public inbox for linux-fsdevel@vger.kernel.org
 help / color / mirror / Atom feed
From: Amir Goldstein <amir73il@gmail.com>
To: Christian Brauner <brauner@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>,
	Miklos Szeredi <miklos@szeredi.hu>,
	Paul Moore <paul@paul-moore.com>, Gao Xiang <xiang@kernel.org>,
	linux-security-module@vger.kernel.org, selinux@vger.kernel.org,
	linux-erofs@lists.ozlabs.org, linux-fsdevel@vger.kernel.org,
	linux-unionfs@vger.kernel.org,
	syzbot+f34aab278bf5d664e2be@syzkaller.appspotmail.com
Subject: [PATCH v2] fs: allow backing file code to open an O_PATH file with negative dentry
Date: Fri, 20 Mar 2026 13:29:18 +0100	[thread overview]
Message-ID: <20260320122918.1726043-1-amir73il@gmail.com> (raw)

The fields f_mapping and f_wb_err are irrelevant for the O_PATH file
in backing_file_user_path_file().

Create a dedicated helper kernel_path_file_open(), which skips all the
generic code in do_dentry_open() and does only the essentials, so that
the internal O_PATH file could be opened with a negative dentry.

This is needed for backing_tmpfile_open() to open a backing O_PATH
tmpfile before instantiating the dentry.

The callers of backing_tmpfile_open() are responsible for calling
backing_tmpfile_finish() after making the path positive.

Reported-by: syzbot+f34aab278bf5d664e2be@syzkaller.appspotmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
---

Christian,

This patch fixes the syzbot report [1] that the
backing_file_user_path_file() patch [2] introduces.

Following your feedback on v1, this version makes an effort to stay
out of the way of main vfs execution paths and restrict the changes
to backing_file users.

This still introduced a temporary state of an O_PATH file with negative
path, but only for a short time and only for backing_file users and
ones that use backing_tmpfile_open() (i.e. only overlayfs), so the
risk is minimal.

WDYT?

Thanks,
Amir.

Changes since v1:
- Create helper for internal O_PATH open with negative path
- Create backing_tmpfile_finish() API to fixup the negative path

[1] https://syzkaller.appspot.com/bug?extid=f34aab278bf5d664e2be
[2] https://lore.kernel.org/linux-fsdevel/20260318131258.1457101-1-amir73il@gmail.com/

 fs/backing-file.c            |  6 ++++++
 fs/file_table.c              | 14 ++++++++++++--
 fs/internal.h                |  7 +++++++
 fs/open.c                    | 34 ++++++++++++++++++++++++++++++----
 fs/overlayfs/dir.c           |  2 ++
 include/linux/backing-file.h |  1 +
 6 files changed, 58 insertions(+), 6 deletions(-)

diff --git a/fs/backing-file.c b/fs/backing-file.c
index d0a64c2103907..3357d624eac96 100644
--- a/fs/backing-file.c
+++ b/fs/backing-file.c
@@ -80,6 +80,12 @@ struct file *backing_tmpfile_open(const struct path *user_path,
 }
 EXPORT_SYMBOL(backing_tmpfile_open);
 
+void backing_tmpfile_finish(struct file *file)
+{
+	backing_file_set_user_path_inode(file);
+}
+EXPORT_SYMBOL_GPL(backing_tmpfile_finish);
+
 struct backing_aio {
 	struct kiocb iocb;
 	refcount_t ref;
diff --git a/fs/file_table.c b/fs/file_table.c
index e8b4eb2bbff85..a4d1064d50896 100644
--- a/fs/file_table.c
+++ b/fs/file_table.c
@@ -69,11 +69,21 @@ EXPORT_SYMBOL_GPL(backing_file_user_path_file);
 
 int backing_file_open_user_path(struct file *f, const struct path *path)
 {
-	/* open an O_PATH file to reference the user path - should not fail */
-	return WARN_ON(vfs_open(path, &backing_file(f)->user_path_file));
+	if (WARN_ON(!(f->f_mode & FMODE_BACKING)))
+		return -EIO;
+	kernel_path_file_open(&backing_file(f)->user_path_file, path);
+	return 0;
 }
 EXPORT_SYMBOL_GPL(backing_file_open_user_path);
 
