[LibreQoS] Fwd: xmit_more?

[LibreQoS] Fwd: xmit_more?

[Dave Taht]

I was looking over some very old work, here: https://lwn.net/Articles/615238/

about BQL and xmit_more in the context  of the mvpp2 work going on over here:

https://forum.openwrt.org/t/add-support-for-mikrotik-rb5009ug/104391/831


Courtesy, jesper:

"I'm on PTO and traveling, so i don't have access to my bpftrace
scripts (here from the hotel lobby).

And yes I usually just "code" a oneliner with bpftrace to measure the
bulking effect.

I coded the qdisc tx bulking and added a tracepoint for this purpose.

You can also use the pktgen scripts in samples/pktgen/ 03 to burst
packets to see the xmit more effect."

--Jesper

man. 17. okt. 2022 16.40 skrev Toke Høiland-Jørgensen <toke@redhat.com>:
>
> Dave Taht <dave.taht@gmail.com> writes:
>
> > On Mon, Oct 17, 2022 at 7:00 AM Toke Høiland-Jørgensen <toke@redhat.com> wrote:
> >>
> >> Dave Taht <dave.taht@gmail.com> writes:
> >>
> >> > This is an old blog from jesper from ages back. Is xmit_more on on
> >> > most drivers nowadays? Was it subsumed by something else?
> >>
> >> Definitely still a thing. As for driver support:
> >
> > Well, marcin just added bql support to the mvpp2, but not xmit_more,
> > apparently. He's still working out
> > ways to measure the improvement better, but it did work correctly in
> > the openwrt 5.15 build.
> >
> > There a good way to see xmit_more doing magic?
>
> Jesper has a bpftrace invocation that spits out a nice histogram (for
> both BQL and xmit_more). Hopefully he's willing to share that ;)
>
> -Toke
>


-- 
This song goes out to all the folk that thought Stadia would work:
https://www.linkedin.com/posts/dtaht_the-mushroom-song-activity-6981366665607352320-FXtz
Dave Täht CEO, TekLibre, LLC