Analogous to other important ECN information, export TCPI_OPT_ECN_LOW
in tcp_info tcpi_options field.
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Change-Id: I08d8d8c7e8780e6e37df54038ee50301ac5a0320
Signed-off-by: Juhyung Park <qkrwngud825@gmail.com>
The RCU read lock isn't necessary in list_lru_count_one() when the
condition that requires RCU (CONFIG_MEMCG && !CONFIG_SLOB) isn't met.
The highly-frequent RCU lock and unlock adds measurable overhead to the
shrink_slab() path when it isn't needed. As such, we can simply omit the
RCU read lock in this case to improve performance.
Signed-off-by: Sultan Alsawaf <sultan@kerneltoast.com>
Signed-off-by: priiii1808 <priyanshusinghal0818@gmail.com>
All callers of d_alloc_parallel() must make sure that resulting
in-lookup dentry (if any) will encounter __d_lookup_done() before
the final dput(). d_add_ci() might end up creating in-lookup
dentries; they are fed to d_splice_alias(), which will normally
make sure they meet __d_lookup_done(). However, it is possible
to end up with d_splice_alias() failing with ERR_PTR(-ELOOP)
without having done so. It takes a corrupted ntfs or case-insensitive
xfs image, but neither should end up with memory corruption...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
I have a RT task X at a high priority and cyclictest on each CPU with
lower priority than X's. If X is active and each CPU wakes their own
cylictest thread then it ends in a longer rto_push storm.
A random CPU determines via balance_rt() that the CPU on which X is
running needs to push tasks. X has the highest priority, cyclictest is
next in line so there is nothing that can be done since the task with
the higher priority is not touched.
tell_cpu_to_push() increments rto_loop_next and schedules
rto_push_irq_work_func() on X's CPU. The other CPUs also increment the
loop counter and do the same. Once rto_push_irq_work_func() is active it
does nothing because it has _no_ pushable tasks on its runqueue. Then
checks rto_next_cpu() and decides to queue irq_work on the local CPU
because another CPU requested a push by incrementing the counter.
I have traces where ~30 CPUs request this ~3 times each before it
finally ends. This greatly increases X's runtime while X isn't making
much progress.
Teach rto_next_cpu() to only return CPUs which also have tasks on their
runqueue which can be pushed away. This does not reduce the
tell_cpu_to_push() invocations (rto_loop_next counter increments) but
reduces the amount of issued rto_push_irq_work_func() if nothing can be
done. As the result the overloaded CPU is blocked less often.
There are still cases where the "same job" is repeated several times
(for instance the current CPU needs to resched but didn't yet because
the irq-work is repeated a few times and so the old task remains on the
CPU) but the majority of request end in tell_cpu_to_push() before an IPI
is issued.
Reviewed-by: "Steven Rostedt (Google)" <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20230801152648._y603AS_@linutronix.de
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Commit "rcu: Create RCU-specific workqueues with rescuers" switched RCU
to using local workqueses and removed power efficiency flag from them.
This caused a performance regression that can be observed in Geekbench 5
after enabling CONFIG_WQ_POWER_EFFICIENT_DEFAULT: score went down from
760/2500 to 620/2300 (single/multi core respectively).
Add WQ_POWER_EFFICIENT flag to avoid this regression.
Change-Id: I2c4f41faa55548f9e81a1c0cbe10703948062d89
When a binder reference is cleaned up, any freeze work queued in the
associated process should also be removed. Otherwise, the reference is
freed while its ref->freeze.work is still queued in proc->work leading
to a use-after-free issue as shown by the following KASAN report:
==================================================================
BUG: KASAN: slab-use-after-free in binder_release_work+0x398/0x3d0
Read of size 8 at addr ffff31600ee91488 by task kworker/5:1/211
CPU: 5 UID: 0 PID: 211 Comm: kworker/5:1 Not tainted 6.11.0-rc7-00382-gfc6c92196396 #22
Hardware name: linux,dummy-virt (DT)
Workqueue: events binder_deferred_func
Call trace:
binder_release_work+0x398/0x3d0
binder_deferred_func+0xb60/0x109c
process_one_work+0x51c/0xbd4
worker_thread+0x608/0xee8
Allocated by task 703:
__kmalloc_cache_noprof+0x130/0x280
binder_thread_write+0xdb4/0x42a0
binder_ioctl+0x18f0/0x25ac
__arm64_sys_ioctl+0x124/0x190
invoke_syscall+0x6c/0x254
Freed by task 211:
kfree+0xc4/0x230
binder_deferred_func+0xae8/0x109c
process_one_work+0x51c/0xbd4
worker_thread+0x608/0xee8
==================================================================
This commit fixes the issue by ensuring any queued freeze work is removed
when cleaning up a binder reference.
