Previously, to probe the working set of a task, we'd use
a very simple and crude method: mark all of its address
space PROT_NONE.
That method has various (obvious) disadvantages:
- it samples the working set at dissimilar rates,
giving some tasks a sampling quality advantage
over others.
- creates performance problems for tasks with very
large working sets
- over-samples processes with large address spaces but
which only very rarely execute
Improve that method by keeping a rotating offset into the
address space that marks the current position of the scan,
and advance it by a constant rate (in a CPU cycles execution
proportional manner). If the offset reaches the last mapped
address of the mm then it then it starts over at the first
address.
The per-task nature of the working set sampling functionality in this tree
allows such constant rate, per task, execution-weight proportional sampling
of the working set, with an adaptive sampling interval/frequency that
goes from once per 100ms up to just once per 8 seconds. The current
sampling volume is 256 MB per interval.
As tasks mature and converge their working set, so does the
sampling rate slow down to just a trickle, 256 MB per 8
seconds of CPU time executed.
This, beyond being adaptive, also rate-limits rarely
executing systems and does not over-sample on overloaded
systems.
[ In AutoNUMA speak, this patch deals with the effective sampling
rate of the 'hinting page fault'. AutoNUMA's scanning is
currently rate-limited, but it is also fundamentally
single-threaded, executing in the knuma_scand kernel thread,
so the limit in AutoNUMA is global and does not scale up with
the number of CPUs, nor does it scan tasks in an execution
proportional manner.
So the idea of rate-limiting the scanning was first implemented
in the AutoNUMA tree via a global rate limit. This patch goes
beyond that by implementing an execution rate proportional
working set sampling rate that is not implemented via a single
global scanning daemon. ]
[ Dan Carpenter pointed out a possible NULL pointer dereference in the
first version of this patch. ]
Based-on-idea-by: Andrea Arcangeli <aarcange@redhat.com>
Bug-Found-By: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
[ Wrote changelog and fixed bug. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
NOTE: This patch is based on "sched, numa, mm: Add fault driven
placement and migration policy" but as it throws away all the policy
to just leave a basic foundation I had to drop the signed-offs-by.
This patch creates a bare-bones method for setting PTEs pte_numa in the
context of the scheduler that when faulted later will be faulted onto the
node the CPU is running on. In itself this does nothing useful but any
placement policy will fundamentally depend on receiving hints on placement
from fault context and doing something intelligent about it.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: Rik van Riel <riel@redhat.com>
Pull scheduler changes from Ingo Molnar:
"Continued quest to clean up and enhance the cputime code by Frederic
Weisbecker, in preparation for future tickless kernel features.
Other than that, smallish changes."
Fix up trivial conflicts due to additions next to each other in arch/{x86/}Kconfig
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (24 commits)
cputime: Make finegrained irqtime accounting generally available
cputime: Gather time/stats accounting config options into a single menu
ia64: Reuse system and user vtime accounting functions on task switch
ia64: Consolidate user vtime accounting
vtime: Consolidate system/idle context detection
cputime: Use a proper subsystem naming for vtime related APIs
sched: cpu_power: enable ARCH_POWER
sched/nohz: Clean up select_nohz_load_balancer()
sched: Fix load avg vs. cpu-hotplug
sched: Remove __ARCH_WANT_INTERRUPTS_ON_CTXSW
sched: Fix nohz_idle_balance()
sched: Remove useless code in yield_to()
sched: Add time unit suffix to sched sysctl knobs
sched/debug: Limit sd->*_idx range on sysctl
sched: Remove AFFINE_WAKEUPS feature flag
s390: Remove leftover account_tick_vtime() header
cputime: Consolidate vtime handling on context switch
sched: Move cputime code to its own file
cputime: Generalize CONFIG_VIRT_CPU_ACCOUNTING
tile: Remove SD_PREFER_LOCAL leftover
...
This reverts commit 970e178985.
Nikolay Ulyanitsky reported thatthe 3.6-rc5 kernel has a 15-20%
performance drop on PostgreSQL 9.2 on his machine (running "pgbench").
Borislav Petkov was able to reproduce this, and bisected it to this
commit 970e178985 ("sched: Improve scalability via 'CPU buddies' ...")
apparently because the new single-idle-buddy model simply doesn't find
idle CPU's to reschedule on aggressively enough.
Mike Galbraith suspects that it is likely due to the user-mode spinlocks
in PostgreSQL not reacting well to preemption, but we don't really know
the details - I'll just revert the commit for now.
