Re: [PATCH v5 02/10] target/ppc: Disconnect hflags from MSR

[Greg Kurz]

On Wed, 31 Mar 2021 15:47:26 +1100
David Gibson <david@gibson.dropbear.id.au> wrote:

> On Wed, Mar 31, 2021 at 06:04:27AM +0200, Greg Kurz wrote:
> > On Wed, 31 Mar 2021 11:09:06 +1100
> > David Gibson <david@gibson.dropbear.id.au> wrote:
> > 
> > > On Tue, Mar 30, 2021 at 08:01:13AM -0700, Richard Henderson wrote:
> > > > On 3/29/21 10:54 PM, David Gibson wrote:
> > > > >    B) Just the hflags patches from my / Richard's tree
> > > > >       https://gitlab.com/dgibson/qemu/-/pipelines/278497244
> > > > 
> > > > Look closer at this one -- it's an s390x test that's failing:
> > > > 
> > 
> > I've been seeing errors with s390x as well in CI but I couldn't
> > reproduce locally... and of course, now it seems I cannot
> > reproduce locally with ppc64abi32 either :-\
> 
> Huh.  Well supporting the idea that the issues I've seen on gitlab
> were just bad luck, I've now gotten a clean check with the hflags
> patches... bug only on my ppc-for-6.1 branch.
> 
> The ppc64 bug that Greg was seeing still makes me nervous, as does the
> failures which we saw at one point which showed that new hflags assert
> explicitly failing.
> 
> Since the hflags stuff is of moderate complexity and is a bug fix,
> it's not a regression fix.  So, I'm going to postpone that until
> ppc-for-6.1, and move ahead with this PR without it.
> 
> Richard - the remaining possible problem with the hflags stuff seems
> to manifest with the assert failing in the last patch.  However, I'm
> guess that's just exposing some more subtle problem introduced by an
> earlier patch.  Any chance you could re-order the series to insert the
> assert near the beginning, which might give us a better way of
> bisecting if this shows up again.
> 
> Greg, if this shows up again for you locally, can you please try to
> track it down.
> 

Will do.

Cheers,

--
Greg

pgp87932c0iE6.pgp
Description: OpenPGP digital signature