On Mon, Mar 02, 2020 at 05:00:23PM -0800, Ralph Campbell wrote:
When memory is migrated to the GPU, it is likely to be accessed by GPU code soon afterwards. Instead of waiting for a GPU fault, map the migrated memory into the GPU page tables with the same access permissions as the source CPU page table entries. This preserves copy on write semantics.
Signed-off-by: Ralph Campbell rcampbell@nvidia.com Cc: Christoph Hellwig hch@lst.de Cc: Jason Gunthorpe jgg@mellanox.com Cc: "Jérôme Glisse" jglisse@redhat.com Cc: Ben Skeggs bskeggs@redhat.com
Originally this patch was targeted for Jason's rdma tree since other HMM related changes were queued there. Now that those have been merged, this patch just contains changes to nouveau so it could go through any tree. I guess Ben Skeggs' tree would be appropriate.
Yep
+static inline struct nouveau_pfnmap_args * +nouveau_pfns_to_args(void *pfns)
don't use static inline inside C files
+{
- struct nvif_vmm_pfnmap_v0 *p =
container_of(pfns, struct nvif_vmm_pfnmap_v0, phys);- return container_of(p, struct nouveau_pfnmap_args, p);
And this should just be
return container_of(pfns, struct nouveau_pfnmap_args, p.phys);
+static struct nouveau_svmm * +nouveau_find_svmm(struct nouveau_svm *svm, struct mm_struct *mm) +{
- struct nouveau_ivmm *ivmm;
- list_for_each_entry(ivmm, &svm->inst, head) {
if (ivmm->svmm->notifier.mm == mm)return ivmm->svmm;- }
- return NULL;
+}
Is this re-implementing mmu_notifier_get() ?
Jason