[Intel-gfx] [PATCH v3] drm/i915: Remove the preempted request from the execution queue
Chris Wilson
chris at chris-wilson.co.uk
Wed Feb 22 18:53:17 UTC 2017
After the request is cancelled, we then need to remove it from the
global execution timeline and return it to the context timeline, the
inverse of submit_request().
v2: Move manipulation of struct intel_wait to helpers
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_gem_request.c | 66 ++++++++++++++++++++--
drivers/gpu/drm/i915/i915_gem_request.h | 3 +
drivers/gpu/drm/i915/intel_breadcrumbs.c | 17 +++++-
drivers/gpu/drm/i915/intel_ringbuffer.h | 37 +++++++++++-
drivers/gpu/drm/i915/selftests/intel_breadcrumbs.c | 6 +-
5 files changed, 117 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index d18f450977e0..76e31cd7840e 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -441,6 +441,55 @@ void i915_gem_request_submit(struct drm_i915_gem_request *request)
spin_unlock_irqrestore(&engine->timeline->lock, flags);
}
+void __i915_gem_request_unsubmit(struct drm_i915_gem_request *request)
+{
+ struct intel_engine_cs *engine = request->engine;
+ struct intel_timeline *timeline;
+
+ assert_spin_locked(&engine->timeline->lock);
+
+ /* Only unwind in reverse order, required so that the per-context list
+ * is kept in seqno/ring order.
+ */
+ GEM_BUG_ON(request->global_seqno != engine->timeline->seqno);
+ engine->timeline->seqno--;
+
+ /* We may be recursing from the signal callback of another i915 fence */
+ spin_lock_nested(&request->lock, SINGLE_DEPTH_NESTING);
+ request->global_seqno = 0;
+ if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags))
+ intel_engine_cancel_signaling(request);
+ spin_unlock(&request->lock);
+
+ /* Transfer back from the global per-engine timeline to per-context */
+ timeline = request->timeline;
+ GEM_BUG_ON(timeline == engine->timeline);
+
+ spin_lock(&timeline->lock);
+ list_move(&request->link, &timeline->requests);
+ spin_unlock(&timeline->lock);
+
+ /* We don't need to wake_up any waiters on request->execute, they
+ * will get woken by any other event or us re-adding this request
+ * to the engine timeline (__i915_gem_request_submit()). The waiters
+ * should be quite adapt at finding that the request now has a new
+ * global_seqno to the one they went to sleep on.
+ */
+}
+
+void i915_gem_request_unsubmit(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_unsubmit(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)
{
@@ -1034,15 +1083,15 @@ long i915_wait_request(struct drm_i915_gem_request *req,
if (flags & I915_WAIT_LOCKED)
add_wait_queue(errq, &reset);
- intel_wait_init(&wait, i915_gem_request_global_seqno(req));
+ intel_wait_init(&wait);
+restart:
reset_wait_queue(&req->execute, &exec);
- if (!wait.seqno) {
+ if (!intel_wait_update_request(&wait, req)) {
do {
set_current_state(state);
- wait.seqno = i915_gem_request_global_seqno(req);
- if (wait.seqno)
+ if (intel_wait_update_request(&wait, req))
break;
if (flags & I915_WAIT_LOCKED &&
@@ -1070,7 +1119,7 @@ long i915_wait_request(struct drm_i915_gem_request *req,
if (timeout < 0)
goto complete;
- GEM_BUG_ON(!wait.seqno);
+ GEM_BUG_ON(!intel_wait_has_seqno(&wait));
}
GEM_BUG_ON(!i915_sw_fence_signaled(&req->submit));
@@ -1100,7 +1149,7 @@ long i915_wait_request(struct drm_i915_gem_request *req,
timeout = io_schedule_timeout(timeout);
if (intel_wait_complete(&wait) &&
- i915_gem_request_global_seqno(req) == wait.seqno)
+ intel_wait_check_request(&wait, req))
break;
set_current_state(state);
@@ -1135,6 +1184,11 @@ long i915_wait_request(struct drm_i915_gem_request *req,
/* Only spin if we know the GPU is processing this request */
if (i915_spin_request(req, state, 2))
break;
+
+ if (!intel_wait_check_request(&wait, req)) {
+ intel_engine_remove_wait(req->engine, &wait);
+ goto restart;
+ }
}
intel_engine_remove_wait(req->engine, &wait);
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index b81f6709905c..5f73d8c0a38a 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -274,6 +274,9 @@ void __i915_add_request(struct drm_i915_gem_request *req, bool flush_caches);
void __i915_gem_request_submit(struct drm_i915_gem_request *request);
void i915_gem_request_submit(struct drm_i915_gem_request *request);
+void __i915_gem_request_unsubmit(struct drm_i915_gem_request *request);
+void i915_gem_request_unsubmit(struct drm_i915_gem_request *request);
+
struct intel_rps_client;
#define NO_WAITBOOST ERR_PTR(-1)
#define IS_RPS_CLIENT(p) (!IS_ERR(p))
diff --git a/drivers/gpu/drm/i915/intel_breadcrumbs.c b/drivers/gpu/drm/i915/intel_breadcrumbs.c
index 46e7fca4b189..ac8663d8e73f 100644
--- a/drivers/gpu/drm/i915/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c
@@ -453,7 +453,12 @@ void intel_engine_remove_wait(struct intel_engine_cs *engine,
spin_unlock_irq(&b->lock);
}
-static bool signal_complete(struct drm_i915_gem_request *request)
+static bool signal_valid(const struct drm_i915_gem_request *request)
+{
+ return intel_wait_check_request(&request->signaling.wait, request);
+}
+
+static bool signal_complete(const struct drm_i915_gem_request *request)
{
if (!request)
return false;
@@ -462,7 +467,7 @@ static bool signal_complete(struct drm_i915_gem_request *request)
* signalled that this wait is already completed.
