Re: [PATCH net-next v2] net: pppoe: implement GRO support

[Felix Fietkau]

On 22.07.25 10:36, Paolo Abeni wrote:
> On 7/16/25 10:14 AM, Felix Fietkau wrote:
> > +static struct sk_buff *pppoe_gro_receive(struct list_head *head,
> > +					 struct sk_buff *skb)
> > +{
> > +	const struct packet_offload *ptype;
> > +	unsigned int hlen, off_pppoe;
> > +	struct sk_buff *pp = NULL;
> > +	struct pppoe_hdr *phdr;
> > +	struct sk_buff *p;
> > +	__be16 type;
> > +	int flush = 1;
>
> Minor nit: please respect the reverse christmas tree order above

Will do

> > +	off_pppoe = skb_gro_offset(skb);
> > +	hlen = off_pppoe + sizeof(*phdr) + 2;
> > +	phdr = skb_gro_header(skb, hlen, off_pppoe);
> > +	if (unlikely(!phdr))
> > +		goto out;
> > +
> > +	/* ignore packets with padding or invalid length */
> > +	if (skb_gro_len(skb) != be16_to_cpu(phdr->length) + hlen - 2)
> > +		goto out;
> > +
> > +	NAPI_GRO_CB(skb)->network_offsets[NAPI_GRO_CB(skb)->encap_mark] = hlen;
> > +
> > +	type = pppoe_hdr_proto(phdr);
> > +	if (!type)
> > +		goto out;
> > +
> > +	ptype = gro_find_receive_by_type(type);
> > +	if (!ptype)
> > +		goto out;
> > +
> > +	flush = 0;
> > +
> > +	list_for_each_entry(p, head, list) {
> > +		struct pppoe_hdr *phdr2;
> > +
> > +		if (!NAPI_GRO_CB(p)->same_flow)
> > +			continue;
> > +
> > +		phdr2 = (struct pppoe_hdr *)(p->data + off_pppoe);
> > +		if (compare_pppoe_header(phdr, phdr2))
> > +			NAPI_GRO_CB(p)->same_flow = 0;
> > +	}
> > +
> > +	skb_gro_pull(skb, sizeof(*phdr) + 2);
> > +	skb_gro_postpull_rcsum(skb, phdr, sizeof(*phdr) + 2);
> > +
> > +	pp = ptype->callbacks.gro_receive(head, skb);
>
> Here you can use INDIRECT_CALL_INET()

I did that in the initial version, but then I got reports of build 
failures with the patch:

ERROR: modpost: "inet_gro_receive" [drivers/net/ppp/pppoe.ko] undefined!
ERROR: modpost: "inet_gro_complete" [drivers/net/ppp/pppoe.ko] undefined!

Should I leave it out, or export inet_gro_receive/complete?
> > +
> > +out:
> > +	skb_gro_flush_final(skb, pp, flush);
> > +
> > +	return pp;
> > +}
> > +
> > +static int pppoe_gro_complete(struct sk_buff *skb, int nhoff)
> > +{
> > +	struct pppoe_hdr *phdr = (struct pppoe_hdr *)(skb->data + nhoff);
> > +	__be16 type = pppoe_hdr_proto(phdr);
> > +	struct packet_offload *ptype;
> > +	int err = -ENOENT;
> > +
> > +	ptype = gro_find_complete_by_type(type);
> > +	if (ptype)
> > +		err = ptype->callbacks.gro_complete(skb, nhoff +
> > +						    sizeof(*phdr) + 2);
>
> Possibly even here but it's less relevant.
>
> > +
> > +	return err;
> > +}
> > +
> > +static struct packet_offload pppoe_packet_offload __read_mostly = {
> > +	.type = cpu_to_be16(ETH_P_PPP_SES),
> > +	.priority = 10,
>
> The priority value should be IMHO greater then the exiting ones to avoid
> possible regressions on other protocols. i.e. 20 should do.
I'll fix it in v3.

Thanks,

- Felix