drm/amdgpu: consistenly use VM moved flag
authorChristian König <christian.koenig@amd.com>
Thu, 19 Apr 2018 12:22:56 +0000 (14:22 +0200)
committerAlex Deucher <alexander.deucher@amd.com>
Thu, 24 May 2018 15:07:54 +0000 (10:07 -0500)
Instead of sometimes checking if the vm_status is empty use the moved
flag and also reset it when the BO leaves the state machine.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

index a31afac..f5dee4c 100644 (file)
@@ -902,8 +902,8 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_device *adev,
                if (!entry->base.bo)
                        continue;
 
-               if (list_empty(&entry->base.vm_status))
-                       list_add(&entry->base.vm_status, &vm->relocated);
+               if (!entry->base.moved)
+                       list_move(&entry->base.vm_status, &vm->relocated);
                amdgpu_vm_invalidate_level(adev, vm, entry, level + 1);
        }
 }
@@ -964,6 +964,7 @@ restart:
                bo_base = list_first_entry(&vm->relocated,
                                           struct amdgpu_vm_bo_base,
                                           vm_status);
+               bo_base->moved = false;
                list_del_init(&bo_base->vm_status);
 
                bo = bo_base->bo->parent;
@@ -1877,10 +1878,10 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
        if (mapping->flags & AMDGPU_PTE_PRT)
                amdgpu_vm_prt_get(adev);
 
-       if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+       if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv &&
+           !bo_va->base.moved) {
                spin_lock(&vm->moved_lock);
-               if (list_empty(&bo_va->base.vm_status))
-                       list_add(&bo_va->base.vm_status, &vm->moved);
+               list_move(&bo_va->base.vm_status, &vm->moved);
                spin_unlock(&vm->moved_lock);
        }
        trace_amdgpu_vm_bo_map(bo_va, mapping);
@@ -2233,6 +2234,7 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
 
        list_for_each_entry(bo_base, &bo->va, bo_list) {
                struct amdgpu_vm *vm = bo_base->vm;
+               bool was_moved = bo_base->moved;
 
                bo_base->moved = true;
                if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
@@ -2244,16 +2246,16 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
                        continue;
                }
 
-               if (bo->tbo.type == ttm_bo_type_kernel) {
-                       if (list_empty(&bo_base->vm_status))
-                               list_add(&bo_base->vm_status, &vm->relocated);
+               if (was_moved)
                        continue;
-               }
 
-               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->moved_lock);
+               if (bo->tbo.type == ttm_bo_type_kernel) {
+                       list_move(&bo_base->vm_status, &vm->relocated);
+               } else {
+                       spin_lock(&bo_base->vm->moved_lock);
+                       list_move(&bo_base->vm_status, &vm->moved);
+                       spin_unlock(&bo_base->vm->moved_lock);
+               }
        }
 }