FAILED: patch "[PATCH] sunrpc: fix use-after-free in __rpc_clnt_handle_event and" failed to apply to 6.12-stable tree

PATCH New [email protected] 2 participants 5 messages Sep 3 – Sep 8

Subsystems (4)

Networking Stable branch NFS client NFS server

lore

Share & tools

lore

[email protected]

author maintainer

2026-09-03 14:30 UTC #0 lore raw

The patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkout FETCH_HEAD
git cherry-pick -x 932a8cf6abb2b2f8677b79153a823108d8861fe2
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026090317-curled-track-17d1@gregkh' --subject-prefix 'PATCH 6.12.y' 'HEAD^..'

Possible dependencies:

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 932a8cf6abb2b2f8677b79153a823108d8861fe2 Mon Sep 17 00:00:00 2001
From: Luxiao Xu <[email protected]>
Date: Tue, 7 Jul 2026 13:20:47 +0800
Subject: [PATCH] sunrpc: fix use-after-free in __rpc_clnt_handle_event and
__rpc_clnt_remove_pipedir

Normal client creation goes through rpc_setup_pipedir(), which records clnt->pipefs_sb, but the mount-event path in __rpc_clnt_handle_event() calls rpc_setup_pipedir_sb() directly and never refreshes that field. The umount path also removes the directory without clearing clnt->pipefs_sb.

After a late pipefs mount or any remount, rpc_clnt_remove_pipedir() compares the current superblock against a stale pipefs_sb pointer and skips cleanup, leaving pipefs dentries whose inode private data still points at a freed rpc_clnt, leading to a potential use-after-free during subsequent rpc_info_open() or rpc_show_info() calls.

Fix this by properly updating clnt->pipefs_sb upon mount events and clearing it during unmount or failure paths.

Fixes bfca5fb4e97c ("SUNRPC: Fix RPC client cleaned up the freed pipefs dentries")

Cc [email protected]

Reported-by Yuan Tan <[email protected]>

Reported-by Xin Liu <[email protected]>

Reviewed-by Ren Wei <[email protected]>

Assisted-by Codex:gpt-5.4

Signed-off-by Luxiao Xu <[email protected]>

Signed-off-by Trond Myklebust <[email protected]>

net/sunrpc/clnt.c +17 -5

index efa26899bc7d..6cedc824cf82 100644
@@ -96,7 +96,10 @@ static void rpc_unregister_client(struct rpc_clnt *clnt)
 
 static void __rpc_clnt_remove_pipedir(struct rpc_clnt *clnt)
 {
-	rpc_remove_client_dir(clnt);
+	if (clnt->pipefs_sb) {
+		rpc_remove_client_dir(clnt);
+		clnt->pipefs_sb = NULL;
+	}
 }
 
 static void rpc_clnt_remove_pipedir(struct rpc_clnt *clnt)

@@ -177,19 +180,28 @@ static int rpc_clnt_skip_event(struct rpc_clnt *clnt, unsigned long event)
 }
 
 static int __rpc_clnt_handle_event(struct rpc_clnt *clnt, unsigned long event,
-				   struct super_block *sb)
+				    struct super_block *sb)
 {
+	int err = 0;
+
 	switch (event) {
 	case RPC_PIPEFS_MOUNT:
-		return rpc_setup_pipedir_sb(sb, clnt);
+		clnt->pipefs_sb = sb;
+		err = rpc_setup_pipedir_sb(sb, clnt);
+		if (err)
+			clnt->pipefs_sb = NULL;
+		break;
 	case RPC_PIPEFS_UMOUNT:
-		__rpc_clnt_remove_pipedir(clnt);
+		if (clnt->pipefs_sb == sb) {
+			__rpc_clnt_remove_pipedir(clnt);
+			clnt->pipefs_sb = NULL;
+		}
 		break;
 	default:
 		printk(KERN_ERR "%s: unknown event: %ld\n", __func__, event);
 		return -ENOTSUPP;
 	}
-	return 0;
+	return err;
 }
 
 static int __rpc_pipefs_event(struct rpc_clnt *clnt, unsigned long event,

Sasha Levin

maintainer

2026-09-08 22:37 UTC #1 lore raw

PATCH 1/4 rpc_populate(): lift cleanup into callers

From: Al Viro <[email protected]>

[ Upstream commit 8be22c49646e47452e74aa0b97ea50fb04c271ed ]

rpc_populate() is called either from fill_super (where we don't need to remove any files on failure - rpc_kill_sb() will take them all out anyway) or from rpc_mkdir_populate(), where we need to remove the directory we'd been trying to populate along with whatever we'd put into it before we failed. Simpler to combine that into simple_recursive_removal() there.

