The kick_idle variable is no longer used, this commit therefore remove
it and also remove associated code in the do_pick_task_scx().
Signed-off-by: Zqiang <qiang.zhang@linux.dev>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
move_local_task_to_local_dsq() is used when moving a task from a non-local
DSQ to a local DSQ on the same CPU. It directly manipulates the local DSQ
without going through dispatch_enqueue() and was missing the post-enqueue
handling that triggers preemption when SCX_ENQ_PREEMPT is set or the idle
task is running.
The function is used by move_task_between_dsqs() which backs
scx_bpf_dsq_move() and may be called while the CPU is busy.
Add local_dsq_post_enq() call to move_local_task_to_local_dsq(). As the
dispatch path doesn't need post-enqueue handling, add SCX_RQ_IN_BALANCE
early exit to keep consume_dispatch_q() behavior unchanged and avoid
triggering unnecessary resched when scx_bpf_dsq_move() is used from the
dispatch path.
Fixes: 4c30f5ce4f ("sched_ext: Implement scx_bpf_dispatch[_vtime]_from_dsq()")
Cc: stable@vger.kernel.org # v6.12+
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Factor out local_dsq_post_enq() which performs post-enqueue handling for
local DSQs - triggering resched_curr() if SCX_ENQ_PREEMPT is specified or if
the current CPU is idle. No functional change.
This will be used by the next patch to fix move_local_task_to_local_dsq().
Cc: stable@vger.kernel.org # v6.12+
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
scx_enable() calls scx_bypass(true) to initialize in bypass mode and then
scx_bypass(false) on success to exit. If scx_enable() fails during task
initialization - e.g. scx_cgroup_init() or scx_init_task() returns an error -
it jumps to err_disable while bypass is still active. scx_disable_workfn()
then calls scx_bypass(true/false) for its own bypass, leaving the bypass depth
at 1 instead of 0. This causes the system to remain permanently in bypass mode
after a failed scx_enable().
Failures after task initialization is complete - e.g. scx_tryset_enable_state()
at the end - already call scx_bypass(false) before reaching the error path and
are not affected. This only affects a subset of failure modes.
Fix it by tracking whether scx_enable() called scx_bypass(true) in a bool and
having scx_disable_workfn() call an extra scx_bypass(false) to clear it. This
is a temporary measure as the bypass depth will be moved into the sched
instance, which will make this tracking unnecessary.
Fixes: 8c2090c504 ("sched_ext: Initialize in bypass mode")
Cc: stable@vger.kernel.org # v6.12+
Reported-by: Chris Mason <clm@meta.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Link: https://lore.kernel.org/stable/286e6f7787a81239e1ce2989b52391ce%40kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
Early when trying to get sched_ext and proxy-exe working together,
I kept tripping over NULL ptr in put_prev_task_scx() on the line:
if (sched_class_above(&ext_sched_class, next->sched_class)) {
Which was due to put_prev_task() passes a NULL next, calling:
prev->sched_class->put_prev_task(rq, prev, NULL);
put_prev_task_scx() already guards for a NULL next in the
switch_class case, but doesn't seem to have a guard for
sched_class_above() check.
I can't say I understand why this doesn't trip usually without
proxy-exec. And in newer kernels there are way fewer
put_prev_task(), and I can't easily reproduce the issue now
even with proxy-exec.
But we still have one put_prev_task() call left in core.c that
seems like it could trip this, so I wanted to send this out for
consideration.
tj: put_prev_task() can be called with NULL @next; however, when @p is
queued, that doesn't happen, so this condition shouldn't currently be
triggerable. The connection isn't straightforward or necessarily reliable,
so add the NULL check even if it can't currently be triggered.
Link: http://lkml.kernel.org/r/20251206022218.1541878-1-jstultz@google.com
Signed-off-by: John Stultz <jstultz@google.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Pull scheduler fixes from Ingo Molnar:
"Miscellaneous scheduler fixes/cleanups:
- Fix psi_dequeue() for Proxy Execution
- Fix hrtick() vs. scheduling context bug
- Fix unfairness caused by stalled tg_load_avg_contrib when the last
task migrates out
- Fix whitespace noise in headers
- Remove a preempt-disable section in rt_mutex_setprio()"
* tag 'sched-urgent-2025-12-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/core: Fix psi_dequeue() for Proxy Execution
sched/fair: Fix unfairness caused by stalled tg_load_avg_contrib when the last task migrates out
sched/rt: Remove a preempt-disable section in rt_mutex_setprio()
sched/hrtick: Fix hrtick() vs. scheduling context
sched/headers: Remove whitespace noise from kernel/sched/sched.h
Currently, if the sleep flag is set, psi_dequeue() doesn't
change any of the psi_flags.
This is because psi_task_switch() will clear TSK_ONCPU as well
as other potential flags (TSK_RUNNING), and the assumption is
that a voluntary sleep always consists of a task being dequeued
followed shortly there after with a psi_sched_switch() call.
Proxy Execution changes this expectation, as mutex-blocked tasks
that would normally sleep stay on the runqueue. But in the case
where the mutex-owning task goes to sleep, or the owner is on a
remote cpu, we will then deactivate the blocked task shortly
after.
In that situation, the mutex-blocked task will have had its
TSK_ONCPU cleared when it was switched off the cpu, but it will
stay TSK_RUNNING. Then if we later dequeue it (as currently done
if we hit a case find_proxy_task() can't yet handle, such as the
case of the owner being on another rq or a sleeping owner)
psi_dequeue() won't change any state (leaving it TSK_RUNNING),
as it incorrectly expects a psi_task_switch() call to
immediately follow.
Later on when the task get woken/re-enqueued, and psi_flags are
set for TSK_RUNNING, we hit an error as the task is already
TSK_RUNNING:
psi: inconsistent task state! task=188:kworker/28:0 cpu=28 psi_flags=4 clear=0 set=4
To resolve this, extend the logic in psi_dequeue() so that
if the sleep flag is set, we also check if psi_flags have
TSK_ONCPU set (meaning the psi_task_switch is imminent) before
we do the shortcut return.
If TSK_ONCPU is not set, that means we've already switched away,
and this psi_dequeue call needs to clear the flags.
Fixes: be41bde4c3 ("sched: Add an initial sketch of the find_proxy_task() function")
Reported-by: K Prateek Nayak <kprateek.nayak@amd.com>
Signed-off-by: John Stultz <jstultz@google.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: K Prateek Nayak <kprateek.nayak@amd.com>
Tested-by: Haiyue Wang <haiyuewa@163.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://patch.msgid.link/20251205012721.756394-1-jstultz@google.com
Closes: https://lore.kernel.org/lkml/20251117185550.365156-1-kprateek.nayak@amd.com/
When a task is migrated out, there is a probability that the tg->load_avg
value will become abnormal. The reason is as follows:
1. Due to the 1ms update period limitation in update_tg_load_avg(), there
is a possibility that the reduced load_avg is not updated to tg->load_avg
when a task migrates out.
2. Even though __update_blocked_fair() traverses the leaf_cfs_rq_list and
calls update_tg_load_avg() for cfs_rqs that are not fully decayed, the key
function cfs_rq_is_decayed() does not check whether
cfs->tg_load_avg_contrib is null. Consequently, in some cases,
__update_blocked_fair() removes cfs_rqs whose avg.load_avg has not been
updated to tg->load_avg.
Add a check of cfs_rq->tg_load_avg_contrib in cfs_rq_is_decayed(),
which fixes the case (2.) mentioned above.
Fixes: 1528c661c2 ("sched/fair: Ratelimit update to tg->load_avg")
Signed-off-by: xupengbo <xupengbo@oppo.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Aaron Lu <ziqianlu@bytedance.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-by: Aaron Lu <ziqianlu@bytedance.com>
Link: https://patch.msgid.link/20250827022208.14487-1-xupengbo@oppo.com
rt_mutex_setprio() has only one caller: rt_mutex_adjust_prio(). It
expects that task_struct::pi_lock and rt_mutex_base::wait_lock are held.
Both locks are raw_spinlock_t and are acquired with disabled interrupts.
Nevertheless rt_mutex_setprio() disables preemption while invoking
__balance_callbacks() and raw_spin_rq_unlock(). Even if one of the
balance callbacks unlocks the rq then it must not enable interrupts
because rt_mutex_base::wait_lock is still locked.