Fixes: d579b04a52a1 ("binder: frozen notification")
Cc: stable@vger.kernel.org
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Bug: 366003708
Link: https://lore.kernel.org/all/20240924184401.76043-4-cmllamas@google.com/
Change-Id: Icc40e7dd6157981f4adbea7243e55be118552321
[cmllamas: drop BINDER_STAT_FREEZE as it's not supported here]
Signed-off-by: Carlos Llamas <cmllamas@google.com>
If a freeze notification is cleared with BC_CLEAR_FREEZE_NOTIFICATION
before calling binder_freeze_notification_done(), then it is detached
from its reference (e.g. ref->freeze) but the work remains queued in
proc->delivered_freeze. This leads to a memory leak when the process
exits as any pending entries in proc->delivered_freeze are not freed:
unreferenced object 0xffff38e8cfa36180 (size 64):
comm "binder-util", pid 655, jiffies 4294936641
hex dump (first 32 bytes):
b8 e9 9e c8 e8 38 ff ff b8 e9 9e c8 e8 38 ff ff .....8.......8..
0b 00 00 00 00 00 00 00 3c 1f 4b 00 00 00 00 00 ........<.K.....
backtrace (crc 95983b32):
[<000000000d0582cf>] kmemleak_alloc+0x34/0x40
[<000000009c99a513>] __kmalloc_cache_noprof+0x208/0x280
[<00000000313b1704>] binder_thread_write+0xdec/0x439c
[<000000000cbd33bb>] binder_ioctl+0x1b68/0x22cc
[<000000002bbedeeb>] __arm64_sys_ioctl+0x124/0x190
[<00000000b439adee>] invoke_syscall+0x6c/0x254
[<00000000173558fc>] el0_svc_common.constprop.0+0xac/0x230
[<0000000084f72311>] do_el0_svc+0x40/0x58
[<000000008b872457>] el0_svc+0x38/0x78
[<00000000ee778653>] el0t_64_sync_handler+0x120/0x12c
[<00000000a8ec61bf>] el0t_64_sync+0x190/0x194
This patch fixes the leak by ensuring that any pending entries in
proc->delivered_freeze are freed during binder_deferred_release().
Fixes: d579b04a52a1 ("binder: frozen notification")
Cc: stable@vger.kernel.org
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Acked-by: Todd Kjos <tkjos@google.com>
Link: https://lore.kernel.org/r/20240926233632.821189-8-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bug: 366003708
(cherry picked from commit 1db76ec2b4b206ff943e292a0b55e68ff3443598
git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc.git
char-misc-next)
Change-Id: Iafdec3421c521b4b591b94455deba7ee5102c8ca
[cmllamas: drop BINDER_STAT_FREEZE and use binder_proc_ext_entry()]
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Add the pending proc->delivered_freeze work to the debugfs output. This
information was omitted in the original implementation of the freeze
notification and can be valuable for debugging issues.
Fixes: d579b04a52a1 ("binder: frozen notification")
Cc: stable@vger.kernel.org
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Acked-by: Todd Kjos <tkjos@google.com>
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240926233632.821189-9-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bug: 366003708
(cherry picked from commit cb2aeb2ec25884133110ffe5a67ff3cf7dee5ceb
git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc.git
char-misc-next)
Change-Id: Ifc9a22b52e38c35af661732486fa1f154adb34de
[cmllamas: fix KMI break with binder_proc_ext_entry()]
Signed-off-by: Carlos Llamas <cmllamas@google.com>
For us, it's most helpful to have the round-robin timeslice as low as is
allowed by the scheduler to reduce latency. Since it's limited by the
scheduler tick rate, just set the default to 1 jiffy, which is the
lowest possible value.
Signed-off-by: Sultan Alsawaf <sultan@kerneltoast.com>
Change-Id: I6c9f6bb5bbadf363efb719d3e30b0b073654d688
During load balance, we try at most env->loop_max time to move a task.
But it can happen that the loop_max LRU tasks (ie tail of
the cfs_tasks list) can't be moved to dst_cpu because of affinity.
In this case, loop in the list until we found at least one.
The maximum of detached tasks remained the same as before.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220825122726.20819-2-vincent.guittot@linaro.org
This is a forward port of pa/890483 with modifications from the original
patch due to changes in sched/softirq.c which applies the same logic.
We're finding audio glitches caused by audio-producing RT tasks
that are either interrupted to handle softirq's or that are
scheduled onto cpu's that are handling softirq's.
In a previous patch, we attempted to catch many cases of the
latter problem, but it's clear that we are still losing
significant numbers of races in some apps.
This patch attempts to address the following problem::
It attempts to reduce the most common windows in which
we lose the race between scheduling an RT task on a remote
core and starting to handle softirq's on that core.
We still lose some races, but we lose significantly fewer.
(And we don't want to introduce any heavyweight forms
of synchronization on these paths.)
Bug: 64912585
Bug: 136771796
Bug: 144961676
Change-Id: Ida89a903be0f1965552dd0e84e67ef1d3158c7d8
Signed-off-by: Miguel de Dios <migueldedios@google.com>
Signed-off-by: Jimmy Shiu <jimmyshiu@google.com>
(cherry picked from commit 3c2569f21be6b7c457b389821b098d30bd03b84c)
(cherry picked from commit f9cf84966a315169f3f505993b9e15b6c53f2bbb)
(cherry picked from commit e6edaa29efee9e2f2451ab50030d8cde19ab227b)
(cherry picked from commit db52e23658eaf304ba2333f04d910313c0c21e05)
(cherry picked from commit aa2d4d07929d1c06c6b99ed8718394cf8e471d21)
(cherry picked from commit cefdb96edccdc3f788275d1bcef89c8dfeeb8b11)
(cherry picked from commit 5a45a0208069ac5704713b35a0858fcd5997f97a)
(cherry picked from commit 86ac137eb7399547c6cd20d60063d244091e920f)
When a new threshold breaching stall happens after a psi event was
generated and within the window duration, the new event is not
generated because the events are rate-limited to one per window. If
after that no new stall is recorded then the event will not be
generated even after rate-limiting duration has passed. This is
happening because with no new stall, window_update will not be called
even though threshold was previously breached. To fix this, record
threshold breaching occurrence and generate the event once window
duration is passed.
Suggested-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Suren Baghdasaryan <surenb@google.com>
Link: https://lore.kernel.org/r/1643093818-19835-1-git-send-email-huangzhaoyang@gmail.com
(cherry picked from commit e6df4ead85d9da1b07dd40bd4c6d2182f3e210c4)
(cherry picked from commit 007db79713251fd67cc5ce7142942772f6074aca)
(cherry picked from commit 129c4324ce9ea8be9c528629bca72e1e1f80d3fb)
When a trigger being created, its win.start_value and win.start_time are
reset to zero. If group->total[PSI_POLL][t->state] has accumulated before,
this trigger will be fired unexpectedly in the next period, even if its
growth time does not reach its threshold.
So set the window of the new trigger to the current state value.
Signed-off-by: Hailong Liu <liuhailong@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Suren Baghdasaryan <surenb@google.com>
Link: https://lore.kernel.org/r/1648789811-3788971-1-git-send-email-liuhailong@linux.alibaba.com
(cherry picked from commit 915a087e4c47334a2f7ba2a4092c4bade0873769)
(cherry picked from commit 168af79cd0d6750ddf6c775c79833059da606c34)
(cherry picked from commit 9b3b0d7ba99e41a1159a0b9c58ff138bb51d52a1)
Martin find it confusing when look at the /proc/pressure/cpu output,
and found no hint about that CPU "full" line in psi Documentation.
% cat /proc/pressure/cpu
some avg10=0.92 avg60=0.91 avg300=0.73 total=933490489
full avg10=0.22 avg60=0.23 avg300=0.16 total=358783277
The PSI_CPU_FULL state is introduced by commit e7fcd7622823
("psi: Add PSI_CPU_FULL state"), which mainly for cgroup level,
but also counted at the system level as a side effect.
