I am not sure about this patch, I am asking Ingo to take a decision.
task_struct->state == EXIT_DEAD is a very special case, to avoid a confusion
it makes sense to introduce a new state, TASK_DEAD, while EXIT_DEAD should
live only in ->exit_state as documented in sched.h.
Note that this state is not visible to user-space, get_task_state() masks off
unsuitable states.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
schedule() checks PF_DEAD on every context switch and sets ->state = EXIT_DEAD
to ensure that the exiting task will be deactivated. Note that this EXIT_DEAD
is in fact a "random" value, we can use any bit except normal TASK_XXX values.
It is better to set this state in do_exit() along with PF_DEAD flag and remove
that check in schedule().
We are safe wrt concurrent try_to_wake_up() (for example ptrace, tkill), it
can not change task's ->state: the 'state' argument of try_to_wake_up() can't
have EXIT_DEAD bit. And in case when try_to_wake_up() sees a stale value of
->state == TASK_RUNNING it will do nothing.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I am not sure this patch is correct: I can't understand what the current
code does, and I don't know what it was supposed to do.
The comment says:
* can't change policy, except between SCHED_NORMAL
* and SCHED_BATCH:
The code:
if (((policy != SCHED_NORMAL && p->policy != SCHED_BATCH) &&
(policy != SCHED_BATCH && p->policy != SCHED_NORMAL)) &&
But this is equivalent to:
if ( (is_rt_policy(policy) && has_rt_policy(p)) &&
which means something different. We can't _decrease_ the current
->rt_priority with such a check (if rlim[RLIMIT_RTPRIO] == 0).
Probably, it was supposed to be:
if ( !(policy == SCHED_NORMAL && p->policy == SCHED_BATCH) &&
!(policy == SCHED_BATCH && p->policy == SCHED_NORMAL)
this matches the comment, but strange: it doesn't allow to _drop_ the
realtime priority when rlim[RLIMIT_RTPRIO] == 0.
I think the right check would be:
/* can't set/change rt policy */
if (is_rt_policy(policy) &&
policy != p->policy &&
!rlim_rtprio)
return -EPERM;
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Imho, makes the code a bit easier to read.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Spawing ksoftirqd, migration, or watchdog, and calling init_timers_cpu()
may fail with small memory. If it happens in initcalls, kernel NULL
pointer dereference happens later. This patch makes crash happen
immediately in such cases. It seems a bit better than getting kernel NULL
pointer dereference later.
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Akinobu Mita <mita@miraclelinux.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The scheduler will stop load balancing if the most busy processor contains
processes pinned via processor affinity.
The scheduler currently only does one search for busiest cpu. If it cannot
pull any tasks away from the busiest cpu because they were pinned then the
scheduler goes into a corner and sulks leaving the idle processors idle.
F.e. If you have processor 0 busy running four tasks pinned via taskset,
there are none on processor 1 and one just started two processes on
processor 2 then the scheduler will not move one of the two processes away
from processor 2.
This patch fixes that issue by forcing the scheduler to come out of its
corner and retrying the load balancing by considering other processors for
load balancing.
This patch was originally developed by John Hawkes and discussed at
http://marc.theaimsgroup.com/?l=linux-kernel&m=113901368523205&w=2.
I have removed extraneous material and gone back to equipping struct rq
with the cpu the queue is associated with since this makes the patch much
easier and it is likely that others in the future will have the same
difficulty of figuring out which processor owns which runqueue.
The overhead added through these patches is a single word on the stack if
the kernel is configured to support 32 cpus or less (32 bit). For 32 bit
environments the maximum number of cpus that can be configued is 255 which
would result in the use of 32 bytes additional on the stack. On IA64 up to
1k cpus can be configured which will result in the use of 128 additional
bytes on the stack. The maximum additional cache footprint is one
cacheline. Typically memory use will be much less than a cacheline and the
additional cpumask will be placed on the stack in a cacheline that already
contains other local variable.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: John Hawkes <hawkes@sgi.com>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Peter Williams <pwil3058@bigpond.net.au>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
sched_setscheduler() looks at ->signal->rlim[]. It is unsafe do
dereference ->signal unless tasklist_lock or ->siglock is held (or p ==
current). We pin the task structure, but this can't prevent from
release_task()->__exit_signal() which sets ->signal = NULL.
