pidfs: preserve thread pidfds reopened by file handle
PATCH 1/1 pidfs: preserve thread pidfds reopened by file handle
PIDFD_THREAD shares O_EXCL. do_dentry_open() clears O_EXCL after pidfs_export_open() validates the flags, so open_by_handle_at() silently turns a thread pidfd into a process pidfd.
Restore PIDFD_THREAD on the opened file, matching pidfs_alloc_file(), and cover the restored flag with F_GETFL.
--- fs/pidfs.c | 8 +++++++- tools/testing/selftests/pidfd/pidfd_file_handle_test.c | 1 + 2 files changed, 8 insertions(+), 1 deletion(-)
fs/pidfs.c +7 -1
@@ -939,12 +939,18 @@ static int pidfs_export_permission(struct handle_to_path_ctx *ctx,static struct file *pidfs_export_open (const struct path *path,unsigned int oflags) { +struct file *file; +/* * Clear O_LARGEFILE as open_by_handle_at() forces it and raise * O_RDWR as pidfds always are. */ oflags &= ~O_LARGEFILE ; -return dentry_open (path, oflags |O_RDWR ,current_cred ()); + file =dentry_open (path, oflags |O_RDWR ,current_cred ()); +/* do_dentry_open() strips O_EXCL, which encodes PIDFD_THREAD. */ +if (!IS_ERR (file)) + file->f_flags |= oflags &PIDFD_THREAD ; +return file; }static const struct export_operations pidfs_export_operations = {
tools/testing/selftests/pidfd/pidfd_file_handle_test.c +1 -0
@@ -373,6 +373,7 @@ TEST_F(file_handle, open_by_handle_at_valid_flags)O_CLOEXEC | O_EXCL);ASSERT_GE (pidfd,0 ); +ASSERT_NE (fcntl (pidfd,F_GETFL ) &PIDFD_THREAD ,0 );ASSERT_EQ (fstat (pidfd, &st2),0 );ASSERT_TRUE (st1.st_dev == st2.st_dev && st1.st_ino == st2.st_ino);
Signature · 1 line
2.52.0
Quoted 8 lines · Li Chen
PIDFD_THREADsharesO_EXCL.do_dentry_open()clearsO_EXCLafterpidfs_export_open()validates the flags, soopen_by_handle_at()silently turns a thread pidfd into a process pidfd.Restore
PIDFD_THREADon the opened file, matchingpidfs_alloc_file(), and cover the restored flag withF_GETFL.Signed-off-by Li Chen <[email protected]>
Makes sense. Just looking at this code, what about PIDFD_AUTOKILL (aka O_TRUNC?)? Reading the commit introducing it 07c3ef58223e it appears it was never meant to be directly requested by userspace so probably pidfs_export_open() needs to reject it. But that's unrelated issue to your patch so feel free to add:
Honza
Quoted 44 lines
--- fs/pidfs.c | 8 +++++++- tools/testing/selftests/pidfd/pidfd_file_handle_test.c | 1 + 2 files changed, 8 insertions(+), 1 deletion(-)fs/pidfs.c +7
-1@@ -939,12 +939,18 @@ static int pidfs_export_permission(struct handle_to_path_ctx *ctx,static struct file *pidfs_export_open (const struct path *path,unsigned int oflags) { +struct file *file; +/* * Clear O_LARGEFILE as open_by_handle_at() forces it and raise * O_RDWR as pidfds always are. */ oflags &= ~O_LARGEFILE ; -return dentry_open (path, oflags |O_RDWR ,current_cred ()); + file =dentry_open (path, oflags |O_RDWR ,current_cred ()); +/* do_dentry_open() strips O_EXCL, which encodes PIDFD_THREAD. */ +if (!IS_ERR (file)) + file->f_flags |= oflags &PIDFD_THREAD ; +return file; }static const struct export_operations pidfs_export_operations = {tools/testing/selftests/pidfd/pidfd_file_handle_test.c +1
-0@@ -373,6 +373,7 @@ TEST_F(file_handle, open_by_handle_at_valid_flags)O_CLOEXEC | O_EXCL);ASSERT_GE (pidfd,0 ); +ASSERT_NE (fcntl (pidfd,F_GETFL ) &PIDFD_THREAD ,0 );ASSERT_EQ (fstat (pidfd, &st2),0 );ASSERT_TRUE (st1.st_dev == st2.st_dev && st1.st_ino == st2.st_ino);Signature · 1 line
2.52.0
Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR
Christian Brauner
maintainerQuoted 8 lines · Jan Kara
Quoted 7 lines · Li Chen
PIDFD_THREADsharesO_EXCL.do_dentry_open()clearsO_EXCLafterpidfs_export_open()validates the flags, soopen_by_handle_at()silently turns a thread pidfd into a process pidfd.Restore
PIDFD_THREADon the opened file, matchingpidfs_alloc_file(), and cover the restored flag withF_GETFL.
