diff options
| author | Felix Kuehling <Felix.Kuehling@amd.com> | 2021-07-16 22:46:21 -0400 |
|---|---|---|
| committer | Alex Deucher <alexander.deucher@amd.com> | 2021-08-06 16:12:32 -0400 |
| commit | a43e2a0e11491b73e2acaa27ee74d6c3b86deac0 (patch) | |
| tree | bd3d9c29ba1f9d91437d1683b76bfb37c27d24ab /drivers/gpu | |
| parent | ed7c28c77103e1441ed22eb492b26b9b035403f7 (diff) | |
drm/amdkfd: Allow querying SVM attributes that are clear
Currently the SVM get_attr call allows querying, which flags are set
in the entire address range. Add the opposite query, which flags are
clear in the entire address range. Both queries can be combined in a
single get_attr call, which allows answering questions such as, "is
this address range coherent, non-coherent, or a mix of both"?
Proposed userspace for UAPI:
https://github.com/RadeonOpenCompute/ROCR-Runtime/tree/memory_model_queries
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Philip Yand <philip.yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Diffstat (limited to 'drivers/gpu')
| -rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 19 |
1 files changed, 13 insertions, 6 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c index c7b364e4a287..7df69b77bc5c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -3019,7 +3019,8 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size, struct svm_range *prange; uint32_t prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED; uint32_t location = KFD_IOCTL_SVM_LOCATION_UNDEFINED; - uint32_t flags = 0xffffffff; + uint32_t flags_and = 0xffffffff; + uint32_t flags_or = 0; int gpuidx; uint32_t i; @@ -3046,12 +3047,12 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size, get_accessible = true; break; case KFD_IOCTL_SVM_ATTR_SET_FLAGS: + case KFD_IOCTL_SVM_ATTR_CLR_FLAGS: get_flags = true; break; case KFD_IOCTL_SVM_ATTR_GRANULARITY: get_granularity = true; break; - case KFD_IOCTL_SVM_ATTR_CLR_FLAGS: case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE: case KFD_IOCTL_SVM_ATTR_NO_ACCESS: fallthrough; @@ -3069,7 +3070,8 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size, if (!node) { pr_debug("range attrs not found return default values\n"); svm_range_set_default_attributes(&location, &prefetch_loc, - &granularity, &flags); + &granularity, &flags_and); + flags_or = flags_and; if (p->xnack_enabled) bitmap_copy(bitmap_access, svms->bitmap_supported, MAX_GPU_INSTANCE); @@ -3115,8 +3117,10 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size, bitmap_and(bitmap_aip, bitmap_aip, prange->bitmap_aip, MAX_GPU_INSTANCE); } - if (get_flags) - flags &= prange->flags; + if (get_flags) { + flags_and &= prange->flags; + flags_or |= prange->flags; + } if (get_granularity && prange->granularity < granularity) granularity = prange->granularity; @@ -3150,7 +3154,10 @@ fill_values: attrs[i].type = KFD_IOCTL_SVM_ATTR_NO_ACCESS; break; case KFD_IOCTL_SVM_ATTR_SET_FLAGS: - attrs[i].value = flags; + attrs[i].value = flags_and; + break; + case KFD_IOCTL_SVM_ATTR_CLR_FLAGS: + attrs[i].value = ~flags_or; break; case KFD_IOCTL_SVM_ATTR_GRANULARITY: attrs[i].value = (uint32_t)granularity; |