[PATCH 0/4] Expose encoding information of volumes

[PATCH 0/4] Expose encoding information of volumes

[Gabriel Krisman Bertazi]

Daniel Tang reported [1] some inconsistencies in detecting filesystem
casefold support on shmem.  The problem is that we don't expose the
casefold arguments in /proc/mounts for shmem, requiring you to probe the
the filesystem to discover it.  More interestingly, it can be a problem
for programs that use /proc/mounts to remount filesystems, making them
to remount "case-sensitively'.  This apparently is a real use case in
linux utils, according to my favorite LLM.  shmem is fixed by patch 1.

This made me go look at how other casefolding filesystems expose this
information and found it to be very inconsistent.  f2fs is the only one
doing it properly over /sys/fs/<vol>/encoding.  Ext4 and ovl don't have
nice ways to expose it, so the remaining patches replicate the f2fs
interface for them.

I considered a /sys/fs interface for shmem as wel. It makes sense.  But
since shmem is the only filesystem where casefolding is a mount option,
it also makes sense for it to appear in /proc/mounts.  Considering it
doesn't have /sys/fs at the moment, it does fell like overkill to add it
just for this file.

ovl also doesn't have a /sys/fs/ entry.  But in its case, casefold is
derived from the lower/upper directories, so it cannot not be set in
/proc/mounts.  For this reason, I ended up plumbling /sys/fs/overlay and
wiring the encoding file just like ext4 and f2fs.

[1] https://lore.kernel.org/linux-fsdevel/5193732.31r3eYUQgx@daniel-desktop3/#

---
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jan Kara <jack@suse.cz>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Hugh Dickins <hughd@google.com>
Cc: Daniel Tang <danielzgtg.opensource@gmail.com>
Cc: linux-ext4@vger.kernel.org
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Amir Goldstein <amir73il@gmail.com>
CC: linux-unionfs@vger.kernel.org
CC: linux-mm@kvack.org

Gabriel Krisman Bertazi (4):
  shmem: Report casefold setting in mount options
  ext4: Expose the filesystem encoding through sysfs
  ovl: Add /sys/fs/overlay plumbing
  ovl: Expose the filesystem encoding through sysfs

 Documentation/ABI/testing/sysfs-fs-ext4    |  21 ++++
 Documentation/ABI/testing/sysfs-fs-overlay |  13 +++
 MAINTAINERS                                |   1 +
 fs/ext4/sysfs.c                            |  29 ++++++
 fs/overlayfs/Makefile                      |   2 +-
 fs/overlayfs/overlayfs.h                   |   5 +
 fs/overlayfs/ovl_entry.h                   |   7 ++
 fs/overlayfs/super.c                       |  26 ++++-
 fs/overlayfs/sysfs.c                       | 116 +++++++++++++++++++++
 mm/shmem.c                                 |  11 ++
 10 files changed, 227 insertions(+), 4 deletions(-)
 create mode 100644 Documentation/ABI/testing/sysfs-fs-overlay
 create mode 100644 fs/overlayfs/sysfs.c

-- 
2.55.0

[PATCH 1/4] shmem: Report casefold setting in mount options

[Gabriel Krisman Bertazi]

For shmem, casefolding is a mount option.  Report the current encoding
and casefolding flags in /proc/mounts, so userspace can easily check if
casefolding is supported withour relying on parsing dmesg or trying to
set +F.

Reported-by: Daniel Tang <danielzgtg.opensource@gmail.com>
Link: https://lore.kernel.org/linux-fsdevel/5193732.31r3eYUQgx@daniel-desktop3/#R
Assisted-by: LLM
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 mm/shmem.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/mm/shmem.c b/mm/shmem.c
index 897fa2b61346..2b93fcbf5127 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -4916,6 +4916,17 @@ static int shmem_show_options(struct seq_file *seq, struct dentry *root)
 	if (sbinfo->qlimits.grpquota_ihardlimit)
 		seq_printf(seq, ",grpquota_inode_hardlimit=%lld",
 			   sbinfo->qlimits.grpquota_ihardlimit);
+#endif
+#if IS_ENABLED(CONFIG_UNICODE)
+	if (root->d_sb->s_encoding) {
+		unsigned int version = root->d_sb->s_encoding->version;
+
+		seq_printf(seq, ",casefold=utf8-%u.%u.%u",
+			   unicode_major(version), unicode_minor(version),
+			   unicode_rev(version));
+		if (sb_has_strict_encoding(root->d_sb))
+			seq_puts(seq, ",strict_encoding");
+	}
 #endif
 	return 0;
 }
-- 
2.55.0

Re: [PATCH 1/4] shmem: Report casefold setting in mount options

[sashiko-bot]

> For shmem, casefolding is a mount option.  Report the current encoding
> and casefolding flags in /proc/mounts, so userspace can easily check if
> casefolding is supported withour relying on parsing dmesg or trying to
> set +F.
> 
> Reported-by: Daniel Tang <danielzgtg.opensource@gmail.com>
> Link: https://lore.kernel.org/linux-fsdevel/5193732.31r3eYUQgx@daniel-desktop3/#R
> Assisted-by: LLM
> Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260928213155.2612039-1-krisman@suse.de?part=1

