Re: Warnings during the virtio-net-failover test

[Laurent Vivier]

On 20/12/2021 11:16, Thomas Huth wrote:
>  Hi Laurent,
>
> today I noticed that there are quite a bunch of warnings during the virtio-net-failover 
> test:
>
> $ QTEST_QEMU_BINARY=./qemu-system-x86_64 tests/qtest/virtio-net-failover
> /x86_64/failover-virtio-net/hotplug_1: qemu-system-x86_64: warning: Primary device not 
> found
> Virtio-net failover will not work. Make sure primary device has parameter 
> failover_pair_id=standby0
> OK
> /x86_64/failover-virtio-net/hotplug_1_reverse: OK
> /x86_64/failover-virtio-net/hotplug_2: qemu-system-x86_64: warning: Primary device not 
> found
> Virtio-net failover will not work. Make sure primary device has parameter 
> failover_pair_id=standby0
> OK
> [...]
> /x86_64/failover-virtio-net/migrate/out: qemu-system-x86_64: warning: Primary device not 
> found
> Virtio-net failover will not work. Make sure primary device has parameter 
> failover_pair_id=standby0
> OK
> /x86_64/failover-virtio-net/migrate/in: OK
> /x86_64/failover-virtio-net/migrate/abort/wait-unplug: qemu-system-x86_64: warning: 
> Primary device not found
> Virtio-net failover will not work. Make sure primary device has parameter 
> failover_pair_id=standby0
> OK
> /x86_64/failover-virtio-net/migrate/abort/active: qemu-system-x86_64: warning: Primary 
> device not found
> Virtio-net failover will not work. Make sure primary device has parameter 
> failover_pair_id=standby0
> OK
> /x86_64/failover-virtio-net/multi/out: OK
> /x86_64/failover-virtio-net/multi/in: OK
>
> I assume they are false positives? If so, could you please come up with a patch to 
> silence them, since they are quite confusing...?
In fact they are the results of the test itself as we try to plug the standby 
device first.

They are produced by a error_setg(), but as the caller is not attached to a user interface 
the errp is used with warn_report_err().

In the end it uses error_vprintf() that seems that could be silence using 
QTEST_SILENT_ERRORS environment variable.

I'm going to write a patch using that.

Thanks,

Laurent