public inbox for linux-btrfs@vger.kernel.org
 help / color / mirror / Atom feed
From: ZhengYuan Huang <gality369@gmail.com>
To: dsterba@suse.com, clm@fb.com
Cc: wqu@suse.com, linux-btrfs@vger.kernel.org,
	linux-kernel@vger.kernel.org, baijiaju1990@gmail.com,
	r33s3n6@gmail.com, zzzccc427@gmail.com,
	ZhengYuan Huang <gality369@gmail.com>
Subject: [PATCH v2 0/2] btrfs: verify cached extent buffers against tree parent checks
Date: Fri, 13 Mar 2026 17:19:22 +0800	[thread overview]
Message-ID: <20260313091924.570554-1-gality369@gmail.com> (raw)

This series fixes a btrfs crash caused by reusing a cached extent buffer
without re-running the caller supplied tree-parent verification.

The problem happens when a tree block is first read and validated with one
expected level, then later looked up again through a path that derives a
different expected level from corrupted metadata. If the extent buffer is
already marked EXTENT_BUFFER_UPTODATE, the cached-hit path returns it
without re-validating the supplied btrfs_tree_parent_check. This can allow
an inconsistent btrfs_root to be constructed and later lead to a
null-ptr-deref during backref walking.

Patch 1/2 is a preparatory change that extends
btrfs_buffer_uptodate() to support tree-parent verification on cached
buffers. Patch 2/2 uses that support on the cached-hit path and contains
the actual fix.

Together, these changes make cache hits and fresh reads follow the same
tree-parent verification rules, turning the corruption into a read failure
instead of constructing an inconsistent root object and crashing later.

For reference, a more detailed analysis of the trigger path is available at:
https://lore.kernel.org/all/CAOmEq9U14a=pwN_dw2M70gfujhMKki434cfmegoxcyUpkYs5bQ@mail.gmail.com/

Changes since v1:
  - drop the adhoc root-specific consistency check in read_tree_root_path()
  - move the validation into the cached-hit path as suggested by Qu Wenruo
  - extend btrfs_buffer_uptodate() with an optional tree-parent check
  - make read_tree_root_path() pass its check when validating a cached root

ZhengYuan Huang (2):
  btrfs: add tree parent check to btrfs_buffer_uptodate()
  btrfs: revalidate cached tree blocks on the uptodate path

 fs/btrfs/ctree.c       |  2 +-
 fs/btrfs/disk-io.c     | 18 ++++++++++++++----
 fs/btrfs/disk-io.h     |  3 ++-
 fs/btrfs/extent-tree.c |  2 +-
 fs/btrfs/extent_io.c   | 12 ++++++++++--
 fs/btrfs/tree-log.c    |  2 +-
 6 files changed, 29 insertions(+), 10 deletions(-)

-- 
2.43.0


             reply	other threads:[~2026-03-13  9:19 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-13  9:19 ZhengYuan Huang [this message]
2026-03-13  9:19 ` [PATCH v2 1/2] btrfs: add tree parent check to btrfs_buffer_uptodate() ZhengYuan Huang
2026-03-13  9:19 ` [PATCH v2 2/2] btrfs: revalidate cached tree blocks on the uptodate path ZhengYuan Huang
2026-03-15 21:17   ` Qu Wenruo
2026-03-13 23:49 ` [PATCH v2 0/2] btrfs: verify cached extent buffers against tree parent checks Qu Wenruo

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=20260313091924.570554-1-gality369@gmail.com \
    --to=gality369@gmail.com \
    --cc=baijiaju1990@gmail.com \
    --cc=clm@fb.com \
    --cc=dsterba@suse.com \
    --cc=linux-btrfs@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=r33s3n6@gmail.com \
    --cc=wqu@suse.com \
    --cc=zzzccc427@gmail.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