* [RESEND][PATCH v3 1/1] sched: Report the different kinds of imbalances in /proc/schedstat
2024-05-14 4:44 [RESEND][PATCH v3 0/1] Fix lb_imbalance reporting in /proc/schedstat Swapnil Sapkal
@ 2024-05-14 4:44 ` Swapnil Sapkal
2024-05-14 11:14 ` Chen Yu
0 siblings, 1 reply; 6+ messages in thread
From: Swapnil Sapkal @ 2024-05-14 4:44 UTC (permalink / raw)
To: linux-kernel, mingo, peterz
Cc: torvalds, juri.lelli, dietmar.eggemann, rostedt, mgorman, bristot,
vschneid, vincent.guittot, gautham.shenoy, sshegde,
kprateek.nayak, Swapnil Sapkal
In /proc/schedstat, lb_imbalance reports the sum of imbalances
discovered in sched domains with each call to sched_balance_rq(), which is
not very useful because lb_imbalance does not mention whether the imbalance
is due to load, utilization, nr_tasks or misfit_tasks. Remove this field
from /proc/schedstat.
Currently there is no field in /proc/schedstat to report different types
of imbalances. Introduce new fields in /proc/schedstat to report the
total imbalances in load, utilization, nr_tasks or misfit_tasks.
Added fields to /proc/schedstat:
- lb_imbalance_load: Total imbalance due to load.
- lb_imbalance_util: Total imbalance due to utilization.
- lb_imbalance_task: Total imbalance due to number of tasks.
- lb_imbalance_misfit: Total imbalance due to misfit tasks.
Reviewed-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Signed-off-by: Swapnil Sapkal <swapnil.sapkal@amd.com>
---
Documentation/scheduler/sched-stats.rst | 121 ++++++++++++++----------
include/linux/sched/topology.h | 5 +-
kernel/sched/fair.c | 21 +++-
kernel/sched/stats.c | 7 +-
4 files changed, 99 insertions(+), 55 deletions(-)
diff --git a/Documentation/scheduler/sched-stats.rst b/Documentation/scheduler/sched-stats.rst
index 7c2b16c4729d..0dd47bdd5608 100644
--- a/Documentation/scheduler/sched-stats.rst
+++ b/Documentation/scheduler/sched-stats.rst
@@ -5,11 +5,16 @@ Scheduler Statistics
Version 16 of schedstats changed the order of definitions within
'enum cpu_idle_type', which changed the order of [CPU_MAX_IDLE_TYPES]
columns in show_schedstat(). In particular the position of CPU_IDLE
-and __CPU_NOT_IDLE changed places. The size of the array is unchanged.
+and __CPU_NOT_IDLE changed places. The size of the enum is unchanged.
+Also stop reporting 'lb_imbalance' as it has no significance anymore
+and instead add more relevant fields namely 'lb_imbalance_load',
+'lb_imbalance_util', 'lb_imbalance_task' and 'lb_imbalance_misfit'.
Version 15 of schedstats dropped counters for some sched_yield:
yld_exp_empty, yld_act_empty and yld_both_empty. Otherwise, it is
-identical to version 14.
+identical to version 14. Details are available at
+
+ https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/scheduler/sched-stats.txt?id=1e1dbb259c79b
Version 14 of schedstats includes support for sched_domains, which hit the
mainline kernel in 2.6.20 although it is identical to the stats from version
@@ -26,7 +31,14 @@ cpus on the machine, while domain0 is the most tightly focused domain,
sometimes balancing only between pairs of cpus. At this time, there
are no architectures which need more than three domain levels. The first
field in the domain stats is a bit map indicating which cpus are affected
-by that domain.
+by that domain. Details are available at
+
+ https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/sched-stats.txt?id=b762f3ffb797c
+
+The schedstat documentation is maintained version 10 onwards and is not
+updated for version 11 and 12. The details for version 10 are available at
+
+ https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/sched-stats.txt?id=1da177e4c3f4
These fields are counters, and only increment. Programs which make use
of these will need to start with a baseline observation and then calculate
@@ -73,86 +85,93 @@ One of these is produced per domain for each cpu described. (Note that if
CONFIG_SMP is not defined, *no* domains are utilized and these lines
will not appear in the output.)
-domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
+domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
The first field is a bit mask indicating what cpus this domain operates over.
