Re: [virtio-dev] Re: [VIRTIO PCI PATCH v5 1/1] transport-pci: Add freeze_mode to virtio_pci_common_cfg

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Sep 26, 2023 at 11:37:37AM +0000, Parav Pandit wrote:
> Hi Jiqian,
> 
> > From: Chen, Jiqian <Jiqian.Chen@amd.com>
> > Sent: Tuesday, September 26, 2023 4:30 PM
> 
> > It is not to trigger resume behavior to device by setting unfreeze in my scenario.
> > You may misunderstand my intention.
> > In current kernel and qemu code, it will trigger reset device behavior by setting
> > device_status to 0 during resuming, and I found some operations of reset is not
> > reasonable during resuming, like virtio_gpu_reset, it will destroy render
> > resources and then caused the display gone after guest finishing reusming.
> 
> When suspend is done, resume must not be done.
> I donât see a need to mix these operations from guest specially when suspend is done.
> 
> Did I miss your response for my suggestion few days ago in [1] for not mixing reset with suspend?
> 
> When new feature bit of suspend is offered, guest will _not_ do reset, hence no need to mix reset with suspend.
> 
> [1] https://lists.oasis-open.org/archives/virtio-comment/202309/msg00260.html

I frankly don't understand what you proposed there.

reset is a simple operation, it just erases all virtio
state from the device. It is very commonly is performed at
OS/driver startup. Making it not reset some virtio state
is a bad idea since suddenly you kexec a guest
that attempts a reset and device is stuck in some weird state.

-- 
MST