sched/balancing: Standardize the naming of scheduler load-balancing functions
Subsystems (8)
cover sched/balancing: Standardize the naming of scheduler load-balancing functions
Over the years we've grown a colorful zoo of scheduler load-balancing function names - both following random, idiosyncratic patterns, and gaining historic misnomers that are not accurate anymore.
We have 'newidle_balance()' to rebalance newly idle tasks, but 'balance_domains()' to rebalance domains. We have a find_idlest_cpu() function whose purpose is not to find the idlest CPU anymore, and a find_busiest_queue() function whose purpose is not to find the busiest runqueue anymore.
Fix most of the misnomers and organize the functions along the sched_balance_*() namespace:
scheduler_tick() => sched_tick()
run_rebalance_domains() => sched_balance_softirq()
trigger_load_balance() => sched_balance_trigger()
rebalance_domains() => sched_balance_domains()
load_balance() => sched_balance_rq()
newidle_balance() => sched_balance_newidle()
find_busiest_queue() => sched_balance_find_src_rq()
find_busiest_group() => sched_balance_find_src_group()
find_idlest_group_cpu() => sched_balance_find_dst_group_cpu()
find_idlest_group() => sched_balance_find_dst_group()
find_idlest_cpu() => sched_balance_find_dst_cpu()
update_blocked_averages() => sched_balance_update_blocked_averages()
I think the visual improvement of left vs. right column demonstrates the goal nicely.
While the function names got a bit longer, another advantage of the common prefix, beyond readability, is that now a:
git grep sched_balance_
... will show most of the balancing code nicely.
( I have a few more patches that standardize the NOHZ balancing
code along the nohz_balance_*() nomenclature as well. )
Thanks,
Ingo
==================>
Ingo Molnar (13):
sched/balancing: Rename run_rebalance_domains() => sched_balance_softirq() sched/balancing: Rename scheduler_tick() => sched_tick()
sched/balancing: Rename trigger_load_balance() => sched_balance_trigger() sched/balancing: Rename rebalance_domains() => sched_balance_domains() sched/balancing: Rename load_balance() => sched_balance_rq()
sched/balancing: Rename find_busiest_queue() => find_src_rq() sched/balancing: Rename find_src_rq() => sched_balance_find_src_rq() sched/balancing: Rename find_busiest_group() => sched_balance_find_src_group() sched/balancing: Rename update_blocked_averages() => sched_balance_update_blocked_averages() sched/balancing: Rename newidle_balance() => sched_balance_newidle() sched/balancing: Rename find_idlest_group_cpu() => sched_balance_find_dst_group_cpu() sched/balancing: Rename find_idlest_group() => sched_balance_find_dst_group() sched/balancing: Rename find_idlest_cpu() => sched_balance_find_dst_cpu()
Documentation/scheduler/sched-domains.rst | 12 ++-- Documentation/scheduler/sched-stats.rst | 32 +++++------ Documentation/translations/zh_CN/scheduler/sched-domains.rst | 10 ++-- Documentation/translations/zh_CN/scheduler/sched-stats.rst | 30 +++++----- arch/arm/kernel/topology.c | 2 +- include/linux/sched.h | 2 +- include/linux/sched/topology.h | 2 +- kernel/sched/core.c | 6 +- kernel/sched/fair.c | 88 ++++++++++++++--------------- kernel/sched/loadavg.c | 2 +- kernel/sched/pelt.c | 2 +- kernel/sched/sched.h | 4 +- kernel/time/timer.c | 2 +- kernel/workqueue.c | 2 +- .../selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc | 2 +- 15 files changed, 99 insertions(+), 99 deletions(-)
Signature · 1 line
2.40.1
PATCH 1/13 sched/balancing: Rename run_rebalance_domains() => sched_balance_softirq()
run_rebalance_domains() is a misnomer, as it doesn't only run rebalance_domains(), but since the introduction of the NOHZ code it also runs nohz_idle_balance().
Rename it to sched_balance_softirq(), reflecting its more generic purpose and that it's a softirq handler.
--- Documentation/scheduler/sched-domains.rst | 2 +- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 2 +- kernel/sched/fair.c | 4 ++-- 3 files changed, 4 insertions(+), 4 deletions(-)
Documentation/scheduler/sched-domains.rst +1 -1
@@ -34,7 +34,7 @@ out of balance are tasks moved between groups. In kernel/sched/core.c, trigger_load_balance() is run periodically on each CPU through scheduler_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load -balancing workhorse, run_rebalance_domains()->rebalance_domains(), is then run +balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run in softirq context (SCHED_SOFTIRQ). The latter function takes two arguments: the runqueue of current CPU and whether
Documentation/translations/zh_CN/scheduler/sched-domains.rst +1 -1
@@ -36,7 +36,7 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这 在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过scheduler_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 -的工作由run_rebalance_domains()->rebalance_domains()完成,在软中断上下文中执行 +的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。 后一个函数有两个入参:当前CPU的运行队列、它在scheduler_tick()调用时是否空闲。函数会从
kernel/sched/fair.c +2 -2
@@ -12415,7 +12415,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)* - indirectly from a remote scheduler_tick() for NOHZ idle balancing * through the SMP cross-call nohz_csd_func() */ -static __latent_entropy void run_rebalance_domains(struct softirq_action *h) +static __latent_entropy void sched_balance_softirq(struct softirq_action *h) {struct rq *this_rq =this_rq ();enum cpu_idle_type idle = this_rq->idle_balance;
@@ -13216,7 +13216,7 @@ __init void init_sched_fair_class(void)#endif } -open_softirq (SCHED_SOFTIRQ , run_rebalance_domains); +open_softirq (SCHED_SOFTIRQ , sched_balance_softirq);#ifdef CONFIG_NO_HZ_COMMON nohz.next_balance = jiffies;
Signature · 1 line
2.40.1
PATCH 2/13 sched/balancing: Rename scheduler_tick() => sched_tick()
In reply to Ingo Molnar- Standardize on prefixing scheduler-internal functions defined in <linux/sched.h> with sched_*() prefix. scheduler_tick() was the only function using the scheduler_ prefix. Harmonize it.
- The other reason to rename it is the NOHZ scheduler tick handling functions are already named sched_tick_*(). Make the 'git grep sched_tick' more meaningful.
--- Documentation/scheduler/sched-domains.rst | 4 ++-- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 4 ++-- include/linux/sched.h | 2 +- kernel/sched/core.c | 4 ++-- kernel/sched/loadavg.c | 2 +- kernel/time/timer.c | 2 +- kernel/workqueue.c | 2 +- tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc | 2 +- 8 files changed, 11 insertions(+), 11 deletions(-)
Documentation/scheduler/sched-domains.rst +2 -2
@@ -32,13 +32,13 @@ load of each of its member CPUs, and only when the load of a group becomes out of balance are tasks moved between groups. In kernel/sched/core.c, trigger_load_balance() is run periodically on each CPU -through scheduler_tick(). It raises a softirq after the next regularly scheduled +through sched_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run in softirq context (SCHED_SOFTIRQ). The latter function takes two arguments: the runqueue of current CPU and whether -the CPU was idle at the time the scheduler_tick() happened and iterates over all +the CPU was idle at the time the sched_tick() happened and iterates over all sched domains our CPU is on, starting from its base domain and going up the ->parent chain. While doing that, it checks to see if the current domain has exhausted its rebalance interval. If so, it runs load_balance() on that domain. It then checks
Documentation/translations/zh_CN/scheduler/sched-domains.rst +2 -2
@@ -34,12 +34,12 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这 调度域中的负载均衡发生在调度组中。也就是说,每个组被视为一个实体。组的负载被定义为它 管辖的每个CPU的负载之和。仅当组的负载不均衡后,任务才在组之间发生迁移。 -在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过scheduler_tick() +在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过sched_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。 -后一个函数有两个入参:当前CPU的运行队列、它在scheduler_tick()调用时是否空闲。函数会从 +后一个函数有两个入参:当前CPU的运行队列、它在sched_tick()调用时是否空闲。函数会从 当前CPU所在的基调度域开始迭代执行,并沿着parent指针链向上进入更高层级的调度域。在迭代 过程中,函数会检查当前调度域是否已经耗尽了再平衡的时间间隔,如果是,它在该调度域运行 load_balance()。接下来它检查父调度域(如果存在),再后来父调度域的父调度域,以此类推。
include/linux/sched.h +1 -1
@@ -301,7 +301,7 @@ enum { TASK_COMM_LEN = 16, }; -extern void scheduler_tick (void ); +extern void sched_tick (void );#define MAX_SCHEDULE_TIMEOUT LONG_MAX
kernel/sched/core.c +2 -2
@@ -5651,7 +5651,7 @@ static inline u64 cpu_resched_latency(struct rq *rq) { return 0; }* This function gets called by the timer code, with HZ frequency. * We call it with interrupts disabled. */ -void scheduler_tick (void ) +void sched_tick (void ) {int cpu =smp_processor_id ();struct rq *rq =cpu_rq (cpu);
@@ -6574,7 +6574,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)* paths. For example, see arch/x86/entry_64.S. * * To drive preemption between tasks, the scheduler sets the flag in timer -* interrupt handler scheduler_tick(). +* interrupt handler sched_tick(). * * 3. Wakeups don't really cause entry into schedule(). They add a * task to the run-queue and that's it.
