drm/radeon: cope with foreign fences inside the reservation object

Not the whole world is a radeon! :-)
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/radeon/cik.c

@@ -3993,7 +3993,7 @@ struct radeon_fence *cik_copy_cpdma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	for (i = 0; i < num_loops; i++) {

drivers/gpu/drm/radeon/cik_sdma.c

@@ -571,7 +571,7 @@ struct radeon_fence *cik_copy_dma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	for (i = 0; i < num_loops; i++) {

drivers/gpu/drm/radeon/evergreen_dma.c

@@ -133,7 +133,7 @@ struct radeon_fence *evergreen_copy_dma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	for (i = 0; i < num_loops; i++) {

drivers/gpu/drm/radeon/r600.c

@@ -2912,7 +2912,7 @@ struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));

drivers/gpu/drm/radeon/r600_dma.c

@@ -470,7 +470,7 @@ struct radeon_fence *r600_copy_dma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	for (i = 0; i < num_loops; i++) {

drivers/gpu/drm/radeon/radeon.h

@@ -589,7 +589,8 @@ bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ring,
 				struct radeon_semaphore *semaphore);
 void radeon_semaphore_sync_fence(struct radeon_semaphore *semaphore,
 				 struct radeon_fence *fence);
-void radeon_semaphore_sync_resv(struct radeon_semaphore *semaphore,
+int radeon_semaphore_sync_resv(struct radeon_device *rdev,
+			       struct radeon_semaphore *semaphore,
 			       struct reservation_object *resv,
 			       bool shared);
 int radeon_semaphore_sync_rings(struct radeon_device *rdev,

drivers/gpu/drm/radeon/radeon_cs.c

@@ -249,9 +249,9 @@ static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority
 	return 0;
 }
 
-static void radeon_cs_sync_rings(struct radeon_cs_parser *p)
+static int radeon_cs_sync_rings(struct radeon_cs_parser *p)
 {
-	int i;
+	int i, r = 0;
 
 	for (i = 0; i < p->nrelocs; i++) {
 		struct reservation_object *resv;
@@ -260,9 +260,13 @@ static void radeon_cs_sync_rings(struct radeon_cs_parser *p)
 			continue;
 
 		resv = p->relocs[i].robj->tbo.resv;
-		radeon_semaphore_sync_resv(p->ib.semaphore, resv,
+		r = radeon_semaphore_sync_resv(p->rdev, p->ib.semaphore, resv,
 					       p->relocs[i].tv.shared);
+
+		if (r)
+			break;
 	}
+	return r;
 }
 
 /* XXX: note that this is called from the legacy UMS CS ioctl as well */
@@ -472,13 +476,19 @@ static int radeon_cs_ib_chunk(struct radeon_device *rdev,
 		return r;
 	}
 
+	r = radeon_cs_sync_rings(parser);
+	if (r) {
+		if (r != -ERESTARTSYS)
+			DRM_ERROR("Failed to sync rings: %i\n", r);
+		return r;
+	}
+
 	if (parser->ring == R600_RING_TYPE_UVD_INDEX)
 		radeon_uvd_note_usage(rdev);
 	else if ((parser->ring == TN_RING_TYPE_VCE1_INDEX) ||
 		 (parser->ring == TN_RING_TYPE_VCE2_INDEX))
 		radeon_vce_note_usage(rdev);
 
-	radeon_cs_sync_rings(parser);
 	r = radeon_ib_schedule(rdev, &parser->ib, NULL, true);
 	if (r) {
 		DRM_ERROR("Failed to schedule IB !\n");
@@ -565,7 +575,13 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev,
 	if (r) {
 		goto out;
 	}
-	radeon_cs_sync_rings(parser);
+
+	r = radeon_cs_sync_rings(parser);
+	if (r) {
+		if (r != -ERESTARTSYS)
+			DRM_ERROR("Failed to sync rings: %i\n", r);
+		goto out;
+	}
 	radeon_semaphore_sync_fence(parser->ib.semaphore, vm->fence);
 
 	if ((rdev->family >= CHIP_TAHITI) &&

drivers/gpu/drm/radeon/radeon_fence.c

@@ -541,6 +541,15 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr)
 	uint64_t seq[RADEON_NUM_RINGS] = {};
 	long r;
 
