RE: [PATCH net-next] net: fec: add XDP_TX feature support

[Wei Fang]

> -----Original Message-----
> From: Jakub Kicinski <kuba@xxxxxxxxxx>
> Sent: 2023年7月20日 23:25
> To: Wei Fang <wei.fang@xxxxxxx>
> Cc: davem@xxxxxxxxxxxxx; edumazet@xxxxxxxxxx; pabeni@xxxxxxxxxx;
> ast@xxxxxxxxxx; daniel@xxxxxxxxxxxxx; hawk@xxxxxxxxxx;
> john.fastabend@xxxxxxxxx; Clark Wang <xiaoning.wang@xxxxxxx>; Shenwei
> Wang <shenwei.wang@xxxxxxx>; netdev@xxxxxxxxxxxxxxx; dl-linux-imx
> <linux-imx@xxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx; bpf@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH net-next] net: fec: add XDP_TX feature support
> 
> On Thu, 20 Jul 2023 07:06:05 +0000 Wei Fang wrote:
> > > Are you taking budget into account? When NAPI is called with budget
> > > of 0 we are *not* in napi / softirq context. You can't be processing
> > > any XDP tx under such conditions (it may be a netpoll call from IRQ
> context).
> >
> > Actually, the fec driver never takes the budget into account for
> > cleaning up tx BD ring. The budget is only valid for rx.
> 
> I know, that's what I'm complaining about. XDP can only run in normal NAPI
> context, i.e. when NAPI is called with budget != 0. That works out without any
> changes on Rx, if budget is zero drivers already don't process Rx. But similar
> change must be done on Tx when adding XDP support. You can still process all
> normal skb packets on Tx when budget is 0 (in fact you should), but you
> _can't_ process any XDP Tx frame.
Sorry, I did not realize that we can not process any tx XDP packet if the "budget"
is 0. I noticed your latest clarification [1] in napi.rst, I believe it will help many
people avoid this problem like me. Thank you very much.
[1]: https://lore.kernel.org/netdev/20230720161323.2025379-1-kuba@xxxxxxxxxx/T/