Skip to contentSkip to conversation

general protection fault in kernel_sock_shutdown (4)

syzbot Applied syzbot 11 participants 43 messages Apr 23 – May 30

Subsystems (5)
lore
Share & tools
lore

syzbot

author syzbot
#0 lore raw

Hello,

syzbot found the following issue on:

HEAD commit:    6596a02b2078 Merge tag 'drm-next-2026-04-22' of https://gi..
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=10d5acce580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=3e19fa1907a3dfda
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       gcc (Debian 14.2.0-19) 14.2.0, GNU ld (GNU Binutils for Debian) 2.44

Unfortunately, I don't have any reproducer for this issue yet.

Downloadable assets:
disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/d900f083ada3/non_bootable_disk-6596a02b.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/1e78cc4623b5/vmlinux-6596a02b.xz
kernel image: https://storage.googleapis.com/syzbot-assets/7ae78b19bafa/bzImage-6596a02b.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:

Reported-by [email protected]
Oops: general protection fault, probably for non-canonical address 0xdffffc000000000d: 0000 [#1] SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f]
CPU: 3 UID: 0 PID: 12652 Comm: syz.7.1709 Tainted: G             L      syzkaller #0 PREEMPT(full) 
Tainted: [L]=SOFTLOCKUP
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
RIP: 0010:kernel_sock_shutdown+0x47/0x70 net/socket.c:3785
Code: fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 75 33 48 b8 00 00 00 00 00 fc ff df 4c 8b 63 20 49 8d 7c 24 68 48 89 fa 48 c1 ea 03 <80> 3c 02 00 75 1a 49 8b 44 24 68 89 ee 48 89 df 5b 5d 41 5c e9 46
RSP: 0018:ffffc9000566f180 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: ffff888058587240 RCX: 0000000000000000
RDX: 000000000000000d RSI: ffffffff895ced12 RDI: 0000000000000068
RBP: 0000000000000002 R08: 0000000000000001 R09: ffffed1006d98945
R10: ffff888036cc4a2b R11: 0000003683c25c00 R12: 0000000000000000
R13: ffff88805c998000 R14: 0000000000000002 R15: 0000000000000018
FS:  00007f1306d976c0(0000) GS:ffff8880d65db000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f1306d97d58 CR3: 00000000404f1000 CR4: 0000000000352ef0
DR0: ffffffffffffffff DR1: 00000000000001f8 DR2: 0000000000000002
DR3: ffffffffefffff15 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Call Trace:
 <TASK>
 udp_tunnel_sock_release+0x68/0x80 net/ipv4/udp_tunnel_core.c:202
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put+0xae/0x130 drivers/infiniband/sw/rxe/rxe_net.c:639
 rxe_net_del+0x83/0x120 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
 nldev_dellink+0x289/0x3c0 drivers/infiniband/core/nldev.c:1849
 rdma_nl_rcv_msg+0x392/0x6f0 drivers/infiniband/core/netlink.c:195
 rdma_nl_rcv_skb.constprop.0.isra.0+0x2cb/0x410 drivers/infiniband/core/netlink.c:239
 netlink_unicast_kernel net/netlink/af_netlink.c:1318 [inline]
 netlink_unicast+0x585/0x850 net/netlink/af_netlink.c:1344
 netlink_sendmsg+0x8b0/0xda0 net/netlink/af_netlink.c:1894
 sock_sendmsg_nosec net/socket.c:787 [inline]
 __sock_sendmsg net/socket.c:802 [inline]
 ____sys_sendmsg+0x9e1/0xb70 net/socket.c:2698
 ___sys_sendmsg+0x190/0x1e0 net/socket.c:2752
 __sys_sendmsg+0x170/0x220 net/socket.c:2784
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0x10b/0xf80 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f1305f9c819
Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 e8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f1306d97028 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f1306216090 RCX: 00007f1305f9c819
RDX: 0000000000000000 RSI: 00002000000002c0 RDI: 0000000000000003
RBP: 00007f1306032c91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007f1306216128 R14: 00007f1306216090 R15: 00007ffd8ecad288
 </TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:kernel_sock_shutdown+0x47/0x70 net/socket.c:3785
Code: fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 75 33 48 b8 00 00 00 00 00 fc ff df 4c 8b 63 20 49 8d 7c 24 68 48 89 fa 48 c1 ea 03 <80> 3c 02 00 75 1a 49 8b 44 24 68 89 ee 48 89 df 5b 5d 41 5c e9 46
RSP: 0018:ffffc9000566f180 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: ffff888058587240 RCX: 0000000000000000
RDX: 000000000000000d RSI: ffffffff895ced12 RDI: 0000000000000068
RBP: 0000000000000002 R08: 0000000000000001 R09: ffffed1006d98945
R10: ffff888036cc4a2b R11: 0000003683c25c00 R12: 0000000000000000
R13: ffff88805c998000 R14: 0000000000000002 R15: 0000000000000018
FS:  00007f1306d976c0(0000) GS:ffff8880d64db000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000200000317000 CR3: 00000000404f1000 CR4: 0000000000352ef0
DR0: ffffffffffffffff DR1: 00000000000001f8 DR2: 0000000000000002
DR3: ffffffffefffff15 DR6: 00000000ffff0ff0 DR7: 0000000000000400
----------------
Code disassembly (best guess):
   0:	fc                   	cld
   1:	ff                   	lcall  (bad)
   2:	df 48 89             	fisttps -0x77(%rax)
   5:	fa                   	cli
   6:	48 c1 ea 03          	shr    $0x3,%rdx
   a:	80 3c 02 00          	cmpb   $0x0,(%rdx,%rax,1)
   e:	75 33                	jne    0x43
  10:	48 b8 00 00 00 00 00 	movabs $0xdffffc0000000000,%rax
  17:	fc ff df
  1a:	4c 8b 63 20          	mov    0x20(%rbx),%r12
  1e:	49 8d 7c 24 68       	lea    0x68(%r12),%rdi
  23:	48 89 fa             	mov    %rdi,%rdx
  26:	48 c1 ea 03          	shr    $0x3,%rdx
* 2a:	80 3c 02 00          	cmpb   $0x0,(%rdx,%rax,1) <-- trapping instruction
  2e:	75 1a                	jne    0x4a
  30:	49 8b 44 24 68       	mov    0x68(%r12),%rax
  35:	89 ee                	mov    %ebp,%esi
  37:	48 89 df             	mov    %rbx,%rdi
  3a:	5b                   	pop    %rbx
  3b:	5d                   	pop    %rbp
  3c:	41 5c                	pop    %r12
  3e:	e9                   	.byte 0xe9
  3f:	46                   	rex.RX

---
This report is generated by a bot. It may contain errors. See https://goo.gl/tpsmEJ for more information about syzbot. syzbot engineers can be reached at [email protected].

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

If the report is already addressed, let syzbot know by replying with:
#syz fix: exact-commit-title

If you want to overwrite report's subsystems, reply with:
#syz set subsystems: new-subsystem
(See the list of subsystem names on the web dashboard)

If the report is a duplicate of another one, reply with:
#syz dup: exact-subject-of-another-report

If you want to undo deduplication, reply with:
#syz undup

maintainer
#1 lore raw
Quoted 1 line · On Thu, 23 Apr 2026 08:01:35 -0700 syzbot wrote:

Subject: [syzbot] [net?] general protection fault in kernel_sock_shutdown (4)

#syz set subsystems: rdma

#2 lore raw
In reply to syzbot

This report was analysed with the help of an automated kernel crash analysis assistant. The analysis below is tentative and should be reviewed by a human before any action is taken.

Decoded Backtrace
-----------------

1. kernel_sock_shutdown -- crash site (net/socket.c:3785)

  3783  int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
  3784  {
  3785      return READ_ONCE(sock->ops)->shutdown(sock, how);
              /* CRASH: sock->ops is NULL (R12 = 0x0); KASAN traps
                 null-ptr-deref at offset 0x68 = offsetof(proto_ops, shutdown) */
  3786  }

Register context at crash:

    RBX = 0xffff888058587240  (struct socket *sock)
    R12 = 0x0000000000000000  (sock->ops, loaded from RBX+0x20 -- NULL)
    RDI = 0x0000000000000068  (= NULL + 0x68, address of shutdown fn ptr)
    RBP = 0x0000000000000002  (how = SHUT_RDWR)

2. udp_tunnel_sock_release (net/ipv4/udp_tunnel_core.c:196-202)

  196  void udp_tunnel_sock_release(struct socket *sock)
  197  {
  198      rcu_assign_sk_user_data(sock->sk, NULL);
  199      synchronize_rcu();
  200      kernel_sock_shutdown(sock, SHUT_RDWR);  /* <- calls crash site */
  201      sock_release(sock);
  202  }

3. rxe_release_udp_tunnel inlined (drivers/infiniband/sw/rxe/rxe_net.c:290-293)

  290  static void rxe_release_udp_tunnel(struct socket *sk)
  291  {
  292      if (sk)
  293          udp_tunnel_sock_release(sk);
  294  }

4. rxe_sock_put (drivers/infiniband/sw/rxe/rxe_net.c:632-643)

  632  static void rxe_sock_put(struct sock *sk,
  633                            void (*set_sk)(struct net *, struct sock *),
  634                            struct net *net)
  635  {
  636      if (refcount_read(&sk->sk_refcnt) > SK_REF_FOR_TUNNEL) {
  637          __sock_put(sk);
  638      } else {
  639          rxe_release_udp_tunnel(sk->sk_socket);  /* <- release BEFORE clear */
  640          sk = NULL;
  641          set_sk(net, sk);                         /* <- clear AFTER (too late) */
  642      }
  643  }

Caller: rxe_net_del (rxe_net.c:644-666), triggered via:

    nldev_dellink -> rxe_dellink -> rxe_net_del -> rxe_sock_put

Tentative Analysis
------------------

sock->ops is set to NULL by sock_release() (net/socket.c:726) after calling ops->release(sock). The crash in kernel_sock_shutdown() means the socket was already passed to sock_release() before this call.

Two independent code paths can release the same UDP tunnel socket stored in the per-network-namespace rxe_ns_sock structure:

Path 1 -- namespace teardown (rxe_ns.c, rxe_ns_exit()):
rcu_assign_pointer(ns_sk->rxe_sk4, NULL); /* clears pointer FIRST */
udp_tunnel_sock_release(sk->sk_socket); /* then releases */

Path 2 -- RDMA link delete (rxe_net.c, rxe_net_del() -> rxe_sock_put()):

   sk = rxe_ns_pernet_sk4(net);                 /* reads pointer (no ownership) */
   rxe_release_udp_tunnel(sk->sk_socket);       /* releases FIRST */
   set_sk(net, NULL);                           /* clears AFTER */

The following TOCTOU (time-of-check time-of-use) race is possible when namespace teardown and RDMA link deletion occur concurrently:

Thread A (rxe_net_del):

    rxe_ns_pernet_sk4() -> sk = X  (non-NULL)

Thread B (rxe_ns_exit):

    rcu_assign_pointer(sk4, NULL)
    udp_tunnel_sock_release(X->sk_socket)
      sock_release(X->sk_socket)
        X->sk_socket->ops = NULL       <- clears ops

Thread A (rxe_net_del) continues:

    rxe_sock_put(sk=X, ...)
      rxe_release_udp_tunnel(X->sk_socket)
        kernel_sock_shutdown(X->sk_socket, SHUT_RDWR)
          READ_ONCE(sock->ops)->shutdown(...)
                                       <- CRASH: sock->ops == NULL

The bug was introduced by two commits in March 2026 that added per-network-namespace support to the Soft RoCE (RXE) driver:

13f2a53c2a71e RDMA/rxe: Add net namespace support for IPv4/IPv6 sockets f1327abd6abed RDMA/rxe: Support RDMA link creation and destruction per
net namespace

Neither commit provides synchronisation between the two teardown paths.

Potential Solution
------------------

Replace rxe_ns_pernet_sk4() calls in rxe_net_del() (and rxe_notify()) with an atomic exchange that simultaneously reads and clears the pernet pointer, so only one of the two teardown paths can ever obtain a non-NULL socket pointer:

  struct sock *rxe_ns_pernet_take_sk4(struct net *net)
  {
      struct rxe_ns_sock *ns_sk = net_generic(net, rxe_pernet_id);
      return unrcu_pointer(xchg(&ns_sk->rxe_sk4, RCU_INITIALIZER(NULL)));
  }

Whichever path (rxe_ns_exit or rxe_net_del) wins the xchg gets the socket and releases it; the loser gets NULL and skips the release.

More information
----------------

Oops-Analysis: https://lore.kernel.org/r/[email protected]

Assisted-by linux-kernel-oops-x86 skill (Claude Sonnet 4.6)
#3 lore raw
In reply to syzbot

Unfortunately the AI had a burp and did not write out the proper URL for analysis data; it should have been

http://oops.fenrus.org/reports/lkml/69ea344f.a00a0220.17a17.0040.GAE_google.com/report.html

and in addition, it made a candidate patch (below)

From: Arjan van de Ven <[email protected]>
Subject: [PATCH] RDMA/rxe: fix double-release race on UDP tunnel socket teardown

This patch is based on a BUG as reported at
https://lore.kernel.org/r/[email protected].

