drm/i915: Replace global breadcrumbs with per-context interrupt tracking

A few years ago, see commit 688e6c725816 ("drm/i915: Slaughter the
thundering i915_wait_request herd"), the issue of handling multiple
clients waiting in parallel was brought to our attention. The
requirement was that every client should be woken immediately upon its
request being signaled, without incurring any cpu overhead.

To handle certain fragility of our hw meant that we could not do a
simple check inside the irq handler (some generations required almost
unbounded delays before we could be sure of seqno coherency) and so
request completion checking required delegation.

Before commit 688e6c725816, the solution was simple. Every client
waiting on a request would be woken on every interrupt and each would do
a heavyweight check to see if their request was complete. Commit
688e6c725816 introduced an rbtree so that only the earliest waiter on
the global timeline would woken, and would wake the next and so on.
(Along with various complications to handle requests being reordered
along the global timeline, and also a requirement for kthread to provide
a delegate for fence signaling that had no process context.)

The global rbtree depends on knowing the execution timeline (and global
seqno). Without knowing that order, we must instead check all contexts
queued to the HW to see which may have advanced. We trim that list by
only checking queued contexts that are being waited on, but still we
keep a list of all active contexts and their active signalers that we
inspect from inside the irq handler. By moving the waiters onto the fence
signal list, we can combine the client wakeup with the dma_fence
signaling (a dramatic reduction in complexity, but does require the HW
being coherent, the seqno must be visible from the cpu before the
interrupt is raised - we keep a timer backup just in case).

Having previously fixed all the issues with irq-seqno serialisation (by
inserting delays onto the GPU after each request instead of random delays
on the CPU after each interrupt), we can rely on the seqno state to
perfom direct wakeups from the interrupt handler. This allows us to
preserve our single context switch behaviour of the current routine,
with the only downside that we lose the RT priority sorting of wakeups.
In general, direct wakeup latency of multiple clients is about the same
(about 10% better in most cases) with a reduction in total CPU time spent
in the waiter (about 20-50% depending on gen). Average herd behaviour is
improved, but at the cost of not delegating wakeups on task_prio.

v2: Capture fence signaling state for error state and add comments to
warm even the most cold of hearts.
v3: Check if the request is still active before busywaiting
v4: Reduce the amount of pointer misdirection with list_for_each_safe
and using a local i915_request variable inside the loops
v5: Add a missing pluralisation to a purely informative selftest message.

References: 688e6c725816 ("drm/i915: Slaughter the thundering i915_wait_request herd")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190129205230.19056-2-chris@chris-wilson.co.uk

1 files changed, 8 insertions, 75 deletions

diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 898e06014295..304a7ef7f7fb 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -447,9 +447,14 @@ static void error_print_request(struct drm_i915_error_state_buf *m,
 	if (!erq->seqno)
 		return;
 
-	err_printf(m, "%s pid %d, ban score %d, seqno %8x:%08x, prio %d, emitted %dms, start %08x, head %08x, tail %08x\n",
+	err_printf(m, "%s pid %d, ban score %d, seqno %8x:%08x%s%s, prio %d, emitted %dms, start %08x, head %08x, tail %08x\n",
 		   prefix, erq->pid, erq->ban_score,
-		   erq->context, erq->seqno, erq->sched_attr.priority,
+		   erq->context, erq->seqno,
+		   test_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
+			    &erq->flags) ? "!" : "",
+		   test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
+			    &erq->flags) ? "+" : "",
+		   erq->sched_attr.priority,
 		   jiffies_to_msecs(erq->jiffies - epoch),
 		   erq->start, erq->head, erq->tail);
 }
@@ -530,7 +535,6 @@ static void error_print_engine(struct drm_i915_error_state_buf *m,
 	}
 	err_printf(m, "  seqno: 0x%08x\n", ee->seqno);
 	err_printf(m, "  last_seqno: 0x%08x\n", ee->last_seqno);
-	err_printf(m, "  waiting: %s\n", yesno(ee->waiting));
 	err_printf(m, "  ring->head: 0x%08x\n", ee->cpu_ring_head);
 	err_printf(m, "  ring->tail: 0x%08x\n", ee->cpu_ring_tail);
 	err_printf(m, "  hangcheck timestamp: %dms (%lu%s)\n",
@@ -804,21 +808,6 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m,
 						    error->epoch);
 		}
 
