CVE-2026-90261: btrfs: zoned: flush active metadata block group at btree_writepages() start
In the Linux kernel, the following vulnerability has been resolved:
btrfs: zoned: flush active metadata block group at btreewritepages() start
btreewritepages() writes the btree inode's dirty metadata in ascending logical address order. On a zoned filesystem only one metadata and one system block group is active for writing at a time, and checkbgisactive() (via btrfscheckmetawritepointer()) pivots the active block group as writeback moves from one block group to the next.
If the active block group sits at a higher logical address than another block group that also holds dirty metadata, the ascending walk reaches the lower one first and, to write it, has to finish the active block group and activate the lower one. It cannot finish a block group that still has unsent IO, and during WBSYNCALL && !forsync (commit) writeback it deliberately refuses to wait for that IO under fsinfo->zonedmetaiolock, as that can deadlock. The pivot thus cannot issue the submission itself either, so it gives up: btrfscheckmetawritepointer() returns -EAGAIN, which btrfswriteandwaittransaction() treats as fatal and aborts the transaction, forcing the filesystem read-only. This happens intermittently under metadata-heavy relocation (e.g. fstests btrfs/187).
Flush the active metadata and system block groups at the start of btreewritepages(), under the fsinfo->zonedmetaiolock it already holds, so they have no unsent IO left and the later pivot can finish them and make forward progress.
Affected Software
Event History
Frequently Asked Questions
Which systems are exposed to this failure mode?
The issue applies to Btrfs filesystems operating in zoned mode. It depends on metadata writeback spanning block groups whose logical-address ordering conflicts with the currently active metadata block group.
What conditions trigger the filesystem becoming read-only?
During commit writeback using WB_SYNC_ALL without for_sync, writeback can need to switch from a higher-address active metadata block group to a lower-address block group containing dirty metadata. If the active group still has unsent I/O, the switch returns -EAGAIN; transaction writeback treats that as fatal, aborts the transaction, and forces the filesystem read-only.
How can administrators recognize that the issue has occurred?
The observable outcome described is a fatal transaction abort followed by the filesystem being forced into read-only mode. The failure is associated with btrfs_check_meta_write_pointer() returning -EAGAIN during transaction writeback.