The Soft RoCE (RXE) driver stores per-network-namespace UDP tunnel sockets for IPv4 and IPv6 encapsulation. Two independent code paths tear these sockets down: rxe_ns_exit(), called when a network namespace is destroyed, and rxe_net_del(), called when an RDMA link is deleted via netlink. Both paths read the per-namespace socket pointer and call udp_tunnel_sock_release() on it.

A time-of-check/time-of-use (TOCTOU) race exists in rxe_net_del(). It reads the socket pointer via rxe_ns_pernet_sk4(), then passes it to rxe_sock_put() for release. If rxe_ns_exit() runs concurrently between the read and the release, it clears the pointer and calls udp_tunnel_sock_release() first, causing sock_release() to set sock->ops = NULL. When rxe_net_del() then calls udp_tunnel_sock_release() on the same socket, kernel_sock_shutdown() dereferences the now-NULL sock->ops, triggering a KASAN null-ptr-deref at offset 0x68 (the shutdown function pointer in struct proto_ops).

A minimal alternative would guard against NULL sock->ops inside udp_tunnel_sock_release() before calling kernel_sock_shutdown(). That treats the symptom rather than the root cause and leaves the double-release of socket state intact.

Add rxe_ns_pernet_take_sk4() and rxe_ns_pernet_take_sk6() which use xchg() to atomically swap the per-namespace socket pointer to NULL and return the old value. Replace the non-atomic reads in rxe_net_del() with these take variants, and release the socket directly via udp_tunnel_sock_release() without going through rxe_sock_put().

Whichever teardown path executes take first claims ownership of the socket; the second caller gets NULL and skips the release, closing the double-release window.

Oops-Analysis: http://oops.fenrus.org/reports/lkml/69ea344f.a00a0220.17a17.0040.GAE_google.com/report.html

Fixes 13f2a53c2a71 ("RDMA/rxe: Add net namespace support for IPv4/IPv6 sockets")
Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Assisted-by GitHub Copilot patcher:claude linux-kernel-oops-x86.
Cc Zhu Yanjun <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Leon Romanovsky <[email protected]>
---
 drivers/infiniband/sw/rxe/rxe_net.c |    8 ++++----
 drivers/infiniband/sw/rxe/rxe_ns.c  |   14 ++++++++++++++
 drivers/infiniband/sw/rxe/rxe_ns.h  |    7 +++++++
 3 files changed, 25 insertions(+), 4 deletions(-)
drivers/infiniband/sw/rxe/rxe_net.c +4 -4
index 50a2cb5405e22..4f604636cb7b4 100644
@@ -655,13 +655,13 @@ void rxe_net_del(struct ib_device *dev)
 
 	net = dev_net(ndev);
 
-	sk = rxe_ns_pernet_sk4(net);
+	sk = rxe_ns_pernet_take_sk4(net);
 	if (sk)
-		rxe_sock_put(sk, rxe_ns_pernet_set_sk4, net);
+		udp_tunnel_sock_release(sk->sk_socket);
 
-	sk = rxe_ns_pernet_sk6(net);
+	sk = rxe_ns_pernet_take_sk6(net);
 	if (sk)
-		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
+		udp_tunnel_sock_release(sk->sk_socket);
 
 	dev_put(ndev);
 }
drivers/infiniband/sw/rxe/rxe_ns.c +14 -0
index 8b9d734229b24..d9d376e3c670f 100644
@@ -91,6 +91,13 @@ void rxe_ns_pernet_set_sk4(struct net *net, struct sock *sk)
 	synchronize_rcu();
 }
 
+struct sock *rxe_ns_pernet_take_sk4(struct net *net)
+{
+	struct rxe_ns_sock *ns_sk = net_generic(net, rxe_pernet_id);
+
+	return xchg((__force struct sock **)&ns_sk->rxe_sk4, NULL);
+}
+
 #if IS_ENABLED(CONFIG_IPV6)
 struct sock *rxe_ns_pernet_sk6(struct net *net)
 {
@@ -111,6 +118,13 @@ void rxe_ns_pernet_set_sk6(struct net *net, struct sock *sk)
 	rcu_assign_pointer(ns_sk->rxe_sk6, sk);
 	synchronize_rcu();
 }
+
+struct sock *rxe_ns_pernet_take_sk6(struct net *net)
+{
+	struct rxe_ns_sock *ns_sk = net_generic(net, rxe_pernet_id);
+
+	return xchg((__force struct sock **)&ns_sk->rxe_sk6, NULL);
+}
 #endif /* IPV6 */
 
 int rxe_namespace_init(void)
drivers/infiniband/sw/rxe/rxe_ns.h +6 -0
index 4da2709e6b714..9d9a5106b77c8 100644
@@ -5,10 +5,17 @@
 
 struct sock *rxe_ns_pernet_sk4(struct net *net);
 void rxe_ns_pernet_set_sk4(struct net *net, struct sock *sk);
+struct sock *rxe_ns_pernet_take_sk4(struct net *net);
 
 #if IS_ENABLED(CONFIG_IPV6)
 void rxe_ns_pernet_set_sk6(struct net *net, struct sock *sk);
 struct sock *rxe_ns_pernet_sk6(struct net *net);
+struct sock *rxe_ns_pernet_take_sk6(struct net *net);
 #else /* IPv6 */
 static inline struct sock *rxe_ns_pernet_sk6(struct net *net)
 {
@@ -18,6 +25,10 @@ static inline void rxe_ns_pernet_set_sk6(struct net *net, struct sock *sk)
 {
 }
 
+static inline struct sock *rxe_ns_pernet_take_sk6(struct net *net)
+{
+	return NULL;
+}
 #endif /* IPv6 */
 
 int rxe_namespace_init(void);
#4 lore raw
In reply to syzbot

Unfortunately the AI had a burp and did not write out the proper URL for analysis data; it should have been

http://oops.fenrus.org/reports/lkml/69ea344f.a00a0220.17a17.0040.GAE_google.com/report.html

and in addition, it made a candidate patch (below)

From: Arjan van de Ven <[email protected]>
Subject: [PATCH] RDMA/rxe: fix double-release race on UDP tunnel socket teardown

This patch is based on a BUG as reported at
https://lore.kernel.org/r/[email protected].

The Soft RoCE (RXE) driver stores per-network-namespace UDP tunnel sockets for IPv4 and IPv6 encapsulation. Two independent code paths tear these sockets down: rxe_ns_exit(), called when a network namespace is destroyed, and rxe_net_del(), called when an RDMA link is deleted via netlink. Both paths read the per-namespace socket pointer and call udp_tunnel_sock_release() on it.

A time-of-check/time-of-use (TOCTOU) race exists in rxe_net_del(). It reads the socket pointer via rxe_ns_pernet_sk4(), then passes it to rxe_sock_put() for release. If rxe_ns_exit() runs concurrently between the read and the release, it clears the pointer and calls udp_tunnel_sock_release() first, causing sock_release() to set sock->ops = NULL. When rxe_net_del() then calls udp_tunnel_sock_release() on the same socket, kernel_sock_shutdown() dereferences the now-NULL sock->ops, triggering a KASAN null-ptr-deref at offset 0x68 (the shutdown function pointer in struct proto_ops).

A minimal alternative would guard against NULL sock->ops inside udp_tunnel_sock_release() before calling kernel_sock_shutdown(). That treats the symptom rather than the root cause and leaves the double-release of socket state intact.

Add rxe_ns_pernet_take_sk4() and rxe_ns_pernet_take_sk6() which use xchg() to atomically swap the per-namespace socket pointer to NULL and return the old value. Replace the non-atomic reads in rxe_net_del() with these take variants, and release the socket directly via udp_tunnel_sock_release() without going through rxe_sock_put().

Whichever teardown path executes take first claims ownership of the socket; the second caller gets NULL and skips the release, closing the double-release window.

Oops-Analysis: http://oops.fenrus.org/reports/lkml/69ea344f.a00a0220.17a17.0040.GAE_google.com/report.html

Fixes 13f2a53c2a71 ("RDMA/rxe: Add net namespace support for IPv4/IPv6 sockets")
Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Assisted-by GitHub Copilot patcher:claude linux-kernel-oops-x86.
Cc Zhu Yanjun <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Leon Romanovsky <[email protected]>
---
 drivers/infiniband/sw/rxe/rxe_net.c |    8 ++++----
 drivers/infiniband/sw/rxe/rxe_ns.c  |   14 ++++++++++++++
 drivers/infiniband/sw/rxe/rxe_ns.h  |    7 +++++++
 3 files changed, 25 insertions(+), 4 deletions(-)
drivers/infiniband/sw/rxe/rxe_net.c +4 -4
index 50a2cb5405e22..4f604636cb7b4 100644
@@ -655,13 +655,13 @@ void rxe_net_del(struct ib_device *dev)
 
 	net = dev_net(ndev);
 
-	sk = rxe_ns_pernet_sk4(net);
+	sk = rxe_ns_pernet_take_sk4(net);
 	if (sk)
-		rxe_sock_put(sk, rxe_ns_pernet_set_sk4, net);
+		udp_tunnel_sock_release(sk->sk_socket);
 
-	sk = rxe_ns_pernet_sk6(net);
+	sk = rxe_ns_pernet_take_sk6(net);
 	if (sk)
-		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
+		udp_tunnel_sock_release(sk->sk_socket);
 
 	dev_put(ndev);
 }
drivers/infiniband/sw/rxe/rxe_ns.c +14 -0
index 8b9d734229b24..d9d376e3c670f 100644
@@ -91,6 +91,13 @@ void rxe_ns_pernet_set_sk4(struct net *net, struct sock *sk)
 	synchronize_rcu();
 }
 
+struct sock *rxe_ns_pernet_take_sk4(struct net *net)
+{
+	struct rxe_ns_sock *ns_sk = net_generic(net, rxe_pernet_id);
+
+	return xchg((__force struct sock **)&ns_sk->rxe_sk4, NULL);
+}
+
 #if IS_ENABLED(CONFIG_IPV6)
 struct sock *rxe_ns_pernet_sk6(struct net *net)
 {
@@ -111,6 +118,13 @@ void rxe_ns_pernet_set_sk6(struct net *net, struct sock *sk)
 	rcu_assign_pointer(ns_sk->rxe_sk6, sk);
 	synchronize_rcu();
 }
+
+struct sock *rxe_ns_pernet_take_sk6(struct net *net)
+{
+	struct rxe_ns_sock *ns_sk = net_generic(net, rxe_pernet_id);
+
+	return xchg((__force struct sock **)&ns_sk->rxe_sk6, NULL);
+}
 #endif /* IPV6 */
 
 int rxe_namespace_init(void)
drivers/infiniband/sw/rxe/rxe_ns.h +6 -0
index 4da2709e6b714..9d9a5106b77c8 100644
@@ -5,10 +5,17 @@
 
 struct sock *rxe_ns_pernet_sk4(struct net *net);
 void rxe_ns_pernet_set_sk4(struct net *net, struct sock *sk);
+struct sock *rxe_ns_pernet_take_sk4(struct net *net);
 
 #if IS_ENABLED(CONFIG_IPV6)
 void rxe_ns_pernet_set_sk6(struct net *net, struct sock *sk);
 struct sock *rxe_ns_pernet_sk6(struct net *net);
+struct sock *rxe_ns_pernet_take_sk6(struct net *net);
 #else /* IPv6 */
 static inline struct sock *rxe_ns_pernet_sk6(struct net *net)
 {
@@ -18,6 +25,10 @@ static inline void rxe_ns_pernet_set_sk6(struct net *net, struct sock *sk)
 {
 }
 
+static inline struct sock *rxe_ns_pernet_take_sk6(struct net *net)
+{
+	return NULL;
+}
 #endif /* IPv6 */
 
 int rxe_namespace_init(void);
#5 lore raw
Quoted 81 lines · Arjan van de Ven

Unfortunately the AI had a burp and did not write out the proper URL for analysis data; it should have been

http://oops.fenrus.org/reports/lkml/69ea344f.a00a0220.17a17.0040.GAE_google.com/report.html

and in addition, it made a candidate patch (below)

From: Arjan van de Ven <[email protected]>
Subject: [PATCH] RDMA/rxe: fix double-release race on UDP tunnel socket teardown

This patch is based on a BUG as reported at
https://lore.kernel.org/r/[email protected].

The Soft RoCE (RXE) driver stores per-network-namespace UDP tunnel sockets for IPv4 and IPv6 encapsulation. Two independent code paths tear these sockets down: rxe_ns_exit(), called when a network namespace is destroyed, and rxe_net_del(), called when an RDMA link is deleted via netlink. Both paths read the per-namespace socket pointer and call udp_tunnel_sock_release() on it.

A time-of-check/time-of-use (TOCTOU) race exists in rxe_net_del(). It reads the socket pointer via rxe_ns_pernet_sk4(), then passes it to rxe_sock_put() for release. If rxe_ns_exit() runs concurrently between the read and the release, it clears the pointer and calls udp_tunnel_sock_release() first, causing sock_release() to set sock->ops = NULL. When rxe_net_del() then calls udp_tunnel_sock_release() on the same socket, kernel_sock_shutdown() dereferences the now-NULL sock->ops, triggering a KASAN null-ptr-deref at offset 0x68 (the shutdown function pointer in struct proto_ops).

