IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
The time spent executing IRQ handlers can be significant but it is not reflected
in the utilization of CPU when deciding to choose an OPP. Now that we have
access to this metric, schedutil can take it into account when selecting
the OPP for a CPU.
RQS utilization don't see the time spend under interrupt context and report
their value in the normal context time window. We need to compensate this when
adding interrupt utilization
The CPU utilization is:
IRQ util_avg + (1 - IRQ util_avg / max capacity ) * /Sum rq util_avg
A test with iperf on hikey (octo arm64) gives the following speedup:
iperf -c server_address -r -t 5
w/o patch w/ patch
Tx 276 Mbits/sec 304 Mbits/sec +10%
Rx 299 Mbits/sec 328 Mbits/sec +9%
8 iterations
stdev is lower than 1%
Only WFI idle state is enabled (shallowest idle state).
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten.Rasmussen@arm.com
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: claudio@evidence.eu.com
Cc: daniel.lezcano@linaro.org
Cc: dietmar.eggemann@arm.com
Cc: joel@joelfernandes.org
Cc: juri.lelli@redhat.com
Cc: luca.abeni@santannapisa.it
Cc: patrick.bellasi@arm.com
Cc: quentin.perret@arm.com
Cc: rjw@rjwysocki.net
Cc: valentin.schneider@arm.com
Link: http://lkml.kernel.org/r/1530200714-4504-8-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
interrupt and steal time are the only remaining activities tracked by
rt_avg. Like for sched classes, we can use PELT to track their average
utilization of the CPU. But unlike sched class, we don't track when
entering/leaving interrupt; Instead, we take into account the time spent
under interrupt context when we update rqs' clock (rq_clock_task).
This also means that we have to decay the normal context time and account
for interrupt time during the update.
That's also important to note that because:
rq_clock == rq_clock_task + interrupt time
and rq_clock_task is used by a sched class to compute its utilization, the
util_avg of a sched class only reflects the utilization of the time spent
in normal context and not of the whole time of the CPU. The utilization of
interrupt gives an more accurate level of utilization of CPU.
The CPU utilization is:
avg_irq + (1 - avg_irq / max capacity) * /Sum avg_rq
Most of the time, avg_irq is small and neglictible so the use of the
approximation CPU utilization = /Sum avg_rq was enough.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten.Rasmussen@arm.com
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: claudio@evidence.eu.com
Cc: daniel.lezcano@linaro.org
Cc: dietmar.eggemann@arm.com
Cc: joel@joelfernandes.org
Cc: juri.lelli@redhat.com
Cc: luca.abeni@santannapisa.it
Cc: patrick.bellasi@arm.com
Cc: quentin.perret@arm.com
Cc: rjw@rjwysocki.net
Cc: valentin.schneider@arm.com
Cc: viresh.kumar@linaro.org
Link: http://lkml.kernel.org/r/1530200714-4504-7-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Add both CFS and RT utilization when selecting an OPP for CFS tasks as RT
can preempt and steal CFS's running time.
RT util_avg is used to take into account the utilization of RT tasks
on the CPU when selecting OPP. If a RT task migrate, the RT utilization
will not migrate but will decay over time. On an overloaded CPU, CFS
utilization reflects the remaining utilization avialable on CPU. When RT
task migrates, the CFS utilization will increase when tasks will start to
use the newly available capacity. At the same pace, RT utilization will
decay and both variations will compensate each other to keep unchanged
overall utilization and will prevent any OPP drop.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten.Rasmussen@arm.com
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: claudio@evidence.eu.com
Cc: daniel.lezcano@linaro.org
Cc: dietmar.eggemann@arm.com
Cc: joel@joelfernandes.org
Cc: juri.lelli@redhat.com
Cc: luca.abeni@santannapisa.it
Cc: patrick.bellasi@arm.com
Cc: quentin.perret@arm.com
Cc: rjw@rjwysocki.net
Cc: valentin.schneider@arm.com
Link: http://lkml.kernel.org/r/1530200714-4504-4-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When a new task wakes-up for the first time, its initial utilization
is set to half of the spare capacity of its CPU. The current
implementation of post_init_entity_util_avg() uses SCHED_CAPACITY_SCALE
directly as a capacity reference. As a result, on a big.LITTLE system, a
new task waking up on an idle little CPU will be given ~512 of util_avg,
even if the CPU's capacity is significantly less than that.
Fix this by computing the spare capacity with arch_scale_cpu_capacity().
Signed-off-by: Quentin Perret <quentin.perret@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dietmar.eggemann@arm.com
Cc: morten.rasmussen@arm.com
Cc: patrick.bellasi@arm.com
Link: http://lkml.kernel.org/r/20180612112215.25448-1-quentin.perret@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Mark noticed that syzkaller is able to reliably trigger the following warning:
dl_rq->running_bw > dl_rq->this_bw
WARNING: CPU: 1 PID: 153 at kernel/sched/deadline.c:124 switched_from_dl+0x454/0x608
Kernel panic - not syncing: panic_on_warn set ...
CPU: 1 PID: 153 Comm: syz-executor253 Not tainted 4.18.0-rc3+ #29
Hardware name: linux,dummy-virt (DT)
Call trace:
dump_backtrace+0x0/0x458
show_stack+0x20/0x30
dump_stack+0x180/0x250
panic+0x2dc/0x4ec
__warn_printk+0x0/0x150
report_bug+0x228/0x2d8
bug_handler+0xa0/0x1a0
brk_handler+0x2f0/0x568
do_debug_exception+0x1bc/0x5d0
el1_dbg+0x18/0x78
switched_from_dl+0x454/0x608
__sched_setscheduler+0x8cc/0x2018
sys_sched_setattr+0x340/0x758
el0_svc_naked+0x30/0x34
syzkaller reproducer runs a bunch of threads that constantly switch
between DEADLINE and NORMAL classes while interacting through futexes.
The splat above is caused by the fact that if a DEADLINE task is setattr
back to NORMAL while in non_contending state (blocked on a futex -
inactive timer armed), its contribution to running_bw is not removed
before sub_rq_bw() gets called (!task_on_rq_queued() branch) and the
latter sees running_bw > this_bw.
Fix it by removing a task contribution from running_bw if the task is
not queued and in non_contending state while switched to a different
class.
Reported-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Reviewed-by: Luca Abeni <luca.abeni@santannapisa.it>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: claudio@evidence.eu.com
Cc: rostedt@goodmis.org
Link: http://lkml.kernel.org/r/20180711072948.27061-1-juri.lelli@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Gaurav reports that commit:
85f1abe001 ("kthread, sched/wait: Fix kthread_parkme() completion issue")
isn't working for him. Because of the following race:
> controller Thread CPUHP Thread
> takedown_cpu
> kthread_park
> kthread_parkme
> Set KTHREAD_SHOULD_PARK
> smpboot_thread_fn
> set Task interruptible
>
>
> wake_up_process
> if (!(p->state & state))
> goto out;
>
> Kthread_parkme
> SET TASK_PARKED
> schedule
> raw_spin_lock(&rq->lock)
> ttwu_remote
> waiting for __task_rq_lock
> context_switch
>
> finish_lock_switch
>
>
>
> Case TASK_PARKED
> kthread_park_complete
>
>
> SET Running
Furthermore, Oleg noticed that the whole scheduler TASK_PARKED
handling is buggered because the TASK_DEAD thing is done with
preemption disabled, the current code can still complete early on
preemption :/
So basically revert that earlier fix and go with a variant of the
alternative mentioned in the commit. Promote TASK_PARKED to special
state to avoid the store-store issue on task->state leading to the
WARN in kthread_unpark() -> __kthread_bind().
But in addition, add wait_task_inactive() to kthread_park() to ensure
the task really is PARKED when we return from kthread_park(). This
avoids the whole kthread still gets migrated nonsense -- although it
would be really good to get this done differently.
Reported-by: Gaurav Kohli <gkohli@codeaurora.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 85f1abe001 ("kthread, sched/wait: Fix kthread_parkme() completion issue")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When a cfs_rq is throttled, parent cfs_rq->nr_running is decreased and
everything happens at cfs_rq level. Currently util_est stays unchanged
in such case and it keeps accounting the utilization of throttled tasks.
This can somewhat make sense as we don't dequeue tasks but only throttled
cfs_rq.
If a task of another group is enqueued/dequeued and root cfs_rq becomes
idle during the dequeue, util_est will be cleared whereas it was
accounting util_est of throttled tasks before. So the behavior of util_est
is not always the same regarding throttled tasks and depends of side
activity. Furthermore, util_est will not be updated when the cfs_rq is
unthrottled as everything happens at cfs_rq level. Main results is that
util_est will stay null whereas we now have running tasks. We have to wait
for the next dequeue/enqueue of the previously throttled tasks to get an
up to date util_est.
Remove the assumption that cfs_rq's estimated utilization of a CPU is 0
if there is no running task so the util_est of a task remains until the
latter is dequeued even if its cfs_rq has been throttled.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 7f65ea42eb ("sched/fair: Add util_est on top of PELT")
Link: http://lkml.kernel.org/r/1528972380-16268-1-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When period gets restarted after some idle time, start_cfs_bandwidth()
doesn't update the expiration information, expire_cfs_rq_runtime() will
see cfs_rq->runtime_expires smaller than rq clock and go to the clock
drift logic, wasting needless CPU cycles on the scheduler hot path.
Update the global expiration in start_cfs_bandwidth() to avoid frequent
expire_cfs_rq_runtime() calls once a new period begins.
Signed-off-by: Xunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Ben Segall <bsegall@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180620101834.24455-2-xlpang@linux.alibaba.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
I noticed that cgroup task groups constantly get throttled even
if they have low CPU usage, this causes some jitters on the response
time to some of our business containers when enabling CPU quotas.
It's very simple to reproduce:
mkdir /sys/fs/cgroup/cpu/test
cd /sys/fs/cgroup/cpu/test
echo 100000 > cpu.cfs_quota_us
echo $$ > tasks
then repeat:
cat cpu.stat | grep nr_throttled # nr_throttled will increase steadily
After some analysis, we found that cfs_rq::runtime_remaining will
be cleared by expire_cfs_rq_runtime() due to two equal but stale
"cfs_{b|q}->runtime_expires" after period timer is re-armed.
The current condition to judge clock drift in expire_cfs_rq_runtime()
is wrong, the two runtime_expires are actually the same when clock
drift happens, so this condtion can never hit. The orginal design was
correctly done by this commit:
a9cf55b286 ("sched: Expire invalid runtime")
... but was changed to be the current implementation due to its locking bug.
This patch introduces another way, it adds a new field in both structures
cfs_rq and cfs_bandwidth to record the expiration update sequence, and
uses them to figure out if clock drift happens (true if they are equal).
Signed-off-by: Xunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Ben Segall <bsegall@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 51f2176d74 ("sched/fair: Fix unlocked reads of some cfs_b->quota/period")
Link: http://lkml.kernel.org/r/20180620101834.24455-1-xlpang@linux.alibaba.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
With commit:
8f111bc357 ("cpufreq/schedutil: Rewrite CPUFREQ_RT support")
the schedutil governor uses rq->rt.rt_nr_running to detect whether an
RT task is currently running on the CPU and to set frequency to max
if necessary.
cpufreq_update_util() is called in enqueue/dequeue_top_rt_rq() but
rq->rt.rt_nr_running has not been updated yet when dequeue_top_rt_rq() is
called so schedutil still considers that an RT task is running when the
last task is dequeued. The update of rq->rt.rt_nr_running happens later
in dequeue_rt_stack().
In fact, we can take advantage of the sequence that the dequeue then
re-enqueue rt entities when a rt task is enqueued or dequeued;
As a result enqueue_top_rt_rq() is always called when a task is
enqueued or dequeued and also when groups are throttled or unthrottled.
The only place that not use enqueue_top_rt_rq() is when root rt_rq is
throttled.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: efault@gmx.de
Cc: juri.lelli@redhat.com
Cc: patrick.bellasi@arm.com
Cc: viresh.kumar@linaro.org
Fixes: 8f111bc357 ('cpufreq/schedutil: Rewrite CPUFREQ_RT support')
Link: http://lkml.kernel.org/r/1530021202-21695-1-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Some people have reported that the warning in sched_tick_remote()
occasionally triggers, especially in favour of some RCU-Torture
pressure:
WARNING: CPU: 11 PID: 906 at kernel/sched/core.c:3138 sched_tick_remote+0xb6/0xc0
Modules linked in:
CPU: 11 PID: 906 Comm: kworker/u32:3 Not tainted 4.18.0-rc2+ #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
Workqueue: events_unbound sched_tick_remote
RIP: 0010:sched_tick_remote+0xb6/0xc0
Code: e8 0f 06 b8 00 c6 03 00 fb eb 9d 8b 43 04 85 c0 75 8d 48 8b 83 e0 0a 00 00 48 85 c0 75 81 eb 88 48 89 df e8 bc fe ff ff eb aa <0f> 0b eb
+c5 66 0f 1f 44 00 00 bf 17 00 00 00 e8 b6 2e fe ff 0f b6
Call Trace:
process_one_work+0x1df/0x3b0
worker_thread+0x44/0x3d0
kthread+0xf3/0x130
? set_worker_desc+0xb0/0xb0
? kthread_create_worker_on_cpu+0x70/0x70
ret_from_fork+0x35/0x40
This happens when the remote tick applies on an idle task. Usually the
idle_cpu() check avoids that, but it is performed before we lock the
runqueue and it is therefore racy. It was intended to be that way in
order to prevent from useless runqueue locks since idle task tick
callback is a no-op.
Now if the racy check slips out of our hands and we end up remotely
ticking an idle task, the empty task_tick_idle() is harmless. Still
it won't pass the WARN_ON_ONCE() test that ensures rq_clock_task() is
not too far from curr->se.exec_start because update_curr_idle() doesn't
update the exec_start value like other scheduler policies. Hence the
reported false positive.
So let's have another check, while the rq is locked, to make sure we
don't remote tick on an idle task. The lockless idle_cpu() still applies
to avoid unecessary rq lock contention.
Reported-by: Jacek Tomaka <jacekt@dug.com>
Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reported-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1530203381-31234-1-git-send-email-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
After commit:
82958366cf ("sched: Replace update_shares weight distribution with per-entity computation")
tg_unthrottle_up() did not update the weight.
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/lkml/1523423816-18322-1-git-send-email-lirongqing@baidu.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
match_string() returns the index of an array for a matching string,
which can be used instead of the open coded variant.
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/lkml/1527765086-19873-15-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The static key sched_smt_present is only updated at boot time when SMT
siblings have been detected. Booting with maxcpus=1 and bringing the
siblings online after boot rebuilds the scheduling domains correctly but
does not update the static key, so the SMT code is not enabled.
Let the key be updated in the scheduler CPU hotplug code to fix this.
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Linus noted that swait basically implements exclusive mode -- because
swake_up() only wakes a single waiter. And because of that it should
take care to properly deal with the interruptible case.
In short, the problem is that swake_up() can race with a signal. In
this this case it is possible the swake_up() 'wakes' the waiter that
is already on the way out because it just got a signal and the wakeup
gets lost.
The normal wait code is very careful and avoids this situation, make
sure we do too.
Copy the exact exclusive semantics from wait.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: bigeasy@linutronix.de
Cc: oleg@redhat.com
Cc: paulmck@linux.vnet.ibm.com
Cc: pbonzini@redhat.com
Link: https://lkml.kernel.org/r/20180612083909.209762413@infradead.org
During a context switch, we first switch_mm() to the next task's mm,
then switch_to() that new task. This means that vmalloc'd regions which
had previously been faulted in can transiently disappear in the context
of the prev task.
Functions instrumented by KCOV may try to access a vmalloc'd kcov_area
during this window, and as the fault handling code is instrumented, this
results in a recursive fault.
We must avoid accessing any kcov_area during this window. We can do so
with a new flag in kcov_mode, set prior to switching the mm, and cleared
once the new task is live. Since task_struct::kcov_mode isn't always a
specific enum kcov_mode value, this is made an unsigned int.
The manipulation is hidden behind kcov_{prepare,finish}_switch() helpers,
which are empty for !CONFIG_KCOV kernels.
The code uses macros because I can't use static inline functions without a
circular include dependency between <linux/sched.h> and <linux/kcov.h>,
since the definition of task_struct uses things defined in <linux/kcov.h>
Link: http://lkml.kernel.org/r/20180504135535.53744-4-mark.rutland@arm.com
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull restartable sequence support from Thomas Gleixner:
"The restartable sequences syscall (finally):
After a lot of back and forth discussion and massive delays caused by
the speculative distraction of maintainers, the core set of
restartable sequences has finally reached a consensus.
