[PATCH 00/15 v3] Batch submission via GuC

[PATCH 00/15 v3] Batch submission via GuC

[Dave Gordon]

This patch series enables command submission via the GuC. In this mode,
instead of the host CPU driving the execlist port directly, it hands
over work items to the GuC, using a doorbell mechanism to tell the GuC
that new items have been added to its work queue. The GuC then dispatches
contexts to the various GPU engines, and manages the resulting context-
switch interrupts. Completion of a batch is however still signalled to
the CPU; the GuC is not involved in handling user interrupts.

There are three subsequences within the patch series:

  drm/i915: Add i915_gem_object_create_from_data()
  drm/i915: Embedded microcontroller (uC) firmware loading support

These first two patches provide a common framework for fetching the
firmware that may be required by any embedded microcontroller from a
file, and saving some or all of it in a pageable (shmfs-backed) GEM
object (or otherwise, if required by the uC-specific code).  It is
hoped that this framework is sufficiently general that it can be used
for all curent and future microcontrollers.

  drm/i915: Add GuC-related module parameters
  drm/i915: Add GuC-related header files
  drm/i915: GuC-specific firmware loader
  drm/i915: Debugfs interface to read GuC load status

These four patches complete the GuC loader. At this point in the sequence
we can load and activate the GuC firmware, but not submit any batches
through it. (This is nonetheless a potentially useful state, as the GuC
could do other useful work even when not handling batch submissions).

  drm/i915: Expose two LRC functions for GuC submission mode
  drm/i915: GuC submission setup, phase 1
  drm/i915: Enable GuC firmware log
  drm/i915: Implementation of GuC client
  drm/i915: Interrupt routing for GuC submission
  drm/i915: Integrate GuC-based command submission
  drm/i915: Debugfs interface for GuC submission statistics
  Documentation/drm: kerneldoc for GuC
  drm/i915: Enable GuC submission, where supported

In the final section, we implement the GuC submission mechanism, link
it into the (execlist-based) submission path, and finally enable it
(on supported platforms). On platforms where there is no GuC, or if
the GuC firmware cannot be found or is invalid, batch submission will
revert to using the execlist mechanism directly.

The GuC firmware itself is not included in this patchset; it is or will
be available for download from https://01.org/linuxgraphics/downloads/
This driver works with and requires GuC firmware revision 3.x. It will
not work with any firmware version 1.x, as the GuC protocol in those
revisions was incompatible and is no longer supported.

Prerequisites: GuC submission will expose existing inadequacies in
some of the existing codepaths unless certain other patches are applied.
In particular we require Michel Thierry's patch:
  drm/i915/lrc: Update PDPx registers with LRI commands
(because the GuC support light-restore, which execlist mode doesn't).
Unfortunately the version of this recently committed to drm-intel-nightly
    7a01a0a drm/i915/lrc: Update PDPx registers with lri commands
was incomplete in a way that was mostly harmless, but incompatible with
GuC usage, so the missing part has been folded into patch 7 below:
  drm/i915: Expose two LRC functions for GuC submission mode
The previous prerequisite patch
  drm/i915: Allocate OLR more safely (workaround until OLR goes away)
is no longer required as the OLR has now gone away :)

Alex Dai (7):
  drm/i915: Add GuC-related module parameters
  drm/i915: GuC-specific firmware loader
  drm/i915: Debugfs interface to read GuC load status
  drm/i915: GuC submission setup, phase 1
  drm/i915: Enable GuC firmware log
  drm/i915: Integrate GuC-based command submission
  Documentation/drm: kerneldoc for GuC

Dave Gordon (8):
  drm/i915: Add i915_gem_object_create_from_data()
  drm/i915: Embedded microcontroller (uC) firmware loading support
  drm/i915: Add GuC-related header files
  drm/i915: Expose two LRC functions for GuC submission mode
  drm/i915: Implementation of GuC client
  drm/i915: Interrupt routing for GuC submission
  drm/i915: Debugfs interface for GuC submission statistics
  drm/i915: Enable GuC submission, where supported

Ben Widawsky (0):
Vinit Azad (0):
Michael H. Nguyen (0):
  created the original versions on which some of these patches are based.

 Documentation/DocBook/drm.tmpl             |  19 +
 drivers/gpu/drm/i915/Makefile              |   7 +
 drivers/gpu/drm/i915/i915_debugfs.c        | 116 +++-
 drivers/gpu/drm/i915/i915_dma.c            |   4 +
 drivers/gpu/drm/i915/i915_drv.h            |  15 +
 drivers/gpu/drm/i915/i915_gem.c            |  48 ++
 drivers/gpu/drm/i915/i915_guc_reg.h        | 102 ++++
 drivers/gpu/drm/i915/i915_guc_submission.c | 861 +++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_params.c         |   9 +
 drivers/gpu/drm/i915/i915_reg.h            |  15 +-
 drivers/gpu/drm/i915/intel_guc.h           | 101 ++++
 drivers/gpu/drm/i915/intel_guc_fwif.h      | 236 ++++++++
 drivers/gpu/drm/i915/intel_guc_loader.c    | 530 ++++++++++++++++++
 drivers/gpu/drm/i915/intel_lrc.c           |  85 +--
 drivers/gpu/drm/i915/intel_lrc.h           |  11 +
 drivers/gpu/drm/i915/intel_uc_loader.c     | 310 +++++++++++
 drivers/gpu/drm/i915/intel_uc_loader.h     |  92 +++
 17 files changed, 2520 insertions(+), 41 deletions(-)
 create mode 100644 drivers/gpu/drm/i915/i915_guc_reg.h
 create mode 100644 drivers/gpu/drm/i915/i915_guc_submission.c
 create mode 100644 drivers/gpu/drm/i915/intel_guc.h
 create mode 100644 drivers/gpu/drm/i915/intel_guc_fwif.h
 create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c
 create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c
 create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h

-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 01/15] drm/i915: Add i915_gem_object_create_from_data()

[Dave Gordon]

i915_gem_object_create_from_data() is a generic function to save data
from a plain linear buffer in a new pageable gem object that can later
be accessed by the CPU and/or GPU.

We will need this for the microcontroller firmware loading support code.

Derived from i915_gem_object_write(), originally by Alex Dai

v2:
    Change of function: now allocates & fills a new object, rather than
        writing to an existing object
    New name courtesy of Chris Wilson
    Explicit domain-setting and other improvements per review comments
        by Chris Wilson & Daniel Vetter

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h |  2 ++
 drivers/gpu/drm/i915/i915_gem.c | 40 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 42 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 1dbd957..b0bb07d 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2748,6 +2748,8 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj,
 			 const struct drm_i915_gem_object_ops *ops);
 struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
 						  size_t size);
+struct drm_i915_gem_object *i915_gem_object_create_from_data(
+		struct drm_device *dev, const void *data, size_t size);
 void i915_init_vm(struct drm_i915_private *dev_priv,
 		  struct i915_address_space *vm);
 void i915_gem_free_object(struct drm_gem_object *obj);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 49016e0..aa8f4c3 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -5472,3 +5472,43 @@ bool i915_gem_obj_is_pinned(struct drm_i915_gem_object *obj)
 
 	return false;
 }
+
+/* Allocate a new GEM object and fill it with the supplied data */
+struct drm_i915_gem_object *
+i915_gem_object_create_from_data(struct drm_device *dev,
+			         const void *data, size_t size)
+{
+	struct drm_i915_gem_object *obj;
+	struct sg_table *sg;
+	size_t bytes;
+	int ret;
+
+	obj = i915_gem_alloc_object(dev, round_up(size, PAGE_SIZE));
+	if (IS_ERR_OR_NULL(obj))
+		return obj;
+
+	ret = i915_gem_object_set_to_cpu_domain(obj, true);
+	if (ret)
+		goto fail;
+
+	ret = i915_gem_object_get_pages(obj);
+	if (ret)
+		goto fail;
+
+	i915_gem_object_pin_pages(obj);
+	sg = obj->pages;
+	bytes = sg_copy_from_buffer(sg->sgl, sg->nents, (void *)data, size);
+	i915_gem_object_unpin_pages(obj);
+
+	if (WARN_ON(bytes != size)) {
+		DRM_ERROR("Incomplete copy, wrote %zu of %zu", bytes, size);
+		ret = -EFAULT;
+		goto fail;
+	}
+
+	return obj;
+
+fail:
+	drm_gem_object_unreference(&obj->base);
+	return ERR_PTR(ret);
+}
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 02/15] drm/i915: Embedded microcontroller (uC) firmware loading support

[Dave Gordon]

Current devices may contain one or more programmable microcontrollers
that need to have a firmware image (aka "binary blob") loaded from an
external medium and transferred to the device's memory.

This file provides common support functions for doing this; they can
then be used by each uC-specific loader, thus reducing code duplication
and testing effort.

v2:
    Local functions should pass dev_priv rather than dev [Chris Wilson]
    Various other improvements per Chris Wilson's review comments

v3:
    Defeatured version without async prefetch [Daniel Vetter]

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Alex Dai <yu.dai@intel.com>
---
 drivers/gpu/drm/i915/Makefile          |   3 +
 drivers/gpu/drm/i915/intel_uc_loader.c | 310 +++++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_uc_loader.h |  92 ++++++++++
 3 files changed, 405 insertions(+)
 create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c
 create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h

diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index de21965..f1f80fc 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -39,6 +39,9 @@ i915-y += i915_cmd_parser.o \
 	  intel_ringbuffer.o \
 	  intel_uncore.o
 
+# generic ancilliary microcontroller support
+i915-y += intel_uc_loader.o
+
 # autogenerated null render state
 i915-y += intel_renderstate_gen6.o \
 	  intel_renderstate_gen7.o \
diff --git a/drivers/gpu/drm/i915/intel_uc_loader.c b/drivers/gpu/drm/i915/intel_uc_loader.c
new file mode 100644
index 0000000..a8fc1dd
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_uc_loader.c
@@ -0,0 +1,310 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Author:
+ *	Dave Gordon <david.s.gordon@intel.com>
+ */
+#include <linux/firmware.h>
+#include "i915_drv.h"
+#include "intel_uc_loader.h"
+
+/**
+ * DOC: Common functions for embedded microcontroller (uC) firmware loading
+ *
+ * The functions in this file provide common support code for loading the
+ * firmware that may be required by an embedded microcontroller (uC).
+ *
+ * The function intel_uc_fw_init() should be called first; it requires no
+ * locking, and can be called even before GEM has been initialised. It just
+ * initialises the tracking data and stores its parameters for the subsequent
+ * call to intel_uc_fw_fetch().
+ *
+ * At some convenient point after GEM initialisation, the driver should call
+ * intel_uc_fw_fetch(). The first time, this will use the Linux kernel's
+ * request_firmware() call to open and read the firmware image into memory.
+ * (On subsequent calls, this is skipped, as either the firmware has already
+ * been fetched into memory, or it is already known that no valid firmware
+ * image could be found).
+ *
+ * The callback() function passed to intel_uc_fw_fetch() can further check
+ * the firmware image before it is saved. This function can reject the image
+ * by returning a negative error code (e.g. -ENOEXEC), or accept it. In the
+ * latter case, it can return INTEL_UC_FW_GOOD (which is also the default if
+ * no callback() is supplied), and the common code here will save the whole
+ * of the firmware image in a (pageable, shmfs-backed) GEM object.
+ *
+ * If saving the whole image unmodified is not appropriate (for example, if
+ * only a small part of the image is needed later, or the data needs to be
+ * reorganised before saving), the callback() function can instead make its
+ * own arrangements for saving the required data in a GEM object or otherwise
+ * and then return INTEL_UC_FW_SAVED.
+ *
+ * (If such a callback does stash (some of) the image data in a GEM object,
+ * it can use the uc_fw_obj and uc_fw_size fields; the common framework will
+ * then handle deallocating the object on failure or finalisation. Otherwise
+ * any allocated resources will have to be managed by the uC-specific code).
+ *
+ * After a successful call to intel_uc_fw_fetch(), the uC-specific code can
+ * transfer the data in the GEM object (or its own alternative) to the uC's
+ * memory (in some uC-specific way, not handled here).
+ *
+ * During driver shutdown, or if driver load is aborted, intel_uc_fw_fini()
+ * should be called to release any remaining resources.
+ */
+
+/* User-friendly representation of an enum */
+const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status)
+{
+	switch (status) {
+	case INTEL_UC_FIRMWARE_FAIL:
+		return "FAIL";
+	case INTEL_UC_FIRMWARE_NONE:
+		return "NONE";
+	case INTEL_UC_FIRMWARE_PENDING:
+		return "PENDING";
+	case INTEL_UC_FIRMWARE_SUCCESS:
+		return "SUCCESS";
+	default:
+		return "UNKNOWN!";
+	}
+};
+
+/*
+ * Called once per uC, late in driver initialisation, after GEM is set up.
+ * First, we fetch the firmware image using request_firmware(), then allow
+ * the optional callback() function to check it. If it's good, and callback()
+ * doesn't say it's already saved the image, we will save it here by copying
+ * the whole thing into a (pageable, shmfs-backed) GEM object,
+ */
+static void
+uc_fw_fetch(struct intel_uc_fw *uc_fw,
+	    int callback(struct intel_uc_fw *))
+{
+	struct drm_device *dev = uc_fw->uc_dev;
+	struct drm_i915_gem_object *obj;
+	const struct firmware *fw;
+	int cb_status = INTEL_UC_FW_GOOD;
+
+	DRM_DEBUG_DRIVER("before waiting: %s fw fetch status %s, fw %p\n",
+		uc_fw->uc_name,
+		intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status),
+		uc_fw->uc_fw_blob);
+
+	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+	WARN_ON(uc_fw->uc_fw_fetch_status != INTEL_UC_FIRMWARE_PENDING);
+
+	if (request_firmware(&fw, uc_fw->uc_fw_path, &dev->pdev->dev))
+		goto fail;
+	if (!fw)
+		goto fail;
+
+	DRM_DEBUG_DRIVER("fetch %s fw from %s succeeded, fw %p\n",
+		uc_fw->uc_name, uc_fw->uc_fw_path, fw);
+	uc_fw->uc_fw_blob = fw;
+
+	/* Callback to the optional uC-specific function, if supplied */
+	if (callback)
+		cb_status = callback(uc_fw);
+	if (cb_status < 0)
+		goto fail;
+	switch (cb_status) {
+	default:
+		WARN(1, "Invalid status %d from fw checking callback %pf\n",
+				cb_status, callback);
+		goto fail;
+
+	case INTEL_UC_FW_SAVED:
+		// Good, already saved, nothing to do
+		break;
+
+	case INTEL_UC_FW_GOOD:
+		// Good, save in GEM object
+		obj = i915_gem_object_create_from_data(dev, fw->data, fw->size);
+		if (!obj)
+			goto fail;
+
+		uc_fw->uc_fw_obj = obj;
+		uc_fw->uc_fw_size = fw->size;
+	}
+
+	DRM_DEBUG_DRIVER("%s fw fetch status SUCCESS, cb %d, obj %p\n",
+			uc_fw->uc_name, cb_status, uc_fw->uc_fw_obj);
+
+	release_firmware(fw);
+	uc_fw->uc_fw_blob = NULL;
+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_SUCCESS;
+	return;
+
+fail:
+	DRM_DEBUG_DRIVER("%s fw fetch status FAIL; fw %p, obj %p\n",
+		uc_fw->uc_name, fw, uc_fw->uc_fw_obj);
+	DRM_ERROR("Failed to fetch %s firmware from %s\n",
+		  uc_fw->uc_name, uc_fw->uc_fw_path);
+
+	obj = uc_fw->uc_fw_obj;
+	if (obj)
+		drm_gem_object_unreference(&obj->base);
+	uc_fw->uc_fw_obj = NULL;
+
+	release_firmware(fw);		/* OK even if fw is NULL */
+	uc_fw->uc_fw_blob = NULL;
+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_FAIL;
+}
+
+/**
+ * intel_uc_fw_fetch() - fetch the firmware image
+ * @uc_fw:	intel_uc_fw structure
+ * @callback:	optional callback function to validate and/or save the image
+ *
+ * If the fetch is not PENDING (i.e. on the second and subsequent calls), this
+ * function will just return an approriate errno, based on the previously-set
+ * status.
+ *
+ * On the first call only, it will try to retrieve the firmaware image using
+ * the parameters set earlier. If an image is found, the optional @callback()
+ * will be called to further validate it.
+ *
+ * When it is called, @uc_fw->uc_fw_blob refers to the fetched firmware image,
+ * and @uc_fw->uc_fw_obj is NULL. The @callback() function can return an error
+ * code (any negative value), in which case the image will be rejected.  The
+ * fetch status will be set to FAIL, and this function will return -EIO.
+ *
+ * Or, @callback() can return INTEL_UC_FW_GOOD. The image is considered good,
+ * and it will be saved in a GEM object as described above. In this case,
+ * @uc_fw->uc_fw_obj will be set to point to the GEM object, and the size of
+ * the image will be in @uc_fw->uc_fw_size. This is also the default if no
+ * @callback() is supplied.
+ *
+ * Finally, @callback() can return INTEL_UC_FW_SAVED. The image is considered
+ * good, but @callback() has already deal with saving the content, so this
+ * common code will not allocate and fill a GEM object. @callback() may use
+ * @uc_fw->uc_fw_obj to hold a reference to its own GEM object, if it has
+ * allocated one, and in this case the common code will deal with disposing
+ * of it on error or finalisation; otherwise it can make its own arragements.
+ *
+ * After this call, @uc_fw->uc_fw_fetch_status will show whether the firmware
+ * image was successfully fetched and saved.
+ *
+ * In all cases the firmware blob is released before this function returns.
+ *
+ * Return:	non-zero code on error
+ */
+int
+intel_uc_fw_fetch(struct intel_uc_fw *uc_fw,
+		  int callback(struct intel_uc_fw *))
+{
+	WARN_ON(!mutex_is_locked(&uc_fw->uc_dev->struct_mutex));
+
+	if (uc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_PENDING) {
+		/* We only come here once */
+		uc_fw_fetch(uc_fw, callback);
+		/* status must now be FAIL or SUCCESS */
+	}
+
+	DRM_DEBUG_DRIVER("%s fw fetch status %s\n", uc_fw->uc_name,
+		intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status));
+
+	switch (uc_fw->uc_fw_fetch_status) {
+	case INTEL_UC_FIRMWARE_FAIL:
+		/* something went wrong :( */
+		return -EIO;
+
+	case INTEL_UC_FIRMWARE_NONE:
+		/* no firmware, nothing to do (not an error) */
+		return 0;
+
+	case INTEL_UC_FIRMWARE_PENDING:
+	default:
+		/* "can't happen" */
+		WARN_ONCE(1, "%s fw %s invalid uc_fw_fetch_status %s [%d]\n",
+			uc_fw->uc_name, uc_fw->uc_fw_path,
+			intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status),
+			uc_fw->uc_fw_fetch_status);
+		return -ENXIO;
+
+	case INTEL_UC_FIRMWARE_SUCCESS:
+		return 0;
+	}
+}
+
+/**
+ * intel_uc_fw_init() - define parameters for fetching firmware
+ * @dev:	drm device
+ * @uc_fw:	intel_uc_fw structure
+ * @name:	human-readable device name (e.g. "GuC") for messages
+ * @fw_path:	(trailing parts of) path to firmware (e.g. "i915/guc_fw.bin")
+ * 		@fw_path == NULL means "no firmware expected" (not an error),
+ * 		@fw_path == "" (empty string) means "firmware unknown" i.e.
+ * 		the uC requires firmware, but the driver doesn't know where
+ * 		to find the proper version. This will be logged as an error.
+ *
+ * This is called just once per uC, during driver loading. It is therefore
+ * automatically single-threaded and does not need to acquire any mutexes
+ * or spinlocks. OTOH, GEM is not yet fully initialised, so we can't do
+ * very much here.
+ *
+ * The main task here is to save the parameters for later. The actual fetching
+ * will happen when intel_uc_fw_init() is called, after GEM initialisation is
+ * complete.
+ */
+void
+intel_uc_fw_init(struct drm_device *dev, struct intel_uc_fw *uc_fw,
+		 const char *name, const char *fw_path)
+{
+	uc_fw->uc_dev = dev;
+	uc_fw->uc_name = name;
+	uc_fw->uc_fw_path = fw_path;
+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_NONE;
+	uc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_NONE;
+
+	if (fw_path == NULL)
+		return;
+
+	if (*fw_path == '\0') {
+		DRM_ERROR("No %s firmware known for this platform\n", name);
+		uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_FAIL;
+		return;
+	}
+
+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_PENDING;
+	DRM_DEBUG_DRIVER("%s firmware pending, path %s\n",
+		name, fw_path);
+}
+
+/**
+ * intel_uc_fw_fini() - clean up all uC firmware-related data
+ * @uc_fw:	intel_uc_fw structure
+ */
+void
+intel_uc_fw_fini(struct intel_uc_fw *uc_fw)
+{
+	WARN_ON(!mutex_is_locked(&uc_fw->uc_dev->struct_mutex));
+
+	if (uc_fw->uc_fw_obj)
+		drm_gem_object_unreference(&uc_fw->uc_fw_obj->base);
+	uc_fw->uc_fw_obj = NULL;
+
+	release_firmware(uc_fw->uc_fw_blob);	/* OK even if NULL */
+	uc_fw->uc_fw_blob = NULL;
+
+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_NONE;
+}
diff --git a/drivers/gpu/drm/i915/intel_uc_loader.h b/drivers/gpu/drm/i915/intel_uc_loader.h
new file mode 100644
index 0000000..b710406
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_uc_loader.h
@@ -0,0 +1,92 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Author:
+ *	Dave Gordon <david.s.gordon@intel.com>
+ */
+#ifndef _INTEL_UC_LOADER_H
+#define _INTEL_UC_LOADER_H
+
+/*
+ * Microcontroller (uC) firmware loading support
+ */
+
+/*
+ * These values are used to track the stages of getting the required firmware
+ * into an onboard microcontroller. The common code tracks the phases of
+ * fetching the firmware (aka "binary blob") from an external file into a GEM
+ * object in the 'uc_fw_fetch_status' field below; the uC-specific DMA code
+ * uses the 'uc_fw_load_status' field to track the transfer from GEM object
+ * to uC memory.
+ *
+ * For the first (fetch) stage, the interpretation of the values is:
+ * NONE - no firmware is being fetched e.g. because there is no uC
+ * PENDING - parameters initialised, pending call to intel_uc_fw_fetch()
+ * SUCCESS - uC firmware fetched into a GEM object and ready for use
+ * FAIL - something went wrong; uC firmware is not available
+ *
+ * The second (load) stage is similar:
+ * NONE - no firmware is being loaded e.g. because there is no uC
+ * PENDING - firmware DMA load in progress
+ * SUCCESS - uC firmware loaded into uC memory and ready for use
+ * FAIL - something went wrong; uC firmware is not available
+ *
+ * The function intel_uc_fw_status_repr() will convert this enum to a
+ * string representation suitable for use in log messages.
+ */
+enum intel_uc_fw_status {
+	INTEL_UC_FIRMWARE_FAIL = -1,
+	INTEL_UC_FIRMWARE_NONE = 0,
+	INTEL_UC_FIRMWARE_PENDING,
+	INTEL_UC_FIRMWARE_SUCCESS
+};
+const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status);
+
+/*
+ * This structure encapsulates all the data needed during the process of
+ * fetching, caching, and loading the firmware image into the uC.
+ */
+struct intel_uc_fw {
+	struct drm_device *		uc_dev;
+	const char *			uc_name;
+	const char *			uc_fw_path;
+	const struct firmware *		uc_fw_blob;
+	size_t				uc_fw_size;
+	struct drm_i915_gem_object *	uc_fw_obj;
+	enum intel_uc_fw_status		uc_fw_fetch_status;
+	enum intel_uc_fw_status		uc_fw_load_status;
+};
+
+void intel_uc_fw_init(struct drm_device *dev, struct intel_uc_fw *uc_fw,
+		const char *uc_name, const char *fw_path);
+int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw,
+		int callback(struct intel_uc_fw *));
+void intel_uc_fw_fini(struct intel_uc_fw *uc_fw);
+
+/*
+ * The callback() function passed to intel_uc_fw_fetch() above can return
+ * a negative number (a standard error code), or one of the values below:
+ */
+#define	INTEL_UC_FW_GOOD	1	/* f/w good, save in GEM object	*/
+#define	INTEL_UC_FW_SAVED	2	/* f/w good, already saved	*/
+
+#endif
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 03/15] drm/i915: Add GuC-related module parameters

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

Two new module parameters: "enable_guc_submission" which will turn
on submission of batchbuffers via the GuC (when implemented), and
"guc_log_level" which controls the level of debugging logged by the
GuC and captured by the host.

Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h    | 2 ++
 drivers/gpu/drm/i915/i915_params.c | 9 +++++++++
 2 files changed, 11 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index b0bb07d..9618f57 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2599,6 +2599,8 @@ struct i915_params {
 	bool reset;
 	bool disable_display;
 	bool disable_vtd_wa;
+	bool enable_guc_submission;
+	int guc_log_level;
 	int use_mmio_flip;
 	int mmio_debug;
 	bool verbose_state_checks;
diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c
index 7983fe4..a8f48dd 100644
--- a/drivers/gpu/drm/i915/i915_params.c
+++ b/drivers/gpu/drm/i915/i915_params.c
@@ -53,6 +53,8 @@ struct i915_params i915 __read_mostly = {
 	.verbose_state_checks = 1,
 	.nuclear_pageflip = 0,
 	.edp_vswing = 0,
+	.enable_guc_submission = false,
+	.guc_log_level = -1,
 };
 
 module_param_named(modeset, i915.modeset, int, 0400);
@@ -186,3 +188,10 @@ MODULE_PARM_DESC(edp_vswing,
 		 "Ignore/Override vswing pre-emph table selection from VBT "
 		 "(0=use value from vbt [default], 1=low power swing(200mV),"
 		 "2=default swing(400mV))");
+
+module_param_named(enable_guc_submission, i915.enable_guc_submission, bool, 0400);
+MODULE_PARM_DESC(enable_guc_submission, "Enable GuC submission (default:false)");
+
+module_param_named(guc_log_level, i915.guc_log_level, int, 0400);
+MODULE_PARM_DESC(guc_log_level,
+	"GuC firmware logging level (-1:disabled (default), 0-3:enabled)");
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 04/15] drm/i915: Add GuC-related header files

[Dave Gordon]

intel_guc_fwif.h contains the subset of the GuC interface that we
will need for submission of commands through the GuC. These MUST
be kept in sync with the definitions used by the GuC firmware, and
updates to this file will (or should) be autogenerated from the
source files used to build the firmware. Editing this file is
therefore not recommended.

i915_guc_reg.h contains definitions of GuC-related hardware:
registers, bitmasks, etc. These should match the BSpec.

v2:
    Files renamed & resliced per review comments by Chris Wilson

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_guc_reg.h   | 102 +++++++++++++++
 drivers/gpu/drm/i915/intel_guc_fwif.h | 236 ++++++++++++++++++++++++++++++++++
 2 files changed, 338 insertions(+)
 create mode 100644 drivers/gpu/drm/i915/i915_guc_reg.h
 create mode 100644 drivers/gpu/drm/i915/intel_guc_fwif.h

diff --git a/drivers/gpu/drm/i915/i915_guc_reg.h b/drivers/gpu/drm/i915/i915_guc_reg.h
new file mode 100644
index 0000000..ccdc6c8
--- /dev/null
+++ b/drivers/gpu/drm/i915/i915_guc_reg.h
@@ -0,0 +1,102 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+#ifndef _I915_GUC_REG_H_
+#define _I915_GUC_REG_H_
+
+/* Definitions of GuC H/W registers, bits, etc */
+
+#define GUC_STATUS			0xc000
+#define   GS_BOOTROM_SHIFT		1
+#define   GS_BOOTROM_MASK		  (0x7F << GS_BOOTROM_SHIFT)
+#define   GS_BOOTROM_RSA_FAILED		  (0x50 << GS_BOOTROM_SHIFT)
+#define   GS_UKERNEL_SHIFT		8
+#define   GS_UKERNEL_MASK		  (0xFF << GS_UKERNEL_SHIFT)
+#define   GS_UKERNEL_LAPIC_DONE		  (0x30 << GS_UKERNEL_SHIFT)
+#define   GS_UKERNEL_DPC_ERROR		  (0x60 << GS_UKERNEL_SHIFT)
+#define   GS_UKERNEL_READY		  (0xF0 << GS_UKERNEL_SHIFT)
+#define   GS_MIA_SHIFT			16
+#define   GS_MIA_MASK			  (0x07 << GS_MIA_SHIFT)
+
+#define GUC_WOPCM_SIZE			0xc050
+#define   GUC_WOPCM_SIZE_VALUE  	  (0x80 << 12)	/* 512KB */
+#define GUC_WOPCM_OFFSET		0x80000		/* 512KB */
+
+#define SOFT_SCRATCH(n)			(0xc180 + ((n) * 4))
+
+#define UOS_RSA_SCRATCH_0		0xc200
+#define DMA_ADDR_0_LOW			0xc300
+#define DMA_ADDR_0_HIGH			0xc304
+#define DMA_ADDR_1_LOW			0xc308
+#define DMA_ADDR_1_HIGH			0xc30c
+#define   DMA_ADDRESS_SPACE_WOPCM	  (7 << 16)
+#define   DMA_ADDRESS_SPACE_GTT		  (8 << 16)
+#define DMA_COPY_SIZE			0xc310
+#define DMA_CTRL			0xc314
+#define   UOS_MOVE			  (1<<4)
+#define   START_DMA			  (1<<0)
+#define DMA_GUC_WOPCM_OFFSET		0xc340
+
+#define GEN8_GT_PM_CONFIG		0x138140
+#define GEN9_GT_PM_CONFIG		0x13816c
+#define   GEN8_GT_DOORBELL_ENABLE	  (1<<0)
+
+#define GEN8_GTCR			0x4274
+#define   GEN8_GTCR_INVALIDATE		  (1<<0)
+
+#define GUC_ARAT_C6DIS			0xA178
+
+#define GUC_SHIM_CONTROL		0xc064
+#define   GUC_DISABLE_SRAM_INIT_TO_ZEROES	(1<<0)
+#define   GUC_ENABLE_READ_CACHE_LOGIC		(1<<1)
+#define   GUC_ENABLE_MIA_CACHING		(1<<2)
+#define   GUC_GEN10_MSGCH_ENABLE		(1<<4)
+#define   GUC_ENABLE_READ_CACHE_FOR_SRAM_DATA	(1<<9)
+#define   GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA	(1<<10)
+#define   GUC_ENABLE_MIA_CLOCK_GATING		(1<<15)
+#define   GUC_GEN10_SHIM_WC_ENABLE		(1<<21)
+
+#define GUC_SHIM_CONTROL_VALUE	(GUC_DISABLE_SRAM_INIT_TO_ZEROES	| \
+				 GUC_ENABLE_READ_CACHE_LOGIC		| \
+				 GUC_ENABLE_MIA_CACHING			| \
+				 GUC_ENABLE_READ_CACHE_FOR_SRAM_DATA	| \
+				 GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA)
+
+#define HOST2GUC_INTERRUPT		0xc4c8
+#define   HOST2GUC_TRIGGER		  (1<<0)
+
+#define DRBMISC1			0x1984
+#define   DOORBELL_ENABLE		  (1<<0)
+
+#define GEN8_DRBREGL(x)			(0x1000 + (x) * 8)
+#define   GEN8_DRB_VALID		  (1<<0)
+#define GEN8_DRBREGU(x)			(GEN8_DRBREGL(x) + 4)
+
+#define DE_GUCRMR			0x44054
+
+#define GUC_BCS_RCS_IER			0xC550
+#define GUC_VCS2_VCS1_IER		0xC554
+#define GUC_WD_VECS_IER			0xC558
+#define GUC_PM_P24C_IER			0xC55C
+
+#endif
diff --git a/drivers/gpu/drm/i915/intel_guc_fwif.h b/drivers/gpu/drm/i915/intel_guc_fwif.h
new file mode 100644
index 0000000..f6bb825
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_guc_fwif.h
@@ -0,0 +1,236 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+#ifndef _INTEL_GUC_FWIF_H
+#define _INTEL_GUC_FWIF_H
+
+#define GFXCORE_FAMILY_GEN8		11
+#define GFXCORE_FAMILY_GEN9		12
+#define GFXCORE_FAMILY_FORCE_ULONG	0x7fffffff
+
+#define GUC_CTX_PRIORITY_CRITICAL	0
+#define GUC_CTX_PRIORITY_HIGH		1
+#define GUC_CTX_PRIORITY_NORMAL		2
+#define GUC_CTX_PRIORITY_LOW		3
+
+#define GUC_MAX_GPU_CONTEXTS		1024
+#define	GUC_INVALID_CTX_ID		(GUC_MAX_GPU_CONTEXTS + 1)
+
+/* Work queue item header definitions */
+#define WQ_STATUS_ACTIVE		1
+#define WQ_STATUS_SUSPENDED		2
+#define WQ_STATUS_CMD_ERROR		3
+#define WQ_STATUS_ENGINE_ID_NOT_USED	4
+#define WQ_STATUS_SUSPENDED_FROM_RESET	5
+#define WQ_TYPE_SHIFT			0
+#define   WQ_TYPE_BATCH_BUF		(0x1 << WQ_TYPE_SHIFT)
+#define   WQ_TYPE_PSEUDO		(0x2 << WQ_TYPE_SHIFT)
+#define   WQ_TYPE_INORDER		(0x3 << WQ_TYPE_SHIFT)
+#define WQ_TARGET_SHIFT			10
+#define WQ_LEN_SHIFT			16
+#define WQ_NO_WCFLUSH_WAIT		(1 << 27)
+#define WQ_PRESENT_WORKLOAD		(1 << 28)
+#define WQ_WORKLOAD_SHIFT		29
+#define   WQ_WORKLOAD_GENERAL		(0 << WQ_WORKLOAD_SHIFT)
+#define   WQ_WORKLOAD_GPGPU		(1 << WQ_WORKLOAD_SHIFT)
+#define   WQ_WORKLOAD_TOUCH		(2 << WQ_WORKLOAD_SHIFT)
+
+#define WQ_RING_TAIL_SHIFT		20
+#define WQ_RING_TAIL_MASK		(0x7FF << WQ_RING_TAIL_SHIFT)
+
+#define GUC_DOORBELL_ENABLED		1
+#define GUC_DOORBELL_DISABLED		0
+
+#define GUC_CTX_DESC_ATTR_ACTIVE	(1 << 0)
+#define GUC_CTX_DESC_ATTR_PENDING_DB	(1 << 1)
+#define GUC_CTX_DESC_ATTR_KERNEL	(1 << 2)
+#define GUC_CTX_DESC_ATTR_PREEMPT	(1 << 3)
+#define GUC_CTX_DESC_ATTR_RESET		(1 << 4)
+#define GUC_CTX_DESC_ATTR_WQLOCKED	(1 << 5)
+#define GUC_CTX_DESC_ATTR_PCH		(1 << 6)
+
+/* The guc control data is 10 DWORDs */
+#define GUC_CTL_CTXINFO			0
+#define   GUC_CTL_CTXNUM_IN16_SHIFT	0
+#define   GUC_CTL_BASE_ADDR_SHIFT	12
+#define GUC_CTL_ARAT_HIGH		1
+#define GUC_CTL_ARAT_LOW		2
+#define GUC_CTL_DEVICE_INFO		3
+#define   GUC_CTL_GTTYPE_SHIFT		0
+#define   GUC_CTL_COREFAMILY_SHIFT	7
+#define GUC_CTL_LOG_PARAMS		4
+#define   GUC_LOG_VALID			(1 << 0)
+#define   GUC_LOG_NOTIFY_ON_HALF_FULL	(1 << 1)
+#define   GUC_LOG_ALLOC_IN_MEGABYTE	(1 << 3)
+#define   GUC_LOG_CRASH_PAGES		1
+#define   GUC_LOG_CRASH_SHIFT		4
+#define   GUC_LOG_DPC_PAGES		3
+#define   GUC_LOG_DPC_SHIFT		6
+#define   GUC_LOG_ISR_PAGES		3
+#define   GUC_LOG_ISR_SHIFT		9
+#define   GUC_LOG_BUF_ADDR_SHIFT	12
+#define GUC_CTL_PAGE_FAULT_CONTROL	5
+#define GUC_CTL_WA			6
+#define   GUC_CTL_WA_UK_BY_DRIVER	(1 << 3)
+#define GUC_CTL_FEATURE			7
+#define   GUC_CTL_VCS2_ENABLED		(1 << 0)
+#define   GUC_CTL_KERNEL_SUBMISSIONS	(1 << 1)
+#define   GUC_CTL_FEATURE2		(1 << 2)
+#define   GUC_CTL_POWER_GATING		(1 << 3)
+#define   GUC_CTL_DISABLE_SCHEDULER	(1 << 4)
+#define   GUC_CTL_PREEMPTION_LOG	(1 << 5)
+#define   GUC_CTL_ENABLE_SLPC		(1 << 7)
+#define GUC_CTL_DEBUG			8
+#define   GUC_LOG_VERBOSITY_SHIFT	0
+#define   GUC_LOG_VERBOSITY_LOW		(0 << GUC_LOG_VERBOSITY_SHIFT)
+#define   GUC_LOG_VERBOSITY_MED		(1 << GUC_LOG_VERBOSITY_SHIFT)
+#define   GUC_LOG_VERBOSITY_HIGH	(2 << GUC_LOG_VERBOSITY_SHIFT)
+#define   GUC_LOG_VERBOSITY_ULTRA	(3 << GUC_LOG_VERBOSITY_SHIFT)
+/* Verbosity range-check limits, without the shift */
+#define	  GUC_LOG_VERBOSITY_MIN		0
+#define	  GUC_LOG_VERBOSITY_MAX		3
+
+#define GUC_CTL_MAX_DWORDS		(GUC_CTL_DEBUG + 1)
+
+struct guc_doorbell_info {
+	u32 db_status;
+	u32 cookie;
+	u32 reserved[14];
+} __packed;
+
+union guc_doorbell_qw {
+	struct {
+		u32 db_status;
+		u32 cookie;
+	};
+	u64 value_qw;
+} __packed;
+
+#define GUC_MAX_DOORBELLS		256
+#define GUC_INVALID_DOORBELL_ID		(GUC_MAX_DOORBELLS)
+
+#define GUC_DB_SIZE			(PAGE_SIZE)
+#define GUC_WQ_SIZE			(PAGE_SIZE * 2)
+
+/* Work item for submitting workloads into work queue of GuC. */
+struct guc_wq_item {
+	u32 header;
+	u32 context_desc;
+	u32 ring_tail;
+	u32 fence_id;
+} __packed;
+
+struct guc_process_desc {
+	u32 context_id;
+	u64 db_base_addr;
+	u32 head;
+	u32 tail;
+	u32 error_offset;
+	u64 wq_base_addr;
+	u32 wq_size_bytes;
+	u32 wq_status;
+	u32 engine_presence;
+	u32 priority;
+	u32 reserved[30];
+} __packed;
+
+/* engine id and context id is packed into guc_execlist_context.context_id*/
+#define GUC_ELC_CTXID_OFFSET		0
+#define GUC_ELC_ENGINE_OFFSET		29
+
+/* The execlist context including software and HW information */
+struct guc_execlist_context {
+	u32 context_desc;
+	u32 context_id;
+	u32 ring_status;
+	u32 ring_lcra;
+	u32 ring_begin;
+	u32 ring_end;
+	u32 ring_next_free_location;
+	u32 ring_current_tail_pointer_value;
+	u8 engine_state_submit_value;
+	u8 engine_state_wait_value;
+	u16 pagefault_count;
+	u16 engine_submit_queue_count;
+} __packed;
+
+/*Context descriptor for communicating between uKernel and Driver*/
+struct guc_context_desc {
+	u32 sched_common_area;
+	u32 context_id;
+	u32 pas_id;
+	u8 engines_used;
+	u64 db_trigger_cpu;
+	u32 db_trigger_uk;
+	u64 db_trigger_phy;
+	u16 db_id;
+
+	struct guc_execlist_context lrc[I915_NUM_RINGS];
+
+	u8 attribute;
+
+	u32 priority;
+
+	u32 wq_sampled_tail_offset;
+	u32 wq_total_submit_enqueues;
+
+	u32 process_desc;
+	u32 wq_addr;
+	u32 wq_size;
+
+	u32 engine_presence;
+
+	u32 reserved0[1];
+	u64 reserved1[1];
+
+	u64 desc_private;
+} __packed;
+
+/* This Action will be programmed in C180 - SOFT_SCRATCH_O_REG */
+enum host2guc_action {
+	HOST2GUC_ACTION_DEFAULT = 0x0,
+	HOST2GUC_ACTION_SAMPLE_FORCEWAKE = 0x6,
+	HOST2GUC_ACTION_ALLOCATE_DOORBELL = 0x10,
+	HOST2GUC_ACTION_DEALLOCATE_DOORBELL = 0x20,
+	HOST2GUC_ACTION_SLPC_REQUEST = 0x3003,
+	HOST2GUC_ACTION_LIMIT
+};
+
+/*
+ * The GuC sends its response to a command by overwriting the
+ * command in SS0. The response is distinguishable from a command
+ * by the fact that all the MASK bits are set. The remaining bits
+ * give more detail.
+ */
+#define	GUC2HOST_RESPONSE_MASK		((u32)0xF0000000)
+#define	GUC2HOST_IS_RESPONSE(x) 	((u32)(x) >= GUC2HOST_RESPONSE_MASK)
+#define	GUC2HOST_STATUS(x)		(GUC2HOST_RESPONSE_MASK | (x))
+
+/* GUC will return status back to SOFT_SCRATCH_O_REG */
+enum guc2host_status {
+	GUC2HOST_STATUS_SUCCESS = GUC2HOST_STATUS(0x0),
+	GUC2HOST_STATUS_ALLOCATE_DOORBELL_FAIL = GUC2HOST_STATUS(0x10),
+	GUC2HOST_STATUS_DEALLOCATE_DOORBELL_FAIL = GUC2HOST_STATUS(0x20),
+	GUC2HOST_STATUS_GENERIC_FAIL = GUC2HOST_STATUS(0x0000F000)
+};
+
+#endif
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 05/15] drm/i915: GuC-specific firmware loader

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

This uses the common firmware loader to fetch the firmware image,
then loads it into the GuC's memory via a dedicated DMA engine.

