aboutsummaryrefslogtreecommitdiffstats
path: root/common/recipes-kernel/linux/linux-yocto-4.14.71/4505-drm-amdgpu-rework-VM-state-machine-lock-handling-v2.patch
diff options
context:
space:
mode:
Diffstat (limited to 'common/recipes-kernel/linux/linux-yocto-4.14.71/4505-drm-amdgpu-rework-VM-state-machine-lock-handling-v2.patch')
-rw-r--r--common/recipes-kernel/linux/linux-yocto-4.14.71/4505-drm-amdgpu-rework-VM-state-machine-lock-handling-v2.patch299
1 files changed, 299 insertions, 0 deletions
diff --git a/common/recipes-kernel/linux/linux-yocto-4.14.71/4505-drm-amdgpu-rework-VM-state-machine-lock-handling-v2.patch b/common/recipes-kernel/linux/linux-yocto-4.14.71/4505-drm-amdgpu-rework-VM-state-machine-lock-handling-v2.patch
new file mode 100644
index 00000000..5cbd53ed
--- /dev/null
+++ b/common/recipes-kernel/linux/linux-yocto-4.14.71/4505-drm-amdgpu-rework-VM-state-machine-lock-handling-v2.patch
@@ -0,0 +1,299 @@
+From 28e3a7d57dba306b92dd09cf999e0cad064fcbf5 Mon Sep 17 00:00:00 2001
+From: christian koenig <christian.koenig@amd.com>
+Date: Thu, 19 Apr 2018 10:56:02 +0200
+Subject: [PATCH 4505/5725] drm/amdgpu: rework VM state machine lock handling
+ v2
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+Only the moved state needs a separate spin lock protection. All other
+states are protected by reserving the VM anyway.
+
+v2: fix some more incorrect cases
+
+Change-Id: I9cd5f107c46eea80091f8366234a6c3fb6f508be
+Signed-off-by: Christian König <christian.koenig@amd.com>
+Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
+Signed-off-by: Kalyan Alle <kalyan.alle@amd.com>
+---
+ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 66 +++++++++++-----------------------
+ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 4 +--
+ 2 files changed, 21 insertions(+), 49 deletions(-)
+
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+index caf5f61..048e39e 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+@@ -119,9 +119,7 @@ static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
+ * is currently evicted. add the bo to the evicted list to make sure it
+ * is validated on next vm use to avoid fault.
+ * */
+- spin_lock(&vm->status_lock);
+ list_move_tail(&base->vm_status, &vm->evicted);
+- spin_unlock(&vm->status_lock);
+ }
+
+ /**
+@@ -228,7 +226,6 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ struct ttm_bo_global *glob = adev->mman.bdev.glob;
+ int r;
+
+- spin_lock(&vm->status_lock);
+ while (!list_empty(&vm->evicted)) {
+ struct amdgpu_vm_bo_base *bo_base;
+ struct amdgpu_bo *bo;
+@@ -236,10 +233,8 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ bo_base = list_first_entry(&vm->evicted,
+ struct amdgpu_vm_bo_base,
+ vm_status);
+- spin_unlock(&vm->status_lock);
+
+ bo = bo_base->bo;
+- BUG_ON(!bo);
+ if (bo->parent) {
+ r = validate(param, bo);
+ if (r)
+@@ -259,13 +254,14 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ return r;
+ }
+
+- spin_lock(&vm->status_lock);
+- if (bo->tbo.type != ttm_bo_type_kernel)
++ if (bo->tbo.type != ttm_bo_type_kernel) {
++ spin_lock(&vm->moved_lock);
+ list_move(&bo_base->vm_status, &vm->moved);
+- else
++ spin_unlock(&vm->moved_lock);
++ } else {
+ list_move(&bo_base->vm_status, &vm->relocated);
++ }
+ }
+- spin_unlock(&vm->status_lock);
+
