Re: [linus:master] [tcp_cubic] 25c1a9ca53: packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k-idle-restart_ipv4-mapped-v6.fail

From: Jason Xing
Date: Mon Jan 27 2025 - 18:31:50 EST


On Mon, Jan 27, 2025 at 11:03 PM Neal Cardwell <ncardwell@xxxxxxxxxx> wrote:
>
> On Mon, Jan 27, 2025 at 5:20 AM Jason Xing <kerneljasonxing@xxxxxxxxx> wrote:
> >
> > On Sun, Jan 26, 2025 at 4:49 PM Jason Xing <kerneljasonxing@xxxxxxxxx> wrote:
> > >
> > > On Sun, Jan 26, 2025 at 2:30 PM kernel test robot <oliver.sang@xxxxxxxxx> wrote:
> > > >
> > > >
> > > >
> > > > Hello,
> > > >
> > > > kernel test robot noticed "packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k-idle-restart_ipv4-mapped-v6.fail" on:
> > > >
> > > > (
> > > > in fact, there are other failed cases which can pass on parent:
> > > >
> > > > 4395a44acb15850e 25c1a9ca53db5780757e7f53e68
> > > > ---------------- ---------------------------
> > > > fail:runs %reproduction fail:runs
> > > > | | |
> > > > :6 100% 6:6 packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k-idle-restart_ipv4-mapped-v6.fail
> > > > :6 100% 6:6 packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k-idle-restart_ipv4.fail
> > > > :6 100% 6:6 packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k-idle-restart_ipv6.fail
> > > > :6 100% 6:6 packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k_ipv4-mapped-v6.fail
> > > > :6 100% 6:6 packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k_ipv4.fail
> > > > :6 100% 6:6 packetdrill.packetdrill/gtests/net/tcp/cubic/cubic-bulk-166k_ipv6.fail
> > >
> > > Thanks for the report. I remembered that Mahdi once modified/adjusted
> > > some of them, please see the link[1].
> > >
> > > [1]: https://lore.kernel.org/all/223960459.607292.1737102176209@xxxxxxxxxxxxxx/
> > >
> > > I think we're supposed to update them altogether?
> >
> > Should the updated pkt scripts target net or net-next tree, BTW?
>
> Those packetdrill scripts are not in the kernel tree yet in "net" or
> "net-next". I suspect they are being pulled from
> https://github.com/google/packetdrill ...
>
> I will try to update those packetdrill scripts ASAP, using Mahdi's
> updates as a starting point.

Thanks, Neal. I can also contribute to it if needed as I've already
started to implement hystart++ one or two days ago (after
communicating with Mahdi Arghavani).

>
> It seems that perhaps we should migrate all the Linux packetdrill
> scripts to the Linux source tree, so we are not in this confusing
> state where some tests are in https://github.com/google/packetdrill
> and some are in the Linux source tree...

Agreed.

Thanks,
Jason