| Message ID | 20200330175748.GA32709@ziepe.ca (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [GIT,PULL] Please pull HMM changes | expand |
On Mon, Mar 30, 2020 at 02:57:48PM -0300, Jason Gunthorpe wrote: > Hi Linus, > > This series arose from a review of hmm_range_fault() by Christoph, Ralph and > myself. Several bug fixes and some general clarity. > > hmm_range_fault() is being used by these 'SVM' style drivers to > non-destructively read the page tables. It is very similar to get_user_pages() > except that the output is an array of PFNs and per-pfn flags, and it has > various modes of reading. > > This is necessary before RDMA ODP can be converted, as we don't want to have > weird corner case regressions, which is still a looking forward item. Ralph > has a nice tester for this routine, but it is waiting for feedback from the > selftests maintainers. > > Regards, > Jason > > The following changes since commit f8788d86ab28f61f7b46eb6be375f8a726783636: > > Linux 5.6-rc3 (2020-02-23 16:17:42 -0800) > > are available in the Git repository at: > > git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma.git tags/for-linus-hmm > > for you to fetch changes up to 9cee0e8c6f1eb4b5e56d3eb7f5d47b05637bab4f: > > mm/hmm: return error for non-vma snapshots (2020-03-27 20:19:25 -0300) > > ---------------------------------------------------------------- > hmm related patches for 5.7 > > This series focuses on corner case bug fixes and general clarity > improvements to hmm_range_fault(). > > - 9 bug fixes > > - Allow pgmap to track the 'owner' of a DEVICE_PRIVATE - in this case the > owner tells the driver if it can understand the DEVICE_PRIVATE page or > not. Use this to resolve a bug in nouveau where it could touch > DEVICE_PRIVATE pages from other drivers. > > - Remove a bunch of dead, redundant or unused code and flags > > - Clarity improvements to hmm_range_fault() > > ---------------------------------------------------------------- > Christoph Hellwig (9): > mm/hmm: don't provide a stub for hmm_range_fault() > mm/hmm: remove the unused HMM_FAULT_ALLOW_RETRY flag > mm/hmm: simplify hmm_vma_walk_hugetlb_entry() > mm/hmm: don't handle the non-fault case in hmm_vma_walk_hole_() > mm: merge hmm_vma_do_fault into into hmm_vma_walk_hole_ > memremap: add an owner field to struct dev_pagemap > mm: handle multiple owners of device private pages in migrate_vma > mm: simplify device private page handling in hmm_range_fault > mm/hmm: check the device private page owner in hmm_range_fault() > > Jason Gunthorpe (17): > mm/hmm: add missing unmaps of the ptep during hmm_vma_handle_pte() > mm/hmm: do not call hmm_vma_walk_hole() while holding a spinlock > mm/hmm: add missing pfns set to hmm_vma_walk_pmd() > mm/hmm: add missing call to hmm_range_need_fault() before returning EFAULT > mm/hmm: reorganize how !pte_present is handled in hmm_vma_handle_pte() > mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages > mm/hmm: add missing call to hmm_pte_need_fault in HMM_PFN_SPECIAL handling > mm/hmm: do not check pmd_protnone twice in hmm_vma_handle_pmd() > mm/hmm: remove pgmap checking for devmap pages > mm/hmm: return the fault type from hmm_pte_need_fault() > mm/hmm: remove unused code and tidy comments > mm/hmm: remove HMM_FAULT_SNAPSHOT > mm/hmm: remove the CONFIG_TRANSPARENT_HUGEPAGE #ifdef This patch causes an error on arm32 all{mod,yes}config because pmd_pfn is only defined when CONFIG_ARM_LPAE is set, which is a dependency of CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE and CONFIG_TRANSPARENT_HUGEPAGE. $ make -j$(nproc) -s ARCH=arm CC=clang CROSS_COMPILE=arm-linux-gnueabi- O=out/arm32 distclean allyesconfig mm/hmm.o mm/hmm.c:207:8: error: implicit declaration of function 'pmd_pfn' [-Werror,-Wimplicit-function-declaration] pfn = pmd_pfn(pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT); ^ 1 error generated. https://elixir.bootlin.com/linux/v5.6/source/arch/arm/include/asm/pgtable-3level.h#L236 https://elixir.bootlin.com/linux/v5.6/source/arch/arm/include/asm/pgtable.h#L29 https://elixir.bootlin.com/linux/v5.6/source/arch/arm/Kconfig#L1579 No idea how to rectify that but thought I would let you know. Cheers, Nathan
