| Message ID | 20210616105937.23201-1-apopple@nvidia.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Add support for SVM atomics in Nouveau | expand |
On Wed, 16 Jun 2021 20:59:27 +1000 Alistair Popple <apopple@nvidia.com> wrote:
> This is my series to add support for SVM atomics in Nouveau
Can we please have a nice [0/n] overview for this patchset?
Introduction ============ Some devices have features such as atomic PTE bits that can be used to implement atomic access to system memory. To support atomic operations to a shared virtual memory page such a device needs access to that page which is exclusive of the CPU. This series introduces a mechanism to temporarily unmap pages granting exclusive access to a device. These changes are required to support OpenCL atomic operations in Nouveau to shared virtual memory (SVM) regions allocated with the CL_MEM_SVM_ATOMICS clSVMAlloc flag. A more complete description of the OpenCL SVM feature is available at https://www.khronos.org/registry/OpenCL/specs/3.0-unified/html/ OpenCL_API.html#_shared_virtual_memory . Implementation ============== Exclusive device access is implemented by adding a new swap entry type (SWAP_DEVICE_EXCLUSIVE) which is similar to a migration entry. The main difference is that on fault the original entry is immediately restored by the fault handler instead of waiting. Restoring the entry triggers calls to MMU notifers which allows a device driver to revoke the atomic access permission from the GPU prior to the CPU finalising the entry. Patches ======= Patches 1 & 2 refactor existing migration and device private entry functions. Patches 3 & 4 rework try_to_unmap_one() by splitting out unrelated functionality into separate functions - try_to_migrate_one() and try_to_munlock_one(). Patch 5 renames some existing code but does not introduce functionality. Patch 6 is a small clean-up to swap entry handling in copy_pte_range(). Patch 7 contains the bulk of the implementation for device exclusive memory. Patch 8 contains some additions to the HMM selftests to ensure everything works as expected. Patch 9 is a cleanup for the Nouveau SVM implementation. Patch 10 contains the implementation of atomic access for the Nouveau driver. Testing ======= This has been tested with upstream Mesa 21.1.0 and a simple OpenCL program which checks that GPU atomic accesses to system memory are atomic. Without this series the test fails as there is no way of write-protecting the page mapping which results in the device clobbering CPU writes. For reference the test is available at https://ozlabs.org/~apopple/opencl_svm_atomics/ Further testing has been performed by adding support for testing exclusive access to the hmm-tests kselftests. On Wednesday, 16 June 2021 8:59:27 PM AEST Alistair Popple wrote: Alistair Popple (10): mm: Remove special swap entry functions mm/swapops: Rework swap entry manipulation code mm/rmap: Split try_to_munlock from try_to_unmap mm/rmap: Split migration into its own function mm: Rename migrate_pgmap_owner mm/memory.c: Allow different return codes for copy_nonpresent_pte() mm: Device exclusive memory access mm: Selftests for exclusive device memory nouveau/svm: Refactor nouveau_range_fault nouveau/svm: Implement atomic SVM access Documentation/vm/hmm.rst | 19 +- Documentation/vm/unevictable-lru.rst | 33 +- arch/s390/mm/pgtable.c | 2 +- drivers/gpu/drm/nouveau/include/nvif/if000c.h | 1 + drivers/gpu/drm/nouveau/nouveau_svm.c | 156 ++++- drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h | 1 + .../drm/nouveau/nvkm/subdev/mmu/vmmgp100.c | 6 + fs/proc/task_mmu.c | 23 +- include/linux/mmu_notifier.h | 26 +- include/linux/rmap.h | 11 +- include/linux/swap.h | 13 +- include/linux/swapops.h | 123 ++-- lib/test_hmm.c | 127 +++- lib/test_hmm_uapi.h | 2 + mm/debug_vm_pgtable.c | 12 +- mm/hmm.c | 12 +- mm/huge_memory.c | 48 +- mm/hugetlb.c | 10 +- mm/memcontrol.c | 2 +- mm/memory.c | 175 ++++- mm/migrate.c | 51 +- mm/mlock.c | 12 +- mm/mprotect.c | 18 +- mm/page_vma_mapped.c | 15 +- mm/rmap.c | 612 +++++++++++++++--- tools/testing/selftests/vm/hmm-tests.c | 158 +++++ 26 files changed, 1341 insertions(+), 327 deletions(-)