This patch is derived from GuC loading work originally done by
Vinit Azad and Ben Widawsky. It has been reconstructed to accord
with the common firmware loading mechanism by Dave Gordon as well
as new firmware layout etc.

v2:
    Various improvements per review comments by Chris Wilson

v3:
    Removed 'wait' parameter to intel_guc_ucode_load() as prefetch
        is no longer supported in the common firmware loader, per
	Daniel Vetter's request.
    F/w checker callback fn now returns errno rather than bool.

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/Makefile           |   3 +
 drivers/gpu/drm/i915/i915_dma.c         |   4 +
 drivers/gpu/drm/i915/i915_drv.h         |  11 +
 drivers/gpu/drm/i915/i915_gem.c         |   8 +
 drivers/gpu/drm/i915/i915_reg.h         |   4 +-
 drivers/gpu/drm/i915/intel_guc.h        |  49 ++++
 drivers/gpu/drm/i915/intel_guc_loader.c | 448 ++++++++++++++++++++++++++++++++
 7 files changed, 526 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/i915/intel_guc.h
 create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c

diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index f1f80fc..62a8c83 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -42,6 +42,9 @@ i915-y += i915_cmd_parser.o \
 # generic ancilliary microcontroller support
 i915-y += intel_uc_loader.o
 
+# general-purpose microcontroller (GuC) support
+i915-y += intel_guc_loader.o
+
 # autogenerated null render state
 i915-y += intel_renderstate_gen6.o \
 	  intel_renderstate_gen7.o \
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index c5349fa..730d91b 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -469,6 +469,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
 
 cleanup_gem:
 	mutex_lock(&dev->struct_mutex);
+	intel_guc_ucode_fini(dev);
 	i915_gem_cleanup_ringbuffer(dev);
 	i915_gem_context_fini(dev);
 	mutex_unlock(&dev->struct_mutex);
@@ -866,6 +867,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 
 	intel_uncore_init(dev);
 
+	intel_guc_ucode_init(dev);
+
 	/* Load CSR Firmware for SKL */
 	intel_csr_ucode_init(dev);
 
@@ -1117,6 +1120,7 @@ int i915_driver_unload(struct drm_device *dev)
 	flush_workqueue(dev_priv->wq);
 
 	mutex_lock(&dev->struct_mutex);
+	intel_guc_ucode_fini(dev);
 	i915_gem_cleanup_ringbuffer(dev);
 	i915_gem_context_fini(dev);
 	mutex_unlock(&dev->struct_mutex);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 9618f57..a7ccac5 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -50,6 +50,7 @@
 #include <linux/intel-iommu.h>
 #include <linux/kref.h>
 #include <linux/pm_qos.h>
+#include "intel_guc.h"
 
 /* General customization:
  */
@@ -1687,6 +1688,8 @@ struct drm_i915_private {
 
 	struct i915_virtual_gpu vgpu;
 
+	struct intel_guc guc;
+
 	struct intel_csr csr;
 
 	/* Display CSR-related protection */
@@ -1931,6 +1934,11 @@ static inline struct drm_i915_private *dev_to_i915(struct device *dev)
 	return to_i915(dev_get_drvdata(dev));
 }
 
+static inline struct drm_i915_private *guc_to_i915(struct intel_guc *guc)
+{
+	return container_of(guc, struct drm_i915_private, guc);
+}
+
 /* Iterate over initialised rings */
 #define for_each_ring(ring__, dev_priv__, i__) \
 	for ((i__) = 0; (i__) < I915_NUM_RINGS; (i__)++) \
@@ -2539,6 +2547,9 @@ struct drm_i915_cmd_table {
 
 #define HAS_CSR(dev)	(IS_SKYLAKE(dev))
 
+#define HAS_GUC_UCODE(dev)	(IS_GEN9(dev))
+#define HAS_GUC_SCHED(dev)	(IS_GEN9(dev))
+
 #define INTEL_PCH_DEVICE_ID_MASK		0xff00
 #define INTEL_PCH_IBX_DEVICE_ID_TYPE		0x3b00
 #define INTEL_PCH_CPT_DEVICE_ID_TYPE		0x1c00
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index aa8f4c3..80d7890 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -5076,6 +5076,14 @@ i915_gem_init_hw(struct drm_device *dev)
 			goto out;
 	}
 
+	/*
+	 * We can't enable contexts until all firmware is loaded; if this
+	 * fails, disable GuC submissions and fall back to execlist mode
+	 */
+	ret = intel_guc_ucode_load(dev);
+	if (ret)
+		i915.enable_guc_submission = false;
+
 	/* Now it is safe to go back round and do everything else: */
 	for_each_ring(ring, dev_priv, i) {
 		struct drm_i915_gem_request *req;
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 313b1f9..eefb847 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -6837,7 +6837,9 @@ enum skl_disp_power_wells {
 #define   GEN9_PGCTL_SSB_EU311_ACK	(1 << 14)
 
 #define GEN7_MISCCPCTL			(0x9424)
-#define   GEN7_DOP_CLOCK_GATE_ENABLE	(1<<0)
+#define   GEN7_DOP_CLOCK_GATE_ENABLE		(1<<0)
+#define   GEN8_DOP_CLOCK_GATE_CFCLK_ENABLE	(1<<2)
+#define   GEN8_DOP_CLOCK_GATE_GUC_ENABLE	(1<<4)
 
 /* IVYBRIDGE DPF */
 #define GEN7_L3CDERRST1			0xB008 /* L3CD Error Status 1 */
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
new file mode 100644
index 0000000..b38d2b0
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -0,0 +1,49 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+#ifndef _INTEL_GUC_H_
+#define _INTEL_GUC_H_
+
+#include "intel_uc_loader.h"
+#include "intel_guc_fwif.h"
+#include "i915_guc_reg.h"
+
+struct intel_guc {
+	/* Generic uC firmware management */
+	struct intel_uc_fw guc_fw;
+
+	/* GuC-specific additions */
+	uint16_t fw_major_wanted;
+	uint16_t fw_minor_wanted;
+	uint16_t fw_major_found;
+	uint16_t fw_minor_found;
+
+	uint32_t log_flags;
+};
+
+/* intel_guc_loader.c */
+extern void intel_guc_ucode_init(struct drm_device *dev);
+extern int intel_guc_ucode_load(struct drm_device *dev);
+extern void intel_guc_ucode_fini(struct drm_device *dev);
+
+#endif
diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
new file mode 100644
index 0000000..4929838
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
@@ -0,0 +1,448 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Vinit Azad <vinit.azad@intel.com>
+ *    Ben Widawsky <ben@bwidawsk.net>
+ *    Dave Gordon <david.s.gordon@intel.com>
+ *    Alex Dai <yu.dai@intel.com>
+ */
+#include <linux/firmware.h>
+#include "i915_drv.h"
+#include "intel_guc.h"
+
+/**
+ * DOC: GuC
+ *
+ * intel_guc:
+ * Top level structure of guc. It handles firmware loading and manages client
+ * pool and doorbells. intel_guc owns a i915_guc_client to replace the legacy
+ * ExecList submission.
+ *
+ * Firmware versioning:
+ * The firmware build process will generate a version header file with major and
+ * minor version defined. The versions are built into CSS header of firmware.
+ * i915 kernel driver set the minimal firmware version required per platform.
+ * The firmware installation package will install (symbolic link) proper version
+ * of firmware.
+ *
+ * GuC address space:
+ * GuC does not allow any gfx GGTT address that falls into range [0, WOPCM_TOP),
+ * which is reserved for Boot ROM, SRAM and WOPCM. Currently this top address is
+ * 512K. In order to exclude 0-512K address space from GGTT, all gfx objects
+ * used by GuC is pinned with PIN_OFFSET_BIAS along with size of WOPCM.
+ *
+ * Firmware log:
+ * Firmware log is enabled by setting i915.guc_log_level to non-negative level.
+ * Log data is printed out via reading debugfs i915_guc_log_dump. Reading from
+ * i915_guc_load_status will print out firmware loading status and scratch
+ * registers value.
+ *
+ */
+
+#define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
+MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
+
+static u32 get_gttype(struct drm_i915_private *dev_priv)
+{
+	/* XXX: GT type based on PCI device ID? field seems unused by fw */
+	return 0;
+}
+
+static u32 get_core_family(struct drm_i915_private *dev_priv)
+{
+	switch (INTEL_INFO(dev_priv)->gen) {
+	case 8:
+		return GFXCORE_FAMILY_GEN8;
+	case 9:
+		return GFXCORE_FAMILY_GEN9;
+	default:
+		DRM_ERROR("GUC: unknown gen for scheduler init\n");
+		return GFXCORE_FAMILY_FORCE_ULONG;
+	}
+}
+
+static void set_guc_init_params(struct drm_i915_private *dev_priv)
+{
+	struct intel_guc *guc = &dev_priv->guc;
+	u32 params[GUC_CTL_MAX_DWORDS];
+	int i;
+
+	memset(&params, 0, sizeof(params));
+
+	params[GUC_CTL_DEVICE_INFO] |=
+		(get_gttype(dev_priv) << GUC_CTL_GTTYPE_SHIFT) |
+		(get_core_family(dev_priv) << GUC_CTL_COREFAMILY_SHIFT);
+
+	/* GuC ARAT increment is 10 ns. GuC default scheduler quantum is one
+	 * second. This ARAR is calculated by:
+	 * Scheduler-Quantum-in-ns / ARAT-increment-in-ns = 1000000000 / 10
+	 */
+	params[GUC_CTL_ARAT_HIGH] = 0;
+	params[GUC_CTL_ARAT_LOW] = 100000000;
+
+	params[GUC_CTL_WA] |= GUC_CTL_WA_UK_BY_DRIVER;
+
+	params[GUC_CTL_FEATURE] |= GUC_CTL_DISABLE_SCHEDULER |
+			GUC_CTL_VCS2_ENABLED;
+
+	if (i915.guc_log_level >= 0) {
+		params[GUC_CTL_LOG_PARAMS] = guc->log_flags;
+		params[GUC_CTL_DEBUG] =
+			i915.guc_log_level << GUC_LOG_VERBOSITY_SHIFT;
+	}
+
+	I915_WRITE(SOFT_SCRATCH(0), 0);
+
+	for (i = 0; i < GUC_CTL_MAX_DWORDS; i++)
+		I915_WRITE(SOFT_SCRATCH(1 + i), params[i]);
+}
+
+/*
+ * Read the GuC status register (GUC_STATUS) and store it in the
+ * specified location; then return a boolean indicating whether
+ * the value matches either of two values representing completion
+ * of the GuC boot process.
+ *
+ * This is used for polling the GuC status in a wait_for_atomic()
+ * loop below.
+ */
+static inline bool guc_ucode_response(struct drm_i915_private *dev_priv,
+				      u32 *status)
+{
+	u32 val = I915_READ(GUC_STATUS);
+	*status = val;
+	return ((val & GS_UKERNEL_MASK) == GS_UKERNEL_READY ||
+		(val & GS_UKERNEL_MASK) == GS_UKERNEL_LAPIC_DONE);
+}
+
+/*
+ * Transfer the firmware image to RAM for execution by the microcontroller.
+ *
+ * GuC Firmware layout:
+ * +-------------------------------+  ----
+ * |          CSS header           |  128B
+ * | contains major/minor version  |
+ * +-------------------------------+  ----
+ * |             uCode             |
+ * +-------------------------------+  ----
+ * |         RSA signature         |  256B
+ * +-------------------------------+  ----
+ * |         RSA public Key        |  256B
+ * +-------------------------------+  ----
+ * |       Public key modulus      |    4B
+ * +-------------------------------+  ----
+ *
+ * Architecturally, the DMA engine is bidirectional, and can potentially even
+ * transfer between GTT locations. This functionality is left out of the API
+ * for now as there is no need for it.
+ *
+ * Note that GuC needs the CSS header plus uKernel code to be copied by the
+ * DMA engine in one operation, whereas the RSA signature is loaded via MMIO.
+ */
+
+#define UOS_CSS_HEADER_OFFSET		0
+#define UOS_VER_MINOR_OFFSET		0x44
+#define UOS_VER_MAJOR_OFFSET		0x46
+#define UOS_CSS_HEADER_SIZE		0x80
+#define UOS_RSA_SIG_SIZE		0x100
+#define UOS_CSS_SIGNING_SIZE		0x204
+
+static int guc_ucode_xfer_dma(struct drm_i915_private *dev_priv)
+{
+	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
+	struct drm_i915_gem_object *fw_obj = guc_fw->uc_fw_obj;
+	unsigned long offset;
+	struct sg_table *sg = fw_obj->pages;
+	u32 status, ucode_size, rsa[UOS_RSA_SIG_SIZE / sizeof(u32)];
+	int i, ret = 0;
+
+	/* uCode size, also is where RSA signature starts */
+	offset = ucode_size = guc_fw->uc_fw_size - UOS_CSS_SIGNING_SIZE;
+
+	/* Copy RSA signature from the fw image to HW for verification */
+	sg_pcopy_to_buffer(sg->sgl, sg->nents, rsa, UOS_RSA_SIG_SIZE, offset);
+	for (i = 0; i < UOS_RSA_SIG_SIZE / sizeof(u32); i++)
+		I915_WRITE(UOS_RSA_SCRATCH_0 + i * sizeof(u32), rsa[i]);
+
+	/* Set the source address for the new blob */
+	offset = i915_gem_obj_ggtt_offset(fw_obj);
+	I915_WRITE(DMA_ADDR_0_LOW, lower_32_bits(offset));
+	I915_WRITE(DMA_ADDR_0_HIGH, upper_32_bits(offset) & 0xFFFF);
+
+	/* Set the destination. Current uCode expects an 8k stack starting from
+	 * offset 0. */
+	I915_WRITE(DMA_ADDR_1_LOW, 0x2000);
+
+	/* XXX: The image is automatically transfered to SRAM after the RSA
+	 * verification. This is why the address space is chosen as such. */
+	I915_WRITE(DMA_ADDR_1_HIGH, DMA_ADDRESS_SPACE_WOPCM);
+
+	I915_WRITE(DMA_COPY_SIZE, ucode_size);
+
+	/* Finally start the DMA */
+	I915_WRITE(DMA_CTRL, _MASKED_BIT_ENABLE(UOS_MOVE | START_DMA));
+
+	/*
+	 * Spin-wait for the DMA to complete & the GuC to start up.
+	 * NB: Docs recommend not using the interrupt for completion.
+	 * Measurements indicate this should take no more than 20ms, so a
+	 * timeout here indicates that the GuC has failed and is unusable.
+	 * (Higher levels of the driver will attempt to fall back to
+	 * execlist mode if this happens.)
+	 */
+	ret = wait_for_atomic(guc_ucode_response(dev_priv, &status), 100);
+
+	DRM_DEBUG_DRIVER("DMA status 0x%x, GuC status 0x%x\n",
+			I915_READ(DMA_CTRL), status);
+
+	if ((status & GS_BOOTROM_MASK) == GS_BOOTROM_RSA_FAILED) {
+		DRM_ERROR("%s firmware signature verification failed\n",
+			guc_fw->uc_name);
+		ret = -ENOEXEC;
+	}
+
+	DRM_DEBUG_DRIVER("returning %d\n", ret);
+
+	return ret;
+}
+
+/*
+ * Load the GuC firmware blob into the MinuteIA.
+ */
+static int guc_ucode_xfer(struct drm_i915_private *dev_priv)
+{
+	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
+	struct drm_device *dev = dev_priv->dev;
+	int ret;
+
+	ret = i915_gem_object_set_to_gtt_domain(guc_fw->uc_fw_obj, false);
+	if (ret) {
+		DRM_DEBUG_DRIVER("set-domain failed %d\n", ret);
+		return ret;
+	}
+
+	ret = i915_gem_obj_ggtt_pin(guc_fw->uc_fw_obj, 0, 0);
+	if (ret) {
+		DRM_DEBUG_DRIVER("pin failed %d\n", ret);
+		return ret;
+	}
+
+	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
+
+	/* init WOPCM */
+	I915_WRITE(GUC_WOPCM_SIZE, GUC_WOPCM_SIZE_VALUE);
+	I915_WRITE(DMA_GUC_WOPCM_OFFSET, GUC_WOPCM_OFFSET);
+
+	/* Invalidate GuC TLB to let GuC take the latest updates to GTT. */
+	I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
+
+	/* Set MMIO/WA for GuC init */
+	I915_WRITE(DRBMISC1, DOORBELL_ENABLE);
+
+	/* Enable MIA caching. GuC clock gating is disabled. */
+	I915_WRITE(GUC_SHIM_CONTROL, GUC_SHIM_CONTROL_VALUE);
+
+	/* WaC6DisallowByGfxPause*/
+	I915_WRITE(GEN6_GFXPAUSE, 0x30FFF);
+
+	if (IS_SKYLAKE(dev))
+		I915_WRITE(GEN9_GT_PM_CONFIG, GEN8_GT_DOORBELL_ENABLE);
+	else
+		I915_WRITE(GEN8_GT_PM_CONFIG, GEN8_GT_DOORBELL_ENABLE);
+
+	if (IS_GEN9(dev)) {
+		/* DOP Clock Gating Enable for GuC clocks */
+		I915_WRITE(GEN7_MISCCPCTL, (GEN8_DOP_CLOCK_GATE_GUC_ENABLE |
+					    I915_READ(GEN7_MISCCPCTL)));
+
+		/* allows for 5us before GT can go to RC6 */
+		I915_WRITE(GUC_ARAT_C6DIS, 0x1FF);
+	}
+
+	set_guc_init_params(dev_priv);
+
+	ret = guc_ucode_xfer_dma(dev_priv);
+
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
+
+	/*
+	 * We keep the object pages for reuse during resume. But we can unpin it
+	 * now that DMA has completed, so it doesn't continue to take up space.
+	 */
+	i915_gem_object_ggtt_unpin(guc_fw->uc_fw_obj);
+
+	return ret;
+}
+
+/*
+ * Check the firmware that was found; if it's the wrong size or the wrong
+ * version, return a negative error code. If it's OK, return a positive
+ * status value. Here we can just return INTEL_UC_FW_GOOD; the common loader
+ * code will then save the data for later in a pageable (tmpfs-backed) GEM
+ * object.
+ *
+ * Alternatively (for example if we wanted only part of the image) we could
+ * save the required portion here and then return INTEL_UC_FW_SAVED to tell
+ * the common loader that the data is good, and we've already handled saving
+ * anything we need later.
+ *
+ * The GuC firmware image has the version number embedded at a well-known
+ * offset within the firmware blob; note that major / minor version are
+ * TWO bytes each (i.e. u16), although all pointers and offsets are defined
+ * in terms of bytes (u8).
+ */
+static int guc_ucode_check(struct intel_uc_fw *guc_fw)
+{
+	struct intel_guc *guc = container_of(guc_fw, struct intel_guc, guc_fw);
+	const u8 *css_header = guc_fw->uc_fw_blob->data + UOS_CSS_HEADER_OFFSET;
+	const size_t blobsize = guc_fw->uc_fw_blob->size;
+	const size_t minsize = UOS_CSS_HEADER_SIZE + UOS_CSS_SIGNING_SIZE;
+	const size_t maxsize = GUC_WOPCM_SIZE_VALUE + UOS_CSS_SIGNING_SIZE
+			- 0x8000; /* 32k reserved (8K stack + 24k context) */
+
+	DRM_DEBUG_DRIVER("firmware file size %zu (minimum %zu, maximum %zu)\n",
+		blobsize, minsize, maxsize);
+
+	/* Check the size of the blob befoe examining buffer contents */
+	if (blobsize < minsize || blobsize > maxsize)
+		return -ENOEXEC;
+
+	guc->fw_major_found = *(u16 *)(css_header + UOS_VER_MAJOR_OFFSET);
+	guc->fw_minor_found = *(u16 *)(css_header + UOS_VER_MINOR_OFFSET);
+
+	if (guc->fw_major_found != guc->fw_major_wanted ||
+	    guc->fw_minor_found < guc->fw_minor_wanted) {
+		DRM_ERROR("GuC firmware version %d.%d, required %d.%d\n",
+			guc->fw_major_found, guc->fw_minor_found,
+			guc->fw_major_wanted, guc->fw_minor_wanted);
+		return -ENOEXEC;
+	}
+
+	DRM_DEBUG_DRIVER("firmware version %d.%d OK (minimum %d.%d)\n",
+			guc->fw_major_found, guc->fw_minor_found,
+			guc->fw_major_wanted, guc->fw_minor_wanted);
+
+	return INTEL_UC_FW_GOOD;
+}
+
+/**
+ * intel_guc_ucode_init() - initiate a firmware loading request
+ *
+ * Called early during driver load, before GEM is initialised.
+ * Driver is single threaded, so no mutex is required.
+ *
+ * This just sets parameters for use when intel_guc_ucode_load()
+ * is called later, after GEM initialisation is complete.
+ */
+void intel_guc_ucode_init(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc *guc = &dev_priv->guc;
+	struct intel_uc_fw *guc_fw = &guc->guc_fw;
+	const char *path;
+
+	if (!HAS_GUC_SCHED(dev))
+		i915.enable_guc_submission = false;
+
+	if (!HAS_GUC_UCODE(dev)) {
+		path = NULL;
+	} else if (IS_SKYLAKE(dev)) {
+		path = I915_SKL_GUC_UCODE;
+		guc->fw_major_wanted = 3;
+		guc->fw_minor_wanted = 0;
+	} else {
+		i915.enable_guc_submission = false;
+		path = "";	/* unknown device */
+	}
+
+	intel_uc_fw_init(dev, guc_fw, "GuC", path);
+}
+
+/**
+ * intel_guc_ucode_load() - load GuC uCode into the device
+ *
+ * Called from gem_init_hw() during driver loading and also after a GPU reset.
+ *
+ * Calls the common loader to get the firmware registered earlier. On the first
+ * call, this will actually fetch it from the filesystem; thereafter, we will
+ * already either have the blob in a GEM object, or have determined that no
+ * valid firmware image could be found).
+ *
+ * If we have a good firmware image, transfer it to the h/w.
+ */
+int intel_guc_ucode_load(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
+	int err;
+
+	DRM_DEBUG_DRIVER("%s fw status: fetch %s, load %s\n",
+		guc_fw->uc_name,
+		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
+		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
+
+	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
+		return 0;
+
+	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_SUCCESS &&
+	    guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_FAIL)
+		return -ENOEXEC;
+
+	guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_PENDING;
+	err = intel_uc_fw_fetch(guc_fw, guc_ucode_check);
+	if (err)
+		goto fail;
+
+	err = guc_ucode_xfer(dev_priv);
+	if (err)
+		goto fail;
+
+	guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_SUCCESS;
+
+	DRM_DEBUG_DRIVER("%s fw status: fetch %s, load %s\n",
+		guc_fw->uc_name,
+		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
+		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
+
+	return 0;
+
+fail:
+	if (guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_PENDING)
+		guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_FAIL;
+
+	DRM_ERROR("Failed to initialize GuC, error %d\n", err);
+
+	return err;
+}
+
+/**
+ * intel_guc_ucode_fini() - clean up all allocated resources
+ */
+void intel_guc_ucode_fini(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
+
+	intel_uc_fw_fini(guc_fw);
+}
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 06/15] drm/i915: Debugfs interface to read GuC load status

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

The new node provides access to the status of the common uC loader
code and the GuC-specific loader; also the scratch registers used
for communication between the i915 driver and the GuC firmware.

v2:
    Changes to output formats per Chris Wilson's suggestions

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c | 45 +++++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 71ba519..65d40f1 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -2359,6 +2359,50 @@ static int i915_llc(struct seq_file *m, void *data)
 	return 0;
 }
 
+static void i915_uc_load_status_info(struct seq_file *m, struct intel_uc_fw *uc_fw)
+{
+	seq_printf(m, "%s firmware status:\n",
+			uc_fw->uc_name);
+	seq_printf(m, "\tpath: %s\n",
+			uc_fw->uc_fw_path);
+	seq_printf(m, "\tfetch: %s\n",
+			intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status));
+	seq_printf(m, "\tload: %s\n",
+			intel_uc_fw_status_repr(uc_fw->uc_fw_load_status));
+}
+
+static int i915_guc_load_status_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = m->private;
+	struct drm_i915_private *dev_priv = node->minor->dev->dev_private;
+	u32 tmp, i;
+
+	if (!HAS_GUC_UCODE(dev_priv->dev))
+		return 0;
+
+	i915_uc_load_status_info(m, &dev_priv->guc.guc_fw);
+
+	seq_printf(m, "\tversion wanted: %d.%d\n",
+		dev_priv->guc.fw_major_wanted, dev_priv->guc.fw_minor_wanted);
+	seq_printf(m, "\tversion found: %d.%d\n",
+		dev_priv->guc.fw_major_found, dev_priv->guc.fw_minor_found);
+
+	tmp = I915_READ(GUC_STATUS);
+
+	seq_printf(m, "\nGuC status 0x%08x:\n", tmp);
+	seq_printf(m, "\tBootrom status = 0x%x\n",
+		(tmp & GS_BOOTROM_MASK) >> GS_BOOTROM_SHIFT);
+	seq_printf(m, "\tuKernel status = 0x%x\n",
+		(tmp & GS_UKERNEL_MASK) >> GS_UKERNEL_SHIFT);
+	seq_printf(m, "\tMIA Core status = 0x%x\n",
+		(tmp & GS_MIA_MASK) >> GS_MIA_SHIFT);
+	seq_puts(m, "\nScratch registers:\n");
+	for (i = 0; i < 16; i++)
+		seq_printf(m, "\t%2d: \t0x%x\n", i, I915_READ(SOFT_SCRATCH(i)));
+
+	return 0;
+}
+
 static int i915_edp_psr_status(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = m->private;
@@ -5073,6 +5117,7 @@ static const struct drm_info_list i915_debugfs_list[] = {
 	{"i915_gem_hws_bsd", i915_hws_info, 0, (void *)VCS},
 	{"i915_gem_hws_vebox", i915_hws_info, 0, (void *)VECS},
 	{"i915_gem_batch_pool", i915_gem_batch_pool_info, 0},
+	{"i915_guc_load_status", i915_guc_load_status_info, 0},
 	{"i915_frequency_info", i915_frequency_info, 0},
 	{"i915_hangcheck_info", i915_hangcheck_info, 0},
 	{"i915_drpc_info", i915_drpc_info, 0},
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 07/15] drm/i915: Expose two LRC functions for GuC submission mode

[Dave Gordon]

GuC submission is basically execlist submission, but with the GuC
handling the actual writes to the ELSP and the resulting context
switch interrupts. So to prepare a context for submission via the
GuC, we need some of the same functions used in execlist mode.
This commit exposes two such functions, changing their names to
better describe what they do (they're related to logical ring
contexts rather than to execlists per se).

v2:
    Replaces previous "drm/i915: Move execlists defines from .c to .h"

v3:
    Incorporates a change to one of the functions exposed here that was
    previously part of an internal patch, but which was omitted from
    the version recently committed to drm-intel-nightly:
	7a01a0a drm/i915/lrc: Update PDPx registers with lri commands
    So we reinstate this change here.

Issue: VIZ-4884
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/intel_lrc.c | 37 +++++++++++++------------------------
 drivers/gpu/drm/i915/intel_lrc.h |  5 +++++
 2 files changed, 18 insertions(+), 24 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 9b928ab..164c63a 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -262,8 +262,8 @@ u32 intel_execlists_ctx_id(struct drm_i915_gem_object *ctx_obj)
 	return lrca >> 12;
 }
 
-static uint64_t execlists_ctx_descriptor(struct intel_engine_cs *ring,
-					 struct drm_i915_gem_object *ctx_obj)
+uint64_t intel_lr_context_descriptor(struct intel_engine_cs *ring,
+				     struct drm_i915_gem_object *ctx_obj)
 {
 	struct drm_device *dev = ring->dev;
 	uint64_t desc;
@@ -304,13 +304,13 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
 
 	/* XXX: You must always write both descriptors in the order below. */
 	if (ctx_obj1)
-		temp = execlists_ctx_descriptor(ring, ctx_obj1);
+		temp = intel_lr_context_descriptor(ring, ctx_obj1);
 	else
 		temp = 0;
 	desc[1] = (u32)(temp >> 32);
 	desc[0] = (u32)temp;
 
-	temp = execlists_ctx_descriptor(ring, ctx_obj0);
+	temp = intel_lr_context_descriptor(ring, ctx_obj0);
 	desc[3] = (u32)(temp >> 32);
 	desc[2] = (u32)temp;
 
@@ -329,10 +329,10 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
 	spin_unlock(&dev_priv->uncore.lock);
 }
 
-static int execlists_update_context(struct drm_i915_gem_object *ctx_obj,
-				    struct drm_i915_gem_object *ring_obj,
-				    struct i915_hw_ppgtt *ppgtt,
-				    u32 tail)
+/* Update the ringbuffer pointer and tail offset in a saved context image */
+void intel_lr_context_update(struct drm_i915_gem_object *ctx_obj,
+			     struct drm_i915_gem_object *ring_obj,
+			     u32 tail)
 {
 	struct page *page;
 	uint32_t *reg_state;
@@ -340,22 +340,11 @@ static int execlists_update_context(struct drm_i915_gem_object *ctx_obj,
 	page = i915_gem_object_get_page(ctx_obj, 1);
 	reg_state = kmap_atomic(page);
 
-	reg_state[CTX_RING_TAIL+1] = tail;
 	reg_state[CTX_RING_BUFFER_START+1] = i915_gem_obj_ggtt_offset(ring_obj);
-
-	/* True PPGTT with dynamic page allocation: update PDP registers and
-	 * point the unallocated PDPs to the scratch page
-	 */
-	if (ppgtt) {
-		ASSIGN_CTX_PDP(ppgtt, reg_state, 3);
-		ASSIGN_CTX_PDP(ppgtt, reg_state, 2);
-		ASSIGN_CTX_PDP(ppgtt, reg_state, 1);
-		ASSIGN_CTX_PDP(ppgtt, reg_state, 0);
-	}
+	if (tail != ~0u)
+		reg_state[CTX_RING_TAIL+1] = tail;
 
 	kunmap_atomic(reg_state);
-
-	return 0;
 }
 
 static void execlists_submit_contexts(struct intel_engine_cs *ring,
@@ -371,7 +360,7 @@ static void execlists_submit_contexts(struct intel_engine_cs *ring,
 	WARN_ON(!i915_gem_obj_is_pinned(ctx_obj0));
 	WARN_ON(!i915_gem_obj_is_pinned(ringbuf0->obj));
 
-	execlists_update_context(ctx_obj0, ringbuf0->obj, to0->ppgtt, tail0);
+	intel_lr_context_update(ctx_obj0, ringbuf0->obj, tail0);
 
 	if (to1) {
 		ringbuf1 = to1->engine[ring->id].ringbuf;
@@ -380,7 +369,7 @@ static void execlists_submit_contexts(struct intel_engine_cs *ring,
 		WARN_ON(!i915_gem_obj_is_pinned(ctx_obj1));
 		WARN_ON(!i915_gem_obj_is_pinned(ringbuf1->obj));
 
-		execlists_update_context(ctx_obj1, ringbuf1->obj, to1->ppgtt, tail1);
+		intel_lr_context_update(ctx_obj1, ringbuf1->obj, tail1);
 	}
 
 	execlists_elsp_write(ring, ctx_obj0, ctx_obj1);
@@ -2032,7 +2021,7 @@ populate_lr_context(struct intel_context *ctx, struct drm_i915_gem_object *ctx_o
 	reg_state[CTX_RING_TAIL+1] = 0;
 	reg_state[CTX_RING_BUFFER_START] = RING_START(ring->mmio_base);
 	/* Ring buffer start address is not known until the buffer is pinned.
-	 * It is written to the context image in execlists_update_context()
+	 * It is written to the context image in intel_lr_context_update()
 	 */
 	reg_state[CTX_RING_BUFFER_CONTROL] = RING_CTL(ring->mmio_base);
 	reg_state[CTX_RING_BUFFER_CONTROL+1] =
diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h
index f59940a..b3659a1 100644
--- a/drivers/gpu/drm/i915/intel_lrc.h
+++ b/drivers/gpu/drm/i915/intel_lrc.h
@@ -73,6 +73,11 @@ void intel_lr_context_unpin(struct intel_engine_cs *ring,
 		struct intel_context *ctx);
 void intel_lr_context_reset(struct drm_device *dev,
 			struct intel_context *ctx);
+void intel_lr_context_update(struct drm_i915_gem_object *ctx_obj,
+			     struct drm_i915_gem_object *ring_obj,
+			     u32 tail);
+uint64_t intel_lr_context_descriptor(struct intel_engine_cs *ring,
+				     struct drm_i915_gem_object *ctx_obj);
 
 /* Execlists */
 int intel_sanitize_enable_execlists(struct drm_device *dev, int enable_execlists);
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 08/15] drm/i915: GuC submission setup, phase 1

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

This adds the first of the data structures used to communicate with the
GuC (the pool of guc_context structures).

We create a GuC-specific wrapper round the GEM object allocator as all
GEM objects shared with the GuC must be pinned into GGTT space at an
address that is NOT in the range [0..WOPCM_SIZE), as that range of GGTT
addresses is not accessible to the GuC (from the GuC's point of view,
it's permanently reserved for other objects such as the BootROM & SRAM).

