maple_tree: remove GFP_ZERO from kmem_cache_alloc() and kmem_cache_alloc_bulk()

Preallocations are common in the VMA code to avoid allocating under
certain locking conditions.  The preallocations must also cover the
worst-case scenario.  Removing the GFP_ZERO flag from the
kmem_cache_alloc() (and bulk variant) calls will reduce the amount of time
spent zeroing memory that may not be used.  Only zero out the necessary
area to keep track of the allocations in the maple state.  Zero the entire
node prior to using it in the tree.

This required internal changes to node counting on allocation, so the test
code is also updated.

This restores some micro-benchmark performance: up to +9% in mmtests mmap1
by my testing +10% to +20% in mmap, mmapaddr, mmapmany tests reported by
Red Hat

Link: https://bugzilla.redhat.com/show_bug.cgi?id=2149636
Link: https://lkml.kernel.org/r/20230105160427.2988454-1-Liam.Howlett@oracle.com
Signed-off-by: Liam Howlett <Liam.Howlett@oracle.com>
Reported-by: Jirka Hladky <jhladky@redhat.com>
Suggested-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: invert logic for early page initialisation checks

Rename early_page_uninitialised() to early_page_initialised() and invert
its logic to make the code more readable.

Link: https://lkml.kernel.org/r/20230104191805.2535864-1-rppt@kernel.org
Signed-off-by: Mike Rapoport (IBM) <rppt@kernel.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

workingset: fix confusion around eviction vs refault container

Refault decisions are made based on the lruvec where the page was evicted,
as that determined its LRU order while it was alive.  Stats and workingset
aging must then occur on the lruvec of the new page, as that's the node
and cgroup that experience the refault and that's the lruvec whose
nonresident info ages out by a new resident page.  Those lruvecs could be
different when a page is shared between cgroups, or the refaulting page is
allocated on a different node.

There are currently two mix-ups:

1. When swap is available, the resident anon set must be considered
   when comparing the refault distance. The comparison is made against
   the right anon set, but the check for swap is not. When pages get
   evicted from a cgroup with swap, and refault in one without, this
   can incorrectly consider a hot refault as cold - and vice
   versa. Fix that by using the eviction cgroup for the swap check.

2. The stats and workingset age are updated against the wrong lruvec
   altogether: the right cgroup but the wrong NUMA node. When a page
   refaults on a different NUMA node, this will have confusing stats
   and distort the workingset age on a different lruvec - again
   possibly resulting in hot/cold misclassifications down the line.

Fix the swap check and the refault pgdat to address both concerns.

This was found during code review.  It hasn't caused notable issues in
production, suggesting that those refault-migrations are relatively rare
in practice.

Link: https://lkml.kernel.org/r/20230104222944.2380117-1-nphamcs@gmail.com
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Co-developed-by: Nhat Pham <nphamcs@gmail.com>
Signed-off-by: Nhat Pham <nphamcs@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/uffd: detect pgtable allocation failures

Before this patch, when there's any pgtable allocation issues happened
during change_protection(), the error will be ignored from the syscall.
For shmem, there will be an error dumped into the host dmesg.  Two issues
with that:

  (1) Doing a trace dump when allocation fails is not anything close to
      grace.

  (2) The user should be notified with any kind of such error, so the user
      can trap it and decide what to do next, either by retrying, or stop
      the process properly, or anything else.

For userfault users, this will change the API of UFFDIO_WRITEPROTECT when
pgtable allocation failure happened.  It should not normally break anyone,
though.  If it breaks, then in good ways.

One man-page update will be on the way to introduce the new -ENOMEM for
UFFDIO_WRITEPROTECT.  Not marking stable so we keep the old behavior on
the 5.19-till-now kernels.

[akpm@linux-foundation.org: coding-style cleanups]
Link: https://lkml.kernel.org/r/20230104225207.1066932-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reported-by: James Houghton <jthoughton@google.com>
Acked-by: James Houghton <jthoughton@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/mprotect: use long for page accountings and retval

Switch to use type "long" for page accountings and retval across the whole
procedure of change_protection().

The change should have shrinked the possible maximum page number to be
half comparing to previous (ULONG_MAX / 2), but it shouldn't overflow on
any system either because the maximum possible pages touched by change
protection should be ULONG_MAX / PAGE_SIZE.

Two reasons to switch from "unsigned long" to "long":

  1. It suites better on count_vm_numa_events(), whose 2nd parameter takes
     a long type.

  2. It paves way for returning negative (error) values in the future.

Currently the only caller that consumes this retval is change_prot_numa(),
where the unsigned long was converted to an int.  Since at it, touching up
the numa code to also take a long, so it'll avoid any possible overflow
too during the int-size convertion.

Link: https://lkml.kernel.org/r/20230104225207.1066932-3-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: James Houghton <jthoughton@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/vaddr: convert hugetlb related functions to use a folio

Convert damon_hugetlb_mkold() and damon_young_hugetlb_entry() to
use a folio.

Link: https://lkml.kernel.org/r/20221230070849.63358-9-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon: remove unneeded damon_get_page()

After all damon_get_page() callers are converted to damon_get_folio(),
remove unneeded wrapper damon_get_page().

Link: https://lkml.kernel.org/r/20221230070849.63358-8-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/vaddr: convert damon_young_pmd_entry() to use a folio

With damon_get_folio(), let's convert damon_young_pmd_entry()
to use a folio.

Link: https://lkml.kernel.org/r/20221230070849.63358-7-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/paddr: convert damon_pa_*() to use a folio

With damon_get_folio(), let's convert all the damon_pa_*() to use a folio.

Link: https://lkml.kernel.org/r/20221230070849.63358-6-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon: convert damon_ptep/pmdp_mkold() to use a folio

With damon_get_folio(), let's convert damon_ptep_mkold() and
damon_pmdp_mkold() to use a folio.

Link: https://lkml.kernel.org/r/20221230070849.63358-5-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon: introduce damon_get_folio()

Introduce damon_get_folio(), and the temporary wrapper function
damon_get_page(), which help us to convert damon related functions to use
folios, and it will be dropped once the conversion is completed.

Link: https://lkml.kernel.org/r/20221230070849.63358-4-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: page_idle: convert page idle to use a folio

Firstly, make page_idle_get_page() return a folio, also rename it to
page_idle_get_folio(), then, use it to convert page_idle_bitmap_read() and
page_idle_bitmap_write() functions.

Link: https://lkml.kernel.org/r/20221230070849.63358-3-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memcg: add folio_memcg_check()

Patch series "mm: convert page_idle/damon to use folios", v4.

This patch (of 8):

Convert page_memcg_check() into folio_memcg_check() and add a
page_memcg_check() wrapper.  The behaviour of page_memcg_check() is
unchanged; tail pages always had a NULL ->memcg_data.

Link: https://lkml.kernel.org/r/20221230070849.63358-1-wangkefeng.wang@huawei.com
Link: https://lkml.kernel.org/r/20221230070849.63358-2-wangkefeng.wang@huawei.com
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

zram: fix typos in comments

- The double `range` is duplicated in comment, remove one.
 - change `syfs` to `sysfs`

Link: https://lkml.kernel.org/r/20221223040331.4194-1-jhs2.lee@samsung.com
Signed-off-by: JeongHyeon Lee <jhs2.lee@samsung.com>
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: huge_memory: convert split_huge_pages_all() to use a folio

Straightforwardly convert split_huge_pages_all() to use a folio.

Link: https://lkml.kernel.org/r/20221229122503.149083-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: remove generic_writepages

Now that all external callers are gone, just fold it into do_writepages.

Link: https://lkml.kernel.org/r/20221229161031.391878-7-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

ocfs2: use filemap_fdatawrite_wbc instead of generic_writepages

filemap_fdatawrite_wbc is a fairly thing wrapper around do_writepages, and
the big difference there is support for cgroup writeback, which is not
supported by ocfs2, and the potential to use ->writepages instead of
->writepage, which ocfs2 does not currently implement but eventually
should.

Link: https://lkml.kernel.org/r/20221229161031.391878-6-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

jbd2,ocfs2: move jbd2_journal_submit_inode_data_buffers to ocfs2

jbd2_journal_submit_inode_data_buffers is only used by ocfs2, so move it
there to prepare for removing generic_writepages.

Link: https://lkml.kernel.org/r/20221229161031.391878-5-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: Jan Kara <jack@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

ntfs3: remove ->writepage

->writepage is a very inefficient method to write back data, and only used
through write_cache_pages or a a fallback when no ->migrate_folio method
is present.

Set ->migrate_folio to the generic buffer_head based helper, and remove
the ->writepage implementation.

Link: https://lkml.kernel.org/r/20221229161031.391878-4-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: Jan Kara <jack@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

ntfs3: stop using generic_writepages

Open code the resident inode handling in ntfs_writepages by directly using
write_cache_pages to prepare removing the ->writepage handler in ntfs3.

Link: https://lkml.kernel.org/r/20221229161031.391878-3-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: Jan Kara <jack@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fs: remove an outdated comment on mpage_writepages

Patch series "remove generic_writepages"

This series removes generic_writepages by open coding the current
functionality in the three remaining callers.  Besides removing some
code the main benefit is that one of the few remaining ->writepage
callers from outside the core page cache code go away.

This patch (of 6):

mpage_writepages doesn't do any of the page locking itself, so remove and
outdated comment on the locking pattern there.

