linux/mm
Jason Gunthorpe 99cb252f5e mm/mmu_notifier: add an interval tree notifier
Of the 13 users of mmu_notifiers, 8 of them use only
invalidate_range_start/end() and immediately intersect the
mmu_notifier_range with some kind of internal list of VAs.  4 use an
interval tree (i915_gem, radeon_mn, umem_odp, hfi1). 4 use a linked list
of some kind (scif_dma, vhost, gntdev, hmm)

And the remaining 5 either don't use invalidate_range_start() or do some
special thing with it.

It turns out that building a correct scheme with an interval tree is
pretty complicated, particularly if the use case is synchronizing against
another thread doing get_user_pages().  Many of these implementations have
various subtle and difficult to fix races.

This approach puts the interval tree as common code at the top of the mmu
notifier call tree and implements a shareable locking scheme.

It includes:
 - An interval tree tracking VA ranges, with per-range callbacks
 - A read/write locking scheme for the interval tree that avoids
   sleeping in the notifier path (for OOM killer)
 - A sequence counter based collision-retry locking scheme to tell
   device page fault that a VA range is being concurrently invalidated.

This is based on various ideas:
- hmm accumulates invalidated VA ranges and releases them when all
  invalidates are done, via active_invalidate_ranges count.
  This approach avoids having to intersect the interval tree twice (as
  umem_odp does) at the potential cost of a longer device page fault.

- kvm/umem_odp use a sequence counter to drive the collision retry,
  via invalidate_seq

- a deferred work todo list on unlock scheme like RTNL, via deferred_list.
  This makes adding/removing interval tree members more deterministic

- seqlock, except this version makes the seqlock idea multi-holder on the
  write side by protecting it with active_invalidate_ranges and a spinlock

To minimize MM overhead when only the interval tree is being used, the
entire SRCU and hlist overheads are dropped using some simple
branches. Similarly the interval tree overhead is dropped when in hlist
mode.

The overhead from the mandatory spinlock is broadly the same as most of
existing users which already had a lock (or two) of some sort on the
invalidation path.