Restore tasklist_lock across the setscheduler call.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initialize init task's pi_waiters plist. Otherwise cpu hotplug of cpu 0
might crash, since rt_mutex_getprio() accesses an uninitialized list head.
call chain which led to crash:
take_cpu_down
sched_idle_next
__setscheduler
rt_mutex_getprio
Using PLIST_HEAD_INIT in the INIT_TASK macro doesn't work unfortunately,
since the pi_waiters member is only conditionally present.
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
In cond_resched_lock() it calls __resched_legal() before dropping the spin
lock. __resched_legal() will always finds the preempt_count non-zero and
will prevent the call to __cond_resched().
The attached patch adds a parameter to __resched_legal() with the expected
preempt_count value.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Use the correct groups while initializing sched groups power for
allnodes_domain. This fixes the crash observed while creating exclusive
cpusets.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Reported-and-tested-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Make the task-related schedstats functions callable by delay accounting even
if schedstats collection isn't turned on. This removes the dependency of
delay accounting on schedstats.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Balbir Singh <balbir@in.ibm.com>
Cc: Jes Sorensen <jes@sgi.com>
Cc: Peter Chubb <peterc@gelato.unsw.edu.au>
Cc: Erich Focht <efocht@ess.nec.de>
Cc: Levent Serinol <lserinol@gmail.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Unlike earlier iterations of the delay accounting patches, now delays are only
collected for the actual I/O waits rather than try and cover the delays seen
in I/O submission paths.
Account separately for block I/O delays incurred as a result of swapin page
faults whose frequency can be affected by the task/process' rss limit. Hence
swapin delays can act as feedback for rss limit changes independent of I/O
priority changes.
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Balbir Singh <balbir@in.ibm.com>
Cc: Jes Sorensen <jes@sgi.com>
Cc: Peter Chubb <peterc@gelato.unsw.edu.au>
Cc: Erich Focht <efocht@ess.nec.de>
Cc: Levent Serinol <lserinol@gmail.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
On platforms that have __ARCH_WANT_UNLOCKED_CTXSW set and want to implement
lock validator support there's a bug in rq->lock handling: in this case we
dont 'carry over' the runqueue lock into another task - but still we did a
spinlock_release() of it. Fix this by making the spinlock_release() in
context_switch() dependent on !__ARCH_WANT_UNLOCKED_CTXSW.
(Reported by Ralf Baechle on MIPS, which has __ARCH_WANT_UNLOCKED_CTXSW.
This fixes a lockdep-internal BUG message on such platforms.)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
- constify and optimize stat_nam (thanks to Michael Tokarev!)
- spelling and comment fixes
Signed-off-by: Andreas Mohr <andi@lisas.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Problem:
In the function __migrate_task(), deactivate_task() followed by
activate_task() is used to move the task from one run queue to
another. This has two undesirable effects:
1. The task's priority is recalculated. (Nowhere else in the
scheduler code is the priority recalculated for a change of CPU.)
2. The task's time stamp is set to the current time. At the very least,
this makes the adjustment of the time stamp before the call to
deactivate_task() redundant but I believe the problem is more serious
as the time stamp now holds the time of the queue change instead of
the time at which the task was woken. In addition, unless dest_rq is
the same queue as "current" is on the time stamp could be inaccurate
due to inter CPU drift.
Solution:
Replace the call to activate_task() with one to __activate_task().
Signed-off-by: Peter Williams <pwil3058@bigpond.net.au>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
convert:
- runqueue_t to 'struct rq'
- prio_array_t to 'struct prio_array'
- migration_req_t to 'struct migration_req'
I was the one who added these but they are both against the kernel coding
style and also were used inconsistently at places. So just get rid of them at
once, now that we are flushing the scheduler patch-queue anyway.
Conversion was mostly scripted, the result was reviewed and all secondary
whitespace and style impact (if any) was fixed up by hand.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
cleanup: remove task_t and convert all the uses to struct task_struct. I
introduced it for the scheduler anno and it was a mistake.
Conversion was mostly scripted, the result was reviewed and all
secondary whitespace and style impact (if any) was fixed up by hand.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Clean up some of the impact of recent (and not so recent) scheduler
changes:
- turning macros into nice inline functions
- sanitizing and unifying variable definitions
- whitespace, style consistency, 80-lines, comment correctness, spelling
and curly braces police
Due to the macro hell and variable placement simplifications there's even 26
bytes of .text saved:
text data bss dec hex filename
25510 4153 192 29855 749f sched.o.before
25484 4153 192 29829 7485 sched.o.after
[akpm@osdl.org: build fix]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Teach per-CPU runqueue locks and recursive locking code to the lock validator.
Has no effect on non-lockdep kernels.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Use the lock validator framework to prove spinlock and rwlock locking
correctness.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Accurate hard-IRQ-flags and softirq-flags state tracing.
This allows us to attach extra functionality to IRQ flags on/off
events (such as trace-on/off).
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Generic lock debugging:
- generalized lock debugging framework. For example, a bug in one lock
subsystem turns off debugging in all lock subsystems.
- got rid of the caller address passing (__IP__/__IP_DECL__/etc.) from
the mutex/rtmutex debugging code: it caused way too much prototype
hackery, and lockdep will give the same information anyway.
- ability to do silent tests
- check lock freeing in vfree too.
- more finegrained debugging options, to allow distributions to
turn off more expensive debugging features.
There's no separate 'held mutexes' list anymore - but there's a 'held locks'
stack within lockdep, which unifies deadlock detection across all lock
classes. (this is independent of the lockdep validation stuff - lockdep first
checks whether we are holding a lock already)
Here are the current debugging options:
CONFIG_DEBUG_MUTEXES=y
CONFIG_DEBUG_LOCK_ALLOC=y
which do:
config DEBUG_MUTEXES
bool "Mutex debugging, basic checks"
config DEBUG_LOCK_ALLOC
bool "Detect incorrect freeing of live mutexes"
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Fix a bug identified by Zou Nan hai <nanhai.zou@intel.com>:
If the system is in state SYSTEM_BOOTING, and need_resched() is true,
cond_resched() returns true even though it didn't reschedule. Consequently
need_resched() remains true and JBD locks up.
Fix that by teaching cond_resched() to only return true if it really did call
schedule().
cond_resched_lock() and cond_resched_softirq() have a problem too. If we're
in SYSTEM_BOOTING state and need_resched() is true, these functions will drop
the lock and will then try to call schedule(), but the SYSTEM_BOOTING state
will prevent schedule() from being called. So on return, need_resched() will
still be true, but cond_resched_lock() has to return 1 to tell the caller that
the lock was dropped. The caller will probably lock up.
Bottom line: if these functions dropped the lock, they _must_ call schedule()
to clear need_resched(). Make it so.
Also, uninline __cond_resched(). It's largeish, and slowpath.
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
When the priority of a task, which is blocked on a lock, changes we must
propagate this change into the PI lock chain. Therefor the chain walk code
is changed to get rid of the references to current to avoid false positives
in the deadlock detector, as setscheduler might be called by a task which
holds the lock on which the task whose priority is changed is blocked.
