[PATCH net] kcm: avoid potential race in kcm_tx_work

[Eric Dumazet]

syzbot found that kcm_tx_work() could crash [1] in:

/* Primarily for SOCK_SEQPACKET sockets */
if (likely(sk->sk_socket) &&
test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
<<*>> clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
sk->sk_write_space(sk);
}

I think the reason is that another thread might concurrently
run in kcm_release() and call sock_orphan(sk) while sk is not
locked. kcm_tx_work() find sk->sk_socket being NULL.

[1]
BUG: KASAN: null-ptr-deref in instrument_atomic_write include/linux/instrumented.h:86 [inline]
BUG: KASAN: null-ptr-deref in clear_bit include/asm-generic/bitops/instrumented-atomic.h:41 [inline]
BUG: KASAN: null-ptr-deref in kcm_tx_work+0xff/0x160 net/kcm/kcmsock.c:742
Write of size 8 at addr 0000000000000008 by task kworker/u4:3/53

CPU: 0 PID: 53 Comm: kworker/u4:3 Not tainted 5.19.0-rc3-next-20220621-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: kkcmd kcm_tx_work
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
kasan_report+0xbe/0x1f0 mm/kasan/report.c:495
check_region_inline mm/kasan/generic.c:183 [inline]
kasan_check_range+0x13d/0x180 mm/kasan/generic.c:189
instrument_atomic_write include/linux/instrumented.h:86 [inline]
clear_bit include/asm-generic/bitops/instrumented-atomic.h:41 [inline]
kcm_tx_work+0xff/0x160 net/kcm/kcmsock.c:742
process_one_work+0x996/0x1610 kernel/workqueue.c:2289
worker_thread+0x665/0x1080 kernel/workqueue.c:2436
kthread+0x2e9/0x3a0 kernel/kthread.c:376
ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:302
</TASK>

Fixes: ab7ac4eb9832 ("kcm: Kernel Connection Multiplexor module")
Reported-by: syzbot <syzk...@googlegroups.com>
Signed-off-by: Eric Dumazet <edum...@google.com>
Cc: Tom Herbert <t...@herbertland.com>
---
net/kcm/kcmsock.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/kcm/kcmsock.c b/net/kcm/kcmsock.c
index 1215c863e1c410fa9ba5b9c3706152decfb3ebac..27725464ec08fe2b5f2e86202636cbc895568098 100644
--- a/net/kcm/kcmsock.c
+++ b/net/kcm/kcmsock.c
@@ -1838,10 +1838,10 @@ static int kcm_release(struct socket *sock)
kcm = kcm_sk(sk);
mux = kcm->mux;

+ lock_sock(sk);
sock_orphan(sk);
kfree_skb(kcm->seq_skb);

- lock_sock(sk);
/* Purge queue under lock to avoid race condition with tx_work trying
* to act when queue is nonempty. If tx_work runs after this point
* it will just return.
--
2.38.0.rc1.362.ged0d419d3c-goog

[Dmitry Vyukov]

On Wed, 12 Oct 2022 at 15:34, 'Eric Dumazet' via syzkaller
<syzk...@googlegroups.com> wrote:
>
> syzbot found that kcm_tx_work() could crash [1] in:
>
> /* Primarily for SOCK_SEQPACKET sockets */
> if (likely(sk->sk_socket) &&
> test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
> <<*>> clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
> sk->sk_write_space(sk);
> }
>
> I think the reason is that another thread might concurrently
> run in kcm_release() and call sock_orphan(sk) while sk is not
> locked. kcm_tx_work() find sk->sk_socket being NULL.

Does it make sense to add some lockdep annotations to sock_orphan()
and maybe some other similar functions to catch such cases earlier?

> --
> You received this message because you are subscribed to the Google Groups "syzkaller" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to syzkaller+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/syzkaller/20221012133412.519394-1-edumazet%40google.com.

[Eric Dumazet]

On Wed, Oct 12, 2022 at 7:00 AM Dmitry Vyukov <dvy...@google.com> wrote:
>
> On Wed, 12 Oct 2022 at 15:34, 'Eric Dumazet' via syzkaller
> <syzk...@googlegroups.com> wrote:
> >
> > syzbot found that kcm_tx_work() could crash [1] in:
> >
> > /* Primarily for SOCK_SEQPACKET sockets */
> > if (likely(sk->sk_socket) &&
> > test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
> > <<*>> clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
> > sk->sk_write_space(sk);
> > }
> >
> > I think the reason is that another thread might concurrently
> > run in kcm_release() and call sock_orphan(sk) while sk is not
> > locked. kcm_tx_work() find sk->sk_socket being NULL.
>
> Does it make sense to add some lockdep annotations to sock_orphan()
> and maybe some other similar functions to catch such cases earlier?

I thought about that, but this seems net-next material.

[patchwork-b...@kernel.org]

Hello:

This patch was applied to netdev/net.git (master)
by Jakub Kicinski <ku...@kernel.org>:

On Wed, 12 Oct 2022 13:34:12 +0000 you wrote:
> syzbot found that kcm_tx_work() could crash [1] in:
>
> /* Primarily for SOCK_SEQPACKET sockets */
> if (likely(sk->sk_socket) &&
> test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
> <<*>> clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
> sk->sk_write_space(sk);
> }
>

> [...]

Here is the summary with links:
- [net] kcm: avoid potential race in kcm_tx_work
https://git.kernel.org/netdev/net/c/ec7eede369fe

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html