Re: [PATCH-next v2 6/6] selftests/cgroup: Skip test_cpuset_prs.sh test if some required CPUs are offline

[Waiman Long]

On 10/11/26 5:00 AM, Guopeng Zhang wrote:
> 在 2026/10/11 06:19, Waiman Long 写道:
> > The test_cpuset_prs.sh test assumes that all the relevant CPUs being
> > tested are online. If some of them are offline, some of the test cases
> > may fail. So skip the test if the required CPUs are either offline or
> > not present.
> >
> > Signed-off-by: Waiman Long <longman@xxxxxxxxxx>
> > ---
> >   .../selftests/cgroup/test_cpuset_prs.sh       | 20 +++++++++++++++++++
> >   1 file changed, 20 insertions(+)
> >
> > diff --git a/tools/testing/selftests/cgroup/test_cpuset_prs.sh b/tools/testing/selftests/cgroup/test_cpuset_prs.sh
> > index 3d392e8c66aa..f57971ebb90e 100755
> > --- a/tools/testing/selftests/cgroup/test_cpuset_prs.sh
> > +++ b/tools/testing/selftests/cgroup/test_cpuset_prs.sh
> > @@ -118,6 +118,26 @@ TEST_ISOLCPUS=
> >   		skip_test "Current isolated CPUs ($CURRENT_ISOLCPUS) don't match boot isolated CPUs!"
> >   }
> >   
> > +#
> > +# This test assumes that all the relevant CPUs being tested are online.
> > +# If some of them are offline, some of the tests may fail. So we are going to
> > +# skip running this test script if some of the required CPUs may be offline.
> > +# The required CPUs include CPUs 0-7 and the first boot time isolated CPU if
> > +# present. However the check below is not exhaustive and can be a false
> > +# positive, but users should not run this test if some CPUs are offline.
> > +#
> > +ONLINE_CPUS=$(cat /sys/devices/system/cpu/online | sed -e "s/,.*//")
> > +ONLINE_FIRST=$(echo $ONLINE_CPUS | sed -e "s/-.*//")
> > +ONLINE_LAST=$(echo $ONLINE_CPUS | sed -e "s/.*-//")
> > +if [[ -n "$FIRST_ISOLCPUS" ]]
> > +then
> > +	LAST_REQUIRED_CPU=$FIRST_ISOLCPUS
> > +else
> > +	LAST_REQUIRED_CPU=7
> > +fi
> > +[[ $ONLINE_FIRST != 0 || $ONLINE_LAST -lt $LAST_REQUIRED_CPU ]] &&
> > +	skip_test "Some of the required CPUs are offline or not present!"
> > +
> Could this check run before the setup that creates the test cgroup and
> changes sched/verbose?
>
> skip_test() exits directly, and the cleanup trap hasn't been installed
> here. We have already created test; with -v, sched/verbose has also been
> set to Y. This exit leaves the cgroup behind and bypasses restoring
> verbose.

Good point. I am going to move the trap statement earlier to make sure 
that it will be run when there is an early abort.

Cheers,
Longman