kernel/sched/loadavg.c +1 -1
@@ -379,7 +379,7 @@ void calc_global_load(void) }/* -* Called from scheduler_tick() to periodically update this CPU's +* Called from sched_tick() to periodically update this CPU's * active count. */ void calc_global_load_tick (struct rq *this_rq)
kernel/time/timer.c +1 -1
@@ -2089,7 +2089,7 @@ void update_process_times(int user_tick)if (in_irq ())irq_work_tick ();#endif -scheduler_tick (); +sched_tick ();if (IS_ENABLED (CONFIG_POSIX_TIMERS ))run_posix_cpu_timers (); }
kernel/workqueue.c +1 -1
@@ -1324,7 +1324,7 @@ void wq_worker_sleeping(struct task_struct *task)* wq_worker_tick - a scheduler tick occurred while a kworker is running * @task: task currently running * -* Called from scheduler_tick(). We're in the IRQ context and the current +* Called from sched_tick(). We're in the IRQ context and the current * worker's fields which follow the 'K' locking rule can be accessed safely. */ void wq_worker_tick (struct task_struct *task)
tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc +1 -1
@@ -19,7 +19,7 @@ fail() { # mesg FILTER=set_ftrace_filter FUNC1="schedule" -FUNC2="scheduler_tick" +FUNC2="sched_tick" ALL_FUNCS="#### all functions enabled ####"
Signature · 1 line
2.40.1
PATCH 3/13 sched/balancing: Rename trigger_load_balance() => sched_balance_trigger()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
--- Documentation/scheduler/sched-domains.rst | 2 +- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 2 +- kernel/sched/core.c | 2 +- kernel/sched/fair.c | 2 +- kernel/sched/sched.h | 2 +- 5 files changed, 5 insertions(+), 5 deletions(-)
Documentation/scheduler/sched-domains.rst +1 -1
@@ -31,7 +31,7 @@ is treated as one entity. The load of a group is defined as the sum of the load of each of its member CPUs, and only when the load of a group becomes out of balance are tasks moved between groups. -In kernel/sched/core.c, trigger_load_balance() is run periodically on each CPU +In kernel/sched/core.c, sched_balance_trigger() is run periodically on each CPU through sched_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run
Documentation/translations/zh_CN/scheduler/sched-domains.rst +1 -1
@@ -34,7 +34,7 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这 调度域中的负载均衡发生在调度组中。也就是说,每个组被视为一个实体。组的负载被定义为它 管辖的每个CPU的负载之和。仅当组的负载不均衡后,任务才在组之间发生迁移。 -在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过sched_tick() +在kernel/sched/core.c中,sched_balance_trigger()在每个CPU上通过sched_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。
kernel/sched/core.c +1 -1
@@ -5689,7 +5689,7 @@ void sched_tick(void)#ifdef CONFIG_SMP rq->idle_balance =idle_cpu (cpu); -trigger_load_balance (rq); +sched_balance_trigger (rq);#endif }
kernel/sched/fair.c +1 -1
@@ -12438,7 +12438,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)/* * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. */ -void trigger_load_balance (struct rq *rq) +void sched_balance_trigger (struct rq *rq) {/* * Don't need to rebalance while attached to NULL domain or
kernel/sched/sched.h +1 -1
@@ -2397,7 +2397,7 @@ extern struct task_struct *pick_next_task_idle(struct rq *rq);extern void update_group_capacity (struct sched_domain *sd,int cpu); -extern void trigger_load_balance (struct rq *rq); +extern void sched_balance_trigger (struct rq *rq);extern void set_cpus_allowed_common (struct task_struct *p,struct affinity_context *ctx);
Signature · 1 line
2.40.1
PATCH 4/13 sched/balancing: Rename rebalance_domains() => sched_balance_domains()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
--- Documentation/scheduler/sched-domains.rst | 2 +- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 2 +- arch/arm/kernel/topology.c | 2 +- kernel/sched/fair.c | 8 ++++---- kernel/sched/sched.h | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)
Documentation/scheduler/sched-domains.rst +1 -1
@@ -34,7 +34,7 @@ out of balance are tasks moved between groups. In kernel/sched/core.c, sched_balance_trigger() is run periodically on each CPU through sched_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load -balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run +balancing workhorse, sched_balance_softirq()->sched_balance_domains(), is then run in softirq context (SCHED_SOFTIRQ). The latter function takes two arguments: the runqueue of current CPU and whether
Documentation/translations/zh_CN/scheduler/sched-domains.rst +1 -1
@@ -36,7 +36,7 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这 在kernel/sched/core.c中,sched_balance_trigger()在每个CPU上通过sched_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 -的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 +的工作由sched_balance_softirq()->sched_balance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。 后一个函数有两个入参:当前CPU的运行队列、它在sched_tick()调用时是否空闲。函数会从
arch/arm/kernel/topology.c +1 -1
@@ -42,7 +42,7 @@* can take this difference into account during load balance. A per cpu * structure is preferred because each CPU updates its own cpu_capacity field * during the load balance except for idle cores. One idle core is selected -* to run the rebalance_domains for all idle cores and the cpu_capacity can be +* to run the sched_balance_domains for all idle cores and the cpu_capacity can be * updated during this sequence. */
kernel/sched/fair.c +4 -4
@@ -11685,7 +11685,7 @@ static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost)* * Balancing parameters are set up in init_sched_domains. */ -static void rebalance_domains (struct rq *rq,enum cpu_idle_type idle) +static void sched_balance_domains (struct rq *rq,enum cpu_idle_type idle) {int continue_balancing =1 ;int cpu = rq->cpu;
@@ -12161,7 +12161,7 @@ static void _nohz_idle_balance(struct rq *this_rq, unsigned int flags)rq_unlock_irqrestore (rq, &rf);if (flags &NOHZ_BALANCE_KICK ) -rebalance_domains (rq,CPU_IDLE ); +sched_balance_domains (rq,CPU_IDLE ); }if (time_after (next_balance, rq->next_balance)) {
@@ -12422,7 +12422,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)/* * If this CPU has a pending NOHZ_BALANCE_KICK, then do the * balancing on behalf of the other idle CPUs whose ticks are -* stopped. Do nohz_idle_balance *before* rebalance_domains to +* stopped. Do nohz_idle_balance *before* sched_balance_domains to * give the idle CPUs a chance to load balance. Else we may * load balance only within the local sched_domain hierarchy * and abort nohz_idle_balance altogether if we pull some load.
@@ -12432,7 +12432,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)/* normal load balance */ update_blocked_averages (this_rq->cpu); -rebalance_domains (this_rq, idle); +sched_balance_domains (this_rq, idle); }/*
kernel/sched/sched.h +1 -1
@@ -2904,7 +2904,7 @@ extern void cfs_bandwidth_usage_dec(void);#define NOHZ_NEWILB_KICK_BIT 2#define NOHZ_NEXT_KICK_BIT 3 -/* Run rebalance_domains() */ +/* Run sched_balance_domains() */ #define NOHZ_BALANCE_KICK BIT(NOHZ_BALANCE_KICK_BIT)/* Update blocked load */ #define NOHZ_STATS_KICK BIT(NOHZ_STATS_KICK_BIT)
Signature · 1 line
2.40.1
PATCH 5/13 sched/balancing: Rename load_balance() => sched_balance_rq()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
Also load_balance() has become somewhat of a misnomer: historically it was the first and primary load-balancing function that was called, but with the introduction of sched domains, it's become a lower layer function that balances runqueues.
Rename it to sched_balance_rq() accordingly.
--- Documentation/scheduler/sched-domains.rst | 4 ++-- Documentation/scheduler/sched-stats.rst | 32 ++++++++++++++++---------------- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 4 ++-- Documentation/translations/zh_CN/scheduler/sched-stats.rst | 30 +++++++++++++++--------------- include/linux/sched/topology.h | 2 +- kernel/sched/fair.c | 10 +++++----- 6 files changed, 41 insertions(+), 41 deletions(-)
Documentation/scheduler/sched-domains.rst +2 -2
@@ -41,11 +41,11 @@ The latter function takes two arguments: the runqueue of current CPU and whether the CPU was idle at the time the sched_tick() happened and iterates over all sched domains our CPU is on, starting from its base domain and going up the ->parent chain. While doing that, it checks to see if the current domain has exhausted its -rebalance interval. If so, it runs load_balance() on that domain. It then checks +rebalance interval. If so, it runs sched_balance_rq() on that domain. It then checks the parent sched_domain (if it exists), and the parent of the parent and so forth. -Initially, load_balance() finds the busiest group in the current sched domain. +Initially, sched_balance_rq() finds the busiest group in the current sched domain. If it succeeds, it looks for the busiest runqueue of all the CPUs' runqueues in that group. If it manages to find such a runqueue, it locks both our initial CPU's runqueue and the newly found busiest one and starts moving tasks from it
Documentation/scheduler/sched-stats.rst +16 -16
@@ -72,53 +72,53 @@ domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 The first field is a bit mask indicating what cpus this domain operates over. -The next 24 are a variety of load_balance() statistics in grouped into types +The next 24 are a variety of sched_balance_rq() statistics in grouped into types of idleness (idle, busy, and newly idle): - 1) # of times in this domain load_balance() was called when the + 1) # of times in this domain sched_balance_rq() was called when the cpu was idle - 2) # of times in this domain load_balance() checked but found + 2) # of times in this domain sched_balance_rq() checked but found the load did not require balancing when the cpu was idle - 3) # of times in this domain load_balance() tried to move one or + 3) # of times in this domain sched_balance_rq() tried to move one or more tasks and failed, when the cpu was idle 4) sum of imbalances discovered (if any) with each call to - load_balance() in this domain when the cpu was idle + sched_balance_rq() in this domain when the cpu was idle 5) # of times in this domain pull_task() was called when the cpu was idle 6) # of times in this domain pull_task() was called even though the target task was cache-hot when idle - 7) # of times in this domain load_balance() was called but did + 7) # of times in this domain sched_balance_rq() was called but did not find a busier queue while the cpu was idle 8) # of times in this domain a busier queue was found while the cpu was idle but no busier group was found - 9) # of times in this domain load_balance() was called when the + 9) # of times in this domain sched_balance_rq() was called when the cpu was busy - 10) # of times in this domain load_balance() checked but found the + 10) # of times in this domain sched_balance_rq() checked but found the load did not require balancing when busy - 11) # of times in this domain load_balance() tried to move one or + 11) # of times in this domain sched_balance_rq() tried to move one or more tasks and failed, when the cpu was busy 12) sum of imbalances discovered (if any) with each call to - load_balance() in this domain when the cpu was busy + sched_balance_rq() in this domain when the cpu was busy 13) # of times in this domain pull_task() was called when busy 14) # of times in this domain pull_task() was called even though the target task was cache-hot when busy - 15) # of times in this domain load_balance() was called but did not + 15) # of times in this domain sched_balance_rq() was called but did not find a busier queue while the cpu was busy 16) # of times in this domain a busier queue was found while the cpu was busy but no busier group was found - 17) # of times in this domain load_balance() was called when the + 17) # of times in this domain sched_balance_rq() was called when the cpu was just becoming idle - 18) # of times in this domain load_balance() checked but found the + 18) # of times in this domain sched_balance_rq() checked but found the load did not require balancing when the cpu was just becoming idle - 19) # of times in this domain load_balance() tried to move one or more + 19) # of times in this domain sched_balance_rq() tried to move one or more tasks and failed, when the cpu was just becoming idle 20) sum of imbalances discovered (if any) with each call to - load_balance() in this domain when the cpu was just becoming idle + sched_balance_rq() in this domain when the cpu was just becoming idle 21) # of times in this domain pull_task() was called when newly idle 22) # of times in this domain pull_task() was called even though the target task was cache-hot when just becoming idle - 23) # of times in this domain load_balance() was called but did not + 23) # of times in this domain sched_balance_rq() was called but did not find a busier queue while the cpu was just becoming idle 24) # of times in this domain a busier queue was found while the cpu was just becoming idle but no busier group was found
Documentation/translations/zh_CN/scheduler/sched-domains.rst +2 -2
@@ -42,9 +42,9 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这 后一个函数有两个入参:当前CPU的运行队列、它在sched_tick()调用时是否空闲。函数会从 当前CPU所在的基调度域开始迭代执行,并沿着parent指针链向上进入更高层级的调度域。在迭代 过程中,函数会检查当前调度域是否已经耗尽了再平衡的时间间隔,如果是,它在该调度域运行 -load_balance()。接下来它检查父调度域(如果存在),再后来父调度域的父调度域,以此类推。 +sched_balance_rq()。接下来它检查父调度域(如果存在),再后来父调度域的父调度域,以此类推。 -起初,load_balance()查找当前调度域中最繁忙的调度组。如果成功,在该调度组管辖的全部CPU +起初,sched_balance_rq()查找当前调度域中最繁忙的调度组。如果成功,在该调度组管辖的全部CPU 的运行队列中找出最繁忙的运行队列。如能找到,对当前的CPU运行队列和新找到的最繁忙运行 队列均加锁,并把任务从最繁忙队列中迁移到当前CPU上。被迁移的任务数量等于在先前迭代执行 中计算出的该调度域的调度组的不均衡值。
Documentation/translations/zh_CN/scheduler/sched-stats.rst +15 -15
