All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v2 00/16] Add MHI bus support
@ 2020-01-31 13:49 Manivannan Sadhasivam
  2020-01-31 13:49 ` [PATCH v2 01/16] docs: Add documentation for MHI bus Manivannan Sadhasivam
                   ` (15 more replies)
  0 siblings, 16 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

Hello,

This is the second attempt at adding the MHI (Modem Host Interface) bus
interface to Linux kernel. MHI is a communication protocol used by the
host processors to control and communicate with modems over a high
speed peripheral bus or shared memory. The MHI protocol has been
designed and developed by Qualcomm Innovation Center, Inc., for use
in their modems.

The first submission was made by Sujeev Dias of Qualcomm:

https://lkml.org/lkml/2018/4/26/1159
https://lkml.org/lkml/2018/7/9/987

This series addresses most of the review comments by Greg and Arnd for
the initial patchset. Furthermore, in order to ease the review process
I've splitted the patches logically and dropped few of them which were
not required for this initial submission.

Below is the high level changelog:

1. Removed all DT related code
2. Got rid of pci specific struct members from top level mhi structs
3. Moved device specific callbacks like ul_xfer() to driver struct. It
   doesn’t make sense to have callbacks in device struct as suggested by
   Greg
4. Used priv data of `struct device` instead of own priv data in
   `mhi_device` as suggested by Greg. This will allow us to use
    dev_set{get}_drvdata() APIs in client drivers
5. Removed all debugfs related code
6. Changes to the APIs to look uniform
7. Converted the documentation to .rst and placed in its own subdirectory
8. Changes to the MHI device naming
9. Converted all uppercase variable names to appropriate lowercase ones
10. Removed custom debug code and used the dev_* ones where applicable
11. Dropped timesync, DTR, UCI, and Qcom controller related codes
12. Added QRTR client driver patch
13. Added modalias support for the MHI stack as well as client driver for
    autoloading of modules (client drivers) by udev once the MHI devices
    are created

This series includes the MHI stack as well as the QRTR client driver which
falls under the networking subsystem.

The reference controller implementation is here:
https://git.linaro.org/people/manivannan.sadhasivam/linux.git/tree/drivers/net/wireless/ath/ath11k/mhi.c?h=ath11k-qca6390-mhi
It will be submitted later along with ath11k patches.

Following developers deserve explicit acknowledgements for their
contributions to the MHI code:

Sujeev Dias
Siddartha Mohanadoss
Hemant Kumar
Jeff Hugo

Thanks,
Mani

Changes in v2:

* Added put_device to mhi_dealloc_device
* Removed unused members from struct mhi_controller
* Removed the atomicity of dev_wake in struct mhi_device as it is not required
* Reordered MHI structs to avoid holes
* Used struct device name for the controller device
* Marked the required and optional mhi_controller members for helping the
  controller driver implementation
* Cleanups to the MHI doc
* Removed _relaxed variants and used readl/writel
* Added comments for MHI specific acronyms
* Removed the usage of bitfields and used bitmasks for mhi_event_ctxt and
  mhi_chan_ctxt
* Used __64/__u32 types for structures representing hw states
* Added Hemant as a co-maintainer of MHI bus. He is from the MHI team of
  Qualcomm and he will take up reviews and if possible, maintainership
  in future.

Manivannan Sadhasivam (16):
  docs: Add documentation for MHI bus
  bus: mhi: core: Add support for registering MHI controllers
  bus: mhi: core: Add support for registering MHI client drivers
  bus: mhi: core: Add support for creating and destroying MHI devices
  bus: mhi: core: Add support for ringing channel/event ring doorbells
  bus: mhi: core: Add support for PM state transitions
  bus: mhi: core: Add support for basic PM operations
  bus: mhi: core: Add support for downloading firmware over BHIe
  bus: mhi: core: Add support for downloading RDDM image during panic
  bus: mhi: core: Add support for processing events from client device
  bus: mhi: core: Add support for data transfer
  bus: mhi: core: Add uevent support for module autoloading
  MAINTAINERS: Add entry for MHI bus
  net: qrtr: Add MHI transport layer
  net: qrtr: Do not depend on ARCH_QCOM
  soc: qcom: Do not depend on ARCH_QCOM for QMI helpers

 Documentation/index.rst           |    1 +
 Documentation/mhi/index.rst       |   18 +
 Documentation/mhi/mhi.rst         |  218 ++++
 Documentation/mhi/topology.rst    |   60 ++
 MAINTAINERS                       |   10 +
 drivers/bus/Kconfig               |    1 +
 drivers/bus/Makefile              |    3 +
 drivers/bus/mhi/Kconfig           |   14 +
 drivers/bus/mhi/Makefile          |    2 +
 drivers/bus/mhi/core/Makefile     |    3 +
 drivers/bus/mhi/core/boot.c       |  508 ++++++++++
 drivers/bus/mhi/core/init.c       | 1301 ++++++++++++++++++++++++
 drivers/bus/mhi/core/internal.h   |  699 +++++++++++++
 drivers/bus/mhi/core/main.c       | 1576 +++++++++++++++++++++++++++++
 drivers/bus/mhi/core/pm.c         |  967 ++++++++++++++++++
 drivers/soc/qcom/Kconfig          |    1 -
 include/linux/mhi.h               |  661 ++++++++++++
 include/linux/mod_devicetable.h   |   13 +
 net/qrtr/Kconfig                  |    8 +-
 net/qrtr/Makefile                 |    2 +
 net/qrtr/mhi.c                    |  207 ++++
 scripts/mod/devicetable-offsets.c |    3 +
 scripts/mod/file2alias.c          |   10 +
 23 files changed, 6284 insertions(+), 2 deletions(-)
 create mode 100644 Documentation/mhi/index.rst
 create mode 100644 Documentation/mhi/mhi.rst
 create mode 100644 Documentation/mhi/topology.rst
 create mode 100644 drivers/bus/mhi/Kconfig
 create mode 100644 drivers/bus/mhi/Makefile
 create mode 100644 drivers/bus/mhi/core/Makefile
 create mode 100644 drivers/bus/mhi/core/boot.c
 create mode 100644 drivers/bus/mhi/core/init.c
 create mode 100644 drivers/bus/mhi/core/internal.h
 create mode 100644 drivers/bus/mhi/core/main.c
 create mode 100644 drivers/bus/mhi/core/pm.c
 create mode 100644 include/linux/mhi.h
 create mode 100644 net/qrtr/mhi.c

-- 
2.17.1


^ permalink raw reply	[flat|nested] 62+ messages in thread

* [PATCH v2 01/16] docs: Add documentation for MHI bus
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
@ 2020-01-31 13:49 ` Manivannan Sadhasivam
  2020-01-31 22:57   ` Jeffrey Hugo
  2020-01-31 13:49 ` [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers Manivannan Sadhasivam
                   ` (14 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam, Jonathan Corbet, linux-doc

MHI (Modem Host Interface) is a communication protocol used by the
host processors to control and communicate with modems over a high
speed peripheral bus or shared memory. The MHI protocol has been
designed and developed by Qualcomm Innovation Center, Inc., for use
in their modems. This commit adds the documentation for the bus and
the implementation in Linux kernel.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/987

Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-doc@vger.kernel.org
Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: converted to .rst and splitted the patch]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 Documentation/index.rst        |   1 +
 Documentation/mhi/index.rst    |  18 +++
 Documentation/mhi/mhi.rst      | 218 +++++++++++++++++++++++++++++++++
 Documentation/mhi/topology.rst |  60 +++++++++
 4 files changed, 297 insertions(+)
 create mode 100644 Documentation/mhi/index.rst
 create mode 100644 Documentation/mhi/mhi.rst
 create mode 100644 Documentation/mhi/topology.rst

diff --git a/Documentation/index.rst b/Documentation/index.rst
index e99d0bd2589d..edc9b211bbff 100644
--- a/Documentation/index.rst
+++ b/Documentation/index.rst
@@ -133,6 +133,7 @@ needed).
    misc-devices/index
    mic/index
    scheduler/index
+   mhi/index
 
 Architecture-agnostic documentation
 -----------------------------------
diff --git a/Documentation/mhi/index.rst b/Documentation/mhi/index.rst
new file mode 100644
index 000000000000..1d8dec302780
--- /dev/null
+++ b/Documentation/mhi/index.rst
@@ -0,0 +1,18 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+===
+MHI
+===
+
+.. toctree::
+   :maxdepth: 1
+
+   mhi
+   topology
+
+.. only::  subproject and html
+
+   Indices
+   =======
+
+   * :ref:`genindex`
diff --git a/Documentation/mhi/mhi.rst b/Documentation/mhi/mhi.rst
new file mode 100644
index 000000000000..803ff84f7d7b
--- /dev/null
+++ b/Documentation/mhi/mhi.rst
@@ -0,0 +1,218 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+==========================
+MHI (Modem Host Interface)
+==========================
+
+This document provides information about the MHI protocol.
+
+Overview
+========
+
+MHI is a protocol developed by Qualcomm Innovation Center, Inc. It is used
+by the host processors to control and communicate with modem devices over high
+speed peripheral buses or shared memory. Even though MHI can be easily adapted
+to any peripheral buses, it is primarily used with PCIe based devices. MHI
+provides logical channels over the physical buses and allows transporting the
+modem protocols, such as IP data packets, modem control messages, and
+diagnostics over at least one of those logical channels. Also, the MHI
+protocol provides data acknowledgment feature and manages the power state of the
+modems via one or more logical channels.
+
+MHI Internals
+=============
+
+MMIO
+----
+
+MMIO (Memory mapped IO) consists of a set of registers in the device hardware,
+which are mapped to the host memory space by the peripheral buses like PCIe.
+Following are the major components of MMIO register space:
+
+MHI control registers: Access to MHI configurations registers
+
+MHI BHI registers: BHI (Boot Host Interface) registers are used by the host
+for downloading the firmware to the device before MHI initialization.
+
+Channel Doorbell array: Channel Doorbell (DB) registers used by the host to
+notify the device when there is new work to do.
+
+Event Doorbell array: Associated with event context array, the Event Doorbell
+(DB) registers are used by the host to notify the device when new events are
+available.
+
+Debug registers: A set of registers and counters used by the device to expose
+debugging information like performance, functional, and stability to the host.
+
+Data structures
+---------------
+
+All data structures used by MHI are in the host system memory. Using the
+physical interface, the device accesses those data structures. MHI data
+structures and data buffers in the host system memory regions are mapped for
+the device.
+
+Channel context array: All channel configurations are organized in channel
+context data array.
+
+Transfer rings: Used by the host to schedule work items for a channel. The
+transfer rings are organized as a circular queue of Transfer Descriptors (TD).
+
+Event context array: All event configurations are organized in the event context
+data array.
+
+Event rings: Used by the device to send completion and state transition messages
+to the host
+
+Command context array: All command configurations are organized in command
+context data array.
+
+Command rings: Used by the host to send MHI commands to the device. The command
+rings are organized as a circular queue of Command Descriptors (CD).
+
+Channels
+--------
+
+MHI channels are logical, unidirectional data pipes between a host and a device.
+The concept of channels in MHI is similar to endpoints in USB. MHI supports up
+to 256 channels. However, specific device implementations may support less than
+the maximum number of channels allowed.
+
+Two unidirectional channels with their associated transfer rings form a
+bidirectional data pipe, which can be used by the upper-layer protocols to
+transport application data packets (such as IP packets, modem control messages,
+diagnostics messages, and so on). Each channel is associated with a single
+transfer ring.
+
+Transfer rings
+--------------
+
+Transfers between the host and device are organized by channels and defined by
+Transfer Descriptors (TD). TDs are managed through transfer rings, which are
+defined for each channel between the device and host and reside in the host
+memory. TDs consist of one or more ring elements (or transfer blocks)::
+
+        [Read Pointer (RP)] ----------->[Ring Element] } TD
+        [Write Pointer (WP)]-           [Ring Element]
+                             -          [Ring Element]
+                              --------->[Ring Element]
+                                        [Ring Element]
+
+Below is the basic usage of transfer rings:
+
+* Host allocates memory for transfer ring.
+* Host sets the base pointer, read pointer, and write pointer in corresponding
+  channel context.
+* Ring is considered empty when RP == WP.
+* Ring is considered full when WP + 1 == RP.
+* RP indicates the next element to be serviced by the device.
+* When the host has a new buffer to send, it updates the ring element with
+  buffer information, increments the WP to the next element and rings the
+  associated channel DB.
+
+Event rings
+-----------
+
+Events from the device to host are organized in event rings and defined by Event
+Descriptors (ED). Event rings are used by the device to report events such as
+data transfer completion status, command completion status, and state changes
+to the host. Event rings are the array of EDs that resides in the host
+memory. EDs consist of one or more ring elements (or transfer blocks)::
+
+        [Read Pointer (RP)] ----------->[Ring Element] } ED
+        [Write Pointer (WP)]-           [Ring Element]
+                             -          [Ring Element]
+                              --------->[Ring Element]
+                                        [Ring Element]
+
+Below is the basic usage of event rings:
+
+* Host allocates memory for event ring.
+* Host sets the base pointer, read pointer, and write pointer in corresponding
+  channel context.
+* Both host and device has a local copy of RP, WP.
+* Ring is considered empty (no events to service) when WP + 1 == RP.
+* Ring is considered full of events when RP == WP.
+* When there is a new event the device needs to send, the device updates ED
+  pointed by RP, increments the RP to the next element and triggers the
+  interrupt.
+
+Ring Element
+------------
+
+A Ring Element is a data structure used to transfer a single block
+of data between the host and the device. Transfer ring element types contain a
+single buffer pointer, the size of the buffer, and additional control
+information. Other ring element types may only contain control and status
+information. For single buffer operations, a ring descriptor is composed of a
+single element. For large multi-buffer operations (such as scatter and gather),
+elements can be chained to form a longer descriptor.
+
+MHI Operations
+==============
+
+MHI States
+----------
+
+MHI_STATE_RESET
+~~~~~~~~~~~~~~~
+MHI is in reset state after power-up or hardware reset. The host is not allowed
+to access device MMIO register space.
+
+MHI_STATE_READY
+~~~~~~~~~~~~~~~
+MHI is ready for initialization. The host can start MHI initialization by
+programming MMIO registers.
+
+MHI_STATE_M0
+~~~~~~~~~~~~
+MHI is running and operational in the device. The host can start channels by
+issuing channel start command.
+
+MHI_STATE_M1
+~~~~~~~~~~~~
+MHI operation is suspended by the device. This state is entered when the
+device detects inactivity at the physical interface within a preset time.
+
+MHI_STATE_M2
+~~~~~~~~~~~~
+MHI is in low power state. MHI operation is suspended and the device may
+enter lower power mode.
+
+MHI_STATE_M3
+~~~~~~~~~~~~
+MHI operation stopped by the host. This state is entered when the host suspends
+MHI operation.
+
+MHI Initialization
+------------------
+
+After system boots, the device is enumerated over the physical interface.
+In the case of PCIe, the device is enumerated and assigned BAR-0 for
+the device's MMIO register space. To initialize the MHI in a device,
+the host performs the following operations:
+
+* Allocates the MHI context for event, channel and command arrays.
+* Initializes the context array, and prepares interrupts.
+* Waits until the device enters READY state.
+* Programs MHI MMIO registers and sets device into MHI_M0 state.
+* Waits for the device to enter M0 state.
+
+MHI Data Transfer
+-----------------
+
+MHI data transfer is initiated by the host to transfer data to the device.
+Following are the sequence of operations performed by the host to transfer
+data to device:
+
+* Host prepares TD with buffer information.
+* Host increments the WP of the corresponding channel transfer ring.
+* Host rings the channel DB register.
+* Device wakes up to process the TD.
+* Device generates a completion event for the processed TD by updating ED.
+* Device increments the RP of the corresponding event ring.
+* Device triggers IRQ to wake up the host.
+* Host wakes up and checks the event ring for completion event.
+* Host updates the WP of the corresponding event ring to indicate that the
+  data transfer has been completed successfully.
+
diff --git a/Documentation/mhi/topology.rst b/Documentation/mhi/topology.rst
new file mode 100644
index 000000000000..dc7799d03294
--- /dev/null
+++ b/Documentation/mhi/topology.rst
@@ -0,0 +1,60 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+============
+MHI Topology
+============
+
+This document provides information about the MHI topology modeling and
+representation in the kernel.
+
+MHI Controller
+--------------
+
+MHI controller driver manages the interaction with the MHI client devices
+such as the external modems and WiFi chipsets. It is also the MHI bus master
+which is in charge of managing the physical link between the host and device.
+It is however not involved in the actual data transfer as the data transfer
+is taken care by the physical bus such as PCIe. Each controller driver exposes
+channels and events based on the client device type.
+
+Below are the roles of the MHI controller driver:
+
+* Turns on the physical bus and establishes the link to the device
+* Configures IRQs, IOMMU, and IOMEM
+* Allocates struct mhi_controller and registers with the MHI bus framework
+  with channel and event configurations using mhi_register_controller.
+* Initiates power on and shutdown sequence
+* Initiates suspend and resume power management operations of the device.
+
+MHI Device
+----------
+
+MHI device is the logical device which binds to a maximum of two MHI channels
+for bi-directional communication. Once MHI is in powered on state, the MHI
+core will create MHI devices based on the channel configuration exposed
+by the controller. There can be a single MHI device for each channel or for a
+couple of channels.
+
+Each supported device is enumerated in::
+
+        /sys/bus/mhi/devices/
+
+MHI Driver
+----------
+
+MHI driver is the client driver which binds to one or more MHI devices. The MHI
+driver sends and receives the upper-layer protocol packets like IP packets,
+modem control messages, and diagnostics messages over MHI. The MHI core will
+bind the MHI devices to the MHI driver.
+
+Each supported driver is enumerated in::
+
+        /sys/bus/mhi/drivers/
+
+Below are the roles of the MHI driver:
+
+* Registers the driver with the MHI bus framework using mhi_driver_register.
+* Prepares the device for transfer by calling mhi_prepare_for_transfer.
+* Initiates data transfer by calling mhi_queue_transfer.
+* Once the data transfer is finished, calls mhi_unprepare_from_transfer to
+  end data transfer.
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
  2020-01-31 13:49 ` [PATCH v2 01/16] docs: Add documentation for MHI bus Manivannan Sadhasivam
@ 2020-01-31 13:49 ` Manivannan Sadhasivam
  2020-02-06 16:56   ` Greg KH
                     ` (2 more replies)
  2020-01-31 13:49 ` [PATCH v2 03/16] bus: mhi: core: Add support for registering MHI client drivers Manivannan Sadhasivam
                   ` (13 subsequent siblings)
  15 siblings, 3 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for registering MHI controller drivers with
the MHI stack. MHI controller drivers manages the interaction with the
MHI client devices such as the external modems and WiFi chipsets. They
are also the MHI bus master in charge of managing the physical link
between the host and client device.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/987

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[jhugo: added static config for controllers and fixed several bugs]
Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
[mani: removed DT dependency, splitted and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/Kconfig             |   1 +
 drivers/bus/Makefile            |   3 +
 drivers/bus/mhi/Kconfig         |  14 ++
 drivers/bus/mhi/Makefile        |   2 +
 drivers/bus/mhi/core/Makefile   |   3 +
 drivers/bus/mhi/core/init.c     | 407 +++++++++++++++++++++++++++++++
 drivers/bus/mhi/core/internal.h | 163 +++++++++++++
 include/linux/mhi.h             | 420 ++++++++++++++++++++++++++++++++
 include/linux/mod_devicetable.h |  12 +
 9 files changed, 1025 insertions(+)
 create mode 100644 drivers/bus/mhi/Kconfig
 create mode 100644 drivers/bus/mhi/Makefile
 create mode 100644 drivers/bus/mhi/core/Makefile
 create mode 100644 drivers/bus/mhi/core/init.c
 create mode 100644 drivers/bus/mhi/core/internal.h
 create mode 100644 include/linux/mhi.h

diff --git a/drivers/bus/Kconfig b/drivers/bus/Kconfig
index 50200d1c06ea..383934e54786 100644
--- a/drivers/bus/Kconfig
+++ b/drivers/bus/Kconfig
@@ -202,5 +202,6 @@ config DA8XX_MSTPRI
 	  peripherals.
 
 source "drivers/bus/fsl-mc/Kconfig"
+source "drivers/bus/mhi/Kconfig"
 
 endmenu
diff --git a/drivers/bus/Makefile b/drivers/bus/Makefile
index 1320bcf9fa9d..05f32cd694a4 100644
--- a/drivers/bus/Makefile
+++ b/drivers/bus/Makefile
@@ -34,3 +34,6 @@ obj-$(CONFIG_UNIPHIER_SYSTEM_BUS)	+= uniphier-system-bus.o
 obj-$(CONFIG_VEXPRESS_CONFIG)	+= vexpress-config.o
 
 obj-$(CONFIG_DA8XX_MSTPRI)	+= da8xx-mstpri.o
+
+# MHI
+obj-$(CONFIG_MHI_BUS)		+= mhi/
diff --git a/drivers/bus/mhi/Kconfig b/drivers/bus/mhi/Kconfig
new file mode 100644
index 000000000000..a8bd9bd7db7c
--- /dev/null
+++ b/drivers/bus/mhi/Kconfig
@@ -0,0 +1,14 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# MHI bus
+#
+# Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+#
+
+config MHI_BUS
+       tristate "Modem Host Interface (MHI) bus"
+       help
+	 Bus driver for MHI protocol. Modem Host Interface (MHI) is a
+	 communication protocol used by the host processors to control
+	 and communicate with modem devices over a high speed peripheral
+	 bus or shared memory.
diff --git a/drivers/bus/mhi/Makefile b/drivers/bus/mhi/Makefile
new file mode 100644
index 000000000000..19e6443b72df
--- /dev/null
+++ b/drivers/bus/mhi/Makefile
@@ -0,0 +1,2 @@
+# core layer
+obj-y += core/
diff --git a/drivers/bus/mhi/core/Makefile b/drivers/bus/mhi/core/Makefile
new file mode 100644
index 000000000000..2db32697c67f
--- /dev/null
+++ b/drivers/bus/mhi/core/Makefile
@@ -0,0 +1,3 @@
+obj-$(CONFIG_MHI_BUS) := mhi.o
+
+mhi-y := init.o
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
new file mode 100644
index 000000000000..a00f1547d252
--- /dev/null
+++ b/drivers/bus/mhi/core/init.c
@@ -0,0 +1,407 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ *
+ */
+
+#define dev_fmt(fmt) "MHI: " fmt
+
+#include <linux/device.h>
+#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/mhi.h>
+#include <linux/mod_devicetable.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/wait.h>
+#include "internal.h"
+
+static int parse_ev_cfg(struct mhi_controller *mhi_cntrl,
+			struct mhi_controller_config *config)
+{
+	int i, num;
+	struct mhi_event *mhi_event;
+	struct mhi_event_config *event_cfg;
+
+	num = config->num_events;
+	mhi_cntrl->total_ev_rings = num;
+	mhi_cntrl->mhi_event = kcalloc(num, sizeof(*mhi_cntrl->mhi_event),
+				       GFP_KERNEL);
+	if (!mhi_cntrl->mhi_event)
+		return -ENOMEM;
+
+	/* Populate event ring */
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < num; i++) {
+		event_cfg = &config->event_cfg[i];
+
+		mhi_event->er_index = i;
+		mhi_event->ring.elements = event_cfg->num_elements;
+		mhi_event->intmod = event_cfg->irq_moderation_ms;
+		mhi_event->irq = event_cfg->irq;
+
+		if (event_cfg->channel != U32_MAX) {
+			/* This event ring has a dedicated channel */
+			mhi_event->chan = event_cfg->channel;
+			if (mhi_event->chan >= mhi_cntrl->max_chan) {
+				dev_err(mhi_cntrl->dev,
+					"Event Ring channel not available\n");
+				goto error_ev_cfg;
+			}
+
+			mhi_event->mhi_chan =
+				&mhi_cntrl->mhi_chan[mhi_event->chan];
+		}
+
+		/* Priority is fixed to 1 for now */
+		mhi_event->priority = 1;
+
+		mhi_event->db_cfg.brstmode = event_cfg->mode;
+		if (MHI_INVALID_BRSTMODE(mhi_event->db_cfg.brstmode))
+			goto error_ev_cfg;
+
+		mhi_event->data_type = event_cfg->data_type;
+
+		mhi_event->hw_ring = event_cfg->hardware_event;
+		if (mhi_event->hw_ring)
+			mhi_cntrl->hw_ev_rings++;
+		else
+			mhi_cntrl->sw_ev_rings++;
+
+		mhi_event->cl_manage = event_cfg->client_managed;
+		mhi_event->offload_ev = event_cfg->offload_channel;
+		mhi_event++;
+	}
+
+	/* We need IRQ for each event ring + additional one for BHI */
+	mhi_cntrl->nr_irqs_req = mhi_cntrl->total_ev_rings + 1;
+
+	return 0;
+
+error_ev_cfg:
+
+	kfree(mhi_cntrl->mhi_event);
+	return -EINVAL;
+}
+
+static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
+			struct mhi_controller_config *config)
+{
+	int i;
+	u32 chan;
+	struct mhi_channel_config *ch_cfg;
+
+	mhi_cntrl->max_chan = config->max_channels;
+
+	/*
+	 * The allocation of MHI channels can exceed 32KB in some scenarios,
+	 * so to avoid any memory possible allocation failures, vzalloc is
+	 * used here
+	 */
+	mhi_cntrl->mhi_chan = vzalloc(mhi_cntrl->max_chan *
+				      sizeof(*mhi_cntrl->mhi_chan));
+	if (!mhi_cntrl->mhi_chan)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&mhi_cntrl->lpm_chans);
+
+	/* Populate channel configurations */
+	for (i = 0; i < config->num_channels; i++) {
+		struct mhi_chan *mhi_chan;
+
+		ch_cfg = &config->ch_cfg[i];
+
+		chan = ch_cfg->num;
+		if (chan >= mhi_cntrl->max_chan) {
+			dev_err(mhi_cntrl->dev,
+				"Channel %d not available\n", chan);
+			goto error_chan_cfg;
+		}
+
+		mhi_chan = &mhi_cntrl->mhi_chan[chan];
+		mhi_chan->name = ch_cfg->name;
+		mhi_chan->chan = chan;
+
+		mhi_chan->tre_ring.elements = ch_cfg->num_elements;
+		if (!mhi_chan->tre_ring.elements)
+			goto error_chan_cfg;
+
+		/*
+		 * For some channels, local ring length should be bigger than
+		 * the transfer ring length due to internal logical channels
+		 * in device. So host can queue much more buffers than transfer
+		 * ring length. Example, RSC channels should have a larger local
+		 * channel length than transfer ring length.
+		 */
+		mhi_chan->buf_ring.elements = ch_cfg->local_elements;
+		if (!mhi_chan->buf_ring.elements)
+			mhi_chan->buf_ring.elements = mhi_chan->tre_ring.elements;
+		mhi_chan->er_index = ch_cfg->event_ring;
+		mhi_chan->dir = ch_cfg->dir;
+
+		/*
+		 * For most channels, chtype is identical to channel directions.
+		 * So, if it is not defined then assign channel direction to
+		 * chtype
+		 */
+		mhi_chan->type = ch_cfg->type;
+		if (!mhi_chan->type)
+			mhi_chan->type = (enum mhi_ch_type)mhi_chan->dir;
+
+		mhi_chan->ee_mask = ch_cfg->ee_mask;
+
+		mhi_chan->db_cfg.pollcfg = ch_cfg->pollcfg;
+		mhi_chan->xfer_type = ch_cfg->data_type;
+
+		mhi_chan->lpm_notify = ch_cfg->lpm_notify;
+		mhi_chan->offload_ch = ch_cfg->offload_channel;
+		mhi_chan->db_cfg.reset_req = ch_cfg->doorbell_mode_switch;
+		mhi_chan->pre_alloc = ch_cfg->auto_queue;
+		mhi_chan->auto_start = ch_cfg->auto_start;
+
+		/*
+		 * If MHI host allocates buffers, then the channel direction
+		 * should be DMA_FROM_DEVICE and the buffer type should be
+		 * MHI_BUF_RAW
+		 */
+		if (mhi_chan->pre_alloc && (mhi_chan->dir != DMA_FROM_DEVICE ||
+				mhi_chan->xfer_type != MHI_BUF_RAW)) {
+			dev_err(mhi_cntrl->dev,
+				"Invalid channel configuration\n");
+			goto error_chan_cfg;
+		}
+
+		/*
+		 * Bi-directional and direction less channel must be an
+		 * offload channel
+		 */
+		if ((mhi_chan->dir == DMA_BIDIRECTIONAL ||
+		     mhi_chan->dir == DMA_NONE) && !mhi_chan->offload_ch) {
+			dev_err(mhi_cntrl->dev,
+				"Invalid channel configuration\n");
+			goto error_chan_cfg;
+		}
+
+		if (!mhi_chan->offload_ch) {
+			mhi_chan->db_cfg.brstmode = ch_cfg->doorbell;
+			if (MHI_INVALID_BRSTMODE(mhi_chan->db_cfg.brstmode)) {
+				dev_err(mhi_cntrl->dev,
+					"Invalid Door bell mode\n");
+				goto error_chan_cfg;
+			}
+		}
+
+		mhi_chan->configured = true;
+
+		if (mhi_chan->lpm_notify)
+			list_add_tail(&mhi_chan->node, &mhi_cntrl->lpm_chans);
+	}
+
+	return 0;
+
+error_chan_cfg:
+	vfree(mhi_cntrl->mhi_chan);
+
+	return -EINVAL;
+}
+
+static int parse_config(struct mhi_controller *mhi_cntrl,
+			struct mhi_controller_config *config)
+{
+	int ret;
+
+	/* Parse MHI channel configuration */
+	ret = parse_ch_cfg(mhi_cntrl, config);
+	if (ret)
+		return ret;
+
+	/* Parse MHI event configuration */
+	ret = parse_ev_cfg(mhi_cntrl, config);
+	if (ret)
+		goto error_ev_cfg;
+
+	mhi_cntrl->timeout_ms = config->timeout_ms;
+	if (!mhi_cntrl->timeout_ms)
+		mhi_cntrl->timeout_ms = MHI_TIMEOUT_MS;
+
+	mhi_cntrl->bounce_buf = config->use_bounce_buf;
+	mhi_cntrl->buffer_len = config->buf_len;
+	if (!mhi_cntrl->buffer_len)
+		mhi_cntrl->buffer_len = MHI_MAX_MTU;
+
+	return 0;
+
+error_ev_cfg:
+	vfree(mhi_cntrl->mhi_chan);
+
+	return ret;
+}
+
+int mhi_register_controller(struct mhi_controller *mhi_cntrl,
+			    struct mhi_controller_config *config)
+{
+	int ret;
+	int i;
+	struct mhi_event *mhi_event;
+	struct mhi_chan *mhi_chan;
+	struct mhi_cmd *mhi_cmd;
+	struct mhi_device *mhi_dev;
+
+	if (!mhi_cntrl)
+		return -EINVAL;
+
+	if (!mhi_cntrl->runtime_get || !mhi_cntrl->runtime_put)
+		return -EINVAL;
+
+	if (!mhi_cntrl->status_cb || !mhi_cntrl->link_status)
+		return -EINVAL;
+
+	ret = parse_config(mhi_cntrl, config);
+	if (ret)
+		return -EINVAL;
+
+	mhi_cntrl->mhi_cmd = kcalloc(NR_OF_CMD_RINGS,
+				     sizeof(*mhi_cntrl->mhi_cmd), GFP_KERNEL);
+	if (!mhi_cntrl->mhi_cmd) {
+		ret = -ENOMEM;
+		goto error_alloc_cmd;
+	}
+
+	INIT_LIST_HEAD(&mhi_cntrl->transition_list);
+	spin_lock_init(&mhi_cntrl->transition_lock);
+	spin_lock_init(&mhi_cntrl->wlock);
+	init_waitqueue_head(&mhi_cntrl->state_event);
+
+	mhi_cmd = mhi_cntrl->mhi_cmd;
+	for (i = 0; i < NR_OF_CMD_RINGS; i++, mhi_cmd++)
+		spin_lock_init(&mhi_cmd->lock);
+
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		/* Skip for offload events */
+		if (mhi_event->offload_ev)
+			continue;
+
+		mhi_event->mhi_cntrl = mhi_cntrl;
+		spin_lock_init(&mhi_event->lock);
+	}
+
+	mhi_chan = mhi_cntrl->mhi_chan;
+	for (i = 0; i < mhi_cntrl->max_chan; i++, mhi_chan++) {
+		mutex_init(&mhi_chan->mutex);
+		init_completion(&mhi_chan->completion);
+		rwlock_init(&mhi_chan->lock);
+	}
+
+	/* Register controller with MHI bus */
+	mhi_dev = mhi_alloc_device(mhi_cntrl);
+	if (IS_ERR(mhi_dev)) {
+		dev_err(mhi_cntrl->dev, "Failed to allocate device\n");
+		ret = PTR_ERR(mhi_dev);
+		goto error_alloc_dev;
+	}
+
+	mhi_dev->dev_type = MHI_DEVICE_CONTROLLER;
+	mhi_dev->mhi_cntrl = mhi_cntrl;
+	dev_set_name(&mhi_dev->dev, "%s", dev_name(mhi_cntrl->dev));
+
+	/* Init wakeup source */
+	device_init_wakeup(&mhi_dev->dev, true);
+
+	ret = device_add(&mhi_dev->dev);
+	if (ret)
+		goto error_add_dev;
+
+	mhi_cntrl->mhi_dev = mhi_dev;
+
+	return 0;
+
+error_add_dev:
+	mhi_dealloc_device(mhi_cntrl, mhi_dev);
+
+error_alloc_dev:
+	kfree(mhi_cntrl->mhi_cmd);
+
+error_alloc_cmd:
+	vfree(mhi_cntrl->mhi_chan);
+	kfree(mhi_cntrl->mhi_event);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_register_controller);
+
+void mhi_unregister_controller(struct mhi_controller *mhi_cntrl)
+{
+	struct mhi_device *mhi_dev = mhi_cntrl->mhi_dev;
+
+	kfree(mhi_cntrl->mhi_cmd);
+	kfree(mhi_cntrl->mhi_event);
+	vfree(mhi_cntrl->mhi_chan);
+
+	device_del(&mhi_dev->dev);
+	put_device(&mhi_dev->dev);
+}
+EXPORT_SYMBOL_GPL(mhi_unregister_controller);
+
+static void mhi_release_device(struct device *dev)
+{
+	struct mhi_device *mhi_dev = to_mhi_device(dev);
+
+	if (mhi_dev->ul_chan)
+		mhi_dev->ul_chan->mhi_dev = NULL;
+
+	if (mhi_dev->dl_chan)
+		mhi_dev->dl_chan->mhi_dev = NULL;
+
+	kfree(mhi_dev);
+}
+
+struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
+{
+	struct mhi_device *mhi_dev;
+	struct device *dev;
+
+	mhi_dev = kzalloc(sizeof(*mhi_dev), GFP_KERNEL);
+	if (!mhi_dev)
+		return ERR_PTR(-ENOMEM);
+
+	dev = &mhi_dev->dev;
+	device_initialize(dev);
+	dev->bus = &mhi_bus_type;
+	dev->release = mhi_release_device;
+	dev->parent = mhi_cntrl->dev;
+	mhi_dev->mhi_cntrl = mhi_cntrl;
+	mhi_dev->dev_wake = 0;
+
+	return mhi_dev;
+}
+
+static int mhi_match(struct device *dev, struct device_driver *drv)
+{
+	return 0;
+};
+
+struct bus_type mhi_bus_type = {
+	.name = "mhi",
+	.dev_name = "mhi",
+	.match = mhi_match,
+};
+
+static int __init mhi_init(void)
+{
+	return bus_register(&mhi_bus_type);
+}
+
+static void __exit mhi_exit(void)
+{
+	bus_unregister(&mhi_bus_type);
+}
+
+postcore_initcall(mhi_init);
+module_exit(mhi_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("MHI Host Interface");
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
new file mode 100644
index 000000000000..86a07be2e038
--- /dev/null
+++ b/drivers/bus/mhi/core/internal.h
@@ -0,0 +1,163 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ *
+ */
+
+#ifndef _MHI_INT_H
+#define _MHI_INT_H
+
+#include <linux/mhi.h>
+
+extern struct bus_type mhi_bus_type;
+
+/* MHI transfer completion events */
+enum mhi_ev_ccs {
+	MHI_EV_CC_INVALID = 0x0,
+	MHI_EV_CC_SUCCESS = 0x1,
+	MHI_EV_CC_EOT = 0x2, /* End of transfer event */
+	MHI_EV_CC_OVERFLOW = 0x3,
+	MHI_EV_CC_EOB = 0x4, /* End of block event */
+	MHI_EV_CC_OOB = 0x5, /* Out of block event */
+	MHI_EV_CC_DB_MODE = 0x6,
+	MHI_EV_CC_UNDEFINED_ERR = 0x10,
+	MHI_EV_CC_BAD_TRE = 0x11,
+};
+
+enum mhi_ch_state {
+	MHI_CH_STATE_DISABLED = 0x0,
+	MHI_CH_STATE_ENABLED = 0x1,
+	MHI_CH_STATE_RUNNING = 0x2,
+	MHI_CH_STATE_SUSPENDED = 0x3,
+	MHI_CH_STATE_STOP = 0x4,
+	MHI_CH_STATE_ERROR = 0x5,
+};
+
+#define MHI_INVALID_BRSTMODE(mode) (mode != MHI_DB_BRST_DISABLE && \
+				    mode != MHI_DB_BRST_ENABLE)
+
+#define NR_OF_CMD_RINGS			1
+#define CMD_EL_PER_RING			128
+#define PRIMARY_CMD_RING		0
+#define MHI_MAX_MTU			0xffff
+
+enum mhi_er_type {
+	MHI_ER_TYPE_INVALID = 0x0,
+	MHI_ER_TYPE_VALID = 0x1,
+};
+
+struct db_cfg {
+	bool reset_req;
+	bool db_mode;
+	u32 pollcfg;
+	enum mhi_db_brst_mode brstmode;
+	dma_addr_t db_val;
+	void (*process_db)(struct mhi_controller *mhi_cntrl,
+			   struct db_cfg *db_cfg, void __iomem *io_addr,
+			   dma_addr_t db_val);
+};
+
+struct mhi_ring {
+	dma_addr_t dma_handle;
+	dma_addr_t iommu_base;
+	u64 *ctxt_wp; /* point to ctxt wp */
+	void *pre_aligned;
+	void *base;
+	void *rp;
+	void *wp;
+	size_t el_size;
+	size_t len;
+	size_t elements;
+	size_t alloc_size;
+	void __iomem *db_addr;
+};
+
+struct mhi_cmd {
+	struct mhi_ring ring;
+	spinlock_t lock;
+};
+
+struct mhi_buf_info {
+	void *v_addr;
+	void *bb_addr;
+	void *wp;
+	void *cb_buf;
+	dma_addr_t p_addr;
+	size_t len;
+	enum dma_data_direction dir;
+};
+
+struct mhi_event {
+	struct mhi_controller *mhi_cntrl;
+	struct mhi_chan *mhi_chan; /* dedicated to channel */
+	u32 er_index;
+	u32 intmod;
+	u32 irq;
+	int chan; /* this event ring is dedicated to a channel (optional) */
+	u32 priority;
+	enum mhi_er_data_type data_type;
+	struct mhi_ring ring;
+	struct db_cfg db_cfg;
+	struct tasklet_struct task;
+	spinlock_t lock;
+	int (*process_event)(struct mhi_controller *mhi_cntrl,
+			     struct mhi_event *mhi_event,
+			     u32 event_quota);
+	bool hw_ring;
+	bool cl_manage;
+	bool offload_ev; /* managed by a device driver */
+};
+
+struct mhi_chan {
+	const char *name;
+	/*
+	 * Important: When consuming, increment tre_ring first and when
+	 * releasing, decrement buf_ring first. If tre_ring has space, buf_ring
+	 * is guranteed to have space so we do not need to check both rings.
+	 */
+	struct mhi_ring buf_ring;
+	struct mhi_ring tre_ring;
+	u32 chan;
+	u32 er_index;
+	u32 intmod;
+	enum mhi_ch_type type;
+	enum dma_data_direction dir;
+	struct db_cfg db_cfg;
+	enum mhi_ch_ee_mask ee_mask;
+	enum mhi_buf_type xfer_type;
+	enum mhi_ch_state ch_state;
+	enum mhi_ev_ccs ccs;
+	int (*gen_tre)(struct mhi_controller *mhi_cntrl,
+		       struct mhi_chan *mhi_chan, void *buf, void *cb,
+		       size_t len, enum mhi_flags flags);
+	int (*queue_xfer)(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+			  void *buf, size_t len, enum mhi_flags mflags);
+	struct mhi_device *mhi_dev;
+	void (*xfer_cb)(struct mhi_device *mhi_dev, struct mhi_result *result);
+	struct mutex mutex;
+	struct completion completion;
+	rwlock_t lock;
+	struct list_head node;
+	bool lpm_notify;
+	bool configured;
+	bool offload_ch;
+	bool pre_alloc;
+	bool auto_start;
+	bool wake_capable;
+};
+
+/* Default MHI timeout */
+#define MHI_TIMEOUT_MS (1000)
+
+struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl);
+static inline void mhi_dealloc_device(struct mhi_controller *mhi_cntrl,
+				      struct mhi_device *mhi_dev)
+{
+	put_device(&mhi_dev->dev);
+	kfree(mhi_dev);
+}
+
+int mhi_destroy_device(struct device *dev, void *data);
+void mhi_create_devices(struct mhi_controller *mhi_cntrl);
+
+#endif /* _MHI_INT_H */
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
new file mode 100644
index 000000000000..02330236ced8
--- /dev/null
+++ b/include/linux/mhi.h
@@ -0,0 +1,420 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ *
+ */
+#ifndef _MHI_H_
+#define _MHI_H_
+
+#include <linux/device.h>
+#include <linux/dma-direction.h>
+#include <linux/mutex.h>
+#include <linux/rwlock_types.h>
+#include <linux/slab.h>
+#include <linux/spinlock_types.h>
+#include <linux/wait.h>
+#include <linux/workqueue.h>
+
+struct mhi_chan;
+struct mhi_event;
+struct mhi_ctxt;
+struct mhi_cmd;
+struct mhi_buf_info;
+
+/**
+ * enum mhi_callback - MHI callback
+ * @MHI_CB_IDLE: MHI entered idle state
+ * @MHI_CB_PENDING_DATA: New data available for client to process
+ * @MHI_CB_LPM_ENTER: MHI host entered low power mode
+ * @MHI_CB_LPM_EXIT: MHI host about to exit low power mode
+ * @MHI_CB_EE_RDDM: MHI device entered RDDM exec env
+ * @MHI_CB_EE_MISSION_MODE: MHI device entered Mission Mode exec env
+ * @MHI_CB_SYS_ERROR: MHI device entered error state (may recover)
+ * @MHI_CB_FATAL_ERROR: MHI device entered fatal error state
+ */
+enum mhi_callback {
+	MHI_CB_IDLE,
+	MHI_CB_PENDING_DATA,
+	MHI_CB_LPM_ENTER,
+	MHI_CB_LPM_EXIT,
+	MHI_CB_EE_RDDM,
+	MHI_CB_EE_MISSION_MODE,
+	MHI_CB_SYS_ERROR,
+	MHI_CB_FATAL_ERROR,
+};
+
+/**
+ * enum mhi_flags - Transfer flags
+ * @MHI_EOB: End of buffer for bulk transfer
+ * @MHI_EOT: End of transfer
+ * @MHI_CHAIN: Linked transfer
+ */
+enum mhi_flags {
+	MHI_EOB,
+	MHI_EOT,
+	MHI_CHAIN,
+};
+
+/**
+ * enum mhi_device_type - Device types
+ * @MHI_DEVICE_XFER: Handles data transfer
+ * @MHI_DEVICE_TIMESYNC: Use for timesync feature
+ * @MHI_DEVICE_CONTROLLER: Control device
+ */
+enum mhi_device_type {
+	MHI_DEVICE_XFER,
+	MHI_DEVICE_TIMESYNC,
+	MHI_DEVICE_CONTROLLER,
+};
+
+/**
+ * enum mhi_ch_type - Channel types
+ * @MHI_CH_TYPE_INVALID: Invalid channel type
+ * @MHI_CH_TYPE_OUTBOUND: Outbound channel to the device
+ * @MHI_CH_TYPE_INBOUND: Inbound channel from the device
+ * @MHI_CH_TYPE_INBOUND_COALESCED: Coalesced channel for the device to combine
+ *				   multiple packets and send them as a single
+ *				   large packet to reduce CPU consumption
+ */
+enum mhi_ch_type {
+	MHI_CH_TYPE_INVALID = 0,
+	MHI_CH_TYPE_OUTBOUND = DMA_TO_DEVICE,
+	MHI_CH_TYPE_INBOUND = DMA_FROM_DEVICE,
+	MHI_CH_TYPE_INBOUND_COALESCED = 3,
+};
+
+/**
+ * enum mhi_ee_type - Execution environment types
+ * @MHI_EE_PBL: Primary Bootloader
+ * @MHI_EE_SBL: Secondary Bootloader
+ * @MHI_EE_AMSS: Modem, aka the primary runtime EE
+ * @MHI_EE_RDDM: Ram dump download mode
+ * @MHI_EE_WFW: WLAN firmware mode
+ * @MHI_EE_PTHRU: Passthrough
+ * @MHI_EE_EDL: Embedded downloader
+ */
+enum mhi_ee_type {
+	MHI_EE_PBL,
+	MHI_EE_SBL,
+	MHI_EE_AMSS,
+	MHI_EE_RDDM,
+	MHI_EE_WFW,
+	MHI_EE_PTHRU,
+	MHI_EE_EDL,
+	MHI_EE_MAX_SUPPORTED = MHI_EE_EDL,
+	MHI_EE_DISABLE_TRANSITION, /* local EE, not related to mhi spec */
+	MHI_EE_NOT_SUPPORTED,
+	MHI_EE_MAX,
+};
+
+/**
+ * enum mhi_buf_type - Accepted buffer type for the channel
+ * @MHI_BUF_RAW: Raw buffer
+ * @MHI_BUF_SKB: SKB struct
+ * @MHI_BUF_SCLIST: Scatter-gather list
+ * @MHI_BUF_NOP: CPU offload channel, host does not accept transfer
+ * @MHI_BUF_DMA: Receive DMA address mapped by client
+ * @MHI_BUF_RSC_DMA: Reserved Side Coalesced (RSC) type premapped buffer
+ */
+enum mhi_buf_type {
+	MHI_BUF_RAW,
+	MHI_BUF_SKB,
+	MHI_BUF_SCLIST,
+	MHI_BUF_NOP,
+	MHI_BUF_DMA,
+	MHI_BUF_RSC_DMA,
+};
+
+/**
+ * enum mhi_ch_ee_mask - Execution environment mask for channel
+ * @MHI_CH_EE_PBL: Allow channel to be used in PBL EE
+ * @MHI_CH_EE_SBL: Allow channel to be used in SBL EE
+ * @MHI_CH_EE_AMSS: Allow channel to be used in AMSS EE
+ * @MHI_CH_EE_RDDM: Allow channel to be used in RDDM EE
+ * @MHI_CH_EE_PTHRU: Allow channel to be used in PTHRU EE
+ * @MHI_CH_EE_WFW: Allow channel to be used in WFW EE
+ * @MHI_CH_EE_EDL: Allow channel to be used in EDL EE
+ */
+enum mhi_ch_ee_mask {
+	MHI_CH_EE_PBL = BIT(MHI_EE_PBL),
+	MHI_CH_EE_SBL = BIT(MHI_EE_SBL),
+	MHI_CH_EE_AMSS = BIT(MHI_EE_AMSS),
+	MHI_CH_EE_RDDM = BIT(MHI_EE_RDDM),
+	MHI_CH_EE_PTHRU = BIT(MHI_EE_PTHRU),
+	MHI_CH_EE_WFW = BIT(MHI_EE_WFW),
+	MHI_CH_EE_EDL = BIT(MHI_EE_EDL),
+};
+
+/**
+ * enum mhi_er_data_type - Event ring data types
+ * @MHI_ER_DATA: Only client data over this ring
+ * @MHI_ER_CTRL: MHI control data and client data
+ * @MHI_ER_TSYNC: Time sync events
+ */
+enum mhi_er_data_type {
+	MHI_ER_DATA,
+	MHI_ER_CTRL,
+	MHI_ER_TSYNC,
+};
+
+/**
+ * enum mhi_db_brst_mode - Doorbell mode
+ * @MHI_DB_BRST_DISABLE: Burst mode disable
+ * @MHI_DB_BRST_ENABLE: Burst mode enable
+ */
+enum mhi_db_brst_mode {
+	MHI_DB_BRST_DISABLE = 0x2,
+	MHI_DB_BRST_ENABLE = 0x3,
+};
+
+/**
+ * struct mhi_channel_config - Channel configuration structure for controller
+ * @name: The name of this channel
+ * @num: The number assigned to this channel
+ * @num_elements: The number of elements that can be queued to this channel
+ * @local_elements: The local ring length of the channel
+ * @event_ring: The event rung index that services this channel
+ * @dir: Direction that data may flow on this channel
+ * @type: Channel type
+ * @ee_mask: Execution Environment mask for this channel
+ * @pollcfg: Polling configuration for burst mode.  0 is default.  milliseconds
+	     for UL channels, multiple of 8 ring elements for DL channels
+ * @data_type: Data type accepted by this channel
+ * @doorbell: Doorbell mode
+ * @lpm_notify: The channel master requires low power mode notifications
+ * @offload_channel: The client manages the channel completely
+ * @doorbell_mode_switch: Channel switches to doorbell mode on M0 transition
+ * @auto_queue: Framework will automatically queue buffers for DL traffic
+ * @auto_start: Automatically start (open) this channel
+ */
+struct mhi_channel_config {
+	char *name;
+	u32 num;
+	u32 num_elements;
+	u32 local_elements;
+	u32 event_ring;
+	enum dma_data_direction dir;
+	enum mhi_ch_type type;
+	u32 ee_mask;
+	u32 pollcfg;
+	enum mhi_buf_type data_type;
+	enum mhi_db_brst_mode doorbell;
+	bool lpm_notify;
+	bool offload_channel;
+	bool doorbell_mode_switch;
+	bool auto_queue;
+	bool auto_start;
+};
+
+/**
+ * struct mhi_event_config - Event ring configuration structure for controller
+ * @num_elements: The number of elements that can be queued to this ring
+ * @irq_moderation_ms: Delay irq for additional events to be aggregated
+ * @irq: IRQ associated with this ring
+ * @channel: Dedicated channel number. U32_MAX indicates a non-dedicated ring
+ * @priority: Priority of this ring. Use 1 for now
+ * @mode: Doorbell mode
+ * @data_type: Type of data this ring will process
+ * @hardware_event: This ring is associated with hardware channels
+ * @client_managed: This ring is client managed
+ * @offload_channel: This ring is associated with an offloaded channel
+ */
+struct mhi_event_config {
+	u32 num_elements;
+	u32 irq_moderation_ms;
+	u32 irq;
+	u32 channel;
+	u32 priority;
+	enum mhi_db_brst_mode mode;
+	enum mhi_er_data_type data_type;
+	bool hardware_event;
+	bool client_managed;
+	bool offload_channel;
+};
+
+/**
+ * struct mhi_controller_config - Root MHI controller configuration
+ * @max_channels: Maximum number of channels supported
+ * @timeout_ms: Timeout value for operations. 0 means use default
+ * @buf_len: Size of automatically allocated buffers. 0 means use default
+ * @num_channels: Number of channels defined in @ch_cfg
+ * @ch_cfg: Array of defined channels
+ * @num_events: Number of event rings defined in @event_cfg
+ * @event_cfg: Array of defined event rings
+ * @use_bounce_buf: Use a bounce buffer pool due to limited DDR access
+ * @m2_no_db: Host is not allowed to ring DB in M2 state
+ */
+struct mhi_controller_config {
+	u32 max_channels;
+	u32 timeout_ms;
+	u32 buf_len;
+	u32 num_channels;
+	struct mhi_channel_config *ch_cfg;
+	u32 num_events;
+	struct mhi_event_config *event_cfg;
+	bool use_bounce_buf;
+	bool m2_no_db;
+};
+
+/**
+ * struct mhi_controller - Master MHI controller structure
+ * @dev: Driver model device node for the controller (required)
+ * @mhi_dev: MHI device instance for the controller
+ * @regs: Base address of MHI MMIO register space (required)
+ * @iova_start: IOMMU starting address for data (required)
+ * @iova_stop: IOMMU stop address for data (required)
+ * @fw_image: Firmware image name for normal booting (required)
+ * @edl_image: Firmware image name for emergency download mode (optional)
+ * @sbl_size: SBL image size downloaded through BHIe (optional)
+ * @seg_len: BHIe vector size (optional)
+ * @mhi_chan: Points to the channel configuration table
+ * @lpm_chans: List of channels that require LPM notifications
+ * @irq: base irq # to request (required)
+ * @max_chan: Maximum number of channels the controller supports
+ * @total_ev_rings: Total # of event rings allocated
+ * @hw_ev_rings: Number of hardware event rings
+ * @sw_ev_rings: Number of software event rings
+ * @nr_irqs_req: Number of IRQs required to operate (optional)
+ * @nr_irqs: Number of IRQ allocated by bus master (required)
+ * @mhi_event: MHI event ring configurations table
+ * @mhi_cmd: MHI command ring configurations table
+ * @mhi_ctxt: MHI device context, shared memory between host and device
+ * @pm_mutex: Mutex for suspend/resume operation
+ * @pm_lock: Lock for protecting MHI power management state
+ * @timeout_ms: Timeout in ms for state transitions
+ * @pm_state: MHI power management state
+ * @db_access: DB access states
+ * @ee: MHI device execution environment
+ * @dev_wake: Device wakeup count
+ * @pending_pkts: Pending packets for the controller
+ * @transition_list: List of MHI state transitions
+ * @transition_lock: Lock for protecting MHI state transition list
+ * @wlock: Lock for protecting device wakeup
+ * @st_worker: State transition worker
+ * @fw_worker: Firmware download worker
+ * @syserr_worker: System error worker
+ * @state_event: State change event
+ * @status_cb: CB function to notify power states of the device (required)
+ * @link_status: CB function to query link status of the device (required)
+ * @wake_get: CB function to assert device wake (optional)
+ * @wake_put: CB function to de-assert device wake (optional)
+ * @wake_toggle: CB function to assert and de-assert device wake (optional)
+ * @runtime_get: CB function to controller runtime resume (required)
+ * @runtimet_put: CB function to decrement pm usage (required)
+ * @buffer_len: Bounce buffer length
+ * @bounce_buf: Use of bounce buffer
+ * @fbc_download: MHI host needs to do complete image transfer (optional)
+ * @pre_init: MHI host needs to do pre-initialization before power up
+ * @wake_set: Device wakeup set flag
+ */
+struct mhi_controller {
+	struct device *dev;
+	struct mhi_device *mhi_dev;
+	void __iomem *regs;
+	dma_addr_t iova_start;
+	dma_addr_t iova_stop;
+	const char *fw_image;
+	const char *edl_image;
+	size_t sbl_size;
+	size_t seg_len;
+	struct mhi_chan *mhi_chan;
+	struct list_head lpm_chans;
+	int *irq;
+	u32 max_chan;
+	u32 total_ev_rings;
+	u32 hw_ev_rings;
+	u32 sw_ev_rings;
+	u32 nr_irqs_req;
+	u32 nr_irqs;
+
+	struct mhi_event *mhi_event;
+	struct mhi_cmd *mhi_cmd;
+	struct mhi_ctxt *mhi_ctxt;
+
+	struct mutex pm_mutex;
+	rwlock_t pm_lock;
+	u32 timeout_ms;
+	u32 pm_state;
+	u32 db_access;
+	enum mhi_ee_type ee;
+	atomic_t dev_wake;
+	atomic_t pending_pkts;
+	struct list_head transition_list;
+	spinlock_t transition_lock;
+	spinlock_t wlock;
+	struct work_struct st_worker;
+	struct work_struct fw_worker;
+	struct work_struct syserr_worker;
+	wait_queue_head_t state_event;
+
+	void (*status_cb)(struct mhi_controller *mhi_cntrl, enum mhi_callback cb);
+	int (*link_status)(struct mhi_controller *mhi_cntrl);
+	void (*wake_get)(struct mhi_controller *mhi_cntrl, bool override);
+	void (*wake_put)(struct mhi_controller *mhi_cntrl, bool override);
+	void (*wake_toggle)(struct mhi_controller *mhi_cntrl);
+	int (*runtime_get)(struct mhi_controller *mhi_cntrl);
+	void (*runtime_put)(struct mhi_controller *mhi_cntrl);
+
+	size_t buffer_len;
+	bool bounce_buf;
+	bool fbc_download;
+	bool pre_init;
+	bool wake_set;
+};
+
+/**
+ * struct mhi_device - Structure representing a MHI device which binds
+ *                     to channels
+ * @id: Pointer to MHI device ID struct
+ * @chan_name: Name of the channel to which the device binds
+ * @mhi_cntrl: Controller the device belongs to
+ * @ul_chan: UL channel for the device
+ * @dl_chan: DL channel for the device
+ * @dev: Driver model device node for the MHI device
+ * @dev_type: MHI device type
+ * @tiocm: Device current terminal settings
+ * @dev_wake: Device wakeup counter
+ */
+struct mhi_device {
+	const struct mhi_device_id *id;
+	const char *chan_name;
+	struct mhi_controller *mhi_cntrl;
+	struct mhi_chan *ul_chan;
+	struct mhi_chan *dl_chan;
+	struct device dev;
+	enum mhi_device_type dev_type;
+	u32 tiocm;
+	u32 dev_wake;
+};
+
+/**
+ * struct mhi_result - Completed buffer information
+ * @buf_addr: Address of data buffer
+ * @bytes_xferd: # of bytes transferred
+ * @dir: Channel direction
+ * @transaction_status: Status of last transaction
+ */
+struct mhi_result {
+	void *buf_addr;
+	size_t bytes_xferd;
+	enum dma_data_direction dir;
+	int transaction_status;
+};
+
+#define to_mhi_device(dev) container_of(dev, struct mhi_device, dev)
+
+/**
+ * mhi_register_controller - Register MHI controller
+ * @mhi_cntrl: MHI controller to register
+ * @config: Configuration to use for the controller
+ */
+int mhi_register_controller(struct mhi_controller *mhi_cntrl,
+			    struct mhi_controller_config *config);
+
+/**
+ * mhi_unregister_controller - Unregister MHI controller
+ * @mhi_cntrl: MHI controller to unregister
+ */
+void mhi_unregister_controller(struct mhi_controller *mhi_cntrl);
+
+#endif /* _MHI_H_ */
diff --git a/include/linux/mod_devicetable.h b/include/linux/mod_devicetable.h
index e3596db077dc..be15e997fe39 100644
--- a/include/linux/mod_devicetable.h
+++ b/include/linux/mod_devicetable.h
@@ -821,4 +821,16 @@ struct wmi_device_id {
 	const void *context;
 };
 
