When comparing the current struct sched_group with the yet-busiest
domain in update_sd_pick_busiest(), if the two groups have the same
group type, we're currently doing a bit of unnecessary work for any
group >= group_misfit_task. We're comparing the two groups, and then
returning only if false (the group in question is not the busiest).
Otherwise, we break out, do an extra unnecessary conditional check that's
vacuously false for any group type > group_fully_busy, and then always
return true.
Let's just return directly in the switch statement instead. This doesn't
change the size of vmlinux with llvm 17 (not surprising given that all
of this is inlined in load_balance()), but it does shrink load_balance()
by 88 bytes on x86. Given that it also improves readability, this seems
worth doing.
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Link: https://lore.kernel.org/r/20240206043921.850302-4-void@manifault.com
In update_sd_pick_busiest(), when comparing two sched groups that are
both of type group_misfit_task, we currently consider the new group as
busier than the current busiest group even if the new group has the
same misfit task load as the current busiest group. We can avoid some
unnecessary writes if we instead only consider the newest group to be
the busiest if it has a higher load than the current busiest. This
matches the behavior of other group types where we compare load, such as
two groups that are both overloaded.
Let's update the group_misfit_task type comparison to also only update
the busiest group in the event of strict inequality.
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Link: https://lore.kernel.org/r/20240206043921.850302-3-void@manifault.com
In update_sd_lb_stats(), when we're iterating over the sched groups that
comprise a sched domain, we're skipping the call to
update_sd_pick_busiest() for the sched group that contains the local /
destination CPU. We use a goto to skip the call, but we could just as
easily check !local_group, as there's no other logic that we need to
skip with the goto. Let's remove the goto, and check for !local_group in
the if statement instead.
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Link: https://lore.kernel.org/r/20240206043921.850302-2-void@manifault.com
When picking a CPU on task wakeup, select_idle_core() has to take
into account the scheduling domain where the function looks for the CPU.
This is because the "isolcpus" kernel command line option can remove CPUs
from the domain to isolate them from other SMT siblings.
This change replaces the set of CPUs allowed to run the task from
p->cpus_ptr by the intersection of p->cpus_ptr and sched_domain_span(sd)
which is stored in the 'cpus' argument provided by select_idle_cpu().
Fixes: 9fe1f127b9 ("sched/fair: Merge select_idle_core/cpu()")
Signed-off-by: Keisuke Nishimura <keisuke.nishimura@inria.fr>
Signed-off-by: Julia Lawall <julia.lawall@inria.fr>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20240110131707.437301-2-keisuke.nishimura@inria.fr
When picking a CPU on task wakeup, select_idle_smt() has to take
into account the scheduling domain of @target. This is because the
"isolcpus" kernel command line option can remove CPUs from the domain to
isolate them from other SMT siblings.
This fix checks if the candidate CPU is in the target scheduling domain.
Commit:
df3cb4ea1f ("sched/fair: Fix wrong cpu selecting from isolated domain")
... originally introduced this fix by adding the check of the scheduling
domain in the loop.
However, commit:
3e6efe87cd ("sched/fair: Remove redundant check in select_idle_smt()")
... accidentally removed the check. Bring it back.
Fixes: 3e6efe87cd ("sched/fair: Remove redundant check in select_idle_smt()")
Signed-off-by: Keisuke Nishimura <keisuke.nishimura@inria.fr>
Signed-off-by: Julia Lawall <julia.lawall@inria.fr>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20240110131707.437301-1-keisuke.nishimura@inria.fr
There are helper functions called cpu_util_dl() and cpu_util_rt() which give
the average utilization of DL and RT respectively. But there are a few
places in code where access to these variables is open-coded.
Instead use the helper function so that code becomes simpler and easier to
maintain later on.
No functional changes intended.
Signed-off-by: Shrikanth Hegde <sshegde@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20240101154624.100981-2-sshegde@linux.vnet.ibm.com
The timekeeping duty is handed over from the outgoing CPU on stop
machine, then the oneshot tick is stopped right after. Therefore it's
guaranteed that the current CPU isn't the timekeeper upon its last call
to idle.
Besides, calling tick_nohz_idle_stop_tick() while the dying CPU goes
into idle suggests that the tick is going to be stopped while it is
actually stopped already from the appropriate CPU hotplug state.
Remove the confusing call and the obsolete case handling and convert it
to a sanity check that verifies the above assumption.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240225225508.11587-16-frederic@kernel.org
On some systems, sys_membarrier can be very expensive, causing overall
slowdowns for everything. So put a lock on the path in order to
serialize the accesses to prevent the ability for this to be called at
too high of a frequency and saturate the machine.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-and-tested-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Fixes: 22e4ebb975 ("membarrier: Provide expedited private command")
Fixes: c5f58bd58f ("membarrier: Provide GLOBAL_EXPEDITED command")
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
RISC-V uses xRET instructions on return from interrupt and to go back
to user-space; the xRET instruction is not core serializing.
Use FENCE.I for providing core serialization as follows:
- by calling sync_core_before_usermode() on return from interrupt (cf.
ipi_sync_core()),
- via switch_mm() and sync_core_before_usermode() (respectively, for
uthread->uthread and kthread->uthread transitions) before returning
to user-space.
