Messages in this thread | | | Date | Thu, 10 Mar 2022 13:30:14 -0300 | From | Marcelo Tosatti <> | Subject | Re: [patch v11 07/13] task isolation: sync vmstats conditional on changes |
| |
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.
| |