+ return 0;
+ }
+@@ -279,13 +275,7 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ */
+ bool amdgpu_vm_ready(struct amdgpu_vm *vm)
+ {
+- bool ready;
+-
+- spin_lock(&vm->status_lock);
+- ready = list_empty(&vm->evicted);
+- spin_unlock(&vm->status_lock);
+-
+- return ready;
++ return list_empty(&vm->evicted);
+ }
+
+ /**
+@@ -478,9 +468,7 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
+ pt->parent = amdgpu_bo_ref(parent->base.bo);
+
+ amdgpu_vm_bo_base_init(&entry->base, vm, pt);
+- spin_lock(&vm->status_lock);
+ list_move(&entry->base.vm_status, &vm->relocated);
+- spin_unlock(&vm->status_lock);
+ }
+
+ if (level < AMDGPU_VM_PTB) {
+@@ -927,10 +915,8 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_device *adev,
+ if (!entry->base.bo)
+ continue;
+
+- spin_lock(&vm->status_lock);
+ if (list_empty(&entry->base.vm_status))
+ list_add(&entry->base.vm_status, &vm->relocated);
+- spin_unlock(&vm->status_lock);
+ amdgpu_vm_invalidate_level(adev, vm, entry, level + 1);
+ }
+ }
+@@ -975,7 +961,6 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev,
+ params.func = amdgpu_vm_do_set_ptes;
+ }
+
+- spin_lock(&vm->status_lock);
+ while (!list_empty(&vm->relocated)) {
+ struct amdgpu_vm_bo_base *bo_base, *parent;
+ struct amdgpu_vm_pt *pt, *entry;
+@@ -985,13 +970,10 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev,
+ struct amdgpu_vm_bo_base,
+ vm_status);
+ list_del_init(&bo_base->vm_status);
+- spin_unlock(&vm->status_lock);
+
+ bo = bo_base->bo->parent;
+- if (!bo) {
+- spin_lock(&vm->status_lock);
++ if (!bo)
+ continue;
+- }
+
+ parent = list_first_entry(&bo->va, struct amdgpu_vm_bo_base,
+ bo_list);
+@@ -1000,12 +982,10 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev,
+
+ amdgpu_vm_update_pde(&params, vm, pt, entry);
+
+- spin_lock(&vm->status_lock);
+ if (!vm->use_cpu_for_update &&
+ (ndw - params.ib->length_dw) < 32)
+ break;
+ }
+- spin_unlock(&vm->status_lock);
+
+ if (vm->use_cpu_for_update) {
+ /* Flush HDP */
+@@ -1108,9 +1088,7 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
+ if (entry->huge) {
+ /* Add the entry to the relocated list to update it. */
+ entry->huge = false;
+- spin_lock(&p->vm->status_lock);
+ list_move(&entry->base.vm_status, &p->vm->relocated);
+- spin_unlock(&p->vm->status_lock);
+ }
+ return;
+ }
+@@ -1618,8 +1596,9 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
+ amdgpu_asic_flush_hdp(adev, NULL);
+ }
+
+- spin_lock(&vm->status_lock);
++ spin_lock(&vm->moved_lock);
+ list_del_init(&bo_va->base.vm_status);
++ spin_unlock(&vm->moved_lock);
+
+ /* If the BO is not in its preferred location add it back to
+ * the evicted list so that it gets validated again on the
+@@ -1629,7 +1608,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
+ !(bo->preferred_domains &
+ amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type)))
+ list_add_tail(&bo_va->base.vm_status, &vm->evicted);
+- spin_unlock(&vm->status_lock);
+
+ list_splice_init(&bo_va->invalids, &bo_va->valids);
+ bo_va->cleared = clear;
+@@ -1841,14 +1819,14 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
+ bool clear;
+ int r = 0;
+
+- spin_lock(&vm->status_lock);
++ spin_lock(&vm->moved_lock);
+ while (!list_empty(&vm->moved)) {
+ struct amdgpu_bo_va *bo_va;
+ struct reservation_object *resv;