Link: https://lore.kernel.org/r/20191112202231.3856-3-jgg@ziepe.ca
Acked-by: Christian König <christian.koenig@amd.com>
Tested-by: Philip Yang <Philip.Yang@amd.com>
Tested-by: Ralph Campbell <rcampbell@nvidia.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2019-11-23 19:56:44 -04:00
..
kasan mm: introduce compound_nr() 2019-09-24 15:54:08 -07:00
backing-dev.c bdi: Do not use freezable workqueue 2019-10-06 09:11:35 -06:00
balloon_compaction.c mm/balloon_compaction: suppress allocation warnings 2019-09-04 07:42:01 -04:00
cleancache.c Driver Core and debugfs changes for 5.3-rc1 2019-07-12 12:24:03 -07:00
cma_debug.c mm/cma_debug.c: fix the break condition in cma_maxchunk_get() 2019-05-14 09:47:45 -07:00
cma.c mm/cma.c: fail if fixed declaration can't be honored 2019-07-16 19:23:21 -07:00
cma.h License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
compaction.c mm, compaction: fix wrong pfn handling in __reset_isolation_pfn() 2019-10-14 15:04:01 -07:00
debug_page_ref.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
debug.c mm: update references to page _refcount 2019-05-14 19:52:47 -07:00
dmapool.c mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options 2019-07-12 11:05:46 -07:00
early_ioremap.c mm/early_ioremap: Fix boot hang with earlyprintk=efi,keep 2017-12-11 14:54:44 +01:00
fadvise.c fs: Export generic_fadvise() 2019-08-30 22:43:58 -07:00
failslab.c mm/failslab.c: by default, do not fail allocations with direct reclaim only 2019-07-12 11:05:43 -07:00
filemap.c mm/filemap.c: include <linux/ramfs.h> for generic_file_vm_ops definition 2019-10-19 06:32:32 -04:00
frame_vector.c mm: untag user pointers in get_vaddr_frames 2019-09-25 17:51:41 -07:00
frontswap.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 482 2019-06-19 17:09:52 +02:00
gup_benchmark.c mm/gup: replace get_user_pages_longterm() with FOLL_LONGTERM 2019-05-14 09:47:45 -07:00
gup.c mm/gup: fix a misnamed "write" argument, and a related bug 2019-10-19 06:32:32 -04:00
highmem.c mm: convert totalram_pages and totalhigh_pages variables to atomic 2018-12-28 12:11:47 -08:00
hmm.c mm/hmm: allow snapshot of the special zero page 2019-10-29 14:26:28 -03:00
huge_memory.c mm/thp: fix node page state in split_huge_page_to_list() 2019-10-19 06:32:32 -04:00
hugetlb_cgroup.c mm: introduce compound_nr() 2019-09-24 15:54:08 -07:00
hugetlb.c hugetlbfs: don't access uninitialized memmaps in pfn_range_valid_gigantic() 2019-10-19 06:32:32 -04:00
hwpoison-inject.c hwpoison-inject: no need to check return value of debugfs_create functions 2019-06-03 15:39:40 +02:00
init-mm.c mm/init-mm.c: include <linux/mman.h> for vm_committed_as_batch 2019-10-19 06:32:32 -04:00
internal.h mm: introduce MADV_COLD 2019-09-25 17:51:41 -07:00
interval_tree.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 248 2019-06-19 17:09:08 +02:00
Kconfig mm/mmu_notifier: add an interval tree notifier 2019-11-23 19:56:44 -04:00
Kconfig.debug mm, page_owner, debug_pagealloc: save and dump freeing stack trace 2019-09-24 15:54:08 -07:00
khugepaged.c khugepaged: enable collapse pmd for pte-mapped THP 2019-09-24 15:54:11 -07:00
kmemleak-test.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 333 2019-06-05 17:37:06 +02:00
kmemleak.c kmemleak: Do not corrupt the object_list during clean-up 2019-10-14 08:56:16 -07:00
ksm.c mm: move memcmp_pages() and pages_identical() 2019-09-24 15:54:11 -07:00
list_lru.c mm: memcg/slab: stop setting page->mem_cgroup pointer for slab pages 2019-07-12 11:05:44 -07:00
maccess.c The main changes in this release include: 2019-07-18 11:51:00 -07:00
madvise.c mm: factor out common parts between MADV_COLD and MADV_PAGEOUT 2019-09-25 17:51:41 -07:00
Makefile mm: silence -Woverride-init/initializer-overrides 2019-09-24 15:54:10 -07:00
memblock.c mm: memblock: do not enforce current limit for memblock_phys* family 2019-10-19 06:32:32 -04:00
memcontrol.c mm/memcontrol: update lruvec counters in mem_cgroup_move_account 2019-10-19 06:32:32 -04:00
memfd.c mm: page cache: store only head pages in i_pages 2019-09-24 15:54:08 -07:00
memory_hotplug.c mm/memory_hotplug: don't access uninitialized memmaps in shrink_pgdat_span() 2019-10-19 06:32:31 -04:00
memory-failure.c mm/memory-failure.c: don't access uninitialized memmaps in memory_failure() 2019-10-19 06:32:31 -04:00
memory.c mm: do not hash address in print_bad_pte() 2019-09-24 15:54:09 -07:00
mempolicy.c Merge branch 'hugepage-fallbacks' (hugepatch patches from David Rientjes) 2019-09-28 14:26:47 -07:00
mempool.c docs/core-api/mm: fix return value descriptions in mm/ 2019-03-05 21:07:20 -08:00
memremap.c mm/memunmap: don't access uninitialized memmap in memunmap_pages() 2019-10-19 06:32:32 -04:00
memtest.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
migrate.c mm: untag user pointers passed to memory syscalls 2019-09-25 17:51:41 -07:00
mincore.c mm: untag user pointers passed to memory syscalls 2019-09-25 17:51:41 -07:00
mlock.c mm: untag user pointers passed to memory syscalls 2019-09-25 17:51:41 -07:00
mm_init.c treewide: Add SPDX license identifier for missed files 2019-05-21 10:50:45 +02:00
mmap.c mm: untag user pointers in mmap/munmap/mremap/brk 2019-09-25 17:51:41 -07:00
mmu_context.c sched/headers: Prepare to move the task_lock()/unlock() APIs to <linux/sched/task.h> 2017-03-02 08:42:38 +01:00
