Peter Zijlstra [Fri, 25 Jan 2008 20:08:29 +0000 (21:08 +0100)]
sched: rt time limit
Very simple time limit on the realtime scheduling classes.
Allow the rq's realtime class to consume sched_rt_ratio of every
sched_rt_period slice. If the class exceeds this quota the fair class
will preempt the realtime class.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Peter Zijlstra [Fri, 25 Jan 2008 20:08:29 +0000 (21:08 +0100)]
sched: high-res preemption tick
Use HR-timers (when available) to deliver an accurate preemption tick.
The regular scheduler tick that runs at 1/HZ can be too coarse when nice
level are used. The fairness system will still keep the cpu utilisation 'fair'
by then delaying the task that got an excessive amount of CPU time but try to
minimize this by delivering preemption points spot-on.
The average frequency of this extra interrupt is sched_latency / nr_latency.
Which need not be higher than 1/HZ, its just that the distribution within the
sched_latency period is important.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Peter Zijlstra [Fri, 25 Jan 2008 20:08:27 +0000 (21:08 +0100)]
sched: SCHED_FIFO/SCHED_RR watchdog timer
Introduce a new rlimit that allows the user to set a runtime timeout on
real-time tasks their slice. Once this limit is exceeded the task will receive
SIGXCPU.
So it measures runtime since the last sleep.
Input and ideas by Thomas Gleixner and Lennart Poettering.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> CC: Lennart Poettering <mzxreary@0pointer.de> CC: Michael Kerrisk <mtk.manpages@googlemail.com> CC: Ulrich Drepper <drepper@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Pavel Emelyanov [Fri, 25 Jan 2008 20:08:26 +0000 (21:08 +0100)]
uids: merge multiple error paths in alloc_uid() into one
There are already 4 error paths in alloc_uid() that do incremental rollbacks.
I think it's time to merge them. This costs us 8 lines of code :)
Maybe it would be better to merge this patch with the previous one, but I
remember that some time ago I sent a similar patch (fixing the error path and
cleaning it), but I was told to make two patches in such cases.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Acked-by: Dhaval Giani <dhaval@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gregory Haskins [Fri, 25 Jan 2008 20:08:26 +0000 (21:08 +0100)]
sched: dynamically update the root-domain span/online maps
The baseline code statically builds the span maps when the domain is formed.
Previous attempts at dynamically updating the maps caused a suspend-to-ram
regression, which should now be fixed.
Paul E. McKenney [Fri, 25 Jan 2008 20:08:25 +0000 (21:08 +0100)]
Preempt-RCU: update RCU Documentation.
This patch updates the RCU documentation to reflect preemptible RCU as
well as recent publications.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Reviewed-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Paul E. McKenney [Fri, 25 Jan 2008 20:08:25 +0000 (21:08 +0100)]
Preempt-RCU: CPU Hotplug handling
This patch allows preemptible RCU to tolerate CPU-hotplug operations.
It accomplishes this by maintaining a local copy of a map of online
CPUs, which it accesses under its own lock.
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Reviewed-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Paul E. McKenney [Fri, 25 Jan 2008 20:08:24 +0000 (21:08 +0100)]
Preempt-RCU: implementation
This patch implements a new version of RCU which allows its read-side
critical sections to be preempted. It uses a set of counter pairs
to keep track of the read-side critical sections and flips them
when all tasks exit read-side critical section. The details
of this implementation can be found in this paper -
This patch was developed as a part of the -rt kernel development and
meant to provide better latencies when read-side critical sections of
RCU don't disable preemption. As a consequence of keeping track of RCU
readers, the readers have a slight overhead (optimizations in the paper).
This implementation co-exists with the "classic" RCU implementations
and can be switched to at compiler.
Also includes RCU tracing summarized in debugfs.
[ akpm@linux-foundation.org: build fixes on non-preempt architectures ]
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com> Reviewed-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Paul E. McKenney [Fri, 25 Jan 2008 20:08:24 +0000 (21:08 +0100)]
Preempt-RCU: fix rcu_barrier for preemptive environment.
Fix rcu_barrier() to work properly in preemptive kernel environment.
Also, the ordering of callback must be preserved while moving
callbacks to another CPU during CPU hotplug.
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Reviewed-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Paul E. McKenney [Fri, 25 Jan 2008 20:08:24 +0000 (21:08 +0100)]
Preempt-RCU: reorganize RCU code into rcuclassic.c and rcupdate.c
This patch re-organizes the RCU code to enable multiple implementations
of RCU. Users of RCU continues to include rcupdate.h and the
RCU interfaces remain the same. This is in preparation for
subsequently merging the preemptible RCU implementation.