It comes with the basic non disputed core implementation along with
support for arm, powerpc and x86 and a full set of selftests
It was exposed to linux-next earlier this week, so it does not fully
comply with the merge window requirements, but there is really no
point to drag it out for yet another cycle"
* 'core-rseq-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
rseq/selftests: Provide Makefile, scripts, gitignore
rseq/selftests: Provide parametrized tests
rseq/selftests: Provide basic percpu ops test
rseq/selftests: Provide basic test
rseq/selftests: Provide rseq library
selftests/lib.mk: Introduce OVERRIDE_TARGETS
powerpc: Wire up restartable sequences system call
powerpc: Add syscall detection for restartable sequences
powerpc: Add support for restartable sequences
x86: Wire up restartable sequence system call
x86: Add support for restartable sequences
arm: Wire up restartable sequences system call
arm: Add syscall detection for restartable sequences
arm: Add restartable sequences support
rseq: Introduce restartable sequences system call
uapi/headers: Provide types_32_64.h
Expose a new system call allowing each thread to register one userspace
memory area to be used as an ABI between kernel and user-space for two
purposes: user-space restartable sequences and quick access to read the
current CPU number value from user-space.
* Restartable sequences (per-cpu atomics)
Restartables sequences allow user-space to perform update operations on
per-cpu data without requiring heavy-weight atomic operations.
The restartable critical sections (percpu atomics) work has been started
by Paul Turner and Andrew Hunter. It lets the kernel handle restart of
critical sections. [1] [2] The re-implementation proposed here brings a
few simplifications to the ABI which facilitates porting to other
architectures and speeds up the user-space fast path.
Here are benchmarks of various rseq use-cases.
Test hardware:
arm32: ARMv7 Processor rev 4 (v7l) "Cubietruck", 2-core
x86-64: Intel E5-2630 v3@2.40GHz, 16-core, hyperthreading
The following benchmarks were all performed on a single thread.
* Per-CPU statistic counter increment
getcpu+atomic (ns/op) rseq (ns/op) speedup
arm32: 344.0 31.4 11.0
x86-64: 15.3 2.0 7.7
* LTTng-UST: write event 32-bit header, 32-bit payload into tracer
per-cpu buffer
getcpu+atomic (ns/op) rseq (ns/op) speedup
arm32: 2502.0 2250.0 1.1
x86-64: 117.4 98.0 1.2
* liburcu percpu: lock-unlock pair, dereference, read/compare word
getcpu+atomic (ns/op) rseq (ns/op) speedup
arm32: 751.0 128.5 5.8
x86-64: 53.4 28.6 1.9
* jemalloc memory allocator adapted to use rseq
Using rseq with per-cpu memory pools in jemalloc at Facebook (based on
rseq 2016 implementation):
The production workload response-time has 1-2% gain avg. latency, and
the P99 overall latency drops by 2-3%.
* Reading the current CPU number
Speeding up reading the current CPU number on which the caller thread is
running is done by keeping the current CPU number up do date within the
cpu_id field of the memory area registered by the thread. This is done
by making scheduler preemption set the TIF_NOTIFY_RESUME flag on the
current thread. Upon return to user-space, a notify-resume handler
updates the current CPU value within the registered user-space memory
area. User-space can then read the current CPU number directly from
memory.
Keeping the current cpu id in a memory area shared between kernel and
user-space is an improvement over current mechanisms available to read
the current CPU number, which has the following benefits over
alternative approaches:
- 35x speedup on ARM vs system call through glibc
- 20x speedup on x86 compared to calling glibc, which calls vdso
executing a "lsl" instruction,
- 14x speedup on x86 compared to inlined "lsl" instruction,
- Unlike vdso approaches, this cpu_id value can be read from an inline
assembly, which makes it a useful building block for restartable
sequences.
- The approach of reading the cpu id through memory mapping shared
between kernel and user-space is portable (e.g. ARM), which is not the
case for the lsl-based x86 vdso.
On x86, yet another possible approach would be to use the gs segment
selector to point to user-space per-cpu data. This approach performs
similarly to the cpu id cache, but it has two disadvantages: it is
not portable, and it is incompatible with existing applications already
using the gs segment selector for other purposes.
Benchmarking various approaches for reading the current CPU number:
ARMv7 Processor rev 4 (v7l)
Machine model: Cubietruck
- Baseline (empty loop): 8.4 ns
- Read CPU from rseq cpu_id: 16.7 ns
- Read CPU from rseq cpu_id (lazy register): 19.8 ns
- glibc 2.19-0ubuntu6.6 getcpu: 301.8 ns
- getcpu system call: 234.9 ns
x86-64 Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz:
- Baseline (empty loop): 0.8 ns
- Read CPU from rseq cpu_id: 0.8 ns
- Read CPU from rseq cpu_id (lazy register): 0.8 ns
- Read using gs segment selector: 0.8 ns
- "lsl" inline assembly: 13.0 ns
- glibc 2.19-0ubuntu6 getcpu: 16.6 ns
- getcpu system call: 53.9 ns
- Speed (benchmark taken on v8 of patchset)
Running 10 runs of hackbench -l 100000 seems to indicate, contrary to
expectations, that enabling CONFIG_RSEQ slightly accelerates the
scheduler:
Configuration: 2 sockets * 8-core Intel(R) Xeon(R) CPU E5-2630 v3 @
2.40GHz (directly on hardware, hyperthreading disabled in BIOS, energy
saving disabled in BIOS, turboboost disabled in BIOS, cpuidle.off=1
kernel parameter), with a Linux v4.6 defconfig+localyesconfig,
restartable sequences series applied.
* CONFIG_RSEQ=n
avg.: 41.37 s
std.dev.: 0.36 s
* CONFIG_RSEQ=y
avg.: 40.46 s
std.dev.: 0.33 s
- Size
On x86-64, between CONFIG_RSEQ=n/y, the text size increase of vmlinux is
567 bytes, and the data size increase of vmlinux is 5696 bytes.
[1] https://lwn.net/Articles/650333/
[2] http://www.linuxplumbersconf.org/2013/ocw/system/presentations/1695/original/LPC%20-%20PerCpu%20Atomics.pdf
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Watson <davejwatson@fb.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: "H . Peter Anvin" <hpa@zytor.com>
Cc: Chris Lameter <cl@linux.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Andrew Hunter <ahh@google.com>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: "Paul E . McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ben Maurer <bmaurer@fb.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: linux-api@vger.kernel.org
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20151027235635.16059.11630.stgit@pjt-glaptop.roam.corp.google.com
Link: http://lkml.kernel.org/r/20150624222609.6116.86035.stgit@kitami.mtv.corp.google.com
Link: https://lkml.kernel.org/r/20180602124408.8430-3-mathieu.desnoyers@efficios.com
These include a significant update of the generic power domains (genpd)
and Operating Performance Points (OPP) frameworks, mostly related to
the introduction of power domain performance levels, cpufreq updates
(new driver for Qualcomm Kryo processors, updates of the existing
drivers, some core fixes, schedutil governor improvements), PCI power
management fixes, ACPI workaround for EC-based wakeup events handling
on resume from suspend-to-idle, and major updates of the turbostat
and pm-graph utilities.
Specifics:
- Introduce power domain performance levels into the the generic
power domains (genpd) and Operating Performance Points (OPP)
frameworks (Viresh Kumar, Rajendra Nayak, Dan Carpenter).
- Fix two issues in the runtime PM framework related to the
initialization and removal of devices using device links (Ulf
Hansson).
- Clean up the initialization of drivers for devices in PM domains
(Ulf Hansson, Geert Uytterhoeven).
- Fix a cpufreq core issue related to the policy sysfs interface
causing CPU online to fail for CPUs sharing one cpufreq policy in
some situations (Tao Wang).
- Make it possible to use platform-specific suspend/resume hooks
in the cpufreq-dt driver and make the Armada 37xx DVFS use that
feature (Viresh Kumar, Miquel Raynal).
- Optimize policy transition notifications in cpufreq (Viresh Kumar).
- Improve the iowait boost mechanism in the schedutil cpufreq
governor (Patrick Bellasi).
- Improve the handling of deferred frequency updates in the
schedutil cpufreq governor (Joel Fernandes, Dietmar Eggemann,
Rafael Wysocki, Viresh Kumar).
- Add a new cpufreq driver for Qualcomm Kryo (Ilia Lin).
- Fix and clean up some cpufreq drivers (Colin Ian King, Dmitry
Osipenko, Doug Smythies, Luc Van Oostenryck, Simon Horman,
Viresh Kumar).
- Fix the handling of PCI devices with the DPM_SMART_SUSPEND flag
set and update stale comments in the PCI core PM code (Rafael
Wysocki).
- Work around an issue related to the handling of EC-based wakeup
events in the ACPI PM core during resume from suspend-to-idle if
the EC has been put into the low-power mode (Rafael Wysocki).
- Improve the handling of wakeup source objects in the PM core (Doug
Berger, Mahendran Ganesh, Rafael Wysocki).
- Update the driver core to prevent deferred probe from breaking
suspend/resume ordering (Feng Kan).
- Clean up the PM core somewhat (Bjorn Helgaas, Ulf Hansson, Rafael
Wysocki).
- Make the core suspend/resume code and cpufreq support the RT patch
(Sebastian Andrzej Siewior, Thomas Gleixner).
- Consolidate the PM QoS handling in cpuidle governors (Rafael
Wysocki).
- Fix a possible crash in the hibernation core (Tetsuo Handa).
- Update the rockchip-io Adaptive Voltage Scaling (AVS) driver
(David Wu).
- Update the turbostat utility (fixes, cleanups, new CPU IDs, new
command line options, built-in "Low Power Idle" counters support,
new POLL and POLL% columns) and add an entry for it to MAINTAINERS
(Len Brown, Artem Bityutskiy, Chen Yu, Laura Abbott, Matt Turner,
Prarit Bhargava, Srinivas Pandruvada).
- Update the pm-graph to version 5.1 (Todd Brandt).
- Update the intel_pstate_tracer utility (Doug Smythies).
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQIcBAABCAAGBQJbFRzjAAoJEILEb/54YlRxREQQAKD7IjnLA86ZDkmwiwzFa9Cz
OJ0qlKAcMZGjeWH6LYq7lqWtaJ5PcFkBwNB4sRyKFdGPQOX3Ph8ZzILm2j8hhma4
Azn9632P6CoYHABa8Vof+A1BZ/j0aWtvtJEfqXhtF6rAYyWQlF0UmOIRsMs+54a+
Z/w4WuLaX8qYq3JlR60TogNtTIbdUjkjfvxMGrE9OSQ8n4oEhqoF/v0WoTHYLpWw
fu81M378axOu0Sgq1ZQ8GPUdblUqIO97iWwF7k2YUl7D9n5dm4wOhXDz3CLI8Cdb
RkoFFdp8bJIthbc5desKY2XFU1ClY8lxEVMXewFzTGwWMw0OyWgQP0/ZiG+Mujq3
CSbstg8GGpbwQoWU+VrluYa0FtqofV2UaGk1gOuPaojMqaIchRU4Nmbd2U6naNwp
XN7A1DzrOVGEt0ny8ztKH2Oqmj+NOCcRsChlYzdhLQ1wlqG54iCGwAML2ZJF9/Nw
0Sx8hm6eyWLzjSa0L384Msb+v5oqCoac66gPHCl2x7W+3F+jmqx1KbmkI2SRNUAL
7CS9lcImpvC4uZB54Aqya104vfqHiDse7WP0GrKqOmNVucD7hYCPiq/pycLwez+b
V3zLyvly8PsuBIa4AOQGGiK45HGpaKuB4TkRqRyFO0Fb5uL1M+Ld6kJiWlacl4az
STEUjY/90SRQvX3ocGyB
=wqBV
-----END PGP SIGNATURE-----
Merge tag 'pm-4.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management updates from Rafael Wysocki:
"These include a significant update of the generic power domains
(genpd) and Operating Performance Points (OPP) frameworks, mostly
related to the introduction of power domain performance levels,
cpufreq updates (new driver for Qualcomm Kryo processors, updates of
the existing drivers, some core fixes, schedutil governor
improvements), PCI power management fixes, ACPI workaround for
EC-based wakeup events handling on resume from suspend-to-idle, and
major updates of the turbostat and pm-graph utilities.
Specifics:
- Introduce power domain performance levels into the the generic
power domains (genpd) and Operating Performance Points (OPP)
frameworks (Viresh Kumar, Rajendra Nayak, Dan Carpenter).
- Fix two issues in the runtime PM framework related to the
initialization and removal of devices using device links (Ulf
Hansson).
- Clean up the initialization of drivers for devices in PM domains
(Ulf Hansson, Geert Uytterhoeven).
- Fix a cpufreq core issue related to the policy sysfs interface
causing CPU online to fail for CPUs sharing one cpufreq policy in
some situations (Tao Wang).
- Make it possible to use platform-specific suspend/resume hooks in
the cpufreq-dt driver and make the Armada 37xx DVFS use that
feature (Viresh Kumar, Miquel Raynal).
- Optimize policy transition notifications in cpufreq (Viresh Kumar).
- Improve the iowait boost mechanism in the schedutil cpufreq
governor (Patrick Bellasi).
- Improve the handling of deferred frequency updates in the schedutil
cpufreq governor (Joel Fernandes, Dietmar Eggemann, Rafael Wysocki,
Viresh Kumar).
- Add a new cpufreq driver for Qualcomm Kryo (Ilia Lin).
- Fix and clean up some cpufreq drivers (Colin Ian King, Dmitry
Osipenko, Doug Smythies, Luc Van Oostenryck, Simon Horman, Viresh
Kumar).
- Fix the handling of PCI devices with the DPM_SMART_SUSPEND flag set
and update stale comments in the PCI core PM code (Rafael Wysocki).
- Work around an issue related to the handling of EC-based wakeup
events in the ACPI PM core during resume from suspend-to-idle if
the EC has been put into the low-power mode (Rafael Wysocki).
- Improve the handling of wakeup source objects in the PM core (Doug
Berger, Mahendran Ganesh, Rafael Wysocki).
- Update the driver core to prevent deferred probe from breaking
suspend/resume ordering (Feng Kan).
- Clean up the PM core somewhat (Bjorn Helgaas, Ulf Hansson, Rafael
Wysocki).
- Make the core suspend/resume code and cpufreq support the RT patch
(Sebastian Andrzej Siewior, Thomas Gleixner).
- Consolidate the PM QoS handling in cpuidle governors (Rafael
Wysocki).
- Fix a possible crash in the hibernation core (Tetsuo Handa).
- Update the rockchip-io Adaptive Voltage Scaling (AVS) driver (David
Wu).
- Update the turbostat utility (fixes, cleanups, new CPU IDs, new
command line options, built-in "Low Power Idle" counters support,
new POLL and POLL% columns) and add an entry for it to MAINTAINERS
(Len Brown, Artem Bityutskiy, Chen Yu, Laura Abbott, Matt Turner,
Prarit Bhargava, Srinivas Pandruvada).
- Update the pm-graph to version 5.1 (Todd Brandt).
- Update the intel_pstate_tracer utility (Doug Smythies)"
* tag 'pm-4.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (128 commits)
tools/power turbostat: update version number
tools/power turbostat: Add Node in output
tools/power turbostat: add node information into turbostat calculations
tools/power turbostat: remove num_ from cpu_topology struct
tools/power turbostat: rename num_cores_per_pkg to num_cores_per_node
tools/power turbostat: track thread ID in cpu_topology
tools/power turbostat: Calculate additional node information for a package
tools/power turbostat: Fix node and siblings lookup data
tools/power turbostat: set max_num_cpus equal to the cpumask length
tools/power turbostat: if --num_iterations, print for specific number of iterations
tools/power turbostat: Add Cannon Lake support
tools/power turbostat: delete duplicate #defines
x86: msr-index.h: Correct SNB_C1/C3_AUTO_UNDEMOTE defines
tools/power turbostat: Correct SNB_C1/C3_AUTO_UNDEMOTE defines
tools/power turbostat: add POLL and POLL% column
tools/power turbostat: Fix --hide Pk%pc10
tools/power turbostat: Build-in "Low Power Idle" counters support
tools/power turbostat: Don't make man pages executable
tools/power turbostat: remove blank lines
tools/power turbostat: a small C-states dump readability immprovement
...
