[PATCH 26/50] drm/i915: Apply an execution_mask to the virtual_engine

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Allow the user to direct which physical engines of the virtual engine
they wish to execute one, as sometimes it is necessary to override the
load balancing algorithm.

Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
---
 drivers/gpu/drm/i915/gt/intel_lrc.c    |  58 +++++++++++
 drivers/gpu/drm/i915/gt/selftest_lrc.c | 131 +++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_request.c    |   1 +
 drivers/gpu/drm/i915/i915_request.h    |   3 +
 4 files changed, 193 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 43894c2cc5ee..3833c8d2c28c 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -552,6 +552,18 @@ execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
 	intel_engine_context_out(rq->engine);
 	execlists_context_status_change(rq, status);
 	trace_i915_request_out(rq);
+
+	/*
+	 * If this is part of a virtual engine, its next request may have
+	 * been blocked waiting for access to the active context. We have
+	 * to kick all the siblings again in case we need to switch (e.g.
+	 * the next request is not runnable on this engine). Hopefully,
+	 * we will already have submitted the next request before the
+	 * tasklet runs and do not need to rebuild each virtual tree
+	 * and kick everyone again.
+	 */
+	if (rq->engine != rq->hw_context->engine)
+		tasklet_schedule(&rq->hw_context->engine->execlists.tasklet);
 }
 
 static u64 execlists_update_context(struct i915_request *rq)
@@ -779,6 +791,9 @@ static bool virtual_matches(const struct virtual_engine *ve,
 {
 	const struct intel_engine_cs *active;
 
+	if (!(rq->execution_mask & engine->mask)) /* We peeked too soon! */
+		return false;
+
 	/*
 	 * We track when the HW has completed saving the context image
 	 * (i.e. when we have seen the final CS event switching out of
@@ -3138,12 +3153,44 @@ static const struct intel_context_ops virtual_context_ops = {
 	.destroy = virtual_context_destroy,
 };
 
+static intel_engine_mask_t virtual_submission_mask(struct virtual_engine *ve)
+{
+	struct i915_request *rq;
+	intel_engine_mask_t mask;
+
+	rq = READ_ONCE(ve->request);
+	if (!rq)
+		return 0;
+
+	/* The rq is ready for submission; rq->execution_mask is now stable. */
+	mask = rq->execution_mask;
+	if (unlikely(!mask)) {
+		/* Invalid selection, submit to a random engine in error */
+		i915_request_skip(rq, -ENODEV);
+		mask = ve->siblings[0]->mask;
+	}
+
+	GEM_TRACE("%s: rq=%llx:%lld, mask=%x, prio=%d\n",
+		  ve->base.name,
+		  rq->fence.context, rq->fence.seqno,
+		  mask, ve->base.execlists.queue_priority_hint);
+
+	return mask;
+}
+
 static void virtual_submission_tasklet(unsigned long data)
 {
 	struct virtual_engine * const ve = (struct virtual_engine *)data;
 	const int prio = ve->base.execlists.queue_priority_hint;
+	intel_engine_mask_t mask;
 	unsigned int n;
 
+	rcu_read_lock();
+	mask = virtual_submission_mask(ve);
+	rcu_read_unlock();
+	if (unlikely(!mask))
+		return;
+
 	local_irq_disable();
 	for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
 		struct intel_engine_cs *sibling = ve->siblings[n];
@@ -3151,6 +3198,17 @@ static void virtual_submission_tasklet(unsigned long data)
 		struct rb_node **parent, *rb;
 		bool first;
 
+		if (unlikely(!(mask & sibling->mask))) {
+			if (!RB_EMPTY_NODE(&node->rb)) {
+				spin_lock(&sibling->timeline.lock);
+				rb_erase_cached(&node->rb,
+						&sibling->execlists.virtual);
+				RB_CLEAR_NODE(&node->rb);
+				spin_unlock(&sibling->timeline.lock);
+			}
+			continue;
+		}
+
 		spin_lock(&sibling->timeline.lock);
 
 		if (!RB_EMPTY_NODE(&node->rb)) {
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index f34aa9e042a3..209e51ef13e6 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1480,6 +1480,136 @@ static int live_virtual_engine(void *arg)
 	return err;
 }
 
