scheduler changes for v5.14
Linus,
Please pull the latest sched/core git tree from:
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched-core-2021-06-28
# HEAD: adf3c31e18b765ea24eba7b0c1efc076b8ee3d55 sched/doc: Update the CPU capacity asymmetry bits
Scheduler udpates for this cycle:
- Changes to core scheduling facilities:
- Add "Core Scheduling" via CONFIG_SCHED_CORE=y, which enables coordinated scheduling across SMT siblings. This is a much requested feature for cloud computing platforms, to allow the flexible utilization of SMT siblings, without exposing untrusted domains to information leaks & side channels, plus to ensure more deterministic computing performance on SMT systems used by heterogenous workloads.
There's new prctls to set core scheduling groups, which allows more flexible management of workloads that can share siblings.
- Fix task->state access anti-patterns that may result in missed wakeups and rename it to ->__state in the process to catch new abuses.
- Load-balancing changes:
- Tweak newidle_balance for fair-sched, to improve 'memcache'-like workloads.
- "Age" (decay) average idle time, to better track & improve workloads such as 'tbench'.
- Fix & improve energy-aware (EAS) balancing logic & metrics.
- Fix & improve the uclamp metrics.
- Fix task migration (taskset) corner case on !CONFIG_CPUSET.
- Fix RT and deadline utilization tracking across policy changes
- Introduce a "burstable" CFS controller via cgroups, which allows bursty CPU-bound workloads to borrow a bit against their future quota to improve overall latencies & batching. Can be tweaked via /sys/fs/cgroup/cpu/<X>/cpu.cfs_burst_us.
- Rework assymetric topology/capacity detection & handling.
- Scheduler statistics & tooling:
- Disable delayacct by default, but add a sysctl to enable it at runtime if tooling needs it. Use static keys and other optimizations to make it more palatable.
- Use sched_clock() in delayacct, instead of ktime_get_ns().
- Misc cleanups and fixes.
Thanks,
Ingo
------------------>
Alexey Dobriyan (4):
sched: Make nr_running() return 32-bit value
sched: Make nr_iowait() return 32-bit value
sched: Make nr_iowait_cpu() return 32-bit value
sched: Make multiple runqueue task counters 32-bit
Aubrey Li (1):
sched: Migration changes for core scheduling
Beata Michalska (3):
sched/core: Introduce SD_ASYM_CPUCAPACITY_FULL sched_domain flag sched/topology: Rework CPU capacity asymmetry detection sched/doc: Update the CPU capacity asymmetry bits
Chris Hyser (2):
sched: prctl() core-scheduling interface
kselftest: Add test for core sched prctl interface
Dietmar Eggemann (1):
sched/fair: Return early from update_tg_cfs_load() if delta == 0
Eric Dumazet (1):
sched/debug: Remove obsolete init_schedstats()
Frederic Weisbecker (1):
sched: Stop PF_NO_SETAFFINITY from being inherited by various init system threads
Huaixin Chang (1):
sched/fair: Introduce the burstable CFS controller
Ingo Molnar (1):
sched: Fix leftover comment typos
Joel Fernandes (Google) (3):
sched: Fix priority inversion of cookied task with sibling sched/fair: Snapshot the min_vruntime of CPUs on force idle
Documentation: Add usecases, design and interface for core scheduling
Lukasz Luba (3):
thermal/cpufreq_cooling: Update offline CPUs per-cpu thermal_pressure sched/fair: Take thermal pressure into account while estimating energy sched/cpufreq: Consider reduced CPU capacity in energy calculation
Masahiro Yamada (1):
