Re: [RFC PATCH 5/6] sched/isolation: Introduce isolated task work

[Frederic Weisbecker <frederic@kernel.org>]

Le Wed, Jun 26, 2024 at 03:27:59PM +0200, Vlastimil Babka a écrit :
> On 6/25/24 3:52 PM, Frederic Weisbecker wrote:
> > Some asynchronous kernel work may be pending upon resume to userspace
> > and execute later on. On isolated workload this becomes problematic once
> > the process is done with preparatory work involving syscalls and wants
> > to run in userspace without being interrupted.
> > 
> > Provide an infrastructure to queue a work to be executed from the current
> > isolated task context right before resuming to userspace. This goes with
> > the assumption that isolated tasks are pinned to a single nohz_full CPU.
> > 
> > Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
> > ---
> >  include/linux/sched.h           |  1 +
> >  include/linux/sched/isolation.h | 17 +++++++++++++++++
> >  kernel/sched/core.c             |  1 +
> >  kernel/sched/isolation.c        | 31 +++++++++++++++++++++++++++++++
> >  kernel/sched/sched.h            |  1 +
> >  5 files changed, 51 insertions(+)
> > 
> > diff --git a/include/linux/sched.h b/include/linux/sched.h
> > index d531b610c410..f6df21866055 100644
> > --- a/include/linux/sched.h
> > +++ b/include/linux/sched.h
> > @@ -1349,6 +1349,7 @@ struct task_struct {
> >  #endif
> >  
> >  #ifdef CONFIG_NO_HZ_FULL
> > +	struct callback_head		nohz_full_work;
> >  	atomic_t			tick_dep_mask;
> >  #endif
> >  
> > diff --git a/include/linux/sched/isolation.h b/include/linux/sched/isolation.h
> > index 2b461129d1fa..e69ec5ed1d70 100644
> > --- a/include/linux/sched/isolation.h
> > +++ b/include/linux/sched/isolation.h
> > @@ -72,4 +72,21 @@ static inline bool cpu_is_isolated(int cpu)
> >  	       cpuset_cpu_is_isolated(cpu);
> >  }
> >  
> > +#if defined(CONFIG_NO_HZ_FULL)
> > +extern int __isolated_task_work_queue(void);
> > +
> > +static inline int isolated_task_work_queue(void)
> > +{
> > +	if (!housekeeping_test_cpu(raw_smp_processor_id(), HK_TYPE_TICK))
> 
> This is an unconditional call to a function defined in
> kernel/sched/isolation.c, and only there a static_branch_unlikely() test
> happens, but the call overhead is always paid, and the next patch adds that
> to folio_add_lru().
> 
> I notice a housekeeping_cpu() function above that does the static branch
> inline, which is great, except it defaults to return true so not directly
> applicable, but this function could be done the same way to keep the static
> branch inline.

Right, there definetly needs some inlining.

Thanks.