aboutsummaryrefslogtreecommitdiffstats
path: root/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/4723-drm-amdgpu-explicitely-sync-to-VM-updates-v2.patch
diff options
context:
space:
mode:
Diffstat (limited to 'meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/4723-drm-amdgpu-explicitely-sync-to-VM-updates-v2.patch')
-rw-r--r--meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/4723-drm-amdgpu-explicitely-sync-to-VM-updates-v2.patch359
1 files changed, 359 insertions, 0 deletions
diff --git a/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/4723-drm-amdgpu-explicitely-sync-to-VM-updates-v2.patch b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/4723-drm-amdgpu-explicitely-sync-to-VM-updates-v2.patch
new file mode 100644
index 00000000..d97d306d
--- /dev/null
+++ b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/4723-drm-amdgpu-explicitely-sync-to-VM-updates-v2.patch
@@ -0,0 +1,359 @@
+From df1e3b51e41af0ed53e425f748a2a63cefcfa62d Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com>
+Date: Fri, 29 Nov 2019 11:33:54 +0100
+Subject: [PATCH 4723/4736] drm/amdgpu: explicitely sync to VM updates v2
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+Allows us to reduce the overhead while syncing to fences a bit.
+
+v2: also drop adev parameter from the functions
+
+Change-Id: I0828d0691fb87f0b9ae9205b15e18e6509c86d61
+Signed-off-by: Christian König <christian.koenig@amd.com>
+Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
+Signed-off-by: Rahul Kumar <rahul.kumar1@amd.com>
+---
+ .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 8 ++--
+ drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 19 +++-------
+ drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 13 +++----
+ drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 3 +-
+ drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c | 2 +-
+ drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 38 ++++++++++++++-----
+ drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h | 8 ++--
+ drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c | 2 +-
+ 8 files changed, 52 insertions(+), 41 deletions(-)
+
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+index 3d7d6b5f423e..a21201e579b1 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+@@ -392,7 +392,7 @@ static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync)
+ if (ret)
+ return ret;
+
+- return amdgpu_sync_fence(NULL, sync, vm->last_update, false);
++ return amdgpu_sync_fence(sync, vm->last_update, false);
+ }
+
+ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem)
+@@ -807,7 +807,7 @@ static int unmap_bo_from_gpuvm(struct amdgpu_device *adev,
+
+ amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
+
+- amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false);
++ amdgpu_sync_fence(sync, bo_va->last_pt_update, false);
+
+ return 0;
+ }
+@@ -826,7 +826,7 @@ static int update_gpuvm_pte(struct amdgpu_device *adev,
+ return ret;
+ }
+
+- return amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false);
++ return amdgpu_sync_fence(sync, bo_va->last_pt_update, false);
+ }
+
+ static int map_bo_to_gpuvm(struct amdgpu_device *adev,
+@@ -2337,7 +2337,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
+ pr_debug("Memory eviction: Validate BOs failed. Try again\n");
+ goto validate_map_fail;
+ }
+- ret = amdgpu_sync_fence(NULL, &sync_obj, bo->tbo.moving, false);
++ ret = amdgpu_sync_fence(&sync_obj, bo->tbo.moving, false);
+ if (ret) {
+ pr_debug("Memory eviction: Sync BO fence failed. Try again\n");
+ goto validate_map_fail;
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+index e8dfbcfad034..2c570274b5a6 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+@@ -790,29 +790,23 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
+ if (r)
+ return r;
+
+- r = amdgpu_sync_fence(adev, &p->job->sync,
+- fpriv->prt_va->last_pt_update, false);
++ r = amdgpu_sync_vm_fence(&p->job->sync, fpriv->prt_va->last_pt_update);
+ if (r)
+ return r;
+
+ if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) {
+- struct dma_fence *f;
+-
+ bo_va = fpriv->csa_va;
+ BUG_ON(!bo_va);
+ r = amdgpu_vm_bo_update(adev, bo_va, false);
+ if (r)
+ return r;
+
+- f = bo_va->last_pt_update;
+- r = amdgpu_sync_fence(adev, &p->job->sync, f, false);
++ r = amdgpu_sync_vm_fence(&p->job->sync, bo_va->last_pt_update);
+ if (r)
+ return r;
+ }
+
+ amdgpu_bo_list_for_each_entry(e, p->bo_list) {
+- struct dma_fence *f;
+-
+ /* ignore duplicates */
