I suspect a feed-back loop between cpuidle and the aperf/mperf
cpu_power bits, where when we have idle C-states lower the ratio,
which leads to lower cpu_power and then less load, which generates
more idle time, etc..
Put in a knob to disable it.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Provide an ach specific hook for cpufreq based scaling of
cpu_power.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
[ego@in.ibm.com: spotting bugs]
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make the idle balancer more agressive, to improve a
x264 encoding workload provided by Jason Garrett-Glaser:
NEXT_BUDDY NO_LB_BIAS
encoded 600 frames, 252.82 fps, 22096.60 kb/s
encoded 600 frames, 250.69 fps, 22096.60 kb/s
encoded 600 frames, 245.76 fps, 22096.60 kb/s
NO_NEXT_BUDDY LB_BIAS
encoded 600 frames, 344.44 fps, 22096.60 kb/s
encoded 600 frames, 346.66 fps, 22096.60 kb/s
encoded 600 frames, 352.59 fps, 22096.60 kb/s
NO_NEXT_BUDDY NO_LB_BIAS
encoded 600 frames, 425.75 fps, 22096.60 kb/s
encoded 600 frames, 425.45 fps, 22096.60 kb/s
encoded 600 frames, 422.49 fps, 22096.60 kb/s
Peter pointed out that this is better done via newidle_idx,
not via LB_BIAS, newidle balancing should look for where
there is load _now_, not where there was load 2 ticks ago.
Worst-case latencies are improved as well as no buddies
means less vruntime spread. (as per prior lkml discussions)
This change improves kbuild-peak parallelism as well.
Reported-by: Jason Garrett-Glaser <darkshikari@gmail.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1253011667.9128.16.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When merging select_task_rq_fair() and sched_balance_self() we lost
the use of wake_idx, restore that and set them to 0 to make wake
balancing more aggressive.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
While merging select_task_rq_fair() and sched_balance_self() I made
a mistake that leads to testing the wrong task affinty.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
for_each_domain() uses RCU to serialize the sched_domains, except
it doesn't actually use rcu_read_lock() and instead relies on
disabling preemption -> FAIL.
XXX: audit other sched_domain code.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
One of the problems of power-saving balancing is that under certain
scenarios it is too slow and allows tons of real work to pile up.
Avoid this by ignoring the powersave stuff when there's real work
to be done.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The problem with wake_idle() is that is doesn't respect things like
cpu_power, which means it doesn't deal well with SMT nor the recent
RT interaction.
To cure this, it needs to do what sched_balance_self() does, which
leads to the possibility of merging select_task_rq_fair() and
sched_balance_self().
Modify sched_balance_self() to:
- update_shares() when walking up the domain tree,
(it only called it for the top domain, but it should
have done this anyway), which allows us to remove
this ugly bit from try_to_wake_up().
- do wake_affine() on the smallest domain that contains
both this (the waking) and the prev (the wakee) cpu for
WAKE invocations.
Then use the top-down balance steps it had to replace wake_idle().
This leads to the dissapearance of SD_WAKE_BALANCE and
SD_WAKE_IDLE_FAR, with SD_WAKE_IDLE replaced with SD_BALANCE_WAKE.
SD_WAKE_AFFINE needs SD_BALANCE_WAKE to be effective.
Touch all topology bits to replace the old with new SD flags --
platforms might need re-tuning, enabling SD_BALANCE_WAKE
conditionally on a NUMA distance seems like a good additional
feature, magny-core and small nehalem systems would want this
enabled, systems with slow interconnects would not.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We're going to want to drop rq->lock in try_to_wake_up() for a
longer period of time, however we also want to deal with concurrent
waking of the same task, which is currently handled by holding
rq->lock.
So introduce a new TASK state, namely TASK_WAKING, which indicates
someone is already waking the task (other wakers will fail p->state
& state).
We also keep preemption disabled over the whole ttwu().
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Rather ugly patch to fully place the sched_balance_self() code
inside the fair class.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Move the sched_balance_self() code into sched_fair.c
This facilitates the merger of sched_balance_self() and
sched_fair::select_task_rq().
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
In preparation to other code movement, move weighted_cpuload(),
source_load() and target_load() before the class includes.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add a NEXT_BUDDY feature flag to aid in debugging.