Therefore interrupts should remain disabled and disabling preemption is
not needed.
Commit 4c9a4bc89a ("sched: Allow balance callbacks for check_class_changed()")
adds a preempt-disable section to rt_mutex_setprio() and
__sched_setscheduler(). In __sched_setscheduler() the preemption is
disabled before rq is unlocked and interrupts enabled but I don't see
why it makes a difference in rt_mutex_setprio().
Remove the preempt_disable() section from rt_mutex_setprio().
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://patch.msgid.link/20251127155529.t_sTatE4@linutronix.de
The sched_class::task_tick() method is called on the donor
sched_class, and sched_tick() hands it rq->donor as argument,
which is consistent.
However, while hrtick() uses the donor sched_class, it then passes
rq->curr, which is inconsistent. Fix it.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: John Stultz <jstultz@google.com>
Link: https://patch.msgid.link/20250918080205.442967033@infradead.org
Pull sched_ext updates from Tejun Heo:
- Improve recovery from misbehaving BPF schedulers.
When a scheduler puts many tasks with varying affinity restrictions
on a shared DSQ, CPUs scanning through tasks they cannot run can
overwhelm the system, causing lockups.
Bypass mode now uses per-CPU DSQs with a load balancer to avoid this,
and hooks into the hardlockup detector to attempt recovery.
Add scx_cpu0 example scheduler to demonstrate this scenario.
- Add lockless peek operation for DSQs to reduce lock contention for
schedulers that need to query queue state during load balancing.
- Allow scx_bpf_reenqueue_local() to be called from anywhere in
preparation for deprecating cpu_acquire/release() callbacks in favor
of generic BPF hooks.
- Prepare for hierarchical scheduler support: add
scx_bpf_task_set_slice() and scx_bpf_task_set_dsq_vtime() kfuncs,
make scx_bpf_dsq_insert*() return bool, and wrap kfunc args in
structs for future aux__prog parameter.
- Implement cgroup_set_idle() callback to notify BPF schedulers when a
cgroup's idle state changes.
- Fix migration tasks being incorrectly downgraded from
stop_sched_class to rt_sched_class across sched_ext enable/disable.
Applied late as the fix is low risk and the bug subtle but needs
stable backporting.
- Various fixes and cleanups including cgroup exit ordering,
SCX_KICK_WAIT reliability, and backward compatibility improvements.
* tag 'sched_ext-for-6.19' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext: (44 commits)
sched_ext: Fix incorrect sched_class settings for per-cpu migration tasks
sched_ext: tools: Removing duplicate targets during non-cross compilation
sched_ext: Use kvfree_rcu() to release per-cpu ksyncs object
sched_ext: Pass locked CPU parameter to scx_hardlockup() and add docs
sched_ext: Update comments replacing breather with aborting mechanism
sched_ext: Implement load balancer for bypass mode
sched_ext: Factor out abbreviated dispatch dequeue into dispatch_dequeue_locked()
sched_ext: Factor out scx_dsq_list_node cursor initialization into INIT_DSQ_LIST_CURSOR
sched_ext: Add scx_cpu0 example scheduler
sched_ext: Hook up hardlockup detector
sched_ext: Make handle_lockup() propagate scx_verror() result
sched_ext: Refactor lockup handlers into handle_lockup()
sched_ext: Make scx_exit() and scx_vexit() return bool
sched_ext: Exit dispatch and move operations immediately when aborting
sched_ext: Simplify breather mechanism with scx_aborting flag
sched_ext: Use per-CPU DSQs instead of per-node global DSQs in bypass mode
sched_ext: Refactor do_enqueue_task() local and global DSQ paths
sched_ext: Use shorter slice in bypass mode
sched_ext: Mark racy bitfields to prevent adding fields that can't tolerate races
sched_ext: Minor cleanups to scx_task_iter
...
Pull cgroup updates from Tejun Heo:
- Defer task cgroup unlink until after the dying task's final context
switch so that controllers see the cgroup properly populated until
the task is truly gone
- cpuset cleanups and simplifications.
Enforce that domain isolated CPUs stay in root or isolated partitions
and fail if isolated+nohz_full would leave no housekeeping CPU. Fix
sched/deadline root domain handling during CPU hot-unplug and race
for tasks in attaching cpusets
- Misc fixes including memory reclaim protection documentation and
selftest KTAP conformance
* tag 'cgroup-for-6.19' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: (21 commits)
cpuset: Treat cpusets in attaching as populated
sched/deadline: Walk up cpuset hierarchy to decide root domain when hot-unplug
cgroup/cpuset: Introduce cpuset_cpus_allowed_locked()
docs: cgroup: No special handling of unpopulated memcgs
docs: cgroup: Note about sibling relative reclaim protection
docs: cgroup: Explain reclaim protection target
selftests/cgroup: conform test to KTAP format output
cpuset: remove need_rebuild_sched_domains
cpuset: remove global remote_children list
cpuset: simplify node setting on error
cgroup: include missing header for struct irq_work
cgroup: Fix sleeping from invalid context warning on PREEMPT_RT
cgroup/cpuset: Globally track isolated_cpus update
cgroup/cpuset: Ensure domain isolated CPUs stay in root or isolated partition
cgroup/cpuset: Move up prstate_housekeeping_conflict() helper
cgroup/cpuset: Fail if isolated and nohz_full don't leave any housekeeping
cgroup/cpuset: Rename update_unbound_workqueue_cpumask() to update_isolation_cpumasks()
cgroup: Defer task cgroup unlink until after the task is done switching out
cgroup: Move dying_tasks cleanup from cgroup_task_release() to cgroup_task_free()
cgroup: Rename cgroup lifecycle hooks to cgroup_task_*()
...
Pull power management updates from Rafael Wysocki:
"There are quite a few interesting things here, including new hardware
support, new features, some bug fixes and documentation updates. In
addition, there are a usual bunch of minor fixes and cleanups all
over.
In the new hardware support category, there are intel_pstate and
intel_rapl driver updates to support new processors, Panther Lake,
Wildcat Lake, Noval Lake, and Diamond Rapids in the OOB mode, OPP and
bandwidth allocation support in the tegra186 cpufreq driver, and
JH7110S SOC support in dt-platdev cpufreq.
The new features are the PM QoS CPU latency limit for suspend-to-idle,
the netlink support for the energy model management, support for
terminating system suspend via a wakeup event during the sync of file
systems, configurable number of hibernation compression threads, the
runtime PM auto-cleanup macros, and the "poweroff" PM event that is
expected to be used during system shutdown.
Bugs are mostly fixed in cpuidle governors, but there are also fixes
elsewhere, like in the amd-pstate cpufreq driver.
Documentation updates include, but are not limited to, a new doc on
debugging shutdown hangs, cross-referencing fixes and cleanups in the
intel_pstate documentation, and updates of comments in the core
hibernation code.