Link: https://lkml.kernel.org/r/20221229161031.391878-1-hch@lst.de
Link: https://lkml.kernel.org/r/20221229161031.391878-2-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/thp: check and bail out if page in deferred queue already

Kernel build regression with LLVM was reported here:
https://lore.kernel.org/all/Y1GCYXGtEVZbcv%2F5@dev-arch.thelio-3990X/ with
commit f35b5d7d676e ("mm: align larger anonymous mappings on THP
boundaries").  And the commit f35b5d7d676e was reverted.

It turned out the regression is related with madvise(MADV_DONTNEED)
was used by ld.lld. But with none PMD_SIZE aligned parameter len.
trace-bpfcc captured:
531607  531732  ld.lld          do_madvise.part.0 start: 0x7feca9000000, len: 0x7fb000, behavior: 0x4
531607  531793  ld.lld          do_madvise.part.0 start: 0x7fec86a00000, len: 0x7fb000, behavior: 0x4

If the underneath physical page is THP, the madvise(MADV_DONTNEED) can
trigger split_queue_lock contention raised significantly. perf showed
following data:
    14.85%     0.00%  ld.lld           [kernel.kallsyms]           [k]
       entry_SYSCALL_64_after_hwframe
           11.52%
                entry_SYSCALL_64_after_hwframe
                do_syscall_64
                __x64_sys_madvise
                do_madvise.part.0
                zap_page_range
                unmap_single_vma
                unmap_page_range
                page_remove_rmap
                deferred_split_huge_page
                __lock_text_start
                native_queued_spin_lock_slowpath

If THP can't be removed from rmap as whole THP, partial THP will be
removed from rmap by removing sub-pages from rmap.  Even the THP head page
is added to deferred queue already, the split_queue_lock will be acquired
and check whether the THP head page is in the queue already.  Thus, the
contention of split_queue_lock is raised.

Before acquire split_queue_lock, check and bail out early if the THP
head page is in the queue already. The checking without holding
split_queue_lock could race with deferred_split_scan, but it doesn't
impact the correctness here.

Test result of building kernel with ld.lld:
commit 7b5a0b664ebe (parent commit of f35b5d7d676e):
time -f "\t%E real,\t%U user,\t%S sys" make LD=ld.lld -skj96 allmodconfig all
        6:07.99 real,   26367.77 user,  5063.35 sys

commit f35b5d7d676e:
time -f "\t%E real,\t%U user,\t%S sys" make LD=ld.lld -skj96 allmodconfig all
        7:22.15 real,   26235.03 user,  12504.55 sys

commit f35b5d7d676e with the fixing patch:
time -f "\t%E real,\t%U user,\t%S sys" make LD=ld.lld -skj96 allmodconfig all
        6:08.49 real,   26520.15 user,  5047.91 sys

Link: https://lkml.kernel.org/r/20221223135207.2275317-1-fengwei.yin@intel.com
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Feng Tang <feng.tang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_reporting: replace rcu_access_pointer() with rcu_dereference_protected()

Page reporting fetches pr_dev_info using rcu_access_pointer(), which is
for safely fetching a pointer that will not be dereferenced but could
concurrently updated.  The code indeed does not dereference pr_dev_info
after fetching it using rcu_access_pointer(), but it fetches the pointer
while concurrent updates to the pointer is avoided by holding the update
side lock, page_reporting_mutex.

In the case, rcu_dereference_protected() should be used instead because it
provides better readability and performance on some cases, as
rcu_dereference_protected() avoids use of READ_ONCE().  Replace the
rcu_access_pointer() calls with rcu_dereference_protected().

Link: https://lkml.kernel.org/r/20221228175942.149491-1-sj@kernel.org
Fixes: 36e66c554b5c ("mm: introduce Reported pages")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: fix comment of page table counter

Commit af5b0f6a09e42 ("mm: consolidate page table accounting")
consolidates page table accounting to a single counter in struct mm_struct
{} as mm->pgtables_bytes.  So the meanning of this counter should be the
size of all page tables now.

Link: https://lkml.kernel.org/r/20221224060233.417827-1-kele.huang@columbia.edu
Signed-off-by: Kele Huang <kele.huang@columbia.edu>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Colin Cross <ccross@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/mprotect: drop pgprot_t parameter from change_protection()

Being able to provide a custom protection opens the door for
inconsistencies and BUGs: for example, accidentally allowing for more
permissions than desired by other mechanisms (e.g., softdirty tracking).
vma->vm_page_prot should be the single source of truth.

Only PROT_NUMA is special: there is no way we can erroneously allow
for more permissions when removing all permissions. Special-case using
the MM_CP_PROT_NUMA flag.

[david@redhat.com: PAGE_NONE might not be defined without CONFIG_NUMA_BALANCING]
Link: https://lkml.kernel.org/r/5084ff1c-ebb3-f918-6a60-bacabf550a88@redhat.com
Link: https://lkml.kernel.org/r/20221223155616.297723-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/userfaultfd: rely on vma->vm_page_prot in uffd_wp_range()

Patch series "mm: uffd-wp + change_protection() cleanups".

Cleanup page protection handling in uffd-wp when calling
change_protection() and improve unprotecting uffd=wp in private mappings,
trying to set PTEs writable again if possible just like we do during
mprotect() when upgrading write permissions.  Make the change_protection()
interface harder to get wrong :)

I consider both pages primarily cleanups, although patch #1 fixes a corner
case with uffd-wp and softdirty tracking for shmem.  @Peter, please let me
know if we should flag patch #1 as pure cleanup -- I have no idea how
important softdirty tracking on shmem is.

This patch (of 2):

uffd_wp_range() currently calculates page protection manually using
vm_get_page_prot().  This will ignore any other reason for active
writenotify: one mechanism applicable to shmem is softdirty tracking.

For example, the following sequence

1) Write to mapped shmem page
2) Clear softdirty
3) Register uffd-wp covering the mapped page
4) Unregister uffd-wp covering the mapped page
5) Write to page again

will not set the modified page softdirty, because uffd_wp_range() will
ignore that writenotify is required for softdirty tracking and simply map
the page writable again using change_protection().  Similarly, instead of
unregistering, protecting followed by un-protecting the page using uffd-wp
would result in the same situation.

Now that we enable writenotify whenever enabling uffd-wp on a VMA,
vma->vm_page_prot will already properly reflect our requirements: the
default is to write-protect all PTEs.  However, for shared mappings we
would now not remap the PTEs writable if possible when unprotecting, just
like for private mappings (COW).  To compensate, set
MM_CP_TRY_CHANGE_WRITABLE just like mprotect() does to try mapping
individual PTEs writable.

For private mappings, this change implies that we will now always try
setting PTEs writable when un-protecting, just like when upgrading write
permissions using mprotect(), which is an improvement.

For shared mappings, we will only set PTEs writable if
can_change_pte_writable()/can_change_pmd_writable() indicates that it's
ok.  For ordinary shmem, this will be the case when PTEs are dirty, which
should usually be the case -- otherwise we could special-case shmem in
can_change_pte_writable()/can_change_pmd_writable() easily, because shmem
itself doesn't require writenotify.

Note that hugetlb does not yet implement MM_CP_TRY_CHANGE_WRITABLE, so we
won't try setting PTEs writable when unprotecting or when unregistering
uffd-wp.  This can be added later on top by implementing
MM_CP_TRY_CHANGE_WRITABLE.

While commit ffd05793963a ("userfaultfd: wp: support write protection for
userfault vma range") introduced that code, it should only be applicable
to uffd-wp on shared mappings -- shmem (hugetlb does not support softdirty
tracking).  I don't think this corner cases justifies to cc stable.  Let's
just handle it correctly and prepare for change_protection() cleanups.

[david@redhat.com: o need for additional harmless checks if we're wr-protecting either way]
Link: https://lkml.kernel.org/r/71412742-a71f-9c74-865f-773ad83db7a5@redhat.com
Link: https://lkml.kernel.org/r/20221223155616.297723-1-david@redhat.com
Link: https://lkml.kernel.org/r/20221223155616.297723-2-david@redhat.com
Fixes: b1f9e876862d ("mm/uffd: enable write protection for shmem & hugetlbfs")
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: ksm_functional_tests: fix a typo in comment

Fix a typo of "comaring" which should be "comparing".

Link: https://lkml.kernel.org/r/202212231050245952617@zte.com.cn
Signed-off-by: Xu Panda <xu.panda@zte.com.cn>
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: simplify arch_has_hw_pte_young() check

Scanning page tables when hardware does not set the accessed bit has
no real use cases.

Link: https://lkml.kernel.org/r/20221222041905.2431096-9-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: clarify scan_control flags

Among the flags in scan_control:
1. sc->may_swap, which indicates swap constraint due to memsw.max, is
   supported as usual.
2. sc->proactive, which indicates reclaim by memory.reclaim, may not
   opportunistically skip the aging path, since it is considered less
   latency sensitive.
3. !(sc->gfp_mask & __GFP_IO), which indicates IO constraint, lowers
   swappiness to prioritize file LRU, since clean file folios are more
   likely to exist.
4. sc->may_writepage and sc->may_unmap, which indicates opportunistic
   reclaim, are rejected, since unmapped clean folios are already
   prioritized. Scanning for more of them is likely futile and can
   cause high reclaim latency when there is a large number of memcgs.

The rest are handled by the existing code.

Link: https://lkml.kernel.org/r/20221222041905.2431096-8-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: per-node lru_gen_folio lists

For each node, memcgs are divided into two generations: the old and
the young. For each generation, memcgs are randomly sharded into
multiple bins to improve scalability. For each bin, an RCU hlist_nulls
is virtually divided into three segments: the head, the tail and the
default.