+static int mask_virtual_engine(struct drm_i915_private *i915,
+			       struct intel_engine_cs **siblings,
+			       unsigned int nsibling)
+{
+	struct i915_request *request[MAX_ENGINE_INSTANCE + 1];
+	struct i915_gem_context *ctx;
+	struct intel_context *ve;
+	struct igt_live_test t;
+	unsigned int n;
+	int err;
+
+	/*
+	 * Check that by setting the execution mask on a request, we can
+	 * restrict it to our desired engine within the virtual engine.
+	 */
+
+	ctx = kernel_context(i915);
+	if (!ctx)
+		return -ENOMEM;
+
+	ve = intel_execlists_create_virtual(ctx, siblings, nsibling);
+	if (IS_ERR(ve)) {
+		err = PTR_ERR(ve);
+		goto out_close;
+	}
+
+	err = intel_context_pin(ve);
+	if (err)
+		goto out_put;
+
+	err = igt_live_test_begin(&t, i915, __func__, ve->engine->name);
+	if (err)
+		goto out_unpin;
+
+	for (n = 0; n < nsibling; n++) {
+		request[n] = i915_request_create(ve);
+		if (IS_ERR(request)) {
+			err = PTR_ERR(request);
+			nsibling = n;
+			goto out;
+		}
+
+		/* Reverse order as it's more likely to be unnatural */
+		request[n]->execution_mask = siblings[nsibling - n - 1]->mask;
+
+		i915_request_get(request[n]);
+		i915_request_add(request[n]);
+	}
+
+	for (n = 0; n < nsibling; n++) {
+		if (i915_request_wait(request[n], I915_WAIT_LOCKED, HZ / 10) < 0) {
+			pr_err("%s(%s): wait for %llx:%lld timed out\n",
+			       __func__, ve->engine->name,
+			       request[n]->fence.context,
+			       request[n]->fence.seqno);
+
+			GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
+				  __func__, ve->engine->name,
+				  request[n]->fence.context,
+				  request[n]->fence.seqno);
+			GEM_TRACE_DUMP();
+			i915_gem_set_wedged(i915);
+			err = -EIO;
+			goto out;
+		}
+
+		if (request[n]->engine != siblings[nsibling - n - 1]) {
+			pr_err("Executed on wrong sibling '%s', expected '%s'\n",
+			       request[n]->engine->name,
+			       siblings[nsibling - n - 1]->name);
+			err = -EINVAL;
+			goto out;
+		}
+	}
+
+	err = igt_live_test_end(&t);
+	if (err)
+		goto out;
+
+out:
+	if (igt_flush_test(i915, I915_WAIT_LOCKED))
+		err = -EIO;
+
+	for (n = 0; n < nsibling; n++)
+		i915_request_put(request[n]);
+
+out_unpin:
+	intel_context_unpin(ve);
+out_put:
+	intel_context_put(ve);
+out_close:
+	kernel_context_close(ctx);
+	return err;
+}
+
+static int live_virtual_mask(void *arg)
+{
+	struct drm_i915_private *i915 = arg;
+	struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
+	unsigned int class, inst;
+	int err = 0;
+
+	if (USES_GUC_SUBMISSION(i915))
+		return 0;
+
+	mutex_lock(&i915->drm.struct_mutex);
+
+	for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
+		unsigned int nsibling;
+
+		nsibling = 0;
+		for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
+			if (!i915->engine_class[class][inst])
+				break;
+
+			siblings[nsibling++] = i915->engine_class[class][inst];
+		}
+		if (nsibling < 2)
+			continue;
+
+		err = mask_virtual_engine(i915, siblings, nsibling);
+		if (err)
+			goto out_unlock;
+	}
+
+out_unlock:
+	mutex_unlock(&i915->drm.struct_mutex);
+	return err;
+}
+
 int intel_execlists_live_selftests(struct drm_i915_private *i915)
 {
 	static const struct i915_subtest tests[] = {
@@ -1493,6 +1623,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
 		SUBTEST(live_preempt_hang),
 		SUBTEST(live_preempt_smoke),
 		SUBTEST(live_virtual_engine),
+		SUBTEST(live_virtual_mask),
 	};
 
 	if (!HAS_EXECLISTS(i915))
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 891e31d54307..9352406d5504 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -687,6 +687,7 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
 	rq->batch = NULL;
 	rq->capture_list = NULL;
 	rq->waitboost = false;
+	rq->execution_mask = ALL_ENGINES;
 
 	INIT_LIST_HEAD(&rq->active_list);
 	INIT_LIST_HEAD(&rq->execute_cb);
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 8025a89b5999..d7f9b2194568 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -28,6 +28,8 @@
 #include <linux/dma-fence.h>
 #include <linux/lockdep.h>
 
+#include "gt/intel_engine_types.h"
+
 #include "i915_gem.h"
 #include "i915_scheduler.h"
 #include "i915_selftest.h"
@@ -156,6 +158,7 @@ struct i915_request {
 	 */
 	struct i915_sched_node sched;
 	struct i915_dependency dep;
+	intel_engine_mask_t execution_mask;
 
 	/*
 	 * A convenience pointer to the current breadcrumb value stored in
-- 
2.20.1

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




[Index of Archives]     [AMD Graphics]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux