Skip to contentSkip to conversation

fs: Remove NTFS classic

PATCH Applied Matthew Wilcox (Oracle) 10 participants 35 messages Jan 15, 2024 – Apr 18, 2024

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fs: Remove NTFS classic

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

Cc Anton Altaparmakov <[email protected]>
Cc Namjae Jeon <[email protected]>
---
 CREDITS                            |    5 +
 Documentation/filesystems/ntfs.rst |  466 ----
 MAINTAINERS                        |   10 -
 fs/Kconfig                         |    1 -
 fs/Makefile                        |    1 -
 fs/ntfs/Kconfig                    |   81 -
 fs/ntfs/Makefile                   |   15 -
 fs/ntfs/aops.c                     | 1744 ---------------
 fs/ntfs/aops.h                     |   88 -
 fs/ntfs/attrib.c                   | 2624 -----------------------
 fs/ntfs/attrib.h                   |  102 -
 fs/ntfs/bitmap.c                   |  179 --
 fs/ntfs/bitmap.h                   |  104 -
 fs/ntfs/collate.c                  |  110 -
 fs/ntfs/collate.h                  |   36 -
 fs/ntfs/compress.c                 |  950 ---------
 fs/ntfs/debug.c                    |  159 --
 fs/ntfs/debug.h                    |   57 -
 fs/ntfs/dir.c                      | 1540 -------------
 fs/ntfs/dir.h                      |   34 -
 fs/ntfs/endian.h                   |   79 -
 fs/ntfs/file.c                     | 1997 -----------------
 fs/ntfs/index.c                    |  440 ----
 fs/ntfs/index.h                    |  134 --
 fs/ntfs/inode.c                    | 3102 ---------------------------
 fs/ntfs/inode.h                    |  310 ---
 fs/ntfs/layout.h                   | 2421 ---------------------
 fs/ntfs/lcnalloc.c                 | 1000 ---------
 fs/ntfs/lcnalloc.h                 |  131 --
 fs/ntfs/logfile.c                  |  849 --------
 fs/ntfs/logfile.h                  |  295 ---
 fs/ntfs/malloc.h                   |   77 -
 fs/ntfs/mft.c                      | 2907 -------------------------
 fs/ntfs/mft.h                      |  110 -
 fs/ntfs/mst.c                      |  189 --
 fs/ntfs/namei.c                    |  392 ----
 fs/ntfs/ntfs.h                     |  150 --
 fs/ntfs/quota.c                    |  103 -
 fs/ntfs/quota.h                    |   21 -
 fs/ntfs/runlist.c                  | 1893 ----------------
 fs/ntfs/runlist.h                  |   88 -
 fs/ntfs/super.c                    | 3202 ----------------------------
 fs/ntfs/sysctl.c                   |   58 -
 fs/ntfs/sysctl.h                   |   27 -
 fs/ntfs/time.h                     |   89 -
 fs/ntfs/types.h                    |   55 -
 fs/ntfs/unistr.c                   |  384 ----
 fs/ntfs/upcase.c                   |   73 -
 fs/ntfs/usnjrnl.c                  |   70 -
 fs/ntfs/usnjrnl.h                  |  191 --
 fs/ntfs/volume.h                   |  164 --
 51 files changed, 5 insertions(+), 29302 deletions(-)
 delete mode 100644 Documentation/filesystems/ntfs.rst
 delete mode 100644 fs/ntfs/Kconfig
 delete mode 100644 fs/ntfs/Makefile
 delete mode 100644 fs/ntfs/aops.c
 delete mode 100644 fs/ntfs/aops.h
 delete mode 100644 fs/ntfs/attrib.c
 delete mode 100644 fs/ntfs/attrib.h
 delete mode 100644 fs/ntfs/bitmap.c
 delete mode 100644 fs/ntfs/bitmap.h
 delete mode 100644 fs/ntfs/collate.c
 delete mode 100644 fs/ntfs/collate.h
 delete mode 100644 fs/ntfs/compress.c
 delete mode 100644 fs/ntfs/debug.c
 delete mode 100644 fs/ntfs/debug.h
 delete mode 100644 fs/ntfs/dir.c
 delete mode 100644 fs/ntfs/dir.h
 delete mode 100644 fs/ntfs/endian.h
 delete mode 100644 fs/ntfs/file.c
 delete mode 100644 fs/ntfs/index.c
 delete mode 100644 fs/ntfs/index.h
 delete mode 100644 fs/ntfs/inode.c
 delete mode 100644 fs/ntfs/inode.h
 delete mode 100644 fs/ntfs/layout.h
 delete mode 100644 fs/ntfs/lcnalloc.c
 delete mode 100644 fs/ntfs/lcnalloc.h
 delete mode 100644 fs/ntfs/logfile.c
 delete mode 100644 fs/ntfs/logfile.h
 delete mode 100644 fs/ntfs/malloc.h
 delete mode 100644 fs/ntfs/mft.c
 delete mode 100644 fs/ntfs/mft.h
 delete mode 100644 fs/ntfs/mst.c
 delete mode 100644 fs/ntfs/namei.c
 delete mode 100644 fs/ntfs/ntfs.h
 delete mode 100644 fs/ntfs/quota.c
 delete mode 100644 fs/ntfs/quota.h
 delete mode 100644 fs/ntfs/runlist.c
 delete mode 100644 fs/ntfs/runlist.h
 delete mode 100644 fs/ntfs/super.c
 delete mode 100644 fs/ntfs/sysctl.c
 delete mode 100644 fs/ntfs/sysctl.h
 delete mode 100644 fs/ntfs/time.h
 delete mode 100644 fs/ntfs/types.h
 delete mode 100644 fs/ntfs/unistr.c
 delete mode 100644 fs/ntfs/upcase.c
 delete mode 100644 fs/ntfs/usnjrnl.c
 delete mode 100644 fs/ntfs/usnjrnl.h
 delete mode 100644 fs/ntfs/volume.h
CREDITS +5 -0
index 06b177c6af6a..78c0be09041b 100644
@@ -63,6 +63,11 @@ D: dosfs, LILO, some fd features, ATM, various other hacks here and there
 S: Buenos Aires
 S: Argentina
 
+NTFS FILESYSTEM
+N: Anton Altaparmakov
+E: [email protected]
+D: NTFS filesystem
+
 N: Tim Alpaerts
 E: [email protected]
 D: 802.2 class II logical link control layer,
Documentation/filesystems/ntfs.rst deleted +0 -466

466 lines not on this page · raw message

MAINTAINERS +0 -10
index 391bbb855cbe..9d4a1e5a8e7f 100644
@@ -15460,16 +15460,6 @@ W:	https://github.com/davejiang/linux/wiki
 T:	git https://github.com/davejiang/linux.git
 F:	drivers/ntb/hw/intel/
 
-NTFS FILESYSTEM
-M:	Anton Altaparmakov <[email protected]>
-R:	Namjae Jeon <[email protected]>
-L:	[email protected]
-S:	Supported
-W:	http://www.tuxera.com/
-T:	git git://git.kernel.org/pub/scm/linux/kernel/git/aia21/ntfs.git
-F:	Documentation/filesystems/ntfs.rst
-F:	fs/ntfs/
-
 NTFS3 FILESYSTEM
 M:	Konstantin Komarov <[email protected]>
 L:	[email protected]
fs/Kconfig +0 -1
index a3159831ba98..96eb9d31c5a9 100644
@@ -163,7 +163,6 @@ menu "DOS/FAT/EXFAT/NT Filesystems"
 
 source "fs/fat/Kconfig"
 source "fs/exfat/Kconfig"
-source "fs/ntfs/Kconfig"
 source "fs/ntfs3/Kconfig"
 
 endmenu
fs/Makefile +0 -1
index a6962c588962..faeb0f3a1018 100644
@@ -92,7 +92,6 @@ obj-y				+= unicode/
 obj-$(CONFIG_SYSV_FS)		+= sysv/
 obj-$(CONFIG_SMBFS)		+= smb/
 obj-$(CONFIG_HPFS_FS)		+= hpfs/
-obj-$(CONFIG_NTFS_FS)		+= ntfs/
 obj-$(CONFIG_NTFS3_FS)		+= ntfs3/
 obj-$(CONFIG_UFS_FS)		+= ufs/
 obj-$(CONFIG_EFS_FS)		+= efs/
fs/ntfs/Kconfig deleted +0 -81

81 lines not on this page · raw message

fs/ntfs/Makefile deleted +0 -15
deleted file mode 100644
index 3e736572ed00..000000000000
@@ -1,15 +0,0 @@
-# SPDX-License-Identifier: GPL-2.0
-# Rules for making the NTFS driver.
-
-obj-$(CONFIG_NTFS_FS) += ntfs.o
-
-ntfs-y := aops.o attrib.o collate.o compress.o debug.o dir.o file.o \
-	  index.o inode.o mft.o mst.o namei.o runlist.o super.o sysctl.o \
-	  unistr.o upcase.o
-
-ntfs-$(CONFIG_NTFS_RW) += bitmap.o lcnalloc.o logfile.o quota.o usnjrnl.o
-
-ccflags-y := -DNTFS_VERSION=\"2.1.32\"
-ccflags-$(CONFIG_NTFS_DEBUG)	+= -DDEBUG
-ccflags-$(CONFIG_NTFS_RW)	+= -DNTFS_RW
-
fs/ntfs/aops.c deleted +0 -1744

1,744 lines not on this page · raw message

fs/ntfs/aops.h deleted +0 -88

88 lines not on this page · raw message

fs/ntfs/attrib.c deleted +0 -2624

565 lines not on this page · raw message

fs/ntfs/attrib.h deleted +0 -102

102 lines not on this page · raw message

fs/ntfs/bitmap.c deleted +0 -179

179 lines not on this page · raw message

fs/ntfs/bitmap.h deleted +0 -104

104 lines not on this page · raw message

fs/ntfs/collate.c deleted +0 -110

110 lines not on this page · raw message

fs/ntfs/collate.h deleted +0 -36
deleted file mode 100644
index f2255619b4f4..000000000000

Lines of this file are not shown. View the raw message.

