Re: [PATCH] drm/amdgpu: grab the id mgr lock while accessing passid_mapping
Reviewed-by: Chunming Zhou 在 2019/9/10 16:56, Christian König 写道: > Ping! > > Am 09.09.19 um 13:59 schrieb Christian König: >> Need to make sure that we actually dropping the right fence. >> Could be done with RCU as well, but to complicated for a fix. >> >> Signed-off-by: Christian König >> --- >> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 12 +--- >> 1 file changed, 9 insertions(+), 3 deletions(-) >> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c >> index b285ab25146d..e11764164cbf 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c >> @@ -1036,10 +1036,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, >> struct amdgpu_job *job, bool need_ >> id->oa_base != job->oa_base || >> id->oa_size != job->oa_size); >> bool vm_flush_needed = job->vm_needs_flush; >> - bool pasid_mapping_needed = id->pasid != job->pasid || >> - !id->pasid_mapping || >> - !dma_fence_is_signaled(id->pasid_mapping); >> struct dma_fence *fence = NULL; >> + bool pasid_mapping_needed; >> unsigned patch_offset = 0; >> int r; >> @@ -1049,6 +1047,12 @@ int amdgpu_vm_flush(struct amdgpu_ring >> *ring, struct amdgpu_job *job, bool need_ >> pasid_mapping_needed = true; >> } >> + mutex_lock(_mgr->lock); >> + if (id->pasid != job->pasid || !id->pasid_mapping || >> + !dma_fence_is_signaled(id->pasid_mapping)) >> + pasid_mapping_needed = true; >> + mutex_unlock(_mgr->lock); >> + >> gds_switch_needed &= !!ring->funcs->emit_gds_switch; >> vm_flush_needed &= !!ring->funcs->emit_vm_flush && >> job->vm_pd_addr != AMDGPU_BO_INVALID_OFFSET; >> @@ -1088,9 +1092,11 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, >> struct amdgpu_job *job, bool need_ >> } >> if (pasid_mapping_needed) { >> + mutex_lock(_mgr->lock); >> id->pasid = job->pasid; >> dma_fence_put(id->pasid_mapping); >> id->pasid_mapping = dma_fence_get(fence); >> + mutex_unlock(_mgr->lock); >> } >> dma_fence_put(fence); > > ___ > amd-gfx mailing list > amd-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/amd-gfx ___ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Re: [PATCH] drm/amdgpu: grab the id mgr lock while accessing passid_mapping
Ping! Am 09.09.19 um 13:59 schrieb Christian König: Need to make sure that we actually dropping the right fence. Could be done with RCU as well, but to complicated for a fix. Signed-off-by: Christian König --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 12 +--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index b285ab25146d..e11764164cbf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1036,10 +1036,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ id->oa_base != job->oa_base || id->oa_size != job->oa_size); bool vm_flush_needed = job->vm_needs_flush; - bool pasid_mapping_needed = id->pasid != job->pasid || - !id->pasid_mapping || - !dma_fence_is_signaled(id->pasid_mapping); struct dma_fence *fence = NULL; + bool pasid_mapping_needed; unsigned patch_offset = 0; int r; @@ -1049,6 +1047,12 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ pasid_mapping_needed = true; } + mutex_lock(_mgr->lock); + if (id->pasid != job->pasid || !id->pasid_mapping || + !dma_fence_is_signaled(id->pasid_mapping)) + pasid_mapping_needed = true; + mutex_unlock(_mgr->lock); + gds_switch_needed &= !!ring->funcs->emit_gds_switch; vm_flush_needed &= !!ring->funcs->emit_vm_flush && job->vm_pd_addr != AMDGPU_BO_INVALID_OFFSET; @@ -1088,9 +1092,11 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ } if (pasid_mapping_needed) { + mutex_lock(_mgr->lock); id->pasid = job->pasid; dma_fence_put(id->pasid_mapping); id->pasid_mapping = dma_fence_get(fence); + mutex_unlock(_mgr->lock); } dma_fence_put(fence); ___ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
[PATCH] drm/amdgpu: grab the id mgr lock while accessing passid_mapping
Need to make sure that we actually dropping the right fence. Could be done with RCU as well, but to complicated for a fix. Signed-off-by: Christian König --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 12 +--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index b285ab25146d..e11764164cbf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1036,10 +1036,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ id->oa_base != job->oa_base || id->oa_size != job->oa_size); bool vm_flush_needed = job->vm_needs_flush; - bool pasid_mapping_needed = id->pasid != job->pasid || - !id->pasid_mapping || - !dma_fence_is_signaled(id->pasid_mapping); struct dma_fence *fence = NULL; + bool pasid_mapping_needed; unsigned patch_offset = 0; int r; @@ -1049,6 +1047,12 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ pasid_mapping_needed = true; } + mutex_lock(_mgr->lock); + if (id->pasid != job->pasid || !id->pasid_mapping || + !dma_fence_is_signaled(id->pasid_mapping)) + pasid_mapping_needed = true; + mutex_unlock(_mgr->lock); + gds_switch_needed &= !!ring->funcs->emit_gds_switch; vm_flush_needed &= !!ring->funcs->emit_vm_flush && job->vm_pd_addr != AMDGPU_BO_INVALID_OFFSET; @@ -1088,9 +1092,11 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ } if (pasid_mapping_needed) { + mutex_lock(_mgr->lock); id->pasid = job->pasid; dma_fence_put(id->pasid_mapping); id->pasid_mapping = dma_fence_get(fence); + mutex_unlock(_mgr->lock); } dma_fence_put(fence); -- 2.17.1 ___ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx