On 2017-09-19 09:24 AM, Christian König wrote:
> Am 19.09.2017 um 14:59 schrieb Andres Rodriguez:
>> Introduce a flag to signal that access to a BO will be synchronized
>> through an external mechanism.
>>
>> Currently all buffers shared between contexts are subject to implicit
>> synchronization. However, this is only required for protocols that
>> currently don't support an explicit synchronization mechanism (DRI2/3).
>>
>> This patch introduces the AMDGPU_GEM_CREATE_EXPLICIT_SYNC, so that
>> users can specify when it is safe to disable implicit sync.
>>
>> v2: only disable explicit sync in amdgpu_cs_ioctl
>>
>> Signed-off-by: Andres Rodriguez <andresx7 at gmail.com>
>> ---
>>
>> Hey Christian,
>>
>> I kept the amdgpu_bo_explicit_sync() function since it makes it easier
>> to maintain an 80 line wrap in amdgpu_cs_sync_rings()
>
> Looks good to me, but I would like to see the matching user space code
> as well.
>
> Especially I have no idea how you want to have DRI3 compatibility with
> that?
>
No problem. I'm fixing the radv patch atm and I'll re-send it for your
reference.
Regards,
Andres
> Regards,
> Christian.
>
>>
>> Regards,
>> Andres
>>
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c    | 3 ++-
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c   | 4 +++-
>> Â drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 8 ++++++++
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c  | 7 +++++--
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h  | 3 ++-
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c   | 5 +++--
>>  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c    | 8 ++++----
>>  include/uapi/drm/amdgpu_drm.h             | 2 ++
>> Â 8 files changed, 29 insertions(+), 11 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> index db97e78..bc8a403 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> @@ -704,7 +704,8 @@ static int amdgpu_cs_sync_rings(struct
>> amdgpu_cs_parser *p)
>> Â Â Â Â Â list_for_each_entry(e, &p->validated, tv.head) {
>> Â Â Â Â Â Â Â Â Â struct reservation_object *resv = e->robj->tbo.resv;
>> -Â Â Â Â Â Â Â r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp);
>> +Â Â Â Â Â Â Â r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â amdgpu_bo_explicit_sync(e->robj));
>> Â Â Â Â Â Â Â Â Â if (r)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â return r;
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> index b0d45c8..21e9936 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> @@ -212,7 +212,9 @@ int amdgpu_gem_create_ioctl(struct drm_device
>> *dev, void *data,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_GEM_CREATE_CPU_GTT_USWC |
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_GEM_CREATE_VRAM_CLEARED |
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_GEM_CREATE_VM_ALWAYS_VALID))
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_GEM_CREATE_VM_ALWAYS_VALID |
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_GEM_CREATE_EXPLICIT_SYNC))
>> +
>> Â Â Â Â Â Â Â Â Â return -EINVAL;
>> Â Â Â Â Â /* reject invalid gem domains */
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>> index c26ef53..428aae0 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>> @@ -193,6 +193,14 @@ static inline bool
>> amdgpu_bo_gpu_accessible(struct amdgpu_bo *bo)
>> Â Â Â Â Â }
>> Â }
>> +/**
>> + * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced
>> + */
>> +static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo)
>> +{
>> +Â Â Â return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC;
>> +}
>> +
>> Â int amdgpu_bo_create(struct amdgpu_device *adev,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â unsigned long size, int byte_align,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â bool kernel, u32 domain, u64 flags,
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
>> index c586f44..a4bf21f 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
>> @@ -169,14 +169,14 @@ int amdgpu_sync_fence(struct amdgpu_device
>> *adev, struct amdgpu_sync *sync,
>> Â Â *
>> Â Â * @sync: sync object to add fences from reservation object to
>> Â Â * @resv: reservation object with embedded fence
>> - * @shared: true if we should only sync to the exclusive fence
>> + * @explicit_sync: true if we should only sync to the exclusive fence
>> Â Â *
>> Â Â * Sync to the fence
>> Â Â */
>> Â int amdgpu_sync_resv(struct amdgpu_device *adev,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct amdgpu_sync *sync,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct reservation_object *resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â void *owner)
>> +Â Â Â Â Â Â Â Â Â Â Â Â void *owner, bool explicit_sync)
>> Â {
>> Â Â Â Â Â struct reservation_object_list *flist;
>> Â Â Â Â Â struct dma_fence *f;
>> @@ -191,6 +191,9 @@ int amdgpu_sync_resv(struct amdgpu_device *adev,