On RISC-V, the serialization in switch_mm() is activated by resetting
the icache_stale_mask of the mm at prepare_sync_core_cmd().
Suggested-by: Palmer Dabbelt <palmer@dabbelt.com>
Signed-off-by: Andrea Parri <parri.andrea@gmail.com>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/r/20240131144936.29190-5-parri.andrea@gmail.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
The membarrier system call requires a full memory barrier after storing
to rq->curr, before going back to user-space. The barrier is only
needed when switching between processes: the barrier is implied by
mmdrop() when switching from kernel to userspace, and it's not needed
when switching from userspace to kernel.
Rely on the feature/mechanism ARCH_HAS_MEMBARRIER_CALLBACKS and on the
primitive membarrier_arch_switch_mm(), already adopted by the PowerPC
architecture, to insert the required barrier.
Fixes: fab957c11e ("RISC-V: Atomic and Locking Code")
Signed-off-by: Andrea Parri <parri.andrea@gmail.com>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/r/20240131144936.29190-2-parri.andrea@gmail.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Mark the task as having a cached timestamp when set assign it, so we
can efficiently check if it needs updating post being scheduled back in.
This covers both the actual schedule out case, which would've flushed
the plug, and the preemption case which doesn't touch the plugged
requests (for many reasons, one of them being then we'd need to have
preemption disabled around plug state manipulation).
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull scheduler fix from Ingo Molnar:
"Fix a cpufreq related performance regression on certain systems, where
the CPU would remain at the lowest frequency, degrading performance
substantially"
* tag 'sched-urgent-2024-01-18' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/fair: Fix frequency selection for non-invariant case
Linus reported a ~50% performance regression on single-threaded
workloads on his AMD Ryzen system, and bisected it to:
9c0b4bb7f6 ("sched/cpufreq: Rework schedutil governor performance estimation")
When frequency invariance is not enabled, get_capacity_ref_freq(policy)
is supposed to return the current frequency and the performance margin
applied by map_util_perf(), enabling the utilization to go above the
maximum compute capacity and to select a higher frequency than the current one.
After the changes in 9c0b4bb7f6, the performance margin was applied
earlier in the path to take into account utilization clampings and
we couldn't get a utilization higher than the maximum compute capacity,
and the CPU remained 'stuck' at lower frequencies.
To fix this, we must use a frequency above the current frequency to
get a chance to select a higher OPP when the current one becomes fully used.
Apply the same margin and return a frequency 25% higher than the current
one in order to switch to the next OPP before we fully use the CPU
at the current one.
[ mingo: Clarified the changelog. ]
Fixes: 9c0b4bb7f6 ("sched/cpufreq: Rework schedutil governor performance estimation")
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Bisected-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Wyes Karny <wkarny@gmail.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Wyes Karny <wkarny@gmail.com>
Link: https://lore.kernel.org/r/20240114183600.135316-1-vincent.guittot@linaro.org
Pull header cleanups from Kent Overstreet:
"The goal is to get sched.h down to a type only header, so the main
thing happening in this patchset is splitting out various _types.h
headers and dependency fixups, as well as moving some things out of
sched.h to better locations.
This is prep work for the memory allocation profiling patchset which
adds new sched.h interdepencencies"
* tag 'header_cleanup-2024-01-10' of https://evilpiepirate.org/git/bcachefs: (51 commits)
Kill sched.h dependency on rcupdate.h
kill unnecessary thread_info.h include
Kill unnecessary kernel.h include
preempt.h: Kill dependency on list.h
rseq: Split out rseq.h from sched.h
LoongArch: signal.c: add header file to fix build error
restart_block: Trim includes
lockdep: move held_lock to lockdep_types.h
sem: Split out sem_types.h
uidgid: Split out uidgid_types.h
seccomp: Split out seccomp_types.h
refcount: Split out refcount_types.h
uapi/linux/resource.h: fix include
x86/signal: kill dependency on time.h
syscall_user_dispatch.h: split out *_types.h
mm_types_task.h: Trim dependencies
Split out irqflags_types.h
ipc: Kill bogus dependency on spinlock.h
shm: Slim down dependencies
workqueue: Split out workqueue_types.h
...
Pull non-MM updates from Andrew Morton:
"Quite a lot of kexec work this time around. Many singleton patches in
many places. The notable patch series are:
- nilfs2 folio conversion from Matthew Wilcox in 'nilfs2: Folio
conversions for file paths'.
- Additional nilfs2 folio conversion from Ryusuke Konishi in 'nilfs2:
Folio conversions for directory paths'.
- IA64 remnant removal in Heiko Carstens's 'Remove unused code after
IA-64 removal'.
- Arnd Bergmann has enabled the -Wmissing-prototypes warning
everywhere in 'Treewide: enable -Wmissing-prototypes'. This had
some followup fixes:
- Nathan Chancellor has cleaned up the hexagon build in the series
'hexagon: Fix up instances of -Wmissing-prototypes'.
