[PATCH v3 0/2] mm, swap: don't spin on a bad swap entry

Breno Leitao posted 2 patches 1 month, 1 week ago
mm/memory.c      |  9 ++++++---
mm/mincore.c     |  2 +-
mm/shmem.c       |  2 +-
mm/swap_state.c  |  4 ++--
mm/swapfile.c    | 14 +++++++++-----
mm/userfaultfd.c |  4 ++--
mm/zswap.c       |  2 +-
7 files changed, 22 insertions(+), 15 deletions(-)
[PATCH v3 0/2] mm, swap: don't spin on a bad swap entry
Posted by Breno Leitao 1 month, 1 week ago
I've seen some machines at Meta fleet that show the following type of
problem:

1) It gets some weird warning:

  BUG: Bad page map in process khugepaged  pte:f000eef300000017 pmd:00000067
  addr:00007f57c0a01000 vm_flags:20200073 anon_vma:ffff88829af7c340 mapping:0000000000000000 index:7f57c0a01

The corruption is most likely the collapse/PT_RECLAIM race fixed by
commit 366a4532d96f ("mm: fix the race between collapse and PT_RECLAIM
under per-vma lock"). But this series is not about this one.

2) Then the fault never makes progress. do_swap_page() returns 0 when
   get_swap_device() fails, so the fault is retried, reads the same
   entry and faults again. Nothing in the round trip changes the PTE,
   and the same line comes out on every pass:

  get_swap_device: Bad swap offset entry 3ffffffc043c5

Patch 1 makes get_swap_device() return ERR_PTR(-EIO) for a malformed
entry, keeping NULL for a device swapoff is taking away, and converts
the callers. No functional change expected.

Patch 2 uses that to return VM_FAULT_SIGBUS instead of retrying.

The rate limiting patch that used to open this series was split out and
posted on its own as a backportable hotfix [1], per Andrew's request. It
should land first: patch 1 here touches the lines next to it in
get_swap_device(). This patch will probably conflict with [1], but the
merge should be trivial, given the only change in [1] is the
addition of the __ratelimited() suffix. 

	pr_err("%s: %s%08lx\n", __func__, Bad_file, entry.val);
	pr_err_ratelimited("%s: %s%08lx\n", __func__, Bad_file, entry.val);

I am happy to respin, in case you prefer

[1] https://patch.msgid.link/20260818-swap_part_one-v1-1-a4fc58119fc0@debian.org

---
Changes in v3:
- Split the rate limiting patch out and posted it separately as a
  hotfix, with a Fixes: tag and cc:stable (Andrew)
- Drop the si = NULL dance in do_swap_page() and move_pages_ptes(), and
  let the paths that drop the reference skip an error pointer instead
  (Barry)
- Link to v2: https://patch.msgid.link/20260813-swap-v2-0-4a625ccabdae@debian.org

Changes in v2:
- Rate limit swap_dup_entry_direct()'s print too (Andrew)
- Drop "in get_swap_device()" from patch 1's subject, it now covers all
  three prints
- Return ERR_PTR(-EIO) rather than ERR_PTR(-EINVAL) for a malformed
  entry; -EINVAL is too soft for a corrupt page table (David)
- Document the malformed entry case in get_swap_device()'s kerneldoc,
  in patch 2 instead of patch 3 (David)
- Reword patch 2's changelog, "an entry that can never name a slot on
  any device" was unclear (David)
- Link to v1: https://patch.msgid.link/20260810-swap-v1-0-375ef0767206@debian.org

To: Andrew Morton <akpm@linux-foundation.org>
To: David Hildenbrand <david@kernel.org>
To: Lorenzo Stoakes <ljs@kernel.org>
To: "Liam R. Howlett" <liam@infradead.org>
To: Vlastimil Babka <vbabka@kernel.org>
To: Mike Rapoport <rppt@kernel.org>
To: Suren Baghdasaryan <surenb@google.com>
To: Michal Hocko <mhocko@suse.com>
To: Jann Horn <jannh@google.com>
To: Pedro Falcato <pfalcato@suse.de>
To: Hugh Dickins <hughd@google.com>
To: Baolin Wang <baolin.wang@linux.alibaba.com>
To: Chris Li <chrisl@kernel.org>
To: Kairui Song <kasong@tencent.com>
To: Kemeng Shi <shikemeng@huaweicloud.com>
To: Nhat Pham <nphamcs@gmail.com>
To: Baoquan He <baoquan.he@linux.dev>
To: Barry Song <baohua@kernel.org>
To: Youngjun Park <youngjun.park@lge.com>
To: Peter Xu <peterx@redhat.com>
To: Johannes Weiner <hannes@cmpxchg.org>
To: Yosry Ahmed <yosry@kernel.org>
To: Chengming Zhou <chengming.zhou@linux.dev>
Cc: linux-mm@kvack.org
Cc: linux-kernel@vger.kernel.org