Specifics:
- Introduce and document a QoS limit on CPU exit latency during
wakeup from suspend-to-idle (Ulf Hansson)
- Add support for building libcpupower statically (Zuo An)
- Add support for sending netlink notifications to user space on
energy model updates (Changwoo Mini, Peng Fan)
- Minor improvements to the Rust OPP interface (Tamir Duberstein)
- Fixes to scope-based pointers in the OPP library (Viresh Kumar)
- Use residency threshold in polling state override decisions in the
menu cpuidle governor (Aboorva Devarajan)
- Add sanity check for exit latency and target residency in the
cpufreq core (Rafael Wysocki)
- Use this_cpu_ptr() where possible in the teo governor (Christian
Loehle)
- Rework the handling of tick wakeups in the teo cpuidle governor to
increase the likelihood of stopping the scheduler tick in the cases
when tick wakeups can be counted as non-timer ones (Rafael Wysocki)
- Fix a reverse condition in the teo cpuidle governor and drop a
misguided target residency check from it (Rafael Wysocki)
- Clean up multiple minor defects in the teo cpuidle governor (Rafael
Wysocki)
- Update header inclusion to make it follow the Include What You Use
principle (Andy Shevchenko)
- Enable MSR-based RAPL PMU support in the intel_rapl power capping
driver and arrange for using it on the Panther Lake and Wildcat
Lake processors (Kuppuswamy Sathyanarayanan)
- Add support for Nova Lake and Wildcat Lake processors to the
intel_rapl power capping driver (Kaushlendra Kumar, Srinivas
Pandruvada)
- Add OPP and bandwidth support for Tegra186 (Aaron Kling)
- Optimizations for parameter array handling in the amd-pstate
cpufreq driver (Mario Limonciello)
- Fix for mode changes with offline CPUs in the amd-pstate cpufreq
driver (Gautham Shenoy)
- Preserve freq_table_sorted across suspend/hibernate in the cpufreq
core (Zihuan Zhang)
- Adjust energy model rules for Intel hybrid platforms in the
intel_pstate cpufreq driver and improve printing of debug messages
in it (Rafael Wysocki)
- Replace deprecated strcpy() in cpufreq_unregister_governor()
(Thorsten Blum)
- Fix duplicate hyperlink target errors in the intel_pstate cpufreq
driver documentation and use :ref: directive for internal linking
in it (Swaraj Gaikwad, Bagas Sanjaya)
- Add Diamond Rapids OOB mode support to the intel_pstate cpufreq
driver (Kuppuswamy Sathyanarayanan)
- Use mutex guard for driver locking in the intel_pstate driver and
eliminate some code duplication from it (Rafael Wysocki)
- Replace udelay() with usleep_range() in ACPI cpufreq (Kaushlendra
Kumar)
- Minor improvements to various cpufreq drivers (Christian Marangi,
Hal Feng, Jie Zhan, Marco Crivellari, Miaoqian Lin, and Shuhao Fu)
- Replace snprintf() with scnprintf() in show_trace_dev_match()
(Kaushlendra Kumar)
- Fix memory allocation error handling in pm_vt_switch_required()
(Malaya Kumar Rout)
- Introduce CALL_PM_OP() macro and use it to simplify code in generic
PM operations (Kaushlendra Kumar)
- Add module param to backtrace all CPUs in the device power
management watchdog (Sergey Senozhatsky)
- Rework message printing in swsusp_save() (Rafael Wysocki)
- Make it possible to change the number of hibernation compression
threads (Xueqin Luo)
- Clarify that only cgroup1 freezer uses PM freezer (Tejun Heo)
- Add document on debugging shutdown hangs to PM documentation and
correct a mistaken configuration option in it (Mario Limonciello)
- Shut down wakeup source timer before removing the wakeup source
from the list (Kaushlendra Kumar, Rafael Wysocki)
- Introduce new PMSG_POWEROFF event for system shutdown handling with
the help of PM device callbacks (Mario Limonciello)
- Make pm_test delay interruptible by wakeup events (Riwen Lu)
- Clean up kernel-doc comment style usage in the core hibernation
code and remove unuseful comments from it (Sunday Adelodun, Rafael
Wysocki)
- Add support for handling wakeup events and aborting the suspend
process while it is syncing file systems (Samuel Wu, Rafael
Wysocki)
- Add WQ_UNBOUND to pm_wq workqueue (Marco Crivellari)
- Add runtime PM wrapper macros for ACQUIRE()/ACQUIRE_ERR() and use
them in the PCI core and the ACPI TAD driver (Rafael Wysocki)
- Improve runtime PM in the ACPI TAD driver (Rafael Wysocki)
- Update pm_runtime_allow/forbid() documentation (Rafael Wysocki)
- Fix typos in runtime.c comments (Malaya Kumar Rout)
- Move governor.h from devfreq under include/linux/ and rename to
devfreq-governor.h to allow devfreq governor definitions in out of
drivers/devfreq/ (Dmitry Baryshkov)
- Use min() to improve readability in tegra30-devfreq.c (Thorsten
Blum)
- Fix potential use-after-free issue of OPP handling in
hisi_uncore_freq.c (Pengjie Zhang)
- Fix typo in DFSO_DOWNDIFFERENTIAL macro name in
governor_simpleondemand.c in devfreq (Riwen Lu)"
* tag 'pm-6.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (96 commits)
PM / devfreq: Fix typo in DFSO_DOWNDIFFERENTIAL macro name
cpuidle: Warn instead of bailing out if target residency check fails
cpuidle: Update header inclusion
Documentation: power/cpuidle: Document the CPU system wakeup latency QoS
cpuidle: Respect the CPU system wakeup QoS limit for cpuidle
sched: idle: Respect the CPU system wakeup QoS limit for s2idle
pmdomain: Respect the CPU system wakeup QoS limit for cpuidle
pmdomain: Respect the CPU system wakeup QoS limit for s2idle
PM: QoS: Introduce a CPU system wakeup QoS limit
cpuidle: governors: teo: Add missing space to the description
PM: hibernate: Extra cleanup of comments in swap handling code
PM / devfreq: tegra30: use min to simplify actmon_cpu_to_emc_rate
PM / devfreq: hisi: Fix potential UAF in OPP handling
PM / devfreq: Move governor.h to a public header location
powercap: intel_rapl: Enable MSR-based RAPL PMU support
powercap: intel_rapl: Prepare read_raw() interface for atomic-context callers
cpufreq: qcom-nvmem: fix compilation warning for qcom_cpufreq_ipq806x_match_list
PM: sleep: Call pm_sleep_fs_sync() instead of ksys_sync_helper()
PM: sleep: Add support for wakeup during filesystem sync
cpufreq: ACPI: Replace udelay() with usleep_range()
...
Pull timer core updates from Thomas Gleixner:
- Prevent a thundering herd problem when the timekeeper CPU is delayed
and a large number of CPUs compete to acquire jiffies_lock to do the
update. Limit it to one CPU with a separate "uncontended" atomic
variable.
- A set of improvements for the timer migration mechanism:
- Support imbalanced NUMA trees correctly
- Support dynamic exclusion of CPUs from the migrator duty to allow
the cpuset/isolation mechanism to exclude them from handling
timers of remote idle CPUs
- The usual small updates, cleanups and enhancements
* tag 'timers-core-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
timers/migration: Exclude isolated cpus from hierarchy
cpumask: Add initialiser to use cleanup helpers
sched/isolation: Force housekeeping if isolcpus and nohz_full don't leave any
cgroup/cpuset: Rename update_unbound_workqueue_cpumask() to update_isolation_cpumasks()
timers/migration: Use scoped_guard on available flag set/clear
timers/migration: Add mask for CPUs available in the hierarchy
timers/migration: Rename 'online' bit to 'available'
selftests/timers/nanosleep: Add tests for return of remaining time
selftests/timers: Clean up kernel version check in posix_timers
time: Fix a few typos in time[r] related code comments
time: tick-oneshot: Add missing Return and parameter descriptions to kernel-doc
hrtimer: Store time as ktime_t in restart block
timers/migration: Remove dead code handling idle CPU checking for remote timers
timers/migration: Remove unused "cpu" parameter from tmigr_get_group()
timers/migration: Assert that hotplug preparing CPU is part of stable active hierarchy
timers/migration: Fix imbalanced NUMA trees
timers/migration: Remove locking on group connection
timers/migration: Convert "while" loops to use "for"
tick/sched: Limit non-timekeeper CPUs calling jiffies update
Pull irq core updates from Thomas Gleixner:
"Updates for the interrupt core and treewide cleanups:
- Rework of the Per Processor Interrupt (PPI) management on ARM[64]
PPI support was built under the assumption that the systems are
homogenous so that the same CPU local device types are connected to
them. That's unfortunately wishful thinking and created horrible
workarounds.
This rework provides affinity management for PPIs so that they can
be individually configured in the firmware tables and mops up the
related drivers all over the place.
- Prevent CPUSET/isolation changes to arbitrarily affine interrupt
threads to random CPUs, which ignores user or driver settings.
