net: diag: Don't double-free TCP_NEW_SYN_RECV sockets in tcp_abort
authorLorenzo Colitti <lorenzo@google.com>
Sat, 7 Jul 2018 07:31:40 +0000 (16:31 +0900)
committerDavid S. Miller <davem@davemloft.net>
Sun, 8 Jul 2018 01:56:10 +0000 (10:56 +0900)
When tcp_diag_destroy closes a TCP_NEW_SYN_RECV socket, it first
frees it by calling inet_csk_reqsk_queue_drop_and_and_put in
tcp_abort, and then frees it again by calling sock_gen_put.

Since tcp_abort only has one caller, and all the other codepaths
in tcp_abort don't free the socket, just remove the free in that
function.

Cc: David Ahern <dsa@cumulusnetworks.com>
Tested: passes Android sock_diag_test.py, which exercises this codepath
Fixes: d7226c7a4dd1 ("net: diag: Fix refcnt leak in error path destroying socket")
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsa@cumulusnetworks.com>
Tested-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp.c

index e7b53d2..c959bb6 100644 (file)
@@ -3720,8 +3720,7 @@ int tcp_abort(struct sock *sk, int err)
                        struct request_sock *req = inet_reqsk(sk);
 
                        local_bh_disable();
-                       inet_csk_reqsk_queue_drop_and_put(req->rsk_listener,
-                                                         req);
+                       inet_csk_reqsk_queue_drop(req->rsk_listener, req);
                        local_bh_enable();
                        return 0;
                }