- Nathan also addressed some s390 warnings in 's390: A couple of
fixes for -Wmissing-prototypes'.
- Arnd Bergmann addresses the same warnings for MIPS in his series
'mips: address -Wmissing-prototypes warnings'.
- Baoquan He has made kexec_file operate in a top-down-fitting manner
similar to kexec_load in the series 'kexec_file: Load kernel at top
of system RAM if required'
- Baoquan He has also added the self-explanatory 'kexec_file: print
out debugging message if required'.
- Some checkstack maintenance work from Tiezhu Yang in the series
'Modify some code about checkstack'.
- Douglas Anderson has disentangled the watchdog code's logging when
multiple reports are occurring simultaneously. The series is
'watchdog: Better handling of concurrent lockups'.
- Yuntao Wang has contributed some maintenance work on the crash code
in 'crash: Some cleanups and fixes'"
* tag 'mm-nonmm-stable-2024-01-09-10-33' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (157 commits)
crash_core: fix and simplify the logic of crash_exclude_mem_range()
x86/crash: use SZ_1M macro instead of hardcoded value
x86/crash: remove the unused image parameter from prepare_elf_headers()
kdump: remove redundant DEFAULT_CRASH_KERNEL_LOW_SIZE
scripts/decode_stacktrace.sh: strip unexpected CR from lines
watchdog: if panicking and we dumped everything, don't re-enable dumping
watchdog/hardlockup: use printk_cpu_sync_get_irqsave() to serialize reporting
watchdog/softlockup: use printk_cpu_sync_get_irqsave() to serialize reporting
watchdog/hardlockup: adopt softlockup logic avoiding double-dumps
kexec_core: fix the assignment to kimage->control_page
x86/kexec: fix incorrect end address passed to kernel_ident_mapping_init()
lib/trace_readwrite.c:: replace asm-generic/io with linux/io
nilfs2: cpfile: fix some kernel-doc warnings
stacktrace: fix kernel-doc typo
scripts/checkstack.pl: fix no space expression between sp and offset
x86/kexec: fix incorrect argument passed to kexec_dprintk()
x86/kexec: use pr_err() instead of kexec_dprintk() when an error occurs
nilfs2: add missing set_freezable() for freezable kthread
kernel: relay: remove relay_file_splice_read dead code, doesn't work
docs: submit-checklist: remove all of "make namespacecheck"
...
When a CPU is taken offline, the contribution of its cfs_rqs to task_groups'
load may remain and will negatively impact the calculation of the share of
the online CPUs.
To fix this bug, clear the contribution of an offlining CPU to task groups'
load and skip its contribution while it is inactive.
Here's the reproducer of the anomaly, by Imran Khan:
"So far I have encountered only one rather lengthy way of reproducing this issue,
which is as follows:
1. Take a KVM guest (booted with 4 CPUs and can be scaled up to 124 CPUs) and
create 2 custom cgroups: /sys/fs/cgroup/cpu/test_group_1 and /sys/fs/cgroup/
cpu/test_group_2
2. Assign a CPU intensive workload to each of these cgroups and start the
workload.
For my tests I am using following app:
int main(int argc, char *argv[])
{
unsigned long count, i, val;
if (argc != 2) {
printf("usage: ./a.out <number of random nums to generate> \n");
return 0;
}
count = strtoul(argv[1], NULL, 10);
printf("Generating %lu random numbers \n", count);
for (i = 0; i < count; i++) {
val = rand();
val = val % 2;
//usleep(1);
}
printf("Generated %lu random numbers \n", count);
return 0;
}
Also since the system is booted with 4 CPUs, in order to completely load the
system I am also launching 4 instances of same test app under:
/sys/fs/cgroup/cpu/
3. We can see that both of the cgroups get similar CPU time:
# systemd-cgtop --depth 1
Path Tasks %CPU Memory Input/s Output/s
/ 659 - 5.5G - -
/system.slice - - 5.7G - -
/test_group_1 4 - - - -
/test_group_2 3 - - - -
/user.slice 31 - 56.5M - -
Path Tasks %CPU Memory Input/s Output/s
/ 659 394.6 5.5G - -
/test_group_2 3 65.7 - - -
/user.slice 29 55.1 48.0M - -
/test_group_1 4 47.3 - - -
/system.slice - 2.2 5.7G - -
Path Tasks %CPU Memory Input/s Output/s
/ 659 394.8 5.5G - -
/test_group_1 4 62.9 - - -
/user.slice 28 44.9 54.2M - -
/test_group_2 3 44.7 - - -
/system.slice - 0.9 5.7G - -
Path Tasks %CPU Memory Input/s Output/s
/ 659 394.4 5.5G - -
/test_group_2 3 58.8 - - -
/test_group_1 4 51.9 - - -
/user.slice 30 39.3 59.6M - -
/system.slice - 1.9 5.7G - -
Path Tasks %CPU Memory Input/s Output/s
/ 659 394.7 5.5G - -
/test_group_1 4 60.9 - - -
/test_group_2 3 57.9 - - -
/user.slice 28 43.5 36.9M - -
/system.slice - 3.0 5.7G - -
Path Tasks %CPU Memory Input/s Output/s
/ 659 395.0 5.5G - -
/test_group_1 4 66.8 - - -
/test_group_2 3 56.3 - - -
/user.slice 29 43.1 51.8M - -
/system.slice - 0.7 5.7G - -
4. Now move systemd-udevd to one of these test groups, say test_group_1, and
perform scale up to 124 CPUs followed by scale down back to 4 CPUs from the
host side.