Naturally, the FULL state doesn't exist for the CPU resource at
the system level. These "full" numbers can come from CPU idle
schedule latency. For example, t1 is the time when task wakeup
on an idle CPU, t2 is the time when CPU pick and switch to it.
The delta of (t2 - t1) will be in CPU_FULL state.
Another case all processes can be stalled is when all cgroups
have been throttled at the same time, which unlikely to happen.
Anyway, CPU_FULL metric is meaningless and confusing at the
system level. So this patch will report zeroes for CPU full
at the system level, and update psi Documentation accordingly.
Fixes: e7fcd7622823 ("psi: Add PSI_CPU_FULL state")
Reported-by: Martin Steigerwald <Martin.Steigerwald@proact.de>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lore.kernel.org/r/20220408121914.82855-1-zhouchengming@bytedance.com
(cherry picked from commit 890d550d7dbac7a31ecaa78732aa22be282bb6b8)
(cherry picked from commit f5187de2b75d019739caec97e8e7886a27e8554c)
(cherry picked from commit 669718aaede6df28e37f6a11c0e257d18f050b1a)
This patch addresses an issue seen where SCHED_FIFO prio 99
tasks were being woken up on a cpu where a long-running softirq
was executing, and the RT task was not being migrated, causing
long (10+ms wakeup latencies).
Prior to upstream commit 934fc3314b39 ("sched/cpupri: Remap
CPUPRI_NORMAL to MAX_RT_PRIO-1") the task->prio -> cpupri
mapping is a little ackward.
For RT tasks, its calculated as:
cpupri = MAX_RT_PRIO - prio + 1;
See:
https://android.googlesource.com/kernel/common/+/refs/heads/android13-5.10/kernel/sched/cpupri.c#39
This is added ontop of the also ackward detail that the
task->prio is inverted (RT prio99 -> 0), means the cpupri
mapping for RT tasks goes from 2->101. This makes it easy to
evaluate the cpupri incorrectly.
Which it turns out happened In commit 3adfd8e344ac ("ANDROID:
sched: avoid placing RT threads on cores handling softirqs"):
3adfd8e344%5E%21/
With the lines:
int task_pri = convert_prio(p->prio);
bool drop_nopreempts = task_pri <= MAX_RT_PRIO;
Where the added logic to decide to migrate a rt task off of a
cpu depended on this drop_nopreempts being true.
This works properly for rt tasks from prio 99 to 1, but for the
case of task->prio == 0 (userland rt prio 99 tasks) it breaks,
as the cpupri will be MAX_RT_PRIO - 0 + 1, which then gets
checked as <= MAX_RT_PRIO.
This prevents the cpu from being dropped from the scheduling
set and prevents the rt user prio 99 task from migrating, which
results in high priority rt tasks being left on cpus where long
running softirqs are executing, causing long latencies.
This patch fixes the off by one by changing the evaulation
to MAX_RT_PRIO + 1, so that user-prio 99 tasks will also be
migrated if appropriate.
Luckilly this odd cpupri mapping has been fixed upstream, making
this patch no longer necessary in 5.15 and newer kernels.
Fixes: 3adfd8e344ac ("ANDROID: sched: avoid placing RT threads on cores handling softirqs")
Signed-off-by: John Stultz <jstultz@google.com>
Change-Id: Ia2db7cd461eb4c90f5850b791de1ae95582f7530
(cherry picked from commit bec3b2f002cd13d44e32710315de91e636ecef84)
(cherry picked from commit e8e441d4a2f6b6a316d229f8ff1ebe0f40099817)
(cherry picked from commit 1c1f38f1e1fa82e09e0e3f6f941ca81127682ec3)
Use the task_current() function where appropriate.
No functional change.
Signed-off-by: Hui Su <sh_def@163.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Link: https://lkml.kernel.org/r/20201030173223.GA52339@rlk
While doing newidle load balancing, it is possible for new tasks to
arrive, such as with pending wakeups. newidle_balance() already accounts
for this by exiting the sched_domain load_balance() iteration if it
detects these cases. This is very important for minimizing wakeup
latency.