Also add some comments about the get/put_task_struct usage to avoid
confusion.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There is no need to hold tasklist_lock across the setscheduler call, when
we pin the task structure with get_task_struct(). Interrupts are disabled
in setscheduler anyway and the permission checks do not need interrupts
disabled.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add framework to boost/unboost the priority of RT tasks.
This consists of:
- caching the 'normal' priority in ->normal_prio
- providing a functions to set/get the priority of the task
- make sched_setscheduler() aware of boosting
The effective_prio() cleanups also fix a priority-calculation bug pointed out
by Andrey Gelman, in set_user_nice().
has_rt_policy() fix: Peter Williams <pwil3058@bigpond.net.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Andrey Gelman <agelman@012.net.il>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Thomas Gleixner is adding the call to a rtmutex function in setscheduler.
This call grabs a spin_lock that is not always protected by interrupts
disabled. So this means that setscheduler cant be called from interrupt
context.
To prevent this from happening in the future, this patch adds a
BUG_ON(in_interrupt()) in that function. (Thanks to akpm <aka. Andrew
Morton> for this suggestion).
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
sysfs entries 'sched_mc_power_savings' and 'sched_smt_power_savings' in
/sys/devices/system/cpu/ control the MC/SMT power savings policy for the
scheduler.
Based on the values (1-enable, 0-disable) for these controls, sched groups
cpu power will be determined for different domains. When power savings
policy is enabled and under light load conditions, scheduler will minimize
the physical packages/cpu cores carrying the load and thus conserving
power(with a perf impact based on the workload characteristics... see OLS
2005 CMP kernel scheduler paper for more details..)
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Con Kolivas <kernel@kolivas.org>
Cc: "Chen, Kenneth W" <kenneth.w.chen@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
As explained here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=114327539012323&w=2
there is a problem with sharing sched_group structures between two
separate sched_group structures for different sched_domains.
The patch has been tested and found to avoid the kernel lockup problem
described in above URL.
Signed-off-by: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The sched group structures used to represent various nodes need to be
allocated from respective nodes (as suggested here also:
http://uwsg.ucs.indiana.edu/hypermail/linux/kernel/0603.3/0051.html)
Signed-off-by: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Try to handle mem allocation failures in build_sched_domains by bailing out
and cleaning up thus-far allocated memory. The patch has a direct consequence
that we disable load balancing completely (even at sibling level) upon *any*
memory allocation failure.
[Lee.Schermerhorn@hp.com: bugfix]
Signed-off-by: Srivatsa Vaddagir <vatsa@in.ibm.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Problem:
To help distribute high priority tasks evenly across the available CPUs
move_tasks() does not, under some circumstances, skip tasks whose load
weight is bigger than the designated amount. Because the highest priority
task on the busiest queue may be on the expired array it may be moved as a
result of this mechanism. Apart from not being the most desirable way to
redistribute the high priority tasks (we'd rather move the second highest
priority task), there is a risk that this could set up a loop with this
task bouncing backwards and forwards between the two queues. (This latter
possibility can be demonstrated by running a nice==-20 CPU bound task on an
otherwise quiet 2 CPU system.)
Solution:
Modify the mechanism so that it does not override skip for the highest
priority task on the CPU. Of course, if there are more than one tasks at
the highest priority then it will allow the override for one of them as
this is a desirable redistribution of high priority tasks.
Signed-off-by: Peter Williams <pwil3058@bigpond.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Problem:
The move_tasks() function is designed to move UP TO the amount of load it
is asked to move and in doing this it skips over tasks looking for ones
whose load weights are less than or equal to the remaining load to be
moved. This is (in general) a good thing but it has the unfortunate result
of breaking one of the original load balancer's good points: namely, that
(within the limits imposed by the active/expired array model and the fact
the expired is processed first) it moves high priority tasks before low
priority ones and this means there's a good chance (see active/expired
problem for why it's only a chance) that the highest priority task on the
queue but not actually on the CPU will be moved to the other CPU where (as
a high priority task) it may preempt the current task.