fs/ntfs/compress.c deleted +0 -950

950 lines not on this page · raw message

fs/ntfs/debug.c deleted +0 -159

159 lines not on this page · raw message

fs/ntfs/debug.h deleted +0 -57

57 lines not on this page · raw message

fs/ntfs/dir.c deleted +0 -1540

1,540 lines not on this page · raw message

fs/ntfs/dir.h deleted +0 -34
deleted file mode 100644
index 0e326753df40..000000000000

Lines of this file are not shown. View the raw message.

fs/ntfs/endian.h deleted +0 -79

79 lines not on this page · raw message

fs/ntfs/file.c deleted +0 -1997

1,997 lines not on this page · raw message

fs/ntfs/index.c deleted +0 -440

440 lines not on this page · raw message

fs/ntfs/index.h deleted +0 -134

134 lines not on this page · raw message

fs/ntfs/inode.c deleted +0 -3102

3,102 lines not on this page · raw message

fs/ntfs/inode.h deleted +0 -310

310 lines not on this page · raw message

fs/ntfs/layout.h deleted +0 -2421

2,421 lines not on this page · raw message

fs/ntfs/lcnalloc.c deleted +0 -1000

1,000 lines not on this page · raw message

fs/ntfs/lcnalloc.h deleted +0 -131

131 lines not on this page · raw message

fs/ntfs/logfile.c deleted +0 -849

849 lines not on this page · raw message

fs/ntfs/logfile.h deleted +0 -295

295 lines not on this page · raw message

fs/ntfs/malloc.h deleted +0 -77

77 lines not on this page · raw message

fs/ntfs/mft.c deleted +0 -2907

2,907 lines not on this page · raw message

fs/ntfs/mft.h deleted +0 -110

110 lines not on this page · raw message

fs/ntfs/mst.c deleted +0 -189

189 lines not on this page · raw message

fs/ntfs/namei.c deleted +0 -392

392 lines not on this page · raw message

fs/ntfs/ntfs.h deleted +0 -150

150 lines not on this page · raw message

fs/ntfs/quota.c deleted +0 -103

103 lines not on this page · raw message

fs/ntfs/quota.h deleted +0 -21
deleted file mode 100644
index fe3132a3d6d2..000000000000

Lines of this file are not shown. View the raw message.

fs/ntfs/runlist.c deleted +0 -1893

1,893 lines not on this page · raw message

fs/ntfs/runlist.h deleted +0 -88

88 lines not on this page · raw message

fs/ntfs/super.c deleted +0 -3202

3,202 lines not on this page · raw message

fs/ntfs/sysctl.c deleted +0 -58

58 lines not on this page · raw message

fs/ntfs/sysctl.h deleted +0 -27
deleted file mode 100644
index 96bb2299d2d5..000000000000

Lines of this file are not shown. View the raw message.

fs/ntfs/time.h deleted +0 -89

89 lines not on this page · raw message

fs/ntfs/types.h deleted +0 -55

55 lines not on this page · raw message

fs/ntfs/unistr.c deleted +0 -384

384 lines not on this page · raw message

fs/ntfs/upcase.c deleted +0 -73

73 lines not on this page · raw message

fs/ntfs/usnjrnl.c deleted +0 -70

70 lines not on this page · raw message

fs/ntfs/usnjrnl.h deleted +0 -191

191 lines not on this page · raw message

fs/ntfs/volume.h deleted +0 -164

164 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
2.43.0

Part of the diff is not shown here. View the raw message.

#1 lore raw

Hi Matthew,

Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

It was my impression that people are complaining ntfs3 is causing a whole lot of problems including corrupting people's data. Also, it appears the maintainer has basically disappeared after it got merged.

Is it really such a good idea to remove the original ntfs driver which actually works fine and does not cause any problems when the replacement is so poor and unmaintained?

Also, which APIs are you referring to? I can take a look into those.

Best regards,

Anton

Signature · 3 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer
#2 lore raw
Quoted 8 lines · Anton Altaparmakov

Date: Mon, 15 Jan 2024 12:00:35
From: Anton Altaparmakov <[email protected]>
To: "Matthew Wilcox (Oracle)" <[email protected]>,
Christian Brauner <[email protected]>
Cc: [email protected], [email protected],
[email protected], Namjae Jeon <[email protected]>
Subject: Re: [PATCH] fs: Remove NTFS classic

Quoted 8 lines

Hi Matthew,

Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

It was my impression that people are complaining ntfs3 is causing a whole lot of problems including corrupting people's data. Also, it appears the maintainer has basically disappeared after it got merged.

To be fair - it's more like "intermittent" maintenance rather than "no maintenance". Konstantin - would it be OK for you to have a co-maintainer to help out?

Enrico

Quoted 14 lines

Is it really such a good idea to remove the original ntfs driver which actually works fine and does not cause any problems when the replacement is so poor and unmaintained?

Also, which APIs are you referring to? I can take a look into those.

Best regards,

Anton

Signature · 3 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer
maintainer Acked-by
#3 lore raw
In reply to Anton Altaparmakov

2024-01-15 20:00 GMT+09:00, Anton Altaparmakov <[email protected]>:

Quoted 10 lines

Hi Matthew,

Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

It was my impression that people are complaining ntfs3 is causing a whole lot of problems including corrupting people's data. Also, it appears the maintainer has basically disappeared after it got merged.

What can replace this is not only ntfs3 but also ntfs3g. I think read-only ntfs is obsolete.
I am in favor of remove it.

Thanks.

Quoted 15 lines

Is it really such a good idea to remove the original ntfs driver which actually works fine and does not cause any problems when the replacement is so poor and unmaintained?

Also, which APIs are you referring to? I can take a look into those.

Best regards,

Anton
--
Anton Altaparmakov <anton at tuxera.com> (replace at with @) Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer

#4 lore raw
In reply to Anton Altaparmakov
Quoted 10 lines · Anton Altaparmakov

Hi Matthew,

Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

It was my impression that people are complaining ntfs3 is causing a whole lot of problems including corrupting people's data. Also, it appears the maintainer has basically disappeared after it got merged.

I'm not terribly happy with how the maintainer behaves either, but you've also been missing in action for nine years (if we're counting code authored by you) or two years (if a R-b is enough).

According to your documentation, you don't support creating new files or directories, so I'd suggest that your code has never been put through the xfstests wringer in the way that ntfs3 has.

Quoted 1 line

Also, which APIs are you referring to? I can take a look into those.

The biggest one for me is the folio work. Everywhere in your code that use a struct page needs to be converted to a struct folio. Support for large folios is optional, and I wouldn't bother trying to add that. Take a look at, eg, nilfs2, ext4, ext2, iomap, NFS, AFS for some filesystems which have been (at least mostly) converted.

Any functions in mm/folio-compat.c should no longer be called.

If we're being really ambitious, filesystems should stop using the buffer cache and switch to using iomap. There's a lot of work going on and unmaintained filesystems are holding us back.

#5 lore raw

Re: [PATCH] Remove NTFS classic

Hi Matthew,

Quoted 29 lines · Matthew Wilcox
Quoted 10 lines · Anton Altaparmakov

Hi Matthew,

Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

It was my impression that people are complaining ntfs3 is causing a whole lot of problems including corrupting people's data. Also, it appears the maintainer has basically disappeared after it got merged.

I'm not terribly happy with how the maintainer behaves either, but you've also been missing in action for nine years (if we're counting code authored by you) or two years (if a R-b is enough).

According to your documentation, you don't support creating new files or directories, so I'd suggest that your code has never been put through the xfstests wringer in the way that ntfs3 has.

Quoted 1 line

Also, which APIs are you referring to? I can take a look into those.

The biggest one for me is the folio work. Everywhere in your code that use a struct page needs to be converted to a struct folio. Support for large folios is optional, and I wouldn't bother trying to add that. Take a look at, eg, nilfs2, ext4, ext2, iomap, NFS, AFS for some filesystems which have been (at least mostly) converted.

Quoted 5 lines

Any functions in mm/folio-compat.c should no longer be called.

If we're being really ambitious, filesystems should stop using the buffer cache and switch to using iomap. There's a lot of work going on and unmaintained filesystems are holding us back.

Thanks. I would have been happy to do that but it seems everyone else prefers to remove it so I have given my ack for that instead...

Best regards,

Anton

Signature · 3 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer
Acked-by on v1 1/1
#6 lore raw
In reply to Matthew Wilcox (Oracle)
Quoted 7 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

Cc Anton Altaparmakov <[email protected]>
Cc Namjae Jeon <[email protected]>

I agree - this is largely untestable code and given the newer, better featured functionality we have (ntfs3, ntfs3g), we should just get rid of this old code.

Cheers,

Dave.

Signature · 2 lines
Dave Chinner
[email protected]
maintainer Applied
#7 lore raw
In reply to Matthew Wilcox (Oracle)
Quoted 5 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

I see no reason to not at least try and remove it given that we have ntfs3 as a replacement. Worst case is we have to put it back in. Let's try it.

---

Applied to the vfs.fs branch of the vfs/vfs.git tree. Patches in the vfs.fs branch should appear in linux-next soon.

Please report any outstanding bugs that were missed during review in a new review to the original patch series allowing us to drop it.

It's encouraged to provide Acked-bys and Reviewed-bys even though the patch has now been applied. If possible patch trailers will be updated.

Note that commit hashes shown below are subject to change due to rebase, trailer updates or similar. If in doubt, please check the listed branch.

tree:   https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git
branch: vfs.fs

[1/1] fs: Remove NTFS classic

      https://git.kernel.org/vfs/vfs/c/9c67092ed339
#8 lore raw

Re: [PATCH] Remove NTFS classic

Hi,

It seems there is consensus to remove it so please add:

Best regards,

Anton

Quoted 32 lines · Christian Brauner
Quoted 5 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

I see no reason to not at least try and remove it given that we have ntfs3 as a replacement. Worst case is we have to put it back in. Let's try it.

---

Applied to the vfs.fs branch of the vfs/vfs.git tree. Patches in the vfs.fs branch should appear in linux-next soon.

Please report any outstanding bugs that were missed during review in a new review to the original patch series allowing us to drop it.

It's encouraged to provide Acked-bys and Reviewed-bys even though the patch has now been applied. If possible patch trailers will be updated.