Note that rpc_pipe is overlocking directories quite a bit - locked parent is no obstacle to finding a child in dcache, so keeping it locked won't prevent userland observing a partially built subtree.

All we need is to follow minimal VFS requirements; it's not as if clients used directory locking for exclusion - tree changes are serialized, but that's done on ->pipefs_sb_lock.

Reviewed-by Jeff Layton <[email protected]>

Signed-off-by Al Viro <[email protected]>

Stable-dep-of: 932a8cf6abb2 ("sunrpc: fix use-after-free in __rpc_clnt_handle_event and __rpc_clnt_remove_pipedir")

Signed-off-by Sasha Levin <[email protected]>

---
 include/linux/sunrpc/rpc_pipe_fs.h |  2 +-
 net/sunrpc/clnt.c                  | 36 ++++++++++++------------------
 net/sunrpc/rpc_pipe.c              | 26 ++++++++++-----------
 3 files changed, 28 insertions(+), 36 deletions(-)

include/linux/sunrpc/rpc_pipe_fs.h +1 -1

index 3b35b6f6533aa..f4b9ea5c83ea1 100644
@@ -98,7 +98,7 @@ static inline bool rpc_msg_is_inflight(const struct rpc_pipe_msg *msg) {
 }
 
 struct rpc_clnt;
-extern struct dentry *rpc_create_client_dir(struct dentry *, const char *, struct rpc_clnt *);
+extern int rpc_create_client_dir(struct dentry *, const char *, struct rpc_clnt *);
 extern int rpc_remove_client_dir(struct rpc_clnt *);
 
 extern void rpc_init_pipe_dir_head(struct rpc_pipe_dir_head *pdh);

net/sunrpc/clnt.c +14 ~~-22~~

index e88e21c2ab329..1eca77bd03930 100644
@@ -112,47 +112,46 @@ static void rpc_clnt_remove_pipedir(struct rpc_clnt *clnt)
 	}
 }
 
-static struct dentry *rpc_setup_pipedir_sb(struct super_block *sb,
+static int rpc_setup_pipedir_sb(struct super_block *sb,
 				    struct rpc_clnt *clnt)
 {
 	static uint32_t clntid;
 	const char *dir_name = clnt->cl_program->pipe_dir_name;
 	char name[15];
-	struct dentry *dir, *dentry;
+	struct dentry *dir;
+	int err;
 
 	dir = rpc_d_lookup_sb(sb, dir_name);
 	if (dir == NULL) {
 		pr_info("RPC: pipefs directory doesn't exist: %s\n", dir_name);
-		return dir;
+		return -ENOENT;
 	}
 	for (;;) {
 		snprintf(name, sizeof(name), "clnt%x", (unsigned int)clntid++);
 		name[sizeof(name) - 1] = '\0';
-		dentry = rpc_create_client_dir(dir, name, clnt);
-		if (!IS_ERR(dentry))
+		err = rpc_create_client_dir(dir, name, clnt);
+		if (!err)
 			break;
-		if (dentry == ERR_PTR(-EEXIST))
+		if (err == -EEXIST)
 			continue;
 		printk(KERN_INFO "RPC: Couldn't create pipefs entry"
-				" %s/%s, error %ld\n",
-				dir_name, name, PTR_ERR(dentry));
+				" %s/%s, error %d\n",
+				dir_name, name, err);
 		break;
 	}
 	dput(dir);
-	return dentry;
+	return err;
 }
 
 static int
 rpc_setup_pipedir(struct super_block *pipefs_sb, struct rpc_clnt *clnt)
 {
-	struct dentry *dentry;
-
 	clnt->pipefs_sb = pipefs_sb;
 
 	if (clnt->cl_program->pipe_dir_name != NULL) {
-		dentry = rpc_setup_pipedir_sb(pipefs_sb, clnt);
-		if (IS_ERR(dentry))
-			return PTR_ERR(dentry);
+		int err = rpc_setup_pipedir_sb(pipefs_sb, clnt);
+		if (err && err != -ENOENT)
+			return err;
 	}
 	return 0;
 }