@@ -75,42 +75,42 @@ domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 繁忙,新空闲): - 1) 当CPU空闲时,load_balance()在这个调度域中被调用了#次 - 2) 当CPU空闲时,load_balance()在这个调度域中被调用,但是发现负载无需 + 1) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用了#次 + 2) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,但是发现负载无需 均衡#次 - 3) 当CPU空闲时,load_balance()在这个调度域中被调用,试图迁移1个或更多 + 3) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,试图迁移1个或更多 任务且失败了#次 - 4) 当CPU空闲时,load_balance()在这个调度域中被调用,发现不均衡(如果有) + 4) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,发现不均衡(如果有) #次 5) 当CPU空闲时,pull_task()在这个调度域中被调用#次 6) 当CPU空闲时,尽管目标任务是热缓存状态,pull_task()依然被调用#次 - 7) 当CPU空闲时,load_balance()在这个调度域中被调用,未能找到更繁忙的 + 7) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,未能找到更繁忙的 队列#次 8) 当CPU空闲时,在调度域中找到了更繁忙的队列,但未找到更繁忙的调度组 #次 - 9) 当CPU繁忙时,load_balance()在这个调度域中被调用了#次 - 10) 当CPU繁忙时,load_balance()在这个调度域中被调用,但是发现负载无需 + 9) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用了#次 + 10) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,但是发现负载无需 均衡#次 - 11) 当CPU繁忙时,load_balance()在这个调度域中被调用,试图迁移1个或更多 + 11) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,试图迁移1个或更多 任务且失败了#次 - 12) 当CPU繁忙时,load_balance()在这个调度域中被调用,发现不均衡(如果有) + 12) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,发现不均衡(如果有) #次 13) 当CPU繁忙时,pull_task()在这个调度域中被调用#次 14) 当CPU繁忙时,尽管目标任务是热缓存状态,pull_task()依然被调用#次 - 15) 当CPU繁忙时,load_balance()在这个调度域中被调用,未能找到更繁忙的 + 15) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,未能找到更繁忙的 队列#次 16) 当CPU繁忙时,在调度域中找到了更繁忙的队列,但未找到更繁忙的调度组 #次 - 17) 当CPU新空闲时,load_balance()在这个调度域中被调用了#次 - 18) 当CPU新空闲时,load_balance()在这个调度域中被调用,但是发现负载无需 + 17) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用了#次 + 18) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,但是发现负载无需 均衡#次 - 19) 当CPU新空闲时,load_balance()在这个调度域中被调用,试图迁移1个或更多 + 19) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,试图迁移1个或更多 任务且失败了#次 - 20) 当CPU新空闲时,load_balance()在这个调度域中被调用,发现不均衡(如果有) + 20) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,发现不均衡(如果有) #次 21) 当CPU新空闲时,pull_task()在这个调度域中被调用#次 22) 当CPU新空闲时,尽管目标任务是热缓存状态,pull_task()依然被调用#次 - 23) 当CPU新空闲时,load_balance()在这个调度域中被调用,未能找到更繁忙的 + 23) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,未能找到更繁忙的 队列#次 24) 当CPU新空闲时,在调度域中找到了更繁忙的队列,但未找到更繁忙的调度组 #次
include/linux/sched/topology.h +1 -1
@@ -110,7 +110,7 @@ struct sched_domain {unsigned long last_decay_max_lb_cost;#ifdef CONFIG_SCHEDSTATS -/* load_balance() stats */ +/* sched_balance_rq() stats */ unsigned int lb_count[CPU_MAX_IDLE_TYPES ];unsigned int lb_failed[CPU_MAX_IDLE_TYPES ];unsigned int lb_balanced[CPU_MAX_IDLE_TYPES ];
kernel/sched/fair.c +5 -5
@@ -6866,7 +6866,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)#ifdef CONFIG_SMP -/* Working cpumask for: load_balance, load_balance_newidle. */ +/* Working cpumask for: sched_balance_rq, load_balance_newidle. */ static DEFINE_PER_CPU(cpumask_var_t, load_balance_mask); static DEFINE_PER_CPU(cpumask_var_t, select_rq_mask); static DEFINE_PER_CPU(cpumask_var_t, should_we_balance_tmpmask);
@@ -11242,7 +11242,7 @@ static int should_we_balance(struct lb_env *env)* Check this_cpu to ensure it is balanced within domain. Attempt to move * tasks if there is an imbalance. */ -static int load_balance (int this_cpu,struct rq *this_rq, +static int sched_balance_rq (int this_cpu,struct rq *this_rq,struct sched_domain *sd,enum cpu_idle_type idle,int *continue_balancing) {
@@ -11647,7 +11647,7 @@ static int active_load_balance_cpu_stop(void *data)static atomic_t sched_balance_running =ATOMIC_INIT (0 );/* -* Scale the max load_balance interval with the number of CPUs in the system. +* Scale the max sched_balance_rq interval with the number of CPUs in the system. * This trades load-balance latency on larger machines for less cross talk. */ void update_max_interval (void )
@@ -11727,7 +11727,7 @@ static void sched_balance_domains(struct rq *rq, enum cpu_idle_type idle) }if (time_after_eq (jiffies, sd->last_balance + interval)) { -if (load_balance (cpu, rq, sd, idle, &continue_balancing)) { +if (sched_balance_rq (cpu, rq, sd, idle, &continue_balancing)) {/* * The LBF_DST_PINNED logic could have changed * env->dst_cpu, so we can't know our idle
@@ -12353,7 +12353,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)if (sd->flags &SD_BALANCE_NEWIDLE ) { - pulled_task =load_balance (this_cpu, this_rq, + pulled_task =sched_balance_rq (this_cpu, this_rq, sd,CPU_NEWLY_IDLE , &continue_balancing);
Signature · 1 line
2.40.1
PATCH 6/13 sched/balancing: Rename find_busiest_queue() => find_src_rq()
In reply to Ingo MolnarThe find_busiest_queue() naming has two small quirks:
- Scheduler functions that deal with runqueues usually have a rq_ prefix or _rq postfix, but this function has neither.
- Plus the 'busiest' qualifier to this function was historically correct, but has become somewhat of a misnomer: in quite a few cases we will not pick the busiest runqueue - but the best (possible) runqueue we can balance tasks from. So name it a bit more neutrally, similar to the 'src/dst' nomenclature we are already using when moving tasks between runqueues.
To fix both quirks, rename it to find_src_rq().
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -10959,9 +10959,9 @@ static struct sched_group *find_busiest_group(struct lb_env *env) }/* -* find_busiest_queue - find the busiest runqueue among the CPUs in the group. +* find_src_rq - find the busiest runqueue among the CPUs in the group. */ -static struct rq *find_busiest_queue (struct lb_env *env, +static struct rq *find_src_rq (struct lb_env *env,struct sched_group *group) {struct rq *busiest =NULL , *rq;
@@ -11280,7 +11280,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq, goto out_balanced; } - busiest =find_busiest_queue (&env, group); + busiest =find_src_rq (&env, group);if (!busiest) {schedstat_inc (sd->lb_nobusyq[idle]); goto out_balanced;
Signature · 1 line
2.40.1
PATCH 7/13 sched/balancing: Rename find_src_rq() => sched_balance_find_src_rq()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -10959,9 +10959,9 @@ static struct sched_group *find_busiest_group(struct lb_env *env) }/* -* find_src_rq - find the busiest runqueue among the CPUs in the group. +* sched_balance_find_src_rq - find the busiest runqueue among the CPUs in the group. */ -static struct rq *find_src_rq (struct lb_env *env, +static struct rq *sched_balance_find_src_rq (struct lb_env *env,struct sched_group *group) {struct rq *busiest =NULL , *rq;
@@ -11280,7 +11280,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq, goto out_balanced; } - busiest =find_src_rq (&env, group); + busiest =sched_balance_find_src_rq (&env, group);if (!busiest) {schedstat_inc (sd->lb_nobusyq[idle]); goto out_balanced;
Signature · 1 line
2.40.1
PATCH 8/13 sched/balancing: Rename find_busiest_group() => sched_balance_find_src_group()
In reply to Ingo MolnarMake two naming changes:
1)
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
2)
Similar to find_busiest_queue(), the find_busiest_group() naming has become a bit of a misnomer: the 'busiest' qualifier to this function was historically correct but in the current code
in quite a few cases we will not pick the 'busiest' group - but the best (possible) group we can balance from based on a complex set of constraints.
So name it a bit more neutrally, similar to the 'src/dst' nomenclature we are already using when moving tasks between runqueues, and also use the sched_balance_ prefix: sched_balance_find_src_group().
--- kernel/sched/fair.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
kernel/sched/fair.c +7 -7
@@ -9430,7 +9430,7 @@ static void update_blocked_averages(int cpu)rq_unlock_irqrestore (rq, &rf); } -/********** Helpers for find_busiest_group ************************/ +/********** Helpers for sched_balance_find_src_group ************************/ /* * sg_lb_stats - stats of a sched_group required for load-balancing:
@@ -9637,7 +9637,7 @@ static inline int check_misfit_status(struct rq *rq, struct sched_domain *sd)* * When this is so detected; this group becomes a candidate for busiest; see * update_sd_pick_busiest(). And calculate_imbalance() and -* find_busiest_group() avoid some of the usual balance conditions to allow it +* sched_balance_find_src_group() avoid some of the usual balance conditions to allow it * to create an effective group imbalance. * * This is a somewhat tricky proposition since the next run might not find the
@@ -10788,7 +10788,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s ) / SCHED_CAPACITY_SCALE; } -/******* find_busiest_group() helpers end here *********************/ +/******* sched_balance_find_src_group() helpers end here *********************/ /* * Decision matrix according to the local and busiest group type:
@@ -10811,7 +10811,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s*/ /** -* find_busiest_group - Returns the busiest group within the sched_domain +* sched_balance_find_src_group - Returns the busiest group within the sched_domain * if there is an imbalance. * @env: The load balancing environment. *
@@ -10820,7 +10820,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s* * Return: - The busiest group if imbalance exists. */ -static struct sched_group *find_busiest_group (struct lb_env *env) +static struct sched_group *sched_balance_find_src_group (struct lb_env *env) {struct sg_lb_stats *local, *busiest;struct sd_lb_stats sds;
@@ -11274,7 +11274,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq, goto out_balanced; } - group =find_busiest_group (&env); + group =sched_balance_find_src_group (&env);if (!group) {schedstat_inc (sd->lb_nobusyg[idle]); goto out_balanced;
@@ -11298,7 +11298,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq, env.flags |=LBF_ALL_PINNED ;if (busiest->nr_running >1 ) {/* -* Attempt to move tasks. If find_busiest_group has found +* Attempt to move tasks. If sched_balance_find_src_group has found * an imbalance but busiest->nr_running <= 1, the group is * still unbalanced. ld_moved simply stays zero, so it is * correctly treated as an imbalance.
Signature · 1 line
2.40.1
PATCH 9/13 sched/balancing: Rename update_blocked_averages() => sched_balance_update_blocked_averages()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
--- kernel/sched/fair.c | 8 ++++---- kernel/sched/pelt.c | 2 +- 2 files changed, 5 insertions(+), 5 deletions(-)
kernel/sched/fair.c +4 -4
@@ -9411,7 +9411,7 @@ static unsigned long task_h_load(struct task_struct *p) }#endif -static void update_blocked_averages (int cpu) +static void sched_balance_update_blocked_averages (int cpu) {bool decayed = false, done = true;struct rq *rq =cpu_rq (cpu);
@@ -12079,7 +12079,7 @@ static bool update_nohz_stats(struct rq *rq)if (!time_after (jiffies,READ_ONCE (rq->last_blocked_load_update_tick)))return true; -update_blocked_averages (cpu); +sched_balance_update_blocked_averages (cpu);return rq->has_blocked_load; }
@@ -12339,7 +12339,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)raw_spin_rq_unlock (this_rq); t0 =sched_clock_cpu (this_cpu); -update_blocked_averages (this_cpu); +sched_balance_update_blocked_averages (this_cpu);rcu_read_lock ();for_each_domain (this_cpu, sd) {
@@ -12431,7 +12431,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)return ;/* normal load balance */ -update_blocked_averages (this_rq->cpu); +sched_balance_update_blocked_averages (this_rq->cpu);sched_balance_domains (this_rq, idle); }
kernel/sched/pelt.c +1 -1
@@ -209,7 +209,7 @@ ___update_load_sum(u64 now, struct sched_avg *sa,* This means that weight will be 0 but not running for a sched_entity * but also for a cfs_rq if the latter becomes idle. As an example, * this happens during idle_balance() which calls -* update_blocked_averages(). +* sched_balance_update_blocked_averages(). * * Also see the comment in accumulate_sum(). */
Signature · 1 line
2.40.1
PATCH 10/13 sched/balancing: Rename newidle_balance() => sched_balance_newidle()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
--- kernel/sched/fair.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
kernel/sched/fair.c +8 -8
@@ -4816,7 +4816,7 @@ static inline unsigned long cfs_rq_load_avg(struct cfs_rq *cfs_rq)return cfs_rq->avg.load_avg; } -static int newidle_balance (struct rq *this_rq,struct rq_flags *rf); +static int sched_balance_newidle (struct rq *this_rq,struct rq_flags *rf);static inline unsigned long task_util (struct task_struct *p) {
@@ -5136,7 +5136,7 @@ attach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) {}static inline void detach_entity_load_avg (struct cfs_rq *cfs_rq,struct sched_entity *se) {} -static inline int newidle_balance (struct rq *rq,struct rq_flags *rf) +static inline int sched_balance_newidle (struct rq *rq,struct rq_flags *rf) {return 0 ; }
@@ -8253,7 +8253,7 @@ balance_fair(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)if (rq->nr_running)return 1 ; - return newidle_balance(rq, rf) != 0; + return sched_balance_newidle(rq, rf) != 0; }#endif /* CONFIG_SMP */
@@ -8505,10 +8505,10 @@ done: __maybe_unused;if (!rf)return NULL ; - new_tasks =newidle_balance (rq, rf); + new_tasks =sched_balance_newidle (rq, rf);/* -* Because newidle_balance() releases (and re-acquires) rq->lock, it is +* Because sched_balance_newidle() releases (and re-acquires) rq->lock, it is * possible for any higher priority task to appear. In that case we * must re-start the pick_next_entity() loop. */
@@ -11493,7 +11493,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq, ld_moved =0 ;/* -* newidle_balance() disregards balance intervals, so we could +* sched_balance_newidle() disregards balance intervals, so we could * repeatedly reach this code, which would lead to balance_interval * skyrocketing in a short amount of time. Skip the balance_interval * increase logic to avoid that.
