[PATCH 1/2] Bluetooth: RFCOMM: Hold the listener across child unlink
From: Cen Zhang
Date: Sat Nov 21 2026 - 02:03:13 EST
rfcomm_sk_state_change() saves the listener pointer, unlinks a closed
unaccepted child, and then notifies the listener. Once the child is no
longer on the accept queue, listener release can finish without waiting
for that callback:
RFCOMM callback Listener release
lock child socket
save parent
bt_accept_unlink(child)
drain the empty accept queue
drop the last listener reference
parent->sk_data_ready(parent)
Take a listener reference before unlinking and release it after the
notification. Track that reference separately from SOCK_ZAPPED so a
different state does not release a reference it never acquired. Drop
the DLC lock before either socket can be destroyed.
The excerpt below retains the headline and stack traces.
Keep the socket reference across accept-queue unlink, as commit
4bd0b274054f ("Bluetooth: fix UAF in bt_accept_dequeue()") does for
the dequeued socket.
KASAN report as below:
BUG: KASAN: slab-use-after-free in rfcomm_sk_state_change+0x3a4/0x3f0
Read of size 8 at addr ffff88810bc8b188 by task krfcommd/75
Call Trace:
<TASK>
dump_stack_lvl+0x93/0xd0
print_report+0xce/0x630
? rfcomm_sk_state_change+0x3a4/0x3f0
? srso_alias_return_thunk+0x5/0xfbef5
? __virt_addr_valid+0x20d/0x410
? rfcomm_sk_state_change+0x3a4/0x3f0
kasan_report+0xe0/0x110
? rfcomm_sk_state_change+0x3a4/0x3f0
rfcomm_sk_state_change+0x3a4/0x3f0
__rfcomm_dlc_close+0x1a6/0x6d0
rfcomm_run+0x1b7c/0x5390
? __pfx_rfcomm_run+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_woken_wake_function+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? lock_release+0xc8/0x280
? srso_alias_return_thunk+0x5/0xfbef5
? trace_hardirqs_on+0x18/0x160
? srso_alias_return_thunk+0x5/0xfbef5
? __kthread_parkme+0x175/0x220
? __pfx_rfcomm_run+0x10/0x10
kthread+0x34f/0x460
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_kthread+0x10/0x10
ret_from_fork+0x659/0x940
? __pfx_ret_from_fork+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? __switch_to+0x74f/0xf80
? __pfx_kthread+0x10/0x10
ret_from_fork_asm+0x1a/0x30
</TASK>
Allocated by task 513:
kasan_save_stack+0x33/0x60
kasan_save_track+0x14/0x30
__kasan_kmalloc+0xaa/0xb0
__kmalloc_noprof+0x2d7/0x770
sk_prot_alloc+0x13e/0x260
sk_alloc+0x37/0xb40
bt_sock_alloc+0x40/0x3b0
rfcomm_sock_alloc.constprop.0+0x56/0x290
rfcomm_sock_create+0xe6/0x170
bt_sock_create+0x171/0x330
__sock_create+0x2c4/0x730
__sys_socket+0x141/0x210
__x64_sys_socket+0x77/0xc0
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Freed by task 515:
kasan_save_stack+0x33/0x60
kasan_save_track+0x14/0x30
kasan_save_free_info+0x3b/0x60
__kasan_slab_free+0x5f/0x80
kfree+0x307/0x580
__sk_destruct+0x64f/0x790
sk_destruct+0xb3/0xd0
__sk_free+0xdd/0x370
sk_free+0x51/0x80
rfcomm_sock_kill+0x108/0x310
rfcomm_sock_release+0x122/0x1a0
__sock_release+0xb8/0x270
sock_close+0x21/0x30
__fput+0x39f/0xa60
fput_close_sync+0xff/0x200
__x64_sys_close+0x8c/0xf0
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Cen Zhang <zzzccc427@xxxxxxxxx>
---
This patch was written with assistance from Codex (OpenAI).
net/bluetooth/rfcomm/sock.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)
diff --git a/net/bluetooth/rfcomm/sock.c b/net/bluetooth/rfcomm/sock.c
index 3ec2fb4ce5804..948f0da417bdc 100644
--- a/net/bluetooth/rfcomm/sock.c
+++ b/net/bluetooth/rfcomm/sock.c
@@ -65,7 +65,8 @@ static void rfcomm_sk_data_ready(struct rfcomm_dlc *d, struct sk_buff *skb)
static void rfcomm_sk_state_change(struct rfcomm_dlc *d, int err)
__must_hold(&d->lock)
{
- struct sock *sk = d->owner, *parent;
+ struct sock *sk = d->owner, *parent, *parent_ref = NULL;
+ bool kill;
if (!sk)
return;
@@ -82,6 +83,9 @@ static void rfcomm_sk_state_change(struct rfcomm_dlc *d, int err)
parent = bt_sk(sk)->parent;
if (parent) {
if (d->state == BT_CLOSED) {
+ /* Keep the listener alive across unlink. */
+ sock_hold(parent);
+ parent_ref = parent;
sock_set_flag(sk, SOCK_ZAPPED);
bt_accept_unlink(sk);
}
@@ -95,11 +99,14 @@ static void rfcomm_sk_state_change(struct rfcomm_dlc *d, int err)
release_sock(sk);
- if (parent && sock_flag(sk, SOCK_ZAPPED)) {
- /* We have to drop DLC lock here, otherwise
- * rfcomm_sock_destruct() will dead lock. */
+ kill = parent && sock_flag(sk, SOCK_ZAPPED);
+ if (kill || parent_ref) {
+ /* Socket destruction takes a DLC lock. */
rfcomm_dlc_unlock(d);
- rfcomm_sock_kill(sk);
+ if (kill)
+ rfcomm_sock_kill(sk);
+ if (parent_ref)
+ sock_put(parent_ref);
rfcomm_dlc_lock(d);
}
}