Pull scheduler updates from Ingo Molnar:
- power-aware scheduling improvements (Patrick Bellasi)
- NUMA balancing improvements (Mel Gorman)
- vCPU scheduling fixes (Rohit Jain)
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/fair: Update util_est before updating schedutil
sched/cpufreq: Modify aggregate utilization to always include blocked FAIR utilization
sched/deadline/Documentation: Add overrun signal and GRUB-PA documentation
sched/core: Distinguish between idle_cpu() calls based on desired effect, introduce available_idle_cpu()
sched/wait: Include <linux/wait.h> in <linux/swait.h>
sched/numa: Stagger NUMA balancing scan periods for new threads
sched/core: Don't schedule threads on pre-empted vCPUs
sched/fair: Avoid calling sync_entity_load_avg() unnecessarily
sched/fair: Rearrange select_task_rq_fair() to optimize it
Pull RCU updates from Ingo Molnar:
- updates to the handling of expedited grace periods
- updates to reduce lock contention in the rcu_node combining tree
[ These are in preparation for the consolidation of RCU-bh,
RCU-preempt, and RCU-sched into a single flavor, which was
requested by Linus in response to a security flaw whose root cause
included confusion between the multiple flavors of RCU ]
- torture-test updates that save their users some time and effort
- miscellaneous fixes
* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (44 commits)
rcu/x86: Provide early rcu_cpu_starting() callback
torture: Make kvm-find-errors.sh find build warnings
rcutorture: Abbreviate kvm.sh summary lines
rcutorture: Print end-of-test state in kvm.sh summary
rcutorture: Print end-of-test state
torture: Fold parse-torture.sh into parse-console.sh
torture: Add a script to edit output from failed runs
rcu: Update list of rcu_future_grace_period() trace events
rcu: Drop early GP request check from rcu_gp_kthread()
rcu: Simplify and inline cpu_needs_another_gp()
rcu: The rcu_gp_cleanup() function does not need cpu_needs_another_gp()
rcu: Make rcu_start_this_gp() check for out-of-range requests
rcu: Add funnel locking to rcu_start_this_gp()
rcu: Make rcu_start_future_gp() caller select grace period
rcu: Inline rcu_start_gp_advanced() into rcu_start_future_gp()
rcu: Clear request other than RCU_GP_FLAG_INIT at GP end
rcu: Cleanup, don't put ->completed into an int
rcu: Switch __rcu_process_callbacks() to rcu_accelerate_cbs()
rcu: Avoid __call_rcu_core() root rcu_node ->lock acquisition
rcu: Make rcu_migrate_callbacks wake GP kthread when needed
...
A missing clock update is causing the following warning:
rq->clock_update_flags < RQCF_ACT_SKIP
WARNING: CPU: 10 PID: 0 at kernel/sched/sched.h:963 inactive_task_timer+0x5d6/0x720
Call Trace:
<IRQ>
__hrtimer_run_queues+0x10f/0x530
hrtimer_interrupt+0xe5/0x240
smp_apic_timer_interrupt+0x79/0x2b0
apic_timer_interrupt+0xf/0x20
</IRQ>
do_idle+0x203/0x280
cpu_startup_entry+0x6f/0x80
start_secondary+0x1b0/0x200
secondary_startup_64+0xa5/0xb0
hardirqs last enabled at (793919): [<ffffffffa27c5f6e>] cpuidle_enter_state+0x9e/0x360
hardirqs last disabled at (793920): [<ffffffffa2a0096e>] interrupt_entry+0xce/0xe0
softirqs last enabled at (793922): [<ffffffffa20bef78>] irq_enter+0x68/0x70
softirqs last disabled at (793921): [<ffffffffa20bef5d>] irq_enter+0x4d/0x70
This happens because inactive_task_timer() calls sub_running_bw() (if
TASK_DEAD and non_contending) that might trigger a schedutil update,
which might access the clock. Clock is however currently updated only
later in inactive_task_timer() function.
Fix the problem by updating the clock right after task_rq_lock().
Reported-by: kernel test robot <xiaolong.ye@intel.com>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Claudio Scordino <claudio@evidence.eu.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luca Abeni <luca.abeni@santannapisa.it>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180530160809.9074-1-juri.lelli@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
select_task_rq() is used in a few paths to select the CPU upon which a
thread should be run - for example it is used by try_to_wake_up() & by
fork or exec balancing. As-is it allows use of any online CPU that is
present in the task's cpus_allowed mask.
This presents a problem because there is a period whilst CPUs are
brought online where a CPU is marked online, but is not yet fully
initialized - ie. the period where CPUHP_AP_ONLINE_IDLE <= state <
CPUHP_ONLINE. Usually we don't run any user tasks during this window,
but there are corner cases where this can happen. An example observed
is:
- Some user task A, running on CPU X, forks to create task B.
- sched_fork() calls __set_task_cpu() with cpu=X, setting task B's
task_struct::cpu field to X.
- CPU X is offlined.
- Task A, currently somewhere between the __set_task_cpu() in
copy_process() and the call to wake_up_new_task(), is migrated to
CPU Y by migrate_tasks() when CPU X is offlined.
- CPU X is onlined, but still in the CPUHP_AP_ONLINE_IDLE state. The
scheduler is now active on CPU X, but there are no user tasks on
the runqueue.
- Task A runs on CPU Y & reaches wake_up_new_task(). This calls
select_task_rq() with cpu=X, taken from task B's task_struct,
and select_task_rq() allows CPU X to be returned.
- Task A enqueues task B on CPU X's runqueue, via activate_task() &
enqueue_task().
- CPU X now has a user task on its runqueue before it has reached the
CPUHP_ONLINE state.
In most cases, the user tasks that schedule on the newly onlined CPU
have no idea that anything went wrong, but one case observed to be
problematic is if the task goes on to invoke the sched_setaffinity
syscall. The newly onlined CPU reaches the CPUHP_AP_ONLINE_IDLE state
before the CPU that brought it online calls stop_machine_unpark(). This
means that for a portion of the window of time between
CPUHP_AP_ONLINE_IDLE & CPUHP_ONLINE the newly onlined CPU's struct
cpu_stopper has its enabled field set to false. If a user thread is
executed on the CPU during this window and it invokes sched_setaffinity
with a CPU mask that does not include the CPU it's running on, then when
__set_cpus_allowed_ptr() calls stop_one_cpu() intending to invoke
migration_cpu_stop() and perform the actual migration away from the CPU
it will simply return -ENOENT rather than calling migration_cpu_stop().
We then return from the sched_setaffinity syscall back to the user task
that is now running on a CPU which it just asked not to run on, and
which is not present in its cpus_allowed mask.
This patch resolves the problem by having select_task_rq() enforce that
user tasks run on CPUs that are active - the same requirement that
select_fallback_rq() already enforces. This should ensure that newly
onlined CPUs reach the CPUHP_AP_ACTIVE state before being able to
schedule user tasks, and also implies that bringup_wait_for_ap() will
have called stop_machine_unpark() which resolves the sched_setaffinity
issue above.
I haven't yet investigated them, but it may be of interest to review
whether any of the actions performed by hotplug states between
CPUHP_AP_ONLINE_IDLE & CPUHP_AP_ACTIVE could have similar unintended
effects on user tasks that might schedule before they are reached, which
might widen the scope of the problem from just affecting the behaviour
of sched_setaffinity.
Signed-off-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180526154648.11635-2-paul.burton@mips.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
As already enforced by the WARN() in __set_cpus_allowed_ptr(), the rules
for running on an online && !active CPU are stricter than just being a
kthread, you need to be a per-cpu kthread.
If you're not strictly per-CPU, you have better CPUs to run on and
don't need the partially booted one to get your work done.
The exception is to allow smpboot threads to bootstrap the CPU itself
and get kernel 'services' initialized before we allow userspace on it.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 955dbdf4ce ("sched: Allow migrating kthreads into online but inactive CPUs")
Link: http://lkml.kernel.org/r/20170725165821.cejhb7v2s3kecems@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When a task is enqueued the estimated utilization of a CPU is updated
to better support the selection of the required frequency.
However, schedutil is (implicitly) updated by update_load_avg() which
always happens before util_est_{en,de}queue(), thus potentially
introducing a latency between estimated utilization updates and
frequency selections.
Let's update util_est at the beginning of enqueue_task_fair(),
which will ensure that all schedutil updates will see the most
updated estimated utilization value for a CPU.
Reported-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Steve Muckle <smuckle@google.com>
Fixes: 7f65ea42eb ("sched/fair: Add util_est on top of PELT")
Link: http://lkml.kernel.org/r/20180524141023.13765-3-patrick.bellasi@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since the refactoring introduced by:
commit 8f111bc357 ("cpufreq/schedutil: Rewrite CPUFREQ_RT support")
we aggregate FAIR utilization only if this class has runnable tasks.
This was mainly due to avoid the risk to stay on an high frequency just
because of the blocked utilization of a CPU not being properly decayed
while the CPU was idle.
However, since:
commit 31e77c93e4 ("sched/fair: Update blocked load when newly idle")
the FAIR blocked utilization is properly decayed also for IDLE CPUs.
This allows us to use the FAIR blocked utilization as a safe mechanism
to gracefully reduce the frequency only if no FAIR tasks show up on a
CPU for a reasonable period of time.
Moreover, we also reduce the frequency drops of CPUs running periodic
tasks which, depending on the task periodicity and the time required
for a frequency switch, was increasing the chances to introduce some
undesirable performance variations.
Reported-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Steve Muckle <smuckle@google.com>
Link: http://lkml.kernel.org/r/20180524141023.13765-2-patrick.bellasi@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Commit 152db033d7 (schedutil: Allow cpufreq requests to be made
even when kthread kicked) made changes to prevent utilization updates
from being discarded during processing a previous request, but it
left a small window in which that still can happen in the one-CPU
policy case. Namely, updates coming in after setting work_in_progress
in sugov_update_commit() and clearing it in sugov_work() will still
be dropped due to the work_in_progress check in sugov_update_single().
To close that window, rearrange the code so as to acquire the update
lock around the deferred update branch in sugov_update_single()
and drop the work_in_progress check from it.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Juri Lelli <juri.lelli@redhat.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Currently there is a chance of a schedutil cpufreq update request to be
dropped if there is a pending update request. This pending request can
be delayed if there is a scheduling delay of the irq_work and the wake
up of the schedutil governor kthread.
A very bad scenario is when a schedutil request was already just made,
such as to reduce the CPU frequency, then a newer request to increase
CPU frequency (even sched deadline urgent frequency increase requests)
can be dropped, even though the rate limits suggest that its Ok to
process a request. This is because of the way the work_in_progress flag
is used.
This patch improves the situation by allowing new requests to happen
even though the old one is still being processed. Note that in this
approach, if an irq_work was already issued, we just update next_freq
and don't bother to queue another request so there's no extra work being
done to make this happen.
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This routine checks if the CPU running this code belongs to the policy
of the target CPU or if not, can it do remote DVFS for it remotely. But
the current name of it implies as if it is only about doing remote
updates.
Rename it to make it more relevant.
Suggested-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The iowait boosting code has been recently updated to add a progressive
boosting behavior which allows to be less aggressive in boosting tasks
doing only sporadic IO operations, thus being more energy efficient for
example on mobile platforms.
The current code is now however a bit convoluted. Some functionalities
(e.g. iowait boost reset) are replicated in different paths and their
documentation is slightly misaligned.
Let's cleanup the code by consolidating all the IO wait boosting related
functionality within within few dedicated functions and better define
their role:
- sugov_iowait_boost: set/increase the IO wait boost of a CPU
- sugov_iowait_apply: apply/reduce the IO wait boost of a CPU
Both these two function are used at every sugov update and they make
use of a unified IO wait boost reset policy provided by:
- sugov_iowait_reset: reset/disable the IO wait boost of a CPU
if a CPU is not updated for more then one tick
This makes possible a cleaner and more self-contained design for the IO
wait boosting code since the rest of the sugov update routines, both for
single and shared frequency domains, follow the same template:
/* Configure IO boost, if required */
sugov_iowait_boost()
/* Return here if freq change is in progress or throttled */
/* Collect and aggregate utilization information */
sugov_get_util()
sugov_aggregate_util()
/*
* Add IO boost, if currently enabled, on top of the aggregated
* utilization value
*/
sugov_iowait_apply()
As a extra bonus, let's also add the documentation for the new
functions and better align the in-code documentation.
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
A more energy efficient update of the IO wait boosting mechanism has
been introduced in:
commit a5a0809bc5 ("cpufreq: schedutil: Make iowait boost more energy efficient")
where the boost value is expected to be:
- doubled at each successive wakeup from IO
staring from the minimum frequency supported by a CPU
- reset when a CPU is not updated for more then one tick
by either disabling the IO wait boost or resetting its value to the
minimum frequency if this new update requires an IO boost.
This approach is supposed to "ignore" boosting for sporadic wakeups from
IO, while still getting the frequency boosted to the maximum to benefit
long sequence of wakeup from IO operations.
However, these assumptions are not always satisfied.
For example, when an IO boosted CPU enters idle for more the one tick
and then wakes up after an IO wait, since in sugov_set_iowait_boost() we
first check the IOWAIT flag, we keep doubling the iowait boost instead
of restarting from the minimum frequency value.
This misbehavior could happen mainly on non-shared frequency domains,
thus defeating the energy efficiency optimization, but it can also
happen on shared frequency domain systems.
Let fix this issue in sugov_set_iowait_boost() by:
- first check the IO wait boost reset conditions
to eventually reset the boost value
- then applying the correct IO boost value
if required by the caller
Fixes: a5a0809bc5 (cpufreq: schedutil: Make iowait boost more energy efficient)
Reported-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since the grub_reclaim() function can be made static, make it so.
Silences the following GCC warning (W=1):
kernel/sched/deadline.c:1120:5: warning: no previous prototype for ‘grub_reclaim’ [-Wmissing-prototypes]
Signed-off-by: Mathieu Malaterre <malat@debian.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180516200902.959-1-malat@debian.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In the following commit:
6b55c9654f ("sched/debug: Move print_cfs_rq() declaration to kernel/sched/sched.h")
the print_cfs_rq() prototype was added to <kernel/sched/sched.h>,
right next to the prototypes for print_cfs_stats(), print_rt_stats()
and print_dl_stats().
Finish this previous commit and also move related prototypes for
print_rt_rq() and print_dl_rq().
Remove existing extern declarations now that they not needed anymore.
Silences the following GCC warning, triggered by W=1:
kernel/sched/debug.c:573:6: warning: no previous prototype for ‘print_rt_rq’ [-Wmissing-prototypes]
kernel/sched/debug.c:603:6: warning: no previous prototype for ‘print_dl_rq’ [-Wmissing-prototypes]
Signed-off-by: Mathieu Malaterre <malat@debian.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180516195348.30426-1-malat@debian.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
- Updates to the handling of expedited grace periods, perhaps most
notably parallelizing their initialization. Other changes
include fixes from Boqun Feng.
- Miscellaneous fixes. These include an nvme fix from Nitzan Carmi
that I am carrying because it depends on a new SRCU function
cleanup_srcu_struct_quiesced(). This branch also includes fixes
from Byungchul Park and Yury Norov.
- Updates to reduce lock contention in the rcu_node combining tree.
These are in preparation for the consolidation of RCU-bh,
RCU-preempt, and RCU-sched into a single flavor, which was
requested by Linus Torvalds in response to a security flaw
whose root cause included confusion between the multiple flavors
of RCU.
- Torture-test updates that save their users some time and effort.
Conflicts:
drivers/nvme/host/core.c
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Variants of proc_create{,_data} that directly take a struct seq_operations
argument and drastically reduces the boilerplate code in the callers.
All trivial callers converted over.
Signed-off-by: Christoph Hellwig <hch@lst.de>
The cond_resched_softirq() macro is not used anywhere in mainline, so
this commit simplifies the kernel by eliminating it.
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Tested-by: Nicholas Piggin <npiggin@gmail.com>
The schedutil driver sets sg_policy->next_freq to UINT_MAX on certain
occasions to discard the cached value of next freq:
- In sugov_start(), when the schedutil governor is started for a group
of CPUs.
- And whenever we need to force a freq update before rate-limit
duration, which happens when:
- there is an update in cpufreq policy limits.
- Or when the utilization of DL scheduling class increases.
In return, get_next_freq() doesn't return a cached next_freq value but
recalculates the next frequency instead.
But having special meaning for a particular value of frequency makes the
code less readable and error prone. We recently fixed a bug where the
UINT_MAX value was considered as valid frequency in
sugov_update_single().
All we need is a flag which can be used to discard the value of
sg_policy->next_freq and we already have need_freq_update for that. Lets
reuse it instead of setting next_freq to UINT_MAX.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This reverts commit e2cabe48c2.
Lifting the restriction that the sugov kthread is bound to the
policy->related_cpus for a system with a slow switching cpufreq driver,
which is able to perform DVFS from any cpu (e.g. cpufreq-dt), is not
only not beneficial it also harms Enery-Aware Scheduling (EAS) on
systems with asymmetric cpu capacities (e.g. Arm big.LITTLE).
The sugov kthread which does the update for the little cpus could
potentially run on a big cpu. It could prevent that the big cluster goes
into deeper idle states although all the tasks are running on the little
cluster.