Note that commit hashes shown below are subject to change due to rebase, trailer updates or similar. If in doubt, please check the listed branch.

tree:   https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git
branch: vfs.fs

[1/1] fs: Remove NTFS classic

     https://git.kernel.org/vfs/vfs/c/9c67092ed339
Signature · 3 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer
#9 lore raw

Re: [PATCH] Remove NTFS classic

Quoted 3 lines · Anton Altaparmakov

Hi,

It seems there is consensus to remove it so please add:

Well, we'll try. This is one of those cases where we might end up not being able to do it. But imho this is a case where there's sufficient reason to at least try and remove this code precisely because we have an alternative implementation that's been around for a while.

IOW, this isn't like reiserfs where we're actually getting rid of a filesystem completely.

#10 lore raw

Re: [PATCH] Remove NTFS classic

Hi,

Quoted 6 lines · Christian Brauner
Quoted 1 line · Anton Altaparmakov

It seems there is consensus to remove it so please add:

Well, we'll try. This is one of those cases where we might end up not being able to do it.

If that happens, I am happy to update the APIs as Matthew detailed but obviously I do not want to waste my time doing that if it will be removed anyway...

Best regards,

Anton

Quoted 6 lines

But imho this is a case where there's sufficient
reason to at least try and remove this code precisely because we have an alternative implementation that's been around for a while.

IOW, this isn't like reiserfs where we're actually getting rid of a filesystem completely.

Signature · 3 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
Linux NTFS maintainer
#11 lore raw
In reply to Christian Brauner
Quoted 8 lines · Christian Brauner
Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

I see no reason to not at least try and remove it given that we have ntfs3 as a replacement. Worst case is we have to put it back in. Let's try it.

This patch breaks boot of my Lenovo ThinkPad X13s where I have an fstab entry to access firmware from the original Windows partition; ended up in some kind of systemd emergency mode.

Fix is easy enough once I figured out what broke, but requires changing the fs type from 'ntfs' to 'ntfs3' in the fstab (after enabling NTFS3 in the kernel config).

Is it possible to provide an alias or something to avoid breaking people's machines like this? Perhaps something is needed for the Kconfig too.

I also now get a bunch of warning when listing the root directory:

[   68.967970] ntfs3: nvme0n1p3: ino=130d6, Correct links count -> 1.
[   68.969137] ntfs3: nvme0n1p3: ino=13a, Correct links count -> 1.
[   68.969798] ntfs3: nvme0n1p3: ino=3ad, Correct links count -> 1.
[   68.970431] ntfs3: nvme0n1p3: ino=3d9, Correct links count -> 1.
[   68.971150] ntfs3: nvme0n1p3: ino=26, Correct links count -> 1.
[   68.971780] ntfs3: nvme0n1p3: ino=eb, Correct links count -> 1.
[   68.972263] ntfs3: nvme0n1p3: ino=1ce, Correct links count -> 1.
[   68.973062] ntfs3: nvme0n1p3: ino=a608, Correct links count -> 1.

Flagging this as a regression so that Thorsten is aware of it:

#regzbot introduced: 7ffa8f3d3023

Johan

#12 lore raw
Quoted 24 lines · Johan Hovold
Quoted 8 lines · Christian Brauner
Quoted 3 lines · Matthew Wilcox (Oracle)

The replacement, NTFS3, was merged over two years ago. It is now time to remove the original from the tree as it is the last user of several APIs, and it is not worth changing.

I see no reason to not at least try and remove it given that we have ntfs3 as a replacement. Worst case is we have to put it back in. Let's try it.

This patch breaks boot of my Lenovo ThinkPad X13s where I have an fstab entry to access firmware from the original Windows partition; ended up in some kind of systemd emergency mode.

Fix is easy enough once I figured out what broke, but requires changing the fs type from 'ntfs' to 'ntfs3' in the fstab (after enabling NTFS3 in the kernel config).

Is it possible to provide an alias or something to avoid breaking people's machines like this? Perhaps something is needed for the Kconfig too.

I also now get a bunch of warning when listing the root directory:

Quoted 8 lines
[   68.967970] ntfs3: nvme0n1p3: ino=130d6, Correct links count -> 1.
[   68.969137] ntfs3: nvme0n1p3: ino=13a, Correct links count -> 1.
[   68.969798] ntfs3: nvme0n1p3: ino=3ad, Correct links count -> 1.
[   68.970431] ntfs3: nvme0n1p3: ino=3d9, Correct links count -> 1.
[   68.971150] ntfs3: nvme0n1p3: ino=26, Correct links count -> 1.
[   68.971780] ntfs3: nvme0n1p3: ino=eb, Correct links count -> 1.
[   68.972263] ntfs3: nvme0n1p3: ino=1ce, Correct links count -> 1.
[   68.973062] ntfs3: nvme0n1p3: ino=a608, Correct links count -> 1.

This shouldn't warn because it's correcting this on the fly.

Quoted 4 lines

Flagging this as a regression so that Thorsten is aware of it:

#regzbot introduced: 7ffa8f3d3023

Thanks for the report. I'll send a fix.

#13 lore raw

PATCH 1/2 ntfs3: serve as alias for the legacy ntfs driver

In reply to Johan Hovold

Johan Hovold reported that removing the legacy ntfs driver broke boot for him since his fstab uses the legacy ntfs driver to access firmware from the original Windows partition.

Use ntfs3 as an alias for legacy ntfs if CONFIG_NTFS_FS is selected. This is similar to how ext3 is treated.

Fixes 7ffa8f3d3023 ("fs: Remove NTFS classic")
Cc "Matthew Wilcox (Oracle)" <[email protected]>
Cc Johan Hovold <[email protected]>

---
Hey,

This is so far compile tested. It would be great if someone could test this. @Johan?

Thanks!
Christian

---
 fs/ntfs3/Kconfig |  9 +++++++++
 fs/ntfs3/super.c | 31 +++++++++++++++++++++++++++++++
 2 files changed, 40 insertions(+)
fs/ntfs3/Kconfig +9 -0
index cdfdf51e55d7..7bc31d69f680 100644
@@ -46,3 +46,12 @@ config NTFS3_FS_POSIX_ACL
 	  NOTE: this is linux only feature. Windows will ignore these ACLs.
 
 	  If you don't know what Access Control Lists are, say N.
+
+config NTFS_FS
+	tristate "NTFS file system support"
+	select NTFS3_FS
+	select BUFFER_HEAD
+	select NLS
+	help
+	  This config option is here only for backward compatibility. NTFS
+	  filesystem is now handled by the NTFS3 driver.
fs/ntfs3/super.c +31 -0
index 9df7c20d066f..8d2e51bae2cb 100644
@@ -1798,6 +1798,35 @@ static struct file_system_type ntfs_fs_type = {
 	.kill_sb		= ntfs3_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 };
+
+#if IS_ENABLED(CONFIG_NTFS_FS)
+static struct file_system_type ntfs_legacy_fs_type = {
+	.owner			= THIS_MODULE,
+	.name			= "ntfs",
+	.init_fs_context	= ntfs_init_fs_context,
+	.parameters		= ntfs_fs_parameters,
+	.kill_sb		= ntfs3_kill_sb,
+	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
+};
+MODULE_ALIAS_FS("ntfs");
+
+static inline void register_as_ntfs_legacy(void)
+{
+	int err = register_filesystem(&ntfs_legacy_fs_type);
+	if (err)
+		pr_warn("ntfs3: Failed to register legacy ntfs filesystem driver: %d\n", err);
+}
+
+static inline void unregister_as_ntfs_legacy(void)
+{
+	unregister_filesystem(&ntfs_legacy_fs_type);
+}
+#else
+static inline void register_as_ntfs_legacy(void) {}
+static inline void unregister_as_ntfs_legacy(void) {}
+#endif
+
+
 // clang-format on
 
 static int __init init_ntfs_fs(void)
@@ -1832,6 +1861,7 @@ static int __init init_ntfs_fs(void)
 		goto out1;
 	}
 
+	register_as_ntfs_legacy();
 	err = register_filesystem(&ntfs_fs_type);
 	if (err)
 		goto out;
@@ -1849,6 +1879,7 @@ static void __exit exit_ntfs_fs(void)
 	rcu_barrier();
 	kmem_cache_destroy(ntfs_inode_cachep);
 	unregister_filesystem(&ntfs_fs_type);
+	unregister_as_ntfs_legacy();
 	ntfs3_exit_bitmap();
 
 #ifdef CONFIG_PROC_FS
Signature · 1 line
2.43.0
Tested-by on v1 1/2
#14 lore raw

Re: [PATCH 1/2] ntfs3: serve as alias for the legacy ntfs driver

Quoted 12 lines · Christian Brauner

Johan Hovold reported that removing the legacy ntfs driver broke boot for him since his fstab uses the legacy ntfs driver to access firmware from the original Windows partition.

Use ntfs3 as an alias for legacy ntfs if CONFIG_NTFS_FS is selected. This is similar to how ext3 is treated.

Fixes 7ffa8f3d3023 ("fs: Remove NTFS classic")
Cc "Matthew Wilcox (Oracle)" <[email protected]>
Cc Johan Hovold <[email protected]>
Quoted 5 lines

---
Hey,

This is so far compile tested. It would be great if someone could test this. @Johan?

This seems to do the trick. Thanks for the quick fix.

Do we want to do something about the fact that ntfs mounts may now become writable as well?

Johan

#15 lore raw

Re: [PATCH 1/2] ntfs3: serve as alias for the legacy ntfs driver

Quoted 19 lines · Johan Hovold
Quoted 12 lines · Christian Brauner

Johan Hovold reported that removing the legacy ntfs driver broke boot for him since his fstab uses the legacy ntfs driver to access firmware from the original Windows partition.

Use ntfs3 as an alias for legacy ntfs if CONFIG_NTFS_FS is selected. This is similar to how ext3 is treated.

Fixes 7ffa8f3d3023 ("fs: Remove NTFS classic")
Cc "Matthew Wilcox (Oracle)" <[email protected]>
Cc Johan Hovold <[email protected]>
Quoted 5 lines

---
Hey,

This is so far compile tested. It would be great if someone could test this. @Johan?