+ bo = ttm_to_amdgpu_bo(e->tv.bo);
+ if (!bo)
+@@ -826,8 +820,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
+ if (r)
+ return r;
+
+- f = bo_va->last_pt_update;
+- r = amdgpu_sync_fence(adev, &p->job->sync, f, false);
++ r = amdgpu_sync_vm_fence(&p->job->sync, bo_va->last_pt_update);
+ if (r)
+ return r;
+ }
+@@ -840,7 +833,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
+ if (r)
+ return r;
+
+- r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update, false);
++ r = amdgpu_sync_vm_fence(&p->job->sync, vm->last_update);
+ if (r)
+ return r;
+
+@@ -982,7 +975,7 @@ static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p,
+ dma_fence_put(old);
+ }
+
+- r = amdgpu_sync_fence(p->adev, &p->job->sync, fence, true);
++ r = amdgpu_sync_fence(&p->job->sync, fence, true);
+ dma_fence_put(fence);
+ if (r)
+ return r;
+@@ -1003,7 +996,7 @@ static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p,
+
+ return r;
+ }
+- r = amdgpu_sync_fence(p->adev, &p->job->sync, fence, true);
++ r = amdgpu_sync_fence(&p->job->sync, fence, true);
+ dma_fence_put(fence);
+
+ return r;
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
+index dfe155566571..dc2ea2b60ed8 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
+@@ -206,7 +206,7 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm,
+ int r;
+
+ if (ring->vmid_wait && !dma_fence_is_signaled(ring->vmid_wait))
+- return amdgpu_sync_fence(adev, sync, ring->vmid_wait, false);
++ return amdgpu_sync_fence(sync, ring->vmid_wait, false);
+
+ fences = kmalloc_array(sizeof(void *), id_mgr->num_ids, GFP_KERNEL);
+ if (!fences)
+@@ -241,7 +241,7 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm,
+ return -ENOMEM;
+ }
+
+- r = amdgpu_sync_fence(adev, sync, &array->base, false);
++ r = amdgpu_sync_fence(sync, &array->base, false);
+ dma_fence_put(ring->vmid_wait);
+ ring->vmid_wait = &array->base;
+ return r;
+@@ -294,7 +294,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
+ tmp = amdgpu_sync_peek_fence(&(*id)->active, ring);
+ if (tmp) {
+ *id = NULL;
+- r = amdgpu_sync_fence(adev, sync, tmp, false);
++ r = amdgpu_sync_fence(sync, tmp, false);
+ return r;
+ }
+ needs_flush = true;
+@@ -303,7 +303,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
+ /* Good we can use this VMID. Remember this submission as
+ * user of the VMID.
+ */
+- r = amdgpu_sync_fence(ring->adev, &(*id)->active, fence, false);
++ r = amdgpu_sync_fence(&(*id)->active, fence, false);
+ if (r)
+ return r;
+
+@@ -375,7 +375,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
+ /* Good, we can use this VMID. Remember this submission as
+ * user of the VMID.
+ */
+- r = amdgpu_sync_fence(ring->adev, &(*id)->active, fence, false);
++ r = amdgpu_sync_fence(&(*id)->active, fence, false);
+ if (r)
+ return r;
+
+@@ -435,8 +435,7 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
+ id = idle;
+
+ /* Remember this submission as user of the VMID */
+- r = amdgpu_sync_fence(ring->adev, &id->active,
+- fence, false);
++ r = amdgpu_sync_fence(&id->active, fence, false);
+ if (r)
+ goto error;
+
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+index 71fd9bb7ead7..8b7fce7c811f 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+@@ -193,8 +193,7 @@ static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job,
+ fence = amdgpu_sync_get_fence(&job->sync, &explicit);
+ if (fence && explicit) {
+ if (drm_sched_dependency_optimized(fence, s_entity)) {
+- r = amdgpu_sync_fence(ring->adev, &job->sched_sync,
+- fence, false);
++ r = amdgpu_sync_fence(&job->sched_sync, fence, false);
+ if (r)
+ DRM_ERROR("Error adding fence (%d)\n", r);
+ }
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c
+index 3d5beb00b0db..a467b177543b 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c
+@@ -408,7 +408,7 @@ int amdgpu_sem_add_cs(struct amdgpu_ctx *ctx, struct drm_sched_entity *entity,
+ mutex_lock(&centity->sem_lock);
+ list_for_each_entry_safe(dep, tmp, &centity->sem_dep_list,
+ list) {
+- r = amdgpu_sync_fence(ctx->adev, sync, dep->fence, true);
++ r = amdgpu_sync_fence(sync, dep->fence, true);
+ if (r)
+ goto err;
+ dma_fence_put(dep->fence);
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+index 2d6f5ec77a68..17f017fc8fcb 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+@@ -130,7 +130,8 @@ static void amdgpu_sync_keep_later(struct dma_fence **keep,
+ * Tries to add the fence to an existing hash entry. Returns true when an entry
+ * was found, false otherwise.
