Commit bfe8eb3b authored by Linus Torvalds's avatar Linus Torvalds
Browse files

Merge tag 'sched-core-2024-01-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 

Pull scheduler updates from Ingo Molnar:
 "Energy scheduling:

   - Consolidate how the max compute capacity is used in the scheduler
     and how we calculate the frequency for a level of utilization.

   - Rework interface between the scheduler and the schedutil governor

   - Simplify the util_est logic

  Deadline scheduler:

   - Work more towards reducing SCHED_DEADLINE starvation of low
     priority tasks (e.g., SCHED_OTHER) tasks when higher priority tasks
     monopolize CPU cycles, via the introduction of 'deadline servers'
     (nested/2-level scheduling).

     "Fair servers" to make use of this facility are not introduced yet.

  EEVDF:

   - Introduce O(1) fastpath for EEVDF task selection

  NUMA balancing:

   - Tune the NUMA-balancing vma scanning logic some more, to better
     distribute the probability of a particular vma getting scanned.

  Plus misc fixes, cleanups and updates"

* tag 'sched-core-2024-01-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (30 commits)
  sched/fair: Fix tg->load when offlining a CPU
  sched/fair: Remove unused 'next_buddy_marked' local variable in check_preempt_wakeup_fair()
  sched/fair: Use all little CPUs for CPU-bound workloads
  sched/fair: Simplify util_est
  sched/fair: Remove SCHED_FEAT(UTIL_EST_FASTUP, true)
  arm64/amu: Use capacity_ref_freq() to set AMU ratio
  cpufreq/cppc: Set the frequency used for computing the capacity
  cpufreq/cppc: Move and rename cppc_cpufreq_{perf_to_khz|khz_to_perf}()
  energy_model: Use a fixed reference frequency
  cpufreq/schedutil: Use a fixed reference frequency
  cpufreq: Use the fixed and coherent frequency for scaling capacity
  sched/topology: Add a new arch_scale_freq_ref() method
  freezer,sched: Clean saved_state when restoring it during thaw
  sched/fair: Update min_vruntime for reweight_entity() correctly
  sched/doc: Update documentation after renames and synchronize Chinese version
  sched/cpufreq: Rework iowait boost
  sched/cpufreq: Rework schedutil governor performance estimation
  sched/pelt: Avoid underestimation of task utilization
  sched/timers: Explain why idle task schedules out on remote timer enqueue
  sched/cpuidle: Comment about timers requirements VS idle handler
  ...
parents aac4de46 cdb3033e

Documentation/scheduler/sched-design-CFS.rst

+4 −4
Original line number Diff line number Diff line
@@ -180,7 +180,7 @@ This is the (partial) list of the hooks: 
   compat_yield sysctl is turned on; in that case, it places the scheduling

   entity at the right-most end of the red-black tree.



 - check_preempt_curr(...)

 - wakeup_preempt(...)



   This function checks if a task that entered the runnable state should

   preempt the currently running task.
@@ -189,10 +189,10 @@ This is the (partial) list of the hooks: 


   This function chooses the most appropriate task eligible to run next.



 - set_curr_task(...)

 - set_next_task(...)



   This function is called when a task changes its scheduling class or changes

   its task group.

   This function is called when a task changes its scheduling class, changes

   its task group or is scheduled.



 - task_tick(...)

Documentation/scheduler/schedutil.rst

+3 −4
Original line number Diff line number Diff line
@@ -90,8 +90,8 @@ For more detail see: 
 - Documentation/scheduler/sched-capacity.rst:"1. CPU Capacity + 2. Task utilization"





UTIL_EST / UTIL_EST_FASTUP

==========================

UTIL_EST

========



Because periodic tasks have their averages decayed while they sleep, even

though when running their expected utilization will be the same, they suffer a
@@ -99,8 +99,7 @@ though when running their expected utilization will be the same, they suffer a 


To alleviate this (a default enabled option) UTIL_EST drives an Infinite

Impulse Response (IIR) EWMA with the 'running' value on dequeue -- when it is

highest. A further default enabled option UTIL_EST_FASTUP modifies the IIR

filter to instantly increase and only decay on decrease.

highest. UTIL_EST filters to instantly increase and only decay on decrease.



A further runqueue wide sum (of runnable tasks) is maintained of:

Documentation/translations/zh_CN/scheduler/sched-design-CFS.rst

+4 −4
Original line number Diff line number Diff line
@@ -80,7 +80,7 @@ p->se.vruntime。一旦p->se.vruntime变得足够大，其它的任务将成为 
CFS使用纳秒粒度的计时，不依赖于任何jiffies或HZ的细节。因此CFS并不像之前的调度器那样

有“时间片”的概念，也没有任何启发式的设计。唯一可调的参数（你需要打开CONFIG_SCHED_DEBUG）是：



   /sys/kernel/debug/sched/min_granularity_ns

   /sys/kernel/debug/sched/base_slice_ns



它可以用来将调度器从“桌面”模式（也就是低时延）调节为“服务器”（也就是高批处理）模式。

它的默认设置是适合桌面的工作负载。SCHED_BATCH也被CFS调度器模块处理。
@@ -147,7 +147,7 @@ array）。 
   这个函数的行为基本上是出队，紧接着入队，除非compat_yield sysctl被开启。在那种情况下，

   它将调度实体放在红黑树的最右端。



 - check_preempt_curr(...)

 - wakeup_preempt(...)



   这个函数检查进入可运行状态的任务能否抢占当前正在运行的任务。
@@ -155,9 +155,9 @@ array）。 


   这个函数选择接下来最适合运行的任务。



 - set_curr_task(...)

 - set_next_task(...)



   这个函数在任务改变调度类或改变任务组时被调用。

   这个函数在任务改变调度类，改变任务组时，或者任务被调度时被调用。



 - task_tick(...)

Documentation/translations/zh_CN/scheduler/schedutil.rst

+3 −4
Original line number Diff line number Diff line
@@ -89,16 +89,15 @@ r_cpu被定义为当前CPU的最高性能水平与系统中任何其它CPU的最 
 - Documentation/translations/zh_CN/scheduler/sched-capacity.rst:"1. CPU Capacity + 2. Task utilization"





UTIL_EST / UTIL_EST_FASTUP

==========================

UTIL_EST

========



由于周期性任务的平均数在睡眠时会衰减，而在运行时其预期利用率会和睡眠前相同，

因此它们在再次运行后会面临（DVFS）的上涨。



为了缓解这个问题，（一个默认使能的编译选项）UTIL_EST驱动一个无限脉冲响应

（Infinite Impulse Response，IIR）的EWMA，“运行”值在出队时是最高的。

另一个默认使能的编译选项UTIL_EST_FASTUP修改了IIR滤波器，使其允许立即增加，

仅在利用率下降时衰减。

UTIL_EST滤波使其在遇到更高值时立刻增加，而遇到低值时会缓慢衰减。



进一步，运行队列的（可运行任务的）利用率之和由下式计算：

arch/arm/include/asm/topology.h

+1 −0
Original line number Diff line number Diff line
@@ -13,6 +13,7 @@ 
#define arch_set_freq_scale topology_set_freq_scale

#define arch_scale_freq_capacity topology_get_freq_scale

#define arch_scale_freq_invariant topology_scale_freq_invariant

#define arch_scale_freq_ref topology_get_freq_ref

#endif



/* Replace task scheduler's default cpu-invariant accounting */