Include group statistics in /proc/sched_debug.
Since the statistics of a group entity isn't exported directly from the
kernel, it becomes difficult to obtain some of the group statistics.
For example, the current method to obtain exec time of a group entity
is not always accurate. One has to read the exec times of all
the tasks(/proc/<pid>/sched) in the group and add them. This method
fails (or becomes difficult) if we want to collect stats of a group over
a duration where tasks get created and terminated.
This patch makes it easier to obtain group stats by directly including
them in /proc/sched_debug. Stats like group exec time would help user
programs (like LTP) to accurately measure the group fairness.
Signed-off-by: Bharata B Rao <[email protected]>
CC: Peter Zijlstra <[email protected]>
CC: Ingo Molnar <[email protected]>
CC: Srivatsa Vaddagiri <[email protected]>
---
kernel/sched_debug.c | 67 ++++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 66 insertions(+), 1 deletion(-)
--- a/kernel/sched_debug.c
+++ b/kernel/sched_debug.c
@@ -53,6 +53,70 @@ static unsigned long nsec_low(unsigned l
#define SPLIT_NS(x) nsec_high(x), nsec_low(x)
+#ifdef CONFIG_GROUP_SCHED
+
+static void print_group_stats(struct seq_file *m, struct task_group *tg,
+ int cpu)
+{
+ struct sched_entity *se = tg->se[cpu];
+
+#if defined(CONFIG_CGROUP_SCHED)
+ char path[128] = "";
+ struct cgroup *cgroup = NULL;
+
+ if (tg)
+ cgroup = tg->css.cgroup;
+
+ if (cgroup)
+ cgroup_path(cgroup, path, sizeof(path));
+
+ SEQ_printf(m, "\ngroup[%d]:%s\n", cpu, path);
+#else
+ SEQ_printf(m, "\ngroup[%d]:\n", cpu);
+#endif
+#define P(F) \
+ SEQ_printf(m, " .%-30s: %lld\n", #F, (long long)F)
+#define PN(F) \
+ SEQ_printf(m, " .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)F))
+
+ PN(se->exec_start);
+ PN(se->vruntime);
+ PN(se->sum_exec_runtime);
+#ifdef CONFIG_SCHEDSTATS
+ PN(se->wait_start);
+ PN(se->sleep_start);
+ PN(se->block_start);
+ PN(se->sleep_max);
+ PN(se->block_max);
+ PN(se->exec_max);
+ PN(se->slice_max);
+ PN(se->wait_max);
+ PN(se->wait_sum);
+ P(se->wait_count);
+#endif
+ P(se->load.weight);
+#undef PN
+#undef P
+}
+
+static void print_tg_stats(struct seq_file *m, int cpu)
+{
+ struct task_group *tg;
+
+ rcu_read_lock();
+ list_for_each_entry(tg, &task_groups, list) {
+ if (tg->se[cpu])
+ print_group_stats(m, tg, cpu);
+ }
+ rcu_read_unlock();
+}
+#else
+static void print_tg_stats(struct seq_file *m, int cpu)
+{
+ return;
+}
+#endif
+
static void
print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
{
@@ -264,6 +328,7 @@ static void print_cpu(struct seq_file *m
print_rt_stats(m, cpu);
print_rq(m, rq, cpu);
+ print_tg_stats(m, cpu);
}
static int sched_debug_show(struct seq_file *m, void *v)
@@ -271,7 +336,7 @@ static int sched_debug_show(struct seq_f
u64 now = ktime_to_ns(ktime_get());
int cpu;
- SEQ_printf(m, "Sched Debug Version: v0.07, %s %.*s\n",
+ SEQ_printf(m, "Sched Debug Version: v0.08, %s %.*s\n",
init_utsname()->release,
(int)strcspn(init_utsname()->version, " "),
init_utsname()->version);
--
An extract of /proc/sched_debug showing group stats obtained from