An onlining memcg is added to the tail of a random bin in the old
generation. The eviction starts at the head of a random bin in the old
generation. The per-node memcg generation counter, whose reminder (mod
2) indexes the old generation, is incremented when all its bins become
empty.

There are four operations:
1. MEMCG_LRU_HEAD, which moves an memcg to the head of a random bin in
   its current generation (old or young) and updates its "seg" to
   "head";
2. MEMCG_LRU_TAIL, which moves an memcg to the tail of a random bin in
   its current generation (old or young) and updates its "seg" to
   "tail";
3. MEMCG_LRU_OLD, which moves an memcg to the head of a random bin in
   the old generation, updates its "gen" to "old" and resets its "seg"
   to "default";
4. MEMCG_LRU_YOUNG, which moves an memcg to the tail of a random bin
   in the young generation, updates its "gen" to "young" and resets
   its "seg" to "default".

The events that trigger the above operations are:
1. Exceeding the soft limit, which triggers MEMCG_LRU_HEAD;
2. The first attempt to reclaim an memcg below low, which triggers
   MEMCG_LRU_TAIL;
3. The first attempt to reclaim an memcg below reclaimable size
   threshold, which triggers MEMCG_LRU_TAIL;
4. The second attempt to reclaim an memcg below reclaimable size
   threshold, which triggers MEMCG_LRU_YOUNG;
5. Attempting to reclaim an memcg below min, which triggers
   MEMCG_LRU_YOUNG;
6. Finishing the aging on the eviction path, which triggers
   MEMCG_LRU_YOUNG;
7. Offlining an memcg, which triggers MEMCG_LRU_OLD.

Note that memcg LRU only applies to global reclaim, and the
round-robin incrementing of their max_seq counters ensures the
eventual fairness to all eligible memcgs. For memcg reclaim, it still
relies on mem_cgroup_iter().

Link: https://lkml.kernel.org/r/20221222041905.2431096-7-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: shuffle should_run_aging()

Move should_run_aging() next to its only caller left.

Link: https://lkml.kernel.org/r/20221222041905.2431096-6-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: remove aging fairness safeguard

Recall that the aging produces the youngest generation: first it scans
for accessed folios and updates their gen counters; then it increments
lrugen->max_seq.

The current aging fairness safeguard for kswapd uses two passes to
ensure the fairness to multiple eligible memcgs. On the first pass,
which is shared with the eviction, it checks whether all eligible
memcgs are low on cold folios. If so, it requires a second pass, on
which it ages all those memcgs at the same time.

With memcg LRU, the aging, while ensuring eventual fairness, will run
when necessary. Therefore the current aging fairness safeguard for
kswapd will not be needed.

Note that memcg LRU only applies to global reclaim. For memcg reclaim,
the aging can be unfair to different memcgs, i.e., their
lrugen->max_seq can be incremented at different paces.

Link: https://lkml.kernel.org/r/20221222041905.2431096-5-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: remove eviction fairness safeguard

Recall that the eviction consumes the oldest generation: first it
bucket-sorts folios whose gen counters were updated by the aging and
reclaims the rest; then it increments lrugen->min_seq.

The current eviction fairness safeguard for global reclaim has a
dilemma: when there are multiple eligible memcgs, should it continue
or stop upon meeting the reclaim goal? If it continues, it overshoots
and increases direct reclaim latency; if it stops, it loses fairness
between memcgs it has taken memory away from and those it has yet to.

With memcg LRU, the eviction, while ensuring eventual fairness, will
stop upon meeting its goal. Therefore the current eviction fairness
safeguard for global reclaim will not be needed.

Note that memcg LRU only applies to global reclaim. For memcg reclaim,
the eviction will continue, even if it is overshooting. This becomes
unconditional due to code simplification.

Link: https://lkml.kernel.org/r/20221222041905.2431096-4-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: rename lrugen->lists[] to lrugen->folios[]

lru_gen_folio will be chained into per-node lists by the coming
lrugen->list.

Link: https://lkml.kernel.org/r/20221222041905.2431096-3-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: multi-gen LRU: rename lru_gen_struct to lru_gen_folio

Patch series "mm: multi-gen LRU: memcg LRU", v3.

Overview
========

An memcg LRU is a per-node LRU of memcgs.  It is also an LRU of LRUs,
since each node and memcg combination has an LRU of folios (see
mem_cgroup_lruvec()).

Its goal is to improve the scalability of global reclaim, which is
critical to system-wide memory overcommit in data centers.  Note that
memcg reclaim is currently out of scope.

Its memory bloat is a pointer to each lruvec and negligible to each
pglist_data.  In terms of traversing memcgs during global reclaim, it
improves the best-case complexity from O(n) to O(1) and does not affect
the worst-case complexity O(n).  Therefore, on average, it has a sublinear
complexity in contrast to the current linear complexity.

The basic structure of an memcg LRU can be understood by an analogy to
the active/inactive LRU (of folios):
1. It has the young and the old (generations), i.e., the counterparts
   to the active and the inactive;
2. The increment of max_seq triggers promotion, i.e., the counterpart
   to activation;
3. Other events trigger similar operations, e.g., offlining an memcg
   triggers demotion, i.e., the counterpart to deactivation.

In terms of global reclaim, it has two distinct features:
1. Sharding, which allows each thread to start at a random memcg (in
   the old generation) and improves parallelism;
2. Eventual fairness, which allows direct reclaim to bail out at will
   and reduces latency without affecting fairness over some time.

The commit message in patch 6 details the workflow:
https://lore.kernel.org/r/20221222041905.2431096-7-yuzhao@google.com/

The following is a simple test to quickly verify its effectiveness.

  Test design:
  1. Create multiple memcgs.
  2. Each memcg contains a job (fio).
  3. All jobs access the same amount of memory randomly.
  4. The system does not experience global memory pressure.
  5. Periodically write to the root memory.reclaim.

  Desired outcome:
  1. All memcgs have similar pgsteal counts, i.e., stddev(pgsteal)
     over mean(pgsteal) is close to 0%.
  2. The total pgsteal is close to the total requested through
     memory.reclaim, i.e., sum(pgsteal) over sum(requested) is close
     to 100%.

  Actual outcome [1]:
                                     MGLRU off    MGLRU on
  stddev(pgsteal) / mean(pgsteal)    75%          20%
  sum(pgsteal) / sum(requested)      425%         95%

  ####################################################################
  MEMCGS=128

  for ((memcg = 0; memcg < $MEMCGS; memcg++)); do
      mkdir /sys/fs/cgroup/memcg$memcg
  done

  start() {
      echo $BASHPID > /sys/fs/cgroup/memcg$memcg/cgroup.procs

      fio -name=memcg$memcg --numjobs=1 --ioengine=mmap \
          --filename=/dev/zero --size=1920M --rw=randrw \
          --rate=64m,64m --random_distribution=random \
          --fadvise_hint=0 --time_based --runtime=10h \
          --group_reporting --minimal
  }

  for ((memcg = 0; memcg < $MEMCGS; memcg++)); do
      start &
  done

  sleep 600

  for ((i = 0; i < 600; i++)); do
      echo 256m >/sys/fs/cgroup/memory.reclaim
      sleep 6
  done

  for ((memcg = 0; memcg < $MEMCGS; memcg++)); do
      grep "pgsteal " /sys/fs/cgroup/memcg$memcg/memory.stat
  done
  ####################################################################

[1]: This was obtained from running the above script (touches less
     than 256GB memory) on an EPYC 7B13 with 512GB DRAM for over an
     hour.

This patch (of 8):

The new name lru_gen_folio will be more distinct from the coming
lru_gen_memcg.

Link: https://lkml.kernel.org/r/20221222041905.2431096-1-yuzhao@google.com
Link: https://lkml.kernel.org/r/20221222041905.2431096-2-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: vmalloc: replace BUG_ON() by WARN_ON_ONCE()

Currently a vm_unmap_ram() functions triggers a BUG() if an area is not
found.  Replace it by the WARN_ON_ONCE() error message and keep machine
alive instead of stopping it.

The worst case is a memory leaking.

Link: https://lkml.kernel.org/r/20221222190022.134380-3-urezki@gmail.com
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Lorenzo Stoakes <lstoakes@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Baoquan He <bhe@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: vmalloc: avoid calling __find_vmap_area() twice in __vunmap()

Currently the __vunmap() path calls __find_vmap_area() twice.  Once on
entry to check that the area exists, then inside the remove_vm_area()
function which also performs a new search for the VA.

In order to improvie it from a performance point of view we split
remove_vm_area() into two new parts:
  - find_unlink_vmap_area() that does a search and unlink from tree;
  - __remove_vm_area() that removes without searching.

In this case there is no any functional change for remove_vm_area()
whereas vm_remove_mappings(), where a second search happens, switches to
the __remove_vm_area() variant where the already detached VA is passed as
a parameter, so there is no need to find it again.