A minimal alternative would guard against NULL sock->ops inside udp_tunnel_sock_release() before calling kernel_sock_shutdown(). That treats the symptom rather than the root cause and leaves the double-release of socket state intact.

Add rxe_ns_pernet_take_sk4() and rxe_ns_pernet_take_sk6() which use xchg() to atomically swap the per-namespace socket pointer to NULL and return the old value. Replace the non-atomic reads in rxe_net_del() with these take variants, and release the socket directly via udp_tunnel_sock_release() without going through rxe_sock_put().

Whichever teardown path executes take first claims ownership of the socket; the second caller gets NULL and skips the release, closing the double-release window.

Oops-Analysis: http://oops.fenrus.org/reports/lkml/69ea344f.a00a0220.17a17.0040.GAE_google.com/report.html

Fixes 13f2a53c2a71 ("RDMA/rxe: Add net namespace support for IPv4/IPv6 sockets")
Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Assisted-by GitHub Copilot patcher:claude linux-kernel-oops-x86.
Cc Zhu Yanjun <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Leon Romanovsky <[email protected]>
---
 drivers/infiniband/sw/rxe/rxe_net.c |    8 ++++----
 drivers/infiniband/sw/rxe/rxe_ns.c  |   14 ++++++++++++++
 drivers/infiniband/sw/rxe/rxe_ns.h  |    7 +++++++
 3 files changed, 25 insertions(+), 4 deletions(-)
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e22..4f604636cb7b4 100644

drivers/infiniband/sw/rxe/rxe_net.c void rxe_net_del(struct ib_device *dev) · line 660

Quoted 8 lines
drivers/infiniband/sw/rxe/rxe_net.c +2 -2
@@ -655,13 +655,13 @@ void rxe_net_del(struct ib_device *dev)
 
        net = dev_net(ndev);
 
-       sk = rxe_ns_pernet_sk4(net);
+       sk = rxe_ns_pernet_take_sk4(net);
        if (sk)
-               rxe_sock_put(sk, rxe_ns_pernet_set_sk4, net);
+               udp_tunnel_sock_release(sk->sk_socket);

This leaks sk->sk_refcnt, no AI slop please.

I'm working on the right fix.

Thanks.

author syzbot
#6 lore raw
In reply to syzbot
syzbot has found a reproducer for the following issue on:

syzbot has found a reproducer for the following issue on:

HEAD commit:    74fe02ce122a Merge tag 'wq-for-7.1-rc2-fixes' of git://git..
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=16e895ce580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=59da38148f3a3d24
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       gcc (Debian 14.2.0-19) 14.2.0, GNU ld (GNU Binutils for Debian) 2.44
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=13a613ba580000

Downloadable assets:
disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/d900f083ada3/non_bootable_disk-74fe02ce.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/c0a591d96864/vmlinux-74fe02ce.xz
kernel image: https://storage.googleapis.com/syzbot-assets/9f94fb623cd1/bzImage-74fe02ce.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:

Reported-by [email protected]
Oops: general protection fault, probably for non-canonical address 0xdffffc000000000d: 0000 [#1] SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f]
CPU: 3 UID: 0 PID: 5986 Comm: syz.3.20 Not tainted syzkaller #0 PREEMPT(full) 
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
RIP: 0010:kernel_sock_shutdown+0x47/0x70 net/socket.c:3785
Code: fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 75 33 48 b8 00 00 00 00 00 fc ff df 4c 8b 63 20 49 8d 7c 24 68 48 89 fa 48 c1 ea 03 <80> 3c 02 00 75 1a 49 8b 44 24 68 89 ee 48 89 df 5b 5d 41 5c ff e0
RSP: 0018:ffffc9000391f180 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: ffff88802a2a0040 RCX: ffffffff8b8b72bd
RDX: 000000000000000d RSI: ffffffff89553b32 RDI: 0000000000000068
RBP: 0000000000000002 R08: 0000000000000001 R09: fffff52000723dfc
R10: ffffc9000391efe7 R11: 0000000000000000 R12: 0000000000000000
R13: ffff8880311b8000 R14: 0000000000000002 R15: 0000000000000018
FS:  00007f602d1fe6c0(0000) GS:ffff8880d6675000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000561c522a6000 CR3: 000000002e99e000 CR4: 0000000000352ef0
Call Trace:
 <TASK>
 udp_tunnel_sock_release+0x68/0x80 net/ipv4/udp_tunnel_core.c:202
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put+0xae/0x130 drivers/infiniband/sw/rxe/rxe_net.c:639
 rxe_net_del+0x83/0x120 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
 nldev_dellink+0x289/0x3c0 drivers/infiniband/core/nldev.c:1849
 rdma_nl_rcv_msg+0x392/0x6f0 drivers/infiniband/core/netlink.c:195
 rdma_nl_rcv_skb.constprop.0.isra.0+0x2cb/0x410 drivers/infiniband/core/netlink.c:239
 netlink_unicast_kernel net/netlink/af_netlink.c:1318 [inline]
 netlink_unicast+0x585/0x850 net/netlink/af_netlink.c:1344
 netlink_sendmsg+0x8b0/0xda0 net/netlink/af_netlink.c:1894
 sock_sendmsg_nosec net/socket.c:787 [inline]
 __sock_sendmsg net/socket.c:802 [inline]
 ____sys_sendmsg+0x9e1/0xb70 net/socket.c:2698
 ___sys_sendmsg+0x190/0x1e0 net/socket.c:2752
 __sys_sendmsg+0x170/0x220 net/socket.c:2784
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0x10b/0xf80 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f602db9cdd9
Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 e8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f602d1fe028 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f602de16090 RCX: 00007f602db9cdd9
RDX: 0000000000000000 RSI: 00002000000002c0 RDI: 0000000000000007
RBP: 00007f602dc32d69 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007f602de16128 R14: 00007f602de16090 R15: 00007ffc1d89c428
 </TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:kernel_sock_shutdown+0x47/0x70 net/socket.c:3785
Code: fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 75 33 48 b8 00 00 00 00 00 fc ff df 4c 8b 63 20 49 8d 7c 24 68 48 89 fa 48 c1 ea 03 <80> 3c 02 00 75 1a 49 8b 44 24 68 89 ee 48 89 df 5b 5d 41 5c ff e0
RSP: 0018:ffffc9000391f180 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: ffff88802a2a0040 RCX: ffffffff8b8b72bd
RDX: 000000000000000d RSI: ffffffff89553b32 RDI: 0000000000000068
RBP: 0000000000000002 R08: 0000000000000001 R09: fffff52000723dfc
R10: ffffc9000391efe7 R11: 0000000000000000 R12: 0000000000000000
R13: ffff8880311b8000 R14: 0000000000000002 R15: 0000000000000018
FS:  00007f602d1fe6c0(0000) GS:ffff8880d6675000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000561c522a6000 CR3: 000000002e99e000 CR4: 0000000000352ef0
----------------
Code disassembly (best guess):
   0:	fc                   	cld
   1:	ff                   	lcall  (bad)
   2:	df 48 89             	fisttps -0x77(%rax)
   5:	fa                   	cli
   6:	48 c1 ea 03          	shr    $0x3,%rdx
   a:	80 3c 02 00          	cmpb   $0x0,(%rdx,%rax,1)
   e:	75 33                	jne    0x43
  10:	48 b8 00 00 00 00 00 	movabs $0xdffffc0000000000,%rax
  17:	fc ff df
  1a:	4c 8b 63 20          	mov    0x20(%rbx),%r12
  1e:	49 8d 7c 24 68       	lea    0x68(%r12),%rdi
  23:	48 89 fa             	mov    %rdi,%rdx
  26:	48 c1 ea 03          	shr    $0x3,%rdx
* 2a:	80 3c 02 00          	cmpb   $0x0,(%rdx,%rax,1) <-- trapping instruction
  2e:	75 1a                	jne    0x4a
  30:	49 8b 44 24 68       	mov    0x68(%r12),%rax
  35:	89 ee                	mov    %ebp,%esi
  37:	48 89 df             	mov    %rbx,%rdi
  3a:	5b                   	pop    %rbx
  3b:	5d                   	pop    %rbp
  3c:	41 5c                	pop    %r12
  3e:	ff e0                	jmp    *%rax

---
If you want syzbot to run the reproducer, reply with:
#syz test: git://repo/address.git branch-or-commit-hash
If you attach or paste a git patch, syzbot will apply it before testing.

#7 lore raw
Quoted 15 lines · syzbot

syzbot has found a reproducer for the following issue on:

HEAD commit:    74fe02ce122a Merge tag 'wq-for-7.1-rc2-fixes' of git://git..
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=16e895ce580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=59da38148f3a3d24
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       gcc (Debian 14.2.0-19) 14.2.0, GNU ld (GNU Binutils for Debian) 2.44
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=13a613ba580000

Downloadable assets:
disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/d900f083ada3/non_bootable_disk-74fe02ce.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/c0a591d96864/vmlinux-74fe02ce.xz
kernel image: https://storage.googleapis.com/syzbot-assets/9f94fb623cd1/bzImage-74fe02ce.xz

Quoted 5 lines

IMPORTANT: if you fix the issue, please add the following tag to the commit:

Reported-by [email protected]
Oops: general protection fault, probably for non-canonical address 0xdffffc000000000d: 0000 [#1] SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f]

Thanks a lot. IIRC, this problem is in process. The link is https://patchwork.kernel.org/project/linux-rdma/patch/[email protected]/

Hi, Kuniyuki Iwashima

I think you are fixing this problem. I hope that we can see your commit very soon.

Zhu Yanjun

Quoted 87 lines
CPU: 3 UID: 0 PID: 5986 Comm: syz.3.20 Not tainted syzkaller #0 PREEMPT(full)
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
RIP: 0010:kernel_sock_shutdown+0x47/0x70 net/socket.c:3785
Code: fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 75 33 48 b8 00 00 00 00 00 fc ff df 4c 8b 63 20 49 8d 7c 24 68 48 89 fa 48 c1 ea 03 <80> 3c 02 00 75 1a 49 8b 44 24 68 89 ee 48 89 df 5b 5d 41 5c ff e0
RSP: 0018:ffffc9000391f180 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: ffff88802a2a0040 RCX: ffffffff8b8b72bd
RDX: 000000000000000d RSI: ffffffff89553b32 RDI: 0000000000000068
RBP: 0000000000000002 R08: 0000000000000001 R09: fffff52000723dfc
R10: ffffc9000391efe7 R11: 0000000000000000 R12: 0000000000000000
R13: ffff8880311b8000 R14: 0000000000000002 R15: 0000000000000018
FS:  00007f602d1fe6c0(0000) GS:ffff8880d6675000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000561c522a6000 CR3: 000000002e99e000 CR4: 0000000000352ef0
Call Trace:
  <TASK>
  udp_tunnel_sock_release+0x68/0x80 net/ipv4/udp_tunnel_core.c:202
  rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
  rxe_sock_put+0xae/0x130 drivers/infiniband/sw/rxe/rxe_net.c:639
  rxe_net_del+0x83/0x120 drivers/infiniband/sw/rxe/rxe_net.c:660
  rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
  nldev_dellink+0x289/0x3c0 drivers/infiniband/core/nldev.c:1849
  rdma_nl_rcv_msg+0x392/0x6f0 drivers/infiniband/core/netlink.c:195
  rdma_nl_rcv_skb.constprop.0.isra.0+0x2cb/0x410 drivers/infiniband/core/netlink.c:239
  netlink_unicast_kernel net/netlink/af_netlink.c:1318 [inline]
  netlink_unicast+0x585/0x850 net/netlink/af_netlink.c:1344
  netlink_sendmsg+0x8b0/0xda0 net/netlink/af_netlink.c:1894
  sock_sendmsg_nosec net/socket.c:787 [inline]
  __sock_sendmsg net/socket.c:802 [inline]
  ____sys_sendmsg+0x9e1/0xb70 net/socket.c:2698
  ___sys_sendmsg+0x190/0x1e0 net/socket.c:2752
  __sys_sendmsg+0x170/0x220 net/socket.c:2784
  do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
  do_syscall_64+0x10b/0xf80 arch/x86/entry/syscall_64.c:94
  entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f602db9cdd9
Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 e8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f602d1fe028 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f602de16090 RCX: 00007f602db9cdd9
RDX: 0000000000000000 RSI: 00002000000002c0 RDI: 0000000000000007
RBP: 00007f602dc32d69 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007f602de16128 R14: 00007f602de16090 R15: 00007ffc1d89c428
  </TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:kernel_sock_shutdown+0x47/0x70 net/socket.c:3785
Code: fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 75 33 48 b8 00 00 00 00 00 fc ff df 4c 8b 63 20 49 8d 7c 24 68 48 89 fa 48 c1 ea 03 <80> 3c 02 00 75 1a 49 8b 44 24 68 89 ee 48 89 df 5b 5d 41 5c ff e0
RSP: 0018:ffffc9000391f180 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: ffff88802a2a0040 RCX: ffffffff8b8b72bd
RDX: 000000000000000d RSI: ffffffff89553b32 RDI: 0000000000000068
RBP: 0000000000000002 R08: 0000000000000001 R09: fffff52000723dfc
R10: ffffc9000391efe7 R11: 0000000000000000 R12: 0000000000000000
R13: ffff8880311b8000 R14: 0000000000000002 R15: 0000000000000018
FS:  00007f602d1fe6c0(0000) GS:ffff8880d6675000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000561c522a6000 CR3: 000000002e99e000 CR4: 0000000000352ef0
----------------
Code disassembly (best guess):
    0:	fc                   	cld
    1:	ff                   	lcall  (bad)
    2:	df 48 89             	fisttps -0x77(%rax)
    5:	fa                   	cli
    6:	48 c1 ea 03          	shr    $0x3,%rdx
    a:	80 3c 02 00          	cmpb   $0x0,(%rdx,%rax,1)
    e:	75 33                	jne    0x43
   10:	48 b8 00 00 00 00 00 	movabs $0xdffffc0000000000,%rax
   17:	fc ff df
   1a:	4c 8b 63 20          	mov    0x20(%rbx),%r12
   1e:	49 8d 7c 24 68       	lea    0x68(%r12),%rdi
   23:	48 89 fa             	mov    %rdi,%rdx
   26:	48 c1 ea 03          	shr    $0x3,%rdx
* 2a:	80 3c 02 00          	cmpb   $0x0,(%rdx,%rax,1) <-- trapping instruction
   2e:	75 1a                	jne    0x4a
   30:	49 8b 44 24 68       	mov    0x68(%r12),%rax
   35:	89 ee                	mov    %ebp,%esi
   37:	48 89 df             	mov    %rbx,%rdi
   3a:	5b                   	pop    %rbx
   3b:	5d                   	pop    %rbp
   3c:	41 5c                	pop    %r12
   3e:	ff e0                	jmp    *%rax