On Mon, Mar 30, 2020 at 12:54:03PM -0700, Nathan Chancellor wrote: > > Jason Gunthorpe (17): > > mm/hmm: add missing unmaps of the ptep during hmm_vma_handle_pte() > > mm/hmm: do not call hmm_vma_walk_hole() while holding a spinlock > > mm/hmm: add missing pfns set to hmm_vma_walk_pmd() > > mm/hmm: add missing call to hmm_range_need_fault() before returning EFAULT > > mm/hmm: reorganize how !pte_present is handled in hmm_vma_handle_pte() > > mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages > > mm/hmm: add missing call to hmm_pte_need_fault in HMM_PFN_SPECIAL handling > > mm/hmm: do not check pmd_protnone twice in hmm_vma_handle_pmd() > > mm/hmm: remove pgmap checking for devmap pages > > mm/hmm: return the fault type from hmm_pte_need_fault() > > mm/hmm: remove unused code and tidy comments > > mm/hmm: remove HMM_FAULT_SNAPSHOT > > mm/hmm: remove the CONFIG_TRANSPARENT_HUGEPAGE #ifdef > > This patch causes an error on arm32 all{mod,yes}config because pmd_pfn > is only defined when CONFIG_ARM_LPAE is set, which is a dependency of > CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE and CONFIG_TRANSPARENT_HUGEPAGE. Oh! I'm again surprised 0-day did not catch on to this. linux-next surely would have, but things got there later than I would have preferred due to the world being upside down right now :( > https://elixir.bootlin.com/linux/v5.6/source/arch/arm/include/asm/pgtable-3level.h#L236 > https://elixir.bootlin.com/linux/v5.6/source/arch/arm/include/asm/pgtable.h#L29 > https://elixir.bootlin.com/linux/v5.6/source/arch/arm/Kconfig#L1579 > > No idea how to rectify that but thought I would let you know. I'll just drop the patch. Next cycle we can add some comment here as this requirement is a hard to notice. The for-linus-hmm tag is updated now to reflect this. Thanks a lot, Jason
Hi Linus, This series arose from a review of hmm_range_fault() by Christoph, Ralph and myself. Several bug fixes and some general clarity. hmm_range_fault() is being used by these 'SVM' style drivers to non-destructively read the page tables. It is very similar to get_user_pages() except that the output is an array of PFNs and per-pfn flags, and it has various modes of reading. This is necessary before RDMA ODP can be converted, as we don't want to have weird corner case regressions, which is still a looking forward item. Ralph has a nice tester for this routine, but it is waiting for feedback from the selftests maintainers. Regards, Jason The following changes since commit f8788d86ab28f61f7b46eb6be375f8a726783636: Linux 5.6-rc3 (2020-02-23 16:17:42 -0800) are available in the Git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma.git tags/for-linus-hmm for you to fetch changes up to 9cee0e8c6f1eb4b5e56d3eb7f5d47b05637bab4f: mm/hmm: return error for non-vma snapshots (2020-03-27 20:19:25 -0300) ---------------------------------------------------------------- hmm related patches for 5.7 This series focuses on corner case bug fixes and general clarity improvements to hmm_range_fault(). - 9 bug fixes - Allow pgmap to track the 'owner' of a DEVICE_PRIVATE - in this case the owner tells the driver if it can understand the DEVICE_PRIVATE page or not. Use this to resolve a bug in nouveau where it could touch DEVICE_PRIVATE pages from other drivers. - Remove a bunch of dead, redundant or unused code and flags - Clarity improvements to hmm_range_fault() ---------------------------------------------------------------- Christoph Hellwig (9): mm/hmm: don't provide a stub for hmm_range_fault() mm/hmm: remove the unused HMM_FAULT_ALLOW_RETRY flag mm/hmm: simplify hmm_vma_walk_hugetlb_entry() mm/hmm: don't handle the non-fault case in hmm_vma_walk_hole_() mm: merge hmm_vma_do_fault into into hmm_vma_walk_hole_ memremap: add an owner field to struct dev_pagemap mm: handle multiple owners of device private pages in migrate_vma mm: simplify device private page handling in hmm_range_fault mm/hmm: check the device private page owner in hmm_range_fault() Jason Gunthorpe (17): mm/hmm: add missing unmaps of the ptep during hmm_vma_handle_pte() mm/hmm: do not call hmm_vma_walk_hole() while holding a spinlock mm/hmm: add missing pfns set to hmm_vma_walk_pmd() mm/hmm: add missing call to hmm_range_need_fault() before returning EFAULT mm/hmm: reorganize how !pte_present is handled in hmm_vma_handle_pte() mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages mm/hmm: add missing call to hmm_pte_need_fault in HMM_PFN_SPECIAL handling mm/hmm: do not check pmd_protnone twice in hmm_vma_handle_pmd() mm/hmm: remove pgmap checking for devmap pages mm/hmm: return the fault type from hmm_pte_need_fault() mm/hmm: remove unused code and tidy comments mm/hmm: remove HMM_FAULT_SNAPSHOT mm/hmm: remove the CONFIG_TRANSPARENT_HUGEPAGE #ifdef mm/hmm: use device_private_entry_to_pfn() mm/hmm: do not unconditionally set pfns when returning EBUSY mm/hmm: do not set pfns when returning an error code mm/hmm: return error for non-vma snapshots Documentation/vm/hmm.rst | 12 +- arch/powerpc/kvm/book3s_hv_uvmem.c | 3 + drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 3 +- drivers/gpu/drm/nouveau/nouveau_dmem.c | 19 +- drivers/gpu/drm/nouveau/nouveau_svm.c | 3 +- include/linux/hmm.h | 125 +-------- include/linux/memremap.h | 4 + include/linux/migrate.h | 8 + mm/hmm.c | 476 +++++++++++++------------------- mm/memremap.c | 4 + mm/migrate.c | 9 +- 11 files changed, 227 insertions(+), 439 deletions(-)