Performance wise, i use test_vmalloc.sh with 32 threads doing alloc
free on a 64-CPUs-x86_64-box:

perf without this patch:
-   31.41%     0.50%  vmalloc_test/10  [kernel.vmlinux]    [k] __vunmap
   - 30.92% __vunmap
      - 17.67% _raw_spin_lock
           native_queued_spin_lock_slowpath
      - 12.33% remove_vm_area
         - 11.79% free_vmap_area_noflush
            - 11.18% _raw_spin_lock
                 native_queued_spin_lock_slowpath
        0.76% free_unref_page

perf with this patch:
-   11.35%     0.13%  vmalloc_test/14  [kernel.vmlinux]    [k] __vunmap
   - 11.23% __vunmap
      - 8.28% find_unlink_vmap_area
         - 7.95% _raw_spin_lock
              7.44% native_queued_spin_lock_slowpath
      - 1.93% free_vmap_area_noflush
         - 0.56% _raw_spin_lock
              0.53% native_queued_spin_lock_slowpath
        0.60% __vunmap_range_noflush

__vunmap() consumes around ~20% less CPU cycles on this test.

Also, switch from find_vmap_area() to find_unlink_vmap_area() to prevent a
double access to the vmap_area_lock: one for finding area, second time is
for unlinking from a tree.

[urezki@gmail.com: switch to find_unlink_vmap_area() in vm_unmap_ram()]
Link: https://lkml.kernel.org/r/20221222190022.134380-2-urezki@gmail.com
Link: https://lkml.kernel.org/r/20221222190022.134380-1-urezki@gmail.com
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reported-by: Roman Gushchin <roman.gushchin@linux.dev>
Reviewed-by: Lorenzo Stoakes <lstoakes@gmail.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: move FOLL_* defs to mm_types.h

Move FOLL_* definitions to linux/mm_types.h to make them more accessible
without having to drag in all of linux/mm.h and everything that drags in
too[1].

Link: https://lkml.kernel.org/r/2161258.1671657894@warthog.procyon.org.uk
Signed-off-by: David Howells <dhowells@redhat.com>
Suggested-by: Matthew Wilcox <willy@infradead.org>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: new primitive kvmemdup()

Similar to kmemdup(), but support large amount of bytes with kvmalloc()
and does *not* guarantee that the result will be physically contiguous.
Use only in cases where kvmalloc() is needed and free it with kvfree().
Also adapt policy_unpack.c in case someone bisect into this.

Link: https://lkml.kernel.org/r/20221221144245.27164-1-sunhao.th@gmail.com
Signed-off-by: Hao Sun <sunhao.th@gmail.com>
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Nick Terrell <terrelln@fb.com>
Cc: John Johansen <john.johansen@canonical.com>
Cc: Paul Moore <paul@paul-moore.com>
Cc: James Morris <jmorris@namei.org>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert deactivate_page() to folio_deactivate()

Deactivate_page() has already been converted to use folios, this change
converts it to take in a folio argument instead of calling page_folio().
It also renames the function folio_deactivate() to be more consistent with
other folio functions.

[akpm@linux-foundation.org: fix left-over comments, per Yu Zhao]
Link: https://lkml.kernel.org/r/20221221180848.20774-5-vishal.moola@gmail.com
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon: convert damon_pa_mark_accessed_or_deactivate() to use folios

This change replaces 2 calls to compound_head() from put_page() and 1 call
from mark_page_accessed() with one from page_folio().  This is in
preparation for the conversion of deactivate_page() to folio_deactivate().

Link: https://lkml.kernel.org/r/20221221180848.20774-4-vishal.moola@gmail.com
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

madvise: convert madvise_cold_or_pageout_pte_range() to use folios

This change removes a number of calls to compound_head(), and saves
1729 bytes of kernel text.

Link: https://lkml.kernel.org/r/20221221180848.20774-3-vishal.moola@gmail.com
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/memory: add vm_normal_folio()

Patch series "Convert deactivate_page() to folio_deactivate()", v4.

Deactivate_page() has already been converted to use folios.  This patch
series modifies the callers of deactivate_page() to use folios.  It also
introduces vm_normal_folio() to assist with folio conversions, and
converts deactivate_page() to folio_deactivate() which takes in a folio.

This patch (of 4):

Introduce a wrapper function called vm_normal_folio().  This function
calls vm_normal_page() and returns the folio of the page found, or null if
no page is found.

This function allows callers to get a folio from a pte, which will
eventually allow them to completely replace their struct page variables
with struct folio instead.

Link: https://lkml.kernel.org/r/20221221180848.20774-1-vishal.moola@gmail.com
Link: https://lkml.kernel.org/r/20221221180848.20774-2-vishal.moola@gmail.com
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: refine mab_calc_split function

Invert the conditional judgment of the mid_split, to focus the return
statement in the last statement, which is easier to understand and for
better readability.

Link: https://lkml.kernel.org/r/20221221060058.609003-8-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: refine ma_state init from mas_start()

If mas->node is an MAS_START, there are three cases, and they all assign
different values to mas->node and mas->offset.  So there is no need to set
them to a default value before updating.

Update them directly to make them easier to understand and for better
readability.

Link: https://lkml.kernel.org/r/20221221060058.609003-7-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: remove the redundant code

The macros CONFIG_DEBUG_MAPLE_TREE_VERBOSE no one uses, functions
mas_dup_tree() and mas_dup_store() are not implemented, just function
declaration, so drop it.

Link: https://lkml.kernel.org/r/20221221060058.609003-6-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: use macro MA_ROOT_PARENT instead of number

When you need to compare whether node->parent is parent of the
root node, using macro MA_ROOT_PARENT is easier to understand
and for better readability.

Link: https://lkml.kernel.org/r/20221221060058.609003-5-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: use mt_node_max() instead of direct operations mt_max[]

Use mt_node_max() to get the maximum number of slots for a node,
rather than direct operations mt_max[], makes it better portability.

Link: https://lkml.kernel.org/r/20221221060058.609003-4-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: remove extra return statement

For functions with a return type of void, it is unnecessary to
add a reurn statement at the end of the function, so drop it.

Link: https://lkml.kernel.org/r/20221221060058.609003-3-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

maple_tree: remove extra space and blank line

Patch series "Clean up and refinement for maple tree", v2.

This patchset cleans up and refines some maple tree code.  A few small
changes make the code easier to understand and for better readability.

This patch (of 7):

These extra space and blank lines are unnecessary, so drop them.

Link: https://lkml.kernel.org/r/20221221060058.609003-1-vernon2gm@gmail.com
Link: https://lkml.kernel.org/r/20221221060058.609003-2-vernon2gm@gmail.com
Signed-off-by: Vernon Yang <vernon2gm@gmail.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: vmalloc: correct use of __GFP_NOWARN mask in __vmalloc_area_node()

This function sets __GFP_NOWARN in the gfp_mask rendering the warn_alloc()
invocations no-ops.  Remove this and instead rely on this flag being set
only for the vm_area_alloc_pages() function, ensuring it is cleared for
each of the warn_alloc() calls.

Link: https://lkml.kernel.org/r/20221219123659.90614-1-lstoakes@gmail.com
Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com>
Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

tools/vm/page_owner_sort: free memory before exit

Although when a process terminates, the kernel will removes memory
associated with that process, It's neither good style nor proper design to
leave it to kernel.  This patch free allocated memory before process exit.

Link: https://lkml.kernel.org/r/20221219164917.14132-1-iecedge@gmail.com
Signed-off-by: Jianlin Lv <iecedge@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

kasan: allow sampling page_alloc allocations for HW_TAGS

As Hardware Tag-Based KASAN is intended to be used in production, its
performance impact is crucial.  As page_alloc allocations tend to be big,
tagging and checking all such allocations can introduce a significant
slowdown.

Add two new boot parameters that allow to alleviate that slowdown:

- kasan.page_alloc.sample, which makes Hardware Tag-Based KASAN tag only
  every Nth page_alloc allocation with the order configured by the second
  added parameter (default: tag every such allocation).

- kasan.page_alloc.sample.order, which makes sampling enabled by the first
  parameter only affect page_alloc allocations with the order equal or
  greater than the specified value (default: 3, see below).

The exact performance improvement caused by using the new parameters
depends on their values and the applied workload.

The chosen default value for kasan.page_alloc.sample.order is 3, which
matches both PAGE_ALLOC_COSTLY_ORDER and SKB_FRAG_PAGE_ORDER.  This is
done for two reasons:

1. PAGE_ALLOC_COSTLY_ORDER is "the order at which allocations are deemed
   costly to service", which corresponds to the idea that only large and
   thus costly allocations are supposed to sampled.

2. One of the workloads targeted by this patch is a benchmark that sends
   a large amount of data over a local loopback connection. Most multi-page
   data allocations in the networking subsystem have the order of
   SKB_FRAG_PAGE_ORDER (or PAGE_ALLOC_COSTLY_ORDER).

When running a local loopback test on a testing MTE-enabled device in sync
mode, enabling Hardware Tag-Based KASAN introduces a ~50% slowdown.
Applying this patch and setting kasan.page_alloc.sampling to a value
higher than 1 allows to lower the slowdown.  The performance improvement
saturates around the sampling interval value of 10 with the default
sampling page order of 3.  This lowers the slowdown to ~20%.  The slowdown
in real scenarios involving the network will likely be better.

Enabling page_alloc sampling has a downside: KASAN misses bad accesses to
a page_alloc allocation that has not been tagged.  This lowers the value
of KASAN as a security mitigation.

However, based on measuring the number of page_alloc allocations of
different orders during boot in a test build, sampling with the default
kasan.page_alloc.sample.order value affects only ~7% of allocations.  The
rest ~93% of allocations are still checked deterministically.

Link: https://lkml.kernel.org/r/129da0614123bb85ed4dd61ae30842b2dd7c903f.1671471846.git.andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Reviewed-by: Marco Elver <elver@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Brand <markbrand@google.com>
Cc: Peter Collingbourne <pcc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

swap: avoid holding swap reference in swap_cache_get_folio

All its callers either already hold a reference to, or lock the swap
device while calling this function.  There is only one exception in
shmem_swapin_folio, just make this caller also hold a reference of the
swap device, so this helper can be simplified and saves a few cycles.