Solution:
Modify move_tasks() so that high priority tasks are not skipped when moving
them will make them the highest priority task on their new run queue.
Signed-off-by: Peter Williams <pwil3058@bigpond.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Cc: "Chen, Kenneth W" <kenneth.w.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Problem:
The introduction of separate run queues per CPU has brought with it "nice"
enforcement problems that are best described by a simple example.
For the sake of argument suppose that on a single CPU machine with a
nice==19 hard spinner and a nice==0 hard spinner running that the nice==0
task gets 95% of the CPU and the nice==19 task gets 5% of the CPU. Now
suppose that there is a system with 2 CPUs and 2 nice==19 hard spinners and
2 nice==0 hard spinners running. The user of this system would be entitled
to expect that the nice==0 tasks each get 95% of a CPU and the nice==19
tasks only get 5% each. However, whether this expectation is met is pretty
much down to luck as there are four equally likely distributions of the
tasks to the CPUs that the load balancing code will consider to be balanced
with loads of 2.0 for each CPU. Two of these distributions involve one
nice==0 and one nice==19 task per CPU and in these circumstances the users
expectations will be met. The other two distributions both involve both
nice==0 tasks being on one CPU and both nice==19 being on the other CPU and
each task will get 50% of a CPU and the user's expectations will not be
met.
Solution:
The solution to this problem that is implemented in the attached patch is
to use weighted loads when determining if the system is balanced and, when
an imbalance is detected, to move an amount of weighted load between run
queues (as opposed to a number of tasks) to restore the balance. Once
again, the easiest way to explain why both of these measures are necessary
is to use a simple example. Suppose that (in a slight variation of the
above example) that we have a two CPU system with 4 nice==0 and 4 nice=19
hard spinning tasks running and that the 4 nice==0 tasks are on one CPU and
the 4 nice==19 tasks are on the other CPU. The weighted loads for the two
CPUs would be 4.0 and 0.2 respectively and the load balancing code would
move 2 tasks resulting in one CPU with a load of 2.0 and the other with
load of 2.2. If this was considered to be a big enough imbalance to
justify moving a task and that task was moved using the current
move_tasks() then it would move the highest priority task that it found and
this would result in one CPU with a load of 3.0 and the other with a load
of 1.2 which would result in the movement of a task in the opposite
direction and so on -- infinite loop. If, on the other hand, an amount of
load to be moved is calculated from the imbalance (in this case 0.1) and
move_tasks() skips tasks until it find ones whose contributions to the
weighted load are less than this amount it would move two of the nice==19
tasks resulting in a system with 2 nice==0 and 2 nice=19 on each CPU with
loads of 2.1 for each CPU.
One of the advantages of this mechanism is that on a system where all tasks
have nice==0 the load balancing calculations would be mathematically
identical to the current load balancing code.
Notes:
struct task_struct:
has a new field load_weight which (in a trade off of space for speed)
stores the contribution that this task makes to a CPU's weighted load when
it is runnable.
struct runqueue:
has a new field raw_weighted_load which is the sum of the load_weight
values for the currently runnable tasks on this run queue. This field
always needs to be updated when nr_running is updated so two new inline
functions inc_nr_running() and dec_nr_running() have been created to make
sure that this happens. This also offers a convenient way to optimize away
this part of the smpnice mechanism when CONFIG_SMP is not defined.
int try_to_wake_up():
in this function the value SCHED_LOAD_BALANCE is used to represent the load
contribution of a single task in various calculations in the code that
decides which CPU to put the waking task on. While this would be a valid
on a system where the nice values for the runnable tasks were distributed
evenly around zero it will lead to anomalous load balancing if the
distribution is skewed in either direction. To overcome this problem
SCHED_LOAD_SCALE has been replaced by the load_weight for the relevant task
or by the average load_weight per task for the queue in question (as
appropriate).