+	/*
+	 * This function should not be called on !radeon fences.
+	 * If this is the case, it would mean this function can
+	 * also be called on radeon fences belonging to another card.
+	 * exclusive_lock is not held in that case.
+	 */
+	if (WARN_ON_ONCE(!to_radeon_fence(&fence->base)))
+		return fence_wait(&fence->base, intr);
+
 	seq[fence->ring] = fence->seq;
 	r = radeon_fence_wait_seq_timeout(fence->rdev, seq, intr, MAX_SCHEDULE_TIMEOUT);
 	if (r < 0) {

drivers/gpu/drm/radeon/radeon_semaphore.c

@@ -124,27 +124,42 @@ void radeon_semaphore_sync_fence(struct radeon_semaphore *semaphore,
  *
  * Sync to the fence using this semaphore object
  */
-void radeon_semaphore_sync_resv(struct radeon_semaphore *sema,
+int radeon_semaphore_sync_resv(struct radeon_device *rdev,
+			       struct radeon_semaphore *sema,
 			       struct reservation_object *resv,
 			       bool shared)
 {
 	struct reservation_object_list *flist;
 	struct fence *f;
+	struct radeon_fence *fence;
 	unsigned i;
+	int r = 0;
 
 	/* always sync to the exclusive fence */
 	f = reservation_object_get_excl(resv);
-	radeon_semaphore_sync_fence(sema, (struct radeon_fence*)f);
+	fence = f ? to_radeon_fence(f) : NULL;
+	if (fence && fence->rdev == rdev)
+		radeon_semaphore_sync_fence(sema, fence);
+	else if (f)
+		r = fence_wait(f, true);
 
 	flist = reservation_object_get_list(resv);
-	if (shared || !flist)
-		return;
+	if (shared || !flist || r)
+		return r;
 
 	for (i = 0; i < flist->shared_count; ++i) {
 		f = rcu_dereference_protected(flist->shared[i],
 					      reservation_object_held(resv));
-		radeon_semaphore_sync_fence(sema, (struct radeon_fence*)f);
+		fence = to_radeon_fence(f);
+		if (fence && fence->rdev == rdev)
+			radeon_semaphore_sync_fence(sema, fence);
+		else
+			r = fence_wait(f, true);
+
+		if (r)
+			break;
 	}
+	return r;
 }
 
 /**

drivers/gpu/drm/radeon/radeon_vm.c

@@ -698,7 +698,7 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev,
 	if (ib.length_dw != 0) {
 		radeon_asic_vm_pad_ib(rdev, &ib);
 
-		radeon_semaphore_sync_resv(ib.semaphore, pd->tbo.resv, false);
+		radeon_semaphore_sync_resv(rdev, ib.semaphore, pd->tbo.resv, false);
 		radeon_semaphore_sync_fence(ib.semaphore, vm->last_id_use);
 		WARN_ON(ib.length_dw > ndw);
 		r = radeon_ib_schedule(rdev, &ib, NULL, false);
@@ -825,7 +825,7 @@ static void radeon_vm_update_ptes(struct radeon_device *rdev,
 		unsigned nptes;
 		uint64_t pte;
 
-		radeon_semaphore_sync_resv(ib->semaphore, pt->tbo.resv, false);
+		radeon_semaphore_sync_resv(rdev, ib->semaphore, pt->tbo.resv, false);
 
 		if ((addr & ~mask) == (end & ~mask))
 			nptes = end - addr;

drivers/gpu/drm/radeon/rv770_dma.c

@@ -67,7 +67,7 @@ struct radeon_fence *rv770_copy_dma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	for (i = 0; i < num_loops; i++) {

drivers/gpu/drm/radeon/si_dma.c

@@ -252,7 +252,7 @@ struct radeon_fence *si_copy_dma(struct radeon_device *rdev,
 		return ERR_PTR(r);
 	}
 
-	radeon_semaphore_sync_resv(sem, resv, false);
+	radeon_semaphore_sync_resv(rdev, sem, resv, false);
 	radeon_semaphore_sync_rings(rdev, sem, ring->idx);
 
 	for (i = 0; i < num_loops; i++) {