Re: [PATCH 1/4] shmem: Report casefold setting in mount options

[Mohammed EL Kadiri]

On Mon, Sep 28, 2026 at 05:31:52PM -0400, Gabriel Krisman Bertazi wrote:
> For shmem, casefolding is a mount option.  Report the current encoding
> and casefolding flags in /proc/mounts

This works for me, on 90ddfbd19636 (akpm/mm-unstable), x86_64,
CONFIG_UNICODE=y:

  # mount -t tmpfs -o casefold tmpfs /t
  # grep /t /proc/mounts
  tmpfs /t tmpfs rw,relatime,casefold=utf8-12.1.0 0 0

Tested-by: Mohammed EL Kadiri <med08elkadiri@gmail.com>

Heads-up on one side effect, since mount(8) builds its remount options
from /proc/mounts: once casefold= is listed there, it gets passed back on
every remount.

  # mount -t tmpfs -o casefold tmpfs /t
  [  527.593309] tmpfs: Using encoding : utf8-12.1.0
  # mount -o remount,size=64M /t
  [  527.595716] tmpfs: Using encoding : utf8-12.1.0

That second line is shmem_parse_opt_casefold() running on a remount that
only asked to change size.  It calls utf8_load() each time, and nothing
frees the result - shmem_reconfigure() doesn't look at ctx->encoding and
shmem_free_fc() doesn't unload it.  50000 such remounts grew kmalloc-32 by
~50000 objects; a tmpfs without casefold stays flat.

The leak predates this patch - you can trigger it by passing casefold= on
a remount by hand - so I sent the fix as a separate patch:

https://lore.kernel.org/all/ba38f80f629fd093c77f3a49fdab7b71ee7bbc5f.1790687276.git.med08elkadiri@gmail.com/

Re: [PATCH 1/4] shmem: Report casefold setting in mount options

[Gabriel Krisman Bertazi]

Mohammed EL Kadiri <med08elkadiri@gmail.com> writes:

> On Mon, Sep 28, 2026 at 05:31:52PM -0400, Gabriel Krisman Bertazi wrote:
>> For shmem, casefolding is a mount option.  Report the current encoding
>> and casefolding flags in /proc/mounts
>
> This works for me, on 90ddfbd19636 (akpm/mm-unstable), x86_64,
> CONFIG_UNICODE=y:
>
>   # mount -t tmpfs -o casefold tmpfs /t
>   # grep /t /proc/mounts
>   tmpfs /t tmpfs rw,relatime,casefold=utf8-12.1.0 0 0
>
> Tested-by: Mohammed EL Kadiri <med08elkadiri@gmail.com>

Thanks!

>
> Heads-up on one side effect, since mount(8) builds its remount options
> from /proc/mounts: once casefold= is listed there, it gets passed back on
> every remount.
>
>   # mount -t tmpfs -o casefold tmpfs /t
>   [  527.593309] tmpfs: Using encoding : utf8-12.1.0
>   # mount -o remount,size=64M /t
>   [  527.595716] tmpfs: Using encoding : utf8-12.1.0

This is desired behavior, IMO.  You should not lose casefold during a
remount, at risk of botching your data/splitting unrelated files. Also
this is the reason why I didn't expose it directly in /proc/mounts for
ovl, it would be easier but break the remount.

> That second line is shmem_parse_opt_casefold() running on a remount that
> only asked to change size.  It calls utf8_load() each time, and nothing
> frees the result - shmem_reconfigure() doesn't look at ctx->encoding and
> shmem_free_fc() doesn't unload it.  50000 such remounts grew kmalloc-32 by
> ~50000 objects; a tmpfs without casefold stays flat.
>
> The leak predates this patch - you can trigger it by passing casefold= on
> a remount by hand - so I sent the fix as a separate patch:
>
> https://lore.kernel.org/all/ba38f80f629fd093c77f3a49fdab7b71ee7bbc5f.1790687276.git.med08elkadiri@gmail.com/

Thanks for reporting and fixing this.  feel free to cc me on further
unicode stuff.  Will review it right now.

Thanks,

-- 
Gabriel Krisman Bertazi

[PATCH 2/4] ext4: Expose the filesystem encoding through sysfs

[Gabriel Krisman Bertazi]

Daniel Tang pointed out that shmem doesn't list casefold as a mount
option, even when it is mounted casefolded.  While fixing it, I noticed
f2fs actually already provides that information via sysfs, but ext4
doesn't.  This patch reproduces the same interface in ext4 to have
feature parity, so users don't need to rely on parsing dmesg or trying +F
on a temporary directory just to check for support.

