scheduler changes for v2.6.25
Linus, please pull the latest scheduler git tree for v2.6.25 from:
git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched.git
it contains various enhancements to the scheduler - find the full shortlog is below. 96 commits from 19 authors - scheduler developers have been busy again. :-/
The scheduling behavior of the kernel to normal users should not change over v2.6.24, but there are a good number of new features and enhancements under the hood:
- various instrumentation and debugging enhancements from Arjan van de Ven. LatencyTOP support in the scheduler is one example.
- Peter Zijlstra's RT time limit and RT throttling code for the RT scheduling class. Both audio and RT folks have asked for this.
- Paul E. McKenney's preemptible RCU code.
- refcount based CPU-hotplug rework by Gautham R Shenoy. (The "Lukewarm IQ detected in hotplug locking" printout is now gone, forever i hope! :)
- there's serious interest in running RT tasks on enterprise-class hardware, so Steven Rostedt and Gregory Haskins wrote a large number of enhancements to the RT scheduling class and load-balancer.
- Peter Zijlstra's high-resolution scheduler tick code. We still have a HZ tick running for a few other odd things, but it's dwindling now.
- Srivatsa Vaddagiri's next version of group scheduling. More accurate, more fair.
- make printk robust against self-recursion, and use stable timestamps for printk-timestamps.
- the softlockup detector will now detect hung, uninterruptible tasks automatically, and report them.
and a good number of other, smaller enhancements. Enjoy,
Ingo
------------------>
Alexey Dobriyan (1):
sched, futex: detach sched.h and futex.h
Arjan van de Ven (4):
debug: show being-loaded/being-unloaded indicator for modules
debug: track and print last unloaded module in the oops trace
sched: latencytop support
sched: keep total / count stats in addition to the max for
Dipankar Sarma (1):
Preempt-RCU: Use softirq instead of tasklets for
Dmitry Adamushko (3):
sched: no need for 'affine wakeup' balancing
sched: get rid of 'new_cpu' in try_to_wake_up()
sched: fix goto retry in pick_next_task_rt()
Gautham R Shenoy (3):
cpu-hotplug: refcount based cpu hotplug
cpu-hotplug: replace lock_cpu_hotplug() with get_online_cpus()
cpu-hotplug: replace per-subsystem mutexes with get_online_cpus()
Gregory Haskins (15):
sched: add RT-balance cpu-weight
sched: clean up this_rq use in kernel/sched_rt.c
sched: de-SCHED_OTHER-ize the RT path
sched: break out search for RT tasks
sched: RT balancing: include current CPU
sched: pre-route RT tasks on wakeup
sched: optimize RT affinity
sched: wake-balance fixes
sched: break out early if RT task cannot be migrated
sched: RT-balance, optimize
sched: add sched-domain roots
sched: only balance our RT tasks within our domain
sched: RT-balance, only adjust overload state when changing
sched: remove some old cpuset logic
sched: dynamically update the root-domain span/online maps
Guillaume Chazarain (2):
sched: fix rq->clock warps on frequency changes
sched: monitor clock underflows in /proc/sched_debug
Herbert Xu (1):
sched: do not do cond_resched() when CONFIG_PREEMPT
Ingo Molnar (25):
printk: make printk more robust by not allowing recursion
sched: fix CONFIG_PRINT_TIME's reliance on sched_clock()
sched: remove printk_clock()
sched: remove printk_clock references from ia64
cpu-hotplug: fix build on !CONFIG_SMP
softlockup: automatically detect hung TASK_UNINTERRUPTIBLE tasks
sched: clean up pick_next_highest_task_rt()
sched: clean up find_lock_lowest_rq()
sched: clean up overlong line in kernel/sched_debug.c
sched: clean up kernel/sched_rt.c
sched: remove rt_overload()
sched: remove leftover debugging
sched: clean up pull_rt_task()
sched: clean up schedule_balance_rt()
sched: fix sched_rt.c:join/leave_domain
sched: remove unused JIFFIES_TO_NS() macro
sched: style cleanup, #2
sched: add credits for RT balancing improvements
sched: reactivate fork balancing
sched: whitespace cleanups in topology.h
sched: documentation, whitespace fixes
sched: make PREEMPT_BKL the default
sched: remove the !PREEMPT_BKL code
softlockup: fix signedness
printk: use ktime_get()
Michal Schmidt (1):
sched: fix, always create kernel threads with normal priority
Mike Galbraith (1):
sched: pull_rt_task() cleanup
Nick Piggin (1):
sched: print backtrace of running tasks too
Paolo Ciarrocchi (1):
debug: clean up kernel/profile.c
Paul E. McKenney (5):
Preempt-RCU: reorganize RCU code into rcuclassic.c and rcupdate.c
Preempt-RCU: fix rcu_barrier for preemptive environment.
