On Mon, Aug 26, 2013 at 09:50:38PM +0200, folkert wrote:
> > 2. Get the system into the bad state and then do some deeper. Start
> > with outgoing ping, instrument guest driver and host vhost_net
> > functions to see what the drivers are doing, inspect the transmit
> > vring, etc.
Okay, it's time to understand what is going on at the virtio-net level.
You can check whether the host is being notified of packets ready to be
transmitted:
(host)$ sudo perf probe --module vhost_net handle_tx
(host)$ sudo perf record --pid 24306 -e probe:handle_tx -R
[...attempt to transmit packets inside guest...]
^C
(host)$ sudo perf script
vhost-24305 24306 [000] 101120.817103: probe:handle_tx: (ffffffffa06fd6b0)
Here the qemu-kvm process was pid 24305 and the vhost kernel thread was
24306 (you can find out the correct pid to use using ps aux). If your
host is only running the guest under test you can drop the pid argument
and replace it with -a.
If you see handle_tx probes firing on the host then the issue may be
with the vring or tap device. If you do not see handle_tx probes firing
on the host, then the problem may be inside the guest.
Stefan
--
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
