On Wed, Jul 18, 2007 at 11:11:30AM +0200, Joachim Deguara wrote: > 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 476d139c218e44e045e4bc6d4cc02b010b343939 > but never removed them. > > Thanks to Rick's perl script who I borrowed some of the updated descriptions > from.
Ah, thanks, I actually didn't realise there was such good documentation there. Patch looks good. BTW. I have a simple program to do a basic statistical summary of the multiprocessor balancing if you are interested and haven't seen it. Acked-by: Nick Piggin <[EMAIL PROTECTED]> > > -Joachim > > -- > Updating schedstats documentation from version 10 to 14. > > Signed-off-by: Joachim Deguara <[EMAIL PROTECTED]> > > Index: kernel/Documentation/sched-stats.txt > =================================================================== > --- kernel.orig/Documentation/sched-stats.txt > +++ kernel/Documentation/sched-stats.txt > @@ -1,10 +1,11 @@ > -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. > +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 10 of schedstat, there is at least one level of domain > +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 a > > 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() stat > 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 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_migrate_task() statistic: > - 18) # of times sched_migrate_task() was called > - > -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 > 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: > - > - 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 > - load_balance() in this domain when the cpu was idle > - 8) 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 > - 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 > +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() 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 > + 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 > + 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 > + 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 > - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