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Reviewed-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Dipankar Sarma [Fri, 25 Jan 2008 20:08:23 +0000 (21:08 +0100)]
Preempt-RCU: Use softirq instead of tasklets for
This patch makes RCU use softirq instead of tasklets.
It also adds a memory barrier after raising the softirq
inorder to ensure that the cpu sees the most recently updated
value of rcu->cur while processing callbacks.
The discussion of the related theoretical race pointed out
by James Huang can be found here --> http://lkml.org/lkml/2007/11/20/603
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com> Reviewed-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gregory Haskins [Fri, 25 Jan 2008 20:08:23 +0000 (21:08 +0100)]
sched: RT-balance, only adjust overload state when changing
The overload set/clears were originally idempotent when this logic was first
implemented. But that is no longer true due to the addition of the atomic
counter and this logic was never updated to work properly with that change.
So only adjust the overload state if it is actually changing to avoid
getting out of sync.
Steven Rostedt [Fri, 25 Jan 2008 20:08:22 +0000 (21:08 +0100)]
sched: RT-balance, add new methods to sched_class
Dmitry Adamushko found that the current implementation of the RT
balancing code left out changes to the sched_setscheduler and
rt_mutex_setprio.
This patch addresses this issue by adding methods to the schedule classes
to handle being switched out of (switched_from) and being switched into
(switched_to) a sched_class. Also a method for changing of priorities
is also added (prio_changed).
This patch also removes some duplicate logic between rt_mutex_setprio and
sched_setscheduler.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Steven Rostedt [Fri, 25 Jan 2008 20:08:22 +0000 (21:08 +0100)]
sched: RT-balance, replace hooks with pre/post schedule and wakeup methods
To make the main sched.c code more agnostic to the schedule classes.
Instead of having specific hooks in the schedule code for the RT class
balancing. They are replaced with a pre_schedule, post_schedule
and task_wake_up methods. These methods may be used by any of the classes
but currently, only the sched_rt class implements them.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gregory Haskins [Fri, 25 Jan 2008 20:08:18 +0000 (21:08 +0100)]
sched: only balance our RT tasks within our domain
We move the rt-overload data as the first global to per-domain
reclassification. This limits the scope of overload related cache-line
bouncing to stay with a specified partition instead of affecting all
cpus in the system.
Finally, we limit the scope of find_lowest_cpu searches to the domain
instead of the entire system. Note that we would always respect domain
boundaries even without this patch, but we first would scan potentially
all cpus before whittling the list down. Now we can avoid looking at
RQs that are out of scope, again reducing cache-line hits.
Note: In some cases, task->cpus_allowed will effectively reduce our search
to within our domain. However, I believe there are cases where the
cpus_allowed mask may be all ones and therefore we err on the side of
caution. If it can be optimized later, so be it.
Gregory Haskins [Fri, 25 Jan 2008 20:08:18 +0000 (21:08 +0100)]
sched: add sched-domain roots
We add the notion of a root-domain which will be used later to rescope
global variables to per-domain variables. Each exclusive cpuset
essentially defines an island domain by fully partitioning the member cpus
from any other cpuset. However, we currently still maintain some
policy/state as global variables which transcend all cpusets. Consider,
for instance, rt-overload state.
Whenever a new exclusive cpuset is created, we also create a new
root-domain object and move each cpu member to the root-domain's span.
By default the system creates a single root-domain with all cpus as
members (mimicking the global state we have today).
We add some plumbing for storing class specific data in our root-domain.
Whenever a RQ is switching root-domains (because of repartitioning) we
give each sched_class the opportunity to remove any state from its old
domain and add state to the new one. This logic doesn't have any clients
yet but it will later in the series.
Signed-off-by: Gregory Haskins <ghaskins@novell.com> CC: Christoph Lameter <clameter@sgi.com> CC: Paul Jackson <pj@sgi.com> CC: Simon Derr <simon.derr@bull.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Steven Rostedt [Fri, 25 Jan 2008 20:08:13 +0000 (21:08 +0100)]
sched: RT-balance, optimize cpu search
This patch removes several cpumask operations by keeping track
of the first of the CPUS that is of the lowest priority. When
the search for the lowest priority runqueue is completed, all
the bits up to the first CPU with the lowest priority runqueue
is cleared.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gregory Haskins [Fri, 25 Jan 2008 20:08:13 +0000 (21:08 +0100)]
sched: RT-balance, optimize
We can cheaply track the number of bits set in the cpumask for the lowest
priority CPUs. Therefore, compute the mask's weight and use it to skip
the optimal domain search logic when there is only one CPU available.