However, if we are already in load_balance(), we may stay there for a
while before returning back to newidle_balance(). This is most
exacerbated if we enter a 'goto redo' loop in the LBF_ALL_PINNED case. A
very straightforward workaround to this is to adjust should_we_balance()
to bail out if we're doing a CPU_NEWLY_IDLE balance and new tasks are
detected.
This was tested with the following reproduction:
- two threads that take turns sleeping and waking each other up are
affined to two cores
- a large number of threads with 100% utilization are pinned to all
other cores
Without this patch, wakeup latency was ~120us for the pair of threads,
almost entirely spent in load_balance(). With this patch, wakeup latency
is ~6us.
Signed-off-by: Josh Don <joshdon@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220609025515.2086253-1-joshdon@google.com
Similarly to util_avg and util_sum, don't sync load_sum with the low
bound of load_avg but only ensure that load_sum stays in the correct range.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Link: https://lkml.kernel.org/r/20220111134659.24961-5-vincent.guittot@linaro.org
Similarly to util_avg and util_sum, don't sync runnable_sum with the low
bound of runnable_avg but only ensure that runnable_sum stays in the
correct range.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Link: https://lkml.kernel.org/r/20220111134659.24961-4-vincent.guittot@linaro.org
Rick reported performance regressions in bugzilla because of cpu frequency
being lower than before:
https://bugzilla.kernel.org/show_bug.cgi?id=215045
He bisected the problem to:
commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consistent")
This commit forces util_sum to be synced with the new util_avg after
removing the contribution of a task and before the next periodic sync. By
doing so util_sum is rounded to its lower bound and might lost up to
LOAD_AVG_MAX-1 of accumulated contribution which has not yet been
reflected in util_avg.
update_tg_cfs_util() is not the only place where we round util_sum and
lost some accumulated contributions that are not already reflected in
util_avg. Modify update_tg_cfs_util() and detach_entity_load_avg() to not
sync util_sum with the new util_avg. Instead of always setting util_sum to
the low bound of util_avg, which can significantly lower the utilization,
we propagate the difference. In addition, we also check that cfs's util_sum
always stays above the lower bound for a given util_avg as it has been
observed that sched_entity's util_sum is sometimes above cfs one.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Link: https://lkml.kernel.org/r/20220111134659.24961-3-vincent.guittot@linaro.org
Kevin is reporting crashes which point to a use-after-free of a cfs_rq
in update_blocked_averages(). Initial debugging revealed that we've
live cfs_rq's (on_list=1) in an about to be kfree()'d task group in
free_fair_sched_group(). However, it was unclear how that can happen.
His kernel config happened to lead to a layout of struct sched_entity
that put the 'my_q' member directly into the middle of the object
which makes it incidentally overlap with SLUB's freelist pointer.
That, in combination with SLAB_FREELIST_HARDENED's freelist pointer
mangling, leads to a reliable access violation in form of a #GP which
made the UAF fail fast.
Michal seems to have run into the same issue[1]. He already correctly
diagnosed that commit a7b359fc6a37 ("sched/fair: Correctly insert
cfs_rq's to list on unthrottle") is causing the preconditions for the
UAF to happen by re-adding cfs_rq's also to task groups that have no
more running tasks, i.e. also to dead ones. His analysis, however,
misses the real root cause and it cannot be seen from the crash
backtrace only, as the real offender is tg_unthrottle_up() getting
called via sched_cfs_period_timer() via the timer interrupt at an
inconvenient time.
When unregister_fair_sched_group() unlinks all cfs_rq's from the dying
task group, it doesn't protect itself from getting interrupted. If the
timer interrupt triggers while we iterate over all CPUs or after
unregister_fair_sched_group() has finished but prior to unlinking the
task group, sched_cfs_period_timer() will execute and walk the list of
task groups, trying to unthrottle cfs_rq's, i.e. re-add them to the
dying task group. These will later -- in free_fair_sched_group() -- be
kfree()'ed while still being linked, leading to the fireworks Kevin
and Michal are seeing.
To fix this race, ensure the dying task group gets unlinked first.
However, simply switching the order of unregistering and unlinking the
task group isn't sufficient, as concurrent RCU walkers might still see
it, as can be seen below:
CPU1: CPU2:
: timer IRQ:
: do_sched_cfs_period_timer():
: :
: distribute_cfs_runtime():
: rcu_read_lock();
: :
: unthrottle_cfs_rq():
sched_offline_group(): :
: walk_tg_tree_from(…,tg_unthrottle_up,…):
list_del_rcu(&tg->list); :
(1) : list_for_each_entry_rcu(child, &parent->children, siblings)
: :
(2) list_del_rcu(&tg->siblings); :
: tg_unthrottle_up():
unregister_fair_sched_group(): struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)];
: :
list_del_leaf_cfs_rq(tg->cfs_rq[cpu]); :
: :
: if (!cfs_rq_is_decayed(cfs_rq) || cfs_rq->nr_running)
(3) : list_add_leaf_cfs_rq(cfs_rq);
: :
: :
: :
: :
: :
(4) : rcu_read_unlock();
CPU 2 walks the task group list in parallel to sched_offline_group(),
specifically, it'll read the soon to be unlinked task group entry at
(1). Unlinking it on CPU 1 at (2) therefore won't prevent CPU 2 from
still passing it on to tg_unthrottle_up(). CPU 1 now tries to unlink
all cfs_rq's via list_del_leaf_cfs_rq() in
unregister_fair_sched_group(). Meanwhile CPU 2 will re-add some of
these at (3), which is the cause of the UAF later on.
To prevent this additional race from happening, we need to wait until
walk_tg_tree_from() has finished traversing the task groups, i.e.
after the RCU read critical section ends in (4). Afterwards we're safe
to call unregister_fair_sched_group(), as each new walk won't see the
dying task group any more.
On top of that, we need to wait yet another RCU grace period after
unregister_fair_sched_group() to ensure print_cfs_stats(), which might
run concurrently, always sees valid objects, i.e. not already free'd
ones.
This patch survives Michal's reproducer[2] for 8h+ now, which used to
trigger within minutes before.
[1] https://lore.kernel.org/lkml/20211011172236.11223-1-mkoutny@suse.com/
[2] https://lore.kernel.org/lkml/20211102160228.GA57072@blackbody.suse.cz/
Fixes: a7b359fc6a37 ("sched/fair: Correctly insert cfs_rq's to list on unthrottle")
[peterz: shuffle code around a bit]
Reported-by: Kevin Tanguy <kevin.tanguy@corp.ovh.com>
Signed-off-by: Mathias Krause <minipli@grsecurity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
With a default value of 500us, sysctl_sched_migration_cost is
significanlty higher than the cost of load_balance. Remove the
condition and rely on the sd->max_newidle_lb_cost to abort
newidle_balance.
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>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20211019123537.17146-5-vincent.guittot@linaro.org
Decay max_newidle_lb_cost only when it has not been updated for a while
and ensure to not decay a recently changed value.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20211019123537.17146-4-vincent.guittot@linaro.org
In newidle_balance(), the scheduler skips load balance to the new idle cpu
when the 1st sd of this_rq is:
this_rq->avg_idle < sd->max_newidle_lb_cost
Doing a costly call to update_blocked_averages() will not be useful and
simply adds overhead when this condition is true.
Check the condition early in newidle_balance() to skip
update_blocked_averages() when possible.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20211019123537.17146-3-vincent.guittot@linaro.org
The time spent to update the blocked load can be significant depending of
the complexity fo the cgroup hierarchy. Take this time into account in
the cost of the 1st load balance of a newly idle cpu.
Also reduce the number of call to sched_clock_cpu() and track more actual
work.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20211019123537.17146-2-vincent.guittot@linaro.org
commit 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
reported some inconsitencies between *_avg and *_sum.
commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consistent")
fixed some but one remains when dequeuing load.
sync the cfs's load_sum with its load_avg after dequeuing the load of a
sched_entity.
Fixes: 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Odin Ugedal <odin@uged.al>
Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Link: https://lore.kernel.org/r/20210701171837.32156-1-vincent.guittot@linaro.org
Ensure that a CFS parent will be in the list whenever one of its children is also
in the list.
A warning on rq->tmp_alone_branch != &rq->leaf_cfs_rq_list has been
reported while running LTP test cfs_bandwidth01.