Assisted-by: LLM
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 Documentation/ABI/testing/sysfs-fs-ext4 | 21 ++++++++++++++++++
 fs/ext4/sysfs.c                         | 29 +++++++++++++++++++++++++
 2 files changed, 50 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-fs-ext4 b/Documentation/ABI/testing/sysfs-fs-ext4
index 2edd0a6672d3..8abc5706f459 100644
--- a/Documentation/ABI/testing/sysfs-fs-ext4
+++ b/Documentation/ABI/testing/sysfs-fs-ext4
@@ -116,3 +116,24 @@ Contact:	"Theodore Ts'o" <tytso@mit.edu>
 Description:
 		This file is read-only and shows the pid of journal thread in
 		current pid-namespace or 0 if task is unreachable.
+
+What:		/sys/fs/ext4/<disk>/encoding
+Date:		September 2026
+Contact:	"Gabriel Krisman Bertazi" <krisman@suse.de>
+Description:
+		Displays name and version of the encoding set for the
+		filesystem.  If no encoding is set, displays (none)
+
+What:		/sys/fs/ext4/<disk>/encoding_flags
+Date:		September 2026
+Contact:	"Gabriel Krisman Bertazi" <krisman@suse.de>
+Description:
+		This is a read-only entry to show the value of
+		sb.s_encoding_flags, the value is hexadecimal.
+
+		============================     ==========
+		Flag_Name                        Flag_Value
+		============================     ==========
+		SB_ENC_STRICT_MODE_FL            0x00000001
+		SB_ENC_NO_COMPAT_FALLBACK_FL     0x00000002
+		============================     ==========
diff --git a/fs/ext4/sysfs.c b/fs/ext4/sysfs.c
index 923b375e017f..bc1c59d67eed 100644
--- a/fs/ext4/sysfs.c
+++ b/fs/ext4/sysfs.c
@@ -32,6 +32,8 @@ typedef enum {
 	attr_clusters_in_group,
 	attr_mb_order,
 	attr_feature,
+	attr_encoding,
+	attr_encoding_flags,
 	attr_pointer_pi,
 	attr_pointer_ui,
 	attr_pointer_ul,
@@ -161,6 +163,25 @@ static ssize_t err_report_sec_store(struct ext4_sb_info *sbi,
 	return count;
 }
 
+static ssize_t encoding_show(struct ext4_sb_info *sbi, char *buf)
+{
+#if IS_ENABLED(CONFIG_UNICODE)
+	struct super_block *sb = sbi->s_sb;
+
+	if (ext4_has_feature_casefold(sb))
+		return sysfs_emit(buf, "UTF-8 (%d.%d.%d)\n",
+				  unicode_major(sb->s_encoding->version),
+				  unicode_minor(sb->s_encoding->version),
+				  unicode_rev(sb->s_encoding->version));
+#endif
+	return sysfs_emit(buf, "(none)\n");
+}
+
+static ssize_t encoding_flags_show(struct ext4_sb_info *sbi, char *buf)
+{
+	return sysfs_emit(buf, "%x\n", le16_to_cpu(sbi->s_es->s_encoding_flags));
+}
+
 static ssize_t journal_task_show(struct ext4_sb_info *sbi, char *buf)
 {
 	if (!sbi->s_journal)
@@ -283,6 +304,8 @@ EXT4_RO_ATTR_ES_STRING(last_error_func, s_last_error_func, 32);
 EXT4_ATTR(first_error_time, 0444, first_error_time);
 EXT4_ATTR(last_error_time, 0444, last_error_time);
 EXT4_ATTR(journal_task, 0444, journal_task);
+EXT4_ATTR_FUNC(encoding, 0444);
+EXT4_ATTR_FUNC(encoding_flags, 0444);
 EXT4_RW_ATTR_SBI_UI(mb_prefetch, s_mb_prefetch);
 EXT4_RW_ATTR_SBI_UI(mb_prefetch_limit, s_mb_prefetch_limit);
 EXT4_RW_ATTR_SBI_UL(last_trim_minblks, s_last_trim_minblks);
@@ -333,6 +356,8 @@ static struct attribute *ext4_attrs[] = {
 	ATTR_LIST(first_error_time),
 	ATTR_LIST(last_error_time),
 	ATTR_LIST(journal_task),
+	ATTR_LIST(encoding),
+	ATTR_LIST(encoding_flags),
 #ifdef CONFIG_EXT4_DEBUG
 	ATTR_LIST(simulate_fail),
 #endif
@@ -483,6 +508,10 @@ static ssize_t ext4_attr_show(struct kobject *kobj,
 		return print_tstamp(buf, sbi->s_es, s_last_error_time);
 	case attr_journal_task:
 		return journal_task_show(sbi, buf);
+	case attr_encoding:
+		return encoding_show(sbi, buf);
+	case attr_encoding_flags:
+		return encoding_flags_show(sbi, buf);
 	default:
 		return ext4_generic_attr_show(a, sbi, buf);
 	}
-- 
2.55.0

Re: [PATCH 2/4] ext4: Expose the filesystem encoding through sysfs

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] ext4: NULL pointer deref in encoding_show() via dynamic casefold
--

commit c325498f77eea14582e06218baa713f49fde001a
Author: Gabriel Krisman Bertazi <krisman@suse.de>

ext4: Expose the filesystem encoding through sysfs

This commit introduces new sysfs attributes in ext4 to display the active
filesystem encoding and encoding flags. This provides users with a direct
interface to check casefold support without relying on dmesg parsing.

