Commit b762f3ff authored by Joachim Deguara's avatar Joachim Deguara Committed by Ingo Molnar
Browse files

[PATCH] sched: update Documentation/sched-stats.txt 



While learning about schedstats I found that the documentation in the tree
is old.  I updated it and found some interesting stuff like schedstats
version 14 is the same as version and version 13 never saw a kernel
release!  Also there are 6 fields in the current schedstats that are not
used anymore.  Nick had made them irrelevant in commit
476d139c but never removed them.

Thanks to Rick's perl script who I borrowed some of the updated descriptions
from.

Signed-off-by: default avatarJoachim Deguara <joachim.deguara@amd.com>
Acked-by: default avatarNick Piggin <npiggin@suse.de>
Cc: Rick Lindsley <ricklind@us.ibm.com>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
parent f3373461

Documentation/sched-stats.txt

+99 −96
Original line number Diff line number Diff line 
Version 10 of schedstats includes support for sched_domains, which

hit the mainline kernel in 2.6.7.  Some counters make more sense to be

per-runqueue; other to be per-domain.  Note that domains (and their associated

information) will only be pertinent and available on machines utilizing

CONFIG_SMP.



In version 10 of schedstat, there is at least one level of domain

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

12 which was in the kernel from 2.6.13-2.6.19 (version 13 never saw a kernel

release).  Some counters make more sense to be per-runqueue; other to be

per-domain.  Note that domains (and their associated information) will only

be pertinent and available on machines utilizing CONFIG_SMP.



In version 14 of schedstat, there is at least one level of domain

statistics for each cpu listed, and there may well be more than one

domain.  Domains have no particular names in this implementation, but

the highest numbered one typically arbitrates balancing across all the
@@ -27,7 +28,7 @@ to write their own scripts, the fields are described here. 


CPU statistics

--------------

cpu<N> 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

cpu<N> 1 2 3 4 5 6 7 8 9 10 11 12



NOTE: In the sched_yield() statistics, the active queue is considered empty

    if it has only one process in it, since obviously the process calling
@@ -39,48 +40,20 @@ First four fields are sched_yield() statistics: 
     3) # of times just the expired queue was empty

     4) # of times sched_yield() was called



Next four are schedule() statistics:

     5) # of times the active queue had at least one other process on it

     6) # of times we switched to the expired queue and reused it

     7) # of times schedule() was called

     8) # of times schedule() left the processor idle



Next four are active_load_balance() statistics:

     9) # of times active_load_balance() was called

    10) # of times active_load_balance() caused this cpu to gain a task

    11) # of times active_load_balance() caused this cpu to lose a task

    12) # of times active_load_balance() tried to move a task and failed



Next three are try_to_wake_up() statistics:

    13) # of times try_to_wake_up() was called

    14) # of times try_to_wake_up() successfully moved the awakening task

    15) # of times try_to_wake_up() attempted to move the awakening task



Next two are wake_up_new_task() statistics:

    16) # of times wake_up_new_task() was called

    17) # of times wake_up_new_task() successfully moved the new task



Next one is a sched_migrate_task() statistic:

    18) # of times sched_migrate_task() was called

Next three are schedule() statistics:

     5) # of times we switched to the expired queue and reused it

     6) # of times schedule() was called

     7) # of times schedule() left the processor idle



Next one is a sched_balance_exec() statistic:

    19) # of times sched_balance_exec() was called

Next two are try_to_wake_up() statistics:

     8) # of times try_to_wake_up() was called

     9) # of times try_to_wake_up() was called to wake up the local cpu



Next three are statistics describing scheduling latency:

    20) sum of all time spent running by tasks on this processor (in ms)

    21) sum of all time spent waiting to run by tasks on this processor (in ms)

    22) # of tasks (not necessarily unique) given to the processor



The last six are statistics dealing with pull_task():

    23) # of times pull_task() moved a task to this cpu when newly idle

    24) # of times pull_task() stole a task from this cpu when another cpu

	was newly idle

    25) # of times pull_task() moved a task to this cpu when idle

    26) # of times pull_task() stole a task from this cpu when another cpu

	was idle

    27) # of times pull_task() moved a task to this cpu when busy

    28) # of times pull_task() stole a task from this cpu when another cpu

	was busy

    10) sum of all time spent running by tasks on this processor (in jiffies)

    11) sum of all time spent waiting to run by tasks on this processor (in

        jiffies)

    12) # of timeslices run on this cpu





Domain statistics
@@ -89,65 +62,95 @@ 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> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

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



The first field is a bit mask indicating what cpus this domain operates over.