5. Run the same workload i.e 4 instances of CPU hogger under /sys/fs/cgroup/cpu
and one instance of CPU hogger each in /sys/fs/cgroup/cpu/test_group_1 and
/sys/fs/cgroup/test_group_2.
It can be seen that test_group_1 (the one where systemd-udevd was moved) is getting
much less CPU time than the test_group_2, even though at this point of time both of
these groups have only CPU hogger running:
# systemd-cgtop --depth 1
Path Tasks %CPU Memory Input/s Output/s
/ 1219 - 5.4G - -
/system.slice - - 5.6G - -
/test_group_1 4 - - - -
/test_group_2 3 - - - -
/user.slice 26 - 91.3M - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 394.3 5.4G - -
/test_group_2 3 82.7 - - -
/test_group_1 4 14.3 - - -
/system.slice - 0.8 5.6G - -
/user.slice 26 0.4 91.2M - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 394.6 5.4G - -
/test_group_2 3 67.4 - - -
/system.slice - 24.6 5.6G - -
/test_group_1 4 12.5 - - -
/user.slice 26 0.4 91.2M - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 395.2 5.4G - -
/test_group_2 3 60.9 - - -
/system.slice - 27.9 5.6G - -
/test_group_1 4 12.2 - - -
/user.slice 26 0.4 91.2M - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 395.2 5.4G - -
/test_group_2 3 69.4 - - -
/test_group_1 4 13.9 - - -
/user.slice 28 1.6 92.0M - -
/system.slice - 1.0 5.6G - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 395.6 5.4G - -
/test_group_2 3 59.3 - - -
/test_group_1 4 14.1 - - -
/user.slice 28 1.3 92.2M - -
/system.slice - 0.7 5.6G - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 395.5 5.4G - -
/test_group_2 3 67.2 - - -
/test_group_1 4 11.5 - - -
/user.slice 28 1.3 92.5M - -
/system.slice - 0.6 5.6G - -
Path Tasks %CPU Memory Input/s Output/s
/ 1221 395.1 5.4G - -
/test_group_2 3 76.8 - - -
/test_group_1 4 12.9 - - -
/user.slice 28 1.3 92.8M - -
/system.slice - 1.2 5.6G - -
From sched_debug data it can be seen that in bad case the load.weight of per-CPU
sched entities corresponding to test_group_1 has reduced significantly and
also load_avg of test_group_1 remains much higher than that of test_group_2,
even though systemd-udevd stopped running long time back and at this point of
time both cgroups just have the CPU hogger app as running entity."
[ mingo: Added details from the original discussion, plus minor edits to the patch. ]
Reported-by: Imran Khan <imran.f.khan@oracle.com>
Tested-by: Imran Khan <imran.f.khan@oracle.com>
Tested-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Imran Khan <imran.f.khan@oracle.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Link: https://lore.kernel.org/r/20231223111545.62135-1-vincent.guittot@linaro.org
We're trying to get sched.h down to more or less just types only, not
code - rseq can live in its own header.
This helps us kill the dependency on preempt.h in sched.h.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Running N CPU-bound tasks on an N CPUs platform:
- with asymmetric CPU capacity
- not being a DynamIq system (i.e. having a PKG level sched domain
without the SD_SHARE_PKG_RESOURCES flag set)
.. might result in a task placement where two tasks run on a big CPU
and none on a little CPU. This placement could be more optimal by
using all CPUs.
Testing platform:
Juno-r2:
- 2 big CPUs (1-2), maximum capacity of 1024
- 4 little CPUs (0,3-5), maximum capacity of 383
Testing workload ([1]):
Spawn 6 CPU-bound tasks. During the first 100ms (step 1), each tasks
is affine to a CPU, except for:
- one little CPU which is left idle.
- one big CPU which has 2 tasks affine.
After the 100ms (step 2), remove the cpumask affinity.
Behavior before the patch:
During step 2, the load balancer running from the idle CPU tags sched
domains as:
- little CPUs: 'group_has_spare'. Cf. group_has_capacity() and
group_is_overloaded(), 3 CPU-bound tasks run on a 4 CPUs
sched-domain, and the idle CPU provides enough spare capacity
regarding the imbalance_pct
- big CPUs: 'group_overloaded'. Indeed, 3 tasks run on a 2 CPUs
sched-domain, so the following path is used:
group_is_overloaded()
\-if (sgs->sum_nr_running <= sgs->group_weight) return true;
The following path which would change the migration type to
'migrate_task' is not taken:
calculate_imbalance()
\-if (env->idle != CPU_NOT_IDLE && env->imbalance == 0)
as the local group has some spare capacity, so the imbalance
is not 0.