> diff --git a/fs/ext4/sysfs.c b/fs/ext4/sysfs.c
> index 923b375e017fa..bc1c59d67eed8 100644
> --- a/fs/ext4/sysfs.c
> +++ b/fs/ext4/sysfs.c
[ ... ]
> @@ -161,6 +163,25 @@ static ssize_t err_report_sec_store(struct ext4_sb_info *sbi,
>  	return count;
>  }
>  
> +static ssize_t encoding_show(struct ext4_sb_info *sbi, char *buf)
> +{
> +#if IS_ENABLED(CONFIG_UNICODE)
> +	struct super_block *sb = sbi->s_sb;
> +
> +	if (ext4_has_feature_casefold(sb))
> +		return sysfs_emit(buf, "UTF-8 (%d.%d.%d)\n",
> +				  unicode_major(sb->s_encoding->version),
> +				  unicode_minor(sb->s_encoding->version),
> +				  unicode_rev(sb->s_encoding->version));

[Severity: High]
Does this introduce a regression where reading the sysfs attribute causes a
NULL pointer dereference if casefold is enabled dynamically?

If a privileged user dynamically enables EXT4_FEATURE_INCOMPAT_CASEFOLD on
an online filesystem via the EXT4_IOC_SET_TUNE_SB_PARAM ioctl, it updates
the in-memory on-disk superblock representation but does not load the
unicode map into the VFS superblock:

fs/ext4/ioctl.c:ext4_ioctl_set_tune_sb() {
    ...
    if ((params.set_feature_incompat_mask &
         EXT4_FEATURE_INCOMPAT_CASEFOLD)) {
        enabling_casefold = 1;
    ...
}

In this state, ext4_has_feature_casefold(sb) evaluates to true, but
sb->s_encoding remains NULL. If the sysfs attribute is read,
encoding_show() will unconditionally dereference sb->s_encoding->version,
resulting in a system crash.

> +#endif
> +	return sysfs_emit(buf, "(none)\n");
> +}
> +
[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260928213155.2612039-1-krisman@suse.de?part=2

Re: [PATCH 2/4] ext4: Expose the filesystem encoding through sysfs

[Gabriel Krisman Bertazi]

sashiko-bot@kernel.org writes:

> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [High] ext4: NULL pointer deref in encoding_show() via dynamic casefold
> --
>
> commit c325498f77eea14582e06218baa713f49fde001a
> Author: Gabriel Krisman Bertazi <krisman@suse.de>
>
> ext4: Expose the filesystem encoding through sysfs
>
> This commit introduces new sysfs attributes in ext4 to display the active
> filesystem encoding and encoding flags. This provides users with a direct
> interface to check casefold support without relying on dmesg parsing.
>
>> diff --git a/fs/ext4/sysfs.c b/fs/ext4/sysfs.c
>> index 923b375e017fa..bc1c59d67eed8 100644
>> --- a/fs/ext4/sysfs.c
>> +++ b/fs/ext4/sysfs.c
> [ ... ]
>> @@ -161,6 +163,25 @@ static ssize_t err_report_sec_store(struct ext4_sb_info *sbi,
>>  	return count;
>>  }
>>  
>> +static ssize_t encoding_show(struct ext4_sb_info *sbi, char *buf)
>> +{
>> +#if IS_ENABLED(CONFIG_UNICODE)
>> +	struct super_block *sb = sbi->s_sb;
>> +
>> +	if (ext4_has_feature_casefold(sb))
>> +		return sysfs_emit(buf, "UTF-8 (%d.%d.%d)\n",
>> +				  unicode_major(sb->s_encoding->version),
>> +				  unicode_minor(sb->s_encoding->version),
>> +				  unicode_rev(sb->s_encoding->version));
>
> [Severity: High]
> Does this introduce a regression where reading the sysfs attribute causes a
> NULL pointer dereference if casefold is enabled dynamically?
>
> If a privileged user dynamically enables EXT4_FEATURE_INCOMPAT_CASEFOLD on
> an online filesystem via the EXT4_IOC_SET_TUNE_SB_PARAM ioctl, it updates
> the in-memory on-disk superblock representation but does not load the
> unicode map into the VFS superblock:
>
> fs/ext4/ioctl.c:ext4_ioctl_set_tune_sb() {
>     ...
>     if ((params.set_feature_incompat_mask &
>          EXT4_FEATURE_INCOMPAT_CASEFOLD)) {
>         enabling_casefold = 1;
>     ...
> }
>
> In this state, ext4_has_feature_casefold(sb) evaluates to true, but
> sb->s_encoding remains NULL. If the sysfs attribute is read,
> encoding_show() will unconditionally dereference sb->s_encoding->version,
> resulting in a system crash.

Uh...

If we flip EXT4_FEATURE_INCOMPAT_CASEFOLD without loading the casefold
table, I'm surprised the current code doesn't blow up already.  I
completely missed the introduction of EXT4_IOC_SET_TUNE_SB_PARAM, so I'm
missing context if there is a use case to actually do an online flip of
feat_casefold.  It seems the ioctl actually can set this bit, but I
don't see any provision to load the table.  I'm a bit confused now.  I
can work around it for /sys/fs, but is anyone actually enabling casefold
without a remount?