@@ -12277,7 +12277,7 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }#endif /* CONFIG_NO_HZ_COMMON */ /* -* newidle_balance is called by schedule() if this_cpu is about to become +* sched_balance_newidle is called by schedule() if this_cpu is about to become * idle. Attempts to pull tasks from other CPUs. * * Returns:
@@ -12285,7 +12285,7 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }* 0 - failed, no new tasks * > 0 - success, new (fair) tasks present */ -static int newidle_balance (struct rq *this_rq,struct rq_flags *rf) +static int sched_balance_newidle (struct rq *this_rq,struct rq_flags *rf) {unsigned long next_balance = jiffies +HZ ;int this_cpu = this_rq->cpu;
Signature · 1 line
2.40.1
PATCH 11/13 sched/balancing: Rename find_idlest_group_cpu() => sched_balance_find_dst_group_cpu()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
Also use 'dst' instead of 'idlest': while historically correct, today it's not really true anymore that we return the 'idlest' group or CPU, we sort by idle-exit latency and only return the idlest CPUs from the lowest-latency set of CPUs.
The true 'idlest' CPUs often remain idle for a long time and are never returned as long as the system is under-loaded.
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -7101,10 +7101,10 @@ static struct sched_group * find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu);/* -* find_idlest_group_cpu - find the idlest CPU among the CPUs in the group. +* sched_balance_find_dst_group_cpu - find the idlest CPU among the CPUs in the group. */ static int -find_idlest_group_cpu (struct sched_group *group,struct task_struct *p,int this_cpu) +sched_balance_find_dst_group_cpu (struct sched_group *group,struct task_struct *p,int this_cpu) {unsigned long load, min_load =ULONG_MAX ;unsigned int min_exit_latency =UINT_MAX ;
@@ -7191,7 +7191,7 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *pcontinue ; } - new_cpu =find_idlest_group_cpu (group, p, cpu); + new_cpu =sched_balance_find_dst_group_cpu (group, p, cpu);if (new_cpu == cpu) {/* Now try balancing at a lower domain level of 'cpu': */ sd = sd->child;
Signature · 1 line
2.40.1
PATCH 12/13 sched/balancing: Rename find_idlest_group() => sched_balance_find_dst_group()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
Also use 'dst' instead of 'idlest', because it's not really true that we return the 'idlest' group or CPU, we sort by idle-exit latency and only return the idlest CPUs from the lowest-latency set of CPUs.
The true 'idlest' CPUs often remain idle for a long time and are never returned as long as the system is under-loaded.
--- kernel/sched/fair.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
kernel/sched/fair.c +4 -4
@@ -7098,7 +7098,7 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, }static struct sched_group * -find_idlest_group (struct sched_domain *sd,struct task_struct *p,int this_cpu); +sched_balance_find_dst_group (struct sched_domain *sd,struct task_struct *p,int this_cpu);/* * sched_balance_find_dst_group_cpu - find the idlest CPU among the CPUs in the group.
@@ -7185,7 +7185,7 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *pcontinue ; } - group =find_idlest_group (sd, p, cpu); + group =sched_balance_find_dst_group (sd, p, cpu);if (!group) { sd = sd->child;continue ;
@@ -10296,13 +10296,13 @@ static bool update_pick_idlest(struct sched_group *idlest, }/* -* find_idlest_group() finds and returns the least busy CPU group within the +* sched_balance_find_dst_group() finds and returns the least busy CPU group within the * domain. * * Assumes p is allowed on at least one CPU in sd. */ static struct sched_group * -find_idlest_group (struct sched_domain *sd,struct task_struct *p,int this_cpu) +sched_balance_find_dst_group (struct sched_domain *sd,struct task_struct *p,int this_cpu) {struct sched_group *idlest =NULL , *local =NULL , *group = sd->groups;struct sg_lb_stats local_sgs, tmp_sgs;
Signature · 1 line
2.40.1
PATCH 13/13 sched/balancing: Rename find_idlest_cpu() => sched_balance_find_dst_cpu()
In reply to Ingo MolnarStandardize scheduler load-balancing function names on the sched_balance_() prefix.
Also use 'dst' instead of 'idlest', because it's not really true that we return the 'idlest' group or CPU, we sort by idle-exit latency and only return the idlest CPUs from the lowest-latency set of CPUs.
The true 'idlest' CPUs often remain idle for a long time and are never returned as long as the system is under-loaded.
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -7160,7 +7160,7 @@ sched_balance_find_dst_group_cpu(struct sched_group *group, struct task_struct *return shallowest_idle_cpu !=-1 ? shallowest_idle_cpu : least_loaded_cpu; } -static inline int find_idlest_cpu (struct sched_domain *sd,struct task_struct *p, +static inline int sched_balance_find_dst_cpu (struct sched_domain *sd,struct task_struct *p,int cpu,int prev_cpu,int sd_flag) {int new_cpu = cpu;
@@ -7936,7 +7936,7 @@ compute_energy(struct energy_env *eenv, struct perf_domain *pd,* NOTE: Forkees are not accepted in the energy-aware wake-up path because * they don't have any useful utilization data yet and it's not possible to * forecast their impact on energy consumption. Consequently, they will be -* placed by find_idlest_cpu() on the least loaded CPU, which might turn out +* placed by sched_balance_find_dst_cpu() on the least loaded CPU, which might turn out * to be energy-inefficient in some use-cases. The alternative would be to * bias new tasks towards specific types of CPUs first, or to try to infer * their util_avg from the parent task, but those heuristics could hurt
@@ -8201,7 +8201,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags)if (unlikely (sd)) {/* Slow path */ - new_cpu =find_idlest_cpu (sd, p, cpu, prev_cpu, sd_flag); + new_cpu =sched_balance_find_dst_cpu (sd, p, cpu, prev_cpu, sd_flag); }else if (wake_flags &WF_TTWU ) {/* XXX always ? */ /* Fast path */ new_cpu =select_idle_sibling (p, prev_cpu, new_cpu);
Signature · 1 line
2.40.1
Quoted 14 lines · Ingo Molnar
Over the years we've grown a colorful zoo of scheduler load-balancing function names - both following random, idiosyncratic patterns, and gaining historic misnomers that are not accurate anymore.
We have '
newidle_balance()' to rebalance newly idle tasks, but 'balance_domains()' to rebalance domains. We have afind_idlest_cpu()function whose purpose is not to find the idlest CPU anymore, and afind_busiest_queue()function whose purpose is not to find the busiest runqueue anymore.Fix most of the misnomers and organize the functions along the sched_balance_*() namespace:
Quoted 12 lines
scheduler_tick()=>sched_tick()
run_rebalance_domains()=>sched_balance_softirq()
trigger_load_balance()=>sched_balance_trigger()
rebalance_domains()=>sched_balance_domains()
load_balance()=>sched_balance_rq()
newidle_balance()=>sched_balance_newidle()
find_busiest_queue()=>sched_balance_find_src_rq()
find_busiest_group()=>sched_balance_find_src_group()
find_idlest_group_cpu()=>sched_balance_find_dst_group_cpu()
find_idlest_group()=>sched_balance_find_dst_group()
find_idlest_cpu()=>sched_balance_find_dst_cpu()
update_blocked_averages()=>sched_balance_update_blocked_averages()
Forgot to mention that this series is on top of the scheduler tree (tip:sched/core) plus my other pending queue:
https://lore.kernel.org/r/[email protected]
Thanks,
Ingo
Vincent Guittot
maintainerRe: [PATCH 07/13] sched/balancing: Rename find_src_rq() => sched_balance_find_src_rq()
In reply to Ingo MolnarQuoted 3 lines · Ingo Molnar
Standardize scheduler load-balancing function names on the
sched_balance_()prefix.
This patch renames the renaming done by the previous one. They could be merged in one
sched/balancing: Rename find_busiest_queue() => find_src_rq() sched/balancing: Rename find_src_rq() => sched_balance_find_src_rq()
Quoted 40 lines
Signed-off-by Ingo Molnar <[email protected]>Cc Dietmar Eggemann <[email protected]>Cc Linus Torvalds <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Shrikanth Hegde <[email protected]>Cc Valentin Schneider <[email protected]>Cc Vincent Guittot <[email protected]>--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)kernel/sched/fair.c +3
-4@@ -10959,9 +10959,9 @@ static struct sched_group *find_busiest_group(struct lb_env *env) }/* -* find_src_rq - find the busiest runqueue among the CPUs in the group. +* sched_balance_find_src_rq - find the busiest runqueue among the CPUs in the group. */ -static struct rq *find_src_rq (struct lb_env *env, +static struct rq *sched_balance_find_src_rq (struct lb_env *env,struct sched_group *group) {struct rq *busiest =NULL , *rq;@@ -11280,7 +11280,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq, goto out_balanced; } - busiest =find_src_rq (&env, group); + busiest =sched_balance_find_src_rq (&env, group);if (!busiest) {schedstat_inc (sd->lb_nobusyq[idle]); goto out_balanced; --2.40.1
Re: [PATCH 07/13] sched/balancing: Rename find_src_rq() => sched_balance_find_src_rq()
Quoted 10 lines · Vincent Guittot
Quoted 3 lines · Ingo Molnar
Standardize scheduler load-balancing function names on the
sched_balance_()prefix.This patch renames the renaming done by the previous one. They could be merged in one
sched/balancing: Rename
find_busiest_queue()=>find_src_rq()sched/balancing: Renamefind_src_rq()=>sched_balance_find_src_rq()
Yeah - I already did that in the 00/13 summary description, have done it in the series as well.
Thanks,
Ingo
Re: [PATCH 09/13] sched/balancing: Rename update_blocked_averages() => sched_balance_update_blocked_averages()
In reply to Ingo MolnarQuoted 60 lines · Ingo Molnar
Standardize scheduler load-balancing function names on the
sched_balance_()prefix.Signed-off-by Ingo Molnar <[email protected]>Cc Dietmar Eggemann <[email protected]>Cc Linus Torvalds <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Shrikanth Hegde <[email protected]>Cc Valentin Schneider <[email protected]>Cc Vincent Guittot <[email protected]>--- kernel/sched/fair.c | 8 ++++---- kernel/sched/pelt.c | 2 +- 2 files changed, 5 insertions(+), 5 deletions(-)kernel/sched/fair.c +4
-4@@ -9411,7 +9411,7 @@ static unsigned long task_h_load(struct task_struct *p) }#endif -static void update_blocked_averages (int cpu) +static void sched_balance_update_blocked_averages (int cpu) {bool decayed = false, done = true;struct rq *rq =cpu_rq (cpu);@@ -12079,7 +12079,7 @@ static bool update_nohz_stats(struct rq *rq)if (!time_after (jiffies,READ_ONCE (rq->last_blocked_load_update_tick)))return true; -update_blocked_averages (cpu); +sched_balance_update_blocked_averages (cpu);return rq->has_blocked_load; }@@ -12339,7 +12339,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)raw_spin_rq_unlock (this_rq); t0 =sched_clock_cpu (this_cpu); -update_blocked_averages (this_cpu); +sched_balance_update_blocked_averages (this_cpu);rcu_read_lock ();for_each_domain (this_cpu, sd) {@@ -12431,7 +12431,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)return ;/* normal load balance */ -update_blocked_averages (this_rq->cpu); +sched_balance_update_blocked_averages (this_rq->cpu);sched_balance_domains (this_rq, idle); }kernel/sched/pelt.c +0
-0
kernel/sched/pelt.c ___update_load_sum(u64 now, struct sched_avg *sa, · line 211
Quoted 3 lines
kernel/sched/pelt.c +0
-0@@ -209,7 +209,7 @@ ___update_load_sum(u64 now, struct sched_avg *sa,* This means that weight will be 0 but not running for a sched_entity * but also for a cfs_rq if the latter becomes idle. As an example, * this happens during idle_balance() which calls
Could we also fix this ghost idle_balance() in this serial (maybe in patch 10)?
