Subject: [PATCH] af_iucv: drop skbs rejected by filter
authorUrsula Braun <ubraun@linux.vnet.ibm.com>
Fri, 7 Oct 2016 13:51:48 +0000 (15:51 +0200)
committerDavid S. Miller <davem@davemloft.net>
Wed, 12 Oct 2016 05:56:04 +0000 (01:56 -0400)
A packet filter might be installed for instance with setsockopt
SO_ATTACH_FILTER. af_iucv currently queues skbs rejected by filter
into the backlog queue. This does not make sense, since packets
rejected by filter can be dropped immediately. This patch adds
separate sk_filter return code checking, and dropping of packets
if applicable.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Reported-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/iucv/af_iucv.c

index 88d0eaa18749213b7152ca71dc6b6a18a5a83dfc..cfb9e5f4e28f52c699072eaf41f12f3a6e667160 100644 (file)
@@ -1323,8 +1323,13 @@ static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
        }
 
        IUCV_SKB_CB(skb)->offset = 0;
-       if (sock_queue_rcv_skb(sk, skb))
-               skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
+       if (sk_filter(sk, skb)) {
+               atomic_inc(&sk->sk_drops);      /* skb rejected by filter */
+               kfree_skb(skb);
+               return;
+       }
+       if (__sock_queue_rcv_skb(sk, skb))      /* handle rcv queue full */
+               skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
 }
 
 /* iucv_process_message_q() - Process outstanding IUCV messages
@@ -1438,13 +1443,13 @@ static int iucv_sock_recvmsg(struct socket *sock, struct msghdr *msg,
                rskb = skb_dequeue(&iucv->backlog_skb_q);
                while (rskb) {
                        IUCV_SKB_CB(rskb)->offset = 0;
-                       if (sock_queue_rcv_skb(sk, rskb)) {
+                       if (__sock_queue_rcv_skb(sk, rskb)) {
+                               /* handle rcv queue full */
                                skb_queue_head(&iucv->backlog_skb_q,
                                                rskb);
                                break;
-                       } else {
-                               rskb = skb_dequeue(&iucv->backlog_skb_q);
                        }
+                       rskb = skb_dequeue(&iucv->backlog_skb_q);
                }
                if (skb_queue_empty(&iucv->backlog_skb_q)) {
                        if (!list_empty(&iucv->message_q.list))
@@ -2124,12 +2129,17 @@ static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
        skb_reset_transport_header(skb);
        skb_reset_network_header(skb);
        IUCV_SKB_CB(skb)->offset = 0;
+       if (sk_filter(sk, skb)) {
+               atomic_inc(&sk->sk_drops);      /* skb rejected by filter */
+               kfree_skb(skb);
+               return NET_RX_SUCCESS;
+       }
+
        spin_lock(&iucv->message_q.lock);
        if (skb_queue_empty(&iucv->backlog_skb_q)) {
-               if (sock_queue_rcv_skb(sk, skb)) {
+               if (__sock_queue_rcv_skb(sk, skb))
                        /* handle rcv queue full */
                        skb_queue_tail(&iucv->backlog_skb_q, skb);
-               }
        } else
                skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
        spin_unlock(&iucv->message_q.lock);