- Plug a harmless race in the interrupt affinity proc interface,
which allows to see a half updated mask
- Adjust the priority of secondary interrupt threads on RT, so that
the combination of primary and secondary thread emulates the
hardware interrupt plus thread scenario. Having them at the same
priority can cause starvation issues in some drivers"
* tag 'irq-core-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (33 commits)
genirq: Remove cpumask availability check on kthread affinity setting
genirq: Fix interrupt threads affinity vs. cpuset isolated partitions
genirq: Prevent early spurious wake-ups of interrupt threads
genirq: Use raw_spinlock_irq() in irq_set_affinity_notifier()
genirq/manage: Reduce priority of forced secondary interrupt handler
genirq/proc: Fix race in show_irq_affinity()
genirq: Fix percpu_devid irq affinity documentation
perf: arm_pmu: Kill last use of per-CPU cpu_armpmu pointer
irqdomain: Kill of_node_to_fwnode() helper
genirq: Kill irq_{g,s}et_percpu_devid_partition()
irqchip: Kill irq-partition-percpu
irqchip/apple-aic: Drop support for custom PMU irq partitions
irqchip/gic-v3: Drop support for custom PPI partitions
coresight: trbe: Request specific affinities for per CPU interrupts
perf: arm_spe_pmu: Request specific affinities for per CPU interrupts
perf: arm_pmu: Request specific affinities for per CPU NMIs/interrupts
genirq: Add request_percpu_irq_affinity() helper
genirq: Allow per-cpu interrupt sharing for non-overlapping affinities
genirq: Update request_percpu_nmi() to take an affinity
genirq: Add affinity to percpu_devid interrupt requests
...
Pull rseq updates from Thomas Gleixner:
"A large overhaul of the restartable sequences and CID management:
The recent enablement of RSEQ in glibc resulted in regressions which
are caused by the related overhead. It turned out that the decision to
invoke the exit to user work was not really a decision. More or less
each context switch caused that. There is a long list of small issues
which sums up nicely and results in a 3-4% regression in I/O
benchmarks.
The other detail which caused issues due to extra work in context
switch and task migration is the CID (memory context ID) management.
It also requires to use a task work to consolidate the CID space,
which is executed in the context of an arbitrary task and results in
sporadic uncontrolled exit latencies.
The rewrite addresses this by:
- Removing deprecated and long unsupported functionality
- Moving the related data into dedicated data structures which are
optimized for fast path processing.
- Caching values so actual decisions can be made
- Replacing the current implementation with a optimized inlined
variant.
- Separating fast and slow path for architectures which use the
generic entry code, so that only fault and error handling goes into
the TIF_NOTIFY_RESUME handler.
- Rewriting the CID management so that it becomes mostly invisible in
the context switch path. That moves the work of switching modes
into the fork/exit path, which is a reasonable tradeoff. That work
is only required when a process creates more threads than the
cpuset it is allowed to run on or when enough threads exit after
that. An artificial thread pool benchmarks which triggers this did
not degrade, it actually improved significantly.
The main effect in migration heavy scenarios is that runqueue lock
held time and therefore contention goes down significantly"
* tag 'core-rseq-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (54 commits)
sched/mmcid: Switch over to the new mechanism
sched/mmcid: Implement deferred mode change
irqwork: Move data struct to a types header
sched/mmcid: Provide CID ownership mode fixup functions
sched/mmcid: Provide new scheduler CID mechanism
sched/mmcid: Introduce per task/CPU ownership infrastructure
sched/mmcid: Serialize sched_mm_cid_fork()/exit() with a mutex
sched/mmcid: Provide precomputed maximal value
sched/mmcid: Move initialization out of line
signal: Move MMCID exit out of sighand lock
sched/mmcid: Convert mm CID mask to a bitmap
cpumask: Cache num_possible_cpus()
sched/mmcid: Use cpumask_weighted_or()
cpumask: Introduce cpumask_weighted_or()
sched/mmcid: Prevent pointless work in mm_update_cpus_allowed()
sched/mmcid: Move scheduler code out of global header
sched: Fixup whitespace damage
sched/mmcid: Cacheline align MM CID storage
sched/mmcid: Use proper data structures
sched/mmcid: Revert the complex CID management
...
Pull scheduler updates from Ingo Molnar:
"Scalability and load-balancing improvements:
- Enable scheduler feature NEXT_BUDDY (Mel Gorman)
- Reimplement NEXT_BUDDY to align with EEVDF goals (Mel Gorman)
- Skip sched_balance_running cmpxchg when balance is not due (Tim
Chen)
- Implement generic code for architecture specific sched domain NUMA
distances (Tim Chen)
- Optimize the NUMA distances of the sched-domains builds of Intel
Granite Rapids (GNR) and Clearwater Forest (CWF) platforms (Tim
Chen)
- Implement proportional newidle balance: a randomized algorithm that
runs newidle balancing proportional to its success rate. (Peter
Zijlstra)
Scheduler infrastructure changes:
- Implement the 'sched_change' scoped_guard() pattern for the entire
scheduler (Peter Zijlstra)
- More broadly utilize the sched_change guard (Peter Zijlstra)
- Add support to pick functions to take runqueue-flags (Joel
Fernandes)
- Provide and use set_need_resched_current() (Peter Zijlstra)
Fair scheduling enhancements:
- Forfeit vruntime on yield (Fernand Sieber)
- Only update stats for allowed CPUs when looking for dst group (Adam
Li)
CPU-core scheduling enhancements:
- Optimize core cookie matching check (Fernand Sieber)
Deadline scheduler fixes:
- Only set free_cpus for online runqueues (Doug Berger)
- Fix dl_server time accounting (Peter Zijlstra)
- Fix dl_server stop condition (Peter Zijlstra)
Proxy scheduling fixes:
- Yield the donor task (Fernand Sieber)
Fixes and cleanups:
- Fix do_set_cpus_allowed() locking (Peter Zijlstra)
- Fix migrate_disable_switch() locking (Peter Zijlstra)
- Remove double update_rq_clock() in __set_cpus_allowed_ptr_locked()
(Hao Jia)
- Increase sched_tick_remote timeout (Phil Auld)
- sched/deadline: Use cpumask_weight_and() in dl_bw_cpus() (Shrikanth
Hegde)
- sched/deadline: Clean up select_task_rq_dl() (Shrikanth Hegde)"
* tag 'sched-core-2025-12-01' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (44 commits)
sched: Provide and use set_need_resched_current()
sched/fair: Proportional newidle balance
sched/fair: Small cleanup to update_newidle_cost()
sched/fair: Small cleanup to sched_balance_newidle()
sched/fair: Revert max_newidle_lb_cost bump
sched/fair: Reimplement NEXT_BUDDY to align with EEVDF goals
sched/fair: Enable scheduler feature NEXT_BUDDY
sched: Increase sched_tick_remote timeout
sched/fair: Have SD_SERIALIZE affect newidle balancing
sched/fair: Skip sched_balance_running cmpxchg when balance is not due
sched/deadline: Minor cleanup in select_task_rq_dl()
sched/deadline: Use cpumask_weight_and() in dl_bw_cpus
sched/deadline: Document dl_server
sched/deadline: Fix dl_server stop condition
sched/deadline: Fix dl_server time accounting
sched/core: Remove double update_rq_clock() in __set_cpus_allowed_ptr_locked()
sched/eevdf: Fix min_vruntime vs avg_vruntime
sched/core: Add comment explaining force-idle vruntime snapshots
sched/core: Optimize core cookie matching check
sched/proxy: Yield the donor task
...