Honglei
Quoted 5 lines
kernel/sched/pelt.c +1
-1-* update_blocked_averages(). +* sched_balance_update_blocked_averages(). * * Also see the comment in accumulate_sum(). */
Re: [PATCH 05/13] sched/balancing: Rename load_balance() => sched_balance_rq()
In reply to Ingo MolnarQuoted 9 lines · Ingo Molnar
Standardize scheduler load-balancing function names on the
sched_balance_()prefix.Also
load_balance()has become somewhat of a misnomer: historically it was the first and primary load-balancing function that was called, but with the introduction of sched domains, it's become a lower layer function that balances runqueues.Rename it to
sched_balance_rq()accordingly.
nit: Can this be sched_balance_rqs()? since load balancing happens between two runqeueus.
Quoted 8 lines
Signed-off-by Ingo Molnar <[email protected]>Cc Dietmar Eggemann <[email protected]>Cc Linus Torvalds <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Shrikanth Hegde <[email protected]>Cc Valentin Schneider <[email protected]>Cc Vincent Guittot <[email protected]>---
Though one would have been familiar with names(for someone started recently), given the correct behaviour and historical context helps why the name changes are making sense.
Re: [PATCH 05/13] sched/balancing: Rename load_balance() => sched_balance_rq()
Quoted 11 lines · Shrikanth Hegde
Quoted 8 lines · Ingo Molnar
Standardize scheduler load-balancing function names on the
sched_balance_()prefix.Also
load_balance()has become somewhat of a misnomer: historically it was the first and primary load-balancing function that was called, but with the introduction of sched domains, it's become a lower layer function that balances runqueues.
Quoted 4 lines
Quoted 1 line
Rename it to
sched_balance_rq()accordingly.nit: Can this be
sched_balance_rqs()? since load balancing happens between two runqeueus.
Yeah, but we really are primarily balancing *this* runqueue - because it got potentially out of balance due to a newidle event, or we are checking its balance in the periodic load-balancing tick. So it's really a shortcut for 'balance this runqueue' - singular, although internally it will indeed search for a source runqueue to move tasks from.
So it's a kind of a pull-balancing model, with a singular target (this_cpu).
Quoted 9 lines
Quoted 8 lines
Signed-off-by Ingo Molnar <[email protected]>Cc Dietmar Eggemann <[email protected]>Cc Linus Torvalds <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Shrikanth Hegde <[email protected]>Cc Valentin Schneider <[email protected]>Cc Vincent Guittot <[email protected]>---
Quoted 4 lines
Though one would have been familiar with names(for someone started recently), given the correct behaviour and historical context helps why the name changes are making sense.
Reviewed-by Shrikanth Hegde <[email protected]>
Thanks! I've added your Reviewed-by tags to the series.
Ingo
Re: [PATCH 09/13] sched/balancing: Rename update_blocked_averages() => sched_balance_update_blocked_averages()
In reply to Honglei WangQuoted 9 lines · Honglei Wang
kernel/sched/pelt.c ___update_load_sum(u64 now, struct sched_avg *sa, · line 211
Quoted 6 lines
kernel/sched/pelt.c +0
-0@@ -209,7 +209,7 @@ ___update_load_sum(u64 now, struct sched_avg *sa,* This means that weight will be 0 but not running for a sched_entity * but also for a cfs_rq if the latter becomes idle. As an example, * this happens during idle_balance() which calls Could we also fix this ghost
idle_balance()in this serial (maybe in patch 10)?
Good point - I've added the patch below.
Thanks,
Ingo
===================>
From: Ingo Molnar <[email protected]>
Date: Tue, 12 Mar 2024 11:33:50 +0100
Subject: [PATCH] sched/balancing: Fix a couple of outdated function names in comments
The 'idle_balance()' function hasn't existed for years, and there's no load_balance_newidle() either - both are sched_balance_newidle() today.
--- kernel/sched/fair.c | 2 +- kernel/sched/pelt.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
kernel/sched/fair.c +1 -1
@@ -6866,7 +6866,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)#ifdef CONFIG_SMP -/* Working cpumask for: sched_balance_rq, load_balance_newidle. */ +/* Working cpumask for: sched_balance_rq(), sched_balance_newidle(). */ static DEFINE_PER_CPU(cpumask_var_t, load_balance_mask); static DEFINE_PER_CPU(cpumask_var_t, select_rq_mask); static DEFINE_PER_CPU(cpumask_var_t, should_we_balance_tmpmask);
kernel/sched/pelt.c +1 -1
@@ -208,7 +208,7 @@ ___update_load_sum(u64 now, struct sched_avg *sa,* se has been already dequeued but cfs_rq->curr still points to it. * This means that weight will be 0 but not running for a sched_entity * but also for a cfs_rq if the latter becomes idle. As an example, -* this happens during idle_balance() which calls +* this happens during sched_balance_newidle() which calls * sched_balance_update_blocked_averages(). * * Also see the comment in accumulate_sum().
[tip: sched/core] sched/balancing: Rename find_idlest_cpu() => sched_balance_find_dst_cpu()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 686d148cbb5a1c2891914b8d11147d3c5556a29a Gitweb: https://git.kernel.org/tip/686d148cbb5a1c2891914b8d11147d3c5556a29a Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:19 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename find_idlest_cpu() => sched_balance_find_dst_cpu()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
Also use 'dst' instead of 'idlest', because it's not really true that we return the 'idlest' group or CPU, we sort by idle-exit latency and only return the idlest CPUs from the lowest-latency set of CPUs.
The true 'idlest' CPUs often remain idle for a long time and are never returned as long as the system is under-loaded.
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -7160,7 +7160,7 @@ sched_balance_find_dst_group_cpu(struct sched_group *group, struct task_struct *return shallowest_idle_cpu !=-1 ? shallowest_idle_cpu : least_loaded_cpu; } -static inline int find_idlest_cpu (struct sched_domain *sd,struct task_struct *p, +static inline int sched_balance_find_dst_cpu (struct sched_domain *sd,struct task_struct *p,int cpu,int prev_cpu,int sd_flag) {int new_cpu = cpu;
@@ -7936,7 +7936,7 @@ compute_energy(struct energy_env *eenv, struct perf_domain *pd,* NOTE: Forkees are not accepted in the energy-aware wake-up path because * they don't have any useful utilization data yet and it's not possible to * forecast their impact on energy consumption. Consequently, they will be -* placed by find_idlest_cpu() on the least loaded CPU, which might turn out +* placed by sched_balance_find_dst_cpu() on the least loaded CPU, which might turn out * to be energy-inefficient in some use-cases. The alternative would be to * bias new tasks towards specific types of CPUs first, or to try to infer * their util_avg from the parent task, but those heuristics could hurt
@@ -8201,7 +8201,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags)if (unlikely (sd)) {/* Slow path */ - new_cpu =find_idlest_cpu (sd, p, cpu, prev_cpu, sd_flag); + new_cpu =sched_balance_find_dst_cpu (sd, p, cpu, prev_cpu, sd_flag); }else if (wake_flags &WF_TTWU ) {/* XXX always ? */ /* Fast path */ new_cpu =select_idle_sibling (p, prev_cpu, new_cpu);
[tip: sched/core] sched/balancing: Rename find_idlest_group() => sched_balance_find_dst_group()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: a88b17080294f735c4124acccfa2d803a6a7d46f Gitweb: https://git.kernel.org/tip/a88b17080294f735c4124acccfa2d803a6a7d46f Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:18 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename find_idlest_group() => sched_balance_find_dst_group()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
Also use 'dst' instead of 'idlest', because it's not really true that we return the 'idlest' group or CPU, we sort by idle-exit latency and only return the idlest CPUs from the lowest-latency set of CPUs.
The true 'idlest' CPUs often remain idle for a long time and are never returned as long as the system is under-loaded.
--- kernel/sched/fair.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
kernel/sched/fair.c +4 -4
@@ -7098,7 +7098,7 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, }static struct sched_group * -find_idlest_group (struct sched_domain *sd,struct task_struct *p,int this_cpu); +sched_balance_find_dst_group (struct sched_domain *sd,struct task_struct *p,int this_cpu);/* * sched_balance_find_dst_group_cpu - find the idlest CPU among the CPUs in the group.
@@ -7185,7 +7185,7 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *pcontinue ; } - group =find_idlest_group (sd, p, cpu); + group =sched_balance_find_dst_group (sd, p, cpu);if (!group) { sd = sd->child;continue ;
@@ -10296,13 +10296,13 @@ static bool update_pick_idlest(struct sched_group *idlest, }/* -* find_idlest_group() finds and returns the least busy CPU group within the +* sched_balance_find_dst_group() finds and returns the least busy CPU group within the * domain. * * Assumes p is allowed on at least one CPU in sd. */ static struct sched_group * -find_idlest_group (struct sched_domain *sd,struct task_struct *p,int this_cpu) +sched_balance_find_dst_group (struct sched_domain *sd,struct task_struct *p,int this_cpu) {struct sched_group *idlest =NULL , *local =NULL , *group = sd->groups;struct sg_lb_stats local_sgs, tmp_sgs;
[tip: sched/core] sched/balancing: Rename newidle_balance() => sched_balance_newidle()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 7d058285cd77cc1411c91efd1b1673530bb1bee8 Gitweb: https://git.kernel.org/tip/7d058285cd77cc1411c91efd1b1673530bb1bee8 Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:16 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename newidle_balance() => sched_balance_newidle()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
--- kernel/sched/fair.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
kernel/sched/fair.c +8 -8
@@ -4816,7 +4816,7 @@ static inline unsigned long cfs_rq_load_avg(struct cfs_rq *cfs_rq)return cfs_rq->avg.load_avg; } -static int newidle_balance (struct rq *this_rq,struct rq_flags *rf); +static int sched_balance_newidle (struct rq *this_rq,struct rq_flags *rf);static inline unsigned long task_util (struct task_struct *p) {
@@ -5136,7 +5136,7 @@ attach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) {}static inline void detach_entity_load_avg (struct cfs_rq *cfs_rq,struct sched_entity *se) {} -static inline int newidle_balance (struct rq *rq,struct rq_flags *rf) +static inline int sched_balance_newidle (struct rq *rq,struct rq_flags *rf) {return 0 ; }
@@ -8253,7 +8253,7 @@ balance_fair(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)if (rq->nr_running)return 1 ; - return newidle_balance(rq, rf) != 0; + return sched_balance_newidle(rq, rf) != 0; }#endif /* CONFIG_SMP */
@@ -8505,10 +8505,10 @@ idle:if (!rf)return NULL ; - new_tasks =newidle_balance (rq, rf); + new_tasks =sched_balance_newidle (rq, rf);/* -* Because newidle_balance() releases (and re-acquires) rq->lock, it is +* Because sched_balance_newidle() releases (and re-acquires) rq->lock, it is * possible for any higher priority task to appear. In that case we * must re-start the pick_next_entity() loop. */
@@ -11493,7 +11493,7 @@ out_one_pinned: ld_moved =0 ;/* -* newidle_balance() disregards balance intervals, so we could +* sched_balance_newidle() disregards balance intervals, so we could * repeatedly reach this code, which would lead to balance_interval * skyrocketing in a short amount of time. Skip the balance_interval * increase logic to avoid that.