Quoted 7 lines
Quoted 1 line
Signed-off-by Li Chen <[email protected]>Makes sense. Just looking at this code, what about
PIDFD_AUTOKILL(akaO_TRUNC?)? Reading the commit introducing it07c3ef58223eit appears it was never meant to be directly requested by userspace so probablypidfs_export_open()needs to reject it. But that's unrelated issue to your patch so feel free to add:
Good point. But pidfs_export_open() is guarded bypidfs_export_permission() which does:
#define VALID_FILE_HANDLE_OPEN_FLAGS \
(O_RDONLY | O_WRONLY | O_RDWR | O_NONBLOCK | O_CLOEXEC | O_EXCL)
if (oflags & ~(VALID_FILE_HANDLE_OPEN_FLAGS | O_LARGEFILE))
return -EINVAL;
and so is safe. But I've added the following patch on top which also splats if that is passed!
From 99a859485394eca8cf6271de09ae72ac3b846bda Mon Sep 17 00:00:00 2001
From: Christian Brauner <[email protected]>
Date: Wed, 22 Jul 2026 13:56:28 +0200
Subject: [PATCH] pidfs: add pidfs_dentry_open() helper
Both pidfs_alloc_file() and pidfs_export_open() need to force O_RDWR and reapply the pidfd flags that do_dentry_open() strips. Move the common logic into a helper.
PIDFD_AUTOKILL is now part of the restore mask in the file handle path as well, but pidfs_export_permission() rejects O_TRUNC, so this is a no-op there. But warn nonetheless.
--- fs/pidfs.c | 42 +++++++++++++++++++++--------------------- 1 file changed, 21 insertions(+), 21 deletions(-)
fs/pidfs.c +21 -21
@@ -915,6 +915,20 @@ static struct dentry *pidfs_fh_to_dentry(struct super_block *sb,return path.dentry; } +static struct file *pidfs_dentry_open (const struct path *path, +unsigned int flags, +const struct cred *cred) +{ +struct file *file; + +/* pidfds are always O_RDWR. */ + file =dentry_open (path, flags |O_RDWR , cred); +/* do_dentry_open() strips O_EXCL and O_TRUNC. */ +if (!IS_ERR (file)) + file->f_flags |= flags & (PIDFD_THREAD |PIDFD_AUTOKILL ); +return file; +} +/* * Make sure that we reject any nonsensical flags that users pass via * open_by_handle_at(). Note that PIDFD_THREAD is defined as O_EXCL, and
@@ -939,18 +953,14 @@ static int pidfs_export_permission(struct handle_to_path_ctx *ctx,static struct file *pidfs_export_open (const struct path *path,unsigned int oflags) { -struct file *file; -/* -* Clear O_LARGEFILE as open_by_handle_at() forces it and raise -* O_RDWR as pidfds always are. +* Opening via file handle may never raise PIDFD_AUTOKILL. That can +* only be done at task creation! */ - oflags &= ~O_LARGEFILE ; - file =dentry_open (path, oflags |O_RDWR ,current_cred ()); -/* do_dentry_open() strips O_EXCL, which encodes PIDFD_THREAD. */ -if (!IS_ERR (file)) - file->f_flags |= oflags &PIDFD_THREAD ; -return file; +if (WARN_ON_ONCE (oflags &PIDFD_AUTOKILL )) +return ERR_PTR (-EINVAL ); +/* Clear O_LARGEFILE as open_by_handle_at() forces it. */ +return pidfs_dentry_open (path, oflags & ~O_LARGEFILE ,current_cred ()); }static const struct export_operations pidfs_export_operations = {
@@ -1114,7 +1124,6 @@ static struct file_system_type pidfs_type = {struct file *pidfs_alloc_file (struct pid *pid,unsigned int flags) { -struct file *pidfd_file; struct path path __free(path_put) = {};int ret;
@@ -1132,16 +1141,7 @@ struct file *pidfs_alloc_file(struct pid *pid, unsigned int flags)VFS_WARN_ON_ONCE (!pid->attr); flags &= ~PIDFD_STALE ; - flags |=O_RDWR ; - pidfd_file =dentry_open (&path, flags,current_cred ()); -/* -* Raise PIDFD_THREAD and PIDFD_AUTOKILL explicitly as -* do_dentry_open() strips O_EXCL and O_TRUNC. -*/ -if (!IS_ERR (pidfd_file)) - pidfd_file->f_flags |= (flags & (PIDFD_THREAD |PIDFD_AUTOKILL )); - -return pidfd_file; +return pidfs_dentry_open (&path, flags,current_cred ()); }void __init pidfs_init (void )