+#define MHI_NAME_SIZE 32
+
+/**
+ * struct mhi_device_id - MHI device identification
+ * @chan: MHI channel name
+ * @driver_data: driver data;
+ */
+struct mhi_device_id {
+	const char chan[MHI_NAME_SIZE];
+	kernel_ulong_t driver_data;
+};
+
 #endif /* LINUX_MOD_DEVICETABLE_H */
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 03/16] bus: mhi: core: Add support for registering MHI client drivers
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
  2020-01-31 13:49 ` [PATCH v2 01/16] docs: Add documentation for MHI bus Manivannan Sadhasivam
  2020-01-31 13:49 ` [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers Manivannan Sadhasivam
@ 2020-01-31 13:49 ` Manivannan Sadhasivam
  2020-01-31 23:00   ` Jeffrey Hugo
  2020-01-31 13:49 ` [PATCH v2 04/16] bus: mhi: core: Add support for creating and destroying MHI devices Manivannan Sadhasivam
                   ` (12 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for registering MHI client drivers with the
MHI stack. MHI client drivers binds to one or more MHI devices inorder
to sends and receive the upper-layer protocol packets like IP packets,
modem control messages, and diagnostics messages over MHI bus.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/987

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/init.c | 149 ++++++++++++++++++++++++++++++++++++
 include/linux/mhi.h         |  39 ++++++++++
 2 files changed, 188 insertions(+)

diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index a00f1547d252..999a675a5f0e 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -379,8 +379,157 @@ struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
 	return mhi_dev;
 }
 
+static int mhi_driver_probe(struct device *dev)
+{
+	struct mhi_device *mhi_dev = to_mhi_device(dev);
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct device_driver *drv = dev->driver;
+	struct mhi_driver *mhi_drv = to_mhi_driver(drv);
+	struct mhi_event *mhi_event;
+	struct mhi_chan *ul_chan = mhi_dev->ul_chan;
+	struct mhi_chan *dl_chan = mhi_dev->dl_chan;
+
+	if (ul_chan) {
+		/*
+		 * If channel supports LPM notifications then status_cb should
+		 * be provided
+		 */
+		if (ul_chan->lpm_notify && !mhi_drv->status_cb)
+			return -EINVAL;
+
+		/* For non-offload channels then xfer_cb should be provided */
+		if (!ul_chan->offload_ch && !mhi_drv->ul_xfer_cb)
+			return -EINVAL;
+
+		ul_chan->xfer_cb = mhi_drv->ul_xfer_cb;
+	}
+
+	if (dl_chan) {
+		/*
+		 * If channel supports LPM notifications then status_cb should
+		 * be provided
+		 */
+		if (dl_chan->lpm_notify && !mhi_drv->status_cb)
+			return -EINVAL;
+
+		/* For non-offload channels then xfer_cb should be provided */
+		if (!dl_chan->offload_ch && !mhi_drv->dl_xfer_cb)
+			return -EINVAL;
+
+		mhi_event = &mhi_cntrl->mhi_event[dl_chan->er_index];
+
+		/*
+		 * If the channel event ring is managed by client, then
+		 * status_cb must be provided so that the framework can
+		 * notify pending data
+		 */
+		if (mhi_event->cl_manage && !mhi_drv->status_cb)
+			return -EINVAL;
+
+		dl_chan->xfer_cb = mhi_drv->dl_xfer_cb;
+	}
+
+	/* Call the user provided probe function */
+	return mhi_drv->probe(mhi_dev, mhi_dev->id);
+}
+
+static int mhi_driver_remove(struct device *dev)
+{
+	struct mhi_device *mhi_dev = to_mhi_device(dev);
+	struct mhi_driver *mhi_drv = to_mhi_driver(dev->driver);
+	struct mhi_chan *mhi_chan;
+	enum mhi_ch_state ch_state[] = {
+		MHI_CH_STATE_DISABLED,
+		MHI_CH_STATE_DISABLED
+	};
+	int dir;
+
+	/* Skip if it is a controller device */
+	if (mhi_dev->dev_type == MHI_DEVICE_CONTROLLER)
+		return 0;
+
+	/* Reset both channels */
+	for (dir = 0; dir < 2; dir++) {
+		mhi_chan = dir ? mhi_dev->ul_chan : mhi_dev->dl_chan;
+
+		if (!mhi_chan)
+			continue;
+
+		/* Wake all threads waiting for completion */
+		write_lock_irq(&mhi_chan->lock);
+		mhi_chan->ccs = MHI_EV_CC_INVALID;
+		complete_all(&mhi_chan->completion);
+		write_unlock_irq(&mhi_chan->lock);
+
+		/* Set the channel state to disabled */
+		mutex_lock(&mhi_chan->mutex);
+		write_lock_irq(&mhi_chan->lock);
+		ch_state[dir] = mhi_chan->ch_state;
+		mhi_chan->ch_state = MHI_CH_STATE_SUSPENDED;
+		write_unlock_irq(&mhi_chan->lock);
+
+		mutex_unlock(&mhi_chan->mutex);
+	}
+
+	mhi_drv->remove(mhi_dev);
+
+	/* De-init channel if it was enabled */
+	for (dir = 0; dir < 2; dir++) {
+		mhi_chan = dir ? mhi_dev->ul_chan : mhi_dev->dl_chan;
+
+		if (!mhi_chan)
+			continue;
+
+		mutex_lock(&mhi_chan->mutex);
+
+		mhi_chan->ch_state = MHI_CH_STATE_DISABLED;
+
+		mutex_unlock(&mhi_chan->mutex);
+	}
+
+	return 0;
+}
+
+int mhi_driver_register(struct mhi_driver *mhi_drv)
+{
+	struct device_driver *driver = &mhi_drv->driver;
+
+	if (!mhi_drv->probe || !mhi_drv->remove)
+		return -EINVAL;
+
+	driver->bus = &mhi_bus_type;
+	driver->probe = mhi_driver_probe;
+	driver->remove = mhi_driver_remove;
+
+	return driver_register(driver);
+}
+EXPORT_SYMBOL_GPL(mhi_driver_register);
+
+void mhi_driver_unregister(struct mhi_driver *mhi_drv)
+{
+	driver_unregister(&mhi_drv->driver);
+}
+EXPORT_SYMBOL_GPL(mhi_driver_unregister);
+
 static int mhi_match(struct device *dev, struct device_driver *drv)
 {
+	struct mhi_device *mhi_dev = to_mhi_device(dev);
+	struct mhi_driver *mhi_drv = to_mhi_driver(drv);
+	const struct mhi_device_id *id;
+
+	/*
+	 * If the device is a controller type then there is no client driver
+	 * associated with it
+	 */
+	if (mhi_dev->dev_type == MHI_DEVICE_CONTROLLER)
+		return 0;
+
+	for (id = mhi_drv->id_table; id->chan[0]; id++)
+		if (!strcmp(mhi_dev->chan_name, id->chan)) {
+			mhi_dev->id = id;
+			return 1;
+		}
+
 	return 0;
 };
 
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 02330236ced8..137f1891701f 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -372,6 +372,8 @@ struct mhi_controller {
  * @dl_chan: DL channel for the device
  * @dev: Driver model device node for the MHI device
  * @dev_type: MHI device type
+ * @ul_chan_id: MHI channel id for UL transfer
+ * @dl_chan_id: MHI channel id for DL transfer
  * @tiocm: Device current terminal settings
  * @dev_wake: Device wakeup counter
  */
@@ -383,6 +385,8 @@ struct mhi_device {
 	struct mhi_chan *dl_chan;
 	struct device dev;
 	enum mhi_device_type dev_type;
+	int ul_chan_id;
+	int dl_chan_id;
 	u32 tiocm;
 	u32 dev_wake;
 };
@@ -401,6 +405,29 @@ struct mhi_result {
 	int transaction_status;
 };
 
+/**
+ * struct mhi_driver - Structure representing a MHI client driver
+ * @probe: CB function for client driver probe function
+ * @remove: CB function for client driver remove function
+ * @ul_xfer_cb: CB function for UL data transfer
+ * @dl_xfer_cb: CB function for DL data transfer
+ * @status_cb: CB functions for asynchronous status
+ * @driver: Device driver model driver
+ */
+struct mhi_driver {
+	const struct mhi_device_id *id_table;
+	int (*probe)(struct mhi_device *mhi_dev,
+		     const struct mhi_device_id *id);
+	void (*remove)(struct mhi_device *mhi_dev);
+	void (*ul_xfer_cb)(struct mhi_device *mhi_dev,
+			   struct mhi_result *result);
+	void (*dl_xfer_cb)(struct mhi_device *mhi_dev,
+			   struct mhi_result *result);
+	void (*status_cb)(struct mhi_device *mhi_dev, enum mhi_callback mhi_cb);
+	struct device_driver driver;
+};
+
+#define to_mhi_driver(drv) container_of(drv, struct mhi_driver, driver)
 #define to_mhi_device(dev) container_of(dev, struct mhi_device, dev)
 
 /**
@@ -417,4 +444,16 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
  */
 void mhi_unregister_controller(struct mhi_controller *mhi_cntrl);
 
+/**
+ * mhi_driver_register - Register driver with MHI framework
+ * @mhi_drv: Driver associated with the device
+ */
+int mhi_driver_register(struct mhi_driver *mhi_drv);
+
+/**
+ * mhi_driver_unregister - Unregister a driver for mhi_devices
+ * @mhi_drv: Driver associated with the device
+ */
+void mhi_driver_unregister(struct mhi_driver *mhi_drv);
+
 #endif /* _MHI_H_ */
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 04/16] bus: mhi: core: Add support for creating and destroying MHI devices
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (2 preceding siblings ...)
  2020-01-31 13:49 ` [PATCH v2 03/16] bus: mhi: core: Add support for registering MHI client drivers Manivannan Sadhasivam
@ 2020-01-31 13:49 ` Manivannan Sadhasivam
  2020-02-06 16:16   ` Jeffrey Hugo
  2020-01-31 13:49 ` [PATCH v2 05/16] bus: mhi: core: Add support for ringing channel/event ring doorbells Manivannan Sadhasivam
                   ` (11 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for creating and destroying MHI devices. The
MHI devices binds to the MHI channels and are used to transfer data
between MHI host and client device.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/989

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted from pm patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/Makefile |   2 +-
 drivers/bus/mhi/core/main.c   | 123 ++++++++++++++++++++++++++++++++++
 include/linux/mhi.h           |   2 +
 3 files changed, 126 insertions(+), 1 deletion(-)
 create mode 100644 drivers/bus/mhi/core/main.c

diff --git a/drivers/bus/mhi/core/Makefile b/drivers/bus/mhi/core/Makefile
index 2db32697c67f..77f7730da4bf 100644
--- a/drivers/bus/mhi/core/Makefile
+++ b/drivers/bus/mhi/core/Makefile
@@ -1,3 +1,3 @@
 obj-$(CONFIG_MHI_BUS) := mhi.o
 
-mhi-y := init.o
+mhi-y := init.o main.o
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
new file mode 100644
index 000000000000..216fd8691140
--- /dev/null
+++ b/drivers/bus/mhi/core/main.c
@@ -0,0 +1,123 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ *
+ */
+
+#define dev_fmt(fmt) "MHI: " fmt
+
+#include <linux/device.h>
+#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/mhi.h>
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/slab.h>
+#include "internal.h"
+
+int mhi_destroy_device(struct device *dev, void *data)
+{
+	struct mhi_device *mhi_dev;
+	struct mhi_controller *mhi_cntrl;
+
+	if (dev->bus != &mhi_bus_type)
+		return 0;
+
+	mhi_dev = to_mhi_device(dev);
+	mhi_cntrl = mhi_dev->mhi_cntrl;
+
+	/* Only destroy virtual devices thats attached to bus */
+	if (mhi_dev->dev_type ==  MHI_DEVICE_CONTROLLER)
+		return 0;
+
+	dev_dbg(mhi_cntrl->dev, "destroy device for chan:%s\n",
+		 mhi_dev->chan_name);
+
+	/* Notify the client and remove the device from MHI bus */
+	device_del(dev);
+	put_device(dev);
+
+	return 0;
+}
+
+static void mhi_notify(struct mhi_device *mhi_dev, enum mhi_callback cb_reason)
+{
+	struct mhi_driver *mhi_drv;
+
+	if (!mhi_dev->dev.driver)
+		return;
+
+	mhi_drv = to_mhi_driver(mhi_dev->dev.driver);
+
+	if (mhi_drv->status_cb)
+		mhi_drv->status_cb(mhi_dev, cb_reason);
+}
+
+/* Bind MHI channels to MHI devices */
+void mhi_create_devices(struct mhi_controller *mhi_cntrl)
+{
+	int i;
+	struct mhi_chan *mhi_chan;
+	struct mhi_device *mhi_dev;
+	int ret;
+
+	mhi_chan = mhi_cntrl->mhi_chan;
+	for (i = 0; i < mhi_cntrl->max_chan; i++, mhi_chan++) {
+		if (!mhi_chan->configured || mhi_chan->mhi_dev ||
+		    !(mhi_chan->ee_mask & BIT(mhi_cntrl->ee)))
+			continue;
+		mhi_dev = mhi_alloc_device(mhi_cntrl);
+		if (!mhi_dev)
+			return;
+
+		mhi_dev->dev_type = MHI_DEVICE_XFER;
+		switch (mhi_chan->dir) {
+		case DMA_TO_DEVICE:
+			mhi_dev->ul_chan = mhi_chan;
+			mhi_dev->ul_chan_id = mhi_chan->chan;
+			break;
+		case DMA_FROM_DEVICE:
+			/* We use dl_chan as offload channels */
+			mhi_dev->dl_chan = mhi_chan;
+			mhi_dev->dl_chan_id = mhi_chan->chan;
+			break;
+		default:
+			dev_err(mhi_cntrl->dev, "Direction not supported\n");
+			mhi_dealloc_device(mhi_cntrl, mhi_dev);
+			return;
+		}
+
+		mhi_chan->mhi_dev = mhi_dev;
+
+		/* Check next channel if it matches */
+		if ((i + 1) < mhi_cntrl->max_chan && mhi_chan[1].configured) {
+			if (!strcmp(mhi_chan[1].name, mhi_chan->name)) {
+				i++;
+				mhi_chan++;
+				if (mhi_chan->dir == DMA_TO_DEVICE) {
+					mhi_dev->ul_chan = mhi_chan;
+					mhi_dev->ul_chan_id = mhi_chan->chan;
+				} else {
+					mhi_dev->dl_chan = mhi_chan;
+					mhi_dev->dl_chan_id = mhi_chan->chan;
+				}
+				mhi_chan->mhi_dev = mhi_dev;
+			}
+		}
+
+		/* Channel name is same for both UL and DL */
+		mhi_dev->chan_name = mhi_chan->name;
+		dev_set_name(&mhi_dev->dev, "%04x_%s", mhi_chan->chan,
+			     mhi_dev->chan_name);
+
+		/* Init wakeup source if available */
+		if (mhi_dev->dl_chan && mhi_dev->dl_chan->wake_capable)
+			device_init_wakeup(&mhi_dev->dev, true);
+
+		ret = device_add(&mhi_dev->dev);
+		if (ret)
+			mhi_dealloc_device(mhi_cntrl, mhi_dev);
+	}
+}
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 137f1891701f..5cbc1e33829f 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -186,6 +186,7 @@ enum mhi_db_brst_mode {
  * @doorbell_mode_switch: Channel switches to doorbell mode on M0 transition
  * @auto_queue: Framework will automatically queue buffers for DL traffic
  * @auto_start: Automatically start (open) this channel
+ * @wake-capable: Channel capable of waking up the system
  */
 struct mhi_channel_config {
 	char *name;
@@ -204,6 +205,7 @@ struct mhi_channel_config {
 	bool doorbell_mode_switch;
 	bool auto_queue;
 	bool auto_start;
+	bool wake_capable;
 };
 
 /**
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 05/16] bus: mhi: core: Add support for ringing channel/event ring doorbells
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (3 preceding siblings ...)
  2020-01-31 13:49 ` [PATCH v2 04/16] bus: mhi: core: Add support for creating and destroying MHI devices Manivannan Sadhasivam
@ 2020-01-31 13:49 ` Manivannan Sadhasivam
  2020-02-06 20:14   ` Jeffrey Hugo
  2020-01-31 13:49 ` [PATCH v2 06/16] bus: mhi: core: Add support for PM state transitions Manivannan Sadhasivam
                   ` (10 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for ringing channel and event ring doorbells
by MHI host. The MHI host can use the channel and event ring doorbells
for notifying the client device about processing transfer and event
rings which it has queued using MMIO registers.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/989

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted from pm patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/init.c     | 140 ++++++++++++++++
 drivers/bus/mhi/core/internal.h | 282 ++++++++++++++++++++++++++++++++
 drivers/bus/mhi/core/main.c     | 118 +++++++++++++
 include/linux/mhi.h             |   4 +
 4 files changed, 544 insertions(+)

diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index 999a675a5f0e..dec996fa1d1d 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -19,6 +19,136 @@
 #include <linux/wait.h>
 #include "internal.h"
 
+int mhi_init_mmio(struct mhi_controller *mhi_cntrl)
+{
+	u32 val;
+	int i, ret;
+	struct mhi_chan *mhi_chan;
+	struct mhi_event *mhi_event;
+	void __iomem *base = mhi_cntrl->regs;
+	struct {
+		u32 offset;
+		u32 mask;
+		u32 shift;
+		u32 val;
+	} reg_info[] = {
+		{
+			CCABAP_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->mhi_ctxt->chan_ctxt_addr),
+		},
+		{
+			CCABAP_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->mhi_ctxt->chan_ctxt_addr),
+		},
+		{
+			ECABAP_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->mhi_ctxt->er_ctxt_addr),
+		},
+		{
+			ECABAP_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->mhi_ctxt->er_ctxt_addr),
+		},
+		{
+			CRCBAP_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->mhi_ctxt->cmd_ctxt_addr),
+		},
+		{
+			CRCBAP_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->mhi_ctxt->cmd_ctxt_addr),
+		},
+		{
+			MHICFG, MHICFG_NER_MASK, MHICFG_NER_SHIFT,
+			mhi_cntrl->total_ev_rings,
+		},
+		{
+			MHICFG, MHICFG_NHWER_MASK, MHICFG_NHWER_SHIFT,
+			mhi_cntrl->hw_ev_rings,
+		},
+		{
+			MHICTRLBASE_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->iova_start),
+		},
+		{
+			MHICTRLBASE_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->iova_start),
+		},
+		{
+			MHIDATABASE_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->iova_start),
+		},
+		{
+			MHIDATABASE_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->iova_start),
+		},
+		{
+			MHICTRLLIMIT_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->iova_stop),
+		},
+		{
+			MHICTRLLIMIT_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->iova_stop),
+		},
+		{
+			MHIDATALIMIT_HIGHER, U32_MAX, 0,
+			upper_32_bits(mhi_cntrl->iova_stop),
+		},
+		{
+			MHIDATALIMIT_LOWER, U32_MAX, 0,
+			lower_32_bits(mhi_cntrl->iova_stop),
+		},
+		{ 0, 0, 0 }
+	};
+
+	dev_dbg(mhi_cntrl->dev, "Initializing MHI registers\n");
+
+	/* Read channel db offset */
+	ret = mhi_read_reg_field(mhi_cntrl, base, CHDBOFF, CHDBOFF_CHDBOFF_MASK,
+				 CHDBOFF_CHDBOFF_SHIFT, &val);
+	if (ret) {
+		dev_err(mhi_cntrl->dev, "Unable to read CHDBOFF register\n");
+		return -EIO;
+	}
+
+	/* Setup wake db */
+	mhi_cntrl->wake_db = base + val + (8 * MHI_DEV_WAKE_DB);
+	mhi_write_reg(mhi_cntrl, mhi_cntrl->wake_db, 4, 0);
+	mhi_write_reg(mhi_cntrl, mhi_cntrl->wake_db, 0, 0);
+	mhi_cntrl->wake_set = false;
+
+	/* Setup channel db address for each channel in tre_ring */
+	mhi_chan = mhi_cntrl->mhi_chan;
+	for (i = 0; i < mhi_cntrl->max_chan; i++, val += 8, mhi_chan++)
+		mhi_chan->tre_ring.db_addr = base + val;
+
+	/* Read event ring db offset */
+	ret = mhi_read_reg_field(mhi_cntrl, base, ERDBOFF, ERDBOFF_ERDBOFF_MASK,
+				 ERDBOFF_ERDBOFF_SHIFT, &val);
+	if (ret) {
+		dev_err(mhi_cntrl->dev, "Unable to read ERDBOFF register\n");
+		return -EIO;
+	}
+
+	/* Setup event db address for each ev_ring */
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, val += 8, mhi_event++) {
+		if (mhi_event->offload_ev)
+			continue;
+
+		mhi_event->ring.db_addr = base + val;
+	}
+
+	/* Setup DB register for primary CMD rings */
+	mhi_cntrl->mhi_cmd[PRIMARY_CMD_RING].ring.db_addr = base + CRDB_LOWER;
+
+	/* Write to MMIO registers */
+	for (i = 0; reg_info[i].offset; i++)
+		mhi_write_reg_field(mhi_cntrl, base, reg_info[i].offset,
+				    reg_info[i].mask, reg_info[i].shift,
+				    reg_info[i].val);
+
+	return 0;
+}
+
 static int parse_ev_cfg(struct mhi_controller *mhi_cntrl,
 			struct mhi_controller_config *config)
 {
@@ -63,6 +193,11 @@ static int parse_ev_cfg(struct mhi_controller *mhi_cntrl,
 		if (MHI_INVALID_BRSTMODE(mhi_event->db_cfg.brstmode))
 			goto error_ev_cfg;
 
+		if (mhi_event->db_cfg.brstmode == MHI_DB_BRST_ENABLE)
+			mhi_event->db_cfg.process_db = mhi_db_brstmode;
+		else
+			mhi_event->db_cfg.process_db = mhi_db_brstmode_disable;
+
 		mhi_event->data_type = event_cfg->data_type;
 
 		mhi_event->hw_ring = event_cfg->hardware_event;
@@ -194,6 +329,11 @@ static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
 			}
 		}
 
+		if (mhi_chan->db_cfg.brstmode == MHI_DB_BRST_ENABLE)
+			mhi_chan->db_cfg.process_db = mhi_db_brstmode;
+		else
+			mhi_chan->db_cfg.process_db = mhi_db_brstmode_disable;
+
 		mhi_chan->configured = true;
 
 		if (mhi_chan->lpm_notify)
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index 86a07be2e038..960dc5cf9343 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -11,6 +11,262 @@
 
 extern struct bus_type mhi_bus_type;
 
+/* MHI MMIO register mapping */
+#define PCI_INVALID_READ(val) (val == U32_MAX)
+
+#define MHIREGLEN (0x0)
+#define MHIREGLEN_MHIREGLEN_MASK (0xFFFFFFFF)
+#define MHIREGLEN_MHIREGLEN_SHIFT (0)
+
+#define MHIVER (0x8)
+#define MHIVER_MHIVER_MASK (0xFFFFFFFF)
+#define MHIVER_MHIVER_SHIFT (0)
+
+#define MHICFG (0x10)
+#define MHICFG_NHWER_MASK (0xFF000000)
+#define MHICFG_NHWER_SHIFT (24)
+#define MHICFG_NER_MASK (0xFF0000)
+#define MHICFG_NER_SHIFT (16)
+#define MHICFG_NHWCH_MASK (0xFF00)
+#define MHICFG_NHWCH_SHIFT (8)
+#define MHICFG_NCH_MASK (0xFF)
+#define MHICFG_NCH_SHIFT (0)
+
+#define CHDBOFF (0x18)
+#define CHDBOFF_CHDBOFF_MASK (0xFFFFFFFF)
+#define CHDBOFF_CHDBOFF_SHIFT (0)
+
+#define ERDBOFF (0x20)
+#define ERDBOFF_ERDBOFF_MASK (0xFFFFFFFF)
+#define ERDBOFF_ERDBOFF_SHIFT (0)
+
+#define BHIOFF (0x28)
+#define BHIOFF_BHIOFF_MASK (0xFFFFFFFF)
+#define BHIOFF_BHIOFF_SHIFT (0)
+
+#define BHIEOFF (0x2C)
+#define BHIEOFF_BHIEOFF_MASK (0xFFFFFFFF)
+#define BHIEOFF_BHIEOFF_SHIFT (0)
+
+#define DEBUGOFF (0x30)
+#define DEBUGOFF_DEBUGOFF_MASK (0xFFFFFFFF)
+#define DEBUGOFF_DEBUGOFF_SHIFT (0)
+
+#define MHICTRL (0x38)
+#define MHICTRL_MHISTATE_MASK (0x0000FF00)
+#define MHICTRL_MHISTATE_SHIFT (8)
+#define MHICTRL_RESET_MASK (0x2)
+#define MHICTRL_RESET_SHIFT (1)
+
+#define MHISTATUS (0x48)
+#define MHISTATUS_MHISTATE_MASK (0x0000FF00)
+#define MHISTATUS_MHISTATE_SHIFT (8)
+#define MHISTATUS_SYSERR_MASK (0x4)
+#define MHISTATUS_SYSERR_SHIFT (2)
+#define MHISTATUS_READY_MASK (0x1)
+#define MHISTATUS_READY_SHIFT (0)
+
+#define CCABAP_LOWER (0x58)
+#define CCABAP_LOWER_CCABAP_LOWER_MASK (0xFFFFFFFF)
+#define CCABAP_LOWER_CCABAP_LOWER_SHIFT (0)
+
+#define CCABAP_HIGHER (0x5C)
+#define CCABAP_HIGHER_CCABAP_HIGHER_MASK (0xFFFFFFFF)
+#define CCABAP_HIGHER_CCABAP_HIGHER_SHIFT (0)
+
+#define ECABAP_LOWER (0x60)
+#define ECABAP_LOWER_ECABAP_LOWER_MASK (0xFFFFFFFF)
+#define ECABAP_LOWER_ECABAP_LOWER_SHIFT (0)
+
+#define ECABAP_HIGHER (0x64)
+#define ECABAP_HIGHER_ECABAP_HIGHER_MASK (0xFFFFFFFF)
+#define ECABAP_HIGHER_ECABAP_HIGHER_SHIFT (0)
+
+#define CRCBAP_LOWER (0x68)
+#define CRCBAP_LOWER_CRCBAP_LOWER_MASK (0xFFFFFFFF)
+#define CRCBAP_LOWER_CRCBAP_LOWER_SHIFT (0)
+
+#define CRCBAP_HIGHER (0x6C)
+#define CRCBAP_HIGHER_CRCBAP_HIGHER_MASK (0xFFFFFFFF)
+#define CRCBAP_HIGHER_CRCBAP_HIGHER_SHIFT (0)
+
+#define CRDB_LOWER (0x70)
+#define CRDB_LOWER_CRDB_LOWER_MASK (0xFFFFFFFF)
+#define CRDB_LOWER_CRDB_LOWER_SHIFT (0)
+
+#define CRDB_HIGHER (0x74)
+#define CRDB_HIGHER_CRDB_HIGHER_MASK (0xFFFFFFFF)
+#define CRDB_HIGHER_CRDB_HIGHER_SHIFT (0)
+
+#define MHICTRLBASE_LOWER (0x80)
+#define MHICTRLBASE_LOWER_MHICTRLBASE_LOWER_MASK (0xFFFFFFFF)
+#define MHICTRLBASE_LOWER_MHICTRLBASE_LOWER_SHIFT (0)
+
+#define MHICTRLBASE_HIGHER (0x84)
+#define MHICTRLBASE_HIGHER_MHICTRLBASE_HIGHER_MASK (0xFFFFFFFF)
+#define MHICTRLBASE_HIGHER_MHICTRLBASE_HIGHER_SHIFT (0)
+
+#define MHICTRLLIMIT_LOWER (0x88)
+#define MHICTRLLIMIT_LOWER_MHICTRLLIMIT_LOWER_MASK (0xFFFFFFFF)
+#define MHICTRLLIMIT_LOWER_MHICTRLLIMIT_LOWER_SHIFT (0)
+
+#define MHICTRLLIMIT_HIGHER (0x8C)
+#define MHICTRLLIMIT_HIGHER_MHICTRLLIMIT_HIGHER_MASK (0xFFFFFFFF)
+#define MHICTRLLIMIT_HIGHER_MHICTRLLIMIT_HIGHER_SHIFT (0)
+
+#define MHIDATABASE_LOWER (0x98)
+#define MHIDATABASE_LOWER_MHIDATABASE_LOWER_MASK (0xFFFFFFFF)
+#define MHIDATABASE_LOWER_MHIDATABASE_LOWER_SHIFT (0)
+
+#define MHIDATABASE_HIGHER (0x9C)
+#define MHIDATABASE_HIGHER_MHIDATABASE_HIGHER_MASK (0xFFFFFFFF)
+#define MHIDATABASE_HIGHER_MHIDATABASE_HIGHER_SHIFT (0)
+
+#define MHIDATALIMIT_LOWER (0xA0)
+#define MHIDATALIMIT_LOWER_MHIDATALIMIT_LOWER_MASK (0xFFFFFFFF)
+#define MHIDATALIMIT_LOWER_MHIDATALIMIT_LOWER_SHIFT (0)
+
+#define MHIDATALIMIT_HIGHER (0xA4)
+#define MHIDATALIMIT_HIGHER_MHIDATALIMIT_HIGHER_MASK (0xFFFFFFFF)
+#define MHIDATALIMIT_HIGHER_MHIDATALIMIT_HIGHER_SHIFT (0)
+
+/* Host request register */
+#define MHI_SOC_RESET_REQ_OFFSET (0xB0)
+#define MHI_SOC_RESET_REQ BIT(0)
+
+/* MHI BHI offfsets */
+#define BHI_BHIVERSION_MINOR (0x00)
+#define BHI_BHIVERSION_MAJOR (0x04)
+#define BHI_IMGADDR_LOW (0x08)
+#define BHI_IMGADDR_HIGH (0x0C)
+#define BHI_IMGSIZE (0x10)
+#define BHI_RSVD1 (0x14)
+#define BHI_IMGTXDB (0x18)
+#define BHI_TXDB_SEQNUM_BMSK (0x3FFFFFFF)
+#define BHI_TXDB_SEQNUM_SHFT (0)
+#define BHI_RSVD2 (0x1C)
+#define BHI_INTVEC (0x20)
+#define BHI_RSVD3 (0x24)
+#define BHI_EXECENV (0x28)
+#define BHI_STATUS (0x2C)
+#define BHI_ERRCODE (0x30)
+#define BHI_ERRDBG1 (0x34)
+#define BHI_ERRDBG2 (0x38)
+#define BHI_ERRDBG3 (0x3C)
+#define BHI_SERIALNU (0x40)
+#define BHI_SBLANTIROLLVER (0x44)
+#define BHI_NUMSEG (0x48)
+#define BHI_MSMHWID(n) (0x4C + (0x4 * n))
+#define BHI_OEMPKHASH(n) (0x64 + (0x4 * n))
+#define BHI_RSVD5 (0xC4)
+#define BHI_STATUS_MASK (0xC0000000)
+#define BHI_STATUS_SHIFT (30)
+#define BHI_STATUS_ERROR (3)
+#define BHI_STATUS_SUCCESS (2)
+#define BHI_STATUS_RESET (0)
+
+/* MHI BHIE offsets */
+#define BHIE_MSMSOCID_OFFS (0x0000)
+#define BHIE_TXVECADDR_LOW_OFFS (0x002C)
+#define BHIE_TXVECADDR_HIGH_OFFS (0x0030)
+#define BHIE_TXVECSIZE_OFFS (0x0034)
+#define BHIE_TXVECDB_OFFS (0x003C)
+#define BHIE_TXVECDB_SEQNUM_BMSK (0x3FFFFFFF)
+#define BHIE_TXVECDB_SEQNUM_SHFT (0)
+#define BHIE_TXVECSTATUS_OFFS (0x0044)
+#define BHIE_TXVECSTATUS_SEQNUM_BMSK (0x3FFFFFFF)
+#define BHIE_TXVECSTATUS_SEQNUM_SHFT (0)
+#define BHIE_TXVECSTATUS_STATUS_BMSK (0xC0000000)
+#define BHIE_TXVECSTATUS_STATUS_SHFT (30)
+#define BHIE_TXVECSTATUS_STATUS_RESET (0x00)
+#define BHIE_TXVECSTATUS_STATUS_XFER_COMPL (0x02)
+#define BHIE_TXVECSTATUS_STATUS_ERROR (0x03)
+#define BHIE_RXVECADDR_LOW_OFFS (0x0060)
+#define BHIE_RXVECADDR_HIGH_OFFS (0x0064)
+#define BHIE_RXVECSIZE_OFFS (0x0068)
+#define BHIE_RXVECDB_OFFS (0x0070)
+#define BHIE_RXVECDB_SEQNUM_BMSK (0x3FFFFFFF)
+#define BHIE_RXVECDB_SEQNUM_SHFT (0)
+#define BHIE_RXVECSTATUS_OFFS (0x0078)
+#define BHIE_RXVECSTATUS_SEQNUM_BMSK (0x3FFFFFFF)
+#define BHIE_RXVECSTATUS_SEQNUM_SHFT (0)
+#define BHIE_RXVECSTATUS_STATUS_BMSK (0xC0000000)
+#define BHIE_RXVECSTATUS_STATUS_SHFT (30)
+#define BHIE_RXVECSTATUS_STATUS_RESET (0x00)
+#define BHIE_RXVECSTATUS_STATUS_XFER_COMPL (0x02)
+#define BHIE_RXVECSTATUS_STATUS_ERROR (0x03)
+
+#define EV_CTX_RESERVED_MASK GENMASK(7, 0)
+#define EV_CTX_INTMODC_MASK GENMASK(15, 8)
+#define EV_CTX_INTMODC_SHIFT 8
+#define EV_CTX_INTMODT_MASK GENMASK(31, 16)
+#define EV_CTX_INTMODT_SHIFT 16
+struct mhi_event_ctxt {
+	__u32 intmod;
+	__u32 ertype;
+	__u32 msivec;
+
+	__u64 rbase __packed __aligned(4);
+	__u64 rlen __packed __aligned(4);
+	__u64 rp __packed __aligned(4);
+	__u64 wp __packed __aligned(4);
+};
+
+#define CHAN_CTX_CHSTATE_MASK GENMASK(7, 0)
+#define CHAN_CTX_CHSTATE_SHIFT 0
+#define CHAN_CTX_BRSTMODE_MASK GENMASK(9, 8)
+#define CHAN_CTX_BRSTMODE_SHIFT 8
+#define CHAN_CTX_POLLCFG_MASK GENMASK(15, 10)
+#define CHAN_CTX_POLLCFG_SHIFT 10
+#define CHAN_CTX_RESERVED_MASK GENMASK(31, 16)
+struct mhi_chan_ctxt {
+	__u32 chcfg;
+	__u32 chtype;
+	__u32 erindex;
+
+	__u64 rbase __packed __aligned(4);
+	__u64 rlen __packed __aligned(4);
+	__u64 rp __packed __aligned(4);
+	__u64 wp __packed __aligned(4);
+};
+
+struct mhi_cmd_ctxt {
+	__u32 reserved0;
+	__u32 reserved1;
+	__u32 reserved2;
+
+	__u64 rbase __packed __aligned(4);
+	__u64 rlen __packed __aligned(4);
+	__u64 rp __packed __aligned(4);
+	__u64 wp __packed __aligned(4);
+};
+
+struct mhi_ctxt {
+	struct mhi_event_ctxt *er_ctxt;
+	struct mhi_chan_ctxt *chan_ctxt;
+	struct mhi_cmd_ctxt *cmd_ctxt;
+	dma_addr_t er_ctxt_addr;
+	dma_addr_t chan_ctxt_addr;
+	dma_addr_t cmd_ctxt_addr;
+};
+
+struct mhi_tre {
+	u64 ptr;
+	u32 dword[2];
+};
+
+struct bhi_vec_entry {
+	u64 dma_addr;
+	u64 size;
+};
+
+enum mhi_cmd_type {
+	MHI_CMD_NOP = 1,
+	MHI_CMD_RESET_CHAN = 16,
+	MHI_CMD_STOP_CHAN = 17,
+	MHI_CMD_START_CHAN = 18,
+};
+
 /* MHI transfer completion events */
 enum mhi_ev_ccs {
 	MHI_EV_CC_INVALID = 0x0,
@@ -39,6 +295,7 @@ enum mhi_ch_state {
 #define NR_OF_CMD_RINGS			1
 #define CMD_EL_PER_RING			128
 #define PRIMARY_CMD_RING		0
+#define MHI_DEV_WAKE_DB			127
 #define MHI_MAX_MTU			0xffff
 
 enum mhi_er_type {
@@ -160,4 +417,29 @@ static inline void mhi_dealloc_device(struct mhi_controller *mhi_cntrl,
 int mhi_destroy_device(struct device *dev, void *data);
 void mhi_create_devices(struct mhi_controller *mhi_cntrl);
 
+/* Register access methods */
+void mhi_db_brstmode(struct mhi_controller *mhi_cntrl, struct db_cfg *db_cfg,
+		     void __iomem *db_addr, dma_addr_t db_val);
+void mhi_db_brstmode_disable(struct mhi_controller *mhi_cntrl,
+			     struct db_cfg *db_mode, void __iomem *db_addr,
+			     dma_addr_t db_val);
+int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl,
+			      void __iomem *base, u32 offset, u32 *out);
+int __must_check mhi_read_reg_field(struct mhi_controller *mhi_cntrl,
+				    void __iomem *base, u32 offset, u32 mask,
+				    u32 shift, u32 *out);
+void mhi_write_reg(struct mhi_controller *mhi_cntrl, void __iomem *base,
+		   u32 offset, u32 val);
+void mhi_write_reg_field(struct mhi_controller *mhi_cntrl, void __iomem *base,
+			 u32 offset, u32 mask, u32 shift, u32 val);
+void mhi_ring_er_db(struct mhi_event *mhi_event);
+void mhi_write_db(struct mhi_controller *mhi_cntrl, void __iomem *db_addr,
+		  dma_addr_t db_val);
+void mhi_ring_cmd_db(struct mhi_controller *mhi_cntrl, struct mhi_cmd *mhi_cmd);
+void mhi_ring_chan_db(struct mhi_controller *mhi_cntrl,
+		      struct mhi_chan *mhi_chan);
+
+/* Initialization methods */
+int mhi_init_mmio(struct mhi_controller *mhi_cntrl);
+
 #endif /* _MHI_INT_H */
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
index 216fd8691140..59d29470bf73 100644
--- a/drivers/bus/mhi/core/main.c
+++ b/drivers/bus/mhi/core/main.c
@@ -17,6 +17,124 @@
 #include <linux/slab.h>
 #include "internal.h"
 
+int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl,
+			      void __iomem *base, u32 offset, u32 *out)
+{
+	u32 tmp = readl(base + offset);
+
+	/* If there is any unexpected value, query the link status */
+	if (PCI_INVALID_READ(tmp) &&
+	    mhi_cntrl->link_status(mhi_cntrl))
+		return -EIO;
+
+	*out = tmp;
+
+	return 0;
+}
+
+int __must_check mhi_read_reg_field(struct mhi_controller *mhi_cntrl,
+				    void __iomem *base, u32 offset,
+				    u32 mask, u32 shift, u32 *out)
+{
+	u32 tmp;
+	int ret;
+
+	ret = mhi_read_reg(mhi_cntrl, base, offset, &tmp);
+	if (ret)
+		return ret;
+
+	*out = (tmp & mask) >> shift;
+
+	return 0;
+}
+
+void mhi_write_reg(struct mhi_controller *mhi_cntrl, void __iomem *base,
+		   u32 offset, u32 val)
+{
+	writel(val, base + offset);
+}
+
+void mhi_write_reg_field(struct mhi_controller *mhi_cntrl, void __iomem *base,
+			 u32 offset, u32 mask, u32 shift, u32 val)
+{
+	int ret;
+	u32 tmp;
+
+	ret = mhi_read_reg(mhi_cntrl, base, offset, &tmp);
+	if (ret)
+		return;
+
+	tmp &= ~mask;
+	tmp |= (val << shift);
+	mhi_write_reg(mhi_cntrl, base, offset, tmp);
+}
+
+void mhi_write_db(struct mhi_controller *mhi_cntrl, void __iomem *db_addr,
+		  dma_addr_t db_val)
+{
+	mhi_write_reg(mhi_cntrl, db_addr, 4, upper_32_bits(db_val));
+	mhi_write_reg(mhi_cntrl, db_addr, 0, lower_32_bits(db_val));
+}
+
+void mhi_db_brstmode(struct mhi_controller *mhi_cntrl,
+		     struct db_cfg *db_cfg,
+		     void __iomem *db_addr,
+		     dma_addr_t db_val)
+{
+	if (db_cfg->db_mode) {
+		db_cfg->db_val = db_val;
+		mhi_write_db(mhi_cntrl, db_addr, db_val);
+		db_cfg->db_mode = 0;
+	}
+}
+
+void mhi_db_brstmode_disable(struct mhi_controller *mhi_cntrl,
+			     struct db_cfg *db_cfg,
+			     void __iomem *db_addr,
+			     dma_addr_t db_val)
+{
+	db_cfg->db_val = db_val;
+	mhi_write_db(mhi_cntrl, db_addr, db_val);
+}
+
+void mhi_ring_er_db(struct mhi_event *mhi_event)
+{
+	struct mhi_ring *ring = &mhi_event->ring;
+
+	mhi_event->db_cfg.process_db(mhi_event->mhi_cntrl, &mhi_event->db_cfg,
+				     ring->db_addr, *ring->ctxt_wp);
+}
+
+void mhi_ring_cmd_db(struct mhi_controller *mhi_cntrl, struct mhi_cmd *mhi_cmd)
+{
+	dma_addr_t db;
+	struct mhi_ring *ring = &mhi_cmd->ring;
+
+	db = ring->iommu_base + (ring->wp - ring->base);
+	*ring->ctxt_wp = db;
+	mhi_write_db(mhi_cntrl, ring->db_addr, db);
+}
+
+void mhi_ring_chan_db(struct mhi_controller *mhi_cntrl,
+		      struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *ring = &mhi_chan->tre_ring;
+	dma_addr_t db;
+
+	db = ring->iommu_base + (ring->wp - ring->base);
+	*ring->ctxt_wp = db;
+	mhi_chan->db_cfg.process_db(mhi_cntrl, &mhi_chan->db_cfg,
+				    ring->db_addr, db);
+}
+
+enum mhi_ee_type mhi_get_exec_env(struct mhi_controller *mhi_cntrl)
+{
+	u32 exec;
+	int ret = mhi_read_reg(mhi_cntrl, mhi_cntrl->bhi, BHI_EXECENV, &exec);
+
+	return (ret) ? MHI_EE_MAX : exec;
+}
+
 int mhi_destroy_device(struct device *dev, void *data)
 {
 	struct mhi_device *mhi_dev;
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 5cbc1e33829f..5c4b77fa0fb4 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -263,6 +263,8 @@ struct mhi_controller_config {
  * @dev: Driver model device node for the controller (required)
  * @mhi_dev: MHI device instance for the controller
  * @regs: Base address of MHI MMIO register space (required)
+ * @bhi: Points to base of MHI BHI register space
+ * @wake_db: MHI WAKE doorbell register address
  * @iova_start: IOMMU starting address for data (required)
  * @iova_stop: IOMMU stop address for data (required)
  * @fw_image: Firmware image name for normal booting (required)
@@ -313,6 +315,8 @@ struct mhi_controller {
 	struct device *dev;
 	struct mhi_device *mhi_dev;
 	void __iomem *regs;
+	void __iomem *bhi;
+	void __iomem *wake_db;
 	dma_addr_t iova_start;
 	dma_addr_t iova_stop;
 	const char *fw_image;
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 06/16] bus: mhi: core: Add support for PM state transitions
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (4 preceding siblings ...)
  2020-01-31 13:49 ` [PATCH v2 05/16] bus: mhi: core: Add support for ringing channel/event ring doorbells Manivannan Sadhasivam
@ 2020-01-31 13:49 ` Manivannan Sadhasivam
  2020-02-06 20:15   ` Jeffrey Hugo
  2020-01-31 13:50 ` [PATCH v2 07/16] bus: mhi: core: Add support for basic PM operations Manivannan Sadhasivam
                   ` (9 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:49 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for transitioning the MHI states as a
part of the power management operations. Helpers functions are
provided for the state transitions, which will be consumed by the
actual power management routines.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/989