Pull locking updates from Ingo Molnar:
"Mutexes:
- Redo __mutex_init() to reduce generated code size (Sebastian
Andrzej Siewior)
Seqlocks:
- Introduce scoped_seqlock_read() (Peter Zijlstra)
- Change thread_group_cputime() to use scoped_seqlock_read() (Oleg
Nesterov)
- Change do_task_stat() to use scoped_seqlock_read() (Oleg Nesterov)
- Change do_io_accounting() to use scoped_seqlock_read() (Oleg
Nesterov)
- Fix the incorrect documentation of read_seqbegin_or_lock() /
need_seqretry() (Oleg Nesterov)
- Allow KASAN to fail optimizing (Peter Zijlstra)
Local lock updates:
- Fix all kernel-doc warnings (Randy Dunlap)
- Add the <linux/local_lock*.h> headers to MAINTAINERS (Sebastian
Andrzej Siewior)
- Reduce the risk of shadowing via s/l/__l/ and s/tl/__tl/ (Vincent
Mailhol)
Lock debugging:
- spinlock/debug: Fix data-race in do_raw_write_lock (Alexander
Sverdlin)
Atomic primitives infrastructure:
- atomic: Skip alignment check for try_cmpxchg() old arg (Arnd
Bergmann)
Rust runtime integration:
- sync: atomic: Enable generated Atomic<T> usage (Boqun Feng)
- sync: atomic: Implement Debug for Atomic<Debug> (Boqun Feng)
- debugfs: Remove Rust native atomics and replace them with Linux
versions (Boqun Feng)
- debugfs: Implement Reader for Mutex<T> only when T is Unpin (Boqun
Feng)
- lock: guard: Add T: Unpin bound to DerefMut (Daniel Almeida)
- lock: Pin the inner data (Daniel Almeida)
- lock: Add a Pin<&mut T> accessor (Daniel Almeida)"
* tag 'locking-core-2025-12-01' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
locking/local_lock: Fix all kernel-doc warnings
locking/local_lock: s/l/__l/ and s/tl/__tl/ to reduce the risk of shadowing
locking/local_lock: Add the <linux/local_lock*.h> headers to MAINTAINERS
locking/mutex: Redo __mutex_init() to reduce generated code size
rust: debugfs: Replace the usage of Rust native atomics
rust: sync: atomic: Implement Debug for Atomic<Debug>
rust: sync: atomic: Make Atomic*Ops pub(crate)
seqlock: Allow KASAN to fail optimizing
rust: debugfs: Implement Reader for Mutex<T> only when T is Unpin
seqlock: Change do_io_accounting() to use scoped_seqlock_read()
seqlock: Change do_task_stat() to use scoped_seqlock_read()
seqlock: Change thread_group_cputime() to use scoped_seqlock_read()
seqlock: Introduce scoped_seqlock_read()
documentation: seqlock: fix the wrong documentation of read_seqbegin_or_lock/need_seqretry
atomic: Skip alignment check for try_cmpxchg() old arg
rust: lock: Add a Pin<&mut T> accessor
rust: lock: Pin the inner data
rust: lock: guard: Add T: Unpin bound to DerefMut
locking/spinlock/debug: Fix data-race in do_raw_write_lock
When loading the ebpf scheduler, the tasks in the scx_tasks list will
be traversed and invoke __setscheduler_class() to get new sched_class.
however, this would also incorrectly set the per-cpu migration
task's->sched_class to rt_sched_class, even after unload, the per-cpu
migration task's->sched_class remains sched_rt_class.
The log for this issue is as follows:
./scx_rustland --stats 1
[ 199.245639][ T630] sched_ext: "rustland" does not implement cgroup cpu.weight
[ 199.269213][ T630] sched_ext: BPF scheduler "rustland" enabled
04:25:09 [INFO] RustLand scheduler attached
bpftrace -e 'iter:task /strcontains(ctx->task->comm, "migration")/
{ printf("%s:%d->%pS\n", ctx->task->comm, ctx->task->pid, ctx->task->sched_class); }'
Attaching 1 probe...
migration/0:24->rt_sched_class+0x0/0xe0
migration/1:27->rt_sched_class+0x0/0xe0
migration/2:33->rt_sched_class+0x0/0xe0
migration/3:39->rt_sched_class+0x0/0xe0
migration/4:45->rt_sched_class+0x0/0xe0
migration/5:52->rt_sched_class+0x0/0xe0
migration/6:58->rt_sched_class+0x0/0xe0
migration/7:64->rt_sched_class+0x0/0xe0
sched_ext: BPF scheduler "rustland" disabled (unregistered from user space)
EXIT: unregistered from user space
04:25:21 [INFO] Unregister RustLand scheduler
bpftrace -e 'iter:task /strcontains(ctx->task->comm, "migration")/
{ printf("%s:%d->%pS\n", ctx->task->comm, ctx->task->pid, ctx->task->sched_class); }'
Attaching 1 probe...
migration/0:24->rt_sched_class+0x0/0xe0
migration/1:27->rt_sched_class+0x0/0xe0
migration/2:33->rt_sched_class+0x0/0xe0
migration/3:39->rt_sched_class+0x0/0xe0
migration/4:45->rt_sched_class+0x0/0xe0
migration/5:52->rt_sched_class+0x0/0xe0
migration/6:58->rt_sched_class+0x0/0xe0
migration/7:64->rt_sched_class+0x0/0xe0
This commit therefore generate a new scx_setscheduler_class() and
add check for stop_sched_class to replace __setscheduler_class().
Fixes: f0e1a0643a ("sched_ext: Implement BPF extensible scheduler class")
Cc: stable@vger.kernel.org # v6.12+
Signed-off-by: Zqiang <qiang.zhang@linux.dev>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Now that all pieces are in place, change the implementations of
sched_mm_cid_fork() and sched_mm_cid_exit() to adhere to the new strict
ownership scheme and switch context_switch() over to use the new
mm_cid_schedin() functionality.
The common case is that there is no mode change required, which makes
fork() and exit() just update the user count and the constraints.
In case that a new user would exceed the CID space limit the fork() context
handles the transition to per CPU mode with mm::mm_cid::mutex held. exit()
handles the transition back to per task mode when the user count drops
below the switch back threshold. fork() might also be forced to handle a
deferred switch back to per task mode, when a affinity change increased the
number of allowed CPUs enough.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172550.280380631@linutronix.de
When affinity changes cause an increase of the number of CPUs allowed for
tasks which are related to a MM, that might results in a situation where
the ownership mode can go back from per CPU mode to per task mode.
As affinity changes happen with runqueue lock held there is no way to do
the actual mode change and required fixup right there.
Add the infrastructure to defer it to a workqueue. The scheduled work can
race with a fork() or exit(). Whatever happens first takes care of it.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172550.216484739@linutronix.de
CIDs are either owned by tasks or by CPUs. The ownership mode depends on
the number of tasks related to a MM and the number of CPUs on which these
tasks are theoretically allowed to run on. Theoretically because that
number is the superset of CPU affinities of all tasks which only grows and
never shrinks.
Switching to per CPU mode happens when the user count becomes greater than
the maximum number of CIDs, which is calculated by:
opt_cids = min(mm_cid::nr_cpus_allowed, mm_cid::users);
max_cids = min(1.25 * opt_cids, nr_cpu_ids);
The +25% allowance is useful for tight CPU masks in scenarios where only a
few threads are created and destroyed to avoid frequent mode
switches. Though this allowance shrinks, the closer opt_cids becomes to
nr_cpu_ids, which is the (unfortunate) hard ABI limit.
At the point of switching to per CPU mode the new user is not yet visible
in the system, so the task which initiated the fork() runs the fixup
function: mm_cid_fixup_tasks_to_cpu() walks the thread list and either
transfers each tasks owned CID to the CPU the task runs on or drops it into
the CID pool if a task is not on a CPU at that point in time. Tasks which
schedule in before the task walk reaches them do the handover in
mm_cid_schedin(). When mm_cid_fixup_tasks_to_cpus() completes it's
guaranteed that no task related to that MM owns a CID anymore.
Switching back to task mode happens when the user count goes below the
threshold which was recorded on the per CPU mode switch:
pcpu_thrs = min(opt_cids - (opt_cids / 4), nr_cpu_ids / 2);
This threshold is updated when a affinity change increases the number of
allowed CPUs for the MM, which might cause a switch back to per task mode.
If the switch back was initiated by a exiting task, then that task runs the
fixup function. If it was initiated by a affinity change, then it's run
either in the deferred update function in context of a workqueue or by a
task which forks a new one or by a task which exits. Whatever happens
first. mm_cid_fixup_cpus_to_task() walks through the possible CPUs and
either transfers the CPU owned CIDs to a related task which runs on the CPU
or drops it into the pool. Tasks which schedule in on a CPU which the walk
did not cover yet do the handover themselves.
This transition from CPU to per task ownership happens in two phases:
1) mm:mm_cid.transit contains MM_CID_TRANSIT. This is OR'ed on the task
CID and denotes that the CID is only temporarily owned by the
task. When it schedules out the task drops the CID back into the
pool if this bit is set.
2) The initiating context walks the per CPU space and after completion
clears mm:mm_cid.transit. After that point the CIDs are strictly
task owned again.
This two phase transition is required to prevent CID space exhaustion
during the transition as a direct transfer of ownership would fail if
two tasks are scheduled in on the same CPU before the fixup freed per
CPU CIDs.
When mm_cid_fixup_cpus_to_tasks() completes it's guaranteed that no CID
related to that MM is owned by a CPU anymore.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172550.088189028@linutronix.de
The MM CID management has two fundamental requirements:
1) It has to guarantee that at no given point in time the same CID is
used by concurrent tasks in userspace.
