[OS-BUILD PATCH 0/0] mm: FOLL_UNSHARE and COR fault

[Andrea Arcangeli (via Email Bridge)]

From: Andrea Arcangeli on gitlab.com
Merge Request: https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124
NOTE: Truncated patchset due to missing @redhat.com email
      address on your GitLab profile at https://gitlab.com/-/profile.
      Once that is fixed, close and reopen the merge request to
      retrigger sending the emails.

Tested: fixed https://github.com/aagit/kernel-testcases-
for-v5.11/blob/main/vmsplice-v5.11.c
        no-regression https://github.com/aagit/kernel-testcases-
for-v5.11/blob/main/page_count_do_wp_page.c
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1958742
Upstream Status: RHEL-only

3b807e3e4d34 (Andrea Arcangeli)
   mm: FOLL_UNSHARE|FOLL_MM_SYNC: CentOS Stream kABI

2a5c44451e60 (Andrea Arcangeli)
   mm: use_mm: fix for arches checking mm_users to optimize TLB flushes

488425ad09ab (Andrea Arcangeli)
   mm: thp: optimize total_mapcount() with head_compound_mapcount

d2a5d1e916ad (Andrea Arcangeli)
   mm: thp: page_mapcount_lock: optimize the lock_page_memcg

c2256f6642e4 (Andrea Arcangeli)
   mm: thp: page_mapcount_lock: optimize the migrate path

7e6dfb703d3e (Andrea Arcangeli)
   mm: cacheline alignment for page_table_lock and mmap_lock

8468ed6b2b93 (Andrea Arcangeli)
   mm: gup: pack has_pinned in MMF_HAS_PINNED

f6c91952e805 (Andrea Arcangeli)
   mm: gup: allow FOLL_PIN to scale in SMP

61c4edba6dbc (Andrea Arcangeli)
   mm: gup: document FOLL_MM_SYNC

ea651b9dc415 (Andrea Arcangeli)
   mm: gup: FOLL_UNSHARE|FOLL_MM_SYNC: zeropage and MAP_PRIVATE
pagecache

92ebfc1323e3 (Andrea Arcangeli)
   mm: gup: FOLL_UNSHARE|FOLL_MM_SYNC: enable COR on PageKsm

23928d6acc9e (Andrea Arcangeli)
   mm: gup: introduce FOLL_MM_SYNC and FAULT_FLAG_MM_SYNC

e124d38951e9 (Andrea Arcangeli)
   mm: COW: skip the page lock in the COW copy path

ace0bed51c7d (Andrea Arcangeli)
   mm: COW: restore full accuracy in page reuse

66bc25dc1340 (Andrea Arcangeli)
   mm: gup: FOLL_UNSHARE: optimize mmu notifier

b982d1124362 (Andrea Arcangeli)
   mm: gup: FOLL_UNSHARE

a7051d10b1aa (Andrea Arcangeli)
   mm: gup: gup_page_unshare()

eacdacdef0ff (Andrea Arcangeli)
   mm: gup: COR: copy-on-read fault

549c420a5d41 (Andrea Arcangeli)
   mm: thp: introduce irqsafe methods to check if anonymous pages are
shared

114b743e88c4 (Andrea Arcangeli)
   mm: thp: introduce page_mapcount_seq irqsafe version

54a91db05c7c (Andrea Arcangeli)
   mm: thp: introduce page_trans_huge_anon_shared

ff10b15e05db (Andrea Arcangeli)
   mm: thp: stabilize the THP mapcount in page_remove_anon_compound_rmap

207e26b93bf8 (Andrea Arcangeli)
   mm: thp: replace the page lock with the seqlock for the THP mapcount

95cbac18cf57 (Andrea Arcangeli)
   mm: thp: make the THP mapcount atomic with a seqlock

 fs/proc/task_mmu.c             |  51 ++++----
 include/linux/huge_mm.h        | 128 ++++++++++++++++++++
 include/linux/ksm.h            |   7 ++
 include/linux/mm.h             |  20 +++-
 include/linux/mm_types.h       |  18 ++-
 include/linux/sched/coredump.h |   1 +
 kernel/fork.c                  |   1 -
 kernel/kthread.c               |   2 +
 mm/gup.c                       | 133 ++++++++++++++++++++-
 mm/huge_memory.c               | 160 +++++++++++++++++--------
 mm/ksm.c                       |  25 ++++
 mm/memory.c                    | 261
+++++++++++++++++++++++++++++++++++------
 mm/rmap.c                      |  24 +++-
 mm/swapfile.c                  |  14 ++-
 mm/util.c                      |  32 ++++-
 15 files changed, 746 insertions(+), 131 deletions(-)

[Don Zickus (via Email Bridge)]

From: Don Zickus on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124#note_5716...

@aarcange - Any reason why this isn't going upstream?

[Andrea Arcangeli (via Email Bridge)]

From: Andrea Arcangeli on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124#note_5717...

Hi Don,

as fast as I know, upstream was clear about
09854ba94c6aad7886996bfbee2530b3d8a7f4f4 being correct, so trying to
argue about that point again doesn't seem a productive undertaking, we
simply agree to disagree about it.

Unfortunately I don't see a way to fix https://github.com/aagit/kernel-
testcases-for-v5.11/blob/main/vmsplice-v5.11.c without introducing this
regression: https://github.com/aagit/kernel-testcases-
for-v5.11/blob/main/page_count_do_wp_page.c unless I revert
09854ba94c6aad7886996bfbee2530b3d8a7f4f4 first.

Last Friday Peter already posted 2 of these patches through his tree
upstream and they're getting merged, so I'll rebase and update. The
question is what else could go upstream, unfortunately I don't know
given the above premise.

[Don Zickus (via Email Bridge)]

From: Don Zickus on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124#note_5718...

@aarcange - The concern is fedora/ark tree git merges linus tree almost
daily.  This patchset looks ripe for potential merge conflicts.  We
would like to reduce that otherwise we are stuck reverting this all the
time until the conflicts are resolved.

Thoughts?

[Andrea Arcangeli (via Email Bridge)]

From: Andrea Arcangeli on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124#note_5718...

That's a valid concern indeed, but one I have no solution for since I
also deal with the rejects. The best I can do is to repost it later if
needed, and no problem with me if it cannot be merged right now.

[Waiman Long (via Email Bridge)]

From: Waiman Long on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124#note_5719...

Perhaps, this patchset can wait until the final merge to v5.14 after
which the code will be more stable.

[Andrea Arcangeli (via Email Bridge)]

From: Andrea Arcangeli on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1124#note_5719...

Hi Waiman,

Agreed, re-evaluating it later sounds the best. Thanks for checking!