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[jhugo: removed dma_zalloc_coherent() and fixed several bugs]
Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
[mani: splitted the pm patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/Makefile   |   2 +-
 drivers/bus/mhi/core/init.c     |  65 +++
 drivers/bus/mhi/core/internal.h | 175 +++++++++
 drivers/bus/mhi/core/main.c     |   9 +
 drivers/bus/mhi/core/pm.c       | 678 ++++++++++++++++++++++++++++++++
 include/linux/mhi.h             |  52 +++
 6 files changed, 980 insertions(+), 1 deletion(-)
 create mode 100644 drivers/bus/mhi/core/pm.c

diff --git a/drivers/bus/mhi/core/Makefile b/drivers/bus/mhi/core/Makefile
index 77f7730da4bf..a0070f9cdfcd 100644
--- a/drivers/bus/mhi/core/Makefile
+++ b/drivers/bus/mhi/core/Makefile
@@ -1,3 +1,3 @@
 obj-$(CONFIG_MHI_BUS) := mhi.o
 
-mhi-y := init.o main.o
+mhi-y := init.o main.o pm.o
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index dec996fa1d1d..e8cf1692ccba 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -19,6 +19,62 @@
 #include <linux/wait.h>
 #include "internal.h"
 
+const char * const mhi_ee_str[MHI_EE_MAX] = {
+	[MHI_EE_PBL] = "PBL",
+	[MHI_EE_SBL] = "SBL",
+	[MHI_EE_AMSS] = "AMSS",
+	[MHI_EE_RDDM] = "RDDM",
+	[MHI_EE_WFW] = "WFW",
+	[MHI_EE_PTHRU] = "PASS THRU",
+	[MHI_EE_EDL] = "EDL",
+	[MHI_EE_DISABLE_TRANSITION] = "DISABLE",
+	[MHI_EE_NOT_SUPPORTED] = "NOT SUPPORTED",
+};
+
+const char * const dev_state_tran_str[DEV_ST_TRANSITION_MAX] = {
+	[DEV_ST_TRANSITION_PBL] = "PBL",
+	[DEV_ST_TRANSITION_READY] = "READY",
+	[DEV_ST_TRANSITION_SBL] = "SBL",
+	[DEV_ST_TRANSITION_MISSION_MODE] = "MISSION_MODE",
+};
+
+const char * const mhi_state_str[MHI_STATE_MAX] = {
+	[MHI_STATE_RESET] = "RESET",
+	[MHI_STATE_READY] = "READY",
+	[MHI_STATE_M0] = "M0",
+	[MHI_STATE_M1] = "M1",
+	[MHI_STATE_M2] = "M2",
+	[MHI_STATE_M3] = "M3",
+	[MHI_STATE_M3_FAST] = "M3_FAST",
+	[MHI_STATE_BHI] = "BHI",
+	[MHI_STATE_SYS_ERR] = "SYS_ERR",
+};
+
+static const char * const mhi_pm_state_str[] = {
+	[MHI_PM_STATE_DISABLE] = "DISABLE",
+	[MHI_PM_STATE_POR] = "POR",
+	[MHI_PM_STATE_M0] = "M0",
+	[MHI_PM_STATE_M2] = "M2",
+	[MHI_PM_STATE_M3_ENTER] = "M?->M3",
+	[MHI_PM_STATE_M3] = "M3",
+	[MHI_PM_STATE_M3_EXIT] = "M3->M0",
+	[MHI_PM_STATE_FW_DL_ERR] = "FW DL Error",
+	[MHI_PM_STATE_SYS_ERR_DETECT] = "SYS_ERR Detect",
+	[MHI_PM_STATE_SYS_ERR_PROCESS] = "SYS_ERR Process",
+	[MHI_PM_STATE_SHUTDOWN_PROCESS] = "SHUTDOWN Process",
+	[MHI_PM_STATE_LD_ERR_FATAL_DETECT] = "LD or Error Fatal Detect",
+};
+
+const char *to_mhi_pm_state_str(enum mhi_pm_state state)
+{
+	int index = find_last_bit((unsigned long *)&state, 32);
+
+	if (index >= ARRAY_SIZE(mhi_pm_state_str))
+		return "Invalid State";
+
+	return mhi_pm_state_str[index];
+}
+
 int mhi_init_mmio(struct mhi_controller *mhi_cntrl)
 {
 	u32 val;
@@ -372,6 +428,11 @@ static int parse_config(struct mhi_controller *mhi_cntrl,
 	if (!mhi_cntrl->buffer_len)
 		mhi_cntrl->buffer_len = MHI_MAX_MTU;
 
+	/* By default, host is allowed to ring DB in both M0 and M2 states */
+	mhi_cntrl->db_access = MHI_PM_M0 | MHI_PM_M2;
+	if (config->m2_no_db)
+		mhi_cntrl->db_access &= ~MHI_PM_M2;
+
 	return 0;
 
 error_ev_cfg:
@@ -411,8 +472,12 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
 	}
 
 	INIT_LIST_HEAD(&mhi_cntrl->transition_list);
+	mutex_init(&mhi_cntrl->pm_mutex);
+	rwlock_init(&mhi_cntrl->pm_lock);
 	spin_lock_init(&mhi_cntrl->transition_lock);
 	spin_lock_init(&mhi_cntrl->wlock);
+	INIT_WORK(&mhi_cntrl->st_worker, mhi_pm_st_worker);
+	INIT_WORK(&mhi_cntrl->syserr_worker, mhi_pm_sys_err_worker);
 	init_waitqueue_head(&mhi_cntrl->state_event);
 
 	mhi_cmd = mhi_cntrl->mhi_cmd;
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index 960dc5cf9343..a0db8b5d7424 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -267,6 +267,79 @@ enum mhi_cmd_type {
 	MHI_CMD_START_CHAN = 18,
 };
 
+/* No operation command */
+#define MHI_TRE_CMD_NOOP_PTR (0)
+#define MHI_TRE_CMD_NOOP_DWORD0 (0)
+#define MHI_TRE_CMD_NOOP_DWORD1 (MHI_CMD_NOP << 16)
+
+/* Channel reset command */
+#define MHI_TRE_CMD_RESET_PTR (0)
+#define MHI_TRE_CMD_RESET_DWORD0 (0)
+#define MHI_TRE_CMD_RESET_DWORD1(chid) ((chid << 24) | \
+					(MHI_CMD_RESET_CHAN << 16))
+
+/* Channel stop command */
+#define MHI_TRE_CMD_STOP_PTR (0)
+#define MHI_TRE_CMD_STOP_DWORD0 (0)
+#define MHI_TRE_CMD_STOP_DWORD1(chid) ((chid << 24) | \
+				       (MHI_CMD_STOP_CHAN << 16))
+
+/* Channel start command */
+#define MHI_TRE_CMD_START_PTR (0)
+#define MHI_TRE_CMD_START_DWORD0 (0)
+#define MHI_TRE_CMD_START_DWORD1(chid) ((chid << 24) | \
+					(MHI_CMD_START_CHAN << 16))
+
+#define MHI_TRE_GET_CMD_CHID(tre) (((tre)->dword[1] >> 24) & 0xFF)
+#define MHI_TRE_GET_CMD_TYPE(tre) (((tre)->dword[1] >> 16) & 0xFF)
+
+/* Event descriptor macros */
+#define MHI_TRE_EV_PTR(ptr) (ptr)
+#define MHI_TRE_EV_DWORD0(code, len) ((code << 24) | len)
+#define MHI_TRE_EV_DWORD1(chid, type) ((chid << 24) | (type << 16))
+#define MHI_TRE_GET_EV_PTR(tre) ((tre)->ptr)
+#define MHI_TRE_GET_EV_CODE(tre) (((tre)->dword[0] >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_LEN(tre) ((tre)->dword[0] & 0xFFFF)
+#define MHI_TRE_GET_EV_CHID(tre) (((tre)->dword[1] >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_TYPE(tre) (((tre)->dword[1] >> 16) & 0xFF)
+#define MHI_TRE_GET_EV_STATE(tre) (((tre)->dword[0] >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_EXECENV(tre) (((tre)->dword[0] >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_SEQ(tre) ((tre)->dword[0])
+#define MHI_TRE_GET_EV_TIME(tre) ((tre)->ptr)
+#define MHI_TRE_GET_EV_COOKIE(tre) lower_32_bits((tre)->ptr)
+#define MHI_TRE_GET_EV_VEID(tre) (((tre)->dword[0] >> 16) & 0xFF)
+#define MHI_TRE_GET_EV_LINKSPEED(tre) (((tre)->dword[1] >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_LINKWIDTH(tre) ((tre)->dword[0] & 0xFF)
+
+/* Transfer descriptor macros */
+#define MHI_TRE_DATA_PTR(ptr) (ptr)
+#define MHI_TRE_DATA_DWORD0(len) (len & MHI_MAX_MTU)
+#define MHI_TRE_DATA_DWORD1(bei, ieot, ieob, chain) ((2 << 16) | (bei << 10) \
+	| (ieot << 9) | (ieob << 8) | chain)
+
+/* RSC transfer descriptor macros */
+#define MHI_RSCTRE_DATA_PTR(ptr, len) (((u64)len << 48) | ptr)
+#define MHI_RSCTRE_DATA_DWORD0(cookie) (cookie)
+#define MHI_RSCTRE_DATA_DWORD1 (MHI_PKT_TYPE_COALESCING << 16)
+
+enum mhi_pkt_type {
+	MHI_PKT_TYPE_INVALID = 0x0,
+	MHI_PKT_TYPE_NOOP_CMD = 0x1,
+	MHI_PKT_TYPE_TRANSFER = 0x2,
+	MHI_PKT_TYPE_COALESCING = 0x8,
+	MHI_PKT_TYPE_RESET_CHAN_CMD = 0x10,
+	MHI_PKT_TYPE_STOP_CHAN_CMD = 0x11,
+	MHI_PKT_TYPE_START_CHAN_CMD = 0x12,
+	MHI_PKT_TYPE_STATE_CHANGE_EVENT = 0x20,
+	MHI_PKT_TYPE_CMD_COMPLETION_EVENT = 0x21,
+	MHI_PKT_TYPE_TX_EVENT = 0x22,
+	MHI_PKT_TYPE_RSC_TX_EVENT = 0x28,
+	MHI_PKT_TYPE_EE_EVENT = 0x40,
+	MHI_PKT_TYPE_TSYNC_EVENT = 0x48,
+	MHI_PKT_TYPE_BW_REQ_EVENT = 0x50,
+	MHI_PKT_TYPE_STALE_EVENT, /* internal event */
+};
+
 /* MHI transfer completion events */
 enum mhi_ev_ccs {
 	MHI_EV_CC_INVALID = 0x0,
@@ -292,6 +365,81 @@ enum mhi_ch_state {
 #define MHI_INVALID_BRSTMODE(mode) (mode != MHI_DB_BRST_DISABLE && \
 				    mode != MHI_DB_BRST_ENABLE)
 
+extern const char * const mhi_ee_str[MHI_EE_MAX];
+#define TO_MHI_EXEC_STR(ee) (((ee) >= MHI_EE_MAX) ? \
+			     "INVALID_EE" : mhi_ee_str[ee])
+
+#define MHI_IN_PBL(ee) (ee == MHI_EE_PBL || ee == MHI_EE_PTHRU || \
+			ee == MHI_EE_EDL)
+
+#define MHI_IN_MISSION_MODE(ee) (ee == MHI_EE_AMSS || ee == MHI_EE_WFW)
+
+enum dev_st_transition {
+	DEV_ST_TRANSITION_PBL,
+	DEV_ST_TRANSITION_READY,
+	DEV_ST_TRANSITION_SBL,
+	DEV_ST_TRANSITION_MISSION_MODE,
+	DEV_ST_TRANSITION_MAX,
+};
+
+extern const char * const dev_state_tran_str[DEV_ST_TRANSITION_MAX];
+#define TO_DEV_STATE_TRANS_STR(state) (((state) >= DEV_ST_TRANSITION_MAX) ? \
+				"INVALID_STATE" : dev_state_tran_str[state])
+
+extern const char * const mhi_state_str[MHI_STATE_MAX];
+#define TO_MHI_STATE_STR(state) ((state >= MHI_STATE_MAX || \
+				  !mhi_state_str[state]) ? \
+				"INVALID_STATE" : mhi_state_str[state])
+
+/* internal power states */
+enum mhi_pm_state {
+	MHI_PM_STATE_DISABLE,
+	MHI_PM_STATE_POR,
+	MHI_PM_STATE_M0,
+	MHI_PM_STATE_M2,
+	MHI_PM_STATE_M3_ENTER,
+	MHI_PM_STATE_M3,
+	MHI_PM_STATE_M3_EXIT,
+	MHI_PM_STATE_FW_DL_ERR,
+	MHI_PM_STATE_SYS_ERR_DETECT,
+	MHI_PM_STATE_SYS_ERR_PROCESS,
+	MHI_PM_STATE_SHUTDOWN_PROCESS,
+	MHI_PM_STATE_LD_ERR_FATAL_DETECT,
+	MHI_PM_STATE_MAX
+};
+
+#define MHI_PM_DISABLE			BIT(0)
+#define MHI_PM_POR			BIT(1)
+#define MHI_PM_M0			BIT(2)
+#define MHI_PM_M2			BIT(3)
+#define MHI_PM_M3_ENTER			BIT(4)
+#define MHI_PM_M3			BIT(5)
+#define MHI_PM_M3_EXIT			BIT(6)
+/* firmware download failure state */
+#define MHI_PM_FW_DL_ERR		BIT(7)
+#define MHI_PM_SYS_ERR_DETECT		BIT(8)
+#define MHI_PM_SYS_ERR_PROCESS		BIT(9)
+#define MHI_PM_SHUTDOWN_PROCESS		BIT(10)
+/* link not accessible */
+#define MHI_PM_LD_ERR_FATAL_DETECT	BIT(11)
+
+#define MHI_REG_ACCESS_VALID(pm_state) ((pm_state & (MHI_PM_POR | MHI_PM_M0 | \
+		MHI_PM_M2 | MHI_PM_M3_ENTER | MHI_PM_M3_EXIT | \
+		MHI_PM_SYS_ERR_DETECT | MHI_PM_SYS_ERR_PROCESS | \
+		MHI_PM_SHUTDOWN_PROCESS | MHI_PM_FW_DL_ERR)))
+#define MHI_PM_IN_ERROR_STATE(pm_state) (pm_state >= MHI_PM_FW_DL_ERR)
+#define MHI_PM_IN_FATAL_STATE(pm_state) (pm_state == MHI_PM_LD_ERR_FATAL_DETECT)
+#define MHI_DB_ACCESS_VALID(mhi_cntrl) (mhi_cntrl->pm_state & \
+					mhi_cntrl->db_access)
+#define MHI_WAKE_DB_CLEAR_VALID(pm_state) (pm_state & (MHI_PM_M0 | \
+						MHI_PM_M2 | MHI_PM_M3_EXIT))
+#define MHI_WAKE_DB_SET_VALID(pm_state) (pm_state & MHI_PM_M2)
+#define MHI_WAKE_DB_FORCE_SET_VALID(pm_state) MHI_WAKE_DB_CLEAR_VALID(pm_state)
+#define MHI_EVENT_ACCESS_INVALID(pm_state) (pm_state == MHI_PM_DISABLE || \
+					    MHI_PM_IN_ERROR_STATE(pm_state))
+#define MHI_PM_IN_SUSPEND_STATE(pm_state) (pm_state & \
+					   (MHI_PM_M3_ENTER | MHI_PM_M3))
+
 #define NR_OF_CMD_RINGS			1
 #define CMD_EL_PER_RING			128
 #define PRIMARY_CMD_RING		0
@@ -314,6 +462,16 @@ struct db_cfg {
 			   dma_addr_t db_val);
 };
 
+struct mhi_pm_transitions {
+	enum mhi_pm_state from_state;
+	u32 to_states;
+};
+
+struct state_transition {
+	struct list_head node;
+	enum dev_st_transition state;
+};
+
 struct mhi_ring {
 	dma_addr_t dma_handle;
 	dma_addr_t iommu_base;
@@ -417,6 +575,23 @@ static inline void mhi_dealloc_device(struct mhi_controller *mhi_cntrl,
 int mhi_destroy_device(struct device *dev, void *data);
 void mhi_create_devices(struct mhi_controller *mhi_cntrl);
 
+/* Power management APIs */
+enum mhi_pm_state __must_check mhi_tryset_pm_state(
+					struct mhi_controller *mhi_cntrl,
+					enum mhi_pm_state state);
+const char *to_mhi_pm_state_str(enum mhi_pm_state state);
+enum mhi_ee_type mhi_get_exec_env(struct mhi_controller *mhi_cntrl);
+int mhi_queue_state_transition(struct mhi_controller *mhi_cntrl,
+			       enum dev_st_transition state);
+void mhi_pm_st_worker(struct work_struct *work);
+void mhi_pm_sys_err_worker(struct work_struct *work);
+int mhi_ready_state_transition(struct mhi_controller *mhi_cntrl);
+void mhi_ctrl_ev_task(unsigned long data);
+int mhi_pm_m0_transition(struct mhi_controller *mhi_cntrl);
+void mhi_pm_m1_transition(struct mhi_controller *mhi_cntrl);
+int mhi_pm_m3_transition(struct mhi_controller *mhi_cntrl);
+int __mhi_device_get_sync(struct mhi_controller *mhi_cntrl);
+
 /* Register access methods */
 void mhi_db_brstmode(struct mhi_controller *mhi_cntrl, struct db_cfg *db_cfg,
 		     void __iomem *db_addr, dma_addr_t db_val);
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
index 59d29470bf73..d46ecbb97a28 100644
--- a/drivers/bus/mhi/core/main.c
+++ b/drivers/bus/mhi/core/main.c
@@ -135,6 +135,15 @@ enum mhi_ee_type mhi_get_exec_env(struct mhi_controller *mhi_cntrl)
 	return (ret) ? MHI_EE_MAX : exec;
 }
 
+enum mhi_state mhi_get_mhi_state(struct mhi_controller *mhi_cntrl)
+{
+	u32 state;
+	int ret = mhi_read_reg_field(mhi_cntrl, mhi_cntrl->regs, MHISTATUS,
+				     MHISTATUS_MHISTATE_MASK,
+				     MHISTATUS_MHISTATE_SHIFT, &state);
+	return ret ? MHI_STATE_MAX : state;
+}
+
 int mhi_destroy_device(struct device *dev, void *data)
 {
 	struct mhi_device *mhi_dev;
diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
new file mode 100644
index 000000000000..9b2a0fe158a5
--- /dev/null
+++ b/drivers/bus/mhi/core/pm.c
@@ -0,0 +1,678 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ *
+ */
+
+#define dev_fmt(fmt) "MHI: " fmt
+
+#include <linux/delay.h>
+#include <linux/device.h>
+#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/mhi.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/wait.h>
+#include "internal.h"
+
+/*
+ * Not all MHI state transitions are synchronous. Transitions like Linkdown,
+ * SYS_ERR, and shutdown can happen anytime asynchronously. This function will
+ * transition to a new state only if we're allowed to.
+ *
+ * Priority increases as we go down. For instance, from any state in L0, the
+ * transition can be made to states in L1, L2 and L3. A notable exception to
+ * this rule is state DISABLE.  From DISABLE state we can only transition to
+ * POR state. Also, while in L2 state, user cannot jump back to previous
+ * L1 or L0 states.
+ *
+ * Valid transitions:
+ * L0: DISABLE <--> POR
+ *     POR <--> POR
+ *     POR -> M0 -> M2 --> M0
+ *     POR -> FW_DL_ERR
+ *     FW_DL_ERR <--> FW_DL_ERR
+ *     M0 <--> M0
+ *     M0 -> FW_DL_ERR
+ *     M0 -> M3_ENTER -> M3 -> M3_EXIT --> M0
+ * L1: SYS_ERR_DETECT -> SYS_ERR_PROCESS --> POR
+ * L2: SHUTDOWN_PROCESS -> DISABLE
+ * L3: LD_ERR_FATAL_DETECT <--> LD_ERR_FATAL_DETECT
+ *     LD_ERR_FATAL_DETECT -> SHUTDOWN_PROCESS
+ */
+static struct mhi_pm_transitions const dev_state_transitions[] = {
+	/* L0 States */
+	{
+		MHI_PM_DISABLE,
+		MHI_PM_POR
+	},
+	{
+		MHI_PM_POR,
+		MHI_PM_POR | MHI_PM_DISABLE | MHI_PM_M0 |
+		MHI_PM_SYS_ERR_DETECT | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT | MHI_PM_FW_DL_ERR
+	},
+	{
+		MHI_PM_M0,
+		MHI_PM_M0 | MHI_PM_M2 | MHI_PM_M3_ENTER |
+		MHI_PM_SYS_ERR_DETECT | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT | MHI_PM_FW_DL_ERR
+	},
+	{
+		MHI_PM_M2,
+		MHI_PM_M0 | MHI_PM_SYS_ERR_DETECT | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	{
+		MHI_PM_M3_ENTER,
+		MHI_PM_M3 | MHI_PM_SYS_ERR_DETECT | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	{
+		MHI_PM_M3,
+		MHI_PM_M3_EXIT | MHI_PM_SYS_ERR_DETECT |
+		MHI_PM_SHUTDOWN_PROCESS | MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	{
+		MHI_PM_M3_EXIT,
+		MHI_PM_M0 | MHI_PM_SYS_ERR_DETECT | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	{
+		MHI_PM_FW_DL_ERR,
+		MHI_PM_FW_DL_ERR | MHI_PM_SYS_ERR_DETECT |
+		MHI_PM_SHUTDOWN_PROCESS | MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	/* L1 States */
+	{
+		MHI_PM_SYS_ERR_DETECT,
+		MHI_PM_SYS_ERR_PROCESS | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	{
+		MHI_PM_SYS_ERR_PROCESS,
+		MHI_PM_POR | MHI_PM_SHUTDOWN_PROCESS |
+		MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	/* L2 States */
+	{
+		MHI_PM_SHUTDOWN_PROCESS,
+		MHI_PM_DISABLE | MHI_PM_LD_ERR_FATAL_DETECT
+	},
+	/* L3 States */
+	{
+		MHI_PM_LD_ERR_FATAL_DETECT,
+		MHI_PM_LD_ERR_FATAL_DETECT | MHI_PM_SHUTDOWN_PROCESS
+	},
+};
+
+enum mhi_pm_state __must_check mhi_tryset_pm_state(struct mhi_controller *mhi_cntrl,
+						   enum mhi_pm_state state)
+{
+	unsigned long cur_state = mhi_cntrl->pm_state;
+	int index = find_last_bit(&cur_state, 32);
+
+	if (unlikely(index >= ARRAY_SIZE(dev_state_transitions)))
+		return cur_state;
+
+	if (unlikely(dev_state_transitions[index].from_state != cur_state))
+		return cur_state;
+
+	if (unlikely(!(dev_state_transitions[index].to_states & state)))
+		return cur_state;
+
+	mhi_cntrl->pm_state = state;
+	return mhi_cntrl->pm_state;
+}
+
+void mhi_set_mhi_state(struct mhi_controller *mhi_cntrl, enum mhi_state state)
+{
+	if (state == MHI_STATE_RESET) {
+		mhi_write_reg_field(mhi_cntrl, mhi_cntrl->regs, MHICTRL,
+				    MHICTRL_RESET_MASK, MHICTRL_RESET_SHIFT, 1);
+	} else {
+		mhi_write_reg_field(mhi_cntrl, mhi_cntrl->regs, MHICTRL,
+				    MHICTRL_MHISTATE_MASK,
+				    MHICTRL_MHISTATE_SHIFT, state);
+	}
+}
+
+/* Handle device ready state transition */
+int mhi_ready_state_transition(struct mhi_controller *mhi_cntrl)
+{
+	void __iomem *base = mhi_cntrl->regs;
+	u32 reset = 1, ready = 0;
+	struct mhi_event *mhi_event;
+	enum mhi_pm_state cur_state;
+	int ret, i;
+
+	/* Wait for RESET to be cleared and READY bit to be set by the device */
+	wait_event_timeout(mhi_cntrl->state_event,
+			   MHI_PM_IN_FATAL_STATE(mhi_cntrl->pm_state) ||
+			   mhi_read_reg_field(mhi_cntrl, base, MHICTRL,
+					      MHICTRL_RESET_MASK,
+					      MHICTRL_RESET_SHIFT, &reset) ||
+			   mhi_read_reg_field(mhi_cntrl, base, MHISTATUS,
+					      MHISTATUS_READY_MASK,
+					      MHISTATUS_READY_SHIFT, &ready) ||
+			   (!reset && ready),
+			   msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	/* Check if device entered error state */
+	if (MHI_PM_IN_FATAL_STATE(mhi_cntrl->pm_state)) {
+		dev_err(mhi_cntrl->dev, "Device link is not accessible\n");
+		return -EIO;
+	}
+
+	/* Timeout if device did not transition to ready state */
+	if (reset || !ready) {
+		dev_err(mhi_cntrl->dev, "Device Ready timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	dev_dbg(mhi_cntrl->dev, "Device in READY State\n");
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	cur_state = mhi_tryset_pm_state(mhi_cntrl, MHI_PM_POR);
+	mhi_cntrl->dev_state = MHI_STATE_READY;
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+
+	if (cur_state != MHI_PM_POR) {
+		dev_err(mhi_cntrl->dev, "Error moving to state %s from %s\n",
+			to_mhi_pm_state_str(MHI_PM_POR),
+			to_mhi_pm_state_str(cur_state));
+		return -EIO;
+	}
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (!MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
+		dev_err(mhi_cntrl->dev, "Device registers not accessible\n");
+		goto error_mmio;
+	}
+
+	/* Configure MMIO registers */
+	ret = mhi_init_mmio(mhi_cntrl);
+	if (ret) {
+		dev_err(mhi_cntrl->dev, "Error configuring MMIO registers\n");
+		goto error_mmio;
+	}
+
+	/* Add elements to all SW event rings */
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		struct mhi_ring *ring = &mhi_event->ring;
+
+		/* Skip if this is an offload or HW event */
+		if (mhi_event->offload_ev || mhi_event->hw_ring)
+			continue;
+
+		ring->wp = ring->base + ring->len - ring->el_size;
+		*ring->ctxt_wp = ring->iommu_base + ring->len - ring->el_size;
+		/* Update all cores */
+		smp_wmb();
+
+		/* Ring the event ring db */
+		spin_lock_irq(&mhi_event->lock);
+		mhi_ring_er_db(mhi_event);
+		spin_unlock_irq(&mhi_event->lock);
+	}
+
+	/* Set MHI to M0 state */
+	mhi_set_mhi_state(mhi_cntrl, MHI_STATE_M0);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return 0;
+
+error_mmio:
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return -EIO;
+}
+
+int mhi_pm_m0_transition(struct mhi_controller *mhi_cntrl)
+{
+	enum mhi_pm_state cur_state;
+	struct mhi_chan *mhi_chan;
+	int i;
+
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	mhi_cntrl->dev_state = MHI_STATE_M0;
+	cur_state = mhi_tryset_pm_state(mhi_cntrl, MHI_PM_M0);
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+	if (unlikely(cur_state != MHI_PM_M0)) {
+		dev_err(mhi_cntrl->dev, "Unable to transition to M0 state\n");
+		return -EIO;
+	}
+
+	/* Wake up the device */
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	mhi_cntrl->wake_get(mhi_cntrl, true);
+
+	/* Ring all event rings and CMD ring only if we're in mission mode */
+	if (MHI_IN_MISSION_MODE(mhi_cntrl->ee)) {
+		struct mhi_event *mhi_event = mhi_cntrl->mhi_event;
+		struct mhi_cmd *mhi_cmd =
+			&mhi_cntrl->mhi_cmd[PRIMARY_CMD_RING];
+
+		for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+			if (mhi_event->offload_ev)
+				continue;
+
+			spin_lock_irq(&mhi_event->lock);
+			mhi_ring_er_db(mhi_event);
+			spin_unlock_irq(&mhi_event->lock);
+		}
+
+		/* Only ring primary cmd ring if ring is not empty */
+		spin_lock_irq(&mhi_cmd->lock);
+		if (mhi_cmd->ring.rp != mhi_cmd->ring.wp)
+			mhi_ring_cmd_db(mhi_cntrl, mhi_cmd);
+		spin_unlock_irq(&mhi_cmd->lock);
+	}
+
+	/* Ring channel DB registers */
+	mhi_chan = mhi_cntrl->mhi_chan;
+	for (i = 0; i < mhi_cntrl->max_chan; i++, mhi_chan++) {
+		struct mhi_ring *tre_ring = &mhi_chan->tre_ring;
+
+		write_lock_irq(&mhi_chan->lock);
+		if (mhi_chan->db_cfg.reset_req)
+			mhi_chan->db_cfg.db_mode = true;
+
+		/* Only ring DB if ring is not empty */
+		if (tre_ring->base && tre_ring->wp  != tre_ring->rp)
+			mhi_ring_chan_db(mhi_cntrl, mhi_chan);
+		write_unlock_irq(&mhi_chan->lock);
+	}
+
+	mhi_cntrl->wake_put(mhi_cntrl, false);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+	wake_up_all(&mhi_cntrl->state_event);
+
+	return 0;
+}
+
+/*
+ * After receiving the MHI state change event from the device indicating the
+ * transition to M1 state, the host can transition the device to M2 state
+ * for keeping it in low power state.
+ */
+void mhi_pm_m1_transition(struct mhi_controller *mhi_cntrl)
+{
+	enum mhi_pm_state state;
+
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	state = mhi_tryset_pm_state(mhi_cntrl, MHI_PM_M2);
+	if (state == MHI_PM_M2) {
+		mhi_set_mhi_state(mhi_cntrl, MHI_STATE_M2);
+		mhi_cntrl->dev_state = MHI_STATE_M2;
+
+		write_unlock_irq(&mhi_cntrl->pm_lock);
+		wake_up_all(&mhi_cntrl->state_event);
+
+		/* If there are any pending resources, exit M2 immediately */
+		if (unlikely(atomic_read(&mhi_cntrl->pending_pkts) ||
+			     atomic_read(&mhi_cntrl->dev_wake))) {
+			dev_dbg(mhi_cntrl->dev,
+				 "Exiting M2, pending_pkts: %d dev_wake: %d\n",
+				 atomic_read(&mhi_cntrl->pending_pkts),
+				 atomic_read(&mhi_cntrl->dev_wake));
+			read_lock_bh(&mhi_cntrl->pm_lock);
+			mhi_cntrl->wake_get(mhi_cntrl, true);
+			mhi_cntrl->wake_put(mhi_cntrl, true);
+			read_unlock_bh(&mhi_cntrl->pm_lock);
+		} else {
+			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_IDLE);
+		}
+	} else {
+		write_unlock_irq(&mhi_cntrl->pm_lock);
+	}
+}
+
+/* MHI M3 completion handler */
+int mhi_pm_m3_transition(struct mhi_controller *mhi_cntrl)
+{
+	enum mhi_pm_state state;
+
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	mhi_cntrl->dev_state = MHI_STATE_M3;
+	state = mhi_tryset_pm_state(mhi_cntrl, MHI_PM_M3);
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+	if (state != MHI_PM_M3) {
+		dev_err(mhi_cntrl->dev, "Unable to transition to M3 state\n");
+		return -EIO;
+	}
+
+	wake_up_all(&mhi_cntrl->state_event);
+
+	return 0;
+}
+
+/* Handle device Mission Mode transition */
+static int mhi_pm_mission_mode_transition(struct mhi_controller *mhi_cntrl)
+{
+	int i, ret;
+	struct mhi_event *mhi_event;
+
+	dev_dbg(mhi_cntrl->dev, "Processing Mission Mode transition\n");
+
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	if (MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state))
+		mhi_cntrl->ee = mhi_get_exec_env(mhi_cntrl);
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+
+	if (!MHI_IN_MISSION_MODE(mhi_cntrl->ee))
+		return -EIO;
+
+	wake_up_all(&mhi_cntrl->state_event);
+
+	mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_MISSION_MODE);
+
+	/* Force MHI to be in M0 state before continuing */
+	ret = __mhi_device_get_sync(mhi_cntrl);
+	if (ret)
+		return ret;
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+		ret = -EIO;
+		goto error_mission_mode;
+	}
+
+	/* Add elements to all HW event rings */
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		struct mhi_ring *ring = &mhi_event->ring;
+
+		if (mhi_event->offload_ev || !mhi_event->hw_ring)
+			continue;
+
+		ring->wp = ring->base + ring->len - ring->el_size;
+		*ring->ctxt_wp = ring->iommu_base + ring->len - ring->el_size;
+		/* Update to all cores */
+		smp_wmb();
+
+		spin_lock_irq(&mhi_event->lock);
+		if (MHI_DB_ACCESS_VALID(mhi_cntrl))
+			mhi_ring_er_db(mhi_event);
+		spin_unlock_irq(&mhi_event->lock);
+	}
+
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	/*
+	 * The MHI devices are only created when the client device switches its
+	 * Execution Environment (EE) to either SBL or AMSS states
+	 */
+	mhi_create_devices(mhi_cntrl);
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+
+error_mission_mode:
+	mhi_cntrl->wake_put(mhi_cntrl, false);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return ret;
+}
+
+/* Handle SYS_ERR and Shutdown transitions */
+static void mhi_pm_disable_transition(struct mhi_controller *mhi_cntrl,
+				      enum mhi_pm_state transition_state)
+{
+	enum mhi_pm_state cur_state, prev_state;
+	struct mhi_event *mhi_event;
+	struct mhi_cmd_ctxt *cmd_ctxt;
+	struct mhi_cmd *mhi_cmd;
+	struct mhi_event_ctxt *er_ctxt;
+	int ret, i;
+
+	dev_dbg(mhi_cntrl->dev,
+		 "Transitioning from PM state: %s to: %s\n",
+		 to_mhi_pm_state_str(mhi_cntrl->pm_state),
+		 to_mhi_pm_state_str(transition_state));
+
+	/* We must notify MHI control driver so it can clean up first */
+	if (transition_state == MHI_PM_SYS_ERR_PROCESS) {
+		mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_SYS_ERROR);
+	}
+
+	mutex_lock(&mhi_cntrl->pm_mutex);
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	prev_state = mhi_cntrl->pm_state;
+	cur_state = mhi_tryset_pm_state(mhi_cntrl, transition_state);
+	if (cur_state == transition_state) {
+		mhi_cntrl->ee = MHI_EE_DISABLE_TRANSITION;
+		mhi_cntrl->dev_state = MHI_STATE_RESET;
+	}
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+
+	/* Wake up threads waiting for state transition */
+	wake_up_all(&mhi_cntrl->state_event);
+
+	if (cur_state != transition_state) {
+		dev_err(mhi_cntrl->dev,
+			 "Failed to transition to state: %s from: %s\n",
+			 to_mhi_pm_state_str(transition_state),
+			 to_mhi_pm_state_str(cur_state));
+		mutex_unlock(&mhi_cntrl->pm_mutex);
+		return;
+	}
+
+	/* Trigger MHI RESET so that the device will not access host memory */
+	if (MHI_REG_ACCESS_VALID(prev_state)) {
+		u32 in_reset = -1;
+		unsigned long timeout = msecs_to_jiffies(mhi_cntrl->timeout_ms);
+
+		dev_dbg(mhi_cntrl->dev, "Triggering MHI Reset in device\n");
+		mhi_set_mhi_state(mhi_cntrl, MHI_STATE_RESET);
+
+		/* Wait for the reset bit to be cleared by the device */
+		ret = wait_event_timeout(mhi_cntrl->state_event,
+					 mhi_read_reg_field(mhi_cntrl,
+							    mhi_cntrl->regs,
+							    MHICTRL,
+							    MHICTRL_RESET_MASK,
+							    MHICTRL_RESET_SHIFT,
+							    &in_reset) ||
+					!in_reset, timeout);
+		if ((!ret || in_reset) && cur_state == MHI_PM_SYS_ERR_PROCESS) {
+			dev_err(mhi_cntrl->dev,
+				"Device failed to exit MHI Reset state\n");
+			mutex_unlock(&mhi_cntrl->pm_mutex);
+			return;
+		}
+
+		/*
+		 * Device will clear BHI_INTVEC as a part of RESET processing,
+		 * hence re-program it
+		 */
+		mhi_write_reg(mhi_cntrl, mhi_cntrl->bhi, BHI_INTVEC, 0);
+	}
+
+	dev_dbg(mhi_cntrl->dev,
+		 "Waiting for all pending event ring processing to complete\n");
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		if (mhi_event->offload_ev)
+			continue;
+		tasklet_kill(&mhi_event->task);
+	}
+
+	/* Release lock and wait for all pending threads to complete */
+	mutex_unlock(&mhi_cntrl->pm_mutex);
+	dev_dbg(mhi_cntrl->dev,
+		 "Waiting for all pending threads to complete\n");
+	wake_up_all(&mhi_cntrl->state_event);
+	flush_work(&mhi_cntrl->st_worker);
+	flush_work(&mhi_cntrl->fw_worker);
+
+	dev_dbg(mhi_cntrl->dev,
+		 "Reset all active channels and remove MHI devices\n");
+	device_for_each_child(mhi_cntrl->dev, NULL, mhi_destroy_device);
+
+	mutex_lock(&mhi_cntrl->pm_mutex);
+
+	WARN_ON(atomic_read(&mhi_cntrl->dev_wake));
+	WARN_ON(atomic_read(&mhi_cntrl->pending_pkts));
+
+	/* Reset the ev rings and cmd rings */
+	dev_dbg(mhi_cntrl->dev, "Resetting EV CTXT and CMD CTXT\n");
+	mhi_cmd = mhi_cntrl->mhi_cmd;
+	cmd_ctxt = mhi_cntrl->mhi_ctxt->cmd_ctxt;
+	for (i = 0; i < NR_OF_CMD_RINGS; i++, mhi_cmd++, cmd_ctxt++) {
+		struct mhi_ring *ring = &mhi_cmd->ring;
+
+		ring->rp = ring->base;
+		ring->wp = ring->base;
+		cmd_ctxt->rp = cmd_ctxt->rbase;
+		cmd_ctxt->wp = cmd_ctxt->rbase;
+	}
+
+	mhi_event = mhi_cntrl->mhi_event;
+	er_ctxt = mhi_cntrl->mhi_ctxt->er_ctxt;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, er_ctxt++,
+		     mhi_event++) {
+		struct mhi_ring *ring = &mhi_event->ring;
+
+		/* Skip offload events */
+		if (mhi_event->offload_ev)
+			continue;
+
+		ring->rp = ring->base;
+		ring->wp = ring->base;
+		er_ctxt->rp = er_ctxt->rbase;
+		er_ctxt->wp = er_ctxt->rbase;
+	}
+
+	if (cur_state == MHI_PM_SYS_ERR_PROCESS) {
+		mhi_ready_state_transition(mhi_cntrl);
+	} else {
+		/* Move to disable state */
+		write_lock_irq(&mhi_cntrl->pm_lock);
+		cur_state = mhi_tryset_pm_state(mhi_cntrl, MHI_PM_DISABLE);
+		write_unlock_irq(&mhi_cntrl->pm_lock);
+		if (unlikely(cur_state != MHI_PM_DISABLE))
+			dev_err(mhi_cntrl->dev,
+				"Error moving from PM state: %s to: %s\n",
+				to_mhi_pm_state_str(cur_state),
+				to_mhi_pm_state_str(MHI_PM_DISABLE));
+	}
+
+	dev_dbg(mhi_cntrl->dev, "Exiting with PM state: %s, MHI state: %s\n",
+		 to_mhi_pm_state_str(mhi_cntrl->pm_state),
+		 TO_MHI_STATE_STR(mhi_cntrl->dev_state));
+
+	mutex_unlock(&mhi_cntrl->pm_mutex);
+}
+
+/* Queue a new work item and schedule work */
+int mhi_queue_state_transition(struct mhi_controller *mhi_cntrl,
+			       enum dev_st_transition state)
+{
+	struct state_transition *item = kmalloc(sizeof(*item), GFP_ATOMIC);
+	unsigned long flags;
+
+	if (!item)
+		return -ENOMEM;
+
+	item->state = state;
+	spin_lock_irqsave(&mhi_cntrl->transition_lock, flags);
+	list_add_tail(&item->node, &mhi_cntrl->transition_list);
+	spin_unlock_irqrestore(&mhi_cntrl->transition_lock, flags);
+
+	schedule_work(&mhi_cntrl->st_worker);
+
+	return 0;
+}
+
+/* SYS_ERR worker */
+void mhi_pm_sys_err_worker(struct work_struct *work)
+{
+	struct mhi_controller *mhi_cntrl = container_of(work,
+							struct mhi_controller,
+							syserr_worker);
+
+	mhi_pm_disable_transition(mhi_cntrl, MHI_PM_SYS_ERR_PROCESS);
+}
+
+/* Device State Transition worker */
+void mhi_pm_st_worker(struct work_struct *work)
+{
+	struct state_transition *itr, *tmp;
+	LIST_HEAD(head);
+	struct mhi_controller *mhi_cntrl = container_of(work,
+							struct mhi_controller,
+							st_worker);
+	spin_lock_irq(&mhi_cntrl->transition_lock);
+	list_splice_tail_init(&mhi_cntrl->transition_list, &head);
+	spin_unlock_irq(&mhi_cntrl->transition_lock);
+
+	list_for_each_entry_safe(itr, tmp, &head, node) {
+		list_del(&itr->node);
+		dev_dbg(mhi_cntrl->dev, "Handling state transition: %s\n",
+			TO_DEV_STATE_TRANS_STR(itr->state));
+
+		switch (itr->state) {
+		case DEV_ST_TRANSITION_PBL:
+			write_lock_irq(&mhi_cntrl->pm_lock);
+			if (MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state))
+				mhi_cntrl->ee = mhi_get_exec_env(mhi_cntrl);
+			write_unlock_irq(&mhi_cntrl->pm_lock);
+			if (MHI_IN_PBL(mhi_cntrl->ee))
+				wake_up_all(&mhi_cntrl->state_event);
+			break;
+		case DEV_ST_TRANSITION_SBL:
+			write_lock_irq(&mhi_cntrl->pm_lock);
+			mhi_cntrl->ee = MHI_EE_SBL;
+			write_unlock_irq(&mhi_cntrl->pm_lock);
+			/*
+			 * The MHI devices are only created when the client
+			 * device switches its Execution Environment (EE) to
+			 * either SBL or AMSS states
+			 */
+			mhi_create_devices(mhi_cntrl);
+			break;
+		case DEV_ST_TRANSITION_MISSION_MODE:
+			mhi_pm_mission_mode_transition(mhi_cntrl);
+			break;
+		case DEV_ST_TRANSITION_READY:
+			mhi_ready_state_transition(mhi_cntrl);
+			break;
+		default:
+			break;
+		}
+		kfree(itr);
+	}
+}
+
+int __mhi_device_get_sync(struct mhi_controller *mhi_cntrl)
+{
+	int ret;
+
+	/* Wake up the device */
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	mhi_cntrl->wake_get(mhi_cntrl, true);
+	if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
+		pm_wakeup_event(&mhi_cntrl->mhi_dev->dev, 0);
+		mhi_cntrl->runtime_get(mhi_cntrl);
+		mhi_cntrl->runtime_put(mhi_cntrl);
+	}
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	ret = wait_event_timeout(mhi_cntrl->state_event,
+				 mhi_cntrl->pm_state == MHI_PM_M0 ||
+				 MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state),
+				 msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	if (!ret || MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+		read_lock_bh(&mhi_cntrl->pm_lock);
+		mhi_cntrl->wake_put(mhi_cntrl, false);
+		read_unlock_bh(&mhi_cntrl->pm_lock);
+		return -EIO;
+	}
+
+	return 0;
+}
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 5c4b77fa0fb4..1164f379f54b 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -107,6 +107,31 @@ enum mhi_ee_type {
 	MHI_EE_MAX,
 };
 
+/**
+ * enum mhi_state - MHI states
+ * @MHI_STATE_RESET: Reset state
+ * @MHI_STATE_READY: Ready state
+ * @MHI_STATE_M0: M0 state
+ * @MHI_STATE_M1: M1 state
+ * @MHI_STATE_M2: M2 state
+ * @MHI_STATE_M3: M3 state
+ * @MHI_STATE_M3_FAST: M3 Fast state
+ * @MHI_STATE_BHI: BHI state
+ * @MHI_STATE_SYS_ERR: System Error state
+ */
+enum mhi_state {
+	MHI_STATE_RESET = 0x0,
+	MHI_STATE_READY = 0x1,
+	MHI_STATE_M0 = 0x2,
+	MHI_STATE_M1 = 0x3,
+	MHI_STATE_M2 = 0x4,
+	MHI_STATE_M3 = 0x5,
+	MHI_STATE_M3_FAST = 0x6,
+	MHI_STATE_BHI = 0x7,
+	MHI_STATE_SYS_ERR = 0xFF,
+	MHI_STATE_MAX,
+};
+
 /**
  * enum mhi_buf_type - Accepted buffer type for the channel
  * @MHI_BUF_RAW: Raw buffer
@@ -289,6 +314,7 @@ struct mhi_controller_config {
  * @pm_state: MHI power management state
  * @db_access: DB access states
  * @ee: MHI device execution environment
+ * @dev_state: MHI device state
  * @dev_wake: Device wakeup count
  * @pending_pkts: Pending packets for the controller
  * @transition_list: List of MHI state transitions
@@ -317,6 +343,7 @@ struct mhi_controller {
 	void __iomem *regs;
 	void __iomem *bhi;
 	void __iomem *wake_db;
+
 	dma_addr_t iova_start;
 	dma_addr_t iova_stop;
 	const char *fw_image;
@@ -343,6 +370,7 @@ struct mhi_controller {
 	u32 pm_state;
 	u32 db_access;
 	enum mhi_ee_type ee;
+	enum mhi_state dev_state;
 	atomic_t dev_wake;
 	atomic_t pending_pkts;
 	struct list_head transition_list;
@@ -411,6 +439,22 @@ struct mhi_result {
 	int transaction_status;
 };
 
+/**
+ * struct mhi_buf - MHI Buffer description
+ * @buf: Virtual address of the buffer
+ * @name: Buffer label. For offload channel, configurations name must be:
+ *        ECA - Event context array data
+ *        CCA - Channel context array data
+ * @dma_addr: IOMMU address of the buffer
+ * @len: # of bytes
+ */
+struct mhi_buf {
+	void *buf;
+	const char *name;
+	dma_addr_t dma_addr;
+	size_t len;
+};
+
 /**
  * struct mhi_driver - Structure representing a MHI client driver
  * @probe: CB function for client driver probe function
@@ -462,4 +506,12 @@ int mhi_driver_register(struct mhi_driver *mhi_drv);
  */
 void mhi_driver_unregister(struct mhi_driver *mhi_drv);
 
+/**
+ * mhi_set_mhi_state - Set MHI device state
+ * @mhi_cntrl: MHI controller
+ * @state: State to set
+ */
+void mhi_set_mhi_state(struct mhi_controller *mhi_cntrl,
+		       enum mhi_state state);
+
 #endif /* _MHI_H_ */
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 07/16] bus: mhi: core: Add support for basic PM operations
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (5 preceding siblings ...)
  2020-01-31 13:49 ` [PATCH v2 06/16] bus: mhi: core: Add support for PM state transitions Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-06 20:15   ` Jeffrey Hugo
  2020-01-31 13:50 ` [PATCH v2 08/16] bus: mhi: core: Add support for downloading firmware over BHIe Manivannan Sadhasivam
                   ` (8 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for basic MHI PM operations such as
mhi_async_power_up, mhi_sync_power_up, and mhi_power_down. These
routines places the MHI bus into respective power domain states
and calls the state_transition APIs when necessary. The MHI
controller driver is expected to call these PM routines for
MHI powerup and powerdown.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/989

