Re: [PATCH bpf 2/2] veth: update mem type in xdp_buff

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Jakub Kicinski <kuba@kernel.org> writes:

> On Wed, 8 Oct 2025 12:37:22 +0200 Maciej Fijalkowski wrote:
>> > > I guess we're slipping into a philosophical discussion but I'd say 
>> > > that the problem is that rxq stores part of what is de facto xdp buff
>> > > state. It is evacuated into the xdp frame when frame is constructed,
>> > > as packet is detached from driver context. We need to reconstitute it
>> > > when we convert frame (skb, or anything else) back info an xdp buff.  
>> > 
>> > So let us have mem type per xdp_buff then. Feels clunky anyways to change
>> > it on whole rxq on xdp_buff basis. Maybe then everyone will be happy?  
>> 
>> ...however would we be fine with taking a potential performance hit?
>
> I'd think the perf hit will be a blocker, supposedly it's in rxq for
> a reason. We are updating it per packet in the few places that are
> coded up correctly (cpumap) so while it is indeed kinda weird we're
> not making it any worse?
>
> Maybe others disagree. I don't feel super strongly. My gut feeling is
> what I drafted is best we can do in a fix.

I'd tend to agree, although I don't have a good intuition for how much
of a performance hit this would end up being.

-Toke