Steven Rostedt [Fri, 25 Jan 2008 20:08:12 +0000 (21:08 +0100)]
sched: RT-balance, avoid overloading
This patch changes the searching for a run queue by a waking RT task
to try to pick another runqueue if the currently running task
is an RT task.
The reason is that RT tasks behave different than normal
tasks. Preempting a normal task to run a RT task to keep
its cache hot is fine, because the preempted non-RT task
may wait on that same runqueue to run again unless the
migration thread comes along and pulls it off.
RT tasks behave differently. If one is preempted, it makes
an active effort to continue to run. So by having a high
priority task preempt a lower priority RT task, that lower
RT task will then quickly try to run on another runqueue.
This will cause that lower RT task to replace its nice
hot cache (and TLB) with a completely cold one. This is
for the hope that the new high priority RT task will keep
its cache hot.
Remeber that this high priority RT task was just woken up.
So it may likely have been sleeping for several milliseconds,
and will end up with a cold cache anyway. RT tasks run till
they voluntarily stop, or are preempted by a higher priority
task. This means that it is unlikely that the woken RT task
will have a hot cache to wake up to. So pushing off a lower
RT task is just killing its cache for no good reason.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gregory Haskins [Fri, 25 Jan 2008 20:08:12 +0000 (21:08 +0100)]
sched: wake-balance fixes
We have logic to detect whether the system has migratable tasks, but we are
not using it when deciding whether to push tasks away. So we add support
for considering this new information.
Gregory Haskins [Fri, 25 Jan 2008 20:08:11 +0000 (21:08 +0100)]
sched: optimize RT affinity
The current code base assumes a relatively flat CPU/core topology and will
route RT tasks to any CPU fairly equally. In the real world, there are
various toplogies and affinities that govern where a task is best suited to
run with the smallest amount of overhead. NUMA and multi-core CPUs are
prime examples of topologies that can impact cache performance.
Fortunately, linux is already structured to represent these topologies via
the sched_domains interface. So we change our RT router to consult a
combination of topology and affinity policy to best place tasks during
migration.
Gregory Haskins [Fri, 25 Jan 2008 20:08:10 +0000 (21:08 +0100)]
sched: pre-route RT tasks on wakeup
In the original patch series that Steven Rostedt and I worked on together,
we both took different approaches to low-priority wakeup path. I utilized
"pre-routing" (push the task away to a less important RQ before activating)
approach, while Steve utilized a "post-routing" approach. The advantage of
my approach is that you avoid the overhead of a wasted activate/deactivate
cycle and peripherally related burdens. The advantage of Steve's method is
that it neatly solves an issue preventing a "pull" optimization from being
deployed.
In the end, we ended up deploying Steve's idea. But it later dawned on me
that we could get the best of both worlds by deploying both ideas together,
albeit slightly modified.
The idea is simple: Use a "light-weight" lookup for pre-routing, since we
only need to approximate a good home for the task. And we also retain the
post-routing push logic to clean up any inaccuracies caused by a condition
of "priority mistargeting" caused by the lightweight lookup. Most of the
time, the pre-routing should work and yield lower overhead. In the cases
where it doesnt, the post-router will bat cleanup.
Gregory Haskins [Fri, 25 Jan 2008 20:08:10 +0000 (21:08 +0100)]
sched: RT balancing: include current CPU
It doesn't hurt if we allow the current CPU to be included in the
search. We will just simply skip it later if the current CPU turns out
to be the lowest.
Gregory Haskins [Fri, 25 Jan 2008 20:08:09 +0000 (21:08 +0100)]
sched: de-SCHED_OTHER-ize the RT path
The current wake-up code path tries to determine if it can optimize the
wake-up to "this_cpu" by computing load calculations. The problem is that
these calculations are only relevant to SCHED_OTHER tasks where load is king.
For RT tasks, priority is king. So the load calculation is completely wasted
bandwidth.
Therefore, we create a new sched_class interface to help with
pre-wakeup routing decisions and move the load calculation as a function
of CFS task's class.
Gregory Haskins [Fri, 25 Jan 2008 20:08:09 +0000 (21:08 +0100)]
sched: clean up this_rq use in kernel/sched_rt.c
"this_rq" is normally used to denote the RQ on the current cpu
(i.e. "cpu_rq(this_cpu)"). So clean up the usage of this_rq to be
more consistent with the rest of the code.
Gregory Haskins [Fri, 25 Jan 2008 20:08:07 +0000 (21:08 +0100)]
sched: add RT-balance cpu-weight
Some RT tasks (particularly kthreads) are bound to one specific CPU.
It is fairly common for two or more bound tasks to get queued up at the
same time. Consider, for instance, softirq_timer and softirq_sched. A
timer goes off in an ISR which schedules softirq_thread to run at RT50.
