Re: [PATCH v2] sched,livepatch: Untangle cond_resched() and live-patching

[Petr Mladek <pmladek@suse.com>]

On Tue 2025-05-13 15:34:50, Miroslav Benes wrote:
> On Fri, 9 May 2025, Sebastian Andrzej Siewior wrote:
> 
> > From: Peter Zijlstra <peterz@infradead.org>
> > 
> > With the goal of deprecating / removing VOLUNTARY preempt, live-patch
> > needs to stop relying on cond_resched() to make forward progress.
> > 
> > Instead, rely on schedule() with TASK_FREEZABLE set. Just like
> > live-patching, the freezer needs to be able to stop tasks in a safe /
> > known state.
> > 
> > @@ -365,27 +356,20 @@ static bool klp_try_switch_task(struct task_struct *task)
> >  
> >  void __klp_sched_try_switch(void)
> >  {
> > +	/*
> > +	 * This function is called from __schedule() while a context switch is
> > +	 * about to happen. Preemption is already disabled and klp_mutex
> > +	 * can't be acquired.
> > +	 * Disabled preemption is used to prevent racing with other callers of
> > +	 * klp_try_switch_task(). Thanks to task_call_func() they won't be
> > +	 * able to switch to this task while it's running.
> > +	 */
> > +	lockdep_assert_preemption_disabled();
> > +
> > +	/* Make sure current didn't get patched */
> >       if (likely(!klp_patch_pending(current)))
> >                return;
> 
> This last comment is not precise. If !klp_patch_pending(), there is 
> nothing to do. Fast way out. So if it was up to me, I would remove the 
> line all together.

I think that this is not just a speedup. Right below is a read memory
barrier. It says that we need to read TIF_PATCH_PENDING here to
make sure that klp_target_state is correct after the barrier.

Anyway, I agree that the comment is confusing and might be removed.
The comment above the memory barrier might/should be enough.

Best Regards,
Petr