sched: Fix a stale comment in pick_next_task()
Mel Gorman (1):
delayacct: Document task_delayacct sysctl
Odin Ugedal (1):
sched/fair: Fix ascii art by relpacing tabs
Oleg Nesterov (1):
sched/core: Remove the pointless BUG_ON(!task) from wake_up_q()
Paul Gortmaker (1):
sched/isolation: Reconcile rcu_nocbs= and nohz_full=
Peter Zijlstra (31):
delayacct: Use sched_clock()
sched: Rename sched_info_{queued,dequeued}
sched: Simplify sched_info_on()
kvm: Select SCHED_INFO instead of TASK_DELAY_ACCT
delayacct: Add static_branch in scheduler hooks
delayacct: Default disabled
delayacct: Add sysctl to enable at runtime sched/fair: Add a few assertions
sched: Provide raw_spin_rq_*lock*() helpers
sched: Wrap rq::lock access
sched: Prepare for Core-wide rq->lock
sched: Core-wide rq->lock
sched: Optimize rq_lockp() usage
sched: Allow sched_core_put() from atomic context
sched: Introduce sched_class::pick_task()
sched: Basic tracking of matching tasks
sched: Add core wide task selection and scheduling
sched: Trivial forced-newidle balancer
sched: Trivial core scheduling cookie management
sched: Inherit task cookie on fork() sched,stats: Further simplify sched_info
sched: Add CONFIG_SCHED_CORE help text sched,init: Fix DEBUG_PREEMPT vs early boot sched/fair: Age the average idle time
sched: Unbreak wakeups
sched: Introduce task_is_running() sched,perf,kvm: Fix preemption condition
sched: Add get_current_state() sched,timer: Use __set_current_state() sched,arch: Remove unused TASK_STATE offsets
sched: Change task_struct::state
Pierre Gondois (2):
sched/fair: Only compute base_energy_pd if necessary sched/fair: Fix negative energy delta in find_energy_efficient_cpu()
Qais Yousef (3):
sched/uclamp: Fix wrong implementation of cpu.uclamp.min sched/uclamp: Fix locking around cpu_util_update_eff() sched/uclamp: Fix uclamp_tg_restrict()
Rik van Riel (1):
sched,fair: Skip newidle_balance if a wakeup is pending
Valentin Schneider (3):
sched/core: Initialize the idle task with preemption disabled
sched: Make the idle task quack like a per-CPU kthread
sched: Don't defer CPU pick to migration_cpu_stop()
Vincent Donnefort (2):
sched/rt: Fix RT utilization tracking during policy change sched/rt: Fix Deadline utilization tracking during policy change
Vincent Guittot (1):
sched/pelt: Check that *_avg are null when *_sum are
Vineeth Pillai (1):
sched/fair: Fix forced idle sibling starvation corner case
Yejune Deng (1):
lib/smp_processor_id: Use is_percpu_thread() instead of nr_cpus_allowed
Zhaoyang Huang (1):
psi: Fix race between psi_trigger_create/destroy
Documentation/accounting/delay-accounting.rst | 12 +- .../admin-guide/hw-vuln/core-scheduling.rst | 223 ++++ Documentation/admin-guide/hw-vuln/index.rst | 1 + Documentation/admin-guide/kernel-parameters.txt | 2 +- Documentation/admin-guide/sysctl/kernel.rst | 7 + Documentation/scheduler/sched-capacity.rst | 6 +- Documentation/scheduler/sched-energy.rst | 2 +- arch/alpha/kernel/process.c | 2 +- arch/alpha/kernel/smp.c | 1 - arch/arc/kernel/smp.c | 1 - arch/arc/kernel/stacktrace.c | 2 +- arch/arm/kernel/process.c | 2 +- arch/arm/kernel/smp.c | 1 - arch/arm64/include/asm/preempt.h | 2 +- arch/arm64/kernel/process.c | 2 +- arch/arm64/kernel/smp.c | 1 - arch/arm64/kvm/Kconfig | 5 +- arch/csky/kernel/asm-offsets.c | 1 - arch/csky/kernel/smp.c | 1 - arch/csky/kernel/stacktrace.c | 2 +- arch/h8300/kernel/asm-offsets.c | 1 - arch/h8300/kernel/process.c | 2 +- arch/hexagon/kernel/process.c | 2 +- arch/ia64/kernel/mca.c | 2 +- arch/ia64/kernel/process.c | 4 +- arch/ia64/kernel/ptrace.c | 8 +- arch/ia64/kernel/smpboot.c | 1 - arch/m68k/kernel/process.c | 2 +- arch/microblaze/kernel/asm-offsets.c | 1 - arch/mips/kernel/asm-offsets.c | 1 - arch/mips/kernel/process.c | 2 +- arch/mips/kernel/smp.c | 1 - arch/nds32/kernel/process.c | 2 +- arch/nios2/kernel/process.c | 2 +- arch/openrisc/kernel/asm-offsets.c | 1 - arch/openrisc/kernel/smp.c | 2 - arch/parisc/kernel/asm-offsets.c | 1 - arch/parisc/kernel/process.c | 4 +- arch/parisc/kernel/smp.c | 1 - arch/powerpc/kernel/process.c | 4 +- arch/powerpc/kernel/smp.c | 1 - arch/powerpc/xmon/xmon.c | 13 +- arch/riscv/kernel/smpboot.c | 1 - arch/riscv/kernel/stacktrace.c | 2 +- arch/s390/include/asm/preempt.h | 4 +- arch/s390/kernel/process.c | 2 +- arch/s390/kernel/smp.c | 1 - arch/s390/mm/fault.c | 2 +- arch/sh/kernel/process_32.c | 2 +- arch/sh/kernel/smp.c | 2 - arch/sparc/kernel/process_32.c | 3 +- arch/sparc/kernel/process_64.c | 3 +- arch/sparc/kernel/smp_32.c | 1 - arch/sparc/kernel/smp_64.c | 3 - arch/um/kernel/process.c | 2 +- arch/x86/include/asm/preempt.h | 2 +- arch/x86/kernel/process.c | 4 +- arch/x86/kernel/smpboot.c | 1 - arch/x86/kvm/Kconfig | 5 +- arch/xtensa/kernel/process.c | 2 +- arch/xtensa/kernel/smp.c | 1 - block/blk-mq.c | 6 +- drivers/cpuidle/governors/menu.c | 6 +- drivers/md/dm.c | 6 +- drivers/net/ethernet/qualcomm/qca_spi.c | 6 +- drivers/thermal/cpufreq_cooling.c | 2 +- drivers/usb/gadget/udc/max3420_udc.c | 15 +- drivers/usb/host/max3421-hcd.c | 3 +- fs/binfmt_elf.c | 8 +- fs/binfmt_elf_fdpic.c | 4 +- fs/proc/loadavg.c | 2 +- fs/proc/stat.c | 4 +- fs/userfaultfd.c | 4 +- include/asm-generic/preempt.h | 2 +- include/linux/delayacct.h | 22 +- include/linux/energy_model.h | 16 +- include/linux/kthread.h | 2 + include/linux/sched.h | 50 +- include/linux/sched/cpufreq.h | 2 +- include/linux/sched/debug.h | 2 +- include/linux/sched/sd_flags.h | 10 + include/linux/sched/signal.h | 2 +- include/linux/sched/stat.h | 16 +- include/linux/sched_clock.h | 2 +- include/uapi/linux/prctl.h | 8 + init/init_task.c | 2 +- init/main.c | 17 +- kernel/Kconfig.preempt | 20 + kernel/cgroup/cgroup-v1.c | 2 +- kernel/debug/kdb/kdb_support.c | 18 +- kernel/delayacct.c | 71 +- kernel/events/core.c | 7 +- kernel/fork.c | 12 +- kernel/freezer.c | 2 +- kernel/hung_task.c | 2 +- kernel/kcsan/report.c | 2 +- kernel/kthread.c | 34 +- kernel/locking/lockdep.c | 2 +- kernel/locking/mutex.c | 6 +- kernel/locking/rtmutex.c | 4 +- kernel/locking/rwsem.c | 2 +- kernel/ptrace.c | 12 +- kernel/rcu/rcutorture.c | 4 +- kernel/rcu/tree_plugin.h | 2 +- kernel/rcu/tree_stall.h | 12 +- kernel/sched/Makefile | 1 + kernel/sched/core.c | 1138 +++++++++++++++++--- kernel/sched/core_sched.c | 229 ++++ kernel/sched/cpuacct.c | 12 +- kernel/sched/cpufreq_schedutil.c | 1 + kernel/sched/deadline.c | 50 +- kernel/sched/debug.c | 4 +- kernel/sched/fair.c | 440 +++++--- kernel/sched/idle.c | 13 +- kernel/sched/isolation.c | 4 +- kernel/sched/loadavg.c | 2 +- kernel/sched/pelt.h | 2 +- kernel/sched/psi.c | 12 +- kernel/sched/rt.c | 48 +- kernel/sched/sched.h | 437 ++++++-- kernel/sched/stats.h | 68 +- kernel/sched/stop_task.c | 14 +- kernel/sched/topology.c | 213 ++-- kernel/signal.c | 2 +- kernel/smpboot.c | 1 - kernel/softirq.c | 5 +- kernel/sys.c | 5 + kernel/sysctl.c | 12 + kernel/time/timer.c | 2 +- lib/smp_processor_id.c | 6 +- lib/syscall.c | 4 +- mm/compaction.c | 2 +- net/core/dev.c | 2 +- tools/include/uapi/linux/prctl.h | 8 + tools/testing/selftests/sched/.gitignore | 1 + tools/testing/selftests/sched/Makefile | 14 + tools/testing/selftests/sched/config | 1 + tools/testing/selftests/sched/cs_prctl_test.c | 338 ++++++ virt/kvm/kvm_main.c | 2 +- 139 files changed, 3122 insertions(+), 781 deletions(-) create mode 100644 Documentation/admin-guide/hw-vuln/core-scheduling.rst create mode 100644 kernel/sched/core_sched.c create mode 100644 tools/testing/selftests/sched/.gitignore create mode 100644 tools/testing/selftests/sched/Makefile create mode 100644 tools/testing/selftests/sched/config create mode 100644 tools/testing/selftests/sched/cs_prctl_test.c
Quoted 2 lines · Ingo Molnar
- Add "Core Scheduling" via
CONFIG_SCHED_CORE=y,
Grr.
Why is this feature "default y"? I see no reason why anybody would enable it outside of cloud providers, which doesn't argue for it being on by default.
Linus
The pull request you sent on Mon, 28 Jun 2021 08:51:28 +0200:
The pull request you sent on Mon, 28 Jun 2021 08:51:28 +0200:
Quoted 1 line
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.gitsched-core-2021-06-28
has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/54a728dc5e4feb0a9278ad62b19f34ad21ed0ee4
Thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/prtracker.html
PATCH 1/1 sched/core: Disable CONFIG_SCHED_CORE by default
In reply to Linus TorvaldsQuoted 9 lines · Linus Torvalds
Quoted 2 lines · Ingo Molnar
- Add "Core Scheduling" via
CONFIG_SCHED_CORE=y,Grr.
Why is this feature "default y"? I see no reason why anybody would enable it outside of cloud providers, which doesn't argue for it being on by default.
You are completely right, I missed this. Find below the patch to fix it - feel free to apply it directly, will rebase sched/urgent accordingly.
Thanks,
Ingo
=====================>
From: Ingo Molnar <[email protected]>
Date: Mon, 28 Jun 2021 21:55:16 +0200
Subject: [PATCH] sched/core: Disable CONFIG_SCHED_CORE by default
This option adds extra overhead to the scheduler, and most users wouldn't want it.
--- kernel/Kconfig.preempt | 1 - 1 file changed, 1 deletion(-)
kernel/Kconfig.preempt +0 -1
@@ -102,7 +102,6 @@ config PREEMPT_DYNAMIC config SCHED_CORE bool "Core Scheduling for SMT" - default y depends on SCHED_SMT help This option permits Core Scheduling, a means of coordinated task
Re: [PATCH] sched/core: Disable CONFIG_SCHED_CORE by default
Quoted 2 lines · Ingo Molnar
You are completely right, I missed this. Find below the patch to fix it -
I think you should update the helptext too, which currently says that it's enabled if SCHED_SMT is enabled.
And yes, it may be that there is no measurable performance downside, but this feature has been discussed and under development for a long time now, and I'd like the default to try to limit the impact since it makes little sense to most people.
Linus
PATCH 1/1 sched/core: Disable CONFIG_SCHED_CORE by default
Quoted 11 lines · Linus Torvalds
Quoted 2 lines · Ingo Molnar
You are completely right, I missed this. Find below the patch to fix it -
I think you should update the helptext too, which currently says that it's enabled if
SCHED_SMTis enabled.And yes, it may be that there is no measurable performance downside, but this feature has been discussed and under development for a long time now, and I'd like the default to try to limit the impact since it makes little sense to most people.