---
If you want syzbot to run the reproducer, reply with:
#syz test: git://repo/address.git branch-or-commit-hash
If you attach or paste a git patch, syzbot will apply it before testing.

Signature · 2 lines
Best Regards,
Yanjun.Zhu
#8 lore raw
Quoted 24 lines · Zhu Yanjun
Quoted 15 lines · syzbot

syzbot has found a reproducer for the following issue on:

HEAD commit:    74fe02ce122a Merge tag 'wq-for-7.1-rc2-fixes' of git://git..
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=16e895ce580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=59da38148f3a3d24
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       gcc (Debian 14.2.0-19) 14.2.0, GNU ld (GNU Binutils for Debian) 2.44
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=13a613ba580000

Downloadable assets:
disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/d900f083ada3/non_bootable_disk-74fe02ce.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/c0a591d96864/vmlinux-74fe02ce.xz
kernel image: https://storage.googleapis.com/syzbot-assets/9f94fb623cd1/bzImage-74fe02ce.xz

Quoted 5 lines

IMPORTANT: if you fix the issue, please add the following tag to the commit:

Reported-by [email protected]
Oops: general protection fault, probably for non-canonical address 0xdffffc000000000d: 0000 [#1] SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f]
Quoted 7 lines

Thanks a lot. IIRC, this problem is in process. The link is
https://patchwork.kernel.org/project/linux-rdma/patch/[email protected]/

Hi, Kuniyuki Iwashima

I think you are fixing this problem. I hope that we can see your commit very soon.

Yes, I was sidetracked but will respin v3 this week.

#9 lore raw
In reply to syzbot
Quoted 10 lines

Date: Wed, 06 May 2026 06:48:30 -0700 [thread overview]
syzbot has found a reproducer for the following issue on:

HEAD commit:    74fe02ce122a Merge tag 'wq-for-7.1-rc2-fixes' of git://git..
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=16e895ce580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=59da38148f3a3d24
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       gcc (Debian 14.2.0-19) 14.2.0, GNU ld (GNU Binutils for Debian) 2.44
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=13a613ba580000

#syz test

net/socket.c +5 -1
@@ -3782,7 +3782,11 @@ EXPORT_SYMBOL(kernel_getpeername);
 
 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
 {
-	return READ_ONCE(sock->ops)->shutdown(sock, how);
+	const struct proto_ops *ops = READ_ONCE(sock->ops);
+	if (ops)
+		return ops->shutdown(sock, how);
+	else
+		return 0;
 }
 EXPORT_SYMBOL(kernel_sock_shutdown);
 

--

author syzbot
#10 lore raw
syzbot tried to test the proposed patch but the build/boot failed:

Hello,

syzbot tried to test the proposed patch but the build/boot failed:

lost connection to test machine

syzkaller login: qemu-system-x86_64: ahci: PRDT length for NCQ command (0x0) is smaller than the requested size (0x1cc000)
Warning: Permanently added '[localhost]:12337' (ED25519) to the list of known hosts.

[   97.096781][   T10] cfg80211: failed to load regulatory.db
[  152.152171][ T1025] ata1.00: exception Emask 0x0 SAct 0x800 SErr 0x0 action 0x6 frozen
[  152.155707][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.158413][ T1025] ata1.00: cmd 61/60:58:36:81:04/0e:00:00:00:00/40 tag 11 ncq dma 1884160 ou
[  152.158413][ T1025]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
[  152.167703][ T1025] ata1.00: status: { DRDY }
[  152.169908][ T1025] ata1: hard resetting link
[  152.494622][ T1025] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
[  152.499274][ T1025] ata1.00: configured for UDMA/100
[  152.502602][ T1025] ata1: EH complete
qemu-system-x86_64: ahci: PRDT length for NCQ command (0x0) is smaller than the requested size (0xc2000)
[  152.536671][ T1025] ata1.00: Read log 0x10 page 0x00 failed, Emask 0x1
[  152.539788][ T1025] ata1: failed to read log page 10h (errno=-5)
[  152.543143][ T1025] ata1.00: NCQ disabled due to excessive errors
[  152.546416][ T1025] ata1.00: exception Emask 0x1 SAct 0xfc00 SErr 0x0 action 0x0
[  152.549623][ T1025] ata1.00: irq_stat 0x41000000
[  152.552697][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.555276][ T1025] ata1.00: cmd 61/10:50:36:01:05/0c:00:00:00:00/40 tag 10 ncq dma 1581056 ou
[  152.555276][ T1025]          res 50/00:00:00:00:00/00:00:00:00:00/00 Emask 0x1 (device error)
[  152.564262][ T1025] ata1.00: status: { DRDY }
[  152.567499][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.570300][ T1025] ata1.00: cmd 61/b0:58:46:0d:05/03:00:00:00:00/40 tag 11 ncq dma 483328 out
[  152.570300][ T1025]          res 50/00:00:00:00:00/00:00:00:00:00/00 Emask 0x1 (device error)
[  152.578750][ T1025] ata1.00: status: { DRDY }
[  152.580940][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.584242][ T1025] ata1.00: cmd 61/c8:60:f6:10:05/05:00:00:00:00/40 tag 12 ncq dma 757760 out
[  152.584242][ T1025]          res 50/00:00:00:00:00/00:00:00:00:00/00 Emask 0x1 (device error)
[  152.591624][ T1025] ata1.00: status: { DRDY }
[  152.594145][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.597218][ T1025] ata1.00: cmd 61/f0:68:be:16:05/02:00:00:00:00/40 tag 13 ncq dma 385024 out
[  152.597218][ T1025]          res 50/00:00:00:00:00/00:00:00:00:00/00 Emask 0x1 (device error)
[  152.605201][ T1025] ata1.00: status: { DRDY }
[  152.607461][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.610119][ T1025] ata1.00: cmd 61/10:70:ae:19:05/06:00:00:00:00/40 tag 14 ncq dma 794624 out
[  152.610119][ T1025]          res 50/00:00:00:00:00/00:00:00:00:00/00 Emask 0x1 (device error)
[  152.617854][ T1025] ata1.00: status: { DRDY }
[  152.620399][ T1025] ata1.00: failed command: WRITE FPDMA QUEUED
[  152.623555][ T1025] ata1.00: cmd 61/f8:78:be:1f:05/02:00:00:00:00/40 tag 15 ncq dma 389120 out
[  152.623555][ T1025]          res 50/00:00:00:00:00/00:00:00:00:00/00 Emask 0x1 (device error)
[  152.632127][ T1025] ata1.00: status: { DRDY }
[  152.635455][ T1025] ata1.00: configured for UDMA/100
[  152.638343][ T1025] ata1: EH complete

qemu-system-x86_64: hw/ide/core.c:934: ide_dma_cb: Assertion `prep_size >= 0 && prep_size <= n * 512' failed. Connection to localhost closed by remote host.

syzkaller build log:
go env (err=<nil>)
AR='ar'
CC='gcc'
CGO_CFLAGS='-O2 -g'
CGO_CPPFLAGS=''
CGO_CXXFLAGS='-O2 -g'
CGO_ENABLED='1'
CGO_FFLAGS='-O2 -g'
CGO_LDFLAGS='-O2 -g'
CXX='g++'
GCCGO='gccgo'
GO111MODULE='auto'
GOAMD64='v1'
GOARCH='amd64'
GOAUTH='netrc'
GOBIN=''
GOCACHE='/syzkaller/.cache/go-build'
GOCACHEPROG=''
GODEBUG=''
GOENV='/syzkaller/.config/go/env'
GOEXE=''
GOEXPERIMENT=''
GOFIPS140='off'
GOFLAGS=''
GOGCCFLAGS='-fPIC -m64 -pthread -Wl,--no-gc-sections -fmessage-length=0 -ffile-prefix-map=/tmp/go-build2539546880=/tmp/go-build -gno-record-gcc-switches' GOHOSTARCH='amd64'
GOHOSTOS='linux'
GOINSECURE=''
GOMOD='/syzkaller/jobs/linux/gopath/src/github.com/google/syzkaller/go.mod'
GOMODCACHE='/syzkaller/jobs/linux/gopath/pkg/mod'
GONOPROXY=''
GONOSUMDB=''
GOOS='linux'
GOPATH='/syzkaller/jobs/linux/gopath'
GOPRIVATE=''
GOPROXY='https://proxy.golang.org,direct'
GOROOT='/usr/local/go'
GOSUMDB='sum.golang.org'
GOTELEMETRY='local'
GOTELEMETRYDIR='/syzkaller/.config/go/telemetry'
GOTMPDIR=''
GOTOOLCHAIN='auto'
GOTOOLDIR='/usr/local/go/pkg/tool/linux_amd64'
GOVCS=''
GOVERSION='go1.26.0'
GOWORK=''
PKG_CONFIG='pkg-config'

git status (err=<nil>)
HEAD detached at 23ad3581d162
nothing to commit, working tree clean

tput: No value for $TERM and no -T specified
tput: No value for $TERM and no -T specified
Makefile:31: run command via tools/syz-env for best compatibility, see:
Makefile:32: https://github.com/google/syzkaller/blob/master/docs/contributing.md#using-syz-env
go list -f '{{.Stale}}' -ldflags="-s -w -X github.com/google/syzkaller/prog.GitRevision=23ad3581d162728720256cdd0a99f8702ec9c4c5 -X github.com/google/syzkaller/prog.gitRevisionDate=20260506-081407" ./sys/syz-sysgen | grep -q false || go install -ldflags="-s -w -X github.com/google/syzkaller/prog.GitRevision=23ad3581d162728720256cdd0a99f8702ec9c4c5 -X github.com/google/syzkaller/prog.gitRevisionDate=20260506-081407" ./sys/syz-sysgen make .descriptions
tput: No value for $TERM and no -T specified
tput: No value for $TERM and no -T specified
Makefile:31: run command via tools/syz-env for best compatibility, see:
Makefile:32: https://github.com/google/syzkaller/blob/master/docs/contributing.md#using-syz-env
bin/syz-sysgen
touch .descriptions
GOOS=linux GOARCH=amd64 go build -ldflags="-s -w -X github.com/google/syzkaller/prog.GitRevision=23ad3581d162728720256cdd0a99f8702ec9c4c5 -X github.com/google/syzkaller/prog.gitRevisionDate=20260506-081407" -o ./bin/linux_amd64/syz-execprog github.com/google/syzkaller/tools/syz-execprog
mkdir -p ./bin/linux_amd64

g++ -o ./bin/linux_amd64/syz-executor executor/executor.cc \
	-m64 -O2 -pthread -Wall -Werror -Wparentheses -Wunused-const-variable -Wframe-larger-than=16384 -Wno-stringop-overflow -Wno-array-bounds -Wno-format-overflow -Wno-unused-but-set-variable -Wno-unused-command-line-argument -static-pie -std=c++17 -I. -Iexecutor/_include   -DGOOS_linux=1 -DGOARCH_amd64=1 \
	-DHOSTGOOS_linux=1 -DGIT_REVISION=\"23ad3581d162728720256cdd0a99f8702ec9c4c5\"

go: downloading golang.org/x/sync v0.20.0
go: downloading go.opentelemetry.io/otel/sdk v1.43.0
go: downloading google.golang.org/grpc v1.80.0
go: downloading go.opentelemetry.io/otel v1.43.0
go: downloading go.opentelemetry.io/otel/trace v1.43.0
go: downloading google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9
go: downloading golang.org/x/net v0.52.0
go: downloading google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9
go: downloading github.com/ianlancetaylor/demangle v0.0.0-20260505044615-1ff4bf46051f
go: downloading go.opentelemetry.io/otel/sdk/metric v1.43.0
go: downloading go.opentelemetry.io/otel/metric v1.43.0
go: downloading golang.org/x/crypto v0.49.0
go: downloading golang.org/x/text v0.35.0
go: downloading github.com/go-jose/go-jose/v4 v4.1.4
/usr/bin/ld: /tmp/ccMZeaB8.o: in function `Connection::Connect(char const*, char const*)':
executor.cc:(.text._ZN10Connection7ConnectEPKcS1_[_ZN10Connection7ConnectEPKcS1_]+0x386): warning: Using 'gethostbyname' in statically linked applications requires at runtime the shared libraries from the glibc version used for linking ./tools/check-syzos.sh 2>/dev/null