this patch:
group[1]:/3/a/1
.se->exec_start : 256484.781577
.se->vruntime : 12868.176994
.se->sum_exec_runtime : 3243.669709
.se->wait_start : 0.000000
.se->sleep_start : 0.000000
.se->block_start : 0.000000
.se->sleep_max : 0.000000
.se->block_max : 0.000000
.se->exec_max : 1.002095
.se->slice_max : 13.997073
.se->wait_max : 67.978322
.se->wait_sum : 7141.676906
.se->wait_count : 203
.se->load.weight : 255
group[1]:/3/a
.se->exec_start : 256484.781577
.se->vruntime : 18649.700858
.se->sum_exec_runtime : 3302.827997
.se->wait_start : 0.000000
.se->sleep_start : 0.000000
.se->block_start : 0.000000
.se->sleep_max : 0.000000
.se->block_max : 0.000000
.se->exec_max : 1.009411
.se->slice_max : 7.015594
.se->wait_max : 67.978322
.se->wait_sum : 7082.536617
.se->wait_count : 214
.se->load.weight : 255
group[1]:/3
.se->exec_start : 256484.781577
.se->vruntime : 224861.311011
.se->sum_exec_runtime : 3302.827997
.se->wait_start : 0.000000
.se->sleep_start : 0.000000
.se->block_start : 0.000000
.se->sleep_max : 0.000000
.se->block_max : 0.000000
.se->exec_max : 1.002095
.se->slice_max : 19.007546
.se->wait_max : 67.978322
.se->wait_sum : 7082.536617
.se->wait_count : 214
.se->load.weight : 24
Regards,
Bharata.
On Mon, 2008-11-10 at 14:53 +0530, Bharata B Rao wrote:
> Include group statistics in /proc/sched_debug.
>
> Since the statistics of a group entity isn't exported directly from the
> kernel, it becomes difficult to obtain some of the group statistics.
> For example, the current method to obtain exec time of a group entity
> is not always accurate. One has to read the exec times of all
> the tasks(/proc/<pid>/sched) in the group and add them. This method
> fails (or becomes difficult) if we want to collect stats of a group over
> a duration where tasks get created and terminated.
>
> This patch makes it easier to obtain group stats by directly including
> them in /proc/sched_debug. Stats like group exec time would help user
> programs (like LTP) to accurately measure the group fairness.
>
> Signed-off-by: Bharata B Rao <[email protected]>
> CC: Peter Zijlstra <[email protected]>
> CC: Ingo Molnar <[email protected]>
> CC: Srivatsa Vaddagiri <[email protected]>
> ---
<snip patch>
> An extract of /proc/sched_debug showing group stats obtained from
> this patch:
>
> group[1]:/3/a/1
> .se->exec_start : 256484.781577
> .se->vruntime : 12868.176994
> .se->sum_exec_runtime : 3243.669709
> .se->wait_start : 0.000000
> .se->sleep_start : 0.000000
> .se->block_start : 0.000000
> .se->sleep_max : 0.000000
> .se->block_max : 0.000000
> .se->exec_max : 1.002095
> .se->slice_max : 13.997073
> .se->wait_max : 67.978322
> .se->wait_sum : 7141.676906
> .se->wait_count : 203
> .se->load.weight : 255
Why not simply add them to the cfs_rq[n]:/path sections we already have?
Subject: sched: cleanup debug info
From: Peter Zijlstra <[email protected]>
Date: Thu Oct 30 23:57:11 CET 2008
While looking over the sched_debug code I noticed that we printed the rq
schedstats for every cfs_rq, ammend this.
Also change nr_spead_over into an int, and fix a little buglet in
min_vruntime printing.
