On Mon, Jun 22, 2020 at 5:02 AM Lorenzo Bianconi <lore...@kernel.org> wrote:
>
> > On Fri, Jun 19, 2020 at 4:48 AM Lorenzo Bianconi <lore...@kernel.org> wrote:
> > >
> > > ovs connection tracking module performs de-fragmentation on incoming
> > > fragmented traffic. Take info account if traffic has been de-fragmented
> > > in execute_check_pkt_len action otherwise we will perform the wrong
> > > nested action considering the original packet size. This issue typically
> > > occurs if ovs-vswitchd adds a rule in the pipeline that requires
> > > connection
> > > tracking (e.g. OVN stateful ACLs) before execute_check_pkt_len action.
> > >
> > > Fixes: 4d5ec89fc8d14 ("net: openvswitch: Add a new action check_pkt_len")
> > > Signed-off-by: Lorenzo Bianconi <lore...@kernel.org>
> > > ---
> > > net/openvswitch/actions.c | 6 ++++--
> > > 1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/net/openvswitch/actions.c b/net/openvswitch/actions.c
> > > index fc0efd8833c8..9f4dd64e53bb 100644
> > > --- a/net/openvswitch/actions.c
> > > +++ b/net/openvswitch/actions.c
> > > @@ -1169,9 +1169,10 @@ static int execute_check_pkt_len(struct datapath
> > > *dp, struct sk_buff *skb,
> > > struct sw_flow_key *key,
> > > const struct nlattr *attr, bool last)
> > > {
> > > + struct ovs_skb_cb *ovs_cb = OVS_CB(skb);
> > > const struct nlattr *actions, *cpl_arg;
> > > const struct check_pkt_len_arg *arg;
> > > - int rem = nla_len(attr);
> > > + int len, rem = nla_len(attr);
> > > bool clone_flow_key;
> > >
> > > /* The first netlink attribute in 'attr' is always
> > > @@ -1180,7 +1181,8 @@ static int execute_check_pkt_len(struct datapath
> > > *dp, struct sk_buff *skb,
> > > cpl_arg = nla_data(attr);
> > > arg = nla_data(cpl_arg);
> > >
> > > - if (skb->len <= arg->pkt_len) {
> > > + len = ovs_cb->mru ? ovs_cb->mru : skb->len;
> > > + if (len <= arg->pkt_len) {
> >
> > We could also check for the segmented packet and use segment length
> > for this check.
>
> Hi Pravin,
>
> thx for review.
> By 'segmented packet' and 'segment length', do you mean 'fragment' and
> 'fragment length'?
I am actually talking about GSO packets.
Thanks.
> If so I guess we can't retrieve the original fragment length if we exec
> OVS_ACTION_ATTR_CT before OVS_ACTION_ATTR_CHECK_PKT_LEN (e.g if we have a
> stateful ACL in the ingress pipeline) since handle_fragments() will
> reconstruct
> the whole IP datagram and it will store frag_max_size in OVS_CB(skb)->mru.
> Am I missing something?
>
> Regards,
> Lorenzo
>
