Currently sched_avg_update() (which updates rt_avg stats in the rq)
is getting called from scale_rt_power() (in the load balance context)
which doesn't take rq->lock.
Fix it by moving the sched_avg_update() to more appropriate
update_cpu_load() where the CFS load gets updated as well.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1282596171.2694.3.camel@sbsiddha-MOBL3>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
There is a scalability issue for current implementation of optimistic
mutex spin in the kernel. It is found on a 8 node 64 core Nehalem-EX
system (HT mode).
The intention of the optimistic mutex spin is to busy wait and spin on a
mutex if the owner of the mutex is running, in the hope that the mutex
will be released soon and be acquired, without the thread trying to
acquire mutex going to sleep. However, when we have a large number of
threads, contending for the mutex, we could have the mutex grabbed by
other thread, and then another ……, and we will keep spinning, wasting cpu
cycles and adding to the contention. One possible fix is to quit
spinning and put the current thread on wait-list if mutex lock switch to
a new owner while we spin, indicating heavy contention (see the patch
included).
I did some testing on a 8 socket Nehalem-EX system with a total of 64
cores. Using Ingo's test-mutex program that creates/delete files with 256
threads (http://lkml.org/lkml/2006/1/8/50) , I see the following speed up
after putting in the mutex spin fix:
./mutex-test V 256 10
Ops/sec
2.6.34 62864
With fix 197200
Repeating the test with Aim7 fserver workload, again there is a speed up
with the fix:
Jobs/min
2.6.34 91657
With fix 149325
To look at the impact on the distribution of mutex acquisition time, I
collected the mutex acquisition time on Aim7 fserver workload with some
instrumentation. The average acquisition time is reduced by 48% and
number of contentions reduced by 32%.
#contentions Time to acquire mutex (cycles)
2.6.34 72973 44765791
With fix 49210 23067129
The histogram of mutex acquisition time is listed below. The acquisition
time is in 2^bin cycles. We see that without the fix, the acquisition
time is mostly around 2^26 cycles. With the fix, we the distribution get
spread out a lot more towards the lower cycles, starting from 2^13.
However, there is an increase of the tail distribution with the fix at
2^28 and 2^29 cycles. It seems a small price to pay for the reduced
average acquisition time and also getting the cpu to do useful work.
Mutex acquisition time distribution (acq time = 2^bin cycles):
2.6.34 With Fix
bin #occurrence % #occurrence %
11 2 0.00% 120 0.24%
12 10 0.01% 790 1.61%
13 14 0.02% 2058 4.18%
14 86 0.12% 3378 6.86%
15 393 0.54% 4831 9.82%
16 710 0.97% 4893 9.94%
17 815 1.12% 4667 9.48%
18 790 1.08% 5147 10.46%
19 580 0.80% 6250 12.70%
20 429 0.59% 6870 13.96%
21 311 0.43% 1809 3.68%
22 255 0.35% 2305 4.68%
23 317 0.44% 916 1.86%
24 610 0.84% 233 0.47%
25 3128 4.29% 95 0.19%
26 63902 87.69% 122 0.25%
27 619 0.85% 286 0.58%
28 0 0.00% 3536 7.19%
29 0 0.00% 903 1.83%
30 0 0.00% 0 0.00%
I've done similar experiments with 2.6.35 kernel on smaller boxes as
well. One is on a dual-socket Westmere box (12 cores total, with HT).
Another experiment is on an old dual-socket Core 2 box (4 cores total, no
HT)
On the 12-core Westmere box, I see a 250% increase for Ingo's mutex-test
program with my mutex patch but no significant difference in aim7's
fserver workload.
On the 4-core Core 2 box, I see the difference with the patch for both
mutex-test and aim7 fserver are negligible.
So far, it seems like the patch has not caused regression on smaller
systems.
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: <stable@kernel.org> # .35.x
LKML-Reference: <1282168827.9542.72.camel@schen9-DESK>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (27 commits)
sched: Use correct macro to display sched_child_runs_first in /proc/sched_debug
sched: No need for bootmem special cases
sched: Revert nohz_ratelimit() for now
sched: Reduce update_group_power() calls
sched: Update rq->clock for nohz balanced cpus
sched: Fix spelling of sibling
sched, cpuset: Drop __cpuexit from cpu hotplug callbacks
sched: Fix the racy usage of thread_group_cputimer() in fastpath_timer_check()
sched: run_posix_cpu_timers: Don't check ->exit_state, use lock_task_sighand()
sched: thread_group_cputime: Simplify, document the "alive" check
sched: Remove the obsolete exit_state/signal hacks
sched: task_tick_rt: Remove the obsolete ->signal != NULL check
sched: __sched_setscheduler: Read the RLIMIT_RTPRIO value lockless
sched: Fix comments to make them DocBook happy
sched: Fix fix_small_capacity
powerpc: Exclude arch_sd_sibiling_asym_packing() on UP
powerpc: Enable asymmetric SMT scheduling on POWER7
sched: Add asymmetric group packing option for sibling domain
sched: Fix capacity calculations for SMT4
sched: Change nohz idle load balancing logic to push model
...
