sched: Make tunable scaling style configurable

As scaling now takes place on all kind of cpu add/remove events a user
that configures values via proc should be able to configure if his set
values are still rescaled or kept whatever happens.

As the comments state that log2 was just a second guess that worked the
interface is not just designed for on/off, but to choose a scaling type.
Currently this allows none, log and linear, but more important it allwos
us to keep the interface even if someone has an even better idea how to
scale the values.

Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1259579808-11357-3-git-send-email-ehrhardt@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

5 files changed, 61 insertions, 2 deletions

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 4b1ebd3280c..ee9f200d12d 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1902,13 +1902,22 @@ extern unsigned int sysctl_sched_wakeup_granularity;
 extern unsigned int sysctl_sched_shares_ratelimit;
 extern unsigned int sysctl_sched_shares_thresh;
 extern unsigned int sysctl_sched_child_runs_first;
+
+enum sched_tunable_scaling {
+	SCHED_TUNABLESCALING_NONE,
+	SCHED_TUNABLESCALING_LOG,
+	SCHED_TUNABLESCALING_LINEAR,
+	SCHED_TUNABLESCALING_END,
+};
+extern enum sched_tunable_scaling sysctl_sched_tunable_scaling;
+
 #ifdef CONFIG_SCHED_DEBUG
 extern unsigned int sysctl_sched_migration_cost;
 extern unsigned int sysctl_sched_nr_migrate;
 extern unsigned int sysctl_sched_time_avg;
 extern unsigned int sysctl_timer_migration;
 
-int sched_nr_latency_handler(struct ctl_table *table, int write,
+int sched_proc_update_handler(struct ctl_table *table, int write,
 		void __user *buffer, size_t *length,
 		loff_t *ppos);
 #endif
diff --git a/kernel/sched.c b/kernel/sched.c
index b54ecf84b6b..116efed962c 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -7033,7 +7033,20 @@ cpumask_var_t nohz_cpu_mask;
 static void update_sysctl(void)
 {
 	unsigned int cpus = min(num_online_cpus(), 8U);
-	unsigned int factor = 1 + ilog2(cpus);
+	unsigned int factor;
+
+	switch (sysctl_sched_tunable_scaling) {
+	case SCHED_TUNABLESCALING_NONE:
+		factor = 1;
+		break;
+	case SCHED_TUNABLESCALING_LINEAR:
+		factor = cpus;
+		break;
+	case SCHED_TUNABLESCALING_LOG:
+	default:
+		factor = 1 + ilog2(cpus);
+		break;
+	}
 
 #define SET_SYSCTL(name) \
 	(sysctl_##name = (factor) * normalized_sysctl_##name)
diff --git a/kernel/sched_debug.c b/kernel/sched_debug.c
index 5fda66615fe..0fc5287fe80 100644
--- a/kernel/sched_debug.c
+++ b/kernel/sched_debug.c
@@ -309,6 +309,12 @@ static void print_cpu(struct seq_file *m, int cpu)
 	print_rq(m, rq, cpu);
 }
 
+static const char *sched_tunable_scaling_names[] = {
+	"none",
+	"logaritmic",
+	"linear"
+};
+
 static int sched_debug_show(struct seq_file *m, void *v)
 {
 	u64 now = ktime_to_ns(ktime_get());
@@ -334,6 +340,10 @@ static int sched_debug_show(struct seq_file *m, void *v)
 #undef PN
 #undef P
 
+	SEQ_printf(m, "  .%-40s: %d (%s)\n", "sysctl_sched_tunable_scaling",
+		sysctl_sched_tunable_scaling,
+		sched_tunable_scaling_names[sysctl_sched_tunable_scaling]);
+
 	for_each_online_cpu(cpu)
 		print_cpu(m, cpu);
 
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 71b3458245e..455106d318a 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -21,6 +21,7 @@
  */
 
 #include <linux/latencytop.h>
+#include <linux/sched.h>
 
 /*
  * Targeted preemption latency for CPU-bound tasks:
@@ -38,6 +39,18 @@ unsigned int sysctl_sched_latency = 5000000ULL;
 unsigned int normalized_sysctl_sched_latency = 5000000ULL;
 
 /*
+ * The initial- and re-scaling of tunables is configurable
+ * (default SCHED_TUNABLESCALING_LOG = *(1+ilog(ncpus))
+ *
+ * Options are:
+ * SCHED_TUNABLESCALING_NONE - unscaled, always *1
+ * SCHED_TUNABLESCALING_LOG - scaled logarithmical, *1+ilog(ncpus)
+ * SCHED_TUNABLESCALING_LINEAR - scaled linear, *ncpus
+ */
+enum sched_tunable_scaling sysctl_sched_tunable_scaling
+	= SCHED_TUNABLESCALING_LOG;
+
+/*
  * Minimal preemption granularity for CPU-bound tasks:
  * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds)
  */
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index e5cc53514ca..d10406e5fdf 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -251,6 +251,8 @@ static int min_sched_granularity_ns = 100000;		/* 100 usecs */
 static int max_sched_granularity_ns = NSEC_PER_SEC;	/* 1 second */
 static int min_wakeup_granularity_ns;			/* 0 usecs */
 static int max_wakeup_granularity_ns = NSEC_PER_SEC;	/* 1 second */
+static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
+static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 #endif
 
 static struct ctl_table kern_table[] = {
@@ -306,6 +308,18 @@ static struct ctl_table kern_table[] = {
 	},
 	{
 		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "sched_tunable_scaling",
+		.data		= &sysctl_sched_tunable_scaling,
+		.maxlen		= sizeof(enum sched_tunable_scaling),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec_minmax,
+		.strategy	= &sysctl_intvec,
+		.extra1		= &min_sched_tunable_scaling,
+		.extra2		= &max_sched_tunable_scaling,
+	},
+
+	{
+		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "sched_shares_thresh",
 		.data		= &sysctl_sched_shares_thresh,
 		.maxlen		= sizeof(unsigned int),