Linux Security Modules development
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Cai Xinchen" <caixinchen1@huawei.com>
Cc: bpf@vger.kernel.org, linux-security-module@vger.kernel.org,
	linux-trace-kernel@vger.kernel.org, selinux@vger.kernel.org,
	dsterba@suse.cz, "Chuck Lever" <cel@kernel.org>,
	"Jeff Layton" <jlayton@kernel.org>,
	"Anna Schumaker" <anna@kernel.org>
Subject: Re: [PATCH -next -v4 03/17] fs: pass struct path to xattr helpers
Date: Tue, 17 Nov 2026 14:59:14 +0000	[thread overview]
Message-ID: <sashiko-outbox-219006@kernel.org> (raw)
In-Reply-To: <20261117150542.1169451-4-caixinchen1@huawei.com>

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Medium] ksmbd: dotdot mount walk in
  ksmbd_populate_dot_dotdot_entries() escapes share_root
- [Low] nfsd: reverse Christmas tree order violated in nfsd_getxattr()
  and nfsd_listxattr()
--

commit 659e941625d287aa476faa3056e5fd5cc15c2d66
Author: Cai Xinchen <caixinchen1@huawei.com>

fs: pass struct path to xattr helpers

This commit updates vfs_setxattr(), vfs_getxattr(), vfs_listxattr(),
vfs_removexattr(), and the __vfs_*_locked() helpers along with their
callers in ksmbd, nfsd, overlayfs, cachefiles, ecryptfs, and LSMs to
pass a struct path instead of a struct dentry.

> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
> index 1c686b82e70a0..aef4097e9ef37 100644
> --- a/fs/nfsd/vfs.c
> +++ b/fs/nfsd/vfs.c
[ ... ]
> @@ -2676,23 +2678,24 @@ __be32
>  nfsd_getxattr(struct svc_rqst *rqstp, struct svc_fh *fhp, char *name,
>  	      void **bufp, int *lenp)
>  {
> +	struct path path;
> +	struct inode *inode;
> +	char *buf;
>  	ssize_t len;
>  	__be32 err;
> -	char *buf;
> -	struct inode *inode;
> -	struct dentry *dentry;
[ ... ]
> @@ -2749,23 +2752,24 @@ __be32
>  nfsd_listxattr(struct svc_rqst *rqstp, struct svc_fh *fhp, char **bufp,
>  	       int *lenp)
>  {
> +	struct path path;
> +	struct inode *inode;
> +	char *buf;
>  	ssize_t len;
>  	__be32 err;
> -	char *buf;
> -	struct inode *inode;
> -	struct dentry *dentry;

[Severity: Low]
This isn't a bug, but NFSD keeps local variable declarations in reverse
Christmas tree order, so struct inode *inode; should come before struct
path path; in both nfsd_getxattr() and nfsd_listxattr().

> diff --git a/fs/smb/server/smb_common.c b/fs/smb/server/smb_common.c
> index e7a987af64ba5..a6da27a7d7cbb 100644
> --- a/fs/smb/server/smb_common.c
> +++ b/fs/smb/server/smb_common.c
> @@ -467,37 +467,82 @@ int ksmbd_populate_dot_dotdot_entries(struct ksmbd_work *work, int info_level,
>  {
>  	int i, rc = 0;
>  	struct ksmbd_conn *conn = work->conn;
> -	const struct mnt_idmap *idmap = file_mnt_idmap(dir->filp);
> +	const struct path *const share_root = &work->tcon->share_conf->vfs_path;
>  
>  	for (i = 0; i < 2; i++) {
>  		struct kstat kstat;
>  		struct ksmbd_kstat ksmbd_kstat;
> -		struct dentry *dentry;
> +		struct path path = dir->filp->f_path;
> +		const struct mnt_idmap *idmap;
> +		struct dentry *parent;
>  
>  		if (!dir->dot_dotdot[i]) { /* fill dot entry info */
>  			if (i == 0) {
>  				d_info->name = ".";
>  				d_info->name_len = 1;
> -				dentry = dget(dir->filp->f_path.dentry);
> +				path_get(&path);
>  			} else {
>  				d_info->name = "..";
>  				d_info->name_len = 2;
> -				dentry = dget_parent(dir->filp->f_path.dentry);
> +				path_get(&path);
> +				/*
> +				 * A ".." entry must not walk above the
> +				 * share root, so it designates the share
> +				 * root itself in that case.
> +				 */
> +				if (!path_equal(&path, share_root)) {
> +					/*
> +					 * The parent of a mount root dentry
> +					 * is not covered by the same mount,
> +					 * so follow the mount stack up like
> +					 * a ".." path walk would, to keep
> +					 * the path valid.
> +					 */
> +					while (path.dentry == path.mnt->mnt_root &&
> +					       follow_up(&path))
> +						;
> +					if (path.dentry == path.mnt->mnt_root) {
> +						/*
> +						 * The mount is detached, for
> +						 * example after a lazy umount
> +						 * which the open file pins
> +						 * down, so the parent would
> +						 * not be covered by path.mnt:
> +						 * report the share root
> +						 * instead, as for any other
> +						 * ".." above the share root.
> +						 */
> +						path_put(&path);
> +						path_get(share_root);
> +						path = *share_root;
> +					} else {
> +						parent = dget_parent(path.dentry);
> +						dput(path.dentry);
> +						path.dentry = parent;
> +					}
> +				}
>  			}