As of commit dcce284 ("mm: Extend gfp masking to the page
allocator") and commit 7e85ee0 ("slab,slub: don't enable
interrupts during early boot"), the slab allocator makes
sure we don't attempt to sleep during boot.
Therefore, remove bootmem special cases from the scheduler
and use plain GFP_KERNEL instead.
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1279225102-2572-1-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Norbert reported that nohz_ratelimit() causes his laptop to burn about
4W (40%) extra. For now back out the change and see if we can adjust
the power management code to make better decisions.
Reported-by: Norbert Preining <preining@logic.at>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Mike Galbraith <efault@gmx.de>
Cc: Arjan van de Ven <arjan@infradead.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
GCC 4.4.1 on ARM has been observed to replace the while loop in
sched_avg_update with a call to uldivmod, resulting in the
following build failure at link-time:
kernel/built-in.o: In function `sched_avg_update':
kernel/sched.c:1261: undefined reference to `__aeabi_uldivmod'
kernel/sched.c:1261: undefined reference to `__aeabi_uldivmod'
make: *** [.tmp_vmlinux1] Error 1
This patch introduces a fake data hazard to the loop body to
prevent the compiler optimising the loop away.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Because cgroup_fork() is ran before sched_fork() [ from copy_process() ]
and the child's pid is not yet visible the child is pinned to its
cgroup. Therefore we can silence this warning.
A nicer solution would be moving cgroup_fork() to right after
dup_task_struct() and exclude PF_STARTING from task_subsys_state().
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Commit 3a101d05 (sched: adjust when cpu_active and cpuset
configurations are updated during cpu on/offlining) added
hotplug notifiers marked with __cpuexit; however, ia64 drops
text in __cpuexit during link unlike x86.
This means that functions which are referenced during init but used
only for cpu hot unplugging afterwards shouldn't be marked with
__cpuexit. Drop __cpuexit from those functions.
Reported-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Tony Luck <tony.luck@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <4C1FDF5B.1040301@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Commit e70971591 ("sched: Optimize unused cgroup configuration") introduced
an imbalanced scheduling bug.
If we do not use CGROUP, function update_h_load won't update h_load. When the
system has a large number of tasks far more than logical CPU number, the
incorrect cfs_rq[cpu]->h_load value will cause load_balance() to pull too
many tasks to the local CPU from the busiest CPU. So the busiest CPU keeps
going in a round robin. That will hurt performance.
The issue was found originally by a scientific calculation workload that
developed by Yanmin. With that commit, the workload performance drops
about 40%.
CPU before after
00 : 2 : 7
01 : 1 : 7
02 : 11 : 6
03 : 12 : 7
04 : 6 : 6
05 : 11 : 7
06 : 10 : 6
07 : 12 : 7
08 : 11 : 6
09 : 12 : 6
10 : 1 : 6
11 : 1 : 6
12 : 6 : 6
13 : 2 : 6
14 : 2 : 6
15 : 1 : 6
Reviewed-by: Yanmin zhang <yanmin.zhang@intel.com>
Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1276754893.9452.5442.camel@debian>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
In the new push model, all idle CPUs indeed go into nohz mode. There is
still the concept of idle load balancer (performing the load balancing
on behalf of all the idle cpu's in the system). Busy CPU kicks the nohz
balancer when any of the nohz CPUs need idle load balancing.
The kickee CPU does the idle load balancing on behalf of all idle CPUs
instead of the normal idle balance.
This addresses the below two problems with the current nohz ilb logic:
* the idle load balancer continued to have periodic ticks during idle and
wokeup frequently, even though it did not have any rebalancing to do on
behalf of any of the idle CPUs.
* On x86 and CPUs that have APIC timer stoppage on idle CPUs, this
periodic wakeup can result in a periodic additional interrupt on a CPU
doing the timer broadcast.
Also currently we are migrating the unpinned timers from an idle to the cpu
doing idle load balancing (when all the cpus in the system are idle,
there is no idle load balancing cpu and timers get added to the same idle cpu
where the request was made. So the existing optimization works only on semi idle
system).