2) The CID space must not exceed the number of possible CPUs in a
system. While most allocators (glibc, tcmalloc, jemalloc) do not
care about that, there seems to be at least some LTTng library
depending on it.
The CID space compaction itself is not a functional correctness
requirement, it is only a useful optimization mechanism to reduce the
memory foot print in unused user space pools.
The optimal CID space is:
min(nr_tasks, nr_cpus_allowed);
Where @nr_tasks is the number of actual user space threads associated to
the mm and @nr_cpus_allowed is the superset of all task affinities. It is
growth only as it would be insane to take a racy snapshot of all task
affinities when the affinity of one task changes just do redo it 2
milliseconds later when the next task changes it's affinity.
That means that as long as the number of tasks is lower or equal than the
number of CPUs allowed, each task owns a CID. If the number of tasks
exceeds the number of CPUs allowed it switches to per CPU mode, where the
CPUs own the CIDs and the tasks borrow them as long as they are scheduled
in.
For transition periods CIDs can go beyond the optimal space as long as they
don't go beyond the number of possible CPUs.
The current upstream implementation adds overhead into task migration to
keep the CID with the task. It also has to do the CID space consolidation
work from a task work in the exit to user space path. As that work is
assigned to a random task related to a MM this can inflict unwanted exit
latencies.
Implement the context switch parts of a strict ownership mechanism to
address this.
This removes most of the work from the task which schedules out. Only
during transitioning from per CPU to per task ownership it is required to
drop the CID when leaving the CPU to prevent CID space exhaustion. Other
than that scheduling out is just a single check and branch.
The task which schedules in has to check whether:
1) The ownership mode changed
2) The CID is within the optimal CID space
In stable situations this results in zero work. The only short disruption
is when ownership mode changes or when the associated CID is not in the
optimal CID space. The latter only happens when tasks exit and therefore
the optimal CID space shrinks.
That mechanism is strictly optimized for the common case where no change
happens. The only case where it actually causes a temporary one time spike
is on mode changes when and only when a lot of tasks related to a MM
schedule exactly at the same time and have eventually to compete on
allocating a CID from the bitmap.
In the sysbench test case which triggered the spinlock contention in the
initial CID code, __schedule() drops significantly in perf top on a 128
Core (256 threads) machine when running sysbench with 255 threads, which
fits into the task mode limit of 256 together with the parent thread:
Upstream rseq/perf branch +CID rework
0.42% 0.37% 0.32% [k] __schedule
Increasing the number of threads to 256, which puts the test process into
per CPU mode looks about the same.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172550.023984859@linutronix.de
The MM CID management has two fundamental requirements:
1) It has to guarantee that at no given point in time the same CID is
used by concurrent tasks in userspace.
2) The CID space must not exceed the number of possible CPUs in a
system. While most allocators (glibc, tcmalloc, jemalloc) do not care
about that, there seems to be at least librseq depending on it.
The CID space compaction itself is not a functional correctness
requirement, it is only a useful optimization mechanism to reduce the
memory foot print in unused user space pools.
The optimal CID space is:
min(nr_tasks, nr_cpus_allowed);
Where @nr_tasks is the number of actual user space threads associated to
the mm and @nr_cpus_allowed is the superset of all task affinities. It is
growth only as it would be insane to take a racy snapshot of all task
affinities when the affinity of one task changes just do redo it 2
milliseconds later when the next task changes its affinity.
That means that as long as the number of tasks is lower or equal than the
number of CPUs allowed, each task owns a CID. If the number of tasks
exceeds the number of CPUs allowed it switches to per CPU mode, where the
CPUs own the CIDs and the tasks borrow them as long as they are scheduled
in.
For transition periods CIDs can go beyond the optimal space as long as they
don't go beyond the number of possible CPUs.
The current upstream implementation adds overhead into task migration to
keep the CID with the task. It also has to do the CID space consolidation
work from a task work in the exit to user space path. As that work is
assigned to a random task related to a MM this can inflict unwanted exit
latencies.
This can be done differently by implementing a strict CID ownership
mechanism. Either the CIDs are owned by the tasks or by the CPUs. The
latter provides less locality when tasks are heavily migrating, but there
is no justification to optimize for overcommit scenarios and thereby
penalizing everyone else.
Provide the basic infrastructure to implement this:
- Change the UNSET marker to BIT(31) from ~0U
- Add the ONCPU marker as BIT(30)
- Add the TRANSIT marker as BIT(29)
That allows to check for ownership trivially and provides a simple check for
UNSET as well. The TRANSIT marker is required to prevent CID space
exhaustion when switching from per CPU to per task mode.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251119172549.960252358@linutronix.de
Prepare for the new CID management scheme which puts the CID ownership
transition into the fork() and exit() slow path by serializing
sched_mm_cid_fork()/exit() with it, so task list and cpu mask walks can be
done in interruptible and preemptible code.
The contention on it is not worse than on other concurrency controls in the
fork()/exit() machinery.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172549.895826703@linutronix.de
Reading mm::mm_users and mm:::mm_cid::nr_cpus_allowed every time to compute
the maximal CID value is just wasteful as that value is only changing on
fork(), exit() and eventually when the affinity changes.
So it can be easily precomputed at those points and provided in mm::mm_cid
for consumption in the hot path.
But there is an issue with using mm::mm_users for accounting because that
does not necessarily reflect the number of user space tasks as other kernel
code can take temporary references on the MM which skew the picture.
Solve that by adding a users counter to struct mm_mm_cid, which is modified
by fork() and exit() and used for precomputing under mm_mm_cid::lock.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172549.832764634@linutronix.de
Currently the user can set up isolcpus and nohz_full in such a way that
leaves no housekeeping CPU (i.e. no CPU that is neither domain isolated
nor nohz full). This can be a problem for other subsystems (e.g. the
timer wheel imgration).
Prevent this configuration by invalidating the last setting in case the
union of isolcpus (domain) and nohz_full covers all CPUs.
Signed-off-by: Gabriele Monaco <gmonaco@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Waiman Long <longman@redhat.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://patch.msgid.link/20251120145653.296659-6-gmonaco@redhat.com
Pull sched_ext fix from Tejun Heo:
"One low risk and obvious fix: scx_enable() was dereferencing an error
pointer on helper kthread creation failure. Fixed"
* tag 'sched_ext-for-6.18-rc6-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext:
sched_ext: Fix scx_enable() crash on helper kthread creation failure
A crash was observed when the sched_ext selftests runner was
terminated with Ctrl+\ while test 15 was running:
NIP [c00000000028fa58] scx_enable.constprop.0+0x358/0x12b0
LR [c00000000028fa2c] scx_enable.constprop.0+0x32c/0x12b0
Call Trace:
scx_enable.constprop.0+0x32c/0x12b0 (unreliable)
bpf_struct_ops_link_create+0x18c/0x22c
__sys_bpf+0x23f8/0x3044
sys_bpf+0x2c/0x6c
system_call_exception+0x124/0x320
system_call_vectored_common+0x15c/0x2ec
kthread_run_worker() returns an ERR_PTR() on failure rather than NULL,
but the current code in scx_alloc_and_add_sched() only checks for a NULL
helper. Incase of failure on SIGQUIT, the error is not handled in
scx_alloc_and_add_sched() and scx_enable() ends up dereferencing an
error pointer.
Error handling is fixed in scx_alloc_and_add_sched() to propagate
PTR_ERR() into ret, so that scx_enable() jumps to the existing error
path, avoiding random dereference on failure.
Fixes: bff3b5aec1 ("sched_ext: Move disable machinery into scx_sched")
Cc: stable@vger.kernel.org # v6.16+
Reported-and-tested-by: Samir Mulani <samir@linux.ibm.com>
Signed-off-by: Saket Kumar Bhaskar <skb99@linux.ibm.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Vishal Chourasia <vishalc@linux.ibm.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
*** Bug description ***
When testing kexec-reboot on a 144 cpus machine with
isolcpus=managed_irq,domain,1-71,73-143 in kernel command line, I
encounter the following bug:
[ 97.114759] psci: CPU142 killed (polled 0 ms)
[ 97.333236] Failed to offline CPU143 - error=-16
[ 97.333246] ------------[ cut here ]------------
[ 97.342682] kernel BUG at kernel/cpu.c:1569!
