From: sashiko-bot@kernel.org
To: "Pedro Falcato" <pfalcato@suse.de>,
"David Hildenbrand" <david@kernel.org>,
"Andrew Morton" <akpm@linux-foundation.org>,
"Zi Yan" <ziy@nvidia.com>,
"Matthew Wilcox (Oracle)" <willy@infradead.org>,
"Jan Kara" <jack@suse.cz>, "Lorenzo Stoakes" <ljs@kernel.org>,
"John Hubbard" <jhubbard@nvidia.com>,
"Peter Xu" <peterx@redhat.com>,
"Johannes Weiner" <hannes@cmpxchg.org>
Cc: linux-ext4@vger.kernel.org,
"Liam R. Howlett" <liam@infradead.org>,
"Vlastimil Babka" <vbabka@kernel.org>,
"Mike Rapoport" <rppt@kernel.org>,
"Suren Baghdasaryan" <surenb@google.com>,
"Michal Hocko" <mhocko@suse.com>,
"Baolin Wang" <baolin.wang@linux.alibaba.com>,
"Nico Pache" <nico.pache@linux.dev>,
"Ryan Roberts" <ryan.roberts@arm.com>,
"Dev Jain" <dev.jain@arm.com>, "Barry Song" <baohua@kernel.org>,
"Lance Yang" <lance.yang@linux.dev>,
"Usama Arif" <usama.arif@linux.dev>,
"Kiryl Shutsemau" <kas@kernel.org>,
linux-mm@kvack.org, "Kairui Song" <kasong@tencent.com>,
"Qi Zheng" <qi.zheng@linux.dev>,
"Shakeel Butt" <shakeel.butt@linux.dev>,
"Axel Rasmussen" <axelrasmussen@google.com>,
"Yuanchu Xie" <yuanchu@google.com>, "Wei Xu" <weixugc@google.com>,
"Baoquan He" <baoquan.he@linux.dev>,
"Michal Hocko" <mhocko@kernel.org>,
tytso@mit.edu, "Chuck Lever" <cel@kernel.org>,
"Jeff Layton" <jlayton@kernel.org>,
"Anna Schumaker" <anna@kernel.org>
Subject: Re: [PATCH RFC 5/5] mm/filemap: refuse releasing maybe_dma_pinned folios
Date: Fri, 20 Nov 2026 20:11:46 +0000 [thread overview]
Message-ID: <sashiko-outbox-226011@kernel.org> (raw)
In-Reply-To: <20261119-gup-dirty-what-v1-5-236d69f25c24@suse.de>
Thank you for your contribution! Sashiko AI review found 3 potential
issues to consider:
- [High] mm/filemap: filemap_release_folio() dma-pin check breaks
folio_invalidate()
- [Medium] mm/filemap: VM_WARN_ON_FOLIO on mapped folio hit via
mext_move_extent()
- [Medium] mm/filemap: VM_WARN_ON_FOLIO in filemap_release_folio()
precedes needs_release
--
commit 38a290b76cc80ed81b7d5a26e8b50ef48dfba24c
Author: Pedro Falcato <pfalcato@suse.de>
mm/filemap: refuse releasing maybe_dma_pinned folios
This commit updates filemap_release_folio() to warn if the folio is mapped
and refuse to release filesystem metadata when folio_maybe_dma_pinned() is
true.
> diff --git a/mm/filemap.c b/mm/filemap.c
> index f9e34144c53c7..4fa75663b7f3e 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -4571,8 +4571,11 @@ bool filemap_release_folio(struct folio *folio, gfp_t gfp)
> struct address_space * const mapping = folio->mapping;
>
> BUG_ON(!folio_test_locked(folio));
> + VM_WARN_ON_FOLIO(folio_mapped(folio), folio);
[Severity: Medium]
Can EXT4_IOC_MOVE_EXT hit this VM_WARN_ON_FOLIO() on a memory-mapped file?
In fs/ext4/move_extent.c, mext_move_extent() locks the origin and donor
folios and calls filemap_release_folio() on them without unmapping them
from user page tables or checking folio_mapped() first:
fs/ext4/move_extent.c:mext_move_extent() {
...
if (!filemap_release_folio(folio[0], 0) ||
!filemap_release_folio(folio[1], 0)) {
ret = -EBUSY;
goto unlock;
}
...
}
If either file is memory-mapped with resident PTEs in the moved range, an
unprivileged EXT4_IOC_MOVE_EXT ioctl triggers this warning when
CONFIG_DEBUG_VM is enabled.
> if (!folio_needs_release(folio))
> return true;
[Severity: Medium]
Should VM_WARN_ON_FOLIO(folio_mapped(folio), folio) come after the
!folio_needs_release(folio) check to match filemap_may_release_folio()?
filemap_may_release_folio() returns true when !folio_needs_release(folio)
before checking folio_mapped(folio):
mm/internal.h:filemap_may_release_folio() {
VM_WARN_ON_FOLIO(!folio_test_locked(folio), folio);
if (!folio_needs_release(folio))
return true;
/* There's no way to know whether these are write-mapped, so give up. */
if (folio_mapped(folio))
return false;
...
}
Callers such as collapse_file() (which calls try_to_unmap() without
TTU_SYNC) and __folio_freeze_split_file() (which calls unmap_folio())
can leave a folio mapped if unmapping races with zap_pte_range(), then
rely on filemap_may_release_folio() to gate filemap_release_folio().