Later, we will need to allocate additional GuC-sharable objects for the
submission client(s) and the GuC's debug log.

v2:
    Remove redundant initialisation [Chris Wilson]
    Defer adding struct members until needed [Chris Wilson]
    Local functions should pass dev_priv rather than dev [Chris Wilson]

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/Makefile              |   3 +-
 drivers/gpu/drm/i915/i915_guc_submission.c | 114 +++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_guc.h           |   7 ++
 drivers/gpu/drm/i915/intel_guc_loader.c    |  21 ++++++
 4 files changed, 144 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/i915/i915_guc_submission.c

diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 62a8c83..00f89d26 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -43,7 +43,8 @@ i915-y += i915_cmd_parser.o \
 i915-y += intel_uc_loader.o
 
 # general-purpose microcontroller (GuC) support
-i915-y += intel_guc_loader.o
+i915-y += intel_guc_loader.o \
+	  i915_guc_submission.o
 
 # autogenerated null render state
 i915-y += intel_renderstate_gen6.o \
diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
new file mode 100644
index 0000000..70a0405
--- /dev/null
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -0,0 +1,114 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+#include <linux/circ_buf.h>
+#include "i915_drv.h"
+#include "intel_guc.h"
+
+/**
+ * gem_allocate_guc_obj() - Allocate gem object for GuC usage
+ * @dev:	drm device
+ * @size:	size of object
+ *
+ * This is a wrapper to create a gem obj. In order to use it inside GuC, the
+ * object needs to be pinned lifetime. Also we must pin it to gtt space other
+ * than [0, GUC_WOPCM_SIZE] because this range is reserved inside GuC.
+ *
+ * Return:	A drm_i915_gem_object if successful, otherwise NULL.
+ */
+static struct drm_i915_gem_object *gem_allocate_guc_obj(struct drm_device *dev,
+							u32 size)
+{
+	struct drm_i915_gem_object *obj;
+
+	obj = i915_gem_alloc_object(dev, size);
+	if (!obj)
+		return NULL;
+
+	if (i915_gem_object_get_pages(obj)) {
+		drm_gem_object_unreference(&obj->base);
+		return NULL;
+	}
+
+	if (i915_gem_obj_ggtt_pin(obj, PAGE_SIZE,
+			PIN_OFFSET_BIAS | GUC_WOPCM_SIZE_VALUE)) {
+		drm_gem_object_unreference(&obj->base);
+		return NULL;
+	}
+
+	return obj;
+}
+
+/**
+ * gem_release_guc_obj() - Release gem object allocated for GuC usage
+ * @obj:	gem obj to be released
+  */
+static void gem_release_guc_obj(struct drm_i915_gem_object *obj)
+{
+	if (!obj)
+		return;
+
+	if (i915_gem_obj_is_pinned(obj))
+		i915_gem_object_ggtt_unpin(obj);
+
+	drm_gem_object_unreference(&obj->base);
+}
+
+/*
+ * Set up the memory resources to be shared with the GuC.  At this point,
+ * we require just one object that can be mapped through the GGTT.
+ */
+int i915_guc_submission_init(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	const size_t ctxsize = sizeof(struct guc_context_desc);
+	const size_t poolsize = GUC_MAX_GPU_CONTEXTS * ctxsize;
+	const size_t gemsize = round_up(poolsize, PAGE_SIZE);
+	struct intel_guc *guc = &dev_priv->guc;
+
+	if (!i915.enable_guc_submission)
+		return 0; /* not enabled  */
+
+	if (guc->ctx_pool_obj)
+		return 0; /* already allocated */
+
+	guc->ctx_pool_obj = gem_allocate_guc_obj(dev_priv->dev, gemsize);
+	if (!guc->ctx_pool_obj)
+		return -ENOMEM;
+
+	ida_init(&guc->ctx_ids);
+
+	return 0;
+}
+
+void i915_guc_submission_fini(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc *guc = &dev_priv->guc;
+
+	if (guc->ctx_pool_obj)
+		ida_destroy(&guc->ctx_ids);
+	gem_release_guc_obj(guc->ctx_pool_obj);
+	guc->ctx_pool_obj = NULL;
+}
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
index b38d2b0..f9e5e74 100644
--- a/drivers/gpu/drm/i915/intel_guc.h
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -39,6 +39,9 @@ struct intel_guc {
 	uint16_t fw_minor_found;
 
 	uint32_t log_flags;
+
+	struct drm_i915_gem_object *ctx_pool_obj;
+	struct ida ctx_ids;
 };
 
 /* intel_guc_loader.c */
@@ -46,4 +49,8 @@ extern void intel_guc_ucode_init(struct drm_device *dev);
 extern int intel_guc_ucode_load(struct drm_device *dev);
 extern void intel_guc_ucode_fini(struct drm_device *dev);
 
+/* i915_guc_submission.c */
+int i915_guc_submission_init(struct drm_device *dev);
+void i915_guc_submission_fini(struct drm_device *dev);
+
 #endif
diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
index 4929838..deeaa20 100644
--- a/drivers/gpu/drm/i915/intel_guc_loader.c
+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
@@ -111,6 +111,21 @@ static void set_guc_init_params(struct drm_i915_private *dev_priv)
 			i915.guc_log_level << GUC_LOG_VERBOSITY_SHIFT;
 	}
 
+	/* If GuC scheduling is enabled, setup params here. */
+	if (i915.enable_guc_submission) {
+		u32 pgs = i915_gem_obj_ggtt_offset(dev_priv->guc.ctx_pool_obj);
+		u32 ctx_in_16 = GUC_MAX_GPU_CONTEXTS / 16;
+
+		pgs >>= PAGE_SHIFT;
+		params[GUC_CTL_CTXINFO] = (pgs << GUC_CTL_BASE_ADDR_SHIFT) |
+			(ctx_in_16 << GUC_CTL_CTXNUM_IN16_SHIFT);
+
+		params[GUC_CTL_FEATURE] |= GUC_CTL_KERNEL_SUBMISSIONS;
+
+		/* Unmask this bit to enable GuC scheduler */
+		params[GUC_CTL_FEATURE] &= ~GUC_CTL_DISABLE_SCHEDULER;
+	}
+
 	I915_WRITE(SOFT_SCRATCH(0), 0);
 
 	for (i = 0; i < GUC_CTL_MAX_DWORDS; i++)
@@ -414,6 +429,10 @@ int intel_guc_ucode_load(struct drm_device *dev)
 	if (err)
 		goto fail;
 
+	err = i915_guc_submission_init(dev);
+	if (err)
+		goto fail;
+
 	err = guc_ucode_xfer(dev_priv);
 	if (err)
 		goto fail;
@@ -444,5 +463,7 @@ void intel_guc_ucode_fini(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
 
+	i915_guc_submission_fini(dev);
+
 	intel_uc_fw_fini(guc_fw);
 }
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 09/15] drm/i915: Enable GuC firmware log

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

Allocate a GEM object to hold GuC log data. A debugfs interface
(i915_guc_log_dump) is provided to print out the log content.

v2:
    Add struct members at point of use [Chris Wilson]

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c        | 29 +++++++++++++++++++
 drivers/gpu/drm/i915/i915_guc_submission.c | 46 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_guc.h           |  1 +
 3 files changed, 76 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 65d40f1..ad909f5 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -2403,6 +2403,34 @@ static int i915_guc_load_status_info(struct seq_file *m, void *data)
 	return 0;
 }
 
+static int i915_guc_log_dump(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_i915_gem_object *log_obj = dev_priv->guc.log_obj;
+	u32 *log;
+	int i = 0, pg;
+
+	if (!log_obj)
+		return 0;
+
+	for (pg = 0; pg < log_obj->base.size / PAGE_SIZE; pg++) {
+		log = kmap_atomic(i915_gem_object_get_page(log_obj, pg));
+
+		for (i = 0; i < PAGE_SIZE / sizeof(u32); i += 4)
+			seq_printf(m, "0x%08x 0x%08x 0x%08x 0x%08x\n",
+				   *(log + i), *(log + i + 1),
+				   *(log + i + 2), *(log + i + 3));
+
+		kunmap_atomic(log);
+	}
+
+	seq_putc(m, '\n');
+
+	return 0;
+}
+
 static int i915_edp_psr_status(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = m->private;
@@ -5118,6 +5146,7 @@ static const struct drm_info_list i915_debugfs_list[] = {
 	{"i915_gem_hws_vebox", i915_hws_info, 0, (void *)VECS},
 	{"i915_gem_batch_pool", i915_gem_batch_pool_info, 0},
 	{"i915_guc_load_status", i915_guc_load_status_info, 0},
+	{"i915_guc_log_dump", i915_guc_log_dump, 0},
 	{"i915_frequency_info", i915_frequency_info, 0},
 	{"i915_hangcheck_info", i915_hangcheck_info, 0},
 	{"i915_drpc_info", i915_drpc_info, 0},
diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
index 70a0405..e9d46d6 100644
--- a/drivers/gpu/drm/i915/i915_guc_submission.c
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -75,6 +75,47 @@ static void gem_release_guc_obj(struct drm_i915_gem_object *obj)
 	drm_gem_object_unreference(&obj->base);
 }
 
+static void guc_create_log(struct intel_guc *guc)
+{
+	struct drm_i915_private *dev_priv = guc_to_i915(guc);
+	struct drm_i915_gem_object *obj;
+	unsigned long offset;
+	uint32_t size, flags;
+
+	if (i915.guc_log_level < GUC_LOG_VERBOSITY_MIN)
+		return;
+
+	if (i915.guc_log_level > GUC_LOG_VERBOSITY_MAX)
+		i915.guc_log_level = GUC_LOG_VERBOSITY_MAX;
+
+	/* The first page is to save log buffer state. Allocate one
+	 * extra page for others in case for overlap */
+	size = (1 + GUC_LOG_DPC_PAGES + 1 +
+		GUC_LOG_ISR_PAGES + 1 +
+		GUC_LOG_CRASH_PAGES + 1) << PAGE_SHIFT;
+
+	obj = guc->log_obj;
+	if (!obj) {
+		obj = gem_allocate_guc_obj(dev_priv->dev, size);
+		if (!obj) {
+			/* logging will be off */
+			i915.guc_log_level = -1;
+			return;
+		}
+
+		guc->log_obj = obj;
+	}
+
+	/* each allocated unit is a page */
+	flags = GUC_LOG_VALID | GUC_LOG_NOTIFY_ON_HALF_FULL |
+		(GUC_LOG_DPC_PAGES << GUC_LOG_DPC_SHIFT) |
+		(GUC_LOG_ISR_PAGES << GUC_LOG_ISR_SHIFT) |
+		(GUC_LOG_CRASH_PAGES << GUC_LOG_CRASH_SHIFT);
+
+	offset = i915_gem_obj_ggtt_offset(obj) >> PAGE_SHIFT; /* in pages */
+	guc->log_flags = (offset << GUC_LOG_BUF_ADDR_SHIFT) | flags;
+}
+
 /*
  * Set up the memory resources to be shared with the GuC.  At this point,
  * we require just one object that can be mapped through the GGTT.
@@ -99,6 +140,8 @@ int i915_guc_submission_init(struct drm_device *dev)
 
 	ida_init(&guc->ctx_ids);
 
+	guc_create_log(guc);
+
 	return 0;
 }
 
@@ -107,6 +150,9 @@ void i915_guc_submission_fini(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_guc *guc = &dev_priv->guc;
 
+	gem_release_guc_obj(dev_priv->guc.log_obj);
+	guc->log_obj = NULL;
+
 	if (guc->ctx_pool_obj)
 		ida_destroy(&guc->ctx_ids);
 	gem_release_guc_obj(guc->ctx_pool_obj);
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
index f9e5e74..3de9096 100644
--- a/drivers/gpu/drm/i915/intel_guc.h
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -39,6 +39,7 @@ struct intel_guc {
 	uint16_t fw_minor_found;
 
 	uint32_t log_flags;
+	struct drm_i915_gem_object *log_obj;
 
 	struct drm_i915_gem_object *ctx_pool_obj;
 	struct ida ctx_ids;
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 10/15] drm/i915: Implementation of GuC client

[Dave Gordon]

A GuC client has its own doorbell and workqueue. It maintains the
doorbell cache line, process description object and work queue item.

A default guc_client is created for the i915 driver to use for
normal-priority in-order submission.

Note that the created client is not yet ready for use; doorbell
allocation will fail as we haven't yet linked the GuC's context
descriptor to the default contexts for each ring (see later patch).
However the driver will still be usable, automatically falling
back to execlist mode.

v2:
    Defer adding structure members until needed [Chris Wilson]
    Rationalise type declarations [Chris Wilson]

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_guc_submission.c | 661 +++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_guc.h           |  43 ++
 drivers/gpu/drm/i915/intel_guc_loader.c    |  10 +
 3 files changed, 714 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
index e9d46d6..3c28c8b 100644
--- a/drivers/gpu/drm/i915/i915_guc_submission.c
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -27,6 +27,524 @@
 #include "intel_guc.h"
 
 /**
+ * DOC: GuC Client
+ *
+ * i915_guc_client:
+ * We use the term client to avoid confusion with contexts. A i915_guc_client is
+ * equivalent to GuC object guc_context_desc. This context descriptor is
+ * allocated from a pool of 1024 entries. Kernel driver will allocate doorbell
+ * and workqueue for it. Also the process descriptor (guc_process_desc), which
+ * is mapped to client space. So the client can write Work Item then ring the
+ * doorbell.
+ *
+ * To simplify the implementation, we allocate one gem object that contains all
+ * pages for doorbell, process descriptor and workqueue.
+ *
+ * The Scratch registers:
+ * There are 16 MMIO-based registers start from 0xC180. The kernel driver writes
+ * a value to the action register (SOFT_SCRATCH_0) along with any data. It then
+ * triggers an interrupt on the GuC via another register write (0xC4C8).
+ * Firmware writes a success/fail code back to the action register after
+ * processes the request. The kernel driver polls waiting for this update and
+ * then proceeds.
+ * See host2guc_action()
+ *
+ * Doorbells:
+ * Doorbells are interrupts to uKernel. A doorbell is a single cache line (QW)
+ * mapped into process space.
+ *
+ * Work Items:
+ * There are several types of work items that the host may place into a
+ * workqueue, each with its own requirements and limitations. Currently only
+ * WQ_TYPE_INORDER is needed to support legacy submission via GuC, which
+ * represents in-order queue. The kernel driver packs ring tail pointer and an
+ * ELSP context descriptor dword into Work Item.
+ * See guc_add_workqueue_item()
+ *
+ */
+
+/*
+ * Read GuC command/status register (SOFT_SCRATCH_0)
+ * Return true if it contains a response rather than a command
+ */
+static inline bool host2guc_action_response(struct drm_i915_private *dev_priv,
+					    u32 *status)
+{
+	u32 val = I915_READ(SOFT_SCRATCH(0));
+	*status = val;
+	return GUC2HOST_IS_RESPONSE(val);
+}
+
+static int host2guc_action(struct intel_guc *guc, u32 *data, u32 len)
+{
+	struct drm_i915_private *dev_priv = guc_to_i915(guc);
+	u32 status;
+	int i;
+	int ret;
+
+	if (WARN_ON(len < 1 || len > 15))
+		return -EINVAL;
+
+	spin_lock(&dev_priv->guc.host2guc_lock);
+
+	dev_priv->guc.action_count += 1;
+	dev_priv->guc.action_cmd = data[0];
+
+	for (i = 0; i < len; i++)
+		I915_WRITE(SOFT_SCRATCH(i), data[i]);
+
+	POSTING_READ(SOFT_SCRATCH(i - 1));
+
+	I915_WRITE(HOST2GUC_INTERRUPT, HOST2GUC_TRIGGER);
+
+	ret = wait_for_atomic(host2guc_action_response(dev_priv, &status), 10);
+	if (status != GUC2HOST_STATUS_SUCCESS) {
+		/* either GuC doesn't respond, which is a TIMEOUT,
+		 * or a failure code is returned. */
+		if (ret != -ETIMEDOUT)
+			ret = -EIO;
+
+		DRM_ERROR("GUC: host2guc action 0x%X failed. ret=%d "
+				"status=0x%08X response=0x%08X\n",
+				data[0], ret, status,
+				I915_READ(SOFT_SCRATCH(15)));
+
+		dev_priv->guc.action_fail += 1;
+		dev_priv->guc.action_err = ret;
+	}
+	dev_priv->guc.action_status = status;
+
+	spin_unlock(&dev_priv->guc.host2guc_lock);
+
+	return ret;
+}
+
+/*
+ * Tell the GuC to allocate or deallocate a specific doorbell
+ */
+
+static int host2guc_allocate_doorbell(struct intel_guc *guc,
+				      struct i915_guc_client *client)
+{
+	u32 data[2];
+
+	data[0] = HOST2GUC_ACTION_ALLOCATE_DOORBELL;
+	data[1] = client->ctx_index;
+
+	return host2guc_action(guc, data, 2);
+}
+
+static int host2guc_release_doorbell(struct intel_guc *guc,
+				     struct i915_guc_client *client)
+{
+	u32 data[2];
+
+	data[0] = HOST2GUC_ACTION_DEALLOCATE_DOORBELL;
+	data[1] = client->ctx_index;
+
+	return host2guc_action(guc, data, 2);
+}
+
+/*
+ * Initialise, update, or clear doorbell data shared with the GuC
+ *
+ * These functions modify shared data and so need access to the mapped
+ * client object which contains the page being used for the doorbell
+ */
+
+static void guc_init_doorbell(struct intel_guc *guc,
+			      struct i915_guc_client *client)
+{
+	struct guc_doorbell_info *doorbell;
+	void *base;
+
+	base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0));
+	doorbell = base + client->doorbell_offset;
+
+	doorbell->db_status = 1;
+	doorbell->cookie = 0;
+
+	kunmap_atomic(base);
+}
+
+static int guc_ring_doorbell(struct i915_guc_client *gc)
+{
+	struct guc_process_desc *desc;
+	union guc_doorbell_qw db_cmp, db_exc, db_ret;
+	union guc_doorbell_qw *db;
+	void *base;
+	int attempt = 2, ret = -EAGAIN;
+
+	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0));
+	desc = base + gc->proc_desc_offset;
+
+	/* Update the tail so it is visible to GuC */
+	desc->tail = gc->wq_tail;
+
+	/* current cookie */
+	db_cmp.db_status = GUC_DOORBELL_ENABLED;
+	db_cmp.cookie = gc->cookie;
+
+	/* cookie to be updated */
+	db_exc.db_status = GUC_DOORBELL_ENABLED;
+	db_exc.cookie = gc->cookie + 1;
+	if (db_exc.cookie == 0)
+		db_exc.cookie = 1;
+
+	/* pointer of current doorbell cacheline */
+	db = base + gc->doorbell_offset;
+
+	while (attempt--) {
+		/* lets ring the doorbell */
+		db_ret.value_qw = atomic64_cmpxchg((atomic64_t *)db,
+			db_cmp.value_qw, db_exc.value_qw);
+
+		/* if the exchange was successfully executed */
+		if (db_ret.value_qw == db_cmp.value_qw) {
+			/* db was successfully rung */
+			gc->cookie = db_exc.cookie;
+			ret = 0;
+			break;
+		}
+
+		/* XXX: doorbell was lost and need to acquire it again */
+		if (db_ret.db_status == GUC_DOORBELL_DISABLED)
+			break;
+
+		DRM_ERROR("Cookie mismatch. Expected %d, returned %d\n",
+			  db_cmp.cookie, db_ret.cookie);
+
+		/* update the cookie to newly read cookie from GuC */
+		db_cmp.cookie = db_ret.cookie;
+		db_exc.cookie = db_ret.cookie + 1;
+		if (db_exc.cookie == 0)
+			db_exc.cookie = 1;
+	}
+
+	kunmap_atomic(base);
+	return ret;
+}
+
+static void guc_disable_doorbell(struct intel_guc *guc,
+				 struct i915_guc_client *client)
+{
+	struct drm_i915_private *dev_priv = guc_to_i915(guc);
+	struct guc_doorbell_info *doorbell;
+	void *base;
+	int drbreg = GEN8_DRBREGL(client->doorbell_id);
+	int value;
+
+	base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0));
+	doorbell = base + client->doorbell_offset;
+
+	doorbell->db_status = 0;
+
+	kunmap_atomic(base);
+
+	I915_WRITE(drbreg, I915_READ(drbreg) & ~GEN8_DRB_VALID);
+
+	value = I915_READ(drbreg);
+	WARN_ON((value & GEN8_DRB_VALID) != 0);
+
+	I915_WRITE(GEN8_DRBREGU(client->doorbell_id), 0);
+	I915_WRITE(drbreg, 0);
+
+	/* XXX: wait for any interrupts */
+	/* XXX: wait for workqueue to drain */
+}
+
+/*
+ * Select, assign and relase doorbell cachelines
+ *
+ * These functions track which doorbell cachelines are in use.
+ * The data they manipulate is protected by the host2guc lock.
+ */
+
+static uint32_t select_doorbell_cacheline(struct intel_guc *guc)
+{
+	const uint32_t cacheline_size = boot_cpu_data.x86_clflush_size;
+	uint32_t offset;
+
+	spin_lock(&guc->host2guc_lock);
+
+	/* Doorbell uses a single cache line within a page */
+	offset = guc->db_cacheline & PAGE_MASK;
+
+	/* Moving to next cache line to reduce contention */
+	guc->db_cacheline += cacheline_size;
+
+	spin_unlock(&guc->host2guc_lock);
+
+	return offset;
+}
+
+static uint16_t assign_doorbell(struct intel_guc *guc, uint32_t priority)
+{
+	/* The bitmap is split into two halves - high and normal priority. */
+	const uint16_t half = GUC_MAX_DOORBELLS / 2;
+	const uint16_t start = (priority <= GUC_CTX_PRIORITY_HIGH) ? half : 0;
+	const uint16_t end = start + half;
+	uint16_t id;
+
+	spin_lock(&guc->host2guc_lock);
+	id = find_next_zero_bit(guc->doorbell_bitmap, end, start);
+	if (id == end)
+		id = GUC_INVALID_DOORBELL_ID;
+	else
+		bitmap_set(guc->doorbell_bitmap, id, 1);
+	spin_unlock(&guc->host2guc_lock);
+
+	return id;
+}
+
+static void release_doorbell(struct intel_guc *guc, uint16_t id)
+{
+	spin_lock(&guc->host2guc_lock);
+	bitmap_clear(guc->doorbell_bitmap, id, 1);
+	spin_unlock(&guc->host2guc_lock);
+}
+
+/*
+ * Initialise the process descriptor shared with the GuC firmware.
+ */
+static void guc_init_proc_desc(struct intel_guc *guc,
+			       struct i915_guc_client *client)
+{
+	struct guc_process_desc *desc;
+	void *base;
+
+	base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0));
+	desc = base + client->proc_desc_offset;
+
+	memset(desc, 0, sizeof(*desc));
+
+	/*
+	 * XXX: pDoorbell and WQVBaseAddress are pointers in process address
+	 * space for ring3 clients (set them as in mmap_ioctl) or kernel
+	 * space for kernel clients (map on demand instead? May make debug
+	 * easier to have it mapped).
+	 */
+	desc->wq_base_addr = 0;
+	desc->db_base_addr = 0;
+
+	desc->context_id = client->ctx_index;
+	desc->wq_size_bytes = client->wq_size;
+	desc->wq_status = WQ_STATUS_ACTIVE;
+	desc->priority = client->priority;
+
+	kunmap_atomic(base);
+}
+
+/*
+ * Initialise/clear the context descriptor shared with the GuC firmware.
+ *
+ * This descriptor tells the GuC where (in GGTT space) to find the important
+ * data structures relating to this client (doorbell, process descriptor,
+ * write queue, etc).
+ */
+
+static void guc_init_ctx_desc(struct intel_guc *guc,
+			      struct i915_guc_client *client)
+{
+	struct guc_context_desc desc;
+	struct sg_table *sg;
+
+	memset(&desc, 0, sizeof(desc));
+
+	desc.attribute = GUC_CTX_DESC_ATTR_ACTIVE | GUC_CTX_DESC_ATTR_KERNEL;
+	desc.context_id = client->ctx_index;
+	desc.priority = client->priority;
+	desc.engines_used = (1 << RCS) | (1 << VCS) | (1 << BCS) |
+			    (1 << VECS) | (1 << VCS2); /* all engines */
+	desc.db_id = client->doorbell_id;
+
+	/*
+	 * The CPU address is only needed at certain points, so kmap_atomic on
+	 * demand instead of storing it in the ctx descriptor.
+	 * XXX: May make debug easier to have it mapped
+	 */
+	desc.db_trigger_cpu = 0;
+	desc.db_trigger_uk = client->doorbell_offset +
+		i915_gem_obj_ggtt_offset(client->client_obj);
+	desc.db_trigger_phy = client->doorbell_offset +
+		sg_dma_address(client->client_obj->pages->sgl);
+
+	desc.process_desc = client->proc_desc_offset +
+		i915_gem_obj_ggtt_offset(client->client_obj);
+
+	desc.wq_addr = client->wq_offset +
+		i915_gem_obj_ggtt_offset(client->client_obj);
+
+	desc.wq_size = client->wq_size;
+
+	/*
+	 * XXX: Take LRCs from an existing intel_context if this is not an
+	 * IsKMDCreatedContext client
+	 */
+	desc.desc_private = (uintptr_t)client;
+
+	/* Pool context is pinned already */
+	sg = guc->ctx_pool_obj->pages;
+	sg_pcopy_from_buffer(sg->sgl, sg->nents, &desc, sizeof(desc),
+			     sizeof(desc) * client->ctx_index);
+}
+
+static void guc_fini_ctx_desc(struct intel_guc *guc,
+			      struct i915_guc_client *client)
+{
+	struct guc_context_desc desc;
+	struct sg_table *sg;
+
+	memset(&desc, 0, sizeof(desc));
+
+	sg = guc->ctx_pool_obj->pages;
+	sg_pcopy_from_buffer(sg->sgl, sg->nents, &desc, sizeof(desc),
+			     sizeof(desc) * client->ctx_index);
+}
+
+/* Get valid workqueue item and return it back to offset */
+static int guc_get_workqueue_space(struct i915_guc_client *gc, u32 *offset)
+{
+	struct guc_process_desc *desc;
+	void *base;
+	u32 size = sizeof(struct guc_wq_item);
+	int ret = 0, timeout_counter = 200;
+	unsigned long flags;
+
+	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0));
+	desc = base + gc->proc_desc_offset;
+
+	while (timeout_counter-- > 0) {
+		spin_lock_irqsave(&gc->wq_lock, flags);
+
+		ret = wait_for_atomic(CIRC_SPACE(gc->wq_tail, desc->head,
+				gc->wq_size) >= size, 1);
+
+		if (!ret) {
+			*offset = gc->wq_tail;
+
+			/* advance the tail for next workqueue item */
+			gc->wq_tail += size;
+			gc->wq_tail &= gc->wq_size - 1;
+
+			/* this will break the loop */
+			timeout_counter = 0;
+		}
+
+		spin_unlock_irqrestore(&gc->wq_lock, flags);
+	};
+
+	kunmap_atomic(base);
+
+	return ret;
+}
+
+static int guc_add_workqueue_item(struct i915_guc_client *gc,
+				  struct intel_context *ctx,
+				  struct intel_engine_cs *ring)
+{
+	struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;
+	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
+	struct guc_wq_item *wqi;
+	void *base;
+	u32 tail, wq_len, wq_off = 0;
+	int ret;
+
+	ret = guc_get_workqueue_space(gc, &wq_off);
+	if (ret)
+		return ret;
+
+	/* For now workqueue item is 4 DWs; workqueue buffer is 2 pages. So we
+	 * should not have the case where structure wqi is across page, neither
+	 * wrapped to the beginning. This simplifies the implementation below.
+	 *
+	 * XXX: if not the case, we need save data to a temp wqi and copy it to
+	 * workqueue buffer dw by dw.
+	 */
+	WARN_ON(sizeof(struct guc_wq_item) != 16);
+	WARN_ON(wq_off & 3);
+
+	/* wq starts from the page after doorbell / process_desc */
+	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj,
+			(wq_off + GUC_DB_SIZE) >> PAGE_SHIFT));
+	wq_off &= PAGE_SIZE - 1;
+	wqi = (struct guc_wq_item *)((char *)base + wq_off);
+
+	/* len does not include the header */
+	wq_len = sizeof(struct guc_wq_item) / sizeof(u32) - 1;
+	wqi->header = WQ_TYPE_INORDER |
+			(wq_len << WQ_LEN_SHIFT) |
+			(ring->id << WQ_TARGET_SHIFT) |
+			WQ_NO_WCFLUSH_WAIT;
+
+	/* The GuC wants only the low-order word of the context descriptor */
+	wqi->context_desc = (u32)intel_lr_context_descriptor(ring, ctx_obj);
+
+	/* The GuC firmware wants the tail index in QWords, not bytes */
+	tail = ringbuf->tail >> 3;
+	wqi->ring_tail = tail << WQ_RING_TAIL_SHIFT;
+	wqi->fence_id = 0; /*XXX: what fence to be here */
+
+	kunmap_atomic(base);
+
+	return 0;
+}
+
+static void update_context_image(struct intel_context *ctx,
+				 struct intel_engine_cs *ring)
+{
+	struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;
+	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
+
+	intel_lr_context_update(ctx_obj, ringbuf->obj, ~0u);
+}
+
+/**
+ * i915_guc_submit() - Submit commands through GuC
+ * @client:	the guc client where commands will go through
+ * @ctx:	LRC where commands come from
+ * @ring:	HW engine that will excute the commands
+ *
+ * Return:	0 if succeed
+ */
+int i915_guc_submit(struct i915_guc_client *client,
+		    struct intel_context *ctx,
+		    struct intel_engine_cs *ring)
+{
+	int q_ret, b_ret;
+	unsigned long flags;
+
+	/* Need this because of the deferred pin ctx and ring */
+	/* Shall we move this right after ring is pinned? */
+	update_context_image(ctx, ring);
+
+	q_ret = guc_add_workqueue_item(client, ctx, ring);
+	if (q_ret == 0)
+		b_ret = guc_ring_doorbell(client);
+
+	spin_lock_irqsave(&client->wq_lock, flags);
+	client->submissions += 1;
+	if (q_ret) {
+		client->q_fail += 1;
+		client->retcode = q_ret;
+	} else if (b_ret) {
+		client->b_fail += 1;
+		client->retcode = q_ret = b_ret;
+	} else {
+		client->retcode = 0;
+	}
+	spin_unlock_irqrestore(&client->wq_lock, flags);
+
+	return q_ret;
+}
+
+/*
+ * Everything below here is concerned with setup & teardown, and is
+ * therefore not part of the somewhat time-critical batch-submission
+ * path of i915_guc_submit() above.
+ */
+
+/**
  * gem_allocate_guc_obj() - Allocate gem object for GuC usage
  * @dev:	drm device
  * @size:	size of object
@@ -75,6 +593,121 @@ static void gem_release_guc_obj(struct drm_i915_gem_object *obj)
 	drm_gem_object_unreference(&obj->base);
 }
 
+static void guc_client_free(struct drm_device *dev,
+			    struct i915_guc_client *client)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc *guc = &dev_priv->guc;
+
+	if (!client)
+		return;
+
+	if (client->doorbell_id != GUC_INVALID_DOORBELL_ID) {
+		/*
+		 * First disable the doorbell, then tell the GuC we've
+		 * finished with it, finally deallocate it in our bitmap
+		 */
+		guc_disable_doorbell(guc, client);
+		host2guc_release_doorbell(guc, client);
+		release_doorbell(guc, client->doorbell_id);
+	}
+
+	/*
+	 * XXX: wait for any outstanding submissions before freeing memory.
+	 * Be sure to drop any locks
+	 */
+
+	gem_release_guc_obj(client->client_obj);
+
+	if (client->ctx_index != GUC_INVALID_CTX_ID) {
+		guc_fini_ctx_desc(guc, client);
+		ida_simple_remove(&guc->ctx_ids, client->ctx_index);
+	}
+
+	kfree(client);
+}
+
+/**
+ * guc_client_alloc() - Allocate an i915_guc_client
+ * @dev:	drm device
+ * @priority:	four levels priority _CRITICAL, _HIGH, _NORMAL and _LOW
+ * 		The kernel client to replace ExecList submission is created with
+ * 		NORMAL priority. Priority of a client for scheduler can be HIGH,
+ * 		while a preemption context can use CRITICAL.
+ *
+ * Return:	An i915_guc_client object if success.
+ */
+static struct i915_guc_client *guc_client_alloc(struct drm_device *dev,
+						uint32_t priority)
+{
+	struct i915_guc_client *client;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc *guc = &dev_priv->guc;
+	struct drm_i915_gem_object *obj;
+
+	client = kzalloc(sizeof(*client), GFP_KERNEL);
+	if (!client)
+		return NULL;
+
+	client->doorbell_id = GUC_INVALID_DOORBELL_ID;
+	client->priority = priority;
+
+	client->ctx_index = (uint32_t)ida_simple_get(&guc->ctx_ids, 0,
+			GUC_MAX_GPU_CONTEXTS, GFP_KERNEL);
+	if (client->ctx_index >= GUC_MAX_GPU_CONTEXTS) {
+		client->ctx_index = GUC_INVALID_CTX_ID;
+		goto err;
+	}
+
+	/* The first page is doorbell/proc_desc. Two followed pages are wq. */
+	obj = gem_allocate_guc_obj(dev, GUC_DB_SIZE + GUC_WQ_SIZE);
+	if (!obj)
+		goto err;
+
+	client->client_obj = obj;
+	client->wq_offset = GUC_DB_SIZE;
+	client->wq_size = GUC_WQ_SIZE;
+	spin_lock_init(&client->wq_lock);
+
+	client->doorbell_offset = select_doorbell_cacheline(guc);
+
+	/*
+	 * Since the doorbell only requires a single cacheline, we can save
+	 * space by putting the application process descriptor in the same
+	 * page. Use the half of the page that doesn't include the doorbell.
+	 */
+	if (client->doorbell_offset >= (GUC_DB_SIZE / 2))
+		client->proc_desc_offset = 0;
+	else
+		client->proc_desc_offset = (GUC_DB_SIZE / 2);
+
+	client->doorbell_id = assign_doorbell(guc, client->priority);
+	if (client->doorbell_id == GUC_INVALID_DOORBELL_ID)
+		/* XXX: evict a doorbell instead */
+		goto err;
+
+	guc_init_proc_desc(guc, client);
+	guc_init_ctx_desc(guc, client);
+	guc_init_doorbell(guc, client);
+
+	/* Invalidate GuC TLB to let GuC take the latest updates to GTT. */
+	I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
+
+	/* XXX: Any cache flushes needed? General domain mgmt calls? */
+
+	if (host2guc_allocate_doorbell(guc, client))
+		goto err;
+
+	DRM_DEBUG_DRIVER("new priority %u client %p: ctx_index %u db_id %u\n",
+		priority, client, client->ctx_index, client->doorbell_id);
+
+	return client;
+
+err:
+	guc_client_free(dev, client);
+	return NULL;
+}
+
 static void guc_create_log(struct intel_guc *guc)
 {
 	struct drm_i915_private *dev_priv = guc_to_i915(guc);
@@ -138,6 +771,8 @@ int i915_guc_submission_init(struct drm_device *dev)
 	if (!guc->ctx_pool_obj)
 		return -ENOMEM;
 
+	spin_lock_init(&dev_priv->guc.host2guc_lock);
+
 	ida_init(&guc->ctx_ids);
 
 	guc_create_log(guc);
@@ -145,6 +780,32 @@ int i915_guc_submission_init(struct drm_device *dev)
 	return 0;
 }
 
+int i915_guc_submission_enable(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc *guc = &dev_priv->guc;
+	struct i915_guc_client *client;
+
+	/* client for execbuf submission */
+	client = guc_client_alloc(dev, GUC_CTX_PRIORITY_NORMAL);
+	if (!client) {
+		DRM_ERROR("Failed to create execbuf guc_client\n");
+		return -ENOMEM;
+	}
+
+	guc->execbuf_client = client;
+	return 0;
+}
+
+void i915_guc_submission_disable(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc *guc = &dev_priv->guc;
+
+	guc_client_free(dev, guc->execbuf_client);
+	guc->execbuf_client = NULL;
+}
+
 void i915_guc_submission_fini(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
index 3de9096..5bed529 100644
--- a/drivers/gpu/drm/i915/intel_guc.h
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -28,6 +28,30 @@
 #include "intel_guc_fwif.h"
 #include "i915_guc_reg.h"
 
+struct i915_guc_client {
+	struct drm_i915_gem_object *client_obj;
+	uint32_t priority;
+	uint32_t ctx_index;
+
+	uint32_t proc_desc_offset;
+	uint32_t doorbell_offset;
+	uint32_t cookie;
+	uint16_t doorbell_id;
+	uint16_t padding;		/* Maintain alignment		*/
+
+	uint32_t wq_offset;
+	uint32_t wq_size;
+
+	spinlock_t wq_lock;		/* Protects all data below	*/
+	uint32_t wq_tail;
+
+	/* GuC submission statistics & status */
+	uint64_t submissions;
+	uint32_t q_fail;
+	uint32_t b_fail;
+	int retcode;
+};
+
 struct intel_guc {
 	/* Generic uC firmware management */
 	struct intel_uc_fw guc_fw;
@@ -43,6 +67,20 @@ struct intel_guc {
 
 	struct drm_i915_gem_object *ctx_pool_obj;
 	struct ida ctx_ids;
+
+	struct i915_guc_client *execbuf_client;
+
+	spinlock_t host2guc_lock;	/* Protects all data below	*/
+
+	DECLARE_BITMAP(doorbell_bitmap, GUC_MAX_DOORBELLS);
+	int db_cacheline;
+
+	/* Action status & statistics */
+	uint64_t action_count;		/* Total commands issued	*/
+	uint32_t action_cmd;		/* Last command word		*/
+	uint32_t action_status;		/* Last return status		*/
+	uint32_t action_fail;		/* Total number of failures	*/
+	int32_t action_err;		/* Last error code		*/
 };
 
 /* intel_guc_loader.c */
@@ -52,6 +90,11 @@ extern void intel_guc_ucode_fini(struct drm_device *dev);
 
 /* i915_guc_submission.c */
 int i915_guc_submission_init(struct drm_device *dev);
+int i915_guc_submission_enable(struct drm_device *dev);
+int i915_guc_submit(struct i915_guc_client *client,
+		    struct intel_context *ctx,
+		    struct intel_engine_cs *ring);
+void i915_guc_submission_disable(struct drm_device *dev);
 void i915_guc_submission_fini(struct drm_device *dev);
 
 #endif
diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
index deeaa20..eb76c34 100644
--- a/drivers/gpu/drm/i915/intel_guc_loader.c
+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
@@ -417,6 +417,8 @@ int intel_guc_ucode_load(struct drm_device *dev)
 		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
 		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
 
+	i915_guc_submission_disable(dev);
+
 	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
 		return 0;
 
@@ -444,12 +446,20 @@ int intel_guc_ucode_load(struct drm_device *dev)
 		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
 		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
 
+	if (i915.enable_guc_submission) {
+		err = i915_guc_submission_enable(dev);
+		if (err)
+			goto fail;
+	}
+
 	return 0;
 
 fail:
 	if (guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_PENDING)
 		guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_FAIL;
 
+	i915_guc_submission_disable(dev);
+
 	DRM_ERROR("Failed to initialize GuC, error %d\n", err);
 
 	return err;
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Dave Gordon]

Turn on interrupt steering to route necessary interrupts to GuC.

