From: Mika Penttilä <[email protected]> Currently, the way device page faulting and migration works is not optimal, if you want to do both fault handling and migration at once.
Being able to migrate not present pages (or pages mapped with incorrect permissions, eg. COW) to the GPU requires doing either of the following sequences: 1. hmm_range_fault() - fault in non-present pages with correct permissions, etc. 2. migrate_vma_*() - migrate the pages Or: 1. migrate_vma_*() - migrate present pages 2. If non-present pages detected by migrate_vma_*(): a) call hmm_range_fault() to fault pages in b) call migrate_vma_*() again to migrate now present pages The problem with the first sequence is that you always have to do two page walks even when most of the time the pages are present or zero page mappings so the common case takes a performance hit. The second sequence is better for the common case, but far worse if pages aren't present because now you have to walk the page tables three times (once to find the page is not present, once so hmm_range_fault() can find a non-present page to fault in and once again to setup the migration). It is also tricky to code correctly. One page table walk could costs over 1000 cpu cycles on X86-64, which is a significant hit. We should be able to walk the page table once, faulting pages in as required and replacing them with migration entries if requested. Add a new flag to HMM APIs, HMM_PFN_REQ_MIGRATE, which tells to prepare for migration also during fault handling. Also, for the migrate_vma_setup() call paths, a flag, MIGRATE_VMA_FAULT, is added to tell to add fault handling to migrate. One extra benefit of migrating with hmm_range_fault() path is the migrate_vma.vma gets populated, so no need to retrieve that separataly. Tested in X86-64 VM with HMM test device, passing the selftests. For performance, the migrate throughput tests from the selftests show similar numbers (within error margin) as unmodified kernel. Tested also rebased on the "Remove device private pages from physical address space" series: https://lore.kernel.org/linux-mm/[email protected]/ plus a small patch to adjust with no problems. Changes v8-v9 - rebase on drm-tip - fixed uaf around migrate_vma_split_folio() usage - added missing pmd unlock Changes v7-v8 - rebase on 7.0 - fixed subject in two patches - enhanced commit messages - squashed patch 6 into patch 4 to fix kernel test robot warning - readded dropped Cc block from cover letter - fixed white space Changes v6-v7 - rebase on 7.0.0-rc6 - added documentation and comments - denote to be migrated zero page as HMM_PFN_MIGRATE alone - got rid of HMM_PFN_INOUT_FLAGS movement in patch 2 - picked up Acked-By from David for patch 1 Changes v5-v6 - rebase on 7.0.0-rc4 - use range based TLB flushing while unmapping ptes - gate migration behind HMM_PFN_REQ_MIGRATE for fault and migrate paths - always infer migration flags from migrate->flags only Changes v4-v5 - rebase on 6.19 - fixed David's email address - fixed link issue without CONFIG_TRANSPARENT_HUGEPAGE - refactored into smaller commits - added more comments to code Changes v3-v4: - rebase on 6.19-rc8 - fixed issues found by kernel test robot with random configs - fixed typos Changes v2-v3: - rebase on 6.19-rc7 - fixed issues found by kernel test robot - fixed smatch issues reported by Dan Carpenter <[email protected]> - fixes to lock handling (pmd/pte) on errors - added assertions for pmd/pte lock states - other issues discovered by Matthew, thanks! Changes v1-v2: - rebase on 6.19-rc6 - fixed issues found by kernel test robot - fixed locking (pmd/ptl) to cover handle_ and prepare_ regions parts if migrating - other issues discovered by Matthew, thanks! Changes RFC-v1: - rebase on 6.19-rc5 - adjust for the device THP - changes from feedback Revisions: - RFC https://lore.kernel.org/linux-mm/[email protected]/ - v1: https://lore.kernel.org/all/[email protected]/ - v2: https://lore.kernel.org/all/[email protected]/ - v3: https://lore.kernel.org/all/[email protected]/ - v4: https://lore.kernel.org/all/[email protected]/ - v5: https://lore.kernel.org/linux-mm/[email protected]/ - v6: https://lore.kernel.org/linux-mm/[email protected]/ - v7: https://lore.kernel.org/linux-mm/[email protected]/ - v8: https://lore.kernel.org/linux-mm/[email protected]/ Cc: David Hildenbrand <[email protected]> Cc: Jason Gunthorpe <[email protected]> Cc: Leon Romanovsky <[email protected]> Cc: Alistair Popple <[email protected]> Cc: Balbir Singh <[email protected]> Cc: Zi Yan <[email protected]> Cc: Matthew Brost <[email protected]> Cc: Andrew Morton <[email protected]> Cc: Lorenzo Stoakes <[email protected]> Cc: "Liam R. Howlett" <[email protected]> Cc: Vlastimil Babka <[email protected]> Cc: Mike Rapoport <[email protected]> Cc: Suren Baghdasaryan <[email protected]> Cc: Michal Hocko <[email protected]> Mika Penttilä (5): mm/Kconfig: changes for migrate on fault for device pages mm: Add helper to convert HMM pfn to migrate pfn mm/hmm: do the plumbing for HMM to participate in migration mm: setup device page migration in HMM pagewalk lib/test_hmm:: add a new testcase for the migrate on fault include/linux/hmm.h | 19 +- include/linux/migrate.h | 26 +- lib/test_hmm.c | 101 ++- lib/test_hmm_uapi.h | 19 +- mm/Kconfig | 2 + mm/hmm.c | 835 +++++++++++++++++++++++-- mm/migrate_device.c | 583 +++-------------- tools/testing/selftests/mm/hmm-tests.c | 54 ++ 8 files changed, 1066 insertions(+), 573 deletions(-) drm-tip base-commit: 94d56a898a2db27f841b17f6966a81ba502fe63c -- 2.50.0
