OpenCloudOS-Kernel/drivers/gpu/drm/i915/i915_gem_request.c

1193 lines
34 KiB
C
Raw Normal View History

/*
* Copyright © 2008-2015 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.
*
*/
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
#include <linux/prefetch.h>
#include <linux/dma-fence-array.h>
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
#include "i915_drv.h"
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static const char *i915_fence_get_driver_name(struct dma_fence *fence)
{
return "i915";
}
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static const char *i915_fence_get_timeline_name(struct dma_fence *fence)
{
return to_request(fence)->timeline->common->name;
}
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static bool i915_fence_signaled(struct dma_fence *fence)
{
return i915_gem_request_completed(to_request(fence));
}
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static bool i915_fence_enable_signaling(struct dma_fence *fence)
{
if (i915_fence_signaled(fence))
return false;
intel_engine_enable_signaling(to_request(fence));
return true;
}
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static signed long i915_fence_wait(struct dma_fence *fence,
bool interruptible,
signed long timeout)
{
return i915_wait_request(to_request(fence), interruptible, timeout);
}
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static void i915_fence_release(struct dma_fence *fence)
{
struct drm_i915_gem_request *req = to_request(fence);
kmem_cache_free(req->i915->requests, req);
}
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
const struct dma_fence_ops i915_fence_ops = {
.get_driver_name = i915_fence_get_driver_name,
.get_timeline_name = i915_fence_get_timeline_name,
.enable_signaling = i915_fence_enable_signaling,
.signaled = i915_fence_signaled,
.wait = i915_fence_wait,
.release = i915_fence_release,
};
int i915_gem_request_add_to_client(struct drm_i915_gem_request *req,
struct drm_file *file)
{
struct drm_i915_private *dev_private;
struct drm_i915_file_private *file_priv;
WARN_ON(!req || !file || req->file_priv);
if (!req || !file)
return -EINVAL;
if (req->file_priv)
return -EINVAL;
dev_private = req->i915;
file_priv = file->driver_priv;
spin_lock(&file_priv->mm.lock);
req->file_priv = file_priv;
list_add_tail(&req->client_list, &file_priv->mm.request_list);
spin_unlock(&file_priv->mm.lock);
return 0;
}
static inline void
i915_gem_request_remove_from_client(struct drm_i915_gem_request *request)
{
struct drm_i915_file_private *file_priv = request->file_priv;
if (!file_priv)
return;
spin_lock(&file_priv->mm.lock);
list_del(&request->client_list);
request->file_priv = NULL;
spin_unlock(&file_priv->mm.lock);
}
static struct i915_dependency *
i915_dependency_alloc(struct drm_i915_private *i915)
{
return kmem_cache_alloc(i915->dependencies, GFP_KERNEL);
}
static void
i915_dependency_free(struct drm_i915_private *i915,
struct i915_dependency *dep)
{
kmem_cache_free(i915->dependencies, dep);
}
static void
__i915_priotree_add_dependency(struct i915_priotree *pt,
struct i915_priotree *signal,
struct i915_dependency *dep,
unsigned long flags)
{
drm/i915/scheduler: Execute requests in order of priorities Track the priority of each request and use it to determine the order in which we submit requests to the hardware via execlists. The priority of the request is determined by the user (eventually via the context) but may be overridden at any time by the driver. When we set the priority of the request, we bump the priority of all of its dependencies to match - so that a high priority drawing operation is not stuck behind a background task. When the request is ready to execute (i.e. we have signaled the submit fence following completion of all its dependencies, including third party fences), we put the request into a priority sorted rbtree to be submitted to the hardware. If the request is higher priority than all pending requests, it will be submitted on the next context-switch interrupt as soon as the hardware has completed the current request. We do not currently preempt any current execution to immediately run a very high priority request, at least not yet. One more limitation, is that this is first implementation is for execlists only so currently limited to gen8/gen9. v2: Replace recursive priority inheritance bumping with an iterative depth-first search list. v3: list_next_entry() for walking lists v4: Explain how the dfs solves the recursion problem with PI. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161114204105.29171-8-chris@chris-wilson.co.uk
2016-11-15 04:41:03 +08:00
INIT_LIST_HEAD(&dep->dfs_link);
list_add(&dep->wait_link, &signal->waiters_list);
list_add(&dep->signal_link, &pt->signalers_list);
dep->signaler = signal;
dep->flags = flags;
}
static int
i915_priotree_add_dependency(struct drm_i915_private *i915,
struct i915_priotree *pt,
struct i915_priotree *signal)
{
struct i915_dependency *dep;
dep = i915_dependency_alloc(i915);
if (!dep)
return -ENOMEM;
__i915_priotree_add_dependency(pt, signal, dep, I915_DEPENDENCY_ALLOC);
return 0;
}
static void
i915_priotree_fini(struct drm_i915_private *i915, struct i915_priotree *pt)
{
struct i915_dependency *dep, *next;
drm/i915/scheduler: Execute requests in order of priorities Track the priority of each request and use it to determine the order in which we submit requests to the hardware via execlists. The priority of the request is determined by the user (eventually via the context) but may be overridden at any time by the driver. When we set the priority of the request, we bump the priority of all of its dependencies to match - so that a high priority drawing operation is not stuck behind a background task. When the request is ready to execute (i.e. we have signaled the submit fence following completion of all its dependencies, including third party fences), we put the request into a priority sorted rbtree to be submitted to the hardware. If the request is higher priority than all pending requests, it will be submitted on the next context-switch interrupt as soon as the hardware has completed the current request. We do not currently preempt any current execution to immediately run a very high priority request, at least not yet. One more limitation, is that this is first implementation is for execlists only so currently limited to gen8/gen9. v2: Replace recursive priority inheritance bumping with an iterative depth-first search list. v3: list_next_entry() for walking lists v4: Explain how the dfs solves the recursion problem with PI. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161114204105.29171-8-chris@chris-wilson.co.uk
2016-11-15 04:41:03 +08:00
GEM_BUG_ON(!RB_EMPTY_NODE(&pt->node));
/* Everyone we depended upon (the fences we wait to be signaled)
* should retire before us and remove themselves from our list.