The migration type requested is 'migrate_util' and the busiest
runqueue is the big CPU's runqueue having 2 tasks (each having a
utilization of 512). The idle little CPU cannot pull one of these
task as its capacity is too small for the task. The following path
is used:
detach_tasks()
\-case migrate_util:
\-if (util > env->imbalance) goto next;
After the patch:
As the number of failed balancing attempts grows (with
'nr_balance_failed'), progressively make it easier to migrate
a big task to the idling little CPU. A similar mechanism is
used for the 'migrate_load' migration type.
Improvement:
Running the testing workload [1] with the step 2 representing
a ~10s load for a big CPU:
Before patch: ~19.3s
After patch: ~18s (-6.7%)
Similar issue reported at:
https://lore.kernel.org/lkml/20230716014125.139577-1-qyousef@layalina.io/
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Pierre Gondois <pierre.gondois@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Qais Yousef <qyousef@layalina.io>
Link: https://lore.kernel.org/r/20231206090043.634697-1-pierre.gondois@arm.com
The current method to take into account uclamp hints when estimating the
target frequency can end in a situation where the selected target
frequency is finally higher than uclamp hints, whereas there are no real
needs. Such cases mainly happen because we are currently mixing the
traditional scheduler utilization signal with the uclamp performance
hints. By adding these 2 metrics, we loose an important information when
it comes to select the target frequency, and we have to make some
assumptions which can't fit all cases.
Rework the interface between the scheduler and schedutil governor in order
to propagate all information down to the cpufreq governor.
effective_cpu_util() interface changes and now returns the actual
utilization of the CPU with 2 optional inputs:
- The minimum performance for this CPU; typically the capacity to handle
the deadline task and the interrupt pressure. But also uclamp_min
request when available.
- The maximum targeting performance for this CPU which reflects the
maximum level that we would like to not exceed. By default it will be
the CPU capacity but can be reduced because of some performance hints
set with uclamp. The value can be lower than actual utilization and/or
min performance level.
A new sugov_effective_cpu_perf() interface is also available to compute
the final performance level that is targeted for the CPU, after applying
some cpufreq headroom and taking into account all inputs.
With these 2 functions, schedutil is now able to decide when it must go
above uclamp hints. It now also has a generic way to get the min
performance level.
The dependency between energy model and cpufreq governor and its headroom
policy doesn't exist anymore.
eenv_pd_max_util() asks schedutil for the targeted performance after
applying the impact of the waking task.
[ mingo: Refined the changelog & C comments. ]
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Link: https://lore.kernel.org/r/20231122133904.446032-2-vincent.guittot@linaro.org
Lukasz Luba reported that a thread's util_est can significantly decrease as
a result of sharing the CPU with other threads.
The use case can be easily reproduced with a periodic task TA that runs 1ms
and sleeps 100us. When the task is alone on the CPU, its max utilization and
its util_est is around 888. If another similar task starts to run on the
same CPU, TA will have to share the CPU runtime and its maximum utilization
will decrease around half the CPU capacity (512) then TA's util_est will
follow this new maximum trend which is only the result of sharing the CPU
with others tasks.
Such situation can be detected with runnable_avg wich is close or
equal to util_avg when TA is alone, but increases above util_avg when TA
shares the CPU with other threads and wait on the runqueue.
[ We prefer an util_est that overestimate rather than under estimate
because in 1st case we will not provide enough performance to the
task which will remain under-provisioned, whereas in the other case we
will create some idle time which will enable to reduce contention and
as a result reduces the util_est so the overestimate will be transient
whereas the underestimate will remain. ]
[ mingo: Refined the changelog, added comments from the LKML discussion. ]
Reported-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/lkml/CAKfTPtDd-HhF-YiNTtL9i5k0PfJbF819Yxu4YquzfXgwi7voyw@mail.gmail.com/#t
Link: https://lore.kernel.org/r/20231122140119.472110-1-vincent.guittot@linaro.org
Cc: Hongyan Xia <hongyan.xia2@arm.com>
Low priority tasks (e.g., SCHED_OTHER) can suffer starvation if tasks
with higher priority (e.g., SCHED_FIFO) monopolize CPU(s).
RT Throttling has been introduced a while ago as a (mostly debug)
countermeasure one can utilize to reserve some CPU time for low priority
tasks (usually background type of work, e.g. workqueues, timers, etc.).
It however has its own problems (see documentation) and the undesired
effect of unconditionally throttling FIFO tasks even when no lower
priority activity needs to run (there are mechanisms to fix this issue
as well, but, again, with their own problems).
Introduce deadline servers to service low priority tasks needs under
starvation conditions. Deadline servers are built extending SCHED_DEADLINE
implementation to allow 2-level scheduling (a sched_deadline entity
becomes a container for lower priority scheduling entities).
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/4968601859d920335cf85822eb573a5f179f04b8.1699095159.git.bristot@kernel.org
Now that trace_sched_stat_runtime() no longer takes a vruntime
argument, the task specific bits are identical between
update_curr_common() and update_curr().
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tracing the runtime delta makes sense, observer can sum over time.
Tracing the absolute vruntime makes less sense, inconsistent:
absolute-vs-delta, but also vruntime delta can be computed from
runtime delta.