-The next 24 are a variety of sched_balance_rq() statistics in grouped into types
-of idleness (idle, busy, and newly idle):
+The next 33 are a variety of sched_balance_rq() statistics in grouped into types
+of idleness (busy, idle and newly idle):
1) # of times in this domain sched_balance_rq() was called when the
+ cpu was busy
+ 2) # of times in this domain sched_balance_rq() checked but found the
+ load did not require balancing when busy
+ 3) # of times in this domain sched_balance_rq() tried to move one or
+ more tasks and failed, when the cpu was busy
+ 4) Total imbalance in load when the cpu was busy
+ 5) Total imbalance in utilization when the cpu was busy
+ 6) Total imbalance in number of tasks when the cpu was busy
+ 7) Total imbalance due to misfit tasks when the cpu was busy
+ 8) # of times in this domain pull_task() was called when busy
+ 9) # of times in this domain pull_task() was called even though the
+ target task was cache-hot when busy
+ 10) # of times in this domain sched_balance_rq() was called but did not
+ find a busier queue while the cpu was busy
+ 11) # of times in this domain a busier queue was found while the cpu
+ was busy but no busier group was found
+
+ 12) # of times in this domain sched_balance_rq() was called when the
cpu was idle
- 2) # of times in this domain sched_balance_rq() checked but found
+ 13) # of times in this domain sched_balance_rq() checked but found
the load did not require balancing when the cpu was idle
- 3) # of times in this domain sched_balance_rq() tried to move one or
+ 14) # of times in this domain sched_balance_rq() tried to move one or
more tasks and failed, when the cpu was idle
- 4) sum of imbalances discovered (if any) with each call to
- sched_balance_rq() in this domain when the cpu was idle
- 5) # of times in this domain pull_task() was called when the cpu
+ 15) Total imbalance in load when the cpu was idle
+ 16) Total imbalance in utilization when the cpu was idle
+ 17) Total imbalance in number of tasks when the cpu was idle
+ 18) Total imbalance due to misfit tasks when the cpu was idle
+ 19) # of times in this domain pull_task() was called when the cpu
was idle
- 6) # of times in this domain pull_task() was called even though
+ 20) # of times in this domain pull_task() was called even though
the target task was cache-hot when idle
- 7) # of times in this domain sched_balance_rq() was called but did
+ 21) # of times in this domain sched_balance_rq() was called but did
not find a busier queue while the cpu was idle
- 8) # of times in this domain a busier queue was found while the
+ 22) # of times in this domain a busier queue was found while the
cpu was idle but no busier group was found
- 9) # of times in this domain sched_balance_rq() was called when the
- cpu was busy
- 10) # of times in this domain sched_balance_rq() checked but found the
- load did not require balancing when busy
- 11) # of times in this domain sched_balance_rq() tried to move one or
- more tasks and failed, when the cpu was busy
- 12) sum of imbalances discovered (if any) with each call to
- sched_balance_rq() in this domain when the cpu was busy
- 13) # of times in this domain pull_task() was called when busy
- 14) # of times in this domain pull_task() was called even though the
- target task was cache-hot when busy
- 15) # of times in this domain sched_balance_rq() was called but did not
- find a busier queue while the cpu was busy
- 16) # of times in this domain a busier queue was found while the cpu
- was busy but no busier group was found
- 17) # of times in this domain sched_balance_rq() was called when the
- cpu was just becoming idle
- 18) # of times in this domain sched_balance_rq() checked but found the
+ 23) # of times in this domain sched_balance_rq() was called when the
+ was just becoming idle
+ 24) # of times in this domain sched_balance_rq() checked but found the
load did not require balancing when the cpu was just becoming idle
- 19) # of times in this domain sched_balance_rq() tried to move one or more
+ 25) # of times in this domain sched_balance_rq() tried to move one or more
tasks and failed, when the cpu was just becoming idle
- 20) sum of imbalances discovered (if any) with each call to
- sched_balance_rq() in this domain when the cpu was just becoming idle
- 21) # of times in this domain pull_task() was called when newly idle
- 22) # of times in this domain pull_task() was called even though the
+ 26) Total imbalance in load when the cpu was just becoming idle
+ 27) Total imbalance in utilization when the cpu was just becoming idle
+ 28) Total imbalance in number of tasks when the cpu was just becoming idle
