Re: [Qemu-devel] [PATCH v2 2/2] pci: Add INTx routing notifier

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

On Mon, 2012-07-02 at 16:31 +0200, Jan Kiszka wrote:
> On 2012-07-02 16:26, Alex Williamson wrote:
> > On Mon, 2012-07-02 at 14:38 +0200, Jan Kiszka wrote:
> >> This per-device notifier shall be triggered by any interrupt router
> >> along the path of a device's legacy interrupt signal on routing changes.
> >> For simplicity reasons and as this is a slow path anyway, no further
> >> details on the routing changes are provided. Instead, the callback is
> >> expected to use pci_device_route_intx_to_irq to check the effect of the
> >> change.
> >>
> >> Will be used by KVM PCI device assignment and VFIO.
> >>
> >> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> >> ---
> >>  hw/pci.c      |   23 +++++++++++++++++++++++
> >>  hw/pci.h      |    7 +++++++
> >>  hw/piix_pci.c |    2 ++
> >>  3 files changed, 32 insertions(+), 0 deletions(-)
> >>
> >> diff --git a/hw/pci.c b/hw/pci.c
> >> index 311ba09..772141a 100644
> >> --- a/hw/pci.c
> >> +++ b/hw/pci.c
> >> @@ -1095,6 +1095,29 @@ PCIINTxRoute pci_device_route_intx_to_irq(PCIDevice *dev, int pin)
> >>      return bus->route_intx_to_irq(bus->irq_opaque, pin);
> >>  }
> >>  
> >> +void pci_bus_fire_intx_routing_notifier(PCIBus *bus)
> >> +{
> >> +    PCIDevice *dev;
> >> +    PCIBus *sec;
> >> +    int i;
> >> +
> >> +    for (i = 0; i < ARRAY_SIZE(bus->devices); ++i) {
> >> +        dev = bus->devices[i];
> >> +        if (dev && dev->intx_routing_notifier) {
> >> +            dev->intx_routing_notifier(dev);
> >> +        }
> >> +        QLIST_FOREACH(sec, &bus->child, sibling) {
> >> +            pci_bus_fire_intx_routing_notifier(sec);
> >> +        }
> >> +    }
> >> +}
> >> +
> >> +void pci_device_set_intx_routing_notifier(PCIDevice *dev,
> >> +                                          PCIINTxRoutingNotifier notifier)
> >> +{
> >> +    dev->intx_routing_notifier = notifier;
> >> +}
> >> +
> > 
> > nit, wish there was also an unset here too.  If we switch to MSI mode,
> > there's not much point in calling the notifier, so the driver could
> > unset it.
> 
> We could add
> 
> static inline pci_device_unset_intx_routing_notifier(PCIDevice *dev)
> {
>     dev->intx_routing_notifier = NULL;
> }
> 
> - or simply do pci_device_set_intx_routing_notifier(dev, NULL);.

Duh... sorry, too early yet for me ;)  This is fine.

Acked-by: Alex Williamson <alex.williamson@redhat.com>

> >  Also nice to have a set in the initfn and matching unset in
> > exitfn.  There's potentially a race between the driver freeing data and
> > the pci device going away, but I don't know if we can hit it in qemu.
> 
> Everything should be synchronized by the BQL, so far. Or not?

Yes, I imagine the lock protects getting called in that gap.  Thanks,

Alex