Quoted 6 lines

This seems to do the trick. Thanks for the quick fix.

Do we want to do something about the fact that ntfs mounts may now become writable as well?

We can enforce that mounting as ntfs means that it's read-only unless rw support is compiled in most likely. @Anton or other maintainers?

#16 lore raw

PATCH 2/2 ntfs3: remove warning

In reply to Johan Hovold

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

---
 fs/ntfs3/inode.c | 1 -
 1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -1
index eb7a8c9fba01..8cc94a6a97ed 100644
@@ -424,7 +424,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
 
 	if (names != le16_to_cpu(rec->hard_links)) {
 		/* Correct minor error on the fly. Do not mark inode as dirty. */
-		ntfs_inode_warn(inode, "Correct links count -> %u.", names);
 		rec->hard_links = cpu_to_le16(names);
 		ni->mi.dirty = true;
 	}
Signature · 1 line
2.43.0
Tested-by on v1 2/2
#17 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 8 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

I also see a

ntfs3: Max link count 4000

message on mount which wasn't there with NTFS legacy. Is that benign and should be suppressed too perhaps?

Johan

#18 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 10 lines · Johan Hovold
Quoted 8 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

Quoted 8 lines

I also see a

ntfs3: Max link count 4000

message on mount which wasn't there with NTFS legacy. Is that benign and should be suppressed too perhaps?

We need a reply from the ntfs3 maintainers here.

#19 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 13 lines · Christian Brauner
Quoted 12 lines · Johan Hovold
Quoted 8 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

Quoted 8 lines
Quoted 6 lines

I also see a

ntfs3: Max link count 4000

message on mount which wasn't there with NTFS legacy. Is that benign and should be suppressed too perhaps?

We need a reply from the ntfs3 maintainers here.

Those are known for taken their time to respond -- like we see here, as nothing happened for 10 days now. Which makes we wonder: should we at least bring the first patch from this series onto the track towards mainline?

FWIW, quick side note: just noticed there was another report about the "Correct links count -> 1." messages:
https://lore.kernel.org/all/[email protected]/

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#regzbot duplicate:
https://lore.kernel.org/all/[email protected]/
#regzbot poke

#20 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 35 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 12 lines · Christian Brauner
Quoted 11 lines · Johan Hovold
Quoted 8 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

Quoted 7 lines
Quoted 6 lines

I also see a

ntfs3: Max link count 4000

message on mount which wasn't there with NTFS legacy. Is that benign and should be suppressed too perhaps?

We need a reply from the ntfs3 maintainers here.

Those are known for taken their time to respond -- like we see here, as nothing happened for 10 days now. Which makes we wonder: should we at least bring the first patch from this series onto the track towards mainline?

FWIW, quick side note: just noticed there was another report about the "Correct links count -> 1." messages:
https://lore.kernel.org/all/[email protected]/

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

Quoted 4 lines

#regzbot duplicate:
https://lore.kernel.org/all/[email protected]/
#regzbot poke

Hello Christian, Johan, all,

There is no problem in suppressing the output of any messages during mounting, like:

fs/ntfs3/super.c +0 -0
index cef5467fd928..4643b06b1550 100644
@@ -1804,8 +1804,6 @@ static int __init init_ntfs_fs(void)

{
        int err;
-       pr_info("ntfs3: Max link count %u\n", NTFS_LINK_MAX); -
        if (IS_ENABLED(CONFIG_NTFS3_FS_POSIX_ACL)) pr_info("ntfs3: Enabled Linux POSIX ACLs support\n"); if (IS_ENABLED(CONFIG_NTFS3_64BIT_CLUSTER))

Messages like this:

fs/ntfs3/inode.c +0 -0
index eb7a8c9fba01..8cc94a6a97ed 100644
@@ -424,7 +424,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,

    if (names != le16_to_cpu(rec->hard_links)) {
        /* Correct minor error on the fly. Do not mark inode as dirty. */ -        ntfs_inode_warn(inode, "Correct links count -> %u.", names); rec->hard_links = cpu_to_le16(names);
        ni->mi.dirty = true;
    }

can also be suppressed for the sake of seamless transition from a remote NTFS driver. However, I believe that file system corrections should be reported to the user.

#21 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 39 lines · Konstantin Komarov
Quoted 13 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 12 lines · Christian Brauner
Quoted 11 lines · Johan Hovold
Quoted 8 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines

I also see a

ntfs3: Max link count 4000

message on mount which wasn't there with NTFS legacy. Is that benign and should be suppressed too perhaps?

We need a reply from the ntfs3 maintainers here.

There is no problem in suppressing the output of any messages during mounting, like:

Messages like this:

fs/ntfs3/inode.c +0 -1
index eb7a8c9fba01..8cc94a6a97ed 100644
@@ -424,7 +424,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
     if (names != le16_to_cpu(rec->hard_links)) {
         /* Correct minor error on the fly. Do not mark inode as dirty. */
-        ntfs_inode_warn(inode, "Correct links count -> %u.", names);
         rec->hard_links = cpu_to_le16(names);
         ni->mi.dirty = true;
     }
 
Quoted 4 lines

can also be suppressed for the sake of seamless transition from a remote NTFS driver.
However, I believe that file system corrections should be reported to the user.

A colleague of mine also tracked down a failed boot to the removal of the ntfs driver and reported seeing similar warnings with the ntfs3 driver.

We're both accessing an NTFS partition on a Windows on Arm device, but it makes you wonder whether these warnings (corrections) are correct or indicative of a problem in the driver?

Johan

#22 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 25 lines · Johan Hovold
Quoted 14 lines · Konstantin Komarov

Messages like this:

fs/ntfs3/inode.c +0 -1
index eb7a8c9fba01..8cc94a6a97ed 100644
@@ -424,7 +424,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
     if (names != le16_to_cpu(rec->hard_links)) {
         /* Correct minor error on the fly. Do not mark inode as dirty. */
-        ntfs_inode_warn(inode, "Correct links count -> %u.", names);
         rec->hard_links = cpu_to_le16(names);
         ni->mi.dirty = true;
     }
 
Quoted 4 lines

can also be suppressed for the sake of seamless transition from a remote NTFS driver.
However, I believe that file system corrections should be reported to the user.

A colleague of mine also tracked down a failed boot to the removal of the ntfs driver and reported seeing similar warnings with the ntfs3 driver.

Quoted 3 lines

We're both accessing an NTFS partition on a Windows on Arm device, but it makes you wonder whether these warnings (corrections) are correct or indicative of a problem in the driver?

This is what I see with a recursive ls on that partition (I've added rec->hard_links in parentheses):

