* Pekka Enberg <penberg@xxxxxxxxxx> wrote: > On Sat, May 7, 2011 at 6:01 PM, Anthony Liguori <aliguori@xxxxxxxxxx> wrote: > > On 05/07/2011 09:50 AM, Pekka Enberg wrote: > >> > >> On Sat, 2011-05-07 at 08:14 -0500, Anthony Liguori wrote: > >>> > >>> It seems like this patch is simply avoiding raising the interrupt line > >>> if the ISR has not been acknowledged yet. I don't think there's a > >>> functional issue here but I'm surprised that it's a win. There should > >>> be a very short window when the interrupt is lowered in the APIC but > >>> still not acknowledged in the ISR. > >>> > >>> You should just be saving a pretty cheap system call. I wonder if the > >>> system call is taking longer than it should.. > >> > >> The patch seems to fix a bug where the guest kernel breaks down under > >> interrupt storm and stops doing VIRTIO_PCI_QUEUE_NOTIFY. We assumed it > >> was something with our code but your comments make me wonder if there's > >> a real problem in KVM_IRQ_LINE. > > > > Stops doing it for a short period of time or entirely? > > Seems to be entirely. The test case is doing "ping -f" from host to > guest and vice versa and it takes 30-60 seconds to trigger for me. here's the condition as described by Asias: " I also found when network hangs, guest kernel refuse to give any avail buffers in rx queue to device. At that time, vq->last_used_idx equals vq->vring.used->idx in rx queue, so even with manual IRQ injection using a debug key ctrl-a-i, the network still hangs. " Thanks, Ingo -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html