The problem is that step 3 is still ongoing when step 4 happens, this is because same_ring is true and so we don't bother to wait for the last_flush. Alternatively we could drop same_ring and just check if last_flush is either signaled or emitted on the same ring as the job we currently search an id for. Thinking about it, this might be the better approach. Christian. Am 08.07.2016 um 15:11 schrieb Zhou, David(ChunMing): > What will the situation result in? > Before step3, the step2 must have completed. step4 is expected from step3. > > Regards, > David Zhou > > Sent from my Huawei Mobile > > -------- å??å§?é?®ä»¶ -------- > 主é¢?ï¼?[PATCH 1/3] drm/amdgpu: use one client id per ring again > å??件人ï¼?Christian Ké°?ig > æ?¶ä»¶äººï¼?amd-gfx at lists.freedesktop.org > æ??é??ï¼? > > From: Christian König <christian.koenig at amd.com> > > Otherwise we can run into the following situation: > > 1. Process A grabs ID 1 for ring 0. > 2. Process B grabs ID 1 for ring 0. > 3. Process A grabs ID 1 for ring 1. > 4. Process A tries to reuse ID1 for ring 0 but things he doesn't need > to flush. > > Signed-off-by: Christian König <christian.koenig at amd.com> > Cc: stable at vger.kernel.org > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 7 ++++--- > 1 file changed, 4 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c > b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c > index f8615a4..7ac79d2 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c > @@ -256,7 +256,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct > amdgpu_ring *ring, > if (amdgpu_vm_is_gpu_reset(adev, id)) > continue; > > - if (atomic64_read(&id->owner) != vm->client_id) > + if (atomic64_read(&id->owner) != (vm->client_id + i)) > continue; > > if (job->vm_pd_addr != id->pd_gpu_addr) > @@ -310,7 +310,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct > amdgpu_ring *ring, > id->pd_gpu_addr = job->vm_pd_addr; > id->current_gpu_reset_count = > atomic_read(&adev->gpu_reset_counter); > list_move_tail(&id->list, &adev->vm_manager.ids_lru); > - atomic64_set(&id->owner, vm->client_id); > + atomic64_set(&id->owner, vm->client_id + ring->idx); > vm->ids[ring->idx] = id; > > job->vm_id = id - adev->vm_manager.ids; > @@ -1496,7 +1496,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, > struct amdgpu_vm *vm) > for (i = 0; i < AMDGPU_MAX_RINGS; ++i) > vm->ids[i] = NULL; > vm->va = RB_ROOT; > - vm->client_id = > atomic64_inc_return(&adev->vm_manager.client_counter); > + vm->client_id = atomic64_add_return(AMDGPU_MAX_RINGS, > + &adev->vm_manager.client_counter); > spin_lock_init(&vm->status_lock); > INIT_LIST_HEAD(&vm->invalidated); > INIT_LIST_HEAD(&vm->cleared); > -- > 2.5.0 > > _______________________________________________ > amd-gfx mailing list > amd-gfx at lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/amd-gfx > > > _______________________________________________ > amd-gfx mailing list > amd-gfx at lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/amd-gfx -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20160708/f5fcdc5c/attachment.html>