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
 drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
 2 files changed, 60 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index eefb847..9f32c6c 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
 #define GFX_MODE_GEN7	0x0229c
 #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
 #define   GFX_RUN_LIST_ENABLE		(1<<15)
+#define   GFX_INTERRUPT_STEERING	(1<<14)
 #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
 #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
 #define   GFX_REPLAY_MODE		(1<<11)
 #define   GFX_PSMI_GRANULARITY		(1<<10)
 #define   GFX_PPGTT_ENABLE		(1<<9)
 
+#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
+#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
+#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
+#define   GFX_FORWARD_VBLANK_COND	(2<<5)
+
 #define VLV_DISPLAY_BASE 0x180000
 #define VLV_MIPI_BASE VLV_DISPLAY_BASE
 
@@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
 #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
 #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
 
-#define GEN8_BCS_IRQ_SHIFT 16
 #define GEN8_RCS_IRQ_SHIFT 0
-#define GEN8_VCS2_IRQ_SHIFT 16
+#define GEN8_BCS_IRQ_SHIFT 16
 #define GEN8_VCS1_IRQ_SHIFT 0
+#define GEN8_VCS2_IRQ_SHIFT 16
 #define GEN8_VECS_IRQ_SHIFT 0
+#define GEN8_WD_IRQ_SHIFT 16
 
 #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
 #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
index eb76c34..ef5f3d5 100644
--- a/drivers/gpu/drm/i915/intel_guc_loader.c
+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
@@ -62,6 +62,53 @@
 #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
 MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
 
+static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
+{
+	struct intel_engine_cs *ring;
+	int i, irqs;
+
+	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
+	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
+	for_each_ring(ring, dev_priv, i)
+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
+
+	/* tell DE to send nothing to GuC */
+	I915_WRITE(DE_GUCRMR, ~0);
+
+	/* route all GT interrupts to the host */
+	I915_WRITE(GUC_BCS_RCS_IER, 0);
+	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
+	I915_WRITE(GUC_WD_VECS_IER, 0);
+}
+
+static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
+{
+	struct intel_engine_cs *ring;
+	int i, irqs;
+
+	/* tell all command streamers to forward interrupts and vblank to GuC */
+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
+	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
+	for_each_ring(ring, dev_priv, i)
+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
+
+	/* tell DE to send (all) flip_done to GuC */
+	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
+	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
+	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
+	/* Unmasked bits will cause GuC response message to be sent */
+	I915_WRITE(DE_GUCRMR, ~irqs);
+
+	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
+	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
+	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
+	/* These three registers have the same bit definitions */
+	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
+	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
+	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
+}
+
 static u32 get_gttype(struct drm_i915_private *dev_priv)
 {
 	/* XXX: GT type based on PCI device ID? field seems unused by fw */
@@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
 		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
 		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
 
+	direct_interrupts_to_host(dev_priv);
 	i915_guc_submission_disable(dev);
 
 	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
@@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
 		err = i915_guc_submission_enable(dev);
 		if (err)
 			goto fail;
+		direct_interrupts_to_guc(dev_priv);
 	}
 
 	return 0;
@@ -458,6 +507,7 @@ fail:
 	if (guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_PENDING)
 		guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_FAIL;
 
+	direct_interrupts_to_host(dev_priv);
 	i915_guc_submission_disable(dev);
 
 	DRM_ERROR("Failed to initialize GuC, error %d\n", err);
@@ -473,6 +523,7 @@ void intel_guc_ucode_fini(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
 
+	direct_interrupts_to_host(dev_priv);
 	i915_guc_submission_fini(dev);
 
 	intel_uc_fw_fini(guc_fw);
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 12/15] drm/i915: Integrate GuC-based command submission

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

GuC-based submission is mostly the same as execlist mode, up to
intel_logical_ring_advance_and_submit(), where the context being
dispatched would be added to the execlist queue; at this point
we submit the context to the GuC backend instead.

There are, however, a few other changes also required, notably:
1.  Contexts must be pinned at GGTT addresses accessible by the GuC
    i.e. NOT in the range [0..WOPCM_SIZE), so we have to add the
    PIN_OFFSET_BIAS flag to the relevant GGTT-pinning calls.

2.  The GuC's TLB must be invalidated after a context is pinned at
    a new GGTT address.

3.  GuC firmware uses the one page before Ring Context as shared data.
    Therefore, whenever driver wants to get base address of LRC, we
    will offset one page for it. LRC_PPHWSP_PN is defined as the page
    number of LRCA.

4.  In the work queue used to pass requests to the GuC, the GuC
    firmware requires the ring-tail-offset to be represented as an
    11-bit value, expressed in QWords. Therefore, the ringbuffer
    size must be reduced to the representable range (4 pages).

v2:
    Defer adding #defines until needed [Chris Wilson]
    Rationalise type declarations [Chris Wilson]

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c        |  2 +-
 drivers/gpu/drm/i915/i915_guc_submission.c | 48 +++++++++++++++++++++++++++---
 drivers/gpu/drm/i915/intel_guc.h           |  1 +
 drivers/gpu/drm/i915/intel_lrc.c           | 48 ++++++++++++++++++++++--------
 drivers/gpu/drm/i915/intel_lrc.h           |  6 ++++
 5 files changed, 87 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index ad909f5..9d35550 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1982,7 +1982,7 @@ static void i915_dump_lrc_obj(struct seq_file *m,
 		return;
 	}
 
-	page = i915_gem_object_get_page(ctx_obj, 1);
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 	if (!WARN_ON(page == NULL)) {
 		reg_state = kmap_atomic(page);
 
diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
index 3c28c8b..099f053 100644
--- a/drivers/gpu/drm/i915/i915_guc_submission.c
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -348,16 +348,52 @@ static void guc_init_ctx_desc(struct intel_guc *guc,
 {
 	struct guc_context_desc desc;
 	struct sg_table *sg;
+	int i;
 
 	memset(&desc, 0, sizeof(desc));
 
 	desc.attribute = GUC_CTX_DESC_ATTR_ACTIVE | GUC_CTX_DESC_ATTR_KERNEL;
 	desc.context_id = client->ctx_index;
 	desc.priority = client->priority;
-	desc.engines_used = (1 << RCS) | (1 << VCS) | (1 << BCS) |
-			    (1 << VECS) | (1 << VCS2); /* all engines */
 	desc.db_id = client->doorbell_id;
 
+	for (i = 0; i < I915_NUM_RINGS; i++) {
+		struct guc_execlist_context *lrc = &desc.lrc[i];
+		struct intel_engine_cs *ring;
+		struct drm_i915_gem_object *obj;
+
+		/* TODO: We have a design issue to be solved here. Only when we
+		 * receive the first batch, we know which engine is used by the
+		 * user. But here GuC expects the lrc and ring to be pinned. It
+		 * is not an issue for default context, which is the only one
+		 * for now who owns a GuC client. But for future owner of GuC
+		 * client, need to make sure lrc is pinned prior to enter here.
+		 */
+		obj = client->owner->engine[i].state;
+		if (!obj)
+			break;
+
+		ring = client->owner->engine[i].ringbuf->ring;
+
+		lrc->context_desc = (u32)intel_lr_context_descriptor(ring, obj);
+		/* The state page is after PPHWSP */
+		lrc->ring_lcra = i915_gem_obj_ggtt_offset(obj) +
+				LRC_STATE_PN * PAGE_SIZE;
+		lrc->context_id = (client->ctx_index << GUC_ELC_CTXID_OFFSET) |
+				(ring->id << GUC_ELC_ENGINE_OFFSET);
+
+		obj = client->owner->engine[i].ringbuf->obj;
+
+		lrc->ring_begin = i915_gem_obj_ggtt_offset(obj);
+		lrc->ring_end = lrc->ring_begin + obj->base.size - 1;
+		lrc->ring_next_free_location = lrc->ring_begin;
+		lrc->ring_current_tail_pointer_value = 0;
+
+		desc.engines_used |= (1 << ring->id);
+	}
+
+	WARN_ON(desc.engines_used == 0);
+
 	/*
 	 * The CPU address is only needed at certain points, so kmap_atomic on
 	 * demand instead of storing it in the ctx descriptor.
@@ -634,11 +670,13 @@ static void guc_client_free(struct drm_device *dev,
  * 		The kernel client to replace ExecList submission is created with
  * 		NORMAL priority. Priority of a client for scheduler can be HIGH,
  * 		while a preemption context can use CRITICAL.
+ * @ctx		the context to own the client (we use the default render context)
  *
  * Return:	An i915_guc_client object if success.
  */
 static struct i915_guc_client *guc_client_alloc(struct drm_device *dev,
-						uint32_t priority)
+						uint32_t priority,
+						struct intel_context *ctx)
 {
 	struct i915_guc_client *client;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -651,6 +689,7 @@ static struct i915_guc_client *guc_client_alloc(struct drm_device *dev,
 
 	client->doorbell_id = GUC_INVALID_DOORBELL_ID;
 	client->priority = priority;
+	client->owner = ctx;
 
 	client->ctx_index = (uint32_t)ida_simple_get(&guc->ctx_ids, 0,
 			GUC_MAX_GPU_CONTEXTS, GFP_KERNEL);
@@ -784,10 +823,11 @@ int i915_guc_submission_enable(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_guc *guc = &dev_priv->guc;
+	struct intel_context *ctx = dev_priv->ring[RCS].default_context;
 	struct i915_guc_client *client;
 
 	/* client for execbuf submission */
-	client = guc_client_alloc(dev, GUC_CTX_PRIORITY_NORMAL);
+	client = guc_client_alloc(dev, GUC_CTX_PRIORITY_NORMAL, ctx);
 	if (!client) {
 		DRM_ERROR("Failed to create execbuf guc_client\n");
 		return -ENOMEM;
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
index 5bed529..7da61df 100644
--- a/drivers/gpu/drm/i915/intel_guc.h
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -30,6 +30,7 @@
 
 struct i915_guc_client {
 	struct drm_i915_gem_object *client_obj;
+	struct intel_context *owner;
 	uint32_t priority;
 	uint32_t ctx_index;
 
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 164c63a..9df89ac 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -255,7 +255,8 @@ int intel_sanitize_enable_execlists(struct drm_device *dev, int enable_execlists
  */
 u32 intel_execlists_ctx_id(struct drm_i915_gem_object *ctx_obj)
 {
-	u32 lrca = i915_gem_obj_ggtt_offset(ctx_obj);
+	u32 lrca = i915_gem_obj_ggtt_offset(ctx_obj) +
+			LRC_PPHWSP_PN * PAGE_SIZE;
 
 	/* LRCA is required to be 4K aligned so the more significant 20 bits
 	 * are globally unique */
@@ -267,7 +268,8 @@ uint64_t intel_lr_context_descriptor(struct intel_engine_cs *ring,
 {
 	struct drm_device *dev = ring->dev;
 	uint64_t desc;
-	uint64_t lrca = i915_gem_obj_ggtt_offset(ctx_obj);
+	uint64_t lrca = i915_gem_obj_ggtt_offset(ctx_obj) +
+			LRC_PPHWSP_PN * PAGE_SIZE;
 
 	WARN_ON(lrca & 0xFFFFFFFF00000FFFULL);
 
@@ -337,7 +339,7 @@ void intel_lr_context_update(struct drm_i915_gem_object *ctx_obj,
 	struct page *page;
 	uint32_t *reg_state;
 
-	page = i915_gem_object_get_page(ctx_obj, 1);
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 	reg_state = kmap_atomic(page);
 
 	reg_state[CTX_RING_BUFFER_START+1] = i915_gem_obj_ggtt_offset(ring_obj);
@@ -696,13 +698,17 @@ static void
 intel_logical_ring_advance_and_submit(struct drm_i915_gem_request *request)
 {
 	struct intel_engine_cs *ring = request->ring;
+	struct drm_i915_private *dev_priv = request->i915;
 
 	intel_logical_ring_advance(request->ringbuf);
 
 	if (intel_ring_stopped(ring))
 		return;
 
-	execlists_context_queue(request);
+	if (dev_priv->guc.execbuf_client)
+		i915_guc_submit(dev_priv->guc.execbuf_client, request->ctx, ring);
+	else
+		execlists_context_queue(request);
 }
 
 static void __wrap_ring_buffer(struct intel_ringbuffer *ringbuf)
@@ -996,18 +1002,23 @@ static int intel_lr_context_pin(struct intel_engine_cs *ring,
 {
 	struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;
 	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
+	struct drm_i915_private *dev_priv = ring->dev->dev_private;
 	int ret = 0;
 
 	WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex));
 	if (ctx->engine[ring->id].pin_count++ == 0) {
-		ret = i915_gem_obj_ggtt_pin(ctx_obj,
-				GEN8_LR_CONTEXT_ALIGN, 0);
+		ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN,
+				PIN_OFFSET_BIAS | GUC_WOPCM_SIZE_VALUE);
 		if (ret)
 			goto reset_pin_count;
 
 		ret = intel_pin_and_map_ringbuffer_obj(ring->dev, ringbuf);
 		if (ret)
 			goto unpin_ctx_obj;
+
+		/* Invalidate GuC TLB. */
+		if (i915.enable_guc_submission)
+			I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
 	}
 
 	return ret;
@@ -1640,8 +1651,13 @@ out:
 
 static int gen8_init_rcs_context(struct drm_i915_gem_request *req)
 {
+	struct drm_i915_private *dev_priv = req->i915;
 	int ret;
 
+	/* Invalidate GuC TLB. */
+	if (i915.enable_guc_submission)
+		I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
+
 	ret = intel_logical_ring_workarounds_emit(req);
 	if (ret)
 		return ret;
@@ -1998,7 +2014,7 @@ populate_lr_context(struct intel_context *ctx, struct drm_i915_gem_object *ctx_o
 
 	/* The second page of the context object contains some fields which must
 	 * be set up prior to the first execution. */
-	page = i915_gem_object_get_page(ctx_obj, 1);
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 	reg_state = kmap_atomic(page);
 
 	/* A context is actually a big batch buffer with several MI_LOAD_REGISTER_IMM
@@ -2156,12 +2172,14 @@ static void lrc_setup_hardware_status_page(struct intel_engine_cs *ring,
 		struct drm_i915_gem_object *default_ctx_obj)
 {
 	struct drm_i915_private *dev_priv = ring->dev->dev_private;
+	struct page *page;
 
 	/* The status page is offset 0 from the default context object
 	 * in LRC mode. */
-	ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(default_ctx_obj);
-	ring->status_page.page_addr =
-			kmap(sg_page(default_ctx_obj->pages->sgl));
+	ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(default_ctx_obj)
+			+ LRC_PPHWSP_PN * PAGE_SIZE;
+	page = i915_gem_object_get_page(default_ctx_obj, LRC_PPHWSP_PN);
+	ring->status_page.page_addr = kmap(page);
 	ring->status_page.obj = default_ctx_obj;
 
 	I915_WRITE(RING_HWS_PGA(ring->mmio_base),
@@ -2197,6 +2215,9 @@ int intel_lr_context_deferred_create(struct intel_context *ctx,
 
 	context_size = round_up(get_lr_context_size(ring), 4096);
 
+	/* One extra page as the sharing data between driver and GuC */
+	context_size += PAGE_SIZE * LRC_PPHWSP_PN;
+
 	ctx_obj = i915_gem_alloc_object(dev, context_size);
 	if (!ctx_obj) {
 		DRM_DEBUG_DRIVER("Alloc LRC backing obj failed.\n");
@@ -2204,7 +2225,8 @@ int intel_lr_context_deferred_create(struct intel_context *ctx,
 	}
 
 	if (is_global_default_ctx) {
-		ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN, 0);
+		ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN,
+				PIN_OFFSET_BIAS | GUC_WOPCM_SIZE_VALUE);
 		if (ret) {
 			DRM_DEBUG_DRIVER("Pin LRC backing obj failed: %d\n",
 					ret);
@@ -2223,7 +2245,7 @@ int intel_lr_context_deferred_create(struct intel_context *ctx,
 
 	ringbuf->ring = ring;
 
-	ringbuf->size = 32 * PAGE_SIZE;
+	ringbuf->size = 4 * PAGE_SIZE;
 	ringbuf->effective_size = ringbuf->size;
 	ringbuf->head = 0;
 	ringbuf->tail = 0;
@@ -2323,7 +2345,7 @@ void intel_lr_context_reset(struct drm_device *dev,
 			WARN(1, "Failed get_pages for context obj\n");
 			continue;
 		}
-		page = i915_gem_object_get_page(ctx_obj, 1);
+		page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 		reg_state = kmap_atomic(page);
 
 		reg_state[CTX_RING_HEAD+1] = 0;
diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h
index b3659a1..467157c 100644
--- a/drivers/gpu/drm/i915/intel_lrc.h
+++ b/drivers/gpu/drm/i915/intel_lrc.h
@@ -66,6 +66,12 @@ static inline void intel_logical_ring_emit(struct intel_ringbuffer *ringbuf,
 }
 
 /* Logical Ring Contexts */
+
+/* One extra page is added before LRC for GuC as shared data */
+#define LRC_GUCSHR_PN	(0)
+#define LRC_PPHWSP_PN	(LRC_GUCSHR_PN + 1)
+#define LRC_STATE_PN	(LRC_PPHWSP_PN + 1)
+
 void intel_lr_context_free(struct intel_context *ctx);
 int intel_lr_context_deferred_create(struct intel_context *ctx,
 				     struct intel_engine_cs *ring);
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 13/15] drm/i915: Debugfs interface for GuC submission statistics

[Dave Gordon]

This provides a means of reading status and counts relating
to GuC actions and submissions.

v2:
    Remove surplus blank line in output [Chris Wilson]

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Alex Dai <yu.dai@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c | 40 +++++++++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 9d35550..9fba9c9 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -2403,6 +2403,45 @@ static int i915_guc_load_status_info(struct seq_file *m, void *data)
 	return 0;
 }
 
+static int i915_guc_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_guc guc;
+	struct i915_guc_client client = { .client_obj = 0 };
+
+	if (!HAS_GUC_SCHED(dev_priv->dev))
+		return 0;
+
+	/* Take a local copy of the GuC data, so we can dump it at leisure */
+	spin_lock(&dev_priv->guc.host2guc_lock);
+	guc = dev_priv->guc;
+	if (guc.execbuf_client) {
+		spin_lock(&guc.execbuf_client->wq_lock);
+		client = *guc.execbuf_client;
+		spin_unlock(&guc.execbuf_client->wq_lock);
+	}
+	spin_unlock(&dev_priv->guc.host2guc_lock);
+
+	seq_printf(m, "GuC total action count: %llu\n", guc.action_count);
+	seq_printf(m, "GuC last action command: 0x%x\n", guc.action_cmd);
+	seq_printf(m, "GuC last action status: 0x%x\n", guc.action_status);
+
+	seq_printf(m, "GuC action failure count: %u\n", guc.action_fail);
+	seq_printf(m, "GuC last action error code: %d\n", guc.action_err);
+
+	seq_printf(m, "\nGuC execbuf client @ %p:\n", guc.execbuf_client);
+	seq_printf(m, "\tTotal submissions: %llu\n", client.submissions);
+	seq_printf(m, "\tFailed to queue: %u\n", client.q_fail);
+	seq_printf(m, "\tFailed doorbell: %u\n", client.b_fail);
+	seq_printf(m, "\tLast submission result: %d\n", client.retcode);
+
+	/* Add more as required ... */
+
+	return 0;
+}
+
 static int i915_guc_log_dump(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = m->private;
@@ -5145,6 +5184,7 @@ static const struct drm_info_list i915_debugfs_list[] = {
 	{"i915_gem_hws_bsd", i915_hws_info, 0, (void *)VCS},
 	{"i915_gem_hws_vebox", i915_hws_info, 0, (void *)VECS},
 	{"i915_gem_batch_pool", i915_gem_batch_pool_info, 0},
+	{"i915_guc_info", i915_guc_info, 0},
 	{"i915_guc_load_status", i915_guc_load_status_info, 0},
 	{"i915_guc_log_dump", i915_guc_log_dump, 0},
 	{"i915_frequency_info", i915_frequency_info, 0},
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 14/15] Documentation/drm: kerneldoc for GuC

[Dave Gordon]

From: Alex Dai <yu.dai@intel.com>

Add overview design of GuC, plus some key points related to
the implementation.

Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 Documentation/DocBook/drm.tmpl | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl
index c0312cb..88b53ee 100644
--- a/Documentation/DocBook/drm.tmpl
+++ b/Documentation/DocBook/drm.tmpl
@@ -4218,6 +4218,25 @@ int num_ioctls;</synopsis>
       </sect2>
     </sect1>
     <sect1>
+      <title>GuC-based Command Submission</title>
+      <sect2>
+        <title>Microcontroller (uC) firmware loading support</title>
+!Pdrivers/gpu/drm/i915/intel_uc_loader.c Microcontroller (uC) firmware loading support
+!Idrivers/gpu/drm/i915/intel_uc_loader.c
+      </sect2>
+      <sect2>
+        <title>GuC</title>
+!Pdrivers/gpu/drm/i915/intel_guc_loader.c GuC-specific firmware loader
+!Idrivers/gpu/drm/i915/intel_guc_loader.c
+      </sect2>
+      <sect2>
+        <title>GuC Client</title>
+!Pdrivers/gpu/drm/i915/intel_guc_submission.c GuC-based command submissison
+!Idrivers/gpu/drm/i915/intel_guc_submission.c
+      </sect2>
+    </sect1>
+
+    <sect1>
       <title> Tracing </title>
       <para>
     This sections covers all things related to the tracepoints implemented in
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 15/15] drm/i915: Enable GuC submission, where supported

[Dave Gordon]

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_params.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c
index a8f48dd..479c97b 100644
--- a/drivers/gpu/drm/i915/i915_params.c
+++ b/drivers/gpu/drm/i915/i915_params.c
@@ -53,7 +53,7 @@ struct i915_params i915 __read_mostly = {
 	.verbose_state_checks = 1,
 	.nuclear_pageflip = 0,
 	.edp_vswing = 0,
-	.enable_guc_submission = false,
+	.enable_guc_submission = true,
 	.guc_log_level = -1,
 };
 
@@ -190,7 +190,7 @@ MODULE_PARM_DESC(edp_vswing,
 		 "2=default swing(400mV))");
 
 module_param_named(enable_guc_submission, i915.enable_guc_submission, bool, 0400);
-MODULE_PARM_DESC(enable_guc_submission, "Enable GuC submission (default:false)");
+MODULE_PARM_DESC(enable_guc_submission, "Enable GuC submission (default:true)");
 
 module_param_named(guc_log_level, i915.guc_log_level, int, 0400);
 MODULE_PARM_DESC(guc_log_level,
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 02/15] drm/i915: Embedded microcontroller (uC) firmware loading support

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:24PM +0100, Dave Gordon wrote:
> Current devices may contain one or more programmable microcontrollers
> that need to have a firmware image (aka "binary blob") loaded from an
> external medium and transferred to the device's memory.
> 
> This file provides common support functions for doing this; they can
> then be used by each uC-specific loader, thus reducing code duplication
> and testing effort.
> 
> v2:
>     Local functions should pass dev_priv rather than dev [Chris Wilson]
>     Various other improvements per Chris Wilson's review comments
> 
> v3:
>     Defeatured version without async prefetch [Daniel Vetter]
> 
> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> Signed-off-by: Alex Dai <yu.dai@intel.com>
> ---
>  drivers/gpu/drm/i915/Makefile          |   3 +
>  drivers/gpu/drm/i915/intel_uc_loader.c | 310 +++++++++++++++++++++++++++++++++
>  drivers/gpu/drm/i915/intel_uc_loader.h |  92 ++++++++++
>  3 files changed, 405 insertions(+)
>  create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c
>  create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h
> 
> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> index de21965..f1f80fc 100644
> --- a/drivers/gpu/drm/i915/Makefile
> +++ b/drivers/gpu/drm/i915/Makefile
> @@ -39,6 +39,9 @@ i915-y += i915_cmd_parser.o \
>  	  intel_ringbuffer.o \
>  	  intel_uncore.o
>  
> +# generic ancilliary microcontroller support
> +i915-y += intel_uc_loader.o
> +
>  # autogenerated null render state
>  i915-y += intel_renderstate_gen6.o \
>  	  intel_renderstate_gen7.o \
> diff --git a/drivers/gpu/drm/i915/intel_uc_loader.c b/drivers/gpu/drm/i915/intel_uc_loader.c
> new file mode 100644
> index 0000000..a8fc1dd
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/intel_uc_loader.c
> @@ -0,0 +1,310 @@
> +/*
> + * Copyright © 2014 Intel Corporation
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> + * IN THE SOFTWARE.
> + *
> + * Author:
> + *	Dave Gordon <david.s.gordon@intel.com>
> + */
> +#include <linux/firmware.h>
> +#include "i915_drv.h"
> +#include "intel_uc_loader.h"
> +
> +/**
> + * DOC: Common functions for embedded microcontroller (uC) firmware loading
> + *
> + * The functions in this file provide common support code for loading the
> + * firmware that may be required by an embedded microcontroller (uC).
> + *
> + * The function intel_uc_fw_init() should be called first; it requires no
> + * locking, and can be called even before GEM has been initialised. It just
> + * initialises the tracking data and stores its parameters for the subsequent
> + * call to intel_uc_fw_fetch().
> + *
> + * At some convenient point after GEM initialisation, the driver should call
> + * intel_uc_fw_fetch(). The first time, this will use the Linux kernel's
> + * request_firmware() call to open and read the firmware image into memory.
> + * (On subsequent calls, this is skipped, as either the firmware has already
> + * been fetched into memory, or it is already known that no valid firmware
> + * image could be found).
> + *
> + * The callback() function passed to intel_uc_fw_fetch() can further check
> + * the firmware image before it is saved. This function can reject the image
> + * by returning a negative error code (e.g. -ENOEXEC), or accept it. In the
> + * latter case, it can return INTEL_UC_FW_GOOD (which is also the default if
> + * no callback() is supplied), and the common code here will save the whole
> + * of the firmware image in a (pageable, shmfs-backed) GEM object.
> + *
> + * If saving the whole image unmodified is not appropriate (for example, if
> + * only a small part of the image is needed later, or the data needs to be
> + * reorganised before saving), the callback() function can instead make its
> + * own arrangements for saving the required data in a GEM object or otherwise
> + * and then return INTEL_UC_FW_SAVED.
> + *
> + * (If such a callback does stash (some of) the image data in a GEM object,
> + * it can use the uc_fw_obj and uc_fw_size fields; the common framework will
> + * then handle deallocating the object on failure or finalisation. Otherwise
> + * any allocated resources will have to be managed by the uC-specific code).
> + *
> + * After a successful call to intel_uc_fw_fetch(), the uC-specific code can
> + * transfer the data in the GEM object (or its own alternative) to the uC's
> + * memory (in some uC-specific way, not handled here).
> + *
> + * During driver shutdown, or if driver load is aborted, intel_uc_fw_fini()
> + * should be called to release any remaining resources.
> + */
> +
> +/* User-friendly representation of an enum */
> +const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status)
> +{
> +	switch (status) {
> +	case INTEL_UC_FIRMWARE_FAIL:
> +		return "FAIL";
> +	case INTEL_UC_FIRMWARE_NONE:
> +		return "NONE";
> +	case INTEL_UC_FIRMWARE_PENDING:
> +		return "PENDING";
> +	case INTEL_UC_FIRMWARE_SUCCESS:
> +		return "SUCCESS";
> +	default:
> +		return "UNKNOWN!";
> +	}
> +};
> +
> +/*
> + * Called once per uC, late in driver initialisation, after GEM is set up.
> + * First, we fetch the firmware image using request_firmware(), then allow
> + * the optional callback() function to check it. If it's good, and callback()
> + * doesn't say it's already saved the image, we will save it here by copying
> + * the whole thing into a (pageable, shmfs-backed) GEM object,
> + */
> +static void
> +uc_fw_fetch(struct intel_uc_fw *uc_fw,
> +	    int callback(struct intel_uc_fw *))
> +{
> +	struct drm_device *dev = uc_fw->uc_dev;
> +	struct drm_i915_gem_object *obj;
> +	const struct firmware *fw;
> +	int cb_status = INTEL_UC_FW_GOOD;
> +
> +	DRM_DEBUG_DRIVER("before waiting: %s fw fetch status %s, fw %p\n",
> +		uc_fw->uc_name,
> +		intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status),
> +		uc_fw->uc_fw_blob);
> +
> +	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
> +	WARN_ON(uc_fw->uc_fw_fetch_status != INTEL_UC_FIRMWARE_PENDING);
> +
> +	if (request_firmware(&fw, uc_fw->uc_fw_path, &dev->pdev->dev))
> +		goto fail;
> +	if (!fw)
> +		goto fail;
> +
> +	DRM_DEBUG_DRIVER("fetch %s fw from %s succeeded, fw %p\n",
> +		uc_fw->uc_name, uc_fw->uc_fw_path, fw);
> +	uc_fw->uc_fw_blob = fw;
> +
> +	/* Callback to the optional uC-specific function, if supplied */
> +	if (callback)
> +		cb_status = callback(uc_fw);
> +	if (cb_status < 0)
> +		goto fail;
> +	switch (cb_status) {
> +	default:
> +		WARN(1, "Invalid status %d from fw checking callback %pf\n",
> +				cb_status, callback);
> +		goto fail;
> +
> +	case INTEL_UC_FW_SAVED:
> +		// Good, already saved, nothing to do
> +		break;
> +
> +	case INTEL_UC_FW_GOOD:
> +		// Good, save in GEM object
> +		obj = i915_gem_object_create_from_data(dev, fw->data, fw->size);
> +		if (!obj)
> +			goto fail;
> +
> +		uc_fw->uc_fw_obj = obj;
> +		uc_fw->uc_fw_size = fw->size;
> +	}
> +
> +	DRM_DEBUG_DRIVER("%s fw fetch status SUCCESS, cb %d, obj %p\n",
> +			uc_fw->uc_name, cb_status, uc_fw->uc_fw_obj);
> +
> +	release_firmware(fw);
> +	uc_fw->uc_fw_blob = NULL;
> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_SUCCESS;
> +	return;
> +
> +fail:
> +	DRM_DEBUG_DRIVER("%s fw fetch status FAIL; fw %p, obj %p\n",
> +		uc_fw->uc_name, fw, uc_fw->uc_fw_obj);
> +	DRM_ERROR("Failed to fetch %s firmware from %s\n",
> +		  uc_fw->uc_name, uc_fw->uc_fw_path);
> +
> +	obj = uc_fw->uc_fw_obj;
> +	if (obj)
> +		drm_gem_object_unreference(&obj->base);
> +	uc_fw->uc_fw_obj = NULL;
> +
> +	release_firmware(fw);		/* OK even if fw is NULL */
> +	uc_fw->uc_fw_blob = NULL;
> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_FAIL;
> +}

So looking at this the entire library now boils down to a glorified
version of

	if (request_firmware != 0) {
		DRM_DEBUG_GEM("guc firmware loading failed\n");
		return fail;
	}

	gem_obj_create_from_date(fw)
	relase_firmware

Can I have that inlined (or as a simple static helper function) into the
callsite please? Yes that unfortunately means I ask you to can the really
great kerneldoc you've done (I'd really like all patches to look this
great wrt docs), but I also think that the abstraction really doesn't pay
off. Firmware-loading using request_firmware is not something tricky
enough to justify added abstraction. Yes there was the
synchronization/completion stuff in there before, but as explained in the
previous discussion I don't expect us to be able to share that part at
all, and I'd like any kind of synchronization to be open-coded using
normal kernel primitives (like flush_work or similar) to make it really
stick out to reviewers.

Thanks, Daniel


> +
> +/**
> + * intel_uc_fw_fetch() - fetch the firmware image
> + * @uc_fw:	intel_uc_fw structure
> + * @callback:	optional callback function to validate and/or save the image
> + *
> + * If the fetch is not PENDING (i.e. on the second and subsequent calls), this
> + * function will just return an approriate errno, based on the previously-set
> + * status.
> + *
> + * On the first call only, it will try to retrieve the firmaware image using
> + * the parameters set earlier. If an image is found, the optional @callback()
> + * will be called to further validate it.
> + *
> + * When it is called, @uc_fw->uc_fw_blob refers to the fetched firmware image,
> + * and @uc_fw->uc_fw_obj is NULL. The @callback() function can return an error
> + * code (any negative value), in which case the image will be rejected.  The
> + * fetch status will be set to FAIL, and this function will return -EIO.
> + *
> + * Or, @callback() can return INTEL_UC_FW_GOOD. The image is considered good,
> + * and it will be saved in a GEM object as described above. In this case,
> + * @uc_fw->uc_fw_obj will be set to point to the GEM object, and the size of
> + * the image will be in @uc_fw->uc_fw_size. This is also the default if no
> + * @callback() is supplied.
> + *
> + * Finally, @callback() can return INTEL_UC_FW_SAVED. The image is considered
> + * good, but @callback() has already deal with saving the content, so this
> + * common code will not allocate and fill a GEM object. @callback() may use
> + * @uc_fw->uc_fw_obj to hold a reference to its own GEM object, if it has
> + * allocated one, and in this case the common code will deal with disposing
> + * of it on error or finalisation; otherwise it can make its own arragements.
> + *
> + * After this call, @uc_fw->uc_fw_fetch_status will show whether the firmware
> + * image was successfully fetched and saved.
> + *
> + * In all cases the firmware blob is released before this function returns.
> + *
> + * Return:	non-zero code on error
> + */
> +int
> +intel_uc_fw_fetch(struct intel_uc_fw *uc_fw,
> +		  int callback(struct intel_uc_fw *))
> +{
> +	WARN_ON(!mutex_is_locked(&uc_fw->uc_dev->struct_mutex));
> +
> +	if (uc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_PENDING) {
> +		/* We only come here once */
> +		uc_fw_fetch(uc_fw, callback);
> +		/* status must now be FAIL or SUCCESS */
> +	}
> +
> +	DRM_DEBUG_DRIVER("%s fw fetch status %s\n", uc_fw->uc_name,
> +		intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status));
> +
> +	switch (uc_fw->uc_fw_fetch_status) {
> +	case INTEL_UC_FIRMWARE_FAIL:
> +		/* something went wrong :( */
> +		return -EIO;
> +
> +	case INTEL_UC_FIRMWARE_NONE:
> +		/* no firmware, nothing to do (not an error) */
> +		return 0;
> +
> +	case INTEL_UC_FIRMWARE_PENDING:
> +	default:
> +		/* "can't happen" */
> +		WARN_ONCE(1, "%s fw %s invalid uc_fw_fetch_status %s [%d]\n",
> +			uc_fw->uc_name, uc_fw->uc_fw_path,
> +			intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status),
> +			uc_fw->uc_fw_fetch_status);
> +		return -ENXIO;
> +
> +	case INTEL_UC_FIRMWARE_SUCCESS:
> +		return 0;
> +	}
> +}
> +
> +/**
> + * intel_uc_fw_init() - define parameters for fetching firmware
> + * @dev:	drm device
> + * @uc_fw:	intel_uc_fw structure
> + * @name:	human-readable device name (e.g. "GuC") for messages
> + * @fw_path:	(trailing parts of) path to firmware (e.g. "i915/guc_fw.bin")
> + * 		@fw_path == NULL means "no firmware expected" (not an error),
> + * 		@fw_path == "" (empty string) means "firmware unknown" i.e.
> + * 		the uC requires firmware, but the driver doesn't know where
> + * 		to find the proper version. This will be logged as an error.
> + *
> + * This is called just once per uC, during driver loading. It is therefore
> + * automatically single-threaded and does not need to acquire any mutexes
> + * or spinlocks. OTOH, GEM is not yet fully initialised, so we can't do
> + * very much here.
> + *
> + * The main task here is to save the parameters for later. The actual fetching
> + * will happen when intel_uc_fw_init() is called, after GEM initialisation is
> + * complete.
> + */
> +void
> +intel_uc_fw_init(struct drm_device *dev, struct intel_uc_fw *uc_fw,
> +		 const char *name, const char *fw_path)
> +{
> +	uc_fw->uc_dev = dev;
> +	uc_fw->uc_name = name;
> +	uc_fw->uc_fw_path = fw_path;
> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_NONE;
> +	uc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_NONE;
> +
> +	if (fw_path == NULL)
> +		return;
> +
> +	if (*fw_path == '\0') {
> +		DRM_ERROR("No %s firmware known for this platform\n", name);
> +		uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_FAIL;
> +		return;
> +	}
> +
> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_PENDING;
> +	DRM_DEBUG_DRIVER("%s firmware pending, path %s\n",
> +		name, fw_path);
> +}
> +
> +/**
> + * intel_uc_fw_fini() - clean up all uC firmware-related data
> + * @uc_fw:	intel_uc_fw structure
> + */
> +void
> +intel_uc_fw_fini(struct intel_uc_fw *uc_fw)
> +{
> +	WARN_ON(!mutex_is_locked(&uc_fw->uc_dev->struct_mutex));
> +
> +	if (uc_fw->uc_fw_obj)
> +		drm_gem_object_unreference(&uc_fw->uc_fw_obj->base);
> +	uc_fw->uc_fw_obj = NULL;
> +
> +	release_firmware(uc_fw->uc_fw_blob);	/* OK even if NULL */
> +	uc_fw->uc_fw_blob = NULL;
> +
> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_NONE;
> +}
> diff --git a/drivers/gpu/drm/i915/intel_uc_loader.h b/drivers/gpu/drm/i915/intel_uc_loader.h
> new file mode 100644
> index 0000000..b710406
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/intel_uc_loader.h
> @@ -0,0 +1,92 @@
> +/*
> + * Copyright © 2014 Intel Corporation
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> + * IN THE SOFTWARE.
> + *
> + * Author:
> + *	Dave Gordon <david.s.gordon@intel.com>
> + */
> +#ifndef _INTEL_UC_LOADER_H
> +#define _INTEL_UC_LOADER_H
> +
> +/*
> + * Microcontroller (uC) firmware loading support
> + */
> +
> +/*
> + * These values are used to track the stages of getting the required firmware
> + * into an onboard microcontroller. The common code tracks the phases of
> + * fetching the firmware (aka "binary blob") from an external file into a GEM
> + * object in the 'uc_fw_fetch_status' field below; the uC-specific DMA code
> + * uses the 'uc_fw_load_status' field to track the transfer from GEM object
> + * to uC memory.
> + *
> + * For the first (fetch) stage, the interpretation of the values is:
> + * NONE - no firmware is being fetched e.g. because there is no uC
> + * PENDING - parameters initialised, pending call to intel_uc_fw_fetch()
> + * SUCCESS - uC firmware fetched into a GEM object and ready for use
> + * FAIL - something went wrong; uC firmware is not available
> + *
> + * The second (load) stage is similar:
> + * NONE - no firmware is being loaded e.g. because there is no uC
> + * PENDING - firmware DMA load in progress
> + * SUCCESS - uC firmware loaded into uC memory and ready for use
> + * FAIL - something went wrong; uC firmware is not available
> + *
> + * The function intel_uc_fw_status_repr() will convert this enum to a
> + * string representation suitable for use in log messages.
> + */
> +enum intel_uc_fw_status {
> +	INTEL_UC_FIRMWARE_FAIL = -1,
> +	INTEL_UC_FIRMWARE_NONE = 0,
> +	INTEL_UC_FIRMWARE_PENDING,
> +	INTEL_UC_FIRMWARE_SUCCESS
> +};
> +const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status);
> +
> +/*
> + * This structure encapsulates all the data needed during the process of
> + * fetching, caching, and loading the firmware image into the uC.
> + */
> +struct intel_uc_fw {
> +	struct drm_device *		uc_dev;
> +	const char *			uc_name;
> +	const char *			uc_fw_path;
> +	const struct firmware *		uc_fw_blob;
> +	size_t				uc_fw_size;
> +	struct drm_i915_gem_object *	uc_fw_obj;
> +	enum intel_uc_fw_status		uc_fw_fetch_status;
> +	enum intel_uc_fw_status		uc_fw_load_status;
> +};
> +
> +void intel_uc_fw_init(struct drm_device *dev, struct intel_uc_fw *uc_fw,
> +		const char *uc_name, const char *fw_path);
> +int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw,
> +		int callback(struct intel_uc_fw *));
> +void intel_uc_fw_fini(struct intel_uc_fw *uc_fw);
> +
> +/*
> + * The callback() function passed to intel_uc_fw_fetch() above can return
> + * a negative number (a standard error code), or one of the values below:
> + */
> +#define	INTEL_UC_FW_GOOD	1	/* f/w good, save in GEM object	*/
> +#define	INTEL_UC_FW_SAVED	2	/* f/w good, already saved	*/
> +
> +#endif
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 03/15] drm/i915: Add GuC-related module parameters

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:25PM +0100, Dave Gordon wrote:
> From: Alex Dai <yu.dai@intel.com>
> 
> Two new module parameters: "enable_guc_submission" which will turn
> on submission of batchbuffers via the GuC (when implemented), and
> "guc_log_level" which controls the level of debugging logged by the
> GuC and captured by the host.
> 
> Signed-off-by: Alex Dai <yu.dai@intel.com>
> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_drv.h    | 2 ++
>  drivers/gpu/drm/i915/i915_params.c | 9 +++++++++
>  2 files changed, 11 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index b0bb07d..9618f57 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -2599,6 +2599,8 @@ struct i915_params {
>  	bool reset;
>  	bool disable_display;
>  	bool disable_vtd_wa;
> +	bool enable_guc_submission;
> +	int guc_log_level;
>  	int use_mmio_flip;
>  	int mmio_debug;
>  	bool verbose_state_checks;
> diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c
> index 7983fe4..a8f48dd 100644
> --- a/drivers/gpu/drm/i915/i915_params.c
> +++ b/drivers/gpu/drm/i915/i915_params.c
> @@ -53,6 +53,8 @@ struct i915_params i915 __read_mostly = {
>  	.verbose_state_checks = 1,
>  	.nuclear_pageflip = 0,
>  	.edp_vswing = 0,
> +	.enable_guc_submission = false,
> +	.guc_log_level = -1,
>  };
>  
>  module_param_named(modeset, i915.modeset, int, 0400);
> @@ -186,3 +188,10 @@ MODULE_PARM_DESC(edp_vswing,
>  		 "Ignore/Override vswing pre-emph table selection from VBT "
>  		 "(0=use value from vbt [default], 1=low power swing(200mV),"
>  		 "2=default swing(400mV))");
> +
> +module_param_named(enable_guc_submission, i915.enable_guc_submission, bool, 0400);

Please mark this as _unsave, this isn't anything end-users should fiddle
with.