And In semi idle system, we no longer have periodic ticks on the idle load
balancer CPU. Using that cpu will add more delays to the timers than intended
(as that cpu's timer base may not be uptodate wrt jiffies etc). This was
causing mysterious slowdowns during boot etc.
For now, in the semi idle case, use the nearest busy cpu for migrating timers
from an idle cpu. This is good for power-savings anyway.
Signed-off-by: Venkatesh Pallipadi <venki@google.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1274486981.2840.46.camel@sbs-t61.sc.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
tickless idle has a negative side effect on update_cpu_load(), which
in turn can affect load balancing behavior.
update_cpu_load() is supposed to be called every tick, to keep track
of various load indicies. With tickless idle, there are no scheduler
ticks called on the idle CPUs. Idle CPUs may still do load balancing
(with idle_load_balance CPU) using the stale cpu_load. It will also
cause problems when all CPUs go idle for a while and become active
again. In this case loads would not degrade as expected.
This is how rq->nr_load_updates change looks like under different
conditions:
<cpu_num> <nr_load_updates change>
All CPUS idle for 10 seconds (HZ=1000)
0 1621
10 496
11 139
12 875
13 1672
14 12
15 21
1 1472
2 2426
3 1161
4 2108
5 1525
6 701
7 249
8 766
9 1967
One CPU busy rest idle for 10 seconds
0 10003
10 601
11 95
12 966
13 1597
14 114
15 98
1 3457
2 93
3 6679
4 1425
5 1479
6 595
7 193
8 633
9 1687
All CPUs busy for 10 seconds
0 10026
10 10026
11 10026
12 10026
13 10025
14 10025
15 10025
1 10026
2 10026
3 10026
4 10026
5 10026
6 10026
7 10026
8 10026
9 10026
That is update_cpu_load works properly only when all CPUs are busy.
If all are idle, all the CPUs get way lower updates. And when few
CPUs are busy and rest are idle, only busy and ilb CPU does proper
updates and rest of the idle CPUs will do lower updates.
The patch keeps track of when a last update was done and fixes up
the load avg based on current time.
On one of my test system SPECjbb with warehouse 1..numcpus, patch
improves throughput numbers by ~1% (average of 6 runs). On another
test system (with different domain hierarchy) there is no noticable
change in perf.
Signed-off-by: Venkatesh Pallipadi <venki@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <AANLkTilLtDWQsAUrIxJ6s04WTgmw9GuOODc5AOrYsaR5@mail.gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
- Contrary to what 6d558c3a says, there is no need to reload
prev = rq->curr after the context switch. You always schedule
back to where you came from, prev must be equal to current
even if cpu/rq was changed.
- This also means reacquire_kernel_lock() can use prev instead
of current.
- No need to reassign switch_count if reacquire_kernel_lock()
reports need_resched(), we can just move the initial assignment
down, under the "need_resched_nonpreemptible:" label.
- Try to update the comment after context_switch().
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20100519125711.GA30199@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
For people who otherwise get to write: cpu_clock(smp_processor_id()),
there is now: local_clock().
Also, as per suggestion from Andrew, provide some documentation on
the various clock interfaces, and minimize the unsigned long long vs
u64 mess.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jens Axboe <jaxboe@fusionio.com>
LKML-Reference: <1275052414.1645.52.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Concurrency managed workqueue needs to know when workers are going to
sleep and waking up. Using these two hooks, cmwq keeps track of the
current concurrency level and throttles execution of new works if it's
too high and wakes up another worker from the sleep hook if it becomes
too low.
This patch introduces PF_WQ_WORKER to identify workqueue workers and
adds the following two hooks.
* wq_worker_waking_up(): called when a worker is woken up.
* wq_worker_sleeping(): called when a worker is going to sleep and may
return a pointer to a local task which should be woken up. The
returned task is woken up using try_to_wake_up_local() which is
simplified ttwu which is called under rq lock and can only wake up
local tasks.
Both hooks are currently defined as noop in kernel/workqueue_sched.h.
Later cmwq implementation will replace them with proper
implementation.
These hooks are hard coded as they'll always be enabled.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Ingo Molnar <mingo@elte.hu>
Factor ttwu_activate() and ttwu_woken_up() out of try_to_wake_up().
The factoring out doesn't affect try_to_wake_up() much
code-generation-wise. Depending on configuration options, it ends up
generating the same object code as before or slightly different one
due to different register assignment.
This is to help future implementation of try_to_wake_up_local().
Mike Galbraith suggested rename to ttwu_post_activation() from
ttwu_woken_up() and comment update in try_to_wake_up().
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Ingo Molnar <mingo@elte.hu>