@@ -12277,7 +12277,7 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }#endif /* CONFIG_NO_HZ_COMMON */ /* -* newidle_balance is called by schedule() if this_cpu is about to become +* sched_balance_newidle is called by schedule() if this_cpu is about to become * idle. Attempts to pull tasks from other CPUs. * * Returns:
@@ -12285,7 +12285,7 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }* 0 - failed, no new tasks * > 0 - success, new (fair) tasks present */ -static int newidle_balance (struct rq *this_rq,struct rq_flags *rf) +static int sched_balance_newidle (struct rq *this_rq,struct rq_flags *rf) {unsigned long next_balance = jiffies +HZ ;int this_cpu = this_rq->cpu;
[tip: sched/core] sched/balancing: Rename find_idlest_group_cpu() => sched_balance_find_dst_group_cpu()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 646ebaf51c64c6416ca89765c20041363fc1b518 Gitweb: https://git.kernel.org/tip/646ebaf51c64c6416ca89765c20041363fc1b518 Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:17 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename find_idlest_group_cpu() => sched_balance_find_dst_group_cpu()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
Also use 'dst' instead of 'idlest': while historically correct, today it's not really true anymore that we return the 'idlest' group or CPU, we sort by idle-exit latency and only return the idlest CPUs from the lowest-latency set of CPUs.
The true 'idlest' CPUs often remain idle for a long time and are never returned as long as the system is under-loaded.
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -7101,10 +7101,10 @@ static struct sched_group * find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu);/* -* find_idlest_group_cpu - find the idlest CPU among the CPUs in the group. +* sched_balance_find_dst_group_cpu - find the idlest CPU among the CPUs in the group. */ static int -find_idlest_group_cpu (struct sched_group *group,struct task_struct *p,int this_cpu) +sched_balance_find_dst_group_cpu (struct sched_group *group,struct task_struct *p,int this_cpu) {unsigned long load, min_load =ULONG_MAX ;unsigned int min_exit_latency =UINT_MAX ;
@@ -7191,7 +7191,7 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *pcontinue ; } - new_cpu =find_idlest_group_cpu (group, p, cpu); + new_cpu =sched_balance_find_dst_group_cpu (group, p, cpu);if (new_cpu == cpu) {/* Now try balancing at a lower domain level of 'cpu': */ sd = sd->child;
[tip: sched/core] sched/balancing: Rename find_busiest_group() => sched_balance_find_src_group()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 82cf921432fc184adbbb9c1bced182564876ec5e Gitweb: https://git.kernel.org/tip/82cf921432fc184adbbb9c1bced182564876ec5e Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:14 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename find_busiest_group() => sched_balance_find_src_group()
Make two naming changes:
1)
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
2)
Similar to find_busiest_queue(), the find_busiest_group() naming has become a bit of a misnomer: the 'busiest' qualifier to this function was historically correct but in the current code
in quite a few cases we will not pick the 'busiest' group - but the best (possible) group we can balance from based on a complex set of constraints.
So name it a bit more neutrally, similar to the 'src/dst' nomenclature we are already using when moving tasks between runqueues, and also use the sched_balance_ prefix: sched_balance_find_src_group().
--- kernel/sched/fair.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
kernel/sched/fair.c +7 -7
@@ -9430,7 +9430,7 @@ static void update_blocked_averages(int cpu)rq_unlock_irqrestore (rq, &rf); } -/********** Helpers for find_busiest_group ************************/ +/********** Helpers for sched_balance_find_src_group ************************/ /* * sg_lb_stats - stats of a sched_group required for load-balancing:
@@ -9637,7 +9637,7 @@ static inline int check_misfit_status(struct rq *rq, struct sched_domain *sd)* * When this is so detected; this group becomes a candidate for busiest; see * update_sd_pick_busiest(). And calculate_imbalance() and -* find_busiest_group() avoid some of the usual balance conditions to allow it +* sched_balance_find_src_group() avoid some of the usual balance conditions to allow it * to create an effective group imbalance. * * This is a somewhat tricky proposition since the next run might not find the
@@ -10788,7 +10788,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s ) / SCHED_CAPACITY_SCALE; } -/******* find_busiest_group() helpers end here *********************/ +/******* sched_balance_find_src_group() helpers end here *********************/ /* * Decision matrix according to the local and busiest group type:
@@ -10811,7 +10811,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s*/ /** -* find_busiest_group - Returns the busiest group within the sched_domain +* sched_balance_find_src_group - Returns the busiest group within the sched_domain * if there is an imbalance. * @env: The load balancing environment. *
@@ -10820,7 +10820,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s* * Return: - The busiest group if imbalance exists. */ -static struct sched_group *find_busiest_group (struct lb_env *env) +static struct sched_group *sched_balance_find_src_group (struct lb_env *env) {struct sg_lb_stats *local, *busiest;struct sd_lb_stats sds;
@@ -11274,7 +11274,7 @@ redo: goto out_balanced; } - group =find_busiest_group (&env); + group =sched_balance_find_src_group (&env);if (!group) {schedstat_inc (sd->lb_nobusyg[idle]); goto out_balanced;
@@ -11298,7 +11298,7 @@ redo: env.flags |=LBF_ALL_PINNED ;if (busiest->nr_running >1 ) {/* -* Attempt to move tasks. If find_busiest_group has found +* Attempt to move tasks. If sched_balance_find_src_group has found * an imbalance but busiest->nr_running <= 1, the group is * still unbalanced. ld_moved simply stays zero, so it is * correctly treated as an imbalance.
[tip: sched/core] sched/balancing: Rename update_blocked_averages() => sched_balance_update_blocked_averages()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 391b7a5335c45b2bafe535cb440836ccd17515aa Gitweb: https://git.kernel.org/tip/391b7a5335c45b2bafe535cb440836ccd17515aa Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:15 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename update_blocked_averages() => sched_balance_update_blocked_averages()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
--- kernel/sched/fair.c | 8 ++++---- kernel/sched/pelt.c | 2 +- 2 files changed, 5 insertions(+), 5 deletions(-)
kernel/sched/fair.c +4 -4
@@ -9411,7 +9411,7 @@ static unsigned long task_h_load(struct task_struct *p) }#endif -static void update_blocked_averages (int cpu) +static void sched_balance_update_blocked_averages (int cpu) {bool decayed = false, done = true;struct rq *rq =cpu_rq (cpu);
@@ -12079,7 +12079,7 @@ static bool update_nohz_stats(struct rq *rq)if (!time_after (jiffies,READ_ONCE (rq->last_blocked_load_update_tick)))return true; -update_blocked_averages (cpu); +sched_balance_update_blocked_averages (cpu);return rq->has_blocked_load; }
@@ -12339,7 +12339,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)raw_spin_rq_unlock (this_rq); t0 =sched_clock_cpu (this_cpu); -update_blocked_averages (this_cpu); +sched_balance_update_blocked_averages (this_cpu);rcu_read_lock ();for_each_domain (this_cpu, sd) {
@@ -12431,7 +12431,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)return ;/* normal load balance */ -update_blocked_averages (this_rq->cpu); +sched_balance_update_blocked_averages (this_rq->cpu);sched_balance_domains (this_rq, idle); }
kernel/sched/pelt.c +1 -1
@@ -209,7 +209,7 @@ ___update_load_sum(u64 now, struct sched_avg *sa,* This means that weight will be 0 but not running for a sched_entity * but also for a cfs_rq if the latter becomes idle. As an example, * this happens during idle_balance() which calls -* update_blocked_averages(). +* sched_balance_update_blocked_averages(). * * Also see the comment in accumulate_sum(). */
[tip: sched/core] sched/balancing: Rename find_busiest_queue() => sched_balance_find_src_rq()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: f1cd2e2e79d283e315356bd403c7f928e994f057 Gitweb: https://git.kernel.org/tip/f1cd2e2e79d283e315356bd403c7f928e994f057 Author: Ingo Molnar <[email protected]> AuthorDate: Mon, 23 Oct 2023 13:04:12 +02:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename find_busiest_queue() => sched_balance_find_src_rq()
The find_busiest_queue() naming has two small quirks:
- Scheduler functions that deal with runqueues usually have a rq_ prefix or _rq postfix, but this function has neither.
- Plus the 'busiest' qualifier to this function was historically correct, but has become somewhat of a misnomer: in quite a few cases we will not pick the busiest runqueue - but the best (possible) runqueue we can balance tasks from. So name it a bit more neutrally, similar to the 'src/dst' nomenclature we are already using when moving tasks between runqueues.
To fix both quirks, and to standardize scheduler load-balancing function names on the sched_balance_() prefix, rename the function to sched_balance_find_src_rq().
--- kernel/sched/fair.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/fair.c +3 -3
@@ -10959,9 +10959,9 @@ out_balanced: }/* -* find_busiest_queue - find the busiest runqueue among the CPUs in the group. +* sched_balance_find_src_rq - find the busiest runqueue among the CPUs in the group. */ -static struct rq *find_busiest_queue (struct lb_env *env, +static struct rq *sched_balance_find_src_rq (struct lb_env *env,struct sched_group *group) {struct rq *busiest =NULL , *rq;
@@ -11280,7 +11280,7 @@ redo: goto out_balanced; } - busiest =find_busiest_queue (&env, group); + busiest =sched_balance_find_src_rq (&env, group);if (!busiest) {schedstat_inc (sd->lb_nobusyq[idle]); goto out_balanced;
[tip: sched/core] sched/balancing: Rename load_balance() => sched_balance_rq()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 4c3e509ea9f249458e8692f8298cceac73105948 Gitweb: https://git.kernel.org/tip/4c3e509ea9f249458e8692f8298cceac73105948 Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:11 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 12:00:00 +01:00
sched/balancing: Rename load_balance() => sched_balance_rq()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
Also load_balance() has become somewhat of a misnomer: historically it was the first and primary load-balancing function that was called, but with the introduction of sched domains, it's become a lower layer function that balances runqueues.
Rename it to sched_balance_rq() accordingly.