* However, retirement is run independently on each timeline and
* so we may be called out-of-order.
*/
list_for_each_entry_safe(dep, next, &pt->signalers_list, signal_link) {
list_del(&dep->wait_link);
if (dep->flags & I915_DEPENDENCY_ALLOC)
i915_dependency_free(i915, dep);
}
/* Remove ourselves from everyone who depends upon us */
list_for_each_entry_safe(dep, next, &pt->waiters_list, wait_link) {
list_del(&dep->signal_link);
if (dep->flags & I915_DEPENDENCY_ALLOC)
i915_dependency_free(i915, dep);
}
}
static void
i915_priotree_init(struct i915_priotree *pt)
{
INIT_LIST_HEAD(&pt->signalers_list);
INIT_LIST_HEAD(&pt->waiters_list);
drm/i915/scheduler: Execute requests in order of priorities Track the priority of each request and use it to determine the order in which we submit requests to the hardware via execlists. The priority of the request is determined by the user (eventually via the context) but may be overridden at any time by the driver. When we set the priority of the request, we bump the priority of all of its dependencies to match - so that a high priority drawing operation is not stuck behind a background task. When the request is ready to execute (i.e. we have signaled the submit fence following completion of all its dependencies, including third party fences), we put the request into a priority sorted rbtree to be submitted to the hardware. If the request is higher priority than all pending requests, it will be submitted on the next context-switch interrupt as soon as the hardware has completed the current request. We do not currently preempt any current execution to immediately run a very high priority request, at least not yet. One more limitation, is that this is first implementation is for execlists only so currently limited to gen8/gen9. v2: Replace recursive priority inheritance bumping with an iterative depth-first search list. v3: list_next_entry() for walking lists v4: Explain how the dfs solves the recursion problem with PI. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161114204105.29171-8-chris@chris-wilson.co.uk
2016-11-15 04:41:03 +08:00
RB_CLEAR_NODE(&pt->node);
pt->priority = INT_MIN;
}
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
void i915_gem_retire_noop(struct i915_gem_active *active,
struct drm_i915_gem_request *request)
{
/* Space left intentionally blank */
}
static void i915_gem_request_retire(struct drm_i915_gem_request *request)
{
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
struct i915_gem_active *active, *next;
lockdep_assert_held(&request->i915->drm.struct_mutex);
GEM_BUG_ON(!i915_sw_fence_signaled(&request->submit));
GEM_BUG_ON(!i915_sw_fence_signaled(&request->execute));
GEM_BUG_ON(!i915_gem_request_completed(request));
GEM_BUG_ON(!request->i915->gt.active_requests);
trace_i915_gem_request_retire(request);
spin_lock_irq(&request->engine->timeline->lock);
list_del_init(&request->link);
spin_unlock_irq(&request->engine->timeline->lock);
/* We know the GPU must have read the request to have
* sent us the seqno + interrupt, so use the position
* of tail of the request to update the last known position
* of the GPU head.
*
* Note this requires that we are always called in request
* completion order.
*/
list_del(&request->ring_link);
request->ring->last_retired_head = request->postfix;
if (!--request->i915->gt.active_requests) {
GEM_BUG_ON(!request->i915->gt.awake);
mod_delayed_work(request->i915->wq,
&request->i915->gt.idle_work,
msecs_to_jiffies(100));
}
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
/* Walk through the active list, calling retire on each. This allows
* objects to track their GPU activity and mark themselves as idle
* when their *last* active request is completed (updating state
* tracking lists for eviction, active references for GEM, etc).
*
* As the ->retire() may free the node, we decouple it first and
* pass along the auxiliary information (to avoid dereferencing
* the node after the callback).
*/
list_for_each_entry_safe(active, next, &request->active_list, link) {
/* In microbenchmarks or focusing upon time inside the kernel,
* we may spend an inordinate amount of time simply handling
* the retirement of requests and processing their callbacks.
* Of which, this loop itself is particularly hot due to the
* cache misses when jumping around the list of i915_gem_active.
* So we try to keep this loop as streamlined as possible and
* also prefetch the next i915_gem_active to try and hide
* the likely cache miss.
