gro: take care of DODGY packets
authorEric Dumazet <edumazet@google.com>
Fri, 6 Jan 2023 14:25:23 +0000 (14:25 +0000)
committerDavid S. Miller <davem@davemloft.net>
Mon, 9 Jan 2023 07:37:07 +0000 (07:37 +0000)
Jaroslav reported a recent throughput regression with virtio_net
caused by blamed commit.

It is unclear if DODGY GSO packets coming from user space
can be accepted by GRO engine in the future with minimal
changes, and if there is any expected gain from it.

In the meantime, make sure to detect and flush DODGY packets.

Fixes: 5eddb24901ee ("gro: add support of (hw)gro packets to gro stack")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-and-bisected-by: Jaroslav Pulchart <jaroslav.pulchart@gooddata.com>
Cc: Coco Li <lixiaoyan@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/core/gro.c

index fd8c6a7..506f83d 100644 (file)
@@ -505,8 +505,9 @@ found_ptype:
        NAPI_GRO_CB(skb)->count = 1;
        if (unlikely(skb_is_gso(skb))) {
                NAPI_GRO_CB(skb)->count = skb_shinfo(skb)->gso_segs;
-               /* Only support TCP at the moment. */
-               if (!skb_is_gso_tcp(skb))
+               /* Only support TCP and non DODGY users. */
+               if (!skb_is_gso_tcp(skb) ||
+                   (skb_shinfo(skb)->gso_type & SKB_GSO_DODGY))
                        NAPI_GRO_CB(skb)->flush = 1;
        }