Then the timer handler determines that it's time to smp-rebalance the
system so it schedules softirq_sched to run. So we are in a situation
where we have two RT50 tasks queued, and the system will go into
rt-overload condition to request other CPUs for help.
This causes two problems in the current code:
1) If a high-priority bound task and a low-priority unbounded task queue
up behind the running task, we will fail to ever relocate the unbounded
task because we terminate the search on the first unmovable task.
2) We spend precious futile cycles in the fast-path trying to pull
overloaded tasks over. It is therefore optimial to strive to avoid the
overhead all together if we can cheaply detect the condition before
overload even occurs.
This patch tries to achieve this optimization by utilizing the hamming
weight of the task->cpus_allowed mask. A weight of 1 indicates that
the task cannot be migrated. We will then utilize this information to
skip non-migratable tasks and to eliminate uncessary rebalance attempts.
We introduce a per-rq variable to count the number of migratable tasks
that are currently running. We only go into overload if we have more
than one rt task, AND at least one of them is migratable.
In addition, we introduce a per-task variable to cache the cpus_allowed
weight, since the hamming calculation is probably relatively expensive.
We only update the cached value when the mask is updated which should be
relatively infrequent, especially compared to scheduling frequency
in the fast path.
Steven Rostedt [Fri, 25 Jan 2008 20:08:07 +0000 (21:08 +0100)]
sched: disable standard balancer for RT tasks
Since we now take an active approach to load balancing, we don't need to
balance RT tasks via the normal task balancer. In fact, this code was
found to pull RT tasks away from CPUS that the active movement performed,
resulting in large latencies.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Steven Rostedt [Fri, 25 Jan 2008 20:08:07 +0000 (21:08 +0100)]
sched: pull RT tasks from overloaded runqueues
This patch adds the algorithm to pull tasks from RT overloaded runqueues.
When a pull RT is initiated, all overloaded runqueues are examined for
a RT task that is higher in prio than the highest prio task queued on the
target runqueue. If another runqueue holds a RT task that is of higher
prio than the highest prio task on the target runqueue is found it is pulled
to the target runqueue.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Steven Rostedt [Fri, 25 Jan 2008 20:08:06 +0000 (21:08 +0100)]
sched: add rt-overload tracking
This patch adds an RT overload accounting system. When a runqueue has
more than one RT task queued, it is marked as overloaded. That is that it
is a candidate to have RT tasks pulled from it.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Steven Rostedt [Fri, 25 Jan 2008 20:08:05 +0000 (21:08 +0100)]
sched: add RT task pushing
This patch adds an algorithm to push extra RT tasks off a run queue to
other CPU runqueues.
When more than one RT task is added to a run queue, this algorithm takes
an assertive approach to push the RT tasks that are not running onto other
run queues that have lower priority. The way this works is that the highest
RT task that is not running is looked at and we examine the runqueues on
the CPUS for that tasks affinity mask. We find the runqueue with the lowest
prio in the CPU affinity of the picked task, and if it is lower in prio than
the picked task, we push the task onto that CPU runqueue.
We continue pushing RT tasks off the current runqueue until we don't push any
more. The algorithm stops when the next highest RT task can't preempt any
other processes on other CPUS.
TODO: The algorithm may stop when there are still RT tasks that can be
migrated. Specifically, if the highest non running RT task CPU affinity
is restricted to CPUs that are running higher priority tasks, there may
be a lower priority task queued that has an affinity with a CPU that is
running a lower priority task that it could be migrated to. This
patch set does not address this issue.
Note: checkpatch reveals two over 80 character instances. I'm not sure
that breaking them up will help visually, so I left them as is.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Steven Rostedt [Fri, 25 Jan 2008 20:08:04 +0000 (21:08 +0100)]
sched: track highest prio task queued
This patch adds accounting to each runqueue to keep track of the
highest prio task queued on the run queue. We only care about
RT tasks, so if the run queue does not contain any active RT tasks
its priority will be considered MAX_RT_PRIO.
This information will be used for later patches.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
the current default timeout is 120 seconds. Such messages are printed
up to 10 times per bootup. If the system has crashed already then the
messages are not printed.
if lockdep is enabled then all held locks are printed as well.
this feature is a natural extension to the softlockup-detector (kernel
locked up without scheduling) and to the NMI watchdog (kernel locked up
with IRQs disabled).
