On 20.09.23 14:08, D. Wythe wrote:
From: "D. Wythe" <alibuda@xxxxxxxxxxxxxxxxx>
Consider the following scenarios:
smc_release
smc_close_active
write_lock_bh(&smc->clcsock->sk->sk_callback_lock);
smc->clcsock->sk->sk_user_data = NULL;
write_unlock_bh(&smc->clcsock->sk->sk_callback_lock);
smc_tcp_syn_recv_sock
smc = smc_clcsock_user_data(sk);
/* now */
/* smc == NULL */
Hence, we may read the a NULL value in smc_tcp_syn_recv_sock(). And
since we only unset sk_user_data during smc_release, it's safe to
drop the incoming tcp reqsock.
Fixes: ("net/smc: net/smc: Limit backlog connections"
Signed-off-by: D. Wythe <alibuda@xxxxxxxxxxxxxxxxx>
---
net/smc/af_smc.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index bacdd97..b4acf47 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -125,6 +125,8 @@ static struct sock *smc_tcp_syn_recv_sock(const struct sock *sk,
struct sock *child;
smc = smc_clcsock_user_data(sk);
+ if (unlikely(!smc))
+ goto drop;
if (READ_ONCE(sk->sk_ack_backlog) + atomic_read(&smc->queued_smc_hs) >
sk->sk_max_ack_backlog)
Hi D.Wythe,
this is unfortunately not sufficient for this fix. You have to make sure
that is not a life-time problem. Even so, READ_ONCE() is also needed in
this case.
Thanks,
Wenjia
