Kyuwon Cho <[email protected]> writes:

> ovs_fragment() only handles IPv4 and IPv6 and drops any other packet
> with a WARN_ONCE(). Since OVS_PACKET_CMD_EXECUTE takes the MRU from
> userspace, any ethertype can get there, e.g. an MPLS packet executed by
> an unprivileged user in its own user and network namespace:
>
>   Failed fragment ->ovs0: eth=8847, MRU=500, MTU=1500.
>   WARNING: net/openvswitch/actions.c:831 at ovs_fragment+0x239/0x840, CPU#1: 
> test/106
>   Call Trace:
>    do_execute_actions+0x580e/0x7380
>    ovs_execute_actions+0xd3/0x3e0
>    ovs_packet_cmd_execute+0xc21/0x13c0
>
> The packet is dropped either way, so use a rate-limited warning instead.
>
> Fixes: 7f8a436eaa2c ("openvswitch: Add conntrack action")
> Closes: 
> https://lore.kernel.org/netdev/[email protected]/
> Signed-off-by: Kyuwon Cho <[email protected]>
> ---
> Aaron, here is the patch, with the rate-limited warning you suggested.
> It was written with LLM assistance; we reviewed and tested it. With it
> applied on net, the reproducer from the report prints the message
> instead of warning.

Thanks - it makes sense to me.  We could consider pr_warn_once instead
if preserving the 'only log once' semantic is important.  I guess the
intent here was to let developers know that there was a really
misbehaving ovs-vswitchd process going on.  Given it can be triggered by
an arbitrary user on many distributions, it would hide that - so I think
net_warn_ratelimited makes sense to call instead.

As a follow on, it's worth consideration to change the mru check to only
submit packets that can be handled.  That's a bigger lift since we might
want to change that action to preserve the drop reason
(OVS_DROP_FRAG_INVALID_PROTO) when we get the 'asked to fragment a
packet that does not make sense'.  Alternatively, we could drop it
entirely and just take the PKT_TOO_BIG drop path.  After all, we know we
can't fragment it, and that the link won't forward it.  For now,

Reviewed-by: Aaron Conole <[email protected]>

>  net/openvswitch/actions.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/net/openvswitch/actions.c b/net/openvswitch/actions.c
> index 12e339cbf791..bf3f2b328b53 100644
> --- a/net/openvswitch/actions.c
> +++ b/net/openvswitch/actions.c
> @@ -828,9 +828,9 @@ static void ovs_fragment(struct net *net, struct vport 
> *vport,
>               ip6_fragment(net, skb->sk, skb, ovs_vport_output);
>               refdst_drop(orig_dst);
>       } else {
> -             WARN_ONCE(1, "Failed fragment ->%s: eth=%04x, MRU=%d, MTU=%d.",
> -                       ovs_vport_name(vport), ntohs(key->eth.type), mru,
> -                       vport->dev->mtu);
> +             net_warn_ratelimited("Failed fragment ->%s: eth=%04x, MRU=%d, 
> MTU=%d.\n",
> +                                  ovs_vport_name(vport), 
> ntohs(key->eth.type),
> +                                  mru, vport->dev->mtu);
>               reason = OVS_DROP_FRAG_INVALID_PROTO;
>               goto err;
>       }

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to