>
>> +#endif
>> +	return sysfs_emit(buf, "(none)\n");
>> +}
>> +
> [ ... ]
>
> -- 
> Sashiko AI review · https://sashiko.dev/#/patchset/20260928213155.2612039-1-krisman@suse.de?part=2

-- 
Gabriel Krisman Bertazi

[PATCH 3/4] ovl: Add /sys/fs/overlay plumbing

[Gabriel Krisman Bertazi]

In preparation to expose encoding settings in sysfs, add a subtree for
overlayfs under /sys/fs.  The only remarkable detail is how we reference
the volume.  Use major:minor, which is not very human-friendly but it is
at least discoverable.  I considered uuid, but that would be a problem
if mounted with uuid=off.

Assisted-by: LLM
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 Documentation/ABI/testing/sysfs-fs-overlay |   6 ++
 MAINTAINERS                                |   1 +
 fs/overlayfs/Makefile                      |   2 +-
 fs/overlayfs/overlayfs.h                   |   5 +
 fs/overlayfs/ovl_entry.h                   |   7 ++
 fs/overlayfs/super.c                       |  26 +++++-
 fs/overlayfs/sysfs.c                       | 101 +++++++++++++++++++++
 7 files changed, 144 insertions(+), 4 deletions(-)
 create mode 100644 Documentation/ABI/testing/sysfs-fs-overlay
 create mode 100644 fs/overlayfs/sysfs.c

diff --git a/Documentation/ABI/testing/sysfs-fs-overlay b/Documentation/ABI/testing/sysfs-fs-overlay
new file mode 100644
index 000000000000..9060c39679ae
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-fs-overlay
@@ -0,0 +1,6 @@
+What:		/sys/fs/overlay/<volume>/
+Date:		September 2026
+Contact:	linux-unionfs@vger.kernel.org
+Description:
+		A directory is created for every mounted overlayfs instance.
+		<volume> is set to the device major:minor.
diff --git a/MAINTAINERS b/MAINTAINERS
index c2414447892c..e49f84c4c691 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -20586,6 +20586,7 @@ M:	Amir Goldstein <amir73il@gmail.com>
 L:	linux-unionfs@vger.kernel.org
 S:	Supported
 T:	git git://git.kernel.org/pub/scm/linux/kernel/git/overlayfs/vfs.git
+F:	Documentation/ABI/testing/sysfs-fs-overlay
 F:	Documentation/filesystems/overlayfs.rst
 F:	fs/overlayfs/
 
diff --git a/fs/overlayfs/Makefile b/fs/overlayfs/Makefile
index 5648954f8588..fd4e747820c2 100644
--- a/fs/overlayfs/Makefile
+++ b/fs/overlayfs/Makefile
@@ -6,4 +6,4 @@
 obj-$(CONFIG_OVERLAY_FS) += overlay.o
 
 overlay-objs := super.o namei.o util.o inode.o file.o dir.o readdir.o \
-		copy_up.o export.o params.o xattrs.o
+		copy_up.o export.o params.o xattrs.o sysfs.o
diff --git a/fs/overlayfs/overlayfs.h b/fs/overlayfs/overlayfs.h
index e0d8c6152e9f..91d3894d9bd2 100644
--- a/fs/overlayfs/overlayfs.h
+++ b/fs/overlayfs/overlayfs.h
@@ -939,6 +939,11 @@ extern const struct export_operations ovl_export_fid_operations;
 /* super.c */
 int ovl_fill_super(struct super_block *sb, struct fs_context *fc);
 
+int __init ovl_sysfs_init(void);
+void ovl_sysfs_exit(void);
+int ovl_sysfs_register(struct ovl_fs *ofs);
+void ovl_sysfs_unregister(struct ovl_fs *ofs);
+
 /* Will this overlay be forced to mount/remount ro? */
 static inline bool ovl_force_readonly(struct ovl_fs *ofs)
 {
diff --git a/fs/overlayfs/ovl_entry.h b/fs/overlayfs/ovl_entry.h
index 80cad4ea96a3..c14253e75826 100644
--- a/fs/overlayfs/ovl_entry.h
+++ b/fs/overlayfs/ovl_entry.h
@@ -5,6 +5,9 @@
  * Copyright (C) 2016 Red Hat, Inc.
  */
 
+#include <linux/completion.h>
+#include <linux/kobject.h>
+
 struct ovl_config {
 	char *upperdir;
 	char *workdir;
@@ -56,6 +59,7 @@ struct ovl_entry {
 
 /* private information held for overlayfs's superblock */
 struct ovl_fs {
+	struct super_block *sb;
 	unsigned int numlayer;
 	/* Number of unique fs among layers including upper fs */
 	unsigned int numfs;
@@ -92,6 +96,9 @@ struct ovl_fs {
 	/* r/o snapshot of upperdir sb's only taken on volatile mounts */
 	errseq_t errseq;
 	bool casefold;
+	struct kobject kobj;
+	struct completion kobj_unregister;
+	bool kobj_registered;
 };
 
 /* Number of lower layers, not including data-only layers */
diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
index bd0a3f9039d2..d195dace135d 100644
--- a/fs/overlayfs/super.c
+++ b/fs/overlayfs/super.c
@@ -226,8 +226,10 @@ static void ovl_put_super(struct super_block *sb)
 {
 	struct ovl_fs *ofs = OVL_FS(sb);
 
-	if (ofs)
+	if (ofs) {
+		ovl_sysfs_unregister(ofs);
 		ovl_free_fs(ofs);
+	}
 }
 