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted the pm patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/Makefile   |   2 +-
 drivers/bus/mhi/core/boot.c     |  89 +++++++++
 drivers/bus/mhi/core/init.c     | 323 ++++++++++++++++++++++++++++++++
 drivers/bus/mhi/core/internal.h |  33 ++++
 drivers/bus/mhi/core/main.c     |  87 +++++++++
 drivers/bus/mhi/core/pm.c       | 218 +++++++++++++++++++++
 include/linux/mhi.h             |  51 +++++
 7 files changed, 802 insertions(+), 1 deletion(-)
 create mode 100644 drivers/bus/mhi/core/boot.c

diff --git a/drivers/bus/mhi/core/Makefile b/drivers/bus/mhi/core/Makefile
index a0070f9cdfcd..66e2700c9032 100644
--- a/drivers/bus/mhi/core/Makefile
+++ b/drivers/bus/mhi/core/Makefile
@@ -1,3 +1,3 @@
 obj-$(CONFIG_MHI_BUS) := mhi.o
 
-mhi-y := init.o main.o pm.o
+mhi-y := init.o main.o pm.o boot.o
diff --git a/drivers/bus/mhi/core/boot.c b/drivers/bus/mhi/core/boot.c
new file mode 100644
index 000000000000..0996f18c4281
--- /dev/null
+++ b/drivers/bus/mhi/core/boot.c
@@ -0,0 +1,89 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ *
+ */
+
+#define dev_fmt(fmt) "MHI: " fmt
+
+#include <linux/delay.h>
+#include <linux/device.h>
+#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/mhi.h>
+#include <linux/module.h>
+#include <linux/random.h>
+#include <linux/slab.h>
+#include <linux/wait.h>
+#include "internal.h"
+
+void mhi_free_bhie_table(struct mhi_controller *mhi_cntrl,
+			 struct image_info *image_info)
+{
+	int i;
+	struct mhi_buf *mhi_buf = image_info->mhi_buf;
+
+	for (i = 0; i < image_info->entries; i++, mhi_buf++)
+		mhi_free_coherent(mhi_cntrl, mhi_buf->len, mhi_buf->buf,
+				  mhi_buf->dma_addr);
+
+	kfree(image_info->mhi_buf);
+	kfree(image_info);
+}
+
+int mhi_alloc_bhie_table(struct mhi_controller *mhi_cntrl,
+			 struct image_info **image_info,
+			 size_t alloc_size)
+{
+	size_t seg_size = mhi_cntrl->seg_len;
+	int segments = DIV_ROUND_UP(alloc_size, seg_size) + 1;
+	int i;
+	struct image_info *img_info;
+	struct mhi_buf *mhi_buf;
+
+	img_info = kzalloc(sizeof(*img_info), GFP_KERNEL);
+	if (!img_info)
+		return -ENOMEM;
+
+	/* Allocate memory for entries */
+	img_info->mhi_buf = kcalloc(segments, sizeof(*img_info->mhi_buf),
+				    GFP_KERNEL);
+	if (!img_info->mhi_buf)
+		goto error_alloc_mhi_buf;
+
+	/* Allocate and populate vector table */
+	mhi_buf = img_info->mhi_buf;
+	for (i = 0; i < segments; i++, mhi_buf++) {
+		size_t vec_size = seg_size;
+
+		/* Vector table is the last entry */
+		if (i == segments - 1)
+			vec_size = sizeof(struct bhi_vec_entry) * i;
+
+		mhi_buf->len = vec_size;
+		mhi_buf->buf = mhi_alloc_coherent(mhi_cntrl, vec_size,
+						  &mhi_buf->dma_addr,
+						  GFP_KERNEL);
+		if (!mhi_buf->buf)
+			goto error_alloc_segment;
+	}
+
+	img_info->bhi_vec = img_info->mhi_buf[segments - 1].buf;
+	img_info->entries = segments;
+	*image_info = img_info;
+
+	return 0;
+
+error_alloc_segment:
+	for (--i, --mhi_buf; i >= 0; i--, mhi_buf--)
+		mhi_free_coherent(mhi_cntrl, mhi_buf->len, mhi_buf->buf,
+				  mhi_buf->dma_addr);
+
+error_alloc_mhi_buf:
+	kfree(img_info);
+
+	return -ENOMEM;
+}
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index e8cf1692ccba..573f7c28b028 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -75,6 +75,294 @@ const char *to_mhi_pm_state_str(enum mhi_pm_state state)
 	return mhi_pm_state_str[index];
 }
 
+/* MHI protocol requires the transfer ring to be aligned with ring length */
+static int mhi_alloc_aligned_ring(struct mhi_controller *mhi_cntrl,
+				  struct mhi_ring *ring,
+				  u64 len)
+{
+	ring->alloc_size = len + (len - 1);
+	ring->pre_aligned = mhi_alloc_coherent(mhi_cntrl, ring->alloc_size,
+					       &ring->dma_handle, GFP_KERNEL);
+	if (!ring->pre_aligned)
+		return -ENOMEM;
+
+	ring->iommu_base = (ring->dma_handle + (len - 1)) & ~(len - 1);
+	ring->base = ring->pre_aligned + (ring->iommu_base - ring->dma_handle);
+
+	return 0;
+}
+
+void mhi_deinit_free_irq(struct mhi_controller *mhi_cntrl)
+{
+	int i;
+	struct mhi_event *mhi_event = mhi_cntrl->mhi_event;
+
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		if (mhi_event->offload_ev)
+			continue;
+
+		free_irq(mhi_cntrl->irq[mhi_event->irq], mhi_event);
+	}
+
+	free_irq(mhi_cntrl->irq[0], mhi_cntrl);
+}
+
+int mhi_init_irq_setup(struct mhi_controller *mhi_cntrl)
+{
+	int i;
+	int ret;
+	struct mhi_event *mhi_event = mhi_cntrl->mhi_event;
+
+	/* Setup BHI_INTVEC IRQ */
+	ret = request_threaded_irq(mhi_cntrl->irq[0], mhi_intvec_handler,
+				   mhi_intvec_threaded_handler,
+				   IRQF_SHARED | IRQF_NO_SUSPEND,
+				   "bhi", mhi_cntrl);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		if (mhi_event->offload_ev)
+			continue;
+
+		ret = request_irq(mhi_cntrl->irq[mhi_event->irq],
+				  mhi_irq_handler,
+				  IRQF_SHARED | IRQF_NO_SUSPEND,
+				  "mhi", mhi_event);
+		if (ret) {
+			dev_err(mhi_cntrl->dev,
+				"Error requesting irq:%d for ev:%d\n",
+				mhi_cntrl->irq[mhi_event->irq], i);
+			goto error_request;
+		}
+	}
+
+	return 0;
+
+error_request:
+	for (--i, --mhi_event; i >= 0; i--, mhi_event--) {
+		if (mhi_event->offload_ev)
+			continue;
+
+		free_irq(mhi_cntrl->irq[mhi_event->irq], mhi_event);
+	}
+	free_irq(mhi_cntrl->irq[0], mhi_cntrl);
+
+	return ret;
+}
+
+void mhi_deinit_dev_ctxt(struct mhi_controller *mhi_cntrl)
+{
+	int i;
+	struct mhi_ctxt *mhi_ctxt = mhi_cntrl->mhi_ctxt;
+	struct mhi_cmd *mhi_cmd;
+	struct mhi_event *mhi_event;
+	struct mhi_ring *ring;
+
+	mhi_cmd = mhi_cntrl->mhi_cmd;
+	for (i = 0; i < NR_OF_CMD_RINGS; i++, mhi_cmd++) {
+		ring = &mhi_cmd->ring;
+		mhi_free_coherent(mhi_cntrl, ring->alloc_size,
+				  ring->pre_aligned, ring->dma_handle);
+		ring->base = NULL;
+		ring->iommu_base = 0;
+	}
+
+	mhi_free_coherent(mhi_cntrl,
+			  sizeof(*mhi_ctxt->cmd_ctxt) * NR_OF_CMD_RINGS,
+			  mhi_ctxt->cmd_ctxt, mhi_ctxt->cmd_ctxt_addr);
+
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, mhi_event++) {
+		if (mhi_event->offload_ev)
+			continue;
+
+		ring = &mhi_event->ring;
+		mhi_free_coherent(mhi_cntrl, ring->alloc_size,
+				  ring->pre_aligned, ring->dma_handle);
+		ring->base = NULL;
+		ring->iommu_base = 0;
+	}
+
+	mhi_free_coherent(mhi_cntrl, sizeof(*mhi_ctxt->er_ctxt) *
+			  mhi_cntrl->total_ev_rings, mhi_ctxt->er_ctxt,
+			  mhi_ctxt->er_ctxt_addr);
+
+	mhi_free_coherent(mhi_cntrl, sizeof(*mhi_ctxt->chan_ctxt) *
+			  mhi_cntrl->max_chan, mhi_ctxt->chan_ctxt,
+			  mhi_ctxt->chan_ctxt_addr);
+
+	kfree(mhi_ctxt);
+	mhi_cntrl->mhi_ctxt = NULL;
+}
+
+int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl)
+{
+	struct mhi_ctxt *mhi_ctxt;
+	struct mhi_chan_ctxt *chan_ctxt;
+	struct mhi_event_ctxt *er_ctxt;
+	struct mhi_cmd_ctxt *cmd_ctxt;
+	struct mhi_chan *mhi_chan;
+	struct mhi_event *mhi_event;
+	struct mhi_cmd *mhi_cmd;
+	u32 tmp;
+	int ret = -ENOMEM, i;
+
+	atomic_set(&mhi_cntrl->dev_wake, 0);
+	atomic_set(&mhi_cntrl->pending_pkts, 0);
+
+	mhi_ctxt = kzalloc(sizeof(*mhi_ctxt), GFP_KERNEL);
+	if (!mhi_ctxt)
+		return -ENOMEM;
+
+	/* Setup channel ctxt */
+	mhi_ctxt->chan_ctxt = mhi_alloc_coherent(mhi_cntrl,
+						 sizeof(*mhi_ctxt->chan_ctxt) *
+						 mhi_cntrl->max_chan,
+						 &mhi_ctxt->chan_ctxt_addr,
+						 GFP_KERNEL);
+	if (!mhi_ctxt->chan_ctxt)
+		goto error_alloc_chan_ctxt;
+
+	mhi_chan = mhi_cntrl->mhi_chan;
+	chan_ctxt = mhi_ctxt->chan_ctxt;
+	for (i = 0; i < mhi_cntrl->max_chan; i++, chan_ctxt++, mhi_chan++) {
+		/* Skip if it is an offload channel */
+		if (mhi_chan->offload_ch)
+			continue;
+
+		tmp = chan_ctxt->chcfg;
+		tmp &= ~CHAN_CTX_CHSTATE_MASK;
+		tmp |= (MHI_CH_STATE_DISABLED << CHAN_CTX_CHSTATE_SHIFT);
+		tmp &= ~CHAN_CTX_BRSTMODE_MASK;
+		tmp |= (mhi_chan->db_cfg.brstmode << CHAN_CTX_BRSTMODE_SHIFT);
+		tmp &= ~CHAN_CTX_POLLCFG_MASK;
+		tmp |= (mhi_chan->db_cfg.pollcfg << CHAN_CTX_POLLCFG_SHIFT);
+		chan_ctxt->chcfg = tmp;
+
+		chan_ctxt->chtype = mhi_chan->type;
+		chan_ctxt->erindex = mhi_chan->er_index;
+
+		mhi_chan->ch_state = MHI_CH_STATE_DISABLED;
+		mhi_chan->tre_ring.db_addr = (void __iomem *)&chan_ctxt->wp;
+	}
+
+	/* Setup event context */
+	mhi_ctxt->er_ctxt = mhi_alloc_coherent(mhi_cntrl,
+					       sizeof(*mhi_ctxt->er_ctxt) *
+					       mhi_cntrl->total_ev_rings,
+					       &mhi_ctxt->er_ctxt_addr,
+					       GFP_KERNEL);
+	if (!mhi_ctxt->er_ctxt)
+		goto error_alloc_er_ctxt;
+
+	er_ctxt = mhi_ctxt->er_ctxt;
+	mhi_event = mhi_cntrl->mhi_event;
+	for (i = 0; i < mhi_cntrl->total_ev_rings; i++, er_ctxt++,
+		     mhi_event++) {
+		struct mhi_ring *ring = &mhi_event->ring;
+
+		/* Skip if it is an offload event */
+		if (mhi_event->offload_ev)
+			continue;
+
+		tmp = er_ctxt->intmod;
+		tmp &= ~EV_CTX_INTMODC_MASK;
+		tmp &= ~EV_CTX_INTMODT_MASK;
+		tmp |= (mhi_event->intmod << EV_CTX_INTMODT_SHIFT);
+		er_ctxt->intmod = tmp;
+
+		er_ctxt->ertype = MHI_ER_TYPE_VALID;
+		er_ctxt->msivec = mhi_event->irq;
+		mhi_event->db_cfg.db_mode = true;
+
+		ring->el_size = sizeof(struct mhi_tre);
+		ring->len = ring->el_size * ring->elements;
+		ret = mhi_alloc_aligned_ring(mhi_cntrl, ring, ring->len);
+		if (ret)
+			goto error_alloc_er;
+
+		/*
+		 * If the read pointer equals to the write pointer, then the
+		 * ring is empty
+		 */
+		ring->rp = ring->wp = ring->base;
+		er_ctxt->rbase = ring->iommu_base;
+		er_ctxt->rp = er_ctxt->wp = er_ctxt->rbase;
+		er_ctxt->rlen = ring->len;
+		ring->ctxt_wp = &er_ctxt->wp;
+	}
+
+	/* Setup cmd context */
+	mhi_ctxt->cmd_ctxt = mhi_alloc_coherent(mhi_cntrl,
+						sizeof(*mhi_ctxt->cmd_ctxt) *
+						NR_OF_CMD_RINGS,
+						&mhi_ctxt->cmd_ctxt_addr,
+						GFP_KERNEL);
+	if (!mhi_ctxt->cmd_ctxt)
+		goto error_alloc_er;
+
+	mhi_cmd = mhi_cntrl->mhi_cmd;
+	cmd_ctxt = mhi_ctxt->cmd_ctxt;
+	for (i = 0; i < NR_OF_CMD_RINGS; i++, mhi_cmd++, cmd_ctxt++) {
+		struct mhi_ring *ring = &mhi_cmd->ring;
+
+		ring->el_size = sizeof(struct mhi_tre);
+		ring->elements = CMD_EL_PER_RING;
+		ring->len = ring->el_size * ring->elements;
+		ret = mhi_alloc_aligned_ring(mhi_cntrl, ring, ring->len);
+		if (ret)
+			goto error_alloc_cmd;
+
+		ring->rp = ring->wp = ring->base;
+		cmd_ctxt->rbase = ring->iommu_base;
+		cmd_ctxt->rp = cmd_ctxt->wp = cmd_ctxt->rbase;
+		cmd_ctxt->rlen = ring->len;
+		ring->ctxt_wp = &cmd_ctxt->wp;
+	}
+
+	mhi_cntrl->mhi_ctxt = mhi_ctxt;
+
+	return 0;
+
+error_alloc_cmd:
+	for (--i, --mhi_cmd; i >= 0; i--, mhi_cmd--) {
+		struct mhi_ring *ring = &mhi_cmd->ring;
+
+		mhi_free_coherent(mhi_cntrl, ring->alloc_size,
+				  ring->pre_aligned, ring->dma_handle);
+	}
+	mhi_free_coherent(mhi_cntrl,
+			  sizeof(*mhi_ctxt->cmd_ctxt) * NR_OF_CMD_RINGS,
+			  mhi_ctxt->cmd_ctxt, mhi_ctxt->cmd_ctxt_addr);
+	i = mhi_cntrl->total_ev_rings;
+	mhi_event = mhi_cntrl->mhi_event + i;
+
+error_alloc_er:
+	for (--i, --mhi_event; i >= 0; i--, mhi_event--) {
+		struct mhi_ring *ring = &mhi_event->ring;
+
+		if (mhi_event->offload_ev)
+			continue;
+
+		mhi_free_coherent(mhi_cntrl, ring->alloc_size,
+				  ring->pre_aligned, ring->dma_handle);
+	}
+	mhi_free_coherent(mhi_cntrl, sizeof(*mhi_ctxt->er_ctxt) *
+			  mhi_cntrl->total_ev_rings, mhi_ctxt->er_ctxt,
+			  mhi_ctxt->er_ctxt_addr);
+
+error_alloc_er_ctxt:
+	mhi_free_coherent(mhi_cntrl, sizeof(*mhi_ctxt->chan_ctxt) *
+			  mhi_cntrl->max_chan, mhi_ctxt->chan_ctxt,
+			  mhi_ctxt->chan_ctxt_addr);
+
+error_alloc_chan_ctxt:
+	kfree(mhi_ctxt);
+
+	return ret;
+}
+
 int mhi_init_mmio(struct mhi_controller *mhi_cntrl)
 {
 	u32 val;
@@ -551,6 +839,41 @@ void mhi_unregister_controller(struct mhi_controller *mhi_cntrl)
 }
 EXPORT_SYMBOL_GPL(mhi_unregister_controller);
 
+int mhi_prepare_for_power_up(struct mhi_controller *mhi_cntrl)
+{
+	int ret;
+
+	mutex_lock(&mhi_cntrl->pm_mutex);
+
+	ret = mhi_init_dev_ctxt(mhi_cntrl);
+	if (ret)
+		goto error_dev_ctxt;
+
+	mhi_cntrl->pre_init = true;
+
+	mutex_unlock(&mhi_cntrl->pm_mutex);
+
+	return 0;
+
+error_dev_ctxt:
+	mutex_unlock(&mhi_cntrl->pm_mutex);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_prepare_for_power_up);
+
+void mhi_unprepare_after_power_down(struct mhi_controller *mhi_cntrl)
+{
+	if (mhi_cntrl->fbc_image) {
+		mhi_free_bhie_table(mhi_cntrl, mhi_cntrl->fbc_image);
+		mhi_cntrl->fbc_image = NULL;
+	}
+
+	mhi_deinit_dev_ctxt(mhi_cntrl);
+	mhi_cntrl->pre_init = false;
+}
+EXPORT_SYMBOL_GPL(mhi_unprepare_after_power_down);
+
 static void mhi_release_device(struct device *dev)
 {
 	struct mhi_device *mhi_dev = to_mhi_device(dev);
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index a0db8b5d7424..e325c840697e 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -575,6 +575,11 @@ static inline void mhi_dealloc_device(struct mhi_controller *mhi_cntrl,
 int mhi_destroy_device(struct device *dev, void *data);
 void mhi_create_devices(struct mhi_controller *mhi_cntrl);
 
+int mhi_alloc_bhie_table(struct mhi_controller *mhi_cntrl,
+			 struct image_info **image_info, size_t alloc_size);
+void mhi_free_bhie_table(struct mhi_controller *mhi_cntrl,
+			 struct image_info *image_info);
+
 /* Power management APIs */
 enum mhi_pm_state __must_check mhi_tryset_pm_state(
 					struct mhi_controller *mhi_cntrl,
@@ -616,5 +621,33 @@ void mhi_ring_chan_db(struct mhi_controller *mhi_cntrl,
 
 /* Initialization methods */
 int mhi_init_mmio(struct mhi_controller *mhi_cntrl);
+int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl);
+void mhi_deinit_dev_ctxt(struct mhi_controller *mhi_cntrl);
+int mhi_init_irq_setup(struct mhi_controller *mhi_cntrl);
+void mhi_deinit_free_irq(struct mhi_controller *mhi_cntrl);
+
+/* Memory allocation methods */
+static inline void *mhi_alloc_coherent(struct mhi_controller *mhi_cntrl,
+				       size_t size,
+				       dma_addr_t *dma_handle,
+				       gfp_t gfp)
+{
+	void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
+
+	return buf;
+}
+
+static inline void mhi_free_coherent(struct mhi_controller *mhi_cntrl,
+				     size_t size,
+				     void *vaddr,
+				     dma_addr_t dma_handle)
+{
+	dma_free_coherent(mhi_cntrl->dev, size, vaddr, dma_handle);
+}
+
+/* ISR handlers */
+irqreturn_t mhi_irq_handler(int irq_number, void *dev);
+irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *dev);
+irqreturn_t mhi_intvec_handler(int irq_number, void *dev);
 
 #endif /* _MHI_INT_H */
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
index d46ecbb97a28..4d767dd97956 100644
--- a/drivers/bus/mhi/core/main.c
+++ b/drivers/bus/mhi/core/main.c
@@ -144,6 +144,11 @@ enum mhi_state mhi_get_mhi_state(struct mhi_controller *mhi_cntrl)
 	return ret ? MHI_STATE_MAX : state;
 }
 
+static void *mhi_to_virtual(struct mhi_ring *ring, dma_addr_t addr)
+{
+	return (addr - ring->iommu_base) + ring->base;
+}
+
 int mhi_destroy_device(struct device *dev, void *data)
 {
 	struct mhi_device *mhi_dev;
@@ -248,3 +253,85 @@ void mhi_create_devices(struct mhi_controller *mhi_cntrl)
 			mhi_dealloc_device(mhi_cntrl, mhi_dev);
 	}
 }
+
+irqreturn_t mhi_irq_handler(int irq_number, void *dev)
+{
+	struct mhi_event *mhi_event = dev;
+	struct mhi_controller *mhi_cntrl = mhi_event->mhi_cntrl;
+	struct mhi_event_ctxt *er_ctxt =
+		&mhi_cntrl->mhi_ctxt->er_ctxt[mhi_event->er_index];
+	struct mhi_ring *ev_ring = &mhi_event->ring;
+	void *dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
+
+	/* Only proceed if event ring has pending events */
+	if (ev_ring->rp == dev_rp)
+		return IRQ_HANDLED;
+
+	/* For client managed event ring, notify pending data */
+	if (mhi_event->cl_manage) {
+		struct mhi_chan *mhi_chan = mhi_event->mhi_chan;
+		struct mhi_device *mhi_dev = mhi_chan->mhi_dev;
+
+		if (mhi_dev)
+			mhi_notify(mhi_dev, MHI_CB_PENDING_DATA);
+	} else {
+		tasklet_schedule(&mhi_event->task);
+	}
+
+	return IRQ_HANDLED;
+}
+
+irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *dev)
+{
+	struct mhi_controller *mhi_cntrl = dev;
+	enum mhi_state state = MHI_STATE_MAX;
+	enum mhi_pm_state pm_state = 0;
+	enum mhi_ee_type ee = 0;
+
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	if (MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
+		state = mhi_get_mhi_state(mhi_cntrl);
+		ee = mhi_cntrl->ee;
+		mhi_cntrl->ee = mhi_get_exec_env(mhi_cntrl);
+	}
+
+	if (state == MHI_STATE_SYS_ERR) {
+		dev_dbg(mhi_cntrl->dev, "System error detected\n");
+		pm_state = mhi_tryset_pm_state(mhi_cntrl,
+					       MHI_PM_SYS_ERR_DETECT);
+	}
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+
+	/* If device in RDDM don't bother processing SYS error */
+	if (mhi_cntrl->ee == MHI_EE_RDDM) {
+		if (mhi_cntrl->ee != ee) {
+			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM);
+			wake_up_all(&mhi_cntrl->state_event);
+		}
+		goto exit_intvec;
+	}
+
+	if (pm_state == MHI_PM_SYS_ERR_DETECT) {
+		wake_up_all(&mhi_cntrl->state_event);
+
+		/* For fatal errors, we let controller decide next step */
+		if (MHI_IN_PBL(ee))
+			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR);
+		else
+			schedule_work(&mhi_cntrl->syserr_worker);
+	}
+
+exit_intvec:
+
+	return IRQ_HANDLED;
+}
+
+irqreturn_t mhi_intvec_handler(int irq_number, void *dev)
+{
+	struct mhi_controller *mhi_cntrl = dev;
+
+	/* Wake up events waiting for state change */
+	wake_up_all(&mhi_cntrl->state_event);
+
+	return IRQ_WAKE_THREAD;
+}
diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
index 9b2a0fe158a5..b4bedf139184 100644
--- a/drivers/bus/mhi/core/pm.c
+++ b/drivers/bus/mhi/core/pm.c
@@ -140,6 +140,17 @@ void mhi_set_mhi_state(struct mhi_controller *mhi_cntrl, enum mhi_state state)
 	}
 }
 
+/* NOP for backward compatibility, host allowed to ring DB in M2 state */
+static void mhi_toggle_dev_wake_nop(struct mhi_controller *mhi_cntrl)
+{
+}
+
+static void mhi_toggle_dev_wake(struct mhi_controller *mhi_cntrl)
+{
+	mhi_cntrl->wake_get(mhi_cntrl, false);
+	mhi_cntrl->wake_put(mhi_cntrl, true);
+}
+
 /* Handle device ready state transition */
 int mhi_ready_state_transition(struct mhi_controller *mhi_cntrl)
 {
@@ -676,3 +687,210 @@ int __mhi_device_get_sync(struct mhi_controller *mhi_cntrl)
 
 	return 0;
 }
+
+/* Assert device wake db */
+static void mhi_assert_dev_wake(struct mhi_controller *mhi_cntrl, bool force)
+{
+	unsigned long flags;
+
+	/*
+	 * If force flag is set, then increment the wake count value and
+	 * ring wake db
+	 */
+	if (unlikely(force)) {
+		spin_lock_irqsave(&mhi_cntrl->wlock, flags);
+		atomic_inc(&mhi_cntrl->dev_wake);
+		if (MHI_WAKE_DB_FORCE_SET_VALID(mhi_cntrl->pm_state) &&
+		    !mhi_cntrl->wake_set) {
+			mhi_write_db(mhi_cntrl, mhi_cntrl->wake_db, 1);
+			mhi_cntrl->wake_set = true;
+		}
+		spin_unlock_irqrestore(&mhi_cntrl->wlock, flags);
+	} else {
+		/*
+		 * If resources are already requested, then just increment
+		 * the wake count value and return
+		 */
+		if (likely(atomic_add_unless(&mhi_cntrl->dev_wake, 1, 0)))
+			return;
+
+		spin_lock_irqsave(&mhi_cntrl->wlock, flags);
+		if ((atomic_inc_return(&mhi_cntrl->dev_wake) == 1) &&
+		    MHI_WAKE_DB_SET_VALID(mhi_cntrl->pm_state) &&
+		    !mhi_cntrl->wake_set) {
+			mhi_write_db(mhi_cntrl, mhi_cntrl->wake_db, 1);
+			mhi_cntrl->wake_set = true;
+		}
+		spin_unlock_irqrestore(&mhi_cntrl->wlock, flags);
+	}
+}
+
+/* De-assert device wake db */
+static void mhi_deassert_dev_wake(struct mhi_controller *mhi_cntrl,
+				  bool override)
+{
+	unsigned long flags;
+
+	/*
+	 * Only continue if there is a single resource, else just decrement
+	 * and return
+	 */
+	if (likely(atomic_add_unless(&mhi_cntrl->dev_wake, -1, 1)))
+		return;
+
+	spin_lock_irqsave(&mhi_cntrl->wlock, flags);
+	if ((atomic_dec_return(&mhi_cntrl->dev_wake) == 0) &&
+	    MHI_WAKE_DB_CLEAR_VALID(mhi_cntrl->pm_state) && !override &&
+	    mhi_cntrl->wake_set) {
+		mhi_write_db(mhi_cntrl, mhi_cntrl->wake_db, 0);
+		mhi_cntrl->wake_set = false;
+	}
+	spin_unlock_irqrestore(&mhi_cntrl->wlock, flags);
+}
+
+int mhi_async_power_up(struct mhi_controller *mhi_cntrl)
+{
+	int ret;
+	u32 val;
+	enum mhi_ee_type current_ee;
+	enum dev_st_transition next_state;
+
+	dev_info(mhi_cntrl->dev, "Requested to power ON\n");
+
+	if (mhi_cntrl->nr_irqs < mhi_cntrl->total_ev_rings)
+		return -EINVAL;
+
+	/* Supply default wake routines if not provided by controller driver */
+	if (!mhi_cntrl->wake_get || !mhi_cntrl->wake_put ||
+	    !mhi_cntrl->wake_toggle) {
+		mhi_cntrl->wake_get = mhi_assert_dev_wake;
+		mhi_cntrl->wake_put = mhi_deassert_dev_wake;
+		mhi_cntrl->wake_toggle = (mhi_cntrl->db_access & MHI_PM_M2) ?
+			mhi_toggle_dev_wake_nop : mhi_toggle_dev_wake;
+	}
+
+	mutex_lock(&mhi_cntrl->pm_mutex);
+	mhi_cntrl->pm_state = MHI_PM_DISABLE;
+
+	if (!mhi_cntrl->pre_init) {
+		/* Setup device context */
+		ret = mhi_init_dev_ctxt(mhi_cntrl);
+		if (ret)
+			goto error_dev_ctxt;
+	}
+
+	ret = mhi_init_irq_setup(mhi_cntrl);
+	if (ret)
+		goto error_setup_irq;
+
+	/* Setup BHI offset & INTVEC */
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	ret = mhi_read_reg(mhi_cntrl, mhi_cntrl->regs, BHIOFF, &val);
+	if (ret) {
+		write_unlock_irq(&mhi_cntrl->pm_lock);
+		goto error_bhi_offset;
+	}
+
+	mhi_cntrl->bhi = mhi_cntrl->regs + val;
+
+	/* Setup BHIE offset */
+	if (mhi_cntrl->fbc_download) {
+		ret = mhi_read_reg(mhi_cntrl, mhi_cntrl->regs, BHIEOFF, &val);
+		if (ret) {
+			write_unlock_irq(&mhi_cntrl->pm_lock);
+			dev_err(mhi_cntrl->dev, "Error reading BHIE offset\n");
+			goto error_bhi_offset;
+		}
+
+		mhi_cntrl->bhie = mhi_cntrl->regs + val;
+	}
+
+	mhi_write_reg(mhi_cntrl, mhi_cntrl->bhi, BHI_INTVEC, 0);
+	mhi_cntrl->pm_state = MHI_PM_POR;
+	mhi_cntrl->ee = MHI_EE_MAX;
+	current_ee = mhi_get_exec_env(mhi_cntrl);
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+
+	/* Confirm that the device is in valid exec env */
+	if (!MHI_IN_PBL(current_ee) && current_ee != MHI_EE_AMSS) {
+		dev_err(mhi_cntrl->dev, "Not a valid EE for power on\n");
+		ret = -EIO;
+		goto error_bhi_offset;
+	}
+
+	/* Transition to next state */
+	next_state = MHI_IN_PBL(current_ee) ?
+		DEV_ST_TRANSITION_PBL : DEV_ST_TRANSITION_READY;
+
+	if (next_state == DEV_ST_TRANSITION_PBL)
+		schedule_work(&mhi_cntrl->fw_worker);
+
+	mhi_queue_state_transition(mhi_cntrl, next_state);
+
+	mutex_unlock(&mhi_cntrl->pm_mutex);
+
+	dev_info(mhi_cntrl->dev, "Power on setup success\n");
+
+	return 0;
+
+error_bhi_offset:
+	mhi_deinit_free_irq(mhi_cntrl);
+
+error_setup_irq:
+	if (!mhi_cntrl->pre_init)
+		mhi_deinit_dev_ctxt(mhi_cntrl);
+
+error_dev_ctxt:
+	mutex_unlock(&mhi_cntrl->pm_mutex);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_async_power_up);
+
+void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful)
+{
+	enum mhi_pm_state cur_state;
+
+	/* If it's not a graceful shutdown, force MHI to linkdown state */
+	if (!graceful) {
+		mutex_lock(&mhi_cntrl->pm_mutex);
+		write_lock_irq(&mhi_cntrl->pm_lock);
+		cur_state = mhi_tryset_pm_state(mhi_cntrl,
+						MHI_PM_LD_ERR_FATAL_DETECT);
+		write_unlock_irq(&mhi_cntrl->pm_lock);
+		mutex_unlock(&mhi_cntrl->pm_mutex);
+		if (cur_state != MHI_PM_LD_ERR_FATAL_DETECT)
+			dev_dbg(mhi_cntrl->dev,
+			"Failed to move to state: %s from: %s\n",
+			to_mhi_pm_state_str(MHI_PM_LD_ERR_FATAL_DETECT),
+			to_mhi_pm_state_str(mhi_cntrl->pm_state));
+	}
+	mhi_pm_disable_transition(mhi_cntrl, MHI_PM_SHUTDOWN_PROCESS);
+	mhi_deinit_free_irq(mhi_cntrl);
+
+	if (!mhi_cntrl->pre_init) {
+		/* Free all allocated resources */
+		if (mhi_cntrl->fbc_image) {
+			mhi_free_bhie_table(mhi_cntrl, mhi_cntrl->fbc_image);
+			mhi_cntrl->fbc_image = NULL;
+		}
+		mhi_deinit_dev_ctxt(mhi_cntrl);
+	}
+}
+EXPORT_SYMBOL_GPL(mhi_power_down);
+
+int mhi_sync_power_up(struct mhi_controller *mhi_cntrl)
+{
+	int ret = mhi_async_power_up(mhi_cntrl);
+
+	if (ret)
+		return ret;
+
+	wait_event_timeout(mhi_cntrl->state_event,
+			   MHI_IN_MISSION_MODE(mhi_cntrl->ee) ||
+			   MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state),
+			   msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	return (MHI_IN_MISSION_MODE(mhi_cntrl->ee)) ? 0 : -EIO;
+}
+EXPORT_SYMBOL(mhi_sync_power_up);
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 1164f379f54b..e6dc0a517b33 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -83,6 +83,17 @@ enum mhi_ch_type {
 	MHI_CH_TYPE_INBOUND_COALESCED = 3,
 };
 
+/**
+ * struct image_info - Firmware and RDDM table table
+ * @mhi_buf - Buffer for firmware and RDDM table
+ * @entries - # of entries in table
+ */
+struct image_info {
+	struct mhi_buf *mhi_buf;
+	struct bhi_vec_entry *bhi_vec;
+	u32 entries;
+};
+
 /**
  * enum mhi_ee_type - Execution environment types
  * @MHI_EE_PBL: Primary Bootloader
@@ -289,6 +300,7 @@ struct mhi_controller_config {
  * @mhi_dev: MHI device instance for the controller
  * @regs: Base address of MHI MMIO register space (required)
  * @bhi: Points to base of MHI BHI register space
+ * @bhie: Points to base of MHI BHIe register space
  * @wake_db: MHI WAKE doorbell register address
  * @iova_start: IOMMU starting address for data (required)
  * @iova_stop: IOMMU stop address for data (required)
@@ -296,6 +308,7 @@ struct mhi_controller_config {
  * @edl_image: Firmware image name for emergency download mode (optional)
  * @sbl_size: SBL image size downloaded through BHIe (optional)
  * @seg_len: BHIe vector size (optional)
+ * @fbc_image: Points to firmware image buffer
  * @mhi_chan: Points to the channel configuration table
  * @lpm_chans: List of channels that require LPM notifications
  * @irq: base irq # to request (required)
@@ -342,6 +355,7 @@ struct mhi_controller {
 	struct mhi_device *mhi_dev;
 	void __iomem *regs;
 	void __iomem *bhi;
+	void __iomem *bhie;
 	void __iomem *wake_db;
 
 	dma_addr_t iova_start;
@@ -350,6 +364,7 @@ struct mhi_controller {
 	const char *edl_image;
 	size_t sbl_size;
 	size_t seg_len;
+	struct image_info *fbc_image;
 	struct mhi_chan *mhi_chan;
 	struct list_head lpm_chans;
 	int *irq;
@@ -514,4 +529,40 @@ void mhi_driver_unregister(struct mhi_driver *mhi_drv);
 void mhi_set_mhi_state(struct mhi_controller *mhi_cntrl,
 		       enum mhi_state state);
 
+/**
+ * mhi_prepare_for_power_up - Do pre-initialization before power up.
+ *                            This is optional, call this before power up if
+ *                            the controller does not want bus framework to
+ *                            automatically free any allocated memory during
+ *                            shutdown process.
+ * @mhi_cntrl: MHI controller
+ */
+int mhi_prepare_for_power_up(struct mhi_controller *mhi_cntrl);
+
+/**
+ * mhi_async_power_up - Start MHI power up sequence
+ * @mhi_cntrl: MHI controller
+ */
+int mhi_async_power_up(struct mhi_controller *mhi_cntrl);
+
+/**
+ * mhi_sync_power_up - Start MHI power up sequence and wait till the device
+ *                     device enters valid EE state
+ * @mhi_cntrl: MHI controller
+ */
+int mhi_sync_power_up(struct mhi_controller *mhi_cntrl);
+
+/**
+ * mhi_power_down - Start MHI power down sequence
+ * @mhi_cntrl: MHI controller
+ * @graceful: Link is still accessible, so do a graceful shutdown process
+ */
+void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful);
+
+/**
+ * mhi_unprepare_after_power_down - Free any allocated memory after power down
+ * @mhi_cntrl: MHI controller
+ */
+void mhi_unprepare_after_power_down(struct mhi_controller *mhi_cntrl);
+
 #endif /* _MHI_H_ */
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 08/16] bus: mhi: core: Add support for downloading firmware over BHIe
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (6 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 07/16] bus: mhi: core: Add support for basic PM operations Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-06 20:15   ` Jeffrey Hugo
  2020-01-31 13:50 ` [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic Manivannan Sadhasivam
                   ` (7 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

MHI supports downloading the device firmware over BHI/BHIe (Boot Host
Interface) protocol. Hence, this commit adds necessary helpers, which
will be called during device power up stage.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/989

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted the data transfer patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/boot.c     | 268 ++++++++++++++++++++++++++++++++
 drivers/bus/mhi/core/init.c     |   1 +
 drivers/bus/mhi/core/internal.h |   1 +
 3 files changed, 270 insertions(+)

diff --git a/drivers/bus/mhi/core/boot.c b/drivers/bus/mhi/core/boot.c
index 0996f18c4281..36956fb6eff2 100644
--- a/drivers/bus/mhi/core/boot.c
+++ b/drivers/bus/mhi/core/boot.c
@@ -20,6 +20,121 @@
 #include <linux/wait.h>
 #include "internal.h"
 
+/* Download AMSS image to device */
+static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
+			    const struct mhi_buf *mhi_buf)
+{
+	void __iomem *base = mhi_cntrl->bhie;
+	rwlock_t *pm_lock = &mhi_cntrl->pm_lock;
+	u32 tx_status, sequence_id;
+
+	read_lock_bh(pm_lock);
+	if (!MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
+		read_unlock_bh(pm_lock);
+		return -EIO;
+	}
+
+	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECADDR_HIGH_OFFS,
+		      upper_32_bits(mhi_buf->dma_addr));
+
+	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECADDR_LOW_OFFS,
+		      lower_32_bits(mhi_buf->dma_addr));
+
+	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECSIZE_OFFS, mhi_buf->len);
+
+	sequence_id = prandom_u32() & BHIE_TXVECSTATUS_SEQNUM_BMSK;
+	mhi_write_reg_field(mhi_cntrl, base, BHIE_TXVECDB_OFFS,
+			    BHIE_TXVECDB_SEQNUM_BMSK, BHIE_TXVECDB_SEQNUM_SHFT,
+			    sequence_id);
+	read_unlock_bh(pm_lock);
+
+	/* Wait for the image download to complete */
+	wait_event_timeout(mhi_cntrl->state_event,
+			   MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state) ||
+			   mhi_read_reg_field(mhi_cntrl, base,
+					      BHIE_TXVECSTATUS_OFFS,
+					      BHIE_TXVECSTATUS_STATUS_BMSK,
+					      BHIE_TXVECSTATUS_STATUS_SHFT,
+					      &tx_status) || tx_status,
+			   msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))
+		return -EIO;
+
+	return (tx_status == BHIE_TXVECSTATUS_STATUS_XFER_COMPL) ? 0 : -EIO;
+}
+
+/* Download SBL image to device */
+static int mhi_fw_load_sbl(struct mhi_controller *mhi_cntrl,
+			   dma_addr_t dma_addr,
+			   size_t size)
+{
+	u32 tx_status, val, session_id;
+	int i, ret;
+	void __iomem *base = mhi_cntrl->bhi;
+	rwlock_t *pm_lock = &mhi_cntrl->pm_lock;
+	struct {
+		char *name;
+		u32 offset;
+	} error_reg[] = {
+		{ "ERROR_CODE", BHI_ERRCODE },
+		{ "ERROR_DBG1", BHI_ERRDBG1 },
+		{ "ERROR_DBG2", BHI_ERRDBG2 },
+		{ "ERROR_DBG3", BHI_ERRDBG3 },
+		{ NULL },
+	};
+
+	read_lock_bh(pm_lock);
+	if (!MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
+		read_unlock_bh(pm_lock);
+		goto invalid_pm_state;
+	}
+
+	/* Start SBL download via BHI protocol */
+	mhi_write_reg(mhi_cntrl, base, BHI_STATUS, 0);
+	mhi_write_reg(mhi_cntrl, base, BHI_IMGADDR_HIGH,
+		      upper_32_bits(dma_addr));
+	mhi_write_reg(mhi_cntrl, base, BHI_IMGADDR_LOW,
+		      lower_32_bits(dma_addr));
+	mhi_write_reg(mhi_cntrl, base, BHI_IMGSIZE, size);
+	session_id = prandom_u32() & BHI_TXDB_SEQNUM_BMSK;
+	mhi_write_reg(mhi_cntrl, base, BHI_IMGTXDB, session_id);
+	read_unlock_bh(pm_lock);
+
+	/* Wait for the image download to complete */
+	ret = wait_event_timeout(mhi_cntrl->state_event,
+			   MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state) ||
+			   mhi_read_reg_field(mhi_cntrl, base, BHI_STATUS,
+					      BHI_STATUS_MASK, BHI_STATUS_SHIFT,
+					      &tx_status) || tx_status,
+			   msecs_to_jiffies(mhi_cntrl->timeout_ms));
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))
+		goto invalid_pm_state;
+
+	if (tx_status == BHI_STATUS_ERROR) {
+		dev_err(mhi_cntrl->dev, "Image transfer failed\n");
+		read_lock_bh(pm_lock);
+		if (MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
+			for (i = 0; error_reg[i].name; i++) {
+				ret = mhi_read_reg(mhi_cntrl, base,
+						   error_reg[i].offset, &val);
+				if (ret)
+					break;
+				dev_err(mhi_cntrl->dev, "Reg: %s value: 0x%x\n",
+					error_reg[i].name, val);
+			}
+		}
+		read_unlock_bh(pm_lock);
+		goto invalid_pm_state;
+	}
+
+	return (!ret) ? -ETIMEDOUT : 0;
+
+invalid_pm_state:
+
+	return -EIO;
+}
+
 void mhi_free_bhie_table(struct mhi_controller *mhi_cntrl,
 			 struct image_info *image_info)
 {
@@ -87,3 +202,156 @@ int mhi_alloc_bhie_table(struct mhi_controller *mhi_cntrl,
 
 	return -ENOMEM;
 }
+
+static void mhi_firmware_copy(struct mhi_controller *mhi_cntrl,
+			      const struct firmware *firmware,
+			      struct image_info *img_info)
+{
+	size_t remainder = firmware->size;
+	size_t to_cpy;
+	const u8 *buf = firmware->data;
+	int i = 0;
+	struct mhi_buf *mhi_buf = img_info->mhi_buf;
+	struct bhi_vec_entry *bhi_vec = img_info->bhi_vec;
+
+	while (remainder) {
+		to_cpy = min(remainder, mhi_buf->len);
+		memcpy(mhi_buf->buf, buf, to_cpy);
+		bhi_vec->dma_addr = mhi_buf->dma_addr;
+		bhi_vec->size = to_cpy;
+
+		buf += to_cpy;
+		remainder -= to_cpy;
+		i++;
+		bhi_vec++;
+		mhi_buf++;
+	}
+}
+
+void mhi_fw_load_worker(struct work_struct *work)
+{
+	int ret;
+	struct mhi_controller *mhi_cntrl;
+	const char *fw_name;
+	const struct firmware *firmware = NULL;
+	struct image_info *image_info;
+	void *buf;
+	dma_addr_t dma_addr;
+	size_t size;
+
+	mhi_cntrl = container_of(work, struct mhi_controller, fw_worker);
+
+	dev_dbg(mhi_cntrl->dev, "Waiting for device to enter PBL from: %s\n",
+		TO_MHI_EXEC_STR(mhi_cntrl->ee));
+
+	ret = wait_event_timeout(mhi_cntrl->state_event,
+				 MHI_IN_PBL(mhi_cntrl->ee) ||
+				 MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state),
+				 msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	if (!ret || MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+		dev_err(mhi_cntrl->dev, "Device MHI is not in valid state\n");
+		return;
+	}
+
+	/* If device is in pass through, do reset to ready state transition */
+	if (mhi_cntrl->ee == MHI_EE_PTHRU)
+		goto fw_load_ee_pthru;
+
+	fw_name = (mhi_cntrl->ee == MHI_EE_EDL) ?
+		mhi_cntrl->edl_image : mhi_cntrl->fw_image;
+
+	if (!fw_name || (mhi_cntrl->fbc_download && (!mhi_cntrl->sbl_size ||
+						     !mhi_cntrl->seg_len))) {
+		dev_err(mhi_cntrl->dev,
+			"No firmware image defined or !sbl_size || !seg_len\n");
+		return;
+	}
+
+	ret = request_firmware(&firmware, fw_name, mhi_cntrl->dev);
+	if (ret) {
+		dev_err(mhi_cntrl->dev, "Error loading firmware: %d\n",
+			ret);
+		return;
+	}
+
+	size = (mhi_cntrl->fbc_download) ? mhi_cntrl->sbl_size : firmware->size;
+
+	/* SBL size provided is maximum size, not necessarily the image size */
+	if (size > firmware->size)
+		size = firmware->size;
+
+	buf = mhi_alloc_coherent(mhi_cntrl, size, &dma_addr, GFP_KERNEL);
+	if (!buf) {
+		release_firmware(firmware);
+		return;
+	}
+
+	/* Download SBL image */
+	memcpy(buf, firmware->data, size);
+	ret = mhi_fw_load_sbl(mhi_cntrl, dma_addr, size);
+	mhi_free_coherent(mhi_cntrl, size, buf, dma_addr);
+
+	if (!mhi_cntrl->fbc_download || ret || mhi_cntrl->ee == MHI_EE_EDL)
+		release_firmware(firmware);
+
+	/* Error or in EDL mode, we're done */
+	if (ret || mhi_cntrl->ee == MHI_EE_EDL)
+		return;
+
+	write_lock_irq(&mhi_cntrl->pm_lock);
+	mhi_cntrl->dev_state = MHI_STATE_RESET;
+	write_unlock_irq(&mhi_cntrl->pm_lock);
+
+	/*
+	 * If we're doing fbc, populate vector tables while
+	 * device transitioning into MHI READY state
+	 */
+	if (mhi_cntrl->fbc_download) {
+		ret = mhi_alloc_bhie_table(mhi_cntrl, &mhi_cntrl->fbc_image,
+					   firmware->size);
+		if (ret)
+			goto error_alloc_fw_table;
+
+		/* Load the firmware into BHIE vec table */
+		mhi_firmware_copy(mhi_cntrl, firmware, mhi_cntrl->fbc_image);
+	}
+
+fw_load_ee_pthru:
+	/* Transitioning into MHI RESET->READY state */
+	ret = mhi_ready_state_transition(mhi_cntrl);
+
+	if (!mhi_cntrl->fbc_download)
+		return;
+
+	if (ret)
+		goto error_read;
+
+	/* Wait for the SBL event */
+	ret = wait_event_timeout(mhi_cntrl->state_event,
+				 mhi_cntrl->ee == MHI_EE_SBL ||
+				 MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state),
+				 msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	if (!ret || MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+		dev_err(mhi_cntrl->dev, "MHI did not enter SBL\n");
+		goto error_read;
+	}
+
+	/* Start full firmware image download */
+	image_info = mhi_cntrl->fbc_image;
+	ret = mhi_fw_load_amss(mhi_cntrl,
+			       /* Vector table is the last entry */
+			       &image_info->mhi_buf[image_info->entries - 1]);
+
+	release_firmware(firmware);
+
+	return;
+
+error_read:
+	mhi_free_bhie_table(mhi_cntrl, mhi_cntrl->fbc_image);
+	mhi_cntrl->fbc_image = NULL;
+
+error_alloc_fw_table:
+	release_firmware(firmware);
+}
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index 573f7c28b028..68071d0130f1 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -766,6 +766,7 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
 	spin_lock_init(&mhi_cntrl->wlock);
 	INIT_WORK(&mhi_cntrl->st_worker, mhi_pm_st_worker);
 	INIT_WORK(&mhi_cntrl->syserr_worker, mhi_pm_sys_err_worker);
+	INIT_WORK(&mhi_cntrl->fw_worker, mhi_fw_load_worker);
 	init_waitqueue_head(&mhi_cntrl->state_event);
 
 	mhi_cmd = mhi_cntrl->mhi_cmd;
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index e325c840697e..01047a8c47e1 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -590,6 +590,7 @@ int mhi_queue_state_transition(struct mhi_controller *mhi_cntrl,
 			       enum dev_st_transition state);
 void mhi_pm_st_worker(struct work_struct *work);
 void mhi_pm_sys_err_worker(struct work_struct *work);
+void mhi_fw_load_worker(struct work_struct *work);
 int mhi_ready_state_transition(struct mhi_controller *mhi_cntrl);
 void mhi_ctrl_ev_task(unsigned long data);
 int mhi_pm_m0_transition(struct mhi_controller *mhi_cntrl);
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (7 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 08/16] bus: mhi: core: Add support for downloading firmware over BHIe Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-06 16:41   ` Jeffrey Hugo
  2020-01-31 13:50 ` [PATCH v2 10/16] bus: mhi: core: Add support for processing events from client device Manivannan Sadhasivam
                   ` (6 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

MHI protocol supports downloading RDDM (RAM Dump) image from the
device through BHIE. This is useful to debugging as the RDDM image
can capture the firmware state.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/989

