Re: [PATCH 1/3] memory: reintroduce BQL-free fine-grained PIO/MMIO

[Igor Mammedov <imammedo@redhat.com>]

On Tue, 24 Jun 2025 12:45:27 +0200
Igor Mammedov <imammedo@redhat.com> wrote:

> On Tue, 24 Jun 2025 09:07:11 +0200
> Gerd Hoffmann <kraxel@redhat.com> wrote:
> 
> >   Hi,
> >   
> > > Gerd mentioned this in the relevant bz:
> > > 
> > >         Note: root cause for the initrd issue noted in comment 5 is seabios
> > >         running into problems with ehci -> io errors -> corrupted initrd.
> > >         Sometimes it doesn't boot at all, probably in case the io errors
> > >         happen to hit the kernel not the initrd.
> > > 
> > > This seems to be the last piece of information we have had that is closest
> > > to the root cause.    
> > 
> > seabios used to prefer pmtimer back then for timekeeping then because it
> > has a fixed frequency.  Doing tsc calibration can easily be /way/ off in
> > a virtual machine on a loaded host.
> > 
> > Meanwhile seabios got support for reading the tsc frequency via cpuid
> > (if invtsc is available) or via kvmclock.  If that works seabios will
> > prefer the tsc for timekeeping.
> > 
> > So, when trying to reproduce the failure for analysis you have to either
> > use an old seabios version, or turn off kvmclock + invtsc support,
> > otherwise seabios will not use the pmtimer in the first place.  
> 
> thanks for the hint (I've been trying to reproduce with current seabios),
> I'll try to reproduce with the old seabios.

not exactly the same config but close 

qemu at offending commit 1beb99f787ba11 + exactly the same SeaBIOS
with guest GA RHEL7.3 (3.10.0-514.el7.x86_64) and making sure
seabios used pmtimer.

Running RHBZ 1322713 reproducer over several days in loop,
haven't been able to reproduce the issue.

Differences are: RHEL9 host and for guest a bit newer kernel.
the reproducer used a bit older (intermediate) one that is nowhere
to be found anymore.

As you've said in comment
https://bugzilla.redhat.com/show_bug.cgi?id=1322713#c6
it's strange that patch causes issues at all, especially with
'-smp 1' as in reproducer.

Also repeated with -smp x>1, it still worked fine.

Perhaps issue was elsewhere after all.

> 
> > 
> > You should have this line in the firmware log:
> > 
> >     Using pmtimer, ioport 0x608
> > 
> > HTH & take care,
> >   Gerd
> >   
>