mmu_gather.c mm: remove quicklist page table caches 2019-09-24 15:54:09 -07:00
mmu_notifier.c mm/mmu_notifier: add an interval tree notifier 2019-11-23 19:56:44 -04:00
mmzone.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
mprotect.c mm: untag user pointers passed to memory syscalls 2019-09-25 17:51:41 -07:00
mremap.c mm: untag user pointers in mmap/munmap/mremap/brk 2019-09-25 17:51:41 -07:00
msync.c mm: untag user pointers passed to memory syscalls 2019-09-25 17:51:41 -07:00
nommu.c mm: introduce page_size() 2019-09-24 15:54:08 -07:00
oom_kill.c mm: introduce MADV_COLD 2019-09-25 17:51:41 -07:00
page_alloc.c mm, hugetlb: allow hugepage allocations to reclaim as needed 2019-10-14 15:04:01 -07:00
page_counter.c memcg: introduce memory.min 2018-06-07 17:34:36 -07:00
page_ext.c mm, page_owner: fix off-by-one error in __set_page_owner_handle() 2019-10-14 15:04:00 -07:00
page_idle.c mm/page_idle.c: fix oops because end_pfn is larger than max_pfn 2019-06-29 16:43:45 +08:00
page_io.c mm, swap: use rbtree for swap_extent 2019-07-12 11:05:43 -07:00
page_isolation.c mm/page_isolation.c: change the prototype of undo_isolate_page_range() 2019-07-12 11:05:43 -07:00
page_owner.c mm/page_owner: don't access uninitialized memmaps when reading /proc/pagetypeinfo 2019-10-19 06:32:31 -04:00
page_poison.c mm/page_poison.c: fix a typo in a comment 2019-09-24 15:54:08 -07:00
page_vma_mapped.c mm: introduce page_size() 2019-09-24 15:54:08 -07:00
page-writeback.c writeback, memcg: Implement foreign dirty flushing 2019-08-27 09:22:38 -06:00
pagewalk.c pagewalk: use lockdep_assert_held for locking validation 2019-09-07 04:28:04 -03:00
percpu-internal.h percpu: convert chunk hints to be based on pcpu_block_md 2019-03-13 12:25:31 -07:00
percpu-km.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 428 2019-06-05 17:37:16 +02:00
percpu-stats.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 428 2019-06-05 17:37:16 +02:00
percpu-vm.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 428 2019-06-05 17:37:16 +02:00
percpu.c percpu: Use struct_size() helper 2019-09-04 13:40:49 -07:00
pgtable-generic.c x86/mm: Page size aware flush_tlb_mm_range() 2018-10-09 16:51:11 +02:00
process_vm_access.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 152 2019-05-30 11:26:32 -07:00
readahead.c treewide: Add SPDX license identifier for missed files 2019-05-21 10:50:45 +02:00
rmap.c mm: include <linux/huge_mm.h> for is_vma_temporary_stack 2019-10-19 06:32:32 -04:00
rodata_test.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 441 2019-06-05 17:37:17 +02:00
shmem.c Merge branch 'work.mount3' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2019-10-10 08:16:44 -07:00
shuffle.c mm: fix -Wmissing-prototypes warnings 2019-10-07 15:47:19 -07:00
shuffle.h mm: maintain randomization of page free lists 2019-05-14 19:52:48 -07:00
slab_common.c mm: memcg/slab: fix panic in __free_slab() caused by premature memcg pointer release 2019-10-19 06:32:32 -04:00
slab.c mm/slab.c: fix kernel-doc warning for __ksize() 2019-10-14 15:04:01 -07:00
slab.h mm, slab: move memcg_cache_params structure to mm/slab.h 2019-09-24 15:54:07 -07:00
slob.c mm, sl[aou]b: guarantee natural alignment for kmalloc(power-of-two) 2019-10-07 15:47:20 -07:00
slub.c mm/slub.c: init_on_free=1 should wipe freelist ptr for bulk allocations 2019-10-14 15:04:01 -07:00
sparse-vmemmap.c mm/sparsemem: convert kmalloc_section_memmap() to populate_section_memmap() 2019-07-18 17:08:07 -07:00
sparse.c mm: fix -Wmissing-prototypes warnings 2019-10-07 15:47:19 -07:00
swap_cgroup.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
swap_slots.c mm, swap, get_swap_pages: use entry_size instead of cluster in parameter 2018-08-22 10:52:44 -07:00
swap_state.c mm: page cache: store only head pages in i_pages 2019-09-24 15:54:08 -07:00
swap.c mm: introduce MADV_COLD 2019-09-25 17:51:41 -07:00
swapfile.c vfs: don't allow writes to swap files 2019-08-20 07:55:16 -07:00
truncate.c mm/thp: allow dropping THP from page cache 2019-10-19 06:32:33 -04:00
usercopy.c usercopy: Avoid HIGHMEM pfn warning 2019-09-17 15:20:17 -07:00
userfaultfd.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 499 2019-06-19 17:09:53 +02:00
util.c arm64, mm: make randomization selected by generic topdown mmap layout 2019-09-24 15:54:11 -07:00
vmacache.c mm: get rid of vmacache_flush_all() entirely 2018-09-13 15:18:04 -10:00
vmalloc.c augmented rbtree: add new RB_DECLARE_CALLBACKS_MAX macro 2019-09-25 17:51:39 -07:00
vmpressure.c mm/vmpressure.c: fix a signedness bug in vmpressure_register_event() 2019-10-07 15:47:19 -07:00
vmscan.c mm/vmscan.c: support removing arbitrary sized pages from mapping 2019-10-19 06:32:32 -04:00
vmstat.c mm,thp: stats for file backed THP 2019-09-24 15:54:11 -07:00
workingset.c mm: workingset: fix vmstat counters for shadow nodes 2019-08-13 16:06:52 -07:00
z3fold.c mm/z3fold.c: claim page in the beginning of free 2019-10-07 15:47:19 -07:00
zbud.c treewide: Add SPDX license identifier for more missed files 2019-05-21 10:50:45 +02:00
zpool.c zpool: add malloc_support_movable to zpool_driver 2019-09-24 15:54:12 -07:00
zsmalloc.c mm/zsmalloc.c: fix a -Wunused-function warning 2019-09-24 15:54:12 -07:00
zswap.c zswap: do not map same object twice 2019-09-24 15:54:12 -07:00