[   38.287555] ntfs3: nvme0n1p3: ino=2e1e7, Correct links count -> 1 (2).
[   38.288593] ntfs3: nvme0n1p3: ino=75ff, Correct links count -> 1 (2).
[   38.289887] ntfs3: nvme0n1p3: ino=1b4e1, Correct links count -> 1 (2).
[   38.290144] ntfs3: nvme0n1p3: ino=78c6, Correct links count -> 1 (2).
[   38.291313] ntfs3: nvme0n1p3: ino=8781b, Correct links count -> 1 (2).
[   38.291823] ntfs3: nvme0n1p3: ino=8781e, Correct links count -> 1 (2).
[   38.292289] ntfs3: nvme0n1p3: ino=87820, Correct links count -> 1 (2).
[   38.292978] ntfs3: nvme0n1p3: ino=87823, Correct links count -> 1 (2).
[   38.300531] ntfs3: nvme0n1p3: ino=a324, Correct links count -> 1 (2).
[   38.312235] ntfs3: nvme0n1p3: ino=882c3, Correct links count -> 1 (2).
[   43.286846] ntfs3: 5479 callbacks suppressed
[   43.286856] ntfs3: nvme0n1p3: ino=14aa, Correct links count -> 1 (2).
[   43.286998] ntfs3: nvme0n1p3: ino=14ac, Correct links count -> 1 (2).
[   43.287194] ntfs3: nvme0n1p3: ino=12cc2, Correct links count -> 1 (2).
[   43.287386] ntfs3: nvme0n1p3: ino=12ccd, Correct links count -> 1 (2).
[   43.287576] ntfs3: nvme0n1p3: ino=12d15, Correct links count -> 1 (2).
[   43.287667] ntfs3: nvme0n1p3: ino=12d19, Correct links count -> 1 (2).
[   43.287877] ntfs3: nvme0n1p3: ino=12d3a, Correct links count -> 1 (2).
[   43.288051] ntfs3: nvme0n1p3: ino=12d88, Correct links count -> 1 (2).
[   43.288265] ntfs3: nvme0n1p3: ino=874, Correct links count -> 1 (2).
[   43.288326] ntfs3: nvme0n1p3: ino=875, Correct links count -> 1 (2).
[   48.288211] ntfs3: 7735 callbacks suppressed
[   48.288220] ntfs3: nvme0n1p3: ino=33391, Correct links count -> 1 (2).
[   48.289115] ntfs3: nvme0n1p3: ino=347a4, Correct links count -> 1 (2).
[   48.290214] ntfs3: nvme0n1p3: ino=3553f, Correct links count -> 1 (2).
[   48.291193] ntfs3: nvme0n1p3: ino=35793, Correct links count -> 1 (2).
[   48.292818] ntfs3: nvme0n1p3: ino=358a0, Correct links count -> 1 (2).
[   48.293529] ntfs3: nvme0n1p3: ino=38f54, Correct links count -> 1 (2).
[   48.293901] ntfs3: nvme0n1p3: ino=391f6, Correct links count -> 1 (2).
[   48.294303] ntfs3: nvme0n1p3: ino=39324, Correct links count -> 1 (2).
[   48.294729] ntfs3: nvme0n1p3: ino=394a0, Correct links count -> 1 (2).
[   48.295063] ntfs3: nvme0n1p3: ino=3956a, Correct links count -> 1 (2).
[   53.289392] ntfs3: 11442 callbacks suppressed
[   53.289401] ntfs3: nvme0n1p3: ino=6293e, Correct links count -> 1 (2).
[   53.289972] ntfs3: nvme0n1p3: ino=61df1, Correct links count -> 1 (2).
[   53.290241] ntfs3: nvme0n1p3: ino=61df3, Correct links count -> 1 (2).
[   53.290578] ntfs3: nvme0n1p3: ino=61f3b, Correct links count -> 1 (2).
[   53.290881] ntfs3: nvme0n1p3: ino=62025, Correct links count -> 1 (2).
[   53.291045] ntfs3: nvme0n1p3: ino=629af, Correct links count -> 1 (2).
[   53.291181] ntfs3: nvme0n1p3: ino=61e3c, Correct links count -> 1 (2).
[   53.291463] ntfs3: nvme0n1p3: ino=61e22, Correct links count -> 1 (2).
[   53.291743] ntfs3: nvme0n1p3: ino=62882, Correct links count -> 1 (2).
[   53.292099] ntfs3: nvme0n1p3: ino=61b3d, Correct links count -> 1 (2).
[   58.291790] ntfs3: 5373 callbacks suppressed
[   58.291799] ntfs3: nvme0n1p3: ino=6d5a5, Correct links count -> 1 (2).
[   58.292106] ntfs3: nvme0n1p3: ino=6d7f6, Correct links count -> 1 (2).
[   58.292372] ntfs3: nvme0n1p3: ino=6db43, Correct links count -> 1 (2).
[   58.292653] ntfs3: nvme0n1p3: ino=72557, Correct links count -> 1 (2).
[   58.293244] ntfs3: nvme0n1p3: ino=728d8, Correct links count -> 1 (2).
[   58.294306] ntfs3: nvme0n1p3: ino=72c6e, Correct links count -> 1 (2).
[   58.294944] ntfs3: nvme0n1p3: ino=72ff7, Correct links count -> 1 (2).
[   58.295666] ntfs3: nvme0n1p3: ino=73271, Correct links count -> 1 (2).
[   58.296281] ntfs3: nvme0n1p3: ino=735fd, Correct links count -> 1 (2).
[   58.296991] ntfs3: nvme0n1p3: ino=73880, Correct links count -> 1 (2).
[   63.295009] ntfs3: 13921 callbacks suppressed
[   63.295017] ntfs3: nvme0n1p3: ino=6be65, Correct links count -> 1 (2).
[   63.295902] ntfs3: nvme0n1p3: ino=6c08e, Correct links count -> 1 (2).
[   63.296252] ntfs3: nvme0n1p3: ino=6c2e3, Correct links count -> 1 (2).
[   63.297581] ntfs3: nvme0n1p3: ino=6c610, Correct links count -> 1 (2).
[   63.298321] ntfs3: nvme0n1p3: ino=6c7f9, Correct links count -> 1 (2).
[   63.298730] ntfs3: nvme0n1p3: ino=6cb24, Correct links count -> 1 (2).
[   63.299079] ntfs3: nvme0n1p3: ino=6ceda, Correct links count -> 1 (2).
[   63.299408] ntfs3: nvme0n1p3: ino=6d288, Correct links count -> 1 (2).
[   63.299727] ntfs3: nvme0n1p3: ino=6d533, Correct links count -> 1 (2).
[   63.300080] ntfs3: nvme0n1p3: ino=6d77b, Correct links count -> 1 (2).
[   68.299457] ntfs3: 8228 callbacks suppressed
[   68.299467] ntfs3: nvme0n1p3: ino=3e248, Correct links count -> 1 (2).
[   68.301391] ntfs3: nvme0n1p3: ino=5d7b7, Correct links count -> 1 (2).
[   68.302440] ntfs3: nvme0n1p3: ino=5853d, Correct links count -> 1 (2).
[   68.303123] ntfs3: nvme0n1p3: ino=3ca2e, Correct links count -> 1 (2).
[   68.303722] ntfs3: nvme0n1p3: ino=59a98, Correct links count -> 1 (2).
[   68.304292] ntfs3: nvme0n1p3: ino=59a9b, Correct links count -> 1 (2).
[   68.304981] ntfs3: nvme0n1p3: ino=59a9e, Correct links count -> 1 (2).
[   68.305629] ntfs3: nvme0n1p3: ino=59aa1, Correct links count -> 1 (2).
[   68.306120] ntfs3: nvme0n1p3: ino=3214f, Correct links count -> 1 (2).
[   68.306539] ntfs3: nvme0n1p3: ino=2077b, Correct links count -> 1 (2).
[   73.302727] ntfs3: 8502 callbacks suppressed
[   73.302736] ntfs3: nvme0n1p3: ino=5aa99, Correct links count -> 1 (2).
[   73.303992] ntfs3: nvme0n1p3: ino=50ee9, Correct links count -> 1 (2).
[   73.304744] ntfs3: nvme0n1p3: ino=20420, Correct links count -> 1 (2).
[   73.305080] ntfs3: nvme0n1p3: ino=258c, Correct links count -> 1 (2).
[   73.305470] ntfs3: nvme0n1p3: ino=5a30d, Correct links count -> 1 (2).
[   73.307283] ntfs3: nvme0n1p3: ino=5a54c, Correct links count -> 1 (2).
[   73.307890] ntfs3: nvme0n1p3: ino=5c9de, Correct links count -> 1 (2).
[   73.308495] ntfs3: nvme0n1p3: ino=3d82d, Correct links count -> 1 (2).
[   73.309581] ntfs3: nvme0n1p3: ino=3d839, Correct links count -> 1 (2).
[   73.310016] ntfs3: nvme0n1p3: ino=25c77, Correct links count -> 1 (2).
[   78.304861] ntfs3: 11462 callbacks suppressed
[   78.304868] ntfs3: nvme0n1p3: ino=5c714, Correct links count -> 1 (2).
[   78.305579] ntfs3: nvme0n1p3: ino=57d11, Correct links count -> 1 (2).
[   78.306151] ntfs3: nvme0n1p3: ino=5842d, Correct links count -> 1 (2).
[   78.307412] ntfs3: nvme0n1p3: ino=34e6, Correct links count -> 1 (3).
[   78.307843] ntfs3: nvme0n1p3: ino=5bb23, Correct links count -> 1 (2).
[   78.308509] ntfs3: nvme0n1p3: ino=5c722, Correct links count -> 1 (2).
[   78.310018] ntfs3: nvme0n1p3: ino=5d761, Correct links count -> 1 (2).
[   78.310717] ntfs3: nvme0n1p3: ino=33d18, Correct links count -> 1 (3).
[   78.311179] ntfs3: nvme0n1p3: ino=5d75b, Correct links count -> 1 (3).
[   78.311605] ntfs3: nvme0n1p3: ino=5c708, Correct links count -> 1 (3).

Johan

#23 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Hi,

Had a look at ntfs3 code and it is corrupting your volume. Every such message you are seeing is damaging a file or directory on your volume.

I would personally suggest you modify your /etc/fstab to mount read-only. If it is getting simple things like this wrong who knows what else it is doing incorrect...

Best regards,

Anton

Quoted 135 lines · Johan Hovold
Quoted 25 lines · Johan Hovold
Quoted 14 lines · Konstantin Komarov

Messages like this:

fs/ntfs3/inode.c +0 -1
index eb7a8c9fba01..8cc94a6a97ed 100644
@@ -424,7 +424,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
    if (names != le16_to_cpu(rec->hard_links)) {
        /* Correct minor error on the fly. Do not mark inode as dirty. */
-        ntfs_inode_warn(inode, "Correct links count -> %u.", names);
        rec->hard_links = cpu_to_le16(names);
        ni->mi.dirty = true;
    }
 
Quoted 4 lines

can also be suppressed for the sake of seamless transition from a remote NTFS driver.
However, I believe that file system corrections should be reported to the user.

A colleague of mine also tracked down a failed boot to the removal of the ntfs driver and reported seeing similar warnings with the ntfs3 driver.

Quoted 3 lines

We're both accessing an NTFS partition on a Windows on Arm device, but it makes you wonder whether these warnings (corrections) are correct or indicative of a problem in the driver?

This is what I see with a recursive ls on that partition (I've added rec->hard_links in parentheses):