[ Gautham R Shenoy <ego@in.ibm.com>: CPU hotplug fixes. ]
[ Andrew Morton <akpm@linux-foundation.org>: build warning fix. ]
Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Gautham R Shenoy [Fri, 25 Jan 2008 20:08:02 +0000 (21:08 +0100)]
cpu-hotplug: replace per-subsystem mutexes with get_online_cpus()
This patch converts the known per-subsystem mutexes to get_online_cpus
put_online_cpus. It also eliminates the CPU_LOCK_ACQUIRE and
CPU_LOCK_RELEASE hotplug notification events.
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gautham R Shenoy [Fri, 25 Jan 2008 20:08:02 +0000 (21:08 +0100)]
cpu-hotplug: replace lock_cpu_hotplug() with get_online_cpus()
Replace all lock_cpu_hotplug/unlock_cpu_hotplug from the kernel and use
get_online_cpus and put_online_cpus instead as it highlights the
refcount semantics in these operations.
The new API guarantees protection against the cpu-hotplug operation, but
it doesn't guarantee serialized access to any of the local data
structures. Hence the changes needs to be reviewed.
In case of pseries_add_processor/pseries_remove_processor, use
cpu_maps_update_begin()/cpu_maps_update_done() as we're modifying the
cpu_present_map there.
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Gautham R Shenoy [Fri, 25 Jan 2008 20:08:01 +0000 (21:08 +0100)]
cpu-hotplug: refcount based cpu hotplug
This patch implements a Refcount + Waitqueue based model for
cpu-hotplug.
Now, a thread which wants to prevent cpu-hotplug, will bump up a global
refcount and the thread which wants to perform a cpu-hotplug operation
will block till the global refcount goes to zero.
The readers, if any, during an ongoing cpu-hotplug operation are blocked
until the cpu-hotplug operation is over.
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com> Signed-off-by: Paul Jackson <pj@sgi.com> [For !CONFIG_HOTPLUG_CPU ] Signed-off-by: Ingo Molnar <mingo@elte.hu>
Col1 -> Group name
Col2 -> Cumulative execution time (in seconds) received by all tasks of that
group in a 60sec window across 8 cpus
Col3 -> CPU bandwidth received by the group in the 60sec window, expressed in
percentage. Col3 data is derived as:
Col3 = 100 * Col2 / (NR_CPUS * 60)
Col4 -> CPU bandwidth received by each individual task of the group.
Col4 = 100 * cpu_time_recd_by_task / 60
[I can share the test case that produces a similar o/p if reqd]
The deviation from desired group fairness is as below:
a = +24.47%
b = -9.13%
c = -15.33%
which is quite high.
After the patch below is applied, here are the results:
Deviation from desired group fairness is as below:
a = +0.67%
b = -0.83%
c = +0.17%
which is far better IMO. Most of other runs have yielded a deviation within
+-2% at the most, which is good.
Why do we see bad (group) fairness with current scheuler?
=========================================================
Currently cpu's weight is just the summation of individual task weights.
This can yield incorrect results. For ex: consider three groups as below
on a 2-cpu system:
CPU0 CPU1
---------------------------
A (10) B(5)
C(5)
---------------------------
Group A has 10 tasks, all on CPU0, Group B and C have 5 tasks each all
of which are on CPU1. Each task has the same weight (NICE_0_LOAD =
1024).
The current scheme would yield a cpu weight of 10240 (10*1024) for each cpu and
the load balancer will think both CPUs are perfectly balanced and won't
move around any tasks. This, however, would yield this bandwidth:
A = 50%
B = 25%
C = 25%
which is not the desired result.
What's changing in the patch?
=============================
- How cpu weights are calculated when CONFIF_FAIR_GROUP_SCHED is
defined (see below)
- API Change
- Two tunables introduced in sysfs (under SCHED_DEBUG) to
control the frequency at which the load balance monitor
thread runs.
The basic change made in this patch is how cpu weight (rq->load.weight) is
calculated. Its now calculated as the summation of group weights on a cpu,
rather than summation of task weights. Weight exerted by a group on a
cpu is dependent on the shares allocated to it and also the number of
tasks the group has on that cpu compared to the total number of
(runnable) tasks the group has in the system.
Let,
W(K,i) = Weight of group K on cpu i
T(K,i) = Task load present in group K's cfs_rq on cpu i
T(K) = Total task load of group K across various cpus
S(K) = Shares allocated to group K
NRCPUS = Number of online cpus in the scheduler domain to
which group K is assigned.
Then,
W(K,i) = S(K) * NRCPUS * T(K,i) / T(K)
A load balance monitor thread is created at bootup, which periodically
runs and adjusts group's weight on each cpu. To avoid its overhead, two
min/max tunables are introduced (under SCHED_DEBUG) to control the rate
at which it runs.
Fixes from: Peter Zijlstra <a.p.zijlstra@chello.nl>
- don't start the load_balance_monitor when there is only a single cpu.
