Hi Linus,
Continuing on the cleanup theme from last merge window, here is another set of hmm and closely related patches that are largely cleanups and bug fixes without any particularly notable functional change.
The diffstat below covers a wide section of the kernel, but a lot of this is caused by Christoph's series to revise the page walker API. After discussion we opted to put that series into this tree to avoid conflicts/etc.
Otherwise this is a pleasant reduction in LOC.
Although we continue to try to get reviews, we once again have not got too many Acks from core mm people. However the users of these APIs have again contributed a fair number of acks, reviews and tests.
I am aware of no conflicts this cycle, however Andrew has several smaller patches queued waiting on changes in here.
Included is a branch shared with rdma.git fixing up ODP enough to allow the get/put transformation.
For next merge window I already see there are patches posted to add a test suite for hmm_range_fault() and fix bugs, I am working on a series to harmonize the mmu notifier & interval tree pattern, and also to get ODP using hmm_range_fault(). I think we will need to do this tree at least one more time.
Thanks, Jason
The following changes since commit 27b7fb1ab7bfad45f5702ff0c78a4822a41b1456:
RDMA/mlx5: Fix MR npages calculation for IB_ACCESS_HUGETLB (2019-08-20 13:44:43 -0400)
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 62974fc389b364d8af70e044836362222bd3ae53:
libnvdimm: Enable unit test infrastructure compile checks (2019-09-07 04:28:05 -0300)
---------------------------------------------------------------- hmm related patches for 5.4
This is more cleanup and consolidation of the hmm APIs and the very strongly related mmu_notifier interfaces. Many places across the tree using these interfaces are touched in the process. Beyond that a cleanup to the page walker API and a few memremap related changes round out the series:
- General improvement of hmm_range_fault() and related APIs, more documentation, bug fixes from testing, API simplification & consolidation, and unused API removal
- Simplify the hmm related kconfigs to HMM_MIRROR and DEVICE_PRIVATE, and make them internal kconfig selects
- Hoist a lot of code related to mmu notifier attachment out of drivers by using a refcount get/put attachment idiom and remove the convoluted mmu_notifier_unregister_no_release() and related APIs.
- General API improvement for the migrate_vma API and revision of its only user in nouveau
- Annotate mmu_notifiers with lockdep and sleeping region debugging
Two series unrelated to HMM or mmu_notifiers came along due to dependencies:
- Allow pagemap's memremap_pages family of APIs to work without providing a struct device
- Make walk_page_range() and related use a constant structure for function pointers
---------------------------------------------------------------- Christoph Hellwig (35): nouveau: return -EBUSY when hmm_range_wait_until_valid fails mm/hmm: remove the legacy hmm_pfn_* APIs mm/hmm: replace the block argument to hmm_range_fault with a flags value mm/hmm: merge hmm_range_snapshot into hmm_range_fault amdgpu: remove -EAGAIN handling for hmm_range_fault amdgpu: don't initialize range->list in amdgpu_hmm_init_range nouveau: pass struct nouveau_svmm to nouveau_range_fault mm/hmm: remove the unused vma argument to hmm_range_dma_unmap mm/hmm: remove superfluous arguments from hmm_range_register mm/hmm: remove the page_shift member from struct hmm_range mm/hmm: remove the mask variable in hmm_vma_walk_hugetlb_entry mm/hmm: don't abuse pte_index() in hmm_vma_handle_pmd mm/hmm: only define hmm_vma_walk_pud if needed mm/hmm: cleanup the hmm_vma_handle_pmd stub mm/hmm: cleanup the hmm_vma_walk_hugetlb_entry stub mm/hmm: allow HMM_MIRROR on all architectures with MMU mm/hmm: make HMM_MIRROR an implicit option mm: turn migrate_vma upside down nouveau: reset dma_nr in nouveau_dmem_migrate_alloc_and_copy nouveau: factor out device memory address calculation nouveau: factor out dmem fence completion nouveau: remove a few function stubs nouveau: simplify nouveau_dmem_migrate_to_ram nouveau: simplify nouveau_dmem_migrate_vma mm: remove the unused MIGRATE_PFN_ERROR flag mm: remove the unused MIGRATE_PFN_DEVICE flag mm: remove CONFIG_MIGRATE_VMA_HELPER resource: add a not device managed request_free_mem_region variant memremap: remove the dev field in struct dev_pagemap memremap: don't use a separate devm action for devmap_managed_enable_get memremap: provide a not device managed memremap_pages mm/mmu_notifiers: remove the __mmu_notifier_invalidate_range_start/end exports mm: split out a new pagewalk.h header from mm.h pagewalk: separate function pointers from iterator data pagewalk: use lockdep_assert_held for locking validation