Signed-off-by: Peter Zijlstra <[email protected]>
---
kernel/sched.c | 2 +-
kernel/sched_debug.c | 41 +++++++++++++++++++++--------------------
2 files changed, 22 insertions(+), 21 deletions(-)
Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -399,7 +399,7 @@ struct cfs_rq {
*/
struct sched_entity *curr, *next, *last;
- unsigned long nr_spread_over;
+ unsigned int nr_spread_over;
#ifdef CONFIG_FAIR_GROUP_SCHED
struct rq *rq; /* cpu runqueue to which this cfs_rq is attached */
Index: linux-2.6/kernel/sched_debug.c
===================================================================
--- linux-2.6.orig/kernel/sched_debug.c
+++ linux-2.6/kernel/sched_debug.c
@@ -139,7 +139,7 @@ void print_cfs_rq(struct seq_file *m, in
last = __pick_last_entity(cfs_rq);
if (last)
max_vruntime = last->vruntime;
- min_vruntime = rq->cfs.min_vruntime;
+ min_vruntime = cfs_rq->min_vruntime;
rq0_min_vruntime = per_cpu(runqueues, 0).cfs.min_vruntime;
spin_unlock_irqrestore(&rq->lock, flags);
SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "MIN_vruntime",
@@ -156,26 +156,8 @@ void print_cfs_rq(struct seq_file *m, in
SPLIT_NS(spread0));
SEQ_printf(m, " .%-30s: %ld\n", "nr_running", cfs_rq->nr_running);
SEQ_printf(m, " .%-30s: %ld\n", "load", cfs_rq->load.weight);
-#ifdef CONFIG_SCHEDSTATS
-#define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n);
-
- P(yld_exp_empty);
- P(yld_act_empty);
- P(yld_both_empty);
- P(yld_count);
- P(sched_switch);
- P(sched_count);
- P(sched_goidle);
-
- P(ttwu_count);
- P(ttwu_local);
-
- P(bkl_count);
-
-#undef P
-#endif
- SEQ_printf(m, " .%-30s: %ld\n", "nr_spread_over",
+ SEQ_printf(m, " .%-30s: %d\n", "nr_spread_over",
cfs_rq->nr_spread_over);
#ifdef CONFIG_FAIR_GROUP_SCHED
#ifdef CONFIG_SMP
@@ -250,6 +232,25 @@ static void print_cpu(struct seq_file *m
#undef P
#undef PN
+#ifdef CONFIG_SCHEDSTATS
+#define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n);
+
+ P(yld_exp_empty);
+ P(yld_act_empty);
+ P(yld_both_empty);
+ P(yld_count);
+
+ P(sched_switch);
+ P(sched_count);
+ P(sched_goidle);
+
+ P(ttwu_count);
+ P(ttwu_local);
+
+ P(bkl_count);
+
+#undef P
+#endif
print_cfs_stats(m, cpu);
print_rt_stats(m, cpu);
* Peter Zijlstra <[email protected]> wrote:
> Subject: sched: cleanup debug info
> From: Peter Zijlstra <[email protected]>
> Date: Thu Oct 30 23:57:11 CET 2008
>
> While looking over the sched_debug code I noticed that we printed the rq
> schedstats for every cfs_rq, ammend this.
>
> Also change nr_spead_over into an int, and fix a little buglet in
> min_vruntime printing.
>
> Signed-off-by: Peter Zijlstra <[email protected]>
> ---
> kernel/sched.c | 2 +-
> kernel/sched_debug.c | 41 +++++++++++++++++++++--------------------
> 2 files changed, 22 insertions(+), 21 deletions(-)
applied to tip/sched/urgent, thanks Peter!
Ingo
On Mon, Nov 10, 2008 at 10:30:07AM +0100, Peter Zijlstra wrote:
> On Mon, 2008-11-10 at 14:53 +0530, Bharata B Rao wrote:
>
> > An extract of /proc/sched_debug showing group stats obtained from
> > this patch:
> >
> > group[1]:/3/a/1
> > .se->exec_start : 256484.781577
> > .se->vruntime : 12868.176994
> > .se->sum_exec_runtime : 3243.669709
> > .se->wait_start : 0.000000
> > .se->sleep_start : 0.000000
> > .se->block_start : 0.000000
> > .se->sleep_max : 0.000000
> > .se->block_max : 0.000000
> > .se->exec_max : 1.002095
> > .se->slice_max : 13.997073
> > .se->wait_max : 67.978322
> > .se->wait_sum : 7141.676906
> > .se->wait_count : 203
> > .se->load.weight : 255
>
> Why not simply add them to the cfs_rq[n]:/path sections we already have?
Makes sense. Here is the updated patch which puts group statistics
under cfs_rq[n]:/path section as you suggest.
Include group statistics in /proc/sched_debug.
Since the statistics of a group entity isn't exported directly from the
kernel, it becomes difficult to obtain some of the group statistics.
For example, the current method to obtain exec time of a group entity
is not always accurate. One has to read the exec times of all
the tasks(/proc/<pid>/sched) in the group and add them. This method
fails (or becomes difficult) if we want to collect stats of a group over
a duration where tasks get created and terminated.
