Re: [Qemu-devel] [TRIVIAL][PATCH v2] libqos virtio: Increase ISR timeout

[Marc Marí]

El Mon, 15 Sep 2014 14:47:41 -0700
Peter Maydell <address@hidden> escribió:
> On 11 September 2014 02:40, Marc Marí <address@hidden>
> wrote:
> > Increase the clock step to avoid Travis failure in some builds due
> > to overagressive timeout.
> >
> > Signed-off-by: Marc Marí <address@hidden>
> 
> Unfortunately Travis is still failing with this:
> 
> ERROR:tests/virtio-blk-test.c:209:pci_basic: assertion failed:
> (qvirtio_wait_queue_isr(&qvirtio_pci, &dev->vdev, &vqpci->vq,
> QVIRTIO_BLK_TIMEOUT))
> 
> GTester: last random seed: R02S490664995052f018e72f25518e90cb81
> 
> **
> 
> ERROR:tests/virtio-blk-test.c:578:pci_idx: assertion failed:
> (qvirtio_wait_queue_isr(&qvirtio_pci, &dev->vdev, &vqpci->vq,
> QVIRTIO_BLK_TIMEOUT))
> 
> GTester: last random seed: R02S5471c60ecfff4007d0adc9b0b7c265b6
> 
> thanks
> -- PMM

This means agressive timeout is not the problem. The problem is the ISR
not being raised at all after sending a read or write request.

If this is the log of the failing case:
https://travis-ci.org/qemu/qemu/builds/35118444
(first time looking at Travis logs, not sure)

The difference between the one failing and the same arch not failing is:
brlapi support yes, linux AIO support yes, uuid support yes, licap-ng
support yes, virtio-blk-data-plane yes, vhdx yes (this configuration
fails).

The difference that is most likely to fail is virtio-blk-data-plane. So
we should start looking there.

Marc