Tested on:

commit:         5862221f Merge tag 'parisc-for-7.1-rc3' of git://git.k..
git tree:       upstream
kernel config:  https://syzkaller.appspot.com/x/.config?x=7f195f6be48c12ec
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       Debian clang version 21.1.8 (++20251221033036+2078da43e25a-1~exp1~20251221153213.50), Debian LLD 21.1.8
patch:          https://syzkaller.appspot.com/x/patch.diff?x=1583eece580000
author syzbot
#11 lore raw
In reply to syzbot
syzbot has found a reproducer for the following issue on:

syzbot has found a reproducer for the following issue on:

HEAD commit:    735d2f48cada Add linux-next specific files for 20260506
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=14f0e56a580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=a88880f0f312e277
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       Debian clang version 21.1.8 (++20251221033036+2078da43e25a-1~exp1~20251221153213.50), Debian LLD 21.1.8
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=125c9f6c580000
C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=166580ec580000

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/e65b731bdb98/disk-735d2f48.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/60db2f3d3f2f/vmlinux-735d2f48.xz
kernel image: https://storage.googleapis.com/syzbot-assets/55da282f7ab4/bzImage-735d2f48.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:

Reported-by [email protected]

rdma_rxe: rxe_newlink: failed to add lo

Oops: gen[  127.022080][ T5982] Oops: general protection fault, probably for non-canonical address 0xdffffc0000000004: 0000 [#1] SMP KASAN PTI
KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
CPU: 1 UID: 0 PID: 5982 Comm: syz.3.20 Not tainted syzkaller #0 PREEMPT_{RT,(full)} 
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 04/18/2026
RIP: 0010:kernel_sock_shutdown+0x2a/0x70 net/socket.c:3803
Code: f3 0f 1e fa 41 57 41 56 41 54 53 89 f3 49 89 fe 49 bc 00 00 00 00 00 fc ff df e8 e1 25 c5 f8 4d 8d 7e 20 4c 89 f8 48 c1 e8 03 <42> 80 3c 20 00 74 08 4c 89 ff e8 27 bf 2e f9 4d 8b 3f 49 83 c7 68
RSP: 0018:ffffc900015ef090 EFLAGS: 00010202
RAX: 0000000000000004 RBX: 0000000000000002 RCX: ffff88802dd89ec0
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
R10: dffffc0000000000 R11: ffffed1007cc8979 R12: dffffc0000000000
R13: dffffc0000000000 R14: 0000000000000000 R15: 0000000000000020
FS:  000055556d432500(0000) GS:ffff888125dca000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000001b34563fff CR3: 0000000042b1c000 CR4: 00000000003526f0
Call Trace:
 <TASK>
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
 nldev_dellink+0x304/0x3d0 drivers/infiniband/core/nldev.c:1849
 rdma_nl_rcv_msg drivers/infiniband/core/netlink.c:-1 [inline]
 rdma_nl_rcv_skb drivers/infiniband/core/netlink.c:239 [inline]
 rdma_nl_rcv+0x6d7/0xa10 drivers/infiniband/core/netlink.c:259
 netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline]
 netlink_unicast+0x780/0x920 net/netlink/af_netlink.c:1345
 netlink_sendmsg+0x813/0xb40 net/netlink/af_netlink.c:1895
 sock_sendmsg_nosec+0x112/0x150 net/socket.c:797
 __sock_sendmsg net/socket.c:812 [inline]
 ____sys_sendmsg+0x55c/0x870 net/socket.c:2716
 ___sys_sendmsg+0x2a5/0x360 net/socket.c:2770
 __sys_sendmsg net/socket.c:2802 [inline]
 __do_sys_sendmsg net/socket.c:2807 [inline]
 __se_sys_sendmsg net/socket.c:2805 [inline]
 __x64_sys_sendmsg+0x1c3/0x2a0 net/socket.c:2805
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0x15f/0xf80 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f89172fcdd9
Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 e8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ffe8bf8c018 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f8917575fa0 RCX: 00007f89172fcdd9
RDX: 0000000000000000 RSI: 00002000000002c0 RDI: 0000000000000006
RBP: 00007f8917392d69 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007f8917575fac R14: 00007f8917575fa0 R15: 00007f8917575fa0
 </TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:kernel_sock_shutdown+0x2a/0x70 net/socket.c:3803
Code: f3 0f 1e fa 41 57 41 56 41 54 53 89 f3 49 89 fe 49 bc 00 00 00 00 00 fc ff df e8 e1 25 c5 f8 4d 8d 7e 20 4c 89 f8 48 c1 e8 03 <42> 80 3c 20 00 74 08 4c 89 ff e8 27 bf 2e f9 4d 8b 3f 49 83 c7 68
RSP: 0018:ffffc900015ef090 EFLAGS: 00010202
RAX: 0000000000000004 RBX: 0000000000000002 RCX: ffff88802dd89ec0
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
R10: dffffc0000000000 R11: ffffed1007cc8979 R12: dffffc0000000000
R13: dffffc0000000000 R14: 0000000000000000 R15: 0000000000000020
FS:  000055556d432500(0000) GS:ffff888125dca000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000008 CR3: 0000000042b1c000 CR4: 00000000003526f0
----------------
Code disassembly (best guess):
   0:	f3 0f 1e fa          	endbr64
   4:	41 57                	push   %r15
   6:	41 56                	push   %r14
   8:	41 54                	push   %r12
   a:	53                   	push   %rbx
   b:	89 f3                	mov    %esi,%ebx
   d:	49 89 fe             	mov    %rdi,%r14
  10:	49 bc 00 00 00 00 00 	movabs $0xdffffc0000000000,%r12
  17:	fc ff df
  1a:	e8 e1 25 c5 f8       	call   0xf8c52600
  1f:	4d 8d 7e 20          	lea    0x20(%r14),%r15
  23:	4c 89 f8             	mov    %r15,%rax
  26:	48 c1 e8 03          	shr    $0x3,%rax
* 2a:	42 80 3c 20 00       	cmpb   $0x0,(%rax,%r12,1) <-- trapping instruction
  2f:	74 08                	je     0x39
  31:	4c 89 ff             	mov    %r15,%rdi
  34:	e8 27 bf 2e f9       	call   0xf92ebf60
  39:	4d 8b 3f             	mov    (%r15),%r15
  3c:	49 83 c7 68          	add    $0x68,%r15

---
If you want syzbot to run the reproducer, reply with:
#syz test: git://repo/address.git branch-or-commit-hash
If you attach or paste a git patch, syzbot will apply it before testing.

#12 lore raw

#syz test

drivers/infiniband/core/nldev.c +4 -0
index 96c745d5bac4..3cb3cb7629fe 100644
@@ -1816,6 +1816,8 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return err;
 }
 
+static DEFINE_MUTEX(nldev_dellink_mutex);
+
 static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {
@@ -1846,7 +1848,9 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	 * implicitly scoped to the driver supporting dynamic link deletion like RXE.
 	 */
 	if (device->link_ops && device->link_ops->dellink) {
+		mutex_lock(&nldev_dellink_mutex);
 		err = device->link_ops->dellink(device);
+		mutex_unlock(&nldev_dellink_mutex);
 		if (err)
 			return err;
 	}
author syzbot Tested-by
#13 lore raw
syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Hello,

syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Tested on:

commit:         735d2f48 Add linux-next specific files for 20260506
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=15d2c196580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=a88880f0f312e277
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       Debian clang version 21.1.8 (++20251221033036+2078da43e25a-1~exp1~20251221153213.50), Debian LLD 21.1.8
patch:          https://syzkaller.appspot.com/x/patch.diff?x=101a5f48580000

Note: testing is done by a robot and is best-effort only.

#14 lore raw

PATCH 1/1 RDMA/nldev: add mutual exclusion in nldev_dellink()

In reply to syzbot

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
Fixes a60e3f3d6fba ("RDMA/nldev: Add dellink function pointer")
Reported-by [email protected]
---
 drivers/infiniband/core/nldev.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/infiniband/core/nldev.c +4 -0
index 96c745d5bac4..3cb3cb7629fe 100644
@@ -1816,6 +1816,8 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return err;
 }
 
+static DEFINE_MUTEX(nldev_dellink_mutex);
+
 static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {
@@ -1846,7 +1848,9 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	 * implicitly scoped to the driver supporting dynamic link deletion like RXE.
 	 */
 	if (device->link_ops && device->link_ops->dellink) {
+		mutex_lock(&nldev_dellink_mutex);
 		err = device->link_ops->dellink(device);
+		mutex_unlock(&nldev_dellink_mutex);
 		if (err)
 			return err;
 	}
Signature · 1 line
2.43.0
#15 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
  udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
  rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
  rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
  rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
  rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
Quoted 5 lines
Fixes a60e3f3d6fba ("RDMA/nldev: Add dellink function pointer")
Reported-by [email protected]

Thanks a lot. This looks like a good solution. Since the issue is reproducible,

have you sent this commit to syzbot for verification?

Thanks,

Zhu Yanjun

Quoted 27 lines
---
  drivers/infiniband/core/nldev.c | 4 ++++
  1 file changed, 4 insertions(+)
drivers/infiniband/core/nldev.c +4 -0
index 96c745d5bac4..3cb3cb7629fe 100644
@@ -1816,6 +1816,8 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
  	return err;
  }
 
+static DEFINE_MUTEX(nldev_dellink_mutex);
+
  static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
  			  struct netlink_ext_ack *extack)
  {
@@ -1846,7 +1848,9 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
  	 * implicitly scoped to the driver supporting dynamic link deletion like RXE.
  	 */
  	if (device->link_ops && device->link_ops->dellink) {
+		mutex_lock(&nldev_dellink_mutex);
  		err = device->link_ops->dellink(device);
+		mutex_unlock(&nldev_dellink_mutex);
  		if (err)
  			return err;
  	}
Signature · 2 lines
Best Regards,
Yanjun.Zhu
#16 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 17 lines · Zhu Yanjun
Quoted 11 lines

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
  udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
  rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
  rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
  rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
  rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
Quoted 5 lines
Fixes a60e3f3d6fba ("RDMA/nldev: Add dellink function pointer")
Reported-by [email protected]
Quoted 4 lines

Thanks a lot. This looks like a good solution. Since the issue is reproducible,

have you sent this commit to syzbot for verification?

The patch has been verified by syzbot.

BR,
Edward

Reviewed-by
#17 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 17 lines · Edward Adam Davis
Quoted 16 lines · Zhu Yanjun
Quoted 16 lines

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
   udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
   rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
   rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
   rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
   rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
Fixes a60e3f3d6fba ("RDMA/nldev: Add dellink function pointer")
Reported-by [email protected]
Quoted 5 lines
Quoted 4 lines

Thanks a lot. This looks like a good solution. Since the issue is reproducible,

have you sent this commit to syzbot for verification?

The patch has been verified by syzbot.

Thanks a lot.

Zhu Yanjun

Quoted 4 lines

BR,
Edward

Signature · 2 lines
Best Regards,
Yanjun.Zhu
#18 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

In reply to Edward Adam Davis
Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

[...]

Applied, thanks!

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

Best regards,

Signature · 1 line
Leon Romanovsky <[email protected]>
#19 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 12 lines · Leon Romanovsky
Quoted 10 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254
Quoted 6 lines
Quoted 1 line

[...]

Applied, thanks!

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

This seems like a rxe bug, I would have expected the lock to be inside rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER...

ie it should not change nldev_dellink().

Jason

#20 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 15 lines · Jason Gunthorpe
Quoted 14 lines · Leon Romanovsky
Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

[...]

Quoted 8 lines
Quoted 4 lines

Applied, thanks!

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

This seems like a rxe bug, I would have expected the lock to be inside rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER...

No, it was triggered by RDMA_NLDEV_CMD_DELLINK, you can see the "call trace".

Quoted 2 lines

ie it should not change nldev_dellink().

While this could be fixed within RXE, the same issue affects all other RXE-like submodules when they subsequently support the "dellink" interface, therefore, handling this within nldev_dellink() is relatively more appropriate.