This also provides finer control of error handling in shmem_swapin_folio,
on race (with swap off), it can just try again.  For invalid swap entry,
it can fail with a proper error code.

Link: https://lkml.kernel.org/r/20221219185840.25441-5-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

swap: fold swap_ra_clamp_pfn into swap_ra_info

This makes the code cleaner.  This helper is made of only two line of self
explanational code and not reused anywhere else.

And this actually make the compiled object smaller by a bit.

bloat-o-meter results on x86_64 of mm/swap_state.o:

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-35 (-35)
Function                                     old     new   delta
swap_ra_info.constprop                       512     477     -35
Total: Before=8388, After=8353, chg -0.42%

Link: https://lkml.kernel.org/r/20221219185840.25441-4-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

swap: avoid a redundant pte map if ra window is 1

Avoid a redundant pte map/unmap when swap readahead window is 1.

Link: https://lkml.kernel.org/r/20221219185840.25441-3-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

swapfile: get rid of volatile and avoid redundant read

Patch series "Clean up and fixes for swap", v2.

This series cleans up some code paths, saves a few cycles and reduces the
object size by a bit.  It also fixes some rare race issue with statistics.

This patch (of 4):

Convert a volatile variable to more readable READ_ONCE.  And this actually
avoids the code from reading the variable twice redundantly when it races.

Link: https://lkml.kernel.org/r/20221219185840.25441-1-ryncsn@gmail.com
Link: https://lkml.kernel.org/r/20221219185840.25441-2-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Docs/ABI/damon: document scheme filters files

Document newly added DAMON sysfs interface files for DAMOS filtering on
the DAMON ABI document.

Link: https://lkml.kernel.org/r/20221205230830.144349-12-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Docs/admin-guide/mm/damon/usage: document DAMOS filters of sysfs

Document about the newly added files for DAMOS filters on the DAMON usage
document.

Link: https://lkml.kernel.org/r/20221205230830.144349-11-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/damon/sysfs: test filters directory

Add simple test cases for scheme filters of DAMON sysfs interface.  The
test cases check if the files are populated as expected, receives some
valid inputs, and refuses some invalid inputs.

Link: https://lkml.kernel.org/r/20221205230830.144349-10-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/sysfs-schemes: implement scheme filters

Implement scheme filters functionality of DAMON sysfs interface by making
the code reads the values of files under the filter directories and pass
that to DAMON using DAMON kernel API.

[sj@kernel.org: fix leaking a filter for wrong cgroup path]
Link: https://lkml.kernel.org/r/20221219171807.55708-2-sj@kernel.org
[sj@kernel.org: return an error for filter memcg path id lookup failure]
Link: https://lkml.kernel.org/r/20221219171807.55708-3-sj@kernel.org
Link: https://lkml.kernel.org/r/20221205230830.144349-9-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/sysfs-schemes: connect filter directory and filters directory

Implement 'nr_filters' file under 'filters' directory, which will be used
to populate specific number of 'filter' directory under the directory,
similar to other 'nr_*' files in DAMON sysfs interface.

Link: https://lkml.kernel.org/r/20221205230830.144349-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/sysfs-schemes: implement filter directory

Implement DAMOS filter directory which will be located under the filters
directory.  The directory provides three files, namely type, matching, and
memcg_path.  'type' and 'matching' will be directly connected to the
fields of 'struct damos_filter' having same name.  'memcg_path' will
receive the path of the memory cgroup of the interest and later converted
to memcg id when it's committed.

Link: https://lkml.kernel.org/r/20221205230830.144349-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/sysfs-schemes: implement filters directory

DAMOS filters are currently supported by only DAMON kernel API.  To expose
the feature to user space, implement a DAMON sysfs directory named
'filters' under each scheme directory.  Please note that this is
implementing only the directory.  Following commits will implement more
files and directories, and finally connect the DAMOS filters feature.

Link: https://lkml.kernel.org/r/20221205230830.144349-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Docs/admin-guide/damon/reclaim: document 'skip_anon' parameter

Document the newly added 'skip_anon' parameter of DAMON_RECLAIM, which can
be used to avoid anonymous pages reclamation.

Link: https://lkml.kernel.org/r/20221205230830.144349-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/reclaim: add a parameter called skip_anon for avoiding anonymous pages reclamation

In some cases, for example if users have confidence at anonymous pages
management or the swap device is too slow, users would want to avoid
DAMON_RECLAIM swapping the anonymous pages out.  For such case, add yet
another DAMON_RECLAIM parameter, namely 'skip_anon'.  When it is set as
'Y', DAMON_RECLAIM will avoid reclaiming anonymous pages using a DAMOS
filter.

Link: https://lkml.kernel.org/r/20221205230830.144349-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/paddr: support DAMOS filters

Implement support of the DAMOS filters in the physical address space
monitoring operations set, for all DAMOS actions that it supports
including 'pageout', 'lru_prio', and 'lru_deprio'.

Link: https://lkml.kernel.org/r/20221205230830.144349-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/core: implement damos filter

Patch series "implement DAMOS filtering for anon pages and/or specific
memory cgroups"

DAMOS let users do system operations in a data access pattern oriented
way.  The data access pattern, which is extracted by DAMON, is somewhat
accurate more than what user space could know in many cases.  However, in
some situation, users could know something more than the kernel about the
pattern or some special requirements for some types of memory or
processes.  For example, some users would have slow swap devices and knows
latency-ciritical processes and therefore want to use DAMON-based
proactive reclamation (DAMON_RECLAIM) for only non-anonymous pages of
non-latency-critical processes.

For such restriction, users could exclude the memory regions from the
initial monitoring regions and use non-dynamic monitoring regions update
monitoring operations set including fvaddr and paddr.  They could also
adjust the DAMOS target access pattern.  For dynamically changing memory
layout and access pattern, those would be not enough.

To help the case, add an interface, namely DAMOS filters, which can be
used to avoid the DAMOS actions be applied to specific types of memory, to
DAMON kernel API (damon.h).  At the moment, it supports filtering
anonymous pages and/or specific memory cgroups in or out for each DAMOS
scheme.

This patchset adds the support for all DAMOS actions that 'paddr'
monitoring operations set supports ('pageout', 'lru_prio', and
'lru_deprio'), and the functionality is exposed via DAMON kernel API
(damon.h) the DAMON sysfs interface (/sys/kernel/mm/damon/admins/), and
DAMON_RECLAIM module parameters.

Patches Sequence
----------------

First patch implements DAMOS filter interface to DAMON kernel API.  Second
patch makes the physical address space monitoring operations set to
support the filters from all supporting DAMOS actions.  Third patch adds
anonymous pages filter support to DAMON_RECLAIM, and the fourth patch
documents the DAMON_RECLAIM's new feature.  Fifth to seventh patches
implement DAMON sysfs files for support of the filters, and eighth patch
connects the file to use DAMOS filters feature.  Ninth patch adds simple
self test cases for DAMOS filters of the sysfs interface.  Finally,
following two patches (tenth and eleventh) document the new features and
interfaces.

This patch (of 11):

DAMOS lets users do system operation in a data access pattern oriented
way.  The data access pattern, which is extracted by DAMON, is somewhat
accurate more than what user space could know in many cases.  However, in
some situation, users could know something more than the kernel about the
pattern or some special requirements for some types of memory or
processes.  For example, some users would have slow swap devices and knows
latency-ciritical processes and therefore want to use DAMON-based
proactive reclamation (DAMON_RECLAIM) for only non-anonymous pages of
non-latency-critical processes.

For such restriction, users could exclude the memory regions from the
initial monitoring regions and use non-dynamic monitoring regions update
monitoring operations set including fvaddr and paddr.  They could also
adjust the DAMOS target access pattern.  For dynamically changing memory
layout and access pattern, those would be not enough.

To help the case, add an interface, namely DAMOS filters, which can be
used to avoid the DAMOS actions be applied to specific types of memory, to
DAMON kernel API (damon.h).  At the moment, it supports filtering
anonymous pages and/or specific memory cgroups in or out for each DAMOS
scheme.

Note that this commit adds only the interface to the DAMON kernel API.
The impelmentation should be made in the monitoring operations sets, and
following commits will add that.

Link: https://lkml.kernel.org/r/20221205230830.144349-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20221205230830.144349-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memcontrol: deprecate charge moving

Charge moving mode in cgroup1 allows memory to follow tasks as they
migrate between cgroups.  This is, and always has been, a questionable
thing to do - for several reasons.

First, it's expensive.  Pages need to be identified, locked and isolated
from various MM operations, and reassigned, one by one.

Second, it's unreliable.  Once pages are charged to a cgroup, there isn't
always a clear owner task anymore.  Cache isn't moved at all, for example.
Mapped memory is moved - but if trylocking or isolating a page fails,
it's arbitrarily left behind.  Frequent moving between domains may leave a
task's memory scattered all over the place.

Third, it isn't really needed.  Launcher tasks can kick off workload tasks
directly in their target cgroup.  Using dedicated per-workload groups
allows fine-grained policy adjustments - no need to move tasks and their
physical pages between control domains.  The feature was never
forward-ported to cgroup2, and it hasn't been missed.

