Am 2021-06-29 um 2:01 p.m. schrieb Alex Sierra: > pgmap owner member at the svm migrate init could be referenced > to either adev or hive, depending on device topology.
Please update the commit description before submitting the change: GPUs in the same XGMI hive have direct access to all members' VRAM. When mapping memory to a GPU, we don't need hmm_range_fault to fault device-private pages in the same hive back to the host. Identifying the page owner as the hive, rather than the individual GPU, accomplishes this. > > Signed-off-by: Alex Sierra <[email protected]> > Reviewed-by: Felix Kuehling <[email protected]> > --- > drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 6 +++--- > drivers/gpu/drm/amd/amdkfd/kfd_svm.h | 3 +++ > 2 files changed, 6 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c > b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c > index 45b5349283af..8ce71c8142aa 100644 > --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c > @@ -427,7 +427,7 @@ svm_migrate_vma_to_vram(struct amdgpu_device *adev, > struct svm_range *prange, > migrate.start = start; > migrate.end = end; > migrate.flags = MIGRATE_VMA_SELECT_SYSTEM; > - migrate.pgmap_owner = adev; > + migrate.pgmap_owner = SVM_ADEV_PGMAP_OWNER(adev); > > size = 2 * sizeof(*migrate.src) + sizeof(uint64_t) + sizeof(dma_addr_t); > size *= npages; > @@ -649,7 +649,7 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct > svm_range *prange, > migrate.start = start; > migrate.end = end; > migrate.flags = MIGRATE_VMA_SELECT_DEVICE_PRIVATE; > - migrate.pgmap_owner = adev; > + migrate.pgmap_owner = SVM_ADEV_PGMAP_OWNER(adev); > > size = 2 * sizeof(*migrate.src) + sizeof(uint64_t) + sizeof(dma_addr_t); > size *= npages; > @@ -921,7 +921,7 @@ int svm_migrate_init(struct amdgpu_device *adev) > pgmap->range.start = res->start; > pgmap->range.end = res->end; > pgmap->ops = &svm_migrate_pgmap_ops; > - pgmap->owner = adev; > + pgmap->owner = SVM_ADEV_PGMAP_OWNER(adev); > pgmap->flags = MIGRATE_VMA_SELECT_DEVICE_PRIVATE; > r = devm_memremap_pages(adev->dev, pgmap); > if (IS_ERR(r)) { > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h > b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h > index a9af03994d1a..1f88bdfdbcc2 100644 > --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h > @@ -35,6 +35,9 @@ > #include "amdgpu.h" > #include "kfd_priv.h" > > +#define SVM_ADEV_PGMAP_OWNER(adev)\ > + ((adev)->hive ? (void *)(adev)->hive : (void *)(adev)) > + > struct svm_range_bo { > struct amdgpu_bo *bo; > struct kref kref; _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