*/
prefetchw(next);
INIT_LIST_HEAD(&active->link);
drm/i915: Enable lockless lookup of request tracking via RCU If we enable RCU for the requests (providing a grace period where we can inspect a "dead" request before it is freed), we can allow callers to carefully perform lockless lookup of an active request. However, by enabling deferred freeing of requests, we can potentially hog a lot of memory when dealing with tens of thousands of requests per second - with a quick insertion of a synchronize_rcu() inside our shrinker callback, that issue disappears. v2: Currently, it is our responsibility to handle reclaim i.e. to avoid hogging memory with the delayed slab frees. At the moment, we wait for a grace period in the shrinker, and block for all RCU callbacks on oom. Suggested alternatives focus on flushing our RCU callback when we have a certain number of outstanding request frees, and blocking on that flush after a second high watermark. (So rather than wait for the system to run out of memory, we stop issuing requests - both are nondeterministic.) Paul E. McKenney wrote: Another approach is synchronize_rcu() after some largish number of requests. The advantage of this approach is that it throttles the production of callbacks at the source. The corresponding disadvantage is that it slows things up. Another approach is to use call_rcu(), but if the previous call_rcu() is still in flight, block waiting for it. Yet another approach is the get_state_synchronize_rcu() / cond_synchronize_rcu() pair. The idea is to do something like this: cond_synchronize_rcu(cookie); cookie = get_state_synchronize_rcu(); You would of course do an initial get_state_synchronize_rcu() to get things going. This would not block unless there was less than one grace period's worth of time between invocations. But this assumes a busy system, where there is almost always a grace period in flight. But you can make that happen as follows: cond_synchronize_rcu(cookie); cookie = get_state_synchronize_rcu(); call_rcu(&my_rcu_head, noop_function); Note that you need additional code to make sure that the old callback has completed before doing a new one. Setting and clearing a flag with appropriate memory ordering control suffices (e.g,. smp_load_acquire() and smp_store_release()). v3: More comments on compiler and processor order of operations within the RCU lookup and discover we can use rcu_access_pointer() here instead. v4: Wrap i915_gem_active_get_rcu() to take the rcu_read_lock itself. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470324762-2545-25-git-send-email-chris@chris-wilson.co.uk
2016-08-04 23:32:41 +08:00
RCU_INIT_POINTER(active->request, NULL);
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
active->retire(active, request);
}
i915_gem_request_remove_from_client(request);
if (request->previous_context) {
if (i915.enable_execlists)
intel_lr_context_unpin(request->previous_context,
request->engine);
}
/* Retirement decays the ban score as it is a sign of ctx progress */
if (request->ctx->ban_score > 0)
request->ctx->ban_score--;
i915_gem_context_put(request->ctx);
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
dma_fence_signal(&request->fence);
i915_priotree_fini(request->i915, &request->priotree);
i915_gem_request_put(request);
}
void i915_gem_request_retire_upto(struct drm_i915_gem_request *req)
{
struct intel_engine_cs *engine = req->engine;
struct drm_i915_gem_request *tmp;
lockdep_assert_held(&req->i915->drm.struct_mutex);
GEM_BUG_ON(!i915_gem_request_completed(req));
if (list_empty(&req->link))
return;
do {
tmp = list_first_entry(&engine->timeline->requests,
typeof(*tmp), link);
i915_gem_request_retire(tmp);
} while (tmp != req);
}
static int i915_gem_check_wedge(struct drm_i915_private *dev_priv)
{
struct i915_gpu_error *error = &dev_priv->gpu_error;
if (i915_terminally_wedged(error))
return -EIO;
if (i915_reset_in_progress(error)) {
/* Non-interruptible callers can't handle -EAGAIN, hence return
* -EIO unconditionally for these.
*/
if (!dev_priv->mm.interruptible)
return -EIO;
return -EAGAIN;
}
return 0;
}
static int i915_gem_init_global_seqno(struct drm_i915_private *i915, u32 seqno)
{
struct i915_gem_timeline *timeline = &i915->gt.global_timeline;
struct intel_engine_cs *engine;
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-14 01:14:48 +08:00
enum intel_engine_id id;
int ret;
/* Carefully retire all requests without writing to the rings */
ret = i915_gem_wait_for_idle(i915,
I915_WAIT_INTERRUPTIBLE |
I915_WAIT_LOCKED);
if (ret)
return ret;
i915_gem_retire_requests(i915);
GEM_BUG_ON(i915->gt.active_requests > 1);
/* If the seqno wraps around, we need to clear the breadcrumb rbtree */
if (!i915_seqno_passed(seqno, atomic_read(&timeline->seqno))) {
while (intel_breadcrumbs_busy(i915))
cond_resched(); /* spin until threads are complete */
}
atomic_set(&timeline->seqno, seqno);
/* Finally reset hw state */
for_each_engine(engine, i915, id)
intel_engine_init_global_seqno(engine, seqno);
list_for_each_entry(timeline, &i915->gt.timelines, link) {
for_each_engine(engine, i915, id) {
struct intel_timeline *tl = &timeline->engine[id];
memset(tl->sync_seqno, 0, sizeof(tl->sync_seqno));
}
}
return 0;
}
int i915_gem_set_global_seqno(struct drm_device *dev, u32 seqno)
{
struct drm_i915_private *dev_priv = to_i915(dev);
lockdep_assert_held(&dev_priv->drm.struct_mutex);
if (seqno == 0)
return -EINVAL;
/* HWS page needs to be set less than what we
* will inject to ring
*/
return i915_gem_init_global_seqno(dev_priv, seqno - 1);
}
static int reserve_global_seqno(struct drm_i915_private *i915)
{
u32 active_requests = ++i915->gt.active_requests;
u32 seqno = atomic_read(&i915->gt.global_timeline.seqno);
int ret;
/* Reservation is fine until we need to wrap around */
if (likely(seqno + active_requests > seqno))
return 0;
ret = i915_gem_init_global_seqno(i915, 0);
if (ret) {
i915->gt.active_requests--;
return ret;
}
return 0;
}
static u32 __timeline_get_seqno(struct i915_gem_timeline *tl)
{
/* seqno only incremented under a mutex */
