[PATCH v3 0/5] move per-vma lock into vm_area_struct

[Suren Baghdasaryan <surenb@google.com>]

Back when per-vma locks were introduces, vm_lock was moved out of
vm_area_struct in [1] because of the performance regression caused by
false cacheline sharing. Recent investigation [2] revealed that the
regressions is limited to a rather old Broadwell microarchitecture and
even there it can be mitigated by disabling adjacent cacheline
prefetching, see [3].
Splitting single logical structure into multiple ones leads to more
complicated management, extra pointer dereferences and overall less
maintainable code. When that split-away part is a lock, it complicates
things even further. With no performance benefits, there are no reasons
for this split. Merging the vm_lock back into vm_area_struct also allows
vm_area_struct to use SLAB_TYPESAFE_BY_RCU later in this patchset.
This patchset:
1. moves vm_lock back into vm_area_struct, aligning it at the cacheline
boundary and changing the cache to be cacheline-aligned to minimize
cacheline sharing;
2. changes vm_area_struct initialization to mark new vma as detached until
it is inserted into vma tree;
3. changes vm_area_struct cache to SLAB_TYPESAFE_BY_RCU to allow for their
reuse and to minimize call_rcu() calls.
Pagefault microbenchmarks do not show noticeable performance change.

Changes since v2 [4]
- Updated vma_start_read_locked and vma_start_read_locked_nested comments,
per Lorenzo Stoakes;
- Undated patch 2 description to include motivation for moving vm_lock
into vm_area_struct, per Lorenzo Stoakes, Vlastimil Babka, Mateusz Guzik;
- Undated description to include a note about memory overhead dependency
on kernel config options, per Lorenzo Stoakes and Vlastimil Babka;
- Updated vma tests;
- Introduced vma_mark_attached();
- Moved vma_mark_attached() calls into vma_iter_store{_gfp},
per Lorenzo Stoakes
- Moved vma_lock_init() into vm_area_cachep constructor, per Hugh Dickins;
- Ensure vm_lock is preserved when clearing or copying vm_area_struct;
- Pass mm to vma_start_read() to avoid use of vma->vm_mm, which might be
changed by a racing vma_init() if the vma gets reused;
- Added vma write-locking in vma_mark_attached() to prevent races with
vma->detached check in lock_vma_under_rcu().

Patch applies over mm-unstable

[1] https://lore.kernel.org/all/20230227173632.3292573-34-surenb@google.com/
[2] https://lore.kernel.org/all/ZsQyI%2F087V34JoIt@xsang-OptiPlex-9020/
[3] https://lore.kernel.org/all/CAJuCfpEisU8Lfe96AYJDZ+OM4NoPmnw9bP53cT_kbfP_pR+-2g@mail.gmail.com/
[4] https://lore.kernel.org/all/20241112194635.444146-1-surenb@google.com/

Suren Baghdasaryan (5):
  mm: introduce vma_start_read_locked{_nested} helpers
  mm: move per-vma lock into vm_area_struct
  mm: mark vma as detached until it's added into vma tree
  mm: make vma cache SLAB_TYPESAFE_BY_RCU
  docs/mm: document latest changes to vm_lock

 Documentation/mm/process_addrs.rst |  10 ++-
 include/linux/mm.h                 | 121 +++++++++++++++++++++++------
 include/linux/mm_types.h           |  19 ++---
 kernel/fork.c                      |  88 ++++++++-------------
 mm/memory.c                        |   9 ++-
 mm/userfaultfd.c                   |  22 ++----
 mm/vma.c                           |   8 +-
 mm/vma.h                           |   2 +
 tools/testing/vma/vma_internal.h   |  55 ++++++-------
 9 files changed, 183 insertions(+), 151 deletions(-)


base-commit: 84f7f75fda11c147979984755011c6e2b71e97e1
-- 
2.47.0.338.g60cca15819-goog