On Wed, Mar 24, 2021 at 11:36:53AM +0100, Auger Eric wrote:
> Hi Cleber,
> On 3/23/21 11:15 PM, Cleber Rosa wrote:
> > For users of the LinuxTest class, let's set up the VM with the port
> > redirection for SSH, instead of requiring each test to set the same
> > arguments.
> > 
> > Signed-off-by: Cleber Rosa <crosa@redhat.com>
> > ---
> >  tests/acceptance/avocado_qemu/__init__.py | 4 +++-
> >  tests/acceptance/virtiofs_submounts.py    | 4 ----
> >  2 files changed, 3 insertions(+), 5 deletions(-)
> > 
> > diff --git a/tests/acceptance/avocado_qemu/__init__.py b/tests/acceptance/avocado_qemu/__init__.py
> > index 67f75f66e5..e75b002c70 100644
> > --- a/tests/acceptance/avocado_qemu/__init__.py
> > +++ b/tests/acceptance/avocado_qemu/__init__.py
> > @@ -309,10 +309,12 @@ class LinuxTest(Test, LinuxSSHMixIn):
> >      timeout = 900
> >      chksum = None
> >  
> > -    def setUp(self, ssh_pubkey=None):
> > +    def setUp(self, ssh_pubkey=None, network_device_type='virtio-net'):
> I would be interested in testing with HW bridging too, when a bridge is
> available. Do you think we could have the netdev configurable too?
> This would be helpful to test vhost for instance.
>

Right, I knew from the start that the user mode network would only
go so far.  TBH, I think it went too further than I expected.  But,
requiring, or supporting, other network modes can add quite a bit
of complexity, depending on how much you want the framework to do.

Anyway, this is a valid point/request.  For the lack of a better place,
and given that this may be a larger effort, I'm tracking it here:

  https://gitlab.com/cleber.gnu/qemu/-/issues/3

> With respect the network device type, I am currently working on SMMU
> test and I need to call the parent setUp-) with the following args now:
> 
> super(IOMMU, self).setUp(pubkey,
> 'virtio-net-pci,iommu_platform=on,disable-modern=off,disable-legacy=on')
> 
> It works but I am not sure you had such kind of scenario in mind?
>

I see where you're coming from, and I share the slight feeling of abusing
the feature... but I think it's OK at this point.  I mean, I believe it's
better to focus on say, the HW bridging support as this at least works.

> Thanks
> 
> Eric
>

Cheers,
- Cleber.