Odin Ugedal found the root cause:
$ tree /sys/fs/cgroup/ltp/ -d --charset=ascii
/sys/fs/cgroup/ltp/
|-- drain
`-- test-6851
`-- level2
|-- level3a
| |-- worker1
| `-- worker2
`-- level3b
`-- worker3
Timeline (ish):
- worker3 gets throttled
- level3b is decayed, since it has no more load
- level2 get throttled
- worker3 get unthrottled
- level2 get unthrottled
- worker3 is added to list
- level3b is not added to list, since nr_running==0 and is decayed
[ Vincent Guittot: Rebased and updated to fix for the reported warning. ]
Fixes: a7b359fc6a37 ("sched/fair: Correctly insert cfs_rq's to list on unthrottle")
Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Rik van Riel <riel@surriel.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Acked-by: Odin Ugedal <odin@uged.al>
Link: https://lore.kernel.org/r/20210621174330.11258-1-vincent.guittot@linaro.org
In case the _avg delta is 0 there is no need to update se's _avg
(level n) nor cfs_rq's _avg (level n-1). These values stay the same.
Since cfs_rq's _avg isn't changed, i.e. no load is propagated down,
cfs_rq's _sum should stay the same as well.
So bail out after se's _sum has been updated.
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20210601083616.804229-1-dietmar.eggemann@arm.com
Fix an issue where fairness is decreased since cfs_rq's can end up not
being decayed properly. For two sibling control groups with the same
priority, this can often lead to a load ratio of 99/1 (!!).
This happens because when a cfs_rq is throttled, all the descendant
cfs_rq's will be removed from the leaf list. When they initial cfs_rq
is unthrottled, it will currently only re add descendant cfs_rq's if
they have one or more entities enqueued. This is not a perfect
heuristic.
Instead, we insert all cfs_rq's that contain one or more enqueued
entities, or it its load is not completely decayed.
Can often lead to situations like this for equally weighted control
groups:
$ ps u -C stress
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 10009 88.8 0.0 3676 100 pts/1 R+ 11:04 0:13 stress --cpu 1
root 10023 3.0 0.0 3676 104 pts/1 R+ 11:04 0:00 stress --cpu 1
Fixes: 31bc6aeaab1d ("sched/fair: Optimize update_blocked_averages()")
[vingo: !SMP build fix]
Signed-off-by: Odin Ugedal <odin@uged.al>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20210612112815.61678-1-odin@uged.al
Rounding in PELT calculation happening when entities are attached/detached
of a cfs_rq can result into situations where util/runnable_avg is not null
but util/runnable_sum is. This is normally not possible so we need to
ensure that util/runnable_sum stays synced with util/runnable_avg.
detach_entity_load_avg() is the last place where we don't sync
util/runnable_sum with util/runnbale_avg when moving some sched_entities
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20210601085832.12626-1-vincent.guittot@linaro.org
when removing a cfs_rq from the list we only check _sum value so we must
ensure that _avg and _sum stay synced so load_sum can't be null whereas
load_avg is not after propagating load in the cgroup hierarchy.
Use load_avg to compute load_sum similarly to what is done for util_sum
and runnable_sum.
Fixes: 0e2d2aaaae52 ("sched/fair: Rewrite PELT migration propagation")
Reported-by: Odin Ugedal <odin@uged.al>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Odin Ugedal <odin@uged.al>
Link: https://lkml.kernel.org/r/20210527122916.27683-2-vincent.guittot@linaro.org
The try_to_wake_up function has an optimization where it can queue
a task for wakeup on its previous CPU, if the task is still in the
middle of going to sleep inside schedule().
Once schedule() re-enables IRQs, the task will be woken up with an
IPI, and placed back on the runqueue.
If we have such a wakeup pending, there is no need to search other
CPUs for runnable tasks. Just skip (or bail out early from) newidle
balancing, and run the just woken up task.
For a memcache like workload test, this reduces total CPU use by
about 2%, proportionally split between user and system time,
and p99 and p95 application response time by 10% on average.
The schedstats run_delay number shows a similar improvement.
