Re: [PATCH net-next v1] net: stmmac: Enable TSO on VLANs

[Furong Xu <0x1207@gmail.com>]

On Thu, 13 Jun 2024 11:45:45 +0100
"Russell King (Oracle)" <linux@armlinux.org.uk> wrote:

> On Thu, Jun 13, 2024 at 10:38:08AM +0800, Furong Xu wrote:
> > @@ -4239,16 +4239,32 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
> >  	struct stmmac_txq_stats *txq_stats;
> >  	int tmp_pay_len = 0, first_tx;
> >  	struct stmmac_tx_queue *tx_q;
> > -	bool has_vlan, set_ic;
> > +	bool set_ic;
> >  	u8 proto_hdr_len, hdr;
> >  	u32 pay_len, mss;
> >  	dma_addr_t des;
> >  	int i;
> > +	struct vlan_ethhdr *veth;
> >  
> >  	tx_q = &priv->dma_conf.tx_queue[queue];
> >  	txq_stats = &priv->xstats.txq_stats[queue];
> >  	first_tx = tx_q->cur_tx;
> >  
> > +	if (skb_vlan_tag_present(skb)) {
> > +		/* Always insert VLAN tag to SKB payload for TSO frames.
> > +		 *
> > +		 * Never insert VLAN tag by HW, since segments splited by
> > +		 * TSO engine will be un-tagged by mistake.
> > +		 */
> > +		skb_push(skb, VLAN_HLEN);
> > +		memmove(skb->data, skb->data + VLAN_HLEN, ETH_ALEN * 2);
> > +
> > +		veth = skb_vlan_eth_hdr(skb);
> > +		veth->h_vlan_proto = skb->vlan_proto;
> > +		veth->h_vlan_TCI = htons(skb_vlan_tag_get(skb));
> > +		__vlan_hwaccel_clear_tag(skb);
> > +	}  
> 
> I think drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c::
> otx2_sq_append_skb() does something similar, but uses a helper
> instead:
> 
>         if (skb_shinfo(skb)->gso_size && !is_hw_tso_supported(pfvf, skb)) {
>                 /* Insert vlan tag before giving pkt to tso */
>                 if (skb_vlan_tag_present(skb))
>                         skb = __vlan_hwaccel_push_inside(skb);
>                 otx2_sq_append_tso(pfvf, sq, skb, qidx);
>                 return true;
>         }
> 
> Maybe __vlan_hwaccel_push_inside() should be used here?
> 

Yes, it should. Thanks for your comments.
I will send a new patch.