Despite it being a niche usecase, the maintenance overhead of supporting
it is enormous.  Because pages are moved while they are live and subject
to various MM operations, the synchronization rules are complicated.
There are lock_page_memcg() in MM and FS code, which non-cgroup people
don't understand.  In some cases we've been able to shift code and cgroup
API calls around such that we can rely on native locking as much as
possible.  But that's fragile, and sometimes we need to hold MM locks for
longer than we otherwise would (pte lock e.g.).

Mark the feature deprecated. Hopefully we can remove it soon.

And backport into -stable kernels so that people who develop against
earlier kernels are warned about this deprecation as early as possible.

[akpm@linux-foundation.org: fix memory.rst underlining]
Link: https://lkml.kernel.org/r/Y5COd+qXwk/S+n8N@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: rmap: remove lock_page_memcg()

The previous patch made sure charge moving only touches pages for which
page_mapped() is stable.  lock_page_memcg() is no longer needed.

Link: https://lkml.kernel.org/r/20221206171340.139790-3-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memcontrol: skip moving non-present pages that are mapped elsewhere

Patch series "mm: push down lock_page_memcg()", v2.

This patch (of 3):

During charge moving, the pte lock and the page lock cover nearly all
cases of stabilizing page_mapped().  The only exception is when we're
looking at a non-present pte and find a page in the page cache or in the
swapcache: if the page is mapped elsewhere, it can become unmapped outside
of our control.  For this reason, rmap needs lock_page_memcg().

We don't like cgroup-specific locks in generic MM code - especially in
performance-critical MM code - and for a legacy feature that's unlikely to
have many users left - if any.

So remove the exception.  Arguably that's better semantics anyway: the
page is shared, and another process seems to be the more active user.

Once we stop moving such pages, rmap doesn't need lock_page_memcg()
anymore.  The next patch will remove it.

Link: https://lkml.kernel.org/r/20221206171340.139790-1-hannes@cmpxchg.org
Link: https://lkml.kernel.org/r/20221206171340.139790-2-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Suggested-by: Hugh Dickins <hughd@google.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb: initialize variable to avoid compiler warning

With the gcc 'maybe-uninitialized' warning enabled, gcc will produce:

  mm/hugetlb.c:6896:20: warning: `chg' may be used uninitialized

This is a false positive, but may be difficult for the compiler to
determine.  maybe-uninitialized is disabled by default, but this gets
flagged as a 0-DAY build regression.

Initialize the variable to silence the warning.

Link: https://lkml.kernel.org/r/20221216224507.106789-1-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: swap: convert mark_page_lazyfree() to folio_mark_lazyfree()

mark_page_lazyfree() and the callers are converted to use folio, this
rename and make it to take in a folio argument instead of calling
page_folio().

Link: https://lkml.kernel.org/r/20221209020618.190306-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: huge_memory: convert madvise_free_huge_pmd to use a folio

Using folios instead of pages removes several calls to compound_head(),

Link: https://lkml.kernel.org/r/20221207023431.151008-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

cma: tracing: print alloc result in trace_cma_alloc_finish

The result of the allocation attempt is not printed in
trace_cma_alloc_finish, but it's important to do it so we can set filters
to catch specific errors on allocation or to trigger some operations on
specific errors.

We have printed the result in log, but the log is conditional and could
not be filtered by tracing events.

It introduces little overhead to print this result.  The result of
allocation is named `errorno' in the trace.

Link: https://lkml.kernel.org/r/20221208142130.1501195-1-haowenchao@huawei.com
Signed-off-by: Wenchao Hao <haowenchao@huawei.com>
Cc: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Cc: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

lib/test_vmalloc.c: add parameter use_huge for fix_size_alloc_test

Add a parameter `use_huge' for fix_size_alloc_test(), which can be used to
test allocation vie vmalloc_huge for both functionality and performance.

Link: https://lkml.kernel.org/r/20221212055657.698420-1-panqinglin2020@iscas.ac.cn
Signed-off-by: Qinglin Pan <panqinglin2020@iscas.ac.cn>
Cc: "Uladzislau Rezki (Sony)" <urezki@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/mempolicy: do not duplicate policy if it is not applicable for set_mempolicy_home_node

set_mempolicy_home_node tries to duplicate a memory policy before checking
it whether it is applicable for the operation.  There is no real reason
for doing that and it might actually be a pointless memory allocation and
deallocation exercise for MPOL_INTERLEAVE.

Not a big problem but we can do better. Simply check the policy before
acting on it.

Link: https://lkml.kernel.org/r/20221216194537.238047-2-mathieu.desnoyers@efficios.com
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Feng Tang <feng.tang@intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mpage: use b_folio in do_mpage_readpage()

Remove this conversion of a folio back to a page.

Link: https://lkml.kernel.org/r/20221215214402.3522366-13-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

reiserfs: replace obvious uses of b_page with b_folio

These places just use b_page to get to the buffer's address_space or call
page_folio() on b_page to get a folio.

Link: https://lkml.kernel.org/r/20221215214402.3522366-12-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

nilfs2: replace obvious uses of b_page with b_folio

These places just use b_page to get to the buffer's address_space or the
index of the page the buffer is in.

Link: https://lkml.kernel.org/r/20221215214402.3522366-11-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

jbd2: replace obvious uses of b_page with b_folio

These places just use b_page to get to the buffer's address_space or have
already been converted to folio.

Link: https://lkml.kernel.org/r/20221215214402.3522366-10-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

gfs2: replace obvious uses of b_page with b_folio

These places just use b_page to get to the buffer's address_space.

Link: https://lkml.kernel.org/r/20221215214402.3522366-9-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

buffer: use b_folio in mark_buffer_dirty()

Removes about four calls to compound_head().  Two of them are inline which
removes 132 bytes from the kernel text.

Link: https://lkml.kernel.org/r/20221215214402.3522366-8-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

page_io: remove buffer_head include

page_io never uses buffer_heads to do I/O.

Link: https://lkml.kernel.org/r/20221215214402.3522366-7-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

buffer: use b_folio in end_buffer_async_write()

Save 76 bytes from avoiding the call to compound_head() in SetPageError().
Also avoid the call to compound_head() in end_page_writeback().

Link: https://lkml.kernel.org/r/20221215214402.3522366-6-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

buffer: use b_folio in end_buffer_async_read()

Removes a call to compound_head() in SetPageError(), saving 76 bytes of
text.

Link: https://lkml.kernel.org/r/20221215214402.3522366-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

buffer: use b_folio in touch_buffer()

Removes a call to compound_head() in this path.

Link: https://lkml.kernel.org/r/20221215214402.3522366-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

buffer: replace obvious uses of b_page with b_folio

These cases just check if it's NULL, or use b_page to get to the page's
address space.  They are assumptions that b_page never points to a tail
page.

Link: https://lkml.kernel.org/r/20221215214402.3522366-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

buffer: add b_folio as an alias of b_page

Patch series "Start converting buffer_heads to use folios".

I was hoping that filesystems would convert from buffer_heads to iomap,
but that's not happening particularly quickly.  So the buffer_head
infrastructure needs to be converted from being page-based to being
folio-based.

This patch (of 12):

Buffer heads point to the allocation (ie the folio), not the page.  This
is currently the same thing for all filesystems that use buffer heads, so
this is a safe transitional step.

Link: https://lkml.kernel.org/r/20221215214402.3522366-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20221215214402.3522366-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: introduce hugetlb_walk()

huge_pte_offset() is the main walker function for hugetlb pgtables.  The
name is not really representing what it does, though.

Instead of renaming it, introduce a wrapper function called hugetlb_walk()
which will use huge_pte_offset() inside.  Assert on the locks when walking
the pgtable.

Note, the vma lock assertion will be a no-op for private mappings.

Document the last special case in the page_vma_mapped_walk() path where we
don't need any more lock to call hugetlb_walk().

Taking vma lock there is not needed because either: (1) potential callers
of hugetlb pvmw holds i_mmap_rwsem already (from one rmap_walk()), or (2)
the caller will not walk a hugetlb vma at all so the hugetlb code path not
reachable (e.g.  in ksm or uprobe paths).

It's slightly implicit for future page_vma_mapped_walk() callers on that
lock requirement.  But anyway, when one day this rule breaks, one will get
a straightforward warning in hugetlb_walk() with lockdep, then there'll be
a way out.

[akpm@linux-foundation.org: coding-style cleanups]
Link: https://lkml.kernel.org/r/20221216155229.2043750-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: make walk_hugetlb_range() safe to pmd unshare

Since walk_hugetlb_range() walks the pgtable, it needs the vma lock to
make sure the pgtable page will not be freed concurrently.

Link: https://lkml.kernel.org/r/20221216155226.2043738-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: make follow_hugetlb_page() safe to pmd unshare

Since follow_hugetlb_page() walks the pgtable, it needs the vma lock to
make sure the pgtable page will not be freed concurrently.

Link: https://lkml.kernel.org/r/20221216155223.2043727-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: make hugetlb_follow_page_mask() safe to pmd unshare

Since hugetlb_follow_page_mask() walks the pgtable, it needs the vma lock
to make sure the pgtable page will not be freed concurrently.

Link: https://lkml.kernel.org/r/20221216155219.2043714-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: make userfaultfd_huge_must_wait() safe to pmd unshare

We can take the hugetlb walker lock, here taking vma lock directly.

Link: https://lkml.kernel.org/r/20221216155217.2043700-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: move swap entry handling into vma lock when faulted

In hugetlb_fault(), there used to have a special path to handle swap entry
at the entrance using huge_pte_offset().  That's unsafe because
huge_pte_offset() for a pmd sharable range can access freed pgtables if
without any lock to protect the pgtable from being freed after pmd
unshare.