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted the data transfer patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/boot.c     | 157 +++++++++++++++++++++++++++++++-
 drivers/bus/mhi/core/init.c     |  38 ++++++++
 drivers/bus/mhi/core/internal.h |   2 +
 drivers/bus/mhi/core/pm.c       |  31 +++++++
 include/linux/mhi.h             |  24 +++++
 5 files changed, 249 insertions(+), 3 deletions(-)

diff --git a/drivers/bus/mhi/core/boot.c b/drivers/bus/mhi/core/boot.c
index 36956fb6eff2..564d61f98257 100644
--- a/drivers/bus/mhi/core/boot.c
+++ b/drivers/bus/mhi/core/boot.c
@@ -20,7 +20,159 @@
 #include <linux/wait.h>
 #include "internal.h"
 
-/* Download AMSS image to device */
+/* Setup RDDM vector table for RDDM transfer and program RXVEC */
+void mhi_rddm_prepare(struct mhi_controller *mhi_cntrl,
+		      struct image_info *img_info)
+{
+	struct mhi_buf *mhi_buf = img_info->mhi_buf;
+	struct bhi_vec_entry *bhi_vec = img_info->bhi_vec;
+	void __iomem *base = mhi_cntrl->bhie;
+	u32 sequence_id;
+	unsigned int i;
+
+	for (i = 0; i < img_info->entries - 1; i++, mhi_buf++, bhi_vec++) {
+		bhi_vec->dma_addr = mhi_buf->dma_addr;
+		bhi_vec->size = mhi_buf->len;
+	}
+
+	dev_dbg(mhi_cntrl->dev, "BHIe programming for RDDM\n");
+
+	mhi_write_reg(mhi_cntrl, base, BHIE_RXVECADDR_HIGH_OFFS,
+		      upper_32_bits(mhi_buf->dma_addr));
+
+	mhi_write_reg(mhi_cntrl, base, BHIE_RXVECADDR_LOW_OFFS,
+		      lower_32_bits(mhi_buf->dma_addr));
+
+	mhi_write_reg(mhi_cntrl, base, BHIE_RXVECSIZE_OFFS, mhi_buf->len);
+	sequence_id = prandom_u32() & BHIE_RXVECSTATUS_SEQNUM_BMSK;
+
+	if (unlikely(!sequence_id))
+		sequence_id = 1;
+
+	mhi_write_reg_field(mhi_cntrl, base, BHIE_RXVECDB_OFFS,
+			    BHIE_RXVECDB_SEQNUM_BMSK, BHIE_RXVECDB_SEQNUM_SHFT,
+			    sequence_id);
+
+	dev_dbg(mhi_cntrl->dev, "Address: %p and len: 0x%lx sequence: %u\n",
+		&mhi_buf->dma_addr, mhi_buf->len, sequence_id);
+}
+
+/* Collect RDDM buffer during kernel panic */
+static int __mhi_download_rddm_in_panic(struct mhi_controller *mhi_cntrl)
+{
+	int ret;
+	u32 rx_status;
+	enum mhi_ee_type ee;
+	const u32 delayus = 2000;
+	u32 retry = (mhi_cntrl->timeout_ms * 1000) / delayus;
+	const u32 rddm_timeout_us = 200000;
+	int rddm_retry = rddm_timeout_us / delayus;
+	void __iomem *base = mhi_cntrl->bhie;
+
+	dev_dbg(mhi_cntrl->dev,
+		"Entered with pm_state:%s dev_state:%s ee:%s\n",
+		to_mhi_pm_state_str(mhi_cntrl->pm_state),
+		TO_MHI_STATE_STR(mhi_cntrl->dev_state),
+		TO_MHI_EXEC_STR(mhi_cntrl->ee));
+
+	/*
+	 * This should only be executing during a kernel panic, we expect all
+	 * other cores to shutdown while we're collecting RDDM buffer. After
+	 * returning from this function, we expect the device to reset.
+	 *
+	 * Normaly, we read/write pm_state only after grabbing the
+	 * pm_lock, since we're in a panic, skipping it. Also there is no
+	 * gurantee that this state change would take effect since
+	 * we're setting it w/o grabbing pm_lock
+	 */
+	mhi_cntrl->pm_state = MHI_PM_LD_ERR_FATAL_DETECT;
+	/* update should take the effect immediately */
+	smp_wmb();
+
+	/*
+	 * Make sure device is not already in RDDM. In case the device asserts
+	 * and a kernel panic follows, device will already be in RDDM.
+	 * Do not trigger SYS ERR again and proceed with waiting for
+	 * image download completion.
+	 */
+	ee = mhi_get_exec_env(mhi_cntrl);
+	if (ee != MHI_EE_RDDM) {
+		dev_dbg(mhi_cntrl->dev,
+			"Trigger device into RDDM mode using SYS ERR\n");
+		mhi_set_mhi_state(mhi_cntrl, MHI_STATE_SYS_ERR);
+
+		dev_dbg(mhi_cntrl->dev, "Waiting for device to enter RDDM\n");
+		while (rddm_retry--) {
+			ee = mhi_get_exec_env(mhi_cntrl);
+			if (ee == MHI_EE_RDDM)
+				break;
+
+			udelay(delayus);
+		}
+
+		if (rddm_retry <= 0) {
+			/* Hardware reset so force device to enter RDDM */
+			dev_dbg(mhi_cntrl->dev,
+				"Did not enter RDDM, do a host req reset\n");
+			mhi_write_reg(mhi_cntrl, mhi_cntrl->regs,
+				      MHI_SOC_RESET_REQ_OFFSET,
+				      MHI_SOC_RESET_REQ);
+			udelay(delayus);
+		}
+
+		ee = mhi_get_exec_env(mhi_cntrl);
+	}
+
+	dev_dbg(mhi_cntrl->dev,
+		"Waiting for image download completion, current EE: %s\n",
+		TO_MHI_EXEC_STR(ee));
+
+	while (retry--) {
+		ret = mhi_read_reg_field(mhi_cntrl, base, BHIE_RXVECSTATUS_OFFS,
+					 BHIE_RXVECSTATUS_STATUS_BMSK,
+					 BHIE_RXVECSTATUS_STATUS_SHFT,
+					 &rx_status);
+		if (ret)
+			return -EIO;
+
+		if (rx_status == BHIE_RXVECSTATUS_STATUS_XFER_COMPL)
+			return 0;
+
+		udelay(delayus);
+	}
+
+	ee = mhi_get_exec_env(mhi_cntrl);
+	ret = mhi_read_reg(mhi_cntrl, base, BHIE_RXVECSTATUS_OFFS, &rx_status);
+
+	dev_err(mhi_cntrl->dev, "Did not complete RDDM transfer\n");
+	dev_err(mhi_cntrl->dev, "Current EE: %s\n", TO_MHI_EXEC_STR(ee));
+	dev_err(mhi_cntrl->dev, "RXVEC_STATUS: 0x%x\n", rx_status);
+
+	return -EIO;
+}
+
+/* Download RDDM image from device */
+int mhi_download_rddm_img(struct mhi_controller *mhi_cntrl, bool in_panic)
+{
+	void __iomem *base = mhi_cntrl->bhie;
+	u32 rx_status;
+
+	if (in_panic)
+		return __mhi_download_rddm_in_panic(mhi_cntrl);
+
+	/* Wait for the image download to complete */
+	wait_event_timeout(mhi_cntrl->state_event,
+			   mhi_read_reg_field(mhi_cntrl, base,
+					      BHIE_RXVECSTATUS_OFFS,
+					      BHIE_RXVECSTATUS_STATUS_BMSK,
+					      BHIE_RXVECSTATUS_STATUS_SHFT,
+					      &rx_status) || rx_status,
+			   msecs_to_jiffies(mhi_cntrl->timeout_ms));
+
+	return (rx_status == BHIE_RXVECSTATUS_STATUS_XFER_COMPL) ? 0 : -EIO;
+}
+EXPORT_SYMBOL_GPL(mhi_download_rddm_img);
+
 static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 			    const struct mhi_buf *mhi_buf)
 {
@@ -64,7 +216,6 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 	return (tx_status == BHIE_TXVECSTATUS_STATUS_XFER_COMPL) ? 0 : -EIO;
 }
 
-/* Download SBL image to device */
 static int mhi_fw_load_sbl(struct mhi_controller *mhi_cntrl,
 			   dma_addr_t dma_addr,
 			   size_t size)
@@ -90,7 +241,7 @@ static int mhi_fw_load_sbl(struct mhi_controller *mhi_cntrl,
 		goto invalid_pm_state;
 	}
 
-	/* Start SBL download via BHI protocol */
+	dev_dbg(mhi_cntrl->dev, "Starting SBL download via BHI\n");
 	mhi_write_reg(mhi_cntrl, base, BHI_STATUS, 0);
 	mhi_write_reg(mhi_cntrl, base, BHI_IMGADDR_HIGH,
 		      upper_32_bits(dma_addr));
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index 68071d0130f1..d0886de1c8bb 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -843,6 +843,7 @@ EXPORT_SYMBOL_GPL(mhi_unregister_controller);
 int mhi_prepare_for_power_up(struct mhi_controller *mhi_cntrl)
 {
 	int ret;
+	u32 bhie_off;
 
 	mutex_lock(&mhi_cntrl->pm_mutex);
 
@@ -850,12 +851,44 @@ int mhi_prepare_for_power_up(struct mhi_controller *mhi_cntrl)
 	if (ret)
 		goto error_dev_ctxt;
 
+	/*
+	 * Allocate RDDM table if specified, this table is for debugging purpose
+	 */
+	if (mhi_cntrl->rddm_size) {
+		mhi_alloc_bhie_table(mhi_cntrl, &mhi_cntrl->rddm_image,
+				     mhi_cntrl->rddm_size);
+
+		/*
+		 * This controller supports RDDM, so we need to manually clear
+		 * BHIE RX registers since POR values are undefined.
+		 */
+		ret = mhi_read_reg(mhi_cntrl, mhi_cntrl->regs, BHIEOFF,
+				   &bhie_off);
+		if (ret) {
+			dev_err(mhi_cntrl->dev, "Error getting BHIE offset\n");
+			goto bhie_error;
+		}
+
+		memset_io(mhi_cntrl->regs + bhie_off + BHIE_RXVECADDR_LOW_OFFS,
+			  0, BHIE_RXVECSTATUS_OFFS - BHIE_RXVECADDR_LOW_OFFS +
+			  4);
+
+		if (mhi_cntrl->rddm_image)
+			mhi_rddm_prepare(mhi_cntrl, mhi_cntrl->rddm_image);
+	}
+
 	mhi_cntrl->pre_init = true;
 
 	mutex_unlock(&mhi_cntrl->pm_mutex);
 
 	return 0;
 
+bhie_error:
+	if (mhi_cntrl->rddm_image) {
+		mhi_free_bhie_table(mhi_cntrl, mhi_cntrl->rddm_image);
+		mhi_cntrl->rddm_image = NULL;
+	}
+
 error_dev_ctxt:
 	mutex_unlock(&mhi_cntrl->pm_mutex);
 
@@ -870,6 +903,11 @@ void mhi_unprepare_after_power_down(struct mhi_controller *mhi_cntrl)
 		mhi_cntrl->fbc_image = NULL;
 	}
 
+	if (mhi_cntrl->rddm_image) {
+		mhi_free_bhie_table(mhi_cntrl, mhi_cntrl->rddm_image);
+		mhi_cntrl->rddm_image = NULL;
+	}
+
 	mhi_deinit_dev_ctxt(mhi_cntrl);
 	mhi_cntrl->pre_init = false;
 }
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index 01047a8c47e1..8ce1f11b8074 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -626,6 +626,8 @@ int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl);
 void mhi_deinit_dev_ctxt(struct mhi_controller *mhi_cntrl);
 int mhi_init_irq_setup(struct mhi_controller *mhi_cntrl);
 void mhi_deinit_free_irq(struct mhi_controller *mhi_cntrl);
+void mhi_rddm_prepare(struct mhi_controller *mhi_cntrl,
+		      struct image_info *img_info);
 
 /* Memory allocation methods */
 static inline void *mhi_alloc_coherent(struct mhi_controller *mhi_cntrl,
diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
index b4bedf139184..19987068b475 100644
--- a/drivers/bus/mhi/core/pm.c
+++ b/drivers/bus/mhi/core/pm.c
@@ -447,6 +447,16 @@ static void mhi_pm_disable_transition(struct mhi_controller *mhi_cntrl,
 
 	/* We must notify MHI control driver so it can clean up first */
 	if (transition_state == MHI_PM_SYS_ERR_PROCESS) {
+		/*
+		 * If controller supports RDDM, we do not process
+		 * SYS error state, instead we will jump directly
+		 * to RDDM state
+		 */
+		if (mhi_cntrl->rddm_image) {
+			dev_dbg(mhi_cntrl->dev,
+				 "Controller supports RDDM, so skip SYS_ERR\n");
+			return;
+		}
 		mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_SYS_ERROR);
 	}
 
@@ -894,3 +904,24 @@ int mhi_sync_power_up(struct mhi_controller *mhi_cntrl)
 	return (MHI_IN_MISSION_MODE(mhi_cntrl->ee)) ? 0 : -EIO;
 }
 EXPORT_SYMBOL(mhi_sync_power_up);
+
+int mhi_force_rddm_mode(struct mhi_controller *mhi_cntrl)
+{
+	int ret;
+
+	/* Check if device is already in RDDM */
+	if (mhi_cntrl->ee == MHI_EE_RDDM)
+		return 0;
+
+	dev_dbg(mhi_cntrl->dev, "Triggering SYS_ERR to force RDDM state\n");
+	mhi_set_mhi_state(mhi_cntrl, MHI_STATE_SYS_ERR);
+
+	/* Wait for RDDM event */
+	ret = wait_event_timeout(mhi_cntrl->state_event,
+				 mhi_cntrl->ee == MHI_EE_RDDM,
+				 msecs_to_jiffies(mhi_cntrl->timeout_ms));
+	ret = ret ? 0 : -EIO;
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_force_rddm_mode);
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index e6dc0a517b33..5ce99939a624 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -306,9 +306,11 @@ struct mhi_controller_config {
  * @iova_stop: IOMMU stop address for data (required)
  * @fw_image: Firmware image name for normal booting (required)
  * @edl_image: Firmware image name for emergency download mode (optional)
+ * @rddm_size: RAM dump size that host should allocate for debugging purpose 
  * @sbl_size: SBL image size downloaded through BHIe (optional)
  * @seg_len: BHIe vector size (optional)
  * @fbc_image: Points to firmware image buffer
+ * @rddm_image: Points to RAM dump buffer
  * @mhi_chan: Points to the channel configuration table
  * @lpm_chans: List of channels that require LPM notifications
  * @irq: base irq # to request (required)
@@ -362,9 +364,11 @@ struct mhi_controller {
 	dma_addr_t iova_stop;
 	const char *fw_image;
 	const char *edl_image;
+	size_t rddm_size;
 	size_t sbl_size;
 	size_t seg_len;
 	struct image_info *fbc_image;
+	struct image_info *rddm_image;
 	struct mhi_chan *mhi_chan;
 	struct list_head lpm_chans;
 	int *irq;
@@ -565,4 +569,24 @@ void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful);
  */
 void mhi_unprepare_after_power_down(struct mhi_controller *mhi_cntrl);
 
+/**
+ * mhi_download_rddm_img - Download ramdump image from device for
+ *                         debugging purpose.
+ * @mhi_cntrl: MHI controller
+ * @in_panic: Download rddm image during kernel panic
+ */
+int mhi_download_rddm_img(struct mhi_controller *mhi_cntrl, bool in_panic);
+
+/**
+ * mhi_force_rddm_mode - Force device into rddm mode
+ * @mhi_cntrl: MHI controller
+ */
+int mhi_force_rddm_mode(struct mhi_controller *mhi_cntrl);
+
+/**
+ * mhi_get_mhi_state - Get MHI state of the device
+ * @mhi_cntrl: MHI controller
+ */
+enum mhi_state mhi_get_mhi_state(struct mhi_controller *mhi_cntrl);
+
 #endif /* _MHI_H_ */
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 10/16] bus: mhi: core: Add support for processing events from client device
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (8 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-06 20:16   ` Jeffrey Hugo
  2020-01-31 13:50 ` [PATCH v2 11/16] bus: mhi: core: Add support for data transfer Manivannan Sadhasivam
                   ` (5 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

This commit adds support for processing the MHI data and control
events from the client device. The client device can report various
events such as EE events, state change events by interrupting the
host through IRQ and adding events to the event rings allocated by
the host during initialization.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/988

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted the data transfer patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/init.c     |  19 ++
 drivers/bus/mhi/core/internal.h |  10 +
 drivers/bus/mhi/core/main.c     | 468 ++++++++++++++++++++++++++++++++
 include/linux/mhi.h             |  14 +
 4 files changed, 511 insertions(+)

diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index d0886de1c8bb..1b5169e20e2b 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -544,6 +544,19 @@ static int parse_ev_cfg(struct mhi_controller *mhi_cntrl,
 
 		mhi_event->data_type = event_cfg->data_type;
 
+		switch (mhi_event->data_type) {
+		case MHI_ER_DATA:
+			mhi_event->process_event = mhi_process_data_event_ring;
+			break;
+		case MHI_ER_CTRL:
+			mhi_event->process_event = mhi_process_ctrl_ev_ring;
+			break;
+		default:
+			dev_err(mhi_cntrl->dev,
+				"Event Ring type not supported\n");
+			goto error_ev_cfg;
+		}
+
 		mhi_event->hw_ring = event_cfg->hardware_event;
 		if (mhi_event->hw_ring)
 			mhi_cntrl->hw_ev_rings++;
@@ -781,6 +794,12 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
 
 		mhi_event->mhi_cntrl = mhi_cntrl;
 		spin_lock_init(&mhi_event->lock);
+		if (mhi_event->data_type == MHI_ER_CTRL)
+			tasklet_init(&mhi_event->task, mhi_ctrl_ev_task,
+				     (ulong)mhi_event);
+		else
+			tasklet_init(&mhi_event->task, mhi_ev_task,
+				     (ulong)mhi_event);
 	}
 
 	mhi_chan = mhi_cntrl->mhi_chan;
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index 8ce1f11b8074..aa8a3433afae 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -500,6 +500,8 @@ struct mhi_buf_info {
 	dma_addr_t p_addr;
 	size_t len;
 	enum dma_data_direction dir;
+	bool used; /* Indicates whether the buffer is used or not */
+	bool pre_mapped; /* Already pre-mapped by client */
 };
 
 struct mhi_event {
@@ -648,6 +650,14 @@ static inline void mhi_free_coherent(struct mhi_controller *mhi_cntrl,
 	dma_free_coherent(mhi_cntrl->dev, size, vaddr, dma_handle);
 }
 
+/* Event processing methods */
+void mhi_ctrl_ev_task(unsigned long data);
+void mhi_ev_task(unsigned long data);
+int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl,
+				struct mhi_event *mhi_event, u32 event_quota);
+int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
+			     struct mhi_event *mhi_event, u32 event_quota);
+
 /* ISR handlers */
 irqreturn_t mhi_irq_handler(int irq_number, void *dev);
 irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *dev);
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
index 4d767dd97956..775d4a8fdefd 100644
--- a/drivers/bus/mhi/core/main.c
+++ b/drivers/bus/mhi/core/main.c
@@ -149,6 +149,16 @@ static void *mhi_to_virtual(struct mhi_ring *ring, dma_addr_t addr)
 	return (addr - ring->iommu_base) + ring->base;
 }
 
+static void mhi_del_ring_element(struct mhi_controller *mhi_cntrl,
+				 struct mhi_ring *ring)
+{
+	ring->rp += ring->el_size;
+	if (ring->rp >= (ring->base + ring->len))
+		ring->rp = ring->base;
+	/* smp update */
+	smp_wmb();
+}
+
 int mhi_destroy_device(struct device *dev, void *data)
 {
 	struct mhi_device *mhi_dev;
@@ -335,3 +345,461 @@ irqreturn_t mhi_intvec_handler(int irq_number, void *dev)
 
 	return IRQ_WAKE_THREAD;
 }
+
+static void mhi_recycle_ev_ring_element(struct mhi_controller *mhi_cntrl,
+					struct mhi_ring *ring)
+{
+	dma_addr_t ctxt_wp;
+
+	/* Update the WP */
+	ring->wp += ring->el_size;
+	ctxt_wp = *ring->ctxt_wp + ring->el_size;
+
+	if (ring->wp >= (ring->base + ring->len)) {
+		ring->wp = ring->base;
+		ctxt_wp = ring->iommu_base;
+	}
+
+	*ring->ctxt_wp = ctxt_wp;
+
+	/* Update the RP */
+	ring->rp += ring->el_size;
+	if (ring->rp >= (ring->base + ring->len))
+		ring->rp = ring->base;
+
+	/* Update to all cores */
+	smp_wmb();
+}
+
+static int parse_xfer_event(struct mhi_controller *mhi_cntrl,
+			    struct mhi_tre *event,
+			    struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *buf_ring, *tre_ring;
+	u32 ev_code;
+	struct mhi_result result;
+	unsigned long flags = 0;
+
+	ev_code = MHI_TRE_GET_EV_CODE(event);
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+
+	result.transaction_status = (ev_code == MHI_EV_CC_OVERFLOW) ?
+		-EOVERFLOW : 0;
+
+	/*
+	 * If it's a DB Event then we need to grab the lock
+	 * with preemption disabled and as a write because we
+	 * have to update db register and there are chances that
+	 * another thread could be doing the same.
+	 */
+	if (ev_code >= MHI_EV_CC_OOB)
+		write_lock_irqsave(&mhi_chan->lock, flags);
+	else
+		read_lock_bh(&mhi_chan->lock);
+
+	if (mhi_chan->ch_state != MHI_CH_STATE_ENABLED)
+		goto end_process_tx_event;
+
+	switch (ev_code) {
+	case MHI_EV_CC_OVERFLOW:
+	case MHI_EV_CC_EOB:
+	case MHI_EV_CC_EOT:
+	{
+		dma_addr_t ptr = MHI_TRE_GET_EV_PTR(event);
+		struct mhi_tre *local_rp, *ev_tre;
+		void *dev_rp;
+		struct mhi_buf_info *buf_info;
+		u16 xfer_len;
+
+		/* Get the TRB this event points to */
+		ev_tre = mhi_to_virtual(tre_ring, ptr);
+
+		/* device rp after servicing the TREs */
+		dev_rp = ev_tre + 1;
+		if (dev_rp >= (tre_ring->base + tre_ring->len))
+			dev_rp = tre_ring->base;
+
+		result.dir = mhi_chan->dir;
+
+		/* local rp */
+		local_rp = tre_ring->rp;
+		while (local_rp != dev_rp) {
+			buf_info = buf_ring->rp;
+			/* if it's last tre get len from the event */
+			if (local_rp == ev_tre)
+				xfer_len = MHI_TRE_GET_EV_LEN(event);
+			else
+				xfer_len = buf_info->len;
+
+			result.buf_addr = buf_info->cb_buf;
+			result.bytes_xferd = xfer_len;
+			mhi_del_ring_element(mhi_cntrl, buf_ring);
+			mhi_del_ring_element(mhi_cntrl, tre_ring);
+			local_rp = tre_ring->rp;
+
+			/* notify client */
+			mhi_chan->xfer_cb(mhi_chan->mhi_dev, &result);
+
+			if (mhi_chan->dir == DMA_TO_DEVICE)
+				atomic_dec(&mhi_cntrl->pending_pkts);
+		}
+		break;
+	} /* CC_EOT */
+	case MHI_EV_CC_OOB:
+	case MHI_EV_CC_DB_MODE:
+	{
+		unsigned long flags;
+
+		mhi_chan->db_cfg.db_mode = 1;
+		read_lock_irqsave(&mhi_cntrl->pm_lock, flags);
+		if (tre_ring->wp != tre_ring->rp &&
+		    MHI_DB_ACCESS_VALID(mhi_cntrl)) {
+			mhi_ring_chan_db(mhi_cntrl, mhi_chan);
+		}
+		read_unlock_irqrestore(&mhi_cntrl->pm_lock, flags);
+		break;
+	}
+	case MHI_EV_CC_BAD_TRE:
+	default:
+		dev_err(mhi_cntrl->dev, "Unknown event 0x%x\n", ev_code);
+		break;
+	} /* switch(MHI_EV_READ_CODE(EV_TRB_CODE,event)) */
+
+end_process_tx_event:
+	if (ev_code >= MHI_EV_CC_OOB)
+		write_unlock_irqrestore(&mhi_chan->lock, flags);
+	else
+		read_unlock_bh(&mhi_chan->lock);
+
+	return 0;
+}
+
+static int parse_rsc_event(struct mhi_controller *mhi_cntrl,
+			   struct mhi_tre *event,
+			   struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *buf_ring, *tre_ring;
+	struct mhi_buf_info *buf_info;
+	struct mhi_result result;
+	int ev_code;
+	u32 cookie; /* offset to local descriptor */
+	u16 xfer_len;
+
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+
+	ev_code = MHI_TRE_GET_EV_CODE(event);
+	cookie = MHI_TRE_GET_EV_COOKIE(event);
+	xfer_len = MHI_TRE_GET_EV_LEN(event);
+
+	/* Received out of bound cookie */
+	WARN_ON(cookie >= buf_ring->len);
+
+	buf_info = buf_ring->base + cookie;
+
+	result.transaction_status = (ev_code == MHI_EV_CC_OVERFLOW) ?
+		-EOVERFLOW : 0;
+	result.bytes_xferd = xfer_len;
+	result.buf_addr = buf_info->cb_buf;
+	result.dir = mhi_chan->dir;
+
+	read_lock_bh(&mhi_chan->lock);
+
+	if (mhi_chan->ch_state != MHI_CH_STATE_ENABLED)
+		goto end_process_rsc_event;
+
+	WARN_ON(!buf_info->used);
+
+	/* notify the client */
+	mhi_chan->xfer_cb(mhi_chan->mhi_dev, &result);
+
+	/*
+	 * Note: We're arbitrarily incrementing RP even though, completion
+	 * packet we processed might not be the same one, reason we can do this
+	 * is because device guaranteed to cache descriptors in order it
+	 * receive, so even though completion event is different we can re-use
+	 * all descriptors in between.
+	 * Example:
+	 * Transfer Ring has descriptors: A, B, C, D
+	 * Last descriptor host queue is D (WP) and first descriptor
+	 * host queue is A (RP).
+	 * The completion event we just serviced is descriptor C.
+	 * Then we can safely queue descriptors to replace A, B, and C
+	 * even though host did not receive any completions.
+	 */
+	mhi_del_ring_element(mhi_cntrl, tre_ring);
+	buf_info->used = false;
+
+end_process_rsc_event:
+	read_unlock_bh(&mhi_chan->lock);
+
+	return 0;
+}
+
+static void mhi_process_cmd_completion(struct mhi_controller *mhi_cntrl,
+				       struct mhi_tre *tre)
+{
+	dma_addr_t ptr = MHI_TRE_GET_EV_PTR(tre);
+	struct mhi_cmd *cmd_ring = &mhi_cntrl->mhi_cmd[PRIMARY_CMD_RING];
+	struct mhi_ring *mhi_ring = &cmd_ring->ring;
+	struct mhi_tre *cmd_pkt;
+	struct mhi_chan *mhi_chan;
+	u32 chan;
+
+	cmd_pkt = mhi_to_virtual(mhi_ring, ptr);
+
+	chan = MHI_TRE_GET_CMD_CHID(cmd_pkt);
+	mhi_chan = &mhi_cntrl->mhi_chan[chan];
+	write_lock_bh(&mhi_chan->lock);
+	mhi_chan->ccs = MHI_TRE_GET_EV_CODE(tre);
+	complete(&mhi_chan->completion);
+	write_unlock_bh(&mhi_chan->lock);
+
+	mhi_del_ring_element(mhi_cntrl, mhi_ring);
+}
+
+int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
+			     struct mhi_event *mhi_event,
+			     u32 event_quota)
+{
+	struct mhi_tre *dev_rp, *local_rp;
+	struct mhi_ring *ev_ring = &mhi_event->ring;
+	struct mhi_event_ctxt *er_ctxt =
+		&mhi_cntrl->mhi_ctxt->er_ctxt[mhi_event->er_index];
+	int count = 0;
+	u32 chan;
+	struct mhi_chan *mhi_chan;
+
+	/*
+	 * This is a quick check to avoid unnecessary event processing
+	 * in case MHI is already in error state, but it's still possible
+	 * to transition to error state while processing events
+	 */
+	if (unlikely(MHI_EVENT_ACCESS_INVALID(mhi_cntrl->pm_state)))
+		return -EIO;
+
+	dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
+	local_rp = ev_ring->rp;
+
+	while (dev_rp != local_rp) {
+		enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp);
+
+		switch (type) {
+		case MHI_PKT_TYPE_BW_REQ_EVENT:
+		{
+			struct mhi_link_info *link_info;
+
+			link_info = &mhi_cntrl->mhi_link_info;
+			write_lock_irq(&mhi_cntrl->pm_lock);
+			link_info->target_link_speed =
+				MHI_TRE_GET_EV_LINKSPEED(local_rp);
+			link_info->target_link_width =
+				MHI_TRE_GET_EV_LINKWIDTH(local_rp);
+			write_unlock_irq(&mhi_cntrl->pm_lock);
+			dev_dbg(mhi_cntrl->dev, "Received BW_REQ event\n");
+			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_BW_REQ);
+			break;
+		}
+		case MHI_PKT_TYPE_STATE_CHANGE_EVENT:
+		{
+			enum mhi_state new_state;
+
+			new_state = MHI_TRE_GET_EV_STATE(local_rp);
+
+			dev_dbg(mhi_cntrl->dev,
+				 "State change event to state: %s\n",
+				 TO_MHI_STATE_STR(new_state));
+
+			switch (new_state) {
+			case MHI_STATE_M0:
+				mhi_pm_m0_transition(mhi_cntrl);
+				break;
+			case MHI_STATE_M1:
+				mhi_pm_m1_transition(mhi_cntrl);
+				break;
+			case MHI_STATE_M3:
+				mhi_pm_m3_transition(mhi_cntrl);
+				break;
+			case MHI_STATE_SYS_ERR:
+			{
+				enum mhi_pm_state new_state;
+
+				dev_dbg(mhi_cntrl->dev,
+					 "System error detected\n");
+				write_lock_irq(&mhi_cntrl->pm_lock);
+				new_state = mhi_tryset_pm_state(mhi_cntrl,
+							MHI_PM_SYS_ERR_DETECT);
+				write_unlock_irq(&mhi_cntrl->pm_lock);
+				if (new_state == MHI_PM_SYS_ERR_DETECT)
+					schedule_work(&mhi_cntrl->syserr_worker);
+				break;
+			}
+			default:
+				dev_err(mhi_cntrl->dev, "Invalid state: %s\n",
+					TO_MHI_STATE_STR(new_state));
+			}
+
+			break;
+		}
+		case MHI_PKT_TYPE_CMD_COMPLETION_EVENT:
+			mhi_process_cmd_completion(mhi_cntrl, local_rp);
+			break;
+		case MHI_PKT_TYPE_EE_EVENT:
+		{
+			enum dev_st_transition st = DEV_ST_TRANSITION_MAX;
+			enum mhi_ee_type event = MHI_TRE_GET_EV_EXECENV(local_rp);
+
+			dev_dbg(mhi_cntrl->dev, "Received EE event: %s\n",
+				 TO_MHI_EXEC_STR(event));
+			switch (event) {
+			case MHI_EE_SBL:
+				st = DEV_ST_TRANSITION_SBL;
+				break;
+			case MHI_EE_WFW:
+			case MHI_EE_AMSS:
+				st = DEV_ST_TRANSITION_MISSION_MODE;
+				break;
+			case MHI_EE_RDDM:
+				mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM);
+				write_lock_irq(&mhi_cntrl->pm_lock);
+				mhi_cntrl->ee = event;
+				write_unlock_irq(&mhi_cntrl->pm_lock);
+				wake_up_all(&mhi_cntrl->state_event);
+				break;
+			default:
+				dev_err(mhi_cntrl->dev,
+					"Unhandled EE event: 0x%x\n", type);
+			}
+			if (st != DEV_ST_TRANSITION_MAX)
+				mhi_queue_state_transition(mhi_cntrl, st);
+
+			break;
+		}
+		case MHI_PKT_TYPE_TX_EVENT:
+			chan = MHI_TRE_GET_EV_CHID(local_rp);
+			mhi_chan = &mhi_cntrl->mhi_chan[chan];
+			parse_xfer_event(mhi_cntrl, local_rp, mhi_chan);
+			event_quota--;
+			break;
+		default:
+			dev_err(mhi_cntrl->dev, "Unhandled event type: %d\n",
+				type);
+			break;
+		}
+
+		mhi_recycle_ev_ring_element(mhi_cntrl, ev_ring);
+		local_rp = ev_ring->rp;
+		dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
+		count++;
+	}
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl)))
+		mhi_ring_er_db(mhi_event);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return count;
+}
+
+int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl,
+				struct mhi_event *mhi_event,
+				u32 event_quota)
+{
+	struct mhi_tre *dev_rp, *local_rp;
+	struct mhi_ring *ev_ring = &mhi_event->ring;
+	struct mhi_event_ctxt *er_ctxt =
+		&mhi_cntrl->mhi_ctxt->er_ctxt[mhi_event->er_index];
+	int count = 0;
+	u32 chan;
+	struct mhi_chan *mhi_chan;
+
+	if (unlikely(MHI_EVENT_ACCESS_INVALID(mhi_cntrl->pm_state)))
+		return -EIO;
+
+	dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
+	local_rp = ev_ring->rp;
+
+	while (dev_rp != local_rp && event_quota > 0) {
+		enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp);
+
+		chan = MHI_TRE_GET_EV_CHID(local_rp);
+		mhi_chan = &mhi_cntrl->mhi_chan[chan];
+
+		if (likely(type == MHI_PKT_TYPE_TX_EVENT)) {
+			parse_xfer_event(mhi_cntrl, local_rp, mhi_chan);
+			event_quota--;
+		} else if (type == MHI_PKT_TYPE_RSC_TX_EVENT) {
+			parse_rsc_event(mhi_cntrl, local_rp, mhi_chan);
+			event_quota--;
+		}
+
+		mhi_recycle_ev_ring_element(mhi_cntrl, ev_ring);
+		local_rp = ev_ring->rp;
+		dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
+		count++;
+	}
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl)))
+		mhi_ring_er_db(mhi_event);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return count;
+}
+
+void mhi_ev_task(unsigned long data)
+{
+	struct mhi_event *mhi_event = (struct mhi_event *)data;
+	struct mhi_controller *mhi_cntrl = mhi_event->mhi_cntrl;
+
+	/* process all pending events */
+	spin_lock_bh(&mhi_event->lock);
+	mhi_event->process_event(mhi_cntrl, mhi_event, U32_MAX);
+	spin_unlock_bh(&mhi_event->lock);
+}
+
+void mhi_ctrl_ev_task(unsigned long data)
+{
+	struct mhi_event *mhi_event = (struct mhi_event *)data;
+	struct mhi_controller *mhi_cntrl = mhi_event->mhi_cntrl;
+	enum mhi_state state;
+	enum mhi_pm_state pm_state = 0;
+	int ret;
+
+	/*
+	 * We can check PM state w/o a lock here because there is no way
+	 * PM state can change from reg access valid to no access while this
+	 * thread being executed.
+	 */
+	if (!MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
+		/*
+		 * We may have a pending event but not allowed to
+		 * process it since we are probably in a suspended state,
+		 * so trigger a resume.
+		 */
+		mhi_cntrl->runtime_get(mhi_cntrl);
+		mhi_cntrl->runtime_put(mhi_cntrl);
+
+		return;
+	}
+
+	/* Process ctrl events events */
+	ret = mhi_event->process_event(mhi_cntrl, mhi_event, U32_MAX);
+
+	/*
+	 * We received an IRQ but no events to process, maybe device went to
+	 * SYS_ERR state? Check the state to confirm.
+	 */
+	if (!ret) {
+		write_lock_irq(&mhi_cntrl->pm_lock);
+		state = mhi_get_mhi_state(mhi_cntrl);
+		if (state == MHI_STATE_SYS_ERR) {
+			dev_dbg(mhi_cntrl->dev, "System error detected\n");
+			pm_state = mhi_tryset_pm_state(mhi_cntrl,
+						       MHI_PM_SYS_ERR_DETECT);
+		}
+		write_unlock_irq(&mhi_cntrl->pm_lock);
+		if (pm_state == MHI_PM_SYS_ERR_DETECT)
+			schedule_work(&mhi_cntrl->syserr_worker);
+	}
+}
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 5ce99939a624..2ea2deb922a1 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -31,6 +31,7 @@ struct mhi_buf_info;
  * @MHI_CB_EE_MISSION_MODE: MHI device entered Mission Mode exec env
  * @MHI_CB_SYS_ERROR: MHI device entered error state (may recover)
  * @MHI_CB_FATAL_ERROR: MHI device entered fatal error state
+ * @MHI_CB_BW_REQ: Received a bandwidth switch request from device
  */
 enum mhi_callback {
 	MHI_CB_IDLE,
@@ -41,6 +42,7 @@ enum mhi_callback {
 	MHI_CB_EE_MISSION_MODE,
 	MHI_CB_SYS_ERROR,
 	MHI_CB_FATAL_ERROR,
+	MHI_CB_BW_REQ,
 };
 
 /**
@@ -94,6 +96,16 @@ struct image_info {
 	u32 entries;
 };
 
+/**
+ * struct mhi_link_info - BW requirement
+ * target_link_speed - Link speed as defined by TLS bits in LinkControl reg
+ * target_link_width - Link width as defined by NLW bits in LinkStatus reg
+ */
+struct mhi_link_info {
+	unsigned int target_link_speed;
+	unsigned int target_link_width;
+};
+
 /**
  * enum mhi_ee_type - Execution environment types
  * @MHI_EE_PBL: Primary Bootloader
@@ -335,6 +347,7 @@ struct mhi_controller_config {
  * @transition_list: List of MHI state transitions
  * @transition_lock: Lock for protecting MHI state transition list
  * @wlock: Lock for protecting device wakeup
+ * @mhi_link_info: Device bandwidth info
  * @st_worker: State transition worker
  * @fw_worker: Firmware download worker
  * @syserr_worker: System error worker
@@ -395,6 +408,7 @@ struct mhi_controller {
 	struct list_head transition_list;
 	spinlock_t transition_lock;
 	spinlock_t wlock;
+	struct mhi_link_info mhi_link_info;
 	struct work_struct st_worker;
 	struct work_struct fw_worker;
 	struct work_struct syserr_worker;
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 11/16] bus: mhi: core: Add support for data transfer
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (9 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 10/16] bus: mhi: core: Add support for processing events from client device Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-06 20:16   ` Jeffrey Hugo
  2020-02-17 16:13   ` Arnd Bergmann
  2020-01-31 13:50 ` [PATCH v2 12/16] bus: mhi: core: Add uevent support for module autoloading Manivannan Sadhasivam
                   ` (4 subsequent siblings)
  15 siblings, 2 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

Add support for transferring data between external modem and host
processor using MHI protocol.

This is based on the patch submitted by Sujeev Dias:
https://lkml.org/lkml/2018/7/9/988

Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
[mani: splitted the data transfer patch and cleaned up for upstream]
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/init.c     | 162 ++++++-
 drivers/bus/mhi/core/internal.h |  33 ++
 drivers/bus/mhi/core/main.c     | 777 +++++++++++++++++++++++++++++++-
 drivers/bus/mhi/core/pm.c       |  40 ++
 include/linux/mhi.h             |  55 +++
 5 files changed, 1058 insertions(+), 9 deletions(-)

diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index 1b5169e20e2b..4d0e5bbb3389 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -493,6 +493,73 @@ int mhi_init_mmio(struct mhi_controller *mhi_cntrl)
 	return 0;
 }
 
+void mhi_deinit_chan_ctxt(struct mhi_controller *mhi_cntrl,
+			  struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *buf_ring;
+	struct mhi_ring *tre_ring;
+	struct mhi_chan_ctxt *chan_ctxt;
+
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+	chan_ctxt = &mhi_cntrl->mhi_ctxt->chan_ctxt[mhi_chan->chan];
+
+	mhi_free_coherent(mhi_cntrl, tre_ring->alloc_size,
+			  tre_ring->pre_aligned, tre_ring->dma_handle);
+	vfree(buf_ring->base);
+
+	buf_ring->base = tre_ring->base = NULL;
+	chan_ctxt->rbase = 0;
+}
+
+int mhi_init_chan_ctxt(struct mhi_controller *mhi_cntrl,
+		       struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *buf_ring;
+	struct mhi_ring *tre_ring;
+	struct mhi_chan_ctxt *chan_ctxt;
+	u32 tmp;
+	int ret;
+
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+	tre_ring->el_size = sizeof(struct mhi_tre);
+	tre_ring->len = tre_ring->el_size * tre_ring->elements;
+	chan_ctxt = &mhi_cntrl->mhi_ctxt->chan_ctxt[mhi_chan->chan];
+	ret = mhi_alloc_aligned_ring(mhi_cntrl, tre_ring, tre_ring->len);
+	if (ret)
+		return -ENOMEM;
+
+	buf_ring->el_size = sizeof(struct mhi_buf_info);
+	buf_ring->len = buf_ring->el_size * buf_ring->elements;
+	buf_ring->base = vzalloc(buf_ring->len);
+
+	if (!buf_ring->base) {
+		mhi_free_coherent(mhi_cntrl, tre_ring->alloc_size,
+				  tre_ring->pre_aligned, tre_ring->dma_handle);
+		return -ENOMEM;
+	}
+
+	tmp = chan_ctxt->chcfg;
+	tmp &= ~CHAN_CTX_CHSTATE_MASK;
+	tmp |= (MHI_CH_STATE_ENABLED << CHAN_CTX_CHSTATE_SHIFT);
+	chan_ctxt->chcfg = tmp;
+
+	chan_ctxt->rbase = tre_ring->iommu_base;
+	chan_ctxt->rp = chan_ctxt->wp = chan_ctxt->rbase;
+	chan_ctxt->rlen = tre_ring->len;
+	tre_ring->ctxt_wp = &chan_ctxt->wp;
+
+	tre_ring->rp = tre_ring->wp = tre_ring->base;
+	buf_ring->rp = buf_ring->wp = buf_ring->base;
+	mhi_chan->db_cfg.db_mode = 1;
+
+	/* Update to all cores */
+	smp_wmb();
+
+	return 0;
+}
+
 static int parse_ev_cfg(struct mhi_controller *mhi_cntrl,
 			struct mhi_controller_config *config)
 {
@@ -648,6 +715,31 @@ static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
 		mhi_chan->db_cfg.pollcfg = ch_cfg->pollcfg;
 		mhi_chan->xfer_type = ch_cfg->data_type;
 
+		switch (mhi_chan->xfer_type) {
+		case MHI_BUF_RAW:
+			mhi_chan->gen_tre = mhi_gen_tre;
+			mhi_chan->queue_xfer = mhi_queue_buf;
+			break;
+		case MHI_BUF_SKB:
+			mhi_chan->queue_xfer = mhi_queue_skb;
+			break;
+		case MHI_BUF_SCLIST:
+			mhi_chan->gen_tre = mhi_gen_tre;
+			mhi_chan->queue_xfer = mhi_queue_sclist;
+			break;
+		case MHI_BUF_NOP:
+			mhi_chan->queue_xfer = mhi_queue_nop;
+			break;
+		case MHI_BUF_DMA:
+		case MHI_BUF_RSC_DMA:
+			mhi_chan->queue_xfer = mhi_queue_dma;
+			break;
+		default:
+			dev_err(mhi_cntrl->dev,
+				"Channel datatype not supported\n");
+			goto error_chan_cfg;
+		}
+
 		mhi_chan->lpm_notify = ch_cfg->lpm_notify;
 		mhi_chan->offload_ch = ch_cfg->offload_channel;
 		mhi_chan->db_cfg.reset_req = ch_cfg->doorbell_mode_switch;
@@ -677,6 +769,13 @@ static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
 			goto error_chan_cfg;
 		}
 
+		/*
+		 * If MHI host pre-allocates buffers then client drivers
+		 * cannot queue
+		 */
+		if (mhi_chan->pre_alloc)
+			mhi_chan->queue_xfer = mhi_queue_nop;
+
 		if (!mhi_chan->offload_ch) {
 			mhi_chan->db_cfg.brstmode = ch_cfg->doorbell;
 			if (MHI_INVALID_BRSTMODE(mhi_chan->db_cfg.brstmode)) {
@@ -809,6 +908,14 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
 		rwlock_init(&mhi_chan->lock);
 	}
 
+	if (mhi_cntrl->bounce_buf) {
+		mhi_cntrl->map_single = mhi_map_single_use_bb;
+		mhi_cntrl->unmap_single = mhi_unmap_single_use_bb;
+	} else {
+		mhi_cntrl->map_single = mhi_map_single_no_bb;
+		mhi_cntrl->unmap_single = mhi_unmap_single_no_bb;
+	}
+
 	/* Register controller with MHI bus */
 	mhi_dev = mhi_alloc_device(mhi_cntrl);
 	if (IS_ERR(mhi_dev)) {
@@ -974,6 +1081,14 @@ static int mhi_driver_probe(struct device *dev)
 	struct mhi_event *mhi_event;
 	struct mhi_chan *ul_chan = mhi_dev->ul_chan;
 	struct mhi_chan *dl_chan = mhi_dev->dl_chan;
+	int ret;
+
+	/* Bring device out of LPM */
+	ret = mhi_device_get_sync(mhi_dev);
+	if (ret)
+		return ret;
+
+	ret = -EINVAL;
 
 	if (ul_chan) {
 		/*
@@ -981,13 +1096,18 @@ static int mhi_driver_probe(struct device *dev)
 		 * be provided
 		 */
 		if (ul_chan->lpm_notify && !mhi_drv->status_cb)
-			return -EINVAL;
+			goto exit_probe;
 
 		/* For non-offload channels then xfer_cb should be provided */
 		if (!ul_chan->offload_ch && !mhi_drv->ul_xfer_cb)
-			return -EINVAL;
+			goto exit_probe;
 
 		ul_chan->xfer_cb = mhi_drv->ul_xfer_cb;
+		if (ul_chan->auto_start) {
+			ret = mhi_prepare_channel(mhi_cntrl, ul_chan);
+			if (ret)
+				goto exit_probe;
+		}
 	}
 
 	if (dl_chan) {
@@ -996,11 +1116,11 @@ static int mhi_driver_probe(struct device *dev)
 		 * be provided
 		 */
 		if (dl_chan->lpm_notify && !mhi_drv->status_cb)
-			return -EINVAL;
+			goto exit_probe;
 
 		/* For non-offload channels then xfer_cb should be provided */
 		if (!dl_chan->offload_ch && !mhi_drv->dl_xfer_cb)
-			return -EINVAL;
+			goto exit_probe;
 
 		mhi_event = &mhi_cntrl->mhi_event[dl_chan->er_index];
 
@@ -1010,19 +1130,36 @@ static int mhi_driver_probe(struct device *dev)
 		 * notify pending data
 		 */
 		if (mhi_event->cl_manage && !mhi_drv->status_cb)
-			return -EINVAL;
+			goto exit_probe;
 
 		dl_chan->xfer_cb = mhi_drv->dl_xfer_cb;
 	}
 
 	/* Call the user provided probe function */
-	return mhi_drv->probe(mhi_dev, mhi_dev->id);
+	ret = mhi_drv->probe(mhi_dev, mhi_dev->id);
+	if (ret)
+		goto exit_probe;
+
+	if (dl_chan && dl_chan->auto_start)
+		mhi_prepare_channel(mhi_cntrl, dl_chan);
+
+	mhi_device_put(mhi_dev);
+
+	return ret;
+
+exit_probe:
+	mhi_unprepare_from_transfer(mhi_dev);
+
+	mhi_device_put(mhi_dev);
+
+	return ret;
 }
 
 static int mhi_driver_remove(struct device *dev)
 {
 	struct mhi_device *mhi_dev = to_mhi_device(dev);
 	struct mhi_driver *mhi_drv = to_mhi_driver(dev->driver);
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
 	struct mhi_chan *mhi_chan;
 	enum mhi_ch_state ch_state[] = {
 		MHI_CH_STATE_DISABLED,
@@ -1054,6 +1191,10 @@ static int mhi_driver_remove(struct device *dev)
 		mhi_chan->ch_state = MHI_CH_STATE_SUSPENDED;
 		write_unlock_irq(&mhi_chan->lock);
 
+		/* Reset the non-offload channel */
+		if (!mhi_chan->offload_ch)
+			mhi_reset_chan(mhi_cntrl, mhi_chan);
+
 		mutex_unlock(&mhi_chan->mutex);
 	}
 
@@ -1068,11 +1209,20 @@ static int mhi_driver_remove(struct device *dev)
 
 		mutex_lock(&mhi_chan->mutex);
 
+		if (ch_state[dir] == MHI_CH_STATE_ENABLED &&
+		    !mhi_chan->offload_ch)
+			mhi_deinit_chan_ctxt(mhi_cntrl, mhi_chan);
+
 		mhi_chan->ch_state = MHI_CH_STATE_DISABLED;
 
 		mutex_unlock(&mhi_chan->mutex);
 	}
 
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	while (mhi_dev->dev_wake)
+		mhi_device_put(mhi_dev);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
 	return 0;
 }
 
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index aa8a3433afae..0b95a1ab6dfd 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -599,6 +599,8 @@ int mhi_pm_m0_transition(struct mhi_controller *mhi_cntrl);
 void mhi_pm_m1_transition(struct mhi_controller *mhi_cntrl);
 int mhi_pm_m3_transition(struct mhi_controller *mhi_cntrl);
 int __mhi_device_get_sync(struct mhi_controller *mhi_cntrl);
+int mhi_send_cmd(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan,
+		 enum mhi_cmd_type cmd);
 
 /* Register access methods */
 void mhi_db_brstmode(struct mhi_controller *mhi_cntrl, struct db_cfg *db_cfg,
@@ -630,6 +632,14 @@ int mhi_init_irq_setup(struct mhi_controller *mhi_cntrl);
 void mhi_deinit_free_irq(struct mhi_controller *mhi_cntrl);
 void mhi_rddm_prepare(struct mhi_controller *mhi_cntrl,
 		      struct image_info *img_info);
+int mhi_prepare_channel(struct mhi_controller *mhi_cntrl,
+			struct mhi_chan *mhi_chan);
+int mhi_init_chan_ctxt(struct mhi_controller *mhi_cntrl,
+		       struct mhi_chan *mhi_chan);
+void mhi_deinit_chan_ctxt(struct mhi_controller *mhi_cntrl,
+			  struct mhi_chan *mhi_chan);
+void mhi_reset_chan(struct mhi_controller *mhi_cntrl,
+		    struct mhi_chan *mhi_chan);
 