- rename the kthread because its currently longer than TASK_COMM_LEN
sched: introduce a mutex and corresponding API to serialize access to doms_curarray
doms_cur[] array represents various scheduling domains which are
mutually exclusive. Currently cpusets code can modify this array (by
calling partition_sched_domains()) as a result of user modifying
sched_load_balance flag for various cpusets.
This patch introduces a mutex and corresponding API (only when
CONFIG_FAIR_GROUP_SCHED is defined) which allows a reader to safely read
the doms_cur[] array w/o worrying abt concurrent modifications to the
array.
The fair group scheduler code (introduced in next patch of this series)
makes use of this mutex to walk thr' doms_cur[] array while rebalancing
shares of task groups across cpus.
sched: group scheduling, change how cpu load is calculated
This patch changes how the cpu load exerted by fair_sched_class tasks
is calculated. Load exerted by fair_sched_class tasks on a cpu is now
a summation of the group weights, rather than summation of task weights.
Weight exerted by a group on a cpu is dependent on the shares allocated
to it.
This version of patch has a minor impact on code size, but should have
no runtime/functional impact for !CONFIG_FAIR_GROUP_SCHED.
- Use a mutex to serialize add/remove of task groups and also when
changing shares of a task group. Use the same mutex when printing
cfs_rq debugging stats for various task groups.
- Use list_for_each_entry_rcu in for_each_leaf_cfs_rq macro (when
walking task group list)
Ingo Molnar [Fri, 25 Jan 2008 20:07:58 +0000 (21:07 +0100)]
printk: make printk more robust by not allowing recursion
make printk more robust by allowing recursion only if there's a crash
going on. Also add recursion detection.
I've tested it with an artificially injected printk recursion - instead
of a lockup or spontaneous reboot or other crash, the output was a well
controlled:
Linus Torvalds [Fri, 25 Jan 2008 16:44:29 +0000 (08:44 -0800)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/selinux-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/selinux-2.6:
selinux: make mls_compute_sid always polyinstantiate
security/selinux: constify function pointer tables and fields
security: add a secctx_to_secid() hook
security: call security_file_permission from rw_verify_area
security: remove security_sb_post_mountroot hook
Security: remove security.h include from mm.h
Security: remove security_file_mmap hook sparse-warnings (NULL as 0).
Security: add get, set, and cloning of superblock security information
security/selinux: Add missing "space"
Linus Torvalds [Fri, 25 Jan 2008 16:40:02 +0000 (08:40 -0800)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6:
[AVR32] extint: Set initial irq type to low level
[AVR32] extint: change set_irq_type() handling
[AVR32] NMI debugging
[AVR32] constify function pointer tables
[AVR32] ATNGW100: Update defconfig
[AVR32] ATSTK1002: Update defconfig
[AVR32] Kconfig: Choose daughterboard instead of CPU
[AVR32] Add support for ATSTK1003 and ATSTK1004
[AVR32] Clean up external DAC setup code
[AVR32] ATSTK1000: Move gpio-leds setup to setup.c
[AVR32] Add support for AT32AP7001 and AT32AP7002
[AVR32] Provide more CPU information in /proc/cpuinfo and dmesg
[AVR32] Oprofile support
[AVR32] Include instrumentation menu
Disable VGA text console for AVR32 architecture
[AVR32] Enable debugging only when needed
ptrace: Call arch_ptrace_attach() when request=PTRACE_TRACEME
[AVR32] Remove redundant try_to_freeze() call from do_signal()
[AVR32] Drop GFP_COMP for DMA memory allocations
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (125 commits)
[CRYPTO] twofish: Merge common glue code
[CRYPTO] hifn_795x: Fixup container_of() usage
[CRYPTO] cast6: inline bloat--
[CRYPTO] api: Set default CRYPTO_MINALIGN to unsigned long long
[CRYPTO] tcrypt: Make xcbc available as a standalone test
[CRYPTO] xcbc: Remove bogus hash/cipher test
[CRYPTO] xcbc: Fix algorithm leak when block size check fails
[CRYPTO] tcrypt: Zero axbuf in the right function
[CRYPTO] padlock: Only reset the key once for each CBC and ECB operation
[CRYPTO] api: Include sched.h for cond_resched in scatterwalk.h
[CRYPTO] salsa20-asm: Remove unnecessary dependency on CRYPTO_SALSA20
[CRYPTO] tcrypt: Add select of AEAD
[CRYPTO] salsa20: Add x86-64 assembly version
[CRYPTO] salsa20_i586: Salsa20 stream cipher algorithm (i586 version)
[CRYPTO] gcm: Introduce rfc4106
[CRYPTO] api: Show async type
[CRYPTO] chainiv: Avoid lock spinning where possible
[CRYPTO] seqiv: Add select AEAD in Kconfig
[CRYPTO] scatterwalk: Handle zero nbytes in scatterwalk_map_and_copy
[CRYPTO] null: Allow setkey on digest_null
...