 /* Sync real dirty inodes in upper filesystem (if it exists) */
@@ -1548,6 +1550,7 @@ int ovl_fill_super(struct super_block *sb, struct fs_context *fc)
 	if (fc->user_ns != current_user_ns())
 		goto out_err;
 
+	ofs->sb = sb;
 	ovl_set_d_op(sb);
 
 	if (!ofs->creator_cred) {
@@ -1559,6 +1562,13 @@ int ovl_fill_super(struct super_block *sb, struct fs_context *fc)
 
 	with_ovl_creds(sb)
 		err = ovl_fill_super_creds(fc, sb);
+	if (err)
+		goto out_err;
+	/*
+	 * ->s_root is now set (by ovl_fill_super_creds), so if
+	 * ovl_sysfs_register fails it will be cleaned by ovl_put_super.
+	 */
+	return ovl_sysfs_register(ofs);
 
 out_err:
 	if (err) {
@@ -1598,10 +1608,19 @@ static int __init ovl_init(void)
 	if (ovl_inode_cachep == NULL)
 		return -ENOMEM;
 
+	err = ovl_sysfs_init();
+	if (err)
+		goto out_destroy_cache;
+
 	err = register_filesystem(&ovl_fs_type);
-	if (!err)
-		return 0;
+	if (err)
+		goto out_sysfs_exit;
+
+	return 0;
 
+out_sysfs_exit:
+	ovl_sysfs_exit();
+out_destroy_cache:
 	kmem_cache_destroy(ovl_inode_cachep);
 
 	return err;
@@ -1610,6 +1629,7 @@ static int __init ovl_init(void)
 static void __exit ovl_exit(void)
 {
 	unregister_filesystem(&ovl_fs_type);
+	ovl_sysfs_exit();
 
 	/*
 	 * Make sure all delayed rcu free inodes are flushed before we
diff --git a/fs/overlayfs/sysfs.c b/fs/overlayfs/sysfs.c
new file mode 100644
index 000000000000..be83b57f637c
--- /dev/null
+++ b/fs/overlayfs/sysfs.c
@@ -0,0 +1,101 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Sysfs interface for overlayfs.
+ */
+#include <linux/fs.h>
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+#include <linux/unicode.h>
+#include <linux/xattr.h>
+#include "overlayfs.h"
+
+static struct kset *ovl_kset;
+
+struct ovl_sysfs_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct ovl_fs *ofs, char *buf);
+};
+
+#define OVL_SYSFS_ATTR_RO(_name)					\
+	static struct ovl_sysfs_attr ovl_sysfs_attr_##_name = {		\
+		.attr	= { .name = __stringify(_name), .mode = 0444 },	\
+		.show	= ovl_##_name##_show,				\
+	}
+
+static struct attribute *ovl_sysfs_attrs[] = {
+	NULL,
+};
+ATTRIBUTE_GROUPS(ovl_sysfs);
+
+static ssize_t ovl_sysfs_show(struct kobject *kobj, struct attribute *attr,
+			      char *buf)
+{
+	struct ovl_fs *ofs = container_of(kobj, struct ovl_fs, kobj);
+	struct ovl_sysfs_attr *a = container_of(attr, struct ovl_sysfs_attr,
+						attr);
+
+	return a->show(ofs, buf);
+}
+
+static const struct sysfs_ops ovl_sysfs_ops = {
+	.show	= ovl_sysfs_show,
+};
+
+static void ovl_sysfs_release(struct kobject *kobj)
+{
+	struct ovl_fs *ofs = container_of(kobj, struct ovl_fs, kobj);
+
+	complete(&ofs->kobj_unregister);
+}
+
+static const struct kobj_type ovl_sysfs_ktype = {
+	.default_groups	= ovl_sysfs_groups,
+	.sysfs_ops	= &ovl_sysfs_ops,
+	.release	= ovl_sysfs_release,
+};
+
+int ovl_sysfs_register(struct ovl_fs *ofs)
+{
+	struct super_block *sb = ofs->sb;
+	int err;
+
+	init_completion(&ofs->kobj_unregister);
+	ofs->kobj.kset = ovl_kset;
+	err = kobject_init_and_add(&ofs->kobj, &ovl_sysfs_ktype, NULL, "%u:%u",
+				   MAJOR(sb->s_dev), MINOR(sb->s_dev));
+	if (err) {
+		kobject_put(&ofs->kobj);
+		wait_for_completion(&ofs->kobj_unregister);
+		pr_warn("failed to create sysfs volume %u:%u (err=%d).\n",
+			MAJOR(sb->s_dev), MINOR(sb->s_dev), err);
+		return err;
+	}
+
+	ofs->kobj_registered = true;
+	kobject_uevent(&ofs->kobj, KOBJ_ADD);
+	return 0;
+}
+
+void ovl_sysfs_unregister(struct ovl_fs *ofs)
+{
+	if (!ofs->kobj_registered)
+		return;
+
+	ofs->kobj_registered = false;
+	kobject_del(&ofs->kobj);
+	kobject_put(&ofs->kobj);
+	wait_for_completion(&ofs->kobj_unregister);
+}
+
+int __init ovl_sysfs_init(void)
+{
+	ovl_kset = kset_create_and_add("overlay", NULL, fs_kobj);
+	if (!ovl_kset)
+		return -ENOMEM;
+	return 0;
+}
+
+void ovl_sysfs_exit(void)
+{
+	kset_unregister(ovl_kset);
+}
-- 
2.55.0

