diff options
Diffstat (limited to 'common/recipes-kernel/linux/files/0399-drm-amdgpu-keep-all-fences-in-an-RCU-protected-array.patch')
| -rw-r--r-- | common/recipes-kernel/linux/files/0399-drm-amdgpu-keep-all-fences-in-an-RCU-protected-array.patch | 113 |
1 files changed, 113 insertions, 0 deletions
diff --git a/common/recipes-kernel/linux/files/0399-drm-amdgpu-keep-all-fences-in-an-RCU-protected-array.patch b/common/recipes-kernel/linux/files/0399-drm-amdgpu-keep-all-fences-in-an-RCU-protected-array.patch new file mode 100644 index 00000000..2d5a9c49 --- /dev/null +++ b/common/recipes-kernel/linux/files/0399-drm-amdgpu-keep-all-fences-in-an-RCU-protected-array.patch @@ -0,0 +1,113 @@ +From 7d2083232b7c288ba4a6c440d5382c010d8a12e9 Mon Sep 17 00:00:00 2001 +From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com> +Date: Sun, 13 Mar 2016 19:19:48 +0100 +Subject: [PATCH 0399/1110] drm/amdgpu: keep all fences in an RCU protected + array v2 +MIME-Version: 1.0 +Content-Type: text/plain; charset=UTF-8 +Content-Transfer-Encoding: 8bit + +Just keep all HW fences in a RCU protected array as a +first step to replace the wait queue. + +v2: update commit message, move fixes into separate patch. + +Signed-off-by: Christian König <christian.koenig@amd.com> +Acked-by: Alex Deucher <alexander.deucher@amd.com> +Reviewed-by: Chunming Zhou <david1.zhou@amd.com> +--- + drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++ + drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c | 32 ++++++++++++++++++++++++++++--- + 2 files changed, 31 insertions(+), 3 deletions(-) + +diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h +index bd556b4..43c948d 100644 +--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h ++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h +@@ -354,6 +354,8 @@ struct amdgpu_fence_driver { + unsigned irq_type; + struct timer_list fallback_timer; + wait_queue_head_t fence_queue; ++ unsigned num_fences_mask; ++ struct fence **fences; + }; + + /* some special values for the owner field */ +diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +index 4654113..44eac91 100644 +--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c ++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +@@ -124,6 +124,8 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct fence **f) + { + struct amdgpu_device *adev = ring->adev; + struct amdgpu_fence *fence; ++ struct fence *old, **ptr; ++ unsigned idx; + + fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_KERNEL); + if (fence == NULL) +@@ -137,7 +139,21 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct fence **f) + fence->seq); + amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr, + fence->seq, AMDGPU_FENCE_FLAG_INT); ++ ++ idx = fence->seq & ring->fence_drv.num_fences_mask; ++ ptr = &ring->fence_drv.fences[idx]; ++ /* This function can't be called concurrently anyway, otherwise ++ * emitting the fence would mess up the hardware ring buffer. ++ */ ++ old = rcu_dereference_protected(*ptr, 1); ++ ++ rcu_assign_pointer(*ptr, fence_get(&fence->base)); ++ ++ BUG_ON(old && !fence_is_signaled(old)); ++ fence_put(old); ++ + *f = &fence->base; ++ + return 0; + } + +@@ -380,6 +396,11 @@ int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring, + (unsigned long)ring); + + init_waitqueue_head(&ring->fence_drv.fence_queue); ++ ring->fence_drv.num_fences_mask = num_hw_submission - 1; ++ ring->fence_drv.fences = kcalloc(num_hw_submission, sizeof(void *), ++ GFP_KERNEL); ++ if (!ring->fence_drv.fences) ++ return -ENOMEM; + + timeout = msecs_to_jiffies(amdgpu_lockup_timeout); + if (timeout == 0) { +@@ -441,10 +462,9 @@ int amdgpu_fence_driver_init(struct amdgpu_device *adev) + */ + void amdgpu_fence_driver_fini(struct amdgpu_device *adev) + { +- int i, r; ++ unsigned i, j; ++ int r; + +- if (atomic_dec_and_test(&amdgpu_fence_slab_ref)) +- kmem_cache_destroy(amdgpu_fence_slab); + for (i = 0; i < AMDGPU_MAX_RINGS; i++) { + struct amdgpu_ring *ring = adev->rings[i]; + +@@ -460,8 +480,14 @@ void amdgpu_fence_driver_fini(struct amdgpu_device *adev) + ring->fence_drv.irq_type); + amd_sched_fini(&ring->sched); + del_timer_sync(&ring->fence_drv.fallback_timer); ++ for (j = 0; j <= ring->fence_drv.num_fences_mask; ++j) ++ fence_put(ring->fence_drv.fences[i]); ++ kfree(ring->fence_drv.fences); + ring->fence_drv.initialized = false; + } ++ ++ if (atomic_dec_and_test(&amdgpu_fence_slab_ref)) ++ kmem_cache_destroy(amdgpu_fence_slab); + } + + /** +-- +2.7.4 + |