Re: [patch v11 07/13] task isolation: sync vmstats conditional on changes

[Marcelo Tosatti <mtosatti@redhat.com>]

On Mon, Feb 07, 2022 at 04:19:15PM +0100, Frederic Weisbecker wrote:
> On Fri, Feb 04, 2022 at 02:35:44PM -0300, Marcelo Tosatti wrote:
> > Rather than syncing VM-stats on every return to userspace
> > (or VM-entry), keep track of changes through a per-CPU bool.
> > 
> > This improves performance when enabling task isolated
> > for vcpu VMs.
> > 
> > Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> > 
> > ---
> > v11:
> > - Add TIF_TASK_ISOL bit to thread info flags and use it
> >   to decide whether to perform task isolation work on
> >   return to userspace     
> > 
> >  include/linux/vmstat.h |   13 ++++++++++++-
> >  mm/vmstat.c            |   29 ++++++++++++++++++++++++++++-
> >  2 files changed, 40 insertions(+), 2 deletions(-)
> > 
> > Index: linux-2.6/include/linux/vmstat.h
> > ===================================================================
> > --- linux-2.6.orig/include/linux/vmstat.h
> > +++ linux-2.6/include/linux/vmstat.h
> > @@ -22,7 +22,18 @@ int sysctl_vm_numa_stat_handler(struct c
> >  #endif
> >  
> >  #if defined(CONFIG_SMP) && defined(CONFIG_TASK_ISOLATION)
> > -void sync_vmstat(void);
> > +DECLARE_PER_CPU_ALIGNED(bool, vmstat_dirty);
> > +
> > +extern struct static_key vmstat_sync_enabled;
> > +
> > +void __sync_vmstat(void);
> > +static inline void sync_vmstat(void)
> > +{
> > +	if (static_key_false(&vmstat_sync_enabled))
> > +		__sync_vmstat();
> > +}
> > +
> > +void init_sync_vmstat(void);
> >  #else
> >  static inline void sync_vmstat(void)
> >  {
> > Index: linux-2.6/mm/vmstat.c
> > ===================================================================
> > --- linux-2.6.orig/mm/vmstat.c
> > +++ linux-2.6/mm/vmstat.c
> > @@ -334,6 +334,31 @@ void set_pgdat_percpu_threshold(pg_data_
> >  	}
> >  }
> >  
> > +#ifdef CONFIG_TASK_ISOLATION
> > +struct static_key vmstat_sync_enabled;
> > +DEFINE_PER_CPU_ALIGNED(bool, vmstat_dirty);
> > +
> > +static inline void mark_vmstat_dirty(void)
> > +{
> > +	if (!static_key_false(&vmstat_sync_enabled))
> > +		return;
> > +
> > +	raw_cpu_write(vmstat_dirty, true);
> 
> Why not __this_cpu_write() ? Shouldn't we make sure we are not
> preemptible and not mark the wrong CPU?

#ifdef CONFIG_HAVE_CMPXCHG_LOCAL
/*
 * If we have cmpxchg_local support then we do not need to incur the overhead
 * that comes with local_irq_save/restore if we use this_cpu_cmpxchg.
 *
 * mod_state() modifies the zone counter state through atomic per cpu
 * operations.
 *
 * Overstep mode specifies how overstep should handled:
 *     0       No overstepping
 *     1       Overstepping half of threshold
 *     -1      Overstepping minus half of threshold
*/
static inline void mod_zone_state(struct zone *zone,
       enum zone_stat_item item, long delta, int overstep_mode)
{
        struct per_cpu_zonestat __percpu *pcp = zone->per_cpu_zonestats;
        s8 __percpu *p = pcp->vm_stat_diff + item;
        long o, n, t, z;

        do {
                z = 0;  /* overflow to zone counters */

Perhaps one can

                        n = -os;
                }
        } while (this_cpu_cmpxchg(*p, o, n) != o);
							<-- migrate
        if (z)
                zone_page_state_add(z, zone, item);
        mark_vmstat_dirty();


But we assume the task which is using task isolation is pinned to a single CPU,
so that should not happen.