int move_tasks():
The modifications to this function were complicated by the fact that
active_load_balance() uses it to move exactly one task without checking
whether an imbalance actually exists. This precluded the simple
overloading of max_nr_move with max_load_move and necessitated the addition
of the latter as an extra argument to the function. The internal
implementation is then modified to move up to max_nr_move tasks and
max_load_move of weighted load. This slightly complicates the code where
move_tasks() is called and if ever active_load_balance() is changed to not
use move_tasks() the implementation of move_tasks() should be simplified
accordingly.
struct sched_group *find_busiest_group():
Similar to try_to_wake_up(), there are places in this function where
SCHED_LOAD_SCALE is used to represent the load contribution of a single
task and the same issues are created. A similar solution is adopted except
that it is now the average per task contribution to a group's load (as
opposed to a run queue) that is required. As this value is not directly
available from the group it is calculated on the fly as the queues in the
groups are visited when determining the busiest group.
A key change to this function is that it is no longer to scale down
*imbalance on exit as move_tasks() uses the load in its scaled form.
void set_user_nice():
has been modified to update the task's load_weight field when it's nice
value and also to ensure that its run queue's raw_weighted_load field is
updated if it was runnable.
From: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
With smpnice, sched groups with highest priority tasks can mask the imbalance
between the other sched groups with in the same domain. This patch fixes some
of the listed down scenarios by not considering the sched groups which are
lightly loaded.
a) on a simple 4-way MP system, if we have one high priority and 4 normal
priority tasks, with smpnice we would like to see the high priority task
scheduled on one cpu, two other cpus getting one normal task each and the
fourth cpu getting the remaining two normal tasks. but with current
smpnice extra normal priority task keeps jumping from one cpu to another
cpu having the normal priority task. This is because of the
busiest_has_loaded_cpus, nr_loaded_cpus logic.. We are not including the
cpu with high priority task in max_load calculations but including that in
total and avg_load calcuations.. leading to max_load < avg_load and load
balance between cpus running normal priority tasks(2 Vs 1) will always show
imbalanace as one normal priority and the extra normal priority task will
keep moving from one cpu to another cpu having normal priority task..
b) 4-way system with HT (8 logical processors). Package-P0 T0 has a
highest priority task, T1 is idle. Package-P1 Both T0 and T1 have 1 normal
priority task each.. P2 and P3 are idle. With this patch, one of the
normal priority tasks on P1 will be moved to P2 or P3..
c) With the current weighted smp nice calculations, it doesn't always make
sense to look at the highest weighted runqueue in the busy group..
Consider a load balance scenario on a DP with HT system, with Package-0
containing one high priority and one low priority, Package-1 containing one
low priority(with other thread being idle).. Package-1 thinks that it need
to take the low priority thread from Package-0. And find_busiest_queue()
returns the cpu thread with highest priority task.. And ultimately(with
help of active load balance) we move high priority task to Package-1. And
same continues with Package-0 now, moving high priority task from package-1
to package-0.. Even without the presence of active load balance, load
balance will fail to balance the above scenario.. Fix find_busiest_queue
to use "imbalance" when it is lightly loaded.
[kernel@kolivas.org: sched: store weighted load on up]
[kernel@kolivas.org: sched: add discrete weighted cpu load function]
[suresh.b.siddha@intel.com: sched: remove dead code]
Signed-off-by: Peter Williams <pwil3058@bigpond.com.au>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Cc: "Chen, Kenneth W" <kenneth.w.chen@intel.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Con Kolivas <kernel@kolivas.org>
Cc: John Hawkes <hawkes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There is a race between set_cpus_allowed() and move_task_off_dead_cpu().
__migrate_task() doesn't report any err code, so task can be left on its
runqueue if its cpus_allowed mask changed so that dest_cpu is not longer a
possible target. Also, chaning cpus_allowed mask requires rq->lock being
held.