Preempt-RCU: implementation
Preempt-RCU: CPU Hotplug handling
Preempt-RCU: update RCU Documentation.
Pavel Emelyanov (1):
uids: merge multiple error paths in alloc_uid() into one
Pavel Machek (1):
timers: don't #error on higher HZ values
Peter Zijlstra (13):
sched: remove do_div() from __sched_slice()
sched: sched_rt_entity
sched: SCHED_FIFO/SCHED_RR watchdog timer
sched: high-res preemption tick
sched: rt time limit
sched: rt group scheduling
sched: rt throttling vs no_hz
hrtimer: clean up cpu->base locking tricks
hrtimer: fixup the HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback
hrtimer: unlock hrtimer_wakeup
sched: rt-group: reduce rescheduling
sched: rt-watchdog: fix .rlim_max = RLIM_INFINITY
sched: fix: don't take a mutex from interrupt context
Srivatsa Vaddagiri (5):
sched: group scheduling code cleanup
sched: group scheduling, minor fixes
sched: group scheduling, change how cpu load is calculated
sched: introduce a mutex and corresponding API to serialize access to doms_curarray
sched: group scheduler, fix fairness of cpu bandwidth allocation for task groups
Steven Rostedt (12):
sched: count # of queued RT tasks
sched: track highest prio task queued
sched: add RT task pushing
sched: add rt-overload tracking
sched: pull RT tasks from overloaded runqueues
sched: push RT tasks from overloaded CPUs
sched: disable standard balancer for RT tasks
sched: RT-balance, avoid overloading
sched: RT-balance, optimize cpu search
sched: RT-balance on new task
sched: RT-balance, replace hooks with pre/post schedule and wakeup methods
sched: RT-balance, add new methods to sched_class
Documentation/RCU/RTFP.txt | 210 +++- Documentation/RCU/rcu.txt | 19 Documentation/RCU/torture.txt | 11 Documentation/cpu-hotplug.txt | 11 arch/arm/kernel/time.c | 11 arch/ia64/kernel/setup.c | 4 arch/ia64/kernel/time.c | 27 arch/ia64/sn/kernel/setup.c | 11 arch/mips/kernel/mips-mt-fpaff.c | 10 arch/powerpc/platforms/pseries/hotplug-cpu.c | 8 arch/powerpc/platforms/pseries/rtasd.c | 8 arch/x86/kernel/cpu/mtrr/main.c | 8 arch/x86/kernel/entry_64.S | 6 arch/x86/kernel/microcode.c | 16 arch/x86/kernel/signal_32.c | 3 arch/x86/kernel/signal_64.c | 3 arch/x86/kernel/stacktrace.c | 27 drivers/lguest/x86/core.c | 8 drivers/s390/char/sclp_config.c | 4 fs/Kconfig | 1 fs/proc/base.c | 78 + include/asm-generic/resource.h | 5 include/asm-x86/thread_info_32.h | 2 include/asm-x86/thread_info_64.h | 5 include/linux/cpu.h | 17 include/linux/debug_locks.h | 5 include/linux/futex.h | 6 include/linux/hardirq.h | 6 include/linux/hrtimer.h | 14 include/linux/init_task.h | 7 include/linux/interrupt.h | 1 include/linux/jiffies.h | 6 include/linux/kernel.h | 4 include/linux/latencytop.h | 44 include/linux/notifier.h | 4 include/linux/rcuclassic.h | 164 +++ include/linux/rcupdate.h | 173 +-- include/linux/rcupreempt.h | 86 + include/linux/rcupreempt_trace.h | 99 + include/linux/sched.h | 83 + include/linux/smp_lock.h | 14 include/linux/stacktrace.h | 3 include/linux/topology.h | 5 init/Kconfig | 28 init/main.c | 1 kernel/Kconfig.hz | 2 kernel/Kconfig.preempt | 13 kernel/Makefile | 6 kernel/cpu.c | 164 ++- kernel/cpuset.c | 14 kernel/fork.c | 11 kernel/hrtimer.c | 256 ++-- kernel/kthread.c | 12 kernel/latencytop.c | 239 ++++ kernel/lockdep.c | 12 kernel/module.c | 27 kernel/posix-cpu-timers.c | 30 kernel/printk.c | 57 - kernel/profile.c | 99 - kernel/rcuclassic.c | 575 +++++++++++ kernel/rcupdate.c | 576 ----------- kernel/rcupreempt.c | 953 ++++++++++++++++++ kernel/rcupreempt_trace.c | 330 ++++++ kernel/rcutorture.c | 6 kernel/sched.c | 1384 +++++++++++++++++++-------- kernel/sched_debug.c | 5 kernel/sched_fair.c | 391 ++++++- kernel/sched_idletask.c | 42 kernel/sched_rt.c | 1112 +++++++++++++++++++-- kernel/softlockup.c | 116 ++ kernel/stop_machine.c | 4 kernel/sysctl.c | 77 + kernel/time/tick-sched.c | 13 kernel/timer.c | 3 kernel/user.c | 47 kernel/workqueue.c | 35 lib/Kconfig.debug | 14 lib/kernel_lock.c | 123 -- mm/oom_kill.c | 2 mm/slab.c | 18 net/core/flow.c | 4 81 files changed, 6262 insertions(+), 1776 deletions(-)
[ delta patch too big to be sent to lkml ]
preemptible RCU bug (was Re: [git pull] scheduler changes for v2.6.25)
Quoted 1 line · Ingo Molnar
- Paul E. McKenney's preemptible RCU code.
BUG: New Kconfig option without help text.
$ make oldconfig RCU implementation type:
Quoted 1 line
1. Classic RCU (
CLASSIC_RCU) (NEW)
2. Preemptible RCU (PREEMPT_RCU) (NEW)
choice[1-2]: ?
Sorry, no help available for this option yet.
Signature · 3 lines
Stefan Richter -=====-==--- ---= ==-=- http://arcgraph.de/sr/
Re: preemptible RCU bug (was Re: [git pull] scheduler changes for v2.6.25)
Quoted 5 lines · Stefan Richter
Quoted 1 line · Ingo Molnar
- Paul E. McKenney's preemptible RCU code.
BUG: New Kconfig option without help text.
Quoted 7 lines
$ make oldconfig RCU implementation type:Quoted 1 line
1. Classic RCU (
CLASSIC_RCU) (NEW)2. Preemptible RCU (
PREEMPT_RCU) (NEW)
choice[1-2]: ?Sorry, no help available for this option yet.
Good catch!!!
This patch supplies help text for the "RCU implementation type" kernel configuration choice.
