[patch v9 06/10] task isolation: sync vmstats conditional on changes

[Marcelo Tosatti <mtosatti@redhat.com>]

The logic to disable vmstat worker thread, when entering
nohz full, does not cover all scenarios. For example, it is possible
for the following to happen:
References: <20220127172319.428529308@fuller.cnet>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8

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>

---
 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
 
 #ifdef CONFIG_SMP
-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 +334,24 @@ void set_pgdat_percpu_threshold(pg_data_
 	}
 }
 
+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);
+}
+
+void init_sync_vmstat(void)
+{
+	raw_cpu_write(vmstat_dirty, true);
+}
+
+EXPORT_SYMBOL_GPL(vmstat_dirty);
+
 /*
  * For use when we know that interrupts are disabled,
  * or when we know that preemption is disabled and that
@@ -366,6 +384,7 @@ void __mod_zone_page_state(struct zone *
 		x = 0;
 	}
 	__this_cpu_write(*p, x);
+	mark_vmstat_dirty();
 
 	if (IS_ENABLED(CONFIG_PREEMPT_RT))
 		preempt_enable();
@@ -404,6 +423,7 @@ void __mod_node_page_state(struct pglist
 		x = 0;
 	}
 	__this_cpu_write(*p, x);
+	mark_vmstat_dirty();
 
 	if (IS_ENABLED(CONFIG_PREEMPT_RT))
 		preempt_enable();
@@ -602,6 +622,7 @@ static inline void mod_zone_state(struct
 
 	if (z)
 		zone_page_state_add(z, zone, item);
+	mark_vmstat_dirty();
 }
 
 void mod_zone_page_state(struct zone *zone, enum zone_stat_item item,
@@ -670,6 +691,7 @@ static inline void mod_node_state(struct
 
 	if (z)
 		node_page_state_add(z, pgdat, item);
+	mark_vmstat_dirty();
 }
 
 void mod_node_page_state(struct pglist_data *pgdat, enum node_stat_item item,
@@ -1087,6 +1109,7 @@ static void fill_contig_page_info(struct
 			info->free_blocks_suitable += blocks <<
 						(order - suitable_order);
 	}
+	mark_vmstat_dirty();
 }
 
 /*
@@ -1443,6 +1466,7 @@ static void walk_zones_in_node(struct se
 		if (!nolock)
 			spin_unlock_irqrestore(&zone->lock, flags);
 	}
+	mark_vmstat_dirty();
 }
 #endif
 
@@ -1512,6 +1536,7 @@ static void pagetypeinfo_showfree_print(
 		}
 		seq_putc(m, '\n');
 	}
+	mark_vmstat_dirty();
 }
 
 /* Print out the free pages at each order for each migatetype */
@@ -1932,6 +1957,7 @@ static void vmstat_update(struct work_st
 				this_cpu_ptr(&vmstat_work),
 				round_jiffies_relative(sysctl_stat_interval));
 	}
+	mark_vmstat_dirty();
 }
 
 /*
@@ -2018,13 +2044,14 @@ static void vmstat_shepherd(struct work_
 		round_jiffies_relative(sysctl_stat_interval));
 }
 
-void sync_vmstat(void)
+void __sync_vmstat(void)
 {
 	int cpu;
 
 	cpu = get_cpu();
 
 	refresh_cpu_vm_stats(false);
+	raw_cpu_write(vmstat_dirty, false);
 	put_cpu();
 
 	/*