Signed-off-by: Rik van Riel <riel@surriel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lkml.kernel.org/r/20210422130236.0bb353df@imladris.surriel.com
During load-balance, groups classified as group_misfit_task are filtered
out if they do not pass
group_smaller_max_cpu_capacity(<candidate group>, <local group>);
which itself employs fits_capacity() to compare the sgc->max_capacity of
both groups.
Due to the underlying margin, fits_capacity(X, 1024) will return false for
any X > 819. Tough luck, the capacity_orig's on e.g. the Pixel 4 are
{261, 871, 1024}. If a CPU-bound task ends up on one of those "medium"
CPUs, misfit migration will never intentionally upmigrate it to a CPU of
higher capacity due to the aforementioned margin.
One may argue the 20% margin of fits_capacity() is excessive in the advent
of counter-enhanced load tracking (APERF/MPERF, AMUs), but one point here
is that fits_capacity() is meant to compare a utilization value to a
capacity value, whereas here it is being used to compare two capacity
values. As CPU capacity and task utilization have different dynamics, a
sensible approach here would be to add a new helper dedicated to comparing
CPU capacities.
Also note that comparing capacity extrema of local and source sched_group's
doesn't make much sense when at the day of the day the imbalance will be
pulled by a known env->dst_cpu, whose capacity can be anywhere within the
local group's capacity extrema.
While at it, replace group_smaller_{min, max}_cpu_capacity() with
comparisons of the source group's min/max capacity and the destination
CPU's capacity.
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Qais Yousef <qais.yousef@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
Link: https://lkml.kernel.org/r/20210407220628.3798191-4-valentin.schneider@arm.com
When triggering an active load balance, sd->nr_balance_failed is set to
such a value that any further can_migrate_task() using said sd will ignore
the output of task_hot().
This behaviour makes sense, as active load balance intentionally preempts a
rq's running task to migrate it right away, but this asynchronous write is
a bit shoddy, as the stopper thread might run active_load_balance_cpu_stop
before the sd->nr_balance_failed write either becomes visible to the
stopper's CPU or even happens on the CPU that appended the stopper work.
Add a struct lb_env flag to denote active balancing, and use it in
can_migrate_task(). Remove the sd->nr_balance_failed write that served the
same purpose. Cleanup the LBF_DST_PINNED active balance special case.
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20210407220628.3798191-3-valentin.schneider@arm.com
A long-tail load balance cost is observed on the newly idle path,
this is caused by a race window between the first nr_running check
of the busiest runqueue and its nr_running recheck in detach_tasks.
Before the busiest runqueue is locked, the tasks on the busiest
runqueue could be pulled by other CPUs and nr_running of the busiest
runqueu becomes 1 or even 0 if the running task becomes idle, this
causes detach_tasks breaks with LBF_ALL_PINNED flag set, and triggers
load_balance redo at the same sched_domain level.
In order to find the new busiest sched_group and CPU, load balance will
recompute and update the various load statistics, which eventually leads
to the long-tail load balance cost.
This patch clears LBF_ALL_PINNED flag for this race condition, and hence
reduces the long-tail cost of newly idle balance.
Signed-off-by: Aubrey Li <aubrey.li@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/1614154549-116078-1-git-send-email-aubrey.li@intel.com
update_idle_core() is only done for the case of sched_smt_present.
but test_idle_cores() is done for all machines even those without
SMT.
This can contribute to up 8%+ hackbench performance loss on a
machine like kunpeng 920 which has no SMT. This patch removes the
redundant test_idle_cores() for !SMT machines.
Hackbench is ran with -g {2..14}, for each g it is ran 10 times to get
an average.
$ numactl -N 0 hackbench -p -T -l 20000 -g $1
The below is the result of hackbench w/ and w/o this patch:
g= 2 4 6 8 10 12 14
w/o: 1.8151 3.8499 5.5142 7.2491 9.0340 10.7345 12.0929
w/ : 1.8428 3.7436 5.4501 6.9522 8.2882 9.9535 11.3367
+4.1% +8.3% +7.3% +6.3%
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lkml.kernel.org/r/20210320221432.924-1-song.bao.hua@hisilicon.com
Reorder the tests and skip useless ones when no load balance has been
performed and rq lock has not been released.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Link: https://lkml.kernel.org/r/20210224133007.28644-6-vincent.guittot@linaro.org
