Commit b2be5e96 authored by Peter Zijlstra's avatar Peter Zijlstra Committed by Ingo Molnar

sched: reintroduce the sched_min_granularity tunable

we lost the sched_min_granularity tunable to a clever optimization
that uses the sched_latency/min_granularity ratio - but the ratio
is quite unintuitive to users and can also crash the kernel if the
ratio is set to 0. So reintroduce the min_granularity tunable,
while keeping the ratio maintained internally.

no functionality changed.

[ mingo@elte.hu: some fixlets. ]
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
parent 2cb8600e
...@@ -1460,12 +1460,16 @@ extern void sched_idle_next(void); ...@@ -1460,12 +1460,16 @@ extern void sched_idle_next(void);
#ifdef CONFIG_SCHED_DEBUG #ifdef CONFIG_SCHED_DEBUG
extern unsigned int sysctl_sched_latency; extern unsigned int sysctl_sched_latency;
extern unsigned int sysctl_sched_nr_latency; extern unsigned int sysctl_sched_min_granularity;
extern unsigned int sysctl_sched_wakeup_granularity; extern unsigned int sysctl_sched_wakeup_granularity;
extern unsigned int sysctl_sched_batch_wakeup_granularity; extern unsigned int sysctl_sched_batch_wakeup_granularity;
extern unsigned int sysctl_sched_child_runs_first; extern unsigned int sysctl_sched_child_runs_first;
extern unsigned int sysctl_sched_features; extern unsigned int sysctl_sched_features;
extern unsigned int sysctl_sched_migration_cost; extern unsigned int sysctl_sched_migration_cost;
int sched_nr_latency_handler(struct ctl_table *table, int write,
struct file *file, void __user *buffer, size_t *length,
loff_t *ppos);
#endif #endif
extern unsigned int sysctl_sched_compat_yield; extern unsigned int sysctl_sched_compat_yield;
......
...@@ -211,7 +211,7 @@ static int sched_debug_show(struct seq_file *m, void *v) ...@@ -211,7 +211,7 @@ static int sched_debug_show(struct seq_file *m, void *v)
#define PN(x) \ #define PN(x) \
SEQ_printf(m, " .%-40s: %Ld.%06ld\n", #x, SPLIT_NS(x)) SEQ_printf(m, " .%-40s: %Ld.%06ld\n", #x, SPLIT_NS(x))
PN(sysctl_sched_latency); PN(sysctl_sched_latency);
PN(sysctl_sched_nr_latency); PN(sysctl_sched_min_granularity);
PN(sysctl_sched_wakeup_granularity); PN(sysctl_sched_wakeup_granularity);
PN(sysctl_sched_batch_wakeup_granularity); PN(sysctl_sched_batch_wakeup_granularity);
PN(sysctl_sched_child_runs_first); PN(sysctl_sched_child_runs_first);
......
...@@ -35,16 +35,21 @@ ...@@ -35,16 +35,21 @@
const_debug unsigned int sysctl_sched_latency = 20000000ULL; const_debug unsigned int sysctl_sched_latency = 20000000ULL;
/* /*
* After fork, child runs first. (default) If set to 0 then * Minimal preemption granularity for CPU-bound tasks:
* parent will (try to) run first. * (default: 1 msec, units: nanoseconds)
*/ */
const_debug unsigned int sysctl_sched_child_runs_first = 1; const_debug unsigned int sysctl_sched_min_granularity = 1000000ULL;
/* /*
* Minimal preemption granularity for CPU-bound tasks: * is kept at sysctl_sched_latency / sysctl_sched_min_granularity
* (default: 2 msec, units: nanoseconds)
*/ */
const_debug unsigned int sysctl_sched_nr_latency = 20; const_debug unsigned int sched_nr_latency = 20;
/*
* After fork, child runs first. (default) If set to 0 then
* parent will (try to) run first.
*/
const_debug unsigned int sysctl_sched_child_runs_first = 1;
/* /*
* sys_sched_yield() compat mode * sys_sched_yield() compat mode
...@@ -212,6 +217,22 @@ static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq) ...@@ -212,6 +217,22 @@ static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
* Scheduling class statistics methods: * Scheduling class statistics methods:
*/ */
#ifdef CONFIG_SCHED_DEBUG
int sched_nr_latency_handler(struct ctl_table *table, int write,
struct file *filp, void __user *buffer, size_t *lenp,
loff_t *ppos)
{
int ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos);
if (ret || !write)
return ret;
sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
sysctl_sched_min_granularity);
return 0;
}
#endif
/* /*
* The idea is to set a period in which each task runs once. * The idea is to set a period in which each task runs once.
...@@ -224,7 +245,7 @@ static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq) ...@@ -224,7 +245,7 @@ static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
static u64 __sched_period(unsigned long nr_running) static u64 __sched_period(unsigned long nr_running)
{ {
u64 period = sysctl_sched_latency; u64 period = sysctl_sched_latency;
unsigned long nr_latency = sysctl_sched_nr_latency; unsigned long nr_latency = sched_nr_latency;
if (unlikely(nr_running > nr_latency)) { if (unlikely(nr_running > nr_latency)) {
period *= nr_running; period *= nr_running;
......
...@@ -235,11 +235,14 @@ static struct ctl_table kern_table[] = { ...@@ -235,11 +235,14 @@ static struct ctl_table kern_table[] = {
#ifdef CONFIG_SCHED_DEBUG #ifdef CONFIG_SCHED_DEBUG
{ {
.ctl_name = CTL_UNNUMBERED, .ctl_name = CTL_UNNUMBERED,
.procname = "sched_nr_latency", .procname = "sched_min_granularity_ns",
.data = &sysctl_sched_nr_latency, .data = &sysctl_sched_min_granularity,
.maxlen = sizeof(unsigned int), .maxlen = sizeof(unsigned int),
.mode = 0644, .mode = 0644,
.proc_handler = &proc_dointvec, .proc_handler = &sched_nr_latency_handler,
.strategy = &sysctl_intvec,
.extra1 = &min_sched_granularity_ns,
.extra2 = &max_sched_granularity_ns,
}, },
{ {
.ctl_name = CTL_UNNUMBERED, .ctl_name = CTL_UNNUMBERED,
...@@ -247,7 +250,7 @@ static struct ctl_table kern_table[] = { ...@@ -247,7 +250,7 @@ static struct ctl_table kern_table[] = {
.data = &sysctl_sched_latency, .data = &sysctl_sched_latency,
.maxlen = sizeof(unsigned int), .maxlen = sizeof(unsigned int),
.mode = 0644, .mode = 0644,
.proc_handler = &proc_dointvec_minmax, .proc_handler = &sched_nr_latency_handler,
.strategy = &sysctl_intvec, .strategy = &sysctl_intvec,
.extra1 = &min_sched_granularity_ns, .extra1 = &min_sched_granularity_ns,
.extra2 = &max_sched_granularity_ns, .extra2 = &max_sched_granularity_ns,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment