tcp: Fix data-races around sysctl_max_syn_backlog.

While reading sysctl_max_syn_backlog, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index de4ccd1..d451248 100644 (file)
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -6956,10 +6956,12 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops,
                tcp_rsk(req)->ts_off = af_ops->init_ts_off(net, skb);
 
        if (!want_cookie && !isn) {
+               int max_syn_backlog = READ_ONCE(net->ipv4.sysctl_max_syn_backlog);
+
                /* Kill the following clause, if you dislike this way. */
                if (!syncookies &&
-                   (net->ipv4.sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
-                    (net->ipv4.sysctl_max_syn_backlog >> 2)) &&
+                   (max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
+                    (max_syn_backlog >> 2)) &&
                    !tcp_peer_is_proven(req, dst)) {
                        /* Without syncookies last quarter of
                         * backlog is filled with destinations,