Example: hikey960 w/ 4.16.0-rc6-+
Arm big.LITTLE with per-cluster DVFS
root@h960:~# cat /proc/cpuinfo | grep "^CPU part"
CPU part : 0xd03 (Cortex-A53, little cpu)
CPU part : 0xd03
CPU part : 0xd03
CPU part : 0xd03
CPU part : 0xd09 (Cortex-A73, big cpu)
CPU part : 0xd09
CPU part : 0xd09
CPU part : 0xd09
root@h960:/sys/devices/system/cpu/cpufreq# ls
policy0 policy4 schedutil
root@h960:/sys/devices/system/cpu/cpufreq# cat policy*/related_cpus
0 1 2 3
4 5 6 7
(1) w/o the revert:
root@h960:~# ps -eo pid,class,rtprio,pri,psr,comm | awk 'NR == 1 ||
/sugov/'
PID CLS RTPRIO PRI PSR COMMAND
1489 #6 0 140 1 sugov:0
1490 #6 0 140 0 sugov:4
The sugov kthread sugov:4 responsible for policy4 runs on cpu0. (In this
case both sugov kthreads run on little cpus).
cross policy (cluster) remote callback example:
...
migration/1-14 [001] enqueue_task_fair: this_cpu=1 cpu_of(rq)=5
migration/1-14 [001] sugov_update_shared: this_cpu=1 sg_cpu->cpu=5
sg_cpu->sg_policy->policy->related_cpus=4-7
sugov:4-1490 [000] sugov_work: this_cpu=0
sg_cpu->sg_policy->policy->related_cpus=4-7
...
The remote callback (this_cpu=1, target_cpu=5) is executed on cpu=0.
(2) w/ the revert:
root@h960:~# ps -eo pid,class,rtprio,pri,psr,comm | awk 'NR == 1 ||
/sugov/'
PID CLS RTPRIO PRI PSR COMMAND
1491 #6 0 140 2 sugov:0
1492 #6 0 140 4 sugov:4
The sugov kthread sugov:4 responsible for policy4 runs on cpu4.
cross policy (cluster) remote callback example:
...
migration/1-14 [001] enqueue_task_fair: this_cpu=1 cpu_of(rq)=7
migration/1-14 [001] sugov_update_shared: this_cpu=1 sg_cpu->cpu=7
sg_cpu->sg_policy->policy->related_cpus=4-7
sugov:4-1492 [004] sugov_work: this_cpu=4
sg_cpu->sg_policy->policy->related_cpus=4-7
...
The remote callback (this_cpu=1, target_cpu=7) is executed on cpu=4.
Now the sugov kthread executes again on the policy (cluster) for which
the Operating Performance Point (OPP) should be changed.
It avoids the problem that an otherwise idle policy (cluster) is running
schedutil (the sugov kthread) for another one.
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In the following commit:
247f2f6f3c ("sched/core: Don't schedule threads on pre-empted vCPUs")
... we distinguish between idle_cpu() when the vCPU is not running for
scheduling threads.
However, the idle_cpu() function is used in other places for
actually checking whether the state of the CPU is idle or not.
Hence split the use of that function based on the desired return value,
by introducing the available_idle_cpu() function.
This fixes a (slight) regression in that initial vCPU commit, because
some code paths (like the load-balancer) don't care and shouldn't care
if the vCPU is preempted or not, they just want to know if there's any
tasks on the CPU.
Signed-off-by: Rohit Jain <rohit.k.jain@oracle.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dhaval.giani@oracle.com
Cc: linux-kernel@vger.kernel.org
Cc: matt@codeblueprint.co.uk
Cc: steven.sistare@oracle.com
Cc: subhra.mazumdar@oracle.com
Link: http://lkml.kernel.org/r/1525883988-10356-1-git-send-email-rohit.k.jain@oracle.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Threads share an address space and each can change the protections of the
same address space to trap NUMA faults. This is redundant and potentially
counter-productive as any thread doing the update will suffice. Potentially
only one thread is required but that thread may be idle or it may not have
any locality concerns and pick an unsuitable scan rate.
This patch uses independent scan period but they are staggered based on
the number of address space users when the thread is created. The intent
is that threads will avoid scanning at the same time and have a chance
to adapt their scan rate later if necessary. This reduces the total scan
activity early in the lifetime of the threads.
The different in headline performance across a range of machines and
workloads is marginal but the system CPU usage is reduced as well as overall
scan activity. The following is the time reported by NAS Parallel Benchmark
using unbound openmp threads and a D size class:
4.17.0-rc1 4.17.0-rc1
vanilla stagger-v1r1
Time bt.D 442.77 ( 0.00%) 419.70 ( 5.21%)
Time cg.D 171.90 ( 0.00%) 180.85 ( -5.21%)
Time ep.D 33.10 ( 0.00%) 32.90 ( 0.60%)
Time is.D 9.59 ( 0.00%) 9.42 ( 1.77%)
Time lu.D 306.75 ( 0.00%) 304.65 ( 0.68%)
Time mg.D 54.56 ( 0.00%) 52.38 ( 4.00%)
Time sp.D 1020.03 ( 0.00%) 903.77 ( 11.40%)
Time ua.D 400.58 ( 0.00%) 386.49 ( 3.52%)
Note it's not a universal win but we have no prior knowledge of which
thread matters but the number of threads created often exceeds the size
of the node when the threads are not bound. However, there is a reducation
of overall system CPU usage:
4.17.0-rc1 4.17.0-rc1
vanilla stagger-v1r1
sys-time-bt.D 48.78 ( 0.00%) 48.22 ( 1.15%)
sys-time-cg.D 25.31 ( 0.00%) 26.63 ( -5.22%)
sys-time-ep.D 1.65 ( 0.00%) 0.62 ( 62.42%)
sys-time-is.D 40.05 ( 0.00%) 24.45 ( 38.95%)
sys-time-lu.D 37.55 ( 0.00%) 29.02 ( 22.72%)
sys-time-mg.D 47.52 ( 0.00%) 34.92 ( 26.52%)
sys-time-sp.D 119.01 ( 0.00%) 109.05 ( 8.37%)
sys-time-ua.D 51.52 ( 0.00%) 45.13 ( 12.40%)
NUMA scan activity is also reduced:
NUMA alloc local 1042828 1342670
NUMA base PTE updates 140481138 93577468
NUMA huge PMD updates 272171 180766
NUMA page range updates 279832690 186129660
NUMA hint faults 1395972 1193897
NUMA hint local faults 877925 855053
NUMA hint local percent 62 71
NUMA pages migrated 12057909 9158023
Similar observations are made for other thread-intensive workloads. System
CPU usage is lower even though the headline gains in performance tend to be
small. For example, specjbb 2005 shows almost no difference in performance
but scan activity is reduced by a third on a 4-socket box. I didn't find
a workload (thread intensive or otherwise) that suffered badly.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/20180504154109.mvrha2qo5wdl65vr@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull x86/pti updates from Thomas Gleixner:
"A mixed bag of fixes and updates for the ghosts which are hunting us.
The scheduler fixes have been pulled into that branch to avoid
conflicts.
- A set of fixes to address a khread_parkme() race which caused lost
wakeups and loss of state.
- A deadlock fix for stop_machine() solved by moving the wakeups
outside of the stopper_lock held region.
- A set of Spectre V1 array access restrictions. The possible
problematic spots were discuvered by Dan Carpenters new checks in
smatch.
- Removal of an unused file which was forgotten when the rest of that
functionality was removed"
* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/vdso: Remove unused file
perf/x86/cstate: Fix possible Spectre-v1 indexing for pkg_msr
perf/x86/msr: Fix possible Spectre-v1 indexing in the MSR driver
perf/x86: Fix possible Spectre-v1 indexing for x86_pmu::event_map()
perf/x86: Fix possible Spectre-v1 indexing for hw_perf_event cache_*
perf/core: Fix possible Spectre-v1 indexing for ->aux_pages[]
sched/autogroup: Fix possible Spectre-v1 indexing for sched_prio_to_weight[]
sched/core: Fix possible Spectre-v1 indexing for sched_prio_to_weight[]
sched/core: Introduce set_special_state()
kthread, sched/wait: Fix kthread_parkme() completion issue
kthread, sched/wait: Fix kthread_parkme() wait-loop
sched/fair: Fix the update of blocked load when newly idle
stop_machine, sched: Fix migrate_swap() vs. active_balance() deadlock
Pull scheduler fix from Thomas Gleixner:
"Revert the new NUMA aware placement approach which turned out to
create more problems than it solved"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
Revert "sched/numa: Delay retrying placement for automatic NUMA balance after wake_affine()"
This reverts commit 7347fc87df.
Srikar Dronamra pointed out that while the commit in question did show
a performance improvement on ppc64, it did so at the cost of disabling
active CPU migration by automatic NUMA balancing which was not the intent.
The issue was that a serious flaw in the logic failed to ever active balance
if SD_WAKE_AFFINE was disabled on scheduler domains. Even when it's enabled,
the logic is still bizarre and against the original intent.
Investigation showed that fixing the patch in either the way he suggested,
using the correct comparison for jiffies values or introducing a new
numa_migrate_deferred variable in task_struct all perform similarly to a
revert with a mix of gains and losses depending on the workload, machine
and socket count.
The original intent of the commit was to handle a problem whereby
wake_affine, idle balancing and automatic NUMA balancing disagree on the
appropriate placement for a task. This was particularly true for cases where
a single task was a massive waker of tasks but where wake_wide logic did
not apply. This was particularly noticeable when a futex (a barrier) woke
all worker threads and tried pulling the wakees to the waker nodes. In that
specific case, it could be handled by tuning MPI or openMP appropriately,
but the behavior is not illogical and was worth attempting to fix. However,
the approach was wrong. Given that we're at rc4 and a fix is not obvious,
it's better to play safe, revert this commit and retry later.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: efault@gmx.de
Cc: ggherdovich@suse.cz
Cc: hpa@zytor.com
Cc: matt@codeblueprint.co.uk
Cc: mpe@ellerman.id.au
Link: http://lkml.kernel.org/r/20180509163115.6fnnyeg4vdm2ct4v@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
If the next_freq field of struct sugov_policy is set to UINT_MAX,
it shouldn't be used for updating the CPU frequency (this is a
special "invalid" value), but after commit b7eaf1aab9 (cpufreq:
schedutil: Avoid reducing frequency of busy CPUs prematurely) it
may be passed as the new frequency to sugov_update_commit() in
sugov_update_single().
Fix that by adding an extra check for the special UINT_MAX value
of next_freq to sugov_update_single().
Fixes: b7eaf1aab9 (cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely)
Reported-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: 4.12+ <stable@vger.kernel.org> # 4.12+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
After commit 794a56ebd9 (sched/cpufreq: Change the worker kthread to
SCHED_DEADLINE) schedutil kthreads are "ignored" for a clock frequency
selection point of view, so the potential corner case for RT tasks is not
possible at all now.
Remove the stale comment mentioning it.
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> kernel/sched/core.c:6921 cpu_weight_nice_write_s64() warn: potential spectre issue 'sched_prio_to_weight'
Userspace controls @nice, so sanitize the value before using it to
index an array.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Call sync_entity_load_avg() directly from find_idlest_cpu() instead of
select_task_rq_fair(), as that's where we need to use task's utilization
value. And call sync_entity_load_avg() only after making sure sched
domain spans over one of the allowed CPUs for the task.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Link: http://lkml.kernel.org/r/cd019d1753824c81130eae7b43e2bbcec47cc1ad.1524738578.git.viresh.kumar@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Rearrange select_task_rq_fair() a bit to avoid executing some
conditional statements in few specific code-paths. That gets rid of the
goto as well.
This shouldn't result in any functional changes.
Tested-by: Rohit Jain <rohit.k.jain@oracle.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Link: http://lkml.kernel.org/r/20831b8d237bf3a20e4e328286f678b425ff04c9.1524738578.git.viresh.kumar@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Gaurav reported a perceived problem with TASK_PARKED, which turned out
to be a broken wait-loop pattern in __kthread_parkme(), but the
reported issue can (and does) in fact happen for states that do not do
condition based sleeps.
When the 'current->state = TASK_RUNNING' store of a previous
(concurrent) try_to_wake_up() collides with the setting of a 'special'
sleep state, we can loose the sleep state.
Normal condition based wait-loops are immune to this problem, but for
sleep states that are not condition based are subject to this problem.
There already is a fix for TASK_DEAD. Abstract that and also apply it
to TASK_STOPPED and TASK_TRACED, both of which are also without
condition based wait-loop.
Reported-by: Gaurav Kohli <gkohli@codeaurora.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Even with the wait-loop fixed, there is a further issue with
kthread_parkme(). Upon hotplug, when we do takedown_cpu(),
smpboot_park_threads() can return before all those threads are in fact
blocked, due to the placement of the complete() in __kthread_parkme().
When that happens, sched_cpu_dying() -> migrate_tasks() can end up
migrating such a still runnable task onto another CPU.
Normally the task will have hit schedule() and gone to sleep by the
time we do kthread_unpark(), which will then do __kthread_bind() to
re-bind the task to the correct CPU.
However, when we loose the initial TASK_PARKED store to the concurrent
wakeup issue described previously, do the complete(), get migrated, it
is possible to either:
- observe kthread_unpark()'s clearing of SHOULD_PARK and terminate
the park and set TASK_RUNNING, or
- __kthread_bind()'s wait_task_inactive() to observe the competing
TASK_RUNNING store.
Either way the WARN() in __kthread_bind() will trigger and fail to
correctly set the CPU affinity.
Fix this by only issuing the complete() when the kthread has scheduled
out. This does away with all the icky 'still running' nonsense.
The alternative is to promote TASK_PARKED to a special state, this
guarantees wait_task_inactive() cannot observe a 'stale' TASK_RUNNING
and we'll end up doing the right thing, but this preserves the whole
icky business of potentially migating the still runnable thing.
Reported-by: Gaurav Kohli <gkohli@codeaurora.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
With commit:
31e77c93e4 ("sched/fair: Update blocked load when newly idle")
... we release the rq->lock when updating blocked load of idle CPUs.
This opens a time window during which another CPU can add a task to this
CPU's cfs_rq.
The check for newly added task of idle_balance() is not in the common path.
Move the out label to include this check.
Reported-by: Heiner Kallweit <hkallweit1@gmail.com>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 31e77c93e4 ("sched/fair: Update blocked load when newly idle")
Link: http://lkml.kernel.org/r/20180426103133.GA6953@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull scheduler fixes from Thomas Gleixner:
"A few scheduler fixes:
- Prevent a bogus warning vs. runqueue clock update flags in
do_sched_rt_period_timer()
- Simplify the helper functions which handle requests for skipping
the runqueue clock updat.
- Do not unlock the tunables mutex in the error path of the cpu
frequency scheduler utils. Its not held.
- Enforce proper alignement for 'struct util_est' in sched_avg to
prevent a misalignment fault on IA64"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/core: Force proper alignment of 'struct util_est'
sched/core: Simplify helpers for rq clock update skip requests
sched/rt: Fix rq->clock_update_flags < RQCF_ACT_SKIP warning
sched/cpufreq/schedutil: Fix error path mutex unlock
- Rework the idle loop in order to prevent CPUs from spending too
much time in shallow idle states by making it stop the scheduler
tick before putting the CPU into an idle state only if the idle
duration predicted by the idle governor is long enough. That
required the code to be reordered to invoke the idle governor
before stopping the tick, among other things (Rafael Wysocki,
Frederic Weisbecker, Arnd Bergmann).
- Add the missing description of the residency sysfs attribute to
the cpuidle documentation (Prashanth Prakash).
- Finalize the cpufreq cleanup moving frequency table validation
from drivers to the core (Viresh Kumar).
- Fix a clock leak regression in the armada-37xx cpufreq driver
(Gregory Clement).
- Fix the initialization of the CPU performance data structures
for shared policies in the CPPC cpufreq driver (Shunyong Yang).
- Clean up the ti-cpufreq, intel_pstate and CPPC cpufreq drivers
a bit (Viresh Kumar, Rafael Wysocki).
- Mark the expected switch fall-throughs in the PM QoS core (Gustavo
Silva).
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=Pnk4
-----END PGP SIGNATURE-----
Merge tag 'pm-4.17-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull more power management updates from Rafael Wysocki:
"These include one big-ticket item which is the rework of the idle loop
in order to prevent CPUs from spending too much time in shallow idle
states. It reduces idle power on some systems by 10% or more and may
improve performance of workloads in which the idle loop overhead
matters. This has been in the works for several weeks and it has been
tested and reviewed quite thoroughly.
Also included are changes that finalize the cpufreq cleanup moving
frequency table validation from drivers to the core, a few fixes and
cleanups of cpufreq drivers, a cpuidle documentation update and a PM
QoS core update to mark the expected switch fall-throughs in it.
Specifics:
- Rework the idle loop in order to prevent CPUs from spending too
much time in shallow idle states by making it stop the scheduler
tick before putting the CPU into an idle state only if the idle
duration predicted by the idle governor is long enough.
That required the code to be reordered to invoke the idle governor
before stopping the tick, among other things (Rafael Wysocki,
Frederic Weisbecker, Arnd Bergmann).
- Add the missing description of the residency sysfs attribute to the
cpuidle documentation (Prashanth Prakash).
- Finalize the cpufreq cleanup moving frequency table validation from
drivers to the core (Viresh Kumar).
- Fix a clock leak regression in the armada-37xx cpufreq driver
(Gregory Clement).
- Fix the initialization of the CPU performance data structures for
shared policies in the CPPC cpufreq driver (Shunyong Yang).
- Clean up the ti-cpufreq, intel_pstate and CPPC cpufreq drivers a
bit (Viresh Kumar, Rafael Wysocki).
- Mark the expected switch fall-throughs in the PM QoS core (Gustavo
Silva)"
* tag 'pm-4.17-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (23 commits)
tick-sched: avoid a maybe-uninitialized warning
cpufreq: Drop cpufreq_table_validate_and_show()
cpufreq: SCMI: Don't validate the frequency table twice
cpufreq: CPPC: Initialize shared perf capabilities of CPUs
cpufreq: armada-37xx: Fix clock leak
cpufreq: CPPC: Don't set transition_latency
cpufreq: ti-cpufreq: Use builtin_platform_driver()
cpufreq: intel_pstate: Do not include debugfs.h
PM / QoS: mark expected switch fall-throughs
cpuidle: Add definition of residency to sysfs documentation
time: hrtimer: Use timerqueue_iterate_next() to get to the next timer
nohz: Avoid duplication of code related to got_idle_tick
nohz: Gather tick_sched booleans under a common flag field
cpuidle: menu: Avoid selecting shallow states with stopped tick
cpuidle: menu: Refine idle state selection for running tick
sched: idle: Select idle state before stopping the tick
time: hrtimer: Introduce hrtimer_next_event_without()
time: tick-sched: Split tick_nohz_stop_sched_tick()
cpuidle: Return nohz hint from cpuidle_select()
jiffies: Introduce USER_TICK_USEC and redefine TICK_USEC
...
In order to address the issue with short idle duration predictions
by the idle governor after the scheduler tick has been stopped,
reorder the code in cpuidle_idle_call() so that the governor idle
state selection runs before tick_nohz_idle_go_idle() and use the
"nohz" hint returned by cpuidle_select() to decide whether or not
to stop the tick.
This isn't straightforward, because menu_select() invokes
tick_nohz_get_sleep_length() to get the time to the next timer
event and the number returned by the latter comes from
__tick_nohz_idle_stop_tick(). Fortunately, however, it is possible
to compute that number without actually stopping the tick and with
the help of the existing code.
Namely, tick_nohz_get_sleep_length() can be made call
tick_nohz_next_event(), introduced earlier, to get the time to the
next non-highres timer event. If that happens, tick_nohz_next_event()
need not be called by __tick_nohz_idle_stop_tick() again.
If it turns out that the scheduler tick cannot be stopped going
forward or the next timer event is too close for the tick to be
stopped, tick_nohz_get_sleep_length() can simply return the time to
the next event currently programmed into the corresponding clock
event device.
In addition to knowing the return value of tick_nohz_next_event(),
however, tick_nohz_get_sleep_length() needs to know the time to the
next highres timer event, but with the scheduler tick timer excluded,
which can be computed with the help of hrtimer_get_next_event().
That minimum of that number and the tick_nohz_next_event() return
value is the total time to the next timer event with the assumption
that the tick will be stopped. It can be returned to the idle
governor which can use it for predicting idle duration (under the
assumption that the tick will be stopped) and deciding whether or
not it makes sense to stop the tick before putting the CPU into the
selected idle state.
With the above, the sleep_length field in struct tick_sched is not
necessary any more, so drop it.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=199227
Reported-by: Doug Smythies <dsmythies@telus.net>
Reported-by: Thomas Ilsche <thomas.ilsche@tu-dresden.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Add a new pointer argument to cpuidle_select() and to the ->select
cpuidle governor callback to allow a boolean value indicating
whether or not the tick should be stopped before entering the
selected state to be returned from there.
Make the ladder governor ignore that pointer (to preserve its
current behavior) and make the menu governor return 'false" through
it if:
(1) the idle exit latency is constrained at 0, or
(2) the selected state is a polling one, or
(3) the expected idle period duration is within the tick period
range.
In addition to that, the correction factor computations in the menu
governor need to take the possibility that the tick may not be
stopped into account to avoid artificially small correction factor
values. To that end, add a mechanism to record tick wakeups, as
suggested by Peter Zijlstra, and use it to modify the menu_update()
behavior when tick wakeup occurs. Namely, if the CPU is woken up by
the tick and the return value of tick_nohz_get_sleep_length() is not
within the tick boundary, the predicted idle duration is likely too
short, so make menu_update() try to compensate for that by updating
the governor statistics as though the CPU was idle for a long time.
Since the value returned through the new argument pointer of
cpuidle_select() is not used by its caller yet, this change by
itself is not expected to alter the functionality of the code.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
KASAN splats indicate that in some cases we free a live mm, then
continue to access it, with potentially disastrous results. This is
likely due to a mismatched mmdrop() somewhere in the kernel, but so far
the culprit remains elusive.
Let's have __mmdrop() verify that the mm isn't live for the current
task, similar to the existing check for init_mm. This way, we can catch
this class of issue earlier, and without requiring KASAN.
Currently, idle_task_exit() leaves active_mm stale after it switches to
init_mm. This isn't harmful, but will trigger the new assertions, so we
must adjust idle_task_exit() to update active_mm.
Link: http://lkml.kernel.org/r/20180312140103.19235-1-mark.rutland@arm.com
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Make cpuidle_idle_call() decide whether or not to stop the tick.
First, the cpuidle_enter_s2idle() path deals with the tick (and with
the entire timekeeping for that matter) by itself and it doesn't need
the tick to be stopped beforehand.
Second, to address the issue with short idle duration predictions
by the idle governor after the tick has been stopped, it will be
necessary to change the ordering of cpuidle_select() with respect
to tick_nohz_idle_stop_tick(). To prepare for that, put a
tick_nohz_idle_stop_tick() call in the same branch in which
cpuidle_select() is called.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Push the decision whether or not to stop the tick somewhat deeper
into the idle loop.
Stopping the tick upfront leads to unpleasant outcomes in case the
idle governor doesn't agree with the nohz code on the duration of the
upcoming idle period. Specifically, if the tick has been stopped and
the idle governor predicts short idle, the situation is bad regardless
of whether or not the prediction is accurate. If it is accurate, the
tick has been stopped unnecessarily which means excessive overhead.
If it is not accurate, the CPU is likely to spend too much time in
the (shallow, because short idle has been predicted) idle state
selected by the governor [1].
As the first step towards addressing this problem, change the code
to make the tick stopping decision inside of the loop in do_idle().
In particular, do not stop the tick in the cpu_idle_poll() code path.
Also don't do that in tick_nohz_irq_exit() which doesn't really have
enough information on whether or not to stop the tick.
Link: https://marc.info/?l=linux-pm&m=150116085925208&w=2 # [1]
Link: https://tu-dresden.de/zih/forschung/ressourcen/dateien/projekte/haec/powernightmares.pdf
Suggested-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Prepare the scheduler tick code for reworking the idle loop to
avoid stopping the tick in some cases.
The idea is to split the nohz idle entry call to decouple the idle
time stats accounting and preparatory work from the actual tick stop
code, in order to later be able to delay the tick stop once we reach
more power-knowledgeable callers.
Move away the tick_nohz_start_idle() invocation from
__tick_nohz_idle_enter(), rename the latter to
__tick_nohz_idle_stop_tick() and define tick_nohz_idle_stop_tick()
as a wrapper around it for calling it from the outside.
Make tick_nohz_idle_enter() only call tick_nohz_start_idle() instead
of calling the entire __tick_nohz_idle_enter(), add another wrapper
disabling and enabling interrupts around tick_nohz_idle_stop_tick()
and make the current callers of tick_nohz_idle_enter() call it too
to retain their current functionality.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
By renaming the functions we can get rid of the skip parameter
and have better code redability. It makes zero sense to have
things such as:
rq_clock_skip_update(rq, false)
When the skip request is in fact not going to happen. Ever. Rename
things such that we end up with:
rq_clock_skip_update(rq)
rq_clock_cancel_skipupdate(rq)
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Cc: matt@codeblueprint.co.uk
Cc: rostedt@goodmis.org
Link: http://lkml.kernel.org/r/20180404161539.nhadkff2aats74jh@linux-n805
Signed-off-by: Ingo Molnar <mingo@kernel.org>
While running rt-tests' pi_stress program I got the following splat:
rq->clock_update_flags < RQCF_ACT_SKIP
WARNING: CPU: 27 PID: 0 at kernel/sched/sched.h:960 assert_clock_updated.isra.38.part.39+0x13/0x20
[...]
<IRQ>
enqueue_top_rt_rq+0xf4/0x150
? cpufreq_dbs_governor_start+0x170/0x170
sched_rt_rq_enqueue+0x65/0x80
sched_rt_period_timer+0x156/0x360
? sched_rt_rq_enqueue+0x80/0x80
__hrtimer_run_queues+0xfa/0x260
hrtimer_interrupt+0xcb/0x220
smp_apic_timer_interrupt+0x62/0x120
apic_timer_interrupt+0xf/0x20
</IRQ>
[...]
do_idle+0x183/0x1e0
cpu_startup_entry+0x5f/0x70
start_secondary+0x192/0x1d0
secondary_startup_64+0xa5/0xb0
We can get rid of it be the "traditional" means of adding an
update_rq_clock() call after acquiring the rq->lock in
do_sched_rt_period_timer().
The case for the RT task throttling (which this workload also hits)
can be ignored in that the skip_update call is actually bogus and
quite the contrary (the request bits are removed/reverted).
By setting RQCF_UPDATED we really don't care if the skip is happening
or not and will therefore make the assert_clock_updated() check happy.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Reviewed-by: Matt Fleming <matt@codeblueprint.co.uk>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dave@stgolabs.net
Cc: linux-kernel@vger.kernel.org
Cc: rostedt@goodmis.org
Link: http://lkml.kernel.org/r/20180402164954.16255-1-dave@stgolabs.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull removal of in-kernel calls to syscalls from Dominik Brodowski:
"System calls are interaction points between userspace and the kernel.
Therefore, system call functions such as sys_xyzzy() or
compat_sys_xyzzy() should only be called from userspace via the
syscall table, but not from elsewhere in the kernel.
At least on 64-bit x86, it will likely be a hard requirement from
v4.17 onwards to not call system call functions in the kernel: It is
better to use use a different calling convention for system calls
there, where struct pt_regs is decoded on-the-fly in a syscall wrapper
which then hands processing over to the actual syscall function. This
means that only those parameters which are actually needed for a
specific syscall are passed on during syscall entry, instead of
filling in six CPU registers with random user space content all the
time (which may cause serious trouble down the call chain). Those
x86-specific patches will be pushed through the x86 tree in the near
future.
Moreover, rules on how data may be accessed may differ between kernel
data and user data. This is another reason why calling sys_xyzzy() is
generally a bad idea, and -- at most -- acceptable in arch-specific
code.
This patchset removes all in-kernel calls to syscall functions in the
kernel with the exception of arch/. On top of this, it cleans up the
three places where many syscalls are referenced or prototyped, namely
kernel/sys_ni.c, include/linux/syscalls.h and include/linux/compat.h"
* 'syscalls-next' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo/linux: (109 commits)
bpf: whitelist all syscalls for error injection
kernel/sys_ni: remove {sys_,sys_compat} from cond_syscall definitions
kernel/sys_ni: sort cond_syscall() entries
syscalls/x86: auto-create compat_sys_*() prototypes
syscalls: sort syscall prototypes in include/linux/compat.h
net: remove compat_sys_*() prototypes from net/compat.h
syscalls: sort syscall prototypes in include/linux/syscalls.h
kexec: move sys_kexec_load() prototype to syscalls.h
x86/sigreturn: use SYSCALL_DEFINE0
x86: fix sys_sigreturn() return type to be long, not unsigned long
x86/ioport: add ksys_ioperm() helper; remove in-kernel calls to sys_ioperm()
mm: add ksys_readahead() helper; remove in-kernel calls to sys_readahead()
mm: add ksys_mmap_pgoff() helper; remove in-kernel calls to sys_mmap_pgoff()
mm: add ksys_fadvise64_64() helper; remove in-kernel call to sys_fadvise64_64()
fs: add ksys_fallocate() wrapper; remove in-kernel calls to sys_fallocate()
fs: add ksys_p{read,write}64() helpers; remove in-kernel calls to syscalls
fs: add ksys_truncate() wrapper; remove in-kernel calls to sys_truncate()
fs: add ksys_sync_file_range helper(); remove in-kernel calls to syscall
kernel: add ksys_setsid() helper; remove in-kernel call to sys_setsid()
kernel: add ksys_unshare() helper; remove in-kernel calls to sys_unshare()
...
Pull wait_var_event updates from Ingo Molnar:
"This introduces the new wait_var_event() API, which is a more flexible
waiting primitive than wait_on_atomic_t().
All wait_on_atomic_t() users are migrated over to the new API and
wait_on_atomic_t() is removed. The migration fixes one bug and should
result in no functional changes for the other usecases"
* 'sched-wait-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/wait: Improve __var_waitqueue() code generation
sched/wait: Remove the wait_on_atomic_t() API
sched/wait, arch/mips: Fix and convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, fs/ocfs2: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, fs/nfs: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, fs/fscache: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, fs/btrfs: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, fs/afs: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, drivers/media: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait, drivers/drm: Convert wait_on_atomic_t() usage to the new wait_var_event() API
sched/wait: Introduce wait_var_event()
Pull scheduler updates from Ingo Molnar:
"The main scheduler changes in this cycle were:
- NUMA balancing improvements (Mel Gorman)
- Further load tracking improvements (Patrick Bellasi)
- Various NOHZ balancing cleanups and optimizations (Peter Zijlstra)
- Improve blocked load handling, in particular we can now reduce and
eventually stop periodic load updates on 'very idle' CPUs. (Vincent
Guittot)
- On isolated CPUs offload the final 1Hz scheduler tick as well, plus
related cleanups and reorganization. (Frederic Weisbecker)
- Core scheduler code cleanups (Ingo Molnar)"
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (45 commits)
sched/core: Update preempt_notifier_key to modern API
sched/cpufreq: Rate limits for SCHED_DEADLINE
sched/fair: Update util_est only on util_avg updates
sched/cpufreq/schedutil: Use util_est for OPP selection
sched/fair: Use util_est in LB and WU paths
sched/fair: Add util_est on top of PELT
sched/core: Remove TASK_ALL
sched/completions: Use bool in try_wait_for_completion()
sched/fair: Update blocked load when newly idle
sched/fair: Move idle_balance()
sched/nohz: Merge CONFIG_NO_HZ_COMMON blocks
sched/fair: Move rebalance_domains()
sched/nohz: Optimize nohz_idle_balance()
sched/fair: Reduce the periodic update duration
sched/nohz: Stop NOHZ stats when decayed
sched/cpufreq: Provide migration hint
sched/nohz: Clean up nohz enter/exit
sched/fair: Update blocked load from NEWIDLE
sched/fair: Add NOHZ stats balancing
sched/fair: Restructure nohz_balance_kick()
...
Using the sched-internal do_sched_yield() helper allows us to get rid of
the sched-internal call to the sys_sched_yield() syscall.
This patch is part of a series which removes in-kernel calls to syscalls.
On this basis, the syscall entry path can be streamlined. For details, see
http://lkml.kernel.org/r/20180325162527.GA17492@light.dominikbrodowski.net
Cc: Ingo Molnar <mingo@redhat.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
This patch prevents the 'global_tunables_lock' mutex from being
unlocked before being locked. This mutex is not locked if the
sugov_kthread_create() function fails.
Signed-off-by: Jules Maselbas <jules.maselbas@arm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Chris Redpath <chris.redpath@arm.com>
Cc: Dietmar Eggermann <dietmar.eggemann@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Stephen Kyle <stephen.kyle@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Cc: nd@arm.com
Link: http://lkml.kernel.org/r/20180329144301.38419-1-jules.maselbas@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull scheduler fixes from Ingo Molnar:
"Two sched debug output related fixes: a console output fix and
formatting fixes"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/debug: Adjust newlines for better alignment
sched/debug: Fix per-task line continuation for console output
When the SCHED_DEADLINE scheduling class increases the CPU utilization, it
should not wait for the rate limit, otherwise it may miss some deadline.