*/
if (intel_wait_complete(&request->signaling.wait))
- return true;
+ return signal_valid(request);
/* Carefully check if the request is complete, giving time for the
* seqno to be visible or if the GPU hung.
@@ -542,13 +547,21 @@ static int intel_breadcrumbs_signaler(void *arg)
i915_gem_request_put(request);
} else {
+ DEFINE_WAIT(exec);
+
if (kthread_should_stop()) {
GEM_BUG_ON(request);
break;
}
+ if (request)
+ add_wait_queue(&request->execute, &exec);
+
schedule();
+ if (request)
+ remove_wait_queue(&request->execute, &exec);
+
if (kthread_should_park())
kthread_parkme();
}
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 45d2c2fa946e..fe724d477362 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -582,10 +582,45 @@ static inline u32 intel_hws_seqno_address(struct intel_engine_cs *engine)
/* intel_breadcrumbs.c -- user interrupt bottom-half for waiters */
int intel_engine_init_breadcrumbs(struct intel_engine_cs *engine);
-static inline void intel_wait_init(struct intel_wait *wait, u32 seqno)
+static inline void intel_wait_init(struct intel_wait *wait)
{
wait->tsk = current;
+}
+
+static inline void intel_wait_init_for_seqno(struct intel_wait *wait, u32 seqno)
+{
+ wait->tsk = current;
+ wait->seqno = seqno;
+}
+
+static inline bool intel_wait_has_seqno(const struct intel_wait *wait)
+{
+ return wait->seqno;
+}
+
+static inline bool intel_wait_update_seqno(struct intel_wait *wait, u32 seqno)
+{
wait->seqno = seqno;
+ return intel_wait_has_seqno(wait);
+}
+
+static inline bool intel_wait_update_request(struct intel_wait *wait,
+ struct drm_i915_gem_request *rq)
+{
+ return intel_wait_update_seqno(wait, i915_gem_request_global_seqno(rq));
+}
+
+static inline bool intel_wait_check_seqno(const struct intel_wait *wait,
+ u32 seqno)
+{
+ return wait->seqno == seqno;
+}
+
+static inline bool
+intel_wait_check_request(const struct intel_wait *wait,
+ const struct drm_i915_gem_request *rq)
+{
+ return intel_wait_check_seqno(wait, i915_gem_request_global_seqno(rq));
}
static inline bool intel_wait_complete(const struct intel_wait *wait)
diff --git a/drivers/gpu/drm/i915/selftests/intel_breadcrumbs.c b/drivers/gpu/drm/i915/selftests/intel_breadcrumbs.c
index 6426acc9fdca..621be1ca53d8 100644
--- a/drivers/gpu/drm/i915/selftests/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/selftests/intel_breadcrumbs.c
@@ -131,7 +131,7 @@ static int igt_random_insert_remove(void *arg)
goto out_bitmap;
for (n = 0; n < count; n++)
- intel_wait_init(&waiters[n], seqno_bias + n);
+ intel_wait_init_for_seqno(&waiters[n], seqno_bias + n);
err = check_rbtree(engine, bitmap, waiters, count);
if (err)
@@ -197,7 +197,7 @@ static int igt_insert_complete(void *arg)
goto out_waiters;
for (n = 0; n < count; n++) {
- intel_wait_init(&waiters[n], n + seqno_bias);
+ intel_wait_init_for_seqno(&waiters[n], n + seqno_bias);
intel_engine_add_wait(engine, &waiters[n]);
__set_bit(n, bitmap);
}
@@ -318,7 +318,7 @@ static int igt_wakeup_thread(void *arg)
while (wait_for_ready(w)) {
GEM_BUG_ON(kthread_should_stop());
- intel_wait_init(&wait, w->seqno);
+ intel_wait_init_for_seqno(&wait, w->seqno);
intel_engine_add_wait(w->engine, &wait);
for (;;) {
set_current_state(TASK_UNINTERRUPTIBLE);
--
2.11.0
More information about the Intel-gfx
mailing list