Signature · 1 line
2.53.0
Quoted 38 lines · Christian Brauner
Quoted 8 lines · Jan Kara
Quoted 7 lines · Li Chen
PIDFD_THREADsharesO_EXCL.do_dentry_open()clearsO_EXCLafterpidfs_export_open()validates the flags, soopen_by_handle_at()silently turns a thread pidfd into a process pidfd.Restore
PIDFD_THREADon the opened file, matchingpidfs_alloc_file(), and cover the restored flag withF_GETFL.Quoted 7 lines
Quoted 1 line
Signed-off-by Li Chen <[email protected]>Makes sense. Just looking at this code, what about
PIDFD_AUTOKILL(akaO_TRUNC?)? Reading the commit introducing it07c3ef58223eit appears it was never meant to be directly requested by userspace so probablypidfs_export_open()needs to reject it. But that's unrelated issue to your patch so feel free to add:Good point. But
pidfs_export_open()is guarded bypidfs_export_permission()which does:#define
VALID_FILE_HANDLE_OPEN_FLAGS\
(O_RDONLY|O_WRONLY|O_RDWR|O_NONBLOCK|O_CLOEXEC|O_EXCL)if (oflags & ~(VALID_FILE_HANDLE_OPEN_FLAGS |
O_LARGEFILE))
return-EINVAL;and so is safe. But I've added the following patch on top which also splats if that is passed!
From
99a859485394eca8cf6271de09ae72ac3b846bdaMon Sep 17 00:00:00 2001
From: Christian Brauner <[email protected]>
Date: Wed, 22 Jul 2026 13:56:28 +0200
Subject: [PATCH] pidfs: addpidfs_dentry_open()helperBoth
pidfs_alloc_file()andpidfs_export_open()need to forceO_RDWRand reapply the pidfd flags thatdo_dentry_open()strips. Move the common logic into a helper.
Quoted 5 lines
PIDFD_AUTOKILLis now part of the restore mask in the file handle path as well, butpidfs_export_permission()rejectsO_TRUNC, so this is a no-op there. But warn nonetheless.Signed-off-by Christian Brauner (Amutable) <[email protected]>
Nice! Feel free to add:
Honza
Quoted 80 lines
fs/pidfs.c +21
-21@@ -915,6 +915,20 @@ static struct dentry *pidfs_fh_to_dentry(struct super_block *sb,return path.dentry; } +static struct file *pidfs_dentry_open (const struct path *path, +unsigned int flags, +const struct cred *cred) +{ +struct file *file; + +/* pidfds are always O_RDWR. */ + file =dentry_open (path, flags |O_RDWR , cred); +/* do_dentry_open() strips O_EXCL and O_TRUNC. */ +if (!IS_ERR (file)) + file->f_flags |= flags & (PIDFD_THREAD |PIDFD_AUTOKILL ); +return file; +} +/* * Make sure that we reject any nonsensical flags that users pass via * open_by_handle_at(). Note that PIDFD_THREAD is defined as O_EXCL, and @@ -939,18 +953,14 @@ static int pidfs_export_permission(struct handle_to_path_ctx *ctx,static struct file *pidfs_export_open (const struct path *path,unsigned int oflags) { -struct file *file; -/* -* Clear O_LARGEFILE as open_by_handle_at() forces it and raise -* O_RDWR as pidfds always are. +* Opening via file handle may never raise PIDFD_AUTOKILL. That can +* only be done at task creation! */ - oflags &= ~O_LARGEFILE ; - file =dentry_open (path, oflags |O_RDWR ,current_cred ()); -/* do_dentry_open() strips O_EXCL, which encodes PIDFD_THREAD. */ -if (!IS_ERR (file)) - file->f_flags |= oflags &PIDFD_THREAD ; -return file; +if (WARN_ON_ONCE (oflags &PIDFD_AUTOKILL )) +return ERR_PTR (-EINVAL ); +/* Clear O_LARGEFILE as open_by_handle_at() forces it. */ +return pidfs_dentry_open (path, oflags & ~O_LARGEFILE ,current_cred ()); }static const struct export_operations pidfs_export_operations = {@@ -1114,7 +1124,6 @@ static struct file_system_type pidfs_type = {struct file *pidfs_alloc_file (struct pid *pid,unsigned int flags) { -struct file *pidfd_file; struct path path __free(path_put) = {};int ret;@@ -1132,16 +1141,7 @@ struct file *pidfs_alloc_file(struct pid *pid, unsigned int flags)VFS_WARN_ON_ONCE (!pid->attr); flags &= ~PIDFD_STALE ; - flags |=O_RDWR ; - pidfd_file =dentry_open (&path, flags,current_cred ()); -/* -* Raise PIDFD_THREAD and PIDFD_AUTOKILL explicitly as -* do_dentry_open() strips O_EXCL and O_TRUNC. -*/ -if (!IS_ERR (pidfd_file)) - pidfd_file->f_flags |= (flags & (PIDFD_THREAD |PIDFD_AUTOKILL )); - -return pidfd_file; +return pidfs_dentry_open (&path, flags,current_cred ()); }void __init pidfs_init (void )Signature · 1 line
2.53.0
Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR
Hi Jan & Christian,
---- On Mon, 27 Jul 2026 23:57:14 +0800 Jan Kara <[email protected]> wrote ---
Quoted 45 lines · Christian Brauner
Quoted 38 lines · Jan Kara
Quoted 8 lines · Li Chen
Quoted 7 lines
PIDFD_THREADsharesO_EXCL.do_dentry_open()clearsO_EXCLafterpidfs_export_open()validates the flags, soopen_by_handle_at()silently turns a thread pidfd into a process pidfd.Restore
PIDFD_THREADon the opened file, matchingpidfs_alloc_file(), and cover the restored flag withF_GETFL.Quoted 7 lines
Quoted 1 line
Signed-off-by Li Chen <[email protected]>Makes sense. Just looking at this code, what about
PIDFD_AUTOKILL(akaO_TRUNC?)? Reading the commit introducing it07c3ef58223eit appears it was never meant to be directly requested by userspace so probablypidfs_export_open()needs to reject it. But that's unrelated issue to your patch so feel free to add:Good point. But
pidfs_export_open()is guarded bypidfs_export_permission()which does:#define
VALID_FILE_HANDLE_OPEN_FLAGS\
(O_RDONLY|O_WRONLY|O_RDWR|O_NONBLOCK|O_CLOEXEC|O_EXCL)if (oflags & ~(VALID_FILE_HANDLE_OPEN_FLAGS |
O_LARGEFILE))
return-EINVAL;and so is safe. But I've added the following patch on top which also splats if that is passed!
From
99a859485394eca8cf6271de09ae72ac3b846bdaMon Sep 17 00:00:00 2001
From: Christian Brauner <[email protected]>
Date: Wed, 22 Jul 2026 13:56:28 +0200
Subject: [PATCH] pidfs: addpidfs_dentry_open()helperBoth
pidfs_alloc_file()andpidfs_export_open()need to forceO_RDWRand reapply the pidfd flags thatdo_dentry_open()strips. Move the common logic into a helper.Quoted 5 lines
PIDFD_AUTOKILLis now part of the restore mask in the file handle path as well, butpidfs_export_permission()rejectsO_TRUNC, so this is a no-op there. But warn nonetheless.Signed-off-by Christian Brauner (Amutable) <[email protected]>
Quoted 8 lines
Nice! Feel free to add:
Reviewed-by Jan Kara <[email protected]>Both
pidfs_alloc_file()andpidfs_export_open()need to forceO_RDWRand reapply the pidfd flags thatdo_dentry_open()strips. Move the common logic into a helper.
Thanks both.
Jan, thanks for pointing out the PIDFD_AUTOKILL angle.
Christian, thanks for factoring this into pidfs_dentry_open() and applying the fix.
Regards,
Li
Quoted 1 line · Li Chen
pidfs: preserve thread pidfds reopened by file handle
The patch is correct but I'm likely massing it so a tiny shared helper is called in both the regular creation and the file handle creation.
---
Applied to the vfs.fixes branch of the vfs/vfs.git tree. Patches in the vfs.fixes 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.fixes
[1/1] pidfs: preserve thread pidfds reopened by file handle
https://git.kernel.org/vfs/vfs/c/c8d2d281934c