Re: [PATCH 3/4] ovl: Add /sys/fs/overlay plumbing

[sashiko-bot]

> In preparation to expose encoding settings in sysfs, add a subtree for
> overlayfs under /sys/fs.  The only remarkable detail is how we reference
> the volume.  Use major:minor, which is not very human-friendly but it is
> at least discoverable.  I considered uuid, but that would be a problem
> if mounted with uuid=off.
> 
> Assisted-by: LLM
> Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260928213155.2612039-1-krisman@suse.de?part=3

[PATCH 4/4] ovl: Expose the filesystem encoding through sysfs

[Gabriel Krisman Bertazi]

This patch reproduces the same interface from ext4/f2fs to have feature
parity, so users don't need to rely on parsing dmesg or trying +F on a
temporary directory just to check for support.  Note there is no
encoding_flags file, as none are supported by ovl.

Assisted-by: LLM
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 Documentation/ABI/testing/sysfs-fs-overlay |  7 +++++++
 fs/overlayfs/sysfs.c                       | 15 +++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-fs-overlay b/Documentation/ABI/testing/sysfs-fs-overlay
index 9060c39679ae..7f3c6403c4f8 100644
--- a/Documentation/ABI/testing/sysfs-fs-overlay
+++ b/Documentation/ABI/testing/sysfs-fs-overlay
@@ -4,3 +4,10 @@ Contact:	linux-unionfs@vger.kernel.org
 Description:
 		A directory is created for every mounted overlayfs instance.
 		<volume> is set to the device major:minor.
+
+What:		/sys/fs/overlay/<volume>/encoding
+Date:		September 2026
+Contact:	linux-unionfs@vger.kernel.org
+Description:
+		Displays name and version of the encoding set for the
+		overlay.  If no encoding is set, displays (none)
diff --git a/fs/overlayfs/sysfs.c b/fs/overlayfs/sysfs.c
index be83b57f637c..2144cbb418bc 100644
--- a/fs/overlayfs/sysfs.c
+++ b/fs/overlayfs/sysfs.c
@@ -22,7 +22,22 @@ struct ovl_sysfs_attr {
 		.show	= ovl_##_name##_show,				\
 	}
 
+static ssize_t ovl_encoding_show(struct ovl_fs *ofs, char *buf)
+{
+	struct unicode_map *um = sb_encoding(ofs->sb);
+
+	if (!um)
+		return sysfs_emit(buf, "(none)\n");
+
+	return sysfs_emit(buf, "UTF-8 (%d.%d.%d)\n",
+			  unicode_major(um->version),
+			  unicode_minor(um->version),
+			  unicode_rev(um->version));
+}
+OVL_SYSFS_ATTR_RO(encoding);
+
 static struct attribute *ovl_sysfs_attrs[] = {
+	&ovl_sysfs_attr_encoding.attr,
 	NULL,
 };
 ATTRIBUTE_GROUPS(ovl_sysfs);
-- 
2.55.0

Re: [PATCH 4/4] ovl: Expose the filesystem encoding through sysfs

[sashiko-bot]

> This patch reproduces the same interface from ext4/f2fs to have feature
> parity, so users don't need to rely on parsing dmesg or trying +F on a
> temporary directory just to check for support.  Note there is no
> encoding_flags file, as none are supported by ovl.
> 
> Assisted-by: LLM
> Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260928213155.2612039-1-krisman@suse.de?part=4

Re: [PATCH 4/4] ovl: Expose the filesystem encoding through sysfs

[Amir Goldstein]

On Mon, Sep 28, 2026 at 11:32 PM Gabriel Krisman Bertazi
<krisman@suse.de> wrote:
>
> This patch reproduces the same interface from ext4/f2fs to have feature
> parity, so users don't need to rely on parsing dmesg or trying +F on a
> temporary directory just to check for support.  Note there is no
> encoding_flags file, as none are supported by ovl.
>

Gabriel,

I am not ok with this API.
It is wrong in several ways.
It is not generic, we have already file_getattr() to get casefold status.
There was a suggestion by Pali [1] to get the supported xflags - statx style,
which can be a way to say if a fs could return FS_XFLAG_CASEFOLD.
In Chuck's v2 [2] FILEATTR_CASEFOLD_TYPE was already suggested
as an extension to file_getattr(). I think we can still do that if we include
proper versioning in the casefold type.

TBH, I don't think anyone requested this - you only mentioned
the need to know if tmpfs was mounted with casefold on and that
doesn't mean that etx4/ovl patches are justified, but ext4 can decide
for themselves.

The point is that this sysfs is not going to be the standard for all the
fs that can return the FS_XFLAG_CASEFOLD, so I see no reason
to add this sysfs interface to ovl.

Thanks,
Amir.