Updated patch attached, which updates the help text and clarifies what overhead there is.
While the worst of the overhead is behind the __sched_core_enabled static key, there's a notable increase in pick_next_task(), which might be measurable with the right microbenchmark.
So the 'no overhead' claim is probably not entirely true, and comes from the context of *horrible* overhead of early iterations of the core scheduling feature ...
Thanks,
Ingo
===================================>
From: Ingo Molnar <[email protected]>
Date: Mon, 28 Jun 2021 21:55:16 +0200
Subject: [PATCH] sched/core: Disable CONFIG_SCHED_CORE by default
This option at minimum adds extra code to the scheduler - even if it's default unused - and most users wouldn't want it.
--- kernel/Kconfig.preempt | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/Kconfig.preempt +3 -3
@@ -102,7 +102,6 @@ config PREEMPT_DYNAMIC config SCHED_CORE bool "Core Scheduling for SMT" - default y depends on SCHED_SMT help This option permits Core Scheduling, a means of coordinated task
@@ -115,7 +114,8 @@ config SCHED_CORE - mitigation of some (not all) SMT side channels; - limiting SMT interference to improve determinism and/or performance. - SCHED_CORE is default enabled when SCHED_SMT is enabled -- when - unused there should be no impact on performance. + SCHED_CORE is default disabled. When it is enabled and unused, + which is the likely usage by Linux distributions, there should + be no measurable impact on performance.
[GIT PULL] scheduler fixes for v5.14
In reply to Ingo MolnarLinus,
Please pull the latest sched/urgent git tree from:
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched-urgent-2021-06-30
# HEAD: a22a5cb81e20657194fde6c835e07d28c4dfddbe Merge branch 'sched/core' into sched/urgent, to pick up fix
- Fix a small inconsistency (bug) in load tracking, caught by a new warning that several people reported.
- Flip CONFIG_SCHED_CORE to default-disabled, and update the Kconfig help text.
Thanks,
Ingo
------------------>
Ingo Molnar (1):
sched/core: Disable CONFIG_SCHED_CORE by default
Odin Ugedal (1):
sched/fair: Ensure _sum and _avg values stay consistent
kernel/Kconfig.preempt | 6 +++--- kernel/sched/fair.c | 6 +++--- 2 files changed, 6 insertions(+), 6 deletions(-)
kernel/Kconfig.preempt +3 -3
@@ -102,7 +102,6 @@ config PREEMPT_DYNAMIC config SCHED_CORE bool "Core Scheduling for SMT" - default y depends on SCHED_SMT help This option permits Core Scheduling, a means of coordinated task
@@ -115,7 +114,8 @@ config SCHED_CORE - mitigation of some (not all) SMT side channels; - limiting SMT interference to improve determinism and/or performance. - SCHED_CORE is default enabled when SCHED_SMT is enabled -- when - unused there should be no impact on performance. + SCHED_CORE is default disabled. When it is enabled and unused, + which is the likely usage by Linux distributions, there should + be no measurable impact on performance.
kernel/sched/fair.c +3 -3
@@ -3657,15 +3657,15 @@ update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) r = removed_load;sub_positive (&sa->load_avg, r); -sub_positive (&sa->load_sum, r * divider); + sa->load_sum = sa->load_avg * divider; r = removed_util;sub_positive (&sa->util_avg, r); -sub_positive (&sa->util_sum, r * divider); + sa->util_sum = sa->util_avg * divider; r = removed_runnable;sub_positive (&sa->runnable_avg, r); -sub_positive (&sa->runnable_sum, r * divider); + sa->runnable_sum = sa->runnable_avg * divider;/* * removed_runnable is the unweighted version of removed_load so we
Re: [GIT PULL] scheduler fixes for v5.14
The pull request you sent on Wed, 30 Jun 2021 12:53:09 +0200:
The pull request you sent on Wed, 30 Jun 2021 12:53:09 +0200:
Quoted 1 line
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.gitsched-urgent-2021-06-30
has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/a6eaf3850cb171c328a8b0db6d3c79286a1eba9d
Thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/prtracker.html