@@ -180,16 +179,9 @@ static int rpc_clnt_skip_event(struct rpc_clnt *clnt, unsigned long event)
 static int __rpc_clnt_handle_event(struct rpc_clnt *clnt, unsigned long event,
 				   struct super_block *sb)
 {
-	struct dentry *dentry;
-
 	switch (event) {
 	case RPC_PIPEFS_MOUNT:
-		dentry = rpc_setup_pipedir_sb(sb, clnt);
-		if (!dentry)
-			return -ENOENT;
-		if (IS_ERR(dentry))
-			return PTR_ERR(dentry);
-		break;
+		return rpc_setup_pipedir_sb(sb, clnt);
 	case RPC_PIPEFS_UMOUNT:
 		__rpc_clnt_remove_pipedir(clnt);
 		break;

net/sunrpc/rpc_pipe.c +13 ~~-13~~

index eadc00410ebc5..43119907010e3 100644
@@ -725,7 +725,6 @@ static int rpc_populate(struct dentry *parent,
 	inode_unlock(dir);
 	return 0;
 out_bad:
-	__rpc_depopulate(parent, files, start, eof);
 	inode_unlock(dir);
 	printk(KERN_WARNING "%s: %s failed to populate directory %pd\n",
 			__FILE__, __func__, parent);

@@ -749,14 +748,15 @@ static struct dentry *rpc_mkdir_populate(struct dentry *parent,
 		goto out_err;
 	if (populate != NULL) {
 		error = populate(dentry, args_populate);
-		if (error)
-			goto err_rmdir;
+		if (error) {
+			inode_unlock(dir);
+			simple_recursive_removal(dentry, NULL);
+			return ERR_PTR(error);
+		}
 	}
 out:
 	inode_unlock(dir);
 	return dentry;
-err_rmdir:
-	__rpc_rmdir(dir, dentry);
 out_err:
 	dentry = ERR_PTR(error);
 	goto out;

@@ -1046,19 +1046,19 @@ static void rpc_clntdir_depopulate(struct dentry *dentry)
  * information about the client, together with any "pipes" that may
  * later be created using rpc_mkpipe().
  */
-struct dentry *rpc_create_client_dir(struct dentry *dentry,
-				   const char *name,
-				   struct rpc_clnt *rpc_client)
+int rpc_create_client_dir(struct dentry *dentry,
+			  const char *name,
+			  struct rpc_clnt *rpc_client)
 {
 	struct dentry *ret;
 
 	ret = rpc_mkdir_populate(dentry, name, 0555, NULL,
 				 rpc_clntdir_populate, rpc_client);
-	if (!IS_ERR(ret)) {
-		rpc_client->cl_pipedir_objects.pdh_dentry = ret;
-		rpc_create_pipe_dir_objects(&rpc_client->cl_pipedir_objects);
-	}
-	return ret;
+	if (IS_ERR(ret))
+		return PTR_ERR(ret);
+	rpc_client->cl_pipedir_objects.pdh_dentry = ret;
+	rpc_create_pipe_dir_objects(&rpc_client->cl_pipedir_objects);
+	return 0;
 }
 
 /**

Signature · 1 line

2.53.0

Sasha Levin

maintainer

2026-09-08 22:37 UTC #2 lore raw

PATCH 2/4 rpc_mkpipe_dentry(): saner calling conventions

From: Al Viro <[email protected]>

[ Upstream commit 19a6314a997f6adde0c100ecf9224d1ab43c9603 ]

Instead of returning a dentry or ERR_PTR(-E...), return 0 and store dentry into pipe->dentry on success and return -E... on failure.

