Re: [PATCH RFC net-next] virtio_net: Relax queue requirement for using XDP

From: Magnus Karlsson <magnus.karlsson@gmail.com>
To: "Toke Høiland-Jørgensen" <toke@redhat.com>
Cc: David Ahern <dahern@digitalocean.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	David Ahern <dsahern@gmail.com>, Jason Wang <jasowang@redhat.com>,
	David Ahern <dsahern@kernel.org>,
	Network Development <netdev@vger.kernel.org>,
	"David S. Miller" <davem@davemloft.net>,
	kuba@kernel.org, Jesper Dangaard Brouer <brouer@redhat.com>,
	Magnus Karlsson <magnus.karlsson@intel.com>
Subject: Re: [PATCH RFC net-next] virtio_net: Relax queue requirement for using XDP
Date: Thu, 27 Feb 2020 11:41:23 +0100	[thread overview]
Message-ID: <CAJ8uoz2++_D_XFwUjFri0HmNaNWKtiPNrJr=Fvc8grj-8hRzfg@mail.gmail.com> (raw)
In-Reply-To: <87wo88wcwi.fsf@toke.dk>

On Thu, Feb 27, 2020 at 11:22 AM Toke Høiland-Jørgensen <toke@redhat.com> wrote:
>
> David Ahern <dahern@digitalocean.com> writes:
>
> > On 2/26/20 1:34 AM, Toke Høiland-Jørgensen wrote:
> >>> OK so basically there would be commands to configure which TX queue is
> >>> used by XDP. With enough resources default is to use dedicated queues.
> >>> With not enough resources default is to fail binding xdp program
> >>> unless queues are specified. Does this sound reasonable?
> >> Yeah, that was the idea. See this talk from LPC last year for more
> >> details: https://linuxplumbersconf.org/event/4/contributions/462/
> >>
> >
> > Can we use one of the breakout timeslots at netdevconf and discuss this
> > proposal and status?
>
> Adding in Magnus and Jesper; I won't be at netdevconf, sadly, but you
> guys go ahead :)
>
> Magnus indicated he may have some preliminary code to share soonish.
> Maybe that means before netdevconf? ;)

I will unfortunately be after Netdevconf due to other commitments. The
plan is to send out the RFC to the co-authors of the Plumbers
presentation first, just to check the sanity of it. And after that
send it to the mailing list. Note that I have taken two shortcuts in
the RFC to be able to make quicker progress. The first on is the
driver implementation of the dynamic queue allocation and
de-allocation. It just does this within a statically pre-allocated set
of queues. The second is that the user space interface is just a
setsockopt instead of a rtnetlink interface. Again, just to save some
time in this initial phase. The information communicated in the
interface is the same though. In the current code, the queue manager
can handle the queues of the networking stack, the XDP_TX queues and
queues allocated by user space and used for AF_XDP. Other uses from
user space is not covered due to my setsockopt shortcut. Hopefully
though, this should be enough for an initial assessment.

/Magnus

> -Toke
>r