Removing the vruntime thing also makes the two tracepoint sites
identical, allowing to unify the code in a later patch.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Since the RB-tree is now sorted by deadline, let's first try the
leftmost entity which has the earliest virtual deadline. I've done
some benchmarks to see its effectiveness.
All the benchmarks are done inside a normal cpu cgroup in a clean
environment with cpu turbo disabled, on a dual-CPU Intel Xeon(R)
Platinum 8260 with 2 NUMA nodes each of which has 24C/48T.
hackbench: process/thread + pipe/socket + 1/2/4/8 groups
netperf: TCP/UDP + STREAM/RR + 24/48/72/96/192 threads
tbench: loopback 24/48/72/96/192 threads
schbench: 1/2/4/8 mthreads
direct: cfs_rq has only one entity
parity: RUN_TO_PARITY
fast: O(1) fastpath
slow: heap search
(%) direct parity fast slow
hackbench 92.95 2.02 4.91 0.12
netperf 68.08 6.60 24.18 1.14
tbench 67.55 11.22 20.61 0.62
schbench 69.91 2.65 25.73 1.71
The above results indicate that this fastpath really makes task
selection more efficient.
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20231115033647.80785-4-wuyun.abel@bytedance.com
Since commit fc137c0dda ("sched/numa: enhance vma scanning logic")
NUMA Balancing allows updating PTEs to trap NUMA hinting faults if the
task had previously accessed VMA. However unconditional scan of VMAs are
allowed during initial phase of VMA creation until process's
mm numa_scan_seq reaches 2 even though current task had not accessed VMA.
Rationale:
- Without initial scan subsequent PTE update may never happen.
- Give fair opportunity to all the VMAs to be scanned and subsequently
understand the access pattern of all the VMAs.
But it has a corner case where, if a VMA is created after some time,
process's mm numa_scan_seq could be already greater than 2.
For e.g., values of mm numa_scan_seq when VMAs are created by running
mmtest autonuma benchmark briefly looks like:
start_seq=0 : 459
start_seq=2 : 138
start_seq=3 : 144
start_seq=4 : 8
start_seq=8 : 1
start_seq=9 : 1
This results in no unconditional PTE updates for those VMAs created after
some time.
Fix:
- Note down the initial value of mm numa_scan_seq in per VMA start_seq.
- Allow unconditional scan till start_seq + 2.
Result:
SUT: AMD EPYC Milan with 2 NUMA nodes 256 cpus.
base kernel: upstream 6.6-rc6 with Mels patches [1] applied.
kernbench
========== base patched %gain
Amean elsp-128 165.09 ( 0.00%) 164.78 * 0.19%*
Duration User 41404.28 41375.08
Duration System 9862.22 9768.48
Duration Elapsed 519.87 518.72
Ops NUMA PTE updates 1041416.00 831536.00
Ops NUMA hint faults 263296.00 220966.00
Ops NUMA pages migrated 258021.00 212769.00
Ops AutoNUMA cost 1328.67 1114.69
autonumabench
NUMA01_THREADLOCAL
==================
Amean elsp-NUMA01_THREADLOCAL 81.79 (0.00%) 67.74 * 17.18%*
Duration User 54832.73 47379.67
Duration System 75.00 185.75
Duration Elapsed 576.72 476.09
Ops NUMA PTE updates 394429.00 11121044.00
Ops NUMA hint faults 1001.00 8906404.00
Ops NUMA pages migrated 288.00 2998694.00
Ops AutoNUMA cost 7.77 44666.84
Signed-off-by: Raghavendra K T <raghavendra.kt@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/2ea7cbce80ac7c62e90cbfb9653a7972f902439f.1697816692.git.raghavendra.kt@amd.com
should_we_balance is called for the decision to do load-balancing.
When sched ticks invoke this function, only one CPU should return
true. However, in the current code, two CPUs can return true. The
following situation, where b means busy and i means idle, is an
example, because CPU 0 and CPU 2 return true.
[0, 1] [2, 3]
b b i b
This fix checks if there exists an idle CPU with busy sibling(s)
after looking for a CPU on an idle core. If some idle CPUs with busy
siblings are found, just the first one should do load-balancing.
Fixes: b1bfeab9b0 ("sched/fair: Consider the idle state of the whole core for load balance")
Signed-off-by: Keisuke Nishimura <keisuke.nishimura@inria.fr>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Chen Yu <yu.c.chen@intel.com>
Reviewed-by: Shrikanth Hegde <sshegde@linux.vnet.ibm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20231031133821.1570861-1-keisuke.nishimura@inria.fr
vruntime of the (on_rq && !0-lag) entity needs to be adjusted when
it gets re-weighted, and the calculations can be simplified based
on the fact that re-weight won't change the w-average of all the
entities. Please check the proofs in comments.
But adjusting vruntime can also cause position change in RB-tree
hence require re-queue to fix up which might be costly. This might
be avoided by deferring adjustment to the time the entity actually
leaves tree (dequeue/pick), but that will negatively affect task
selection and probably not good enough either.
Fixes: 147f3efaa2 ("sched/fair: Implement an EEVDF-like scheduling policy")
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20231107090510.71322-2-wuyun.abel@bytedance.com
Pull non-MM updates from Andrew Morton:
"As usual, lots of singleton and doubleton patches all over the tree
and there's little I can say which isn't in the individual changelogs.
The lengthier patch series are
- 'kdump: use generic functions to simplify crashkernel reservation
in arch', from Baoquan He. This is mainly cleanups and
consolidation of the 'crashkernel=' kernel parameter handling
- After much discussion, David Laight's 'minmax: Relax type checks in
min() and max()' is here. Hopefully reduces some typecasting and
the use of min_t() and max_t()
- A group of patches from Oleg Nesterov which clean up and slightly
fix our handling of reads from /proc/PID/task/... and which remove
task_struct.thread_group"
* tag 'mm-nonmm-stable-2023-11-02-14-08' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (64 commits)
scripts/gdb/vmalloc: disable on no-MMU
scripts/gdb: fix usage of MOD_TEXT not defined when CONFIG_MODULES=n
.mailmap: add address mapping for Tomeu Vizoso
mailmap: update email address for Claudiu Beznea
tools/testing/selftests/mm/run_vmtests.sh: lower the ptrace permissions
.mailmap: map Benjamin Poirier's address
scripts/gdb: add lx_current support for riscv
ocfs2: fix a spelling typo in comment
proc: test ProtectionKey in proc-empty-vm test
proc: fix proc-empty-vm test with vsyscall
fs/proc/base.c: remove unneeded semicolon
do_io_accounting: use sig->stats_lock
do_io_accounting: use __for_each_thread()
ocfs2: replace BUG_ON() at ocfs2_num_free_extents() with ocfs2_error()
ocfs2: fix a typo in a comment
scripts/show_delta: add __main__ judgement before main code
treewide: mark stuff as __ro_after_init
fs: ocfs2: check status values
proc: test /proc/${pid}/statm
compiler.h: move __is_constexpr() to compiler.h
...
Pull MM updates from Andrew Morton:
"Many singleton patches against the MM code. The patch series which are
included in this merge do the following:
- Kemeng Shi has contributed some compation maintenance work in the
series 'Fixes and cleanups to compaction'
- Joel Fernandes has a patchset ('Optimize mremap during mutual
alignment within PMD') which fixes an obscure issue with mremap()'s
pagetable handling during a subsequent exec(), based upon an
implementation which Linus suggested
- More DAMON/DAMOS maintenance and feature work from SeongJae Park i
the following patch series:
mm/damon: misc fixups for documents, comments and its tracepoint
mm/damon: add a tracepoint for damos apply target regions
mm/damon: provide pseudo-moving sum based access rate
mm/damon: implement DAMOS apply intervals
mm/damon/core-test: Fix memory leaks in core-test
mm/damon/sysfs-schemes: Do DAMOS tried regions update for only one apply interval
- In the series 'Do not try to access unaccepted memory' Adrian
Hunter provides some fixups for the recently-added 'unaccepted
memory' feature. To increase the feature's checking coverage. 'Plug
a few gaps where RAM is exposed without checking if it is
unaccepted memory'
- In the series 'cleanups for lockless slab shrink' Qi Zheng has done
some maintenance work which is preparation for the lockless slab
shrinking code
- Qi Zheng has redone the earlier (and reverted) attempt to make slab
shrinking lockless in the series 'use refcount+RCU method to
implement lockless slab shrink'
- David Hildenbrand contributes some maintenance work for the rmap
code in the series 'Anon rmap cleanups'
- Kefeng Wang does more folio conversions and some maintenance work
in the migration code. Series 'mm: migrate: more folio conversion
and unification'
- Matthew Wilcox has fixed an issue in the buffer_head code which was
causing long stalls under some heavy memory/IO loads. Some cleanups
were added on the way. Series 'Add and use bdev_getblk()'
- In the series 'Use nth_page() in place of direct struct page
manipulation' Zi Yan has fixed a potential issue with the direct
manipulation of hugetlb page frames
- In the series 'mm: hugetlb: Skip initialization of gigantic tail
struct pages if freed by HVO' has improved our handling of gigantic
pages in the hugetlb vmmemmep optimizaton code. This provides
significant boot time improvements when significant amounts of
gigantic pages are in use
- Matthew Wilcox has sent the series 'Small hugetlb cleanups' - code
rationalization and folio conversions in the hugetlb code
- Yin Fengwei has improved mlock()'s handling of large folios in the
series 'support large folio for mlock'
- In the series 'Expose swapcache stat for memcg v1' Liu Shixin has
added statistics for memcg v1 users which are available (and
useful) under memcg v2
- Florent Revest has enhanced the MDWE (Memory-Deny-Write-Executable)
prctl so that userspace may direct the kernel to not automatically
propagate the denial to child processes. The series is named 'MDWE
without inheritance'
- Kefeng Wang has provided the series 'mm: convert numa balancing
functions to use a folio' which does what it says
- In the series 'mm/ksm: add fork-exec support for prctl' Stefan
