Re: [Qemu-devel] [PATCH] virtio-pci: Clear IRQ at reset

[Michael S. Tsirkin]

On Fri, Mar 13, 2015 at 02:28:20PM +0800, Fam Zheng wrote:
> On Fri, 03/13 14:07, Fam Zheng wrote:
> > On Thu, 03/12 12:15, Michael S. Tsirkin wrote:
> > > On Thu, Mar 12, 2015 at 11:04:33AM +0000, Peter Maydell wrote:
> > > > On 12 March 2015 at 10:57, Michael S. Tsirkin <address@hidden> wrote:
> > > > > This isn't a device reset though.
> > > > > The function that Fam is touching is called
> > > > > when a special "virtio reset" register to
> > > > > poked by the driver.
> > > > > It only resets part of the device, not all of it,
> > > > > and it seems reasonable to ask that it clear the
> > > > > interrupt.
> > > > 
> > > > Oh, right, sorry. Yes, that should clear the interrupt, then.
> > > > (Is there a similar bug on other virtio transports?)
> > > > 
> > > > -- PMM
> > > 
> > > Hmm interesting.
> > > 
> > > I looked at virtio_reset and that one does:
> > >     vdev->isr = 0;
> > >     vdev->config_vector = VIRTIO_NO_VECTOR;
> > >     virtio_notify_vector(vdev, vdev->config_vector);
> > > 
> > > which in turn would call
> > > static void virtio_pci_notify(DeviceState *d, uint16_t vector)
> > > {
> > >     VirtIOPCIProxy *proxy = to_virtio_pci_proxy_fast(d);
> > > 
> > >     if (msix_enabled(&proxy->pci_dev))
> > >         msix_notify(&proxy->pci_dev, vector);
> > >     else {
> > >         VirtIODevice *vdev = virtio_bus_get_device(&proxy->bus);
> > >         pci_set_irq(&proxy->pci_dev, vdev->isr & 1);
> > >     }
> > > }
> > > 
> > > since isr is 0, and msi is disabled, it looks like
> > > pci_set_irq will get invoked automatically.
> > > 
> > > so at this point I stopped understanding how can
> > > this patch help.
> > > 
> > > Fam, does your patch actually help some guests?
> > > Could you pls investigate why isn't virtio_reset
> > > sufficient?
> > 
> > Because virtio_reset doesn't disable msix, so here
> > "msix_notify(&proxy->pci_dev, 0)" is called.
> 
> s/0/VIRTIO_NO_VECTOR/

My answer still holds - we don't need to clear msix
interrupts, and with msi enabled we know intx is low.

No?

> > 
> > I tested by applying your patch:
> > 
> > http://www.spinics.net/lists/kernel/msg1943182.html
> > 
> > and adding printf's in QEMU's virtio_reset and virtio_pci_notify.
> > 
> > It shows pci_set_irq is never called.
> > 
> > Fam
> > 
> >