Re: [PATCH v2] piix: fix regression during unplug in Xen HVM domUs

[Mark Cave-Ayland <mark.cave-ayland@xxxxxxxxxxxx>]

On 01/07/2023 10:53, Bernhard Beschow wrote:

> Am 30. Juni 2023 08:48:02 UTC schrieb Paolo Bonzini <pbonzini@xxxxxxxxxx>:
> > Il mer 28 giu 2023, 13:28 Bernhard Beschow <shentey@xxxxxxxxx> ha scritto:
> >
> > > Am 27. Juni 2023 12:07:40 UTC schrieb Olaf Hering <olaf@xxxxxxxxx>:
> > > > Tue, 27 Jun 2023 10:12:50 +0000 Bernhard Beschow <shentey@xxxxxxxxx>:
> > > >
> > > > > The BAR is a 32 bit register whose default value is 0x00000001. I think
> > > what's supposed to happen here is a pci_set_long() rather than a
> > > pci_set_byte().
> > > > Indeed, the u32 at that address changes from c121 to c101 with the
> > > current code.
> > >
> > > Neat! Would you mind sending a patch fixing the BMIBA register to be reset
> > > as 32 bit?
> > I think we should also check why writing the command register is not
> > disabling the BAR as well.
> So IIUC the BMIBA register is managed internally by QEMU's PCI code and we 
> shouldn't have to mess with the register at all. We should actually remove the 
> explicit reset of BMIBA, correct?
>
> I've tried debugging the PCI code when working on the VIA IDE controller to 
> understand it better. But despite QEMU being compiled with --enable-debug it 
> seemd to be compiled with -O2 still, making debugging quite hard. I'm not sure 
> if any compile flags leak into my build environment though.
Certainly --enable-debug normally does the right thing when building QEMU. If you 
want to double-check the compiler flags in use to see if anything from CFLAGS/LDFLAGS 
is getting picked up, use "make V=1" after configure which outputs the full command 
being used during the build rather than just the summary.

ATB,

Mark.