[ 97.347049] Internal error: Oops - BUG: 00000000f2000800 [#1] SMP
[...]
In essence, the issue originates from the CPU hot-removal process, not
limited to kexec. It can be reproduced by writing a SCHED_DEADLINE
program that waits indefinitely on a semaphore, spawning multiple
instances to ensure some run on CPU 72, and then offlining CPUs 1–143
one by one. When attempting this, CPU 143 failed to go offline.
bash -c 'taskset -cp 0 $$ && for i in {1..143}; do echo 0 > /sys/devices/system/cpu/cpu$i/online 2>/dev/null; done'
Tracking down this issue, I found that dl_bw_deactivate() returned
-EBUSY, which caused sched_cpu_deactivate() to fail on the last CPU.
But that is not the fact, and contributed by the following factors:
When a CPU is inactive, cpu_rq()->rd is set to def_root_domain. For an
blocked-state deadline task (in this case, "cppc_fie"), it was not
migrated to CPU0, and its task_rq() information is stale. So its rq->rd
points to def_root_domain instead of the one shared with CPU0. As a
result, its bandwidth is wrongly accounted into a wrong root domain
during domain rebuild.
*** Issue ***
The key point is that root_domain is only tracked through active rq->rd.
To avoid using a global data structure to track all root_domains in the
system, there should be a method to locate an active CPU within the
corresponding root_domain.
*** Solution ***
To locate the active cpu, the following rules for deadline
sub-system is useful
-1.any cpu belongs to a unique root domain at a given time
-2.DL bandwidth checker ensures that the root domain has active cpus.
Now, let's examine the blocked-state task P.
If P is attached to a cpuset that is a partition root, it is
straightforward to find an active CPU.
If P is attached to a cpuset that has changed from 'root' to 'member',
the active CPUs are grouped into the parent root domain. Naturally, the
CPUs' capacity and reserved DL bandwidth are taken into account in the
ancestor root domain. (In practice, it may be unsafe to attach P to an
arbitrary root domain, since that domain may lack sufficient DL
bandwidth for P.) Again, it is straightforward to find an active CPU in
the ancestor root domain.
This patch groups CPUs into isolated and housekeeping sets. For the
housekeeping group, it walks up the cpuset hierarchy to find active CPUs
in P's root domain and retrieves the valid rd from cpu_rq(cpu)->rd.
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Cc: Waiman Long <longman@redhat.com>
Cc: Chen Ridong <chenridong@huaweicloud.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Pierre Gondois <pierre.gondois@arm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ben Segall <bsegall@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Valentin Schneider <vschneid@redhat.com>
To: linux-kernel@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
mm_update_cpus_allowed() is not required to be invoked for affinity changes
due to migrate_disable() and migrate_enable().
migrate_disable() restricts the task temporarily to a CPU on which the task
was already allowed to run, so nothing changes. migrate_enable() restores
the actual task affinity mask.
If that mask changed between migrate_disable() and migrate_enable() then
that change was already accounted for.
Move the invocation to the proper place to avoid that.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172549.385208276@linutronix.de
The CID management is a complex beast, which affects both scheduling and
task migration. The compaction mechanism forces random tasks of a process
into task work on exit to user space causing latency spikes.
Revert back to the initial simple bitmap allocating mechanics, which are
known to have scalability issues as that allows to gradually build up a
replacement functionality in a reviewable way.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251119172549.068197830@linutronix.de
Pull sched_ext fixes from Tejun Heo:
"Five fixes addressing PREEMPT_RT compatibility and locking issues.
Three commits fix potential deadlocks and sleeps in atomic contexts on
RT kernels by converting locks to raw spinlocks and ensuring IRQ work
runs in hard-irq context. The remaining two fix unsafe locking in the
debug dump path and a variable dereference typo"
* tag 'sched_ext-for-6.18-rc6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext:
sched_ext: Use IRQ_WORK_INIT_HARD() to initialize rq->scx.kick_cpus_irq_work
sched_ext: Fix possible deadlock in the deferred_irq_workfn()
sched/ext: convert scx_tasks_lock to raw spinlock
sched_ext: Fix unsafe locking in the scx_dump_state()
sched_ext: Fix use of uninitialized variable in scx_bpf_cpuperf_set()
Reimplement NEXT_BUDDY preemption to take into account the deadline and
eligibility of the wakee with respect to the waker. In the event
multiple buddies could be considered, the one with the earliest deadline
is selected.
Sync wakeups are treated differently to every other type of wakeup. The
WF_SYNC assumption is that the waker promises to sleep in the very near
future. This is violated in enough cases that WF_SYNC should be treated
as a suggestion instead of a contract. If a waker does go to sleep almost
immediately then the delay in wakeup is negligible. In other cases, it's
throttled based on the accumulated runtime of the waker so there is a
chance that some batched wakeups have been issued before preemption.
For all other wakeups, preemption happens if the wakee has a earlier
deadline than the waker and eligible to run.
While many workloads were tested, the two main targets were a modified
dbench4 benchmark and hackbench because the are on opposite ends of the
spectrum -- one prefers throughput by avoiding preemption and the other
relies on preemption.
First is the dbench throughput data even though it is a poor metric but
it is the default metric. The test machine is a 2-socket machine and the
backing filesystem is XFS as a lot of the IO work is dispatched to kernel
threads. It's important to note that these results are not representative
across all machines, especially Zen machines, as different bottlenecks
are exposed on different machines and filesystems.
dbench4 Throughput (misleading but traditional)
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v5
Hmean 1 1268.80 ( 0.00%) 1269.74 ( 0.07%)
Hmean 4 3971.74 ( 0.00%) 3950.59 ( -0.53%)
Hmean 7 5548.23 ( 0.00%) 5420.08 ( -2.31%)
Hmean 12 7310.86 ( 0.00%) 7165.57 ( -1.99%)
Hmean 21 8874.53 ( 0.00%) 9149.04 ( 3.09%)
Hmean 30 9361.93 ( 0.00%) 10530.04 ( 12.48%)
Hmean 48 9540.14 ( 0.00%) 11820.40 ( 23.90%)
Hmean 79 9208.74 ( 0.00%) 12193.79 ( 32.42%)
Hmean 110 8573.12 ( 0.00%) 11933.72 ( 39.20%)
Hmean 141 7791.33 ( 0.00%) 11273.90 ( 44.70%)
Hmean 160 7666.60 ( 0.00%) 10768.72 ( 40.46%)
As throughput is misleading, the benchmark is modified to use a short
loadfile report the completion time duration in milliseconds.
dbench4 Loadfile Execution Time
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v5
Amean 1 14.62 ( 0.00%) 14.69 ( -0.46%)
Amean 4 18.76 ( 0.00%) 18.85 ( -0.45%)
Amean 7 23.71 ( 0.00%) 24.38 ( -2.82%)
Amean 12 31.25 ( 0.00%) 31.87 ( -1.97%)
Amean 21 45.12 ( 0.00%) 43.69 ( 3.16%)
Amean 30 61.07 ( 0.00%) 54.33 ( 11.03%)
Amean 48 95.91 ( 0.00%) 77.22 ( 19.49%)
Amean 79 163.38 ( 0.00%) 123.08 ( 24.66%)
Amean 110 243.91 ( 0.00%) 175.11 ( 28.21%)
Amean 141 343.47 ( 0.00%) 239.10 ( 30.39%)
Amean 160 401.15 ( 0.00%) 283.73 ( 29.27%)
Stddev 1 0.52 ( 0.00%) 0.51 ( 2.45%)
Stddev 4 1.36 ( 0.00%) 1.30 ( 4.04%)
Stddev 7 1.88 ( 0.00%) 1.87 ( 0.72%)
Stddev 12 3.06 ( 0.00%) 2.45 ( 19.83%)
Stddev 21 5.78 ( 0.00%) 3.87 ( 33.06%)
Stddev 30 9.85 ( 0.00%) 5.25 ( 46.76%)
Stddev 48 22.31 ( 0.00%) 8.64 ( 61.27%)
Stddev 79 35.96 ( 0.00%) 18.07 ( 49.76%)
Stddev 110 59.04 ( 0.00%) 30.93 ( 47.61%)
Stddev 141 85.38 ( 0.00%) 40.93 ( 52.06%)
Stddev 160 96.38 ( 0.00%) 39.72 ( 58.79%)
That is still looking good and the variance is reduced quite a bit.