--- Documentation/scheduler/sched-domains.rst | 4 +- Documentation/scheduler/sched-stats.rst | 32 +++---- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 4 +- Documentation/translations/zh_CN/scheduler/sched-stats.rst | 30 +++---- include/linux/sched/topology.h | 2 +- kernel/sched/fair.c | 10 +- 6 files changed, 41 insertions(+), 41 deletions(-)
Documentation/scheduler/sched-domains.rst +2 -2
@@ -41,11 +41,11 @@ The latter function takes two arguments: the runqueue of current CPU and whether the CPU was idle at the time the sched_tick() happened and iterates over all sched domains our CPU is on, starting from its base domain and going up the ->parent chain. While doing that, it checks to see if the current domain has exhausted its -rebalance interval. If so, it runs load_balance() on that domain. It then checks +rebalance interval. If so, it runs sched_balance_rq() on that domain. It then checks the parent sched_domain (if it exists), and the parent of the parent and so forth. -Initially, load_balance() finds the busiest group in the current sched domain. +Initially, sched_balance_rq() finds the busiest group in the current sched domain. If it succeeds, it looks for the busiest runqueue of all the CPUs' runqueues in that group. If it manages to find such a runqueue, it locks both our initial CPU's runqueue and the newly found busiest one and starts moving tasks from it
Documentation/scheduler/sched-stats.rst +16 -16
@@ -77,53 +77,53 @@ domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 The first field is a bit mask indicating what cpus this domain operates over. -The next 24 are a variety of load_balance() statistics in grouped into types +The next 24 are a variety of sched_balance_rq() statistics in grouped into types of idleness (idle, busy, and newly idle): - 1) # of times in this domain load_balance() was called when the + 1) # of times in this domain sched_balance_rq() was called when the cpu was idle - 2) # of times in this domain load_balance() checked but found + 2) # of times in this domain sched_balance_rq() checked but found the load did not require balancing when the cpu was idle - 3) # of times in this domain load_balance() tried to move one or + 3) # of times in this domain sched_balance_rq() tried to move one or more tasks and failed, when the cpu was idle 4) sum of imbalances discovered (if any) with each call to - load_balance() in this domain when the cpu was idle + sched_balance_rq() in this domain when the cpu was idle 5) # of times in this domain pull_task() was called when the cpu was idle 6) # of times in this domain pull_task() was called even though the target task was cache-hot when idle - 7) # of times in this domain load_balance() was called but did + 7) # of times in this domain sched_balance_rq() was called but did not find a busier queue while the cpu was idle 8) # of times in this domain a busier queue was found while the cpu was idle but no busier group was found - 9) # of times in this domain load_balance() was called when the + 9) # of times in this domain sched_balance_rq() was called when the cpu was busy - 10) # of times in this domain load_balance() checked but found the + 10) # of times in this domain sched_balance_rq() checked but found the load did not require balancing when busy - 11) # of times in this domain load_balance() tried to move one or + 11) # of times in this domain sched_balance_rq() tried to move one or more tasks and failed, when the cpu was busy 12) sum of imbalances discovered (if any) with each call to - load_balance() in this domain when the cpu was busy + sched_balance_rq() in this domain when the cpu was busy 13) # of times in this domain pull_task() was called when busy 14) # of times in this domain pull_task() was called even though the target task was cache-hot when busy - 15) # of times in this domain load_balance() was called but did not + 15) # of times in this domain sched_balance_rq() was called but did not find a busier queue while the cpu was busy 16) # of times in this domain a busier queue was found while the cpu was busy but no busier group was found - 17) # of times in this domain load_balance() was called when the + 17) # of times in this domain sched_balance_rq() was called when the cpu was just becoming idle - 18) # of times in this domain load_balance() checked but found the + 18) # of times in this domain sched_balance_rq() checked but found the load did not require balancing when the cpu was just becoming idle - 19) # of times in this domain load_balance() tried to move one or more + 19) # of times in this domain sched_balance_rq() tried to move one or more tasks and failed, when the cpu was just becoming idle 20) sum of imbalances discovered (if any) with each call to - load_balance() in this domain when the cpu was just becoming idle + sched_balance_rq() in this domain when the cpu was just becoming idle 21) # of times in this domain pull_task() was called when newly idle 22) # of times in this domain pull_task() was called even though the target task was cache-hot when just becoming idle - 23) # of times in this domain load_balance() was called but did not + 23) # of times in this domain sched_balance_rq() was called but did not find a busier queue while the cpu was just becoming idle 24) # of times in this domain a busier queue was found while the cpu was just becoming idle but no busier group was found
Documentation/translations/zh_CN/scheduler/sched-domains.rst +2 -2
@@ -42,9 +42,9 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这� 后一个函数有两个入参:当前CPU的运行队列、它在sched_tick()调用时是否空闲。函数会从 当前CPU所在的基调度域开始迭代执行,并沿着parent指针链向上进入更高层级的调度域。在迭代 过程中,函数会检查当前调度域是否已经耗尽了再平衡的时间间隔,如果是,它在该调度域运行 -load_balance()。接下来它检查父调度域(如果存在),再后来父调度域的父调度域,以此类推。 +sched_balance_rq()。接下来它检查父调度域(如果存在),再后来父调度域的父调度域,以此类推。 -起初,load_balance()查找当前调度域中最繁忙的调度组。如果成功,在该调度组管辖的全部CPU +起初,sched_balance_rq()查找当前调度域中最繁忙的调度组。如果成功,在该调度组管辖的全部CPU 的运行队列中找出最繁忙的运行队列。如能找到,对当前的CPU运行队列和新找到的最繁忙运行 队列均加锁,并把任务从最繁忙队列中迁移到当前CPU上。被迁移的任务数量等于在先前迭代执行 中计算出的该调度域的调度组的不均衡值。
Documentation/translations/zh_CN/scheduler/sched-stats.rst +15 -15
@@ -75,42 +75,42 @@ domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 繁忙,新空闲): - 1) 当CPU空闲时,load_balance()在这个调度域中被调用了#次 - 2) 当CPU空闲时,load_balance()在这个调度域中被调用,但是发现负载无需 + 1) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用了#次 + 2) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,但是发现负载无需 均衡#次 - 3) 当CPU空闲时,load_balance()在这个调度域中被调用,试图迁移1个或更多 + 3) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,试图迁移1个或更多 任务且失败了#次 - 4) 当CPU空闲时,load_balance()在这个调度域中被调用,发现不均衡(如果有) + 4) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,发现不均衡(如果有) #次 5) 当CPU空闲时,pull_task()在这个调度域中被调用#次 6) 当CPU空闲时,尽管目标任务是热缓存状态,pull_task()依然被调用#次 - 7) 当CPU空闲时,load_balance()在这个调度域中被调用,未能找到更繁忙的 + 7) 当CPU空闲时,sched_balance_rq()在这个调度域中被调用,未能找到更繁忙的 队列#次 8) 当CPU空闲时,在调度域中找到了更繁忙的队列,但未找到更繁忙的调度组 #次 - 9) 当CPU繁忙时,load_balance()在这个调度域中被调用了#次 - 10) 当CPU繁忙时,load_balance()在这个调度域中被调用,但是发现负载无需 + 9) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用了#次 + 10) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,但是发现负载无需 均衡#次 - 11) 当CPU繁忙时,load_balance()在这个调度域中被调用,试图迁移1个或更多 + 11) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,试图迁移1个或更多 任务且失败了#次 - 12) 当CPU繁忙时,load_balance()在这个调度域中被调用,发现不均衡(如果有) + 12) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,发现不均衡(如果有) #次 13) 当CPU繁忙时,pull_task()在这个调度域中被调用#次 14) 当CPU繁忙时,尽管目标任务是热缓存状态,pull_task()依然被调用#次 - 15) 当CPU繁忙时,load_balance()在这个调度域中被调用,未能找到更繁忙的 + 15) 当CPU繁忙时,sched_balance_rq()在这个调度域中被调用,未能找到更繁忙的 队列#次 16) 当CPU繁忙时,在调度域中找到了更繁忙的队列,但未找到更繁忙的调度组 #次 - 17) 当CPU新空闲时,load_balance()在这个调度域中被调用了#次 - 18) 当CPU新空闲时,load_balance()在这个调度域中被调用,但是发现负载无需 + 17) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用了#次 + 18) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,但是发现负载无需 均衡#次 - 19) 当CPU新空闲时,load_balance()在这个调度域中被调用,试图迁移1个或更多 + 19) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,试图迁移1个或更多 任务且失败了#次 - 20) 当CPU新空闲时,load_balance()在这个调度域中被调用,发现不均衡(如果有) + 20) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,发现不均衡(如果有) #次 21) 当CPU新空闲时,pull_task()在这个调度域中被调用#次 22) 当CPU新空闲时,尽管目标任务是热缓存状态,pull_task()依然被调用#次 - 23) 当CPU新空闲时,load_balance()在这个调度域中被调用,未能找到更繁忙的 + 23) 当CPU新空闲时,sched_balance_rq()在这个调度域中被调用,未能找到更繁忙的 队列#次 24) 当CPU新空闲时,在调度域中找到了更繁忙的队列,但未找到更繁忙的调度组 #次
include/linux/sched/topology.h +1 -1
@@ -110,7 +110,7 @@ struct sched_domain {unsigned long last_decay_max_lb_cost;#ifdef CONFIG_SCHEDSTATS -/* load_balance() stats */ +/* sched_balance_rq() stats */ unsigned int lb_count[CPU_MAX_IDLE_TYPES ];unsigned int lb_failed[CPU_MAX_IDLE_TYPES ];unsigned int lb_balanced[CPU_MAX_IDLE_TYPES ];
kernel/sched/fair.c +5 -5
@@ -6866,7 +6866,7 @@ dequeue_throttle:#ifdef CONFIG_SMP -/* Working cpumask for: load_balance, load_balance_newidle. */ +/* Working cpumask for: sched_balance_rq, load_balance_newidle. */ static DEFINE_PER_CPU(cpumask_var_t, load_balance_mask); static DEFINE_PER_CPU(cpumask_var_t, select_rq_mask); static DEFINE_PER_CPU(cpumask_var_t, should_we_balance_tmpmask);
@@ -11242,7 +11242,7 @@ static int should_we_balance(struct lb_env *env)* Check this_cpu to ensure it is balanced within domain. Attempt to move * tasks if there is an imbalance. */ -static int load_balance (int this_cpu,struct rq *this_rq, +static int sched_balance_rq (int this_cpu,struct rq *this_rq,struct sched_domain *sd,enum cpu_idle_type idle,int *continue_balancing) {
@@ -11647,7 +11647,7 @@ out_unlock:static atomic_t sched_balance_running =ATOMIC_INIT (0 );/* -* Scale the max load_balance interval with the number of CPUs in the system. +* Scale the max sched_balance_rq interval with the number of CPUs in the system. * This trades load-balance latency on larger machines for less cross talk. */ void update_max_interval (void )
@@ -11727,7 +11727,7 @@ static void sched_balance_domains(struct rq *rq, enum cpu_idle_type idle) }if (time_after_eq (jiffies, sd->last_balance + interval)) { -if (load_balance (cpu, rq, sd, idle, &continue_balancing)) { +if (sched_balance_rq (cpu, rq, sd, idle, &continue_balancing)) {/* * The LBF_DST_PINNED logic could have changed * env->dst_cpu, so we can't know our idle
@@ -12353,7 +12353,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)if (sd->flags &SD_BALANCE_NEWIDLE ) { - pulled_task =load_balance (this_cpu, this_rq, + pulled_task =sched_balance_rq (this_cpu, this_rq, sd,CPU_NEWLY_IDLE , &continue_balancing);
[tip: sched/core] sched/balancing: Rename rebalance_domains() => sched_balance_domains()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 14ff4dbd34f46cc6b6105f549983321241ccbba9 Gitweb: https://git.kernel.org/tip/14ff4dbd34f46cc6b6105f549983321241ccbba9 Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:10 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 11:59:59 +01:00
sched/balancing: Rename rebalance_domains() => sched_balance_domains()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
--- Documentation/scheduler/sched-domains.rst | 2 +- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 2 +- arch/arm/kernel/topology.c | 2 +- kernel/sched/fair.c | 8 +++---- kernel/sched/sched.h | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)
Documentation/scheduler/sched-domains.rst +1 -1
@@ -34,7 +34,7 @@ out of balance are tasks moved between groups. In kernel/sched/core.c, sched_balance_trigger() is run periodically on each CPU through sched_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load -balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run +balancing workhorse, sched_balance_softirq()->sched_balance_domains(), is then run in softirq context (SCHED_SOFTIRQ). The latter function takes two arguments: the runqueue of current CPU and whether
Documentation/translations/zh_CN/scheduler/sched-domains.rst +1 -1
@@ -36,7 +36,7 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这� 在kernel/sched/core.c中,sched_balance_trigger()在每个CPU上通过sched_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 -的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 +的工作由sched_balance_softirq()->sched_balance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。 后一个函数有两个入参:当前CPU的运行队列、它在sched_tick()调用时是否空闲。函数会从
arch/arm/kernel/topology.c +1 -1
@@ -42,7 +42,7 @@* can take this difference into account during load balance. A per cpu * structure is preferred because each CPU updates its own cpu_capacity field * during the load balance except for idle cores. One idle core is selected -* to run the rebalance_domains for all idle cores and the cpu_capacity can be +* to run the sched_balance_domains for all idle cores and the cpu_capacity can be * updated during this sequence. */
kernel/sched/fair.c +4 -4
@@ -11685,7 +11685,7 @@ static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost)* * Balancing parameters are set up in init_sched_domains. */ -static void rebalance_domains (struct rq *rq,enum cpu_idle_type idle) +static void sched_balance_domains (struct rq *rq,enum cpu_idle_type idle) {int continue_balancing =1 ;int cpu = rq->cpu;
@@ -12161,7 +12161,7 @@ static void _nohz_idle_balance(struct rq *this_rq, unsigned int flags)rq_unlock_irqrestore (rq, &rf);if (flags &NOHZ_BALANCE_KICK ) -rebalance_domains (rq,CPU_IDLE ); +sched_balance_domains (rq,CPU_IDLE ); }if (time_after (next_balance, rq->next_balance)) {
@@ -12422,7 +12422,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)/* * If this CPU has a pending NOHZ_BALANCE_KICK, then do the * balancing on behalf of the other idle CPUs whose ticks are -* stopped. Do nohz_idle_balance *before* rebalance_domains to +* stopped. Do nohz_idle_balance *before* sched_balance_domains to * give the idle CPUs a chance to load balance. Else we may * load balance only within the local sched_domain hierarchy * and abort nohz_idle_balance altogether if we pull some load.