 /* Memory allocation methods */
 static inline void *mhi_alloc_coherent(struct mhi_controller *mhi_cntrl,
@@ -663,4 +673,27 @@ irqreturn_t mhi_irq_handler(int irq_number, void *dev);
 irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *dev);
 irqreturn_t mhi_intvec_handler(int irq_number, void *dev);
 
+/* Queue transfer methods */
+int mhi_queue_dma(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags);
+int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan,
+		void *buf, void *cb, size_t buf_len, enum mhi_flags flags);
+int mhi_queue_buf(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags);
+int mhi_queue_skb(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags);
+int mhi_queue_sclist(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags);
+int mhi_queue_nop(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags);
+
+int mhi_map_single_no_bb(struct mhi_controller *mhi_cntrl,
+			 struct mhi_buf_info *buf_info);
+int mhi_map_single_use_bb(struct mhi_controller *mhi_cntrl,
+			  struct mhi_buf_info *buf_info);
+void mhi_unmap_single_no_bb(struct mhi_controller *mhi_cntrl,
+			    struct mhi_buf_info *buf_info);
+void mhi_unmap_single_use_bb(struct mhi_controller *mhi_cntrl,
+			     struct mhi_buf_info *buf_info);
+
 #endif /* _MHI_INT_H */
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
index 775d4a8fdefd..9891a5c0ce2c 100644
--- a/drivers/bus/mhi/core/main.c
+++ b/drivers/bus/mhi/core/main.c
@@ -144,11 +144,82 @@ enum mhi_state mhi_get_mhi_state(struct mhi_controller *mhi_cntrl)
 	return ret ? MHI_STATE_MAX : state;
 }
 
+int mhi_map_single_no_bb(struct mhi_controller *mhi_cntrl,
+			 struct mhi_buf_info *buf_info)
+{
+	buf_info->p_addr = dma_map_single(mhi_cntrl->dev, buf_info->v_addr,
+					  buf_info->len, buf_info->dir);
+	if (dma_mapping_error(mhi_cntrl->dev, buf_info->p_addr))
+		return -ENOMEM;
+
+	return 0;
+}
+
+int mhi_map_single_use_bb(struct mhi_controller *mhi_cntrl,
+			  struct mhi_buf_info *buf_info)
+{
+	void *buf = mhi_alloc_coherent(mhi_cntrl, buf_info->len,
+				       &buf_info->p_addr, GFP_ATOMIC);
+
+	if (!buf)
+		return -ENOMEM;
+
+	if (buf_info->dir == DMA_TO_DEVICE)
+		memcpy(buf, buf_info->v_addr, buf_info->len);
+
+	buf_info->bb_addr = buf;
+
+	return 0;
+}
+
+void mhi_unmap_single_no_bb(struct mhi_controller *mhi_cntrl,
+			    struct mhi_buf_info *buf_info)
+{
+	dma_unmap_single(mhi_cntrl->dev, buf_info->p_addr, buf_info->len,
+			 buf_info->dir);
+}
+
+void mhi_unmap_single_use_bb(struct mhi_controller *mhi_cntrl,
+			     struct mhi_buf_info *buf_info)
+{
+	if (buf_info->dir == DMA_FROM_DEVICE)
+		memcpy(buf_info->v_addr, buf_info->bb_addr, buf_info->len);
+
+	mhi_free_coherent(mhi_cntrl, buf_info->len, buf_info->bb_addr,
+			  buf_info->p_addr);
+}
+
+static int get_nr_avail_ring_elements(struct mhi_controller *mhi_cntrl,
+				      struct mhi_ring *ring)
+{
+	int nr_el;
+
+	if (ring->wp < ring->rp) {
+		nr_el = ((ring->rp - ring->wp) / ring->el_size) - 1;
+	} else {
+		nr_el = (ring->rp - ring->base) / ring->el_size;
+		nr_el += ((ring->base + ring->len - ring->wp) /
+			  ring->el_size) - 1;
+	}
+
+	return nr_el;
+}
+
 static void *mhi_to_virtual(struct mhi_ring *ring, dma_addr_t addr)
 {
 	return (addr - ring->iommu_base) + ring->base;
 }
 
+static void mhi_add_ring_element(struct mhi_controller *mhi_cntrl,
+				 struct mhi_ring *ring)
+{
+	ring->wp += ring->el_size;
+	if (ring->wp >= (ring->base + ring->len))
+		ring->wp = ring->base;
+	/* smp update */
+	smp_wmb();
+}
+
 static void mhi_del_ring_element(struct mhi_controller *mhi_cntrl,
 				 struct mhi_ring *ring)
 {
@@ -415,23 +486,25 @@ static int parse_xfer_event(struct mhi_controller *mhi_cntrl,
 		/* Get the TRB this event points to */
 		ev_tre = mhi_to_virtual(tre_ring, ptr);
 
-		/* device rp after servicing the TREs */
 		dev_rp = ev_tre + 1;
 		if (dev_rp >= (tre_ring->base + tre_ring->len))
 			dev_rp = tre_ring->base;
 
 		result.dir = mhi_chan->dir;
 
-		/* local rp */
 		local_rp = tre_ring->rp;
 		while (local_rp != dev_rp) {
 			buf_info = buf_ring->rp;
-			/* if it's last tre get len from the event */
+			/* If it's the last TRE, get length from the event */
 			if (local_rp == ev_tre)
 				xfer_len = MHI_TRE_GET_EV_LEN(event);
 			else
 				xfer_len = buf_info->len;
 
+			/* Unmap if it's not pre-mapped by client */
+			if (likely(!buf_info->pre_mapped))
+				mhi_cntrl->unmap_single(mhi_cntrl, buf_info);
+
 			result.buf_addr = buf_info->cb_buf;
 			result.bytes_xferd = xfer_len;
 			mhi_del_ring_element(mhi_cntrl, buf_ring);
@@ -443,6 +516,22 @@ static int parse_xfer_event(struct mhi_controller *mhi_cntrl,
 
 			if (mhi_chan->dir == DMA_TO_DEVICE)
 				atomic_dec(&mhi_cntrl->pending_pkts);
+
+			/*
+			 * Recycle the buffer if buffer is pre-allocated,
+			 * if there is an error, not much we can do apart
+			 * from dropping the packet
+			 */
+			if (mhi_chan->pre_alloc) {
+				if (mhi_queue_buf(mhi_chan->mhi_dev, mhi_chan,
+						  buf_info->cb_buf,
+						  buf_info->len, MHI_EOT)) {
+					dev_err(mhi_cntrl->dev,
+						"Error recycling buffer for chan:%d\n",
+						mhi_chan->chan);
+					kfree(buf_info->cb_buf);
+				}
+			}
 		}
 		break;
 	} /* CC_EOT */
@@ -803,3 +892,685 @@ void mhi_ctrl_ev_task(unsigned long data)
 			schedule_work(&mhi_cntrl->syserr_worker);
 	}
 }
+
+static bool mhi_is_ring_full(struct mhi_controller *mhi_cntrl,
+			     struct mhi_ring *ring)
+{
+	void *tmp = ring->wp + ring->el_size;
+
+	if (tmp >= (ring->base + ring->len))
+		tmp = ring->base;
+
+	return (tmp == ring->rp);
+}
+
+/* TODO: Scatter-Gather transfer not implemented */
+int mhi_queue_sclist(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		     void *buf, size_t len, enum mhi_flags mflags)
+{
+	return -EINVAL;
+}
+
+/*
+ * MHI client drivers are not allowed to queue buffer for pre allocated
+ * channels. Hence, this function just returns -EINVAL.
+ */
+int mhi_queue_nop(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags)
+{
+	return -EINVAL;
+}
+
+int mhi_queue_skb(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags)
+{
+	struct sk_buff *skb = buf;
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_ring *tre_ring = &mhi_chan->tre_ring;
+	struct mhi_ring *buf_ring = &mhi_chan->buf_ring;
+	struct mhi_buf_info *buf_info;
+	struct mhi_tre *mhi_tre;
+	int ret;
+
+	if (mhi_is_ring_full(mhi_cntrl, tre_ring))
+		return -ENOMEM;
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (unlikely(MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))) {
+		read_unlock_bh(&mhi_cntrl->pm_lock);
+		return -EIO;
+	}
+
+	/* we're in M3 or transitioning to M3 */
+	if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
+		mhi_cntrl->runtime_get(mhi_cntrl);
+		mhi_cntrl->runtime_put(mhi_cntrl);
+	}
+
+	/* Toggle wake to exit out of M2 */
+	mhi_cntrl->wake_toggle(mhi_cntrl);
+
+	/* Generate the TRE */
+	buf_info = buf_ring->wp;
+
+	buf_info->v_addr = skb->data;
+	buf_info->cb_buf = skb;
+	buf_info->wp = tre_ring->wp;
+	buf_info->dir = mhi_chan->dir;
+	buf_info->len = len;
+	ret = mhi_cntrl->map_single(mhi_cntrl, buf_info);
+	if (ret)
+		goto map_error;
+
+	mhi_tre = tre_ring->wp;
+
+	mhi_tre->ptr = MHI_TRE_DATA_PTR(buf_info->p_addr);
+	mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(buf_info->len);
+	mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(1, 1, 0, 0);
+
+	/* increment WP */
+	mhi_add_ring_element(mhi_cntrl, tre_ring);
+	mhi_add_ring_element(mhi_cntrl, buf_ring);
+
+	if (mhi_chan->dir == DMA_TO_DEVICE)
+		atomic_inc(&mhi_cntrl->pending_pkts);
+
+	if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl))) {
+		read_lock_bh(&mhi_chan->lock);
+		mhi_ring_chan_db(mhi_cntrl, mhi_chan);
+		read_unlock_bh(&mhi_chan->lock);
+	}
+
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return 0;
+
+map_error:
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return ret;
+}
+
+int mhi_queue_dma(struct mhi_device *mhi_dev,
+		  struct mhi_chan *mhi_chan,
+		  void *buf,
+		  size_t len,
+		  enum mhi_flags mflags)
+{
+	struct mhi_buf *mhi_buf = buf;
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_ring *tre_ring = &mhi_chan->tre_ring;
+	struct mhi_ring *buf_ring = &mhi_chan->buf_ring;
+	struct mhi_buf_info *buf_info;
+	struct mhi_tre *mhi_tre;
+
+	if (mhi_is_ring_full(mhi_cntrl, tre_ring))
+		return -ENOMEM;
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (unlikely(MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))) {
+		dev_err(mhi_cntrl->dev,
+			"MHI is not in activate state, PM state: %s\n",
+			to_mhi_pm_state_str(mhi_cntrl->pm_state));
+		read_unlock_bh(&mhi_cntrl->pm_lock);
+
+		return -EIO;
+	}
+
+	/* we're in M3 or transitioning to M3 */
+	if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
+		mhi_cntrl->runtime_get(mhi_cntrl);
+		mhi_cntrl->runtime_put(mhi_cntrl);
+	}
+
+	/* Toggle wake to exit out of M2 */
+	mhi_cntrl->wake_toggle(mhi_cntrl);
+
+	/* Generate the TRE */
+	buf_info = buf_ring->wp;
+	WARN_ON(buf_info->used);
+	buf_info->p_addr = mhi_buf->dma_addr;
+	buf_info->pre_mapped = true;
+	buf_info->cb_buf = mhi_buf;
+	buf_info->wp = tre_ring->wp;
+	buf_info->dir = mhi_chan->dir;
+	buf_info->len = len;
+
+	mhi_tre = tre_ring->wp;
+
+	if (mhi_chan->xfer_type == MHI_BUF_RSC_DMA) {
+		buf_info->used = true;
+		mhi_tre->ptr =
+			MHI_RSCTRE_DATA_PTR(buf_info->p_addr, buf_info->len);
+		mhi_tre->dword[0] =
+			MHI_RSCTRE_DATA_DWORD0(buf_ring->wp - buf_ring->base);
+		mhi_tre->dword[1] = MHI_RSCTRE_DATA_DWORD1;
+	} else {
+		mhi_tre->ptr = MHI_TRE_DATA_PTR(buf_info->p_addr);
+		mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(buf_info->len);
+		mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(1, 1, 0, 0);
+	}
+
+	/* increment WP */
+	mhi_add_ring_element(mhi_cntrl, tre_ring);
+	mhi_add_ring_element(mhi_cntrl, buf_ring);
+
+	if (mhi_chan->dir == DMA_TO_DEVICE)
+		atomic_inc(&mhi_cntrl->pending_pkts);
+
+	if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl))) {
+		read_lock_bh(&mhi_chan->lock);
+		mhi_ring_chan_db(mhi_cntrl, mhi_chan);
+		read_unlock_bh(&mhi_chan->lock);
+	}
+
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	return 0;
+}
+
+int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan,
+		void *buf, void *cb, size_t buf_len, enum mhi_flags flags)
+{
+	struct mhi_ring *buf_ring, *tre_ring;
+	struct mhi_tre *mhi_tre;
+	struct mhi_buf_info *buf_info;
+	int eot, eob, chain, bei;
+	int ret;
+
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+
+	buf_info = buf_ring->wp;
+	buf_info->v_addr = buf;
+	buf_info->cb_buf = cb;
+	buf_info->wp = tre_ring->wp;
+	buf_info->dir = mhi_chan->dir;
+	buf_info->len = buf_len;
+
+	ret = mhi_cntrl->map_single(mhi_cntrl, buf_info);
+	if (ret)
+		return ret;
+
+	eob = !!(flags & MHI_EOB);
+	eot = !!(flags & MHI_EOT);
+	chain = !!(flags & MHI_CHAIN);
+	bei = !!(mhi_chan->intmod);
+
+	mhi_tre = tre_ring->wp;
+	mhi_tre->ptr = MHI_TRE_DATA_PTR(buf_info->p_addr);
+	mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(buf_len);
+	mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, eot, eob, chain);
+
+	/* increment WP */
+	mhi_add_ring_element(mhi_cntrl, tre_ring);
+	mhi_add_ring_element(mhi_cntrl, buf_ring);
+
+	return 0;
+}
+
+int mhi_queue_transfer(struct mhi_device *mhi_dev,
+		       enum dma_data_direction dir, void *buf, size_t len,
+		       enum mhi_flags mflags)
+{
+	if (dir == DMA_TO_DEVICE)
+		return mhi_dev->ul_chan->queue_xfer(mhi_dev, mhi_dev->ul_chan,
+						    buf, len, mflags);
+	else
+		return mhi_dev->dl_chan->queue_xfer(mhi_dev, mhi_dev->dl_chan,
+						    buf, len, mflags);
+}
+EXPORT_SYMBOL_GPL(mhi_queue_transfer);
+
+int mhi_queue_buf(struct mhi_device *mhi_dev, struct mhi_chan *mhi_chan,
+		  void *buf, size_t len, enum mhi_flags mflags)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_ring *tre_ring;
+	unsigned long flags;
+	int ret;
+
+	/*
+	 * this check here only as a guard, it's always
+	 * possible mhi can enter error while executing rest of function,
+	 * which is not fatal so we do not need to hold pm_lock
+	 */
+	if (unlikely(MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)))
+		return -EIO;
+
+	tre_ring = &mhi_chan->tre_ring;
+	if (mhi_is_ring_full(mhi_cntrl, tre_ring))
+		return -ENOMEM;
+
+	ret = mhi_chan->gen_tre(mhi_cntrl, mhi_chan, buf, buf, len, mflags);
+	if (unlikely(ret))
+		return ret;
+
+	read_lock_irqsave(&mhi_cntrl->pm_lock, flags);
+
+	/* we're in M3 or transitioning to M3 */
+	if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
+		mhi_cntrl->runtime_get(mhi_cntrl);
+		mhi_cntrl->runtime_put(mhi_cntrl);
+	}
+
+	/* Toggle wake to exit out of M2 */
+	mhi_cntrl->wake_toggle(mhi_cntrl);
+
+	if (mhi_chan->dir == DMA_TO_DEVICE)
+		atomic_inc(&mhi_cntrl->pending_pkts);
+
+	if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl))) {
+		unsigned long flags;
+
+		read_lock_irqsave(&mhi_chan->lock, flags);
+		mhi_ring_chan_db(mhi_cntrl, mhi_chan);
+		read_unlock_irqrestore(&mhi_chan->lock, flags);
+	}
+
+	read_unlock_irqrestore(&mhi_cntrl->pm_lock, flags);
+
+	return 0;
+}
+
+int mhi_send_cmd(struct mhi_controller *mhi_cntrl,
+		 struct mhi_chan *mhi_chan,
+		 enum mhi_cmd_type cmd)
+{
+	struct mhi_tre *cmd_tre = NULL;
+	struct mhi_cmd *mhi_cmd = &mhi_cntrl->mhi_cmd[PRIMARY_CMD_RING];
+	struct mhi_ring *ring = &mhi_cmd->ring;
+	int chan = 0;
+
+	if (mhi_chan)
+		chan = mhi_chan->chan;
+
+	spin_lock_bh(&mhi_cmd->lock);
+	if (!get_nr_avail_ring_elements(mhi_cntrl, ring)) {
+		spin_unlock_bh(&mhi_cmd->lock);
+		return -ENOMEM;
+	}
+
+	/* prepare the cmd tre */
+	cmd_tre = ring->wp;
+	switch (cmd) {
+	case MHI_CMD_RESET_CHAN:
+		cmd_tre->ptr = MHI_TRE_CMD_RESET_PTR;
+		cmd_tre->dword[0] = MHI_TRE_CMD_RESET_DWORD0;
+		cmd_tre->dword[1] = MHI_TRE_CMD_RESET_DWORD1(chan);
+		break;
+	case MHI_CMD_START_CHAN:
+		cmd_tre->ptr = MHI_TRE_CMD_START_PTR;
+		cmd_tre->dword[0] = MHI_TRE_CMD_START_DWORD0;
+		cmd_tre->dword[1] = MHI_TRE_CMD_START_DWORD1(chan);
+		break;
+	default:
+		dev_err(mhi_cntrl->dev, "Command not supported\n");
+		break;
+	}
+
+	/* queue to hardware */
+	mhi_add_ring_element(mhi_cntrl, ring);
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (likely(MHI_DB_ACCESS_VALID(mhi_cntrl)))
+		mhi_ring_cmd_db(mhi_cntrl, mhi_cmd);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+	spin_unlock_bh(&mhi_cmd->lock);
+
+	return 0;
+}
+
+static void __mhi_unprepare_channel(struct mhi_controller *mhi_cntrl,
+				    struct mhi_chan *mhi_chan)
+{
+	int ret;
+
+	dev_dbg(mhi_cntrl->dev, "Entered: unprepare channel:%d\n",
+		 mhi_chan->chan);
+
+	/* no more processing events for this channel */
+	mutex_lock(&mhi_chan->mutex);
+	write_lock_irq(&mhi_chan->lock);
+	if (mhi_chan->ch_state != MHI_CH_STATE_ENABLED) {
+		write_unlock_irq(&mhi_chan->lock);
+		mutex_unlock(&mhi_chan->mutex);
+		return;
+	}
+
+	mhi_chan->ch_state = MHI_CH_STATE_DISABLED;
+	write_unlock_irq(&mhi_chan->lock);
+
+	reinit_completion(&mhi_chan->completion);
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+		read_unlock_bh(&mhi_cntrl->pm_lock);
+		goto error_invalid_state;
+	}
+
+	mhi_cntrl->wake_toggle(mhi_cntrl);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+
+	mhi_cntrl->runtime_get(mhi_cntrl);
+	mhi_cntrl->runtime_put(mhi_cntrl);
+	ret = mhi_send_cmd(mhi_cntrl, mhi_chan, MHI_CMD_RESET_CHAN);
+	if (ret)
+		goto error_invalid_state;
+
+	/* even if it fails we will still reset */
+	ret = wait_for_completion_timeout(&mhi_chan->completion,
+				msecs_to_jiffies(mhi_cntrl->timeout_ms));
+	if (!ret || mhi_chan->ccs != MHI_EV_CC_SUCCESS)
+		dev_err(mhi_cntrl->dev,
+			"Failed to receive cmd completion, still resetting\n");
+
+error_invalid_state:
+	if (!mhi_chan->offload_ch) {
+		mhi_reset_chan(mhi_cntrl, mhi_chan);
+		mhi_deinit_chan_ctxt(mhi_cntrl, mhi_chan);
+	}
+	dev_dbg(mhi_cntrl->dev, "chan:%d successfully resetted\n",
+		 mhi_chan->chan);
+	mutex_unlock(&mhi_chan->mutex);
+}
+
+int mhi_prepare_channel(struct mhi_controller *mhi_cntrl,
+			struct mhi_chan *mhi_chan)
+{
+	int ret = 0;
+
+	dev_dbg(mhi_cntrl->dev, "Preparing channel: %d\n",
+		 mhi_chan->chan);
+
+	if (!(BIT(mhi_cntrl->ee) & mhi_chan->ee_mask)) {
+		dev_err(mhi_cntrl->dev,
+			"Current EE: %s Required EE Mask: 0x%x for chan: %s\n",
+			TO_MHI_EXEC_STR(mhi_cntrl->ee), mhi_chan->ee_mask,
+			mhi_chan->name);
+		return -ENOTCONN;
+	}
+
+	mutex_lock(&mhi_chan->mutex);
+
+	/* If channel is not in disable state, do not allow it to start */
+	if (mhi_chan->ch_state != MHI_CH_STATE_DISABLED) {
+		ret = -EIO;
+		dev_dbg(mhi_cntrl->dev,
+			 "channel: %d is not in disabled state\n",
+			 mhi_chan->chan);
+		goto error_init_chan;
+	}
+
+	/* Check of client manages channel context for offload channels */
+	if (!mhi_chan->offload_ch) {
+		ret = mhi_init_chan_ctxt(mhi_cntrl, mhi_chan);
+		if (ret)
+			goto error_init_chan;
+	}
+
+	reinit_completion(&mhi_chan->completion);
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+		read_unlock_bh(&mhi_cntrl->pm_lock);
+		ret = -EIO;
+		goto error_pm_state;
+	}
+
+	mhi_cntrl->wake_toggle(mhi_cntrl);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+	mhi_cntrl->runtime_get(mhi_cntrl);
+	mhi_cntrl->runtime_put(mhi_cntrl);
+
+	ret = mhi_send_cmd(mhi_cntrl, mhi_chan, MHI_CMD_START_CHAN);
+	if (ret)
+		goto error_pm_state;
+
+	ret = wait_for_completion_timeout(&mhi_chan->completion,
+				msecs_to_jiffies(mhi_cntrl->timeout_ms));
+	if (!ret || mhi_chan->ccs != MHI_EV_CC_SUCCESS) {
+		ret = -EIO;
+		goto error_pm_state;
+	}
+
+	write_lock_irq(&mhi_chan->lock);
+	mhi_chan->ch_state = MHI_CH_STATE_ENABLED;
+	write_unlock_irq(&mhi_chan->lock);
+
+	/* Pre-allocate buffer for xfer ring */
+	if (mhi_chan->pre_alloc) {
+		int nr_el = get_nr_avail_ring_elements(mhi_cntrl,
+						       &mhi_chan->tre_ring);
+		size_t len = mhi_cntrl->buffer_len;
+
+		while (nr_el--) {
+			void *buf;
+
+			buf = kmalloc(len, GFP_KERNEL);
+			if (!buf) {
+				ret = -ENOMEM;
+				goto error_pre_alloc;
+			}
+
+			/* Prepare transfer descriptors */
+			ret = mhi_chan->gen_tre(mhi_cntrl, mhi_chan, buf, buf,
+						len, MHI_EOT);
+			if (ret) {
+				kfree(buf);
+				goto error_pre_alloc;
+			}
+		}
+
+		read_lock_bh(&mhi_cntrl->pm_lock);
+		if (MHI_DB_ACCESS_VALID(mhi_cntrl)) {
+			read_lock_irq(&mhi_chan->lock);
+			mhi_ring_chan_db(mhi_cntrl, mhi_chan);
+			read_unlock_irq(&mhi_chan->lock);
+		}
+		read_unlock_bh(&mhi_cntrl->pm_lock);
+	}
+
+	mutex_unlock(&mhi_chan->mutex);
+
+	dev_dbg(mhi_cntrl->dev, "Chan: %d successfully moved to start state\n",
+		 mhi_chan->chan);
+
+	return 0;
+
+error_pm_state:
+	if (!mhi_chan->offload_ch)
+		mhi_deinit_chan_ctxt(mhi_cntrl, mhi_chan);
+
+error_init_chan:
+	mutex_unlock(&mhi_chan->mutex);
+
+	return ret;
+
+error_pre_alloc:
+	mutex_unlock(&mhi_chan->mutex);
+	__mhi_unprepare_channel(mhi_cntrl, mhi_chan);
+
+	return ret;
+}
+
+static void mhi_mark_stale_events(struct mhi_controller *mhi_cntrl,
+				  struct mhi_event *mhi_event,
+				  struct mhi_event_ctxt *er_ctxt,
+				  int chan)
+
+{
+	struct mhi_tre *dev_rp, *local_rp;
+	struct mhi_ring *ev_ring;
+	unsigned long flags;
+
+	dev_dbg(mhi_cntrl->dev,
+		 "Marking all events for chan: %d as stale\n", chan);
+
+	ev_ring = &mhi_event->ring;
+
+	/* mark all stale events related to channel as STALE event */
+	spin_lock_irqsave(&mhi_event->lock, flags);
+	dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
+
+	local_rp = ev_ring->rp;
+	while (dev_rp != local_rp) {
+		if (MHI_TRE_GET_EV_TYPE(local_rp) ==
+		    MHI_PKT_TYPE_TX_EVENT &&
+		    chan == MHI_TRE_GET_EV_CHID(local_rp))
+			local_rp->dword[1] = MHI_TRE_EV_DWORD1(chan,
+					MHI_PKT_TYPE_STALE_EVENT);
+		local_rp++;
+		if (local_rp == (ev_ring->base + ev_ring->len))
+			local_rp = ev_ring->base;
+	}
+
+
+	dev_dbg(mhi_cntrl->dev,
+		 "Finished marking events as stale events\n");
+	spin_unlock_irqrestore(&mhi_event->lock, flags);
+}
+
+static void mhi_reset_data_chan(struct mhi_controller *mhi_cntrl,
+				struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *buf_ring, *tre_ring;
+	struct mhi_result result;
+
+	/* Reset any pending buffers */
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+	result.transaction_status = -ENOTCONN;
+	result.bytes_xferd = 0;
+	while (tre_ring->rp != tre_ring->wp) {
+		struct mhi_buf_info *buf_info = buf_ring->rp;
+
+		if (mhi_chan->dir == DMA_TO_DEVICE)
+			atomic_dec(&mhi_cntrl->pending_pkts);
+
+		if (!buf_info->pre_mapped)
+			mhi_cntrl->unmap_single(mhi_cntrl, buf_info);
+
+		mhi_del_ring_element(mhi_cntrl, buf_ring);
+		mhi_del_ring_element(mhi_cntrl, tre_ring);
+
+		if (mhi_chan->pre_alloc) {
+			kfree(buf_info->cb_buf);
+		} else {
+			result.buf_addr = buf_info->cb_buf;
+			mhi_chan->xfer_cb(mhi_chan->mhi_dev, &result);
+		}
+	}
+}
+
+static void mhi_reset_rsc_chan(struct mhi_controller *mhi_cntrl,
+			       struct mhi_chan *mhi_chan)
+{
+	struct mhi_ring *buf_ring, *tre_ring;
+	struct mhi_result result;
+	struct mhi_buf_info *buf_info;
+
+	/* Reset any pending buffers */
+	buf_ring = &mhi_chan->buf_ring;
+	tre_ring = &mhi_chan->tre_ring;
+	result.transaction_status = -ENOTCONN;
+	result.bytes_xferd = 0;
+
+	buf_info = buf_ring->base;
+	for (; (void *)buf_info < buf_ring->base + buf_ring->len; buf_info++) {
+		if (!buf_info->used)
+			continue;
+
+		result.buf_addr = buf_info->cb_buf;
+		mhi_chan->xfer_cb(mhi_chan->mhi_dev, &result);
+		buf_info->used = false;
+	}
+}
+
+void mhi_reset_chan(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan)
+{
+
+	struct mhi_event *mhi_event;
+	struct mhi_event_ctxt *er_ctxt;
+	int chan = mhi_chan->chan;
+
+	/* Nothing to reset, client doesn't queue buffers */
+	if (mhi_chan->offload_ch)
+		return;
+
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	mhi_event = &mhi_cntrl->mhi_event[mhi_chan->er_index];
+	er_ctxt = &mhi_cntrl->mhi_ctxt->er_ctxt[mhi_chan->er_index];
+
+	mhi_mark_stale_events(mhi_cntrl, mhi_event, er_ctxt, chan);
+
+	if (mhi_chan->xfer_type == MHI_BUF_RSC_DMA)
+		mhi_reset_rsc_chan(mhi_cntrl, mhi_chan);
+	else
+		mhi_reset_data_chan(mhi_cntrl, mhi_chan);
+
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+}
+
+/* Move channel to start state */
+int mhi_prepare_for_transfer(struct mhi_device *mhi_dev)
+{
+	int ret, dir;
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_chan *mhi_chan;
+
+	for (dir = 0; dir < 2; dir++) {
+		mhi_chan = dir ? mhi_dev->dl_chan : mhi_dev->ul_chan;
+
+		if (!mhi_chan)
+			continue;
+
+		ret = mhi_prepare_channel(mhi_cntrl, mhi_chan);
+		if (ret)
+			goto error_open_chan;
+	}
+
+	return 0;
+
+error_open_chan:
+	for (--dir; dir >= 0; dir--) {
+		mhi_chan = dir ? mhi_dev->dl_chan : mhi_dev->ul_chan;
+
+		if (!mhi_chan)
+			continue;
+
+		__mhi_unprepare_channel(mhi_cntrl, mhi_chan);
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_prepare_for_transfer);
+
+void mhi_unprepare_from_transfer(struct mhi_device *mhi_dev)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_chan *mhi_chan;
+	int dir;
+
+	for (dir = 0; dir < 2; dir++) {
+		mhi_chan = dir ? mhi_dev->ul_chan : mhi_dev->dl_chan;
+
+		if (!mhi_chan)
+			continue;
+
+		__mhi_unprepare_channel(mhi_cntrl, mhi_chan);
+	}
+}
+EXPORT_SYMBOL_GPL(mhi_unprepare_from_transfer);
+
+int mhi_poll(struct mhi_device *mhi_dev, u32 budget)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_chan *mhi_chan = mhi_dev->dl_chan;
+	struct mhi_event *mhi_event = &mhi_cntrl->mhi_event[mhi_chan->er_index];
+	int ret;
+
+	spin_lock_bh(&mhi_event->lock);
+	ret = mhi_event->process_event(mhi_cntrl, mhi_event, budget);
+	spin_unlock_bh(&mhi_event->lock);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_poll);
diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
index 19987068b475..e37678037cde 100644
--- a/drivers/bus/mhi/core/pm.c
+++ b/drivers/bus/mhi/core/pm.c
@@ -925,3 +925,43 @@ int mhi_force_rddm_mode(struct mhi_controller *mhi_cntrl)
 	return ret;
 }
 EXPORT_SYMBOL_GPL(mhi_force_rddm_mode);
+
+void mhi_device_get(struct mhi_device *mhi_dev)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+
+	mhi_dev->dev_wake++;
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	mhi_cntrl->wake_get(mhi_cntrl, true);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+}
+EXPORT_SYMBOL_GPL(mhi_device_get);
+
+int mhi_device_get_sync(struct mhi_device *mhi_dev)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	int ret;
+
+	ret = __mhi_device_get_sync(mhi_cntrl);
+	if (!ret)
+		mhi_dev->dev_wake++;
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mhi_device_get_sync);
+
+void mhi_device_put(struct mhi_device *mhi_dev)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+
+	mhi_dev->dev_wake--;
+	read_lock_bh(&mhi_cntrl->pm_lock);
+	if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
+		mhi_cntrl->runtime_get(mhi_cntrl);
+		mhi_cntrl->runtime_put(mhi_cntrl);
+	}
+
+	mhi_cntrl->wake_put(mhi_cntrl, false);
+	read_unlock_bh(&mhi_cntrl->pm_lock);
+}
+EXPORT_SYMBOL_GPL(mhi_device_put);
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 2ea2deb922a1..a32548ac3425 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -359,6 +359,8 @@ struct mhi_controller_config {
  * @wake_toggle: CB function to assert and de-assert device wake (optional)
  * @runtime_get: CB function to controller runtime resume (required)
  * @runtimet_put: CB function to decrement pm usage (required)
+ * @map_single: CB function to create TRE buffer
+ * @unmap_single: CB function to destroy TRE buffer
  * @buffer_len: Bounce buffer length
  * @bounce_buf: Use of bounce buffer
  * @fbc_download: MHI host needs to do complete image transfer (optional)
@@ -421,6 +423,10 @@ struct mhi_controller {
 	void (*wake_toggle)(struct mhi_controller *mhi_cntrl);
 	int (*runtime_get)(struct mhi_controller *mhi_cntrl);
 	void (*runtime_put)(struct mhi_controller *mhi_cntrl);
+	int (*map_single)(struct mhi_controller *mhi_cntrl,
+			  struct mhi_buf_info *buf);
+	void (*unmap_single)(struct mhi_controller *mhi_cntrl,
+			     struct mhi_buf_info *buf);
 
 	size_t buffer_len;
 	bool bounce_buf;
@@ -603,4 +609,53 @@ int mhi_force_rddm_mode(struct mhi_controller *mhi_cntrl);
  */
 enum mhi_state mhi_get_mhi_state(struct mhi_controller *mhi_cntrl);
 
+/**
+ * mhi_device_get - Disable device low power mode
+ * @mhi_dev: Device associated with the channel
+ */
+void mhi_device_get(struct mhi_device *mhi_dev);
+
+/**
+ * mhi_device_get_sync - Disable device low power mode. Synchronously
+ *                       take the controller out of suspended state
+ * @mhi_dev: Device associated with the channel
+ */
+int mhi_device_get_sync(struct mhi_device *mhi_dev);
+
+/**
+ * mhi_device_put - Re-enable device low power mode
+ * @mhi_dev: Device associated with the channel
+ */
+void mhi_device_put(struct mhi_device *mhi_dev);
+
+/**
+ * mhi_prepare_for_transfer - Setup channel for data transfer
+ * @mhi_dev: Device associated with the channels
+ */
+int mhi_prepare_for_transfer(struct mhi_device *mhi_dev);
+
+/**
+ * mhi_unprepare_from_transfer - Unprepare the channels
+ * @mhi_dev: Device associated with the channels
+ */
+void mhi_unprepare_from_transfer(struct mhi_device *mhi_dev);
+
+/**
+ * mhi_poll - Poll for any available data in DL direction
+ * @mhi_dev: Device associated with the channels
+ * @budget: # of events to process
+ */
+int mhi_poll(struct mhi_device *mhi_dev, u32 budget);
+
+/**
+ * mhi_queue_transfer - Send or receive data from client device over MHI channel
+ * @mhi_dev: Device associated with the channels
+ * @dir: DMA direction for the channel
+ * @buf: Buffer for holding the data
+ * @len: Buffer length
+ * @mflags: MHI transfer flags used for the transfer
+ */
+int mhi_queue_transfer(struct mhi_device *mhi_dev, enum dma_data_direction dir,
+		       void *buf, size_t len, enum mhi_flags mflags);
+
 #endif /* _MHI_H_ */
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 12/16] bus: mhi: core: Add uevent support for module autoloading
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (10 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 11/16] bus: mhi: core: Add support for data transfer Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-06 20:16   ` Jeffrey Hugo
  2020-01-31 13:50 ` [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus Manivannan Sadhasivam
                   ` (3 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

Add uevent support to MHI bus so that the client drivers can be autoloaded
by udev when the MHI devices gets created. The client drivers are
expected to provide MODULE_DEVICE_TABLE with the MHI id_table struct so
that the alias can be exported.

Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/init.c       |  9 +++++++++
 include/linux/mod_devicetable.h   |  1 +
 scripts/mod/devicetable-offsets.c |  3 +++
 scripts/mod/file2alias.c          | 10 ++++++++++
 4 files changed, 23 insertions(+)

diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index 4d0e5bbb3389..711ddc661530 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -1247,6 +1247,14 @@ void mhi_driver_unregister(struct mhi_driver *mhi_drv)
 }
 EXPORT_SYMBOL_GPL(mhi_driver_unregister);
 
+static int mhi_uevent(struct device *dev, struct kobj_uevent_env *env)
+{
+	struct mhi_device *mhi_dev = to_mhi_device(dev);
+
+	return add_uevent_var(env, "MODALIAS=" MHI_DEVICE_MODALIAS_FMT,
+					mhi_dev->chan_name);
+}
+
 static int mhi_match(struct device *dev, struct device_driver *drv)
 {
 	struct mhi_device *mhi_dev = to_mhi_device(dev);
@@ -1273,6 +1281,7 @@ struct bus_type mhi_bus_type = {
 	.name = "mhi",
 	.dev_name = "mhi",
 	.match = mhi_match,
+	.uevent = mhi_uevent,
 };
 
 static int __init mhi_init(void)
diff --git a/include/linux/mod_devicetable.h b/include/linux/mod_devicetable.h
index be15e997fe39..f10e779a3fd0 100644
--- a/include/linux/mod_devicetable.h
+++ b/include/linux/mod_devicetable.h
@@ -821,6 +821,7 @@ struct wmi_device_id {
 	const void *context;
 };
 
+#define MHI_DEVICE_MODALIAS_FMT "mhi:%s"
 #define MHI_NAME_SIZE 32
 
 /**
diff --git a/scripts/mod/devicetable-offsets.c b/scripts/mod/devicetable-offsets.c
index 054405b90ba4..fe3f4a95cb21 100644
--- a/scripts/mod/devicetable-offsets.c
+++ b/scripts/mod/devicetable-offsets.c
@@ -231,5 +231,8 @@ int main(void)
 	DEVID(wmi_device_id);
 	DEVID_FIELD(wmi_device_id, guid_string);
 
+	DEVID(mhi_device_id);
+	DEVID_FIELD(mhi_device_id, chan);
+
 	return 0;
 }
diff --git a/scripts/mod/file2alias.c b/scripts/mod/file2alias.c
index c91eba751804..cae6a4e471b5 100644
--- a/scripts/mod/file2alias.c
+++ b/scripts/mod/file2alias.c
@@ -1335,6 +1335,15 @@ static int do_wmi_entry(const char *filename, void *symval, char *alias)
 	return 1;
 }
 
+/* Looks like: mhi:S */
+static int do_mhi_entry(const char *filename, void *symval, char *alias)
+{
+	DEF_FIELD_ADDR(symval, mhi_device_id, chan);
+	sprintf(alias, MHI_DEVICE_MODALIAS_FMT, *chan);
+
+	return 1;
+}
+
 /* Does namelen bytes of name exactly match the symbol? */
 static bool sym_is(const char *name, unsigned namelen, const char *symbol)
 {
@@ -1407,6 +1416,7 @@ static const struct devtable devtable[] = {
 	{"typec", SIZE_typec_device_id, do_typec_entry},
 	{"tee", SIZE_tee_client_device_id, do_tee_entry},
 	{"wmi", SIZE_wmi_device_id, do_wmi_entry},
+	{"mhi", SIZE_mhi_device_id, do_mhi_entry},
 };
 
 /* Create MODULE_ALIAS() statements.
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (11 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 12/16] bus: mhi: core: Add uevent support for module autoloading Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-03 10:16   ` Andy Shevchenko
  2020-01-31 13:50 ` [PATCH v2 14/16] net: qrtr: Add MHI transport layer Manivannan Sadhasivam
                   ` (2 subsequent siblings)
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam

Add MAINTAINERS entry for MHI bus.

Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 MAINTAINERS | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index cf6ccca6e61c..9213d150041c 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -10777,6 +10777,16 @@ M:	Vladimir Vid <vladimir.vid@sartura.hr>
 S:	Maintained
 F:	arch/arm64/boot/dts/marvell/armada-3720-uDPU.dts
 
+MHI BUS
+M:	Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
+M:	Hemant Kumar <hemantk@codeaurora.org>
+L:	linux-arm-msm@vger.kernel.org
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mani/mhi.git
+S:	Maintained
+F:	drivers/bus/mhi/
+F:	include/linux/mhi.h
+F:	Documentation/mhi/
+
 MICROBLAZE ARCHITECTURE
 M:	Michal Simek <monstr@monstr.eu>
 W:	http://www.monstr.eu/fdt/
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 14/16] net: qrtr: Add MHI transport layer
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (12 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-02-03 18:12   ` Jakub Kicinski
  2020-01-31 13:50 ` [PATCH v2 15/16] net: qrtr: Do not depend on ARCH_QCOM Manivannan Sadhasivam
  2020-01-31 13:50 ` [PATCH v2 16/16] soc: qcom: Do not depend on ARCH_QCOM for QMI helpers Manivannan Sadhasivam
  15 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam, David S. Miller, netdev

MHI is the transport layer used for communicating to the external modems.
Hence, this commit adds MHI transport layer support to QRTR for
transferring the QMI messages over IPC Router.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 net/qrtr/Kconfig  |   7 ++
 net/qrtr/Makefile |   2 +
 net/qrtr/mhi.c    | 207 ++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 216 insertions(+)
 create mode 100644 net/qrtr/mhi.c

diff --git a/net/qrtr/Kconfig b/net/qrtr/Kconfig
index 63f89cc6e82c..8eb876471564 100644
--- a/net/qrtr/Kconfig
+++ b/net/qrtr/Kconfig
@@ -29,4 +29,11 @@ config QRTR_TUN
 	  implement endpoints of QRTR, for purpose of tunneling data to other
 	  hosts or testing purposes.
 
+config QRTR_MHI
+	tristate "MHI IPC Router channels"
+	depends on MHI_BUS
+	help
+	  Say Y here to support MHI based ipcrouter channels. MHI is the
+	  transport used for communicating to external modems.
+
 endif # QRTR
diff --git a/net/qrtr/Makefile b/net/qrtr/Makefile
index 1c6d6c120fb7..3dc0a7c9d455 100644
--- a/net/qrtr/Makefile
+++ b/net/qrtr/Makefile
@@ -5,3 +5,5 @@ obj-$(CONFIG_QRTR_SMD) += qrtr-smd.o
 qrtr-smd-y	:= smd.o
 obj-$(CONFIG_QRTR_TUN) += qrtr-tun.o
 qrtr-tun-y	:= tun.o
+obj-$(CONFIG_QRTR_MHI) += qrtr-mhi.o
+qrtr-mhi-y	:= mhi.o
diff --git a/net/qrtr/mhi.c b/net/qrtr/mhi.c
new file mode 100644
index 000000000000..c85041a22f85
--- /dev/null
+++ b/net/qrtr/mhi.c
@@ -0,0 +1,207 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/mhi.h>
+#include <linux/mod_devicetable.h>
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <net/sock.h>
+
+#include "qrtr.h"
+
+struct qrtr_mhi_dev {
+	struct qrtr_endpoint ep;
+	struct mhi_device *mhi_dev;
+	struct device *dev;
+	spinlock_t ul_lock;		/* lock to protect ul_pkts */
+	struct list_head ul_pkts;
+	atomic_t in_reset;
+};
+
+struct qrtr_mhi_pkt {
+	struct list_head node;
+	struct sk_buff *skb;
+	struct kref refcount;
+	struct completion done;
+};
+
+static void qrtr_mhi_pkt_release(struct kref *ref)
+{
+	struct qrtr_mhi_pkt *pkt = container_of(ref, struct qrtr_mhi_pkt,
+						refcount);
+	struct sock *sk = pkt->skb->sk;
+
+	consume_skb(pkt->skb);
+	if (sk)
+		sock_put(sk);
+
+	kfree(pkt);
+}
+
+/* From MHI to QRTR */
+static void qcom_mhi_qrtr_dl_callback(struct mhi_device *mhi_dev,
+				      struct mhi_result *mhi_res)
+{
+	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
+	int rc;
+
+	if (!qdev || mhi_res->transaction_status)
+		return;
+
+	rc = qrtr_endpoint_post(&qdev->ep, mhi_res->buf_addr,
+				mhi_res->bytes_xferd);
+	if (rc == -EINVAL)
+		dev_err(qdev->dev, "invalid ipcrouter packet\n");
+}
+
+/* From QRTR to MHI */
+static void qcom_mhi_qrtr_ul_callback(struct mhi_device *mhi_dev,
+				      struct mhi_result *mhi_res)
+{
+	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
+	struct qrtr_mhi_pkt *pkt;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qdev->ul_lock, flags);
+	pkt = list_first_entry(&qdev->ul_pkts, struct qrtr_mhi_pkt, node);
+	list_del(&pkt->node);
+	complete_all(&pkt->done);
+
+	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);
+	spin_unlock_irqrestore(&qdev->ul_lock, flags);
+}
+
+static void qcom_mhi_qrtr_status_callback(struct mhi_device *mhi_dev,
+					  enum mhi_callback mhi_cb)
+{
+	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
+	struct qrtr_mhi_pkt *pkt;
+	unsigned long flags;
+
+	if (mhi_cb != MHI_CB_FATAL_ERROR)
+		return;
+
+	atomic_inc(&qdev->in_reset);
+	spin_lock_irqsave(&qdev->ul_lock, flags);
+	list_for_each_entry(pkt, &qdev->ul_pkts, node)
+		complete_all(&pkt->done);
+	spin_unlock_irqrestore(&qdev->ul_lock, flags);
+}
+
+/* Send data over MHI */
+static int qcom_mhi_qrtr_send(struct qrtr_endpoint *ep, struct sk_buff *skb)
+{
+	struct qrtr_mhi_dev *qdev = container_of(ep, struct qrtr_mhi_dev, ep);
+	struct qrtr_mhi_pkt *pkt;
+	int rc;
+
+	rc = skb_linearize(skb);
+	if (rc) {
+		kfree_skb(skb);
+		return rc;
+	}
+
+	pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
+	if (!pkt) {
+		kfree_skb(skb);
+		return -ENOMEM;
+	}
+
+	init_completion(&pkt->done);
+	kref_init(&pkt->refcount);
+	kref_get(&pkt->refcount);
+	pkt->skb = skb;
+
+	spin_lock_bh(&qdev->ul_lock);
+	list_add_tail(&pkt->node, &qdev->ul_pkts);
+	rc = mhi_queue_transfer(qdev->mhi_dev, DMA_TO_DEVICE, skb, skb->len,
+				MHI_EOT);
+	if (rc) {
+		list_del(&pkt->node);
+		kfree_skb(skb);
+		kfree(pkt);
+		spin_unlock_bh(&qdev->ul_lock);
+		return rc;
+	}
+
+	spin_unlock_bh(&qdev->ul_lock);
+	if (skb->sk)
+		sock_hold(skb->sk);
+
+	rc = wait_for_completion_interruptible_timeout(&pkt->done, HZ * 5);
+	if (atomic_read(&qdev->in_reset))
+		rc = -ECONNRESET;
+	else if (rc == 0)
+		rc = -ETIMEDOUT;
+	else if (rc > 0)
+		rc = 0;
+
+	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);
+
+	return rc;
+}
+
+static int qcom_mhi_qrtr_probe(struct mhi_device *mhi_dev,
+			       const struct mhi_device_id *id)
+{
+	struct qrtr_mhi_dev *qdev;
+	u32 net_id;
+	int rc;
+
+	qdev = devm_kzalloc(&mhi_dev->dev, sizeof(*qdev), GFP_KERNEL);
+	if (!qdev)
+		return -ENOMEM;
+
+	qdev->mhi_dev = mhi_dev;
+	qdev->dev = &mhi_dev->dev;
+	qdev->ep.xmit = qcom_mhi_qrtr_send;
+	atomic_set(&qdev->in_reset, 0);
+
+	net_id = QRTR_EP_NID_AUTO;
+
+	INIT_LIST_HEAD(&qdev->ul_pkts);
+	spin_lock_init(&qdev->ul_lock);
+
+	dev_set_drvdata(&mhi_dev->dev, qdev);
+	rc = qrtr_endpoint_register(&qdev->ep, net_id);
+	if (rc)
+		return rc;
+
+	dev_dbg(qdev->dev, "Qualcomm MHI QRTR driver probed\n");
+
+	return 0;
+}
+
+static void qcom_mhi_qrtr_remove(struct mhi_device *mhi_dev)
+{
+	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
+
+	qrtr_endpoint_unregister(&qdev->ep);
+	dev_set_drvdata(&mhi_dev->dev, NULL);
+}
+
+static const struct mhi_device_id qcom_mhi_qrtr_id_table[] = {
+	{ .chan = "IPCR" },
+	{}
+};
+MODULE_DEVICE_TABLE(mhi, qcom_mhi_qrtr_id_table);
+
+static struct mhi_driver qcom_mhi_qrtr_driver = {
+	.probe = qcom_mhi_qrtr_probe,
+	.remove = qcom_mhi_qrtr_remove,
+	.dl_xfer_cb = qcom_mhi_qrtr_dl_callback,
+	.ul_xfer_cb = qcom_mhi_qrtr_ul_callback,
+	.status_cb = qcom_mhi_qrtr_status_callback,
+	.id_table = qcom_mhi_qrtr_id_table,
+	.driver = {
+		.name = "qcom_mhi_qrtr",
+	},
+};
+
+module_driver(qcom_mhi_qrtr_driver, mhi_driver_register,
+	      mhi_driver_unregister);
+
+MODULE_DESCRIPTION("Qualcomm IPC-Router MHI interface driver");
+MODULE_LICENSE("GPL v2");
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 15/16] net: qrtr: Do not depend on ARCH_QCOM
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (13 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 14/16] net: qrtr: Add MHI transport layer Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  2020-01-31 13:50 ` [PATCH v2 16/16] soc: qcom: Do not depend on ARCH_QCOM for QMI helpers Manivannan Sadhasivam
  15 siblings, 0 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam, David S. Miller, netdev

IPC Router protocol is also used by external modems for exchanging the QMI
messages. Hence, it doesn't always depend on Qualcomm platforms. As a side
effect of removing the ARCH_QCOM dependency, it is going to miss the
COMPILE_TEST build coverage.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 net/qrtr/Kconfig | 1 -
 1 file changed, 1 deletion(-)

diff --git a/net/qrtr/Kconfig b/net/qrtr/Kconfig
index 8eb876471564..f362ca316015 100644
--- a/net/qrtr/Kconfig
+++ b/net/qrtr/Kconfig
@@ -4,7 +4,6 @@
 
 config QRTR
 	tristate "Qualcomm IPC Router support"
-	depends on ARCH_QCOM || COMPILE_TEST
 	---help---
 	  Say Y if you intend to use Qualcomm IPC router protocol.  The
 	  protocol is used to communicate with services provided by other
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* [PATCH v2 16/16] soc: qcom: Do not depend on ARCH_QCOM for QMI helpers
  2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
                   ` (14 preceding siblings ...)
  2020-01-31 13:50 ` [PATCH v2 15/16] net: qrtr: Do not depend on ARCH_QCOM Manivannan Sadhasivam
@ 2020-01-31 13:50 ` Manivannan Sadhasivam
  15 siblings, 0 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-01-31 13:50 UTC (permalink / raw)
  To: gregkh, arnd
  Cc: smohanad, jhugo, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Manivannan Sadhasivam, Andy Gross

QMI helpers are not always used by Qualcomm platforms. One of the
exceptions is the external modems available in near future. As a
side effect of removing the dependency, it is also going to loose
COMPILE_TEST build coverage.