[   38.287555] ntfs3: nvme0n1p3: ino=2e1e7, Correct links count -> 1 (2).
[   38.288593] ntfs3: nvme0n1p3: ino=75ff, Correct links count -> 1 (2).
[   38.289887] ntfs3: nvme0n1p3: ino=1b4e1, Correct links count -> 1 (2).
[   38.290144] ntfs3: nvme0n1p3: ino=78c6, Correct links count -> 1 (2).
[   38.291313] ntfs3: nvme0n1p3: ino=8781b, Correct links count -> 1 (2).
[   38.291823] ntfs3: nvme0n1p3: ino=8781e, Correct links count -> 1 (2).
[   38.292289] ntfs3: nvme0n1p3: ino=87820, Correct links count -> 1 (2).
[   38.292978] ntfs3: nvme0n1p3: ino=87823, Correct links count -> 1 (2).
[   38.300531] ntfs3: nvme0n1p3: ino=a324, Correct links count -> 1 (2).
[   38.312235] ntfs3: nvme0n1p3: ino=882c3, Correct links count -> 1 (2).
[   43.286846] ntfs3: 5479 callbacks suppressed
[   43.286856] ntfs3: nvme0n1p3: ino=14aa, Correct links count -> 1 (2).
[   43.286998] ntfs3: nvme0n1p3: ino=14ac, Correct links count -> 1 (2).
[   43.287194] ntfs3: nvme0n1p3: ino=12cc2, Correct links count -> 1 (2).
[   43.287386] ntfs3: nvme0n1p3: ino=12ccd, Correct links count -> 1 (2).
[   43.287576] ntfs3: nvme0n1p3: ino=12d15, Correct links count -> 1 (2).
[   43.287667] ntfs3: nvme0n1p3: ino=12d19, Correct links count -> 1 (2).
[   43.287877] ntfs3: nvme0n1p3: ino=12d3a, Correct links count -> 1 (2).
[   43.288051] ntfs3: nvme0n1p3: ino=12d88, Correct links count -> 1 (2).
[   43.288265] ntfs3: nvme0n1p3: ino=874, Correct links count -> 1 (2).
[   43.288326] ntfs3: nvme0n1p3: ino=875, Correct links count -> 1 (2).
[   48.288211] ntfs3: 7735 callbacks suppressed
[   48.288220] ntfs3: nvme0n1p3: ino=33391, Correct links count -> 1 (2).
[   48.289115] ntfs3: nvme0n1p3: ino=347a4, Correct links count -> 1 (2).
[   48.290214] ntfs3: nvme0n1p3: ino=3553f, Correct links count -> 1 (2).
[   48.291193] ntfs3: nvme0n1p3: ino=35793, Correct links count -> 1 (2).
[   48.292818] ntfs3: nvme0n1p3: ino=358a0, Correct links count -> 1 (2).
[   48.293529] ntfs3: nvme0n1p3: ino=38f54, Correct links count -> 1 (2).
[   48.293901] ntfs3: nvme0n1p3: ino=391f6, Correct links count -> 1 (2).
[   48.294303] ntfs3: nvme0n1p3: ino=39324, Correct links count -> 1 (2).
[   48.294729] ntfs3: nvme0n1p3: ino=394a0, Correct links count -> 1 (2).
[   48.295063] ntfs3: nvme0n1p3: ino=3956a, Correct links count -> 1 (2).
[   53.289392] ntfs3: 11442 callbacks suppressed
[   53.289401] ntfs3: nvme0n1p3: ino=6293e, Correct links count -> 1 (2).
[   53.289972] ntfs3: nvme0n1p3: ino=61df1, Correct links count -> 1 (2).
[   53.290241] ntfs3: nvme0n1p3: ino=61df3, Correct links count -> 1 (2).
[   53.290578] ntfs3: nvme0n1p3: ino=61f3b, Correct links count -> 1 (2).
[   53.290881] ntfs3: nvme0n1p3: ino=62025, Correct links count -> 1 (2).
[   53.291045] ntfs3: nvme0n1p3: ino=629af, Correct links count -> 1 (2).
[   53.291181] ntfs3: nvme0n1p3: ino=61e3c, Correct links count -> 1 (2).
[   53.291463] ntfs3: nvme0n1p3: ino=61e22, Correct links count -> 1 (2).
[   53.291743] ntfs3: nvme0n1p3: ino=62882, Correct links count -> 1 (2).
[   53.292099] ntfs3: nvme0n1p3: ino=61b3d, Correct links count -> 1 (2).
[   58.291790] ntfs3: 5373 callbacks suppressed
[   58.291799] ntfs3: nvme0n1p3: ino=6d5a5, Correct links count -> 1 (2).
[   58.292106] ntfs3: nvme0n1p3: ino=6d7f6, Correct links count -> 1 (2).
[   58.292372] ntfs3: nvme0n1p3: ino=6db43, Correct links count -> 1 (2).
[   58.292653] ntfs3: nvme0n1p3: ino=72557, Correct links count -> 1 (2).
[   58.293244] ntfs3: nvme0n1p3: ino=728d8, Correct links count -> 1 (2).
[   58.294306] ntfs3: nvme0n1p3: ino=72c6e, Correct links count -> 1 (2).
[   58.294944] ntfs3: nvme0n1p3: ino=72ff7, Correct links count -> 1 (2).
[   58.295666] ntfs3: nvme0n1p3: ino=73271, Correct links count -> 1 (2).
[   58.296281] ntfs3: nvme0n1p3: ino=735fd, Correct links count -> 1 (2).
[   58.296991] ntfs3: nvme0n1p3: ino=73880, Correct links count -> 1 (2).
[   63.295009] ntfs3: 13921 callbacks suppressed
[   63.295017] ntfs3: nvme0n1p3: ino=6be65, Correct links count -> 1 (2).
[   63.295902] ntfs3: nvme0n1p3: ino=6c08e, Correct links count -> 1 (2).
[   63.296252] ntfs3: nvme0n1p3: ino=6c2e3, Correct links count -> 1 (2).
[   63.297581] ntfs3: nvme0n1p3: ino=6c610, Correct links count -> 1 (2).
[   63.298321] ntfs3: nvme0n1p3: ino=6c7f9, Correct links count -> 1 (2).
[   63.298730] ntfs3: nvme0n1p3: ino=6cb24, Correct links count -> 1 (2).
[   63.299079] ntfs3: nvme0n1p3: ino=6ceda, Correct links count -> 1 (2).
[   63.299408] ntfs3: nvme0n1p3: ino=6d288, Correct links count -> 1 (2).
[   63.299727] ntfs3: nvme0n1p3: ino=6d533, Correct links count -> 1 (2).
[   63.300080] ntfs3: nvme0n1p3: ino=6d77b, Correct links count -> 1 (2).
[   68.299457] ntfs3: 8228 callbacks suppressed
[   68.299467] ntfs3: nvme0n1p3: ino=3e248, Correct links count -> 1 (2).
[   68.301391] ntfs3: nvme0n1p3: ino=5d7b7, Correct links count -> 1 (2).
[   68.302440] ntfs3: nvme0n1p3: ino=5853d, Correct links count -> 1 (2).
[   68.303123] ntfs3: nvme0n1p3: ino=3ca2e, Correct links count -> 1 (2).
[   68.303722] ntfs3: nvme0n1p3: ino=59a98, Correct links count -> 1 (2).
[   68.304292] ntfs3: nvme0n1p3: ino=59a9b, Correct links count -> 1 (2).
[   68.304981] ntfs3: nvme0n1p3: ino=59a9e, Correct links count -> 1 (2).
[   68.305629] ntfs3: nvme0n1p3: ino=59aa1, Correct links count -> 1 (2).
[   68.306120] ntfs3: nvme0n1p3: ino=3214f, Correct links count -> 1 (2).
[   68.306539] ntfs3: nvme0n1p3: ino=2077b, Correct links count -> 1 (2).
[   73.302727] ntfs3: 8502 callbacks suppressed
[   73.302736] ntfs3: nvme0n1p3: ino=5aa99, Correct links count -> 1 (2).
[   73.303992] ntfs3: nvme0n1p3: ino=50ee9, Correct links count -> 1 (2).
[   73.304744] ntfs3: nvme0n1p3: ino=20420, Correct links count -> 1 (2).
[   73.305080] ntfs3: nvme0n1p3: ino=258c, Correct links count -> 1 (2).
[   73.305470] ntfs3: nvme0n1p3: ino=5a30d, Correct links count -> 1 (2).
[   73.307283] ntfs3: nvme0n1p3: ino=5a54c, Correct links count -> 1 (2).
[   73.307890] ntfs3: nvme0n1p3: ino=5c9de, Correct links count -> 1 (2).
[   73.308495] ntfs3: nvme0n1p3: ino=3d82d, Correct links count -> 1 (2).
[   73.309581] ntfs3: nvme0n1p3: ino=3d839, Correct links count -> 1 (2).
[   73.310016] ntfs3: nvme0n1p3: ino=25c77, Correct links count -> 1 (2).
[   78.304861] ntfs3: 11462 callbacks suppressed
[   78.304868] ntfs3: nvme0n1p3: ino=5c714, Correct links count -> 1 (2).
[   78.305579] ntfs3: nvme0n1p3: ino=57d11, Correct links count -> 1 (2).
[   78.306151] ntfs3: nvme0n1p3: ino=5842d, Correct links count -> 1 (2).
[   78.307412] ntfs3: nvme0n1p3: ino=34e6, Correct links count -> 1 (3).
[   78.307843] ntfs3: nvme0n1p3: ino=5bb23, Correct links count -> 1 (2).
[   78.308509] ntfs3: nvme0n1p3: ino=5c722, Correct links count -> 1 (2).
[   78.310018] ntfs3: nvme0n1p3: ino=5d761, Correct links count -> 1 (2).
[   78.310717] ntfs3: nvme0n1p3: ino=33d18, Correct links count -> 1 (3).
[   78.311179] ntfs3: nvme0n1p3: ino=5d75b, Correct links count -> 1 (3).
[   78.311605] ntfs3: nvme0n1p3: ino=5c708, Correct links count -> 1 (3).

Johan

Signature · 2 lines
Anton Altaparmakov <anton at tuxera.com> (replace at with @)
Lead in File System Development, Tuxera Inc., http://www.tuxera.com/
#24 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 2 lines · Anton Altaparmakov

Had a look at ntfs3 code and it is corrupting your volume. Every such message you are seeing is damaging a file or directory on your volume.

That's what I feared, thanks for confirming.

Quoted 3 lines

I would personally suggest you modify your /etc/fstab to mount read-only. If it is getting simple things like this wrong who knows what else it is doing incorrect...

I fully agree and that's partly also why I asked Christian to make sure that the alias fs type is always mounted RO.

But it seems we have a bigger problem here and should just restore the old ntfs driver for now.

Johan

#25 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 11 lines · Johan Hovold
Quoted 2 lines · Anton Altaparmakov

Had a look at ntfs3 code and it is corrupting your volume. Every such message you are seeing is damaging a file or directory on your volume.

That's what I feared, thanks for confirming.

Quoted 3 lines

I would personally suggest you modify your /etc/fstab to mount read-only. If it is getting simple things like this wrong who knows what else it is doing incorrect...

Quoted 5 lines

I fully agree and that's partly also why I asked Christian to make sure that the alias fs type is always mounted RO.

But it seems we have a bigger problem here and should just restore the old ntfs driver for now.

Hey Linus,

A brief summary:

The removal of the legacy ntfs driver has caused a regression for some users that rely on the legacy ntfs driver to be available during boot. The affected user here is Johan (Cc'ed). In addition he's seeing dmesg warnings that he didn't see before with the legacy ntfs driver.

I see the following options to resolve this:

(1) Since the ntfs3 driver is supposed to serve as a drop-in replacement for the legacy ntfs driver we should to it the same way we did it for ext3 and ext4 where ext4 registers itself also for the ext3 driver. In other words, we would register ntfs3 as ntfs3 filesystem type and as legacy ntfs filesystem type. To make it fully compatible we also need to make sure it's persistently mounted read-only.

(2) We revert the ntfs driver (for now) as we are in the middle of the cycle.

If you decide (2) is the right way to go then I would suggest we try the removal once more next cycle with the proposed change in (1) in place if you're up for that.

