Linux-2.6.12-rc2v2.6.12-rc2

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!

1 files changed, 153 insertions, 0 deletions

diff --git a/Documentation/sched-stats.txt b/Documentation/sched-stats.txt
new file mode 100644
index 000000000000..6f72021aae51
--- /dev/null
+++ b/Documentation/sched-stats.txt
@@ -0,0 +1,153 @@
+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
+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
+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.
+
+These fields are counters, and only increment.  Programs which make use
+of these will need to start with a baseline observation and then calculate
+the change in the counters at each subsequent observation.  A perl script
+which does this for many of the fields is available at
+
+    http://eaglet.rain.com/rick/linux/schedstat/
+
+Note that any such script will necessarily be version-specific, as the main
+reason to change versions is changes in the output format.  For those wishing
+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
+
+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
+    sched_yield() is that process.
+
+First four fields are sched_yield() statistics:
+     1) # of times both the active and the expired queue were empty
+     2) # of times just the active queue was empty
+     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 one is a sched_balance_exec() statistic:
+    19) # of times sched_balance_exec() was called
+
+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
+
+
+Domain statistics
+-----------------
+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
+
+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
+
+
+/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.
+
+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