linux-kernel - [PATCH V2] smb: client: fix netns refcount leak after net

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20250717132926.901902-1-wangzhaolong@huaweicloud.com>
Date: Thu, 17 Jul 2025 21:29:26 +0800
From: Wang Zhaolong <wangzhaolong@...weicloud.com>
To: sfrench@...ba.org,
	kuniyu@...gle.com
Cc: linux-cifs@...r.kernel.org,
	samba-technical@...ts.samba.org,
	linux-kernel@...r.kernel.org,
	chengzhihao1@...wei.com,
	wangzhaolong@...weicloud.com,
	yi.zhang@...wei.com,
	yangerkun@...wei.com
Subject: [PATCH V2] smb: client: fix netns refcount leak after net_passive changes

After commit 5c70eb5c593d ("net: better track kernel sockets lifetime"),
kernel sockets now use net_passive reference counting. However, commit
95d2b9f693ff ("Revert "smb: client: fix TCP timers deadlock after rmmod"")
restored the manual socket refcount manipulation without adapting to this
new mechanism, causing a memory leak.

The issue can be reproduced by[1]:
1. Creating a network namespace
2. Mounting and Unmounting CIFS within the namespace
3. Deleting the namespace

Some memory leaks may appear after a period of time following step 3.

unreferenced object 0xffff9951419f6b00 (size 256):
  comm "ip", pid 447, jiffies 4294692389 (age 14.730s)
  hex dump (first 32 bytes):
    1b 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 80 77 c2 44 51 99 ff ff  .........w.DQ...
  backtrace:
    __kmem_cache_alloc_node+0x30e/0x3d0
    __kmalloc+0x52/0x120
    net_alloc_generic+0x1d/0x30
    copy_net_ns+0x86/0x200
    create_new_namespaces+0x117/0x300
    unshare_nsproxy_namespaces+0x60/0xa0
    ksys_unshare+0x148/0x360
    __x64_sys_unshare+0x12/0x20
    do_syscall_64+0x59/0x110
    entry_SYSCALL_64_after_hwframe+0x78/0xe2
...
unreferenced object 0xffff9951442e7500 (size 32):
  comm "mount.cifs", pid 475, jiffies 4294693782 (age 13.343s)
  hex dump (first 32 bytes):
    40 c5 38 46 51 99 ff ff 18 01 96 42 51 99 ff ff  @.8FQ......BQ...
    01 00 00 00 6f 00 c5 07 6f 00 d8 07 00 00 00 00  ....o...o.......
  backtrace:
    __kmem_cache_alloc_node+0x30e/0x3d0
    kmalloc_trace+0x2a/0x90
    ref_tracker_alloc+0x8e/0x1d0
    sk_alloc+0x18c/0x1c0
    inet_create+0xf1/0x370
    __sock_create+0xd7/0x1e0
    generic_ip_connect+0x1d4/0x5a0 [cifs]
    cifs_get_tcp_session+0x5d0/0x8a0 [cifs]
    cifs_mount_get_session+0x47/0x1b0 [cifs]
    dfs_mount_share+0xfa/0xa10 [cifs]
    cifs_mount+0x68/0x2b0 [cifs]
    cifs_smb3_do_mount+0x10b/0x760 [cifs]
    smb3_get_tree+0x112/0x2e0 [cifs]
    vfs_get_tree+0x29/0xf0
    path_mount+0x2d4/0xa00
    __se_sys_mount+0x165/0x1d0

Root cause:
When creating kernel sockets, sk_alloc() calls net_passive_inc() for
sockets with sk_net_refcnt=0. The CIFS code manually converts kernel
sockets to user sockets by setting sk_net_refcnt=1, but doesn't call
the corresponding net_passive_dec(). This creates an imbalance in the
net_passive counter, which prevents the network namespace from being
destroyed when its last user reference is dropped. As a result, the
entire namespace and all its associated resources remain allocated.

Timeline of patches leading to this issue:
- commit ef7134c7fc48 ("smb: client: Fix use-after-free of network
  namespace.") in v6.12 fixed the original netns UAF by manually
  managing socket refcounts
- commit e9f2517a3e18 ("smb: client: fix TCP timers deadlock after
  rmmod") in v6.13 attempted to use kernel sockets but introduced
  TCP timer issues
- commit 5c70eb5c593d ("net: better track kernel sockets lifetime")
  in v6.14-rc5 introduced the net_passive mechanism with
  sk_net_refcnt_upgrade() for proper socket conversion
- commit 95d2b9f693ff ("Revert "smb: client: fix TCP timers deadlock
  after rmmod"") in v6.15-rc3 reverted to manual refcount management
  without adapting to the new net_passive changes

Fix this by using sk_net_refcnt_upgrade() which properly handles the
net_passive counter when converting kernel sockets to user sockets.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=220343 [1]
Fixes: 95d2b9f693ff ("Revert "smb: client: fix TCP timers deadlock after rmmod"")
Cc: stable@...r.kernel.org
Signed-off-by: Wang Zhaolong <wangzhaolong@...weicloud.com>
---
 fs/smb/client/connect.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

V1 -> V2:
- Add a simplified description of the reproduction steps in the
  commit message.

diff --git a/fs/smb/client/connect.c b/fs/smb/client/connect.c
index 205f547ca49e..5eec8957f2a9 100644
--- a/fs/smb/client/connect.c
+++ b/fs/smb/client/connect.c
@@ -3360,22 +3360,19 @@ generic_ip_connect(struct TCP_Server_Info *server)
 		socket = server->ssocket;
 	} else {
 		struct net *net = cifs_net_ns(server);
 		struct sock *sk;
 
-		rc = __sock_create(net, sfamily, SOCK_STREAM,
-				   IPPROTO_TCP, &server->ssocket, 1);
+		rc = sock_create_kern(net, sfamily, SOCK_STREAM,
+				      IPPROTO_TCP, &server->ssocket);
 		if (rc < 0) {
 			cifs_server_dbg(VFS, "Error %d creating socket\n", rc);
 			return rc;
 		}
 
 		sk = server->ssocket->sk;
-		__netns_tracker_free(net, &sk->ns_tracker, false);
-		sk->sk_net_refcnt = 1;
-		get_net_track(net, &sk->ns_tracker, GFP_KERNEL);
-		sock_inuse_add(net, 1);
+		sk_net_refcnt_upgrade(sk);
 
 		/* BB other socket options to set KEEPALIVE, NODELAY? */
 		cifs_dbg(FYI, "Socket created\n");
 		socket = server->ssocket;
 		socket->sk->sk_allocation = GFP_NOFS;
-- 
2.39.2