On Fri, 1 Feb 2019, 00:23 Pete Heist <pete@heistp.net wrote:

>
> > On Feb 1, 2019, at 12:18 AM, Toke Høiland-Jørgensen <toke@redhat.com>
> wrote:
> >
> > (crossed streams; but yeah, basically the same conclusion)
>
> Ok, crossed-crossed streams references it is… :)
>
> >> Do you think this is a correct patch?
> >
> > Well, the "correct patch" would be to get upstream fixed :)
> >
> > But yeah, I don't think it would hurt to carry this as a workaround
> > patch until that happens...
> >
> > I guess you could try running it for a while and see if anything else
> > blows up? You seem to have quite a knack for getting it to fail... ;)
>
> That sounds reasonable, let me see if I can break it in some different
> cases.
>

Cool.

My guess is, past qdiscs didn’t manage to fill the tailroom space and cause
> this problem, but I’m not sure…
>

Wouldn't be the first bug we've exposed in this code:
https://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next.git/commit/?id=d5a672ac9f48f81b20b1cad1d9ed7bbf4e418d4c

-Toke

>