> +MODULE_PARM_DESC(enable_guc_submission, "Enable GuC submission (default:false)");
> +
> +module_param_named(guc_log_level, i915.guc_log_level, int, 0400);

I think this one is ok without _unsave, we want users to be able to
collect logfiles without panicking because it taints their kernel ;-)
-Daniel

> +MODULE_PARM_DESC(guc_log_level,
> +	"GuC firmware logging level (-1:disabled (default), 0-3:enabled)");
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 10/15] drm/i915: Implementation of GuC client

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:32PM +0100, Dave Gordon wrote:
> A GuC client has its own doorbell and workqueue. It maintains the
> doorbell cache line, process description object and work queue item.
> 
> A default guc_client is created for the i915 driver to use for
> normal-priority in-order submission.
> 
> Note that the created client is not yet ready for use; doorbell
> allocation will fail as we haven't yet linked the GuC's context
> descriptor to the default contexts for each ring (see later patch).
> However the driver will still be usable, automatically falling
> back to execlist mode.
> 
> v2:
>     Defer adding structure members until needed [Chris Wilson]
>     Rationalise type declarations [Chris Wilson]
> 
> Issue: VIZ-4884
> Signed-off-by: Alex Dai <yu.dai@intel.com>
> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_guc_submission.c | 661 +++++++++++++++++++++++++++++
>  drivers/gpu/drm/i915/intel_guc.h           |  43 ++
>  drivers/gpu/drm/i915/intel_guc_loader.c    |  10 +
>  3 files changed, 714 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
> index e9d46d6..3c28c8b 100644
> --- a/drivers/gpu/drm/i915/i915_guc_submission.c
> +++ b/drivers/gpu/drm/i915/i915_guc_submission.c
> @@ -27,6 +27,524 @@
>  #include "intel_guc.h"
>  
>  /**
> + * DOC: GuC Client

You need to pull all your shiny new kerneldoc into the drm.tmpl, otherwise
it won't show up. You can build the drm/i915 docbook with

$ make htmldocs

Looking at what we have right now for gem a section next to the execlist
one would fit best I think.

Thanks, Daniel

> + *
> + * i915_guc_client:
> + * We use the term client to avoid confusion with contexts. A i915_guc_client is
> + * equivalent to GuC object guc_context_desc. This context descriptor is
> + * allocated from a pool of 1024 entries. Kernel driver will allocate doorbell
> + * and workqueue for it. Also the process descriptor (guc_process_desc), which
> + * is mapped to client space. So the client can write Work Item then ring the
> + * doorbell.
> + *
> + * To simplify the implementation, we allocate one gem object that contains all
> + * pages for doorbell, process descriptor and workqueue.
> + *
> + * The Scratch registers:
> + * There are 16 MMIO-based registers start from 0xC180. The kernel driver writes
> + * a value to the action register (SOFT_SCRATCH_0) along with any data. It then
> + * triggers an interrupt on the GuC via another register write (0xC4C8).
> + * Firmware writes a success/fail code back to the action register after
> + * processes the request. The kernel driver polls waiting for this update and
> + * then proceeds.
> + * See host2guc_action()
> + *
> + * Doorbells:
> + * Doorbells are interrupts to uKernel. A doorbell is a single cache line (QW)
> + * mapped into process space.
> + *
> + * Work Items:
> + * There are several types of work items that the host may place into a
> + * workqueue, each with its own requirements and limitations. Currently only
> + * WQ_TYPE_INORDER is needed to support legacy submission via GuC, which
> + * represents in-order queue. The kernel driver packs ring tail pointer and an
> + * ELSP context descriptor dword into Work Item.
> + * See guc_add_workqueue_item()
> + *
> + */
> +
> +/*
> + * Read GuC command/status register (SOFT_SCRATCH_0)
> + * Return true if it contains a response rather than a command
> + */
> +static inline bool host2guc_action_response(struct drm_i915_private *dev_priv,
> +					    u32 *status)
> +{
> +	u32 val = I915_READ(SOFT_SCRATCH(0));
> +	*status = val;
> +	return GUC2HOST_IS_RESPONSE(val);
> +}
> +
> +static int host2guc_action(struct intel_guc *guc, u32 *data, u32 len)
> +{
> +	struct drm_i915_private *dev_priv = guc_to_i915(guc);
> +	u32 status;
> +	int i;
> +	int ret;
> +
> +	if (WARN_ON(len < 1 || len > 15))
> +		return -EINVAL;
> +
> +	spin_lock(&dev_priv->guc.host2guc_lock);
> +
> +	dev_priv->guc.action_count += 1;
> +	dev_priv->guc.action_cmd = data[0];
> +
> +	for (i = 0; i < len; i++)
> +		I915_WRITE(SOFT_SCRATCH(i), data[i]);
> +
> +	POSTING_READ(SOFT_SCRATCH(i - 1));
> +
> +	I915_WRITE(HOST2GUC_INTERRUPT, HOST2GUC_TRIGGER);
> +
> +	ret = wait_for_atomic(host2guc_action_response(dev_priv, &status), 10);
> +	if (status != GUC2HOST_STATUS_SUCCESS) {
> +		/* either GuC doesn't respond, which is a TIMEOUT,
> +		 * or a failure code is returned. */
> +		if (ret != -ETIMEDOUT)
> +			ret = -EIO;
> +
> +		DRM_ERROR("GUC: host2guc action 0x%X failed. ret=%d "
> +				"status=0x%08X response=0x%08X\n",
> +				data[0], ret, status,
> +				I915_READ(SOFT_SCRATCH(15)));
> +
> +		dev_priv->guc.action_fail += 1;
> +		dev_priv->guc.action_err = ret;
> +	}
> +	dev_priv->guc.action_status = status;
> +
> +	spin_unlock(&dev_priv->guc.host2guc_lock);
> +
> +	return ret;
> +}
> +
> +/*
> + * Tell the GuC to allocate or deallocate a specific doorbell
> + */
> +
> +static int host2guc_allocate_doorbell(struct intel_guc *guc,
> +				      struct i915_guc_client *client)
> +{
> +	u32 data[2];
> +
> +	data[0] = HOST2GUC_ACTION_ALLOCATE_DOORBELL;
> +	data[1] = client->ctx_index;
> +
> +	return host2guc_action(guc, data, 2);
> +}
> +
> +static int host2guc_release_doorbell(struct intel_guc *guc,
> +				     struct i915_guc_client *client)
> +{
> +	u32 data[2];
> +
> +	data[0] = HOST2GUC_ACTION_DEALLOCATE_DOORBELL;
> +	data[1] = client->ctx_index;
> +
> +	return host2guc_action(guc, data, 2);
> +}
> +
> +/*
> + * Initialise, update, or clear doorbell data shared with the GuC
> + *
> + * These functions modify shared data and so need access to the mapped
> + * client object which contains the page being used for the doorbell
> + */
> +
> +static void guc_init_doorbell(struct intel_guc *guc,
> +			      struct i915_guc_client *client)
> +{
> +	struct guc_doorbell_info *doorbell;
> +	void *base;
> +
> +	base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0));
> +	doorbell = base + client->doorbell_offset;
> +
> +	doorbell->db_status = 1;
> +	doorbell->cookie = 0;
> +
> +	kunmap_atomic(base);
> +}
> +
> +static int guc_ring_doorbell(struct i915_guc_client *gc)
> +{
> +	struct guc_process_desc *desc;
> +	union guc_doorbell_qw db_cmp, db_exc, db_ret;
> +	union guc_doorbell_qw *db;
> +	void *base;
> +	int attempt = 2, ret = -EAGAIN;
> +
> +	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0));
> +	desc = base + gc->proc_desc_offset;
> +
> +	/* Update the tail so it is visible to GuC */
> +	desc->tail = gc->wq_tail;
> +
> +	/* current cookie */
> +	db_cmp.db_status = GUC_DOORBELL_ENABLED;
> +	db_cmp.cookie = gc->cookie;
> +
> +	/* cookie to be updated */
> +	db_exc.db_status = GUC_DOORBELL_ENABLED;
> +	db_exc.cookie = gc->cookie + 1;
> +	if (db_exc.cookie == 0)
> +		db_exc.cookie = 1;
> +
> +	/* pointer of current doorbell cacheline */
> +	db = base + gc->doorbell_offset;
> +
> +	while (attempt--) {
> +		/* lets ring the doorbell */
> +		db_ret.value_qw = atomic64_cmpxchg((atomic64_t *)db,
> +			db_cmp.value_qw, db_exc.value_qw);
> +
> +		/* if the exchange was successfully executed */
> +		if (db_ret.value_qw == db_cmp.value_qw) {
> +			/* db was successfully rung */
> +			gc->cookie = db_exc.cookie;
> +			ret = 0;
> +			break;
> +		}
> +
> +		/* XXX: doorbell was lost and need to acquire it again */
> +		if (db_ret.db_status == GUC_DOORBELL_DISABLED)
> +			break;
> +
> +		DRM_ERROR("Cookie mismatch. Expected %d, returned %d\n",
> +			  db_cmp.cookie, db_ret.cookie);
> +
> +		/* update the cookie to newly read cookie from GuC */
> +		db_cmp.cookie = db_ret.cookie;
> +		db_exc.cookie = db_ret.cookie + 1;
> +		if (db_exc.cookie == 0)
> +			db_exc.cookie = 1;
> +	}
> +
> +	kunmap_atomic(base);
> +	return ret;
> +}
> +
> +static void guc_disable_doorbell(struct intel_guc *guc,
> +				 struct i915_guc_client *client)
> +{
> +	struct drm_i915_private *dev_priv = guc_to_i915(guc);
> +	struct guc_doorbell_info *doorbell;
> +	void *base;
> +	int drbreg = GEN8_DRBREGL(client->doorbell_id);
> +	int value;
> +
> +	base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0));
> +	doorbell = base + client->doorbell_offset;
> +
> +	doorbell->db_status = 0;
> +
> +	kunmap_atomic(base);
> +
> +	I915_WRITE(drbreg, I915_READ(drbreg) & ~GEN8_DRB_VALID);
> +
> +	value = I915_READ(drbreg);
> +	WARN_ON((value & GEN8_DRB_VALID) != 0);
> +
> +	I915_WRITE(GEN8_DRBREGU(client->doorbell_id), 0);
> +	I915_WRITE(drbreg, 0);
> +
> +	/* XXX: wait for any interrupts */
> +	/* XXX: wait for workqueue to drain */
> +}
> +
> +/*
> + * Select, assign and relase doorbell cachelines
> + *
> + * These functions track which doorbell cachelines are in use.
> + * The data they manipulate is protected by the host2guc lock.
> + */
> +
> +static uint32_t select_doorbell_cacheline(struct intel_guc *guc)
> +{
> +	const uint32_t cacheline_size = boot_cpu_data.x86_clflush_size;
> +	uint32_t offset;
> +
> +	spin_lock(&guc->host2guc_lock);
> +
> +	/* Doorbell uses a single cache line within a page */
> +	offset = guc->db_cacheline & PAGE_MASK;
> +
> +	/* Moving to next cache line to reduce contention */
> +	guc->db_cacheline += cacheline_size;
> +
> +	spin_unlock(&guc->host2guc_lock);
> +
> +	return offset;
> +}
> +
> +static uint16_t assign_doorbell(struct intel_guc *guc, uint32_t priority)
> +{
> +	/* The bitmap is split into two halves - high and normal priority. */
> +	const uint16_t half = GUC_MAX_DOORBELLS / 2;
> +	const uint16_t start = (priority <= GUC_CTX_PRIORITY_HIGH) ? half : 0;
> +	const uint16_t end = start + half;
> +	uint16_t id;
> +
> +	spin_lock(&guc->host2guc_lock);
> +	id = find_next_zero_bit(guc->doorbell_bitmap, end, start);
> +	if (id == end)
> +		id = GUC_INVALID_DOORBELL_ID;
> +	else
> +		bitmap_set(guc->doorbell_bitmap, id, 1);
> +	spin_unlock(&guc->host2guc_lock);
> +
> +	return id;
> +}
> +
> +static void release_doorbell(struct intel_guc *guc, uint16_t id)
> +{
> +	spin_lock(&guc->host2guc_lock);
> +	bitmap_clear(guc->doorbell_bitmap, id, 1);
> +	spin_unlock(&guc->host2guc_lock);
> +}
> +
> +/*
> + * Initialise the process descriptor shared with the GuC firmware.
> + */
> +static void guc_init_proc_desc(struct intel_guc *guc,
> +			       struct i915_guc_client *client)
> +{
> +	struct guc_process_desc *desc;
> +	void *base;
> +
> +	base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0));
> +	desc = base + client->proc_desc_offset;
> +
> +	memset(desc, 0, sizeof(*desc));
> +
> +	/*
> +	 * XXX: pDoorbell and WQVBaseAddress are pointers in process address
> +	 * space for ring3 clients (set them as in mmap_ioctl) or kernel
> +	 * space for kernel clients (map on demand instead? May make debug
> +	 * easier to have it mapped).
> +	 */
> +	desc->wq_base_addr = 0;
> +	desc->db_base_addr = 0;
> +
> +	desc->context_id = client->ctx_index;
> +	desc->wq_size_bytes = client->wq_size;
> +	desc->wq_status = WQ_STATUS_ACTIVE;
> +	desc->priority = client->priority;
> +
> +	kunmap_atomic(base);
> +}
> +
> +/*
> + * Initialise/clear the context descriptor shared with the GuC firmware.
> + *
> + * This descriptor tells the GuC where (in GGTT space) to find the important
> + * data structures relating to this client (doorbell, process descriptor,
> + * write queue, etc).
> + */
> +
> +static void guc_init_ctx_desc(struct intel_guc *guc,
> +			      struct i915_guc_client *client)
> +{
> +	struct guc_context_desc desc;
> +	struct sg_table *sg;
> +
> +	memset(&desc, 0, sizeof(desc));
> +
> +	desc.attribute = GUC_CTX_DESC_ATTR_ACTIVE | GUC_CTX_DESC_ATTR_KERNEL;
> +	desc.context_id = client->ctx_index;
> +	desc.priority = client->priority;
> +	desc.engines_used = (1 << RCS) | (1 << VCS) | (1 << BCS) |
> +			    (1 << VECS) | (1 << VCS2); /* all engines */
> +	desc.db_id = client->doorbell_id;
> +
> +	/*
> +	 * The CPU address is only needed at certain points, so kmap_atomic on
> +	 * demand instead of storing it in the ctx descriptor.
> +	 * XXX: May make debug easier to have it mapped
> +	 */
> +	desc.db_trigger_cpu = 0;
> +	desc.db_trigger_uk = client->doorbell_offset +
> +		i915_gem_obj_ggtt_offset(client->client_obj);
> +	desc.db_trigger_phy = client->doorbell_offset +
> +		sg_dma_address(client->client_obj->pages->sgl);
> +
> +	desc.process_desc = client->proc_desc_offset +
> +		i915_gem_obj_ggtt_offset(client->client_obj);
> +
> +	desc.wq_addr = client->wq_offset +
> +		i915_gem_obj_ggtt_offset(client->client_obj);
> +
> +	desc.wq_size = client->wq_size;
> +
> +	/*
> +	 * XXX: Take LRCs from an existing intel_context if this is not an
> +	 * IsKMDCreatedContext client
> +	 */
> +	desc.desc_private = (uintptr_t)client;
> +
> +	/* Pool context is pinned already */
> +	sg = guc->ctx_pool_obj->pages;
> +	sg_pcopy_from_buffer(sg->sgl, sg->nents, &desc, sizeof(desc),
> +			     sizeof(desc) * client->ctx_index);
> +}
> +
> +static void guc_fini_ctx_desc(struct intel_guc *guc,
> +			      struct i915_guc_client *client)
> +{
> +	struct guc_context_desc desc;
> +	struct sg_table *sg;
> +
> +	memset(&desc, 0, sizeof(desc));
> +
> +	sg = guc->ctx_pool_obj->pages;
> +	sg_pcopy_from_buffer(sg->sgl, sg->nents, &desc, sizeof(desc),
> +			     sizeof(desc) * client->ctx_index);
> +}
> +
> +/* Get valid workqueue item and return it back to offset */
> +static int guc_get_workqueue_space(struct i915_guc_client *gc, u32 *offset)
> +{
> +	struct guc_process_desc *desc;
> +	void *base;
> +	u32 size = sizeof(struct guc_wq_item);
> +	int ret = 0, timeout_counter = 200;
> +	unsigned long flags;
> +
> +	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0));
> +	desc = base + gc->proc_desc_offset;
> +
> +	while (timeout_counter-- > 0) {
> +		spin_lock_irqsave(&gc->wq_lock, flags);
> +
> +		ret = wait_for_atomic(CIRC_SPACE(gc->wq_tail, desc->head,
> +				gc->wq_size) >= size, 1);
> +
> +		if (!ret) {
> +			*offset = gc->wq_tail;
> +
> +			/* advance the tail for next workqueue item */
> +			gc->wq_tail += size;
> +			gc->wq_tail &= gc->wq_size - 1;
> +
> +			/* this will break the loop */
> +			timeout_counter = 0;
> +		}
> +
> +		spin_unlock_irqrestore(&gc->wq_lock, flags);
> +	};
> +
> +	kunmap_atomic(base);
> +
> +	return ret;
> +}
> +
> +static int guc_add_workqueue_item(struct i915_guc_client *gc,
> +				  struct intel_context *ctx,
> +				  struct intel_engine_cs *ring)
> +{
> +	struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;
> +	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
> +	struct guc_wq_item *wqi;
> +	void *base;
> +	u32 tail, wq_len, wq_off = 0;
> +	int ret;
> +
> +	ret = guc_get_workqueue_space(gc, &wq_off);
> +	if (ret)
> +		return ret;
> +
> +	/* For now workqueue item is 4 DWs; workqueue buffer is 2 pages. So we
> +	 * should not have the case where structure wqi is across page, neither
> +	 * wrapped to the beginning. This simplifies the implementation below.
> +	 *
> +	 * XXX: if not the case, we need save data to a temp wqi and copy it to
> +	 * workqueue buffer dw by dw.
> +	 */
> +	WARN_ON(sizeof(struct guc_wq_item) != 16);
> +	WARN_ON(wq_off & 3);
> +
> +	/* wq starts from the page after doorbell / process_desc */
> +	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj,
> +			(wq_off + GUC_DB_SIZE) >> PAGE_SHIFT));
> +	wq_off &= PAGE_SIZE - 1;
> +	wqi = (struct guc_wq_item *)((char *)base + wq_off);
> +
> +	/* len does not include the header */
> +	wq_len = sizeof(struct guc_wq_item) / sizeof(u32) - 1;
> +	wqi->header = WQ_TYPE_INORDER |
> +			(wq_len << WQ_LEN_SHIFT) |
> +			(ring->id << WQ_TARGET_SHIFT) |
> +			WQ_NO_WCFLUSH_WAIT;
> +
> +	/* The GuC wants only the low-order word of the context descriptor */
> +	wqi->context_desc = (u32)intel_lr_context_descriptor(ring, ctx_obj);
> +
> +	/* The GuC firmware wants the tail index in QWords, not bytes */
> +	tail = ringbuf->tail >> 3;
> +	wqi->ring_tail = tail << WQ_RING_TAIL_SHIFT;
> +	wqi->fence_id = 0; /*XXX: what fence to be here */
> +
> +	kunmap_atomic(base);
> +
> +	return 0;
> +}
> +
> +static void update_context_image(struct intel_context *ctx,
> +				 struct intel_engine_cs *ring)
> +{
> +	struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;
> +	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
> +
> +	intel_lr_context_update(ctx_obj, ringbuf->obj, ~0u);
> +}
> +
> +/**
> + * i915_guc_submit() - Submit commands through GuC
> + * @client:	the guc client where commands will go through
> + * @ctx:	LRC where commands come from
> + * @ring:	HW engine that will excute the commands
> + *
> + * Return:	0 if succeed
> + */
> +int i915_guc_submit(struct i915_guc_client *client,
> +		    struct intel_context *ctx,
> +		    struct intel_engine_cs *ring)
> +{
> +	int q_ret, b_ret;
> +	unsigned long flags;
> +
> +	/* Need this because of the deferred pin ctx and ring */
> +	/* Shall we move this right after ring is pinned? */
> +	update_context_image(ctx, ring);
> +
> +	q_ret = guc_add_workqueue_item(client, ctx, ring);
> +	if (q_ret == 0)
> +		b_ret = guc_ring_doorbell(client);
> +
> +	spin_lock_irqsave(&client->wq_lock, flags);
> +	client->submissions += 1;
> +	if (q_ret) {
> +		client->q_fail += 1;
> +		client->retcode = q_ret;
> +	} else if (b_ret) {
> +		client->b_fail += 1;
> +		client->retcode = q_ret = b_ret;
> +	} else {
> +		client->retcode = 0;
> +	}
> +	spin_unlock_irqrestore(&client->wq_lock, flags);
> +
> +	return q_ret;
> +}
> +
> +/*
> + * Everything below here is concerned with setup & teardown, and is
> + * therefore not part of the somewhat time-critical batch-submission
> + * path of i915_guc_submit() above.
> + */
> +
> +/**
>   * gem_allocate_guc_obj() - Allocate gem object for GuC usage
>   * @dev:	drm device
>   * @size:	size of object
> @@ -75,6 +593,121 @@ static void gem_release_guc_obj(struct drm_i915_gem_object *obj)
>  	drm_gem_object_unreference(&obj->base);
>  }
>  
> +static void guc_client_free(struct drm_device *dev,
> +			    struct i915_guc_client *client)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_guc *guc = &dev_priv->guc;
> +
> +	if (!client)
> +		return;
> +
> +	if (client->doorbell_id != GUC_INVALID_DOORBELL_ID) {
> +		/*
> +		 * First disable the doorbell, then tell the GuC we've
> +		 * finished with it, finally deallocate it in our bitmap
> +		 */
> +		guc_disable_doorbell(guc, client);
> +		host2guc_release_doorbell(guc, client);
> +		release_doorbell(guc, client->doorbell_id);
> +	}
> +
> +	/*
> +	 * XXX: wait for any outstanding submissions before freeing memory.
> +	 * Be sure to drop any locks
> +	 */
> +
> +	gem_release_guc_obj(client->client_obj);
> +
> +	if (client->ctx_index != GUC_INVALID_CTX_ID) {
> +		guc_fini_ctx_desc(guc, client);
> +		ida_simple_remove(&guc->ctx_ids, client->ctx_index);
> +	}
> +
> +	kfree(client);
> +}
> +
> +/**
> + * guc_client_alloc() - Allocate an i915_guc_client
> + * @dev:	drm device
> + * @priority:	four levels priority _CRITICAL, _HIGH, _NORMAL and _LOW
> + * 		The kernel client to replace ExecList submission is created with
> + * 		NORMAL priority. Priority of a client for scheduler can be HIGH,
> + * 		while a preemption context can use CRITICAL.
> + *
> + * Return:	An i915_guc_client object if success.
> + */
> +static struct i915_guc_client *guc_client_alloc(struct drm_device *dev,
> +						uint32_t priority)
> +{
> +	struct i915_guc_client *client;
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_guc *guc = &dev_priv->guc;
> +	struct drm_i915_gem_object *obj;
> +
> +	client = kzalloc(sizeof(*client), GFP_KERNEL);
> +	if (!client)
> +		return NULL;
> +
> +	client->doorbell_id = GUC_INVALID_DOORBELL_ID;
> +	client->priority = priority;
> +
> +	client->ctx_index = (uint32_t)ida_simple_get(&guc->ctx_ids, 0,
> +			GUC_MAX_GPU_CONTEXTS, GFP_KERNEL);
> +	if (client->ctx_index >= GUC_MAX_GPU_CONTEXTS) {
> +		client->ctx_index = GUC_INVALID_CTX_ID;
> +		goto err;
> +	}
> +
> +	/* The first page is doorbell/proc_desc. Two followed pages are wq. */
> +	obj = gem_allocate_guc_obj(dev, GUC_DB_SIZE + GUC_WQ_SIZE);
> +	if (!obj)
> +		goto err;
> +
> +	client->client_obj = obj;
> +	client->wq_offset = GUC_DB_SIZE;
> +	client->wq_size = GUC_WQ_SIZE;
> +	spin_lock_init(&client->wq_lock);
> +
> +	client->doorbell_offset = select_doorbell_cacheline(guc);
> +
> +	/*
> +	 * Since the doorbell only requires a single cacheline, we can save
> +	 * space by putting the application process descriptor in the same
> +	 * page. Use the half of the page that doesn't include the doorbell.
> +	 */
> +	if (client->doorbell_offset >= (GUC_DB_SIZE / 2))
> +		client->proc_desc_offset = 0;
> +	else
> +		client->proc_desc_offset = (GUC_DB_SIZE / 2);
> +
> +	client->doorbell_id = assign_doorbell(guc, client->priority);
> +	if (client->doorbell_id == GUC_INVALID_DOORBELL_ID)
> +		/* XXX: evict a doorbell instead */
> +		goto err;
> +
> +	guc_init_proc_desc(guc, client);
> +	guc_init_ctx_desc(guc, client);
> +	guc_init_doorbell(guc, client);
> +
> +	/* Invalidate GuC TLB to let GuC take the latest updates to GTT. */
> +	I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
> +
> +	/* XXX: Any cache flushes needed? General domain mgmt calls? */
> +
> +	if (host2guc_allocate_doorbell(guc, client))
> +		goto err;
> +
> +	DRM_DEBUG_DRIVER("new priority %u client %p: ctx_index %u db_id %u\n",
> +		priority, client, client->ctx_index, client->doorbell_id);
> +
> +	return client;
> +
> +err:
> +	guc_client_free(dev, client);
> +	return NULL;
> +}
> +
>  static void guc_create_log(struct intel_guc *guc)
>  {
>  	struct drm_i915_private *dev_priv = guc_to_i915(guc);
> @@ -138,6 +771,8 @@ int i915_guc_submission_init(struct drm_device *dev)
>  	if (!guc->ctx_pool_obj)
>  		return -ENOMEM;
>  
> +	spin_lock_init(&dev_priv->guc.host2guc_lock);
> +
>  	ida_init(&guc->ctx_ids);
>  
>  	guc_create_log(guc);
> @@ -145,6 +780,32 @@ int i915_guc_submission_init(struct drm_device *dev)
>  	return 0;
>  }
>  
> +int i915_guc_submission_enable(struct drm_device *dev)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_guc *guc = &dev_priv->guc;
> +	struct i915_guc_client *client;
> +
> +	/* client for execbuf submission */
> +	client = guc_client_alloc(dev, GUC_CTX_PRIORITY_NORMAL);
> +	if (!client) {
> +		DRM_ERROR("Failed to create execbuf guc_client\n");
> +		return -ENOMEM;
> +	}
> +
> +	guc->execbuf_client = client;
> +	return 0;
> +}
> +
> +void i915_guc_submission_disable(struct drm_device *dev)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_guc *guc = &dev_priv->guc;
> +
> +	guc_client_free(dev, guc->execbuf_client);
> +	guc->execbuf_client = NULL;
> +}
> +
>  void i915_guc_submission_fini(struct drm_device *dev)
>  {
>  	struct drm_i915_private *dev_priv = dev->dev_private;
> diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
> index 3de9096..5bed529 100644
> --- a/drivers/gpu/drm/i915/intel_guc.h
> +++ b/drivers/gpu/drm/i915/intel_guc.h
> @@ -28,6 +28,30 @@
>  #include "intel_guc_fwif.h"
>  #include "i915_guc_reg.h"
>  
> +struct i915_guc_client {
> +	struct drm_i915_gem_object *client_obj;
> +	uint32_t priority;
> +	uint32_t ctx_index;
> +
> +	uint32_t proc_desc_offset;
> +	uint32_t doorbell_offset;
> +	uint32_t cookie;
> +	uint16_t doorbell_id;
> +	uint16_t padding;		/* Maintain alignment		*/
> +
> +	uint32_t wq_offset;
> +	uint32_t wq_size;
> +
> +	spinlock_t wq_lock;		/* Protects all data below	*/
> +	uint32_t wq_tail;
> +
> +	/* GuC submission statistics & status */
> +	uint64_t submissions;
> +	uint32_t q_fail;
> +	uint32_t b_fail;
> +	int retcode;
> +};
> +
>  struct intel_guc {
>  	/* Generic uC firmware management */
>  	struct intel_uc_fw guc_fw;
> @@ -43,6 +67,20 @@ struct intel_guc {
>  
>  	struct drm_i915_gem_object *ctx_pool_obj;
>  	struct ida ctx_ids;
> +
> +	struct i915_guc_client *execbuf_client;
> +
> +	spinlock_t host2guc_lock;	/* Protects all data below	*/
> +
> +	DECLARE_BITMAP(doorbell_bitmap, GUC_MAX_DOORBELLS);
> +	int db_cacheline;
> +
> +	/* Action status & statistics */
> +	uint64_t action_count;		/* Total commands issued	*/
> +	uint32_t action_cmd;		/* Last command word		*/
> +	uint32_t action_status;		/* Last return status		*/
> +	uint32_t action_fail;		/* Total number of failures	*/
> +	int32_t action_err;		/* Last error code		*/
>  };
>  
>  /* intel_guc_loader.c */
> @@ -52,6 +90,11 @@ extern void intel_guc_ucode_fini(struct drm_device *dev);
>  
>  /* i915_guc_submission.c */
>  int i915_guc_submission_init(struct drm_device *dev);
> +int i915_guc_submission_enable(struct drm_device *dev);
> +int i915_guc_submit(struct i915_guc_client *client,
> +		    struct intel_context *ctx,
> +		    struct intel_engine_cs *ring);
> +void i915_guc_submission_disable(struct drm_device *dev);
>  void i915_guc_submission_fini(struct drm_device *dev);
>  
>  #endif
> diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> index deeaa20..eb76c34 100644
> --- a/drivers/gpu/drm/i915/intel_guc_loader.c
> +++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> @@ -417,6 +417,8 @@ int intel_guc_ucode_load(struct drm_device *dev)
>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
>  
> +	i915_guc_submission_disable(dev);
> +
>  	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
>  		return 0;
>  
> @@ -444,12 +446,20 @@ int intel_guc_ucode_load(struct drm_device *dev)
>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
>  
> +	if (i915.enable_guc_submission) {
> +		err = i915_guc_submission_enable(dev);
> +		if (err)
> +			goto fail;
> +	}
> +
>  	return 0;
>  
>  fail:
>  	if (guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_PENDING)
>  		guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_FAIL;
>  
> +	i915_guc_submission_disable(dev);
> +
>  	DRM_ERROR("Failed to initialize GuC, error %d\n", err);
>  
>  	return err;
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:33PM +0100, Dave Gordon wrote:
> Turn on interrupt steering to route necessary interrupts to GuC.
> 
> Issue: VIZ-4884
> Signed-off-by: Alex Dai <yu.dai@intel.com>
> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
>  drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
>  2 files changed, 60 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index eefb847..9f32c6c 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
>  #define GFX_MODE_GEN7	0x0229c
>  #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
>  #define   GFX_RUN_LIST_ENABLE		(1<<15)
> +#define   GFX_INTERRUPT_STEERING	(1<<14)
>  #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
>  #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
>  #define   GFX_REPLAY_MODE		(1<<11)
>  #define   GFX_PSMI_GRANULARITY		(1<<10)
>  #define   GFX_PPGTT_ENABLE		(1<<9)
>  
> +#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
> +#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
> +#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
> +#define   GFX_FORWARD_VBLANK_COND	(2<<5)
> +
>  #define VLV_DISPLAY_BASE 0x180000
>  #define VLV_MIPI_BASE VLV_DISPLAY_BASE
>  
> @@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
>  #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
>  #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
>  
> -#define GEN8_BCS_IRQ_SHIFT 16
>  #define GEN8_RCS_IRQ_SHIFT 0
> -#define GEN8_VCS2_IRQ_SHIFT 16
> +#define GEN8_BCS_IRQ_SHIFT 16
>  #define GEN8_VCS1_IRQ_SHIFT 0
> +#define GEN8_VCS2_IRQ_SHIFT 16
>  #define GEN8_VECS_IRQ_SHIFT 0
> +#define GEN8_WD_IRQ_SHIFT 16
>  
>  #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
>  #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
> diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> index eb76c34..ef5f3d5 100644
> --- a/drivers/gpu/drm/i915/intel_guc_loader.c
> +++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> @@ -62,6 +62,53 @@
>  #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
>  MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
>  
> +static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
> +{
> +	struct intel_engine_cs *ring;
> +	int i, irqs;
> +
> +	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
> +	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
> +	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
> +	for_each_ring(ring, dev_priv, i)
> +		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> +
> +	/* tell DE to send nothing to GuC */
> +	I915_WRITE(DE_GUCRMR, ~0);
> +
> +	/* route all GT interrupts to the host */
> +	I915_WRITE(GUC_BCS_RCS_IER, 0);
> +	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
> +	I915_WRITE(GUC_WD_VECS_IER, 0);
> +}
> +
> +static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
> +{
> +	struct intel_engine_cs *ring;
> +	int i, irqs;
> +
> +	/* tell all command streamers to forward interrupts and vblank to GuC */
> +	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
> +	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
> +	for_each_ring(ring, dev_priv, i)
> +		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> +
> +	/* tell DE to send (all) flip_done to GuC */
> +	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
> +	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
> +	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
> +	/* Unmasked bits will cause GuC response message to be sent */
> +	I915_WRITE(DE_GUCRMR, ~irqs);
> +
> +	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
> +	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
> +	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
> +	/* These three registers have the same bit definitions */
> +	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
> +	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
> +	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
> +}
> +
>  static u32 get_gttype(struct drm_i915_private *dev_priv)
>  {
>  	/* XXX: GT type based on PCI device ID? field seems unused by fw */
> @@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
>  
> +	direct_interrupts_to_host(dev_priv);
>  	i915_guc_submission_disable(dev);
>  
>  	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
> @@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
>  		err = i915_guc_submission_enable(dev);
>  		if (err)
>  			goto fail;
> +		direct_interrupts_to_guc(dev_priv);

This looks like still some remnants from when guc loading could fail and
we needed to switch between cpu and guc submission at runtime. Problem is
that you don't own the interrupt state at this point (since interrupts are
already enabled) and hence there could be problems with races.