Here the simplest solution to make it safe is to move the swap handling to
be after the vma lock being held.  We may need to take the fault mutex on
either migration or hwpoison entries now (also the vma lock, but that's
really needed), however neither of them is hot path.

Note that the vma lock cannot be released in hugetlb_fault() when the
migration entry is detected, because in migration_entry_wait_huge() the
pgtable page will be used again (by taking the pgtable lock), so that also
need to be protected by the vma lock.  Modify migration_entry_wait_huge()
so that it must be called with vma read lock held, and properly release
the lock in __migration_entry_wait_huge().

Link: https://lkml.kernel.org/r/20221216155100.2043537-5-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: document huge_pte_offset usage

huge_pte_offset() is potentially a pgtable walker, looking up pte_t* for a
hugetlb address.

Normally, it's always safe to walk a generic pgtable as long as we're with
the mmap lock held for either read or write, because that guarantees the
pgtable pages will always be valid during the process.

But it's not true for hugetlbfs, especially shared: hugetlbfs can have its
pgtable freed by pmd unsharing, it means that even with mmap lock held for
current mm, the PMD pgtable page can still go away from under us if pmd
unsharing is possible during the walk.

So we have two ways to make it safe even for a shared mapping:

  (1) If we're with the hugetlb vma lock held for either read/write, it's
      okay because pmd unshare cannot happen at all.

  (2) If we're with the i_mmap_rwsem lock held for either read/write, it's
      okay because even if pmd unshare can happen, the pgtable page cannot
      be freed from under us.

Document it.

Link: https://lkml.kernel.org/r/20221216155100.2043537-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: don't wait for migration entry during follow page

That's what the code does with !hugetlb pages, so we should logically do
the same for hugetlb, so migration entry will also be treated as no page.

This is probably also the last piece in follow_page code that may sleep,
the last one should be removed in cf994dd8af27 ("mm/gup: remove
FOLL_MIGRATION", 2022-11-16).

Link: https://lkml.kernel.org/r/20221216155100.2043537-3-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: let vma_offset_start() to return start

Patch series "mm/hugetlb: Make huge_pte_offset() thread-safe for pmd
unshare", v4.

Problem
=======

huge_pte_offset() is a major helper used by hugetlb code paths to walk a
hugetlb pgtable.  It's used mostly everywhere since that's needed even
before taking the pgtable lock.

huge_pte_offset() is always called with mmap lock held with either read or
write.  It was assumed to be safe but it's actually not.  One race
condition can easily trigger by: (1) firstly trigger pmd share on a memory
range, (2) do huge_pte_offset() on the range, then at the meantime, (3)
another thread unshare the pmd range, and the pgtable page is prone to lost
if the other shared process wants to free it completely (by either munmap
or exit mm).

The recent work from Mike on vma lock can resolve most of this already.
It's achieved by forbidden pmd unsharing during the lock being taken, so no
further risk of the pgtable page being freed.  It means if we can take the
vma lock around all huge_pte_offset() callers it'll be safe.

There're already a bunch of them that we did as per the latest mm-unstable,
but also quite a few others that we didn't for various reasons especially
on huge_pte_offset() usage.

One more thing to mention is that besides the vma lock, i_mmap_rwsem can
also be used to protect the pgtable page (along with its pgtable lock) from
being freed from under us.  IOW, huge_pte_offset() callers need to either
hold the vma lock or i_mmap_rwsem to safely walk the pgtables.

A reproducer of such problem, based on hugetlb GUP (NOTE: since the race is
very hard to trigger, one needs to apply another kernel delay patch too,
see below):

======8<=======
  #define _GNU_SOURCE
  #include <stdio.h>
  #include <stdlib.h>
  #include <errno.h>
  #include <unistd.h>
  #include <sys/mman.h>
  #include <fcntl.h>
  #include <linux/memfd.h>
  #include <assert.h>
  #include <pthread.h>

  #define  MSIZE  (1UL << 30)     /* 1GB */
  #define  PSIZE  (2UL << 20)     /* 2MB */

  #define  HOLD_SEC  (1)

  int pipefd[2];
  void *buf;

  void *do_map(int fd)
  {
      unsigned char *tmpbuf, *p;
      int ret;

      ret = posix_memalign((void **)&tmpbuf, MSIZE, MSIZE);
      if (ret) {
          perror("posix_memalign() failed");
          return NULL;
      }

      tmpbuf = mmap(tmpbuf, MSIZE, PROT_READ | PROT_WRITE,
                    MAP_SHARED | MAP_FIXED, fd, 0);
      if (tmpbuf == MAP_FAILED) {
          perror("mmap() failed");
          return NULL;
      }
      printf("mmap() -> %p\n", tmpbuf);

      for (p = tmpbuf; p < tmpbuf + MSIZE; p += PSIZE) {
          *p = 1;
      }

      return tmpbuf;
  }

  void do_unmap(void *buf)
  {
      munmap(buf, MSIZE);
  }

  void proc2(int fd)
  {
      unsigned char c;

      buf = do_map(fd);
      if (!buf)
          return;

      read(pipefd[0], &c, 1);
      /*
       * This frees the shared pgtable page, causing use-after-free in
       * proc1_thread1 when soft walking hugetlb pgtable.
       */
      do_unmap(buf);

      printf("Proc2 quitting\n");
  }

  void *proc1_thread1(void *data)
  {
      /*
       * Trigger follow-page on 1st 2m page.  Kernel hack patch needed to
       * withhold this procedure for easier reproduce.
       */
      madvise(buf, PSIZE, MADV_POPULATE_WRITE);
      printf("Proc1-thread1 quitting\n");
      return NULL;
  }

  void *proc1_thread2(void *data)
  {
      unsigned char c;

      /* Wait a while until proc1_thread1() start to wait */
      sleep(0.5);
      /* Trigger pmd unshare */
      madvise(buf, PSIZE, MADV_DONTNEED);
      /* Kick off proc2 to release the pgtable */
      write(pipefd[1], &c, 1);

      printf("Proc1-thread2 quitting\n");
      return NULL;
  }

  void proc1(int fd)
  {
      pthread_t tid1, tid2;
      int ret;

      buf = do_map(fd);
      if (!buf)
          return;

      ret = pthread_create(&tid1, NULL, proc1_thread1, NULL);
      assert(ret == 0);
      ret = pthread_create(&tid2, NULL, proc1_thread2, NULL);
      assert(ret == 0);

      /* Kick the child to share the PUD entry */
      pthread_join(tid1, NULL);
      pthread_join(tid2, NULL);

      do_unmap(buf);
  }

  int main(void)
  {
      int fd, ret;

      fd = memfd_create("test-huge", MFD_HUGETLB | MFD_HUGE_2MB);
      if (fd < 0) {
          perror("open failed");
          return -1;
      }

      ret = ftruncate(fd, MSIZE);
      if (ret) {
          perror("ftruncate() failed");
          return -1;
      }

      ret = pipe(pipefd);
      if (ret) {
          perror("pipe() failed");
          return -1;
      }

      if (fork()) {
          proc1(fd);
      } else {
          proc2(fd);
      }

      close(pipefd[0]);
      close(pipefd[1]);
      close(fd);

      return 0;
  }
======8<=======

The kernel patch needed to present such a race so it'll trigger 100%:

======8<=======
: diff --git a/mm/hugetlb.c b/mm/hugetlb.c
: index 9d97c9a2a15d..f8d99dad5004 100644
: --- a/mm/hugetlb.c
: +++ b/mm/hugetlb.c
: @@ -38,6 +38,7 @@
:  #include <asm/page.h>
:  #include <asm/pgalloc.h>
:  #include <asm/tlb.h>
: +#include <asm/delay.h>
:
:  #include <linux/io.h>
:  #include <linux/hugetlb.h>
: @@ -6290,6 +6291,7 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
:                 bool unshare = false;
:                 int absent;
:                 struct page *page;
: +               unsigned long c = 0;
:
:                 /*
:                  * If we have a pending SIGKILL, don't keep faulting pages and
: @@ -6309,6 +6311,13 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
:                  */
:                 pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
:                                       huge_page_size(h));
: +
: +               pr_info("%s: withhold 1 sec...\n", __func__);
: +               for (c = 0; c < 100; c++) {
: +                       udelay(10000);
: +               }
: +               pr_info("%s: withhold 1 sec...done\n", __func__);
: +
:                 if (pte)
:                         ptl = huge_pte_lock(h, mm, pte);
:                 absent = !pte || huge_pte_none(huge_ptep_get(pte));
: ======8<=======

It'll trigger use-after-free of the pgtable spinlock:

======8<=======
[   16.959907] follow_hugetlb_page: withhold 1 sec...
[   17.960315] follow_hugetlb_page: withhold 1 sec...done
[   17.960550] ------------[ cut here ]------------
[   17.960742] DEBUG_LOCKS_WARN_ON(1)
[   17.960756] WARNING: CPU: 3 PID: 542 at kernel/locking/lockdep.c:231 __lock_acquire+0x955/0x1fa0
[   17.961264] Modules linked in:
[   17.961394] CPU: 3 PID: 542 Comm: hugetlb-pmd-sha Not tainted 6.1.0-rc4-peterx+ #46
[   17.961704] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
[   17.962266] RIP: 0010:__lock_acquire+0x955/0x1fa0
[   17.962516] Code: c0 0f 84 5f fe ff ff 44 8b 1d 0f 9a 29 02 45 85 db 0f 85 4f fe ff ff 48 c7 c6 75 50 83 82 48 c7 c7 1b 4b 7d 82 e8 d3 22 d8 00 <0f> 0b 31 c0 4c 8b 54 24 08 4c 8b 04 24 e9
[   17.963494] RSP: 0018:ffffc90000e4fba8 EFLAGS: 00010096
[   17.963704] RAX: 0000000000000016 RBX: fffffffffd3925a8 RCX: 0000000000000000
[   17.963989] RDX: 0000000000000002 RSI: ffffffff82863ccf RDI: 00000000ffffffff
[   17.964276] RBP: 0000000000000000 R08: 0000000000000000 R09: ffffc90000e4fa58
[   17.964557] R10: 0000000000000003 R11: ffffffff83162688 R12: 0000000000000000
[   17.964839] R13: 0000000000000001 R14: ffff888105eac748 R15: 0000000000000001
[   17.965123] FS:  00007f17c0a00640(0000) GS:ffff888277cc0000(0000) knlGS:0000000000000000
[   17.965443] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   17.965672] CR2: 00007f17c09ffef8 CR3: 000000010c87a005 CR4: 0000000000770ee0
[   17.965956] PKRU: 55555554
[   17.966068] Call Trace:
[   17.966172]  <TASK>
[   17.966268]  ? tick_nohz_tick_stopped+0x12/0x30
[   17.966455]  lock_acquire+0xbf/0x2b0
[   17.966603]  ? follow_hugetlb_page.cold+0x75/0x5c4
[   17.966799]  ? _printk+0x48/0x4e
[   17.966934]  _raw_spin_lock+0x2f/0x40
[   17.967087]  ? follow_hugetlb_page.cold+0x75/0x5c4
[   17.967285]  follow_hugetlb_page.cold+0x75/0x5c4
[   17.967473]  __get_user_pages+0xbb/0x620
[   17.967635]  faultin_vma_page_range+0x9a/0x100
[   17.967817]  madvise_vma_behavior+0x3c0/0xbd0
[   17.967998]  ? mas_prev+0x11/0x290
[   17.968141]  ? find_vma_prev+0x5e/0xa0
[   17.968304]  ? madvise_vma_anon_name+0x70/0x70
[   17.968486]  madvise_walk_vmas+0xa9/0x120
[   17.968650]  do_madvise.part.0+0xfa/0x270
[   17.968813]  __x64_sys_madvise+0x5a/0x70
[   17.968974]  do_syscall_64+0x37/0x90
[   17.969123]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
[   17.969329] RIP: 0033:0x7f1840f0efdb
[   17.969477] Code: c3 66 0f 1f 44 00 00 48 8b 15 39 6e 0e 00 f7 d8 64 89 02 b8 ff ff ff ff eb bc 0f 1f 44 00 00 f3 0f 1e fa b8 1c 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 0d 68
[   17.970205] RSP: 002b:00007f17c09ffe38 EFLAGS: 00000202 ORIG_RAX: 000000000000001c
[   17.970504] RAX: ffffffffffffffda RBX: 00007f17c0a00640 RCX: 00007f1840f0efdb
[   17.970786] RDX: 0000000000000017 RSI: 0000000000200000 RDI: 00007f1800000000
[   17.971068] RBP: 00007f17c09ffe50 R08: 0000000000000000 R09: 00007ffd3954164f
[   17.971353] R10: 00007f1840e10348 R11: 0000000000000202 R12: ffffffffffffff80
[   17.971709] R13: 0000000000000000 R14: 00007ffd39541550 R15: 00007f17c0200000
[   17.972083]  </TASK>
[   17.972199] irq event stamp: 2353
[   17.972372] hardirqs last  enabled at (2353): [<ffffffff8117fe4e>] __up_console_sem+0x5e/0x70
[   17.972869] hardirqs last disabled at (2352): [<ffffffff8117fe33>] __up_console_sem+0x43/0x70
[   17.973365] softirqs last  enabled at (2330): [<ffffffff810f763d>] __irq_exit_rcu+0xed/0x160
[   17.973857] softirqs last disabled at (2323): [<ffffffff810f763d>] __irq_exit_rcu+0xed/0x160
[   17.974341] ---[ end trace 0000000000000000 ]---
[   17.974614] BUG: kernel NULL pointer dereference, address: 00000000000000b8
[   17.975012] #PF: supervisor read access in kernel mode
[   17.975314] #PF: error_code(0x0000) - not-present page
[   17.975615] PGD 103f7b067 P4D 103f7b067 PUD 106cd7067 PMD 0
[   17.975943] Oops: 0000 [#1] PREEMPT SMP NOPTI
[   17.976197] CPU: 3 PID: 542 Comm: hugetlb-pmd-sha Tainted: G        W          6.1.0-rc4-peterx+ #46
[   17.976712] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
[   17.977370] RIP: 0010:__lock_acquire+0x190/0x1fa0
[   17.977655] Code: 98 00 00 00 41 89 46 24 81 e2 ff 1f 00 00 48 0f a3 15 e4 ba dd 02 0f 83 ff 05 00 00 48 8d 04 52 48 c1 e0 06 48 05 c0 d2 f4 83 <44> 0f b6 a0 b8 00 00 00 41 0f b7 46 20 6f
[   17.979170] RSP: 0018:ffffc90000e4fba8 EFLAGS: 00010046
[   17.979787] RAX: 0000000000000000 RBX: fffffffffd3925a8 RCX: 0000000000000000
[   17.980838] RDX: 0000000000000002 RSI: ffffffff82863ccf RDI: 00000000ffffffff
[   17.982048] RBP: 0000000000000000 R08: ffff888105eac720 R09: ffffc90000e4fa58
[   17.982892] R10: ffff888105eab900 R11: ffffffff83162688 R12: 0000000000000000
[   17.983771] R13: 0000000000000001 R14: ffff888105eac748 R15: 0000000000000001
[   17.984815] FS:  00007f17c0a00640(0000) GS:ffff888277cc0000(0000) knlGS:0000000000000000
[   17.985924] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   17.986265] CR2: 00000000000000b8 CR3: 000000010c87a005 CR4: 0000000000770ee0
[   17.986674] PKRU: 55555554
[   17.986832] Call Trace:
[   17.987012]  <TASK>
[   17.987266]  ? tick_nohz_tick_stopped+0x12/0x30
[   17.987770]  lock_acquire+0xbf/0x2b0
[   17.988118]  ? follow_hugetlb_page.cold+0x75/0x5c4
[   17.988575]  ? _printk+0x48/0x4e
[   17.988889]  _raw_spin_lock+0x2f/0x40
[   17.989243]  ? follow_hugetlb_page.cold+0x75/0x5c4
[   17.989687]  follow_hugetlb_page.cold+0x75/0x5c4
[   17.990119]  __get_user_pages+0xbb/0x620
[   17.990500]  faultin_vma_page_range+0x9a/0x100
[   17.990928]  madvise_vma_behavior+0x3c0/0xbd0
[   17.991354]  ? mas_prev+0x11/0x290
[   17.991678]  ? find_vma_prev+0x5e/0xa0
[   17.992024]  ? madvise_vma_anon_name+0x70/0x70
[   17.992421]  madvise_walk_vmas+0xa9/0x120
[   17.992793]  do_madvise.part.0+0xfa/0x270
[   17.993166]  __x64_sys_madvise+0x5a/0x70
[   17.993539]  do_syscall_64+0x37/0x90
[   17.993879]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
======8<=======

Resolution
==========

This patchset protects all the huge_pte_offset() callers to also take the
vma lock properly.

Patch Layout
============

Patch 1-2:         cleanup, or dependency of the follow up patches
Patch 3:           before fixing, document huge_pte_offset() on lock required
Patch 4-8:         each patch resolves one possible race condition
Patch 9:           introduce hugetlb_walk() to replace huge_pte_offset()

Tests
=====

The series is verified with the above reproducer so the race cannot
trigger anymore.  It also passes all hugetlb kselftests.

This patch (of 9):

Even though vma_offset_start() is named like that, it's not returning "the
start address of the range" but rather the offset we should use to offset
the vma->vm_start address.

Make it return the real value of the start vaddr, and it also helps for
all the callers because whenever the retval is used, it'll be ultimately
added into the vma->vm_start anyway, so it's better.

Link: https://lkml.kernel.org/r/20221216155100.2043537-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20221216155100.2043537-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb: update vma flag check for hugetlb vma lock

The check for whether a hugetlb vma lock exists partially depends on the
vma's flags.  Currently, it checks for either VM_MAYSHARE or VM_SHARED.
The reason both flags are used is because VM_MAYSHARE was previously
cleared in hugetlb vmas as they are tore down.  This is no longer the
case, and only the VM_MAYSHARE check is required.

Link: https://lkml.kernel.org/r/20221212235042.178355-2-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Naoya Horiguchi <naoya.horiguchi@linux.dev>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/memfd: add tests for MFD_NOEXEC_SEAL MFD_EXEC

Tests to verify MFD_NOEXEC, MFD_EXEC and vm.memfd_noexec sysctl.

Link: https://lkml.kernel.org/r/20221215001205.51969-6-jeffxu@google.com
Signed-off-by: Jeff Xu <jeffxu@google.com>
Co-developed-by: Daniel Verkamp <dverkamp@chromium.org>
Signed-off-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Jorge Lucangeli Obes <jorgelo@chromium.org>
Cc: kernel test robot <lkp@intel.com>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>