Re: [dpdk-dev] [dpdk-virtio] DPDK stopped working on virtio

[Clarylin L <clearasu@gmail.com>]

I am working with DPDK 2.0.

I guess it's not DPDK code issue , but more like an environment issue (as
same code has been working fine before. It's even working on another setup
now). Someone might have accidentally changed my setup, and I want to find
out what made dpdk-virtio stop working.

Is DPDK-virtio dependent on any specific modules on the hypervisor? or are
there any configurations on the hypervisor that would impact the
functionality of  virtio? or anything else I need to look into and check?
My VM is running on Ubuntu KVM.

On Mon, Feb 22, 2016 at 7:50 PM, Yuanhan Liu <yuanhan.liu@linux.intel.com>
wrote:

> On Mon, Feb 22, 2016 at 11:15:57AM -0800, Clarylin L wrote:
> > I am running DPDK application (testpmd) within a VM based on virtio. With
> > the same hypervisor and same DPDK code, it used to work well. But it
> > stopped working since last week. The VM's port could not receive
> anything.
> > I ran tcpdump on host's physical port, bridge interface as well as vnet
> > interface, and did see packets coming in. However on VM's port there was
> > nothing.
> >
> > I ran gdb trying to debug, it hit function virtio_recv_mergeable_pkts(),
> > but nb_used=VIRTQUEUE_NUSED(rxvq) always gave 0. I guess it's because the
> > queue was empty.
> >
> > I enabled the PMD debug logging and the only thing that might be an issue
> > was the following part. Other than this I could not see any thing that
> > could indicate potential issues.
> >
> > Thu Feb 18 19:20:10 2016^@PMD: get_uio_dev(): Could not find uio resource
> > Thu Feb 18 19:20:10 2016^@PMD: virtio_resource_init_by_ioports(): PCI
> Port
> > IO found start=0xc040 with size=0x40
>
> That could be normal, when you don't bind the driver to igb_uio.
>
> > If someone can give any pointers that I should further look into, that'd
> be
> > very helpful. Appreciate your help!
>
> What's the last commit you are testing? And what are the steps
> to reproduce it? I have a quick try with vhost-switch example,
> with pkts injected by IXIA; it works fine here.
>
> Or better, mind do a git bisect? There aren't too many commits
> there. It should be a pretty fast bisect.
>
>         --yliu
>