Simplest solution would be to just set up the interrupt routing we want
statically at driver load time in the respective irq callbacks (looking at
guc-or-not-guc module option). That means we bake the guc submission y/n
decision into the driver load sequence very early, but it does simplify
things. And that's one of the reasons I don't like runtime fallbacks.
-Daniel

>  	}
>  
>  	return 0;
> @@ -458,6 +507,7 @@ fail:
>  	if (guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_PENDING)
>  		guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_FAIL;
>  
> +	direct_interrupts_to_host(dev_priv);
>  	i915_guc_submission_disable(dev);
>  
>  	DRM_ERROR("Failed to initialize GuC, error %d\n", err);
> @@ -473,6 +523,7 @@ void intel_guc_ucode_fini(struct drm_device *dev)
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
>  
> +	direct_interrupts_to_host(dev_priv);
>  	i915_guc_submission_fini(dev);
>  
>  	intel_uc_fw_fini(guc_fw);
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 14/15] Documentation/drm: kerneldoc for GuC

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:36PM +0100, Dave Gordon wrote:
> From: Alex Dai <yu.dai@intel.com>
> 
> Add overview design of GuC, plus some key points related to
> the implementation.
> 
> Signed-off-by: Alex Dai <yu.dai@intel.com>
> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>

Oops here it is, you can disregard my ask for kerneldoc integration. Next
time around just integrate the new kerneldoc with the drm.tmpl as you go.
-Daniel

> ---
>  Documentation/DocBook/drm.tmpl | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
> 
> diff --git a/Documentation/DocBook/drm.tmpl b/Documentation/DocBook/drm.tmpl
> index c0312cb..88b53ee 100644
> --- a/Documentation/DocBook/drm.tmpl
> +++ b/Documentation/DocBook/drm.tmpl
> @@ -4218,6 +4218,25 @@ int num_ioctls;</synopsis>
>        </sect2>
>      </sect1>
>      <sect1>
> +      <title>GuC-based Command Submission</title>
> +      <sect2>
> +        <title>Microcontroller (uC) firmware loading support</title>
> +!Pdrivers/gpu/drm/i915/intel_uc_loader.c Microcontroller (uC) firmware loading support
> +!Idrivers/gpu/drm/i915/intel_uc_loader.c
> +      </sect2>
> +      <sect2>
> +        <title>GuC</title>
> +!Pdrivers/gpu/drm/i915/intel_guc_loader.c GuC-specific firmware loader
> +!Idrivers/gpu/drm/i915/intel_guc_loader.c
> +      </sect2>
> +      <sect2>
> +        <title>GuC Client</title>
> +!Pdrivers/gpu/drm/i915/intel_guc_submission.c GuC-based command submissison
> +!Idrivers/gpu/drm/i915/intel_guc_submission.c
> +      </sect2>
> +    </sect1>
> +
> +    <sect1>
>        <title> Tracing </title>
>        <para>
>      This sections covers all things related to the tracepoints implemented in
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 05/15] drm/i915: GuC-specific firmware loader

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:27PM +0100, Dave Gordon wrote:
> From: Alex Dai <yu.dai@intel.com>
> 
> This uses the common firmware loader to fetch the firmware image,
> then loads it into the GuC's memory via a dedicated DMA engine.
> 
> This patch is derived from GuC loading work originally done by
> Vinit Azad and Ben Widawsky. It has been reconstructed to accord
> with the common firmware loading mechanism by Dave Gordon as well
> as new firmware layout etc.
> 
> v2:
>     Various improvements per review comments by Chris Wilson
> 
> v3:
>     Removed 'wait' parameter to intel_guc_ucode_load() as prefetch
>         is no longer supported in the common firmware loader, per
> 	Daniel Vetter's request.
>     F/w checker callback fn now returns errno rather than bool.
> 
> Issue: VIZ-4884
> Signed-off-by: Alex Dai <yu.dai@intel.com>
> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> ---
>  drivers/gpu/drm/i915/Makefile           |   3 +
>  drivers/gpu/drm/i915/i915_dma.c         |   4 +
>  drivers/gpu/drm/i915/i915_drv.h         |  11 +
>  drivers/gpu/drm/i915/i915_gem.c         |   8 +
>  drivers/gpu/drm/i915/i915_reg.h         |   4 +-
>  drivers/gpu/drm/i915/intel_guc.h        |  49 ++++
>  drivers/gpu/drm/i915/intel_guc_loader.c | 448 ++++++++++++++++++++++++++++++++
>  7 files changed, 526 insertions(+), 1 deletion(-)
>  create mode 100644 drivers/gpu/drm/i915/intel_guc.h
>  create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c
> 
> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> index f1f80fc..62a8c83 100644
> --- a/drivers/gpu/drm/i915/Makefile
> +++ b/drivers/gpu/drm/i915/Makefile
> @@ -42,6 +42,9 @@ i915-y += i915_cmd_parser.o \
>  # generic ancilliary microcontroller support
>  i915-y += intel_uc_loader.o
>  
> +# general-purpose microcontroller (GuC) support
> +i915-y += intel_guc_loader.o
> +
>  # autogenerated null render state
>  i915-y += intel_renderstate_gen6.o \
>  	  intel_renderstate_gen7.o \
> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> index c5349fa..730d91b 100644
> --- a/drivers/gpu/drm/i915/i915_dma.c
> +++ b/drivers/gpu/drm/i915/i915_dma.c
> @@ -469,6 +469,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
>  
>  cleanup_gem:
>  	mutex_lock(&dev->struct_mutex);
> +	intel_guc_ucode_fini(dev);
>  	i915_gem_cleanup_ringbuffer(dev);
>  	i915_gem_context_fini(dev);
>  	mutex_unlock(&dev->struct_mutex);
> @@ -866,6 +867,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
>  
>  	intel_uncore_init(dev);
>  
> +	intel_guc_ucode_init(dev);
> +
>  	/* Load CSR Firmware for SKL */
>  	intel_csr_ucode_init(dev);
>  
> @@ -1117,6 +1120,7 @@ int i915_driver_unload(struct drm_device *dev)
>  	flush_workqueue(dev_priv->wq);
>  
>  	mutex_lock(&dev->struct_mutex);
> +	intel_guc_ucode_fini(dev);
>  	i915_gem_cleanup_ringbuffer(dev);
>  	i915_gem_context_fini(dev);
>  	mutex_unlock(&dev->struct_mutex);
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 9618f57..a7ccac5 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -50,6 +50,7 @@
>  #include <linux/intel-iommu.h>
>  #include <linux/kref.h>
>  #include <linux/pm_qos.h>
> +#include "intel_guc.h"
>  
>  /* General customization:
>   */
> @@ -1687,6 +1688,8 @@ struct drm_i915_private {
>  
>  	struct i915_virtual_gpu vgpu;
>  
> +	struct intel_guc guc;
> +
>  	struct intel_csr csr;
>  
>  	/* Display CSR-related protection */
> @@ -1931,6 +1934,11 @@ static inline struct drm_i915_private *dev_to_i915(struct device *dev)
>  	return to_i915(dev_get_drvdata(dev));
>  }
>  
> +static inline struct drm_i915_private *guc_to_i915(struct intel_guc *guc)
> +{
> +	return container_of(guc, struct drm_i915_private, guc);
> +}
> +
>  /* Iterate over initialised rings */
>  #define for_each_ring(ring__, dev_priv__, i__) \
>  	for ((i__) = 0; (i__) < I915_NUM_RINGS; (i__)++) \
> @@ -2539,6 +2547,9 @@ struct drm_i915_cmd_table {
>  
>  #define HAS_CSR(dev)	(IS_SKYLAKE(dev))
>  
> +#define HAS_GUC_UCODE(dev)	(IS_GEN9(dev))
> +#define HAS_GUC_SCHED(dev)	(IS_GEN9(dev))
> +
>  #define INTEL_PCH_DEVICE_ID_MASK		0xff00
>  #define INTEL_PCH_IBX_DEVICE_ID_TYPE		0x3b00
>  #define INTEL_PCH_CPT_DEVICE_ID_TYPE		0x1c00
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index aa8f4c3..80d7890 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -5076,6 +5076,14 @@ i915_gem_init_hw(struct drm_device *dev)
>  			goto out;
>  	}
>  
> +	/*
> +	 * We can't enable contexts until all firmware is loaded; if this
> +	 * fails, disable GuC submissions and fall back to execlist mode
> +	 */
> +	ret = intel_guc_ucode_load(dev);
> +	if (ret)
> +		i915.enable_guc_submission = false;

I want an -EIO or similar here since runtime fallbacks to other modes
really aren't great from a maintainance perspective, see my comments on
the irq routing code.

Yes we can make this work, but givin our stellar track record with keeping
disabled features working it won't work for long. And it will impact us
with additional constraints until we give up and rip it out again. Not
worth it imo - if we decide to use the guc on a given platform we should
imo require it and stick to that decision for at least as long as the
driver is loaded. Developers can still change the option when reloading the
driver, users won't have a chance to cause trouble.
-Daniel

> +
>  	/* Now it is safe to go back round and do everything else: */
>  	for_each_ring(ring, dev_priv, i) {
>  		struct drm_i915_gem_request *req;
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 313b1f9..eefb847 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -6837,7 +6837,9 @@ enum skl_disp_power_wells {
>  #define   GEN9_PGCTL_SSB_EU311_ACK	(1 << 14)
>  
>  #define GEN7_MISCCPCTL			(0x9424)
> -#define   GEN7_DOP_CLOCK_GATE_ENABLE	(1<<0)
> +#define   GEN7_DOP_CLOCK_GATE_ENABLE		(1<<0)
> +#define   GEN8_DOP_CLOCK_GATE_CFCLK_ENABLE	(1<<2)
> +#define   GEN8_DOP_CLOCK_GATE_GUC_ENABLE	(1<<4)
>  
>  /* IVYBRIDGE DPF */
>  #define GEN7_L3CDERRST1			0xB008 /* L3CD Error Status 1 */
> diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
> new file mode 100644
> index 0000000..b38d2b0
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/intel_guc.h
> @@ -0,0 +1,49 @@
> +/*
> + * Copyright © 2014 Intel Corporation
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> + * IN THE SOFTWARE.
> + *
> + */
> +#ifndef _INTEL_GUC_H_
> +#define _INTEL_GUC_H_
> +
> +#include "intel_uc_loader.h"
> +#include "intel_guc_fwif.h"
> +#include "i915_guc_reg.h"
> +
> +struct intel_guc {
> +	/* Generic uC firmware management */
> +	struct intel_uc_fw guc_fw;
> +
> +	/* GuC-specific additions */
> +	uint16_t fw_major_wanted;
> +	uint16_t fw_minor_wanted;
> +	uint16_t fw_major_found;
> +	uint16_t fw_minor_found;
> +
> +	uint32_t log_flags;
> +};
> +
> +/* intel_guc_loader.c */
> +extern void intel_guc_ucode_init(struct drm_device *dev);
> +extern int intel_guc_ucode_load(struct drm_device *dev);
> +extern void intel_guc_ucode_fini(struct drm_device *dev);
> +
> +#endif
> diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> new file mode 100644
> index 0000000..4929838
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> @@ -0,0 +1,448 @@
> +/*
> + * Copyright © 2014 Intel Corporation
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> + * IN THE SOFTWARE.
> + *
> + * Authors:
> + *    Vinit Azad <vinit.azad@intel.com>
> + *    Ben Widawsky <ben@bwidawsk.net>
> + *    Dave Gordon <david.s.gordon@intel.com>
> + *    Alex Dai <yu.dai@intel.com>
> + */
> +#include <linux/firmware.h>
> +#include "i915_drv.h"
> +#include "intel_guc.h"
> +
> +/**
> + * DOC: GuC
> + *
> + * intel_guc:
> + * Top level structure of guc. It handles firmware loading and manages client
> + * pool and doorbells. intel_guc owns a i915_guc_client to replace the legacy
> + * ExecList submission.
> + *
> + * Firmware versioning:
> + * The firmware build process will generate a version header file with major and
> + * minor version defined. The versions are built into CSS header of firmware.
> + * i915 kernel driver set the minimal firmware version required per platform.
> + * The firmware installation package will install (symbolic link) proper version
> + * of firmware.
> + *
> + * GuC address space:
> + * GuC does not allow any gfx GGTT address that falls into range [0, WOPCM_TOP),
> + * which is reserved for Boot ROM, SRAM and WOPCM. Currently this top address is
> + * 512K. In order to exclude 0-512K address space from GGTT, all gfx objects
> + * used by GuC is pinned with PIN_OFFSET_BIAS along with size of WOPCM.
> + *
> + * Firmware log:
> + * Firmware log is enabled by setting i915.guc_log_level to non-negative level.
> + * Log data is printed out via reading debugfs i915_guc_log_dump. Reading from
> + * i915_guc_load_status will print out firmware loading status and scratch
> + * registers value.
> + *
> + */
> +
> +#define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
> +MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
> +
> +static u32 get_gttype(struct drm_i915_private *dev_priv)
> +{
> +	/* XXX: GT type based on PCI device ID? field seems unused by fw */
> +	return 0;
> +}
> +
> +static u32 get_core_family(struct drm_i915_private *dev_priv)
> +{
> +	switch (INTEL_INFO(dev_priv)->gen) {
> +	case 8:
> +		return GFXCORE_FAMILY_GEN8;
> +	case 9:
> +		return GFXCORE_FAMILY_GEN9;
> +	default:
> +		DRM_ERROR("GUC: unknown gen for scheduler init\n");
> +		return GFXCORE_FAMILY_FORCE_ULONG;
> +	}
> +}
> +
> +static void set_guc_init_params(struct drm_i915_private *dev_priv)
> +{
> +	struct intel_guc *guc = &dev_priv->guc;
> +	u32 params[GUC_CTL_MAX_DWORDS];
> +	int i;
> +
> +	memset(&params, 0, sizeof(params));
> +
> +	params[GUC_CTL_DEVICE_INFO] |=
> +		(get_gttype(dev_priv) << GUC_CTL_GTTYPE_SHIFT) |
> +		(get_core_family(dev_priv) << GUC_CTL_COREFAMILY_SHIFT);
> +
> +	/* GuC ARAT increment is 10 ns. GuC default scheduler quantum is one
> +	 * second. This ARAR is calculated by:
> +	 * Scheduler-Quantum-in-ns / ARAT-increment-in-ns = 1000000000 / 10
> +	 */
> +	params[GUC_CTL_ARAT_HIGH] = 0;
> +	params[GUC_CTL_ARAT_LOW] = 100000000;
> +
> +	params[GUC_CTL_WA] |= GUC_CTL_WA_UK_BY_DRIVER;
> +
> +	params[GUC_CTL_FEATURE] |= GUC_CTL_DISABLE_SCHEDULER |
> +			GUC_CTL_VCS2_ENABLED;
> +
> +	if (i915.guc_log_level >= 0) {
> +		params[GUC_CTL_LOG_PARAMS] = guc->log_flags;
> +		params[GUC_CTL_DEBUG] =
> +			i915.guc_log_level << GUC_LOG_VERBOSITY_SHIFT;
> +	}
> +
> +	I915_WRITE(SOFT_SCRATCH(0), 0);
> +
> +	for (i = 0; i < GUC_CTL_MAX_DWORDS; i++)
> +		I915_WRITE(SOFT_SCRATCH(1 + i), params[i]);
> +}
> +
> +/*
> + * Read the GuC status register (GUC_STATUS) and store it in the
> + * specified location; then return a boolean indicating whether
> + * the value matches either of two values representing completion
> + * of the GuC boot process.
> + *
> + * This is used for polling the GuC status in a wait_for_atomic()
> + * loop below.
> + */
> +static inline bool guc_ucode_response(struct drm_i915_private *dev_priv,
> +				      u32 *status)
> +{
> +	u32 val = I915_READ(GUC_STATUS);
> +	*status = val;
> +	return ((val & GS_UKERNEL_MASK) == GS_UKERNEL_READY ||
> +		(val & GS_UKERNEL_MASK) == GS_UKERNEL_LAPIC_DONE);
> +}
> +
> +/*
> + * Transfer the firmware image to RAM for execution by the microcontroller.
> + *
> + * GuC Firmware layout:
> + * +-------------------------------+  ----
> + * |          CSS header           |  128B
> + * | contains major/minor version  |
> + * +-------------------------------+  ----
> + * |             uCode             |
> + * +-------------------------------+  ----
> + * |         RSA signature         |  256B
> + * +-------------------------------+  ----
> + * |         RSA public Key        |  256B
> + * +-------------------------------+  ----
> + * |       Public key modulus      |    4B
> + * +-------------------------------+  ----
> + *
> + * Architecturally, the DMA engine is bidirectional, and can potentially even
> + * transfer between GTT locations. This functionality is left out of the API
> + * for now as there is no need for it.
> + *
> + * Note that GuC needs the CSS header plus uKernel code to be copied by the
> + * DMA engine in one operation, whereas the RSA signature is loaded via MMIO.
> + */
> +
> +#define UOS_CSS_HEADER_OFFSET		0
> +#define UOS_VER_MINOR_OFFSET		0x44
> +#define UOS_VER_MAJOR_OFFSET		0x46
> +#define UOS_CSS_HEADER_SIZE		0x80
> +#define UOS_RSA_SIG_SIZE		0x100
> +#define UOS_CSS_SIGNING_SIZE		0x204
> +
> +static int guc_ucode_xfer_dma(struct drm_i915_private *dev_priv)
> +{
> +	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
> +	struct drm_i915_gem_object *fw_obj = guc_fw->uc_fw_obj;
> +	unsigned long offset;
> +	struct sg_table *sg = fw_obj->pages;
> +	u32 status, ucode_size, rsa[UOS_RSA_SIG_SIZE / sizeof(u32)];
> +	int i, ret = 0;
> +
> +	/* uCode size, also is where RSA signature starts */
> +	offset = ucode_size = guc_fw->uc_fw_size - UOS_CSS_SIGNING_SIZE;
> +
> +	/* Copy RSA signature from the fw image to HW for verification */
> +	sg_pcopy_to_buffer(sg->sgl, sg->nents, rsa, UOS_RSA_SIG_SIZE, offset);
> +	for (i = 0; i < UOS_RSA_SIG_SIZE / sizeof(u32); i++)
> +		I915_WRITE(UOS_RSA_SCRATCH_0 + i * sizeof(u32), rsa[i]);
> +
> +	/* Set the source address for the new blob */
> +	offset = i915_gem_obj_ggtt_offset(fw_obj);
> +	I915_WRITE(DMA_ADDR_0_LOW, lower_32_bits(offset));
> +	I915_WRITE(DMA_ADDR_0_HIGH, upper_32_bits(offset) & 0xFFFF);
> +
> +	/* Set the destination. Current uCode expects an 8k stack starting from
> +	 * offset 0. */
> +	I915_WRITE(DMA_ADDR_1_LOW, 0x2000);
> +
> +	/* XXX: The image is automatically transfered to SRAM after the RSA
> +	 * verification. This is why the address space is chosen as such. */
> +	I915_WRITE(DMA_ADDR_1_HIGH, DMA_ADDRESS_SPACE_WOPCM);
> +
> +	I915_WRITE(DMA_COPY_SIZE, ucode_size);
> +
> +	/* Finally start the DMA */
> +	I915_WRITE(DMA_CTRL, _MASKED_BIT_ENABLE(UOS_MOVE | START_DMA));
> +
> +	/*
> +	 * Spin-wait for the DMA to complete & the GuC to start up.
> +	 * NB: Docs recommend not using the interrupt for completion.
> +	 * Measurements indicate this should take no more than 20ms, so a
> +	 * timeout here indicates that the GuC has failed and is unusable.
> +	 * (Higher levels of the driver will attempt to fall back to
> +	 * execlist mode if this happens.)
> +	 */
> +	ret = wait_for_atomic(guc_ucode_response(dev_priv, &status), 100);
> +
> +	DRM_DEBUG_DRIVER("DMA status 0x%x, GuC status 0x%x\n",
> +			I915_READ(DMA_CTRL), status);
> +
> +	if ((status & GS_BOOTROM_MASK) == GS_BOOTROM_RSA_FAILED) {
> +		DRM_ERROR("%s firmware signature verification failed\n",
> +			guc_fw->uc_name);
> +		ret = -ENOEXEC;
> +	}
> +
> +	DRM_DEBUG_DRIVER("returning %d\n", ret);
> +
> +	return ret;
> +}
> +
> +/*
> + * Load the GuC firmware blob into the MinuteIA.
> + */
> +static int guc_ucode_xfer(struct drm_i915_private *dev_priv)
> +{
> +	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
> +	struct drm_device *dev = dev_priv->dev;
> +	int ret;
> +
> +	ret = i915_gem_object_set_to_gtt_domain(guc_fw->uc_fw_obj, false);
> +	if (ret) {
> +		DRM_DEBUG_DRIVER("set-domain failed %d\n", ret);
> +		return ret;
> +	}
> +
> +	ret = i915_gem_obj_ggtt_pin(guc_fw->uc_fw_obj, 0, 0);
> +	if (ret) {
> +		DRM_DEBUG_DRIVER("pin failed %d\n", ret);
> +		return ret;
> +	}
> +
> +	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
> +	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
> +
> +	/* init WOPCM */
> +	I915_WRITE(GUC_WOPCM_SIZE, GUC_WOPCM_SIZE_VALUE);
> +	I915_WRITE(DMA_GUC_WOPCM_OFFSET, GUC_WOPCM_OFFSET);
> +
> +	/* Invalidate GuC TLB to let GuC take the latest updates to GTT. */
> +	I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
> +
> +	/* Set MMIO/WA for GuC init */
> +	I915_WRITE(DRBMISC1, DOORBELL_ENABLE);
> +
> +	/* Enable MIA caching. GuC clock gating is disabled. */
> +	I915_WRITE(GUC_SHIM_CONTROL, GUC_SHIM_CONTROL_VALUE);
> +
> +	/* WaC6DisallowByGfxPause*/
> +	I915_WRITE(GEN6_GFXPAUSE, 0x30FFF);
> +
> +	if (IS_SKYLAKE(dev))
> +		I915_WRITE(GEN9_GT_PM_CONFIG, GEN8_GT_DOORBELL_ENABLE);
> +	else
> +		I915_WRITE(GEN8_GT_PM_CONFIG, GEN8_GT_DOORBELL_ENABLE);
> +
> +	if (IS_GEN9(dev)) {
> +		/* DOP Clock Gating Enable for GuC clocks */
> +		I915_WRITE(GEN7_MISCCPCTL, (GEN8_DOP_CLOCK_GATE_GUC_ENABLE |
> +					    I915_READ(GEN7_MISCCPCTL)));
> +
> +		/* allows for 5us before GT can go to RC6 */
> +		I915_WRITE(GUC_ARAT_C6DIS, 0x1FF);
> +	}
> +
> +	set_guc_init_params(dev_priv);
> +
> +	ret = guc_ucode_xfer_dma(dev_priv);
> +
> +	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
> +
> +	/*
> +	 * We keep the object pages for reuse during resume. But we can unpin it
> +	 * now that DMA has completed, so it doesn't continue to take up space.
> +	 */
> +	i915_gem_object_ggtt_unpin(guc_fw->uc_fw_obj);
> +
> +	return ret;
> +}
> +
> +/*
> + * Check the firmware that was found; if it's the wrong size or the wrong
> + * version, return a negative error code. If it's OK, return a positive
> + * status value. Here we can just return INTEL_UC_FW_GOOD; the common loader
> + * code will then save the data for later in a pageable (tmpfs-backed) GEM
> + * object.
> + *
> + * Alternatively (for example if we wanted only part of the image) we could
> + * save the required portion here and then return INTEL_UC_FW_SAVED to tell
> + * the common loader that the data is good, and we've already handled saving
> + * anything we need later.
> + *
> + * The GuC firmware image has the version number embedded at a well-known
> + * offset within the firmware blob; note that major / minor version are
> + * TWO bytes each (i.e. u16), although all pointers and offsets are defined
> + * in terms of bytes (u8).
> + */
> +static int guc_ucode_check(struct intel_uc_fw *guc_fw)
> +{
> +	struct intel_guc *guc = container_of(guc_fw, struct intel_guc, guc_fw);
> +	const u8 *css_header = guc_fw->uc_fw_blob->data + UOS_CSS_HEADER_OFFSET;
> +	const size_t blobsize = guc_fw->uc_fw_blob->size;
> +	const size_t minsize = UOS_CSS_HEADER_SIZE + UOS_CSS_SIGNING_SIZE;
> +	const size_t maxsize = GUC_WOPCM_SIZE_VALUE + UOS_CSS_SIGNING_SIZE
> +			- 0x8000; /* 32k reserved (8K stack + 24k context) */
> +
> +	DRM_DEBUG_DRIVER("firmware file size %zu (minimum %zu, maximum %zu)\n",
> +		blobsize, minsize, maxsize);
> +
> +	/* Check the size of the blob befoe examining buffer contents */
> +	if (blobsize < minsize || blobsize > maxsize)
> +		return -ENOEXEC;
> +
> +	guc->fw_major_found = *(u16 *)(css_header + UOS_VER_MAJOR_OFFSET);
> +	guc->fw_minor_found = *(u16 *)(css_header + UOS_VER_MINOR_OFFSET);
> +
> +	if (guc->fw_major_found != guc->fw_major_wanted ||
> +	    guc->fw_minor_found < guc->fw_minor_wanted) {
> +		DRM_ERROR("GuC firmware version %d.%d, required %d.%d\n",
> +			guc->fw_major_found, guc->fw_minor_found,
> +			guc->fw_major_wanted, guc->fw_minor_wanted);
> +		return -ENOEXEC;
> +	}
> +
> +	DRM_DEBUG_DRIVER("firmware version %d.%d OK (minimum %d.%d)\n",
> +			guc->fw_major_found, guc->fw_minor_found,
> +			guc->fw_major_wanted, guc->fw_minor_wanted);
> +
> +	return INTEL_UC_FW_GOOD;
> +}
> +
> +/**
> + * intel_guc_ucode_init() - initiate a firmware loading request
> + *
> + * Called early during driver load, before GEM is initialised.
> + * Driver is single threaded, so no mutex is required.
> + *
> + * This just sets parameters for use when intel_guc_ucode_load()
> + * is called later, after GEM initialisation is complete.
> + */
> +void intel_guc_ucode_init(struct drm_device *dev)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_guc *guc = &dev_priv->guc;
> +	struct intel_uc_fw *guc_fw = &guc->guc_fw;
> +	const char *path;
> +
> +	if (!HAS_GUC_SCHED(dev))
> +		i915.enable_guc_submission = false;
> +
> +	if (!HAS_GUC_UCODE(dev)) {
> +		path = NULL;
> +	} else if (IS_SKYLAKE(dev)) {
> +		path = I915_SKL_GUC_UCODE;
> +		guc->fw_major_wanted = 3;
> +		guc->fw_minor_wanted = 0;
> +	} else {
> +		i915.enable_guc_submission = false;
> +		path = "";	/* unknown device */
> +	}
> +
> +	intel_uc_fw_init(dev, guc_fw, "GuC", path);
> +}
> +
> +/**
> + * intel_guc_ucode_load() - load GuC uCode into the device
> + *
> + * Called from gem_init_hw() during driver loading and also after a GPU reset.
> + *
> + * Calls the common loader to get the firmware registered earlier. On the first
> + * call, this will actually fetch it from the filesystem; thereafter, we will
> + * already either have the blob in a GEM object, or have determined that no
> + * valid firmware image could be found).
> + *
> + * If we have a good firmware image, transfer it to the h/w.
> + */
> +int intel_guc_ucode_load(struct drm_device *dev)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
> +	int err;
> +
> +	DRM_DEBUG_DRIVER("%s fw status: fetch %s, load %s\n",
> +		guc_fw->uc_name,
> +		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
> +		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
> +
> +	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
> +		return 0;
> +
> +	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_SUCCESS &&
> +	    guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_FAIL)
> +		return -ENOEXEC;
> +
> +	guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_PENDING;
> +	err = intel_uc_fw_fetch(guc_fw, guc_ucode_check);
> +	if (err)
> +		goto fail;
> +
> +	err = guc_ucode_xfer(dev_priv);
> +	if (err)
> +		goto fail;
> +
> +	guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_SUCCESS;
> +
> +	DRM_DEBUG_DRIVER("%s fw status: fetch %s, load %s\n",
> +		guc_fw->uc_name,
> +		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
> +		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
> +
> +	return 0;
> +
> +fail:
> +	if (guc_fw->uc_fw_load_status == INTEL_UC_FIRMWARE_PENDING)
> +		guc_fw->uc_fw_load_status = INTEL_UC_FIRMWARE_FAIL;
> +
> +	DRM_ERROR("Failed to initialize GuC, error %d\n", err);
> +
> +	return err;
> +}
> +
> +/**
> + * intel_guc_ucode_fini() - clean up all allocated resources
> + */
> +void intel_guc_ucode_fini(struct drm_device *dev)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;
> +
> +	intel_uc_fw_fini(guc_fw);
> +}
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 00/15 v3] Batch submission via GuC

[Daniel Vetter]

On Fri, Jul 03, 2015 at 01:30:22PM +0100, Dave Gordon wrote:
> This patch series enables command submission via the GuC. In this mode,
> instead of the host CPU driving the execlist port directly, it hands
> over work items to the GuC, using a doorbell mechanism to tell the GuC
> that new items have been added to its work queue. The GuC then dispatches
> contexts to the various GPU engines, and manages the resulting context-
> switch interrupts. Completion of a batch is however still signalled to
> the CPU; the GuC is not involved in handling user interrupts.
> 
> There are three subsequences within the patch series:
> 
>   drm/i915: Add i915_gem_object_create_from_data()
>   drm/i915: Embedded microcontroller (uC) firmware loading support
> 
> These first two patches provide a common framework for fetching the
> firmware that may be required by any embedded microcontroller from a
> file, and saving some or all of it in a pageable (shmfs-backed) GEM
> object (or otherwise, if required by the uC-specific code).  It is
> hoped that this framework is sufficiently general that it can be used
> for all curent and future microcontrollers.
> 
>   drm/i915: Add GuC-related module parameters
>   drm/i915: Add GuC-related header files
>   drm/i915: GuC-specific firmware loader
>   drm/i915: Debugfs interface to read GuC load status
> 
> These four patches complete the GuC loader. At this point in the sequence
> we can load and activate the GuC firmware, but not submit any batches
> through it. (This is nonetheless a potentially useful state, as the GuC
> could do other useful work even when not handling batch submissions).
> 
>   drm/i915: Expose two LRC functions for GuC submission mode
>   drm/i915: GuC submission setup, phase 1
>   drm/i915: Enable GuC firmware log
>   drm/i915: Implementation of GuC client
>   drm/i915: Interrupt routing for GuC submission
>   drm/i915: Integrate GuC-based command submission
>   drm/i915: Debugfs interface for GuC submission statistics
>   Documentation/drm: kerneldoc for GuC
>   drm/i915: Enable GuC submission, where supported
> 
> In the final section, we implement the GuC submission mechanism, link
> it into the (execlist-based) submission path, and finally enable it
> (on supported platforms). On platforms where there is no GuC, or if
> the GuC firmware cannot be found or is invalid, batch submission will
> revert to using the execlist mechanism directly.
> 
> The GuC firmware itself is not included in this patchset; it is or will
> be available for download from https://01.org/linuxgraphics/downloads/
> This driver works with and requires GuC firmware revision 3.x. It will
> not work with any firmware version 1.x, as the GuC protocol in those
> revisions was incompatible and is no longer supported.
> 
> Prerequisites: GuC submission will expose existing inadequacies in
> some of the existing codepaths unless certain other patches are applied.
> In particular we require Michel Thierry's patch:
>   drm/i915/lrc: Update PDPx registers with LRI commands
> (because the GuC support light-restore, which execlist mode doesn't).
> Unfortunately the version of this recently committed to drm-intel-nightly
>     7a01a0a drm/i915/lrc: Update PDPx registers with lri commands
> was incomplete in a way that was mostly harmless, but incompatible with
> GuC usage, so the missing part has been folded into patch 7 below:
>   drm/i915: Expose two LRC functions for GuC submission mode
> The previous prerequisite patch
>   drm/i915: Allocate OLR more safely (workaround until OLR goes away)
> is no longer required as the OLR has now gone away :)
> 
> Alex Dai (7):
>   drm/i915: Add GuC-related module parameters
>   drm/i915: GuC-specific firmware loader
>   drm/i915: Debugfs interface to read GuC load status
>   drm/i915: GuC submission setup, phase 1
>   drm/i915: Enable GuC firmware log
>   drm/i915: Integrate GuC-based command submission
>   Documentation/drm: kerneldoc for GuC
> 
> Dave Gordon (8):
>   drm/i915: Add i915_gem_object_create_from_data()
>   drm/i915: Embedded microcontroller (uC) firmware loading support
>   drm/i915: Add GuC-related header files
>   drm/i915: Expose two LRC functions for GuC submission mode
>   drm/i915: Implementation of GuC client
>   drm/i915: Interrupt routing for GuC submission
>   drm/i915: Debugfs interface for GuC submission statistics
>   drm/i915: Enable GuC submission, where supported
> 
> Ben Widawsky (0):
> Vinit Azad (0):
> Michael H. Nguyen (0):
>   created the original versions on which some of these patches are based.

A few minor things still (I guess you expected my comment on the firmware
loader) but looks good overall with those addressed and the detailed
review done on top.
-Daniel

> 
>  Documentation/DocBook/drm.tmpl             |  19 +
>  drivers/gpu/drm/i915/Makefile              |   7 +
>  drivers/gpu/drm/i915/i915_debugfs.c        | 116 +++-
>  drivers/gpu/drm/i915/i915_dma.c            |   4 +
>  drivers/gpu/drm/i915/i915_drv.h            |  15 +
>  drivers/gpu/drm/i915/i915_gem.c            |  48 ++
>  drivers/gpu/drm/i915/i915_guc_reg.h        | 102 ++++
>  drivers/gpu/drm/i915/i915_guc_submission.c | 861 +++++++++++++++++++++++++++++
>  drivers/gpu/drm/i915/i915_params.c         |   9 +
>  drivers/gpu/drm/i915/i915_reg.h            |  15 +-
>  drivers/gpu/drm/i915/intel_guc.h           | 101 ++++
>  drivers/gpu/drm/i915/intel_guc_fwif.h      | 236 ++++++++
>  drivers/gpu/drm/i915/intel_guc_loader.c    | 530 ++++++++++++++++++
>  drivers/gpu/drm/i915/intel_lrc.c           |  85 +--
>  drivers/gpu/drm/i915/intel_lrc.h           |  11 +
>  drivers/gpu/drm/i915/intel_uc_loader.c     | 310 +++++++++++
>  drivers/gpu/drm/i915/intel_uc_loader.h     |  92 +++
>  17 files changed, 2520 insertions(+), 41 deletions(-)
>  create mode 100644 drivers/gpu/drm/i915/i915_guc_reg.h
>  create mode 100644 drivers/gpu/drm/i915/i915_guc_submission.c
>  create mode 100644 drivers/gpu/drm/i915/intel_guc.h
>  create mode 100644 drivers/gpu/drm/i915/intel_guc_fwif.h
>  create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c
>  create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c
>  create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h
> 
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Dave Gordon]

On 06/07/15 15:14, Daniel Vetter wrote:
> On Fri, Jul 03, 2015 at 01:30:33PM +0100, Dave Gordon wrote:
>> Turn on interrupt steering to route necessary interrupts to GuC.
>>
>> Issue: VIZ-4884
>> Signed-off-by: Alex Dai <yu.dai@intel.com>
>> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
>> ---
>>   drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
>>   drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
>>   2 files changed, 60 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
>> index eefb847..9f32c6c 100644
>> --- a/drivers/gpu/drm/i915/i915_reg.h
>> +++ b/drivers/gpu/drm/i915/i915_reg.h
>> @@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
>>   #define GFX_MODE_GEN7	0x0229c
>>   #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
>>   #define   GFX_RUN_LIST_ENABLE		(1<<15)
>> +#define   GFX_INTERRUPT_STEERING	(1<<14)
>>   #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
>>   #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
>>   #define   GFX_REPLAY_MODE		(1<<11)
>>   #define   GFX_PSMI_GRANULARITY		(1<<10)
>>   #define   GFX_PPGTT_ENABLE		(1<<9)
>>
>> +#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
>> +#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
>> +#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
>> +#define   GFX_FORWARD_VBLANK_COND	(2<<5)
>> +
>>   #define VLV_DISPLAY_BASE 0x180000
>>   #define VLV_MIPI_BASE VLV_DISPLAY_BASE
>>
>> @@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
>>   #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
>>   #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
>>
>> -#define GEN8_BCS_IRQ_SHIFT 16
>>   #define GEN8_RCS_IRQ_SHIFT 0
>> -#define GEN8_VCS2_IRQ_SHIFT 16
>> +#define GEN8_BCS_IRQ_SHIFT 16
>>   #define GEN8_VCS1_IRQ_SHIFT 0
>> +#define GEN8_VCS2_IRQ_SHIFT 16
>>   #define GEN8_VECS_IRQ_SHIFT 0
>> +#define GEN8_WD_IRQ_SHIFT 16
>>
>>   #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
>>   #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
>> diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
>> index eb76c34..ef5f3d5 100644
>> --- a/drivers/gpu/drm/i915/intel_guc_loader.c
>> +++ b/drivers/gpu/drm/i915/intel_guc_loader.c
>> @@ -62,6 +62,53 @@
>>   #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
>>   MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
>>
>> +static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
>> +{
>> +	struct intel_engine_cs *ring;
>> +	int i, irqs;
>> +
>> +	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
>> +	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
>> +	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
>> +	for_each_ring(ring, dev_priv, i)
>> +		I915_WRITE(RING_MODE_GEN7(ring), irqs);
>> +
>> +	/* tell DE to send nothing to GuC */
>> +	I915_WRITE(DE_GUCRMR, ~0);
>> +
>> +	/* route all GT interrupts to the host */
>> +	I915_WRITE(GUC_BCS_RCS_IER, 0);
>> +	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
>> +	I915_WRITE(GUC_WD_VECS_IER, 0);
>> +}
>> +
>> +static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
>> +{
>> +	struct intel_engine_cs *ring;
>> +	int i, irqs;
>> +
>> +	/* tell all command streamers to forward interrupts and vblank to GuC */
>> +	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
>> +	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
>> +	for_each_ring(ring, dev_priv, i)
>> +		I915_WRITE(RING_MODE_GEN7(ring), irqs);
>> +
>> +	/* tell DE to send (all) flip_done to GuC */
>> +	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
>> +	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
>> +	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
>> +	/* Unmasked bits will cause GuC response message to be sent */
>> +	I915_WRITE(DE_GUCRMR, ~irqs);
>> +
>> +	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
>> +	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
>> +	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
>> +	/* These three registers have the same bit definitions */
>> +	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
>> +	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
>> +	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
>> +}
>> +
>>   static u32 get_gttype(struct drm_i915_private *dev_priv)
>>   {
>>   	/* XXX: GT type based on PCI device ID? field seems unused by fw */
>> @@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
>>   		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
>>   		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
>>
>> +	direct_interrupts_to_host(dev_priv);
>>   	i915_guc_submission_disable(dev);
>>
>>   	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
>> @@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
>>   		err = i915_guc_submission_enable(dev);
>>   		if (err)
>>   			goto fail;
>> +		direct_interrupts_to_guc(dev_priv);
>
> This looks like still some remnants from when guc loading could fail and
> we needed to switch between cpu and guc submission at runtime. Problem is
> that you don't own the interrupt state at this point (since interrupts are
> already enabled) and hence there could be problems with races.
>
> Simplest solution would be to just set up the interrupt routing we want
> statically at driver load time in the respective irq callbacks (looking at
> guc-or-not-guc module option). That means we bake the guc submission y/n
> decision into the driver load sequence very early, but it does simplify
> things. And that's one of the reasons I don't like runtime fallbacks.
> -Daniel

We can't send interrupts to the GuC before its firmware is ready for 
them, hence we /can't/ set it up statically. We /must/ start with all 
interrupts directed to the host, and switch them to the GuC only once 
the firmware is loaded and running. (I suppose it would be simpler if 
the GuC firmware *took* control of the interrupts when it started, 
rather than having i915 *give* the interrupts to the GuC, but that would 
leave even more opportunities for confusion).

This isn't really "runtime", this *is* still in the driver load stage, 
although rather later than irq setup.

For the important interrupts that we're switching to the GuC (i.e. 
context switch) we DO know that there aren't any in flight already, 
because we can't submit anything that would result in a context switch 
until the GuC is up & running. For others (e.g. VBLANK, FLIP_DONE) it 
doesn't matter; the GuC doesn't initiate these classes of interrupts 
anyway, and it doesn't keep track of specific occurences or try to pair 
up actions and resulting events. AFAIK, it's just listening in to those 
interrupts to see what's going on, presumably for power-management purposes.

.Dave.