>> Â Â Â Â Â f = reservation_object_get_excl(resv);
>> Â Â Â Â Â r = amdgpu_sync_fence(adev, sync, f);
>> +Â Â Â if (explicit_sync)
>> +Â Â Â Â Â Â Â return r;
>> +
>> Â Â Â Â Â flist = reservation_object_get_list(resv);
>> Â Â Â Â Â if (!flist || r)
>> Â Â Â Â Â Â Â Â Â return r;
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
>> index dc76879..70d7e3a 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
>> @@ -45,7 +45,8 @@ int amdgpu_sync_fence(struct amdgpu_device *adev,
>> struct amdgpu_sync *sync,
>> Â int amdgpu_sync_resv(struct amdgpu_device *adev,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct amdgpu_sync *sync,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct reservation_object *resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â void *owner);
>> +Â Â Â Â Â Â Â Â Â Â Â Â void *owner,
>> +Â Â Â Â Â Â Â Â Â Â Â Â bool explicit_sync);
>> Â struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct amdgpu_ring *ring);
>> Â struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync);
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> index c9c059d..bd0b03a 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> @@ -1510,7 +1510,8 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring,
>> uint64_t src_offset,
>> Â Â Â Â Â job->vm_needs_flush = vm_needs_flush;
>> Â Â Â Â Â if (resv) {
>> Â Â Â Â Â Â Â Â Â r = amdgpu_sync_resv(adev, &job->sync, resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_UNDEFINED);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_UNDEFINED,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â false);
>> Â Â Â Â Â Â Â Â Â if (r) {
>> Â Â Â Â Â Â Â Â Â Â Â Â Â DRM_ERROR("sync failed (%d).\n", r);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â goto error_free;
>> @@ -1602,7 +1603,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
>> Â Â Â Â Â if (resv) {
>> Â Â Â Â Â Â Â Â Â r = amdgpu_sync_resv(adev, &job->sync, resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_UNDEFINED);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_UNDEFINED, false);
>> Â Â Â Â Â Â Â Â Â if (r) {
>> Â Â Â Â Â Â Â Â Â Â Â Â Â DRM_ERROR("sync failed (%d).\n", r);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â goto error_free;
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> index 2df254c..9965d68 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> @@ -987,7 +987,7 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device
>> *adev, struct amdgpu_vm *vm,
>> Â Â Â Â Â int r;
>> Â Â Â Â Â amdgpu_sync_create(&sync);
>> -Â Â Â amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner);
>> +Â Â Â amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner,
>> false);
>> Â Â Â Â Â r = amdgpu_sync_wait(&sync, true);
>> Â Â Â Â Â amdgpu_sync_free(&sync);
>> @@ -1128,11 +1128,11 @@ static int amdgpu_vm_update_level(struct
>> amdgpu_device *adev,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â amdgpu_ring_pad_ib(ring, params.ib);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â amdgpu_sync_resv(adev, &job->sync,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â parent->base.bo->tbo.resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_VM);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_VM, false);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â if (shadow)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â amdgpu_sync_resv(adev, &job->sync,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â shadow->tbo.resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_VM);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AMDGPU_FENCE_OWNER_VM, false);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â WARN_ON(params.ib->length_dw > ndw);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â r = amdgpu_job_submit(job, ring, &vm->entity,
>> @@ -1595,7 +1595,7 @@ static int amdgpu_vm_bo_update_mapping(struct
>> amdgpu_device *adev,
>> Â Â Â Â Â Â Â Â Â goto error_free;
>> Â Â Â Â Â r = amdgpu_sync_resv(adev, &job->sync, vm->root.base.bo->tbo.resv,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â owner);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â owner, false);
>> Â Â Â Â Â if (r)
>> Â Â Â Â Â Â Â Â Â goto error_free;
>> diff --git a/include/uapi/drm/amdgpu_drm.h
>> b/include/uapi/drm/amdgpu_drm.h
>> index 69d64e5..96fcde8 100644
>> --- a/include/uapi/drm/amdgpu_drm.h
>> +++ b/include/uapi/drm/amdgpu_drm.h
>> @@ -91,6 +91,8 @@ extern "C" {
>> Â #define AMDGPU_GEM_CREATE_VRAM_CONTIGUOUSÂ Â Â (1 << 5)
>> Â /* Flag that BO is always valid in this VM */
>> Â #define AMDGPU_GEM_CREATE_VM_ALWAYS_VALIDÂ Â Â (1 << 6)
>> +/* Flag that BO sharing will be explicitly synchronized */
>> +#define AMDGPU_GEM_CREATE_EXPLICIT_SYNCÂ Â Â Â Â Â Â (1 << 7)
>>  struct drm_amdgpu_gem_create_in {
>> Â Â Â Â Â /** the requested memory size */
>
>
