* [f2fs-dev] [PATCH 2/2] f2fs: fix to detect recoverable inode during dryrun of find_fsync_dnodes()
2025-11-05 6:50 [f2fs-dev] [PATCH 1/2] f2fs: fix return value of f2fs_recover_fsync_data() Chao Yu via Linux-f2fs-devel
@ 2025-11-05 6:50 ` Chao Yu via Linux-f2fs-devel
0 siblings, 0 replies; 2+ messages in thread
From: Chao Yu via Linux-f2fs-devel @ 2025-11-05 6:50 UTC (permalink / raw)
To: jaegeuk; +Cc: stable, linux-kernel, linux-f2fs-devel
mkfs.f2fs -f /dev/vdd
mount /dev/vdd /mnt/f2fs
touch /mnt/f2fs/foo
sync # avoid CP_UMOUNT_FLAG in last f2fs_checkpoint.ckpt_flags
touch /mnt/f2fs/bar
f2fs_io fsync /mnt/f2fs/bar
f2fs_io shutdown 2 /mnt/f2fs
umount /mnt/f2fs
blockdev --setro /dev/vdd
mount /dev/vdd /mnt/f2fs
mount: /mnt/f2fs: WARNING: source write-protected, mounted read-only.
For the case if we create and fsync a new inode before sudden power-cut,
without norecovery or disable_roll_forward mount option, the following
mount will succeed w/o recovering last fsynced inode.
The problem here is that we only check inode_list list after
find_fsync_dnodes() in f2fs_recover_fsync_data() to find out whether
there is recoverable data in the iamge, but there is a missed case, if
last fsynced inode is not existing in last checkpoint, then, we will
fail to get its inode due to nat of inode node is not existing in last
checkpoint, so the inode won't be linked in inode_list.
Let's detect such case in dyrun mode to fix this issue.
After this change, mount will fail as expected below:
mount: /mnt/f2fs: cannot mount /dev/vdd read-only.
dmesg(1) may have more information after failed mount system call.
demsg:
F2FS-fs (vdd): Need to recover fsync data, but write access unavailable, please try mount w/ disable_roll_forward or norecovery
Cc: stable@kernel.org
Fixes: 6781eabba1bd ("f2fs: give -EINVAL for norecovery and rw mount")
Signed-off-by: Chao Yu <chao@kernel.org>
---
fs/f2fs/recovery.c | 20 ++++++++++++--------
1 file changed, 12 insertions(+), 8 deletions(-)
diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
index 215e442db72c..d7faebaa3c6b 100644
--- a/fs/f2fs/recovery.c
+++ b/fs/f2fs/recovery.c
@@ -399,7 +399,7 @@ static int sanity_check_node_chain(struct f2fs_sb_info *sbi, block_t blkaddr,
}
static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head,
- bool check_only)
+ bool check_only, bool *new_inode)
{
struct curseg_info *curseg;
block_t blkaddr, blkaddr_fast;
@@ -447,16 +447,19 @@ static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head,
quota_inode = true;
}
- /*
- * CP | dnode(F) | inode(DF)
- * For this case, we should not give up now.
- */
entry = add_fsync_inode(sbi, head, ino_of_node(folio),
quota_inode);
if (IS_ERR(entry)) {
err = PTR_ERR(entry);
- if (err == -ENOENT)
+ /*
+ * CP | dnode(F) | inode(DF)
+ * For this case, we should not give up now.
+ */
+ if (err == -ENOENT) {
+ if (check_only)
+ *new_inode = true;
goto next;
+ }
f2fs_folio_put(folio, true);
break;
}
@@ -875,6 +878,7 @@ int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only)
int ret = 0;
unsigned long s_flags = sbi->sb->s_flags;
bool need_writecp = false;
+ bool new_inode = false;
f2fs_notice(sbi, "f2fs_recover_fsync_data: recovery fsync data, "
"check_only: %d", check_only);
@@ -890,8 +894,8 @@ int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only)
f2fs_down_write(&sbi->cp_global_sem);
/* step #1: find fsynced inode numbers */
- err = find_fsync_dnodes(sbi, &inode_list, check_only);
- if (err || list_empty(&inode_list))
+ err = find_fsync_dnodes(sbi, &inode_list, check_only, &new_inode);
+ if (err < 0 || (list_empty(&inode_list) && (!check_only || !new_inode)))
goto skip;
if (check_only) {
--
2.49.0
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
^ permalink raw reply related [flat|nested] 2+ messages in thread