Tests using rt-app on Exynos5422 with up to 10 SCHED_DEADLINE tasks have
shown reductions of even 10% of deadline misses with a negligible
increase of energy consumption (measured through Baylibre Cape).
Signed-off-by: Claudio Scordino <claudio@evidence.eu.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: linux-pm@vger.kernel.org
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Todd Kjos <tkjos@android.com>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Link: https://lkml.kernel.org/r/1520937340-2755-1-git-send-email-claudio@evidence.eu.com
Scheduler debug stats include newlines that display out of alignment
when prefixed by timestamps. For example, the dmesg utility:
% echo t > /proc/sysrq-trigger
% dmesg
...
[ 83.124251]
runnable tasks:
S task PID tree-key switches prio wait-time
sum-exec sum-sleep
-----------------------------------------------------------------------------------------------------------
At the same time, some syslog utilities (like rsyslog by default) don't
like the additional newlines control characters, saving lines like this
to /var/log/messages:
Mar 16 16:02:29 localhost kernel: #012runnable tasks:#012 S task PID tree-key ...
^^^^ ^^^^
Clean these up by moving newline characters to their own SEQ_printf
invocation. This leaves the /proc/sched_debug unchanged, but brings the
entire output into alignment when prefixed:
% echo t > /proc/sysrq-trigger
% dmesg
...
[ 62.410368] runnable tasks:
[ 62.410368] S task PID tree-key switches prio wait-time sum-exec sum-sleep
[ 62.410369] -----------------------------------------------------------------------------------------------------------
[ 62.410369] I kworker/u12:0 5 1932.215593 332 120 0.000000 3.621252 0.000000 0 0 /
and no escaped control characters from rsyslog in /var/log/messages:
Mar 16 16:15:06 localhost kernel: runnable tasks:
Mar 16 16:15:06 localhost kernel: S task PID tree-key ...
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1521484555-8620-3-git-send-email-joe.lawrence@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When the SEQ_printf() macro prints to the console, it runs a simple
printk() without KERN_CONT "continued" line printing. The result of
this is oddly wrapped task info, for example:
% echo t > /proc/sysrq-trigger
% dmesg
...
runnable tasks:
...
[ 29.608611] I
[ 29.608613] rcu_sched 8 3252.013846 4087 120
[ 29.608614] 0.000000 29.090111 0.000000
[ 29.608615] 0 0
[ 29.608616] /
Modify SEQ_printf to use pr_cont() for expected one-line results:
% echo t > /proc/sysrq-trigger
% dmesg
...
runnable tasks:
...
[ 106.716329] S cpuhp/5 37 2006.315026 14 120 0.000000 0.496893 0.000000 0 0 /
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1521484555-8620-2-git-send-email-joe.lawrence@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since we fixed hash_64() to not suck there is no need to play games to
attempt to improve the hash value on 64-bit.
Also, since we don't use the bit value for the variables, use hash_ptr()
directly.
No change in functionality.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: George Spelvin <linux@sciencehorizons.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
There are no users left (everyone got converted to wait_var_event()), remove it.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
As a replacement for the wait_on_atomic_t() API provide the
wait_var_event() API.
The wait_var_event() API is based on the very same hashed-waitqueue
idea, but doesn't care about the type (atomic_t) or the specific
condition (atomic_read() == 0). IOW. it's much more widely
applicable/flexible.
It shares all the benefits/disadvantages of a hashed-waitqueue
approach with the existing wait_on_atomic_t/wait_on_bit() APIs.
The API is modeled after the existing wait_event() API, but instead of
taking a wait_queue_head, it takes an address. This addresses is
hashed to obtain a wait_queue_head from the bit_wait_table.
Similar to the wait_event() API, it takes a condition expression as
second argument and will wait until this expression becomes true.
The following are (mostly) identical replacements:
wait_on_atomic_t(&my_atomic, atomic_t_wait, TASK_UNINTERRUPTIBLE);
wake_up_atomic_t(&my_atomic);
wait_var_event(&my_atomic, !atomic_read(&my_atomic));
wake_up_var(&my_atomic);
The only difference is that wake_up_var() is an unconditional wakeup
and doesn't check the previously hard-coded (atomic_read() == 0)
condition here. This is of little concequence, since most callers are
already conditional on atomic_dec_and_test() and the ones that are
not, are trivial to make so.
Tested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The estimated utilization of a task is currently updated every time the
task is dequeued. However, to keep overheads under control, PELT signals
are effectively updated at maximum once every 1ms.
Thus, for really short running tasks, it can happen that their util_avg
value has not been updates since their last enqueue. If such tasks are
also frequently running tasks (e.g. the kind of workload generated by
hackbench) it can also happen that their util_avg is updated only every
few activations.
This means that updating util_est at every dequeue potentially introduces
not necessary overheads and it's also conceptually wrong if the util_avg
signal has never been updated during a task activation.
Let's introduce a throttling mechanism on task's util_est updates
to sync them with util_avg updates. To make the solution memory
efficient, both in terms of space and load/store operations, we encode a
synchronization flag into the LSB of util_est.enqueued.
This makes util_est an even values only metric, which is still
considered good enough for its purpose.
The synchronization bit is (re)set by __update_load_avg_se() once the
PELT signal of a task has been updated during its last activation.
Such a throttling mechanism allows to keep under control util_est
overheads in the wakeup hot path, thus making it a suitable mechanism
which can be enabled also on high-intensity workload systems.
Thus, this now switches on by default the estimation utilization
scheduler feature.
Suggested-by: Chris Redpath <chris.redpath@arm.com>
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Steve Muckle <smuckle@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Todd Kjos <tkjos@android.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Link: http://lkml.kernel.org/r/20180309095245.11071-5-patrick.bellasi@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When schedutil looks at the CPU utilization, the current PELT value for
that CPU is returned straight away. In certain scenarios this can have
undesired side effects and delays on frequency selection.
For example, since the task utilization is decayed at wakeup time, a
long sleeping big task newly enqueued does not add immediately a
significant contribution to the target CPU. This introduces some latency
before schedutil will be able to detect the best frequency required by
that task.
Moreover, the PELT signal build-up time is a function of the current
frequency, because of the scale invariant load tracking support. Thus,
starting from a lower frequency, the utilization build-up time will
increase even more and further delays the selection of the actual
frequency which better serves the task requirements.
In order to reduce these kind of latencies, we integrate the usage
of the CPU's estimated utilization in the sugov_get_util function.
This allows to properly consider the expected utilization of a CPU which,
for example, has just got a big task running after a long sleep period.
Ultimately this allows to select the best frequency to run a task
right after its wake-up.
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steve Muckle <smuckle@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Todd Kjos <tkjos@android.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Link: http://lkml.kernel.org/r/20180309095245.11071-4-patrick.bellasi@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When the scheduler looks at the CPU utilization, the current PELT value
for a CPU is returned straight away. In certain scenarios this can have
undesired side effects on task placement.
For example, since the task utilization is decayed at wakeup time, when
a long sleeping big task is enqueued it does not add immediately a
significant contribution to the target CPU.
As a result we generate a race condition where other tasks can be placed
on the same CPU while it is still considered relatively empty.
In order to reduce this kind of race conditions, this patch introduces the
required support to integrate the usage of the CPU's estimated utilization
in the wakeup path, via cpu_util_wake(), as well as in the load-balance
path, via cpu_util() which is used by update_sg_lb_stats().
The estimated utilization of a CPU is defined to be the maximum between
its PELT's utilization and the sum of the estimated utilization (at
previous dequeue time) of all the tasks currently RUNNABLE on that CPU.
This allows to properly represent the spare capacity of a CPU which, for
example, has just got a big task running since a long sleep period.
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Steve Muckle <smuckle@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Todd Kjos <tkjos@android.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Link: http://lkml.kernel.org/r/20180309095245.11071-3-patrick.bellasi@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The util_avg signal computed by PELT is too variable for some use-cases.
For example, a big task waking up after a long sleep period will have its
utilization almost completely decayed. This introduces some latency before
schedutil will be able to pick the best frequency to run a task.
The same issue can affect task placement. Indeed, since the task
utilization is already decayed at wakeup, when the task is enqueued in a
CPU, this can result in a CPU running a big task as being temporarily
represented as being almost empty. This leads to a race condition where
other tasks can be potentially allocated on a CPU which just started to run
a big task which slept for a relatively long period.
Moreover, the PELT utilization of a task can be updated every [ms], thus
making it a continuously changing value for certain longer running
tasks. This means that the instantaneous PELT utilization of a RUNNING
task is not really meaningful to properly support scheduler decisions.
For all these reasons, a more stable signal can do a better job of
representing the expected/estimated utilization of a task/cfs_rq.
Such a signal can be easily created on top of PELT by still using it as
an estimator which produces values to be aggregated on meaningful
events.
This patch adds a simple implementation of util_est, a new signal built on
top of PELT's util_avg where:
util_est(task) = max(task::util_avg, f(task::util_avg@dequeue))
This allows to remember how big a task has been reported by PELT in its
previous activations via f(task::util_avg@dequeue), which is the new
_task_util_est(struct task_struct*) function added by this patch.
If a task should change its behavior and it runs longer in a new
activation, after a certain time its util_est will just track the
original PELT signal (i.e. task::util_avg).
The estimated utilization of cfs_rq is defined only for root ones.
That's because the only sensible consumer of this signal are the
scheduler and schedutil when looking for the overall CPU utilization
due to FAIR tasks.
For this reason, the estimated utilization of a root cfs_rq is simply
defined as:
util_est(cfs_rq) = max(cfs_rq::util_avg, cfs_rq::util_est::enqueued)
where:
cfs_rq::util_est::enqueued = sum(_task_util_est(task))
for each RUNNABLE task on that root cfs_rq
It's worth noting that the estimated utilization is tracked only for
objects of interests, specifically:
- Tasks: to better support tasks placement decisions
- root cfs_rqs: to better support both tasks placement decisions as
well as frequencies selection
Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Steve Muckle <smuckle@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Todd Kjos <tkjos@android.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Link: http://lkml.kernel.org/r/20180309095245.11071-2-patrick.bellasi@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull cgroup fixes from Tejun Heo:
"Two commits to fix the following subtle cgroup2 behavior bugs:
- cpu.max was rejecting config when it shouldn't
- thread mode enable was allowed when it shouldn't"
* 'for-4.16-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup: fix rule checking for threaded mode switching
sched, cgroup: Don't reject lower cpu.max on ancestors
Since the return type of the function is bool, the internal
'ret' variable should be bool too.
Signed-off-by: Gaurav Jindal<gauravjindal1104@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180221125407.GA14292@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When NEWLY_IDLE load balance is not triggered, we might need to update the
blocked load anyway. We can kick an ilb so an idle CPU will take care of
updating blocked load or we can try to update them locally before entering
idle. In the latter case, we reuse part of the nohz_idle_balance.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: brendan.jackman@arm.com
Cc: dietmar.eggemann@arm.com
Cc: morten.rasmussen@foss.arm.com
Cc: valentin.schneider@arm.com
Link: http://lkml.kernel.org/r/1518622006-16089-4-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We're going to want to call nohz_idle_balance() or parts thereof from
idle_balance(). Since we already have a forward declaration of
idle_balance() move it down such that it's below nohz_idle_balance()
avoiding the need for a forward declaration for that.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Now that we have two back-to-back NO_HZ_COMMON blocks, merge them.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This pure code movement results in two #ifdef CONFIG_NO_HZ_COMMON
sections landing next to each other.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Avoid calling update_blocked_averages() when it does not in fact have
any by re-using/extending update_nohz_stats().
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Instead of using the cfs_rq_is_decayed() which monitors all *_avg
and *_sum, we create a cfs_rq_has_blocked() which only takes care of
util_avg and load_avg. We are only interested by these 2 values which are
decaying faster than the *_sum so we can stop the periodic update earlier.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: brendan.jackman@arm.com
Cc: dietmar.eggemann@arm.com
Cc: morten.rasmussen@foss.arm.com
Cc: valentin.schneider@arm.com
Link: http://lkml.kernel.org/r/1518517879-2280-3-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Stopped the periodic update of blocked load when all idle CPUs have fully
decayed. We introduce a new nohz.has_blocked that reflect if some idle
CPUs has blocked load that have to be periodiccally updated. nohz.has_blocked
is set everytime that a Idle CPU can have blocked load and it is then clear
when no more blocked load has been detected during an update. We don't need
atomic operation but only to make cure of the right ordering when updating
nohz.idle_cpus_mask and nohz.has_blocked.
Suggested-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: brendan.jackman@arm.com
Cc: dietmar.eggemann@arm.com
Cc: morten.rasmussen@foss.arm.com
Cc: valentin.schneider@arm.com
Link: http://lkml.kernel.org/r/1518517879-2280-2-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
It was suggested that a migration hint might be usefull for the
CPU-freq governors.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The primary observation is that nohz enter/exit is always from the
current CPU, therefore NOHZ_TICK_STOPPED does not in fact need to be
an atomic.
Secondary is that we appear to have 2 nearly identical hooks in the
nohz enter code, set_cpu_sd_state_idle() and
nohz_balance_enter_idle(). Fold the whole set_cpu_sd_state thing into
nohz_balance_{enter,exit}_idle.
Removes an atomic op from both enter and exit paths.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since we already iterate CPUs looking for work on NEWIDLE, use this
iteration to age the blocked load. If the domain for which this is
done completely spand the idle set, we can push the ILB based aging
forward.
Suggested-by: Brendan Jackman <brendan.jackman@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Teach the idle balancer about the need to update statistics which have
a different periodicity from regular balancing.
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The current:
if (nohz_kick_needed())
nohz_balancer_kick()
is pointless complexity, fold them into a single call and avoid the
various conditions at the call site.
When we introduce multiple different needs to kick the ilb, the above
construct also becomes a problem.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Split the NOHZ idle balancer into doing two separate actions:
- update blocked load statistic
- actually load-balance
Since the latter requires the former, ensure this happens. For now
always tag both bits at the same time.
Prepares for a future where we can toggle only the STATS bit.
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Using atomic_t allows us to use the more flexible bitops provided
there. Also its smaller.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Instead of trying to duplicate scheduler state to track if an RT task
is running, directly use the scheduler runqueue state for it.
This vastly simplifies things and fixes a number of bugs related to
sugov and the scheduler getting out of sync wrt this state.
As a consequence we not also update the remove cfs/dl state when
iterating the shared mask.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Due to using GCC defines for configuration, some labels might be unused in
certain configurations. While adding a __maybe_unused to the label is
fine in general, the line has to be terminated with ';'. This is also
reflected in the GCC documentation, but GCC parsed the previous variant
without an error message.
This has been spotted while compiling with goto-cc, the compiler for the
CPROVER tool suite.
Signed-off-by: Norbert Manthey <nmanthey@amazon.de>
Signed-off-by: Michael Tautschnig <tautschn@amazon.co.uk>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1519717660-16157-1-git-send-email-nmanthey@amazon.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Make it easier to concatenate all the scheduler .c files for single-module
compilation.
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Merge these two small .c modules as they implement two aspects
of idle task handling.
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Do the following cleanups and simplifications:
- sched/sched.h already includes <asm/paravirt.h>, so no need to
include it in sched/core.c again.
- order the <linux/sched/*.h> headers alphabetically
- add all <linux/sched/*.h> headers to kernel/sched/sched.h
- remove all unnecessary includes from the .c files that
are already included in kernel/sched/sched.h.
Finally, make all scheduler .c files use a single common header:
#include "sched.h"
... which now contains a union of the relied upon headers.
This makes the various .c files easier to read and easier to handle.
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
A good number of small style inconsistencies have accumulated
in the scheduler core, so do a pass over them to harmonize
all these details:
- fix speling in comments,
- use curly braces for multi-line statements,
- remove unnecessary parentheses from integer literals,
- capitalize consistently,
- remove stray newlines,
- add comments where necessary,
- remove invalid/unnecessary comments,
- align structure definitions and other data types vertically,
- add missing newlines for increased readability,
- fix vertical tabulation where it's misaligned,
- harmonize preprocessor conditional block labeling
and vertical alignment,
- remove line-breaks where they uglify the code,
- add newline after local variable definitions,
No change in functionality:
md5:
1191fa0a890cfa8132156d2959d7e9e2 built-in.o.before.asm
1191fa0a890cfa8132156d2959d7e9e2 built-in.o.after.asm
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
- Fixed style error: Missing space before the open parenthesis
- Fixed style warnings: 2x Missing blank line after declaration
One warning left: else after return
(I don't feel comfortable fixing that without side effects)
Signed-off-by: Mario Leinweber <marioleinweber@web.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/20180302182007.28691-1-marioleinweber@web.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Now that the 1Hz tick is offloaded to workqueues, we can safely remove
the residual code that used to handle it locally.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Chris Metcalf <cmetcalf@mellanox.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Link: http://lkml.kernel.org/r/1519186649-3242-7-git-send-email-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When a CPU runs in full dynticks mode, a 1Hz tick remains in order to
keep the scheduler stats alive. However this residual tick is a burden
for bare metal tasks that can't stand any interruption at all, or want
to minimize them.