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 05/15] drm/i915: GuC-specific firmware loader

[Dave Gordon]

On 06/07/15 15:28, Daniel Vetter wrote:
> On Fri, Jul 03, 2015 at 01:30:27PM +0100, Dave Gordon wrote:
>> From: Alex Dai <yu.dai@intel.com>
>>
>> This uses the common firmware loader to fetch the firmware image,
>> then loads it into the GuC's memory via a dedicated DMA engine.
>>
>> This patch is derived from GuC loading work originally done by
>> Vinit Azad and Ben Widawsky. It has been reconstructed to accord
>> with the common firmware loading mechanism by Dave Gordon as well
>> as new firmware layout etc.
>>
>> v2:
>>      Various improvements per review comments by Chris Wilson
>>
>> v3:
>>      Removed 'wait' parameter to intel_guc_ucode_load() as prefetch
>>          is no longer supported in the common firmware loader, per
>> 	Daniel Vetter's request.
>>      F/w checker callback fn now returns errno rather than bool.
>>
>> Issue: VIZ-4884
>> Signed-off-by: Alex Dai <yu.dai@intel.com>
>> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
>> ---
>>   drivers/gpu/drm/i915/Makefile           |   3 +
>>   drivers/gpu/drm/i915/i915_dma.c         |   4 +
>>   drivers/gpu/drm/i915/i915_drv.h         |  11 +
>>   drivers/gpu/drm/i915/i915_gem.c         |   8 +
>>   drivers/gpu/drm/i915/i915_reg.h         |   4 +-
>>   drivers/gpu/drm/i915/intel_guc.h        |  49 ++++
>>   drivers/gpu/drm/i915/intel_guc_loader.c | 448 ++++++++++++++++++++++++++++++++
>>   7 files changed, 526 insertions(+), 1 deletion(-)
>>   create mode 100644 drivers/gpu/drm/i915/intel_guc.h
>>   create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c
>>
>> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
>> index f1f80fc..62a8c83 100644
>> --- a/drivers/gpu/drm/i915/Makefile
>> +++ b/drivers/gpu/drm/i915/Makefile
>> @@ -42,6 +42,9 @@ i915-y += i915_cmd_parser.o \
>>   # generic ancilliary microcontroller support
>>   i915-y += intel_uc_loader.o
>>
>> +# general-purpose microcontroller (GuC) support
>> +i915-y += intel_guc_loader.o
>> +
>>   # autogenerated null render state
>>   i915-y += intel_renderstate_gen6.o \
>>   	  intel_renderstate_gen7.o \
>> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
>> index c5349fa..730d91b 100644
>> --- a/drivers/gpu/drm/i915/i915_dma.c
>> +++ b/drivers/gpu/drm/i915/i915_dma.c
>> @@ -469,6 +469,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
>>
>>   cleanup_gem:
>>   	mutex_lock(&dev->struct_mutex);
>> +	intel_guc_ucode_fini(dev);
>>   	i915_gem_cleanup_ringbuffer(dev);
>>   	i915_gem_context_fini(dev);
>>   	mutex_unlock(&dev->struct_mutex);
>> @@ -866,6 +867,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
>>
>>   	intel_uncore_init(dev);
>>
>> +	intel_guc_ucode_init(dev);
>> +
>>   	/* Load CSR Firmware for SKL */
>>   	intel_csr_ucode_init(dev);
>>
>> @@ -1117,6 +1120,7 @@ int i915_driver_unload(struct drm_device *dev)
>>   	flush_workqueue(dev_priv->wq);
>>
>>   	mutex_lock(&dev->struct_mutex);
>> +	intel_guc_ucode_fini(dev);
>>   	i915_gem_cleanup_ringbuffer(dev);
>>   	i915_gem_context_fini(dev);
>>   	mutex_unlock(&dev->struct_mutex);
>> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
>> index 9618f57..a7ccac5 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.h
>> +++ b/drivers/gpu/drm/i915/i915_drv.h
>> @@ -50,6 +50,7 @@
>>   #include <linux/intel-iommu.h>
>>   #include <linux/kref.h>
>>   #include <linux/pm_qos.h>
>> +#include "intel_guc.h"
>>
>>   /* General customization:
>>    */
>> @@ -1687,6 +1688,8 @@ struct drm_i915_private {
>>
>>   	struct i915_virtual_gpu vgpu;
>>
>> +	struct intel_guc guc;
>> +
>>   	struct intel_csr csr;
>>
>>   	/* Display CSR-related protection */
>> @@ -1931,6 +1934,11 @@ static inline struct drm_i915_private *dev_to_i915(struct device *dev)
>>   	return to_i915(dev_get_drvdata(dev));
>>   }
>>
>> +static inline struct drm_i915_private *guc_to_i915(struct intel_guc *guc)
>> +{
>> +	return container_of(guc, struct drm_i915_private, guc);
>> +}
>> +
>>   /* Iterate over initialised rings */
>>   #define for_each_ring(ring__, dev_priv__, i__) \
>>   	for ((i__) = 0; (i__) < I915_NUM_RINGS; (i__)++) \
>> @@ -2539,6 +2547,9 @@ struct drm_i915_cmd_table {
>>
>>   #define HAS_CSR(dev)	(IS_SKYLAKE(dev))
>>
>> +#define HAS_GUC_UCODE(dev)	(IS_GEN9(dev))
>> +#define HAS_GUC_SCHED(dev)	(IS_GEN9(dev))
>> +
>>   #define INTEL_PCH_DEVICE_ID_MASK		0xff00
>>   #define INTEL_PCH_IBX_DEVICE_ID_TYPE		0x3b00
>>   #define INTEL_PCH_CPT_DEVICE_ID_TYPE		0x1c00
>> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
>> index aa8f4c3..80d7890 100644
>> --- a/drivers/gpu/drm/i915/i915_gem.c
>> +++ b/drivers/gpu/drm/i915/i915_gem.c
>> @@ -5076,6 +5076,14 @@ i915_gem_init_hw(struct drm_device *dev)
>>   			goto out;
>>   	}
>>
>> +	/*
>> +	 * We can't enable contexts until all firmware is loaded; if this
>> +	 * fails, disable GuC submissions and fall back to execlist mode
>> +	 */
>> +	ret = intel_guc_ucode_load(dev);
>> +	if (ret)
>> +		i915.enable_guc_submission = false;
>
> I want an -EIO or similar here since runtime fallbacks to other modes
> really aren't great from a maintainance perspective, see my comments on
> the irq routing code.
>
> Yes we can make this work, but givin our stellar track record with keeping
> disabled features working it won't work for long. And it will impact us
> with additional constraints until we give up and rip it out again. Not
> worth it imo - if we decide to use the guc on a given platform we should
> imo require it and stick to that decision for at least as long as the
> driver is loaded. Developers can still change the option when reloading the
> driver, users won't have a chance to cause trouble.
> -Daniel

Again, this isn't really "runtime" -- we're still in the driver loading 
stage here. This is analogous to the various "sanitize" functions where 
we cross-check what options have been set and decide which to override, 
except that here we can't determine whether we're going to respect the 
default or user-specified request for GuC submission mode until we know 
whether we have valid firmware for the GuC.

At this point, we haven't submitted any batches, so the main point of 
use of this flag -- in the submission path, to switch between execlist 
and GuC modes -- has never yet been executed. So there should be no 
problem with changing the value before it's first used.

And this is a one-way switch; you (or the default config) asked for GuC 
submission, but we can't support it so we disable the option. There's no 
way to switch it back on without reloading the driver. So this /is/ the 
point at which we decide to use the GuC on a given platform and then 
stick to that decision for at least as long as the driver is loaded.

We have to support execlist mode for the foreseeable future anyway, so 
using it on a machine which (we think) ought to be GuC-capable doesn't 
add /any/ extra maintenance overhead at all.

Why break the user's machine unnecessarily? With real "end-users", 
especially those who have never used Linux before, you only get one 
chance. Sometimes I've installed Linux on a (Windows-using) friend's 
machine, and it hasn't worked first time. Then I switch to another VT, 
type some magic incantations, and 10 minutes later we have a usable 
login screen. Will they adopt Linux? Unlikely :( No matter how good it 
looks thereafter, if the machine's hardware doesn't work with the distro 
straight out of the box, they're just not going to believe it's 
something they can use. So it's very important that everything essential 
to the first-time experience works even when misconfigured -- and 
nothing is more essential than the display driver (networking and wi-fi 
are the next things that will put the user off if they don't work -- and 
they're also drivers that commonly rely on firmware blobs).

.Dave.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 05/15] drm/i915: GuC-specific firmware loader

[Daniel Vetter]

On Mon, Jul 06, 2015 at 05:37:57PM +0100, Dave Gordon wrote:
> On 06/07/15 15:28, Daniel Vetter wrote:
> >On Fri, Jul 03, 2015 at 01:30:27PM +0100, Dave Gordon wrote:
> >>From: Alex Dai <yu.dai@intel.com>
> >>
> >>This uses the common firmware loader to fetch the firmware image,
> >>then loads it into the GuC's memory via a dedicated DMA engine.
> >>
> >>This patch is derived from GuC loading work originally done by
> >>Vinit Azad and Ben Widawsky. It has been reconstructed to accord
> >>with the common firmware loading mechanism by Dave Gordon as well
> >>as new firmware layout etc.
> >>
> >>v2:
> >>     Various improvements per review comments by Chris Wilson
> >>
> >>v3:
> >>     Removed 'wait' parameter to intel_guc_ucode_load() as prefetch
> >>         is no longer supported in the common firmware loader, per
> >>	Daniel Vetter's request.
> >>     F/w checker callback fn now returns errno rather than bool.
> >>
> >>Issue: VIZ-4884
> >>Signed-off-by: Alex Dai <yu.dai@intel.com>
> >>Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> >>---
> >>  drivers/gpu/drm/i915/Makefile           |   3 +
> >>  drivers/gpu/drm/i915/i915_dma.c         |   4 +
> >>  drivers/gpu/drm/i915/i915_drv.h         |  11 +
> >>  drivers/gpu/drm/i915/i915_gem.c         |   8 +
> >>  drivers/gpu/drm/i915/i915_reg.h         |   4 +-
> >>  drivers/gpu/drm/i915/intel_guc.h        |  49 ++++
> >>  drivers/gpu/drm/i915/intel_guc_loader.c | 448 ++++++++++++++++++++++++++++++++
> >>  7 files changed, 526 insertions(+), 1 deletion(-)
> >>  create mode 100644 drivers/gpu/drm/i915/intel_guc.h
> >>  create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c
> >>
> >>diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> >>index f1f80fc..62a8c83 100644
> >>--- a/drivers/gpu/drm/i915/Makefile
> >>+++ b/drivers/gpu/drm/i915/Makefile
> >>@@ -42,6 +42,9 @@ i915-y += i915_cmd_parser.o \
> >>  # generic ancilliary microcontroller support
> >>  i915-y += intel_uc_loader.o
> >>
> >>+# general-purpose microcontroller (GuC) support
> >>+i915-y += intel_guc_loader.o
> >>+
> >>  # autogenerated null render state
> >>  i915-y += intel_renderstate_gen6.o \
> >>  	  intel_renderstate_gen7.o \
> >>diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> >>index c5349fa..730d91b 100644
> >>--- a/drivers/gpu/drm/i915/i915_dma.c
> >>+++ b/drivers/gpu/drm/i915/i915_dma.c
> >>@@ -469,6 +469,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
> >>
> >>  cleanup_gem:
> >>  	mutex_lock(&dev->struct_mutex);
> >>+	intel_guc_ucode_fini(dev);
> >>  	i915_gem_cleanup_ringbuffer(dev);
> >>  	i915_gem_context_fini(dev);
> >>  	mutex_unlock(&dev->struct_mutex);
> >>@@ -866,6 +867,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
> >>
> >>  	intel_uncore_init(dev);
> >>
> >>+	intel_guc_ucode_init(dev);
> >>+
> >>  	/* Load CSR Firmware for SKL */
> >>  	intel_csr_ucode_init(dev);
> >>
> >>@@ -1117,6 +1120,7 @@ int i915_driver_unload(struct drm_device *dev)
> >>  	flush_workqueue(dev_priv->wq);
> >>
> >>  	mutex_lock(&dev->struct_mutex);
> >>+	intel_guc_ucode_fini(dev);
> >>  	i915_gem_cleanup_ringbuffer(dev);
> >>  	i915_gem_context_fini(dev);
> >>  	mutex_unlock(&dev->struct_mutex);
> >>diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> >>index 9618f57..a7ccac5 100644
> >>--- a/drivers/gpu/drm/i915/i915_drv.h
> >>+++ b/drivers/gpu/drm/i915/i915_drv.h
> >>@@ -50,6 +50,7 @@
> >>  #include <linux/intel-iommu.h>
> >>  #include <linux/kref.h>
> >>  #include <linux/pm_qos.h>
> >>+#include "intel_guc.h"
> >>
> >>  /* General customization:
> >>   */
> >>@@ -1687,6 +1688,8 @@ struct drm_i915_private {
> >>
> >>  	struct i915_virtual_gpu vgpu;
> >>
> >>+	struct intel_guc guc;
> >>+
> >>  	struct intel_csr csr;
> >>
> >>  	/* Display CSR-related protection */
> >>@@ -1931,6 +1934,11 @@ static inline struct drm_i915_private *dev_to_i915(struct device *dev)
> >>  	return to_i915(dev_get_drvdata(dev));
> >>  }
> >>
> >>+static inline struct drm_i915_private *guc_to_i915(struct intel_guc *guc)
> >>+{
> >>+	return container_of(guc, struct drm_i915_private, guc);
> >>+}
> >>+
> >>  /* Iterate over initialised rings */
> >>  #define for_each_ring(ring__, dev_priv__, i__) \
> >>  	for ((i__) = 0; (i__) < I915_NUM_RINGS; (i__)++) \
> >>@@ -2539,6 +2547,9 @@ struct drm_i915_cmd_table {
> >>
> >>  #define HAS_CSR(dev)	(IS_SKYLAKE(dev))
> >>
> >>+#define HAS_GUC_UCODE(dev)	(IS_GEN9(dev))
> >>+#define HAS_GUC_SCHED(dev)	(IS_GEN9(dev))
> >>+
> >>  #define INTEL_PCH_DEVICE_ID_MASK		0xff00
> >>  #define INTEL_PCH_IBX_DEVICE_ID_TYPE		0x3b00
> >>  #define INTEL_PCH_CPT_DEVICE_ID_TYPE		0x1c00
> >>diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> >>index aa8f4c3..80d7890 100644
> >>--- a/drivers/gpu/drm/i915/i915_gem.c
> >>+++ b/drivers/gpu/drm/i915/i915_gem.c
> >>@@ -5076,6 +5076,14 @@ i915_gem_init_hw(struct drm_device *dev)
> >>  			goto out;
> >>  	}
> >>
> >>+	/*
> >>+	 * We can't enable contexts until all firmware is loaded; if this
> >>+	 * fails, disable GuC submissions and fall back to execlist mode
> >>+	 */
> >>+	ret = intel_guc_ucode_load(dev);
> >>+	if (ret)
> >>+		i915.enable_guc_submission = false;
> >
> >I want an -EIO or similar here since runtime fallbacks to other modes
> >really aren't great from a maintainance perspective, see my comments on
> >the irq routing code.
> >
> >Yes we can make this work, but givin our stellar track record with keeping
> >disabled features working it won't work for long. And it will impact us
> >with additional constraints until we give up and rip it out again. Not
> >worth it imo - if we decide to use the guc on a given platform we should
> >imo require it and stick to that decision for at least as long as the
> >driver is loaded. Developers can still change the option when reloading the
> >driver, users won't have a chance to cause trouble.
> >-Daniel
> 
> Again, this isn't really "runtime" -- we're still in the driver loading
> stage here. This is analogous to the various "sanitize" functions where we
> cross-check what options have been set and decide which to override, except
> that here we can't determine whether we're going to respect the default or
> user-specified request for GuC submission mode until we know whether we have
> valid firmware for the GuC.

The problem is that a pile of code has run already (specifically irq
setup). Not a problem from the correctness pov it can all be made to work,
but from a longer-term maintainance pov. It'll just bitrot until it's
ripped out. But until that happens everyone has to keep it in mind and
try not to break this fallback. Too much trouble imo for no real benefit.

> At this point, we haven't submitted any batches, so the main point of use of
> this flag -- in the submission path, to switch between execlist and GuC
> modes -- has never yet been executed. So there should be no problem with
> changing the value before it's first used.
> 
> And this is a one-way switch; you (or the default config) asked for GuC
> submission, but we can't support it so we disable the option. There's no way
> to switch it back on without reloading the driver. So this /is/ the point at
> which we decide to use the GuC on a given platform and then stick to that
> decision for at least as long as the driver is loaded.
> 
> We have to support execlist mode for the foreseeable future anyway, so using
> it on a machine which (we think) ought to be GuC-capable doesn't add /any/
> extra maintenance overhead at all.
> 
> Why break the user's machine unnecessarily? With real "end-users",
> especially those who have never used Linux before, you only get one chance.
> Sometimes I've installed Linux on a (Windows-using) friend's machine, and it
> hasn't worked first time. Then I switch to another VT, type some magic
> incantations, and 10 minutes later we have a usable login screen. Will they
> adopt Linux? Unlikely :( No matter how good it looks thereafter, if the
> machine's hardware doesn't work with the distro straight out of the box,
> they're just not going to believe it's something they can use. So it's very
> important that everything essential to the first-time experience works even
> when misconfigured -- and nothing is more essential than the display driver
> (networking and wi-fi are the next things that will put the user off if they
> don't work -- and they're also drivers that commonly rely on firmware
> blobs).

Those end-users install fedora or ubuntu which get the firmware blob
loading just right. None of the big other drivers bother with falling back
to some other mode if the firmware they need isn't there: radeon just
outright bails out, nouveau just disables accel or runtime pm if the
firmware blob is only needed for these features.

The only user I can see are people allergic to firmware blobs (they can
use execlist and we don't care) or developers (can change mod options
too). Misconfigured systems from newbies is not a target market, neither
for intel nor for upstream (yes we WONTFIX bugs of people with funky
configs if the problem doesn't exist on a properly configured distro).
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Daniel Vetter]

On Mon, Jul 06, 2015 at 05:07:31PM +0100, Dave Gordon wrote:
> On 06/07/15 15:14, Daniel Vetter wrote:
> >On Fri, Jul 03, 2015 at 01:30:33PM +0100, Dave Gordon wrote:
> >>Turn on interrupt steering to route necessary interrupts to GuC.
> >>
> >>Issue: VIZ-4884
> >>Signed-off-by: Alex Dai <yu.dai@intel.com>
> >>Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> >>---
> >>  drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
> >>  drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
> >>  2 files changed, 60 insertions(+), 2 deletions(-)
> >>
> >>diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> >>index eefb847..9f32c6c 100644
> >>--- a/drivers/gpu/drm/i915/i915_reg.h
> >>+++ b/drivers/gpu/drm/i915/i915_reg.h
> >>@@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
> >>  #define GFX_MODE_GEN7	0x0229c
> >>  #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
> >>  #define   GFX_RUN_LIST_ENABLE		(1<<15)
> >>+#define   GFX_INTERRUPT_STEERING	(1<<14)
> >>  #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
> >>  #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
> >>  #define   GFX_REPLAY_MODE		(1<<11)
> >>  #define   GFX_PSMI_GRANULARITY		(1<<10)
> >>  #define   GFX_PPGTT_ENABLE		(1<<9)
> >>
> >>+#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
> >>+#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
> >>+#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
> >>+#define   GFX_FORWARD_VBLANK_COND	(2<<5)
> >>+
> >>  #define VLV_DISPLAY_BASE 0x180000
> >>  #define VLV_MIPI_BASE VLV_DISPLAY_BASE
> >>
> >>@@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
> >>  #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
> >>  #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
> >>
> >>-#define GEN8_BCS_IRQ_SHIFT 16
> >>  #define GEN8_RCS_IRQ_SHIFT 0
> >>-#define GEN8_VCS2_IRQ_SHIFT 16
> >>+#define GEN8_BCS_IRQ_SHIFT 16
> >>  #define GEN8_VCS1_IRQ_SHIFT 0
> >>+#define GEN8_VCS2_IRQ_SHIFT 16
> >>  #define GEN8_VECS_IRQ_SHIFT 0
> >>+#define GEN8_WD_IRQ_SHIFT 16
> >>
> >>  #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
> >>  #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
> >>diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> >>index eb76c34..ef5f3d5 100644
> >>--- a/drivers/gpu/drm/i915/intel_guc_loader.c
> >>+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> >>@@ -62,6 +62,53 @@
> >>  #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
> >>  MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
> >>
> >>+static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
> >>+{
> >>+	struct intel_engine_cs *ring;
> >>+	int i, irqs;
> >>+
> >>+	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
> >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
> >>+	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
> >>+	for_each_ring(ring, dev_priv, i)
> >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> >>+
> >>+	/* tell DE to send nothing to GuC */
> >>+	I915_WRITE(DE_GUCRMR, ~0);
> >>+
> >>+	/* route all GT interrupts to the host */
> >>+	I915_WRITE(GUC_BCS_RCS_IER, 0);
> >>+	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
> >>+	I915_WRITE(GUC_WD_VECS_IER, 0);
> >>+}
> >>+
> >>+static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
> >>+{
> >>+	struct intel_engine_cs *ring;
> >>+	int i, irqs;
> >>+
> >>+	/* tell all command streamers to forward interrupts and vblank to GuC */
> >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
> >>+	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
> >>+	for_each_ring(ring, dev_priv, i)
> >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> >>+
> >>+	/* tell DE to send (all) flip_done to GuC */
> >>+	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
> >>+	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
> >>+	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
> >>+	/* Unmasked bits will cause GuC response message to be sent */
> >>+	I915_WRITE(DE_GUCRMR, ~irqs);
> >>+
> >>+	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
> >>+	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
> >>+	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
> >>+	/* These three registers have the same bit definitions */
> >>+	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
> >>+	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
> >>+	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
> >>+}
> >>+
> >>  static u32 get_gttype(struct drm_i915_private *dev_priv)
> >>  {
> >>  	/* XXX: GT type based on PCI device ID? field seems unused by fw */
> >>@@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
> >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
> >>
> >>+	direct_interrupts_to_host(dev_priv);
> >>  	i915_guc_submission_disable(dev);
> >>
> >>  	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
> >>@@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> >>  		err = i915_guc_submission_enable(dev);
> >>  		if (err)
> >>  			goto fail;
> >>+		direct_interrupts_to_guc(dev_priv);
> >
> >This looks like still some remnants from when guc loading could fail and
> >we needed to switch between cpu and guc submission at runtime. Problem is
> >that you don't own the interrupt state at this point (since interrupts are
> >already enabled) and hence there could be problems with races.
> >
> >Simplest solution would be to just set up the interrupt routing we want
> >statically at driver load time in the respective irq callbacks (looking at
> >guc-or-not-guc module option). That means we bake the guc submission y/n
> >decision into the driver load sequence very early, but it does simplify
> >things. And that's one of the reasons I don't like runtime fallbacks.
> >-Daniel
> 
> We can't send interrupts to the GuC before its firmware is ready for them,
> hence we /can't/ set it up statically. We /must/ start with all interrupts
> directed to the host, and switch them to the GuC only once the firmware is
> loaded and running. (I suppose it would be simpler if the GuC firmware
> *took* control of the interrupts when it started, rather than having i915
> *give* the interrupts to the GuC, but that would leave even more
> opportunities for confusion).
> 
> This isn't really "runtime", this *is* still in the driver load stage,
> although rather later than irq setup.
> 
> For the important interrupts that we're switching to the GuC (i.e. context
> switch) we DO know that there aren't any in flight already, because we can't
> submit anything that would result in a context switch until the GuC is up &
> running. For others (e.g. VBLANK, FLIP_DONE) it doesn't matter; the GuC
> doesn't initiate these classes of interrupts anyway, and it doesn't keep
> track of specific occurences or try to pair up actions and resulting events.
> AFAIK, it's just listening in to those interrupts to see what's going on,
> presumably for power-management purposes.

Yeah I misread the patch, somehow I thought that we are touching some
cpu-side irq control registers too. That was my concern since Paulo spent
an awful lot of time unifying the setup for those to make them mesh well
with runtime pm. But this is indeed all just guc private register state.
One exception is the RING_MODE registers, but since that's a masked
bitfield and this code is the only bit that touches it we should be fine
too.

One thing aside: Will the guc receive copies of these (i.e. will the cpu
still be able to mask/receive/ack/clear/whatever them on its own) or is
this somewhat exclusive and it's only either guc or cpu that gets certain
interrupts? If it's just an enable/disable not affecting the cpu side
really I'd suggest we call them enable/disable (more in line with other
irq handling code) instead of the direct_interrupts_to_*. But that's
really just a bikeshed to avoid confusion here.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 02/15] drm/i915: Embedded microcontroller (uC) firmware loading support

[Dave Gordon]

On 06/07/15 15:06, Daniel Vetter wrote:
> On Fri, Jul 03, 2015 at 01:30:24PM +0100, Dave Gordon wrote:
>> Current devices may contain one or more programmable microcontrollers
>> that need to have a firmware image (aka "binary blob") loaded from an
>> external medium and transferred to the device's memory.
>>
>> This file provides common support functions for doing this; they can
>> then be used by each uC-specific loader, thus reducing code duplication
>> and testing effort.
>>
>> v2:
>>      Local functions should pass dev_priv rather than dev [Chris Wilson]
>>      Various other improvements per Chris Wilson's review comments
>>
>> v3:
>>      Defeatured version without async prefetch [Daniel Vetter]
>>
>> Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
>> Signed-off-by: Alex Dai <yu.dai@intel.com>
>> ---
>>   drivers/gpu/drm/i915/Makefile          |   3 +
>>   drivers/gpu/drm/i915/intel_uc_loader.c | 310 +++++++++++++++++++++++++++++++++
>>   drivers/gpu/drm/i915/intel_uc_loader.h |  92 ++++++++++
>>   3 files changed, 405 insertions(+)
>>   create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c
>>   create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h
>>
>> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
>> index de21965..f1f80fc 100644
>> --- a/drivers/gpu/drm/i915/Makefile
>> +++ b/drivers/gpu/drm/i915/Makefile
>> @@ -39,6 +39,9 @@ i915-y += i915_cmd_parser.o \
>>   	  intel_ringbuffer.o \
>>   	  intel_uncore.o
>>
>> +# generic ancilliary microcontroller support
>> +i915-y += intel_uc_loader.o
>> +
>>   # autogenerated null render state
>>   i915-y += intel_renderstate_gen6.o \
>>   	  intel_renderstate_gen7.o \
>> diff --git a/drivers/gpu/drm/i915/intel_uc_loader.c b/drivers/gpu/drm/i915/intel_uc_loader.c
>> new file mode 100644
>> index 0000000..a8fc1dd
>> --- /dev/null
>> +++ b/drivers/gpu/drm/i915/intel_uc_loader.c
>> @@ -0,0 +1,310 @@
>> +/*
>> + * Copyright © 2014 Intel Corporation
>> + *
>> + * Permission is hereby granted, free of charge, to any person obtaining a
>> + * copy of this software and associated documentation files (the "Software"),
>> + * to deal in the Software without restriction, including without limitation
>> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
>> + * and/or sell copies of the Software, and to permit persons to whom the
>> + * Software is furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice (including the next
>> + * paragraph) shall be included in all copies or substantial portions of the
>> + * Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
>> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
>> + * IN THE SOFTWARE.
>> + *
>> + * Author:
>> + *	Dave Gordon <david.s.gordon@intel.com>
>> + */
>> +#include <linux/firmware.h>
>> +#include "i915_drv.h"
>> +#include "intel_uc_loader.h"
>> +
>> +/**
>> + * DOC: Common functions for embedded microcontroller (uC) firmware loading
>> + *
>> + * The functions in this file provide common support code for loading the
>> + * firmware that may be required by an embedded microcontroller (uC).
>> + *
>> + * The function intel_uc_fw_init() should be called first; it requires no
>> + * locking, and can be called even before GEM has been initialised. It just
>> + * initialises the tracking data and stores its parameters for the subsequent
>> + * call to intel_uc_fw_fetch().
>> + *
>> + * At some convenient point after GEM initialisation, the driver should call
>> + * intel_uc_fw_fetch(). The first time, this will use the Linux kernel's
>> + * request_firmware() call to open and read the firmware image into memory.
>> + * (On subsequent calls, this is skipped, as either the firmware has already
>> + * been fetched into memory, or it is already known that no valid firmware
>> + * image could be found).
>> + *
>> + * The callback() function passed to intel_uc_fw_fetch() can further check
>> + * the firmware image before it is saved. This function can reject the image
>> + * by returning a negative error code (e.g. -ENOEXEC), or accept it. In the
>> + * latter case, it can return INTEL_UC_FW_GOOD (which is also the default if
>> + * no callback() is supplied), and the common code here will save the whole
>> + * of the firmware image in a (pageable, shmfs-backed) GEM object.
>> + *
>> + * If saving the whole image unmodified is not appropriate (for example, if
>> + * only a small part of the image is needed later, or the data needs to be
>> + * reorganised before saving), the callback() function can instead make its
>> + * own arrangements for saving the required data in a GEM object or otherwise
>> + * and then return INTEL_UC_FW_SAVED.
>> + *
>> + * (If such a callback does stash (some of) the image data in a GEM object,
>> + * it can use the uc_fw_obj and uc_fw_size fields; the common framework will
>> + * then handle deallocating the object on failure or finalisation. Otherwise
>> + * any allocated resources will have to be managed by the uC-specific code).
>> + *
>> + * After a successful call to intel_uc_fw_fetch(), the uC-specific code can
>> + * transfer the data in the GEM object (or its own alternative) to the uC's
>> + * memory (in some uC-specific way, not handled here).
>> + *
>> + * During driver shutdown, or if driver load is aborted, intel_uc_fw_fini()
>> + * should be called to release any remaining resources.
>> + */
>> +
>> +/* User-friendly representation of an enum */
>> +const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status)
>> +{
>> +	switch (status) {
>> +	case INTEL_UC_FIRMWARE_FAIL:
>> +		return "FAIL";
>> +	case INTEL_UC_FIRMWARE_NONE:
>> +		return "NONE";
>> +	case INTEL_UC_FIRMWARE_PENDING:
>> +		return "PENDING";
>> +	case INTEL_UC_FIRMWARE_SUCCESS:
>> +		return "SUCCESS";
>> +	default:
>> +		return "UNKNOWN!";
>> +	}
>> +};
>> +
>> +/*
>> + * Called once per uC, late in driver initialisation, after GEM is set up.
>> + * First, we fetch the firmware image using request_firmware(), then allow
>> + * the optional callback() function to check it. If it's good, and callback()
>> + * doesn't say it's already saved the image, we will save it here by copying
>> + * the whole thing into a (pageable, shmfs-backed) GEM object,
>> + */
>> +static void
>> +uc_fw_fetch(struct intel_uc_fw *uc_fw,
>> +	    int callback(struct intel_uc_fw *))
>> +{
>> +	struct drm_device *dev = uc_fw->uc_dev;
>> +	struct drm_i915_gem_object *obj;
>> +	const struct firmware *fw;
>> +	int cb_status = INTEL_UC_FW_GOOD;
>> +
>> +	DRM_DEBUG_DRIVER("before waiting: %s fw fetch status %s, fw %p\n",
>> +		uc_fw->uc_name,
>> +		intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status),
>> +		uc_fw->uc_fw_blob);
>> +
>> +	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
>> +	WARN_ON(uc_fw->uc_fw_fetch_status != INTEL_UC_FIRMWARE_PENDING);
>> +
>> +	if (request_firmware(&fw, uc_fw->uc_fw_path, &dev->pdev->dev))
>> +		goto fail;
>> +	if (!fw)
>> +		goto fail;
>> +
>> +	DRM_DEBUG_DRIVER("fetch %s fw from %s succeeded, fw %p\n",
>> +		uc_fw->uc_name, uc_fw->uc_fw_path, fw);
>> +	uc_fw->uc_fw_blob = fw;
>> +
>> +	/* Callback to the optional uC-specific function, if supplied */
>> +	if (callback)
>> +		cb_status = callback(uc_fw);
>> +	if (cb_status < 0)
>> +		goto fail;
>> +	switch (cb_status) {
>> +	default:
>> +		WARN(1, "Invalid status %d from fw checking callback %pf\n",
>> +				cb_status, callback);
>> +		goto fail;
>> +
>> +	case INTEL_UC_FW_SAVED:
>> +		// Good, already saved, nothing to do
>> +		break;
>> +
>> +	case INTEL_UC_FW_GOOD:
>> +		// Good, save in GEM object
>> +		obj = i915_gem_object_create_from_data(dev, fw->data, fw->size);
>> +		if (!obj)
>> +			goto fail;
>> +
>> +		uc_fw->uc_fw_obj = obj;
>> +		uc_fw->uc_fw_size = fw->size;
>> +	}
>> +
>> +	DRM_DEBUG_DRIVER("%s fw fetch status SUCCESS, cb %d, obj %p\n",
>> +			uc_fw->uc_name, cb_status, uc_fw->uc_fw_obj);
>> +
>> +	release_firmware(fw);
>> +	uc_fw->uc_fw_blob = NULL;
>> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_SUCCESS;
>> +	return;
>> +
>> +fail:
>> +	DRM_DEBUG_DRIVER("%s fw fetch status FAIL; fw %p, obj %p\n",
>> +		uc_fw->uc_name, fw, uc_fw->uc_fw_obj);
>> +	DRM_ERROR("Failed to fetch %s firmware from %s\n",
>> +		  uc_fw->uc_name, uc_fw->uc_fw_path);
>> +
>> +	obj = uc_fw->uc_fw_obj;
>> +	if (obj)
>> +		drm_gem_object_unreference(&obj->base);
>> +	uc_fw->uc_fw_obj = NULL;
>> +
>> +	release_firmware(fw);		/* OK even if fw is NULL */
>> +	uc_fw->uc_fw_blob = NULL;
>> +	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_FAIL;
>> +}
>
> So looking at this the entire library now boils down to a glorified
> version of
>
> 	if (request_firmware != 0) {
> 		DRM_DEBUG_GEM("guc firmware loading failed\n");
> 		return fail;
> 	}
>
> 	gem_obj_create_from_date(fw)
> 	relase_firmware
>
> Can I have that inlined (or as a simple static helper function) into the
> callsite please? Yes that unfortunately means I ask you to can the really
> great kerneldoc you've done (I'd really like all patches to look this
> great wrt docs), but I also think that the abstraction really doesn't pay
> off. Firmware-loading using request_firmware is not something tricky
> enough to justify added abstraction. Yes there was the
> synchronization/completion stuff in there before, but as explained in the
> previous discussion I don't expect us to be able to share that part at
> all, and I'd like any kind of synchronization to be open-coded using
> normal kernel primitives (like flush_work or similar) to make it really
> stick out to reviewers.
>
> Thanks, Daniel

Well, I /could/ squash these functions into the GuC-specific loader and 
put a "g" on the front of all the names, but that would just result in 
the whole lot getting copypasted when we come to do the next letter of 
the alphabet :(

Or worse still, someone else would write a completely new version from 
scratch, which would be a waste of effort both for them and for those 
who would then have to review the new code, spot the similarities and 
differences, and ask for it all to be rewritten to be more like the 
by-then-existing GuC code.

It's also not quite as simple as your paraphrase; even without the 
prefetch code, it also deals with the do-this-once logic, and correctly 
freeing resources in all error paths; not enormously complicated but why 
repeat similar code in several places when you can just share the one 
instance (and there /will/ be other users for this code). And we may 
well add other common functionality here, should it turn out useful for 
more than one uC. You're just asking to increase the maintenance effort 
by duplicating the loader code rather than sharing it :(

Conversely, if you think it adds little functionality, it also adds 
little overhead. I wouldn't much want to inline the currently-generic 
functions into their callers also because that would bloat them too 
much; at the moment, with the split between GuC-specific and common 
functions, the biggest functions turn out to be about one screenful. If 
we "open-code" it all we end up with monstrous 300-line-plus functions 
with any number of error paths, which is just not nice.

So I'd really prefer to keep the utility code and the GuC-specific stuff 
tidily separated even before we have the other potential users of this 
support code in upstream.

.Dave.