This can be broken down into these major areas:
- Documentation updates (language translations and fixes, as
well as kobject and kset documenatation updates.)
- major kset/kobject/ktype rework and fixes. This cleans up the
kset and kobject and ktype relationship and architecture,
making sense of things now, and good documenation and samples
are provided for others to use. Also the attributes for
kobjects are much easier to handle now. This cleaned up a LOT
of code all through the kernel, making kobjects easier to use
if you want to.
- struct bus_type has been reworked to now handle the lifetime
rules properly, as the kobject is properly dynamic.
- struct driver has also been reworked, and now the lifetime
issues are resolved.
- the block subsystem has been converted to use struct device
now, and not "raw" kobjects. This patch has been in the -mm
tree for over a year now, and finally all the issues are
worked out with it. Older distros now properly work with new
kernels, and no userspace updates are needed at all.
- nozomi driver is added. This has also been in -mm for a long
time, and many people have asked for it to go in. It is now
in good enough shape to do so.
- lots of class_device conversions to use struct device instead.
The tree is almost all cleaned up now, only SCSI and IB is the
remaining code to fix up...
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-2.6: (196 commits)
Driver core: coding style fixes
Kobject: fix coding style issues in kobject c files
Kobject: fix coding style issues in kobject.h
Driver core: fix coding style issues in device.h
spi: use class iteration api
scsi: use class iteration api
rtc: use class iteration api
power supply : use class iteration api
ieee1394: use class iteration api
Driver Core: add class iteration api
Driver core: Cleanup get_device_parent() in device_add() and device_move()
UIO: constify function pointer tables
Driver Core: constify the name passed to platform_device_register_simple
driver core: fix build with SYSFS=n
sysfs: make SYSFS_DEPRECATED depend on SYSFS
Driver core: use LIST_HEAD instead of call to INIT_LIST_HEAD in __init
kobject: add sample code for how to use ksets/ktypes/kobjects
kobject: add sample code for how to use kobjects in a simple manner.
kobject: update the kobject/kset documentation
kobject: remove old, outdated documentation.
...
Pekka Enberg [Fri, 25 Jan 2008 06:20:51 +0000 (08:20 +0200)]
slab: fix bootstrap on memoryless node
If the node we're booting on doesn't have memory, bootstrapping kmalloc()
caches resorts to fallback_alloc() which requires ->nodelists set for all
nodes. Fix that by calling set_up_list3s() for CACHE_CACHE in
kmem_cache_init().
As kmem_getpages() is called with GFP_THISNODE set, this used to work before
because of breakage in 2.6.22 and before with GFP_THISNODE returning pages from
the wrong node if a node had no memory. So it may have worked accidentally and
in an unsafe manner because the pages would have been associated with the wrong
node which could trigger bug ons and locking troubles.
Tested-by: Mel Gorman <mel@csn.ul.ie> Tested-by: Olaf Hering <olaf@aepfle.de> Reviewed-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
[ With additional one-liner by Olaf Hering - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Karsten Keil [Fri, 25 Jan 2008 10:55:28 +0000 (11:55 +0100)]
fix oops on rmmod capidrv
Fix overwriting the stack with the version string
(it is currently 10 bytes + zero) when unloading the
capidrv module. Safeguard against overwriting it
should the version string grow in the future.
Abhijith Das [Fri, 18 Jan 2008 20:06:37 +0000 (14:06 -0600)]
[GFS2] Allow journal recovery on read-only mount
This patch allows gfs2 to perform journal recovery even if it is mounted
read-only. Strictly speaking, a read-only mount should not be writing to
the filesystem, but we do this only to perform journal recovery. A
read-only mount will fail if we don't recover the dirty journal. Also,
when gfs2 is used as a root filesystem, it will be mounted read-only
before being mounted read-write during the boot sequence. A failed
read-only mount will panic the machine during bootup.
Signed-off-by: Abhijith Das <adas@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
There was a bug in the truncation/invalidation race path for
->page_mkwrite for gfs2. It ought to return 0 so that the effect is the
same as if the page was truncated at any of the other points at which
the page_lock is dropped. This will result in the restart of the whole
page fault path. If it was due to a real truncation (as opposed to an
invalidate because we let a glock go) then the ->fault path will pick
that up when it gets called again.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson [Fri, 11 Jan 2008 19:44:50 +0000 (13:44 -0600)]
[GFS2] gfs2_alloc_required performance
This is a small I/O performance enhancement to gfs2. (Actually, it is a rework of
an earlier version I got wrong). The idea here is to check if the write extends
past the last block in the file. If so, the function can save itself a lot of
time and trouble because it knows an allocate will be required. Benchmarks like
iozone should see better performance.
