Re: [RFC PATCH net-next] selftests: forwarding: add a test for MAC Merge layer

[Vladimir Oltean]

Hi Petr,

On Tue, Feb 14, 2023 at 12:53:06PM +0100, Petr Machata wrote:
> 
> Vladimir Oltean <vladimir.oltean@xxxxxxx> writes:
> 
> > Hi Petr,
> >
> > On Mon, Feb 13, 2023 at 11:51:37AM +0100, Petr Machata wrote:
> >> Vladimir Oltean <vladimir.oltean@xxxxxxx> writes:
> >> 
> >> > +# Borrowed from mlxsw/qos_lib.sh, message adapted.
> >> > +bail_on_lldpad()
> >> > +{
> >> > +	if systemctl is-active --quiet lldpad; then
> >> > +		cat >/dev/stderr <<-EOF
> >> > +		WARNING: lldpad is running
> >> > +
> >> > +			lldpad will likely autoconfigure the MAC Merge layer,
> >> > +			while this test will configure it manually. One of them
> >> > +			is arbitrarily going to overwrite the other. That will
> >> > +			cause spurious failures (or, unlikely, passes) of this
> >> > +			test.
> >> > +		EOF
> >> > +		exit 1
> >> > +	fi
> >> > +}
> >> 
> >> This would be good to have unified. Can you make the function reusable,
> >> with a generic or parametrized message? I should be able to carve a bit
> >> of time later to move it to lib.sh, migrate the mlxsw selftests, and
> >> drop the qos_lib.sh copy.
> >
> > Maybe like this?
> 
> Yes, for most of them, but the issue is that...
> 
> > diff --git a/tools/testing/selftests/drivers/net/mlxsw/sch_tbf_ets.sh b/tools/testing/selftests/drivers/net/mlxsw/sch_tbf_ets.sh
> > index c6ce0b448bf3..bf57400e14ee 100755
> > --- a/tools/testing/selftests/drivers/net/mlxsw/sch_tbf_ets.sh
> > +++ b/tools/testing/selftests/drivers/net/mlxsw/sch_tbf_ets.sh
> > @@ -2,7 +2,7 @@
> >  # SPDX-License-Identifier: GPL-2.0
> >  
> >  source qos_lib.sh
> > -bail_on_lldpad
> > +bail_on_lldpad "configure DCB" "configure Qdiscs"
> 
> ... lib.sh isn't sourced at this point yet. `source
> $lib_dir/sch_tbf_ets.sh' brings that in later in the file, so the bail
> would need to be below that. But if it is, it won't run until after the
> test, which is useless.
> 
> Maybe all it takes is to replace that `source qos_lib.sh' with
> `NUM_NETIFS=0 source $lib_dir/lib.sh', as we do in in_ns(), but I'll
> need to check.
> 
> That's why I proposed to do it myself, some of it is fiddly and having a
> switch handy is going to be... um, handy.

Are you planning to take a look and see whether the callers of bail_on_lldpad()
can also source $lib_dir/lib.sh? Should I resend a blind change here?