[PATCH v1] ntfs: limit MFT mirror comparisons to system records

From: Baolin Liu

Date: Fri Nov 20 2026 - 10:51:49 EST


From: Baolin Liu <liubaolin@xxxxxxxxxx>

Windows 10 can leave $MFTMirr records from record 16 onwards stale on
volumes with large clusters. Comparing them can falsely report
corruption.

Commit 749357bd48c1 ("ntfs: mount hibernated volumes read-only regardless
of errors=") made these mismatches force a read-only mount even with
errors=continue.

Limit the record checks and comparison to the first 16 records within
the mirror, following ntfs-3g. Keep reading the full mirror for runlist
mapping and location checks, and leave mirror writeback unchanged.

Fixes: 2beaa98b46c4 ("ntfs: restore $MFT mirror contents check")
Cc: stable@xxxxxxxxxxxxxxx
Link: https://github.com/tuxera/ntfs-3g/commit/85e208176ffd16ce9a51a9f534b194f0ebbde6e3
Signed-off-by: Baolin Liu <liubaolin@xxxxxxxxxx>
---
fs/ntfs/super.c | 9 +++++++++
1 file changed, 9 insertions(+)

diff --git a/fs/ntfs/super.c b/fs/ntfs/super.c
index 98e6449fd79f..103d2b1b630b 100644
--- a/fs/ntfs/super.c
+++ b/fs/ntfs/super.c
@@ -1207,6 +1207,14 @@ static bool check_mft_mirror(struct ntfs_volume *vol)
++index;
}

+ /*
+ * Windows 10 may leave mirror records beyond the system files
+ * stale. Read the full mirror to map its runlist, but only
+ * compare the system records, as ntfs-3g does.
+ */
+ if (i >= FILE_first_user)
+ goto next_record;
+
/* Do not check the record if it is not in use. */
if (((struct mft_record *)kmft)->flags & MFT_RECORD_IN_USE) {
/* Make sure the record is ok. */
@@ -1250,6 +1258,7 @@ static bool check_mft_mirror(struct ntfs_volume *vol)
i);
goto mm_unmap_out;
}
+next_record:
kmft += vol->mft_record_size;
kmirr += vol->mft_record_size;
} while (++i < vol->mftmirr_size);

base-commit: 4862a81576e3b20ba40cb94a38c5df0450047ffe