+ */
+-static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f, bool explicit)
++static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f,
++ bool explicit)
+ {
+ struct amdgpu_sync_entry *e;
+
+@@ -152,19 +153,18 @@ static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f,
+ * amdgpu_sync_fence - remember to sync to this fence
+ *
+ * @sync: sync object to add fence to
+- * @fence: fence to sync to
++ * @f: fence to sync to
++ * @explicit: if this is an explicit dependency
+ *
++ * Add the fence to the sync object.
+ */
+-int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
+- struct dma_fence *f, bool explicit)
++int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f,
++ bool explicit)
+ {
+ struct amdgpu_sync_entry *e;
+
+ if (!f)
+ return 0;
+- if (amdgpu_sync_same_dev(adev, f) &&
+- amdgpu_sync_get_owner(f) == AMDGPU_FENCE_OWNER_VM)
+- amdgpu_sync_keep_later(&sync->last_vm_update, f);
+
+ if (amdgpu_sync_add_later(sync, f, explicit))
+ return 0;
+@@ -180,6 +180,24 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
+ return 0;
+ }
+
++/**
++ * amdgpu_sync_vm_fence - remember to sync to this VM fence
++ *
++ * @adev: amdgpu device
++ * @sync: sync object to add fence to
++ * @fence: the VM fence to add
++ *
++ * Add the fence to the sync object and remember it as VM update.
++ */
++int amdgpu_sync_vm_fence(struct amdgpu_sync *sync, struct dma_fence *fence)
++{
++ if (!fence)
++ return 0;
++
++ amdgpu_sync_keep_later(&sync->last_vm_update, fence);
++ return amdgpu_sync_fence(sync, fence, false);
++}
++
+ /**
+ * amdgpu_sync_resv - sync to a reservation object
+ *
+@@ -205,7 +223,7 @@ int amdgpu_sync_resv(struct amdgpu_device *adev,
+
+ /* always sync to the exclusive fence */
+ f = reservation_object_get_excl(resv);
+- r = amdgpu_sync_fence(adev, sync, f, false);
++ r = amdgpu_sync_fence(sync, f, false);
+
+ flist = reservation_object_get_list(resv);
+ if (!flist || r)
+@@ -240,7 +258,7 @@ int amdgpu_sync_resv(struct amdgpu_device *adev,
+ continue;
+ }
+
+- r = amdgpu_sync_fence(adev, sync, f, false);
++ r = amdgpu_sync_fence(sync, f, false);
+ if (r)
+ break;
+ }
+@@ -341,7 +359,7 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
+ hash_for_each_safe(source->fences, i, tmp, e, node) {
+ f = e->fence;
+ if (!dma_fence_is_signaled(f)) {
+- r = amdgpu_sync_fence(NULL, clone, f, e->explicit);
++ r = amdgpu_sync_fence(clone, f, e->explicit);
+ if (r)
+ return r;
+ } else {
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
+index 10cf23a57f17..7ca71b306301 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
+@@ -40,8 +40,9 @@ struct amdgpu_sync {
+ };
+
+ void amdgpu_sync_create(struct amdgpu_sync *sync);
+-int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
+- struct dma_fence *f, bool explicit);
++int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f,
++ bool explicit);
++int amdgpu_sync_vm_fence(struct amdgpu_sync *sync, struct dma_fence *fence);
+ int amdgpu_sync_resv(struct amdgpu_device *adev,
+ struct amdgpu_sync *sync,
+ struct reservation_object *resv,
+@@ -49,7 +50,8 @@ int amdgpu_sync_resv(struct amdgpu_device *adev,
+ bool explicit_sync);
+ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
+ struct amdgpu_ring *ring);
+-struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync, bool *explicit);
++struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync,
++ bool *explicit);
+ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone);
+ int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr);
+ void amdgpu_sync_free(struct amdgpu_sync *sync);
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
+index e8db1467a71d..107def9c3611 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
+@@ -71,7 +71,7 @@ static int amdgpu_vm_sdma_prepare(struct amdgpu_vm_update_params *p,
+ p->num_dw_left = ndw;
+
+ /* Wait for moves to be completed */
+- r = amdgpu_sync_fence(p->adev, &p->job->sync, exclusive, false);
++ r = amdgpu_sync_fence(&p->job->sync, exclusive, false);
+ if (r)
+ return r;
+
+--
+2.17.1
+
generated by cgit 1.2.3-korg (git 2.39.0) at 2024-04-27 18:48:47 +0000