Christian

#26 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 6 lines · Christian Brauner

(1) Since the ntfs3 driver is supposed to serve as a drop-in replacement for the legacy ntfs driver we should to it the same way we did it for ext3 and ext4 where ext4 registers itself also for the ext3 driver. In other words, we would register ntfs3 as ntfs3 filesystem type and as legacy ntfs filesystem type.

I think that if just registering it under the same name solves the immediate issue, that's the one we should just go for.

Quoted 2 lines

To make it fully compatible
we also need to make sure it's persistently mounted read-only.

My reaction to that is "only if it turns out we really need to".

It sounds unlikely that somebody has an old ntfs setup and then tries to mount things rw which didn't use to work and things go sideways if that then suddenly works.

But "unlikely" isn't "impossible", of course - it's just that I'd suggest we actually wait for that report to happen and ask what the heck they were doing and why they were doing that...

Linus

#27 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 10 lines · Linus Torvalds
Quoted 6 lines · Christian Brauner

(1) Since the ntfs3 driver is supposed to serve as a drop-in replacement for the legacy ntfs driver we should to it the same way we did it for ext3 and ext4 where ext4 registers itself also for the ext3 driver. In other words, we would register ntfs3 as ntfs3 filesystem type and as legacy ntfs filesystem type.

I think that if just registering it under the same name solves the immediate issue, that's the one we should just go for.

I agree.

Quoted 4 lines
Quoted 2 lines

To make it fully compatible
we also need to make sure it's persistently mounted read-only.

My reaction to that is "only if it turns out we really need to".

Unfortunately, we do. It seems that ntfs3 has some bugs, and (according to Anton, the ntfs-classic maintainer) it's actually corrupting perfectly good filesystems. I'd expected that this kind of bug would have been shaken out by now (it's been in the kernel for over two and a half years!) but it seems like people just haven't been using it.

#28 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

In reply to Linus Torvalds
Quoted 10 lines · Linus Torvalds
Quoted 6 lines · Christian Brauner

(1) Since the ntfs3 driver is supposed to serve as a drop-in replacement for the legacy ntfs driver we should to it the same way we did it for ext3 and ext4 where ext4 registers itself also for the ext3 driver. In other words, we would register ntfs3 as ntfs3 filesystem type and as legacy ntfs filesystem type.

I think that if just registering it under the same name solves the immediate issue, that's the one we should just go for.

I also tend to agree, but...

Quoted 5 lines
Quoted 2 lines

To make it fully compatible
we also need to make sure it's persistently mounted read-only.

My reaction to that is "only if it turns out we really need to".

Quoted 7 lines

It sounds unlikely that somebody has an old ntfs setup and then tries to mount things rw which didn't use to work and things go sideways if that then suddenly works.

But "unlikely" isn't "impossible", of course - it's just that I'd suggest we actually wait for that report to happen and ask what the heck they were doing and why they were doing that...

I think the "ntfs" alias must always be mounted read-only because you can currently have an fstab entry which does not specify "ro" and this mount would suddenly become writeable when updating to 6.9 (possibly by a non-privileged user, etc).

We also need to do something about the ntfs3 driver spamming the logs about broken corrections also when mounted read-only even if it doesn't eat your filesystem then.

And it seems write-support should be disabled in the driver by default until someone has tracked down why listing a directory can currently corrupt your filesystem.

Johan

#29 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 5 lines · Johan Hovold

I think the "ntfs" alias must always be mounted read-only because you can currently have an fstab entry which does not specify "ro" and this mount would suddenly become writeable when updating to 6.9 (possibly by a non-privileged user, etc).

Well, it would be fairly easy to do particularly if we just do it for the old legacy case.

Of course, even the legacy case had that CONFIG_NTFS_RW option, so people who depended on _that_ would want to be able to remount...

Linus

#30 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 7 lines · Linus Torvalds
Quoted 5 lines · Johan Hovold

I think the "ntfs" alias must always be mounted read-only because you can currently have an fstab entry which does not specify "ro" and this mount would suddenly become writeable when updating to 6.9 (possibly by a non-privileged user, etc).

Quoted 5 lines

Well, it would be fairly easy to do particularly if we just do it for the old legacy case.

Of course, even the legacy case had that CONFIG_NTFS_RW option, so people who depended on _that_ would want to be able to remount...

Ah, right, I forgot about CONFIG_NTFS_RW as I've never enabled it.

Judging from the now removed Kconfig entry perhaps not that many people did:

The only supported operation is overwriting existing files, without changing the file length. No file or directory creation, deletion or renaming is possible.

but I guess it still makes my argument above mostly moot.

At least if we disable write support in ntfs3 by default for now...

Johan

#31 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 19 lines · Johan Hovold
Quoted 7 lines · Linus Torvalds
Quoted 5 lines · Johan Hovold

I think the "ntfs" alias must always be mounted read-only because you can currently have an fstab entry which does not specify "ro" and this mount would suddenly become writeable when updating to 6.9 (possibly by a non-privileged user, etc).

Quoted 5 lines

Well, it would be fairly easy to do particularly if we just do it for the old legacy case.

Of course, even the legacy case had that CONFIG_NTFS_RW option, so people who depended on _that_ would want to be able to remount...

Ah, right, I forgot about CONFIG_NTFS_RW as I've never enabled it.

Judging from the now removed Kconfig entry perhaps not that many people did:

Quoted 7 lines

The only supported operation is overwriting existing files, without changing the file length. No file or directory creation, deletion or renaming is possible.

but I guess it still makes my argument above mostly moot.

At least if we disable write support in ntfs3 by default for now...

I think we can disable write support in ntfs3 for now. I've picked up the patch to make ntfs3 serve I sent some time ago that Johan tested now.

The only thing left is to disable write support for ntfs3 as legacy ntfs driver for now. I took a stab at this. The following two patches I'm appending _should_ be enough iiuc. Johan, please take a look and please test.

This is also available on:

https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git vfs.fixes

So feel free to pull from there and look there as well.

From 1a2af5ca9b66bd3d4040f9987c78faff128e552f Mon Sep 17 00:00:00 2001
From: Christian Brauner <[email protected]>
Date: Tue, 16 Apr 2024 12:08:51 +0200
Subject: [PATCH 1/2] ntfs3: enforce read-only when used as legacy ntfs driver

Ensure that ntfs3 is mounted read-only when it is used to provide the legacy ntfs driver.

---
 fs/ntfs3/ntfs_fs.h |  2 ++
 fs/ntfs3/super.c   | 34 ++++++++++++++++++++++++++++++----
 2 files changed, 32 insertions(+), 4 deletions(-)
fs/ntfs3/ntfs_fs.h +2 -0
index 79356fd29a14..184c8bc76b92 100644
@@ -1154,4 +1154,6 @@ static inline void le64_sub_cpu(__le64 *var, u64 val)
 	*var = cpu_to_le64(le64_to_cpu(*var) - val);
 }
 
+bool is_legacy_ntfs(struct super_block *sb);
+
 #endif /* _LINUX_NTFS3_NTFS_FS_H */
fs/ntfs3/super.c +30 -4
index 8d2e51bae2cb..d7f3e03a4010 100644
@@ -408,6 +408,12 @@ static int ntfs_fs_reconfigure(struct fs_context *fc)
 	struct ntfs_mount_options *new_opts = fc->fs_private;
 	int ro_rw;
 
+	/* If ntfs3 is used as legacy ntfs enforce read-only mode. */
+	if (is_legacy_ntfs(sb)) {
+		fc->sb_flags |= SB_RDONLY;
+		goto out;
+	}
+
 	ro_rw = sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY);
 	if (ro_rw && (sbi->flags & NTFS_FLAGS_NEED_REPLAY)) {
 		errorf(fc,
@@ -427,8 +433,6 @@ static int ntfs_fs_reconfigure(struct fs_context *fc)
 			fc,
 			"ntfs3: Cannot use different iocharset when remounting!");
 
-	sync_filesystem(sb);
-
 	if (ro_rw && (sbi->volume.flags & VOLUME_FLAG_DIRTY) &&
 	    !new_opts->force) {
 		errorf(fc,
@@ -436,6 +440,8 @@ static int ntfs_fs_reconfigure(struct fs_context *fc)
 		return -EINVAL;
 	}
 
+out:
+	sync_filesystem(sb);
 	swap(sbi->options, fc->fs_private);
 
 	return 0;
@@ -1730,7 +1736,7 @@ static const struct fs_context_operations ntfs_context_ops = {
  * This will called when mount/remount. We will first initialize
  * options so that if remount we can use just that.
  */
-static int ntfs_init_fs_context(struct fs_context *fc)
+static int __ntfs_init_fs_context(struct fs_context *fc)
 {
 	struct ntfs_mount_options *opts;
 	struct ntfs_sb_info *sbi;
@@ -1778,6 +1784,11 @@ static int ntfs_init_fs_context(struct fs_context *fc)
 	return -ENOMEM;
 }
 
+static int ntfs_init_fs_context(struct fs_context *fc)
+{
+	return __ntfs_init_fs_context(fc);
+}
+
 static void ntfs3_kill_sb(struct super_block *sb)
 {
 	struct ntfs_sb_info *sbi = sb->s_fs_info;
@@ -1800,10 +1811,20 @@ static struct file_system_type ntfs_fs_type = {
 };
 
 #if IS_ENABLED(CONFIG_NTFS_FS)
+static int ntfs_legacy_init_fs_context(struct fs_context *fc)
+{
+	int ret;
+
+	ret = __ntfs_init_fs_context(fc);
+	/* If ntfs3 is used as legacy ntfs enforce read-only mode. */
+	fc->sb_flags |= SB_RDONLY;
+	return ret;
+}
+
 static struct file_system_type ntfs_legacy_fs_type = {
 	.owner			= THIS_MODULE,
 	.name			= "ntfs",
-	.init_fs_context	= ntfs_init_fs_context,
+	.init_fs_context	= ntfs_legacy_init_fs_context,
 	.parameters		= ntfs_fs_parameters,
 	.kill_sb		= ntfs3_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
@@ -1821,9 +1842,14 @@ static inline void unregister_as_ntfs_legacy(void)
 {
 	unregister_filesystem(&ntfs_legacy_fs_type);
 }
+bool is_legacy_ntfs(struct super_block *sb)
+{
+	return sb->s_type == &ntfs_legacy_fs_type;
+}
 #else
 static inline void register_as_ntfs_legacy(void) {}
 static inline void unregister_as_ntfs_legacy(void) {}
+bool is_legacy_ntfs(struct super_block *sb) { return false; }
 #endif
 
 
Signature · 1 line
2.43.0

From 5e646f22f0e9d5268d902dfd33a39154c0b5f578 Mon Sep 17 00:00:00 2001
From: Christian Brauner <[email protected]>
Date: Tue, 16 Apr 2024 12:20:50 +0200
Subject: [PATCH 2/2] ntfs3: add legacy ntfs file operations

To ensure that ioctl()s can't be used to circumvent write restrictions.