Callers are happier that way...

NOTE: dummy rpc_pipe is getting ->dentry set; we never access that, since we
1) never call rpc_unlink() for it (dentry is taken out by
->kill_sb())
2) never call rpc_queue_upcall() for it (writing to that
sucker fails; no downcalls are ever submitted, so no replies are going to arrive)
IOW, having that ->dentry set (and left dangling) is harmless, if ugly; cleaner solution will take more massage.

Reviewed-by Jeff Layton <[email protected]>

Signed-off-by Al Viro <[email protected]>

Stable-dep-of: 932a8cf6abb2 ("sunrpc: fix use-after-free in __rpc_clnt_handle_event and __rpc_clnt_remove_pipedir")

Signed-off-by Sasha Levin <[email protected]>

---
 fs/nfs/blocklayout/rpc_pipefs.c    | 41 ++++++++++++------------------
 fs/nfs/nfs4idmap.c                 |  8 +-----
 fs/nfsd/nfs4recover.c              | 37 ++++++++++-----------------
 include/linux/sunrpc/rpc_pipe_fs.h |  2 +-
 net/sunrpc/auth_gss/auth_gss.c     |  7 +----
 net/sunrpc/rpc_pipe.c              | 31 +++++++++++-----------
 6 files changed, 49 insertions(+), 77 deletions(-)

fs/nfs/blocklayout/rpc_pipefs.c +16 ~~-25~~

index d8d50a88de04f..fda75dfe97b79 100644
@@ -141,17 +141,18 @@ static const struct rpc_pipe_ops bl_upcall_ops = {
 	.destroy_msg	= bl_pipe_destroy_msg,
 };
 
-static struct dentry *nfs4blocklayout_register_sb(struct super_block *sb,
+static int nfs4blocklayout_register_sb(struct super_block *sb,
 					    struct rpc_pipe *pipe)
 {
-	struct dentry *dir, *dentry;
+	struct dentry *dir;
+	int err;
 
 	dir = rpc_d_lookup_sb(sb, NFS_PIPE_DIRNAME);
 	if (dir == NULL)
-		return ERR_PTR(-ENOENT);
-	dentry = rpc_mkpipe_dentry(dir, "blocklayout", NULL, pipe);
+		return -ENOENT;
+	err = rpc_mkpipe_dentry(dir, "blocklayout", NULL, pipe);
 	dput(dir);
-	return dentry;
+	return err;
 }
 
 static void nfs4blocklayout_unregister_sb(struct super_block *sb,

@@ -167,7 +168,6 @@ static int rpc_pipefs_event(struct notifier_block *nb, unsigned long event,
 	struct super_block *sb = ptr;
 	struct net *net = sb->s_fs_info;
 	struct nfs_net *nn = net_generic(net, nfs_net_id);
-	struct dentry *dentry;
 	int ret = 0;
 
 	if (!try_module_get(THIS_MODULE))

@@ -180,12 +180,7 @@ static int rpc_pipefs_event(struct notifier_block *nb, unsigned long event,
 
 	switch (event) {
 	case RPC_PIPEFS_MOUNT:
-		dentry = nfs4blocklayout_register_sb(sb, nn->bl_device_pipe);
-		if (IS_ERR(dentry)) {
-			ret = PTR_ERR(dentry);
-			break;
-		}
-		nn->bl_device_pipe->dentry = dentry;
+		ret = nfs4blocklayout_register_sb(sb, nn->bl_device_pipe);
 		break;
 	case RPC_PIPEFS_UMOUNT:
 		if (nn->bl_device_pipe->dentry)

@@ -203,18 +198,17 @@ static struct notifier_block nfs4blocklayout_block = {
 	.notifier_call = rpc_pipefs_event,
 };
 