Roesch makes is possible for a process to propagate KSM treatment
across exec()
- Huang Ying has enhanced memory tiering's calculation of memory
distances. This is used to permit the dax/kmem driver to use 'high
bandwidth memory' in addition to Optane Data Center Persistent
Memory Modules (DCPMM). The series is named 'memory tiering:
calculate abstract distance based on ACPI HMAT'
- In the series 'Smart scanning mode for KSM' Stefan Roesch has
optimized KSM by teaching it to retain and use some historical
information from previous scans
- Yosry Ahmed has fixed some inconsistencies in memcg statistics in
the series 'mm: memcg: fix tracking of pending stats updates
values'
- In the series 'Implement IOCTL to get and optionally clear info
about PTEs' Peter Xu has added an ioctl to /proc/<pid>/pagemap
which permits us to atomically read-then-clear page softdirty
state. This is mainly used by CRIU
- Hugh Dickins contributed the series 'shmem,tmpfs: general
maintenance', a bunch of relatively minor maintenance tweaks to
this code
- Matthew Wilcox has increased the use of the VMA lock over
file-backed page faults in the series 'Handle more faults under the
VMA lock'. Some rationalizations of the fault path became possible
as a result
- In the series 'mm/rmap: convert page_move_anon_rmap() to
folio_move_anon_rmap()' David Hildenbrand has implemented some
cleanups and folio conversions
- In the series 'various improvements to the GUP interface' Lorenzo
Stoakes has simplified and improved the GUP interface with an eye
to providing groundwork for future improvements
- Andrey Konovalov has sent along the series 'kasan: assorted fixes
and improvements' which does those things
- Some page allocator maintenance work from Kemeng Shi in the series
'Two minor cleanups to break_down_buddy_pages'
- In thes series 'New selftest for mm' Breno Leitao has developed
another MM self test which tickles a race we had between madvise()
and page faults
- In the series 'Add folio_end_read' Matthew Wilcox provides cleanups
and an optimization to the core pagecache code
- Nhat Pham has added memcg accounting for hugetlb memory in the
series 'hugetlb memcg accounting'
- Cleanups and rationalizations to the pagemap code from Lorenzo
Stoakes, in the series 'Abstract vma_merge() and split_vma()'
- Audra Mitchell has fixed issues in the procfs page_owner code's new
timestamping feature which was causing some misbehaviours. In the
series 'Fix page_owner's use of free timestamps'
- Lorenzo Stoakes has fixed the handling of new mappings of sealed
files in the series 'permit write-sealed memfd read-only shared
mappings'
- Mike Kravetz has optimized the hugetlb vmemmap optimization in the
series 'Batch hugetlb vmemmap modification operations'
- Some buffer_head folio conversions and cleanups from Matthew Wilcox
in the series 'Finish the create_empty_buffers() transition'
- As a page allocator performance optimization Huang Ying has added
automatic tuning to the allocator's per-cpu-pages feature, in the
series 'mm: PCP high auto-tuning'
- Roman Gushchin has contributed the patchset 'mm: improve
performance of accounted kernel memory allocations' which improves
their performance by ~30% as measured by a micro-benchmark
- folio conversions from Kefeng Wang in the series 'mm: convert page
cpupid functions to folios'
- Some kmemleak fixups in Liu Shixin's series 'Some bugfix about
kmemleak'
- Qi Zheng has improved our handling of memoryless nodes by keeping
them off the allocation fallback list. This is done in the series
'handle memoryless nodes more appropriately'
- khugepaged conversions from Vishal Moola in the series 'Some
khugepaged folio conversions'"
[ bcachefs conflicts with the dynamically allocated shrinkers have been
resolved as per Stephen Rothwell in
https://lore.kernel.org/all/20230913093553.4290421e@canb.auug.org.au/
with help from Qi Zheng.
The clone3 test filtering conflict was half-arsed by yours truly ]
* tag 'mm-stable-2023-11-01-14-33' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (406 commits)
mm/damon/sysfs: update monitoring target regions for online input commit
mm/damon/sysfs: remove requested targets when online-commit inputs
selftests: add a sanity check for zswap
Documentation: maple_tree: fix word spelling error
mm/vmalloc: fix the unchecked dereference warning in vread_iter()
zswap: export compression failure stats
Documentation: ubsan: drop "the" from article title
mempolicy: migration attempt to match interleave nodes
mempolicy: mmap_lock is not needed while migrating folios
mempolicy: alloc_pages_mpol() for NUMA policy without vma
mm: add page_rmappable_folio() wrapper
mempolicy: remove confusing MPOL_MF_LAZY dead code
mempolicy: mpol_shared_policy_init() without pseudo-vma
mempolicy trivia: use pgoff_t in shared mempolicy tree
mempolicy trivia: slightly more consistent naming
mempolicy trivia: delete those ancient pr_debug()s
mempolicy: fix migrate_pages(2) syscall return nr_failed
kernfs: drop shared NUMA mempolicy hooks
hugetlbfs: drop shared NUMA mempolicy pretence
mm/damon/sysfs-test: add a unit test for damon_sysfs_set_targets()
...