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
It consists of two conditions, split them out in separate toggles
so we can test them independently.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'tracing-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (105 commits)
ring-buffer: only enable ring_buffer_swap_cpu when needed
ring-buffer: check for swapped buffers in start of committing
tracing: report error in trace if we fail to swap latency buffer
tracing: add trace_array_printk for internal tracers to use
tracing: pass around ring buffer instead of tracer
tracing: make tracing_reset safe for external use
tracing: use timestamp to determine start of latency traces
tracing: Remove mentioning of legacy latency_trace file from documentation
tracing/filters: Defer pred allocation, fix memory leak
tracing: remove users of tracing_reset
tracing: disable buffers and synchronize_sched before resetting
tracing: disable update max tracer while reading trace
tracing: print out start and stop in latency traces
ring-buffer: disable all cpu buffers when one finds a problem
ring-buffer: do not count discarded events
ring-buffer: remove ring_buffer_event_discard
ring-buffer: fix ring_buffer_read crossing pages
ring-buffer: remove unnecessary cpu_relax
ring-buffer: do not swap buffers during a commit
ring-buffer: do not reset while in a commit
...
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (64 commits)
sched: Fix sched::sched_stat_wait tracepoint field
sched: Disable NEW_FAIR_SLEEPERS for now
sched: Keep kthreads at default priority
sched: Re-tune the scheduler latency defaults to decrease worst-case latencies
sched: Turn off child_runs_first
sched: Ensure that a child can't gain time over it's parent after fork()
sched: enable SD_WAKE_IDLE
sched: Deal with low-load in wake_affine()
sched: Remove short cut from select_task_rq_fair()
sched: Turn on SD_BALANCE_NEWIDLE
sched: Clean up topology.h
sched: Fix dynamic power-balancing crash
sched: Remove reciprocal for cpu_power
sched: Try to deal with low capacity, fix update_sd_power_savings_stats()
sched: Try to deal with low capacity
sched: Scale down cpu_power due to RT tasks
sched: Implement dynamic cpu_power
sched: Add smt_gain
sched: Update the cpu_power sum during load-balance
sched: Add SD_PREFER_SIBLING
...
* 'perfcounters-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (60 commits)
perf tools: Avoid unnecessary work in directory lookups
perf stat: Clean up statistics calculations a bit more
perf stat: More advanced variance computation
perf stat: Use stddev_mean in stead of stddev
perf stat: Remove the limit on repeat
perf stat: Change noise calculation to use stddev
x86, perf_counter, bts: Do not allow kernel BTS tracing for now
x86, perf_counter, bts: Correct pointer-to-u64 casts
x86, perf_counter, bts: Fail if BTS is not available
perf_counter: Fix output-sharing error path
perf trace: Fix read_string()
perf trace: Print out in nanoseconds
perf tools: Seek to the end of the header area
perf trace: Fix parsing of perf.data
perf trace: Sample timestamps as well
perf_counter: Introduce new (non-)paranoia level to allow raw tracepoint access
perf trace: Sample the CPU too
perf tools: Work around strict aliasing related warnings
perf tools: Clean up warnings list in the Makefile
perf tools: Complete support for dynamic strings
...
* 'irq-threaded-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
genirq: Do not mask oneshot edge type interrupts
genirq: Support nested threaded irq handling
genirq: Add buslock support
genirq: Add oneshot support
* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (28 commits)
rcu: Move end of special early-boot RCU operation earlier
rcu: Changes from reviews: avoid casts, fix/add warnings, improve comments
rcu: Create rcutree plugins to handle hotplug CPU for multi-level trees
rcu: Remove lockdep annotations from RCU's _notrace() API members
rcu: Add #ifdef to suppress __rcu_offline_cpu() warning in !HOTPLUG_CPU builds
rcu: Add CPU-offline processing for single-node configurations
rcu: Add "notrace" to RCU function headers used by ftrace
rcu: Remove CONFIG_PREEMPT_RCU
rcu: Merge preemptable-RCU functionality into hierarchical RCU
rcu: Simplify rcu_pending()/rcu_check_callbacks() API
rcu: Use debugfs_remove_recursive() simplify code.
rcu: Merge per-RCU-flavor initialization into pre-existing macro
rcu: Fix online/offline indication for rcudata.csv trace file
rcu: Consolidate sparse and lockdep declarations in include/linux/rcupdate.h
rcu: Renamings to increase RCU clarity
rcu: Move private definitions from include/linux/rcutree.h to kernel/rcutree.h
rcu: Expunge lingering references to CONFIG_CLASSIC_RCU, optimize on !SMP
rcu: Delay rcu_barrier() wait until beginning of next CPU-hotunplug operation.
rcu: Fix typo in rcu_irq_exit() comment header
rcu: Make rcupreempt_trace.c look at offline CPUs
...