@@ -12432,7 +12432,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)/* normal load balance */ update_blocked_averages (this_rq->cpu); -rebalance_domains (this_rq, idle); +sched_balance_domains (this_rq, idle); }/*
kernel/sched/sched.h +1 -1
@@ -2904,7 +2904,7 @@ extern void cfs_bandwidth_usage_dec(void);#define NOHZ_NEWILB_KICK_BIT 2#define NOHZ_NEXT_KICK_BIT 3 -/* Run rebalance_domains() */ +/* Run sched_balance_domains() */ #define NOHZ_BALANCE_KICK BIT(NOHZ_BALANCE_KICK_BIT)/* Update blocked load */ #define NOHZ_STATS_KICK BIT(NOHZ_STATS_KICK_BIT)
[tip: sched/core] sched/balancing: Rename trigger_load_balance() => sched_balance_trigger()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 983be0628c061989b6cc175d2f5e429b40699fbb Gitweb: https://git.kernel.org/tip/983be0628c061989b6cc175d2f5e429b40699fbb Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:09 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 11:59:59 +01:00
sched/balancing: Rename trigger_load_balance() => sched_balance_trigger()
Standardize scheduler load-balancing function names on the sched_balance_() prefix.
--- Documentation/scheduler/sched-domains.rst | 2 +- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 2 +- kernel/sched/core.c | 2 +- kernel/sched/fair.c | 2 +- kernel/sched/sched.h | 2 +- 5 files changed, 5 insertions(+), 5 deletions(-)
Documentation/scheduler/sched-domains.rst +1 -1
@@ -31,7 +31,7 @@ is treated as one entity. The load of a group is defined as the sum of the load of each of its member CPUs, and only when the load of a group becomes out of balance are tasks moved between groups. -In kernel/sched/core.c, trigger_load_balance() is run periodically on each CPU +In kernel/sched/core.c, sched_balance_trigger() is run periodically on each CPU through sched_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run
Documentation/translations/zh_CN/scheduler/sched-domains.rst +1 -1
@@ -34,7 +34,7 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这� 调度域中的负载均衡发生在调度组中。也就是说,每个组被视为一个实体。组的负载被定义为它 管辖的每个CPU的负载之和。仅当组的负载不均衡后,任务才在组之间发生迁移。 -在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过sched_tick() +在kernel/sched/core.c中,sched_balance_trigger()在每个CPU上通过sched_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。
kernel/sched/core.c +1 -1
@@ -5700,7 +5700,7 @@ void sched_tick(void)#ifdef CONFIG_SMP rq->idle_balance =idle_cpu (cpu); -trigger_load_balance (rq); +sched_balance_trigger (rq);#endif }
kernel/sched/fair.c +1 -1
@@ -12438,7 +12438,7 @@ static __latent_entropy void sched_balance_softirq(struct softirq_action *h)/* * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. */ -void trigger_load_balance (struct rq *rq) +void sched_balance_trigger (struct rq *rq) {/* * Don't need to rebalance while attached to NULL domain or
kernel/sched/sched.h +1 -1
@@ -2397,7 +2397,7 @@ extern struct task_struct *pick_next_task_idle(struct rq *rq);extern void update_group_capacity (struct sched_domain *sd,int cpu); -extern void trigger_load_balance (struct rq *rq); +extern void sched_balance_trigger (struct rq *rq);extern void set_cpus_allowed_common (struct task_struct *p,struct affinity_context *ctx);
[tip: sched/core] sched/balancing: Rename scheduler_tick() => sched_tick()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 86dd6c04ef9f213e14d60c9f64bce1cc019f816e Gitweb: https://git.kernel.org/tip/86dd6c04ef9f213e14d60c9f64bce1cc019f816e Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:08 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 11:59:59 +01:00
sched/balancing: Rename scheduler_tick() => sched_tick()
- Standardize on prefixing scheduler-internal functions defined in <linux/sched.h> with sched_*() prefix. scheduler_tick() was the only function using the scheduler_ prefix. Harmonize it.
- The other reason to rename it is the NOHZ scheduler tick handling functions are already named sched_tick_*(). Make the 'git grep sched_tick' more meaningful.
--- Documentation/scheduler/sched-domains.rst | 4 ++-- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 4 ++-- include/linux/sched.h | 2 +- kernel/sched/core.c | 4 ++-- kernel/sched/loadavg.c | 2 +- kernel/time/timer.c | 2 +- kernel/workqueue.c | 2 +- tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc | 2 +- 8 files changed, 11 insertions(+), 11 deletions(-)
Documentation/scheduler/sched-domains.rst +2 -2
@@ -32,13 +32,13 @@ load of each of its member CPUs, and only when the load of a group becomes out of balance are tasks moved between groups. In kernel/sched/core.c, trigger_load_balance() is run periodically on each CPU -through scheduler_tick(). It raises a softirq after the next regularly scheduled +through sched_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run in softirq context (SCHED_SOFTIRQ). The latter function takes two arguments: the runqueue of current CPU and whether -the CPU was idle at the time the scheduler_tick() happened and iterates over all +the CPU was idle at the time the sched_tick() happened and iterates over all sched domains our CPU is on, starting from its base domain and going up the ->parent chain. While doing that, it checks to see if the current domain has exhausted its rebalance interval. If so, it runs load_balance() on that domain. It then checks
Documentation/translations/zh_CN/scheduler/sched-domains.rst +2 -2
@@ -34,12 +34,12 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这� 调度域中的负载均衡发生在调度组中。也就是说,每个组被视为一个实体。组的负载被定义为它 管辖的每个CPU的负载之和。仅当组的负载不均衡后,任务才在组之间发生迁移。 -在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过scheduler_tick() +在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过sched_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。 -后一个函数有两个入参:当前CPU的运行队列、它在scheduler_tick()调用时是否空闲。函数会从 +后一个函数有两个入参:当前CPU的运行队列、它在sched_tick()调用时是否空闲。函数会从 当前CPU所在的基调度域开始迭代执行,并沿着parent指针链向上进入更高层级的调度域。在迭代 过程中,函数会检查当前调度域是否已经耗尽了再平衡的时间间隔,如果是,它在该调度域运行 load_balance()。接下来它检查父调度域(如果存在),再后来父调度域的父调度域,以此类推。
include/linux/sched.h +1 -1
@@ -301,7 +301,7 @@ enum { TASK_COMM_LEN = 16, }; -extern void scheduler_tick (void ); +extern void sched_tick (void );#define MAX_SCHEDULE_TIMEOUT LONG_MAX
kernel/sched/core.c +2 -2
@@ -5662,7 +5662,7 @@ static inline u64 cpu_resched_latency(struct rq *rq) { return 0; }* This function gets called by the timer code, with HZ frequency. * We call it with interrupts disabled. */ -void scheduler_tick (void ) +void sched_tick (void ) {int cpu =smp_processor_id ();struct rq *rq =cpu_rq (cpu);
@@ -6585,7 +6585,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)* paths. For example, see arch/x86/entry_64.S. * * To drive preemption between tasks, the scheduler sets the flag in timer -* interrupt handler scheduler_tick(). +* interrupt handler sched_tick(). * * 3. Wakeups don't really cause entry into schedule(). They add a * task to the run-queue and that's it.
kernel/sched/loadavg.c +1 -1
@@ -379,7 +379,7 @@ void calc_global_load(void) }/* -* Called from scheduler_tick() to periodically update this CPU's +* Called from sched_tick() to periodically update this CPU's * active count. */ void calc_global_load_tick (struct rq *this_rq)
kernel/time/timer.c +1 -1
@@ -2478,7 +2478,7 @@ void update_process_times(int user_tick)if (in_irq ())irq_work_tick ();#endif -scheduler_tick (); +sched_tick ();if (IS_ENABLED (CONFIG_POSIX_TIMERS ))run_posix_cpu_timers (); }
kernel/workqueue.c +1 -1
@@ -1464,7 +1464,7 @@ void wq_worker_sleeping(struct task_struct *task)* wq_worker_tick - a scheduler tick occurred while a kworker is running * @task: task currently running * -* Called from scheduler_tick(). We're in the IRQ context and the current +* Called from sched_tick(). We're in the IRQ context and the current * worker's fields which follow the 'K' locking rule can be accessed safely. */ void wq_worker_tick (struct task_struct *task)
tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc +1 -1
@@ -19,7 +19,7 @@ fail() { # mesg FILTER=set_ftrace_filter FUNC1="schedule" -FUNC2="scheduler_tick" +FUNC2="sched_tick" ALL_FUNCS="#### all functions enabled ####"
[tip: sched/core] sched/balancing: Rename run_rebalance_domains() => sched_balance_softirq()
In reply to Ingo MolnarThe following commit has been merged into the sched/core branch of tip:
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 70a27d6d1b19392a23bb4a41de7788fbc539f18d Gitweb: https://git.kernel.org/tip/70a27d6d1b19392a23bb4a41de7788fbc539f18d Author: Ingo Molnar <[email protected]> AuthorDate: Fri, 08 Mar 2024 12:18:07 +01:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 12 Mar 2024 11:59:59 +01:00
sched/balancing: Rename run_rebalance_domains() => sched_balance_softirq()
run_rebalance_domains() is a misnomer, as it doesn't only run rebalance_domains(), but since the introduction of the NOHZ code it also runs nohz_idle_balance().
Rename it to sched_balance_softirq(), reflecting its more generic purpose and that it's a softirq handler.
--- Documentation/scheduler/sched-domains.rst | 2 +- Documentation/translations/zh_CN/scheduler/sched-domains.rst | 2 +- kernel/sched/fair.c | 4 ++-- 3 files changed, 4 insertions(+), 4 deletions(-)
Documentation/scheduler/sched-domains.rst +1 -1
@@ -34,7 +34,7 @@ out of balance are tasks moved between groups. In kernel/sched/core.c, trigger_load_balance() is run periodically on each CPU through scheduler_tick(). It raises a softirq after the next regularly scheduled rebalancing event for the current runqueue has arrived. The actual load -balancing workhorse, run_rebalance_domains()->rebalance_domains(), is then run +balancing workhorse, sched_balance_softirq()->rebalance_domains(), is then run in softirq context (SCHED_SOFTIRQ). The latter function takes two arguments: the runqueue of current CPU and whether
Documentation/translations/zh_CN/scheduler/sched-domains.rst +1 -1
@@ -36,7 +36,7 @@ CPU共享。任意两个组的CPU掩码的交集不一定为空,如果是这� 在kernel/sched/core.c中,trigger_load_balance()在每个CPU上通过scheduler_tick() 周期执行。在当前运行队列下一个定期调度再平衡事件到达后,它引发一个软中断。负载均衡真正 -的工作由run_rebalance_domains()->rebalance_domains()完成,在软中断上下文中执行 +的工作由sched_balance_softirq()->rebalance_domains()完成,在软中断上下文中执行 (SCHED_SOFTIRQ)。 后一个函数有两个入参:当前CPU的运行队列、它在scheduler_tick()调用时是否空闲。函数会从
kernel/sched/fair.c +2 -2
@@ -12415,7 +12415,7 @@ out:* - indirectly from a remote scheduler_tick() for NOHZ idle balancing * through the SMP cross-call nohz_csd_func() */ -static __latent_entropy void run_rebalance_domains(struct softirq_action *h) +static __latent_entropy void sched_balance_softirq(struct softirq_action *h) {struct rq *this_rq =this_rq ();enum cpu_idle_type idle = this_rq->idle_balance;
@@ -13216,7 +13216,7 @@ __init void init_sched_fair_class(void)#endif } -open_softirq (SCHED_SOFTIRQ , run_rebalance_domains); +open_softirq (SCHED_SOFTIRQ , sched_balance_softirq);#ifdef CONFIG_NO_HZ_COMMON nohz.next_balance = jiffies;