From: liubaolin <liubaolin12138@163.com>
To: "d3nchk@outlook.com" <d3nchk@outlook.com>,
"ntfs@lists.linux.dev" <ntfs@lists.linux.dev>
Cc: "linkinjeon@kernel.org" <linkinjeon@kernel.org>,
"hyc.lee@gmail.com" <hyc.lee@gmail.com>,
"zenghongling@kylinos.cn" <zenghongling@kylinos.cn>,
"regressions@lists.linux.dev" <regressions@lists.linux.dev>,
"stable@vger.kernel.org" <stable@vger.kernel.org>
Subject: Re: [REGRESSION][NTFS] Linux 7.3.4 mounts 64K-cluster NTFS read-only after $MFTMirr record 24 mismatch
Date: Fri, 20 Nov 2026 22:26:25 +0800 [thread overview]
Message-ID: <ee94181c-784f-4f9b-8697-cca5abec6b07@163.com> (raw)
In-Reply-To: <VE1P190MB0877D026E2C3905F0324B46F85782@VE1P190MB0877.EURP190.PROD.OUTLOOK.COM>
在 2026/11/20 20:49, d3nchk@outlook.com 写道:
> Hello,
>
> Resending in plain text because the previous HTML-formatted message was rejected by the regressions mailing list. Apologies if this creates duplicates for other recipients.
>
> I am reporting a regression in the in-kernel fs/ntfs driver (NOT ntfs3).
>
> Environment:
> - Distribution: Debian 13, backports kernel
> - Working kernel: 7.2.12
> - Broken kernel: 7.3.4+deb13-amd64
> - Device: /dev/nvme0n1p1
> - Filesystem: NTFS 3.1
> - Cluster size: 65536 bytes
> - Mount options: defaults (including errors=continue)
> - Driver: fs/ntfs/ntfs.ko.xz
>
> Kernel log:
>
> ntfs: (device nvme0n1p1): check_mft_mirror(): $MFT and $MFTMirr record 24 do not match. Run chkdsk.
> ntfs: volume version 3.1, dev nvme0n1p1, cluster size 65536
> ntfs: (device nvme0n1p1): load_system_files(): Errors were recorded during mount. Mounting read-only. Run chkdsk.
>
> This volume worked read-write with kernel 7.2.12. After upgrading to 7.3.4, it is forcibly mounted read-only, making normal file operations impossible.
>
> Windows chkdsk has been run repeatedly, both through the GUI and command line, including dismounted-volume checks. The issue persists after rebooting into Linux.
>
> I suspect commit 749357bd48c16c4cdc09093450deca609492f012 introduced the new read-only behavior for this case.
>
> There is also a potentially relevant NTFS compatibility issue: modern Windows may not keep all $MFTMirr records synchronized on volumes with large clusters. Here the mismatch is reported at record 24, beyond the first 16 records.
>
> Please investigate:
>
> 1. Whether the comparison of MFT mirror record 24 is valid for modern Windows-created NTFS volumes with 64K clusters.
> 2. Whether this is a false-positive filesystem corruption report.
> 3. Whether errors=continue can be honored for noncritical MFT mirror mismatches, while retaining mandatory read-only protection for Windows hibernation and genuinely unsafe filesystem states.
> 4. Whether a fix can be backported to the Linux 7.3 stable series.
>
> This regression prevents normal read-write access to this volume.
>
> Thank you.
Hi,
Thanks for the report and the detailed logs.
I checked 749357bd48c1. It didn't introduce the mirror comparison,
but it did make errors recorded during mount force a read-only mount,
even with errors=continue. That explains the change in behaviour you're
seeing.
There is a known compatibility issue with the comparison itself.
On volumes with large clusters, Windows 10 can leave later records
in $MFTMirr out of date.
ntfs-3g already handles this by comparing only the first 16 records,
or fewer if the mirror is smaller. See commit 85e208176ffd ("Stopped
checking matches of MFTMirr against MFT at record 16").
fs/ntfs still compares the whole mirror. With 64 KiB clusters and
1 KiB MFT records, that means records 0 through 63. A difference
at record 24 could therefore just be a stale mirror copy, rather
than filesystem corruption. Your report looks consistent with
this case.
I'll follow the ntfs-3g approach and limit the record checks and
comparison to the first 16 records within the mirror. We'll still
read the full mirror and check its location, and leave the mirror
writeback range unchanged.
This should allow read-write mounting in this case, provided
there are no other errors. We shouldn't need to relax the general
errors= handling or change the read-only protection for hibernated
volumes and other unsafe states.
I'll include a Fixes tag and Cc stable so the patch can be
considered for the 7.3 stable series after upstream acceptance.
I'll send the patch shortly and CC you. You can then give it a
try and let us know whether it resolves the issue.
Thanks,
Baolin
next prev parent reply other threads:[~2026-11-20 14:27 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-20 12:49 [REGRESSION][NTFS] Linux 7.3.4 mounts 64K-cluster NTFS read-only after $MFTMirr record 24 mismatch d3nchk
2026-11-20 14:26 ` liubaolin [this message]
2026-11-20 20:39 ` Sasha Levin
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=ee94181c-784f-4f9b-8697-cca5abec6b07@163.com \
--to=liubaolin12138@163.com \
--cc=d3nchk@outlook.com \
--cc=hyc.lee@gmail.com \
--cc=linkinjeon@kernel.org \
--cc=ntfs@lists.linux.dev \
--cc=regressions@lists.linux.dev \
--cc=stable@vger.kernel.org \
--cc=zenghongling@kylinos.cn \
/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