-static struct dentry *nfs4blocklayout_register_net(struct net *net,
-						   struct rpc_pipe *pipe)
+static int nfs4blocklayout_register_net(struct net *net, struct rpc_pipe *pipe)
 {
 	struct super_block *pipefs_sb;
-	struct dentry *dentry;
+	int ret;
 
 	pipefs_sb = rpc_get_sb_net(net);
 	if (!pipefs_sb)
-		return NULL;
-	dentry = nfs4blocklayout_register_sb(pipefs_sb, pipe);
+		return 0;
+	ret = nfs4blocklayout_register_sb(pipefs_sb, pipe);
 	rpc_put_sb_net(net);
-	return dentry;
+	return ret;
 }
 
 static void nfs4blocklayout_unregister_net(struct net *net,

@@ -232,20 +226,17 @@ static void nfs4blocklayout_unregister_net(struct net *net,
 static int nfs4blocklayout_net_init(struct net *net)
 {
 	struct nfs_net *nn = net_generic(net, nfs_net_id);
-	struct dentry *dentry;
+	int err;
 
 	mutex_init(&nn->bl_mutex);
 	init_waitqueue_head(&nn->bl_wq);
 	nn->bl_device_pipe = rpc_mkpipe_data(&bl_upcall_ops, 0);
 	if (IS_ERR(nn->bl_device_pipe))
 		return PTR_ERR(nn->bl_device_pipe);
-	dentry = nfs4blocklayout_register_net(net, nn->bl_device_pipe);
-	if (IS_ERR(dentry)) {
+	err = nfs4blocklayout_register_net(net, nn->bl_device_pipe);
+	if (unlikely(err))
 		rpc_destroy_pipe_data(nn->bl_device_pipe);
-		return PTR_ERR(dentry);
-	}
-	nn->bl_device_pipe->dentry = dentry;
-	return 0;
+	return err;
 }
 
 static void nfs4blocklayout_net_exit(struct net *net)

fs/nfs/nfs4idmap.c +1 -7

index 25a7c771cfd89..86d1e91f1e3bd 100644
@@ -436,14 +436,8 @@ static int nfs_idmap_pipe_create(struct dentry *dir,
 		struct rpc_pipe_dir_object *pdo)
 {
 	struct idmap *idmap = pdo->pdo_data;
-	struct rpc_pipe *pipe = idmap->idmap_pipe;
-	struct dentry *dentry;
 
-	dentry = rpc_mkpipe_dentry(dir, "idmap", idmap, pipe);
-	if (IS_ERR(dentry))
-		return PTR_ERR(dentry);
-	pipe->dentry = dentry;
-	return 0;
+	return rpc_mkpipe_dentry(dir, "idmap", idmap, idmap->idmap_pipe);
 }
 
 static const struct rpc_pipe_dir_object_ops nfs_idmap_pipe_dir_object_ops = {

fs/nfsd/nfs4recover.c +14 ~~-23~~

index a8182962a2861..5ec98a95db8f7 100644
@@ -926,17 +926,18 @@ static const struct rpc_pipe_ops cld_upcall_ops = {
 	.destroy_msg	= cld_pipe_destroy_msg,
 };
 
-static struct dentry *
+static int
 nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe)
 {
-	struct dentry *dir, *dentry;
+	struct dentry *dir;
+	int err;
 
 	dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR);
 	if (dir == NULL)
-		return ERR_PTR(-ENOENT);
-	dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe);
+		return -ENOENT;
+	err = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe);
 	dput(dir);
-	return dentry;
+	return err;
 }
 
 static void

@@ -946,18 +947,18 @@ nfsd4_cld_unregister_sb(struct rpc_pipe *pipe)
 		rpc_unlink(pipe->dentry);
 }
 
-static struct dentry *
+static int
 nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe)
 {
 	struct super_block *sb;
-	struct dentry *dentry;
+	int err;
 
 	sb = rpc_get_sb_net(net);
 	if (!sb)
-		return NULL;
-	dentry = nfsd4_cld_register_sb(sb, pipe);
+		return 0;
+	err = nfsd4_cld_register_sb(sb, pipe);
 	rpc_put_sb_net(net);
-	return dentry;
+	return err;
 }
 
 static void

@@ -977,7 +978,6 @@ static int
 __nfsd4_init_cld_pipe(struct net *net)
 {
 	int ret;
-	struct dentry *dentry;
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 	struct cld_net *cn;

