Re: [PATCH 0/2] target/ppc: Easy parts of the POWER chiptod series

[Cédric Le Goater]

> > > When you start building the next PPC PR, I think you can also take
> > > this patch :
> > >
> > >     [4/4] target/ppc: Implement core timebase state machine and TFMR
> > >     
> > > https://patchwork.ozlabs.org/project/qemu-ppc/patch/20230603233612.125879-5-npiggin@gmail.com/
> > >
> > > It belongs to the same series.
> >
> > That doesn't apply cleanly with these 2 patches and it didn't look trivial to
> > me. As if there were some code missing.
>
> Yeah, it actually uses some signals from the nest chiptod patch.
>
> > If Nick re-send it rebased on top of ppc-next I can queue it. Since it's a
> > rebase you can keep the r-b. Thanks,
>
> I'll see how it goes, it may have to wait for next merge. SMT for
> powernv is a little more important since it's more user-facing and it
> would be nice to introduce SMT for both pseries and powernv together.

Yes. Please resend.

I gave the series a good try with an install of 23.04 on a powernv9
machine with 2*2*4 CPUs and running KVM SMP guests with libvirt.
It was performing quite well with MTTCG on a 32 CPU ryzen box. No
hangs, no crash. This is a great addition. Thanks for it.

I'll try VFIO after merge.

Cheers,

C.



 
> Chiptod for doesn't really do much except step through skiboot init
> code, and getting more useful things wired up like TB fault / HMI
> injection won't be ready before freeze.
>
> Thanks,
> Nick