Re: [PATCH net-next v2 07/12] net: vxlan: add skb drop reasons to vxlan_rcv()

From: Menglong Dong
Date: Sun Sep 01 2024 - 09:02:29 EST


On Fri, Aug 30, 2024 at 11:04 PM Alexander Lobakin
<aleksander.lobakin@xxxxxxxxx> wrote:
>
> From: Menglong Dong <menglong8.dong@xxxxxxxxx>
> Date: Fri, 30 Aug 2024 09:59:56 +0800
>
> > Introduce skb drop reasons to the function vxlan_rcv(). Following new
> > vxlan drop reasons are added:
> >
> > VXLAN_DROP_INVALID_HDR
> > VXLAN_DROP_VNI_NOT_FOUND
> >
> > And Following core skb drop reason is added:
>
> "the following", lowercase + "the".
>

Okay!

> >
> > SKB_DROP_REASON_IP_TUNNEL_ECN
> >
> > Signed-off-by: Menglong Dong <dongml2@xxxxxxxxxxxxxxx>
>
> [...]
>
> > @@ -23,6 +25,14 @@ enum vxlan_drop_reason {
> > * one pointing to a nexthop
> > */
> > VXLAN_DROP_ENTRY_EXISTS,
> > + /**
> > + * @VXLAN_DROP_INVALID_HDR: the vxlan header is invalid, such as:
>
> Same as before, "VXLAN" in uppercase I'd say.
>
> > + * 1) the reserved fields are not zero
> > + * 2) the "I" flag is not set
> > + */
> > + VXLAN_DROP_INVALID_HDR,
> > + /** @VXLAN_DROP_VNI_NOT_FOUND: no vxlan device found for the vni */
>
> ^
>
> > + VXLAN_DROP_VNI_NOT_FOUND,
> > };
>
> [...]
>
> > if (!raw_proto) {
> > - if (!vxlan_set_mac(vxlan, vs, skb, vni))
> > + reason = vxlan_set_mac(vxlan, vs, skb, vni);
> > + if (reason)
> > goto drop;
>
> This piece must go in the previous patch, see my comment there.
>

Yeah, I'll do it.

> [...]
>
> > @@ -1814,8 +1830,9 @@ static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
> > return 0;
> >
> > drop:
> > + reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED;
>
> Is this possible that @reason will be 0 (NOT_DROPPED_YET) here? At the
> beginning of the function, it's not initialized, then each error path
> sets it to a specific value. In most paths, you check for it being != 0
> as a sign of error, so I doubt it can be 0 here.
>

It can be 0 here, as we don't set a reason for every "goto drop"
path. For example, in the line:

if (!vs)
goto drop;

we don't set a reason, and the "reason" is 0 when we "goto drop",
as I don't think that it is worth introducing a reason here.

Thanks!
Menglong Dong

> > /* Consume bad packet */
> > - kfree_skb(skb);
> > + kfree_skb_reason(skb, reason);
> > return 0;
> > }
>
> Thanks,
> Olek