Re: [PATCH 1/2] sched_ext: Track currently locked rq

[Breno Leitao]

On Wed, Jul 16, 2025 at 02:40:14PM +0200, Andrea Righi wrote:
> Hi Breno,
> 
> On Wed, Jul 16, 2025 at 03:47:38AM -0700, Breno Leitao wrote:
> > Hello Andrea,
> > 
> > On Tue, Jul 15, 2025 at 07:20:28PM +0200, Andrea Righi wrote:
> > > > On Tue, Apr 22, 2025 at 10:26:32AM +0200, Andrea Righi wrote:
> > 
> > > > 
> > > > > +		lockdep_assert_rq_held(rq);
> > > > > +	__this_cpu_write(locked_rq, rq);
> > > > 
> > > > This is hitting the following BUG() on some of my debug kernels:
> > > > 
> > > > 	BUG: using __this_cpu_write() in preemptible [00000000] code: scx_layered_6-9/68770
> > > > 
> > > > I have lockdep enabled, and I don't see the assert above. I am wondering
> > > > if rq is locked but preemption continues to be enabled (!?)
> > > 
> > > Interesting. And it makes sense, because we may have callbacks called from
> > > a preemptible context (especially when rq == NULL).
> > > 
> > > I think we can just put a preempt_disable() / preempt_enable() around
> > > __this_cpu_write(). If we jump to another CPU during the callback it's
> > > fine, since we would track the rq state on the other CPU with its own local
> > > variable. And if we were able to jump there, it means that preemption was
> > > disabled as well.
> > 
> > First of all thanks for the suggestion!
> > 
> > What about a patch like the following:
> 
> Looks good to me, feel free to add my:
> 
> Acked-by: Andrea Righi <arighi@xxxxxxxxxx>
 
Thanks. I will send it to the mailing list them.
--breno