video pipeline, which in this case can even
>>>>>>>>> go into multiple VMs.
>>>>>>>>>
>>>>>>>>> One additional thing to keep in mind for the future: V4L2_EVENT_SOURCE_CHANGE
>>>>>>>>> has a flags field that tells userspace what changed. Right now that is just the
>>>>>>>>> resolution, but in the future you can expect flags for cases where just the
>>>>>>>>> colorspace information changes, but not the resolution.
>>>>>>>>>
>>>>>>>>> Which reminds me of two important missing pieces of information in your protocol:
>>>>>>>>>
>>>>>>>>> 1) You need to communicate the colorspace data:
>>>>>>>>>
>>>>>>>>> - colorspace
>>>>>>>>> - xfer_func
>>>>>>>>> - ycbcr_enc/hsv_enc (unlikely you ever want to support HSV pixelformats, so I
>>>>>>>>> think you can ignore hsv_enc)
>>>>>>>>> - quantization
>>>>>>>>>
>>>>>>>>> See https://hverkuil.home.xs4all.nl/spec/uapi/v4l/pixfmt-v4l2.html#c.v4l2_pix_format
>>>>>>>>> and the links to the colorspace sections in the V4L2 spec for details).
>>>>>>>>>
>>>>>>>>> This information is part of the format, it is reported by the driver.
>>>>>>>> I'll take a look and think what can be put and how into the protocol,
>>>>>>>> do you think I'll have to implement all the above for
>>>>>>>> this stage?
>>>>>>> Yes. Without it VMs will have no way of knowing how to reproduce the right colors.
>>>>>>> They don't *have* to use this information, but it should be there. For cameras
>>>>>>> this isn't all that important, for SDTV/HDTV sources this becomes more relevant
>>>>>>> (esp. the quantization and ycbcr_enc information) and for sources with BT.2020/HDR
>>>>>>> formats this is critical.
>>>>>> ok, then I'll add the following to the set_config request/response:
>>>>>>
>>>>>> uint32_t colorspace;
>>>>>> uint32_t xfer_func;
>>>>>> uint32_t ycbcr_enc;
>>>>>> uint32_t quantization;
>>>> Yet another question here: are the above (color space, xfer etc.) and
>>>> display aspect ratio defined per pixel_format or per pixel_format +
>>>> resolution?
>>>>
>>>> If per pixel_format then
>>>>
>>>> .../vcamera/1/formats/YUYV/display-aspect-ratio = "59/58"
>>>>
>>>> or if per resolution
>>>>
>>>> .../vcamera/1/formats/YUYV/640x480/display-aspect-ratio = "59/58"
>>> They are totally independent of resolution or pixelformat, with the
>>> exception of ycbcr_enc which is of course ignored for RGB pixelformats.
>>>
>>> They are set by the driver, never by the application.
>>>
>>> For HDMI sources these values can change depending on what source is
>>> connected, so they are not fixed and you need to query them whenever
>>> a new source is connected. In fact, then can change midstream, but we
>>> do not have good support for that at the moment.
>> Ah, great, then I'll define colorspace, xfer_func, quantization
>> and display aspect ratio as part of virtual camera device configuration
>> (as vcamera represents a single source) and ycbcr_enc as a part
>> of pixel format configuration (one ycbcr_enc per each
>> pixel format)
>>
>> Does this sound ok?
> Uh, no :-)
>
> ycbcr_enc is not tied to specific pixel formats. The Y'CbCr encoding tells
> you how the Y'CbCr values were derived from the R'G'B' values. So this only
> makes sense if you are in fact receiving Y'CbCr pixels, otherwise you just
> ignore it.
>
> It's up to you what value to assign to ycbcr_enc in that case: V4L2 doesn't
> have any hard requirements for that AFAIK, although it will most likely be
> set to 0 (V4L2_YCBCR_ENC_DEFAULT).
Thank you for the explanation
> Regards,
>
> Hans
Thank you,
Oleksandr
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
^ permalink raw reply [flat|nested] 56+ messages in thread
end of thread, other threads:[~2018-09-11 13:07 UTC | newest]
Thread overview: 56+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-31 9:31 [Xen-devel][PATCH 0/1] cameraif: Add ABI for para-virtualized Oleksandr Andrushchenko
2018-07-31 9:31 ` [PATCH " Oleksandr Andrushchenko
2018-07-31 9:31 ` [Xen-devel][PATCH 1/1] cameraif: add ABI for para-virtual camera Oleksandr Andrushchenko
2018-07-31 9:31 ` [PATCH " Oleksandr Andrushchenko
2018-08-14 8:30 ` [Xen-devel][PATCH " Juergen Gross
2018-08-14 8:30 ` [PATCH " Juergen Gross
2018-08-21 5:54 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-08-21 5:54 ` [PATCH " Oleksandr Andrushchenko
2018-09-03 10:16 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-03 10:16 ` [PATCH " Oleksandr Andrushchenko
2018-09-03 15:25 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-03 15:25 ` [PATCH " Hans Verkuil
2018-09-04 6:56 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-04 6:56 ` [PATCH " Oleksandr Andrushchenko
2018-09-09 10:42 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-09 10:42 ` [PATCH " Hans Verkuil
2018-09-10 5:59 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 5:59 ` [PATCH " Oleksandr Andrushchenko
2018-09-10 8:14 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-10 8:14 ` [PATCH " Hans Verkuil
2018-09-10 8:34 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 8:34 ` [PATCH " Oleksandr Andrushchenko
2018-09-09 10:31 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-09 10:31 ` [PATCH " Hans Verkuil
2018-09-10 7:16 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 7:16 ` [PATCH " Oleksandr Andrushchenko
2018-09-10 7:53 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-10 7:53 ` [PATCH " Hans Verkuil
2018-09-10 8:24 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 8:24 ` [PATCH " Oleksandr Andrushchenko
2018-09-10 9:04 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-10 9:04 ` [PATCH " Hans Verkuil
2018-09-10 9:52 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 9:52 ` [PATCH " Oleksandr Andrushchenko
2018-09-10 11:09 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-10 11:09 ` [PATCH " Hans Verkuil
2018-09-10 11:49 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 11:49 ` [PATCH " Oleksandr Andrushchenko
2018-09-10 12:26 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-10 12:26 ` [PATCH " Hans Verkuil
2018-09-10 13:16 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 13:16 ` [PATCH " Oleksandr Andrushchenko
2018-09-11 6:52 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-11 6:52 ` [PATCH " Oleksandr Andrushchenko
2018-09-11 7:04 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-11 7:04 ` [PATCH " Hans Verkuil
2018-09-11 7:14 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-11 7:14 ` [PATCH " Oleksandr Andrushchenko
2018-09-11 7:52 ` [Xen-devel][PATCH " Hans Verkuil
2018-09-11 7:52 ` [PATCH " Hans Verkuil
2018-09-11 8:09 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-11 8:09 ` [PATCH " Oleksandr Andrushchenko
2018-09-10 12:48 ` [Xen-devel][PATCH 0/1] cameraif: Add ABI for para-virtualized Laurent Pinchart
2018-09-10 12:48 ` [PATCH " Laurent Pinchart
2018-09-10 13:02 ` [Xen-devel][PATCH " Oleksandr Andrushchenko
2018-09-10 13:02 ` [PATCH " Oleksandr Andrushchenko
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.