+void backing_file_set_user_path_inode(struct file *f)
+{
+	if (WARN_ON(!(f->f_mode & FMODE_BACKING)))
+		return;
+	file_set_d_inode(&backing_file(f)->user_path_file);
+}
+EXPORT_SYMBOL_GPL(backing_file_set_user_path_inode);
+
 static void destroy_file(struct file *f)
 {
 	security_file_free(f);
diff --git a/fs/internal.h b/fs/internal.h
index 7c44a58627ba3..4a9e5e00678d9 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -109,6 +109,13 @@ struct file *alloc_empty_file_noaccount(int flags, const struct cred *cred);
 struct file *alloc_empty_backing_file(int flags, const struct cred *cred,
 				      const struct cred *user_cred);
 int backing_file_open_user_path(struct file *f, const struct path *path);
+void backing_file_set_user_path_inode(struct file *f);
+void kernel_path_file_open(struct file *f, const struct path *path);
+
+static inline void file_set_d_inode(struct file *f)
+{
+	f->f_inode = d_inode(f->f_path.dentry);
+}
 
 static inline void file_put_write_access(struct file *file)
 {
diff --git a/fs/open.c b/fs/open.c
index 91f1139591abe..a7b3b04cd9ae7 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -884,10 +884,38 @@ static inline int file_get_write_access(struct file *f)
 	return error;
 }
 
+static const struct file_operations empty_fops = {};
+
+static void do_path_file_open(struct file *f)
+{
+	f->f_mode = FMODE_PATH | FMODE_OPENED;
+	file_set_fsnotify_mode(f, FMODE_NONOTIFY);
+	f->f_op = &empty_fops;
+}
+
+/**
+ * kernel_path_file_open - open an O_PATH file for kernel internal use
+ * @f:		pre-allocated file with f_flags and f_cred initialized
+ * @path:	path to reference (may have a negative dentry)
+ *
+ * Open a minimal O_PATH file that only references a path.
+ * Unlike vfs_open(), this does not require a positive dentry and does not
+ * set up f_mapping and other fields not needed for O_PATH.
+ * If path is negative at the time of this call, the caller is responsible for
+ * callingn backing_file_set_user_path_inode() after making the path positive.
+
+ */
+void kernel_path_file_open(struct file *f, const struct path *path)
+{
+	f->__f_path = *path;
+	path_get(&f->f_path);
+	file_set_d_inode(f);
+	do_path_file_open(f);
+}
+
 static int do_dentry_open(struct file *f,
 			  int (*open)(struct inode *, struct file *))
 {
-	static const struct file_operations empty_fops = {};
 	struct inode *inode = f->f_path.dentry->d_inode;
 	int error;
 
@@ -898,9 +926,7 @@ static int do_dentry_open(struct file *f,
 	f->f_sb_err = file_sample_sb_err(f);
 
 	if (unlikely(f->f_flags & O_PATH)) {
-		f->f_mode = FMODE_PATH | FMODE_OPENED;
-		file_set_fsnotify_mode(f, FMODE_NONOTIFY);
-		f->f_op = &empty_fops;
+		do_path_file_open(f);
 		return 0;
 	}
 
diff --git a/fs/overlayfs/dir.c b/fs/overlayfs/dir.c
index 5fd32ccc134d2..4010c87e10196 100644
--- a/fs/overlayfs/dir.c
+++ b/fs/overlayfs/dir.c
@@ -1408,6 +1408,8 @@ static int ovl_create_tmpfile(struct file *file, struct dentry *dentry,
 			err = ovl_instantiate(dentry, inode, newdentry, false, file);
 			if (!err) {
 				file->private_data = of;
+				/* user_path_file was opened with a negative path */
+				backing_tmpfile_finish(realfile);
 			} else {
 				dput(newdentry);
 				ovl_file_free(of);
diff --git a/include/linux/backing-file.h b/include/linux/backing-file.h
index 8afba93f3ce07..52ac51ada6ff9 100644
--- a/include/linux/backing-file.h
+++ b/include/linux/backing-file.h
@@ -47,6 +47,7 @@ struct file *backing_tmpfile_open(const struct path *user_path,
 				  const struct cred *user_cred, int flags,
 				  const struct path *real_parentpath,
 				  umode_t mode, const struct cred *cred);
+void backing_tmpfile_finish(struct file *file);
 ssize_t backing_file_read_iter(struct file *file, struct iov_iter *iter,
 			       struct kiocb *iocb, int flags,
 			       struct backing_file_ctx *ctx);
-- 
2.53.0


             reply	other threads:[~2026-03-20 12:29 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-20 12:29 Amir Goldstein [this message]
2026-03-20 12:33 ` [PATCH v2] fs: allow backing file code to open an O_PATH file with negative dentry Amir Goldstein

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260320122918.1726043-1-amir73il@gmail.com \
    --to=amir73il@gmail.com \
    --cc=brauner@kernel.org \
    --cc=linux-erofs@lists.ozlabs.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-security-module@vger.kernel.org \
    --cc=linux-unionfs@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=paul@paul-moore.com \
    --cc=selinux@vger.kernel.org \
    --cc=syzbot+f34aab278bf5d664e2be@syzkaller.appspotmail.com \
    --cc=viro@zeniv.linux.org.uk \
    --cc=xiang@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox