[Codel] QDISC_DROP_REASON project

Dave Taht dave.taht at gmail.com
Fri Oct 28 16:34:54 EDT 2022


There were, at last count, 2600+ places where packets could be dropped
in the linux kernel, and that doesn't account for just dropping
packets at the physical layer and (I don't think) on the rx ring.
DROP_REASON support has been migrating into the kernel, but not yet
for qdiscs, and were it there, it would provide a convenient
tracepoint for why it happened, be it congestive, overflow, or self
defense.

You can pull apart the packets and see where they were going or where
they came from, and so on.

And it's kind of bad to be dropping packets for any other reason,
elsewhere in the system.

Similarly I am seeing a LOT of ecn marking in the field that I am not
sure is correct or not, and there's not a good way to track that
presently.

I had a student lined up for this, but she dropped out. I'd much
rather teach someone how to do this pretty basic job inside the kernel
than do it myself, so if you know anyone with even modest kernel
skills willing to take it on with me, I'd appreciate it.

- https://github.com/rchac/LibreQoS/issues/143

On Fri, Oct 28, 2022 at 1:12 PM Dave Taht <dave.taht at gmail.com> wrote:
>
> There is a ton of grant money going around, and various funds are
> closing at the end of this month or the next. If you know talented
> people that are being laid off, or just want to practice their craft
> in any way to make for a better internet, please pass these links
> along. If you know of any other funding sources, please post? I'd like
> to get a stable floor back under make-wifi-fast in particular, and
> find folk willing to fund or help out (and test) on the libreqos.io
> project.
>
> NLNET is eu only and has two funds focused on privacy and security.
> They are typically good for 30-50k eu and run for a year, a very short
> (3 pages or less) application flies well with them. Closing december 1st:
>
> https://nlnet.nl/assure/
> https://nlnet.nl/entrust/
>
> NLNET has been a huge supporter of bufferbloat.net over the years,
> most recently funding my "cerowrt ii" project, which was a
> constructive failure, in that I lost way too much time to dealing with
> regressions in the stack to make the slightest bit of forward
> progress.
>
> In germany there' s this:
>
> https://forum.openwrt.org/t/germanys-sovereign-tech-fund/141089
>
> And in america, ARDC is very focused on wireless applications.
>
> https://www.ampr.org/grants/
>
> The NSF POSE grants program just closed (we didn't qualify), pouring
> 21m into various open source orgs, or so I hope.
>
> --
> 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



-- 
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


More information about the Codel mailing list