return ++tl->seqno.counter;
}
static u32 timeline_get_seqno(struct i915_gem_timeline *tl)
{
return atomic_inc_return(&tl->seqno);
}
void __i915_gem_request_submit(struct drm_i915_gem_request *request)
{
struct intel_engine_cs *engine = request->engine;
struct intel_timeline *timeline;
u32 seqno;
/* Transfer from per-context onto the global per-engine timeline */
timeline = engine->timeline;
GEM_BUG_ON(timeline == request->timeline);
assert_spin_locked(&timeline->lock);
seqno = timeline_get_seqno(timeline->common);
GEM_BUG_ON(!seqno);
GEM_BUG_ON(i915_seqno_passed(intel_engine_get_seqno(engine), seqno));
GEM_BUG_ON(i915_seqno_passed(timeline->last_submitted_seqno, seqno));
request->previous_seqno = timeline->last_submitted_seqno;
timeline->last_submitted_seqno = seqno;
/* We may be recursing from the signal callback of another i915 fence */
spin_lock_nested(&request->lock, SINGLE_DEPTH_NESTING);
request->global_seqno = seqno;
if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags))
intel_engine_enable_signaling(request);
spin_unlock(&request->lock);
GEM_BUG_ON(!request->global_seqno);
engine->emit_breadcrumb(request,
request->ring->vaddr + request->postfix);
spin_lock(&request->timeline->lock);
list_move_tail(&request->link, &timeline->requests);
spin_unlock(&request->timeline->lock);
i915_sw_fence_commit(&request->execute);
}
void i915_gem_request_submit(struct drm_i915_gem_request *request)
{
struct intel_engine_cs *engine = request->engine;
unsigned long flags;
/* Will be called from irq-context when using foreign fences. */
spin_lock_irqsave(&engine->timeline->lock, flags);
__i915_gem_request_submit(request);
spin_unlock_irqrestore(&engine->timeline->lock, flags);
}
static int __i915_sw_fence_call
submit_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
{
struct drm_i915_gem_request *request =
container_of(fence, typeof(*request), submit);
switch (state) {
case FENCE_COMPLETE:
request->engine->submit_request(request);
break;
case FENCE_FREE:
i915_gem_request_put(request);
break;
}
return NOTIFY_DONE;
}
static int __i915_sw_fence_call
execute_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
{
struct drm_i915_gem_request *request =
container_of(fence, typeof(*request), execute);
switch (state) {
case FENCE_COMPLETE:
break;
case FENCE_FREE:
i915_gem_request_put(request);
break;
}
return NOTIFY_DONE;
}
/**
* i915_gem_request_alloc - allocate a request structure
*
* @engine: engine that we wish to issue the request on.
* @ctx: context that the request will be associated with.
* This can be NULL if the request is not directly related to
* any specific user context, in which case this function will
* choose an appropriate context to use.
*
* Returns a pointer to the allocated request if successful,
* or an error code if not.
*/
struct drm_i915_gem_request *
i915_gem_request_alloc(struct intel_engine_cs *engine,
struct i915_gem_context *ctx)
{
struct drm_i915_private *dev_priv = engine->i915;
struct drm_i915_gem_request *req;
int ret;
lockdep_assert_held(&dev_priv->drm.struct_mutex);
/* ABI: Before userspace accesses the GPU (e.g. execbuffer), report
* EIO if the GPU is already wedged, or EAGAIN to drop the struct_mutex
* and restart.
*/
ret = i915_gem_check_wedge(dev_priv);
if (ret)
return ERR_PTR(ret);
ret = reserve_global_seqno(dev_priv);
if (ret)
return ERR_PTR(ret);
/* Move the oldest request to the slab-cache (if not in use!) */
req = list_first_entry_or_null(&engine->timeline->requests,
typeof(*req), link);
if (req && __i915_gem_request_completed(req))
i915_gem_request_retire(req);
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
/* Beware: Dragons be flying overhead.
*
* We use RCU to look up requests in flight. The lookups may
* race with the request being allocated from the slab freelist.
* That is the request we are writing to here, may be in the process
* of being read by __i915_gem_active_get_rcu(). As such,
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
* we have to be very careful when overwriting the contents. During
* the RCU lookup, we change chase the request->engine pointer,
* read the request->global_seqno and increment the reference count.
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
*
* The reference count is incremented atomically. If it is zero,
* the lookup knows the request is unallocated and complete. Otherwise,
* it is either still in use, or has been reallocated and reset
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
* with dma_fence_init(). This increment is safe for release as we
* check that the request we have a reference to and matches the active
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
* request.
*
* Before we increment the refcount, we chase the request->engine
* pointer. We must not call kmem_cache_zalloc() or else we set
* that pointer to NULL and cause a crash during the lookup. If
* we see the request is completed (based on the value of the
* old engine and seqno), the lookup is complete and reports NULL.
* If we decide the request is not completed (new engine or seqno),
* then we grab a reference and double check that it is still the
* active request - which it won't be and restart the lookup.
*
* Do not use kmem_cache_zalloc() here!
*/
req = kmem_cache_alloc(dev_priv->requests, GFP_KERNEL);
if (!req) {
ret = -ENOMEM;
goto err_unreserve;
}
req->timeline = i915_gem_context_lookup_timeline(ctx, engine);
GEM_BUG_ON(req->timeline == engine->timeline);
spin_lock_init(&req->lock);
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
dma_fence_init(&req->fence,
&i915_fence_ops,
&req->lock,
req->timeline->fence_context,
__timeline_get_seqno(req->timeline->common));
/* We bump the ref for the fence chain */
i915_sw_fence_init(&i915_gem_request_get(req)->submit, submit_notify);
i915_sw_fence_init(&i915_gem_request_get(req)->execute, execute_notify);
/* Ensure that the execute fence completes after the submit fence -
* as we complete the execute fence from within the submit fence
* callback, its completion would otherwise be visible first.