+ 29) Total imbalance due to misfit tasks when the cpu was just becoming idle
+ 30) # of times in this domain pull_task() was called when newly idle
+ 31) # of times in this domain pull_task() was called even though the
target task was cache-hot when just becoming idle
- 23) # of times in this domain sched_balance_rq() was called but did not
+ 32) # of times in this domain sched_balance_rq() was called but did not
find a busier queue while the cpu was just becoming idle
- 24) # of times in this domain a busier queue was found while the cpu
+ 33) # of times in this domain a busier queue was found while the cpu
was just becoming idle but no busier group was found
Next three are active_load_balance() statistics:
- 25) # of times active_load_balance() was called
- 26) # of times active_load_balance() tried to move a task and failed
- 27) # of times active_load_balance() successfully moved a task
+ 34) # of times active_load_balance() was called
+ 35) # of times active_load_balance() tried to move a task and failed
+ 36) # of times active_load_balance() successfully moved a task
Next three are sched_balance_exec() statistics:
- 28) sbe_cnt is not used
- 29) sbe_balanced is not used
- 30) sbe_pushed is not used
+ 37) sbe_cnt is not used
+ 38) sbe_balanced is not used
+ 39) sbe_pushed is not used
Next three are sched_balance_fork() statistics:
- 31) sbf_cnt is not used
- 32) sbf_balanced is not used
- 33) sbf_pushed is not used
+ 40) sbf_cnt is not used
+ 41) sbf_balanced is not used
+ 42) sbf_pushed is not used
Next three are try_to_wake_up() statistics:
- 34) # of times in this domain try_to_wake_up() awoke a task that
+ 43) # of times in this domain try_to_wake_up() awoke a task that
last ran on a different cpu in this domain
- 35) # of times in this domain try_to_wake_up() moved a task to the
+ 44) # of times in this domain try_to_wake_up() moved a task to the
waking cpu because it was cache-cold on its own cpu anyway
- 36) # of times in this domain try_to_wake_up() started passive balancing
+ 45) # of times in this domain try_to_wake_up() started passive balancing
/proc/<pid>/schedstat
---------------------
diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h
index c8fe9bab981b..15685c40a713 100644
--- a/include/linux/sched/topology.h
+++ b/include/linux/sched/topology.h
@@ -114,7 +114,10 @@ struct sched_domain {
unsigned int lb_count[CPU_MAX_IDLE_TYPES];
unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
- unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
+ unsigned int lb_imbalance_load[CPU_MAX_IDLE_TYPES];
+ unsigned int lb_imbalance_util[CPU_MAX_IDLE_TYPES];
+ unsigned int lb_imbalance_task[CPU_MAX_IDLE_TYPES];
+ unsigned int lb_imbalance_misfit[CPU_MAX_IDLE_TYPES];
unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a19ea290b790..31077a702e78 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -11238,6 +11238,25 @@ static int should_we_balance(struct lb_env *env)
return group_balance_cpu(sg) == env->dst_cpu;
}
+static void update_lb_imbalance_stat(struct lb_env *env, struct sched_domain *sd,
+ enum cpu_idle_type idle)
+{
+ switch (env->migration_type) {
+ case migrate_load:
+ schedstat_add(sd->lb_imbalance_load[idle], env->imbalance);
+ break;
+ case migrate_util:
+ schedstat_add(sd->lb_imbalance_util[idle], env->imbalance);
+ break;
+ case migrate_task:
+ schedstat_add(sd->lb_imbalance_task[idle], env->imbalance);
+ break;
+ case migrate_misfit:
+ schedstat_add(sd->lb_imbalance_misfit[idle], env->imbalance);
+ break;
+ }
+}
+
/*
* Check this_cpu to ensure it is balanced within domain. Attempt to move
* tasks if there is an imbalance.
@@ -11288,7 +11307,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq,
WARN_ON_ONCE(busiest == env.dst_rq);
- schedstat_add(sd->lb_imbalance[idle], env.imbalance);
+ update_lb_imbalance_stat(&env, sd, idle);
env.src_cpu = busiest->cpu;
env.src_rq = busiest;
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index 78e48f5426ee..a02bc9db2f1c 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -151,11 +151,14 @@ static int show_schedstat(struct seq_file *seq, void *v)
seq_printf(seq, "domain%d %*pb", dcount++,
cpumask_pr_args(sched_domain_span(sd)));
for (itype = 0; itype < CPU_MAX_IDLE_TYPES; itype++) {
- seq_printf(seq, " %u %u %u %u %u %u %u %u",
+ seq_printf(seq, " %u %u %u %u %u %u %u %u %u %u %u",
sd->lb_count[itype],
sd->lb_balanced[itype],
sd->lb_failed[itype],
- sd->lb_imbalance[itype],
+ sd->lb_imbalance_load[itype],
+ sd->lb_imbalance_util[itype],
+ sd->lb_imbalance_task[itype],
+ sd->lb_imbalance_misfit[itype],
sd->lb_gained[itype],
sd->lb_hot_gained[itype],
sd->lb_nobusyq[itype],
--
2.34.1
^ permalink raw reply related [flat|nested] 6+ messages in thread