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 02/15] drm/i915: Embedded microcontroller (uC) firmware loading support

[Daniel Vetter]

On Mon, Jul 06, 2015 at 07:24:37PM +0100, Dave Gordon wrote:
> On 06/07/15 15:06, Daniel Vetter wrote:
> >On Fri, Jul 03, 2015 at 01:30:24PM +0100, Dave Gordon wrote:
> >>Current devices may contain one or more programmable microcontrollers
> >>that need to have a firmware image (aka "binary blob") loaded from an
> >>external medium and transferred to the device's memory.
> >>
> >>This file provides common support functions for doing this; they can
> >>then be used by each uC-specific loader, thus reducing code duplication
> >>and testing effort.
> >>
> >>v2:
> >>     Local functions should pass dev_priv rather than dev [Chris Wilson]
> >>     Various other improvements per Chris Wilson's review comments
> >>
> >>v3:
> >>     Defeatured version without async prefetch [Daniel Vetter]
> >>
> >>Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> >>Signed-off-by: Alex Dai <yu.dai@intel.com>
> >>---
> >>  drivers/gpu/drm/i915/Makefile          |   3 +
> >>  drivers/gpu/drm/i915/intel_uc_loader.c | 310 +++++++++++++++++++++++++++++++++
> >>  drivers/gpu/drm/i915/intel_uc_loader.h |  92 ++++++++++
> >>  3 files changed, 405 insertions(+)
> >>  create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c
> >>  create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h
> >>
> >>diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> >>index de21965..f1f80fc 100644
> >>--- a/drivers/gpu/drm/i915/Makefile
> >>+++ b/drivers/gpu/drm/i915/Makefile
> >>@@ -39,6 +39,9 @@ i915-y += i915_cmd_parser.o \
> >>  	  intel_ringbuffer.o \
> >>  	  intel_uncore.o
> >>
> >>+# generic ancilliary microcontroller support
> >>+i915-y += intel_uc_loader.o
> >>+
> >>  # autogenerated null render state
> >>  i915-y += intel_renderstate_gen6.o \
> >>  	  intel_renderstate_gen7.o \
> >>diff --git a/drivers/gpu/drm/i915/intel_uc_loader.c b/drivers/gpu/drm/i915/intel_uc_loader.c
> >>new file mode 100644
> >>index 0000000..a8fc1dd
> >>--- /dev/null
> >>+++ b/drivers/gpu/drm/i915/intel_uc_loader.c
> >>@@ -0,0 +1,310 @@
> >>+/*
> >>+ * Copyright © 2014 Intel Corporation
> >>+ *
> >>+ * Permission is hereby granted, free of charge, to any person obtaining a
> >>+ * copy of this software and associated documentation files (the "Software"),
> >>+ * to deal in the Software without restriction, including without limitation
> >>+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> >>+ * and/or sell copies of the Software, and to permit persons to whom the
> >>+ * Software is furnished to do so, subject to the following conditions:
> >>+ *
> >>+ * The above copyright notice and this permission notice (including the next
> >>+ * paragraph) shall be included in all copies or substantial portions of the
> >>+ * Software.
> >>+ *
> >>+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> >>+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> >>+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> >>+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> >>+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> >>+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> >>+ * IN THE SOFTWARE.
> >>+ *
> >>+ * Author:
> >>+ *	Dave Gordon <david.s.gordon@intel.com>
> >>+ */
> >>+#include <linux/firmware.h>
> >>+#include "i915_drv.h"
> >>+#include "intel_uc_loader.h"
> >>+
> >>+/**
> >>+ * DOC: Common functions for embedded microcontroller (uC) firmware loading
> >>+ *
> >>+ * The functions in this file provide common support code for loading the
> >>+ * firmware that may be required by an embedded microcontroller (uC).
> >>+ *
> >>+ * The function intel_uc_fw_init() should be called first; it requires no
> >>+ * locking, and can be called even before GEM has been initialised. It just
> >>+ * initialises the tracking data and stores its parameters for the subsequent
> >>+ * call to intel_uc_fw_fetch().
> >>+ *
> >>+ * At some convenient point after GEM initialisation, the driver should call
> >>+ * intel_uc_fw_fetch(). The first time, this will use the Linux kernel's
> >>+ * request_firmware() call to open and read the firmware image into memory.
> >>+ * (On subsequent calls, this is skipped, as either the firmware has already
> >>+ * been fetched into memory, or it is already known that no valid firmware
> >>+ * image could be found).
> >>+ *
> >>+ * The callback() function passed to intel_uc_fw_fetch() can further check
> >>+ * the firmware image before it is saved. This function can reject the image
> >>+ * by returning a negative error code (e.g. -ENOEXEC), or accept it. In the
> >>+ * latter case, it can return INTEL_UC_FW_GOOD (which is also the default if
> >>+ * no callback() is supplied), and the common code here will save the whole
> >>+ * of the firmware image in a (pageable, shmfs-backed) GEM object.
> >>+ *
> >>+ * If saving the whole image unmodified is not appropriate (for example, if
> >>+ * only a small part of the image is needed later, or the data needs to be
> >>+ * reorganised before saving), the callback() function can instead make its
> >>+ * own arrangements for saving the required data in a GEM object or otherwise
> >>+ * and then return INTEL_UC_FW_SAVED.
> >>+ *
> >>+ * (If such a callback does stash (some of) the image data in a GEM object,
> >>+ * it can use the uc_fw_obj and uc_fw_size fields; the common framework will
> >>+ * then handle deallocating the object on failure or finalisation. Otherwise
> >>+ * any allocated resources will have to be managed by the uC-specific code).
> >>+ *
> >>+ * After a successful call to intel_uc_fw_fetch(), the uC-specific code can
> >>+ * transfer the data in the GEM object (or its own alternative) to the uC's
> >>+ * memory (in some uC-specific way, not handled here).
> >>+ *
> >>+ * During driver shutdown, or if driver load is aborted, intel_uc_fw_fini()
> >>+ * should be called to release any remaining resources.
> >>+ */
> >>+
> >>+/* User-friendly representation of an enum */
> >>+const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status)
> >>+{
> >>+	switch (status) {
> >>+	case INTEL_UC_FIRMWARE_FAIL:
> >>+		return "FAIL";
> >>+	case INTEL_UC_FIRMWARE_NONE:
> >>+		return "NONE";
> >>+	case INTEL_UC_FIRMWARE_PENDING:
> >>+		return "PENDING";
> >>+	case INTEL_UC_FIRMWARE_SUCCESS:
> >>+		return "SUCCESS";
> >>+	default:
> >>+		return "UNKNOWN!";
> >>+	}
> >>+};
> >>+
> >>+/*
> >>+ * Called once per uC, late in driver initialisation, after GEM is set up.
> >>+ * First, we fetch the firmware image using request_firmware(), then allow
> >>+ * the optional callback() function to check it. If it's good, and callback()
> >>+ * doesn't say it's already saved the image, we will save it here by copying
> >>+ * the whole thing into a (pageable, shmfs-backed) GEM object,
> >>+ */
> >>+static void
> >>+uc_fw_fetch(struct intel_uc_fw *uc_fw,
> >>+	    int callback(struct intel_uc_fw *))
> >>+{
> >>+	struct drm_device *dev = uc_fw->uc_dev;
> >>+	struct drm_i915_gem_object *obj;
> >>+	const struct firmware *fw;
> >>+	int cb_status = INTEL_UC_FW_GOOD;
> >>+
> >>+	DRM_DEBUG_DRIVER("before waiting: %s fw fetch status %s, fw %p\n",
> >>+		uc_fw->uc_name,
> >>+		intel_uc_fw_status_repr(uc_fw->uc_fw_fetch_status),
> >>+		uc_fw->uc_fw_blob);
> >>+
> >>+	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
> >>+	WARN_ON(uc_fw->uc_fw_fetch_status != INTEL_UC_FIRMWARE_PENDING);
> >>+
> >>+	if (request_firmware(&fw, uc_fw->uc_fw_path, &dev->pdev->dev))
> >>+		goto fail;
> >>+	if (!fw)
> >>+		goto fail;
> >>+
> >>+	DRM_DEBUG_DRIVER("fetch %s fw from %s succeeded, fw %p\n",
> >>+		uc_fw->uc_name, uc_fw->uc_fw_path, fw);
> >>+	uc_fw->uc_fw_blob = fw;
> >>+
> >>+	/* Callback to the optional uC-specific function, if supplied */
> >>+	if (callback)
> >>+		cb_status = callback(uc_fw);
> >>+	if (cb_status < 0)
> >>+		goto fail;
> >>+	switch (cb_status) {
> >>+	default:
> >>+		WARN(1, "Invalid status %d from fw checking callback %pf\n",
> >>+				cb_status, callback);
> >>+		goto fail;
> >>+
> >>+	case INTEL_UC_FW_SAVED:
> >>+		// Good, already saved, nothing to do
> >>+		break;
> >>+
> >>+	case INTEL_UC_FW_GOOD:
> >>+		// Good, save in GEM object
> >>+		obj = i915_gem_object_create_from_data(dev, fw->data, fw->size);
> >>+		if (!obj)
> >>+			goto fail;
> >>+
> >>+		uc_fw->uc_fw_obj = obj;
> >>+		uc_fw->uc_fw_size = fw->size;
> >>+	}
> >>+
> >>+	DRM_DEBUG_DRIVER("%s fw fetch status SUCCESS, cb %d, obj %p\n",
> >>+			uc_fw->uc_name, cb_status, uc_fw->uc_fw_obj);
> >>+
> >>+	release_firmware(fw);
> >>+	uc_fw->uc_fw_blob = NULL;
> >>+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_SUCCESS;
> >>+	return;
> >>+
> >>+fail:
> >>+	DRM_DEBUG_DRIVER("%s fw fetch status FAIL; fw %p, obj %p\n",
> >>+		uc_fw->uc_name, fw, uc_fw->uc_fw_obj);
> >>+	DRM_ERROR("Failed to fetch %s firmware from %s\n",
> >>+		  uc_fw->uc_name, uc_fw->uc_fw_path);
> >>+
> >>+	obj = uc_fw->uc_fw_obj;
> >>+	if (obj)
> >>+		drm_gem_object_unreference(&obj->base);
> >>+	uc_fw->uc_fw_obj = NULL;
> >>+
> >>+	release_firmware(fw);		/* OK even if fw is NULL */
> >>+	uc_fw->uc_fw_blob = NULL;
> >>+	uc_fw->uc_fw_fetch_status = INTEL_UC_FIRMWARE_FAIL;
> >>+}
> >
> >So looking at this the entire library now boils down to a glorified
> >version of
> >
> >	if (request_firmware != 0) {
> >		DRM_DEBUG_GEM("guc firmware loading failed\n");
> >		return fail;
> >	}
> >
> >	gem_obj_create_from_date(fw)
> >	relase_firmware
> >
> >Can I have that inlined (or as a simple static helper function) into the
> >callsite please? Yes that unfortunately means I ask you to can the really
> >great kerneldoc you've done (I'd really like all patches to look this
> >great wrt docs), but I also think that the abstraction really doesn't pay
> >off. Firmware-loading using request_firmware is not something tricky
> >enough to justify added abstraction. Yes there was the
> >synchronization/completion stuff in there before, but as explained in the
> >previous discussion I don't expect us to be able to share that part at
> >all, and I'd like any kind of synchronization to be open-coded using
> >normal kernel primitives (like flush_work or similar) to make it really
> >stick out to reviewers.
> >
> >Thanks, Daniel
> 
> Well, I /could/ squash these functions into the GuC-specific loader and put
> a "g" on the front of all the names, but that would just result in the whole
> lot getting copypasted when we come to do the next letter of the alphabet :(
> 
> Or worse still, someone else would write a completely new version from
> scratch, which would be a waste of effort both for them and for those who
> would then have to review the new code, spot the similarities and
> differences, and ask for it all to be rewritten to be more like the
> by-then-existing GuC code.
> 
> It's also not quite as simple as your paraphrase; even without the prefetch
> code, it also deals with the do-this-once logic, and correctly freeing
> resources in all error paths; not enormously complicated but why repeat
> similar code in several places when you can just share the one instance (and
> there /will/ be other users for this code). And we may well add other common
> functionality here, should it turn out useful for more than one uC. You're
> just asking to increase the maintenance effort by duplicating the loader
> code rather than sharing it :(

do-this-once is solved easily by moving the do-it-once parts into
gem_init, which is called once at driver load. The complexity seems to
easy to reduce by dropping all the fancy loader status logic - the driver
only cares about success y/n. And for "no" we just declare the gpu dead.
Anything in addition to that can be dumped into dmesg using appropriate
DRM_DEBUG* for developers consumption. No need for debugfs or anything
else really imo.

That leaves a small functions which computes the firmware name per
platfrom, does the request_firmware, allocates a gem object, copies the
firmware data over, releases the firmware and then returns the resulting
gem object (or NULL when anything went wrong). guc init in gem_init_hw can
then consume that.

Once we have the guc+1 landing we can just export that or rename
intel_guc_* to intel_gt_* and that's it - the loader here isn't generic
because it copies the firmware to gem objects, which isn't actually what
the dmc loader needs.

> Conversely, if you think it adds little functionality, it also adds little
> overhead. I wouldn't much want to inline the currently-generic functions
> into their callers also because that would bloat them too much; at the
> moment, with the split between GuC-specific and common functions, the
> biggest functions turn out to be about one screenful. If we "open-code" it
> all we end up with monstrous 300-line-plus functions with any number of
> error paths, which is just not nice.

It adds a fairly complex interface (and a non-standard one on top with all
the custom state defines), and it starts out with generic code instead of
the more common extract common code after the fact. Desinging interface
before the users are really clear tends to result in overfitted design
(you wanted to use this for dmc too after all where it doesn't really fit
that well).

> So I'd really prefer to keep the utility code and the GuC-specific stuff
> tidily separated even before we have the other potential users of this
> support code in upstream.

I'm not opposed to a small helper which essentially would do
request_firmware+copy to gem bo. This here is a lot more than that.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Yu Dai]

On 07/06/2015 11:21 AM, Daniel Vetter wrote:
> On Mon, Jul 06, 2015 at 05:07:31PM +0100, Dave Gordon wrote:
> > On 06/07/15 15:14, Daniel Vetter wrote:
> > >On Fri, Jul 03, 2015 at 01:30:33PM +0100, Dave Gordon wrote:
> > >>Turn on interrupt steering to route necessary interrupts to GuC.
> > >>
> > >>Issue: VIZ-4884
> > >>Signed-off-by: Alex Dai <yu.dai@intel.com>
> > >>Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> > >>---
> > >>  drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
> > >>  drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
> > >>  2 files changed, 60 insertions(+), 2 deletions(-)
> > >>
> > >>diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> > >>index eefb847..9f32c6c 100644
> > >>--- a/drivers/gpu/drm/i915/i915_reg.h
> > >>+++ b/drivers/gpu/drm/i915/i915_reg.h
> > >>@@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
> > >>  #define GFX_MODE_GEN7	0x0229c
> > >>  #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
> > >>  #define   GFX_RUN_LIST_ENABLE		(1<<15)
> > >>+#define   GFX_INTERRUPT_STEERING	(1<<14)
> > >>  #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
> > >>  #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
> > >>  #define   GFX_REPLAY_MODE		(1<<11)
> > >>  #define   GFX_PSMI_GRANULARITY		(1<<10)
> > >>  #define   GFX_PPGTT_ENABLE		(1<<9)
> > >>
> > >>+#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
> > >>+#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
> > >>+#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
> > >>+#define   GFX_FORWARD_VBLANK_COND	(2<<5)
> > >>+
> > >>  #define VLV_DISPLAY_BASE 0x180000
> > >>  #define VLV_MIPI_BASE VLV_DISPLAY_BASE
> > >>
> > >>@@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
> > >>  #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
> > >>  #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
> > >>
> > >>-#define GEN8_BCS_IRQ_SHIFT 16
> > >>  #define GEN8_RCS_IRQ_SHIFT 0
> > >>-#define GEN8_VCS2_IRQ_SHIFT 16
> > >>+#define GEN8_BCS_IRQ_SHIFT 16
> > >>  #define GEN8_VCS1_IRQ_SHIFT 0
> > >>+#define GEN8_VCS2_IRQ_SHIFT 16
> > >>  #define GEN8_VECS_IRQ_SHIFT 0
> > >>+#define GEN8_WD_IRQ_SHIFT 16
> > >>
> > >>  #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
> > >>  #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
> > >>diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> > >>index eb76c34..ef5f3d5 100644
> > >>--- a/drivers/gpu/drm/i915/intel_guc_loader.c
> > >>+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> > >>@@ -62,6 +62,53 @@
> > >>  #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
> > >>  MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
> > >>
> > >>+static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
> > >>+{
> > >>+	struct intel_engine_cs *ring;
> > >>+	int i, irqs;
> > >>+
> > >>+	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
> > >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
> > >>+	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
> > >>+	for_each_ring(ring, dev_priv, i)
> > >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> > >>+
> > >>+	/* tell DE to send nothing to GuC */
> > >>+	I915_WRITE(DE_GUCRMR, ~0);
> > >>+
> > >>+	/* route all GT interrupts to the host */
> > >>+	I915_WRITE(GUC_BCS_RCS_IER, 0);
> > >>+	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
> > >>+	I915_WRITE(GUC_WD_VECS_IER, 0);
> > >>+}
> > >>+
> > >>+static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
> > >>+{
> > >>+	struct intel_engine_cs *ring;
> > >>+	int i, irqs;
> > >>+
> > >>+	/* tell all command streamers to forward interrupts and vblank to GuC */
> > >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
> > >>+	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
> > >>+	for_each_ring(ring, dev_priv, i)
> > >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> > >>+
> > >>+	/* tell DE to send (all) flip_done to GuC */
> > >>+	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
> > >>+	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
> > >>+	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
> > >>+	/* Unmasked bits will cause GuC response message to be sent */
> > >>+	I915_WRITE(DE_GUCRMR, ~irqs);
> > >>+
> > >>+	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
> > >>+	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
> > >>+	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
> > >>+	/* These three registers have the same bit definitions */
> > >>+	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
> > >>+	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
> > >>+	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
> > >>+}
> > >>+
> > >>  static u32 get_gttype(struct drm_i915_private *dev_priv)
> > >>  {
> > >>  	/* XXX: GT type based on PCI device ID? field seems unused by fw */
> > >>@@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> > >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
> > >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
> > >>
> > >>+	direct_interrupts_to_host(dev_priv);
> > >>  	i915_guc_submission_disable(dev);
> > >>
> > >>  	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
> > >>@@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> > >>  		err = i915_guc_submission_enable(dev);
> > >>  		if (err)
> > >>  			goto fail;
> > >>+		direct_interrupts_to_guc(dev_priv);
> > >
> > >This looks like still some remnants from when guc loading could fail and
> > >we needed to switch between cpu and guc submission at runtime. Problem is
> > >that you don't own the interrupt state at this point (since interrupts are
> > >already enabled) and hence there could be problems with races.
> > >
> > >Simplest solution would be to just set up the interrupt routing we want
> > >statically at driver load time in the respective irq callbacks (looking at
> > >guc-or-not-guc module option). That means we bake the guc submission y/n
> > >decision into the driver load sequence very early, but it does simplify
> > >things. And that's one of the reasons I don't like runtime fallbacks.
> > >-Daniel
> >
> > We can't send interrupts to the GuC before its firmware is ready for them,
> > hence we /can't/ set it up statically. We /must/ start with all interrupts
> > directed to the host, and switch them to the GuC only once the firmware is
> > loaded and running. (I suppose it would be simpler if the GuC firmware
> > *took* control of the interrupts when it started, rather than having i915
> > *give* the interrupts to the GuC, but that would leave even more
> > opportunities for confusion).
> >
> > This isn't really "runtime", this *is* still in the driver load stage,
> > although rather later than irq setup.
> >
> > For the important interrupts that we're switching to the GuC (i.e. context
> > switch) we DO know that there aren't any in flight already, because we can't
> > submit anything that would result in a context switch until the GuC is up &
> > running. For others (e.g. VBLANK, FLIP_DONE) it doesn't matter; the GuC
> > doesn't initiate these classes of interrupts anyway, and it doesn't keep
> > track of specific occurences or try to pair up actions and resulting events.
> > AFAIK, it's just listening in to those interrupts to see what's going on,
> > presumably for power-management purposes.
>
> Yeah I misread the patch, somehow I thought that we are touching some
> cpu-side irq control registers too. That was my concern since Paulo spent
> an awful lot of time unifying the setup for those to make them mesh well
> with runtime pm. But this is indeed all just guc private register state.
> One exception is the RING_MODE registers, but since that's a masked
> bitfield and this code is the only bit that touches it we should be fine
> too.
>
> One thing aside: Will the guc receive copies of these (i.e. will the cpu
> still be able to mask/receive/ack/clear/whatever them on its own) or is
> this somewhat exclusive and it's only either guc or cpu that gets certain
> interrupts? If it's just an enable/disable not affecting the cpu side
> really I'd suggest we call them enable/disable (more in line with other
> irq handling code) instead of the direct_interrupts_to_*. But that's
> really just a bikeshed to avoid confusion here.
> -Daniel
Right, code and comments are not clear here. For Display interrupts 
(VBLANK, FLIP_DONE etc.), GuC receives copies, which is "forwarding". 
But for those from GT (Context Switch, User Interrupt etc.), the cpu 
side won't receive it if we "route" or "direct" them to GuC.

Thanks,
Alex
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Daniel Vetter]

On Mon, Jul 06, 2015 at 05:00:52PM -0700, Yu Dai wrote:
> 
> 
> On 07/06/2015 11:21 AM, Daniel Vetter wrote:
> >On Mon, Jul 06, 2015 at 05:07:31PM +0100, Dave Gordon wrote:
> >> On 06/07/15 15:14, Daniel Vetter wrote:
> >> >On Fri, Jul 03, 2015 at 01:30:33PM +0100, Dave Gordon wrote:
> >> >>Turn on interrupt steering to route necessary interrupts to GuC.
> >> >>
> >> >>Issue: VIZ-4884
> >> >>Signed-off-by: Alex Dai <yu.dai@intel.com>
> >> >>Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> >> >>---
> >> >>  drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
> >> >>  drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
> >> >>  2 files changed, 60 insertions(+), 2 deletions(-)
> >> >>
> >> >>diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> >> >>index eefb847..9f32c6c 100644
> >> >>--- a/drivers/gpu/drm/i915/i915_reg.h
> >> >>+++ b/drivers/gpu/drm/i915/i915_reg.h
> >> >>@@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
> >> >>  #define GFX_MODE_GEN7	0x0229c
> >> >>  #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
> >> >>  #define   GFX_RUN_LIST_ENABLE		(1<<15)
> >> >>+#define   GFX_INTERRUPT_STEERING	(1<<14)
> >> >>  #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
> >> >>  #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
> >> >>  #define   GFX_REPLAY_MODE		(1<<11)
> >> >>  #define   GFX_PSMI_GRANULARITY		(1<<10)
> >> >>  #define   GFX_PPGTT_ENABLE		(1<<9)
> >> >>
> >> >>+#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
> >> >>+#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
> >> >>+#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
> >> >>+#define   GFX_FORWARD_VBLANK_COND	(2<<5)
> >> >>+
> >> >>  #define VLV_DISPLAY_BASE 0x180000
> >> >>  #define VLV_MIPI_BASE VLV_DISPLAY_BASE
> >> >>
> >> >>@@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
> >> >>  #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
> >> >>  #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
> >> >>
> >> >>-#define GEN8_BCS_IRQ_SHIFT 16
> >> >>  #define GEN8_RCS_IRQ_SHIFT 0
> >> >>-#define GEN8_VCS2_IRQ_SHIFT 16
> >> >>+#define GEN8_BCS_IRQ_SHIFT 16
> >> >>  #define GEN8_VCS1_IRQ_SHIFT 0
> >> >>+#define GEN8_VCS2_IRQ_SHIFT 16
> >> >>  #define GEN8_VECS_IRQ_SHIFT 0
> >> >>+#define GEN8_WD_IRQ_SHIFT 16
> >> >>
> >> >>  #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
> >> >>  #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
> >> >>diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> >> >>index eb76c34..ef5f3d5 100644
> >> >>--- a/drivers/gpu/drm/i915/intel_guc_loader.c
> >> >>+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> >> >>@@ -62,6 +62,53 @@
> >> >>  #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
> >> >>  MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
> >> >>
> >> >>+static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
> >> >>+{
> >> >>+	struct intel_engine_cs *ring;
> >> >>+	int i, irqs;
> >> >>+
> >> >>+	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
> >> >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
> >> >>+	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
> >> >>+	for_each_ring(ring, dev_priv, i)
> >> >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> >> >>+
> >> >>+	/* tell DE to send nothing to GuC */
> >> >>+	I915_WRITE(DE_GUCRMR, ~0);
> >> >>+
> >> >>+	/* route all GT interrupts to the host */
> >> >>+	I915_WRITE(GUC_BCS_RCS_IER, 0);
> >> >>+	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
> >> >>+	I915_WRITE(GUC_WD_VECS_IER, 0);
> >> >>+}
> >> >>+
> >> >>+static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
> >> >>+{
> >> >>+	struct intel_engine_cs *ring;
> >> >>+	int i, irqs;
> >> >>+
> >> >>+	/* tell all command streamers to forward interrupts and vblank to GuC */
> >> >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
> >> >>+	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
> >> >>+	for_each_ring(ring, dev_priv, i)
> >> >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> >> >>+
> >> >>+	/* tell DE to send (all) flip_done to GuC */
> >> >>+	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
> >> >>+	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
> >> >>+	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
> >> >>+	/* Unmasked bits will cause GuC response message to be sent */
> >> >>+	I915_WRITE(DE_GUCRMR, ~irqs);
> >> >>+
> >> >>+	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
> >> >>+	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
> >> >>+	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
> >> >>+	/* These three registers have the same bit definitions */
> >> >>+	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
> >> >>+	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
> >> >>+	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
> >> >>+}
> >> >>+
> >> >>  static u32 get_gttype(struct drm_i915_private *dev_priv)
> >> >>  {
> >> >>  	/* XXX: GT type based on PCI device ID? field seems unused by fw */
> >> >>@@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> >> >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
> >> >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
> >> >>
> >> >>+	direct_interrupts_to_host(dev_priv);
> >> >>  	i915_guc_submission_disable(dev);
> >> >>
> >> >>  	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
> >> >>@@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> >> >>  		err = i915_guc_submission_enable(dev);
> >> >>  		if (err)
> >> >>  			goto fail;
> >> >>+		direct_interrupts_to_guc(dev_priv);
> >> >
> >> >This looks like still some remnants from when guc loading could fail and
> >> >we needed to switch between cpu and guc submission at runtime. Problem is
> >> >that you don't own the interrupt state at this point (since interrupts are
> >> >already enabled) and hence there could be problems with races.
> >> >
> >> >Simplest solution would be to just set up the interrupt routing we want
> >> >statically at driver load time in the respective irq callbacks (looking at
> >> >guc-or-not-guc module option). That means we bake the guc submission y/n
> >> >decision into the driver load sequence very early, but it does simplify
> >> >things. And that's one of the reasons I don't like runtime fallbacks.
> >> >-Daniel
> >>
> >> We can't send interrupts to the GuC before its firmware is ready for them,
> >> hence we /can't/ set it up statically. We /must/ start with all interrupts
> >> directed to the host, and switch them to the GuC only once the firmware is
> >> loaded and running. (I suppose it would be simpler if the GuC firmware
> >> *took* control of the interrupts when it started, rather than having i915
> >> *give* the interrupts to the GuC, but that would leave even more
> >> opportunities for confusion).
> >>
> >> This isn't really "runtime", this *is* still in the driver load stage,
> >> although rather later than irq setup.
> >>
> >> For the important interrupts that we're switching to the GuC (i.e. context
> >> switch) we DO know that there aren't any in flight already, because we can't
> >> submit anything that would result in a context switch until the GuC is up &
> >> running. For others (e.g. VBLANK, FLIP_DONE) it doesn't matter; the GuC
> >> doesn't initiate these classes of interrupts anyway, and it doesn't keep
> >> track of specific occurences or try to pair up actions and resulting events.
> >> AFAIK, it's just listening in to those interrupts to see what's going on,
> >> presumably for power-management purposes.
> >
> >Yeah I misread the patch, somehow I thought that we are touching some
> >cpu-side irq control registers too. That was my concern since Paulo spent
> >an awful lot of time unifying the setup for those to make them mesh well
> >with runtime pm. But this is indeed all just guc private register state.
> >One exception is the RING_MODE registers, but since that's a masked
> >bitfield and this code is the only bit that touches it we should be fine
> >too.
> >
> >One thing aside: Will the guc receive copies of these (i.e. will the cpu
> >still be able to mask/receive/ack/clear/whatever them on its own) or is
> >this somewhat exclusive and it's only either guc or cpu that gets certain
> >interrupts? If it's just an enable/disable not affecting the cpu side
> >really I'd suggest we call them enable/disable (more in line with other
> >irq handling code) instead of the direct_interrupts_to_*. But that's
> >really just a bikeshed to avoid confusion here.
> >-Daniel
> Right, code and comments are not clear here. For Display interrupts (VBLANK,
> FLIP_DONE etc.), GuC receives copies, which is "forwarding". But for those
> from GT (Context Switch, User Interrupt etc.), the cpu side won't receive it
> if we "route" or "direct" them to GuC.

Hm, I guess then it'd make sense to just never enable the cpu side
interrupts when in guc mode. Less confusing at least for sure. Then it
would really just be an enable/disable since it will no longer steal
anything we care about.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 11/15] drm/i915: Interrupt routing for GuC submission

[Yu Dai]

On 07/07/2015 02:06 AM, Daniel Vetter wrote:
> On Mon, Jul 06, 2015 at 05:00:52PM -0700, Yu Dai wrote:
> >
> >
> > On 07/06/2015 11:21 AM, Daniel Vetter wrote:
> > >On Mon, Jul 06, 2015 at 05:07:31PM +0100, Dave Gordon wrote:
> > >> On 06/07/15 15:14, Daniel Vetter wrote:
> > >> >On Fri, Jul 03, 2015 at 01:30:33PM +0100, Dave Gordon wrote:
> > >> >>Turn on interrupt steering to route necessary interrupts to GuC.
> > >> >>
> > >> >>Issue: VIZ-4884
> > >> >>Signed-off-by: Alex Dai <yu.dai@intel.com>
> > >> >>Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
> > >> >>---
> > >> >>  drivers/gpu/drm/i915/i915_reg.h         | 11 +++++--
> > >> >>  drivers/gpu/drm/i915/intel_guc_loader.c | 51 +++++++++++++++++++++++++++++++++
> > >> >>  2 files changed, 60 insertions(+), 2 deletions(-)
> > >> >>
> > >> >>diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> > >> >>index eefb847..9f32c6c 100644
> > >> >>--- a/drivers/gpu/drm/i915/i915_reg.h
> > >> >>+++ b/drivers/gpu/drm/i915/i915_reg.h
> > >> >>@@ -1659,12 +1659,18 @@ enum skl_disp_power_wells {
> > >> >>  #define GFX_MODE_GEN7	0x0229c
> > >> >>  #define RING_MODE_GEN7(ring)	((ring)->mmio_base+0x29c)
> > >> >>  #define   GFX_RUN_LIST_ENABLE		(1<<15)
> > >> >>+#define   GFX_INTERRUPT_STEERING	(1<<14)
> > >> >>  #define   GFX_TLB_INVALIDATE_EXPLICIT	(1<<13)
> > >> >>  #define   GFX_SURFACE_FAULT_ENABLE	(1<<12)
> > >> >>  #define   GFX_REPLAY_MODE		(1<<11)
> > >> >>  #define   GFX_PSMI_GRANULARITY		(1<<10)
> > >> >>  #define   GFX_PPGTT_ENABLE		(1<<9)
> > >> >>
> > >> >>+#define   GFX_FORWARD_VBLANK_MASK	(3<<5)
> > >> >>+#define   GFX_FORWARD_VBLANK_NEVER	(0<<5)
> > >> >>+#define   GFX_FORWARD_VBLANK_ALWAYS	(1<<5)
> > >> >>+#define   GFX_FORWARD_VBLANK_COND	(2<<5)
> > >> >>+
> > >> >>  #define VLV_DISPLAY_BASE 0x180000
> > >> >>  #define VLV_MIPI_BASE VLV_DISPLAY_BASE
> > >> >>
> > >> >>@@ -5677,11 +5683,12 @@ enum skl_disp_power_wells {
> > >> >>  #define GEN8_GT_IIR(which) (0x44308 + (0x10 * (which)))
> > >> >>  #define GEN8_GT_IER(which) (0x4430c + (0x10 * (which)))
> > >> >>
> > >> >>-#define GEN8_BCS_IRQ_SHIFT 16
> > >> >>  #define GEN8_RCS_IRQ_SHIFT 0
> > >> >>-#define GEN8_VCS2_IRQ_SHIFT 16
> > >> >>+#define GEN8_BCS_IRQ_SHIFT 16
> > >> >>  #define GEN8_VCS1_IRQ_SHIFT 0
> > >> >>+#define GEN8_VCS2_IRQ_SHIFT 16
> > >> >>  #define GEN8_VECS_IRQ_SHIFT 0
> > >> >>+#define GEN8_WD_IRQ_SHIFT 16
> > >> >>
> > >> >>  #define GEN8_DE_PIPE_ISR(pipe) (0x44400 + (0x10 * (pipe)))
> > >> >>  #define GEN8_DE_PIPE_IMR(pipe) (0x44404 + (0x10 * (pipe)))
> > >> >>diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c
> > >> >>index eb76c34..ef5f3d5 100644
> > >> >>--- a/drivers/gpu/drm/i915/intel_guc_loader.c
> > >> >>+++ b/drivers/gpu/drm/i915/intel_guc_loader.c
> > >> >>@@ -62,6 +62,53 @@
> > >> >>  #define I915_SKL_GUC_UCODE "i915/skl_guc_ver3.bin"
> > >> >>  MODULE_FIRMWARE(I915_SKL_GUC_UCODE);
> > >> >>
> > >> >>+static void direct_interrupts_to_host(struct drm_i915_private *dev_priv)
> > >> >>+{
> > >> >>+	struct intel_engine_cs *ring;
> > >> >>+	int i, irqs;
> > >> >>+
> > >> >>+	/* tell all command streamers NOT to forward interrupts and vblank to GuC */
> > >> >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
> > >> >>+	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
> > >> >>+	for_each_ring(ring, dev_priv, i)
> > >> >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> > >> >>+
> > >> >>+	/* tell DE to send nothing to GuC */
> > >> >>+	I915_WRITE(DE_GUCRMR, ~0);
> > >> >>+
> > >> >>+	/* route all GT interrupts to the host */
> > >> >>+	I915_WRITE(GUC_BCS_RCS_IER, 0);
> > >> >>+	I915_WRITE(GUC_VCS2_VCS1_IER, 0);
> > >> >>+	I915_WRITE(GUC_WD_VECS_IER, 0);
> > >> >>+}
> > >> >>+
> > >> >>+static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv)
> > >> >>+{
> > >> >>+	struct intel_engine_cs *ring;
> > >> >>+	int i, irqs;
> > >> >>+
> > >> >>+	/* tell all command streamers to forward interrupts and vblank to GuC */
> > >> >>+	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS);
> > >> >>+	irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
> > >> >>+	for_each_ring(ring, dev_priv, i)
> > >> >>+		I915_WRITE(RING_MODE_GEN7(ring), irqs);
> > >> >>+
> > >> >>+	/* tell DE to send (all) flip_done to GuC */
> > >> >>+	irqs = DERRMR_PIPEA_PRI_FLIP_DONE | DERRMR_PIPEA_SPR_FLIP_DONE |
> > >> >>+	       DERRMR_PIPEB_PRI_FLIP_DONE | DERRMR_PIPEB_SPR_FLIP_DONE |
> > >> >>+	       DERRMR_PIPEC_PRI_FLIP_DONE | DERRMR_PIPEC_SPR_FLIP_DONE;
> > >> >>+	/* Unmasked bits will cause GuC response message to be sent */
> > >> >>+	I915_WRITE(DE_GUCRMR, ~irqs);
> > >> >>+
> > >> >>+	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
> > >> >>+	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
> > >> >>+	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
> > >> >>+	/* These three registers have the same bit definitions */
> > >> >>+	I915_WRITE(GUC_BCS_RCS_IER, ~irqs);
> > >> >>+	I915_WRITE(GUC_VCS2_VCS1_IER, ~irqs);
> > >> >>+	I915_WRITE(GUC_WD_VECS_IER, ~irqs);
> > >> >>+}
> > >> >>+
> > >> >>  static u32 get_gttype(struct drm_i915_private *dev_priv)
> > >> >>  {
> > >> >>  	/* XXX: GT type based on PCI device ID? field seems unused by fw */
> > >> >>@@ -417,6 +464,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> > >> >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_fetch_status),
> > >> >>  		intel_uc_fw_status_repr(guc_fw->uc_fw_load_status));
> > >> >>
> > >> >>+	direct_interrupts_to_host(dev_priv);
> > >> >>  	i915_guc_submission_disable(dev);
> > >> >>
> > >> >>  	if (guc_fw->uc_fw_fetch_status == INTEL_UC_FIRMWARE_NONE)
> > >> >>@@ -450,6 +498,7 @@ int intel_guc_ucode_load(struct drm_device *dev)
> > >> >>  		err = i915_guc_submission_enable(dev);
> > >> >>  		if (err)
> > >> >>  			goto fail;
> > >> >>+		direct_interrupts_to_guc(dev_priv);
> > >> >
> > >> >This looks like still some remnants from when guc loading could fail and
> > >> >we needed to switch between cpu and guc submission at runtime. Problem is
> > >> >that you don't own the interrupt state at this point (since interrupts are
> > >> >already enabled) and hence there could be problems with races.
> > >> >
> > >> >Simplest solution would be to just set up the interrupt routing we want
> > >> >statically at driver load time in the respective irq callbacks (looking at
> > >> >guc-or-not-guc module option). That means we bake the guc submission y/n
> > >> >decision into the driver load sequence very early, but it does simplify
> > >> >things. And that's one of the reasons I don't like runtime fallbacks.
> > >> >-Daniel
> > >>
> > >> We can't send interrupts to the GuC before its firmware is ready for them,
> > >> hence we /can't/ set it up statically. We /must/ start with all interrupts
> > >> directed to the host, and switch them to the GuC only once the firmware is
> > >> loaded and running. (I suppose it would be simpler if the GuC firmware
> > >> *took* control of the interrupts when it started, rather than having i915
> > >> *give* the interrupts to the GuC, but that would leave even more
> > >> opportunities for confusion).
> > >>
> > >> This isn't really "runtime", this *is* still in the driver load stage,
> > >> although rather later than irq setup.
> > >>
> > >> For the important interrupts that we're switching to the GuC (i.e. context
> > >> switch) we DO know that there aren't any in flight already, because we can't
> > >> submit anything that would result in a context switch until the GuC is up &
> > >> running. For others (e.g. VBLANK, FLIP_DONE) it doesn't matter; the GuC
> > >> doesn't initiate these classes of interrupts anyway, and it doesn't keep
> > >> track of specific occurences or try to pair up actions and resulting events.
> > >> AFAIK, it's just listening in to those interrupts to see what's going on,
> > >> presumably for power-management purposes.
> > >
> > >Yeah I misread the patch, somehow I thought that we are touching some
> > >cpu-side irq control registers too. That was my concern since Paulo spent
> > >an awful lot of time unifying the setup for those to make them mesh well
> > >with runtime pm. But this is indeed all just guc private register state.
> > >One exception is the RING_MODE registers, but since that's a masked
> > >bitfield and this code is the only bit that touches it we should be fine
> > >too.
> > >
> > >One thing aside: Will the guc receive copies of these (i.e. will the cpu
> > >still be able to mask/receive/ack/clear/whatever them on its own) or is
> > >this somewhat exclusive and it's only either guc or cpu that gets certain
> > >interrupts? If it's just an enable/disable not affecting the cpu side
> > >really I'd suggest we call them enable/disable (more in line with other
> > >irq handling code) instead of the direct_interrupts_to_*. But that's
> > >really just a bikeshed to avoid confusion here.
> > >-Daniel
> > Right, code and comments are not clear here. For Display interrupts (VBLANK,
> > FLIP_DONE etc.), GuC receives copies, which is "forwarding". But for those
> > from GT (Context Switch, User Interrupt etc.), the cpu side won't receive it
> > if we "route" or "direct" them to GuC.
>
> Hm, I guess then it'd make sense to just never enable the cpu side
> interrupts when in guc mode. Less confusing at least for sure. Then it
> would really just be an enable/disable since it will no longer steal
> anything we care about.
Agree. For now only the user interrupt is kept for driver to wake up any 
waiting in queue. We should be able to remove this queue 
(ring->irq_queue) neatly for GuC. So we can completely eliminate CPU 
wake up in handling interrupts from GT. But I don't have a clear picture 
yet. Any ideas are welcome.

Thanks,
Alex
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 00/15 v3] Batch submission via GuC

[O'Rourke, Tom]

On Mon, Jul 06, 2015 at 04:29:47PM +0200, Daniel Vetter wrote:
> On Fri, Jul 03, 2015 at 01:30:22PM +0100, Dave Gordon wrote:
> > This patch series enables command submission via the GuC. In this mode,
[TOR:] ...
> >   created the original versions on which some of these patches are based.
> 
> A few minor things still (I guess you expected my comment on the firmware
> loader) but looks good overall with those addressed and the detailed
> review done on top.
> -Daniel
> 
[TOR:] This patch series has conflicts with Mika's recently
merged "Convert parts of intel_lrc.c to requests" series.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx