Hi,

On Tue, May 12, 2026 at 03:58:28PM +0200, Paolo Abeni wrote:
> On 5/9/26 5:04 AM, Qingfang Deng wrote:
> > +static struct sk_buff *pppoe_gso_segment(struct sk_buff *skb,
> > +					 netdev_features_t features)
> > +{
> > +	struct sk_buff *segs = ERR_PTR(-EINVAL);
> > +	struct packet_offload *ptype;
> > +	struct pppoe_hdr *phdr;
> > +	__be16 orig_type, type;
> > +	int len, nhoff;
> > +
> > +	skb_reset_network_header(skb);
> > +	nhoff = skb_network_header(skb) - skb_mac_header(skb);
> > +
> > +	if (unlikely(!pskb_may_pull(skb, PPPOE_SES_HLEN)))
> > +		goto out;
> > +
> > +	phdr = (struct pppoe_hdr *)skb_network_header(skb);
> > +	type = pppoe_hdr_proto(phdr);
> > +	ptype = gro_find_complete_by_type(type);
> > +	if (!ptype)
> > +		goto out;
> > +
> > +	orig_type = skb->protocol;
> > +	__skb_pull(skb, PPPOE_SES_HLEN);
> > +	features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
> 
> As sashiko points out here you need to use NETIF_F_GSO_SOFTWARE.

I tested this patch using NETIF_F_GSO_SOFTWARE here as Paolo suggests.

Then, I made the following changes to the flowtable (see attached
patch) to use this new pppoe_gso_segment function, and it works for
me:

Tested-by: Pablo Neira Ayuso <pablo@netfilter.org>

I will submit my flowtable patch to nf-next once this patch is applied
to net-next.

Thanks.