Re: [PATCH v2 12/14] torture: Replace cpu_up/down with device_online/offline

From: Qais Yousef
Date: Thu Nov 28 2019 - 11:56:18 EST


On 11/27/19 13:47, Paul E. McKenney wrote:
> On Mon, Nov 25, 2019 at 11:27:52AM +0000, Qais Yousef wrote:
> > The core device API performs extra housekeeping bits that are missing
> > from directly calling cpu_up/down.
> >
> > See commit a6717c01ddc2 ("powerpc/rtas: use device model APIs and
> > serialization during LPM") for an example description of what might go
> > wrong.
> >
> > This also prepares to make cpu_up/down a private interface for anything
> > but the cpu subsystem.
> >
> > Signed-off-by: Qais Yousef <qais.yousef@xxxxxxx>
> > CC: Davidlohr Bueso <dave@xxxxxxxxxxxx>
> > CC: "Paul E. McKenney" <paulmck@xxxxxxxxxx>
> > CC: Josh Triplett <josh@xxxxxxxxxxxxxxxx>
> > CC: linux-kernel@xxxxxxxxxxxxxxx
>
> Looks fine from an rcutorture viewpoint, but why not provide an API
> that pulled lock_device_hotplug() and unlock_device_hotplug() into the
> online/offline calls?

I *think* the right way to do what you say is by doing lock_device_hotplug()
inside device_{online, offline}() - which affects all drivers not just the CPU.

And even then, I think we need to refcount it so nested calls won't deadlock.

I don't know if this can break any rule or not. If Greg thinks it's okay I'd be
happy to post some patches that do just that.

Thanks

--
Qais Yousef

>
> Thanx, Paul
>
> > ---
> > kernel/torture.c | 15 +++++++++++----
> > 1 file changed, 11 insertions(+), 4 deletions(-)
> >
> > diff --git a/kernel/torture.c b/kernel/torture.c
> > index 7c13f5558b71..12115024feb2 100644
> > --- a/kernel/torture.c
> > +++ b/kernel/torture.c
> > @@ -97,7 +97,9 @@ bool torture_offline(int cpu, long *n_offl_attempts, long *n_offl_successes,
> > torture_type, cpu);
> > starttime = jiffies;
> > (*n_offl_attempts)++;
> > - ret = cpu_down(cpu);
> > + lock_device_hotplug();
> > + ret = device_offline(get_cpu_device(cpu));
> > + unlock_device_hotplug();
> > if (ret) {
> > if (verbose)
> > pr_alert("%s" TORTURE_FLAG
> > @@ -148,7 +150,9 @@ bool torture_online(int cpu, long *n_onl_attempts, long *n_onl_successes,
> > torture_type, cpu);
> > starttime = jiffies;
> > (*n_onl_attempts)++;
> > - ret = cpu_up(cpu);
> > + lock_device_hotplug();
> > + ret = device_online(get_cpu_device(cpu));
> > + unlock_device_hotplug();
> > if (ret) {
> > if (verbose)
> > pr_alert("%s" TORTURE_FLAG
> > @@ -192,17 +196,20 @@ torture_onoff(void *arg)
> > for_each_online_cpu(cpu)
> > maxcpu = cpu;
> > WARN_ON(maxcpu < 0);
> > - if (!IS_MODULE(CONFIG_TORTURE_TEST))
> > + if (!IS_MODULE(CONFIG_TORTURE_TEST)) {
> > + lock_device_hotplug();
> > for_each_possible_cpu(cpu) {
> > if (cpu_online(cpu))
> > continue;
> > - ret = cpu_up(cpu);
> > + ret = device_online(get_cpu_device(cpu));
> > if (ret && verbose) {
> > pr_alert("%s" TORTURE_FLAG
> > "%s: Initial online %d: errno %d\n",
> > __func__, torture_type, cpu, ret);
> > }
> > }
> > + unlock_device_hotplug();
> > + }
> >
> > if (maxcpu == 0) {
> > VERBOSE_TOROUT_STRING("Only one CPU, so CPU-hotplug testing is disabled");
> > --
> > 2.17.1
> >