-		if (IS_ERR(ee->waiters)) {
-			err_printf(m, "%s --- ? waiters [unable to acquire spinlock]\n",
-				   m->i915->engine[i]->name);
-		} else if (ee->num_waiters) {
-			err_printf(m, "%s --- %d waiters\n",
-				   m->i915->engine[i]->name,
-				   ee->num_waiters);
-			for (j = 0; j < ee->num_waiters; j++) {
-				err_printf(m, " seqno 0x%08x for %s [%d]\n",
-					   ee->waiters[j].seqno,
-					   ee->waiters[j].comm,
-					   ee->waiters[j].pid);
-			}
-		}
-
 		print_error_obj(m, m->i915->engine[i],
 				"ringbuffer", ee->ringbuffer);
 
@@ -1000,8 +989,6 @@ void __i915_gpu_state_free(struct kref *error_ref)
 		i915_error_object_free(ee->wa_ctx);
 
 		kfree(ee->requests);
-		if (!IS_ERR_OR_NULL(ee->waiters))
-			kfree(ee->waiters);
 	}
 
 	for (i = 0; i < ARRAY_SIZE(error->active_bo); i++)
@@ -1205,59 +1192,6 @@ static void gen6_record_semaphore_state(struct intel_engine_cs *engine,
 			I915_READ(RING_SYNC_2(engine->mmio_base));
 }
 
-static void error_record_engine_waiters(struct intel_engine_cs *engine,
-					struct drm_i915_error_engine *ee)
-{
-	struct intel_breadcrumbs *b = &engine->breadcrumbs;
-	struct drm_i915_error_waiter *waiter;
-	struct rb_node *rb;
-	int count;
-
-	ee->num_waiters = 0;
-	ee->waiters = NULL;
-
-	if (RB_EMPTY_ROOT(&b->waiters))
-		return;
-
-	if (!spin_trylock_irq(&b->rb_lock)) {
-		ee->waiters = ERR_PTR(-EDEADLK);
-		return;
-	}
-
-	count = 0;
-	for (rb = rb_first(&b->waiters); rb != NULL; rb = rb_next(rb))
-		count++;
-	spin_unlock_irq(&b->rb_lock);
-
-	waiter = NULL;
-	if (count)
-		waiter = kmalloc_array(count,
-				       sizeof(struct drm_i915_error_waiter),
-				       GFP_ATOMIC);
-	if (!waiter)
-		return;
-
-	if (!spin_trylock_irq(&b->rb_lock)) {
-		kfree(waiter);
-		ee->waiters = ERR_PTR(-EDEADLK);
-		return;
-	}
-
-	ee->waiters = waiter;
-	for (rb = rb_first(&b->waiters); rb; rb = rb_next(rb)) {
-		struct intel_wait *w = rb_entry(rb, typeof(*w), node);
-
-		strcpy(waiter->comm, w->tsk->comm);
-		waiter->pid = w->tsk->pid;
-		waiter->seqno = w->seqno;
-		waiter++;
-
-		if (++ee->num_waiters == count)
-			break;
-	}
-	spin_unlock_irq(&b->rb_lock);
-}
-
 static void error_record_engine_registers(struct i915_gpu_state *error,
 					  struct intel_engine_cs *engine,
 					  struct drm_i915_error_engine *ee)
@@ -1293,7 +1227,6 @@ static void error_record_engine_registers(struct i915_gpu_state *error,
 
 	intel_engine_get_instdone(engine, &ee->instdone);
 
-	ee->waiting = intel_engine_has_waiter(engine);
 	ee->instpm = I915_READ(RING_INSTPM(engine->mmio_base));
 	ee->acthd = intel_engine_get_active_head(engine);
 	ee->seqno = intel_engine_get_seqno(engine);
@@ -1367,6 +1300,7 @@ static void record_request(struct i915_request *request,
 {
 	struct i915_gem_context *ctx = request->gem_context;
 
+	erq->flags = request->fence.flags;
 	erq->context = ctx->hw_id;
 	erq->sched_attr = request->sched.attr;
 	erq->ban_score = atomic_read(&ctx->ban_score);
@@ -1542,7 +1476,6 @@ static void gem_record_rings(struct i915_gpu_state *error)
 		ee->engine_id = i;
 
 		error_record_engine_registers(error, engine, ee);
-		error_record_engine_waiters(engine, ee);
 		error_record_engine_execlists(engine, ee);
 
 		request = i915_gem_find_active_request(engine);