+
+ bo_va = list_first_entry(&vm->moved,
+ struct amdgpu_bo_va, base.vm_status);
+- spin_unlock(&vm->status_lock);
++ spin_unlock(&vm->moved_lock);
+
+ resv = bo_va->base.bo->tbo.resv;
+
+@@ -1869,9 +1847,9 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
+ if (!clear && resv != vm->root.base.bo->tbo.resv)
+ reservation_object_unlock(resv);
+
+- spin_lock(&vm->status_lock);
++ spin_lock(&vm->moved_lock);
+ }
+- spin_unlock(&vm->status_lock);
++ spin_unlock(&vm->moved_lock);
+
+ return r;
+ }
+@@ -1933,10 +1911,10 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
+ amdgpu_vm_prt_get(adev);
+
+ if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+- spin_lock(&vm->status_lock);
++ spin_lock(&vm->moved_lock);
+ if (list_empty(&bo_va->base.vm_status))
+ list_add(&bo_va->base.vm_status, &vm->moved);
+- spin_unlock(&vm->status_lock);
++ spin_unlock(&vm->moved_lock);
+ }
+ trace_amdgpu_vm_bo_map(bo_va, mapping);
+ }
+@@ -2246,9 +2224,9 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
+
+ list_del(&bo_va->base.bo_list);
+
+- spin_lock(&vm->status_lock);
++ spin_lock(&vm->moved_lock);
+ list_del(&bo_va->base.vm_status);
+- spin_unlock(&vm->status_lock);
++ spin_unlock(&vm->moved_lock);
+
+ list_for_each_entry_safe(mapping, next, &bo_va->valids, list) {
+ list_del(&mapping->list);
+@@ -2291,30 +2269,26 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
+
+ bo_base->moved = true;
+ if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+- spin_lock(&bo_base->vm->status_lock);
+ if (bo->tbo.type == ttm_bo_type_kernel)
+ list_move(&bo_base->vm_status, &vm->evicted);
+ else
+ list_move_tail(&bo_base->vm_status,
+ &vm->evicted);
+- spin_unlock(&bo_base->vm->status_lock);
+ continue;
+ }
+
+ /* Don't add page tables to the moved state */
+ if (bo->tbo.type == ttm_bo_type_kernel) {
+- spin_lock(&bo_base->vm->status_lock);
+ if (list_empty(&bo_base->vm_status))
+ list_add(&bo_base->vm_status, &vm->relocated);
+- spin_unlock(&bo_base->vm->status_lock);
+
+ continue;
+ }
+
+- spin_lock(&bo_base->vm->status_lock);
++ spin_lock(&bo_base->vm->moved_lock);
+ if (list_empty(&bo_base->vm_status))
+ list_add(&bo_base->vm_status, &vm->moved);
+- spin_unlock(&bo_base->vm->status_lock);
++ spin_unlock(&bo_base->vm->moved_lock);
+ }
+ }
+
+@@ -2440,9 +2414,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ vm->va = RB_ROOT_CACHED;
+ for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
+ vm->reserved_vmid[i] = NULL;
+- spin_lock_init(&vm->status_lock);
+ INIT_LIST_HEAD(&vm->evicted);
+ INIT_LIST_HEAD(&vm->relocated);
++ spin_lock_init(&vm->moved_lock);
+ INIT_LIST_HEAD(&vm->moved);
+ INIT_LIST_HEAD(&vm->freed);
+
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+index 3492ab7..7a4c967 100644
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+@@ -171,9 +171,6 @@ struct amdgpu_vm {
+ /* tree of virtual addresses mapped */
+ struct rb_root_cached va;
+
+- /* protecting invalidated */
+- spinlock_t status_lock;
+-
+ /* BOs who needs a validation */
+ struct list_head evicted;
+
+@@ -182,6 +179,7 @@ struct amdgpu_vm {
+
+ /* BOs moved, but not yet updated in the PT */
+ struct list_head moved;
++ spinlock_t moved_lock;
+
+ /* BO mappings freed, but not yet updated in the PT */
+ struct list_head freed;
+--
+2.7.4
+