---
Kconfig.preempt | 8 ++++++++ 1 file changed, 8 insertions(+)
kernel/Kconfig.preempt +8 -0
@@ -131,6 +131,14 @@ choice prompt "RCU implementation type:" default PREEMPT_RCU if PREEMPT_RT default CLASSIC_RCU + help + This allows you to choose either the classic RCU implementation + that is designed for best read-side performance on non-realtime + systems, or the preemptible RCU implementation for best latency + on realtime systems. Note that some kernel preemption modes + will restrict your choice. + + Select the default if you are unsure. config CLASSIC_RCU bool "Classic RCU"
Re: preemptible RCU bug (was Re: [git pull] scheduler changes for v2.6.25)
Quoted 4 lines · Paul E. McKenney
kernel/Kconfig.preempt +0
-0@@ -131,6 +131,14 @@ choice prompt "RCU implementation type:"
kernel/Kconfig.preempt choice · line 141
Quoted 10 lines
kernel/Kconfig.preempt +8
-0default PREEMPT_RCU if PREEMPT_RT default CLASSIC_RCU + help + This allows you to choose either the classic RCU implementation + that is designed for best read-side performance on non-realtime + systems, or the preemptible RCU implementation for best latency + on realtime systems. Note that some kernel preemption modes + will restrict your choice. + + Select the default if you are unsure.
Looks good to me. Serves those people who want to be informed as well as those who want to be comforted. :-) Thanks,
Signature · 3 lines
Stefan Richter -=====-==--- ---= ==-== http://arcgraph.de/sr/
Ingo Molnar
AppliedRe: preemptible RCU bug (was Re: [git pull] scheduler changes for v2.6.25)
In reply to Paul E. McKenneyQuoted 6 lines · Paul E. McKenney
Quoted 1 line
Sorry, no help available for this option yet.
Good catch!!!
This patch supplies help text for the "RCU implementation type" kernel configuration choice.
thanks applied to sched.git. (see the patch below - the config option is in init/Kconfig in mainline, not in kernel/Kconfig.preempt)
Ingo
------------->
Subject: RCU: add help text for "RCU implementation type"
From: "Paul E. McKenney" <[email protected]>
This patch supplies help text for the "RCU implementation type" kernel configuration choice.
---
init/Kconfig | 8 ++++++++ 1 file changed, 8 insertions(+)
init/Kconfig +8 -0
@@ -779,6 +779,14 @@ config PREEMPT_NOTIFIERS choice prompt "RCU implementation type:" default CLASSIC_RCU + help + This allows you to choose either the classic RCU implementation + that is designed for best read-side performance on non-realtime + systems, or the preemptible RCU implementation for best latency + on realtime systems. Note that some kernel preemption modes + will restrict your choice. + + Select the default if you are unsure. config CLASSIC_RCU bool "Classic RCU"
Re: preemptible RCU bug (was Re: [git pull] scheduler changes for v2.6.25)
Quoted 5 lines · Ingo Molnar
Quoted 2 lines · Paul E. McKenney
Quoted 1 line
Sorry, no help available for this option yet.
Quoted 7 lines
Quoted 4 lines
Good catch!!!
This patch supplies help text for the "RCU implementation type" kernel configuration choice.
thanks applied to sched.git. (see the patch below - the config option is in init/Kconfig in mainline, not in
kernel/Kconfig.preempt)
Ah missed that -- thank you for retargeting the patch.
Thanx, Paul
Quoted 36 lines
Ingo
------------->
Subject: RCU: add help text for "RCU implementation type"
From: "Paul E. McKenney" <[email protected]>This patch supplies help text for the "RCU implementation type" kernel configuration choice.
Reported-by Stefan Richter <[email protected]>Signed-off-by Paul E. McKenney <[email protected]>Signed-off-by Ingo Molnar <[email protected]>---
init/Kconfig | 8 ++++++++ 1 file changed, 8 insertions(+)Index: linux/init/Kconfig
===================================================================init/Kconfig +8
-0@@ -779,6 +779,14 @@ config PREEMPT_NOTIFIERS choice prompt "RCU implementation type:" default CLASSIC_RCU + help + This allows you to choose either the classic RCU implementation + that is designed for best read-side performance on non-realtime + systems, or the preemptible RCU implementation for best latency + on realtime systems. Note that some kernel preemption modes + will restrict your choice. + + Select the default if you are unsure. config CLASSIC_RCU bool "Classic RCU"