Re: [PATCH] rcu: Use IRQ_WORK_INIT_HARD for srcu's irq_work

Le Thu, Aug 20, 2026 at 03:55:00PM +0200, Sebastian Andrzej Siewior a écrit :
> On 2026-08-20 14:40:57 [+0200], Frederic Weisbecker wrote:
> > Le Thu, Aug 20, 2026 at 01:00:08PM +0200, Sebastian Andrzej Siewior a écrit :
> > > The irq_work in srcu is used to schedule a delayed work. The work item
> > > is not scheduled directly because it is not always possible wake a
> > > thread directly.
> > > 
> > > On PREEMPT_RT the default irq_work is initialized with IRQ_WORK_LAZY and
> > > is delayed to the irq_work thread. A system booted with the command line
> > > "trace_event=…" will freeze during boot because the irq_work thread is
> > > not yet deployed (and tracing uses synchronize_srcu() in
> > > tp_rcu_cond_sync()).
> > > 
> > > The irq_work performs just a wakeup a thread, there is nothing wrong
> > > with doing this from hardirq context on PREEMPT_RT.
> > > 
> > > Use IRQ_WORK_INIT_HARD for srcu's irq_work.
> > > 
> > > Fixes: 7c405fb3279b3 ("rcu: Use an intermediate irq_work to start process_srcu()")
> > > Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
> > > ---
> > >  kernel/rcu/srcutree.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> > > index 7c2f7cc131f7a..b703233d4a63f 100644
> > > --- a/kernel/rcu/srcutree.c
> > > +++ b/kernel/rcu/srcutree.c
> > > @@ -218,7 +218,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
> > >  	mutex_init(&ssp->srcu_sup->srcu_barrier_mutex);
> > >  	atomic_set(&ssp->srcu_sup->srcu_barrier_cpu_cnt, 0);
> > >  	INIT_DELAYED_WORK(&ssp->srcu_sup->work, process_srcu);
> > > -	init_irq_work(&ssp->srcu_sup->irq_work, srcu_irq_work);
> > > +	ssp->srcu_sup->irq_work = IRQ_WORK_INIT_HARD(srcu_irq_work);
> > 
> > It may need a comment to not lose that.
> 
> We could also make irq_work.o earlier in the Makefile, right at the top
> so it comes before tracing. This was one of the first ideas and makes
> the issue also go away. But given that it is just a wake we avoid waking
> the irq_work/ thread just to schedule a timer/ wake workqueue.
> 
> > >  	ssp->srcu_sup->sda_is_static = is_static;
> > >  	if (!is_static) {
> > >  		ssp->sda = alloc_percpu(struct srcu_data);
> > > -- 
> > > 2.55.0
> > > 
> > >
> > 
> > Perhaps all irq works queued before the relevant kthread is started should
> > temporarily be queued as hard?
> 
> Urgh. irq_work shouldn't become a widespread user ;)
> If the scheduler is up but irq_work thread is not yet created and the
> irq_work uses sleeping locks then it would create warnings.
> In testing right now I see just wake_up_klogd_work_func() which is
> always LAZY but would qualify otherwise for a warning. We wouldn't
> trigger anything yet but you get the idea.

Fair enough! But should we warn if a non-hard IRQ work is queued before
the kthread is created?

Thanks.

> 
> > Thanks.
> 
> Sebastian
> 

-- 
Frederic Weisbecker
SUSE Labs