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson [Fri, 11 Jan 2008 19:31:12 +0000 (13:31 -0600)]
[GFS2] Remove unneeded i_spin
This patch removes a vestigial variable "i_spin" from the gfs2_inode
structure. This not only saves us memory (>300000 of these in memory
for the oom test) it also saves us time because we don't have to
spend time initializing it (i.e. slightly better performance).
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
[GFS2] Reduce inode size by moving i_alloc out of line
It is possible to reduce the size of GFS2 inodes by taking the i_alloc
structure out of the gfs2_inode. This patch allocates the i_alloc
structure whenever its needed, and frees it afterward. This decreases
the amount of low memory we use at the expense of requiring a memory
allocation for each page or partial page that we write. A quick test
with postmark shows that the overhead is not measurable and I also note
that OCFS2 use the same approach.
In the future I'd like to solve the problem by shrinking down the size
of the members of the i_alloc structure, but for now, this reduces the
immediate problem of using too much low-memory on x86 and doesn't add
too much overhead.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Although the values were all being calculated correctly, there was a
race in the assert due to the way it was using atomic variables. This
changes the value we assert on so that we get the same effect by testing
a different variable. This prevents the assert triggering when it shouldn't.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
[GFS2] Fix problems relating to execution of files on GFS2
This patch fixes a couple of problems which affected the execution of files
on GFS2. The first is that there was a corner case where inodes were not
always uptodate at the point at which permissions checks were being carried
out, this was resulting in refusal of execute permission, but only on the
first lookup, subsequent requests worked correctly. The second was a problem
relating to incorrect updating of file sizes which was introduced with the
write_begin/end code for GFS2 a little while back.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Abhijith Das <adas@redhat.com>
Bob Peterson [Thu, 3 Jan 2008 15:24:53 +0000 (09:24 -0600)]
[GFS2] Initialize extent_list earlier
Here is a patch for the latest upstream GFS2 code:
The journal extent map needs to be initialized sooner than it
currently is. Otherwise failed mount attempts (e.g. not enough
journals, etc.) may panic trying to access the uninitialized list.
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
[GFS2] Allow page migration for writeback and ordered pages
To improve performance on NUMA, we use the VM's standard page
migration for writeback and ordered pages. Probably we could
also do the same for journaled data, but that would need a
careful audit of the code, so will be the subject of a later
patch.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson [Wed, 12 Dec 2007 23:52:13 +0000 (17:52 -0600)]
[GFS2] Minor correction
This is a small correction to my previously posted patch1.
It just changes a divide to a shift. It's faster and doesn't
introduce odd dependencies on 32-bit compiles.
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson [Wed, 12 Dec 2007 01:29:17 +0000 (19:29 -0600)]
[GFS2] Function meta_read optimization
This patch optimizes function gfs2_meta_read. Basically, gfs2_meta_wait
was being called regardless of whether a disk read was requested.
This just pulls that wait into the if that triggers the read.
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson [Wed, 12 Dec 2007 01:13:54 +0000 (19:13 -0600)]
[GFS2] Reorganize function gfs2_glmutex_lock
This patch optimizes the function gfs2_glmutex_lock.
The basic theory is: Why bother initializing a holder, setting up
wait bits and then waiting on them, if you know the glock can be
yours. So the holder stuff is placed inside the if checking if the
glock is locked. This one needs careful scrutiny because changing
anything to do with locking should strike terror into one's heart.
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson [Wed, 12 Dec 2007 01:00:16 +0000 (19:00 -0600)]
[GFS2] Run through full bitmaps quicker in gfs2_bitfit
I eliminated the passing of an unused parameter into gfs2_bitfit called rgd.
This also changes the gfs2_bitfit code that searches for free (or used) blocks.
Before, the code was trying to check for bytes that indicated 4 blocks in
the undesired state. The problem is, it was spending more time trying to
do this than it actually was saving. This version only optimizes the case
where we're looking for free blocks, and it checks a machine word at a time.
So on 32-bit machines, it will check 32-bits (16 blocks) and on 64-bit
machines, it will check 64-bits (32 blocks) at a time. The compiler
optimizes that quite well and we save some time, especially when running
through full bitmaps (like the bitmaps allocated for the journals).
There's probably a more elegant or optimized way to do this, but I haven't
thought of it yet. I'm open to suggestions.
Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
projects / karo-tx-linux.git / log