On Thursday, 17 June 2021 9:35:29 AM AEST Andrew Morton wrote: > On Wed, 16 Jun 2021 20:59:27 +1000 Alistair Popple <apopple@nvidia.com> wrote: > > > This is my series to add support for SVM atomics in Nouveau > > Can we please have a nice [0/n] overview for this patchset? > Sorry, I forgot to include that this time. Please see the update I just posted. Let me know if you need me to resend the whole series with the updated cover letter. Thanks. - Alistair
Hi Andrew, This is my series to add support for SVM atomics in Nouveau rebased onto v5.13-rc6-mmots-2021-06-15-20-28. Functionally there are no changes from the previous v10, however some changes were made during the rebase. If anyone would like to see a diff-of-diffs I can post it but my exhaustive (assuming I didn't miss anything in the diff-of-diff) summary below is probably easier to read. Hugh - based on our previous disscussion I'm reasonably confident I haven't missed anything from you in the rebase, but patch 4 ("Split migration into its own function") might be worth looking at as that contained the largest conflicts. Thanks. [PATCH 01/10] mm: Remove special swap entry functions - Fixed migration_entry_to_page() reference in __split_huge_pmd_locked() introduced by Hugh's "mm/thp: fix __split_huge_pmd_locked() on shmem migration entry". - Fixed migration_entry_to_page() reference in page_vma_mapped_walk() changed by Hugh in "mm: page_vma_mapped_walk(): prettify PVMW_MIGRATION block" [PATCH 02/10] mm/swapops: Rework swap entry manipulation code - No changes required. [PATCH 03/10] mm/rmap: Split try_to_munlock from try_to_unmap - No changes required. [PATCH 04/10] mm/rmap: Split migration into its own function - Updated try_to_migrate_one() and try_to_migrate() to accept TTU_SYNC flag. - Updated mmu notifier range calculation in try_to_migrate_one() to use vma_address_end() introduced by Hugh in "mm/thp: fix vma_address() if virtual address below file offset". - Update to unmap_page() in mm/hugh_memory.c to pass TTU_SYNC and check page_mapped() now that try_to_{unmap|migrate} return void. [PATCH 05/10] mm: Rename migrate_pgmap_owner - Added Peter's Reviewed-by. [PATCH 06/10] mm/memory.c: Allow different return codes for copy_non_present_pte() - Updated call to copy_nonpresent_pte() for extra arguments added by Peter in "mm/userfaultfd: fix uffd-wp special cases for fork()". - Added Peter's Reviewed-by. [PATCH 07/10] mm: Device exclusive memory access - s/vma/src_vma/ in copy_nonpresent_pte() due to "mm/userfaultfd: fix uffd-wp special cases for fork()". - Skip calling rmap_walk on tail pages. [PATCH 08/10] mm: Selftests for exclusive device memory - Squashed in a fix from Colin King (see https://lore.kernel.org/kernel-janitors/20210526170530.3766167-1-colin.king@canonical.com/). Not sure what the best way of attributing that is though given it was against next. [PATCH 09/10] nouveau/svm: Refactor nouveau_range_fault - No changes. [PATCH 10/10] nouveau/svm: Implement atomic SVM access - No changes. Alistair Popple (10): mm: Remove special swap entry functions mm/swapops: Rework swap entry manipulation code mm/rmap: Split try_to_munlock from try_to_unmap mm/rmap: Split migration into its own function mm: Rename migrate_pgmap_owner mm/memory.c: Allow different return codes for copy_nonpresent_pte() mm: Device exclusive memory access mm: Selftests for exclusive device memory nouveau/svm: Refactor nouveau_range_fault nouveau/svm: Implement atomic SVM access Documentation/vm/hmm.rst | 19 +- Documentation/vm/unevictable-lru.rst | 33 +- arch/s390/mm/pgtable.c | 2 +- drivers/gpu/drm/nouveau/include/nvif/if000c.h | 1 + drivers/gpu/drm/nouveau/nouveau_svm.c | 156 ++++- drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h | 1 + .../drm/nouveau/nvkm/subdev/mmu/vmmgp100.c | 6 + fs/proc/task_mmu.c | 23 +- include/linux/mmu_notifier.h | 26 +- include/linux/rmap.h | 11 +- include/linux/swap.h | 13 +- include/linux/swapops.h | 123 ++-- lib/test_hmm.c | 127 +++- lib/test_hmm_uapi.h | 2 + mm/debug_vm_pgtable.c | 12 +- mm/hmm.c | 12 +- mm/huge_memory.c | 48 +- mm/hugetlb.c | 10 +- mm/memcontrol.c | 2 +- mm/memory.c | 175 ++++- mm/migrate.c | 51 +- mm/mlock.c | 12 +- mm/mprotect.c | 18 +- mm/page_vma_mapped.c | 15 +- mm/rmap.c | 612 +++++++++++++++--- tools/testing/selftests/vm/hmm-tests.c | 158 +++++ 26 files changed, 1341 insertions(+), 327 deletions(-)