Finally, fairness is a concern so the next report tracks how many
milliseconds does it take for all clients to complete a workfile. This
one is tricky because dbench makes to effort to synchronise clients so
the durations at benchmark start time differ substantially from typical
runtimes. This problem could be mitigated by warming up the benchmark
for a number of minutes but it's a matter of opinion whether that
counts as an evasion of inconvenient results.
dbench4 All Clients Loadfile Execution Time
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v5
Amean 1 15.06 ( 0.00%) 15.07 ( -0.03%)
Amean 4 603.81 ( 0.00%) 524.29 ( 13.17%)
Amean 7 855.32 ( 0.00%) 1331.07 ( -55.62%)
Amean 12 1890.02 ( 0.00%) 2323.97 ( -22.96%)
Amean 21 3195.23 ( 0.00%) 2009.29 ( 37.12%)
Amean 30 13919.53 ( 0.00%) 4579.44 ( 67.10%)
Amean 48 25246.07 ( 0.00%) 5705.46 ( 77.40%)
Amean 79 29701.84 ( 0.00%) 15509.26 ( 47.78%)
Amean 110 22803.03 ( 0.00%) 23782.08 ( -4.29%)
Amean 141 36356.07 ( 0.00%) 25074.20 ( 31.03%)
Amean 160 17046.71 ( 0.00%) 13247.62 ( 22.29%)
Stddev 1 0.47 ( 0.00%) 0.49 ( -3.74%)
Stddev 4 395.24 ( 0.00%) 254.18 ( 35.69%)
Stddev 7 467.24 ( 0.00%) 764.42 ( -63.60%)
Stddev 12 1071.43 ( 0.00%) 1395.90 ( -30.28%)
Stddev 21 1694.50 ( 0.00%) 1204.89 ( 28.89%)
Stddev 30 7945.63 ( 0.00%) 2552.59 ( 67.87%)
Stddev 48 14339.51 ( 0.00%) 3227.55 ( 77.49%)
Stddev 79 16620.91 ( 0.00%) 8422.15 ( 49.33%)
Stddev 110 12912.15 ( 0.00%) 13560.95 ( -5.02%)
Stddev 141 20700.13 ( 0.00%) 14544.51 ( 29.74%)
Stddev 160 9079.16 ( 0.00%) 7400.69 ( 18.49%)
This is more of a mixed bag but it at least shows that fairness
is not crippled.
The hackbench results are more neutral but this is still important.
It's possible to boost the dbench figures by a large amount but only by
crippling the performance of a workload like hackbench. The WF_SYNC
behaviour is important for these workloads and is why the WF_SYNC
changes are not a separate patch.
hackbench-process-pipes
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v5
Amean 1 0.2657 ( 0.00%) 0.2150 ( 19.07%)
Amean 4 0.6107 ( 0.00%) 0.6060 ( 0.76%)
Amean 7 0.7923 ( 0.00%) 0.7440 ( 6.10%)
Amean 12 1.1500 ( 0.00%) 1.1263 ( 2.06%)
Amean 21 1.7950 ( 0.00%) 1.7987 ( -0.20%)
Amean 30 2.3207 ( 0.00%) 2.5053 ( -7.96%)
Amean 48 3.5023 ( 0.00%) 3.9197 ( -11.92%)
Amean 79 4.8093 ( 0.00%) 5.2247 ( -8.64%)
Amean 110 6.1160 ( 0.00%) 6.6650 ( -8.98%)
Amean 141 7.4763 ( 0.00%) 7.8973 ( -5.63%)
Amean 172 8.9560 ( 0.00%) 9.3593 ( -4.50%)
Amean 203 10.4783 ( 0.00%) 10.8347 ( -3.40%)
Amean 234 12.4977 ( 0.00%) 13.0177 ( -4.16%)
Amean 265 14.7003 ( 0.00%) 15.5630 ( -5.87%)
Amean 296 16.1007 ( 0.00%) 17.4023 ( -8.08%)
Processes using pipes are impacted but the variance (not presented) indicates
it's close to noise and the results are not always reproducible. If executed
across multiple reboots, it may show neutral or small gains so the worst
measured results are presented.
Hackbench using sockets is more reliably neutral as the wakeup
mechanisms are different between sockets and pipes.
hackbench-process-sockets
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v2
Amean 1 0.3073 ( 0.00%) 0.3263 ( -6.18%)
Amean 4 0.7863 ( 0.00%) 0.7930 ( -0.85%)
Amean 7 1.3670 ( 0.00%) 1.3537 ( 0.98%)
Amean 12 2.1337 ( 0.00%) 2.1903 ( -2.66%)
Amean 21 3.4683 ( 0.00%) 3.4940 ( -0.74%)
Amean 30 4.7247 ( 0.00%) 4.8853 ( -3.40%)
Amean 48 7.6097 ( 0.00%) 7.8197 ( -2.76%)
Amean 79 14.7957 ( 0.00%) 16.1000 ( -8.82%)
Amean 110 21.3413 ( 0.00%) 21.9997 ( -3.08%)
Amean 141 29.0503 ( 0.00%) 29.0353 ( 0.05%)
Amean 172 36.4660 ( 0.00%) 36.1433 ( 0.88%)
Amean 203 39.7177 ( 0.00%) 40.5910 ( -2.20%)
Amean 234 42.1120 ( 0.00%) 43.5527 ( -3.42%)
Amean 265 45.7830 ( 0.00%) 50.0560 ( -9.33%)
Amean 296 50.7043 ( 0.00%) 54.3657 ( -7.22%)
As schbench has been mentioned in numerous bugs recently, the results
are interesting. A test case that represents the default schbench
behaviour is
schbench Wakeup Latency (usec)
6.18.0-rc1 6.18.0-rc1
vanilla sched-preemptnext-v5
Amean Wakeup-50th-80 7.17 ( 0.00%) 6.00 ( 16.28%)
Amean Wakeup-90th-80 46.56 ( 0.00%) 19.78 ( 57.52%)
Amean Wakeup-99th-80 119.61 ( 0.00%) 89.94 ( 24.80%)
Amean Wakeup-99.9th-80 3193.78 ( 0.00%) 328.22 ( 89.72%)
schbench Requests Per Second (ops/sec)
6.18.0-rc1 6.18.0-rc1
vanilla sched-preemptnext-v5
Hmean RPS-20th-80 8900.91 ( 0.00%) 9176.78 ( 3.10%)
Hmean RPS-50th-80 8987.41 ( 0.00%) 9217.89 ( 2.56%)
Hmean RPS-90th-80 9123.73 ( 0.00%) 9273.25 ( 1.64%)
Hmean RPS-max-80 9193.50 ( 0.00%) 9301.47 ( 1.17%)
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20251112122521.1331238-3-mgorman@techsingularity.net
The NEXT_BUDDY feature reinforces wakeup preemption to encourage the last
wakee to be scheduled sooner on the assumption that the waker/wakee share
cache-hot data. In CFS, it was paired with LAST_BUDDY to switch back on
the assumption that the pair of tasks still share data but also relied
on START_DEBIT and the exact WAKEUP_PREEMPTION implementation to get
good results.
NEXT_BUDDY has been disabled since commit 0ec9fab3d1 ("sched: Improve
latencies and throughput") and LAST_BUDDY was removed in commit 5e963f2bd4
("sched/fair: Commit to EEVDF"). The reasoning is not clear but as vruntime
spread is mentioned so the expectation is that NEXT_BUDDY had an impact
on overall fairness. It was not noted why LAST_BUDDY was removed but it
is assumed that it's very difficult to reason what LAST_BUDDY's correct
and effective behaviour should be while still respecting EEVDFs goals.
Peter Zijlstra noted during review;
I think I was just struggling to make sense of things and figured
less is more and axed it.
I have vague memories trying to work through the dynamics of
a wakeup-stack and the EEVDF latency requirements and getting
a head-ache.
NEXT_BUDDY is easier to reason about given that it's a point-in-time
decision on the wakees deadline and eligibilty relative to the waker. Enable
NEXT_BUDDY as a preparation path to document that the decision to ignore
the current implementation is deliberate. While not presented, the results
were at best neutral and often much more variable.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20251112122521.1331238-2-mgorman@techsingularity.net