Cc: Andy Gross <agross@kernel.org>
Cc: Bjorn Andersson <bjorn.andersson@linaro.org>
Cc: linux-arm-msm@vger.kernel.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/soc/qcom/Kconfig | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/soc/qcom/Kconfig b/drivers/soc/qcom/Kconfig
index 79d826553ac8..ca057bc9aae6 100644
--- a/drivers/soc/qcom/Kconfig
+++ b/drivers/soc/qcom/Kconfig
@@ -88,7 +88,6 @@ config QCOM_PM
 
 config QCOM_QMI_HELPERS
 	tristate
-	depends on ARCH_QCOM || COMPILE_TEST
 	depends on NET
 
 config QCOM_RMTFS_MEM
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 01/16] docs: Add documentation for MHI bus
  2020-01-31 13:49 ` [PATCH v2 01/16] docs: Add documentation for MHI bus Manivannan Sadhasivam
@ 2020-01-31 22:57   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-01-31 22:57 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel, Jonathan Corbet, linux-doc

On 1/31/2020 6:49 AM, Manivannan Sadhasivam wrote:
> MHI (Modem Host Interface) is a communication protocol used by the
> host processors to control and communicate with modems over a high
> speed peripheral bus or shared memory. The MHI protocol has been
> designed and developed by Qualcomm Innovation Center, Inc., for use
> in their modems. This commit adds the documentation for the bus and
> the implementation in Linux kernel.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/987
> 
> Cc: Jonathan Corbet <corbet@lwn.net>
> Cc: linux-doc@vger.kernel.org
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: converted to .rst and splitted the patch]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 03/16] bus: mhi: core: Add support for registering MHI client drivers
  2020-01-31 13:49 ` [PATCH v2 03/16] bus: mhi: core: Add support for registering MHI client drivers Manivannan Sadhasivam
@ 2020-01-31 23:00   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-01-31 23:00 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:49 AM, Manivannan Sadhasivam wrote:
> This commit adds support for registering MHI client drivers with the
> MHI stack. MHI client drivers binds to one or more MHI devices inorder
> to sends and receive the upper-layer protocol packets like IP packets,
> modem control messages, and diagnostics messages over MHI bus.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/987
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus
  2020-01-31 13:50 ` [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus Manivannan Sadhasivam
@ 2020-02-03 10:16   ` Andy Shevchenko
  2020-02-04  7:05     ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Andy Shevchenko @ 2020-02-03 10:16 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: Greg Kroah-Hartman, Arnd Bergmann, smohanad, Jeffrey Hugo,
	Kalle Valo, Bjorn Andersson, hemantk, linux-arm-msm,
	Linux Kernel Mailing List

On Fri, Jan 31, 2020 at 3:53 PM Manivannan Sadhasivam
<manivannan.sadhasivam@linaro.org> wrote:
>
> Add MAINTAINERS entry for MHI bus.

> +MHI BUS
> +M:     Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> +M:     Hemant Kumar <hemantk@codeaurora.org>
> +L:     linux-arm-msm@vger.kernel.org
> +T:     git git://git.kernel.org/pub/scm/linux/kernel/git/mani/mhi.git
> +S:     Maintained
> +F:     drivers/bus/mhi/
> +F:     include/linux/mhi.h
> +F:     Documentation/mhi/

Had you run parse-maintainers.pl afterwards to see if everything is okay?

-- 
With Best Regards,
Andy Shevchenko

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 14/16] net: qrtr: Add MHI transport layer
  2020-01-31 13:50 ` [PATCH v2 14/16] net: qrtr: Add MHI transport layer Manivannan Sadhasivam
@ 2020-02-03 18:12   ` Jakub Kicinski
  2020-02-04  8:19     ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Jakub Kicinski @ 2020-02-03 18:12 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: gregkh, arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel, David S. Miller, netdev

On Fri, 31 Jan 2020 19:20:07 +0530, Manivannan Sadhasivam wrote:
> +/* From QRTR to MHI */
> +static void qcom_mhi_qrtr_ul_callback(struct mhi_device *mhi_dev,
> +				      struct mhi_result *mhi_res)
> +{
> +	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
> +	struct qrtr_mhi_pkt *pkt;
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&qdev->ul_lock, flags);
> +	pkt = list_first_entry(&qdev->ul_pkts, struct qrtr_mhi_pkt, node);
> +	list_del(&pkt->node);
> +	complete_all(&pkt->done);
> +
> +	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);

Which kref_get() does this pair with?

Looks like qcom_mhi_qrtr_send() will release a reference after
completion, too.

> +	spin_unlock_irqrestore(&qdev->ul_lock, flags);
> +}

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus
  2020-02-03 10:16   ` Andy Shevchenko
@ 2020-02-04  7:05     ` Manivannan Sadhasivam
  0 siblings, 0 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-04  7:05 UTC (permalink / raw)
  To: Andy Shevchenko
  Cc: Greg Kroah-Hartman, Arnd Bergmann, smohanad, Jeffrey Hugo,
	Kalle Valo, Bjorn Andersson, hemantk, linux-arm-msm,
	Linux Kernel Mailing List

Hi Andy,

On Mon, Feb 03, 2020 at 12:16:16PM +0200, Andy Shevchenko wrote:
> On Fri, Jan 31, 2020 at 3:53 PM Manivannan Sadhasivam
> <manivannan.sadhasivam@linaro.org> wrote:
> >
> > Add MAINTAINERS entry for MHI bus.
> 
> > +MHI BUS
> > +M:     Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> > +M:     Hemant Kumar <hemantk@codeaurora.org>
> > +L:     linux-arm-msm@vger.kernel.org
> > +T:     git git://git.kernel.org/pub/scm/linux/kernel/git/mani/mhi.git
> > +S:     Maintained
> > +F:     drivers/bus/mhi/
> > +F:     include/linux/mhi.h
> > +F:     Documentation/mhi/
> 
> Had you run parse-maintainers.pl afterwards to see if everything is okay?
> 

Hmm. I didn't run it before but now looks like the entries need to be sorted
as below:

M:      Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
M:      Hemant Kumar <hemantk@codeaurora.org>
L:      linux-arm-msm@vger.kernel.org
S:      Maintained
T:      git git://git.kernel.org/pub/scm/linux/kernel/git/mani/mhi.git
F:      Documentation/mhi/
F:      drivers/bus/mhi/
F:      include/linux/mhi.h

Will address this in next revision.

Thanks,
Mani

> -- 
> With Best Regards,
> Andy Shevchenko

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 14/16] net: qrtr: Add MHI transport layer
  2020-02-03 18:12   ` Jakub Kicinski
@ 2020-02-04  8:19     ` Manivannan Sadhasivam
  2020-02-07  0:14       ` Chris Lew
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-04  8:19 UTC (permalink / raw)
  To: Jakub Kicinski
  Cc: gregkh, arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel, David S. Miller, netdev

Hi Jakub,

On Mon, Feb 03, 2020 at 10:12:25AM -0800, Jakub Kicinski wrote:
> On Fri, 31 Jan 2020 19:20:07 +0530, Manivannan Sadhasivam wrote:
> > +/* From QRTR to MHI */
> > +static void qcom_mhi_qrtr_ul_callback(struct mhi_device *mhi_dev,
> > +				      struct mhi_result *mhi_res)
> > +{
> > +	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
> > +	struct qrtr_mhi_pkt *pkt;
> > +	unsigned long flags;
> > +
> > +	spin_lock_irqsave(&qdev->ul_lock, flags);
> > +	pkt = list_first_entry(&qdev->ul_pkts, struct qrtr_mhi_pkt, node);
> > +	list_del(&pkt->node);
> > +	complete_all(&pkt->done);
> > +
> > +	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);
> 
> Which kref_get() does this pair with?
> 
> Looks like qcom_mhi_qrtr_send() will release a reference after
> completion, too.
> 

Yikes, there is some issue here...

Acutally the issue is not in what you referred above but the overall kref
handling itself. Please see below.

kref_put() should be present in qcom_mhi_qrtr_ul_callback() as it will
decrement the refcount which got incremented in qcom_mhi_qrtr_send(). It
should be noted that kref_init() will fix the refcount to 1 and kref_get() will
increment to 2. So for properly releasing the refcount to 0, we need to call
kref_put() twice.

So if all goes well, the refcount will get decremented twice in
qcom_mhi_qrtr_ul_callback() as well as in qcom_mhi_qrtr_send() and we are good.

But, if the transfer has failed ie., when qcom_mhi_qrtr_ul_callback() doesn't
get called, then we are leaking the refcount. I need to rework the kref handling
code in next iteration.

Thanks for triggering this!

Regards,
Mani

> > +	spin_unlock_irqrestore(&qdev->ul_lock, flags);
> > +}

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 04/16] bus: mhi: core: Add support for creating and destroying MHI devices
  2020-01-31 13:49 ` [PATCH v2 04/16] bus: mhi: core: Add support for creating and destroying MHI devices Manivannan Sadhasivam
@ 2020-02-06 16:16   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 16:16 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:49 AM, Manivannan Sadhasivam wrote:
> This commit adds support for creating and destroying MHI devices. The
> MHI devices binds to the MHI channels and are used to transfer data
> between MHI host and client device.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/989
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted from pm patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic
  2020-01-31 13:50 ` [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic Manivannan Sadhasivam
@ 2020-02-06 16:41   ` Jeffrey Hugo
  2020-02-06 20:17     ` Jeffrey Hugo
  0 siblings, 1 reply; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 16:41 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
> MHI protocol supports downloading RDDM (RAM Dump) image from the
> device through BHIE. This is useful to debugging as the RDDM image
> can capture the firmware state.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/989
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted the data transfer patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> ---
>   drivers/bus/mhi/core/boot.c     | 157 +++++++++++++++++++++++++++++++-
>   drivers/bus/mhi/core/init.c     |  38 ++++++++
>   drivers/bus/mhi/core/internal.h |   2 +
>   drivers/bus/mhi/core/pm.c       |  31 +++++++
>   include/linux/mhi.h             |  24 +++++
>   5 files changed, 249 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/bus/mhi/core/boot.c b/drivers/bus/mhi/core/boot.c
> index 36956fb6eff2..564d61f98257 100644
> --- a/drivers/bus/mhi/core/boot.c
> +++ b/drivers/bus/mhi/core/boot.c
> @@ -20,7 +20,159 @@
>   #include <linux/wait.h>
>   #include "internal.h"
>   
> -/* Download AMSS image to device */

Seems a bit odd to add this line in patch 8, only to remove it here in 
patch 9.  Feels like never adding it in the first place would be better. 
  Not something to warrant a v3 through.

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-01-31 13:49 ` [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers Manivannan Sadhasivam
@ 2020-02-06 16:56   ` Greg KH
  2020-02-11 19:11     ` Manivannan Sadhasivam
  2020-02-06 16:57   ` Greg KH
  2020-02-06 17:08   ` Jeffrey Hugo
  2 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-06 16:56 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> +static void mhi_release_device(struct device *dev)
> +{
> +	struct mhi_device *mhi_dev = to_mhi_device(dev);
> +
> +	if (mhi_dev->ul_chan)
> +		mhi_dev->ul_chan->mhi_dev = NULL;

That looks really odd.  Why didn't you just drop the reference you
should have grabbed here for this pointer?  You did properly increment
it when you saved it, right?  :)

> +
> +	if (mhi_dev->dl_chan)
> +		mhi_dev->dl_chan->mhi_dev = NULL;

Same here.

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-01-31 13:49 ` [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers Manivannan Sadhasivam
  2020-02-06 16:56   ` Greg KH
@ 2020-02-06 16:57   ` Greg KH
  2020-02-11 18:41     ` Manivannan Sadhasivam
  2020-02-06 17:08   ` Jeffrey Hugo
  2 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-06 16:57 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> --- /dev/null
> +++ b/drivers/bus/mhi/core/init.c
> @@ -0,0 +1,407 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> + *
> + */
> +
> +#define dev_fmt(fmt) "MHI: " fmt

This should not be needed, right?  The bus/device name should give you
all you need here from what I can tell.  So why is this needed?

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-01-31 13:49 ` [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers Manivannan Sadhasivam
  2020-02-06 16:56   ` Greg KH
  2020-02-06 16:57   ` Greg KH
@ 2020-02-06 17:08   ` Jeffrey Hugo
  2 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 17:08 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:49 AM, Manivannan Sadhasivam wrote:
> This commit adds support for registering MHI controller drivers with
> the MHI stack. MHI controller drivers manages the interaction with the
> MHI client devices such as the external modems and WiFi chipsets. They
> are also the MHI bus master in charge of managing the physical link
> between the host and client device.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/987
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [jhugo: added static config for controllers and fixed several bugs]
> Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
> [mani: removed DT dependency, splitted and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

> +/**
> + * enum mhi_device_type - Device types
> + * @MHI_DEVICE_XFER: Handles data transfer
> + * @MHI_DEVICE_TIMESYNC: Use for timesync feature
> + * @MHI_DEVICE_CONTROLLER: Control device
> + */
> +enum mhi_device_type {
> +	MHI_DEVICE_XFER,
> +	MHI_DEVICE_TIMESYNC,
> +	MHI_DEVICE_CONTROLLER,
> +};

Time sync support was removed, no?  I don't see MHI_DEVICE_TIMESYNC used 
anywhere in the code.

> +/**
> + * enum mhi_er_data_type - Event ring data types
> + * @MHI_ER_DATA: Only client data over this ring
> + * @MHI_ER_CTRL: MHI control data and client data
> + * @MHI_ER_TSYNC: Time sync events
> + */
> +enum mhi_er_data_type {
> +	MHI_ER_DATA,
> +	MHI_ER_CTRL,
> +	MHI_ER_TSYNC,
> +};

Time sync support was removed, no?  I don't see MHI_ER_TSYNC used 
anywhere in the code.

> +/**
> + * struct mhi_controller - Master MHI controller structure
> + * @dev: Driver model device node for the controller (required)
> + * @mhi_dev: MHI device instance for the controller
> + * @regs: Base address of MHI MMIO register space (required)
> + * @iova_start: IOMMU starting address for data (required)
> + * @iova_stop: IOMMU stop address for data (required)
> + * @fw_image: Firmware image name for normal booting (required)
> + * @edl_image: Firmware image name for emergency download mode (optional)
> + * @sbl_size: SBL image size downloaded through BHIe (optional)
> + * @seg_len: BHIe vector size (optional)
> + * @mhi_chan: Points to the channel configuration table
> + * @lpm_chans: List of channels that require LPM notifications
> + * @irq: base irq # to request (required)
> + * @max_chan: Maximum number of channels the controller supports
> + * @total_ev_rings: Total # of event rings allocated
> + * @hw_ev_rings: Number of hardware event rings
> + * @sw_ev_rings: Number of software event rings
> + * @nr_irqs_req: Number of IRQs required to operate (optional)
> + * @nr_irqs: Number of IRQ allocated by bus master (required)
> + * @mhi_event: MHI event ring configurations table
> + * @mhi_cmd: MHI command ring configurations table
> + * @mhi_ctxt: MHI device context, shared memory between host and device
> + * @pm_mutex: Mutex for suspend/resume operation
> + * @pm_lock: Lock for protecting MHI power management state
> + * @timeout_ms: Timeout in ms for state transitions
> + * @pm_state: MHI power management state
> + * @db_access: DB access states
> + * @ee: MHI device execution environment
> + * @dev_wake: Device wakeup count
> + * @pending_pkts: Pending packets for the controller
> + * @transition_list: List of MHI state transitions
> + * @transition_lock: Lock for protecting MHI state transition list
> + * @wlock: Lock for protecting device wakeup
> + * @st_worker: State transition worker
> + * @fw_worker: Firmware download worker
> + * @syserr_worker: System error worker
> + * @state_event: State change event
> + * @status_cb: CB function to notify power states of the device (required)
> + * @link_status: CB function to query link status of the device (required)
> + * @wake_get: CB function to assert device wake (optional)
> + * @wake_put: CB function to de-assert device wake (optional)
> + * @wake_toggle: CB function to assert and de-assert device wake (optional)
> + * @runtime_get: CB function to controller runtime resume (required)
> + * @runtimet_put: CB function to decrement pm usage (required)
> + * @buffer_len: Bounce buffer length
> + * @bounce_buf: Use of bounce buffer
> + * @fbc_download: MHI host needs to do complete image transfer (optional)
> + * @pre_init: MHI host needs to do pre-initialization before power up
> + * @wake_set: Device wakeup set flag
> + */

I'm happy the optional/required is listed.  However if I look at this 
from the perspective of someone writing a controller for the first time, 
I'm not confident the required/optional annotations are clear enough. 
Perhaps a quick sentance explaining that those annotations indicate 
fields which must be populated prior to mhi_register_controller() ?

> +
> +/**
> + * struct mhi_device - Structure representing a MHI device which binds
> + *                     to channels
> + * @id: Pointer to MHI device ID struct
> + * @chan_name: Name of the channel to which the device binds
> + * @mhi_cntrl: Controller the device belongs to
> + * @ul_chan: UL channel for the device
> + * @dl_chan: DL channel for the device
> + * @dev: Driver model device node for the MHI device
> + * @dev_type: MHI device type
> + * @tiocm: Device current terminal settings
> + * @dev_wake: Device wakeup counter
> + */
> +struct mhi_device {
> +	const struct mhi_device_id *id;
> +	const char *chan_name;
> +	struct mhi_controller *mhi_cntrl;
> +	struct mhi_chan *ul_chan;
> +	struct mhi_chan *dl_chan;
> +	struct device dev;
> +	enum mhi_device_type dev_type;
> +	u32 tiocm;

This was for the old ioctl support, right?  I don't see it used anywhere.

> +	u32 dev_wake;
> +};


-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 05/16] bus: mhi: core: Add support for ringing channel/event ring doorbells
  2020-01-31 13:49 ` [PATCH v2 05/16] bus: mhi: core: Add support for ringing channel/event ring doorbells Manivannan Sadhasivam
@ 2020-02-06 20:14   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:14 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:49 AM, Manivannan Sadhasivam wrote:
> This commit adds support for ringing channel and event ring doorbells
> by MHI host. The MHI host can use the channel and event ring doorbells
> for notifying the client device about processing transfer and event
> rings which it has queued using MMIO registers.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/989
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted from pm patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 06/16] bus: mhi: core: Add support for PM state transitions
  2020-01-31 13:49 ` [PATCH v2 06/16] bus: mhi: core: Add support for PM state transitions Manivannan Sadhasivam
@ 2020-02-06 20:15   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:15 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:49 AM, Manivannan Sadhasivam wrote:
> This commit adds support for transitioning the MHI states as a
> part of the power management operations. Helpers functions are
> provided for the state transitions, which will be consumed by the
> actual power management routines.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/989
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [jhugo: removed dma_zalloc_coherent() and fixed several bugs]
> Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
> [mani: splitted the pm patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 07/16] bus: mhi: core: Add support for basic PM operations
  2020-01-31 13:50 ` [PATCH v2 07/16] bus: mhi: core: Add support for basic PM operations Manivannan Sadhasivam
@ 2020-02-06 20:15   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:15 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
> This commit adds support for basic MHI PM operations such as
> mhi_async_power_up, mhi_sync_power_up, and mhi_power_down. These
> routines places the MHI bus into respective power domain states
> and calls the state_transition APIs when necessary. The MHI
> controller driver is expected to call these PM routines for
> MHI powerup and powerdown.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/989
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted the pm patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 08/16] bus: mhi: core: Add support for downloading firmware over BHIe
  2020-01-31 13:50 ` [PATCH v2 08/16] bus: mhi: core: Add support for downloading firmware over BHIe Manivannan Sadhasivam
@ 2020-02-06 20:15   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:15 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
> MHI supports downloading the device firmware over BHI/BHIe (Boot Host
> Interface) protocol. Hence, this commit adds necessary helpers, which
> will be called during device power up stage.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/989
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted the data transfer patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 10/16] bus: mhi: core: Add support for processing events from client device
  2020-01-31 13:50 ` [PATCH v2 10/16] bus: mhi: core: Add support for processing events from client device Manivannan Sadhasivam
@ 2020-02-06 20:16   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:16 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
> This commit adds support for processing the MHI data and control
> events from the client device. The client device can report various
> events such as EE events, state change events by interrupting the
> host through IRQ and adding events to the event rings allocated by
> the host during initialization.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/988
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted the data transfer patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 11/16] bus: mhi: core: Add support for data transfer
  2020-01-31 13:50 ` [PATCH v2 11/16] bus: mhi: core: Add support for data transfer Manivannan Sadhasivam
@ 2020-02-06 20:16   ` Jeffrey Hugo
  2020-02-17 16:13   ` Arnd Bergmann
  1 sibling, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:16 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
> Add support for transferring data between external modem and host
> processor using MHI protocol.
> 
> This is based on the patch submitted by Sujeev Dias:
> https://lkml.org/lkml/2018/7/9/988
> 
> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
> [mani: splitted the data transfer patch and cleaned up for upstream]
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 12/16] bus: mhi: core: Add uevent support for module autoloading
  2020-01-31 13:50 ` [PATCH v2 12/16] bus: mhi: core: Add uevent support for module autoloading Manivannan Sadhasivam
@ 2020-02-06 20:16   ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:16 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
> Add uevent support to MHI bus so that the client drivers can be autoloaded
> by udev when the MHI devices gets created. The client drivers are
> expected to provide MODULE_DEVICE_TABLE with the MHI id_table struct so
> that the alias can be exported.
> 
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic
  2020-02-06 16:41   ` Jeffrey Hugo
@ 2020-02-06 20:17     ` Jeffrey Hugo
  0 siblings, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-06 20:17 UTC (permalink / raw)
  To: Manivannan Sadhasivam, gregkh, arnd
  Cc: smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm, linux-kernel

On 2/6/2020 9:41 AM, Jeffrey Hugo wrote:
> On 1/31/2020 6:50 AM, Manivannan Sadhasivam wrote:
>> MHI protocol supports downloading RDDM (RAM Dump) image from the
>> device through BHIE. This is useful to debugging as the RDDM image
>> can capture the firmware state.
>>
>> This is based on the patch submitted by Sujeev Dias:
>> https://lkml.org/lkml/2018/7/9/989
>>
>> Signed-off-by: Sujeev Dias <sdias@codeaurora.org>
>> Signed-off-by: Siddartha Mohanadoss <smohanad@codeaurora.org>
>> [mani: splitted the data transfer patch and cleaned up for upstream]
>> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
>> ---
>>   drivers/bus/mhi/core/boot.c     | 157 +++++++++++++++++++++++++++++++-
>>   drivers/bus/mhi/core/init.c     |  38 ++++++++
>>   drivers/bus/mhi/core/internal.h |   2 +
>>   drivers/bus/mhi/core/pm.c       |  31 +++++++
>>   include/linux/mhi.h             |  24 +++++
>>   5 files changed, 249 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/bus/mhi/core/boot.c b/drivers/bus/mhi/core/boot.c
>> index 36956fb6eff2..564d61f98257 100644
>> --- a/drivers/bus/mhi/core/boot.c
>> +++ b/drivers/bus/mhi/core/boot.c
>> @@ -20,7 +20,159 @@
>>   #include <linux/wait.h>
>>   #include "internal.h"
>> -/* Download AMSS image to device */
> 
> Seems a bit odd to add this line in patch 8, only to remove it here in 
> patch 9.  Feels like never adding it in the first place would be better. 
>   Not something to warrant a v3 through.
> 

Since my nit is minor-

Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Tested-by: Jeffrey Hugo <jhugo@codeaurora.org>

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 14/16] net: qrtr: Add MHI transport layer
  2020-02-04  8:19     ` Manivannan Sadhasivam
@ 2020-02-07  0:14       ` Chris Lew
  2020-02-11  3:50         ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Chris Lew @ 2020-02-07  0:14 UTC (permalink / raw)
  To: Manivannan Sadhasivam, Jakub Kicinski
  Cc: gregkh, arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel, David S. Miller, netdev


On 2/4/2020 12:19 AM, Manivannan Sadhasivam wrote:
> Hi Jakub,
>
> On Mon, Feb 03, 2020 at 10:12:25AM -0800, Jakub Kicinski wrote:
>> On Fri, 31 Jan 2020 19:20:07 +0530, Manivannan Sadhasivam wrote:
>>> +/* From QRTR to MHI */
>>> +static void qcom_mhi_qrtr_ul_callback(struct mhi_device *mhi_dev,
>>> +				      struct mhi_result *mhi_res)
>>> +{
>>> +	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
>>> +	struct qrtr_mhi_pkt *pkt;
>>> +	unsigned long flags;
>>> +
>>> +	spin_lock_irqsave(&qdev->ul_lock, flags);
>>> +	pkt = list_first_entry(&qdev->ul_pkts, struct qrtr_mhi_pkt, node);
>>> +	list_del(&pkt->node);
>>> +	complete_all(&pkt->done);
>>> +
>>> +	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);
>> Which kref_get() does this pair with?
>>
>> Looks like qcom_mhi_qrtr_send() will release a reference after
>> completion, too.
>>
> Yikes, there is some issue here...
>
> Acutally the issue is not in what you referred above but the overall kref
> handling itself. Please see below.
>
> kref_put() should be present in qcom_mhi_qrtr_ul_callback() as it will
> decrement the refcount which got incremented in qcom_mhi_qrtr_send(). It
> should be noted that kref_init() will fix the refcount to 1 and kref_get() will
> increment to 2. So for properly releasing the refcount to 0, we need to call
> kref_put() twice.
>
> So if all goes well, the refcount will get decremented twice in
> qcom_mhi_qrtr_ul_callback() as well as in qcom_mhi_qrtr_send() and we are good.
>
> But, if the transfer has failed ie., when qcom_mhi_qrtr_ul_callback() doesn't
> get called, then we are leaking the refcount. I need to rework the kref handling
> code in next iteration.
>
> Thanks for triggering this!
>
> Regards,
> Mani
>
>>> +	spin_unlock_irqrestore(&qdev->ul_lock, flags);
>>> +}

Hi Mani,

I'm not sure if this was changed in your patches but MHI is supposed to give a
ul_callback() for any packet that is successfully queued. In the case of the
transfer failing, the ul_callback() should still be called so there should
be no refcount leaking. It is an essential assumption I made, if that no longer
holds true then the entire driver needs to be reworked.

Thanks,
Chris

-- 

Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 14/16] net: qrtr: Add MHI transport layer
  2020-02-07  0:14       ` Chris Lew
@ 2020-02-11  3:50         ` Manivannan Sadhasivam
  2020-02-12  1:01           ` Chris Lew
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-11  3:50 UTC (permalink / raw)
  To: Chris Lew
  Cc: Jakub Kicinski, gregkh, arnd, smohanad, jhugo, kvalo,
	bjorn.andersson, hemantk, linux-arm-msm, linux-kernel,
	David S. Miller, netdev

Hi Chris,

On Thu, Feb 06, 2020 at 04:14:19PM -0800, Chris Lew wrote:
> 
> On 2/4/2020 12:19 AM, Manivannan Sadhasivam wrote:
> > Hi Jakub,
> > 
> > On Mon, Feb 03, 2020 at 10:12:25AM -0800, Jakub Kicinski wrote:
> > > On Fri, 31 Jan 2020 19:20:07 +0530, Manivannan Sadhasivam wrote:
> > > > +/* From QRTR to MHI */
> > > > +static void qcom_mhi_qrtr_ul_callback(struct mhi_device *mhi_dev,
> > > > +				      struct mhi_result *mhi_res)
> > > > +{
> > > > +	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
> > > > +	struct qrtr_mhi_pkt *pkt;
> > > > +	unsigned long flags;
> > > > +
> > > > +	spin_lock_irqsave(&qdev->ul_lock, flags);
> > > > +	pkt = list_first_entry(&qdev->ul_pkts, struct qrtr_mhi_pkt, node);
> > > > +	list_del(&pkt->node);
> > > > +	complete_all(&pkt->done);
> > > > +
> > > > +	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);
> > > Which kref_get() does this pair with?
> > > 
> > > Looks like qcom_mhi_qrtr_send() will release a reference after
> > > completion, too.
> > > 
> > Yikes, there is some issue here...
> > 
> > Acutally the issue is not in what you referred above but the overall kref
> > handling itself. Please see below.
> > 
> > kref_put() should be present in qcom_mhi_qrtr_ul_callback() as it will
> > decrement the refcount which got incremented in qcom_mhi_qrtr_send(). It
> > should be noted that kref_init() will fix the refcount to 1 and kref_get() will
> > increment to 2. So for properly releasing the refcount to 0, we need to call
> > kref_put() twice.
> > 
> > So if all goes well, the refcount will get decremented twice in
> > qcom_mhi_qrtr_ul_callback() as well as in qcom_mhi_qrtr_send() and we are good.
> > 
> > But, if the transfer has failed ie., when qcom_mhi_qrtr_ul_callback() doesn't
> > get called, then we are leaking the refcount. I need to rework the kref handling
> > code in next iteration.
> > 
> > Thanks for triggering this!
> > 
> > Regards,
> > Mani
> > 
> > > > +	spin_unlock_irqrestore(&qdev->ul_lock, flags);
> > > > +}
> 
> Hi Mani,
> 
> I'm not sure if this was changed in your patches but MHI is supposed to give a
> ul_callback() for any packet that is successfully queued. In the case of the
> transfer failing, the ul_callback() should still be called so there should
> be no refcount leaking. It is an essential assumption I made, if that no longer
> holds true then the entire driver needs to be reworked.
> 

Your assumption is correct. Only when the packet gets queued into the transfer
ring, the ul_xfer_cb will be called irrespective of the transfer state (success
or failure). But when the mhi_queue_transfer() returns even before queuing any
packet, then we need to decrease the refcount in the error path.

Please correct me if I'm wrong.

Thanks,
Mani

> Thanks,
> Chris
> 
> -- 
> 
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-06 16:57   ` Greg KH
@ 2020-02-11 18:41     ` Manivannan Sadhasivam
  2020-02-11 19:20       ` Greg KH
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-11 18:41 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

Hi Greg,

On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > --- /dev/null
> > +++ b/drivers/bus/mhi/core/init.c
> > @@ -0,0 +1,407 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > + *
> > + */
> > +
> > +#define dev_fmt(fmt) "MHI: " fmt
> 
> This should not be needed, right?  The bus/device name should give you
> all you need here from what I can tell.  So why is this needed?
> 

The log will have only the device name as like PCI-E. But that won't specify
where the error is coming from. Having "MHI" prefix helps the users to
quickly identify that the error is coming from MHI stack.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-06 16:56   ` Greg KH
@ 2020-02-11 19:11     ` Manivannan Sadhasivam
  2020-02-11 19:22       ` Greg KH
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-11 19:11 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

Hi Greg,

On Thu, Feb 06, 2020 at 05:56:06PM +0100, Greg KH wrote:
> On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > +static void mhi_release_device(struct device *dev)
> > +{
> > +	struct mhi_device *mhi_dev = to_mhi_device(dev);
> > +
> > +	if (mhi_dev->ul_chan)
> > +		mhi_dev->ul_chan->mhi_dev = NULL;
> 
> That looks really odd.  Why didn't you just drop the reference you
> should have grabbed here for this pointer?  You did properly increment
> it when you saved it, right?  :)
> 

Well, there is no reference count (kref) exist for mhi_dev. And we really
needed to NULL the mhi_dev to avoid any dangling reference to it. The reason for
not having kref is that, each mhi_dev will be used by maximum of 2 channels
only. So thought that refcounting is not needed. Please correct me if I'm
wrong.

Thanks,
Mani

> > +
> > +	if (mhi_dev->dl_chan)
> > +		mhi_dev->dl_chan->mhi_dev = NULL;
> 
> Same here.
> 
> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-11 18:41     ` Manivannan Sadhasivam
@ 2020-02-11 19:20       ` Greg KH
  2020-02-13 15:20         ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-11 19:20 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> Hi Greg,
> 
> On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > --- /dev/null
> > > +++ b/drivers/bus/mhi/core/init.c
> > > @@ -0,0 +1,407 @@
> > > +// SPDX-License-Identifier: GPL-2.0
> > > +/*
> > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > + *
> > > + */
> > > +
> > > +#define dev_fmt(fmt) "MHI: " fmt
> > 
> > This should not be needed, right?  The bus/device name should give you
> > all you need here from what I can tell.  So why is this needed?
> > 
> 
> The log will have only the device name as like PCI-E. But that won't specify
> where the error is coming from. Having "MHI" prefix helps the users to
> quickly identify that the error is coming from MHI stack.

If the driver binds properly to the device, the name of the driver will
be there in the message, so I suggest using that please.

No need for this prefix...

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-11 19:11     ` Manivannan Sadhasivam
@ 2020-02-11 19:22       ` Greg KH
  2020-02-13 15:22         ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-11 19:22 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Wed, Feb 12, 2020 at 12:41:47AM +0530, Manivannan Sadhasivam wrote:
> Hi Greg,
> 
> On Thu, Feb 06, 2020 at 05:56:06PM +0100, Greg KH wrote:
> > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > +static void mhi_release_device(struct device *dev)
> > > +{
> > > +	struct mhi_device *mhi_dev = to_mhi_device(dev);
> > > +
> > > +	if (mhi_dev->ul_chan)
> > > +		mhi_dev->ul_chan->mhi_dev = NULL;
> > 
> > That looks really odd.  Why didn't you just drop the reference you
> > should have grabbed here for this pointer?  You did properly increment
> > it when you saved it, right?  :)
> > 
> 
> Well, there is no reference count (kref) exist for mhi_dev.

Then something is wrong with your model :(

You can't save pointers off to things without reference counting, that
is going to cause you real problems.  See the coding style document for
all the details.

> And we really needed to NULL the mhi_dev to avoid any dangling
> reference to it.

Again, that's not how to do this correctly.

> The reason for not having kref is that, each mhi_dev will be used by
> maximum of 2 channels only. So thought that refcounting is not needed.
> Please correct me if I'm wrong.

Please read section 11 of Documentation/process/coding-style.rst

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 14/16] net: qrtr: Add MHI transport layer
  2020-02-11  3:50         ` Manivannan Sadhasivam
@ 2020-02-12  1:01           ` Chris Lew
  0 siblings, 0 replies; 62+ messages in thread
From: Chris Lew @ 2020-02-12  1:01 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: Jakub Kicinski, gregkh, arnd, smohanad, jhugo, kvalo,
	bjorn.andersson, hemantk, linux-arm-msm, linux-kernel,
	David S. Miller, netdev


On 2/10/2020 7:50 PM, Manivannan Sadhasivam wrote:
> Hi Chris,
>
> On Thu, Feb 06, 2020 at 04:14:19PM -0800, Chris Lew wrote:
>> On 2/4/2020 12:19 AM, Manivannan Sadhasivam wrote:
>>> Hi Jakub,
>>>
>>> On Mon, Feb 03, 2020 at 10:12:25AM -0800, Jakub Kicinski wrote:
>>>> On Fri, 31 Jan 2020 19:20:07 +0530, Manivannan Sadhasivam wrote:
>>>>> +/* From QRTR to MHI */
>>>>> +static void qcom_mhi_qrtr_ul_callback(struct mhi_device *mhi_dev,
>>>>> +				      struct mhi_result *mhi_res)
>>>>> +{
>>>>> +	struct qrtr_mhi_dev *qdev = dev_get_drvdata(&mhi_dev->dev);
>>>>> +	struct qrtr_mhi_pkt *pkt;
>>>>> +	unsigned long flags;
>>>>> +
>>>>> +	spin_lock_irqsave(&qdev->ul_lock, flags);
>>>>> +	pkt = list_first_entry(&qdev->ul_pkts, struct qrtr_mhi_pkt, node);
>>>>> +	list_del(&pkt->node);
>>>>> +	complete_all(&pkt->done);
>>>>> +
>>>>> +	kref_put(&pkt->refcount, qrtr_mhi_pkt_release);
>>>> Which kref_get() does this pair with?
>>>>
>>>> Looks like qcom_mhi_qrtr_send() will release a reference after
>>>> completion, too.
>>>>
>>> Yikes, there is some issue here...
>>>
>>> Acutally the issue is not in what you referred above but the overall kref
>>> handling itself. Please see below.
>>>
>>> kref_put() should be present in qcom_mhi_qrtr_ul_callback() as it will
>>> decrement the refcount which got incremented in qcom_mhi_qrtr_send(). It
>>> should be noted that kref_init() will fix the refcount to 1 and kref_get() will
>>> increment to 2. So for properly releasing the refcount to 0, we need to call
>>> kref_put() twice.
>>>
>>> So if all goes well, the refcount will get decremented twice in
>>> qcom_mhi_qrtr_ul_callback() as well as in qcom_mhi_qrtr_send() and we are good.
>>>
>>> But, if the transfer has failed ie., when qcom_mhi_qrtr_ul_callback() doesn't
>>> get called, then we are leaking the refcount. I need to rework the kref handling
>>> code in next iteration.
>>>
>>> Thanks for triggering this!
>>>
>>> Regards,
>>> Mani
>>>
>>>>> +	spin_unlock_irqrestore(&qdev->ul_lock, flags);
>>>>> +}
>> Hi Mani,
>>
>> I'm not sure if this was changed in your patches but MHI is supposed to give a
>> ul_callback() for any packet that is successfully queued. In the case of the
>> transfer failing, the ul_callback() should still be called so there should
>> be no refcount leaking. It is an essential assumption I made, if that no longer
>> holds true then the entire driver needs to be reworked.
>>
> Your assumption is correct. Only when the packet gets queued into the transfer
> ring, the ul_xfer_cb will be called irrespective of the transfer state (success
> or failure). But when the mhi_queue_transfer() returns even before queuing any
> packet, then we need to decrease the refcount in the error path.
>
> Please correct me if I'm wrong.

The error path for mhi_queue_transfer directly frees the packet structure since no
other context has a reference to those structs. If you wanted to clean it up and converge
using kref release to free, I think that would work. There are some things you'll have to
re-arrange like at what point the packet is added to the ul pkts list.

> Thanks,
> Mani
>
>> Thanks,
>> Chris
>>
>> -- 
>>
>> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-11 19:20       ` Greg KH
@ 2020-02-13 15:20         ` Manivannan Sadhasivam
  2020-02-13 15:34           ` Greg KH
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-13 15:20 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > Hi Greg,
> > 
> > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > --- /dev/null
> > > > +++ b/drivers/bus/mhi/core/init.c
> > > > @@ -0,0 +1,407 @@
> > > > +// SPDX-License-Identifier: GPL-2.0
> > > > +/*
> > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > + *
> > > > + */
> > > > +
> > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > 
> > > This should not be needed, right?  The bus/device name should give you
> > > all you need here from what I can tell.  So why is this needed?
> > > 
> > 
> > The log will have only the device name as like PCI-E. But that won't specify
> > where the error is coming from. Having "MHI" prefix helps the users to
> > quickly identify that the error is coming from MHI stack.
> 
> If the driver binds properly to the device, the name of the driver will
> be there in the message, so I suggest using that please.
> 
> No need for this prefix...
> 

So the driver name will be in the log but that won't help identifying where
the log is coming from. This is more important for MHI since it reuses the
`struct device` of the transport device like PCI-E. For instance, below is
the log without MHI prefix:

[   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
[   47.355724] ath11k_pci 0000:01:00.0: Power on setup success

As you can see, this gives the assumption that the log is coming from the
ath11k_pci driver. But the reality is, it is coming from MHI bus.

With the prefix added, we will get below:

[   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
[   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success

IMO, the prefix will give users a clear idea of logs and that will be very
useful for debugging.

Hope this clarifies.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-11 19:22       ` Greg KH
@ 2020-02-13 15:22         ` Manivannan Sadhasivam
  0 siblings, 0 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-13 15:22 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

Hi Greg,

On Tue, Feb 11, 2020 at 11:22:36AM -0800, Greg KH wrote:
> On Wed, Feb 12, 2020 at 12:41:47AM +0530, Manivannan Sadhasivam wrote:
> > Hi Greg,
> > 
> > On Thu, Feb 06, 2020 at 05:56:06PM +0100, Greg KH wrote:
> > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > +static void mhi_release_device(struct device *dev)
> > > > +{
> > > > +	struct mhi_device *mhi_dev = to_mhi_device(dev);
> > > > +
> > > > +	if (mhi_dev->ul_chan)
> > > > +		mhi_dev->ul_chan->mhi_dev = NULL;
> > > 
> > > That looks really odd.  Why didn't you just drop the reference you
> > > should have grabbed here for this pointer?  You did properly increment
> > > it when you saved it, right?  :)
> > > 
> > 
> > Well, there is no reference count (kref) exist for mhi_dev.
> 
> Then something is wrong with your model :(
> 
> You can't save pointers off to things without reference counting, that
> is going to cause you real problems.  See the coding style document for
> all the details.
> 
> > And we really needed to NULL the mhi_dev to avoid any dangling
> > reference to it.
> 
> Again, that's not how to do this correctly.
> 
> > The reason for not having kref is that, each mhi_dev will be used by
> > maximum of 2 channels only. So thought that refcounting is not needed.
> > Please correct me if I'm wrong.
> 
> Please read section 11 of Documentation/process/coding-style.rst
> 

I have fixed it and will send the next iteration soon.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-13 15:20         ` Manivannan Sadhasivam
@ 2020-02-13 15:34           ` Greg KH
  2020-02-13 15:48             ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-13 15:34 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > Hi Greg,
> > > 
> > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > --- /dev/null
> > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > @@ -0,0 +1,407 @@
> > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > +/*
> > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > + *
> > > > > + */
> > > > > +
> > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > 
> > > > This should not be needed, right?  The bus/device name should give you
> > > > all you need here from what I can tell.  So why is this needed?
> > > > 
> > > 
> > > The log will have only the device name as like PCI-E. But that won't specify
> > > where the error is coming from. Having "MHI" prefix helps the users to
> > > quickly identify that the error is coming from MHI stack.
> > 
> > If the driver binds properly to the device, the name of the driver will
> > be there in the message, so I suggest using that please.
> > 
> > No need for this prefix...
> > 
> 
> So the driver name will be in the log but that won't help identifying where
> the log is coming from. This is more important for MHI since it reuses the
> `struct device` of the transport device like PCI-E. For instance, below is
> the log without MHI prefix:
> 
> [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> 
> As you can see, this gives the assumption that the log is coming from the
> ath11k_pci driver. But the reality is, it is coming from MHI bus.

Then you should NOT be trying to "reuse" a struct device.

> With the prefix added, we will get below:
> 
> [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> 
> IMO, the prefix will give users a clear idea of logs and that will be very
> useful for debugging.
> 
> Hope this clarifies.

Don't try to reuse struct devices, if you are a bus, have your own
devices as that's the correct way to do things.

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-13 15:34           ` Greg KH
@ 2020-02-13 15:48             ` Manivannan Sadhasivam
  2020-02-13 15:53               ` Greg KH
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-13 15:48 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

Hi Greg,

On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > Hi Greg,
> > > > 
> > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > --- /dev/null
> > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > @@ -0,0 +1,407 @@
> > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > +/*
> > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > + *
> > > > > > + */
> > > > > > +
> > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > 
> > > > > This should not be needed, right?  The bus/device name should give you
> > > > > all you need here from what I can tell.  So why is this needed?
> > > > > 
> > > > 
> > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > quickly identify that the error is coming from MHI stack.
> > > 
> > > If the driver binds properly to the device, the name of the driver will
> > > be there in the message, so I suggest using that please.
> > > 
> > > No need for this prefix...
> > > 
> > 
> > So the driver name will be in the log but that won't help identifying where
> > the log is coming from. This is more important for MHI since it reuses the
> > `struct device` of the transport device like PCI-E. For instance, below is
> > the log without MHI prefix:
> > 
> > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > 
> > As you can see, this gives the assumption that the log is coming from the
> > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> 
> Then you should NOT be trying to "reuse" a struct device.
> 
> > With the prefix added, we will get below:
> > 
> > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > 
> > IMO, the prefix will give users a clear idea of logs and that will be very
> > useful for debugging.
> > 
> > Hope this clarifies.
> 
> Don't try to reuse struct devices, if you are a bus, have your own
> devices as that's the correct way to do things.
> 

I assumed that the buses relying on a different physical interface for the
actual communication can reuse the `struct device`. I can see that the MOXTET
bus driver already doing it. It reuses the `struct device` of SPI.

And this assumption has deep rooted in MHI bus design.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-13 15:48             ` Manivannan Sadhasivam
@ 2020-02-13 15:53               ` Greg KH
  2020-02-14 16:41                 ` Jeffrey Hugo
  2020-02-17  5:27                 ` Manivannan Sadhasivam
  0 siblings, 2 replies; 62+ messages in thread
From: Greg KH @ 2020-02-13 15:53 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> Hi Greg,
> 
> On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > Hi Greg,
> > > > > 
> > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > --- /dev/null
> > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > @@ -0,0 +1,407 @@
> > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > +/*
> > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > + *
> > > > > > > + */
> > > > > > > +
> > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > 
> > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > 
> > > > > 
> > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > quickly identify that the error is coming from MHI stack.
> > > > 
> > > > If the driver binds properly to the device, the name of the driver will
> > > > be there in the message, so I suggest using that please.
> > > > 
> > > > No need for this prefix...
> > > > 
> > > 
> > > So the driver name will be in the log but that won't help identifying where
> > > the log is coming from. This is more important for MHI since it reuses the
> > > `struct device` of the transport device like PCI-E. For instance, below is
> > > the log without MHI prefix:
> > > 
> > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > 
> > > As you can see, this gives the assumption that the log is coming from the
> > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > 
> > Then you should NOT be trying to "reuse" a struct device.
> > 
> > > With the prefix added, we will get below:
> > > 
> > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > 
> > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > useful for debugging.
> > > 
> > > Hope this clarifies.
> > 
> > Don't try to reuse struct devices, if you are a bus, have your own
> > devices as that's the correct way to do things.
> > 
> 
> I assumed that the buses relying on a different physical interface for the
> actual communication can reuse the `struct device`. I can see that the MOXTET
> bus driver already doing it. It reuses the `struct device` of SPI.

How can you reuse anything?

> And this assumption has deep rooted in MHI bus design.

Maybe I do not understand what this is at all, but a device can only be
on one "bus" at a time.  How is that being broken here?

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-13 15:53               ` Greg KH
@ 2020-02-14 16:41                 ` Jeffrey Hugo
  2020-02-17  5:27                 ` Manivannan Sadhasivam
  1 sibling, 0 replies; 62+ messages in thread
From: Jeffrey Hugo @ 2020-02-14 16:41 UTC (permalink / raw)
  To: Greg KH, Manivannan Sadhasivam
  Cc: arnd, smohanad, kvalo, bjorn.andersson, hemantk, linux-arm-msm,
	linux-kernel

On 2/13/2020 8:53 AM, Greg KH wrote:
> On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
>> Hi Greg,
>>
>> On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
>>> On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
>>>> On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
>>>>> On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
>>>>>> Hi Greg,
>>>>>>
>>>>>> On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
>>>>>>> On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
>>>>>>>> --- /dev/null
>>>>>>>> +++ b/drivers/bus/mhi/core/init.c
>>>>>>>> @@ -0,0 +1,407 @@
>>>>>>>> +// SPDX-License-Identifier: GPL-2.0
>>>>>>>> +/*
>>>>>>>> + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
>>>>>>>> + *
>>>>>>>> + */
>>>>>>>> +
>>>>>>>> +#define dev_fmt(fmt) "MHI: " fmt
>>>>>>>
>>>>>>> This should not be needed, right?  The bus/device name should give you
>>>>>>> all you need here from what I can tell.  So why is this needed?
>>>>>>>
>>>>>>
>>>>>> The log will have only the device name as like PCI-E. But that won't specify
>>>>>> where the error is coming from. Having "MHI" prefix helps the users to
>>>>>> quickly identify that the error is coming from MHI stack.
>>>>>
>>>>> If the driver binds properly to the device, the name of the driver will
>>>>> be there in the message, so I suggest using that please.
>>>>>
>>>>> No need for this prefix...
>>>>>
>>>>
>>>> So the driver name will be in the log but that won't help identifying where
>>>> the log is coming from. This is more important for MHI since it reuses the
>>>> `struct device` of the transport device like PCI-E. For instance, below is
>>>> the log without MHI prefix:
>>>>
>>>> [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
>>>> [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
>>>>
>>>> As you can see, this gives the assumption that the log is coming from the
>>>> ath11k_pci driver. But the reality is, it is coming from MHI bus.
>>>
>>> Then you should NOT be trying to "reuse" a struct device.
>>>
>>>> With the prefix added, we will get below:
>>>>
>>>> [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
>>>> [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
>>>>
>>>> IMO, the prefix will give users a clear idea of logs and that will be very
>>>> useful for debugging.
>>>>
>>>> Hope this clarifies.
>>>
>>> Don't try to reuse struct devices, if you are a bus, have your own
>>> devices as that's the correct way to do things.
>>>
>>
>> I assumed that the buses relying on a different physical interface for the
>> actual communication can reuse the `struct device`. I can see that the MOXTET
>> bus driver already doing it. It reuses the `struct device` of SPI.
> 
> How can you reuse anything?
> 
>> And this assumption has deep rooted in MHI bus design.
> 
> Maybe I do not understand what this is at all, but a device can only be
> on one "bus" at a time.  How is that being broken here?