@@ -998,13 +998,10 @@ __nfsd4_init_cld_pipe(struct net *net)
 	spin_lock_init(&cn->cn_lock);
 	INIT_LIST_HEAD(&cn->cn_list);
 
-	dentry = nfsd4_cld_register_net(net, cn->cn_pipe);
-	if (IS_ERR(dentry)) {
-		ret = PTR_ERR(dentry);
+	ret = nfsd4_cld_register_net(net, cn->cn_pipe);
+	if (unlikely(ret))
 		goto err_destroy_data;
-	}
 
-	cn->cn_pipe->dentry = dentry;
 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
 	cn->cn_has_legacy = false;
 #endif

@@ -2132,7 +2129,6 @@ rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr)
 	struct net *net = sb->s_fs_info;
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 	struct cld_net *cn = nn->cld_net;
-	struct dentry *dentry;
 	int ret = 0;
 
 	if (!try_module_get(THIS_MODULE))

@@ -2145,12 +2141,7 @@ rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr)
 
 	switch (event) {
 	case RPC_PIPEFS_MOUNT:
-		dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe);
-		if (IS_ERR(dentry)) {
-			ret = PTR_ERR(dentry);
-			break;
-		}
-		cn->cn_pipe->dentry = dentry;
+		ret = nfsd4_cld_register_sb(sb, cn->cn_pipe);
 		break;
 	case RPC_PIPEFS_UMOUNT:
 		if (cn->cn_pipe->dentry)

include/linux/sunrpc/rpc_pipe_fs.h +1 -1

index f4b9ea5c83ea1..c3bab2313e8cc 100644
@@ -127,7 +127,7 @@ extern void rpc_remove_cache_dir(struct dentry *);
 
 struct rpc_pipe *rpc_mkpipe_data(const struct rpc_pipe_ops *ops, int flags);
 void rpc_destroy_pipe_data(struct rpc_pipe *pipe);
-extern struct dentry *rpc_mkpipe_dentry(struct dentry *, const char *, void *,
+extern int rpc_mkpipe_dentry(struct dentry *, const char *, void *,
 					struct rpc_pipe *);
 extern int rpc_unlink(struct dentry *);
 extern int register_rpc_pipefs(void);

net/sunrpc/auth_gss/auth_gss.c +1 -6

index 759d1e03f8c31..8407a4464680e 100644
@@ -902,13 +902,8 @@ static int gss_pipe_dentry_create(struct dentry *dir,
 		struct rpc_pipe_dir_object *pdo)
 {
 	struct gss_pipe *p = pdo->pdo_data;
-	struct dentry *dentry;
 
-	dentry = rpc_mkpipe_dentry(dir, p->name, p->clnt, p->pipe);
-	if (IS_ERR(dentry))
-		return PTR_ERR(dentry);
-	p->pipe->dentry = dentry;
-	return 0;
+	return rpc_mkpipe_dentry(dir, p->name, p->clnt, p->pipe);
 }
 
 static const struct rpc_pipe_dir_object_ops gss_pipe_dir_object_ops = {

net/sunrpc/rpc_pipe.c +16 ~~-15~~

index 43119907010e3..5556e25ffc35a 100644
@@ -799,7 +799,7 @@ static int rpc_rmdir_depopulate(struct dentry *dentry,
  * The @private argument passed here will be available to all these methods
  * from the file pointer, via RPC_I(file_inode(file))->private.
  */
-struct dentry *rpc_mkpipe_dentry(struct dentry *parent, const char *name,
+int rpc_mkpipe_dentry(struct dentry *parent, const char *name,
 				 void *private, struct rpc_pipe *pipe)
 {
 	struct dentry *dentry;