Dan Williams (1): libnvdimm: Enable unit test infrastructure compile checks
Daniel Vetter (6): mm/mmu_notifiers: check if mmu notifier callbacks are allowed to fail mm/mmu_notifiers: add a lockdep map for invalidate_range_start/end mm/mmu_notifiers: prime lockdep mm/mmu_notifiers: annotate with might_sleep() kernel.h: Add non_block_start/end() mm, notifier: Catch sleeping/blocking for !blockable
Jason Gunthorpe (27): mm/hmm: comment on VM_FAULT_RETRY semantics in handle_mm_fault mm/mmu_notifiers: hoist do_mmu_notifier_register down_write to the caller mm/mmu_notifiers: do not speculatively allocate a mmu_notifier_mm mm/mmu_notifiers: add a get/put scheme for the registration misc/sgi-gru: use mmu_notifier_get/put for struct gru_mm_struct hmm: use mmu_notifier_get/put for 'struct hmm' drm/radeon: use mmu_notifier_get/put for struct radeon_mn drm/amdkfd: fix a use after free race with mmu_notifer unregister drm/amdkfd: use mmu_notifier_put Merge 'notifier_get_put' into hmm.git RDMA/odp: Use the common interval tree library instead of generic RDMA/odp: Iterate over the whole rbtree directly RDMA/odp: Make it clearer when a umem is an implicit ODP umem RMDA/odp: Consolidate umem_odp initialization RDMA/odp: Make the three ways to create a umem_odp clear RDMA/odp: Split creating a umem_odp from ib_umem_get RDMA/odp: Provide ib_umem_odp_release() to undo the allocs RDMA/odp: Check for overflow when computing the umem_odp end RDMA/odp: Use kvcalloc for the dma_list and page_list RDMA/mlx5: Use ib_umem_start instead of umem.address RDMA/mlx5: Use odp instead of mr->umem in pagefault_mr Merge branch 'odp_fixes' into hmm.git RDMA/odp: use mmu_notifier_get/put for 'struct ib_ucontext_per_mm' RDMA/odp: remove ib_ucontext from ib_umem mm/mmu_notifiers: remove unregister_no_release csky: add missing brackets in a macro for tlb.h drm/radeon: guard against calling an unpaired radeon_mn_unregister()
Moni Shoua (1): RDMA/core: Make invalidate_range a device operation
Ralph Campbell (6): mm/hmm: replace hmm_update with mmu_notifier_range mm/hmm: a few more C style and comment clean ups mm/hmm: remove hugetlbfs check in hmm_vma_walk_pmd mm/hmm: remove hmm_range vma mm/hmm: hmm_range_fault() NULL pointer bug mm/hmm: hmm_range_fault() infinite loop
Yang, Philip (1): mm/hmm: fix hmm_range_fault()'s handling of swapped out pages
Documentation/vm/hmm.rst | 73 +---- arch/csky/include/asm/tlb.h | 8 +- arch/openrisc/kernel/dma.c | 23 +- arch/powerpc/mm/book3s64/subpage_prot.c | 12 +- arch/s390/mm/gmap.c | 35 +-- drivers/gpu/drm/amd/amdgpu/Kconfig | 4 +- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c | 15 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 31 +- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 3 - drivers/gpu/drm/amd/amdkfd/kfd_process.c | 88 +++--- drivers/gpu/drm/nouveau/Kconfig | 5 +- drivers/gpu/drm/nouveau/nouveau_dmem.c | 456 +++++++++------------------ drivers/gpu/drm/nouveau/nouveau_dmem.h | 11 - drivers/gpu/drm/nouveau/nouveau_drm.c | 3 + drivers/gpu/drm/nouveau/nouveau_svm.c | 23 +- drivers/gpu/drm/radeon/radeon.h | 3 - drivers/gpu/drm/radeon/radeon_device.c | 2 - drivers/gpu/drm/radeon/radeon_drv.c | 2 + drivers/gpu/drm/radeon/radeon_mn.c | 156 +++------ drivers/infiniband/Kconfig | 1 + drivers/infiniband/core/device.c | 1 + drivers/infiniband/core/umem.c | 54 +--- drivers/infiniband/core/umem_odp.c | 524 +++++++++++++++---------------- drivers/infiniband/core/uverbs_cmd.c | 5 - drivers/infiniband/core/uverbs_main.c | 1 + drivers/infiniband/hw/mlx5/main.c | 9 - drivers/infiniband/hw/mlx5/mem.c | 13 - drivers/infiniband/hw/mlx5/mr.c | 38 ++- drivers/infiniband/hw/mlx5/odp.c | 88 +++--- drivers/misc/sgi-gru/grufile.c | 1 + drivers/misc/sgi-gru/grutables.h | 2 - drivers/misc/sgi-gru/grutlbpurge.c | 84 ++--- drivers/nvdimm/Kconfig | 12 + drivers/nvdimm/Makefile | 4 + fs/proc/task_mmu.c | 80 ++--- include/linux/hmm.h | 125 ++------ include/linux/ioport.h | 2 + include/linux/kernel.h | 23 +- include/linux/memremap.h | 3 +- include/linux/migrate.h | 120 ++----- include/linux/mm.h | 46 --- include/linux/mm_types.h | 6 - include/linux/mmu_notifier.h | 59 +++- include/linux/pagewalk.h | 66 ++++ include/linux/sched.h | 4 + include/rdma/ib_umem.h | 2 +- include/rdma/ib_umem_odp.h | 58 ++-- include/rdma/ib_verbs.h | 7 +- kernel/fork.c | 1 - kernel/resource.c | 45 ++- kernel/sched/core.c | 19 +- mm/Kconfig | 20 +- mm/hmm.c | 490 +++++++++-------------------- mm/madvise.c | 42 +-- mm/memcontrol.c | 25 +- mm/mempolicy.c | 17 +- mm/memremap.c | 105 ++++--- mm/migrate.c | 276 ++++++++-------- mm/mincore.c | 17 +- mm/mmu_notifier.c | 263 ++++++++++++---- mm/mprotect.c | 26 +- mm/page_alloc.c | 2 +- mm/pagewalk.c | 126 ++++---- tools/testing/nvdimm/test/iomap.c | 1 - 65 files changed, 1684 insertions(+), 2184 deletions(-) create mode 100644 include/linux/pagewalk.h