RDS-TCP: Recover correctly from pskb_pull()/pksb_trim() failure in rds_tcp_data_recv
authorSowmini Varadhan <sowmini.varadhan@oracle.com>
Mon, 26 Oct 2015 16:46:37 +0000 (12:46 -0400)
committerDavid S. Miller <davem@davemloft.net>
Wed, 28 Oct 2015 02:46:34 +0000 (19:46 -0700)
Either of pskb_pull() or pskb_trim() may fail under low memory conditions.
If rds_tcp_data_recv() ignores such failures, the application will
receive corrupted data because the skb has not been correctly
carved to the RDS datagram size.

Avoid this by handling pskb_pull/pskb_trim failure in the same
manner as the skb_clone failure: bail out of rds_tcp_data_recv(), and
retry via the deferred call to rds_send_worker() that gets set up on
ENOMEM from rds_tcp_read_sock()

Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/rds/tcp_recv.c

index fbc5ef88bc0e692ea4cf9cb59f6163905510928e..27a992154804c685d983206b599ef7a5bf96e8af 100644 (file)
@@ -214,8 +214,15 @@ static int rds_tcp_data_recv(read_descriptor_t *desc, struct sk_buff *skb,
                        }
 
                        to_copy = min(tc->t_tinc_data_rem, left);
-                       pskb_pull(clone, offset);
-                       pskb_trim(clone, to_copy);
+                       if (!pskb_pull(clone, offset) ||
+                           pskb_trim(clone, to_copy)) {
+                               pr_warn("rds_tcp_data_recv: pull/trim failed "
+                                       "left %zu data_rem %zu skb_len %d\n",
+                                       left, tc->t_tinc_data_rem, skb->len);
+                               kfree_skb(clone);
+                               desc->error = -ENOMEM;
+                               goto out;
+                       }
                        skb_queue_tail(&tinc->ti_skb_list, clone);
 
                        rdsdebug("skb %p data %p len %d off %u to_copy %zu -> "