Re: Kernel-managed IRQ affinity (cont)

[Peter Xu]

On Thu, Dec 19, 2019 at 01:09:17PM -0500, Peter Xu wrote:
> On Fri, Dec 20, 2019 at 12:11:15AM +0800, Ming Lei wrote:
> > OK, please try the following patch:
> > 
> > 
> > diff --git a/include/linux/sched/isolation.h b/include/linux/sched/isolation.h
> > index 6c8512d3be88..0fbcbacd1b29 100644
> > --- a/include/linux/sched/isolation.h
> > +++ b/include/linux/sched/isolation.h
> > @@ -13,6 +13,7 @@ enum hk_flags {
> >  	HK_FLAG_TICK		= (1 << 4),
> >  	HK_FLAG_DOMAIN		= (1 << 5),
> >  	HK_FLAG_WQ		= (1 << 6),
> > +	HK_FLAG_MANAGED_IRQ	= (1 << 7),
> >  };
> >  
> >  #ifdef CONFIG_CPU_ISOLATION
> > diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
> > index 1753486b440c..0a75a09cc4e8 100644
> > --- a/kernel/irq/manage.c
> > +++ b/kernel/irq/manage.c
> > @@ -20,6 +20,7 @@
> >  #include <linux/sched/task.h>
> >  #include <uapi/linux/sched/types.h>
> >  #include <linux/task_work.h>
> > +#include <linux/sched/isolation.h>
> >  
> >  #include "internals.h"
> >  
> > @@ -212,12 +213,33 @@ int irq_do_set_affinity(struct irq_data *data, const struct cpumask *mask,
> >  {
> >  	struct irq_desc *desc = irq_data_to_desc(data);
> >  	struct irq_chip *chip = irq_data_get_irq_chip(data);
> > +	const struct cpumask *housekeeping_mask =
> > +		housekeeping_cpumask(HK_FLAG_MANAGED_IRQ);
> >  	int ret;
> > +	cpumask_var_t tmp_mask;
> >  
> >  	if (!chip || !chip->irq_set_affinity)
> >  		return -EINVAL;
> >  
> > -	ret = chip->irq_set_affinity(data, mask, force);
> > +	if (!zalloc_cpumask_var(&tmp_mask, GFP_KERNEL))
> > +		return -EINVAL;
> > +
> > +	/*
> > +	 * Userspace can't change managed irq's affinity, make sure
> > +	 * that isolated CPU won't be selected as the effective CPU
> > +	 * if this irq's affinity includes both isolated CPU and
> > +	 * housekeeping CPU.
> > +	 *
> > +	 * This way guarantees that isolated CPU won't be interrupted
> > +	 * by IO submitted from housekeeping CPU.
> > +	 */
> > +	if (irqd_affinity_is_managed(data) &&
> > +			cpumask_intersects(mask, housekeeping_mask))
> > +		cpumask_and(tmp_mask, mask, housekeeping_mask);
> > +	else
> > +		cpumask_copy(tmp_mask, mask);
> > +
> > +	ret = chip->irq_set_affinity(data, tmp_mask, force);
> >  	switch (ret) {
> >  	case IRQ_SET_MASK_OK:
> >  	case IRQ_SET_MASK_OK_DONE:
> > @@ -229,6 +251,8 @@ int irq_do_set_affinity(struct irq_data *data, const struct cpumask *mask,
> >  		ret = 0;
> >  	}
> >  
> > +	free_cpumask_var(tmp_mask);
> > +
> >  	return ret;
> >  }
> >  
> > diff --git a/kernel/sched/isolation.c b/kernel/sched/isolation.c
> > index 9fcb2a695a41..008d6ac2342b 100644
> > --- a/kernel/sched/isolation.c
> > +++ b/kernel/sched/isolation.c
> > @@ -163,6 +163,12 @@ static int __init housekeeping_isolcpus_setup(char *str)
> >  			continue;
> >  		}
> >  
> > +		if (!strncmp(str, "managed_irq,", 12)) {
> > +			str += 12;
> > +			flags |= HK_FLAG_MANAGED_IRQ;
> > +			continue;
> > +		}
> > +
> >  		pr_warn("isolcpus: Error, unknown flag\n");
> >  		return 0;
> >  	}
> 
> Thanks for the quick patch.  I'll test after my current round of tests
> finish and update.  I'll probably believe this will work for us as
> long as it "functionally" works :) (after all it won't even need a RT
> environment because it's really about where to put some IRQs).  So
> IMHO the more important thing is whether such a solution could be
> acceptable by the upstream.

I've tested this patch, it works for us.  "isolcpus=managed_irq,2-9"
gives me:

[root@rt-vm 32]# pwd
/proc/irq/32
[root@rt-vm 32]# cat smp_affinity
003
[root@rt-vm 32]# cat effective_affinity
001

Thomas, do you think Ming's solution could be accepted?

Thanks,

-- 
Peter Xu