[1] https://lore.kernel.org/linux-fsdevel/20250216164029.20673-4-pali@kernel.org/
[2] https://lore.kernel.org/linux-fsdevel/20251211152116.480799-2-cel@kernel.org/

Re: [PATCH 4/4] ovl: Expose the filesystem encoding through sysfs

[Pali Rohár]

On Tuesday 29 September 2026 16:15:04 Amir Goldstein wrote:
> On Mon, Sep 28, 2026 at 11:32 PM Gabriel Krisman Bertazi
> <krisman@suse.de> wrote:
> >
> > This patch reproduces the same interface from ext4/f2fs to have feature
> > parity, so users don't need to rely on parsing dmesg or trying +F on a
> > temporary directory just to check for support.  Note there is no
> > encoding_flags file, as none are supported by ovl.
> >
> 
> Gabriel,
> 
> I am not ok with this API.
> It is wrong in several ways.
> It is not generic, we have already file_getattr() to get casefold status.
> There was a suggestion by Pali [1] to get the supported xflags - statx style,
> which can be a way to say if a fs could return FS_XFLAG_CASEFOLD.

Yea, sorry for that. I have not finished it, I did not have time. But I
would like to address this as this is something was was asked more times...

> In Chuck's v2 [2] FILEATTR_CASEFOLD_TYPE was already suggested
> as an extension to file_getattr(). I think we can still do that if we include
> proper versioning in the casefold type.
> 
> TBH, I don't think anyone requested this - you only mentioned
> the need to know if tmpfs was mounted with casefold on and that
> doesn't mean that etx4/ovl patches are justified, but ext4 can decide
> for themselves.
> 
> The point is that this sysfs is not going to be the standard for all the
> fs that can return the FS_XFLAG_CASEFOLD, so I see no reason
> to add this sysfs interface to ovl.
> 
> Thanks,
> Amir.
> 
> [1] https://lore.kernel.org/linux-fsdevel/20250216164029.20673-4-pali@kernel.org/
> [2] https://lore.kernel.org/linux-fsdevel/20251211152116.480799-2-cel@kernel.org/

Re: [PATCH 4/4] ovl: Expose the filesystem encoding through sysfs

[Gabriel Krisman Bertazi]

Pali Rohár <pali@kernel.org> writes:

> On Tuesday 29 September 2026 16:15:04 Amir Goldstein wrote:
>> On Mon, Sep 28, 2026 at 11:32 PM Gabriel Krisman Bertazi
>> <krisman@suse.de> wrote:
>> >
>> > This patch reproduces the same interface from ext4/f2fs to have feature
>> > parity, so users don't need to rely on parsing dmesg or trying +F on a
>> > temporary directory just to check for support.  Note there is no
>> > encoding_flags file, as none are supported by ovl.
>> >
>> 
>> Gabriel,
>> 
>> I am not ok with this API.
>> It is wrong in several ways.
>> It is not generic, we have already file_getattr() to get casefold status.
>> There was a suggestion by Pali [1] to get the supported xflags - statx style,
>> which can be a way to say if a fs could return FS_XFLAG_CASEFOLD.
>
> Yea, sorry for that. I have not finished it, I did not have time. But I
> would like to address this as this is something was was asked more
> times...

please cc me when sending it. 

-- 
Gabriel Krisman Bertazi

Re: [PATCH 4/4] ovl: Expose the filesystem encoding through sysfs

[Gabriel Krisman Bertazi]

Amir Goldstein <amir73il@gmail.com> writes:

> On Mon, Sep 28, 2026 at 11:32 PM Gabriel Krisman Bertazi
> <krisman@suse.de> wrote:

> I am not ok with this API.
> It is wrong in several ways.
> It is not generic, we have already file_getattr() to get casefold status.
> There was a suggestion by Pali [1] to get the supported xflags - statx style,
> which can be a way to say if a fs could return FS_XFLAG_CASEFOLD.
> In Chuck's v2 [2] FILEATTR_CASEFOLD_TYPE was already suggested
> as an extension to file_getattr(). I think we can still do that if we include
> proper versioning in the casefold type.

Ok. I'm not fond of the overlay part of this series at all, in
particular due to the major:minor design.  I was looking for an
alternative way to do it and didn't find it, thus /sys/fs.

> TBH, I don't think anyone requested this - you only mentioned
> the need to know if tmpfs was mounted with casefold on and that
> doesn't mean that etx4/ovl patches are justified, but ext4 can decide
> for themselves.

There is a need for a proper probing API when deciding to fallback to
userspace casefolding workarounds, and that is not shmem specific at
all.  I'm sure the Wine folks that right now use ext4/overlayfs will
also benefit from a smarter way to fetch the information.

> The point is that this sysfs is not going to be the standard for all the
> fs that can return the FS_XFLAG_CASEFOLD, so I see no reason
> to add this sysfs interface to ovl.

Ack.  I'll drop patches 3 and 4 and wait on ext4 people to see if they want
patch 2 to match f2fs. Will also keep an eye on FS_XFLAG_CASEFOLD.

> Thanks,
> Amir.
>
> [1] https://lore.kernel.org/linux-fsdevel/20250216164029.20673-4-pali@kernel.org/
> [2] https://lore.kernel.org/linux-fsdevel/20251211152116.480799-2-cel@kernel.org/

-- 
Gabriel Krisman Bertazi