Re: [Qemu-devel] vfio in the guest: no available reset mechanism

[Michael S. Tsirkin]

On Sat, Aug 02, 2014 at 07:54:30AM +0200, Jan Kiszka wrote:
> On 2014-08-01 19:16, Alex Williamson wrote:
> >>  Also, it may let some of our device
> >> models deviate from their real versions (I suppose, e.g., none of the
> >> e1000 devices we currently emulate exposed FLR).
> > 
> > Of course, but what are the chances that the driver will care?
> 
> No drivers of GPOSes, but special or legacy OSes may do so. Keep in mind
> that Intel e.g. is documenting their PCI devices with fixed config space
> addresses for their capability.
> 
> Also, we completely lack PM caps so far. Adding them would already have
> the value of increasing emulation accuracy. And here I think we are free
> to always implement reset behavior behind D3->D0 transitions. So my
> believe is that this will be the better path.
> 
> Jan
> 
> 


That's true but is PM reset per-function or per-device?
I'll have to check the spec - do you happen to rememeber?