The next fifteen are a variety of load_balance() statistics:

The next 24 are a variety of load_balance() statistics in grouped into types

of idleness (idle, busy, and newly idle):



     1) # of times in this domain load_balance() was called when the cpu

	was idle

     2) # of times in this domain load_balance() was called when the cpu

	was busy

     3) # of times in this domain load_balance() was called when the cpu

	was just becoming idle

     4) # of times in this domain load_balance() tried to move one or more

	tasks and failed, when the cpu was idle

     5) # of times in this domain load_balance() tried to move one or more

	tasks and failed, when the cpu was busy

     6) # of times in this domain load_balance() tried to move one or more

	tasks and failed, when the cpu was just becoming idle

     7) sum of imbalances discovered (if any) with each call to

     1) # of times in this domain load_balance() was called when the

        cpu was idle

     2) # of times in this domain load_balance() checked but found

        the load did not require balancing when the cpu was idle

     3) # of times in this domain load_balance() 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

        load_balance() in this domain when the cpu was idle

     8) sum of imbalances discovered (if any) with each call to

     5) # 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

        the target task was cache-hot when idle

     7) # of times in this domain load_balance() 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

        cpu was idle but no busier group was found



     9) # of times in this domain load_balance() was called when the

        cpu was busy

    10) # of times in this domain load_balance() checked but found the

        load did not require balancing when busy

    11) # of times in this domain load_balance() 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

        load_balance() in this domain when the cpu was busy

     9) sum of imbalances discovered (if any) with each call to

    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 load_balance() 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 load_balance() was called when the

        cpu was just becoming idle

    18) # of times in this domain load_balance() checked but found the

        load did not require balancing when the cpu was just becoming idle

    19) # of times in this domain load_balance() 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

        load_balance() in this domain when the cpu was just becoming idle

    10) # of times in this domain load_balance() was called but did not find

	a busier queue while the cpu was idle

    11) # of times in this domain load_balance() was called but did not find

	a busier queue while the cpu was busy

    12) # of times in this domain load_balance() was called but did not find

	a busier queue while the cpu was just becoming idle

    13) # of times in this domain a busier queue was found while the cpu was

	idle but no busier group was found

    14) # of times in this domain a busier queue was found while the cpu was

	busy but no busier group was found

    15) # of times in this domain a busier queue was found while the cpu was

	just becoming idle but no busier group was found



Next two are sched_balance_exec() statistics:

    17) # of times in this domain sched_balance_exec() successfully pushed

	a task to a new cpu

    18) # of times in this domain sched_balance_exec() tried but failed to

	push a task to a new cpu



Next two are try_to_wake_up() statistics:

    19) # of times in this domain try_to_wake_up() tried to move a task based

	on affinity and cache warmth

    20) # of times in this domain try_to_wake_up() tried to move a task based

	on load balancing

    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

        target task was cache-hot when just becoming idle

    23) # of times in this domain load_balance() 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

        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



   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



   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



   Next three are try_to_wake_up() statistics:

    34) # 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

        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



/proc/<pid>/schedstat

----------------

schedstats also adds a new /proc/<pid/schedstat file to include some of

the same information on a per-process level.  There are three fields in

this file correlating to fields 20, 21, and 22 in the CPU fields, but

they only apply for that process.

this file correlating for that process to:

     1) time spent on the cpu

     2) time spent waiting on a runqueue

     3) # of timeslices run on this cpu



A program could be easily written to make use of these extra fields to

report on how well a particular process or set of processes is faring

under the scheduler's policies.  A simple version of such a program is

available at

    http://eaglet.rain.com/rick/linux/schedstat/v10/latency.c
 
    http://eaglet.rain.com/rick/linux/schedstat/v12/latency.c

CRA Git | Maintained and supported by SUSTech CRA and CCSE