Re: [PATCH] Add a new PIIX option to control PCI hot unplugging of devices on non-root buses

[Philippe Mathieu-Daudé]

On 4/20/20 11:24 AM, Daniel P. Berrangé wrote:
> On Fri, Apr 17, 2020 at 12:09:00PM -0400, Michael S. Tsirkin wrote:
> > On Fri, Apr 17, 2020 at 03:36:14PM +0000, Ani Sinha wrote:
> > > > On Apr 17, 2020, at 8:57 PM, Michael S. Tsirkin <address@hidden> wrote:
> > > >
> > > > Is there a real reason to do this? Can't we just limit the
> > > > hotplug control to pcie ports? At some point I'd like us to
> > > > start leaving piix alone..
> > >
> > > Yes we really need this feature as want to be able to hot plug devices into the 
> > > guest but prevent customers from hot unplugging them from say Windows system 
> > > tray.
> > >
> > > ani
> >
> > Problem is, I think this is not something we can support with pcie or shpc.
> > I'm reluctant to add features that only ACPI can support,
> > we are trying to phase that out.
>
>  From the upstream POV, there's been no decision / agreement to phase
> out PIIX, this is purely a RHEL downstream decision & plan. If other
> distros / users have a different POV, and find the feature useful, we
> should accept the patch if it meets the normal QEMU patch requirements.

Remotely related, this thread suggest to rename the current Frankenstein 
PIIX as 'virt_southbridge' /s
https://www.mail-archive.com/address@hidden/msg691232.html

Seriously, apparently thinking to modify PIIX scares various people.
Is there a virt equivalent that could work?
On one hand people want to use Windows guest unmodified (or with 
digitally signed drivers, which seems impossible to do with virtio 
package [*]). OTOH we are not sure how the PIIX model works because it 
is not based on specs, and does not match its datasheet, and do not want 
to modify it.

[*] If you click on 'How to Release-Sign File System Drivers' on
https://www.linux-kvm.org/page/WindowsGuestDrivers/Download_Drivers
it opens https://www.microsoft.com/whdc/driver/tips/IFS_Signing.mspx
and displays "Your request has been blocked."...

> Regards,
> Daniel