Edward

#21 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 24 lines · Edward Adam Davis
Quoted 15 lines · Jason Gunthorpe
Quoted 14 lines · Leon Romanovsky
Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

[...]

Quoted 8 lines
Quoted 4 lines

Applied, thanks!

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

This seems like a rxe bug, I would have expected the lock to be inside rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER...

Quoted 6 lines

No, it was triggered by RDMA_NLDEV_CMD_DELLINK, you can see the "call trace".

Quoted 2 lines

ie it should not change nldev_dellink().

While this could be fixed within RXE, the same issue affects all other RXE-like submodules when they subsequently support the "dellink" interface, therefore, handling this within nldev_dellink() is relatively more appropriate.

Why would other modules have an issue? The problem is rxe's racey refcounting scheme for its lazy socket creation. There is nothing wrong with nldev, and now you've created some nasty BKL in the nldev code to fix rxe while ignoring its other races.

Jason

#22 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 19 lines · Jason Gunthorpe
Quoted 18 lines · Edward Adam Davis
Quoted 17 lines · Jason Gunthorpe
Quoted 16 lines · Leon Romanovsky
Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

[...]

Applied, thanks!

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

This seems like a rxe bug, I would have expected the lock to be inside rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER...

No, it was triggered by RDMA_NLDEV_CMD_DELLINK, you can see the "call trace".

Not that Jason's point. Code wise

rxe_dellink -> rxe_net_del

netdev NETDEV_UNREGISTER:
rxe_notify -> rxe_net_del

both can lead to the same problem

Quoted 10 lines
Quoted 5 lines
Quoted 2 lines

ie it should not change nldev_dellink().

While this could be fixed within RXE, the same issue affects all other RXE-like submodules when they subsequently support the "dellink" interface, therefore, handling this within nldev_dellink() is relatively more appropriate.

Why would other modules have an issue? The problem is rxe's racey refcounting scheme for its lazy socket creation. There is nothing wrong with nldev, and now you've created some nasty BKL in the nldev code to fix rxe while ignoring its other races.

+1

#23 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 43 lines · David Ahern
Quoted 19 lines · Jason Gunthorpe
Quoted 18 lines · Edward Adam Davis
Quoted 17 lines · Jason Gunthorpe
Quoted 16 lines · Leon Romanovsky
Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

[...]

Applied, thanks!

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

This seems like a rxe bug, I would have expected the lock to be inside rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER...

No, it was triggered by RDMA_NLDEV_CMD_DELLINK, you can see the "call trace".

Not that Jason's point. Code wise

rxe_dellink -> rxe_net_del

netdev NETDEV_UNREGISTER:
rxe_notify -> rxe_net_del

both can lead to the same problem

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

ie it should not change nldev_dellink().

While this could be fixed within RXE, the same issue affects all other RXE-like submodules when they subsequently support the "dellink" interface, therefore, handling this within nldev_dellink() is relatively more appropriate.

Quoted 6 lines
Quoted 4 lines

Why would other modules have an issue? The problem is rxe's racey refcounting scheme for its lazy socket creation. There is nothing wrong with nldev, and now you've created some nasty BKL in the nldev code to fix rxe while ignoring its other races.

+1

Edward, please come with a fixup on top of this since it was already applied

Jason

#24 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 4 lines · Jason Gunthorpe

Edward, please come with a fixup on top of this since it was already applied

Zhu Yanjun: As author of the patch that introduced the bug and maintainer of the rxe code, why have you not addressed this problem? It has been well known for many weeks now and multiple people have attempted fixes. Seems like you need to step up and take care of it.

#25 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

Quoted 7 lines · David Ahern
Quoted 3 lines · Jason Gunthorpe

Edward, please come with a fixup on top of this since it was already applied

Zhu Yanjun: As author of the patch that introduced the bug and maintainer of the rxe code, why have you not addressed this problem? It has been well known for many weeks now and multiple people have

I am aware of the issue and have been following the discussion and proposed fixes.

I did not want to rush a change without fully understanding the implications on RXE

behavior and existing users. I am currently reviewing the proposed approaches and

working on a proper fix.

I appreciate everyone who helped investigate and test the issue.

Zhu Yanjun

Quoted 2 lines

attempted fixes. Seems like you need to step up and take care of it.

#26 lore raw

Re: [PATCH] RDMA/nldev: add mutual exclusion in nldev_dellink()

In reply to Jason Gunthorpe
Quoted 49 lines · Jason Gunthorpe
Quoted 48 lines · David Ahern
Quoted 19 lines · Jason Gunthorpe
Quoted 18 lines · Edward Adam Davis
Quoted 17 lines · Jason Gunthorpe
Quoted 16 lines · Leon Romanovsky
Quoted 11 lines · Edward Adam Davis

We must serialize calls to nldev_dellink() or risk a crash as syzbot reported:

Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

[...]

Applied, thanks!

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

[1/1] RDMA/nldev: add mutual exclusion in nldev_dellink()

      https://git.kernel.org/rdma/rdma/c/0b28000b64f40d

This seems like a rxe bug, I would have expected the lock to be inside rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER...

No, it was triggered by RDMA_NLDEV_CMD_DELLINK, you can see the "call trace".

Not that Jason's point. Code wise

rxe_dellink -> rxe_net_del

netdev NETDEV_UNREGISTER:
rxe_notify -> rxe_net_del

both can lead to the same problem

Quoted 10 lines
Quoted 5 lines
Quoted 2 lines

ie it should not change nldev_dellink().

While this could be fixed within RXE, the same issue affects all other RXE-like submodules when they subsequently support the "dellink" interface, therefore, handling this within nldev_dellink() is relatively more appropriate.

Why would other modules have an issue? The problem is rxe's racey refcounting scheme for its lazy socket creation. There is nothing wrong with nldev, and now you've created some nasty BKL in the nldev code to fix rxe while ignoring its other races.

Quoted 4 lines
Quoted 1 line

+1

Edward, please come with a fixup on top of this since it was already applied

OK.

Edward

#27 lore raw

PATCH v2 1/1 RDMA/rxe: add mutual exclusion in rxe_net_del()

We must serialize calls to rxe_net_del() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
 udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
 rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
 rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
 rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
 rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

Jason Gunthorpe suggest placing the lock within rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER.

The patch addressing this issue in nldev_dellink() has already been applied(0b28000b64f4); however, since the fix has now been relocated to rxe, the corresponding remedial code in nldev has been removed.

Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Fixes 0b28000b64f4 ("RDMA/nldev: Add mutual exclusion in nldev_dellink()")
Reported-by [email protected]

---
v1 -> v2: serialize calls to rxe net del

 drivers/infiniband/core/nldev.c     | 4 ----
 drivers/infiniband/sw/rxe/rxe_net.c | 7 ++++++-
 2 files changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/core/nldev.c +0 -4
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return err;
 }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
 static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	 * implicitly scoped to the driver supporting dynamic link deletion like RXE.
 	 */
 	if (device->link_ops && device->link_ops->dellink) {
-		mutex_lock(&nldev_dellink_mutex);
 		err = device->link_ops->dellink(device);
-		mutex_unlock(&nldev_dellink_mutex);
 		if (err)
 			return err;
 	}
drivers/infiniband/sw/rxe/rxe_net.c +6 -1
index 50a2cb5405e2..92847e955ca2 100644
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
 	}
 }
 
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
 void rxe_net_del(struct ib_device *dev)
 {
 	struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
 	struct sock *sk;
 	struct net *net;
 
+	mutex_lock(&rxe_net_del_mutex);
 	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
 	if (!ndev)
-		return;
+		goto out;
 
 	net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
 		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
 	dev_put(ndev);
+out:
+	mutex_unlock(&rxe_net_del_mutex);
 }
 
 static void rxe_port_event(struct rxe_dev *rxe,
Signature · 1 line
2.43.0
#28 lore raw

Re: [PATCH RDMA v2] RDMA/rxe: add mutual exclusion in rxe_net_del()

Quoted 19 lines · Edward Adam Davis

We must serialize calls to rxe_net_del() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
  udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
  rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
  rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
  rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
  rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

Jason Gunthorpe suggest placing the lock within rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER.

The patch addressing this issue in nldev_dellink() has already been applied(0b28000b64f4); however, since the fix has now been relocated to rxe, the corresponding remedial code in nldev has been removed.

Quoted 7 lines
Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Fixes 0b28000b64f4 ("RDMA/nldev: Add mutual exclusion in nldev_dellink()")
Reported-by [email protected]

---
v1 -> v2: serialize calls to rxe net del

I looked through the commit. I am not sure if this commit should be sent to syzbot to verify.

Zhu Yanjun

Quoted 62 lines
  drivers/infiniband/core/nldev.c     | 4 ----
  drivers/infiniband/sw/rxe/rxe_net.c | 7 ++++++-
  2 files changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/core/nldev.c +0 -4
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
  	return err;
  }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
  static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
  			  struct netlink_ext_ack *extack)
  {
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
  	 * implicitly scoped to the driver supporting dynamic link deletion like RXE.
  	 */
  	if (device->link_ops && device->link_ops->dellink) {
-		mutex_lock(&nldev_dellink_mutex);
  		err = device->link_ops->dellink(device);
-		mutex_unlock(&nldev_dellink_mutex);
  		if (err)
  			return err;
  	}
drivers/infiniband/sw/rxe/rxe_net.c +6 -1
index 50a2cb5405e2..92847e955ca2 100644
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
  	}
  }
 
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
  void rxe_net_del(struct ib_device *dev)
  {
  	struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
  	struct sock *sk;
  	struct net *net;
 
+	mutex_lock(&rxe_net_del_mutex);
  	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
  	if (!ndev)
-		return;
+		goto out;
 
  	net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
  		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
  	dev_put(ndev);
+out:
+	mutex_unlock(&rxe_net_del_mutex);
  }
 
  static void rxe_port_event(struct rxe_dev *rxe,
#29 lore raw

Re: [PATCH RDMA v2] RDMA/rxe: add mutual exclusion in rxe_net_del()

In reply to Edward Adam Davis
Quoted 59 lines · Edward Adam Davis

We must serialize calls to rxe_net_del() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
  udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
  rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
  rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
  rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
  rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

Jason Gunthorpe suggest placing the lock within rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER.

The patch addressing this issue in nldev_dellink() has already been applied(0b28000b64f4); however, since the fix has now been relocated to rxe, the corresponding remedial code in nldev has been removed.

Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Fixes 0b28000b64f4 ("RDMA/nldev: Add mutual exclusion in nldev_dellink()")
Reported-by [email protected]

---
v1 -> v2: serialize calls to rxe net del

  drivers/infiniband/core/nldev.c     | 4 ----
  drivers/infiniband/sw/rxe/rxe_net.c | 7 ++++++-
  2 files changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/core/nldev.c +0 -4
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
  	return err;
  }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
  static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
  			  struct netlink_ext_ack *extack)
  {
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
  	 * implicitly scoped to the driver supporting dynamic link deletion like RXE.
  	 */
  	if (device->link_ops && device->link_ops->dellink) {
-		mutex_lock(&nldev_dellink_mutex);
  		err = device->link_ops->dellink(device);
-		mutex_unlock(&nldev_dellink_mutex);
  		if (err)
  			return err;
  	}
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..92847e955ca2 100644

drivers/infiniband/sw/rxe/rxe_net.c static void rxe_sock_put(struct sock *sk, · line 644

Quoted 3 lines
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
  	}
  }
 

I read this commit carefully. There are two paths that can invoke rxe_net_del().

One is through the rdma link del xxx command, while the other is through the netdevice notification chain.

In the netdevice notification chain path, rtnl_lock is already held, and rxe_net_del() is called under that lock.

However, in the rdma link del xxx path, no rtnl_lock is taken.

Because of this, I would like to use the existing rtnl_lock to serialize calls to rxe_net_del().

My proposed commit is shown below. I am not sure whether it fully resolves the problem.

drivers/infiniband/sw/rxe/rxe.c +0 -0
index b0714f9abe3d..84266dc416c4 100644
@@ -251,7 +251,9 @@ static int rxe_newlink(const char *ibdev_name, struct net_device *ndev)
 

 static int rxe_dellink(struct ib_device *dev)
 {
+       rtnl_lock();
        rxe_net_del(dev);
+       rtnl_unlock();

        return 0;
 }

drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..ac53ea73996d 100644
@@ -649,6 +649,8 @@ void rxe_net_del(struct ib_device *dev)

        struct sock *sk;
        struct net *net;

+       ASSERT_RTNL();
+
        ndev = rxe_ib_device_get_netdev(&rxe->ib_dev); if (!ndev)
                return;

Zhu Yanjun

Quoted 26 lines
drivers/infiniband/sw/rxe/rxe_net.c +6 -1
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
  void rxe_net_del(struct ib_device *dev)
  {
  	struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
  	struct sock *sk;
  	struct net *net;
 
+	mutex_lock(&rxe_net_del_mutex);
  	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
  	if (!ndev)
-		return;
+		goto out;
 
  	net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
  		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
  	dev_put(ndev);
+out:
+	mutex_unlock(&rxe_net_del_mutex);
  }
 
  static void rxe_port_event(struct rxe_dev *rxe,
#30 lore raw

Re: [PATCH RDMA v2] RDMA/rxe: add mutual exclusion in rxe_net_del()

Quoted 22 lines · Zhu Yanjun
drivers/infiniband/sw/rxe/rxe.c +2 -0
index b0714f9abe3d..84266dc416c4 100644
@@ -251,7 +251,9 @@ static int rxe_newlink(const char *ibdev_name, 
 struct net_device *ndev)
 
  static int rxe_dellink(struct ib_device *dev)
  {
+       rtnl_lock();
         rxe_net_del(dev);
+       rtnl_unlock();
 
         return 0;
  }
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..ac53ea73996d 100644

drivers/infiniband/sw/rxe/rxe_net.c void rxe_net_del(struct ib_device *dev) · line 656

Quoted 8 lines
drivers/infiniband/sw/rxe/rxe_net.c +2 -0
@@ -649,6 +649,8 @@ void rxe_net_del(struct ib_device *dev)
         struct sock *sk;
         struct net *net;
 
+       ASSERT_RTNL();
+
         ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
         if (!ndev)
                 return;

Since the solution is the same, and requires no new locks, I favor your approach.