* 'core-printk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
printk: Fix "printk: Enable the use of more than one CON_BOOT (early console)"
printk: Restore previous console_loglevel when re-enabling logging
printk: Ensure that "console enabled" messages are printed on the console
printk: Enable the use of more than one CON_BOOT (early console)
* 'writeback' of git://git.kernel.dk/linux-2.6-block:
writeback: check for registered bdi in flusher add and inode dirty
writeback: add name to backing_dev_info
writeback: add some debug inode list counters to bdi stats
writeback: get rid of pdflush completely
writeback: switch to per-bdi threads for flushing data
writeback: move dirty inodes from super_block to backing_dev_info
writeback: get rid of generic_sync_sb_inodes() export
This enables us to track who does what and print info. Its main use
is catching dirty inodes on the default_backing_dev_info, so we can
fix that up.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
This weird perf trace output:
cc1-9943 [001] 2802.059479616: sched_stat_wait: task: as:9944 wait: 2801938766276 [ns]
Is caused by setting one component field of the delta to zero
a bit too early. Move it to later.
( Note, this does not affect the NEW_FAIR_SLEEPERS interactivity bug,
it's just a reporting bug in essence. )
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Nikos Chantziaras <realnc@arcor.de>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <4AA93D34.8040500@arcor.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Nikos Chantziaras and Jens Axboe reported that turning off
NEW_FAIR_SLEEPERS improves desktop interactivity visibly.
Nikos described his experiences the following way:
" With this setting, I can do "nice -n 19 make -j20" and
still have a very smooth desktop and watch a movie at
the same time. Various other annoyances (like the
"logout/shutdown/restart" dialog of KDE not appearing
at all until the background fade-out effect has finished)
are also gone. So this seems to be the single most
important setting that vastly improves desktop behavior,
at least here. "
Jens described it the following way, referring to a 10-seconds
xmodmap scheduling delay he was trying to debug:
" Then I tried switching NO_NEW_FAIR_SLEEPERS on, and then
I get:
Performance counter stats for 'xmodmap .xmodmap-carl':
9.009137 task-clock-msecs # 0.447 CPUs
18 context-switches # 0.002 M/sec
1 CPU-migrations # 0.000 M/sec
315 page-faults # 0.035 M/sec
0.020167093 seconds time elapsed
Woot! "
So disable it for now. In perf trace output i can see weird
delta timestamps:
cc1-9943 [001] 2802.059479616: sched_stat_wait: task: as:9944 wait: 2801938766276 [ns]
That nsec field is not supposed to be that large. More digging
is needed - but lets turn it off while the real bug is found.
Reported-by: Nikos Chantziaras <realnc@arcor.de>
Tested-by: Nikos Chantziaras <realnc@arcor.de>
Reported-by: Jens Axboe <jens.axboe@oracle.com>
Tested-by: Jens Axboe <jens.axboe@oracle.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <4AA93D34.8040500@arcor.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reduce the latency target from 20 msecs to 5 msecs.
Why? Larger latencies increase spread, which is good for scaling,
but bad for worst case latency.
We still have the ilog(nr_cpus) rule to scale up on bigger
server boxes.
Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1252486344.28645.18.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Set child_runs_first default to off.
It hurts 'optimal' make -j<NR_CPUS> workloads as make jobs
get preempted by child tasks, reducing parallelism.
Note, this patch might make existing races in user
applications more prominent than before - so breakages
might be bisected to this commit.
Child-runs-first is broken on SMP to begin with, and we
already had it off briefly in v2.6.23 so most of the
offenders ought to be fixed. Would be nice not to revert
this commit but fix those apps finally ...
Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1252486344.28645.18.camel@marge.simson.net>
[ made the sysctl independent of CONFIG_SCHED_DEBUG, in case
people want to work around broken apps. ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
A fork/exec load is usually "pass the baton", so the child
should never be placed behind the parent. With START_DEBIT we
make room for the new task, but with child_runs_first, that
room comes out of the _parent's_ hide. There's nothing to say
that the parent wasn't ahead of min_vruntime at fork() time,
which means that the "baton carrier", who is essentially the
parent in drag, can gain time and increase scheduling latencies
for waiters.
With NEW_FAIR_SLEEPERS + START_DEBIT + child_runs_first
enabled, we essentially pass the sleeper fairness off to the
child, which is fine, but if we don't base placement on the
parent's updated vruntime, we can end up compounding latency
woes if the child itself then does fork/exec. The debit
incurred at fork doesn't hurt the parent who is then going to
sleep and maybe exit, but the child who acquires the error
harms all comers.
This improves latencies of make -j<n> kernel build workloads.
Reported-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
wake_affine() would always fail under low-load situations where
both prev and this were idle, because adding a single task will
always be a significant imbalance, even if there's nothing
around that could balance it.
Deal with this by allowing imbalance when there's nothing you
can do about it.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
select_task_rq_fair() incorrectly skips the wake_affine()
logic, remove this.
When prev_cpu == this_cpu, the code jumps straight to the
wake_idle() logic, this doesn't give the wake_affine() logic
the chance to pin the task to this cpu.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
fix the following 'make includecheck' warning:
kernel/sysctl.c: linux/security.h is included more than once.
Signed-off-by: Jaswinder Singh Rajput <jaswinderrajput@gmail.com>
Signed-off-by: James Morris <jmorris@namei.org>
Since the ability to swap the cpu buffers adds a small overhead to
the recording of a trace, we only want to add it when needed.
Only the irqsoff and preemptoff tracers use this feature, and both are
not recommended for production kernels. This patch disables its use
when neither irqsoff nor preemptoff is configured.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Because the irqsoff tracer can swap an internal CPU buffer, it is possible
that a swap happens between the start of the write and before the committing
bit is set (the committing bit will disable swapping).
This patch adds a check for this and will fail the write if it detects it.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The irqsoff tracer will fail to swap the cpu buffer with the max
buffer if it preempts a commit. Instead of ignoring this, this patch
makes the tracer report it if the last max latency failed due to preempting
a current commit.
The output of the latency tracer will look like this:
# tracer: irqsoff
#
# irqsoff latency trace v1.1.5 on 2.6.31-rc5
# --------------------------------------------------------------------
# latency: 112 us, #1/1, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
# -----------------
# | task: -4281 (uid:0 nice:0 policy:0 rt_prio:0)
# -----------------
# => started at: save_args
# => ended at: __do_softirq
#
#
# _------=> CPU#
# / _-----=> irqs-off
# | / _----=> need-resched
# || / _---=> hardirq/softirq
# ||| / _--=> preempt-depth
# |||| /
# ||||| delay
# cmd pid ||||| time | caller
# \ / ||||| \ | /
bash-4281 1d.s6 265us : update_max_tr_single: Failed to swap buffers due to commit in progress
Note the latency time and the functions that disabled the irqs or preemption
will still be listed.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
This patch adds a trace_array_printk to allow a tracer to use the
trace_printk on its own trace array.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The latency tracers (irqsoff and wakeup) can swap trace buffers
on the fly. If an event is happening and has reserved data on one of
the buffers, and the latency tracer swaps the global buffer with the
max buffer, the result is that the event may commit the data to the
wrong buffer.
This patch changes the API to the trace recording to be recieve the
buffer that was used to reserve a commit. Then this buffer can be passed
in to the commit.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Reseting the trace buffer without first disabling the buffer and
waiting for any writers to complete, can corrupt the ring buffer.
This patch makes the external version of tracing_reset safe from
corruption by disabling the ring buffer and calling synchronize_sched.
This version can no longer be called from interrupt context. But all those
callers have been removed.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Currently the latency tracers reset the ring buffer. Unfortunately
if a commit is in process (due to a trace event), this can corrupt
the ring buffer. When this happens, the ring buffer will detect
the corruption and then permanently disable the ring buffer.
The bug does not crash the system, but it does prevent further tracing
after the bug is hit.
Instead of reseting the trace buffers, the timestamp of the start of
the trace is used instead. The buffers will still contain the previous
data, but the output will not count any data that is before the
timestamp of the trace.
Note, this only affects the static trace output (trace) and not the
runtime trace output (trace_pipe). The runtime trace output does not
make sense for the latency tracers anyway.
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The predicates of an event and their filter structure are allocated
when we create an event filter for the first time.
These objects must be created once but each time we come with a new
filter, we overwrite such pre-existing allocation, if any.
Thus, this patch checks if the filter has already been allocated
before going ahead.
Spotted-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
LKML-Reference: <4A9CB1BA.3060402@cn.fujitsu.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