@@ -814,21 +814,19 @@ struct dentry *rpc_mkpipe_dentry(struct dentry *parent, const char *name,
 
 	inode_lock_nested(dir, I_MUTEX_PARENT);
 	dentry = __rpc_lookup_create_exclusive(parent, name);
-	if (IS_ERR(dentry))
-		goto out;
+	if (IS_ERR(dentry)) {
+		inode_unlock(dir);
+		return PTR_ERR(dentry);
+	}
 	err = __rpc_mkpipe_dentry(dir, dentry, umode, &rpc_pipe_fops,
 				  private, pipe);
-	if (err)
-		goto out_err;
-out:
+	if (unlikely(err))
+		pr_warn("%s() failed to create pipe %pd/%s (errno = %d)\n",
+			__func__, parent, name, err);
+	else
+		pipe->dentry = dentry;
 	inode_unlock(dir);
-	return dentry;
-out_err:
-	dentry = ERR_PTR(err);
-	printk(KERN_WARNING "%s: %s() failed to create pipe %pd/%s (errno = %d)\n",
-			__FILE__, __func__, parent, name,
-			err);
-	goto out;
+	return err;
 }
 EXPORT_SYMBOL_GPL(rpc_mkpipe_dentry);

@@ -1326,10 +1324,13 @@ rpc_gssd_dummy_populate(struct dentry *root, struct rpc_pipe *pipe_data)
 		goto out;
 	}
 
-	pipe_dentry = rpc_mkpipe_dentry(clnt_dentry, "gssd", NULL, pipe_data);
-	if (IS_ERR(pipe_dentry)) {
+	ret = rpc_mkpipe_dentry(clnt_dentry, "gssd", NULL, pipe_data);
+	if (ret) {
 		__rpc_depopulate(clnt_dentry, gssd_dummy_info_file, 0, 1);
 		__rpc_depopulate(gssd_dentry, gssd_dummy_clnt_dir, 0, 1);
+		pipe_dentry = ERR_PTR(ret);
+	} else {
+		pipe_dentry = pipe_data->dentry;
 	}
 out:
 	dput(clnt_dentry);

Signature · 1 line

2.53.0

Sasha Levin

maintainer

2026-09-08 22:37 UTC #3 lore raw

PATCH 3/4 rpc_pipe: don't overdo directory locking

In reply to Sasha Levin

From: Al Viro <[email protected]>

[ Upstream commit 41a6b9e52b21544c334830bfe68fb78d6677caa0 ]

Don't try to hold directories locked more than VFS requires; lock just before getting a child to be made positive (using simple_start_creating()) and unlock as soon as the child is created. There's no benefit in keeping the parent locked while populating the child - it won't stop dcache lookups anyway.

Reviewed-by Jeff Layton <[email protected]>

Signed-off-by Al Viro <[email protected]>

Stable-dep-of: 932a8cf6abb2 ("sunrpc: fix use-after-free in __rpc_clnt_handle_event and __rpc_clnt_remove_pipedir")

Signed-off-by Sasha Levin <[email protected]>

---
 net/sunrpc/rpc_pipe.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)

net/sunrpc/rpc_pipe.c +12 ~~-13~~

index 5556e25ffc35a..76a9d86ab8ac7 100644
@@ -698,12 +698,14 @@ static int rpc_populate(struct dentry *parent,
 	struct dentry *dentry;
 	int i, err;
 
-	inode_lock(dir);
 	for (i = start; i < eof; i++) {
+		inode_lock(dir);
 		dentry = __rpc_lookup_create_exclusive(parent, files[i].name);
 		err = PTR_ERR(dentry);
-		if (IS_ERR(dentry))
+		if (IS_ERR(dentry)) {
+			inode_unlock(dir);
 			goto out_bad;
+		}
 		switch (files[i].mode & S_IFMT) {
 			default:
 				BUG();

@@ -712,20 +714,20 @@ static int rpc_populate(struct dentry *parent,
 						files[i].mode,
 						files[i].i_fop,
 						private);
+				inode_unlock(dir);
 				break;
 			case S_IFDIR:
 				err = __rpc_mkdir(dir, dentry,
 						files[i].mode,
 						NULL,
 						private);
+				inode_unlock(dir);
 		}
 		if (err != 0)
 			goto out_bad;
 	}
-	inode_unlock(dir);
 	return 0;
 out_bad:
-	inode_unlock(dir);
 	printk(KERN_WARNING "%s: %s failed to populate directory %pd\n",
 			__FILE__, __func__, parent);
 	return err;