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Acked-By: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Unless we expect to have more than 2G CPUs, there's no reason to have 'i'
as a long long here.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The relationship between INTERACTIVE_SLEEP and the ceiling is not perfect
and not explicit enough. The sleep boost is not supposed to be any larger
than without this code and the comment is not clear enough about what
exactly it does, just the reason it does it. Fix it.
There is a ceiling to the priority beyond which tasks that only ever sleep
for very long periods cannot surpass. Fix it.
Prevent the on-runqueue bonus logic from defeating the idle sleep logic.
Opportunity to micro-optimise.
Signed-off-by: Con Kolivas <kernel@kolivas.org>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial report and lock contention fix from Chris Mason:
Recent benchmarks showed some performance regressions between 2.6.16 and
2.6.5. We tracked down one of the regressions to lock contention in
schedule heavy workloads (~70,000 context switches per second)
kernel/sched.c:dependent_sleeper() was responsible for most of the lock
contention, hammering on the run queue locks. The patch below is more of a
discussion point than a suggested fix (although it does reduce lock
contention significantly). The dependent_sleeper code looks very expensive
to me, especially for using a spinlock to bounce control between two
different siblings in the same cpu.
It is further optimized:
* perform dependent_sleeper check after next task is determined
* convert wake_sleeping_dependent to use trylock
* skip smt runqueue check if trylock fails
* optimize double_rq_lock now that smt nice is converted to trylock
* early exit in searching first SD_SHARE_CPUPOWER domain
* speedup fast path of dependent_sleeper
[akpm@osdl.org: cleanup]
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Con Kolivas <kernel@kolivas.org>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Chris Mason <mason@suse.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Make notifier_calls associated with cpu_notifier as __cpuinit.
__cpuinit makes sure that the function is init time only unless
CONFIG_HOTPLUG_CPU is defined.
[akpm@osdl.org: section fix]
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
During some profiling I noticed that default_idle causes a lot of
memory traffic. I think that is caused by the atomic operations
to clear/set the polling flag in thread_info. There is actually
no reason to make this atomic - only the idle thread does it
to itself, other CPUs only read it. So I moved it into ti->status.
Converted i386/x86-64/ia64 for now because that was the easiest
way to fix ACPI which also manipulates these flags in its idle
function.
Cc: Nick Piggin <npiggin@novell.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Len Brown <len.brown@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The introduction of SCHED_BATCH scheduling class with a value of 3 means
that the expression (p->policy & SCHED_FIFO) will return true if policy
is SCHED_BATCH or SCHED_FIFO.
Unfortunately, this expression is used in sys_sched_rr_get_interval()
and in the absence of a comment to say that this is intentional I
presume that it is unintentional and erroneous.
The fix is to change the expression to (p->policy == SCHED_FIFO).
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If a cpu hotplug callback fails on CPU_UP_PREPARE, all callbacks will be
called with CPU_UP_CANCELED. A few of these callbacks assume that on
CPU_UP_PREPARE a pointer to task has been stored in a percpu array. This
assumption is not true if CPU_UP_PREPARE fails and the following calls to
kthread_bind() in CPU_UP_CANCELED will cause an addressing exception
because of passing a NULL pointer.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There are several instances of per_cpu(foo, raw_smp_processor_id()), which
is semantically equivalent to __get_cpu_var(foo) but without the warning
that smp_processor_id() can give if CONFIG_DEBUG_PREEMPT is enabled. For
those architectures with optimized per-cpu implementations, namely ia64,
powerpc, s390, sparc64 and x86_64, per_cpu() turns into more and slower
code than __get_cpu_var(), so it would be preferable to use __get_cpu_var
on those platforms.
This defines a __raw_get_cpu_var(x) macro which turns into per_cpu(x,
raw_smp_processor_id()) on architectures that use the generic per-cpu
implementation, and turns into __get_cpu_var(x) on the architectures that
have an optimized per-cpu implementation.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
