On Mon, Jul 27, 2020 at 11:38 AM Dexuan Cui <decui@xxxxxxxxxxxxx> wrote:
>
> > From: Greg KH <greg@xxxxxxxxx>
> > Sent: Friday, July 24, 2020 10:59 PM
> > > [...]
> > > Eric Dumazet made an alternative that performs the csum validation earlier:
> > >
> > > --- a/net/ipv4/udp.c
> > > +++ b/net/ipv4/udp.c
> > > @@ -1589,8 +1589,7 @@ int udp_queue_rcv_skb(struct sock *sk, struct
> > > sk_buff *skb)
> > > }
> > > }
> > >
> > > - if (rcu_access_pointer(sk->sk_filter) &&
> > > - udp_lib_checksum_complete(skb))
> > > + if (udp_lib_checksum_complete(skb))
> > > goto csum_error;
> > >
> > > if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
> > >
> > > I personally like Eric's fix and IMHO we'd better have it in v4.4 rather than
> > > trying to backport 327868212381.
> >
> > Does Eric's fix work with your testing?
>
> Yes, it worked in my testing overnight.
>
> > If so, great, can you turn it
> > into something I can apply to the 4.4.y stable tree and send it to
> > stable@xxxxxxxxxxxxxxx?
> >
> > greg k-h
>
> Will do shortly.
>
Just as a reminder, please also add the IPv6 part.
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index a8d74f44056a681ef9057c4c4abb34016120b44f..13713e0e5779b75de975faaeb4511bef40e097dc
100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -661,8 +661,7 @@ static int udpv6_queue_rcv_one_skb(struct sock
*sk, struct sk_buff *skb)
}
prefetch(&sk->sk_rmem_alloc);
- if (rcu_access_pointer(sk->sk_filter) &&
- udp_lib_checksum_complete(skb))
+ if (udp_lib_checksum_complete(skb))
goto csum_error;
if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