Maybe I can bring some clarity to the discussion.  I hope I do not muddy 
things.

My interest in this series is I am developing a kernel driver for a 
product which uses MHI.  Unfortunately I am currently under embargo, 
otherwise I would post an RFC of my kernel driver which uses this 
series.  Let me try to dance around the embargo issue to hopefully 
illustrate how I aim to use this series.

Note that I expect my usecase is not the "defining" usecase.  I fully 
expect other usecases to exist.  IE what I hope to explain is not to be 
taken as the "one true way" to use MHI.

At the physical layer, my product is attached to a PCIE bus.  Thus my 
driver registers as a PCI driver for a device containing a specific 
VID/PID.  My product uses MHI as the "control path".  When I want to 
give commands to my product, they occur via the MHI protocol.  The "data 
path" occurs outside of MHI, and thus there is direct access to that 
hardware via PCIE.

In my usecase, my driver acts as both MHI Controller, and also MHI 
Client Driver.  Based on the PCIE device probe, my driver registers a 
MHI Controller.  Its going to be important for me that I have some means 
of associating the MHI Controller instance to the PCIE device instance 
(struct device) so that I can then correlate MHI devices to the PCIE 
device (more on that in a bit).  Right now, since the MHI controller 
essentially reuses the struct device from the PCIE device (provided by 
my driver), I can do that.

My MHI Controller exposes a number of channels, but for this discussion, 
the relevant one is the "control" channel.

My kernel driver also registers a MHI Client Driver which binds to this 
control channel.  Since its possible to have multiple instances of my 
product, I need to make a connection between the specific control 
channel device that my MHI Client Driver has probed on, and the PCIE 
device this control channel runs on so that I can store this connection 
in my struct that is in the driver data of the struct device of the pci 
device.  With that connection made, I can then correlate activity on the 
control MHI channel for a particular device with the services that my 
driver provides.

So, what we have, in my case, is a layering of buses, and its important 
to be able to identify the relationships (ie this particular MHI bus is 
associated with this particular PCI device on the PCI bus).

As I said, currently since the MHI bus is reusing the PCI struct device, 
its possible for my driver to make these connections as necessary.  If 
the MHI bus needs to have its own struct device instance, then how do 
you propose those connections be determined?  The MHI bus device is a 
child of the PCI device (thus from the MHI bus device, we dereference 
its parent)?

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-13 15:53               ` Greg KH
  2020-02-14 16:41                 ` Jeffrey Hugo
@ 2020-02-17  5:27                 ` Manivannan Sadhasivam
  2020-02-17 11:45                   ` Greg KH
  2020-02-17 11:59                   ` Greg KH
  1 sibling, 2 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-17  5:27 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

Hi Greg,

On Thu, Feb 13, 2020 at 07:53:02AM -0800, Greg KH wrote:
> On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> > Hi Greg,
> > 
> > On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > > Hi Greg,
> > > > > > 
> > > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > --- /dev/null
> > > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > > @@ -0,0 +1,407 @@
> > > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > > +/*
> > > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > > + *
> > > > > > > > + */
> > > > > > > > +
> > > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > > 
> > > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > > 
> > > > > > 
> > > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > > quickly identify that the error is coming from MHI stack.
> > > > > 
> > > > > If the driver binds properly to the device, the name of the driver will
> > > > > be there in the message, so I suggest using that please.
> > > > > 
> > > > > No need for this prefix...
> > > > > 
> > > > 
> > > > So the driver name will be in the log but that won't help identifying where
> > > > the log is coming from. This is more important for MHI since it reuses the
> > > > `struct device` of the transport device like PCI-E. For instance, below is
> > > > the log without MHI prefix:
> > > > 
> > > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > > 
> > > > As you can see, this gives the assumption that the log is coming from the
> > > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > > 
> > > Then you should NOT be trying to "reuse" a struct device.
> > > 
> > > > With the prefix added, we will get below:
> > > > 
> > > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > > 
> > > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > > useful for debugging.
> > > > 
> > > > Hope this clarifies.
> > > 
> > > Don't try to reuse struct devices, if you are a bus, have your own
> > > devices as that's the correct way to do things.
> > > 
> > 
> > I assumed that the buses relying on a different physical interface for the
> > actual communication can reuse the `struct device`. I can see that the MOXTET
> > bus driver already doing it. It reuses the `struct device` of SPI.
> 
> How can you reuse anything?
> 
> > And this assumption has deep rooted in MHI bus design.
> 
> Maybe I do not understand what this is at all, but a device can only be
> on one "bus" at a time.  How is that being broken here?
> 

Let me share some insight on how it is being used:

The MHI bus sits on top of the actual physical bus like PCI-E and requires
the physical bus for doing activities like allocating I/O virtual address,
runtime PM etc... The part which gets tied to the PCI-E from MHI is called MHI
controller driver. This MHI controller driver is also the actual PCI-E driver
managing the device.

For instance, we have QCA6390 PCI-E WLAN device. For this device, there is a
ath11k PCI-E driver and the same driver also registers as a MHI controller and
acts as a MHI controller driver. This is where I referred to reusing the PCI-E
struct device. It's not that MHI bus itself is reusing the PCI-E struct device
but we need the PCI-E device pointer to do above mentioned IOVA, PM operations
in some places. One of the usage is below:

```
void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
```

There was some discussion about it here: https://lkml.org/lkml/2020/1/27/21

The MHI bus itself has the struct device and it is the child of the physical
bus (PCI-E in this case).

Now coming to your actual question of why using a custom "MHI" prefix for
dev_ APIs. I agree that if we use the struct device of MHI bus it is not at all
needed. The fact that we are using "mhi_cntrl->dev" (which points to PCI-E dev)
is what confusing and it can be avoided.

I'll change it in next iteration.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17  5:27                 ` Manivannan Sadhasivam
@ 2020-02-17 11:45                   ` Greg KH
  2020-02-17 11:53                     ` Manivannan Sadhasivam
  2020-02-17 11:59                   ` Greg KH
  1 sibling, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-17 11:45 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 10:57:43AM +0530, Manivannan Sadhasivam wrote:
> Hi Greg,
> 
> On Thu, Feb 13, 2020 at 07:53:02AM -0800, Greg KH wrote:
> > On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> > > Hi Greg,
> > > 
> > > On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > > > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > > > Hi Greg,
> > > > > > > 
> > > > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > > --- /dev/null
> > > > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > > > @@ -0,0 +1,407 @@
> > > > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > > > +/*
> > > > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > > > + *
> > > > > > > > > + */
> > > > > > > > > +
> > > > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > > > 
> > > > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > > > 
> > > > > > > 
> > > > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > > > quickly identify that the error is coming from MHI stack.
> > > > > > 
> > > > > > If the driver binds properly to the device, the name of the driver will
> > > > > > be there in the message, so I suggest using that please.
> > > > > > 
> > > > > > No need for this prefix...
> > > > > > 
> > > > > 
> > > > > So the driver name will be in the log but that won't help identifying where
> > > > > the log is coming from. This is more important for MHI since it reuses the
> > > > > `struct device` of the transport device like PCI-E. For instance, below is
> > > > > the log without MHI prefix:
> > > > > 
> > > > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > > > 
> > > > > As you can see, this gives the assumption that the log is coming from the
> > > > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > > > 
> > > > Then you should NOT be trying to "reuse" a struct device.
> > > > 
> > > > > With the prefix added, we will get below:
> > > > > 
> > > > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > > > 
> > > > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > > > useful for debugging.
> > > > > 
> > > > > Hope this clarifies.
> > > > 
> > > > Don't try to reuse struct devices, if you are a bus, have your own
> > > > devices as that's the correct way to do things.
> > > > 
> > > 
> > > I assumed that the buses relying on a different physical interface for the
> > > actual communication can reuse the `struct device`. I can see that the MOXTET
> > > bus driver already doing it. It reuses the `struct device` of SPI.
> > 
> > How can you reuse anything?
> > 
> > > And this assumption has deep rooted in MHI bus design.
> > 
> > Maybe I do not understand what this is at all, but a device can only be
> > on one "bus" at a time.  How is that being broken here?
> > 
> 
> Let me share some insight on how it is being used:
> 
> The MHI bus sits on top of the actual physical bus like PCI-E and requires
> the physical bus for doing activities like allocating I/O virtual address,
> runtime PM etc... The part which gets tied to the PCI-E from MHI is called MHI
> controller driver. This MHI controller driver is also the actual PCI-E driver
> managing the device.
> 
> For instance, we have QCA6390 PCI-E WLAN device. For this device, there is a
> ath11k PCI-E driver and the same driver also registers as a MHI controller and
> acts as a MHI controller driver. This is where I referred to reusing the PCI-E
> struct device. It's not that MHI bus itself is reusing the PCI-E struct device
> but we need the PCI-E device pointer to do above mentioned IOVA, PM operations
> in some places. One of the usage is below:
> 
> ```
> void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
> ```
> 
> There was some discussion about it here: https://lkml.org/lkml/2020/1/27/21
> 
> The MHI bus itself has the struct device and it is the child of the physical
> bus (PCI-E in this case).
> 
> Now coming to your actual question of why using a custom "MHI" prefix for
> dev_ APIs. I agree that if we use the struct device of MHI bus it is not at all
> needed. The fact that we are using "mhi_cntrl->dev" (which points to PCI-E dev)
> is what confusing and it can be avoided.

You should also rename "dev" there, as that is really a "pci device".
So use the real pci device and name it as "parent_pci" or something like
that, so we know just by looking at it as to what it really is.

Especially as traditionally "->dev" is the device structure for _this_
device, not another one.

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17 11:45                   ` Greg KH
@ 2020-02-17 11:53                     ` Manivannan Sadhasivam
  0 siblings, 0 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-17 11:53 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 12:45:15PM +0100, Greg KH wrote:
> On Mon, Feb 17, 2020 at 10:57:43AM +0530, Manivannan Sadhasivam wrote:
> > Hi Greg,
> > 
> > On Thu, Feb 13, 2020 at 07:53:02AM -0800, Greg KH wrote:
> > > On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> > > > Hi Greg,
> > > > 
> > > > On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > > > > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > > > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > Hi Greg,
> > > > > > > > 
> > > > > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > > > --- /dev/null
> > > > > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > > > > @@ -0,0 +1,407 @@
> > > > > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > > > > +/*
> > > > > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > > > > + *
> > > > > > > > > > + */
> > > > > > > > > > +
> > > > > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > > > > 
> > > > > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > > > > 
> > > > > > > > 
> > > > > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > > > > quickly identify that the error is coming from MHI stack.
> > > > > > > 
> > > > > > > If the driver binds properly to the device, the name of the driver will
> > > > > > > be there in the message, so I suggest using that please.
> > > > > > > 
> > > > > > > No need for this prefix...
> > > > > > > 
> > > > > > 
> > > > > > So the driver name will be in the log but that won't help identifying where
> > > > > > the log is coming from. This is more important for MHI since it reuses the
> > > > > > `struct device` of the transport device like PCI-E. For instance, below is
> > > > > > the log without MHI prefix:
> > > > > > 
> > > > > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > > > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > > > > 
> > > > > > As you can see, this gives the assumption that the log is coming from the
> > > > > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > > > > 
> > > > > Then you should NOT be trying to "reuse" a struct device.
> > > > > 
> > > > > > With the prefix added, we will get below:
> > > > > > 
> > > > > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > > > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > > > > 
> > > > > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > > > > useful for debugging.
> > > > > > 
> > > > > > Hope this clarifies.
> > > > > 
> > > > > Don't try to reuse struct devices, if you are a bus, have your own
> > > > > devices as that's the correct way to do things.
> > > > > 
> > > > 
> > > > I assumed that the buses relying on a different physical interface for the
> > > > actual communication can reuse the `struct device`. I can see that the MOXTET
> > > > bus driver already doing it. It reuses the `struct device` of SPI.
> > > 
> > > How can you reuse anything?
> > > 
> > > > And this assumption has deep rooted in MHI bus design.
> > > 
> > > Maybe I do not understand what this is at all, but a device can only be
> > > on one "bus" at a time.  How is that being broken here?
> > > 
> > 
> > Let me share some insight on how it is being used:
> > 
> > The MHI bus sits on top of the actual physical bus like PCI-E and requires
> > the physical bus for doing activities like allocating I/O virtual address,
> > runtime PM etc... The part which gets tied to the PCI-E from MHI is called MHI
> > controller driver. This MHI controller driver is also the actual PCI-E driver
> > managing the device.
> > 
> > For instance, we have QCA6390 PCI-E WLAN device. For this device, there is a
> > ath11k PCI-E driver and the same driver also registers as a MHI controller and
> > acts as a MHI controller driver. This is where I referred to reusing the PCI-E
> > struct device. It's not that MHI bus itself is reusing the PCI-E struct device
> > but we need the PCI-E device pointer to do above mentioned IOVA, PM operations
> > in some places. One of the usage is below:
> > 
> > ```
> > void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
> > ```
> > 
> > There was some discussion about it here: https://lkml.org/lkml/2020/1/27/21
> > 
> > The MHI bus itself has the struct device and it is the child of the physical
> > bus (PCI-E in this case).
> > 
> > Now coming to your actual question of why using a custom "MHI" prefix for
> > dev_ APIs. I agree that if we use the struct device of MHI bus it is not at all
> > needed. The fact that we are using "mhi_cntrl->dev" (which points to PCI-E dev)
> > is what confusing and it can be avoided.
> 
> You should also rename "dev" there, as that is really a "pci device".
> So use the real pci device and name it as "parent_pci" or something like
> that, so we know just by looking at it as to what it really is.
> 
> Especially as traditionally "->dev" is the device structure for _this_
> device, not another one.
> 

Okay, sure. I'll rename it as "pdev" to indicate it as the parent device.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17  5:27                 ` Manivannan Sadhasivam
  2020-02-17 11:45                   ` Greg KH
@ 2020-02-17 11:59                   ` Greg KH
  2020-02-17 13:04                     ` Manivannan Sadhasivam
  1 sibling, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-17 11:59 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 10:57:43AM +0530, Manivannan Sadhasivam wrote:
> Hi Greg,
> 
> On Thu, Feb 13, 2020 at 07:53:02AM -0800, Greg KH wrote:
> > On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> > > Hi Greg,
> > > 
> > > On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > > > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > > > Hi Greg,
> > > > > > > 
> > > > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > > --- /dev/null
> > > > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > > > @@ -0,0 +1,407 @@
> > > > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > > > +/*
> > > > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > > > + *
> > > > > > > > > + */
> > > > > > > > > +
> > > > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > > > 
> > > > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > > > 
> > > > > > > 
> > > > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > > > quickly identify that the error is coming from MHI stack.
> > > > > > 
> > > > > > If the driver binds properly to the device, the name of the driver will
> > > > > > be there in the message, so I suggest using that please.
> > > > > > 
> > > > > > No need for this prefix...
> > > > > > 
> > > > > 
> > > > > So the driver name will be in the log but that won't help identifying where
> > > > > the log is coming from. This is more important for MHI since it reuses the
> > > > > `struct device` of the transport device like PCI-E. For instance, below is
> > > > > the log without MHI prefix:
> > > > > 
> > > > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > > > 
> > > > > As you can see, this gives the assumption that the log is coming from the
> > > > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > > > 
> > > > Then you should NOT be trying to "reuse" a struct device.
> > > > 
> > > > > With the prefix added, we will get below:
> > > > > 
> > > > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > > > 
> > > > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > > > useful for debugging.
> > > > > 
> > > > > Hope this clarifies.
> > > > 
> > > > Don't try to reuse struct devices, if you are a bus, have your own
> > > > devices as that's the correct way to do things.
> > > > 
> > > 
> > > I assumed that the buses relying on a different physical interface for the
> > > actual communication can reuse the `struct device`. I can see that the MOXTET
> > > bus driver already doing it. It reuses the `struct device` of SPI.
> > 
> > How can you reuse anything?
> > 
> > > And this assumption has deep rooted in MHI bus design.
> > 
> > Maybe I do not understand what this is at all, but a device can only be
> > on one "bus" at a time.  How is that being broken here?
> > 
> 
> Let me share some insight on how it is being used:
> 
> The MHI bus sits on top of the actual physical bus like PCI-E and requires
> the physical bus for doing activities like allocating I/O virtual address,
> runtime PM etc... The part which gets tied to the PCI-E from MHI is called MHI
> controller driver. This MHI controller driver is also the actual PCI-E driver
> managing the device.
> 
> For instance, we have QCA6390 PCI-E WLAN device. For this device, there is a
> ath11k PCI-E driver and the same driver also registers as a MHI controller and
> acts as a MHI controller driver. This is where I referred to reusing the PCI-E
> struct device. It's not that MHI bus itself is reusing the PCI-E struct device
> but we need the PCI-E device pointer to do above mentioned IOVA, PM operations
> in some places. One of the usage is below:
> 
> ```
> void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
> ```

Wait, why do you need to call this with the parent dev?  Why not with
your struct device?  What does the parent pointer have that yours does
not?  Is it not correctly having whatever dma attributes the parent has
set properly for your device as well?  If not, why not just fix that and
then _your_ device can be doing the allocation?

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17 11:59                   ` Greg KH
@ 2020-02-17 13:04                     ` Manivannan Sadhasivam
  2020-02-17 14:15                       ` Greg KH
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-17 13:04 UTC (permalink / raw)
  To: Greg KH
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 12:59:30PM +0100, Greg KH wrote:
> On Mon, Feb 17, 2020 at 10:57:43AM +0530, Manivannan Sadhasivam wrote:
> > Hi Greg,
> > 
> > On Thu, Feb 13, 2020 at 07:53:02AM -0800, Greg KH wrote:
> > > On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> > > > Hi Greg,
> > > > 
> > > > On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > > > > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > > > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > Hi Greg,
> > > > > > > > 
> > > > > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > > > --- /dev/null
> > > > > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > > > > @@ -0,0 +1,407 @@
> > > > > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > > > > +/*
> > > > > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > > > > + *
> > > > > > > > > > + */
> > > > > > > > > > +
> > > > > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > > > > 
> > > > > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > > > > 
> > > > > > > > 
> > > > > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > > > > quickly identify that the error is coming from MHI stack.
> > > > > > > 
> > > > > > > If the driver binds properly to the device, the name of the driver will
> > > > > > > be there in the message, so I suggest using that please.
> > > > > > > 
> > > > > > > No need for this prefix...
> > > > > > > 
> > > > > > 
> > > > > > So the driver name will be in the log but that won't help identifying where
> > > > > > the log is coming from. This is more important for MHI since it reuses the
> > > > > > `struct device` of the transport device like PCI-E. For instance, below is
> > > > > > the log without MHI prefix:
> > > > > > 
> > > > > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > > > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > > > > 
> > > > > > As you can see, this gives the assumption that the log is coming from the
> > > > > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > > > > 
> > > > > Then you should NOT be trying to "reuse" a struct device.
> > > > > 
> > > > > > With the prefix added, we will get below:
> > > > > > 
> > > > > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > > > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > > > > 
> > > > > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > > > > useful for debugging.
> > > > > > 
> > > > > > Hope this clarifies.
> > > > > 
> > > > > Don't try to reuse struct devices, if you are a bus, have your own
> > > > > devices as that's the correct way to do things.
> > > > > 
> > > > 
> > > > I assumed that the buses relying on a different physical interface for the
> > > > actual communication can reuse the `struct device`. I can see that the MOXTET
> > > > bus driver already doing it. It reuses the `struct device` of SPI.
> > > 
> > > How can you reuse anything?
> > > 
> > > > And this assumption has deep rooted in MHI bus design.
> > > 
> > > Maybe I do not understand what this is at all, but a device can only be
> > > on one "bus" at a time.  How is that being broken here?
> > > 
> > 
> > Let me share some insight on how it is being used:
> > 
> > The MHI bus sits on top of the actual physical bus like PCI-E and requires
> > the physical bus for doing activities like allocating I/O virtual address,
> > runtime PM etc... The part which gets tied to the PCI-E from MHI is called MHI
> > controller driver. This MHI controller driver is also the actual PCI-E driver
> > managing the device.
> > 
> > For instance, we have QCA6390 PCI-E WLAN device. For this device, there is a
> > ath11k PCI-E driver and the same driver also registers as a MHI controller and
> > acts as a MHI controller driver. This is where I referred to reusing the PCI-E
> > struct device. It's not that MHI bus itself is reusing the PCI-E struct device
> > but we need the PCI-E device pointer to do above mentioned IOVA, PM operations
> > in some places. One of the usage is below:
> > 
> > ```
> > void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
> > ```
> 
> Wait, why do you need to call this with the parent dev?  Why not with
> your struct device?  What does the parent pointer have that yours does
> not?  Is it not correctly having whatever dma attributes the parent has
> set properly for your device as well?  If not, why not just fix that and
> then _your_ device can be doing the allocation?
> 

This is _one_ of the usecases of the parent dev. We are also using it to manage
the runtime PM operations of the physical device (pcie) when the MHI stack goes
into respective states. For instance,

```
        if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
                mhi_cntrl->runtime_get(mhi_cntrl);
                mhi_cntrl->runtime_put(mhi_cntrl);
        }
```

These runtime_put() and runtime_get() are the callbacks to be provided by the
controller drivers for managing its runtime PM states.

Also, the MHI devices for the channels will be created later on after the
controller probe, so at that time we need this parent dev to set the MHI device
parent:

```
struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
{
...
dev->parent = mhi_cntrl->dev;
...
```

Hence, having the parent dev pointer really helps.

Thanks,
Mani
> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17 13:04                     ` Manivannan Sadhasivam
@ 2020-02-17 14:15                       ` Greg KH
  2020-02-17 16:04                         ` Arnd Bergmann
  0 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-17 14:15 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: arnd, smohanad, jhugo, kvalo, bjorn.andersson, hemantk,
	linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 06:34:19PM +0530, Manivannan Sadhasivam wrote:
> On Mon, Feb 17, 2020 at 12:59:30PM +0100, Greg KH wrote:
> > On Mon, Feb 17, 2020 at 10:57:43AM +0530, Manivannan Sadhasivam wrote:
> > > Hi Greg,
> > > 
> > > On Thu, Feb 13, 2020 at 07:53:02AM -0800, Greg KH wrote:
> > > > On Thu, Feb 13, 2020 at 09:18:09PM +0530, Manivannan Sadhasivam wrote:
> > > > > Hi Greg,
> > > > > 
> > > > > On Thu, Feb 13, 2020 at 07:34:18AM -0800, Greg KH wrote:
> > > > > > On Thu, Feb 13, 2020 at 08:50:13PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > On Tue, Feb 11, 2020 at 11:20:55AM -0800, Greg KH wrote:
> > > > > > > > On Wed, Feb 12, 2020 at 12:11:30AM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > > Hi Greg,
> > > > > > > > > 
> > > > > > > > > On Thu, Feb 06, 2020 at 05:57:55PM +0100, Greg KH wrote:
> > > > > > > > > > On Fri, Jan 31, 2020 at 07:19:55PM +0530, Manivannan Sadhasivam wrote:
> > > > > > > > > > > --- /dev/null
> > > > > > > > > > > +++ b/drivers/bus/mhi/core/init.c
> > > > > > > > > > > @@ -0,0 +1,407 @@
> > > > > > > > > > > +// SPDX-License-Identifier: GPL-2.0
> > > > > > > > > > > +/*
> > > > > > > > > > > + * Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.
> > > > > > > > > > > + *
> > > > > > > > > > > + */
> > > > > > > > > > > +
> > > > > > > > > > > +#define dev_fmt(fmt) "MHI: " fmt
> > > > > > > > > > 
> > > > > > > > > > This should not be needed, right?  The bus/device name should give you
> > > > > > > > > > all you need here from what I can tell.  So why is this needed?
> > > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > The log will have only the device name as like PCI-E. But that won't specify
> > > > > > > > > where the error is coming from. Having "MHI" prefix helps the users to
> > > > > > > > > quickly identify that the error is coming from MHI stack.
> > > > > > > > 
> > > > > > > > If the driver binds properly to the device, the name of the driver will
> > > > > > > > be there in the message, so I suggest using that please.
> > > > > > > > 
> > > > > > > > No need for this prefix...
> > > > > > > > 
> > > > > > > 
> > > > > > > So the driver name will be in the log but that won't help identifying where
> > > > > > > the log is coming from. This is more important for MHI since it reuses the
> > > > > > > `struct device` of the transport device like PCI-E. For instance, below is
> > > > > > > the log without MHI prefix:
> > > > > > > 
> > > > > > > [   47.355582] ath11k_pci 0000:01:00.0: Requested to power on
> > > > > > > [   47.355724] ath11k_pci 0000:01:00.0: Power on setup success
> > > > > > > 
> > > > > > > As you can see, this gives the assumption that the log is coming from the
> > > > > > > ath11k_pci driver. But the reality is, it is coming from MHI bus.
> > > > > > 
> > > > > > Then you should NOT be trying to "reuse" a struct device.
> > > > > > 
> > > > > > > With the prefix added, we will get below:
> > > > > > > 
> > > > > > > [   47.355582] ath11k_pci 0000:01:00.0: MHI: Requested to power on
> > > > > > > [   47.355724] ath11k_pci 0000:01:00.0: MHI: Power on setup success
> > > > > > > 
> > > > > > > IMO, the prefix will give users a clear idea of logs and that will be very
> > > > > > > useful for debugging.
> > > > > > > 
> > > > > > > Hope this clarifies.
> > > > > > 
> > > > > > Don't try to reuse struct devices, if you are a bus, have your own
> > > > > > devices as that's the correct way to do things.
> > > > > > 
> > > > > 
> > > > > I assumed that the buses relying on a different physical interface for the
> > > > > actual communication can reuse the `struct device`. I can see that the MOXTET
> > > > > bus driver already doing it. It reuses the `struct device` of SPI.
> > > > 
> > > > How can you reuse anything?
> > > > 
> > > > > And this assumption has deep rooted in MHI bus design.
> > > > 
> > > > Maybe I do not understand what this is at all, but a device can only be
> > > > on one "bus" at a time.  How is that being broken here?
> > > > 
> > > 
> > > Let me share some insight on how it is being used:
> > > 
> > > The MHI bus sits on top of the actual physical bus like PCI-E and requires
> > > the physical bus for doing activities like allocating I/O virtual address,
> > > runtime PM etc... The part which gets tied to the PCI-E from MHI is called MHI
> > > controller driver. This MHI controller driver is also the actual PCI-E driver
> > > managing the device.
> > > 
> > > For instance, we have QCA6390 PCI-E WLAN device. For this device, there is a
> > > ath11k PCI-E driver and the same driver also registers as a MHI controller and
> > > acts as a MHI controller driver. This is where I referred to reusing the PCI-E
> > > struct device. It's not that MHI bus itself is reusing the PCI-E struct device
> > > but we need the PCI-E device pointer to do above mentioned IOVA, PM operations
> > > in some places. One of the usage is below:
> > > 
> > > ```
> > > void *buf = dma_alloc_coherent(mhi_cntrl->dev, size, dma_handle, gfp);
> > > ```
> > 
> > Wait, why do you need to call this with the parent dev?  Why not with
> > your struct device?  What does the parent pointer have that yours does
> > not?  Is it not correctly having whatever dma attributes the parent has
> > set properly for your device as well?  If not, why not just fix that and
> > then _your_ device can be doing the allocation?
> > 
> 
> This is _one_ of the usecases of the parent dev. We are also using it to manage
> the runtime PM operations of the physical device (pcie) when the MHI stack goes
> into respective states. For instance,
> 
> ```
>         if (MHI_PM_IN_SUSPEND_STATE(mhi_cntrl->pm_state)) {
>                 mhi_cntrl->runtime_get(mhi_cntrl);
>                 mhi_cntrl->runtime_put(mhi_cntrl);
>         }
> ```
> 
> These runtime_put() and runtime_get() are the callbacks to be provided by the
> controller drivers for managing its runtime PM states.
> 
> Also, the MHI devices for the channels will be created later on after the
> controller probe, so at that time we need this parent dev to set the MHI device
> parent:
> 
> ```
> struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
> {
> ...
> dev->parent = mhi_cntrl->dev;
> ...
> ```
> 
> Hence, having the parent dev pointer really helps.

Yes, saving the parent device is fine, but you should be doing your own
dma calls using _your_ device, not the parents.  Only mess with the
parent pointer if you need to do something "normal" for a parent.

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17 14:15                       ` Greg KH
@ 2020-02-17 16:04                         ` Arnd Bergmann
  2020-02-17 16:32                           ` Greg KH
  0 siblings, 1 reply; 62+ messages in thread
From: Arnd Bergmann @ 2020-02-17 16:04 UTC (permalink / raw)
  To: Greg KH
  Cc: Manivannan Sadhasivam, smohanad, Jeffrey Hugo, Kalle Valo,
	Bjorn Andersson, hemantk, linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 3:15 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> On Mon, Feb 17, 2020 at 06:34:19PM +0530, Manivannan Sadhasivam wrote:
> > On Mon, Feb 17, 2020 at 12:59:30PM +0100, Greg KH wrote:
> > ```
> > struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
> > {
> > ...
> > dev->parent = mhi_cntrl->dev;
> > ...
> > ```
> >
> > Hence, having the parent dev pointer really helps.
>
> Yes, saving the parent device is fine, but you should be doing your own
> dma calls using _your_ device, not the parents.  Only mess with the
> parent pointer if you need to do something "normal" for a parent.

The MHI device is not involved in DMA at all, as it is not a DMA master,
and has no knowledge of the memory management or whether there
is any DMA at all. I think it is the right abstraction for an MHI driver to
pass kernel pointers into the subsystem interfaces, which then get
mapped by the bus driver that owns the DMA master.

This is similar to how e.g. USB drivers pass data into the USB core
interfaces, which then get the HCI driver to map/unmap it into the
DMA masters.

      Arnd

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 11/16] bus: mhi: core: Add support for data transfer
  2020-01-31 13:50 ` [PATCH v2 11/16] bus: mhi: core: Add support for data transfer Manivannan Sadhasivam
  2020-02-06 20:16   ` Jeffrey Hugo
@ 2020-02-17 16:13   ` Arnd Bergmann
  2020-02-17 16:47     ` Manivannan Sadhasivam
  1 sibling, 1 reply; 62+ messages in thread
From: Arnd Bergmann @ 2020-02-17 16:13 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: gregkh, smohanad, Jeffrey Hugo, Kalle Valo, Bjorn Andersson,
	hemantk, linux-arm-msm, linux-kernel

On Fri, Jan 31, 2020 at 2:51 PM Manivannan Sadhasivam
<manivannan.sadhasivam@linaro.org> wrote:

> @@ -648,6 +715,31 @@ static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
>                 mhi_chan->db_cfg.pollcfg = ch_cfg->pollcfg;
>                 mhi_chan->xfer_type = ch_cfg->data_type;
>
> +               switch (mhi_chan->xfer_type) {
> +               case MHI_BUF_RAW:
> +                       mhi_chan->gen_tre = mhi_gen_tre;
> +                       mhi_chan->queue_xfer = mhi_queue_buf;
> +                       break;
> +               case MHI_BUF_SKB:
> +                       mhi_chan->queue_xfer = mhi_queue_skb;
> +                       break;
> +               case MHI_BUF_SCLIST:
> +                       mhi_chan->gen_tre = mhi_gen_tre;
> +                       mhi_chan->queue_xfer = mhi_queue_sclist;
> +                       break;
> +               case MHI_BUF_NOP:
> +                       mhi_chan->queue_xfer = mhi_queue_nop;
> +                       break;
> +               case MHI_BUF_DMA:
> +               case MHI_BUF_RSC_DMA:
> +                       mhi_chan->queue_xfer = mhi_queue_dma;
> +                       break;
> +               default:
> +                       dev_err(mhi_cntrl->dev,
> +                               "Channel datatype not supported\n");
> +                       goto error_chan_cfg;
> +               }
> +

While looking through the driver to see how the DMA gets handled, I came
across the multitude of mhi_queue_* functions, which seems like a
layering violation to me, given that they are all implemented by the
core code as well, and the client driver needs to be aware of
which one to call. Are you able to lift these out of the common interface
and make the client driver call these directly, or maybe provide a direct
interface based on mhi_buf_info to replace these?

      Arnd

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17 16:04                         ` Arnd Bergmann
@ 2020-02-17 16:32                           ` Greg KH
  2020-02-17 17:50                             ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Greg KH @ 2020-02-17 16:32 UTC (permalink / raw)
  To: Arnd Bergmann
  Cc: Manivannan Sadhasivam, smohanad, Jeffrey Hugo, Kalle Valo,
	Bjorn Andersson, hemantk, linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 05:04:52PM +0100, Arnd Bergmann wrote:
> On Mon, Feb 17, 2020 at 3:15 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > On Mon, Feb 17, 2020 at 06:34:19PM +0530, Manivannan Sadhasivam wrote:
> > > On Mon, Feb 17, 2020 at 12:59:30PM +0100, Greg KH wrote:
> > > ```
> > > struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
> > > {
> > > ...
> > > dev->parent = mhi_cntrl->dev;
> > > ...
> > > ```
> > >
> > > Hence, having the parent dev pointer really helps.
> >
> > Yes, saving the parent device is fine, but you should be doing your own
> > dma calls using _your_ device, not the parents.  Only mess with the
> > parent pointer if you need to do something "normal" for a parent.
> 
> The MHI device is not involved in DMA at all, as it is not a DMA master,
> and has no knowledge of the memory management or whether there
> is any DMA at all. I think it is the right abstraction for an MHI driver to
> pass kernel pointers into the subsystem interfaces, which then get
> mapped by the bus driver that owns the DMA master.
> 
> This is similar to how e.g. USB drivers pass data into the USB core
> interfaces, which then get the HCI driver to map/unmap it into the
> DMA masters.

Ok, then this needs to be named a whole lot better than the original
"dev" name had it.  Heck, even "parent" does not show that type of
representation, make it "controller" or something else a whole lot more
descriptive of what it really is please.

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 11/16] bus: mhi: core: Add support for data transfer
  2020-02-17 16:13   ` Arnd Bergmann
@ 2020-02-17 16:47     ` Manivannan Sadhasivam
  2020-02-18  5:51       ` Manivannan Sadhasivam
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-17 16:47 UTC (permalink / raw)
  To: Arnd Bergmann
  Cc: gregkh, smohanad, Jeffrey Hugo, Kalle Valo, Bjorn Andersson,
	hemantk, linux-arm-msm, linux-kernel

Hi Arnd,

On Mon, Feb 17, 2020 at 05:13:37PM +0100, Arnd Bergmann wrote:
> On Fri, Jan 31, 2020 at 2:51 PM Manivannan Sadhasivam
> <manivannan.sadhasivam@linaro.org> wrote:
> 
> > @@ -648,6 +715,31 @@ static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
> >                 mhi_chan->db_cfg.pollcfg = ch_cfg->pollcfg;
> >                 mhi_chan->xfer_type = ch_cfg->data_type;
> >
> > +               switch (mhi_chan->xfer_type) {
> > +               case MHI_BUF_RAW:
> > +                       mhi_chan->gen_tre = mhi_gen_tre;
> > +                       mhi_chan->queue_xfer = mhi_queue_buf;
> > +                       break;
> > +               case MHI_BUF_SKB:
> > +                       mhi_chan->queue_xfer = mhi_queue_skb;
> > +                       break;
> > +               case MHI_BUF_SCLIST:
> > +                       mhi_chan->gen_tre = mhi_gen_tre;
> > +                       mhi_chan->queue_xfer = mhi_queue_sclist;
> > +                       break;
> > +               case MHI_BUF_NOP:
> > +                       mhi_chan->queue_xfer = mhi_queue_nop;
> > +                       break;
> > +               case MHI_BUF_DMA:
> > +               case MHI_BUF_RSC_DMA:
> > +                       mhi_chan->queue_xfer = mhi_queue_dma;
> > +                       break;
> > +               default:
> > +                       dev_err(mhi_cntrl->dev,
> > +                               "Channel datatype not supported\n");
> > +                       goto error_chan_cfg;
> > +               }
> > +
> 
> While looking through the driver to see how the DMA gets handled, I came
> across the multitude of mhi_queue_* functions, which seems like a
> layering violation to me, given that they are all implemented by the
> core code as well, and the client driver needs to be aware of
> which one to call. Are you able to lift these out of the common interface
> and make the client driver call these directly, or maybe provide a direct
> interface based on mhi_buf_info to replace these?
> 

It sounds reasonable to me. Let me discuss this internally with Qcom guys to
see if they have any objections.

Thanks,
Mani

>       Arnd

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers
  2020-02-17 16:32                           ` Greg KH
@ 2020-02-17 17:50                             ` Manivannan Sadhasivam
  0 siblings, 0 replies; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-17 17:50 UTC (permalink / raw)
  To: Greg KH
  Cc: Arnd Bergmann, smohanad, Jeffrey Hugo, Kalle Valo,
	Bjorn Andersson, hemantk, linux-arm-msm, linux-kernel

On Mon, Feb 17, 2020 at 05:32:05PM +0100, Greg KH wrote:
> On Mon, Feb 17, 2020 at 05:04:52PM +0100, Arnd Bergmann wrote:
> > On Mon, Feb 17, 2020 at 3:15 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > > On Mon, Feb 17, 2020 at 06:34:19PM +0530, Manivannan Sadhasivam wrote:
> > > > On Mon, Feb 17, 2020 at 12:59:30PM +0100, Greg KH wrote:
> > > > ```
> > > > struct mhi_device *mhi_alloc_device(struct mhi_controller *mhi_cntrl)
> > > > {
> > > > ...
> > > > dev->parent = mhi_cntrl->dev;
> > > > ...
> > > > ```
> > > >
> > > > Hence, having the parent dev pointer really helps.
> > >
> > > Yes, saving the parent device is fine, but you should be doing your own
> > > dma calls using _your_ device, not the parents.  Only mess with the
> > > parent pointer if you need to do something "normal" for a parent.
> > 
> > The MHI device is not involved in DMA at all, as it is not a DMA master,
> > and has no knowledge of the memory management or whether there
> > is any DMA at all. I think it is the right abstraction for an MHI driver to
> > pass kernel pointers into the subsystem interfaces, which then get
> > mapped by the bus driver that owns the DMA master.
> > 
> > This is similar to how e.g. USB drivers pass data into the USB core
> > interfaces, which then get the HCI driver to map/unmap it into the
> > DMA masters.
> 
> Ok, then this needs to be named a whole lot better than the original
> "dev" name had it.  Heck, even "parent" does not show that type of
> representation, make it "controller" or something else a whole lot more
> descriptive of what it really is please.
> 

Okay. I'll come up with something more relative.

Thanks,
Mani

> thanks,
> 
> greg k-h

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 11/16] bus: mhi: core: Add support for data transfer
  2020-02-17 16:47     ` Manivannan Sadhasivam
@ 2020-02-18  5:51       ` Manivannan Sadhasivam
  2020-02-18 14:34         ` Arnd Bergmann
  0 siblings, 1 reply; 62+ messages in thread
From: Manivannan Sadhasivam @ 2020-02-18  5:51 UTC (permalink / raw)
  To: Arnd Bergmann
  Cc: gregkh, smohanad, Jeffrey Hugo, Kalle Valo, Bjorn Andersson,
	hemantk, linux-arm-msm, linux-kernel

Hi Arnd,

On Mon, Feb 17, 2020 at 10:17:51PM +0530, Manivannan Sadhasivam wrote:
> Hi Arnd,
> 
> On Mon, Feb 17, 2020 at 05:13:37PM +0100, Arnd Bergmann wrote:
> > On Fri, Jan 31, 2020 at 2:51 PM Manivannan Sadhasivam
> > <manivannan.sadhasivam@linaro.org> wrote:
> > 
> > > @@ -648,6 +715,31 @@ static int parse_ch_cfg(struct mhi_controller *mhi_cntrl,
> > >                 mhi_chan->db_cfg.pollcfg = ch_cfg->pollcfg;
> > >                 mhi_chan->xfer_type = ch_cfg->data_type;
> > >
> > > +               switch (mhi_chan->xfer_type) {
> > > +               case MHI_BUF_RAW:
> > > +                       mhi_chan->gen_tre = mhi_gen_tre;
> > > +                       mhi_chan->queue_xfer = mhi_queue_buf;
> > > +                       break;
> > > +               case MHI_BUF_SKB:
> > > +                       mhi_chan->queue_xfer = mhi_queue_skb;
> > > +                       break;
> > > +               case MHI_BUF_SCLIST:
> > > +                       mhi_chan->gen_tre = mhi_gen_tre;
> > > +                       mhi_chan->queue_xfer = mhi_queue_sclist;
> > > +                       break;
> > > +               case MHI_BUF_NOP:
> > > +                       mhi_chan->queue_xfer = mhi_queue_nop;
> > > +                       break;
> > > +               case MHI_BUF_DMA:
> > > +               case MHI_BUF_RSC_DMA:
> > > +                       mhi_chan->queue_xfer = mhi_queue_dma;
> > > +                       break;
> > > +               default:
> > > +                       dev_err(mhi_cntrl->dev,
> > > +                               "Channel datatype not supported\n");
> > > +                       goto error_chan_cfg;
> > > +               }
> > > +
> > 
> > While looking through the driver to see how the DMA gets handled, I came
> > across the multitude of mhi_queue_* functions, which seems like a
> > layering violation to me, given that they are all implemented by the
> > core code as well, and the client driver needs to be aware of
> > which one to call. Are you able to lift these out of the common interface
> > and make the client driver call these directly, or maybe provide a direct
> > interface based on mhi_buf_info to replace these?
> > 
> 
> It sounds reasonable to me. Let me discuss this internally with Qcom guys to
> see if they have any objections.
> 

I looked into this in detail and found that the queue_xfer callbacks are tied
to the MHI channels. For instance, the callback gets attached to ul_chan (uplink
channel) and dl_chan (downlink channel) during controller registration. And
when the device driver calls the callback, the MHI stack calls respective queue
function for the relevant channel. For instance,

```
int mhi_queue_transfer(struct mhi_device *mhi_dev,
                       enum dma_data_direction dir, void *buf, size_t len,
                       enum mhi_flags mflags)
{
        if (dir == DMA_TO_DEVICE)
                return mhi_dev->ul_chan->queue_xfer(mhi_dev, mhi_dev->ul_chan,
                                                    buf, len, mflags);
        else
                return mhi_dev->dl_chan->queue_xfer(mhi_dev, mhi_dev->dl_chan,
                                                    buf, len, mflags);
}
```

If we use the direct queue API's this will become hard to handle. So, I'll keep
it as it is.

Thanks,
Mani

> Thanks,
> Mani
> 
> >       Arnd

^ permalink raw reply	[flat|nested] 62+ messages in thread

* Re: [PATCH v2 11/16] bus: mhi: core: Add support for data transfer
  2020-02-18  5:51       ` Manivannan Sadhasivam
@ 2020-02-18 14:34         ` Arnd Bergmann
  0 siblings, 0 replies; 62+ messages in thread
From: Arnd Bergmann @ 2020-02-18 14:34 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: gregkh, smohanad, Jeffrey Hugo, Kalle Valo, Bjorn Andersson,
	hemantk, linux-arm-msm, linux-kernel

On Tue, Feb 18, 2020 at 6:51 AM Manivannan Sadhasivam
<manivannan.sadhasivam@linaro.org> wrote:
> On Mon, Feb 17, 2020 at 10:17:51PM +0530, Manivannan Sadhasivam wrote:
> > On Mon, Feb 17, 2020 at 05:13:37PM +0100, Arnd Bergmann wrote:
> > > On Fri, Jan 31, 2020 at 2:51 PM Manivannan Sadhasivam
> > > <manivannan.sadhasivam@linaro.org> wrote:
> > > While looking through the driver to see how the DMA gets handled, I came
> > > across the multitude of mhi_queue_* functions, which seems like a
> > > layering violation to me, given that they are all implemented by the
> > > core code as well, and the client driver needs to be aware of
> > > which one to call. Are you able to lift these out of the common interface
> > > and make the client driver call these directly, or maybe provide a direct
> > > interface based on mhi_buf_info to replace these?
> > >
> >
> > It sounds reasonable to me. Let me discuss this internally with Qcom guys to
> > see if they have any objections.
> >
>
> I looked into this in detail and found that the queue_xfer callbacks are tied
> to the MHI channels. For instance, the callback gets attached to ul_chan (uplink
> channel) and dl_chan (downlink channel) during controller registration. And
> when the device driver calls the callback, the MHI stack calls respective queue
> function for the relevant channel. For instance,
>
> ```
> int mhi_queue_transfer(struct mhi_device *mhi_dev,
>                        enum dma_data_direction dir, void *buf, size_t len,
>                        enum mhi_flags mflags)
> {
>         if (dir == DMA_TO_DEVICE)
>                 return mhi_dev->ul_chan->queue_xfer(mhi_dev, mhi_dev->ul_chan,
>                                                     buf, len, mflags);
>         else
>                 return mhi_dev->dl_chan->queue_xfer(mhi_dev, mhi_dev->dl_chan,
>                                                     buf, len, mflags);
> }
> ```
>
> If we use the direct queue API's this will become hard to handle. So, I'll keep
> it as it is.

Please have another look, this is exactly the part of the subsystem
that I think should be replaced. For the caller, there should not be much
difference between passing ul_chan/dl_chan or
DMA_TO_DEVICE/DMA_FROM_DEVICE, so that too can be lifted
to a higher level.

      Arnd

^ permalink raw reply	[flat|nested] 62+ messages in thread

end of thread, other threads:[~2020-02-18 14:34 UTC | newest]

Thread overview: 62+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-01-31 13:49 [PATCH v2 00/16] Add MHI bus support Manivannan Sadhasivam
2020-01-31 13:49 ` [PATCH v2 01/16] docs: Add documentation for MHI bus Manivannan Sadhasivam
2020-01-31 22:57   ` Jeffrey Hugo
2020-01-31 13:49 ` [PATCH v2 02/16] bus: mhi: core: Add support for registering MHI controllers Manivannan Sadhasivam
2020-02-06 16:56   ` Greg KH
2020-02-11 19:11     ` Manivannan Sadhasivam
2020-02-11 19:22       ` Greg KH
2020-02-13 15:22         ` Manivannan Sadhasivam
2020-02-06 16:57   ` Greg KH
2020-02-11 18:41     ` Manivannan Sadhasivam
2020-02-11 19:20       ` Greg KH
2020-02-13 15:20         ` Manivannan Sadhasivam
2020-02-13 15:34           ` Greg KH
2020-02-13 15:48             ` Manivannan Sadhasivam
2020-02-13 15:53               ` Greg KH
2020-02-14 16:41                 ` Jeffrey Hugo
2020-02-17  5:27                 ` Manivannan Sadhasivam
2020-02-17 11:45                   ` Greg KH
2020-02-17 11:53                     ` Manivannan Sadhasivam
2020-02-17 11:59                   ` Greg KH
2020-02-17 13:04                     ` Manivannan Sadhasivam
2020-02-17 14:15                       ` Greg KH
2020-02-17 16:04                         ` Arnd Bergmann
2020-02-17 16:32                           ` Greg KH
2020-02-17 17:50                             ` Manivannan Sadhasivam
2020-02-06 17:08   ` Jeffrey Hugo
2020-01-31 13:49 ` [PATCH v2 03/16] bus: mhi: core: Add support for registering MHI client drivers Manivannan Sadhasivam
2020-01-31 23:00   ` Jeffrey Hugo
2020-01-31 13:49 ` [PATCH v2 04/16] bus: mhi: core: Add support for creating and destroying MHI devices Manivannan Sadhasivam
2020-02-06 16:16   ` Jeffrey Hugo
2020-01-31 13:49 ` [PATCH v2 05/16] bus: mhi: core: Add support for ringing channel/event ring doorbells Manivannan Sadhasivam
2020-02-06 20:14   ` Jeffrey Hugo
2020-01-31 13:49 ` [PATCH v2 06/16] bus: mhi: core: Add support for PM state transitions Manivannan Sadhasivam
2020-02-06 20:15   ` Jeffrey Hugo
2020-01-31 13:50 ` [PATCH v2 07/16] bus: mhi: core: Add support for basic PM operations Manivannan Sadhasivam
2020-02-06 20:15   ` Jeffrey Hugo
2020-01-31 13:50 ` [PATCH v2 08/16] bus: mhi: core: Add support for downloading firmware over BHIe Manivannan Sadhasivam
2020-02-06 20:15   ` Jeffrey Hugo
2020-01-31 13:50 ` [PATCH v2 09/16] bus: mhi: core: Add support for downloading RDDM image during panic Manivannan Sadhasivam
2020-02-06 16:41   ` Jeffrey Hugo
2020-02-06 20:17     ` Jeffrey Hugo
2020-01-31 13:50 ` [PATCH v2 10/16] bus: mhi: core: Add support for processing events from client device Manivannan Sadhasivam
2020-02-06 20:16   ` Jeffrey Hugo
2020-01-31 13:50 ` [PATCH v2 11/16] bus: mhi: core: Add support for data transfer Manivannan Sadhasivam
2020-02-06 20:16   ` Jeffrey Hugo
2020-02-17 16:13   ` Arnd Bergmann
2020-02-17 16:47     ` Manivannan Sadhasivam
2020-02-18  5:51       ` Manivannan Sadhasivam
2020-02-18 14:34         ` Arnd Bergmann
2020-01-31 13:50 ` [PATCH v2 12/16] bus: mhi: core: Add uevent support for module autoloading Manivannan Sadhasivam
2020-02-06 20:16   ` Jeffrey Hugo
2020-01-31 13:50 ` [PATCH v2 13/16] MAINTAINERS: Add entry for MHI bus Manivannan Sadhasivam
2020-02-03 10:16   ` Andy Shevchenko
2020-02-04  7:05     ` Manivannan Sadhasivam
2020-01-31 13:50 ` [PATCH v2 14/16] net: qrtr: Add MHI transport layer Manivannan Sadhasivam
2020-02-03 18:12   ` Jakub Kicinski
2020-02-04  8:19     ` Manivannan Sadhasivam
2020-02-07  0:14       ` Chris Lew
2020-02-11  3:50         ` Manivannan Sadhasivam
2020-02-12  1:01           ` Chris Lew
2020-01-31 13:50 ` [PATCH v2 15/16] net: qrtr: Do not depend on ARCH_QCOM Manivannan Sadhasivam
2020-01-31 13:50 ` [PATCH v2 16/16] soc: qcom: Do not depend on ARCH_QCOM for QMI helpers Manivannan Sadhasivam

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.