This patch makes it easier to obtain group stats by directly including
them in /proc/sched_debug. Stats like group exec time would help user
programs (like LTP) to accurately measure the group fairness.
Signed-off-by: Bharata B Rao <[email protected]>
CC: Peter Zijlstra <[email protected]>
CC: Ingo Molnar <[email protected]>
CC: Srivatsa Vaddagiri <[email protected]>
---
kernel/sched_debug.c | 37 ++++++++++++++++++++++++++++++++++++-
1 file changed, 36 insertions(+), 1 deletion(-)
--- a/kernel/sched_debug.c
+++ b/kernel/sched_debug.c
@@ -53,6 +53,40 @@ static unsigned long nsec_low(unsigned l
#define SPLIT_NS(x) nsec_high(x), nsec_low(x)
+#ifdef CONFIG_FAIR_GROUP_SCHED
+static void print_cfs_group_stats(struct seq_file *m, int cpu,
+ struct task_group *tg)
+{
+ struct sched_entity *se = tg->se[cpu];
+ if (!se)
+ return;
+
+#define P(F) \
+ SEQ_printf(m, " .%-30s: %lld\n", #F, (long long)F)
+#define PN(F) \
+ SEQ_printf(m, " .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)F))
+
+ PN(se->exec_start);
+ PN(se->vruntime);
+ PN(se->sum_exec_runtime);
+#ifdef CONFIG_SCHEDSTATS
+ PN(se->wait_start);
+ PN(se->sleep_start);
+ PN(se->block_start);
+ PN(se->sleep_max);
+ PN(se->block_max);
+ PN(se->exec_max);
+ PN(se->slice_max);
+ PN(se->wait_max);
+ PN(se->wait_sum);
+ P(se->wait_count);
+#endif
+ P(se->load.weight);
+#undef PN
+#undef P
+}
+#endif
+
static void
print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
{
@@ -186,6 +220,7 @@ void print_cfs_rq(struct seq_file *m, in
#ifdef CONFIG_SMP
SEQ_printf(m, " .%-30s: %lu\n", "shares", cfs_rq->shares);
#endif
+ print_cfs_group_stats(m, cpu, cfs_rq->tg);
#endif
}
@@ -271,7 +306,7 @@ static int sched_debug_show(struct seq_f
u64 now = ktime_to_ns(ktime_get());
int cpu;
- SEQ_printf(m, "Sched Debug Version: v0.07, %s %.*s\n",
+ SEQ_printf(m, "Sched Debug Version: v0.08, %s %.*s\n",
init_utsname()->release,
(int)strcspn(init_utsname()->version, " "),
init_utsname()->version);
--
An example output of group stats from /proc/sched_debug:
cfs_rq[3]:/3/a/1
.exec_clock : 89.598007
.MIN_vruntime : 0.000001
.min_vruntime : 256300.970506
.max_vruntime : 0.000001
.spread : 0.000000
.spread0 : -25373.372248
.nr_running : 0
.load : 0
.yld_exp_empty : 0
.yld_act_empty : 0
.yld_both_empty : 0
.yld_count : 4474
.sched_switch : 0
.sched_count : 40507
.sched_goidle : 12686
.ttwu_count : 15114
.ttwu_local : 11950
.bkl_count : 67
.nr_spread_over : 0
.shares : 0
.se->exec_start : 113676.727170
.se->vruntime : 1592.612714
.se->sum_exec_runtime : 89.598007
.se->wait_start : 0.000000
.se->sleep_start : 0.000000
.se->block_start : 0.000000
.se->sleep_max : 0.000000
.se->block_max : 0.000000
.se->exec_max : 1.000282
.se->slice_max : 1.999750
.se->wait_max : 54.981093
.se->wait_sum : 217.610521
.se->wait_count : 50
.se->load.weight : 2
Regards,
Bharata.
* Bharata B Rao <[email protected]> wrote:
> > Why not simply add them to the cfs_rq[n]:/path sections we already
> > have?
>
> Makes sense. Here is the updated patch which puts group statistics
> under cfs_rq[n]:/path section as you suggest.
>
> Include group statistics in /proc/sched_debug.
applied to tip/sched/core, thanks Bharata!
Ingo