I agree using fence_context is same efficiency with original client id,
since vm only has one sdma entity.
But PASID seems be more proper with per-VM-per-process concept, the
comment also said that.
Regards,
David Zhou
On 2017å¹´12æ??20æ?¥ 21:21, Christian König wrote:
> Use the fence context from the scheduler entity.
>
> Signed-off-by: Christian König <christian.koenig at amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 8 ++++----
> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 --
> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 4 ----
> 3 files changed, 4 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
> index d24884b419cb..16884a0b677b 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
> @@ -115,7 +115,7 @@ static int amdgpu_vmid_grab_reserved_locked(struct amdgpu_vm *vm,
>
> flushed = id->flushed_updates;
> if ((amdgpu_vmid_had_gpu_reset(adev, id)) ||
> - (atomic64_read(&id->owner) != vm->client_id) ||
> + (atomic64_read(&id->owner) != vm->entity.fence_context) ||
> (job->vm_pd_addr != id->pd_gpu_addr) ||
> (updates && (!flushed || updates->context != flushed->context ||
> dma_fence_is_later(updates, flushed))) ||
> @@ -144,7 +144,7 @@ static int amdgpu_vmid_grab_reserved_locked(struct amdgpu_vm *vm,
> id->flushed_updates = dma_fence_get(updates);
> }
> id->pd_gpu_addr = job->vm_pd_addr;
> - atomic64_set(&id->owner, vm->client_id);
> + atomic64_set(&id->owner, vm->entity.fence_context);
> job->vm_needs_flush = needs_flush;
> if (needs_flush) {
> dma_fence_put(id->last_flush);
> @@ -242,7 +242,7 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
> if (amdgpu_vmid_had_gpu_reset(adev, id))
> continue;
>
> - if (atomic64_read(&id->owner) != vm->client_id)
> + if (atomic64_read(&id->owner) != vm->entity.fence_context)
> continue;
>
> if (job->vm_pd_addr != id->pd_gpu_addr)
> @@ -291,7 +291,7 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
> id->pd_gpu_addr = job->vm_pd_addr;
> dma_fence_put(id->flushed_updates);
> id->flushed_updates = dma_fence_get(updates);
> - atomic64_set(&id->owner, vm->client_id);
> + atomic64_set(&id->owner, vm->entity.fence_context);
>
> needs_flush:
> job->vm_needs_flush = true;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index 946bc21c6d7d..af7dceb7131e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -2256,7 +2256,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
> uint64_t init_pde_value = 0;
>
> vm->va = RB_ROOT_CACHED;
> - vm->client_id = atomic64_inc_return(&adev->vm_manager.client_counter);
> for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
> vm->reserved_vmid[i] = NULL;
> spin_lock_init(&vm->status_lock);
> @@ -2502,7 +2501,6 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
> adev->vm_manager.seqno[i] = 0;
>
> atomic_set(&adev->vm_manager.vm_pte_next_ring, 0);
> - atomic64_set(&adev->vm_manager.client_counter, 0);
> spin_lock_init(&adev->vm_manager.prt_lock);
> atomic_set(&adev->vm_manager.num_prt_users, 0);
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> index 78296d1a5b2f..21a80f1bb2b9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> @@ -191,8 +191,6 @@ struct amdgpu_vm {
> /* Scheduler entity for page table updates */
> struct drm_sched_entity entity;
>
> - /* client id and PASID (TODO: replace client_id with PASID) */
> - u64 client_id;
> unsigned int pasid;
> /* dedicated to vm */
> struct amdgpu_vmid *reserved_vmid[AMDGPU_MAX_VMHUBS];
> @@ -230,8 +228,6 @@ struct amdgpu_vm_manager {
> struct amdgpu_ring *vm_pte_rings[AMDGPU_MAX_RINGS];
> unsigned vm_pte_num_rings;
> atomic_t vm_pte_next_ring;
> - /* client id counter */
> - atomic64_t client_counter;
>
> /* partial resident texture handling */
> spinlock_t prt_lock;
