Re: [PATCH RT v3 3/5] sched: migrate_dis/enable: Use rt_invol_sleep

[Scott Wood]

On Tue, 2019-09-24 at 18:05 +0200, Sebastian Andrzej Siewior wrote:
> On 2019-09-24 10:47:36 [-0500], Scott Wood wrote:
> > When the stop machine finishes it will do a wake_up_process() via
> > complete().  Since this does not pass WF_LOCK_SLEEPER, saved_state will
> > be
> > cleared, and you'll have TASK_RUNNING when you get to other_func() and
> > schedule(), regardless of whether CPU1 sends wake_up() -- so this change
> > doesn't actually accomplish anything.
> 
> True, I completely missed that part.
> 
> > While as noted in the other thread I don't think these spurious wakeups
> > are
> > a huge problem, we could avoid them by doing stop_one_cpu_nowait() and
> > then
> > schedule() without messing with task state.  Since we're stopping our
> > own
> > cpu, it should be guaranteed that the stopper has finished by the time
> > we
> > exit schedule().
> 
> I remember loosing a state can be a problem. Lets say it is not "just"
> TASK_UNINTERRUPTIBLE -> TASK_RUNNING which sounds harmless but it is
> __TASK_TRACED and you lose it as part of unlocking siglock.

OK, sounds like stop_one_cpu_nowait() is the way to go then.

-Scott