[patch 3/5] mm: vmstat: optionally flush per-CPU vmstat counters on return to userspace

[Marcelo Tosatti]

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:

1) enter nohz_full, which calls refresh_cpu_vm_stats, syncing the stats.
2) app runs mlock, which increases counters for mlock'ed pages.
3) start -RT loop

Since refresh_cpu_vm_stats from nohz_full logic can happen _before_
the mlock, vmstat shepherd can restart vmstat worker thread on 
the CPU in question.

To fix this, optionally sync the vmstat counters when returning
from userspace, controllable by a new "quiesce_on_exit_to_usermode" isolcpus 
flags (default off).

Signed-off-by: Marcelo Tosatti <mtosatti@xxxxxxxxxx>

Index: linux-2.6-vmstat-update/kernel/sched/isolation.c
===================================================================
--- linux-2.6-vmstat-update.orig/kernel/sched/isolation.c
+++ linux-2.6-vmstat-update/kernel/sched/isolation.c
@@ -8,6 +8,7 @@
  *
  */
 #include "sched.h"
+#include <linux/vmstat.h>
 
 DEFINE_STATIC_KEY_FALSE(housekeeping_overridden);
 EXPORT_SYMBOL_GPL(housekeeping_overridden);
@@ -129,6 +130,11 @@ static int __init housekeeping_setup(cha
 		}
 	}
 
+#ifdef CONFIG_SMP
+	if (flags & HK_FLAG_QUIESCE_URET)
+		static_branch_enable(&vmstat_sync_enabled);
+#endif
+
 	housekeeping_flags |= flags;
 
 	free_bootmem_cpumask_var(non_housekeeping_mask);
Index: linux-2.6-vmstat-update/include/linux/vmstat.h
===================================================================
--- linux-2.6-vmstat-update.orig/include/linux/vmstat.h
+++ linux-2.6-vmstat-update/include/linux/vmstat.h
@@ -21,6 +21,23 @@ int sysctl_vm_numa_stat_handler(struct c
 		void *buffer, size_t *length, loff_t *ppos);
 #endif
 
+#ifdef CONFIG_SMP
+DECLARE_STATIC_KEY_FALSE(vmstat_sync_enabled);
+
+extern void __sync_vmstat(void);
+static inline void sync_vmstat(void)
+{
+	if (static_branch_unlikely(&vmstat_sync_enabled))
+		__sync_vmstat();
+}
+#else
+
+static inline void sync_vmstat(void)
+{
+}
+
+#endif
+
 struct reclaim_stat {
 	unsigned nr_dirty;
 	unsigned nr_unqueued_dirty;
Index: linux-2.6-vmstat-update/mm/vmstat.c
===================================================================
--- linux-2.6-vmstat-update.orig/mm/vmstat.c
+++ linux-2.6-vmstat-update/mm/vmstat.c
@@ -28,6 +28,7 @@
 #include <linux/mm_inline.h>
 #include <linux/page_ext.h>
 #include <linux/page_owner.h>
+#include <linux/sched/isolation.h>
 
 #include "internal.h"
 
@@ -308,6 +309,17 @@ void set_pgdat_percpu_threshold(pg_data_
 	}
 }
 
+DEFINE_STATIC_KEY_FALSE(vmstat_sync_enabled);
+static DEFINE_PER_CPU_ALIGNED(bool, vmstat_dirty);
+
+static inline void mark_vmstat_dirty(void)
+{
+	if (!static_branch_unlikely(&vmstat_sync_enabled))
+		return;
+
+	raw_cpu_write(vmstat_dirty, true);
+}
+
 /*
  * For use when we know that interrupts are disabled,
  * or when we know that preemption is disabled and that
@@ -330,6 +342,7 @@ void __mod_zone_page_state(struct zone *
 		x = 0;
 	}
 	__this_cpu_write(*p, x);
+	mark_vmstat_dirty();
 }
 EXPORT_SYMBOL(__mod_zone_page_state);
 
@@ -361,6 +374,7 @@ void __mod_node_page_state(struct pglist
 		x = 0;
 	}
 	__this_cpu_write(*p, x);
+	mark_vmstat_dirty();
 }
 EXPORT_SYMBOL(__mod_node_page_state);
 
@@ -401,6 +415,7 @@ void __inc_zone_state(struct zone *zone,
 		zone_page_state_add(v + overstep, zone, item);
 		__this_cpu_write(*p, -overstep);
 	}
+	mark_vmstat_dirty();
 }
 
 void __inc_node_state(struct pglist_data *pgdat, enum node_stat_item item)
@@ -419,6 +434,7 @@ void __inc_node_state(struct pglist_data
 		node_page_state_add(v + overstep, pgdat, item);
 		__this_cpu_write(*p, -overstep);
 	}
+	mark_vmstat_dirty();
 }
 
 void __inc_zone_page_state(struct page *page, enum zone_stat_item item)
@@ -447,6 +463,7 @@ void __dec_zone_state(struct zone *zone,
 		zone_page_state_add(v - overstep, zone, item);
 		__this_cpu_write(*p, overstep);
 	}
+	mark_vmstat_dirty();
 }
 
 void __dec_node_state(struct pglist_data *pgdat, enum node_stat_item item)
@@ -465,6 +482,7 @@ void __dec_node_state(struct pglist_data
 		node_page_state_add(v - overstep, pgdat, item);
 		__this_cpu_write(*p, overstep);
 	}
+	mark_vmstat_dirty();
 }
 
 void __dec_zone_page_state(struct page *page, enum zone_stat_item item)
@@ -528,6 +546,7 @@ 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,
@@ -596,6 +615,7 @@ 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,
@@ -2006,6 +2026,37 @@ static void vmstat_shepherd(struct work_
 		round_jiffies_relative(sysctl_stat_interval));
 }
 
+void __sync_vmstat(void)
+{
+	int cpu;
+
+	cpu = get_cpu();
+	if (housekeeping_cpu(cpu, HK_FLAG_QUIESCE_URET)) {
+		put_cpu();
+		return;
+	}
+
+	if (!raw_cpu_read(vmstat_dirty)) {
+		put_cpu();
+		return;
+	}
+
+	refresh_cpu_vm_stats(false);
+	raw_cpu_write(vmstat_dirty, false);
+	put_cpu();
+
+	/*
+	 * If task is migrated to another CPU between put_cpu
+	 * and cancel_delayed_work_sync, the code below might
+	 * cancel vmstat_update work for a different cpu
+	 * (than the one from which the vmstats were flushed).
+	 *
+	 * However, vmstat shepherd will re-enable it later,
+	 * so its harmless.
+	 */
+	cancel_delayed_work_sync(&per_cpu(vmstat_work, cpu));
+}
+
 static void __init start_shepherd_timer(void)
 {
 	int cpu;
Index: linux-2.6-vmstat-update/kernel/entry/common.c
===================================================================
--- linux-2.6-vmstat-update.orig/kernel/entry/common.c
+++ linux-2.6-vmstat-update/kernel/entry/common.c
@@ -6,6 +6,7 @@
 #include <linux/livepatch.h>
 #include <linux/audit.h>
 #include <linux/tick.h>
+#include <linux/vmstat.h>
 
 #include "common.h"
 
@@ -290,6 +291,7 @@ static void syscall_exit_to_user_mode_pr
  */
 static void isolation_exit_to_user_mode_prepare(void)
 {
+	sync_vmstat();
 }
 
 static __always_inline void __syscall_exit_to_user_mode_work(struct pt_regs *regs)