The usual boot parameters "nohz_full=" or "isolcpus=nohz" will now
outsource these scheduler ticks to the global workqueue so that a
housekeeping CPU handles those remotely. The sched_class::task_tick()
implementations have been audited and look safe to be called remotely
as the target runqueue and its current task are passed in parameter
and don't seem to be accessed locally.
Note that in the case of using isolcpus, it's still up to the user to
affine the global workqueues to the housekeeping CPUs through
/sys/devices/virtual/workqueue/cpumask or domains isolation
"isolcpus=nohz,domain".
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Chris Metcalf <cmetcalf@mellanox.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Link: http://lkml.kernel.org/r/1519186649-3242-6-git-send-email-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
As we prepare for offloading the residual 1hz scheduler ticks to
workqueue, let's affine those to housekeepers so that they don't
interrupt the CPUs that don't want to be disturbed.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Chris Metcalf <cmetcalf@mellanox.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Link: http://lkml.kernel.org/r/1519186649-3242-5-git-send-email-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Do that rename in order to normalize the hrtick namespace.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Chris Metcalf <cmetcalf@mellanox.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Link: http://lkml.kernel.org/r/1519186649-3242-2-git-send-email-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
If wake_affine() pulls a task to another node for any reason and the node is
no longer preferred then temporarily stop automatic NUMA balancing pulling
the task back. Otherwise, tasks with a strong waker/wakee relationship
may constantly fight automatic NUMA balancing over where a task should
be placed.
Once again netperf is interesting here. The performance barely changes
but automatic NUMA balancing is interesting:
Hmean send-64 354.67 ( 0.00%) 352.15 ( -0.71%)
Hmean send-128 702.91 ( 0.00%) 693.84 ( -1.29%)
Hmean send-256 1350.07 ( 0.00%) 1344.19 ( -0.44%)
Hmean send-1024 5124.38 ( 0.00%) 4941.24 ( -3.57%)
Hmean send-2048 9687.44 ( 0.00%) 9624.45 ( -0.65%)
Hmean send-3312 14577.64 ( 0.00%) 14514.35 ( -0.43%)
Hmean send-4096 16393.62 ( 0.00%) 16488.30 ( 0.58%)
Hmean send-8192 26877.26 ( 0.00%) 26431.63 ( -1.66%)
Hmean send-16384 38683.43 ( 0.00%) 38264.91 ( -1.08%)
Hmean recv-64 354.67 ( 0.00%) 352.15 ( -0.71%)
Hmean recv-128 702.91 ( 0.00%) 693.84 ( -1.29%)
Hmean recv-256 1350.07 ( 0.00%) 1344.19 ( -0.44%)
Hmean recv-1024 5124.38 ( 0.00%) 4941.24 ( -3.57%)
Hmean recv-2048 9687.43 ( 0.00%) 9624.45 ( -0.65%)
Hmean recv-3312 14577.59 ( 0.00%) 14514.35 ( -0.43%)
Hmean recv-4096 16393.55 ( 0.00%) 16488.20 ( 0.58%)
Hmean recv-8192 26876.96 ( 0.00%) 26431.29 ( -1.66%)
Hmean recv-16384 38682.41 ( 0.00%) 38263.94 ( -1.08%)
NUMA alloc hit 1465986 1423090
NUMA alloc miss 0 0
NUMA interleave hit 0 0
NUMA alloc local 1465897 1423003
NUMA base PTE updates 1473 1420
NUMA huge PMD updates 0 0
NUMA page range updates 1473 1420
NUMA hint faults 1383 1312
NUMA hint local faults 451 124
NUMA hint local percent 32 9
There is a slight degrading in performance but there are slightly fewer
NUMA faults. There is a large drop in the percentage of local faults but
the bulk of migrations for netperf are in small shared libraries so it's
reflecting the fact that automatic NUMA balancing has backed off. This is
a case where despite wake_affine() and automatic NUMA balancing fighting
for placement that there is a marginal benefit to rescheduling to local
data quickly. However, it should be noted that wake_affine() and automatic
NUMA balancing fighting each other constantly is undesirable.
However, the benefit in other cases is large. This is the result for NAS
with the D class sizing on a 4-socket machine:
nas-mpi
4.15.0 4.15.0
sdnuma-v1r23 delayretry-v1r23
Time cg.D 557.00 ( 0.00%) 431.82 ( 22.47%)
Time ep.D 77.83 ( 0.00%) 79.01 ( -1.52%)
Time is.D 26.46 ( 0.00%) 26.64 ( -0.68%)
Time lu.D 727.14 ( 0.00%) 597.94 ( 17.77%)
Time mg.D 191.35 ( 0.00%) 146.85 ( 23.26%)
4.15.0 4.15.0
sdnuma-v1r23delayretry-v1r23
User 75665.20 70413.30
System 20321.59 8861.67
Elapsed 766.13 634.92
Minor Faults 16528502 7127941
Major Faults 4553 5068
NUMA alloc local 6963197 6749135
NUMA base PTE updates 366409093 107491434
NUMA huge PMD updates 687556 198880
NUMA page range updates 718437765 209317994
NUMA hint faults 13643410 4601187
NUMA hint local faults 9212593 3063996
NUMA hint local percent 67 66
Note the massive reduction in system CPU usage even though the percentage
of local faults is barely affected. There is a massive reduction in the
number of PTE updates showing that automatic NUMA balancing has backed off.
A critical observation is also that there is a massive reduction in minor
faults which is due to far fewer NUMA hinting faults being trapped.
There were questions on NAS OMP and how it behaved related to threads
being bound to CPUs. First, there are more gains than losses with this
patch applied and a reduction in system CPU usage:
nas-omp
4.16.0-rc1 4.16.0-rc1
sdnuma-v2r1 delayretry-v2r1
Time bt.D 436.71 ( 0.00%) 430.05 ( 1.53%)
Time cg.D 201.02 ( 0.00%) 180.87 ( 10.02%)
Time ep.D 32.84 ( 0.00%) 32.68 ( 0.49%)
Time is.D 9.63 ( 0.00%) 9.64 ( -0.10%)
Time lu.D 331.20 ( 0.00%) 304.80 ( 7.97%)
Time mg.D 54.87 ( 0.00%) 52.72 ( 3.92%)
Time sp.D 1108.78 ( 0.00%) 917.10 ( 17.29%)
Time ua.D 378.81 ( 0.00%) 398.83 ( -5.28%)
4.16.0-rc1 4.16.0-rc1
sdnuma-v2r1delayretry-v2r1
User 305633.08 296751.91
System 451.75 357.80
Elapsed 2595.73 2368.13
However, it does not close the gap between binding and being unbound. There
is negligible difference between the performance of the baseline and a
patched kernel when threads are bound so it is not presented here:
4.16.0-rc1 4.16.0-rc1
delayretry-bind delayretry-unbound
Time bt.D 385.02 ( 0.00%) 430.05 ( -11.70%)
Time cg.D 144.02 ( 0.00%) 180.87 ( -25.59%)
Time ep.D 32.85 ( 0.00%) 32.68 ( 0.52%)
Time is.D 10.52 ( 0.00%) 9.64 ( 8.37%)
Time lu.D 285.31 ( 0.00%) 304.80 ( -6.83%)
Time mg.D 43.21 ( 0.00%) 52.72 ( -22.01%)
Time sp.D 820.24 ( 0.00%) 917.10 ( -11.81%)
Time ua.D 337.09 ( 0.00%) 398.83 ( -18.32%)
4.16.0-rc1 4.16.0-rc1
delayretry-binddelayretry-unbound
User 277731.25 296751.91
System 261.29 357.80
Elapsed 2100.55 2368.13
Unfortunately, while performance is improved by the patch, there is still
quite a long way to go before it's equivalent to hard binding.
Other workloads like hackbench, tbench, dbench and schbench are barely
affected. dbench shows a mix of gains and losses depending on the machine
although in general, the results are more stable.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-7-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
find_idlest_group() compares a local group with each other group to select
the one that is most idle. When comparing groups in different NUMA domains,
a very slight imbalance is enough to select a remote NUMA node even if the
runnable load on both groups is 0 or close to 0. This ignores the cost of
remote accesses entirely and is a problem when selecting the CPU for a
newly forked task to run on. This is problematic when a forking server
is almost guaranteed to run on a remote node incurring numerous remote
accesses and potentially causing automatic NUMA balancing to try migrate
the task back or migrate the data to another node. Similar weirdness is
observed if a basic shell command pipes output to another as each process
in the pipeline is likely to start on different nodes and then get adjusted
later by wake_affine().
This patch adds imbalance to remote domains when considering whether to
select CPUs from remote domains. If the local domain is selected, imbalance
will still be used to try select a CPU from a lower scheduler domain's group
instead of stacking tasks on the same CPU.
A variety of workloads and machines were tested and as expected, there is no
difference on UMA. The difference on NUMA can be dramatic. This is a comparison
of elapsed times running the git regression test suite. It's fork-intensive with
short-lived processes:
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
Elapsed min 1706.06 ( 0.00%) 1435.94 ( 15.83%)
Elapsed mean 1709.53 ( 0.00%) 1436.98 ( 15.94%)
Elapsed stddev 2.16 ( 0.00%) 1.01 ( 53.38%)
Elapsed coeffvar 0.13 ( 0.00%) 0.07 ( 44.54%)
Elapsed max 1711.59 ( 0.00%) 1438.01 ( 15.98%)
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
User 5434.12 5188.41
System 4878.77 3467.09
Elapsed 10259.06 8624.21
That shows a considerable reduction in elapsed times. It's important to
note that automatic NUMA balancing does not affect this load as processes
are too short-lived.
There is also a noticable impact on hackbench such as this example using
processes and pipes:
hackbench-process-pipes
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
Amean 1 1.0973 ( 0.00%) 0.9393 ( 14.40%)
Amean 4 1.3427 ( 0.00%) 1.3730 ( -2.26%)
Amean 7 1.4233 ( 0.00%) 1.6670 ( -17.12%)
Amean 12 3.0250 ( 0.00%) 3.3013 ( -9.13%)
Amean 21 9.0860 ( 0.00%) 9.5343 ( -4.93%)
Amean 30 14.6547 ( 0.00%) 13.2433 ( 9.63%)
Amean 48 22.5447 ( 0.00%) 20.4303 ( 9.38%)
Amean 79 29.2010 ( 0.00%) 26.7853 ( 8.27%)
Amean 110 36.7443 ( 0.00%) 35.8453 ( 2.45%)
Amean 141 45.8533 ( 0.00%) 42.6223 ( 7.05%)
Amean 172 55.1317 ( 0.00%) 50.6473 ( 8.13%)
Amean 203 64.4420 ( 0.00%) 58.3957 ( 9.38%)
Amean 234 73.2293 ( 0.00%) 67.1047 ( 8.36%)
Amean 265 80.5220 ( 0.00%) 75.7330 ( 5.95%)
Amean 296 88.7567 ( 0.00%) 82.1533 ( 7.44%)
It's not a universal win as there are occasions when spreading wide and
quickly is a benefit but it's more of a win than it is a loss. For other
workloads, there is little difference but netperf is interesting. Without
the patch, the server and client starts on different nodes but quickly get
migrated due to wake_affine. Hence, the difference is overall performance
is marginal but detectable:
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
Hmean send-64 349.09 ( 0.00%) 354.67 ( 1.60%)
Hmean send-128 699.16 ( 0.00%) 702.91 ( 0.54%)
Hmean send-256 1316.34 ( 0.00%) 1350.07 ( 2.56%)
Hmean send-1024 5063.99 ( 0.00%) 5124.38 ( 1.19%)
Hmean send-2048 9705.19 ( 0.00%) 9687.44 ( -0.18%)
Hmean send-3312 14359.48 ( 0.00%) 14577.64 ( 1.52%)
Hmean send-4096 16324.20 ( 0.00%) 16393.62 ( 0.43%)
Hmean send-8192 26112.61 ( 0.00%) 26877.26 ( 2.93%)
Hmean send-16384 37208.44 ( 0.00%) 38683.43 ( 3.96%)
Hmean recv-64 349.09 ( 0.00%) 354.67 ( 1.60%)
Hmean recv-128 699.16 ( 0.00%) 702.91 ( 0.54%)
Hmean recv-256 1316.34 ( 0.00%) 1350.07 ( 2.56%)
Hmean recv-1024 5063.99 ( 0.00%) 5124.38 ( 1.19%)
Hmean recv-2048 9705.16 ( 0.00%) 9687.43 ( -0.18%)
Hmean recv-3312 14359.42 ( 0.00%) 14577.59 ( 1.52%)
Hmean recv-4096 16323.98 ( 0.00%) 16393.55 ( 0.43%)
Hmean recv-8192 26111.85 ( 0.00%) 26876.96 ( 2.93%)
Hmean recv-16384 37206.99 ( 0.00%) 38682.41 ( 3.97%)
However, what is very interesting is how automatic NUMA balancing behaves.
Each netperf instance runs long enough for balancing to activate:
NUMA base PTE updates 4620 1473
NUMA huge PMD updates 0 0
NUMA page range updates 4620 1473
NUMA hint faults 4301 1383
NUMA hint local faults 1309 451
NUMA hint local percent 30 32
NUMA pages migrated 1335 491
AutoNUMA cost 21% 6%
There is an unfortunate number of remote faults although tracing indicated
that the vast majority are in shared libraries. However, the tendency to
start tasks on the same node if there is capacity means that there were
far fewer PTE updates and faults incurred overall.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-6-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When a task exits, it notifies the parent that it has exited. This is a
sync wakeup and the exiting task may pull the parent towards the wakers
CPU. For simple workloads like using a shell, it was observed that the
shell is pulled across nodes by exiting processes. This is daft as the
parent may be long-lived and properly placed. This patch special cases a
sync wakeup on exit to avoid pulling tasks across nodes. Testing on a range
of workloads and machines showed very little differences in performance
although there was a small 3% boost on some machines running a shellscript
intensive workload (git regression test suite).
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-5-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
wake_affine_weight() will consider migrating a task to, or near, the current
CPU if there is a load imbalance. If the CPUs share LLC then either CPU
is valid as a search-for-idle-sibling target and equally appropriate for
stacking two tasks on one CPU if an idle sibling is unavailable. If they do
not share cache then a cross-node migration potentially impacts locality
so while they are equal from a CPU capacity point of view, they are not
equal in terms of memory locality. In either case, it's more appropriate
to migrate only if there is a difference in their effective load.
This patch modifies wake_affine_weight() to only consider migrating a task
if there is a load imbalance for normal wakeups but will allow potential
stacking if the loads are equal and it's a sync wakeup.
For the most part, the different in performance is marginal. For example,
on a 4-socket server running netperf UDP_STREAM on localhost the differences
are as follows:
4.15.0 4.15.0
16rc0 noequal-v1r23
Hmean send-64 355.47 ( 0.00%) 349.50 ( -1.68%)
Hmean send-128 697.98 ( 0.00%) 693.35 ( -0.66%)
Hmean send-256 1328.02 ( 0.00%) 1318.77 ( -0.70%)
Hmean send-1024 5051.83 ( 0.00%) 5051.11 ( -0.01%)
Hmean send-2048 9637.02 ( 0.00%) 9601.34 ( -0.37%)
Hmean send-3312 14355.37 ( 0.00%) 14414.51 ( 0.41%)
Hmean send-4096 16464.97 ( 0.00%) 16301.37 ( -0.99%)
Hmean send-8192 26722.42 ( 0.00%) 26428.95 ( -1.10%)
Hmean send-16384 38137.81 ( 0.00%) 38046.11 ( -0.24%)
Hmean recv-64 355.47 ( 0.00%) 349.50 ( -1.68%)
Hmean recv-128 697.98 ( 0.00%) 693.35 ( -0.66%)
Hmean recv-256 1328.02 ( 0.00%) 1318.77 ( -0.70%)
Hmean recv-1024 5051.83 ( 0.00%) 5051.11 ( -0.01%)
Hmean recv-2048 9636.95 ( 0.00%) 9601.30 ( -0.37%)
Hmean recv-3312 14355.32 ( 0.00%) 14414.48 ( 0.41%)
Hmean recv-4096 16464.74 ( 0.00%) 16301.16 ( -0.99%)
Hmean recv-8192 26721.63 ( 0.00%) 26428.17 ( -1.10%)
Hmean recv-16384 38136.00 ( 0.00%) 38044.88 ( -0.24%)
Stddev send-64 7.30 ( 0.00%) 4.75 ( 34.96%)
Stddev send-128 15.15 ( 0.00%) 22.38 ( -47.66%)
Stddev send-256 13.99 ( 0.00%) 19.14 ( -36.81%)
Stddev send-1024 105.73 ( 0.00%) 67.38 ( 36.27%)
Stddev send-2048 294.57 ( 0.00%) 223.88 ( 24.00%)
Stddev send-3312 302.28 ( 0.00%) 271.74 ( 10.10%)
Stddev send-4096 195.92 ( 0.00%) 121.10 ( 38.19%)
Stddev send-8192 399.71 ( 0.00%) 563.77 ( -41.04%)
Stddev send-16384 1163.47 ( 0.00%) 1103.68 ( 5.14%)
Stddev recv-64 7.30 ( 0.00%) 4.75 ( 34.96%)
Stddev recv-128 15.15 ( 0.00%) 22.38 ( -47.66%)
Stddev recv-256 13.99 ( 0.00%) 19.14 ( -36.81%)
Stddev recv-1024 105.73 ( 0.00%) 67.38 ( 36.27%)
Stddev recv-2048 294.59 ( 0.00%) 223.89 ( 24.00%)
Stddev recv-3312 302.24 ( 0.00%) 271.75 ( 10.09%)
Stddev recv-4096 196.03 ( 0.00%) 121.14 ( 38.20%)
Stddev recv-8192 399.86 ( 0.00%) 563.65 ( -40.96%)
Stddev recv-16384 1163.79 ( 0.00%) 1103.86 ( 5.15%)
The difference in overall performance is marginal but note that most
measurements are less variable. There were similar observations for other
netperf comparisons. hackbench with sockets or threads with processes or
threads showed minor difference with some reduction of migration. tbench
showed only marginal differences that were within the noise. dbench,
regardless of filesystem, showed minor differences all of which are
within noise. Multiple machines, both UMA and NUMA were tested without
any regressions showing up.