BR,
Edward

#31 lore raw

Re: [PATCH RDMA v2] RDMA/rxe: add mutual exclusion in rxe_net_del()

In reply to Yanjun.Zhu
Quoted 71 lines · Yanjun.Zhu
Quoted 59 lines · Edward Adam Davis

We must serialize calls to rxe_net_del() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
  udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
  rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
  rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
  rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
  rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

Jason Gunthorpe suggest placing the lock within rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER.

The patch addressing this issue in nldev_dellink() has already been applied(0b28000b64f4); however, since the fix has now been relocated to rxe, the corresponding remedial code in nldev has been removed.

Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Fixes 0b28000b64f4 ("RDMA/nldev: Add mutual exclusion in nldev_dellink()")
Reported-by [email protected]

---
v1 -> v2: serialize calls to rxe net del

  drivers/infiniband/core/nldev.c     | 4 ----
  drivers/infiniband/sw/rxe/rxe_net.c | 7 ++++++-
  2 files changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/core/nldev.c +0 -4
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
      return err;
  }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
  static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
                        struct netlink_ext_ack *extack)
  {
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
       * implicitly scoped to the driver supporting dynamic link deletion like RXE.
       */
      if (device->link_ops && device->link_ops->dellink) {
-             mutex_lock(&nldev_dellink_mutex);
              err = device->link_ops->dellink(device);
-             mutex_unlock(&nldev_dellink_mutex);
              if (err)
                      return err;
      }
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..92847e955ca2 100644

drivers/infiniband/sw/rxe/rxe_net.c static void rxe_sock_put(struct sock *sk, · line 644

Quoted 3 lines
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
      }
  }
 

I read this commit carefully. There are two paths that can invoke rxe_net_del().

One is through the rdma link del xxx command, while the other is through the netdevice notification chain.

Quoted 7 lines

In the netdevice notification chain path, rtnl_lock is already held, and rxe_net_del() is called under that lock.

However, in the rdma link del xxx path, no rtnl_lock is taken.

Because of this, I would like to use the existing rtnl_lock to serialize calls to rxe_net_del().

-1 for this.

It's a global mutex and heavily contended because many components use it without much care. We are working to reduce the RTNL pressure for years by converting such users with a dedicated lock or per-netns RTNL mutex.

RTNL is not needed here at all, so please use a dedicated lock.

Quoted 63 lines

My proposed commit is shown below. I am not sure whether it fully resolves the problem.

drivers/infiniband/sw/rxe/rxe.c +2 -0
index b0714f9abe3d..84266dc416c4 100644
@@ -251,7 +251,9 @@ static int rxe_newlink(const char *ibdev_name,
 struct net_device *ndev)
 
  static int rxe_dellink(struct ib_device *dev)
  {
+       rtnl_lock();
         rxe_net_del(dev);
+       rtnl_unlock();
 
         return 0;
  }
drivers/infiniband/sw/rxe/rxe_net.c +2 -0
index 50a2cb5405e2..ac53ea73996d 100644
@@ -649,6 +649,8 @@ void rxe_net_del(struct ib_device *dev)
         struct sock *sk;
         struct net *net;
 
+       ASSERT_RTNL();
+
         ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
         if (!ndev)
                 return;

Zhu Yanjun

Quoted 26 lines
drivers/infiniband/sw/rxe/rxe_net.c +6 -1
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
  void rxe_net_del(struct ib_device *dev)
  {
      struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
      struct sock *sk;
      struct net *net;
 
+     mutex_lock(&rxe_net_del_mutex);
      ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
      if (!ndev)
-             return;
+             goto out;
 
      net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
              rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
      dev_put(ndev);
+out:
+     mutex_unlock(&rxe_net_del_mutex);
  }
 
  static void rxe_port_event(struct rxe_dev *rxe,
#32 lore raw

Re: [PATCH RDMA v2] RDMA/rxe: add mutual exclusion in rxe_net_del()

Quoted 78 lines · Kuniyuki Iwashima
Quoted 70 lines · Yanjun.Zhu
Quoted 59 lines · Edward Adam Davis

We must serialize calls to rxe_net_del() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
   udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
   rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
   rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
   rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
   rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

Jason Gunthorpe suggest placing the lock within rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER.

The patch addressing this issue in nldev_dellink() has already been applied(0b28000b64f4); however, since the fix has now been relocated to rxe, the corresponding remedial code in nldev has been removed.

Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Fixes 0b28000b64f4 ("RDMA/nldev: Add mutual exclusion in nldev_dellink()")
Reported-by [email protected]

---
v1 -> v2: serialize calls to rxe net del

   drivers/infiniband/core/nldev.c     | 4 ----
   drivers/infiniband/sw/rxe/rxe_net.c | 7 ++++++-
   2 files changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/core/nldev.c +0 -4
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
       return err;
   }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
   static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
                         struct netlink_ext_ack *extack)
   {
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
        * implicitly scoped to the driver supporting dynamic link deletion like RXE.
        */
       if (device->link_ops && device->link_ops->dellink) {
-             mutex_lock(&nldev_dellink_mutex);
               err = device->link_ops->dellink(device);
-             mutex_unlock(&nldev_dellink_mutex);
               if (err)
                       return err;
       }
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..92847e955ca2 100644

drivers/infiniband/sw/rxe/rxe_net.c static void rxe_sock_put(struct sock *sk, · line 644

Quoted 3 lines
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
       }
   }
 

I read this commit carefully. There are two paths that can invoke rxe_net_del().

One is through the rdma link del xxx command, while the other is through the netdevice notification chain.

Quoted 7 lines

In the netdevice notification chain path, rtnl_lock is already held, and rxe_net_del() is called under that lock.

However, in the rdma link del xxx path, no rtnl_lock is taken.

Because of this, I would like to use the existing rtnl_lock to serialize calls to rxe_net_del().

Quoted 8 lines

-1 for this.

It's a global mutex and heavily contended because many components use it without much care. We are working to reduce the RTNL pressure for years by converting such users with a dedicated lock or per-netns RTNL mutex.

RTNL is not needed here at all, so please use a dedicated lock.

Thanks a lot for your review. I think the following commit can fix this problem.

Please review.

From 80525f5b7fb0af18b9759cbde0237aabb76158cc Mon Sep 17 00:00:00 2001

From: Zhu Yanjun <[email protected]>
Date: Sat, 16 May 2026 22:27:35 +0200
Subject: [PATCH 1/1] RDMA/rxe: Fix Use-After-Free problem in rxe_net_del

syzbot reported a general protection fault (KASAN: null-ptr-deref) in kernel_sock_shutdown() called during the software RoCE (rxe) link deletion path (rxe_dellink -> rxe_net_del).

The root cause is a TOCTOU (Time-of-Check to Time-of-Use) race condition in rxe_net_del(). Previously, the function fetched the socket pointer via rxe_ns_pernet_sk4/6() outside the critical section, and then acquired the lock to release it via rxe_sock_put().

In a highly concurrent teardown environment, another thread could close and clear the pernet socket after it was fetched but before the lock was acquired. This causes rxe_sock_put() to operate on a dangling or already cleared socket pointer, leading to a NULL pointer dereference when kernel_sock_shutdown() attempts to access sock->sk.

Fix this by introducing a dedicated, per-device mutex 'release_lock' and extending its scope. The socket pointers are now fetched, checked, and released entirely within the same locked critical section. This ensures the atomicity of the socket lookup and teardown sequence.

Reported-by [email protected]
Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")

---
 drivers/infiniband/sw/rxe/rxe.c       | 2 ++
 drivers/infiniband/sw/rxe/rxe_net.c   | 4 ++++
 drivers/infiniband/sw/rxe/rxe_verbs.h | 1 +

 3 files changed, 7 insertions(+)
drivers/infiniband/sw/rxe/rxe.c +0 -0
index b0714f9abe3d..46967ecdaf7d 100644
@@ -34,6 +34,7 @@ void rxe_dealloc(struct ib_device *ib_dev)

        WARN_ON(!RB_EMPTY_ROOT(&rxe->mcg_tree));

        mutex_destroy(&rxe->usdev_lock);
+       mutex_destroy(&rxe->release_lock);
 }

 static const struct ib_device_ops rxe_ib_dev_odp_ops = {
@@ -186,6 +187,7 @@ static void rxe_init(struct rxe_dev *rxe, struct net_device *ndev) rxe->mcg_tree = RB_ROOT;

        mutex_init(&rxe->usdev_lock);
+       mutex_init(&rxe->release_lock);
 }

 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu)

drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..c3b188538540 100644
@@ -655,6 +655,8 @@ void rxe_net_del(struct ib_device *dev)
 

        net = dev_net(ndev);

+       mutex_lock(&rxe->release_lock);
+
        sk = rxe_ns_pernet_sk4(net);
        if (sk)
                rxe_sock_put(sk, rxe_ns_pernet_set_sk4, net); @@ -663,6 +665,8 @@ void rxe_net_del(struct ib_device *dev) if (sk)
                rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);

+       mutex_unlock(&rxe->release_lock);
+
        dev_put(ndev);
 }

drivers/infiniband/sw/rxe/rxe_verbs.h +0 -0
index d92f80d16f78..3f54aa0a4356 100644
@@ -422,6 +422,7 @@ struct rxe_dev {

        int                     max_ucontext;
        int                     max_inline_data;
        struct mutex            usdev_lock;
+       struct mutex            release_lock;

        char                    raw_gid[ETH_ALEN];

--
2.43.0

Quoted 63 lines
Quoted 63 lines

My proposed commit is shown below. I am not sure whether it fully resolves the problem.

drivers/infiniband/sw/rxe/rxe.c +2 -0
index b0714f9abe3d..84266dc416c4 100644
@@ -251,7 +251,9 @@ static int rxe_newlink(const char *ibdev_name,
 struct net_device *ndev)
 
   static int rxe_dellink(struct ib_device *dev)
   {
+       rtnl_lock();
          rxe_net_del(dev);
+       rtnl_unlock();
 
          return 0;
   }
drivers/infiniband/sw/rxe/rxe_net.c +2 -0
index 50a2cb5405e2..ac53ea73996d 100644
@@ -649,6 +649,8 @@ void rxe_net_del(struct ib_device *dev)
          struct sock *sk;
          struct net *net;
 
+       ASSERT_RTNL();
+
          ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
          if (!ndev)
                  return;

Zhu Yanjun

Quoted 26 lines
drivers/infiniband/sw/rxe/rxe_net.c +6 -1
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
   void rxe_net_del(struct ib_device *dev)
   {
       struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
       struct sock *sk;
       struct net *net;
 
+     mutex_lock(&rxe_net_del_mutex);
       ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
       if (!ndev)
-             return;
+             goto out;
 
       net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
               rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
       dev_put(ndev);
+out:
+     mutex_unlock(&rxe_net_del_mutex);
   }
 
   static void rxe_port_event(struct rxe_dev *rxe,
Signature · 2 lines
Best Regards,
Yanjun.Zhu
#33 lore raw

Re: [PATCH RDMA v2] RDMA/rxe: add mutual exclusion in rxe_net_del()

Quoted 89 lines · Zhu Yanjun
Quoted 87 lines · Kuniyuki Iwashima
Quoted 72 lines · Yanjun.Zhu
Quoted 61 lines · Edward Adam Davis

We must serialize calls to rxe_net_del() or risk a crash as syzbot reported:

KASAN: null-ptr-deref in range [0x0000000000000020-0x0000000000000027]
Call Trace:
   udp_tunnel_sock_release+0x6d/0x80 net/ipv4/udp_tunnel_core.c:197
   rxe_release_udp_tunnel drivers/infiniband/sw/rxe/rxe_net.c:294 [inline]
   rxe_sock_put drivers/infiniband/sw/rxe/rxe_net.c:639 [inline]
   rxe_net_del+0xfb/0x290 drivers/infiniband/sw/rxe/rxe_net.c:660
   rxe_dellink+0x15/0x20 drivers/infiniband/sw/rxe/rxe.c:254

Jason Gunthorpe suggest placing the lock within rxe to protect its racy implementation of rxe_net_del(), which looks like it is possibly also triggered by NETDEV_UNREGISTER.