*/
i915_sw_fence_await_sw_fence(&req->execute, &req->submit, &req->execq);
i915_priotree_init(&req->priotree);
drm/i915: Refactor activity tracking for requests With the introduction of requests, we amplified the number of atomic refcounted objects we use and update every execbuffer; from none to several references, and a set of references that need to be changed. We also introduced interesting side-effects in the order of retiring requests and objects. Instead of independently tracking the last request for an object, track the active objects for each request. The object will reside in the buffer list of its most recent active request and so we reduce the kref interchange to a list_move. Now retirements are entirely driven by the request, dramatically simplifying activity tracking on the object themselves, and removing the ambiguity between retiring objects and retiring requests. Furthermore with the consolidation of managing the activity tracking centrally, we can look forward to using RCU to enable lockless lookup of the current active requests for an object. In the future, we will be able to query the status or wait upon rendering to an object without even touching the struct_mutex BKL. All told, less code, simpler and faster, and more extensible. v2: Add a typedef for the function pointer for convenience later. v3: Make the noop retirement callback explicit. Allow passing NULL to the init_request_active() which is expanded to a common noop function. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1470293567-10811-16-git-send-email-chris@chris-wilson.co.uk
2016-08-04 14:52:35 +08:00
INIT_LIST_HEAD(&req->active_list);
req->i915 = dev_priv;
req->engine = engine;
req->ctx = i915_gem_context_get(ctx);
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
/* No zalloc, must clear what we need by hand */
req->global_seqno = 0;
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
req->previous_context = NULL;
req->file_priv = NULL;
req->batch = NULL;
drm/i915: Do not overwrite the request with zero on reallocation When using RCU lookup for the request, commit 0eafec6d3244 ("drm/i915: Enable lockless lookup of request tracking via RCU"), we acknowledge that we may race with another thread that could have reallocated the request. In order for the first thread not to blow up, the second thread must not clear the request completed before overwriting it. In the RCU lookup, we allow for the engine/seqno to be replaced but we do not allow for it to be zeroed. The choice we make is to either add extra checking to the RCU lookup, or embrace the inherent races (as intended). It is more complicated as we need to manually clear everything we depend upon being zero initialised, but we benefit from not emiting the memset() to clear the entire frequently allocated structure (that memset turns up in throughput profiles). And at the same time, the lookup remains flexible for future adjustments. v2: Old style LRC requires another variable to be initialize. (The danger inherent in not zeroing everything.) v3: request->batch also needs to be cleared v4: signaling.tsk is no long used unset, but pid still exists Fixes: 0eafec6d3244 ("drm/i915: Enable lockless lookup of request...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Goel, Akash" <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
2016-08-09 16:23:34 +08:00
/*
* Reserve space in the ring buffer for all the commands required to
* eventually emit this request. This is to guarantee that the
* i915_add_request() call can't fail. Note that the reserve may need
* to be redone if the request is not actually submitted straight
* away, e.g. because a GPU scheduler has deferred it.
*/
req->reserved_space = MIN_SPACE_FOR_ADD_REQUEST;
GEM_BUG_ON(req->reserved_space < engine->emit_breadcrumb_sz);
if (i915.enable_execlists)
ret = intel_logical_ring_alloc_request_extras(req);
else
ret = intel_ring_alloc_request_extras(req);
if (ret)
goto err_ctx;
/* Record the position of the start of the request so that
* should we detect the updated seqno part-way through the
* GPU processing the request, we never over-estimate the
* position of the head.
*/
req->head = req->ring->tail;
return req;
err_ctx:
/* Make sure we didn't add ourselves to external state before freeing */
GEM_BUG_ON(!list_empty(&req->active_list));
GEM_BUG_ON(!list_empty(&req->priotree.signalers_list));
GEM_BUG_ON(!list_empty(&req->priotree.waiters_list));
i915_gem_context_put(ctx);
kmem_cache_free(dev_priv->requests, req);
err_unreserve:
dev_priv->gt.active_requests--;
return ERR_PTR(ret);
}
static int
i915_gem_request_await_request(struct drm_i915_gem_request *to,
struct drm_i915_gem_request *from)
{
int ret;
GEM_BUG_ON(to == from);
if (to->engine->schedule) {
ret = i915_priotree_add_dependency(to->i915,
&to->priotree,
&from->priotree);
if (ret < 0)
return ret;
}
if (to->timeline == from->timeline)
return 0;
if (to->engine == from->engine) {
ret = i915_sw_fence_await_sw_fence_gfp(&to->submit,
&from->submit,
GFP_KERNEL);
return ret < 0 ? ret : 0;
}
if (!from->global_seqno) {
ret = i915_sw_fence_await_dma_fence(&to->submit,
&from->fence, 0,
GFP_KERNEL);
return ret < 0 ? ret : 0;
}
if (from->global_seqno <= to->timeline->sync_seqno[from->engine->id])
return 0;
trace_i915_gem_ring_sync_to(to, from);
if (!i915.semaphores) {
if (!i915_spin_request(from, TASK_INTERRUPTIBLE, 2)) {
ret = i915_sw_fence_await_dma_fence(&to->submit,
&from->fence, 0,
GFP_KERNEL);
if (ret < 0)
return ret;
}
} else {
ret = to->engine->semaphore.sync_to(to, from);
if (ret)
return ret;
}
to->timeline->sync_seqno[from->engine->id] = from->global_seqno;
return 0;
}
int
i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
struct dma_fence *fence)
{
struct dma_fence_array *array;
int ret;
int i;
if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
return 0;
if (dma_fence_is_i915(fence))
return i915_gem_request_await_request(req, to_request(fence));
if (!dma_fence_is_array(fence)) {
ret = i915_sw_fence_await_dma_fence(&req->submit,
fence, I915_FENCE_TIMEOUT,
GFP_KERNEL);
return ret < 0 ? ret : 0;
}
/* Note that if the fence-array was created in signal-on-any mode,
* we should *not* decompose it into its individual fences. However,
* we don't currently store which mode the fence-array is operating
* in. Fortunately, the only user of signal-on-any is private to
* amdgpu and we should not see any incoming fence-array from
* sync-file being in signal-on-any mode.
*/
array = to_dma_fence_array(fence);
for (i = 0; i < array->num_fences; i++) {
struct dma_fence *child = array->fences[i];
if (dma_fence_is_i915(child))
ret = i915_gem_request_await_request(req,
to_request(child));
else
ret = i915_sw_fence_await_dma_fence(&req->submit,
child, I915_FENCE_TIMEOUT,
GFP_KERNEL);
if (ret < 0)
return ret;
}
return 0;
}
/**
* i915_gem_request_await_object - set this request to (async) wait upon a bo
*
* @to: request we are wishing to use
* @obj: object which may be in use on another ring.
*
* This code is meant to abstract object synchronization with the GPU.
* Conceptually we serialise writes between engines inside the GPU.
* We only allow one engine to write into a buffer at any time, but
* multiple readers. To ensure each has a coherent view of memory, we must:
*
* - If there is an outstanding write request to the object, the new
* request must wait for it to complete (either CPU or in hw, requests
* on the same ring will be naturally ordered).
*
* - If we are a write request (pending_write_domain is set), the new
* request must wait for outstanding read requests to complete.
*
* Returns 0 if successful, else propagates up the lower layer error.
*/
int
i915_gem_request_await_object(struct drm_i915_gem_request *to,
struct drm_i915_gem_object *obj,
bool write)
{
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
struct dma_fence *excl;
int ret = 0;
if (write) {
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
struct dma_fence **shared;
unsigned int count, i;
ret = reservation_object_get_fences_rcu(obj->resv,
&excl, &count, &shared);
if (ret)
return ret;
for (i = 0; i < count; i++) {
ret = i915_gem_request_await_dma_fence(to, shared[i]);
if (ret)
break;
dma_fence_put(shared[i]);
}
for (; i < count; i++)
dma_fence_put(shared[i]);
kfree(shared);
} else {
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
excl = reservation_object_get_excl_rcu(obj->resv);
}
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
if (excl) {
if (ret == 0)
ret = i915_gem_request_await_dma_fence(to, excl);
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
dma_fence_put(excl);
}
drm/i915: Move GEM activity tracking into a common struct reservation_object In preparation to support many distinct timelines, we need to expand the activity tracking on the GEM object to handle more than just a request per engine. We already use the struct reservation_object on the dma-buf to handle many fence contexts, so integrating that into the GEM object itself is the preferred solution. (For example, we can now share the same reservation_object between every consumer/producer using this buffer and skip the manual import/export via dma-buf.) v2: Reimplement busy-ioctl (by walking the reservation object), postpone the ABI change for another day. Similarly use the reservation object to find the last_write request (if active and from i915) for choosing display CS flips. Caveats: * busy-ioctl: busy-ioctl only reports on the native fences, it will not warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is being rendered to by external fences. It also will not report the same busy state as wait-ioctl (or polling on the dma-buf) in the same circumstances. On the plus side, it does retain reporting of which *i915* engines are engaged with this object. * non-blocking atomic modesets take a step backwards as the wait for render completion blocks the ioctl. This is fixed in a subsequent patch to use a fence instead for awaiting on the rendering, see "drm/i915: Restore nonblocking awaits for modesetting" * dynamic array manipulation for shared-fences in reservation is slower than the previous lockless static assignment (e.g. gem_exec_lut_handle runtime on ivb goes from 42s to 66s), mainly due to atomic operations (maintaining the fence refcounts). * loss of object-level retirement callbacks, emulated by VMA retirement tracking. * minor loss of object-level last activity information from debugfs, could be replaced with per-vma information if desired Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
2016-10-28 20:58:44 +08:00
return ret;
}
static void i915_gem_mark_busy(const struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
if (dev_priv->gt.awake)
return;
GEM_BUG_ON(!dev_priv->gt.active_requests);
intel_runtime_pm_get_noresume(dev_priv);
dev_priv->gt.awake = true;
intel_enable_gt_powersave(dev_priv);
i915_update_gfx_val(dev_priv);
if (INTEL_GEN(dev_priv) >= 6)
gen6_rps_busy(dev_priv);
queue_delayed_work(dev_priv->wq,
&dev_priv->gt.retire_work,
round_jiffies_up_relative(HZ));
}
/*
* NB: This function is not allowed to fail. Doing so would mean the the
* request is not being tracked for completion but the work itself is
* going to happen on the hardware. This would be a Bad Thing(tm).
