---

yaml
---
r: 182519
b: refs/heads/master
c: fa535a7
h: refs/heads/master
i:
  182517: 9974d5d
  182515: 54c9bb6
  182511: ce7ff1d
v: v3

[refs]

@@ -1,2 +1,2 @@
 ---
-refs/heads/master: 0c9cf2efd74dbc90354e2ccc7dbd6bad68ec6c4d
+refs/heads/master: fa535a77bd3fa32b9215ba375d6a202fe73e1dd6

trunk/kernel/sched.c

@@ -8997,13 +8997,31 @@ static void cpuacct_charge(struct task_struct *tsk, u64 cputime)
 	rcu_read_unlock();
 }
 
+/*
+ * When CONFIG_VIRT_CPU_ACCOUNTING is enabled one jiffy can be very large
+ * in cputime_t units. As a result, cpuacct_update_stats calls
+ * percpu_counter_add with values large enough to always overflow the
+ * per cpu batch limit causing bad SMP scalability.
+ *
+ * To fix this we scale percpu_counter_batch by cputime_one_jiffy so we
+ * batch the same amount of time with CONFIG_VIRT_CPU_ACCOUNTING disabled
+ * and enabled. We cap it at INT_MAX which is the largest allowed batch value.
+ */
+#ifdef CONFIG_SMP
+#define CPUACCT_BATCH	\
+	min_t(long, percpu_counter_batch * cputime_one_jiffy, INT_MAX)
+#else
+#define CPUACCT_BATCH	0
+#endif
+
 /*
  * Charge the system/user time to the task's accounting group.
  */
 static void cpuacct_update_stats(struct task_struct *tsk,
 		enum cpuacct_stat_index idx, cputime_t val)
 {
 	struct cpuacct *ca;
+	int batch = CPUACCT_BATCH;
 
 	if (unlikely(!cpuacct_subsys.active))
 		return;
@@ -9012,7 +9030,7 @@ static void cpuacct_update_stats(struct task_struct *tsk,
 	ca = task_ca(tsk);
 
 	do {
-		percpu_counter_add(&ca->cpustat[idx], val);
+		__percpu_counter_add(&ca->cpustat[idx], val, batch);
 		ca = ca->parent;
 	} while (ca);
 	rcu_read_unlock();