[Severity: Medium]
Can the ".." resolution walk above share_root when a mount is stacked on
top of share_root?

If a submount is mounted on share_root and entered via follow_down() when
KSMBD_SHARE_FLAG_CROSSMNT is enabled, dir->filp->f_path is the submount
root, so the initial path_equal(&path, share_root) check is false.
follow_up(&path) then steps onto share_root, but neither the while loop
nor the dget_parent() branch rechecks path_equal(&path, share_root).
If share_root is itself a mountpoint, the loop continues calling
follow_up(&path) into the parent mount; otherwise the else branch calls
dget_parent(path.dentry) and steps to the parent directory of share_root,
exposing host directory attributes and xattrs outside the share to the
SMB client via ksmbd_vfs_fill_dentry_attrs().

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117150542.1169451-1-caixinchen1@huawei.com?part=3

  reply	other threads:[~2026-11-17 14:59 UTC|newest]

Thread overview: 33+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17 15:05 [PATCH -next -v4 00/17] landlock: Add READ_METADATA and WRITE_METADATA access rights Cai Xinchen
2026-11-17 15:05 ` [PATCH -next -v4 01/17] fs: pass struct path to notify_change() Cai Xinchen
2026-11-17 14:48   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 02/17] LSM: pass struct path to the inode_setsecctx hook Cai Xinchen
2026-11-17 14:43   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 03/17] fs: pass struct path to xattr helpers Cai Xinchen
2026-11-17 14:59   ` sashiko-bot [this message]
2026-11-17 15:05 ` [PATCH -next -v4 04/17] fs: pass struct path to POSIX ACL helpers Cai Xinchen
2026-11-17 14:45   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 05/17] LSM: pass struct path and struct file to the inode xattr hooks Cai Xinchen
2026-11-17 15:03   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 06/17] LSM: pass struct path and struct file to the inode posix acl hooks Cai Xinchen
2026-11-17 15:07   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 07/17] LSM: pass struct path and struct file to the inode fileattr hooks Cai Xinchen
2026-11-17 14:51   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 08/17] LSM: rename inode_setattr to path_setattr and pass struct path and file Cai Xinchen
2026-11-17 15:11   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 09/17] LSM: rename inode_getattr to path_getattr and pass struct file Cai Xinchen
2026-11-17 15:05 ` [PATCH -next -v4 10/17] LSM: rename the inode xattr hooks to path_xattr hooks Cai Xinchen
2026-11-17 14:55   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 11/17] LSM: rename the inode posix acl hooks to path_*_acl hooks Cai Xinchen
2026-11-17 14:55   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 12/17] LSM: rename the inode fileattr hooks to path_file_* hooks Cai Xinchen
2026-11-17 15:02   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 13/17] LSM: rename inode_setsecctx to path_setsecctx Cai Xinchen
2026-11-17 14:59   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 14/17] landlock: Add READ_METADATA and WRITE_METADATA access rights Cai Xinchen
2026-11-17 15:05 ` [PATCH -next -v4 15/17] selftests/landlock: Add tests for metadata " Cai Xinchen
2026-11-17 15:12   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 16/17] samples/landlock: Add metadata rights to sandboxer Cai Xinchen
2026-11-17 15:04   ` sashiko-bot
2026-11-17 15:05 ` [PATCH -next -v4 17/17] Documentation: Update landlock doc for metadata rights Cai Xinchen
2026-11-17 15:05   ` sashiko-bot

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=sashiko-outbox-219006@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=anna@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=caixinchen1@huawei.com \
    --cc=cel@kernel.org \
    --cc=dsterba@suse.cz \
    --cc=jlayton@kernel.org \
    --cc=linux-security-module@vger.kernel.org \
    --cc=linux-trace-kernel@vger.kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    --cc=selinux@vger.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