drm/i915/guc: Support request cancellation

This adds GuC backend support for i915_request_cancel(), which in turn
makes CONFIG_DRM_I915_REQUEST_TIMEOUT work.

This implementation makes use of fence while there are likely simplier
options. A fence was chosen because of another feature coming soon
which requires a user to block on a context until scheduling is
disabled. In that case we return the fence to the user and the user can
wait on that fence.

v2:
 (Daniele)
  - A comment about locking the blocked incr / decr
  - A comments about the use of the fence
  - Update commit message explaining why fence
  - Delete redundant check blocked count in unblock function
  - Ring buffer implementation
  - Comment about blocked in submission path
  - Shorter rpm path
v3:
 (Checkpatch)
  - Fix typos in commit message
 (Daniel)
  - Rework to simplier locking structure in guc_context_block / unblock

Signed-off-by: Matthew Brost <[email protected]>
Cc: Tvrtko Ursulin <[email protected]>
Reviewed-by: Daniele Ceraolo Spurio <[email protected]>
Signed-off-by: John Harrison <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

Author: mbrost05

drivers/gpu/drm/i915/gt/intel_context.c

@@ -366,6 +366,12 @@ static int __intel_context_active(struct i915_active *active)
 	return 0;
 }
 
+static int sw_fence_dummy_notify(struct i915_sw_fence *sf,
+				 enum i915_sw_fence_notify state)
+{
+	return NOTIFY_DONE;
+}
+
 void
 intel_context_init(struct intel_context *ce, struct intel_engine_cs *engine)
 {
@@ -399,6 +405,13 @@ intel_context_init(struct intel_context *ce, struct intel_engine_cs *engine)
 	ce->guc_id = GUC_INVALID_LRC_ID;
 	INIT_LIST_HEAD(&ce->guc_id_link);
 
+	/*
+	 * Initialize fence to be complete as this is expected to be complete
+	 * unless there is a pending schedule disable outstanding.
+	 */
+	i915_sw_fence_init(&ce->guc_blocked, sw_fence_dummy_notify);
+	i915_sw_fence_commit(&ce->guc_blocked);
+
 	i915_active_init(&ce->active,
 			 __intel_context_active, __intel_context_retire, 0);
 }

drivers/gpu/drm/i915/gt/intel_context.h

@@ -70,6 +70,13 @@ intel_context_is_pinned(struct intel_context *ce)
 	return atomic_read(&ce->pin_count);
 }
 
+static inline void intel_context_cancel_request(struct intel_context *ce,
+						struct i915_request *rq)
+{
+	GEM_BUG_ON(!ce->ops->cancel_request);
+	return ce->ops->cancel_request(ce, rq);
+}
+
 /**
  * intel_context_unlock_pinned - Releases the earlier locking of 'pinned' status
  * @ce - the context

drivers/gpu/drm/i915/gt/intel_context_types.h

@@ -13,6 +13,7 @@
 #include <linux/types.h>
 
 #include "i915_active_types.h"
+#include "i915_sw_fence.h"
 #include "i915_utils.h"
 #include "intel_engine_types.h"
 #include "intel_sseu.h"
@@ -42,6 +43,9 @@ struct intel_context_ops {
 	void (*unpin)(struct intel_context *ce);
 	void (*post_unpin)(struct intel_context *ce);
 
+	void (*cancel_request)(struct intel_context *ce,
+			       struct i915_request *rq);
+
 	void (*enter)(struct intel_context *ce);
 	void (*exit)(struct intel_context *ce);
 
@@ -156,7 +160,7 @@ struct intel_context {
 		 * sched_state: scheduling state of this context using GuC
 		 * submission
 		 */
-		u8 sched_state;
+		u16 sched_state;
 		/*
 		 * fences: maintains of list of requests that have a submit
 		 * fence related to GuC submission
@@ -184,6 +188,9 @@ struct intel_context {
 	 * GuC ID link - in list when unpinned but guc_id still valid in GuC
 	 */
 	struct list_head guc_id_link;
+
+	/* GuC context blocked fence */
+	struct i915_sw_fence guc_blocked;
 };
 
 #endif /* __INTEL_CONTEXT_TYPES__ */

drivers/gpu/drm/i915/gt/intel_execlists_submission.c

@@ -114,6 +114,7 @@
 #include "gen8_engine_cs.h"
 #include "intel_breadcrumbs.h"
 #include "intel_context.h"
+#include "intel_engine_heartbeat.h"
 #include "intel_engine_pm.h"
 #include "intel_engine_stats.h"
 #include "intel_execlists_submission.h"
@@ -2587,11 +2588,26 @@ static int execlists_context_alloc(struct intel_context *ce)
 	return lrc_alloc(ce, ce->engine);
 }
 
+static void execlists_context_cancel_request(struct intel_context *ce,
+					     struct i915_request *rq)
+{
+	struct intel_engine_cs *engine = NULL;
+
+	i915_request_active_engine(rq, &engine);
+
+	if (engine && intel_engine_pulse(engine))
+		intel_gt_handle_error(engine->gt, engine->mask, 0,
+				      "request cancellation by %s",
+				      current->comm);
+}
+
 static const struct intel_context_ops execlists_context_ops = {
 	.flags = COPS_HAS_INFLIGHT,
 
 	.alloc = execlists_context_alloc,
 
+	.cancel_request = execlists_context_cancel_request,
+
 	.pre_pin = execlists_context_pre_pin,
 	.pin = execlists_context_pin,
 	.unpin = lrc_unpin,
@@ -3608,6 +3624,8 @@ static const struct intel_context_ops virtual_context_ops = {
 
 	.alloc = virtual_context_alloc,
 
+	.cancel_request = execlists_context_cancel_request,
+
 	.pre_pin = virtual_context_pre_pin,
 	.pin = virtual_context_pin,
 	.unpin = lrc_unpin,

drivers/gpu/drm/i915/gt/intel_ring_submission.c

@@ -16,6 +16,7 @@
 #include "intel_reset.h"
 #include "intel_ring.h"
 #include "shmem_utils.h"
+#include "intel_engine_heartbeat.h"
 
 /* Rough estimate of the typical request size, performing a flush,
  * set-context and then emitting the batch.
@@ -604,9 +605,24 @@ static void ring_context_ban(struct intel_context *ce,
 		}
 }
 
+static void ring_context_cancel_request(struct intel_context *ce,
+					struct i915_request *rq)
+{
+	struct intel_engine_cs *engine = NULL;
+
+	i915_request_active_engine(rq, &engine);
+
+	if (engine && intel_engine_pulse(engine))
+		intel_gt_handle_error(engine->gt, engine->mask, 0,
+				      "request cancellation by %s",
+				      current->comm);
+}
+
 static const struct intel_context_ops ring_context_ops = {
 	.alloc = ring_context_alloc,
 
+	.cancel_request = ring_context_cancel_request,
+
 	.ban = ring_context_ban,
 
 	.pre_pin = ring_context_pre_pin,