---
 fs/ntfs3/dir.c     |  7 +++++++
 fs/ntfs3/file.c    |  8 ++++++++
 fs/ntfs3/inode.c   | 20 ++++++++++++++++----
 fs/ntfs3/ntfs_fs.h |  2 ++
 4 files changed, 33 insertions(+), 4 deletions(-)
fs/ntfs3/dir.c +7 -0
index 5cf3d9decf64..263635199b60 100644
@@ -616,4 +616,11 @@ const struct file_operations ntfs_dir_operations = {
 	.compat_ioctl   = ntfs_compat_ioctl,
 #endif
 };
+
+const struct file_operations ntfs_legacy_dir_operations = {
+	.llseek		= generic_file_llseek,
+	.read		= generic_read_dir,
+	.iterate_shared	= ntfs_readdir,
+	.open		= ntfs_file_open,
+};
 // clang-format on
fs/ntfs3/file.c +8 -0
index 5418662c80d8..b73969e05052 100644
@@ -1236,4 +1236,12 @@ const struct file_operations ntfs_file_operations = {
 	.fallocate	= ntfs_fallocate,
 	.release	= ntfs_file_release,
 };
+
+const struct file_operations ntfs_legacy_file_operations = {
+	.llseek		= generic_file_llseek,
+	.read_iter	= ntfs_file_read_iter,
+	.splice_read	= ntfs_file_splice_read,
+	.open		= ntfs_file_open,
+	.release	= ntfs_file_release,
+};
 // clang-format on
fs/ntfs3/inode.c +16 -4
index eb7a8c9fba01..d273eda1cf45 100644
@@ -440,7 +440,10 @@ static struct inode *ntfs_read_mft(struct inode *inode,
 		 * Usually a hard links to directories are disabled.
 		 */
 		inode->i_op = &ntfs_dir_inode_operations;
-		inode->i_fop = &ntfs_dir_operations;
+		if (is_legacy_ntfs(inode->i_sb))
+			inode->i_fop = &ntfs_legacy_dir_operations;
+		else
+			inode->i_fop = &ntfs_dir_operations;
 		ni->i_valid = 0;
 	} else if (S_ISLNK(mode)) {
 		ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
@@ -450,7 +453,10 @@ static struct inode *ntfs_read_mft(struct inode *inode,
 	} else if (S_ISREG(mode)) {
 		ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
 		inode->i_op = &ntfs_file_inode_operations;
-		inode->i_fop = &ntfs_file_operations;
+		if (is_legacy_ntfs(inode->i_sb))
+			inode->i_fop = &ntfs_legacy_file_operations;
+		else
+			inode->i_fop = &ntfs_file_operations;
 		inode->i_mapping->a_ops = is_compressed(ni) ? &ntfs_aops_cmpr :
 							      &ntfs_aops;
 		if (ino != MFT_REC_MFT)
@@ -1614,7 +1620,10 @@ struct inode *ntfs_create_inode(struct mnt_idmap *idmap, struct inode *dir,
 
 	if (S_ISDIR(mode)) {
 		inode->i_op = &ntfs_dir_inode_operations;
-		inode->i_fop = &ntfs_dir_operations;
+		if (is_legacy_ntfs(inode->i_sb))
+			inode->i_fop = &ntfs_legacy_dir_operations;
+		else
+			inode->i_fop = &ntfs_dir_operations;
 	} else if (S_ISLNK(mode)) {
 		inode->i_op = &ntfs_link_inode_operations;
 		inode->i_fop = NULL;
@@ -1623,7 +1632,10 @@ struct inode *ntfs_create_inode(struct mnt_idmap *idmap, struct inode *dir,
 		inode_nohighmem(inode);
 	} else if (S_ISREG(mode)) {
 		inode->i_op = &ntfs_file_inode_operations;
-		inode->i_fop = &ntfs_file_operations;
+		if (is_legacy_ntfs(inode->i_sb))
+			inode->i_fop = &ntfs_legacy_file_operations;
+		else
+			inode->i_fop = &ntfs_file_operations;
 		inode->i_mapping->a_ops = is_compressed(ni) ? &ntfs_aops_cmpr :
 							      &ntfs_aops;
 		init_rwsem(&ni->file.run_lock);
fs/ntfs3/ntfs_fs.h +2 -0
index 184c8bc76b92..5f4d288c6adf 100644
@@ -493,6 +493,7 @@ struct inode *dir_search_u(struct inode *dir, const struct cpu_str *uni,
 			   struct ntfs_fnd *fnd);
 bool dir_is_empty(struct inode *dir);
 extern const struct file_operations ntfs_dir_operations;
+extern const struct file_operations ntfs_legacy_dir_operations;
 
 /* Globals from file.c */
 int ntfs_getattr(struct mnt_idmap *idmap, const struct path *path,
@@ -507,6 +508,7 @@ long ntfs_compat_ioctl(struct file *filp, u32 cmd, unsigned long arg);
 extern const struct inode_operations ntfs_special_inode_operations;
 extern const struct inode_operations ntfs_file_inode_operations;
 extern const struct file_operations ntfs_file_operations;
+extern const struct file_operations ntfs_legacy_file_operations;
 
 /* Globals from frecord.c */
 void ni_remove_mi(struct ntfs_inode *ni, struct mft_inode *mi);
Signature · 1 line
2.43.0
#32 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 11 lines · Christian Brauner
Quoted 9 lines · Johan Hovold

Ah, right, I forgot about CONFIG_NTFS_RW as I've never enabled it.

Judging from the now removed Kconfig entry perhaps not that many people did:

The only supported operation is overwriting existing files, without changing the file length. No file or directory creation, deletion or renaming is possible.

Quoted 7 lines
Quoted 3 lines

but I guess it still makes my argument above mostly moot.

At least if we disable write support in ntfs3 by default for now...

I think we can disable write support in ntfs3 for now. I've picked up the patch to make ntfs3 serve I sent some time ago that Johan tested now.

Note that I actually meant that write support should be disabled completely in ntfs3 for now.

After this first encounter I have zero confidence in that driver and pushing people towards using it (by removing the old, read-only one) is just gonna result in further corrupted filesystems. At least make sure it can't modify anything by default and mark write-support as experimental and broken or something as that's apparently what it is.

Quoted 4 lines

The only thing left is to disable write support for ntfs3 as legacy ntfs driver for now. I took a stab at this. The following two patches I'm appending _should_ be enough iiuc. Johan, please take a look and please test.

I skimmed them and gave them a quick spin. It seems that not specifying either "ro" or "rw" in fstab now results in a ro mount, but I can still specify "rw" explicitly (in fstab or command line) and end up with:

	/dev/nvme0n1p3 on /mnt/windows type ntfs (rw,relatime,uid=0,gid=0,iocharset=iso8859-1)

For obvious reasons, I did not dare listing the root directory or write anything, but it looks like it's not read-only.

Using just my naive temporary hack from yesterday:

fs/ntfs3/super.c +3 -0
index 8d2e51bae2cb..26be6c6d1032 100644
@@ -1177,6 +1177,9 @@ static int ntfs_fill_super(struct super_block *sb, struct fs_context *fc)
        sb->s_xattr = ntfs_xattr_handlers;
        sb->s_d_op = options->nocase ? &ntfs_dentry_ops : NULL;
 
+       ntfs_warn(sb, "ntfs3 driver is broken, mounting read only");
+       sb->s_flags |= SB_RDONLY;
+
        options->nls = ntfs_load_nls(options->nls_name);
        if (IS_ERR(options->nls)) {
                options->nls = NULL;

seems to prevent also explicit rw mounts (but judging from your patches it is not necessarily sufficient to prevent all modifications).

Johan

#33 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

In reply to Christian Brauner
Quoted 17 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

---
  fs/ntfs3/inode.c | 1 -
  1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -0
index eb7a8c9fba01..8cc94a6a97ed 100644

fs/ntfs3/inode.c static struct inode *ntfs_read_mft(struct inode *inode, · line 429

Quoted 7 lines
fs/ntfs3/inode.c +0 -1
@@ -424,7 +424,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
 
  	if (names != le16_to_cpu(rec->hard_links)) {
  		/* Correct minor error on the fly. Do not mark inode as dirty. */
-		ntfs_inode_warn(inode, "Correct links count -> %u.", names);
  		rec->hard_links = cpu_to_le16(names);
  		ni->mi.dirty = true;
  	}

Hello Christian, all,

The initial and true reason for multiple warnings is the disregard of short (DOS) names when counting hard links.

This patch should fixes this bug:
https://lore.kernel.org/ntfs3/[email protected]/T/#u

There is no longer a need to suppress the message itself.

#34 lore raw

Re: [PATCH 2/2] ntfs3: remove warning

Quoted 10 lines · Konstantin Komarov
Quoted 3 lines · Christian Brauner

This causes visible changes for users that rely on ntfs3 to serve as an alternative for the legacy ntfs driver. Print statements such as this should probably be made conditional on a debug config option or similar.

The initial and true reason for multiple warnings is the disregard of short (DOS) names when counting hard links.

This patch should fixes this bug:
https://lore.kernel.org/ntfs3/[email protected]/T/#u

As I just replied in that thread, I'm also seeing link counts being reduced from 3 to 1, that is, to not just be decremented by one due to the DOS name bug.

Are you sure there are no further issues here and that the patch is indeed correct?

I did not test the patch, which is white-space damaged, but if it addresses also the remaining warnings then the commit message would need to be updated as well.

Johan