Re: RFC: vfio-pci API for PCI bus/slot (hot) resets

[Alex Williamson <alex.williamson@redhat.com>]

On Sat, 2013-08-03 at 09:15 +1000, Benjamin Herrenschmidt wrote:
> On Fri, 2013-08-02 at 16:49 -0600, Bjorn Helgaas wrote:
> > [+cc linux-pci]
> > 
> > On Fri, Aug 2, 2013 at 3:28 PM, Benjamin Herrenschmidt
> > <benh@kernel.crashing.org> wrote:
> > 
> > > Right. Another use case is, I know of devices that need a fundamental
> > > reset (PERST) after applying a FW update.
> > 
> > This is a tangent from the real discussion here, but the question of
> > resetting a device after a firmware update concerns me.  Many if not
> > all of our current reset interfaces save & restore the architected
> > parts of config space around the reset.  But a reset after a firmware
> > update may change things like the number and type of BARs or even the
> > functionality of the device, so I don't think the restore is safe in
> > general.
> 
> Right.
> 
> > I doubt this is a big problem in general, but I have found reports of
> > people having to do a system reset or reboot after updating, e.g.,
> > FPGA images.  I suppose at least some of these could be worked around
> > with the right hotplug incantations.
> 
> Yes.
> 
> We have that similar issue with error handling, when the driver doesn't
> have the right hooks, we simulate an unplug, reset, then replug.
> 
> Maybe we could provide generic helpers to do that...

Devices going away and coming back is pretty difficult for vfio to
handle.  Perhaps helpers to rescan a device in-place would be easier.
On the QEMU side we'd need to rescan the device after each reset, which
would be rather tedious for the typical case where it doesn't change.
Thanks,

Alex