The biggest risk with a patch like this is affecting wakeup latencies.
However, the schbench load from Facebook which is very sensitive to wakeup
latency showed a mixed result with mostly improvements in wakeup latency:
4.15.0 4.15.0
16rc0 noequal-v1r23
Lat 50.00th-qrtle-1 38.00 ( 0.00%) 38.00 ( 0.00%)
Lat 75.00th-qrtle-1 49.00 ( 0.00%) 41.00 ( 16.33%)
Lat 90.00th-qrtle-1 52.00 ( 0.00%) 50.00 ( 3.85%)
Lat 95.00th-qrtle-1 54.00 ( 0.00%) 51.00 ( 5.56%)
Lat 99.00th-qrtle-1 63.00 ( 0.00%) 60.00 ( 4.76%)
Lat 99.50th-qrtle-1 66.00 ( 0.00%) 61.00 ( 7.58%)
Lat 99.90th-qrtle-1 78.00 ( 0.00%) 65.00 ( 16.67%)
Lat 50.00th-qrtle-2 38.00 ( 0.00%) 38.00 ( 0.00%)
Lat 75.00th-qrtle-2 42.00 ( 0.00%) 43.00 ( -2.38%)
Lat 90.00th-qrtle-2 46.00 ( 0.00%) 48.00 ( -4.35%)
Lat 95.00th-qrtle-2 49.00 ( 0.00%) 50.00 ( -2.04%)
Lat 99.00th-qrtle-2 55.00 ( 0.00%) 57.00 ( -3.64%)
Lat 99.50th-qrtle-2 58.00 ( 0.00%) 60.00 ( -3.45%)
Lat 99.90th-qrtle-2 65.00 ( 0.00%) 68.00 ( -4.62%)
Lat 50.00th-qrtle-4 41.00 ( 0.00%) 41.00 ( 0.00%)
Lat 75.00th-qrtle-4 45.00 ( 0.00%) 46.00 ( -2.22%)
Lat 90.00th-qrtle-4 50.00 ( 0.00%) 50.00 ( 0.00%)
Lat 95.00th-qrtle-4 54.00 ( 0.00%) 53.00 ( 1.85%)
Lat 99.00th-qrtle-4 61.00 ( 0.00%) 61.00 ( 0.00%)
Lat 99.50th-qrtle-4 65.00 ( 0.00%) 64.00 ( 1.54%)
Lat 99.90th-qrtle-4 76.00 ( 0.00%) 82.00 ( -7.89%)
Lat 50.00th-qrtle-8 48.00 ( 0.00%) 46.00 ( 4.17%)
Lat 75.00th-qrtle-8 55.00 ( 0.00%) 54.00 ( 1.82%)
Lat 90.00th-qrtle-8 60.00 ( 0.00%) 59.00 ( 1.67%)
Lat 95.00th-qrtle-8 63.00 ( 0.00%) 63.00 ( 0.00%)
Lat 99.00th-qrtle-8 71.00 ( 0.00%) 69.00 ( 2.82%)
Lat 99.50th-qrtle-8 74.00 ( 0.00%) 73.00 ( 1.35%)
Lat 99.90th-qrtle-8 98.00 ( 0.00%) 90.00 ( 8.16%)
Lat 50.00th-qrtle-16 56.00 ( 0.00%) 55.00 ( 1.79%)
Lat 75.00th-qrtle-16 68.00 ( 0.00%) 67.00 ( 1.47%)
Lat 90.00th-qrtle-16 77.00 ( 0.00%) 78.00 ( -1.30%)
Lat 95.00th-qrtle-16 82.00 ( 0.00%) 84.00 ( -2.44%)
Lat 99.00th-qrtle-16 90.00 ( 0.00%) 93.00 ( -3.33%)
Lat 99.50th-qrtle-16 93.00 ( 0.00%) 97.00 ( -4.30%)
Lat 99.90th-qrtle-16 110.00 ( 0.00%) 110.00 ( 0.00%)
Lat 50.00th-qrtle-32 68.00 ( 0.00%) 62.00 ( 8.82%)
Lat 75.00th-qrtle-32 90.00 ( 0.00%) 83.00 ( 7.78%)
Lat 90.00th-qrtle-32 110.00 ( 0.00%) 100.00 ( 9.09%)
Lat 95.00th-qrtle-32 122.00 ( 0.00%) 111.00 ( 9.02%)
Lat 99.00th-qrtle-32 145.00 ( 0.00%) 133.00 ( 8.28%)
Lat 99.50th-qrtle-32 154.00 ( 0.00%) 143.00 ( 7.14%)
Lat 99.90th-qrtle-32 2316.00 ( 0.00%) 515.00 ( 77.76%)
Lat 50.00th-qrtle-35 69.00 ( 0.00%) 72.00 ( -4.35%)
Lat 75.00th-qrtle-35 92.00 ( 0.00%) 95.00 ( -3.26%)
Lat 90.00th-qrtle-35 111.00 ( 0.00%) 114.00 ( -2.70%)
Lat 95.00th-qrtle-35 122.00 ( 0.00%) 124.00 ( -1.64%)
Lat 99.00th-qrtle-35 142.00 ( 0.00%) 144.00 ( -1.41%)
Lat 99.50th-qrtle-35 150.00 ( 0.00%) 154.00 ( -2.67%)
Lat 99.90th-qrtle-35 6104.00 ( 0.00%) 5640.00 ( 7.60%)
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-4-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
On sync wakeups, the previous CPU effective load may not be used so delay
the calculation until it's needed.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-3-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The only caller of wake_affine() knows the CPU ID. Pass it in instead of
rechecking it.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-2-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since schedutil kernel thread directly set priority to 0, the macro
SUGOV_KTHREAD_PRIORITY is not used. So remove it.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vikram Mulukutla <markivx@codeaurora.org>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Link: http://lkml.kernel.org/r/1518097702-9665-1-git-send-email-leo.yan@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Mark noticed that he had sporadic "spinlock recursion" warnings from
the DEBUG_SPINLOCK code. Now rq->lock is special in that the owner
changes in the middle of a context switch.
It so happens that we fix up the lock.owner too late, @prev can run
(remotely) the moment prev->on_cpu is cleared, this then allows @prev
to again try and acquire this rq->lock and trigger this warning.
So we have to switch lock.owner before clearing prev->on_cpu.
Do this by moving the DEBUG_SPINLOCK annotation from after switch_to()
to before switch_to() and collect all lockdep annotations there into
prepare_lock_switch() to mirror the existing finish_lock_switch().
Debugged-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
rq->clock_task may be updated between the two calls of
rq_clock_task() in update_curr_rt(). Calling rq_clock_task() only
once makes it more accurate and efficient, taking update_curr() as
reference.
Signed-off-by: Wen Yang <wen.yang99@zte.com.cn>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: zhong.weidong@zte.com.cn
Link: http://lkml.kernel.org/r/1517882008-44552-1-git-send-email-wen.yang99@zte.com.cn
Signed-off-by: Ingo Molnar <mingo@kernel.org>
rq->clock_task may be updated between the two calls of
rq_clock_task() in update_curr_dl(). Calling rq_clock_task() only
once makes it more accurate and efficient, taking update_curr() as
reference.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Wen Yang <wen.yang99@zte.com.cn>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: zhong.weidong@zte.com.cn
Link: http://lkml.kernel.org/r/1517882148-44599-1-git-send-email-wen.yang99@zte.com.cn
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Remove a useless space in # ifdef and align it with others.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1518512382-29426-1-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
While adding cgroup2 interface for the cpu controller, 0d5936344f
("sched: Implement interface for cgroup unified hierarchy") forgot to
update input validation and left it to reject cpu.max config if any
descendant has set a higher value.
cgroup2 officially supports delegation and a descendant must not be
able to restrict what its ancestors can configure. For absolute
limits such as cpu.max and memory.max, this means that the config at
each level should only act as the upper limit at that level and
shouldn't interfere with what other cgroups can configure.
This patch updates config validation on cgroup2 so that the cpu
controller follows the same convention.
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: 0d5936344f ("sched: Implement interface for cgroup unified hierarchy")
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org # v4.15+
The select_idle_sibling() (SIS) rewrite in commit:
10e2f1acd0 ("sched/core: Rewrite and improve select_idle_siblings()")
... replaced a domain iteration with a search that broadly speaking
does a wrapped walk of the scheduler domain sharing a last-level-cache.
While this had a number of improvements, one consequence is that two tasks
that share a waker/wakee relationship push each other around a socket. Even
though two tasks may be active, all cores are evenly used. This is great from
a search perspective and spreads a load across individual cores, but it has
adverse consequences for cpufreq. As each CPU has relatively low utilisation,
cpufreq may decide the utilisation is too low to used a higher P-state and
overall computation throughput suffers.
While individual cpufreq and cpuidle drivers may compensate by artifically
boosting P-state (at c0) or avoiding lower C-states (during idle), it does
not help if hardware-based cpufreq (e.g. HWP) is used.
This patch tracks a recently used CPU based on what CPU a task was running
on when it last was a waker a CPU it was recently using when a task is a
wakee. During SIS, the recently used CPU is used as a target if it's still
allowed by the task and is idle.
The benefit may be non-obvious so consider an example of two tasks
communicating back and forth. Task A may be an application doing IO where
task B is a kworker or kthread like journald. Task A may issue IO, wake
B and B wakes up A on completion. With the existing scheme this may look
like the following (potentially different IDs if SMT is in use but similar
principal applies).
A (cpu 0) wake B (wakes on cpu 1)
B (cpu 1) wake A (wakes on cpu 2)
A (cpu 2) wake B (wakes on cpu 3)
etc.
A careful reader may wonder why CPU 0 was not idle when B wakes A the
first time and it's simply due to the fact that A can be rescheduled to
another CPU and the pattern is that prev == target when B tries to wakeup A
and the information about CPU 0 has been lost.
With this patch, the pattern is more likely to be:
A (cpu 0) wake B (wakes on cpu 1)
B (cpu 1) wake A (wakes on cpu 0)
A (cpu 0) wake B (wakes on cpu 1)
etc
i.e. two communicating casts are more likely to use just two cores instead
of all available cores sharing a LLC.
The most dramatic speedup was noticed on dbench using the XFS filesystem on
UMA as clients interact heavily with workqueues in that configuration. Note
that a similar speedup is not observed on ext4 as the wakeup pattern
is different:
4.15.0-rc9 4.15.0-rc9
waprev-v1 biasancestor-v1
Hmean 1 287.54 ( 0.00%) 817.01 ( 184.14%)
Hmean 2 1268.12 ( 0.00%) 1781.24 ( 40.46%)
Hmean 4 1739.68 ( 0.00%) 1594.47 ( -8.35%)
Hmean 8 2464.12 ( 0.00%) 2479.56 ( 0.63%)
Hmean 64 1455.57 ( 0.00%) 1434.68 ( -1.44%)
The results can be less dramatic on NUMA where automatic balancing interferes
with the test. It's also known that network benchmarks running on localhost
also benefit quite a bit from this patch (roughly 10% on netperf RR for UDP
and TCP depending on the machine). Hackbench also seens small improvements
(6-11% depending on machine and thread count). The facebook schbench was also
tested but in most cases showed little or no different to wakeup latencies.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180130104555.4125-5-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
wake_affine_idle() prefers to move a task to the current CPU if the
wakeup is due to an interrupt. The expectation is that the interrupt
data is cache hot and relevant to the waking task as well as avoiding
a search. However, there is no way to determine if there was cache hot
data on the previous CPU that may exceed the interrupt data. Furthermore,
round-robin delivery of interrupts can migrate tasks around a socket where
each CPU is under-utilised. This can interact badly with cpufreq which
makes decisions based on per-cpu data. It has been observed on machines
with HWP that p-states are not boosted to their maximum levels even though
the workload is latency and throughput sensitive.
This patch uses the previous CPU for the task if it's idle and cache-affine
with the current CPU even if the current CPU is idle due to the wakup
being related to the interrupt. This reduces migrations at the cost of
the interrupt data not being cache hot when the task wakes.
A variety of workloads were tested on various machines and no adverse
impact was noticed that was outside noise. dbench on ext4 on UMA showed
roughly 10% reduction in the number of CPU migrations and it is a case
where interrupts are frequent for IO competions. In most cases, the
difference in performance is quite small but variability is often
reduced. For example, this is the result for pgbench running on a UMA
machine with different numbers of clients.
4.15.0-rc9 4.15.0-rc9
baseline waprev-v1
Hmean 1 22096.28 ( 0.00%) 22734.86 ( 2.89%)
Hmean 4 74633.42 ( 0.00%) 75496.77 ( 1.16%)
Hmean 7 115017.50 ( 0.00%) 113030.81 ( -1.73%)
Hmean 12 126209.63 ( 0.00%) 126613.40 ( 0.32%)
Hmean 16 131886.91 ( 0.00%) 130844.35 ( -0.79%)
Stddev 1 636.38 ( 0.00%) 417.11 ( 34.46%)
Stddev 4 614.64 ( 0.00%) 583.24 ( 5.11%)
Stddev 7 542.46 ( 0.00%) 435.45 ( 19.73%)
Stddev 12 173.93 ( 0.00%) 171.50 ( 1.40%)
Stddev 16 671.42 ( 0.00%) 680.30 ( -1.32%)
CoeffVar 1 2.88 ( 0.00%) 1.83 ( 36.26%)
Note that the different in performance is marginal but for low utilisation,
there is less variability.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180130104555.4125-4-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This is a preparation patch that has wake_affine*() return a CPU ID instead of
a boolean. The intent is to allow the wake_affine() helpers to be avoided
if a decision is already made. This patch has no functional change.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180130104555.4125-3-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
wake_affine_idle() takes parameters it never uses so clean it up.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180130104555.4125-2-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
rq->clock_task may be updated between the two calls of
rq_clock_task() in update_curr_rt(). Calling rq_clock_task() only
once makes it more accurate and efficient, taking update_curr() as
reference.
Signed-off-by: Wen Yang <wen.yang99@zte.com.cn>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: zhong.weidong@zte.com.cn
Link: http://lkml.kernel.org/r/1517800721-42092-1-git-send-email-wen.yang99@zte.com.cn
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When issuing an IPI RT push, where an IPI is sent to each CPU that has more
than one RT task scheduled on it, it references the root domain's rto_mask,
that contains all the CPUs within the root domain that has more than one RT
task in the runable state. The problem is, after the IPIs are initiated, the
rq->lock is released. This means that the root domain that is associated to
the run queue could be freed while the IPIs are going around.
Add a sched_get_rd() and a sched_put_rd() that will increment and decrement
the root domain's ref count respectively. This way when initiating the IPIs,
the scheduler will up the root domain's ref count before releasing the
rq->lock, ensuring that the root domain does not go away until the IPI round
is complete.
Reported-by: Pavan Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 4bdced5c9a ("sched/rt: Simplify the IPI based RT balancing logic")
Link: http://lkml.kernel.org/r/CAEU1=PkiHO35Dzna8EQqNSKW1fr1y1zRQ5y66X117MG06sQtNA@mail.gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
