btrfs_repair_io_failure() writes a repaired data sector or tree block
back to the bad copy unless the filesystem is read-only. SB_RDONLY is
not the only state that must stop that write:
- A transaction abort sets BTRFS_FS_STATE_TRANS_ABORTED, reports the
abort (with a stack dump for most errors), and then
__btrfs_handle_fs_error() sets fs_info->fs_error and only after that
SB_RDONLY. A repair that runs in between still writes after the
abort.
- An error during mount, before SB_BORN, sets fs_error without
SB_RDONLY.
Nothing should be written after an abort. Add
btrfs_can_submit_repair_write(), which also checks the error and the
abort bit, and use it in btrfs_repair_io_failure(). That covers data
read repair and tree block repair (btrfs_repair_eb_io_failure()). The
read still gets the good copy's data.
Fixes: 8ceaad6cd6e7 ("btrfs: do not ASSERT() when the fs flips RO inside btrfs_repair_io_failure()")
Assisted-by: Claude:claude-opus-5-5
* The fs either mounted RO or hit critical errors, no need
* to continue repairing.
*/
- if (unlikely(sb_rdonly(fs_info->sb)))
+ if (unlikely(!btrfs_can_submit_repair_write(fs_info)))
return 0;
if (btrfs_repair_one_zone(fs_info, logical))
(unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR, \
&(fs_info)->fs_state)))
+/*
+ * Return true if a repair write (read repair, scrub) may be submitted: never on
+ * a read-only filesystem, nor after a transaction abort or another
+ * filesystem error. A transaction abort sets BTRFS_FS_STATE_TRANS_ABORTED,
+ * then the error, and only then SB_RDONLY, so checking SB_RDONLY alone is not
+ * enough.
+ */
+static inline bool btrfs_can_submit_repair_write(const struct btrfs_fs_info *fs_info)
+{
+ return !sb_rdonly(fs_info->sb) && !BTRFS_FS_ERROR(fs_info) &&
+ !test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state);
+}
+
/*
* We use folio flag owner_2 to indicate there is an ordered extent with
* unfinished IO.