The patch addressing this issue in nldev_dellink() has already been applied(0b28000b64f4); however, since the fix has now been relocated to rxe, the corresponding remedial code in nldev has been removed.

Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
Fixes 0b28000b64f4 ("RDMA/nldev: Add mutual exclusion in nldev_dellink()")
Reported-by [email protected]

---
v1 -> v2: serialize calls to rxe net del

   drivers/infiniband/core/nldev.c     | 4 ---- drivers/infiniband/sw/rxe/rxe_net.c | 7 ++++++-

   2 files changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/core/nldev.c +0 -2
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, 
 struct nlmsghdr *nlh,
        return err;
    }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
    static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
                          struct netlink_ext_ack *extack)

   {

drivers/infiniband/core/nldev.c +0 -2
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, 
 struct nlmsghdr *nlh,
         * implicitly scoped to the driver supporting dynamic link deletion like RXE.
         */
        if (device->link_ops && device->link_ops->dellink) {
-             mutex_lock(&nldev_dellink_mutex);
                err = device->link_ops->dellink(device);
-             mutex_unlock(&nldev_dellink_mutex);
                if (err)
                        return err;

       }

drivers/infiniband/sw/rxe/rxe_net.c +0 -0
index 50a2cb5405e2..92847e955ca2 100644

drivers/infiniband/sw/rxe/rxe_net.c static void rxe_sock_put(struct sock *sk, · line 644

Quoted 3 lines
drivers/infiniband/sw/rxe/rxe_net.c +0 -0
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
        }
    }
 

I read this commit carefully. There are two paths that can invoke rxe_net_del().

One is through the rdma link del xxx command, while the other is through the netdevice notification chain.

Quoted 7 lines

In the netdevice notification chain path, rtnl_lock is already held, and rxe_net_del() is called under that lock.

However, in the rdma link del xxx path, no rtnl_lock is taken.

Because of this, I would like to use the existing rtnl_lock to serialize calls to rxe_net_del().

-1 for this.

It's a global mutex and heavily contended because many components use it without much care.  We are working to reduce the RTNL pressure for years by converting such users with a dedicated lock or per-netns RTNL mutex.

Quoted 5 lines
Quoted 1 line

RTNL is not needed here at all, so please use a dedicated lock.

Thanks a lot for your review. I think the following commit can fix this problem.

Please review.

The root cause is clear. If no one disagrees with this commit, I will send out the official patch.

In the latest revision, I will move the mutex lock into the network namespace.

I think we have discussed this problem thoroughly, and we all understand the root cause now.

Zhu Yanjun

Quoted 159 lines

From 80525f5b7fb0af18b9759cbde0237aabb76158cc Mon Sep 17 00:00:00 2001

From: Zhu Yanjun <[email protected]>
Date: Sat, 16 May 2026 22:27:35 +0200
Subject: [PATCH 1/1] RDMA/rxe: Fix Use-After-Free problem in rxe_net_del

syzbot reported a general protection fault (KASAN: null-ptr-deref) in kernel_sock_shutdown() called during the software RoCE (rxe) link deletion path (rxe_dellink -> rxe_net_del).

The root cause is a TOCTOU (Time-of-Check to Time-of-Use) race condition in rxe_net_del(). Previously, the function fetched the socket pointer via rxe_ns_pernet_sk4/6() outside the critical section, and then acquired the lock to release it via rxe_sock_put().

In a highly concurrent teardown environment, another thread could close and clear the pernet socket after it was fetched but before the lock was acquired. This causes rxe_sock_put() to operate on a dangling or already cleared socket pointer, leading to a NULL pointer dereference when kernel_sock_shutdown() attempts to access sock->sk.

Fix this by introducing a dedicated, per-device mutex 'release_lock' and extending its scope. The socket pointers are now fetched, checked, and released entirely within the same locked critical section. This ensures the atomicity of the socket lookup and teardown sequence.

Reported-by [email protected]
Fixes f1327abd6abe ("RDMA/rxe: Support RDMA link creation and destruction per net namespace")
---
  drivers/infiniband/sw/rxe/rxe.c       | 2 ++
  drivers/infiniband/sw/rxe/rxe_net.c   | 4 ++++
  drivers/infiniband/sw/rxe/rxe_verbs.h | 1 +
  3 files changed, 7 insertions(+)
drivers/infiniband/sw/rxe/rxe.c +2 -0
index b0714f9abe3d..46967ecdaf7d 100644
@@ -34,6 +34,7 @@ void rxe_dealloc(struct ib_device *ib_dev)
         WARN_ON(!RB_EMPTY_ROOT(&rxe->mcg_tree));
 
         mutex_destroy(&rxe->usdev_lock);
+       mutex_destroy(&rxe->release_lock);
  }
 
  static const struct ib_device_ops rxe_ib_dev_odp_ops = {
@@ -186,6 +187,7 @@ static void rxe_init(struct rxe_dev *rxe, struct 
 net_device *ndev)
         rxe->mcg_tree = RB_ROOT;
 
         mutex_init(&rxe->usdev_lock);
+       mutex_init(&rxe->release_lock);
  }
 

 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu)

drivers/infiniband/sw/rxe/rxe_net.c +4 -0
index 50a2cb5405e2..c3b188538540 100644
@@ -655,6 +655,8 @@ void rxe_net_del(struct ib_device *dev)
 
         net = dev_net(ndev);
 
+       mutex_lock(&rxe->release_lock);
+
         sk = rxe_ns_pernet_sk4(net);
         if (sk)
                 rxe_sock_put(sk, rxe_ns_pernet_set_sk4, net);
@@ -663,6 +665,8 @@ void rxe_net_del(struct ib_device *dev)
         if (sk)
                 rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
+       mutex_unlock(&rxe->release_lock);
+
         dev_put(ndev);
  }
 
drivers/infiniband/sw/rxe/rxe_verbs.h +1 -0
index d92f80d16f78..3f54aa0a4356 100644
@@ -422,6 +422,7 @@ struct rxe_dev {
         int                     max_ucontext;
         int                     max_inline_data;
         struct mutex            usdev_lock;
+       struct mutex            release_lock;
 
         char                    raw_gid[ETH_ALEN];
 
Signature · 1 line
2.43.0
Quoted 63 lines
Quoted 63 lines

My proposed commit is shown below. I am not sure whether it fully resolves the problem.

drivers/infiniband/sw/rxe/rxe.c +2 -0
index b0714f9abe3d..84266dc416c4 100644
@@ -251,7 +251,9 @@ static int rxe_newlink(const char *ibdev_name,
 struct net_device *ndev)
 
    static int rxe_dellink(struct ib_device *dev)
    {
+       rtnl_lock();
           rxe_net_del(dev);
+       rtnl_unlock();
 
           return 0;

   }

drivers/infiniband/sw/rxe/rxe_net.c +2 -0
index 50a2cb5405e2..ac53ea73996d 100644
@@ -649,6 +649,8 @@ void rxe_net_del(struct ib_device *dev)
           struct sock *sk;
           struct net *net;
 
+       ASSERT_RTNL();
+
           ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
           if (!ndev)
                   return;

Zhu Yanjun

Quoted 26 lines
drivers/infiniband/sw/rxe/rxe_net.c +6 -1
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
    void rxe_net_del(struct ib_device *dev)
    {
        struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
        struct sock *sk;
        struct net *net;
 
+     mutex_lock(&rxe_net_del_mutex);
        ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
        if (!ndev)
-             return;
+             goto out;
 
        net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
                rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
        dev_put(ndev);
+out:
+     mutex_unlock(&rxe_net_del_mutex);
    }
 
    static void rxe_port_event(struct rxe_dev *rxe,
author syzbot
#36 lore raw
syzbot has tested the proposed patch but the reproducer is still triggering an issue: lost connection to test machine

Hello,

syzbot has tested the proposed patch but the reproducer is still triggering an issue:
lost connection to test machine

Tested on:

commit:         bb9ed28b RDMA/rxe: Fix null-ptr-deref in kernel_sock_s..
git tree:       https://github.com/zhuyj/linux null-ptr-deref_kernel_sock_shutdown
console output: https://syzkaller.appspot.com/x/log.txt?x=16617cc8580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=eeba87c808be946b
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       Debian clang version 21.1.8 (++20251221033036+2078da43e25a-1~exp1~20251221153213.50), Debian LLD 21.1.8

Note: no patches were applied.

#37 lore raw
drivers/infiniband/core/nldev.c +0 -2
index 3cb3cb7629fe..96c745d5bac4 100644
@@ -1816,8 +1816,6 @@ static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return err;
 }
 
-static DEFINE_MUTEX(nldev_dellink_mutex);
-
 static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {
@@ -1848,9 +1846,7 @@ static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	 * implicitly scoped to the driver supporting dynamic link deletion 

like RXE.
*/
if (device->link_ops && device->link_ops->dellink) {
- mutex_lock(&nldev_dellink_mutex);
err = device->link_ops->dellink(device);
- mutex_unlock(&nldev_dellink_mutex);
if (err)
return err;
}

drivers/infiniband/sw/rxe/rxe_net.c +6 -1
index 50a2cb5405e2..92847e955ca2 100644
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
 	}
 }
 
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
 void rxe_net_del(struct ib_device *dev)
 {
 	struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
 	struct sock *sk;
 	struct net *net;
 
+	mutex_lock(&rxe_net_del_mutex);
 	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
 	if (!ndev)
-		return;
+		goto out;
 
 	net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
 		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
 	dev_put(ndev);
+out:
+	mutex_unlock(&rxe_net_del_mutex);
 }
 
 static void rxe_port_event(struct rxe_dev *rxe,
#38 lore raw
In reply to syzbot

#syz test

drivers/infiniband/sw/rxe/rxe_net.c +6 -1
index 50a2cb5405e2..92847e955ca2 100644
@@ -642,6 +642,8 @@ static void rxe_sock_put(struct sock *sk,
 	}
 }
 
+static DEFINE_MUTEX(rxe_net_del_mutex);
+
 void rxe_net_del(struct ib_device *dev)
 {
 	struct rxe_dev *rxe = container_of(dev, struct rxe_dev, ib_dev);
@@ -649,9 +651,10 @@ void rxe_net_del(struct ib_device *dev)
 	struct sock *sk;
 	struct net *net;
 
+	mutex_lock(&rxe_net_del_mutex);
 	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
 	if (!ndev)
-		return;
+		goto out;
 
 	net = dev_net(ndev);
 
@@ -664,6 +667,8 @@ void rxe_net_del(struct ib_device *dev)
 		rxe_sock_put(sk, rxe_ns_pernet_set_sk6, net);
 
 	dev_put(ndev);
+out:
+	mutex_unlock(&rxe_net_del_mutex);
 }
 
 static void rxe_port_event(struct rxe_dev *rxe,
author syzbot
#39 lore raw
syzbot has tested the proposed patch but the reproducer is still triggering an issue: lost connection to test machine

Hello,

syzbot has tested the proposed patch but the reproducer is still triggering an issue:
lost connection to test machine

Tested on:

commit:         e98d21c1 Add linux-next specific files for 20260508
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=17fa86d2580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=59b98218d9b2edf4
dashboard link: https://syzkaller.appspot.com/bug?extid=d8f76778263ab65c2b21
compiler:       Debian clang version 21.1.8 (++20251221033036+2078da43e25a-1~exp1~20251221153213.50), Debian LLD 21.1.8
patch:          https://syzkaller.appspot.com/x/patch.diff?x=105f3fce580000
author syzbot
#40 lore raw

Forwarded: Re: [syzbot] [rdma] general protection fault in kernel_sock_shutdown (4)

In reply to syzbot
For archival purposes, forwarding an incoming command email to [email protected].

For archival purposes, forwarding an incoming command email to [email protected].

***

Subject: Re: [syzbot] [rdma] general protection fault in kernel_sock_shutdown (4)
Author: [email protected]

#syz test: https://github.com/zhuyj/linux
null-ptr-deref_kernel_sock_shutdown

author syzbot
#41 lore raw

Forwarded: Re: [syzbot] [rdma] general protection fault in kernel_sock_shutdown (4)

In reply to syzbot
For archival purposes, forwarding an incoming command email to [email protected].

For archival purposes, forwarding an incoming command email to [email protected].

***

Subject: Re: [syzbot] [rdma] general protection fault in kernel_sock_shutdown (4)
Author: [email protected]

#syz test: https://github.com/zhuyj/linux
null-ptr-deref_kernel_sock_shutdown

author syzbot
#42 lore raw

Forwarded: Re: [syzbot] [rdma] general protection fault in kernel_sock_shutdown (4)

In reply to syzbot
For archival purposes, forwarding an incoming command email to [email protected].

For archival purposes, forwarding an incoming command email to [email protected].

***

Subject: Re: [syzbot] [rdma] general protection fault in kernel_sock_shutdown (4)
Author: [email protected]

#syz test: https://github.com/zhuyj/linux
null-ptr-deref_kernel_sock_shutdown