---
Breno Leitao (2):
      mm, swap: distinguish a malformed swap entry from a dying device
      mm: fail the fault on a malformed swap entry instead of retrying it

 mm/memory.c      |  9 ++++++---
 mm/mincore.c     |  2 +-
 mm/shmem.c       |  2 +-
 mm/swap_state.c  |  4 ++--
 mm/swapfile.c    | 14 +++++++++-----
 mm/userfaultfd.c |  4 ++--
 mm/zswap.c       |  2 +-
 7 files changed, 22 insertions(+), 15 deletions(-)
---
base-commit: 6b8c8af514d739d0335f5579b585e02babe8a727
change-id: 20260810-swap-25420f9c8ba9

Best regards,
--  
Breno Leitao <leitao@debian.org>
Re: [PATCH v3 0/2] mm, swap: don't spin on a bad swap entry
Posted by Andrew Morton 4 weeks, 1 day ago
On Tue, 18 Aug 2026 03:06:22 -0700 Breno Leitao <leitao@debian.org> wrote:

> I've seen some machines at Meta fleet that show the following type of
> problem:
> 
> 1) It gets some weird warning:
> 
>   BUG: Bad page map in process khugepaged  pte:f000eef300000017 pmd:00000067
>   addr:00007f57c0a01000 vm_flags:20200073 anon_vma:ffff88829af7c340 mapping:0000000000000000 index:7f57c0a01
> 
> The corruption is most likely the collapse/PT_RECLAIM race fixed by
> commit 366a4532d96f ("mm: fix the race between collapse and PT_RECLAIM
> under per-vma lock"). But this series is not about this one.
> 
> 2) Then the fault never makes progress. do_swap_page() returns 0 when
>    get_swap_device() fails, so the fault is retried, reads the same
>    entry and faults again. Nothing in the round trip changes the PTE,
>    and the same line comes out on every pass:
> 
>   get_swap_device: Bad swap offset entry 3ffffffc043c5
> 
> Patch 1 makes get_swap_device() return ERR_PTR(-EIO) for a malformed
> entry, keeping NULL for a device swapoff is taking away, and converts
> the callers. No functional change expected.
> 
> Patch 2 uses that to return VM_FAULT_SIGBUS instead of retrying.
> 
> The rate limiting patch that used to open this series was split out and
> posted on its own as a backportable hotfix [1], per Andrew's request. It
> should land first: patch 1 here touches the lines next to it in
> get_swap_device(). This patch will probably conflict with [1], but the
> merge should be trivial, given the only change in [1] is the
> addition of the __ratelimited() suffix. 
> 
> 	pr_err("%s: %s%08lx\n", __func__, Bad_file, entry.val);
> 	pr_err_ratelimited("%s: %s%08lx\n", __func__, Bad_file, entry.val);
> 

Thanks.  The patches have bitrotted a little - get_swap_device() was an
easy fixup but please double-check that I didn't miss anything.

Or perhaps just refresh-retest-resend if there's any doubt.


> The rate limiting patch that used to open this series was split out and
> posted on its own as a backportable hotfix [1], per Andrew's request. It
> should land first: patch 1 here touches the lines next to it in
> get_swap_device(). This patch will probably conflict with [1], but the
> merge should be trivial, given the only change in [1] is the
> addition of the __ratelimited() suffix. 
> 
> 	pr_err("%s: %s%08lx\n", __func__, Bad_file, entry.val);
> 		pr_err_ratelimited("%s: %s%08lx\n", __func__, Bad_file, entry.val);

oh, you already said that.

As usual when inspecting our error-path code, Sashiko said "you all suck":

	https://sashiko.dev/#/patchset/20260818-swap-v3-0-d3fa52598a59@debian.org

These things do seem on-topic for the changes you're proposing here, so
please take a look?