There are hopefully other approaches to improve scheduler scalability
without it causing these kinds of downsides.
Reported-by: Nikolay Ulyanitsky <lystor@gmail.com>
Bisected-by: Borislav Petkov <bp@alien8.de>
Acked-by: Mike Galbraith <efault@gmx.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix two kernel-doc warnings in kernel/sched/fair.c:
Warning(kernel/sched/fair.c:3660): Excess function parameter 'cpus' description in 'update_sg_lb_stats'
Warning(kernel/sched/fair.c:3806): Excess function parameter 'cpus' description in 'update_sd_lb_stats'
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/50303714.3090204@xenotime.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
migrate_tasks() uses _pick_next_task_rt() to get tasks from the
real-time runqueues to be migrated. When rt_rq is throttled
_pick_next_task_rt() won't return anything, in which case
migrate_tasks() can't move all threads over and gets stuck in an
infinite loop.
Instead unthrottle rt runqueues before migrating tasks.
Additionally: move unthrottle_offline_cfs_rqs() to rq_offline_fair()
Signed-off-by: Peter Boonstoppel <pboonstoppel@nvidia.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Turner <pjt@google.com>
Link: http://lkml.kernel.org/r/5FBF8E85CA34454794F0F7ECBA79798F379D3648B7@HQMAIL04.nvidia.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since power saving code was removed from sched now, the implement
code is out of service in this function, and even pollute other logical.
like, 'want_sd' never has chance to be set '0', that remove the effect
of SD_WAKE_AFFINE here.
So, clean up the obsolete code, includes SD_PREFER_LOCAL.
Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/5028F431.6000306@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Peter Portante reported that for large cgroup hierarchies (and or on
large CPU counts) we get immense lock contention on rq->lock and stuff
stops working properly.
His workload was a ton of processes, each in their own cgroup,
everybody idling except for a sporadic wakeup once every so often.
It was found that:
schedule()
idle_balance()
load_balance()
local_irq_save()
double_rq_lock()
update_h_load()
walk_tg_tree(tg_load_down)
tg_load_down()
Results in an entire cgroup hierarchy walk under rq->lock for every
new-idle balance and since new-idle balance isn't throttled this
results in a lot of work while holding the rq->lock.
This patch does two things, it removes the work from under rq->lock
based on the good principle of race and pray which is widely employed
in the load-balancer as a whole. And secondly it throttles the
update_h_load() calculation to max once per jiffy.
I considered excluding update_h_load() for new-idle balance
all-together, but purely relying on regular balance passes to update
this data might not work out under some rare circumstances where the
new-idle busiest isn't the regular busiest for a while (unlikely, but
a nightmare to debug if someone hits it and suffers).
Cc: pjt@google.com
Cc: Larry Woodman <lwoodman@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Reported-by: Peter Portante <pportant@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-aaarrzfpnaam7pqrekofu8a6@git.kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Current load balance scheme requires only one cpu in a
sched_group (balance_cpu) to look at other peer sched_groups for
imbalance and pull tasks towards itself from a busy cpu. Tasks
thus pulled by balance_cpu could later get picked up by cpus
that are in the same sched_group as that of balance_cpu.
This scheme however fails to pull tasks that are not allowed to
run on balance_cpu (but are allowed to run on other cpus in its
sched_group). That can affect fairness and in some worst case
scenarios cause starvation.
Consider a two core (2 threads/core) system running tasks as
below:
Core0 Core1
/ \ / \
C0 C1 C2 C3
| | | |
v v v v
F0 T1 F1 [idle]
T2
F0 = SCHED_FIFO task (pinned to C0)
F1 = SCHED_FIFO task (pinned to C2)
T1 = SCHED_OTHER task (pinned to C1)
T2 = SCHED_OTHER task (pinned to C1 and C2)
F1 could become a cpu hog, which will starve T2 unless C1 pulls
it. Between C0 and C1 however, C0 is required to look for
imbalance between cores, which will fail to pull T2 towards
Core0. T2 will starve eternally in this case. The same scenario
can arise in presence of non-rt tasks as well (say we replace F1
with high irq load).
We tackle this problem by having balance_cpu move pinned tasks
to one of its sibling cpus (where they can run). We first check
if load balance goal can be met by ignoring pinned tasks,
failing which we retry move_tasks() with a new env->dst_cpu.
This patch modifies load balance semantics on who can move load
towards a given cpu in a given sched_domain.
Before this patch, a given_cpu or a ilb_cpu acting on behalf of
an idle given_cpu is responsible for moving load to given_cpu.
With this patch applied, balance_cpu can in addition decide on
moving some load to a given_cpu.
There is a remote possibility that excess load could get moved
as a result of this (balance_cpu and given_cpu/ilb_cpu deciding
*independently* and at *same* time to move some load to a
given_cpu). However we should see less of such conflicting
decisions in practice and moreover subsequent load balance
cycles should correct the excess load moved to given_cpu.
Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Prashanth Nageshappa <prashanth@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/4FE06CDB.2060605@linux.vnet.ibm.com
[ minor edits ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
While load balancing, if all tasks on the source runqueue are pinned,
we retry after excluding the corresponding source cpu. However, loop counters
env.loop and env.loop_break are not reset before retrying, which can lead
to failure in moving the tasks. In this patch we reset env.loop and
env.loop_break to their inital values before we retry.
Signed-off-by: Prashanth Nageshappa <prashanth@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/4FE06EEF.2090709@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Traversing an entire package is not only expensive, it also leads to tasks
bouncing all over a partially idle and possible quite large package. Fix
that up by assigning a 'buddy' CPU to try to motivate. Each buddy may try
to motivate that one other CPU, if it's busy, tough, it may then try its
SMT sibling, but that's all this optimization is allowed to cost.
Sibling cache buddies are cross-wired to prevent bouncing.
4 socket 40 core + SMT Westmere box, single 30 sec tbench runs, higher is better:
clients 1 2 4 8 16 32 64 128
..........................................................................
pre 30 41 118 645 3769 6214 12233 14312
post 299 603 1211 2418 4697 6847 11606 14557
A nice increase in performance.
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1339471112.7352.32.camel@marge.simpson.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Fix lots of new kernel-doc warnings in kernel/sched/fair.c:
Warning(kernel/sched/fair.c:3625): No description found for parameter 'env'
Warning(kernel/sched/fair.c:3625): Excess function parameter 'sd' description in 'update_sg_lb_stats'
Warning(kernel/sched/fair.c:3735): No description found for parameter 'env'
Warning(kernel/sched/fair.c:3735): Excess function parameter 'sd' description in 'update_sd_pick_busiest'
Warning(kernel/sched/fair.c:3735): Excess function parameter 'this_cpu' description in 'update_sd_pick_busiest'
.. more warnings
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Often when we run into mis-shapen topologies the balance iteration
fails to update the cpu power properly and we'll end up in /0 traps.
Always initialize the cpu-power to a semi-sane value so that we can
at least boot the machine, even if the load-balancer might not
function correctly.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-3lbhyj25sr169ha7z3qht5na@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Weird topologies can lead to asymmetric domain setups. This needs
further consideration since these setups are typically non-minimal
too.
For now, make it work by adding an extra mask selecting which CPUs
are allowed to iterate up.
The topology that triggered it is the one from David Rientjes:
10 20 20 30
20 10 20 20
20 20 10 20
30 20 20 10
resulting in boxes that wouldn't even boot.
Reported-by: David Rientjes <rientjes@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-3p86l9cuaqnxz7uxsojmz5rm@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
SD_OVERLAP exists to allow overlapping groups, overlapping groups
appear in NUMA topologies that aren't fully connected.
The typical result of not fully connected NUMA is that each cpu (or
rather node) will have different spans for a particular distance.
However due to how sched domains are traversed -- only the first cpu
in the mask goes one level up -- the next level only cares about the
spans of the cpus that went up.
Due to this two things were observed to be broken:
- build_overlap_sched_groups() -- since its possible the cpu we're
building the groups for exists in multiple (or all) groups, the
selection criteria of the first group didn't ensure there was a cpu
for which is was true that cpumask_first(span) == cpu. Thus load-
balancing would terminate.
- update_group_power() -- assumed that the cpu span of the first
group of the domain was covered by all groups of the child domain.
The above explains why this isn't true, so deal with it.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: David Rientjes <rientjes@google.com>
Link: http://lkml.kernel.org/r/1337788843.9783.14.camel@laptop
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Peter Zijlstra
Linus Torvalds
Alex Shi
Vincent Guittot
Ingo Molnar
Randy Dunlap
Peter Boonstoppel
Michael Wang
Borislav Petkov
Srivatsa Vaddagiri
Prashanth Nageshappa
Mike Galbraith
Peter Zijlstra