@@ -741,25 +743,22 @@ static struct dentry *rpc_mkdir_populate(struct dentry *parent,
 
 	inode_lock_nested(dir, I_MUTEX_PARENT);
 	dentry = __rpc_lookup_create_exclusive(parent, name);
-	if (IS_ERR(dentry))
-		goto out;
+	if (IS_ERR(dentry)) {
+		inode_unlock(dir);
+		return dentry;
+	}
 	error = __rpc_mkdir(dir, dentry, mode, NULL, private);
+	inode_unlock(dir);
 	if (error != 0)
-		goto out_err;
+		return ERR_PTR(error);
 	if (populate != NULL) {
 		error = populate(dentry, args_populate);
 		if (error) {
-			inode_unlock(dir);
 			simple_recursive_removal(dentry, NULL);
 			return ERR_PTR(error);
 		}
 	}
-out:
-	inode_unlock(dir);
 	return dentry;
-out_err:
-	dentry = ERR_PTR(error);
-	goto out;
 }
 
 static int rpc_rmdir_depopulate(struct dentry *dentry,

Signature · 1 line

2.53.0

Sasha Levin

maintainer

2026-09-08 22:37 UTC #4 lore raw

PATCH 4/4 sunrpc: fix use-after-free in __rpc_clnt_handle_event and __rpc_clnt_remove_pipedir

In reply to Sasha Levin

From: Luxiao Xu <[email protected]>

[ Upstream commit 932a8cf6abb2b2f8677b79153a823108d8861fe2 ]

Fix this by properly updating clnt->pipefs_sb upon mount events and clearing it during unmount or failure paths.

Fixes bfca5fb4e97c ("SUNRPC: Fix RPC client cleaned up the freed pipefs dentries")

Cc [email protected]

Reported-by Yuan Tan <[email protected]>

Reported-by Xin Liu <[email protected]>

Reviewed-by Ren Wei <[email protected]>

Assisted-by Codex:gpt-5.4

Signed-off-by Luxiao Xu <[email protected]>

Signed-off-by Trond Myklebust <[email protected]>

Signed-off-by Sasha Levin <[email protected]>

---
 net/sunrpc/clnt.c | 22 +++++++++++++++++-----
 1 file changed, 17 insertions(+), 5 deletions(-)

net/sunrpc/clnt.c +17 -5

index 1eca77bd03930..f23a17ebd0391 100644
@@ -96,7 +96,10 @@ static void rpc_unregister_client(struct rpc_clnt *clnt)
 
 static void __rpc_clnt_remove_pipedir(struct rpc_clnt *clnt)
 {
-	rpc_remove_client_dir(clnt);
+	if (clnt->pipefs_sb) {
+		rpc_remove_client_dir(clnt);
+		clnt->pipefs_sb = NULL;
+	}
 }
 
 static void rpc_clnt_remove_pipedir(struct rpc_clnt *clnt)

@@ -177,19 +180,28 @@ static int rpc_clnt_skip_event(struct rpc_clnt *clnt, unsigned long event)
 }
 
 static int __rpc_clnt_handle_event(struct rpc_clnt *clnt, unsigned long event,
-				   struct super_block *sb)
+				    struct super_block *sb)
 {
+	int err = 0;
+
 	switch (event) {
 	case RPC_PIPEFS_MOUNT:
-		return rpc_setup_pipedir_sb(sb, clnt);
+		clnt->pipefs_sb = sb;
+		err = rpc_setup_pipedir_sb(sb, clnt);
+		if (err)
+			clnt->pipefs_sb = NULL;
+		break;
 	case RPC_PIPEFS_UMOUNT:
-		__rpc_clnt_remove_pipedir(clnt);
+		if (clnt->pipefs_sb == sb) {
+			__rpc_clnt_remove_pipedir(clnt);
+			clnt->pipefs_sb = NULL;
+		}
 		break;
 	default:
 		printk(KERN_ERR "%s: unknown event: %ld\n", __func__, event);
 		return -ENOTSUPP;
 	}
-	return 0;
+	return err;
 }
 
 static int __rpc_pipefs_event(struct rpc_clnt *clnt, unsigned long event,

Signature · 1 line

2.53.0