*/
void __i915_add_request(struct drm_i915_gem_request *request, bool flush_caches)
{
struct intel_engine_cs *engine = request->engine;
struct intel_ring *ring = request->ring;
struct intel_timeline *timeline = request->timeline;
struct drm_i915_gem_request *prev;
int err;
lockdep_assert_held(&request->i915->drm.struct_mutex);
trace_i915_gem_request_add(request);
/*
* To ensure that this call will not fail, space for its emissions
* should already have been reserved in the ring buffer. Let the ring
* know that it is time to use that space up.
*/
request->reserved_space = 0;
/*
* Emit any outstanding flushes - execbuf can fail to emit the flush
* after having emitted the batchbuffer command. Hence we need to fix
* things up similar to emitting the lazy request. The difference here
* is that the flush _must_ happen before the next request, no matter
* what.
*/
if (flush_caches) {
err = engine->emit_flush(request, EMIT_FLUSH);
/* Not allowed to fail! */
WARN(err, "engine->emit_flush() failed: %d!\n", err);
}
/* Record the position of the start of the breadcrumb so that
* should we detect the updated seqno part-way through the
* GPU processing the request, we never over-estimate the
* position of the ring's HEAD.
*/
err = intel_ring_begin(request, engine->emit_breadcrumb_sz);
GEM_BUG_ON(err);
request->postfix = ring->tail;
ring->tail += engine->emit_breadcrumb_sz * sizeof(u32);
/* Seal the request and mark it as pending execution. Note that
* we may inspect this state, without holding any locks, during
* hangcheck. Hence we apply the barrier to ensure that we do not
* see a more recent value in the hws than we are tracking.
*/
prev = i915_gem_active_raw(&timeline->last_request,
&request->i915->drm.struct_mutex);
if (prev) {
i915_sw_fence_await_sw_fence(&request->submit, &prev->submit,
&request->submitq);
if (engine->schedule)
__i915_priotree_add_dependency(&request->priotree,
&prev->priotree,
&request->dep,
0);
}
spin_lock_irq(&timeline->lock);
list_add_tail(&request->link, &timeline->requests);
spin_unlock_irq(&timeline->lock);
GEM_BUG_ON(i915_seqno_passed(timeline->last_submitted_seqno,
request->fence.seqno));
timeline->last_submitted_seqno = request->fence.seqno;
i915_gem_active_set(&timeline->last_request, request);
list_add_tail(&request->ring_link, &ring->request_list);
request->emitted_jiffies = jiffies;
i915_gem_mark_busy(engine);
/* Let the backend know a new request has arrived that may need
* to adjust the existing execution schedule due to a high priority
* request - i.e. we may want to preempt the current request in order
* to run a high priority dependency chain *before* we can execute this
* request.
*
* This is called before the request is ready to run so that we can
* decide whether to preempt the entire chain so that it is ready to
* run at the earliest possible convenience.
*/
if (engine->schedule)
engine->schedule(request, request->ctx->priority);
local_bh_disable();
i915_sw_fence_commit(&request->submit);
local_bh_enable(); /* Kick the execlists tasklet if just scheduled */
}
static void reset_wait_queue(wait_queue_head_t *q, wait_queue_t *wait)
{
unsigned long flags;
spin_lock_irqsave(&q->lock, flags);
if (list_empty(&wait->task_list))
__add_wait_queue(q, wait);
spin_unlock_irqrestore(&q->lock, flags);
}
static unsigned long local_clock_us(unsigned int *cpu)
{
unsigned long t;
/* Cheaply and approximately convert from nanoseconds to microseconds.
* The result and subsequent calculations are also defined in the same
* approximate microseconds units. The principal source of timing
* error here is from the simple truncation.
*
* Note that local_clock() is only defined wrt to the current CPU;
* the comparisons are no longer valid if we switch CPUs. Instead of
* blocking preemption for the entire busywait, we can detect the CPU
* switch and use that as indicator of system load and a reason to
* stop busywaiting, see busywait_stop().
*/
*cpu = get_cpu();
t = local_clock() >> 10;
put_cpu();
return t;
}
static bool busywait_stop(unsigned long timeout, unsigned int cpu)
{
unsigned int this_cpu;
if (time_after(local_clock_us(&this_cpu), timeout))
return true;
return this_cpu != cpu;
}
bool __i915_spin_request(const struct drm_i915_gem_request *req,
int state, unsigned long timeout_us)
{
unsigned int cpu;
/* When waiting for high frequency requests, e.g. during synchronous
* rendering split between the CPU and GPU, the finite amount of time
* required to set up the irq and wait upon it limits the response
* rate. By busywaiting on the request completion for a short while we
* can service the high frequency waits as quick as possible. However,
* if it is a slow request, we want to sleep as quickly as possible.
* The tradeoff between waiting and sleeping is roughly the time it
* takes to sleep on a request, on the order of a microsecond.
*/
timeout_us += local_clock_us(&cpu);
do {
if (__i915_gem_request_completed(req))
return true;
if (signal_pending_state(state, current))
break;
if (busywait_stop(timeout_us, cpu))
break;
cpu_relax_lowlatency();
} while (!need_resched());
return false;
}
static long
__i915_request_wait_for_execute(struct drm_i915_gem_request *request,
unsigned int flags,
long timeout)
{
const int state = flags & I915_WAIT_INTERRUPTIBLE ?
TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE;
wait_queue_head_t *q = &request->i915->gpu_error.wait_queue;
DEFINE_WAIT(reset);
DEFINE_WAIT(wait);
if (flags & I915_WAIT_LOCKED)
add_wait_queue(q, &reset);
do {
prepare_to_wait(&request->execute.wait, &wait, state);
if (i915_sw_fence_done(&request->execute))
break;
if (flags & I915_WAIT_LOCKED &&
i915_reset_in_progress(&request->i915->gpu_error)) {
__set_current_state(TASK_RUNNING);
i915_reset(request->i915);
reset_wait_queue(q, &reset);
continue;
}
if (signal_pending_state(state, current)) {
timeout = -ERESTARTSYS;
break;
}
timeout = io_schedule_timeout(timeout);
} while (timeout);
finish_wait(&request->execute.wait, &wait);
if (flags & I915_WAIT_LOCKED)
remove_wait_queue(q, &reset);
return timeout;
}
/**
* i915_wait_request - wait until execution of request has finished
* @req: the request to wait upon
* @flags: how to wait
* @timeout: how long to wait in jiffies
*
* i915_wait_request() waits for the request to be completed, for a
* maximum of @timeout jiffies (with MAX_SCHEDULE_TIMEOUT implying an
* unbounded wait).
*
* If the caller holds the struct_mutex, the caller must pass I915_WAIT_LOCKED
* in via the flags, and vice versa if the struct_mutex is not held, the caller
* must not specify that the wait is locked.
*
* Returns the remaining time (in jiffies) if the request completed, which may
* be zero or -ETIME if the request is unfinished after the timeout expires.
* May return -EINTR is called with I915_WAIT_INTERRUPTIBLE and a signal is
* pending before the request completes.
*/
long i915_wait_request(struct drm_i915_gem_request *req,
unsigned int flags,
long timeout)
{
const int state = flags & I915_WAIT_INTERRUPTIBLE ?
TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE;
DEFINE_WAIT(reset);
struct intel_wait wait;
might_sleep();
#if IS_ENABLED(CONFIG_LOCKDEP)
GEM_BUG_ON(debug_locks &&
!!lockdep_is_held(&req->i915->drm.struct_mutex) !=
!!(flags & I915_WAIT_LOCKED));
#endif
GEM_BUG_ON(timeout < 0);
if (i915_gem_request_completed(req))
return timeout;
if (!timeout)
return -ETIME;
trace_i915_gem_request_wait_begin(req);
if (!i915_sw_fence_done(&req->execute)) {
timeout = __i915_request_wait_for_execute(req, flags, timeout);
if (timeout < 0)
goto complete;
GEM_BUG_ON(!i915_sw_fence_done(&req->execute));
}
GEM_BUG_ON(!i915_sw_fence_done(&req->submit));
GEM_BUG_ON(!req->global_seqno);
/* Optimistic short spin before touching IRQs */
if (i915_spin_request(req, state, 5))
goto complete;
set_current_state(state);
if (flags & I915_WAIT_LOCKED)
add_wait_queue(&req->i915->gpu_error.wait_queue, &reset);
intel_wait_init(&wait, req->global_seqno);
if (intel_engine_add_wait(req->engine, &wait))
/* In order to check that we haven't missed the interrupt
* as we enabled it, we need to kick ourselves to do a
* coherent check on the seqno before we sleep.
*/
goto wakeup;
for (;;) {
if (signal_pending_state(state, current)) {
timeout = -ERESTARTSYS;
break;
}
if (!timeout) {
timeout = -ETIME;
break;
}
timeout = io_schedule_timeout(timeout);
if (intel_wait_complete(&wait))
break;
set_current_state(state);
wakeup:
/* Carefully check if the request is complete, giving time
* for the seqno to be visible following the interrupt.
* We also have to check in case we are kicked by the GPU
* reset in order to drop the struct_mutex.
*/
if (__i915_request_irq_complete(req))
break;
/* If the GPU is hung, and we hold the lock, reset the GPU
* and then check for completion. On a full reset, the engine's
* HW seqno will be advanced passed us and we are complete.
* If we do a partial reset, we have to wait for the GPU to
* resume and update the breadcrumb.
*
* If we don't hold the mutex, we can just wait for the worker
* to come along and update the breadcrumb (either directly
* itself, or indirectly by recovering the GPU).
*/
if (flags & I915_WAIT_LOCKED &&
i915_reset_in_progress(&req->i915->gpu_error)) {
__set_current_state(TASK_RUNNING);
i915_reset(req->i915);
reset_wait_queue(&req->i915->gpu_error.wait_queue,
&reset);
continue;
}
/* Only spin if we know the GPU is processing this request */
if (i915_spin_request(req, state, 2))
break;
}
intel_engine_remove_wait(req->engine, &wait);
if (flags & I915_WAIT_LOCKED)
remove_wait_queue(&req->i915->gpu_error.wait_queue, &reset);
__set_current_state(TASK_RUNNING);
complete:
trace_i915_gem_request_wait_end(req);
return timeout;
}
static void engine_retire_requests(struct intel_engine_cs *engine)
{
struct drm_i915_gem_request *request, *next;
list_for_each_entry_safe(request, next,
&engine->timeline->requests, link) {
if (!__i915_gem_request_completed(request))
return;
i915_gem_request_retire(request);
}
}
void i915_gem_retire_requests(struct drm_i915_private *dev_priv)
{
struct intel_engine_cs *engine;
enum intel_engine_id id;
lockdep_assert_held(&dev_priv->drm.struct_mutex);
if (!dev_priv->gt.active_requests)
return;
for_each_engine(engine, dev_priv, id)
engine_retire_requests(engine);
}