For a folio that does not need release, filemap_may_release_folio()
returns true and filemap_release_folio() then triggers
VM_WARN_ON_FOLIO(folio_mapped(folio), folio).
> + if (folio_maybe_dma_pinned(folio))
> + return false;
> if (folio_test_writeback(folio))
> return false;
[Severity: High]
Does refusing to release a DMA-pinned folio here break truncation via
folio_invalidate() and leak the folio and its buffer_heads?
During full-folio truncation, truncate_cleanup_folio() calls
folio_invalidate() before removing the folio from the page cache.
block_invalidate_folio() and gfs2_invalidate_folio() first discard all
buffers on the folio and then call filemap_release_folio(folio, 0) to
free the buffer_heads and clear PG_private.
If filemap_release_folio() returns false because the folio is DMA-pinned,
truncation still removes the folio from the page cache and sets
folio->mapping = NULL, leaving PG_private and the discarded buffer_heads
attached. When the DMA pin holder later unpins the folio as dirty (for
example via unpin_user_pages_dirty_lock()), folio_mark_dirty() sees
folio_mapping(folio) == NULL and calls noop_dirty_folio(), setting
PG_dirty on the orphaned folio while PG_private holds a reference. When
shrink_folio_list() later scans this folio on the file LRU, it sees
folio_test_dirty(folio) and jumps to activate_locked before reaching the
orphaned buffer release path, and because folio->mapping is NULL nothing
can ever write it back or clear PG_dirty, permanently leaking the folio
and its buffer_heads on the file LRU.
On GFS2, gfs2_invalidate_folio() -> gfs2_discard() also leaves struct
gfs2_bufdata attached at bh->b_private to be freed by gfs2_release_folio();
once truncation clears folio->mapping, any later filemap_release_folio()
call sees mapping == NULL and falls back to try_to_free_buffers() instead
of mapping->a_ops->release_folio, permanently leaking struct gfs2_bufdata
even if the folio was not dirtied. In ext4's mpage_release_unused_pages(),
a failed delayed allocation calls block_invalidate_folio() and then
folio_clear_uptodate(folio), relying on filemap_release_folio() ->
try_to_free_buffers() -> folio_cancel_dirty() to clear PG_dirty, so
returning false leaves a pinned folio in the page cache with PG_dirty set
and PG_uptodate clear.
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261119-gup-dirty-what-v1-0-236d69f25c24@suse.de?part=5
next prev parent reply other threads:[~2026-11-20 20:11 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-19 20:10 [PATCH RFC 0/5] mm: GUP dirty + ->release_folio() fixes Pedro Falcato
2026-11-19 20:10 ` [PATCH RFC 1/5] mm/khugepaged: fix filemap_release_folio() interaction with GUP dirty Pedro Falcato
2026-11-19 20:10 ` [PATCH RFC 2/5] mm/huge_memory: " Pedro Falcato
2026-11-20 20:11 ` sashiko-bot
2026-11-19 20:10 ` [PATCH RFC 3/5] mm/vmscan: avoid releasing FS metadata for !FOLL_PIN gup folios Pedro Falcato
2026-11-19 20:10 ` [PATCH RFC 4/5] mm/vmscan: be more careful when releasing FS metadata if buffer_heads_over_limit Pedro Falcato
2026-11-19 20:10 ` [PATCH RFC 5/5] mm/filemap: refuse releasing maybe_dma_pinned folios Pedro Falcato
2026-11-20 20:11 ` sashiko-bot [this message]
2026-11-20 0:55 ` [PATCH RFC 0/5] mm: GUP dirty + ->release_folio() fixes John Hubbard
2026-11-20 11:57 ` Jan Kara
2026-11-20 17:39 ` Christoph Hellwig
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=sashiko-outbox-226011@kernel.org \
--to=sashiko-bot@kernel.org \
--cc=akpm@linux-foundation.org \
--cc=anna@kernel.org \
--cc=axelrasmussen@google.com \
--cc=baohua@kernel.org \
--cc=baolin.wang@linux.alibaba.com \
--cc=baoquan.he@linux.dev \
--cc=cel@kernel.org \
--cc=david@kernel.org \
--cc=dev.jain@arm.com \
--cc=hannes@cmpxchg.org \
--cc=jack@suse.cz \
--cc=jhubbard@nvidia.com \
--cc=jlayton@kernel.org \
--cc=kas@kernel.org \
--cc=kasong@tencent.com \
--cc=lance.yang@linux.dev \
--cc=liam@infradead.org \
--cc=linux-ext4@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=ljs@kernel.org \
--cc=mhocko@kernel.org \
--cc=mhocko@suse.com \
--cc=nico.pache@linux.dev \
--cc=peterx@redhat.com \
--cc=pfalcato@suse.de \
--cc=qi.zheng@linux.dev \
--cc=rppt@kernel.org \
--cc=ryan.roberts@arm.com \
--cc=sashiko-reviews@lists.linux.dev \
--cc=shakeel.butt@linux.dev \
--cc=surenb@google.com \
--cc=tytso@mit.edu \
--cc=usama.arif@linux.dev \
--cc=vbabka@kernel.org \
--cc=weixugc@google.com \
--cc=willy@infradead.org \
--cc=yuanchu@google.com \
--cc=ziy@nvidia.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox