From 5ca3726af7f66a8cc71ce4414cfeb86deb784491 Mon Sep 17 00:00:00 2001 From: Zhao Lei Date: Tue, 22 Mar 2016 16:37:07 +0800 Subject: sched/cpuacct: Show all possible CPUs in cpuacct output Current code show stats of online CPUs in cpuacct.statcpus, show stats of present cpus in cpuacct.usage(_percpu), and using present CPUs for setting cpuacct.usage. It will cause inconsistent result when a CPU is online or offline or hotpluged. We should always use possible CPUs to avoid above problem. Here are the contents of a cpuacct.usage_percpu sysfs file, on a 4 CPU system with maxcpus=32: Before the patch: # cat cpuacct.usage_percpu 2456565 411435 1052897 832584 After the patch: # cat cpuacct.usage_percpu 2456565 411435 1052897 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Suggested-by: Peter Zijlstra Signed-off-by: Zhao Lei Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Tejun Heo Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/a11d56cef12d0b4807f8be3a46bf9798c3014d59.1458635566.git.zhaolei@cn.fujitsu.com Signed-off-by: Ingo Molnar --- kernel/sched/cpuacct.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c index 4a811203c04a..e39bd4faf2a8 100644 --- a/kernel/sched/cpuacct.c +++ b/kernel/sched/cpuacct.c @@ -138,7 +138,7 @@ static u64 cpuusage_read(struct cgroup_subsys_state *css, struct cftype *cft) u64 totalcpuusage = 0; int i; - for_each_present_cpu(i) + for_each_possible_cpu(i) totalcpuusage += cpuacct_cpuusage_read(ca, i); return totalcpuusage; @@ -159,7 +159,7 @@ static int cpuusage_write(struct cgroup_subsys_state *css, struct cftype *cft, goto out; } - for_each_present_cpu(i) + for_each_possible_cpu(i) cpuacct_cpuusage_write(ca, i, 0); out: @@ -172,7 +172,7 @@ static int cpuacct_percpu_seq_show(struct seq_file *m, void *V) u64 percpu; int i; - for_each_present_cpu(i) { + for_each_possible_cpu(i) { percpu = cpuacct_cpuusage_read(ca, i); seq_printf(m, "%llu ", (unsigned long long) percpu); } @@ -191,7 +191,7 @@ static int cpuacct_stats_show(struct seq_file *sf, void *v) int cpu; s64 val = 0; - for_each_online_cpu(cpu) { + for_each_possible_cpu(cpu) { struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu); val += kcpustat->cpustat[CPUTIME_USER]; val += kcpustat->cpustat[CPUTIME_NICE]; @@ -200,7 +200,7 @@ static int cpuacct_stats_show(struct seq_file *sf, void *v) seq_printf(sf, "%s %lld\n", cpuacct_stat_desc[CPUACCT_STAT_USER], val); val = 0; - for_each_online_cpu(cpu) { + for_each_possible_cpu(cpu) { struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu); val += kcpustat->cpustat[CPUTIME_SYSTEM]; val += kcpustat->cpustat[CPUTIME_IRQ]; -- cgit v1.2.3 From d740037fac7052e49450f6fa1454f1144a103b55 Mon Sep 17 00:00:00 2001 From: Dongsheng Yang Date: Tue, 22 Mar 2016 16:37:08 +0800 Subject: sched/cpuacct: Split usage accounting into user_usage and sys_usage Sometimes, cpuacct.usage is not detailed enough to see how much CPU usage a group had. We want to know how much time it used in user mode and how much in kernel mode. This patch introduces more files to give this information: # ls /sys/fs/cgroup/cpuacct/cpuacct.usage* /sys/fs/cgroup/cpuacct/cpuacct.usage /sys/fs/cgroup/cpuacct/cpuacct.usage_percpu /sys/fs/cgroup/cpuacct/cpuacct.usage_user /sys/fs/cgroup/cpuacct/cpuacct.usage_percpu_user /sys/fs/cgroup/cpuacct/cpuacct.usage_sys /sys/fs/cgroup/cpuacct/cpuacct.usage_percpu_sys ... while keeping the ABI with the existing counter. Signed-off-by: Dongsheng Yang [ Ported to newer kernels. ] Signed-off-by: Zhao Lei Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Tejun Heo Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/aa171da036b520b51c79549e9b3215d29473f19d.1458635566.git.zhaolei@cn.fujitsu.com Signed-off-by: Ingo Molnar --- kernel/sched/cpuacct.c | 140 +++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 113 insertions(+), 27 deletions(-) diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c index e39bd4faf2a8..df947e07aac1 100644 --- a/kernel/sched/cpuacct.c +++ b/kernel/sched/cpuacct.c @@ -25,11 +25,22 @@ enum cpuacct_stat_index { CPUACCT_STAT_NSTATS, }; +enum cpuacct_usage_index { + CPUACCT_USAGE_USER, /* ... user mode */ + CPUACCT_USAGE_SYSTEM, /* ... kernel mode */ + + CPUACCT_USAGE_NRUSAGE, +}; + +struct cpuacct_usage { + u64 usages[CPUACCT_USAGE_NRUSAGE]; +}; + /* track cpu usage of a group of tasks and its child groups */ struct cpuacct { struct cgroup_subsys_state css; /* cpuusage holds pointer to a u64-type object on every cpu */ - u64 __percpu *cpuusage; + struct cpuacct_usage __percpu *cpuusage; struct kernel_cpustat __percpu *cpustat; }; @@ -49,7 +60,7 @@ static inline struct cpuacct *parent_ca(struct cpuacct *ca) return css_ca(ca->css.parent); } -static DEFINE_PER_CPU(u64, root_cpuacct_cpuusage); +static DEFINE_PER_CPU(struct cpuacct_usage, root_cpuacct_cpuusage); static struct cpuacct root_cpuacct = { .cpustat = &kernel_cpustat, .cpuusage = &root_cpuacct_cpuusage, @@ -68,7 +79,7 @@ cpuacct_css_alloc(struct cgroup_subsys_state *parent_css) if (!ca) goto out; - ca->cpuusage = alloc_percpu(u64); + ca->cpuusage = alloc_percpu(struct cpuacct_usage); if (!ca->cpuusage) goto out_free_ca; @@ -96,20 +107,37 @@ static void cpuacct_css_free(struct cgroup_subsys_state *css) kfree(ca); } -static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu) +static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu, + enum cpuacct_usage_index index) { - u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); + struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); u64 data; + /* + * We allow index == CPUACCT_USAGE_NRUSAGE here to read + * the sum of suages. + */ + BUG_ON(index > CPUACCT_USAGE_NRUSAGE); + #ifndef CONFIG_64BIT /* * Take rq->lock to make 64-bit read safe on 32-bit platforms. */ raw_spin_lock_irq(&cpu_rq(cpu)->lock); - data = *cpuusage; +#endif + + if (index == CPUACCT_USAGE_NRUSAGE) { + int i = 0; + + data = 0; + for (i = 0; i < CPUACCT_USAGE_NRUSAGE; i++) + data += cpuusage->usages[i]; + } else { + data = cpuusage->usages[index]; + } + +#ifndef CONFIG_64BIT raw_spin_unlock_irq(&cpu_rq(cpu)->lock); -#else - data = *cpuusage; #endif return data; @@ -117,69 +145,103 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu) static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val) { - u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); + struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); + int i; #ifndef CONFIG_64BIT /* * Take rq->lock to make 64-bit write safe on 32-bit platforms. */ raw_spin_lock_irq(&cpu_rq(cpu)->lock); - *cpuusage = val; +#endif + + for (i = 0; i < CPUACCT_USAGE_NRUSAGE; i++) + cpuusage->usages[i] = val; + +#ifndef CONFIG_64BIT raw_spin_unlock_irq(&cpu_rq(cpu)->lock); -#else - *cpuusage = val; #endif } /* return total cpu usage (in nanoseconds) of a group */ -static u64 cpuusage_read(struct cgroup_subsys_state *css, struct cftype *cft) +static u64 __cpuusage_read(struct cgroup_subsys_state *css, + enum cpuacct_usage_index index) { struct cpuacct *ca = css_ca(css); u64 totalcpuusage = 0; int i; for_each_possible_cpu(i) - totalcpuusage += cpuacct_cpuusage_read(ca, i); + totalcpuusage += cpuacct_cpuusage_read(ca, i, index); return totalcpuusage; } +static u64 cpuusage_user_read(struct cgroup_subsys_state *css, + struct cftype *cft) +{ + return __cpuusage_read(css, CPUACCT_USAGE_USER); +} + +static u64 cpuusage_sys_read(struct cgroup_subsys_state *css, + struct cftype *cft) +{ + return __cpuusage_read(css, CPUACCT_USAGE_SYSTEM); +} + +static u64 cpuusage_read(struct cgroup_subsys_state *css, struct cftype *cft) +{ + return __cpuusage_read(css, CPUACCT_USAGE_NRUSAGE); +} + static int cpuusage_write(struct cgroup_subsys_state *css, struct cftype *cft, u64 val) { struct cpuacct *ca = css_ca(css); - int err = 0; - int i; + int cpu; /* * Only allow '0' here to do a reset. */ - if (val) { - err = -EINVAL; - goto out; - } + if (val) + return -EINVAL; - for_each_possible_cpu(i) - cpuacct_cpuusage_write(ca, i, 0); + for_each_possible_cpu(cpu) + cpuacct_cpuusage_write(ca, cpu, 0); -out: - return err; + return 0; } -static int cpuacct_percpu_seq_show(struct seq_file *m, void *V) +static int __cpuacct_percpu_seq_show(struct seq_file *m, + enum cpuacct_usage_index index) { struct cpuacct *ca = css_ca(seq_css(m)); u64 percpu; int i; for_each_possible_cpu(i) { - percpu = cpuacct_cpuusage_read(ca, i); + percpu = cpuacct_cpuusage_read(ca, i, index); seq_printf(m, "%llu ", (unsigned long long) percpu); } seq_printf(m, "\n"); return 0; } +static int cpuacct_percpu_user_seq_show(struct seq_file *m, void *V) +{ + return __cpuacct_percpu_seq_show(m, CPUACCT_USAGE_USER); +} + +static int cpuacct_percpu_sys_seq_show(struct seq_file *m, void *V) +{ + return __cpuacct_percpu_seq_show(m, CPUACCT_USAGE_SYSTEM); +} + +static int cpuacct_percpu_seq_show(struct seq_file *m, void *V) +{ + return __cpuacct_percpu_seq_show(m, CPUACCT_USAGE_NRUSAGE); +} + static const char * const cpuacct_stat_desc[] = { [CPUACCT_STAT_USER] = "user", [CPUACCT_STAT_SYSTEM] = "system", @@ -219,10 +281,26 @@ static struct cftype files[] = { .read_u64 = cpuusage_read, .write_u64 = cpuusage_write, }, + { + .name = "usage_user", + .read_u64 = cpuusage_user_read, + }, + { + .name = "usage_sys", + .read_u64 = cpuusage_sys_read, + }, { .name = "usage_percpu", .seq_show = cpuacct_percpu_seq_show, }, + { + .name = "usage_percpu_user", + .seq_show = cpuacct_percpu_user_seq_show, + }, + { + .name = "usage_percpu_sys", + .seq_show = cpuacct_percpu_sys_seq_show, + }, { .name = "stat", .seq_show = cpuacct_stats_show, @@ -238,10 +316,18 @@ static struct cftype files[] = { void cpuacct_charge(struct task_struct *tsk, u64 cputime) { struct cpuacct *ca; + int index; + + if (user_mode(task_pt_regs(tsk))) + index = CPUACCT_USAGE_USER; + else + index = CPUACCT_USAGE_SYSTEM; rcu_read_lock(); + for (ca = task_ca(tsk); ca; ca = parent_ca(ca)) - *this_cpu_ptr(ca->cpuusage) += cputime; + this_cpu_ptr(ca->cpuusage)->usages[index] += cputime; + rcu_read_unlock(); } -- cgit v1.2.3 From d02c071183e1c01a76811c878c8a52322201f81f Mon Sep 17 00:00:00 2001 From: Srikar Dronamraju Date: Wed, 23 Mar 2016 17:54:44 +0530 Subject: sched/fair: Reset nr_balance_failed after active balancing To force a task migration during active balancing, nr_balance_failed is set to cache_nice_tries + 1. However nr_balance_failed is not reset. As a side effect, the next regular load balance under the same sd, a cache hot task might be migrated, just because nr_balance_failed count is high. Resetting nr_balance_failed after a successful active balance ensures that a hot task is not unreasonably migrated. This can be verified by looking at othe number of hot task migrations reported by /proc/schedstat. Signed-off-by: Srikar Dronamraju Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1458735884-30105-1-git-send-email-srikar@linux.vnet.ibm.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 0fe30e66aff1..cbb075e46b2c 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7399,10 +7399,7 @@ more_balance: &busiest->active_balance_work); } - /* - * We've kicked active balancing, reset the failure - * counter. - */ + /* We've kicked active balancing, force task migration. */ sd->nr_balance_failed = sd->cache_nice_tries+1; } } else @@ -7637,10 +7634,13 @@ static int active_load_balance_cpu_stop(void *data) schedstat_inc(sd, alb_count); p = detach_one_task(&env); - if (p) + if (p) { schedstat_inc(sd, alb_pushed); - else + /* Active balancing done, reset the failure counter. */ + sd->nr_balance_failed = 0; + } else { schedstat_inc(sd, alb_failed); + } } rcu_read_unlock(); out_unlock: -- cgit v1.2.3 From bfdb198ccd99472c5bded689699eb30dd06316bb Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 1 Feb 2016 14:47:59 -0800 Subject: sched/numa: Remove unnecessary NUMA dequeue update from non-SMP kernels In account_entity_enqueue(), we do not do account_numa_enqueue() as NUMA balancing is not needed for UP kernels. Hence, we should remove the account_numa_dequeue() call from account_entity_dequeue() for UP kernels. Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Linus Torvalds Cc: Andrew Morton Cc: Peter Zijlstra Cc: Rik van Riel Cc: Mel Gorman Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1454366879.21738.29.camel@schen9-desk2.jf.intel.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index cbb075e46b2c..1fe9916b4557 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2437,10 +2437,12 @@ account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) update_load_sub(&cfs_rq->load, se->load.weight); if (!parent_entity(se)) update_load_sub(&rq_of(cfs_rq)->load, se->load.weight); +#ifdef CONFIG_SMP if (entity_is_task(se)) { account_numa_dequeue(rq_of(cfs_rq), task_of(se)); list_del_init(&se->group_node); } +#endif cfs_rq->nr_running--; } -- cgit v1.2.3 From 47252cfbac03644ee4a3adfa50c77896aa94f2bb Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Mon, 21 Mar 2016 11:23:39 -0400 Subject: sched/core: Add preempt checks in preempt_schedule() code While testing the tracer preemptoff, I hit this strange trace: <...>-259 0...1 0us : schedule <-worker_thread <...>-259 0d..1 0us : rcu_note_context_switch <-__schedule <...>-259 0d..1 0us : rcu_sched_qs <-rcu_note_context_switch <...>-259 0d..1 0us : rcu_preempt_qs <-rcu_note_context_switch <...>-259 0d..1 0us : _raw_spin_lock <-__schedule <...>-259 0d..1 0us : preempt_count_add <-_raw_spin_lock <...>-259 0d..2 0us : do_raw_spin_lock <-_raw_spin_lock <...>-259 0d..2 1us : deactivate_task <-__schedule <...>-259 0d..2 1us : update_rq_clock.part.84 <-deactivate_task <...>-259 0d..2 1us : dequeue_task_fair <-deactivate_task <...>-259 0d..2 1us : dequeue_entity <-dequeue_task_fair <...>-259 0d..2 1us : update_curr <-dequeue_entity <...>-259 0d..2 1us : update_min_vruntime <-update_curr <...>-259 0d..2 1us : cpuacct_charge <-update_curr <...>-259 0d..2 1us : __rcu_read_lock <-cpuacct_charge <...>-259 0d..2 1us : __rcu_read_unlock <-cpuacct_charge <...>-259 0d..2 1us : clear_buddies <-dequeue_entity <...>-259 0d..2 1us : account_entity_dequeue <-dequeue_entity <...>-259 0d..2 2us : update_min_vruntime <-dequeue_entity <...>-259 0d..2 2us : update_cfs_shares <-dequeue_entity <...>-259 0d..2 2us : hrtick_update <-dequeue_task_fair <...>-259 0d..2 2us : wq_worker_sleeping <-__schedule <...>-259 0d..2 2us : kthread_data <-wq_worker_sleeping <...>-259 0d..2 2us : pick_next_task_fair <-__schedule <...>-259 0d..2 2us : check_cfs_rq_runtime <-pick_next_task_fair <...>-259 0d..2 2us : pick_next_entity <-pick_next_task_fair <...>-259 0d..2 2us : clear_buddies <-pick_next_entity <...>-259 0d..2 2us : pick_next_entity <-pick_next_task_fair <...>-259 0d..2 2us : clear_buddies <-pick_next_entity <...>-259 0d..2 2us : set_next_entity <-pick_next_task_fair <...>-259 0d..2 3us : put_prev_entity <-pick_next_task_fair <...>-259 0d..2 3us : check_cfs_rq_runtime <-put_prev_entity <...>-259 0d..2 3us : set_next_entity <-pick_next_task_fair gnome-sh-1031 0d..2 3us : finish_task_switch <-__schedule gnome-sh-1031 0d..2 3us : _raw_spin_unlock_irq <-finish_task_switch gnome-sh-1031 0d..2 3us : do_raw_spin_unlock <-_raw_spin_unlock_irq gnome-sh-1031 0...2 3us!: preempt_count_sub <-_raw_spin_unlock_irq gnome-sh-1031 0...1 582us : do_raw_spin_lock <-_raw_spin_lock gnome-sh-1031 0...1 583us : _raw_spin_unlock <-drm_gem_object_lookup gnome-sh-1031 0...1 583us : do_raw_spin_unlock <-_raw_spin_unlock gnome-sh-1031 0...1 583us : preempt_count_sub <-_raw_spin_unlock gnome-sh-1031 0...1 584us : _raw_spin_unlock <-drm_gem_object_lookup gnome-sh-1031 0...1 584us+: trace_preempt_on <-drm_gem_object_lookup gnome-sh-1031 0...1 603us : => preempt_count_sub => _raw_spin_unlock => drm_gem_object_lookup => i915_gem_madvise_ioctl => drm_ioctl => do_vfs_ioctl => SyS_ioctl => entry_SYSCALL_64_fastpath As I'm tracing preemption disabled, it seemed incorrect that the trace would go across a schedule and report not being in the scheduler. Looking into this I discovered the problem. schedule() calls preempt_disable() but the preempt_schedule() calls preempt_enable_notrace(). What happened above was that the gnome-shell task was preempted on another CPU, migrated over to the idle cpu. The tracer stared with idle calling schedule(), which called preempt_disable(), but then gnome-shell finished, and it enabled preemption with preempt_enable_notrace() that does stop the trace, even though preemption was enabled. The purpose of the preempt_disable_notrace() in the preempt_schedule() is to prevent function tracing from going into an infinite loop. Because function tracing can trace the preempt_enable/disable() calls that are traced. The problem with function tracing is: NEED_RESCHED set preempt_schedule() preempt_disable() preempt_count_inc() function trace (before incrementing preempt count) preempt_disable_notrace() preempt_enable_notrace() sees NEED_RESCHED set preempt_schedule() (repeat) Now by breaking out the preempt off/on tracing into their own code: preempt_disable_check() and preempt_enable_check(), we can add these to the preempt_schedule() code. As preemption would then be disabled, even if they were to be traced by the function tracer, the disabled preemption would prevent the recursion. Signed-off-by: Steven Rostedt Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20160321112339.6dc78ad6@gandalf.local.home Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 59 insertions(+), 9 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8b489fcac37b..b5cf01d2368e 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2958,6 +2958,20 @@ u64 scheduler_tick_max_deferment(void) #if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \ defined(CONFIG_PREEMPT_TRACER)) +/* + * If the value passed in is equal to the current preempt count + * then we just disabled preemption. Start timing the latency. + */ +static inline void preempt_latency_start(int val) +{ + if (preempt_count() == val) { + unsigned long ip = get_lock_parent_ip(); +#ifdef CONFIG_DEBUG_PREEMPT + current->preempt_disable_ip = ip; +#endif + trace_preempt_off(CALLER_ADDR0, ip); + } +} void preempt_count_add(int val) { @@ -2976,17 +2990,21 @@ void preempt_count_add(int val) DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >= PREEMPT_MASK - 10); #endif - if (preempt_count() == val) { - unsigned long ip = get_lock_parent_ip(); -#ifdef CONFIG_DEBUG_PREEMPT - current->preempt_disable_ip = ip; -#endif - trace_preempt_off(CALLER_ADDR0, ip); - } + preempt_latency_start(val); } EXPORT_SYMBOL(preempt_count_add); NOKPROBE_SYMBOL(preempt_count_add); +/* + * If the value passed in equals to the current preempt count + * then we just enabled preemption. Stop timing the latency. + */ +static inline void preempt_latency_stop(int val) +{ + if (preempt_count() == val) + trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip()); +} + void preempt_count_sub(int val) { #ifdef CONFIG_DEBUG_PREEMPT @@ -3003,13 +3021,15 @@ void preempt_count_sub(int val) return; #endif - if (preempt_count() == val) - trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip()); + preempt_latency_stop(val); __preempt_count_sub(val); } EXPORT_SYMBOL(preempt_count_sub); NOKPROBE_SYMBOL(preempt_count_sub); +#else +static inline void preempt_latency_start(int val) { } +static inline void preempt_latency_stop(int val) { } #endif /* @@ -3284,8 +3304,23 @@ void __sched schedule_preempt_disabled(void) static void __sched notrace preempt_schedule_common(void) { do { + /* + * Because the function tracer can trace preempt_count_sub() + * and it also uses preempt_enable/disable_notrace(), if + * NEED_RESCHED is set, the preempt_enable_notrace() called + * by the function tracer will call this function again and + * cause infinite recursion. + * + * Preemption must be disabled here before the function + * tracer can trace. Break up preempt_disable() into two + * calls. One to disable preemption without fear of being + * traced. The other to still record the preemption latency, + * which can also be traced by the function tracer. + */ preempt_disable_notrace(); + preempt_latency_start(1); __schedule(true); + preempt_latency_stop(1); preempt_enable_no_resched_notrace(); /* @@ -3337,7 +3372,21 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void) return; do { + /* + * Because the function tracer can trace preempt_count_sub() + * and it also uses preempt_enable/disable_notrace(), if + * NEED_RESCHED is set, the preempt_enable_notrace() called + * by the function tracer will call this function again and + * cause infinite recursion. + * + * Preemption must be disabled here before the function + * tracer can trace. Break up preempt_disable() into two + * calls. One to disable preemption without fear of being + * traced. The other to still record the preemption latency, + * which can also be traced by the function tracer. + */ preempt_disable_notrace(); + preempt_latency_start(1); /* * Needs preempt disabled in case user_exit() is traced * and the tracer calls preempt_enable_notrace() causing @@ -3347,6 +3396,7 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void) __schedule(true); exception_exit(prev_ctx); + preempt_latency_stop(1); preempt_enable_no_resched_notrace(); } while (need_resched()); } -- cgit v1.2.3 From 1c3de5e19fc96206dd086e634129d08e5f7b1000 Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Wed, 30 Mar 2016 07:07:51 +0800 Subject: sched/fair: Update comments after a variable rename The following commit: ed82b8a1ff76 ("sched/core: Move the sched_to_prio[] arrays out of line") renamed prio_to_weight to sched_prio_to_weight, but the old name was not updated in comments. Signed-off-by: Yuyang Du Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1459292871-22531-1-git-send-email-yuyang.du@intel.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 1fe9916b4557..4bb5ace60dc8 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -204,7 +204,7 @@ static void __update_inv_weight(struct load_weight *lw) * OR * (delta_exec * (weight * lw->inv_weight)) >> WMULT_SHIFT * - * Either weight := NICE_0_LOAD and lw \e prio_to_wmult[], in which case + * Either weight := NICE_0_LOAD and lw \e sched_prio_to_wmult[], in which case * we're guaranteed shift stays positive because inv_weight is guaranteed to * fit 32 bits, and NICE_0_LOAD gives another 10 bits; therefore shift >= 22. * @@ -5656,7 +5656,7 @@ static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preemp * W_i,0 = \Sum_j w_i,j (2) * * Where w_i,j is the weight of the j-th runnable task on cpu i. This weight - * is derived from the nice value as per prio_to_weight[]. + * is derived from the nice value as per sched_prio_to_weight[]. * * The weight average is an exponential decay average of the instantaneous * weight: -- cgit v1.2.3 From 2b8c41daba327c633228169e8bd8ec067ab443f8 Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Wed, 30 Mar 2016 04:30:56 +0800 Subject: sched/fair: Initiate a new task's util avg to a bounded value A new task's util_avg is set to full utilization of a CPU (100% time running). This accelerates a new task's utilization ramp-up, useful to boost its execution in early time. However, it may result in (insanely) high utilization for a transient time period when a flood of tasks are spawned. Importantly, it violates the "fundamentally bounded" CPU utilization, and its side effect is negative if we don't take any measure to bound it. This patch proposes an algorithm to address this issue. It has two methods to approach a sensible initial util_avg: (1) An expected (or average) util_avg based on its cfs_rq's util_avg: util_avg = cfs_rq->util_avg / (cfs_rq->load_avg + 1) * se.load.weight (2) A trajectory of how successive new tasks' util develops, which gives 1/2 of the left utilization budget to a new task such that the additional util is noticeably large (when overall util is low) or unnoticeably small (when overall util is high enough). In the meantime, the aggregate utilization is well bounded: util_avg_cap = (1024 - cfs_rq->avg.util_avg) / 2^n where n denotes the nth task. If util_avg is larger than util_avg_cap, then the effective util is clamped to the util_avg_cap. Reported-by: Andrey Ryabinin Signed-off-by: Yuyang Du Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: bsegall@google.com Cc: morten.rasmussen@arm.com Cc: pjt@google.com Cc: steve.muckle@linaro.org Link: http://lkml.kernel.org/r/1459283456-21682-1-git-send-email-yuyang.du@intel.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 2 ++ kernel/sched/fair.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++++++-- kernel/sched/sched.h | 1 + 3 files changed, 57 insertions(+), 2 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index b5cf01d2368e..11594230ef4d 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2431,6 +2431,8 @@ void wake_up_new_task(struct task_struct *p) */ set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0)); #endif + /* Post initialize new task's util average when its cfs_rq is set */ + post_init_entity_util_avg(&p->se); rq = __task_rq_lock(p); activate_task(rq, p, 0); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 4bb5ace60dc8..b8cc1c35cd7c 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -682,17 +682,68 @@ void init_entity_runnable_average(struct sched_entity *se) sa->period_contrib = 1023; sa->load_avg = scale_load_down(se->load.weight); sa->load_sum = sa->load_avg * LOAD_AVG_MAX; - sa->util_avg = scale_load_down(SCHED_LOAD_SCALE); - sa->util_sum = sa->util_avg * LOAD_AVG_MAX; + /* + * At this point, util_avg won't be used in select_task_rq_fair anyway + */ + sa->util_avg = 0; + sa->util_sum = 0; /* when this task enqueue'ed, it will contribute to its cfs_rq's load_avg */ } +/* + * With new tasks being created, their initial util_avgs are extrapolated + * based on the cfs_rq's current util_avg: + * + * util_avg = cfs_rq->util_avg / (cfs_rq->load_avg + 1) * se.load.weight + * + * However, in many cases, the above util_avg does not give a desired + * value. Moreover, the sum of the util_avgs may be divergent, such + * as when the series is a harmonic series. + * + * To solve this problem, we also cap the util_avg of successive tasks to + * only 1/2 of the left utilization budget: + * + * util_avg_cap = (1024 - cfs_rq->avg.util_avg) / 2^n + * + * where n denotes the nth task. + * + * For example, a simplest series from the beginning would be like: + * + * task util_avg: 512, 256, 128, 64, 32, 16, 8, ... + * cfs_rq util_avg: 512, 768, 896, 960, 992, 1008, 1016, ... + * + * Finally, that extrapolated util_avg is clamped to the cap (util_avg_cap) + * if util_avg > util_avg_cap. + */ +void post_init_entity_util_avg(struct sched_entity *se) +{ + struct cfs_rq *cfs_rq = cfs_rq_of(se); + struct sched_avg *sa = &se->avg; + long cap = (long)(scale_load_down(SCHED_LOAD_SCALE) - cfs_rq->avg.util_avg) / 2; + + if (cap > 0) { + if (cfs_rq->avg.util_avg != 0) { + sa->util_avg = cfs_rq->avg.util_avg * se->load.weight; + sa->util_avg /= (cfs_rq->avg.load_avg + 1); + + if (sa->util_avg > cap) + sa->util_avg = cap; + } else { + sa->util_avg = cap; + } + sa->util_sum = sa->util_avg * LOAD_AVG_MAX; + } +} + static inline unsigned long cfs_rq_runnable_load_avg(struct cfs_rq *cfs_rq); static inline unsigned long cfs_rq_load_avg(struct cfs_rq *cfs_rq); #else void init_entity_runnable_average(struct sched_entity *se) { } +void post_init_entity_util_avg(struct sched_entity *se) +{ +} #endif /* @@ -8384,6 +8435,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) init_cfs_rq(cfs_rq); init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]); init_entity_runnable_average(se); + post_init_entity_util_avg(se); } return 1; diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index ec2e8d23527e..a7cbad7b3ad2 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1313,6 +1313,7 @@ extern void init_dl_task_timer(struct sched_dl_entity *dl_se); unsigned long to_ratio(u64 period, u64 runtime); extern void init_entity_runnable_average(struct sched_entity *se); +extern void post_init_entity_util_avg(struct sched_entity *se); #ifdef CONFIG_NO_HZ_FULL extern bool sched_can_stop_tick(struct rq *rq); -- cgit v1.2.3 From fb90a6e93c0684ab2629a42462400603aa829b9c Mon Sep 17 00:00:00 2001 From: Rabin Vincent Date: Mon, 4 Apr 2016 15:42:02 +0200 Subject: sched/debug: Don't dump sched debug info in SysRq-W sysrq_sched_debug_show() can dump a lot of information. Don't print out all that if we're just trying to get a list of blocked tasks (SysRq-W). The information is still accessible with SysRq-T. Signed-off-by: Rabin Vincent Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Steven Rostedt Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1459777322-30902-1-git-send-email-rabin.vincent@axis.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 11594230ef4d..06efbb9c9544 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5050,7 +5050,8 @@ void show_state_filter(unsigned long state_filter) touch_all_softlockup_watchdogs(); #ifdef CONFIG_SCHED_DEBUG - sysrq_sched_debug_show(); + if (!state_filter) + sysrq_sched_debug_show(); #endif rcu_read_unlock(); /* -- cgit v1.2.3 From c78b17e28cc2c2df74264afc408bdc6aaf3fbcc8 Mon Sep 17 00:00:00 2001 From: Daniel Lezcano Date: Mon, 11 Apr 2016 16:38:33 +0200 Subject: sched/clock: Remove pointless test in cpu_clock/local_clock In case the HAVE_UNSTABLE_SCHED_CLOCK config is set, the cpu_clock() version checks if sched_clock_stable() is not set and calls sched_clock_cpu(), otherwise it calls sched_clock(). sched_clock_cpu() checks also if sched_clock_stable() is set and, if true, calls sched_clock(). sched_clock() will be called in sched_clock_cpu() if sched_clock_stable() is true. Remove the duplicate test by directly calling sched_clock_cpu() and let the static key act in this function instead. Signed-off-by: Daniel Lezcano Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1460385514-14700-1-git-send-email-daniel.lezcano@linaro.org Signed-off-by: Ingo Molnar --- kernel/sched/clock.c | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-) diff --git a/kernel/sched/clock.c b/kernel/sched/clock.c index fedb967a9841..30c4b202f0ba 100644 --- a/kernel/sched/clock.c +++ b/kernel/sched/clock.c @@ -375,10 +375,7 @@ EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); */ u64 cpu_clock(int cpu) { - if (!sched_clock_stable()) - return sched_clock_cpu(cpu); - - return sched_clock(); + return sched_clock_cpu(cpu); } /* @@ -390,10 +387,7 @@ u64 cpu_clock(int cpu) */ u64 local_clock(void) { - if (!sched_clock_stable()) - return sched_clock_cpu(raw_smp_processor_id()); - - return sched_clock(); + return sched_clock_cpu(raw_smp_processor_id()); } #else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ -- cgit v1.2.3 From 2c923e94cd9c6acff3b22f0ae29cfe65e2658b40 Mon Sep 17 00:00:00 2001 From: Daniel Lezcano Date: Mon, 11 Apr 2016 16:38:34 +0200 Subject: sched/clock: Make local_clock()/cpu_clock() inline The local_clock/cpu_clock functions were changed to prevent a double identical test with sched_clock_cpu() when HAVE_UNSTABLE_SCHED_CLOCK is set. That resulted in one line functions. As these functions are in all the cases one line functions and in the hot path, it is useful to specify them as static inline in order to give a strong hint to the compiler. After verification, it appears the compiler does not inline them without this hint. Change those functions to static inline. sched_clock_cpu() is called via the inlined local_clock()/cpu_clock() functions from sched.h. So any module code including sched.h will reference sched_clock_cpu(). Thus it must be exported with the EXPORT_SYMBOL_GPL macro. Signed-off-by: Daniel Lezcano Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1460385514-14700-2-git-send-email-daniel.lezcano@linaro.org Signed-off-by: Ingo Molnar --- include/linux/sched.h | 32 ++++++++++++++++++++++++++++++-- kernel/sched/clock.c | 42 +----------------------------------------- 2 files changed, 31 insertions(+), 43 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index 52c4847b05e2..13c1c1d07270 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -2303,8 +2303,6 @@ extern unsigned long long notrace sched_clock(void); /* * See the comment in kernel/sched/clock.c */ -extern u64 cpu_clock(int cpu); -extern u64 local_clock(void); extern u64 running_clock(void); extern u64 sched_clock_cpu(int cpu); @@ -2323,6 +2321,16 @@ static inline void sched_clock_idle_sleep_event(void) static inline void sched_clock_idle_wakeup_event(u64 delta_ns) { } + +static inline u64 cpu_clock(int cpu) +{ + return sched_clock(); +} + +static inline u64 local_clock(void) +{ + return sched_clock(); +} #else /* * Architectures can set this to 1 if they have specified @@ -2337,6 +2345,26 @@ extern void clear_sched_clock_stable(void); extern void sched_clock_tick(void); extern void sched_clock_idle_sleep_event(void); extern void sched_clock_idle_wakeup_event(u64 delta_ns); + +/* + * As outlined in clock.c, provides a fast, high resolution, nanosecond + * time source that is monotonic per cpu argument and has bounded drift + * between cpus. + * + * ######################### BIG FAT WARNING ########################## + * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can # + * # go backwards !! # + * #################################################################### + */ +static inline u64 cpu_clock(int cpu) +{ + return sched_clock_cpu(cpu); +} + +static inline u64 local_clock(void) +{ + return sched_clock_cpu(raw_smp_processor_id()); +} #endif #ifdef CONFIG_IRQ_TIME_ACCOUNTING diff --git a/kernel/sched/clock.c b/kernel/sched/clock.c index 30c4b202f0ba..e85a725e5c34 100644 --- a/kernel/sched/clock.c +++ b/kernel/sched/clock.c @@ -318,6 +318,7 @@ u64 sched_clock_cpu(int cpu) return clock; } +EXPORT_SYMBOL_GPL(sched_clock_cpu); void sched_clock_tick(void) { @@ -363,33 +364,6 @@ void sched_clock_idle_wakeup_event(u64 delta_ns) } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); -/* - * As outlined at the top, provides a fast, high resolution, nanosecond - * time source that is monotonic per cpu argument and has bounded drift - * between cpus. - * - * ######################### BIG FAT WARNING ########################## - * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can # - * # go backwards !! # - * #################################################################### - */ -u64 cpu_clock(int cpu) -{ - return sched_clock_cpu(cpu); -} - -/* - * Similar to cpu_clock() for the current cpu. Time will only be observed - * to be monotonic if care is taken to only compare timestampt taken on the - * same CPU. - * - * See cpu_clock(). - */ -u64 local_clock(void) -{ - return sched_clock_cpu(raw_smp_processor_id()); -} - #else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ void sched_clock_init(void) @@ -404,22 +378,8 @@ u64 sched_clock_cpu(int cpu) return sched_clock(); } - -u64 cpu_clock(int cpu) -{ - return sched_clock(); -} - -u64 local_clock(void) -{ - return sched_clock(); -} - #endif /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ -EXPORT_SYMBOL_GPL(cpu_clock); -EXPORT_SYMBOL_GPL(local_clock); - /* * Running clock - returns the time that has elapsed while a guest has been * running. -- cgit v1.2.3 From bd92883051a0228cc34996b8e766111ba10c9aac Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Wed, 6 Apr 2016 21:59:50 +1000 Subject: sched/cpuacct: Check for NULL when using task_pt_regs() task_pt_regs() can return NULL for kernel threads, so add a check. This fixes an oops at boot on ppc64. Reported-and-Tested-by: Srikar Dronamraju Tested-by: Zhao Lei Signed-off-by: Anton Blanchard Acked-by: Zhao Lei Cc: Linus Torvalds Cc: Michael Ellerman Cc: Peter Zijlstra Cc: Stephen Rothwell Cc: Thomas Gleixner Cc: efault@gmx.de Cc: htejun@gmail.com Cc: linuxppc-dev@lists.ozlabs.org Cc: tj@kernel.org Cc: yangds.fnst@cn.fujitsu.com Link: http://lkml.kernel.org/r/20160406215950.04bc3f0b@kryten Signed-off-by: Ingo Molnar --- kernel/sched/cpuacct.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c index df947e07aac1..41f85c4d0938 100644 --- a/kernel/sched/cpuacct.c +++ b/kernel/sched/cpuacct.c @@ -316,12 +316,11 @@ static struct cftype files[] = { void cpuacct_charge(struct task_struct *tsk, u64 cputime) { struct cpuacct *ca; - int index; + int index = CPUACCT_USAGE_SYSTEM; + struct pt_regs *regs = task_pt_regs(tsk); - if (user_mode(task_pt_regs(tsk))) + if (regs && user_mode(regs)) index = CPUACCT_USAGE_USER; - else - index = CPUACCT_USAGE_SYSTEM; rcu_read_lock(); -- cgit v1.2.3 From f454bfddf6ba557381d8bf5df50eff778602ff23 Mon Sep 17 00:00:00 2001 From: Alexander Shishkin Date: Thu, 14 Apr 2016 14:59:49 +0300 Subject: perf/core, sched: Don't use clock function pointer to determine clock Now that local_clock() is explicitly inlined in sched.h, taking its pointer would uninline it in the compilation unit where it's done, making (among other things) comparing pointers to this function produce different results in different compilation units. Case in point, x86 perf core's user page updating function compares event's clock against &local_clock to see if it needs to set zero time offset related bits in the page. This patch fixes the latter by looking at the "use_clockid" event attribute instead, to determine whether local clock is used. Fixing the uninlined local_clock() in perf core is left as an exercise for the author of the prior work. Signed-off-by: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Borislav Petkov Cc: Daniel Lezcano Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: eranian@google.com Cc: vince@deater.net Fixes: http://lkml.kernel.org/r/1459541050-13654-1-git-send-email-daniel.lezcano@linaro.org Link: http://lkml.kernel.org/r/1460635189-2320-1-git-send-email-alexander.shishkin@linux.intel.com Signed-off-by: Ingo Molnar --- arch/x86/events/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c index 041e442a3e28..dd39fde66b54 100644 --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -2177,7 +2177,7 @@ void arch_perf_update_userpage(struct perf_event *event, * cap_user_time_zero doesn't make sense when we're using a different * time base for the records. */ - if (event->clock == &local_clock) { + if (!event->attr.use_clockid) { userpg->cap_user_time_zero = 1; userpg->time_zero = data->cyc2ns_offset; } -- cgit v1.2.3 From 1f621e028baf391f6684003e32e009bc934b750f Mon Sep 17 00:00:00 2001 From: Srikar Dronamraju Date: Wed, 6 Apr 2016 18:47:40 +0530 Subject: sched/fair: Fix asym packing to select correct CPU When asymmetric packing is set in the sched_domain and target CPU is busy, update_sd_pick_busiest() may not select the busiest runqueue. When target CPU is busy, find_busiest_group() will ignore checks for asym packing and may continue to load balance using the currently selected not-the-busiest runqueue as source runqueue. Selecting the busiest runqueue as source when the target CPU is busy, should result in achieving much better load balance. Also when target CPU is not busy and asymmetric packing is set in sd, select higher CPU as source CPU for load balancing. While doing this change, move the check to see if target CPU is busy into check_asym_packing(). The extent of performance benefit from this change decreases with the increasing load. However there is benefit in undercommit as well as overcommit conditions. 1. Record per second ebizzy (32 threads) on a 64 CPU power 7 box. (5 iterations) 4.6.0-rc2 Testcase: Min Max Avg StdDev ebizzy: 5223767.00 10368236.00 7946971.00 1753094.76 4.6.0-rc2+asym-changes Testcase: Min Max Avg StdDev %Change ebizzy: 8617191.00 13872356.00 11383980.00 1783400.89 +24.78% 2. Record per second ebizzy (64 threads) on a 64 CPU power 7 box. (5 iterations) 4.6.0-rc2 Testcase: Min Max Avg StdDev ebizzy: 6497666.00 18399783.00 10818093.20 4051452.08 4.6.0-rc2+asym-changes Testcase: Min Max Avg StdDev %Change ebizzy: 7567365.00 19456937.00 11674063.60 4295407.48 +4.40% 3. Record per second ebizzy (128 threads) on a 64 CPU power 7 box. (5 iterations) 4.6.0-rc2 Testcase: Min Max Avg StdDev ebizzy: 37073983.00 40341911.00 38776241.80 1259766.82 4.6.0-rc2+asym-changes Testcase: Min Max Avg StdDev %Change ebizzy: 38030399.00 41333378.00 39827404.40 1255001.86 +2.54% Signed-off-by: Srikar Dronamraju Signed-off-by: Peter Zijlstra (Intel) Cc: Gautham R Shenoy Cc: Michael Neuling Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Vaidyanathan Srinivasan Link: http://lkml.kernel.org/r/1459948660-16073-1-git-send-email-srikar@linux.vnet.ibm.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index b8cc1c35cd7c..6e371f43fc80 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -6679,6 +6679,9 @@ static bool update_sd_pick_busiest(struct lb_env *env, if (!(env->sd->flags & SD_ASYM_PACKING)) return true; + /* No ASYM_PACKING if target cpu is already busy */ + if (env->idle == CPU_NOT_IDLE) + return true; /* * ASYM_PACKING needs to move all the work to the lowest * numbered CPUs in the group, therefore mark all groups @@ -6688,7 +6691,8 @@ static bool update_sd_pick_busiest(struct lb_env *env, if (!sds->busiest) return true; - if (group_first_cpu(sds->busiest) > group_first_cpu(sg)) + /* Prefer to move from highest possible cpu's work */ + if (group_first_cpu(sds->busiest) < group_first_cpu(sg)) return true; } @@ -6834,6 +6838,9 @@ static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds) if (!(env->sd->flags & SD_ASYM_PACKING)) return 0; + if (env->idle == CPU_NOT_IDLE) + return 0; + if (!sds->busiest) return 0; @@ -7026,8 +7033,7 @@ static struct sched_group *find_busiest_group(struct lb_env *env) busiest = &sds.busiest_stat; /* ASYM feature bypasses nice load balance check */ - if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) && - check_asym_packing(env, &sds)) + if (check_asym_packing(env, &sds)) return sds.busiest; /* There is no busy sibling group to pull tasks from */ -- cgit v1.2.3 From 21e96f88776deead303ecd30a17d1d7c2a1776e3 Mon Sep 17 00:00:00 2001 From: Steve Muckle Date: Mon, 21 Mar 2016 17:21:07 -0700 Subject: sched/fair: Move cpufreq hook to update_cfs_rq_load_avg() The cpufreq hook should be called whenever the root cfs_rq utilization changes so update_cfs_rq_load_avg() is a better place for it. The current location is not invoked in the enqueue_entity() or update_blocked_averages() paths. Suggested-by: Vincent Guittot Signed-off-by: Steve Muckle Signed-off-by: Peter Zijlstra (Intel) Cc: Dietmar Eggemann Cc: Juri Lelli Cc: Michael Turquette Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Patrick Bellasi Cc: Peter Zijlstra Cc: Rafael J. Wysocki Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1458606068-7476-1-git-send-email-smuckle@linaro.org Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 50 ++++++++++++++++++++++++++------------------------ 1 file changed, 26 insertions(+), 24 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 6e371f43fc80..6df80d47a525 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2878,7 +2878,9 @@ static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq); static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) { struct sched_avg *sa = &cfs_rq->avg; + struct rq *rq = rq_of(cfs_rq); int decayed, removed = 0; + int cpu = cpu_of(rq); if (atomic_long_read(&cfs_rq->removed_load_avg)) { s64 r = atomic_long_xchg(&cfs_rq->removed_load_avg, 0); @@ -2893,7 +2895,7 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) sa->util_sum = max_t(s32, sa->util_sum - r * LOAD_AVG_MAX, 0); } - decayed = __update_load_avg(now, cpu_of(rq_of(cfs_rq)), sa, + decayed = __update_load_avg(now, cpu, sa, scale_load_down(cfs_rq->load.weight), cfs_rq->curr != NULL, cfs_rq); #ifndef CONFIG_64BIT @@ -2901,28 +2903,6 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) cfs_rq->load_last_update_time_copy = sa->last_update_time; #endif - return decayed || removed; -} - -/* Update task and its cfs_rq load average */ -static inline void update_load_avg(struct sched_entity *se, int update_tg) -{ - struct cfs_rq *cfs_rq = cfs_rq_of(se); - u64 now = cfs_rq_clock_task(cfs_rq); - struct rq *rq = rq_of(cfs_rq); - int cpu = cpu_of(rq); - - /* - * Track task load average for carrying it to new CPU after migrated, and - * track group sched_entity load average for task_h_load calc in migration - */ - __update_load_avg(now, cpu, &se->avg, - se->on_rq * scale_load_down(se->load.weight), - cfs_rq->curr == se, NULL); - - if (update_cfs_rq_load_avg(now, cfs_rq) && update_tg) - update_tg_load_avg(cfs_rq, 0); - if (cpu == smp_processor_id() && &rq->cfs == cfs_rq) { unsigned long max = rq->cpu_capacity_orig; @@ -2943,8 +2923,30 @@ static inline void update_load_avg(struct sched_entity *se, int update_tg) * See cpu_util(). */ cpufreq_update_util(rq_clock(rq), - min(cfs_rq->avg.util_avg, max), max); + min(sa->util_avg, max), max); } + + return decayed || removed; +} + +/* Update task and its cfs_rq load average */ +static inline void update_load_avg(struct sched_entity *se, int update_tg) +{ + struct cfs_rq *cfs_rq = cfs_rq_of(se); + u64 now = cfs_rq_clock_task(cfs_rq); + struct rq *rq = rq_of(cfs_rq); + int cpu = cpu_of(rq); + + /* + * Track task load average for carrying it to new CPU after migrated, and + * track group sched_entity load average for task_h_load calc in migration + */ + __update_load_avg(now, cpu, &se->avg, + se->on_rq * scale_load_down(se->load.weight), + cfs_rq->curr == se, NULL); + + if (update_cfs_rq_load_avg(now, cfs_rq) && update_tg) + update_tg_load_avg(cfs_rq, 0); } static void attach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) -- cgit v1.2.3 From 41e0d37f7ac81297c07ba311e4ad39465b8c8295 Mon Sep 17 00:00:00 2001 From: Steve Muckle Date: Mon, 21 Mar 2016 17:21:08 -0700 Subject: sched/fair: Do not call cpufreq hook unless util changed There's no reason to call the cpufreq hook if the root cfs_rq utilization has not been modified. Signed-off-by: Steve Muckle Signed-off-by: Peter Zijlstra (Intel) Cc: Dietmar Eggemann Cc: Juri Lelli Cc: Michael Turquette Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Patrick Bellasi Cc: Peter Zijlstra Cc: Rafael J. Wysocki Cc: Thomas Gleixner Cc: Vincent Guittot Link: http://lkml.kernel.org/r/1458606068-7476-2-git-send-email-smuckle@linaro.org Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 6df80d47a525..81552819444c 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2879,20 +2879,21 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) { struct sched_avg *sa = &cfs_rq->avg; struct rq *rq = rq_of(cfs_rq); - int decayed, removed = 0; + int decayed, removed_load = 0, removed_util = 0; int cpu = cpu_of(rq); if (atomic_long_read(&cfs_rq->removed_load_avg)) { s64 r = atomic_long_xchg(&cfs_rq->removed_load_avg, 0); sa->load_avg = max_t(long, sa->load_avg - r, 0); sa->load_sum = max_t(s64, sa->load_sum - r * LOAD_AVG_MAX, 0); - removed = 1; + removed_load = 1; } if (atomic_long_read(&cfs_rq->removed_util_avg)) { long r = atomic_long_xchg(&cfs_rq->removed_util_avg, 0); sa->util_avg = max_t(long, sa->util_avg - r, 0); sa->util_sum = max_t(s32, sa->util_sum - r * LOAD_AVG_MAX, 0); + removed_util = 1; } decayed = __update_load_avg(now, cpu, sa, @@ -2903,7 +2904,8 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) cfs_rq->load_last_update_time_copy = sa->last_update_time; #endif - if (cpu == smp_processor_id() && &rq->cfs == cfs_rq) { + if (cpu == smp_processor_id() && &rq->cfs == cfs_rq && + (decayed || removed_util)) { unsigned long max = rq->cpu_capacity_orig; /* @@ -2926,7 +2928,7 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) min(sa->util_avg, max), max); } - return decayed || removed; + return decayed || removed_load; } /* Update task and its cfs_rq load average */ -- cgit v1.2.3 From a2c6c91f98247fef0fe75216d607812485aeb0df Mon Sep 17 00:00:00 2001 From: Steve Muckle Date: Thu, 24 Mar 2016 15:26:07 -0700 Subject: sched/fair: Call cpufreq hook in additional paths The cpufreq hook should be called any time the root CFS rq utilization changes. This can occur when a task is switched to or from the fair class, or a task moves between groups or CPUs, but these paths currently do not call the cpufreq hook. Fix this by adding the hook to attach_entity_load_avg() and detach_entity_load_avg(). Suggested-by: Vincent Guittot Signed-off-by: Steve Muckle [ Added the .update_freq argument to update_cfs_rq_load_avg() to avoid a double cpufreq call. ] Signed-off-by: Peter Zijlstra (Intel) Cc: Byungchul Park Cc: Dietmar Eggemann Cc: Juri Lelli Cc: Michael Turquette Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Patrick Bellasi Cc: Peter Zijlstra Cc: Rafael J. Wysocki Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1458858367-2831-1-git-send-email-smuckle@linaro.org Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 73 ++++++++++++++++++++++++++++++----------------------- 1 file changed, 42 insertions(+), 31 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 81552819444c..c328bd77fe35 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2874,13 +2874,41 @@ static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force) {} static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq); +static inline void cfs_rq_util_change(struct cfs_rq *cfs_rq) +{ + struct rq *rq = rq_of(cfs_rq); + int cpu = cpu_of(rq); + + if (cpu == smp_processor_id() && &rq->cfs == cfs_rq) { + unsigned long max = rq->cpu_capacity_orig; + + /* + * There are a few boundary cases this might miss but it should + * get called often enough that that should (hopefully) not be + * a real problem -- added to that it only calls on the local + * CPU, so if we enqueue remotely we'll miss an update, but + * the next tick/schedule should update. + * + * It will not get called when we go idle, because the idle + * thread is a different class (!fair), nor will the utilization + * number include things like RT tasks. + * + * As is, the util number is not freq-invariant (we'd have to + * implement arch_scale_freq_capacity() for that). + * + * See cpu_util(). + */ + cpufreq_update_util(rq_clock(rq), + min(cfs_rq->avg.util_avg, max), max); + } +} + /* Group cfs_rq's load_avg is used for task_h_load and update_cfs_share */ -static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) +static inline int +update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq, bool update_freq) { struct sched_avg *sa = &cfs_rq->avg; - struct rq *rq = rq_of(cfs_rq); int decayed, removed_load = 0, removed_util = 0; - int cpu = cpu_of(rq); if (atomic_long_read(&cfs_rq->removed_load_avg)) { s64 r = atomic_long_xchg(&cfs_rq->removed_load_avg, 0); @@ -2896,7 +2924,7 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) removed_util = 1; } - decayed = __update_load_avg(now, cpu, sa, + decayed = __update_load_avg(now, cpu_of(rq_of(cfs_rq)), sa, scale_load_down(cfs_rq->load.weight), cfs_rq->curr != NULL, cfs_rq); #ifndef CONFIG_64BIT @@ -2904,29 +2932,8 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) cfs_rq->load_last_update_time_copy = sa->last_update_time; #endif - if (cpu == smp_processor_id() && &rq->cfs == cfs_rq && - (decayed || removed_util)) { - unsigned long max = rq->cpu_capacity_orig; - - /* - * There are a few boundary cases this might miss but it should - * get called often enough that that should (hopefully) not be - * a real problem -- added to that it only calls on the local - * CPU, so if we enqueue remotely we'll miss an update, but - * the next tick/schedule should update. - * - * It will not get called when we go idle, because the idle - * thread is a different class (!fair), nor will the utilization - * number include things like RT tasks. - * - * As is, the util number is not freq-invariant (we'd have to - * implement arch_scale_freq_capacity() for that). - * - * See cpu_util(). - */ - cpufreq_update_util(rq_clock(rq), - min(sa->util_avg, max), max); - } + if (update_freq && (decayed || removed_util)) + cfs_rq_util_change(cfs_rq); return decayed || removed_load; } @@ -2947,7 +2954,7 @@ static inline void update_load_avg(struct sched_entity *se, int update_tg) se->on_rq * scale_load_down(se->load.weight), cfs_rq->curr == se, NULL); - if (update_cfs_rq_load_avg(now, cfs_rq) && update_tg) + if (update_cfs_rq_load_avg(now, cfs_rq, true) && update_tg) update_tg_load_avg(cfs_rq, 0); } @@ -2976,6 +2983,8 @@ skip_aging: cfs_rq->avg.load_sum += se->avg.load_sum; cfs_rq->avg.util_avg += se->avg.util_avg; cfs_rq->avg.util_sum += se->avg.util_sum; + + cfs_rq_util_change(cfs_rq); } static void detach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) @@ -2988,6 +2997,8 @@ static void detach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *s cfs_rq->avg.load_sum = max_t(s64, cfs_rq->avg.load_sum - se->avg.load_sum, 0); cfs_rq->avg.util_avg = max_t(long, cfs_rq->avg.util_avg - se->avg.util_avg, 0); cfs_rq->avg.util_sum = max_t(s32, cfs_rq->avg.util_sum - se->avg.util_sum, 0); + + cfs_rq_util_change(cfs_rq); } /* Add the load generated by se into cfs_rq's load average */ @@ -3005,7 +3016,7 @@ enqueue_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) cfs_rq->curr == se, NULL); } - decayed = update_cfs_rq_load_avg(now, cfs_rq); + decayed = update_cfs_rq_load_avg(now, cfs_rq, !migrated); cfs_rq->runnable_load_avg += sa->load_avg; cfs_rq->runnable_load_sum += sa->load_sum; @@ -6213,7 +6224,7 @@ static void update_blocked_averages(int cpu) if (throttled_hierarchy(cfs_rq)) continue; - if (update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq)) + if (update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, true)) update_tg_load_avg(cfs_rq, 0); } raw_spin_unlock_irqrestore(&rq->lock, flags); @@ -6274,7 +6285,7 @@ static inline void update_blocked_averages(int cpu) raw_spin_lock_irqsave(&rq->lock, flags); update_rq_clock(rq); - update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq); + update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, true); raw_spin_unlock_irqrestore(&rq->lock, flags); } -- cgit v1.2.3 From cee1afce3053e7aa0793fbd5f2e845fa2cef9e33 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 13 Apr 2016 15:56:50 +0200 Subject: sched/fair: Gather CPU load functions under a more conventional namespace The CPU load update related functions have a weak naming convention currently, starting with update_cpu_load_*() which isn't ideal as "update" is a very generic concept. Since two of these functions are public already (and a third is to come) that's enough to introduce a more conventional naming scheme. So let's do the following rename instead: update_cpu_load_*() -> cpu_load_update_*() Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Byungchul Park Cc: Chris Metcalf Cc: Christoph Lameter Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1460555812-25375-2-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- Documentation/trace/ftrace.txt | 10 +++++----- include/linux/sched.h | 4 ++-- kernel/sched/core.c | 2 +- kernel/sched/fair.c | 24 ++++++++++++------------ kernel/sched/sched.h | 4 ++-- kernel/time/tick-sched.c | 2 +- 6 files changed, 23 insertions(+), 23 deletions(-) diff --git a/Documentation/trace/ftrace.txt b/Documentation/trace/ftrace.txt index f52f297cb406..9857606dd7b7 100644 --- a/Documentation/trace/ftrace.txt +++ b/Documentation/trace/ftrace.txt @@ -1562,12 +1562,12 @@ Doing the same with chrt -r 5 and function-trace set. -0 3dN.1 12us : menu_hrtimer_cancel <-tick_nohz_idle_exit -0 3dN.1 12us : ktime_get <-tick_nohz_idle_exit -0 3dN.1 12us : tick_do_update_jiffies64 <-tick_nohz_idle_exit - -0 3dN.1 13us : update_cpu_load_nohz <-tick_nohz_idle_exit - -0 3dN.1 13us : _raw_spin_lock <-update_cpu_load_nohz + -0 3dN.1 13us : cpu_load_update_nohz <-tick_nohz_idle_exit + -0 3dN.1 13us : _raw_spin_lock <-cpu_load_update_nohz -0 3dN.1 13us : add_preempt_count <-_raw_spin_lock - -0 3dN.2 13us : __update_cpu_load <-update_cpu_load_nohz - -0 3dN.2 14us : sched_avg_update <-__update_cpu_load - -0 3dN.2 14us : _raw_spin_unlock <-update_cpu_load_nohz + -0 3dN.2 13us : __cpu_load_update <-cpu_load_update_nohz + -0 3dN.2 14us : sched_avg_update <-__cpu_load_update + -0 3dN.2 14us : _raw_spin_unlock <-cpu_load_update_nohz -0 3dN.2 14us : sub_preempt_count <-_raw_spin_unlock -0 3dN.1 15us : calc_load_exit_idle <-tick_nohz_idle_exit -0 3dN.1 15us : touch_softlockup_watchdog <-tick_nohz_idle_exit diff --git a/include/linux/sched.h b/include/linux/sched.h index 13c1c1d07270..0b7f6028a50b 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -178,9 +178,9 @@ extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load); extern void calc_global_load(unsigned long ticks); #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON) -extern void update_cpu_load_nohz(int active); +extern void cpu_load_update_nohz(int active); #else -static inline void update_cpu_load_nohz(int active) { } +static inline void cpu_load_update_nohz(int active) { } #endif extern void dump_cpu_task(int cpu); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 06efbb9c9544..c98a2688f390 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2917,7 +2917,7 @@ void scheduler_tick(void) raw_spin_lock(&rq->lock); update_rq_clock(rq); curr->sched_class->task_tick(rq, curr, 0); - update_cpu_load_active(rq); + cpu_load_update_active(rq); calc_global_load_tick(rq); raw_spin_unlock(&rq->lock); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index c328bd77fe35..ecd81c4ebb56 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4559,7 +4559,7 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) } /** - * __update_cpu_load - update the rq->cpu_load[] statistics + * __cpu_load_update - update the rq->cpu_load[] statistics * @this_rq: The rq to update statistics for * @this_load: The current load * @pending_updates: The number of missed updates @@ -4594,7 +4594,7 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) * see decay_load_misses(). For NOHZ_FULL we get to subtract and add the extra * term. See the @active paramter. */ -static void __update_cpu_load(struct rq *this_rq, unsigned long this_load, +static void __cpu_load_update(struct rq *this_rq, unsigned long this_load, unsigned long pending_updates, int active) { unsigned long tickless_load = active ? this_rq->cpu_load[0] : 0; @@ -4642,7 +4642,7 @@ static unsigned long weighted_cpuload(const int cpu) } #ifdef CONFIG_NO_HZ_COMMON -static void __update_cpu_load_nohz(struct rq *this_rq, +static void __cpu_load_update_nohz(struct rq *this_rq, unsigned long curr_jiffies, unsigned long load, int active) @@ -4657,7 +4657,7 @@ static void __update_cpu_load_nohz(struct rq *this_rq, * In the NOHZ_FULL case, we were non-idle, we should consider * its weighted load. */ - __update_cpu_load(this_rq, load, pending_updates, active); + __cpu_load_update(this_rq, load, pending_updates, active); } } @@ -4678,7 +4678,7 @@ static void __update_cpu_load_nohz(struct rq *this_rq, * Called from nohz_idle_balance() to update the load ratings before doing the * idle balance. */ -static void update_cpu_load_idle(struct rq *this_rq) +static void cpu_load_update_idle(struct rq *this_rq) { /* * bail if there's load or we're actually up-to-date. @@ -4686,13 +4686,13 @@ static void update_cpu_load_idle(struct rq *this_rq) if (weighted_cpuload(cpu_of(this_rq))) return; - __update_cpu_load_nohz(this_rq, READ_ONCE(jiffies), 0, 0); + __cpu_load_update_nohz(this_rq, READ_ONCE(jiffies), 0, 0); } /* * Called from tick_nohz_idle_exit() -- try and fix up the ticks we missed. */ -void update_cpu_load_nohz(int active) +void cpu_load_update_nohz(int active) { struct rq *this_rq = this_rq(); unsigned long curr_jiffies = READ_ONCE(jiffies); @@ -4702,7 +4702,7 @@ void update_cpu_load_nohz(int active) return; raw_spin_lock(&this_rq->lock); - __update_cpu_load_nohz(this_rq, curr_jiffies, load, active); + __cpu_load_update_nohz(this_rq, curr_jiffies, load, active); raw_spin_unlock(&this_rq->lock); } #endif /* CONFIG_NO_HZ */ @@ -4710,14 +4710,14 @@ void update_cpu_load_nohz(int active) /* * Called from scheduler_tick() */ -void update_cpu_load_active(struct rq *this_rq) +void cpu_load_update_active(struct rq *this_rq) { unsigned long load = weighted_cpuload(cpu_of(this_rq)); /* - * See the mess around update_cpu_load_idle() / update_cpu_load_nohz(). + * See the mess around cpu_load_update_idle() / cpu_load_update_nohz(). */ this_rq->last_load_update_tick = jiffies; - __update_cpu_load(this_rq, load, 1, 1); + __cpu_load_update(this_rq, load, 1, 1); } /* @@ -8031,7 +8031,7 @@ static void nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle) if (time_after_eq(jiffies, rq->next_balance)) { raw_spin_lock_irq(&rq->lock); update_rq_clock(rq); - update_cpu_load_idle(rq); + cpu_load_update_idle(rq); raw_spin_unlock_irq(&rq->lock); rebalance_domains(rq, CPU_IDLE); } diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index a7cbad7b3ad2..32d9e22cfacf 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -31,9 +31,9 @@ extern void calc_global_load_tick(struct rq *this_rq); extern long calc_load_fold_active(struct rq *this_rq); #ifdef CONFIG_SMP -extern void update_cpu_load_active(struct rq *this_rq); +extern void cpu_load_update_active(struct rq *this_rq); #else -static inline void update_cpu_load_active(struct rq *this_rq) { } +static inline void cpu_load_update_active(struct rq *this_rq) { } #endif /* diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 58e3310c9b21..66bdc9acc283 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -806,7 +806,7 @@ static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now, int { /* Update jiffies first */ tick_do_update_jiffies64(now); - update_cpu_load_nohz(active); + cpu_load_update_nohz(active); calc_load_exit_idle(); touch_softlockup_watchdog_sched(); -- cgit v1.2.3 From 1f41906a6fda1114debd3898668bd7ab6470ee41 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 13 Apr 2016 15:56:51 +0200 Subject: sched/fair: Correctly handle nohz ticks CPU load accounting Ticks can happen while the CPU is in dynticks-idle or dynticks-singletask mode. In fact "nohz" or "dynticks" only mean that we exit the periodic mode and we try to minimize the ticks as much as possible. The nohz subsystem uses a confusing terminology with the internal state "ts->tick_stopped" which is also available through its public interface with tick_nohz_tick_stopped(). This is a misnomer as the tick is instead reduced with the best effort rather than stopped. In the best case the tick can indeed be actually stopped but there is no guarantee about that. If a timer needs to fire one second later, a tick will fire while the CPU is in nohz mode and this is a very common scenario. Now this confusion happens to be a problem with CPU load updates: cpu_load_update_active() doesn't handle nohz ticks correctly because it assumes that ticks are completely stopped in nohz mode and that cpu_load_update_active() can't be called in dynticks mode. When that happens, the whole previous tickless load is ignored and the function just records the load for the current tick, ignoring potentially long idle periods behind. In order to solve this, we could account the current load for the previous nohz time but there is a risk that we account the load of a task that got freshly enqueued for the whole nohz period. So instead, lets record the dynticks load on nohz frame entry so we know what to record in case of nohz ticks, then use this record to account the tickless load on nohz ticks and nohz frame end. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Byungchul Park Cc: Chris Metcalf Cc: Christoph Lameter Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1460555812-25375-3-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- include/linux/sched.h | 6 ++- kernel/sched/fair.c | 97 +++++++++++++++++++++++++++++++----------------- kernel/time/tick-sched.c | 9 +++-- 3 files changed, 72 insertions(+), 40 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index 0b7f6028a50b..d894f2d61388 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -178,9 +178,11 @@ extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load); extern void calc_global_load(unsigned long ticks); #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON) -extern void cpu_load_update_nohz(int active); +extern void cpu_load_update_nohz_start(void); +extern void cpu_load_update_nohz_stop(void); #else -static inline void cpu_load_update_nohz(int active) { } +static inline void cpu_load_update_nohz_start(void) { } +static inline void cpu_load_update_nohz_stop(void) { } #endif extern void dump_cpu_task(int cpu); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index ecd81c4ebb56..b70367a3e1ef 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4563,7 +4563,6 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) * @this_rq: The rq to update statistics for * @this_load: The current load * @pending_updates: The number of missed updates - * @active: !0 for NOHZ_FULL * * Update rq->cpu_load[] statistics. This function is usually called every * scheduler tick (TICK_NSEC). @@ -4592,12 +4591,12 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) * load[i]_n = (1 - 1/2^i)^n * load[i]_0 * * see decay_load_misses(). For NOHZ_FULL we get to subtract and add the extra - * term. See the @active paramter. + * term. */ -static void __cpu_load_update(struct rq *this_rq, unsigned long this_load, - unsigned long pending_updates, int active) +static void cpu_load_update(struct rq *this_rq, unsigned long this_load, + unsigned long pending_updates) { - unsigned long tickless_load = active ? this_rq->cpu_load[0] : 0; + unsigned long tickless_load = this_rq->cpu_load[0]; int i, scale; this_rq->nr_load_updates++; @@ -4642,10 +4641,23 @@ static unsigned long weighted_cpuload(const int cpu) } #ifdef CONFIG_NO_HZ_COMMON -static void __cpu_load_update_nohz(struct rq *this_rq, - unsigned long curr_jiffies, - unsigned long load, - int active) +/* + * There is no sane way to deal with nohz on smp when using jiffies because the + * cpu doing the jiffies update might drift wrt the cpu doing the jiffy reading + * causing off-by-one errors in observed deltas; {0,2} instead of {1,1}. + * + * Therefore we need to avoid the delta approach from the regular tick when + * possible since that would seriously skew the load calculation. This is why we + * use cpu_load_update_periodic() for CPUs out of nohz. However we'll rely on + * jiffies deltas for updates happening while in nohz mode (idle ticks, idle + * loop exit, nohz_idle_balance, nohz full exit...) + * + * This means we might still be one tick off for nohz periods. + */ + +static void cpu_load_update_nohz(struct rq *this_rq, + unsigned long curr_jiffies, + unsigned long load) { unsigned long pending_updates; @@ -4657,23 +4669,10 @@ static void __cpu_load_update_nohz(struct rq *this_rq, * In the NOHZ_FULL case, we were non-idle, we should consider * its weighted load. */ - __cpu_load_update(this_rq, load, pending_updates, active); + cpu_load_update(this_rq, load, pending_updates); } } -/* - * There is no sane way to deal with nohz on smp when using jiffies because the - * cpu doing the jiffies update might drift wrt the cpu doing the jiffy reading - * causing off-by-one errors in observed deltas; {0,2} instead of {1,1}. - * - * Therefore we cannot use the delta approach from the regular tick since that - * would seriously skew the load calculation. However we'll make do for those - * updates happening while idle (nohz_idle_balance) or coming out of idle - * (tick_nohz_idle_exit). - * - * This means we might still be one tick off for nohz periods. - */ - /* * Called from nohz_idle_balance() to update the load ratings before doing the * idle balance. @@ -4686,26 +4685,56 @@ static void cpu_load_update_idle(struct rq *this_rq) if (weighted_cpuload(cpu_of(this_rq))) return; - __cpu_load_update_nohz(this_rq, READ_ONCE(jiffies), 0, 0); + cpu_load_update_nohz(this_rq, READ_ONCE(jiffies), 0); } /* - * Called from tick_nohz_idle_exit() -- try and fix up the ticks we missed. + * Record CPU load on nohz entry so we know the tickless load to account + * on nohz exit. cpu_load[0] happens then to be updated more frequently + * than other cpu_load[idx] but it should be fine as cpu_load readers + * shouldn't rely into synchronized cpu_load[*] updates. */ -void cpu_load_update_nohz(int active) +void cpu_load_update_nohz_start(void) { struct rq *this_rq = this_rq(); + + /* + * This is all lockless but should be fine. If weighted_cpuload changes + * concurrently we'll exit nohz. And cpu_load write can race with + * cpu_load_update_idle() but both updater would be writing the same. + */ + this_rq->cpu_load[0] = weighted_cpuload(cpu_of(this_rq)); +} + +/* + * Account the tickless load in the end of a nohz frame. + */ +void cpu_load_update_nohz_stop(void) +{ unsigned long curr_jiffies = READ_ONCE(jiffies); - unsigned long load = active ? weighted_cpuload(cpu_of(this_rq)) : 0; + struct rq *this_rq = this_rq(); + unsigned long load; if (curr_jiffies == this_rq->last_load_update_tick) return; + load = weighted_cpuload(cpu_of(this_rq)); raw_spin_lock(&this_rq->lock); - __cpu_load_update_nohz(this_rq, curr_jiffies, load, active); + cpu_load_update_nohz(this_rq, curr_jiffies, load); raw_spin_unlock(&this_rq->lock); } -#endif /* CONFIG_NO_HZ */ +#else /* !CONFIG_NO_HZ_COMMON */ +static inline void cpu_load_update_nohz(struct rq *this_rq, + unsigned long curr_jiffies, + unsigned long load) { } +#endif /* CONFIG_NO_HZ_COMMON */ + +static void cpu_load_update_periodic(struct rq *this_rq, unsigned long load) +{ + /* See the mess around cpu_load_update_nohz(). */ + this_rq->last_load_update_tick = READ_ONCE(jiffies); + cpu_load_update(this_rq, load, 1); +} /* * Called from scheduler_tick() @@ -4713,11 +4742,11 @@ void cpu_load_update_nohz(int active) void cpu_load_update_active(struct rq *this_rq) { unsigned long load = weighted_cpuload(cpu_of(this_rq)); - /* - * See the mess around cpu_load_update_idle() / cpu_load_update_nohz(). - */ - this_rq->last_load_update_tick = jiffies; - __cpu_load_update(this_rq, load, 1, 1); + + if (tick_nohz_tick_stopped()) + cpu_load_update_nohz(this_rq, READ_ONCE(jiffies), load); + else + cpu_load_update_periodic(this_rq, load); } /* diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 66bdc9acc283..31872bc53bc4 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -776,6 +776,7 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts, if (!ts->tick_stopped) { nohz_balance_enter_idle(cpu); calc_load_enter_idle(); + cpu_load_update_nohz_start(); ts->last_tick = hrtimer_get_expires(&ts->sched_timer); ts->tick_stopped = 1; @@ -802,11 +803,11 @@ out: return tick; } -static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now, int active) +static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now) { /* Update jiffies first */ tick_do_update_jiffies64(now); - cpu_load_update_nohz(active); + cpu_load_update_nohz_stop(); calc_load_exit_idle(); touch_softlockup_watchdog_sched(); @@ -833,7 +834,7 @@ static void tick_nohz_full_update_tick(struct tick_sched *ts) if (can_stop_full_tick(ts)) tick_nohz_stop_sched_tick(ts, ktime_get(), cpu); else if (ts->tick_stopped) - tick_nohz_restart_sched_tick(ts, ktime_get(), 1); + tick_nohz_restart_sched_tick(ts, ktime_get()); #endif } @@ -1024,7 +1025,7 @@ void tick_nohz_idle_exit(void) tick_nohz_stop_idle(ts, now); if (ts->tick_stopped) { - tick_nohz_restart_sched_tick(ts, now, 0); + tick_nohz_restart_sched_tick(ts, now); tick_nohz_account_idle_ticks(ts); } -- cgit v1.2.3 From 9fd81dd5ce0b12341c9f83346f8d32ac68bd3841 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 19 Apr 2016 17:36:51 +0200 Subject: sched/fair: Optimize !CONFIG_NO_HZ_COMMON CPU load updates Some code in CPU load update only concern NO_HZ configs but it is built on all configurations. When NO_HZ isn't built, that code is harmless but just happens to take some useless ressources in CPU and memory: 1) one useless field in struct rq 2) jiffies record on every tick that is never used (cpu_load_update_periodic) 3) decay_load_missed is called two times on every tick to eventually return immediately with no action taken. And that function is dead code. For pure optimization purposes, lets conditionally build the NO_HZ related code. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Byungchul Park Cc: Chris Metcalf Cc: Christoph Lameter Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1461080211-16271-1-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 5 ++--- kernel/sched/fair.c | 9 +++++++-- kernel/sched/sched.h | 6 ++++-- 3 files changed, 13 insertions(+), 7 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index c98a2688f390..71dffbb27ce6 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -7381,8 +7381,6 @@ void __init sched_init(void) for (j = 0; j < CPU_LOAD_IDX_MAX; j++) rq->cpu_load[j] = 0; - rq->last_load_update_tick = jiffies; - #ifdef CONFIG_SMP rq->sd = NULL; rq->rd = NULL; @@ -7401,12 +7399,13 @@ void __init sched_init(void) rq_attach_root(rq, &def_root_domain); #ifdef CONFIG_NO_HZ_COMMON + rq->last_load_update_tick = jiffies; rq->nohz_flags = 0; #endif #ifdef CONFIG_NO_HZ_FULL rq->last_sched_tick = 0; #endif -#endif +#endif /* CONFIG_SMP */ init_rq_hrtick(rq); atomic_set(&rq->nr_iowait, 0); } diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index b70367a3e1ef..b8a33abce650 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4491,7 +4491,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) } #ifdef CONFIG_SMP - +#ifdef CONFIG_NO_HZ_COMMON /* * per rq 'load' arrray crap; XXX kill this. */ @@ -4557,6 +4557,7 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) } return load; } +#endif /* CONFIG_NO_HZ_COMMON */ /** * __cpu_load_update - update the rq->cpu_load[] statistics @@ -4596,7 +4597,7 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) static void cpu_load_update(struct rq *this_rq, unsigned long this_load, unsigned long pending_updates) { - unsigned long tickless_load = this_rq->cpu_load[0]; + unsigned long __maybe_unused tickless_load = this_rq->cpu_load[0]; int i, scale; this_rq->nr_load_updates++; @@ -4609,6 +4610,7 @@ static void cpu_load_update(struct rq *this_rq, unsigned long this_load, /* scale is effectively 1 << i now, and >> i divides by scale */ old_load = this_rq->cpu_load[i]; +#ifdef CONFIG_NO_HZ_COMMON old_load = decay_load_missed(old_load, pending_updates - 1, i); if (tickless_load) { old_load -= decay_load_missed(tickless_load, pending_updates - 1, i); @@ -4619,6 +4621,7 @@ static void cpu_load_update(struct rq *this_rq, unsigned long this_load, */ old_load += tickless_load; } +#endif new_load = this_load; /* * Round up the averaging division if load is increasing. This @@ -4731,8 +4734,10 @@ static inline void cpu_load_update_nohz(struct rq *this_rq, static void cpu_load_update_periodic(struct rq *this_rq, unsigned long load) { +#ifdef CONFIG_NO_HZ_COMMON /* See the mess around cpu_load_update_nohz(). */ this_rq->last_load_update_tick = READ_ONCE(jiffies); +#endif cpu_load_update(this_rq, load, 1); } diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 32d9e22cfacf..69da6fcaa0e8 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -585,11 +585,13 @@ struct rq { #endif #define CPU_LOAD_IDX_MAX 5 unsigned long cpu_load[CPU_LOAD_IDX_MAX]; - unsigned long last_load_update_tick; #ifdef CONFIG_NO_HZ_COMMON +#ifdef CONFIG_SMP + unsigned long last_load_update_tick; +#endif /* CONFIG_SMP */ u64 nohz_stamp; unsigned long nohz_flags; -#endif +#endif /* CONFIG_NO_HZ_COMMON */ #ifdef CONFIG_NO_HZ_FULL unsigned long last_sched_tick; #endif -- cgit v1.2.3 From fec148c000d0f9ac21679601722811eb60b4cc52 Mon Sep 17 00:00:00 2001 From: Xunlei Pang Date: Thu, 14 Apr 2016 20:19:28 +0800 Subject: sched/deadline: Fix a bug in dl_overflow() I got a minus(very big) dl_b->total_bw during my deadline tests. # grep dl /proc/sched_debug dl_rq[0]: .dl_nr_running : 0 .dl_bw->bw : 996147 .dl_bw->total_bw : -222297900 Something unusual must have happened. After some digging, I finally noticed that when changing a deadline task to normal(cfs), and changing it back to deadline immediately, after it died, we will got the wrong dl_bw->total_bw. The root cause is in dl_overflow(), it has: if (new_bw == p->dl.dl_bw) return 0; 1) When a deadline task is changed to !deadline task, it will start dl timer in switched_from_dl(), and retain previous deadline parameter till the timer expires. 2) If we change it back to deadline with the same bandwidth parameter before the timer expires, as it keeps the old bandwidth although it is not a deadline task. dl_overflow() simply returns success without updating the right data, and got the wrong dl_bw->total_bw. The solution is simple, if @p is not deadline, don't return. Signed-off-by: Xunlei Pang Signed-off-by: Peter Zijlstra (Intel) Acked-by: Juri Lelli Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Steven Rostedt Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1460636368-1993-1-git-send-email-xlpang@redhat.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 71dffbb27ce6..9d84d6004745 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2378,7 +2378,8 @@ static int dl_overflow(struct task_struct *p, int policy, u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0; int cpus, err = -1; - if (new_bw == p->dl.dl_bw) + /* !deadline task may carry old deadline bandwidth */ + if (new_bw == p->dl.dl_bw && task_has_dl_policy(p)) return 0; /* -- cgit v1.2.3 From 594dd290cf5403a9a5818619dfff42d8e8e0518e Mon Sep 17 00:00:00 2001 From: Wanpeng Li Date: Fri, 22 Apr 2016 17:07:24 +0800 Subject: sched/cpufreq: Optimize cpufreq update kicker to avoid update multiple times Sometimes delta_exec is 0 due to update_curr() is called multiple times, this is captured by: u64 delta_exec = rq_clock_task(rq) - curr->se.exec_start; This patch optimizes the cpufreq update kicker by bailing out when nothing changed, it will benefit the upcoming schedutil, since otherwise it will (over)react to the special util/max combination. Signed-off-by: Wanpeng Li Signed-off-by: Peter Zijlstra (Intel) Cc: Peter Zijlstra Cc: Rafael J. Wysocki Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1461316044-9520-1-git-send-email-wanpeng.li@hotmail.com Signed-off-by: Ingo Molnar --- kernel/sched/deadline.c | 8 ++++---- kernel/sched/rt.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-) diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index affd97ec9f65..8f9b5af4e857 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -717,10 +717,6 @@ static void update_curr_dl(struct rq *rq) if (!dl_task(curr) || !on_dl_rq(dl_se)) return; - /* Kick cpufreq (see the comment in linux/cpufreq.h). */ - if (cpu_of(rq) == smp_processor_id()) - cpufreq_trigger_update(rq_clock(rq)); - /* * Consumed budget is computed considering the time as * observed by schedulable tasks (excluding time spent @@ -736,6 +732,10 @@ static void update_curr_dl(struct rq *rq) return; } + /* kick cpufreq (see the comment in linux/cpufreq.h). */ + if (cpu_of(rq) == smp_processor_id()) + cpufreq_trigger_update(rq_clock(rq)); + schedstat_set(curr->se.statistics.exec_max, max(curr->se.statistics.exec_max, delta_exec)); diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index c41ea7ac1764..19e13060fcd5 100644 --- a/kernel/sched/rt.c +++ b/kernel/sched/rt.c @@ -953,14 +953,14 @@ static void update_curr_rt(struct rq *rq) if (curr->sched_class != &rt_sched_class) return; - /* Kick cpufreq (see the comment in linux/cpufreq.h). */ - if (cpu_of(rq) == smp_processor_id()) - cpufreq_trigger_update(rq_clock(rq)); - delta_exec = rq_clock_task(rq) - curr->se.exec_start; if (unlikely((s64)delta_exec <= 0)) return; + /* Kick cpufreq (see the comment in linux/cpufreq.h). */ + if (cpu_of(rq) == smp_processor_id()) + cpufreq_trigger_update(rq_clock(rq)); + schedstat_set(curr->se.statistics.exec_max, max(curr->se.statistics.exec_max, delta_exec)); -- cgit v1.2.3 From 88f10e37e150569a390be7a6161fa0f26b7372e9 Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Tue, 26 Apr 2016 09:39:05 -0700 Subject: sched/core, ARM: Include linux/preempt.h from asm/mmu_context.h arm's mmu_context.h uses preempt_enable_no_resched and but doesn't include anything that would pull in the declaration. If I start including from without this, the build breaks. Signed-off-by: Andy Lutomirski Reviewed-by: Borislav Petkov Cc: Borislav Petkov Cc: Catalin Marinas Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Russell King Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/5b95730a70f2dafe12d4fbf38d20eb7330d67ba3.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar --- arch/arm/include/asm/mmu_context.h | 1 + 1 file changed, 1 insertion(+) diff --git a/arch/arm/include/asm/mmu_context.h b/arch/arm/include/asm/mmu_context.h index fa5b42d44985..ed73babc0dc9 100644 --- a/arch/arm/include/asm/mmu_context.h +++ b/arch/arm/include/asm/mmu_context.h @@ -15,6 +15,7 @@ #include #include +#include #include #include #include -- cgit v1.2.3 From 8efd755ac2fe262d4c8d5c9bbe054bb67dae93da Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 28 Apr 2016 11:39:12 +0200 Subject: mm/mmu_context, sched/core: Fix mmu_context.h assumption Some architectures (such as Alpha) rely on include/linux/sched.h definitions in their mmu_context.h files. So include sched.h before mmu_context.h. Cc: Andy Lutomirski Cc: Borislav Petkov Cc: Linus Torvalds Cc: linux-kernel@vger.kernel.org Cc: Peter Zijlstra Cc: Thomas Gleixner Signed-off-by: Ingo Molnar --- mm/mmu_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mm/mmu_context.c b/mm/mmu_context.c index f802c2d216a7..6f4d27c5bb32 100644 --- a/mm/mmu_context.c +++ b/mm/mmu_context.c @@ -4,9 +4,9 @@ */ #include +#include #include #include -#include #include -- cgit v1.2.3 From f98db6013c557c216da5038d9c52045be55cd039 Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Tue, 26 Apr 2016 09:39:06 -0700 Subject: sched/core: Add switch_mm_irqs_off() and use it in the scheduler By default, this is the same thing as switch_mm(). x86 will override it as an optimization. Signed-off-by: Andy Lutomirski Reviewed-by: Borislav Petkov Cc: Borislav Petkov Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/df401df47bdd6be3e389c6f1e3f5310d70e81b2c.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar --- include/linux/mmu_context.h | 7 +++++++ kernel/sched/core.c | 6 +++--- 2 files changed, 10 insertions(+), 3 deletions(-) diff --git a/include/linux/mmu_context.h b/include/linux/mmu_context.h index 70fffeba7495..a4441784503b 100644 --- a/include/linux/mmu_context.h +++ b/include/linux/mmu_context.h @@ -1,9 +1,16 @@ #ifndef _LINUX_MMU_CONTEXT_H #define _LINUX_MMU_CONTEXT_H +#include + struct mm_struct; void use_mm(struct mm_struct *mm); void unuse_mm(struct mm_struct *mm); +/* Architectures that care about IRQ state in switch_mm can override this. */ +#ifndef switch_mm_irqs_off +# define switch_mm_irqs_off switch_mm +#endif + #endif diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 9d84d6004745..adcafda7ffc8 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -33,7 +33,7 @@ #include #include #include -#include +#include #include #include #include @@ -2733,7 +2733,7 @@ context_switch(struct rq *rq, struct task_struct *prev, atomic_inc(&oldmm->mm_count); enter_lazy_tlb(oldmm, next); } else - switch_mm(oldmm, mm, next); + switch_mm_irqs_off(oldmm, mm, next); if (!prev->mm) { prev->active_mm = NULL; @@ -5274,7 +5274,7 @@ void idle_task_exit(void) BUG_ON(cpu_online(smp_processor_id())); if (mm != &init_mm) { - switch_mm(mm, &init_mm, current); + switch_mm_irqs_off(mm, &init_mm, current); finish_arch_post_lock_switch(); } mmdrop(mm); -- cgit v1.2.3 From e1074888c326038340a1ada9129d679e661f2ea6 Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Tue, 26 Apr 2016 09:39:07 -0700 Subject: x86/mm: Build arch/x86/mm/tlb.c even on !SMP Currently all of the functions that live in tlb.c are inlined on !SMP builds. One can debate whether this is a good idea (in many respects the code in tlb.c is better than the inlined UP code). Regardless, I want to add code that needs to be built on UP and SMP kernels and relates to tlb flushing, so arrange for tlb.c to be compiled unconditionally. Signed-off-by: Andy Lutomirski Reviewed-by: Borislav Petkov Cc: Borislav Petkov Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/f0d778f0d828fc46e5d1946bca80f0aaf9abf032.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar --- arch/x86/mm/Makefile | 3 +-- arch/x86/mm/tlb.c | 4 ++++ 2 files changed, 5 insertions(+), 2 deletions(-) diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile index f98913258c63..62c0043a5fd5 100644 --- a/arch/x86/mm/Makefile +++ b/arch/x86/mm/Makefile @@ -2,7 +2,7 @@ KCOV_INSTRUMENT_tlb.o := n obj-y := init.o init_$(BITS).o fault.o ioremap.o extable.o pageattr.o mmap.o \ - pat.o pgtable.o physaddr.o gup.o setup_nx.o + pat.o pgtable.o physaddr.o gup.o setup_nx.o tlb.o # Make sure __phys_addr has no stackprotector nostackp := $(call cc-option, -fno-stack-protector) @@ -12,7 +12,6 @@ CFLAGS_setup_nx.o := $(nostackp) CFLAGS_fault.o := -I$(src)/../include/asm/trace obj-$(CONFIG_X86_PAT) += pat_rbtree.o -obj-$(CONFIG_SMP) += tlb.o obj-$(CONFIG_X86_32) += pgtable_32.o iomap_32.o diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index fe9b9f776361..a4530e2e20d7 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -28,6 +28,8 @@ * Implement flush IPI by CALL_FUNCTION_VECTOR, Alex Shi */ +#ifdef CONFIG_SMP + struct flush_tlb_info { struct mm_struct *flush_mm; unsigned long flush_start; @@ -353,3 +355,5 @@ static int __init create_tlb_single_page_flush_ceiling(void) return 0; } late_initcall(create_tlb_single_page_flush_ceiling); + +#endif /* CONFIG_SMP */ -- cgit v1.2.3 From 69c0319aabba45bcf33178916a2f06967b4adede Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Tue, 26 Apr 2016 09:39:08 -0700 Subject: x86/mm, sched/core: Uninline switch_mm() It's fairly large and it has quite a few callers. This may also help untangle some headers down the road. Signed-off-by: Andy Lutomirski Reviewed-by: Borislav Petkov Cc: Borislav Petkov Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/54f3367803e7f80b2be62c8a21879aa74b1a5f57.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar --- arch/x86/include/asm/mmu_context.h | 98 +---------------------------------- arch/x86/mm/tlb.c | 102 +++++++++++++++++++++++++++++++++++++ 2 files changed, 104 insertions(+), 96 deletions(-) diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index 84280029cafd..bb911dd7cd01 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -115,103 +115,9 @@ static inline void destroy_context(struct mm_struct *mm) destroy_context_ldt(mm); } -static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next, - struct task_struct *tsk) -{ - unsigned cpu = smp_processor_id(); +extern void switch_mm(struct mm_struct *prev, struct mm_struct *next, + struct task_struct *tsk); - if (likely(prev != next)) { -#ifdef CONFIG_SMP - this_cpu_write(cpu_tlbstate.state, TLBSTATE_OK); - this_cpu_write(cpu_tlbstate.active_mm, next); -#endif - cpumask_set_cpu(cpu, mm_cpumask(next)); - - /* - * Re-load page tables. - * - * This logic has an ordering constraint: - * - * CPU 0: Write to a PTE for 'next' - * CPU 0: load bit 1 in mm_cpumask. if nonzero, send IPI. - * CPU 1: set bit 1 in next's mm_cpumask - * CPU 1: load from the PTE that CPU 0 writes (implicit) - * - * We need to prevent an outcome in which CPU 1 observes - * the new PTE value and CPU 0 observes bit 1 clear in - * mm_cpumask. (If that occurs, then the IPI will never - * be sent, and CPU 0's TLB will contain a stale entry.) - * - * The bad outcome can occur if either CPU's load is - * reordered before that CPU's store, so both CPUs must - * execute full barriers to prevent this from happening. - * - * Thus, switch_mm needs a full barrier between the - * store to mm_cpumask and any operation that could load - * from next->pgd. TLB fills are special and can happen - * due to instruction fetches or for no reason at all, - * and neither LOCK nor MFENCE orders them. - * Fortunately, load_cr3() is serializing and gives the - * ordering guarantee we need. - * - */ - load_cr3(next->pgd); - - trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); - - /* Stop flush ipis for the previous mm */ - cpumask_clear_cpu(cpu, mm_cpumask(prev)); - - /* Load per-mm CR4 state */ - load_mm_cr4(next); - -#ifdef CONFIG_MODIFY_LDT_SYSCALL - /* - * Load the LDT, if the LDT is different. - * - * It's possible that prev->context.ldt doesn't match - * the LDT register. This can happen if leave_mm(prev) - * was called and then modify_ldt changed - * prev->context.ldt but suppressed an IPI to this CPU. - * In this case, prev->context.ldt != NULL, because we - * never set context.ldt to NULL while the mm still - * exists. That means that next->context.ldt != - * prev->context.ldt, because mms never share an LDT. - */ - if (unlikely(prev->context.ldt != next->context.ldt)) - load_mm_ldt(next); -#endif - } -#ifdef CONFIG_SMP - else { - this_cpu_write(cpu_tlbstate.state, TLBSTATE_OK); - BUG_ON(this_cpu_read(cpu_tlbstate.active_mm) != next); - - if (!cpumask_test_cpu(cpu, mm_cpumask(next))) { - /* - * On established mms, the mm_cpumask is only changed - * from irq context, from ptep_clear_flush() while in - * lazy tlb mode, and here. Irqs are blocked during - * schedule, protecting us from simultaneous changes. - */ - cpumask_set_cpu(cpu, mm_cpumask(next)); - - /* - * We were in lazy tlb mode and leave_mm disabled - * tlb flush IPI delivery. We must reload CR3 - * to make sure to use no freed page tables. - * - * As above, load_cr3() is serializing and orders TLB - * fills with respect to the mm_cpumask write. - */ - load_cr3(next->pgd); - trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); - load_mm_cr4(next); - load_mm_ldt(next); - } - } -#endif -} #define activate_mm(prev, next) \ do { \ diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index a4530e2e20d7..ce7a0c986975 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -59,6 +59,108 @@ void leave_mm(int cpu) } EXPORT_SYMBOL_GPL(leave_mm); +#endif /* CONFIG_SMP */ + +void switch_mm(struct mm_struct *prev, struct mm_struct *next, + struct task_struct *tsk) +{ + unsigned cpu = smp_processor_id(); + + if (likely(prev != next)) { +#ifdef CONFIG_SMP + this_cpu_write(cpu_tlbstate.state, TLBSTATE_OK); + this_cpu_write(cpu_tlbstate.active_mm, next); +#endif + cpumask_set_cpu(cpu, mm_cpumask(next)); + + /* + * Re-load page tables. + * + * This logic has an ordering constraint: + * + * CPU 0: Write to a PTE for 'next' + * CPU 0: load bit 1 in mm_cpumask. if nonzero, send IPI. + * CPU 1: set bit 1 in next's mm_cpumask + * CPU 1: load from the PTE that CPU 0 writes (implicit) + * + * We need to prevent an outcome in which CPU 1 observes + * the new PTE value and CPU 0 observes bit 1 clear in + * mm_cpumask. (If that occurs, then the IPI will never + * be sent, and CPU 0's TLB will contain a stale entry.) + * + * The bad outcome can occur if either CPU's load is + * reordered before that CPU's store, so both CPUs must + * execute full barriers to prevent this from happening. + * + * Thus, switch_mm needs a full barrier between the + * store to mm_cpumask and any operation that could load + * from next->pgd. TLB fills are special and can happen + * due to instruction fetches or for no reason at all, + * and neither LOCK nor MFENCE orders them. + * Fortunately, load_cr3() is serializing and gives the + * ordering guarantee we need. + * + */ + load_cr3(next->pgd); + + trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); + + /* Stop flush ipis for the previous mm */ + cpumask_clear_cpu(cpu, mm_cpumask(prev)); + + /* Load per-mm CR4 state */ + load_mm_cr4(next); + +#ifdef CONFIG_MODIFY_LDT_SYSCALL + /* + * Load the LDT, if the LDT is different. + * + * It's possible that prev->context.ldt doesn't match + * the LDT register. This can happen if leave_mm(prev) + * was called and then modify_ldt changed + * prev->context.ldt but suppressed an IPI to this CPU. + * In this case, prev->context.ldt != NULL, because we + * never set context.ldt to NULL while the mm still + * exists. That means that next->context.ldt != + * prev->context.ldt, because mms never share an LDT. + */ + if (unlikely(prev->context.ldt != next->context.ldt)) + load_mm_ldt(next); +#endif + } +#ifdef CONFIG_SMP + else { + this_cpu_write(cpu_tlbstate.state, TLBSTATE_OK); + BUG_ON(this_cpu_read(cpu_tlbstate.active_mm) != next); + + if (!cpumask_test_cpu(cpu, mm_cpumask(next))) { + /* + * On established mms, the mm_cpumask is only changed + * from irq context, from ptep_clear_flush() while in + * lazy tlb mode, and here. Irqs are blocked during + * schedule, protecting us from simultaneous changes. + */ + cpumask_set_cpu(cpu, mm_cpumask(next)); + + /* + * We were in lazy tlb mode and leave_mm disabled + * tlb flush IPI delivery. We must reload CR3 + * to make sure to use no freed page tables. + * + * As above, load_cr3() is serializing and orders TLB + * fills with respect to the mm_cpumask write. + */ + load_cr3(next->pgd); + trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); + load_mm_cr4(next); + load_mm_ldt(next); + } + } +#endif +} + +#ifdef CONFIG_SMP + /* * The flush IPI assumes that a thread switch happens in this order: * [cpu0: the cpu that switches] -- cgit v1.2.3 From 078194f8e9fe3cf54c8fd8bded48a1db5bd8eb8a Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Tue, 26 Apr 2016 09:39:09 -0700 Subject: x86/mm, sched/core: Turn off IRQs in switch_mm() Potential races between switch_mm() and TLB-flush or LDT-flush IPIs could be very messy. AFAICT the code is currently okay, whether by accident or by careful design, but enabling PCID will make it considerably more complicated and will no longer be obviously safe. Fix it with a big hammer: run switch_mm() with IRQs off. To avoid a performance hit in the scheduler, we take advantage of our knowledge that the scheduler already has IRQs disabled when it calls switch_mm(). Signed-off-by: Andy Lutomirski Reviewed-by: Borislav Petkov Cc: Borislav Petkov Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/f19baf759693c9dcae64bbff76189db77cb13398.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar --- arch/x86/include/asm/mmu_context.h | 3 +++ arch/x86/mm/tlb.c | 10 ++++++++++ 2 files changed, 13 insertions(+) diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index bb911dd7cd01..396348196aa7 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -118,6 +118,9 @@ static inline void destroy_context(struct mm_struct *mm) extern void switch_mm(struct mm_struct *prev, struct mm_struct *next, struct task_struct *tsk); +extern void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next, + struct task_struct *tsk); +#define switch_mm_irqs_off switch_mm_irqs_off #define activate_mm(prev, next) \ do { \ diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index ce7a0c986975..5643fd0b1a7d 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -63,6 +63,16 @@ EXPORT_SYMBOL_GPL(leave_mm); void switch_mm(struct mm_struct *prev, struct mm_struct *next, struct task_struct *tsk) +{ + unsigned long flags; + + local_irq_save(flags); + switch_mm_irqs_off(prev, next, tsk); + local_irq_restore(flags); +} + +void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next, + struct task_struct *tsk) { unsigned cpu = smp_processor_id(); -- cgit v1.2.3 From 3e71a462dd483ce508a723356b293731e7d788ea Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 28 Apr 2016 16:16:33 +0200 Subject: sched/core: Move task_rq_lock() out of line Its a rather large function, inline doesn't seems to make much sense: $ size defconfig-build/kernel/sched/core.o{.orig,} text data bss dec hex filename 56533 21037 2320 79890 13812 defconfig-build/kernel/sched/core.o.orig 55733 21037 2320 79090 134f2 defconfig-build/kernel/sched/core.o The 'perf bench sched messaging' micro-benchmark shows a visible improvement of 4-5%: $ for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i ; done $ perf stat --null --repeat 25 -- perf bench sched messaging -g 40 -l 5000 pre: 4.582798193 seconds time elapsed ( +- 1.41% ) 4.733374877 seconds time elapsed ( +- 2.10% ) 4.560955136 seconds time elapsed ( +- 1.43% ) 4.631062303 seconds time elapsed ( +- 1.40% ) post: 4.364765213 seconds time elapsed ( +- 0.91% ) 4.454442734 seconds time elapsed ( +- 1.18% ) 4.448893817 seconds time elapsed ( +- 1.41% ) 4.424346872 seconds time elapsed ( +- 0.97% ) Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 65 ++++++++++++++++++++++++++++++++++++++++++++++++++ kernel/sched/sched.h | 67 ++++------------------------------------------------ 2 files changed, 69 insertions(+), 63 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index c82ca6eccfec..1b609a886795 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -170,6 +170,71 @@ static struct rq *this_rq_lock(void) return rq; } +/* + * __task_rq_lock - lock the rq @p resides on. + */ +struct rq *__task_rq_lock(struct task_struct *p) + __acquires(rq->lock) +{ + struct rq *rq; + + lockdep_assert_held(&p->pi_lock); + + for (;;) { + rq = task_rq(p); + raw_spin_lock(&rq->lock); + if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { + lockdep_pin_lock(&rq->lock); + return rq; + } + raw_spin_unlock(&rq->lock); + + while (unlikely(task_on_rq_migrating(p))) + cpu_relax(); + } +} + +/* + * task_rq_lock - lock p->pi_lock and lock the rq @p resides on. + */ +struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) + __acquires(p->pi_lock) + __acquires(rq->lock) +{ + struct rq *rq; + + for (;;) { + raw_spin_lock_irqsave(&p->pi_lock, *flags); + rq = task_rq(p); + raw_spin_lock(&rq->lock); + /* + * move_queued_task() task_rq_lock() + * + * ACQUIRE (rq->lock) + * [S] ->on_rq = MIGRATING [L] rq = task_rq() + * WMB (__set_task_cpu()) ACQUIRE (rq->lock); + * [S] ->cpu = new_cpu [L] task_rq() + * [L] ->on_rq + * RELEASE (rq->lock) + * + * If we observe the old cpu in task_rq_lock, the acquire of + * the old rq->lock will fully serialize against the stores. + * + * If we observe the new cpu in task_rq_lock, the acquire will + * pair with the WMB to ensure we must then also see migrating. + */ + if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { + lockdep_pin_lock(&rq->lock); + return rq; + } + raw_spin_unlock(&rq->lock); + raw_spin_unlock_irqrestore(&p->pi_lock, *flags); + + while (unlikely(task_on_rq_migrating(p))) + cpu_relax(); + } +} + #ifdef CONFIG_SCHED_HRTICK /* * Use HR-timers to deliver accurate preemption points. diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 69da6fcaa0e8..aab4cf05d48a 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1451,70 +1451,11 @@ static inline void sched_rt_avg_update(struct rq *rq, u64 rt_delta) { } static inline void sched_avg_update(struct rq *rq) { } #endif -/* - * __task_rq_lock - lock the rq @p resides on. - */ -static inline struct rq *__task_rq_lock(struct task_struct *p) - __acquires(rq->lock) -{ - struct rq *rq; - - lockdep_assert_held(&p->pi_lock); - - for (;;) { - rq = task_rq(p); - raw_spin_lock(&rq->lock); - if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { - lockdep_pin_lock(&rq->lock); - return rq; - } - raw_spin_unlock(&rq->lock); - - while (unlikely(task_on_rq_migrating(p))) - cpu_relax(); - } -} - -/* - * task_rq_lock - lock p->pi_lock and lock the rq @p resides on. - */ -static inline struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) +struct rq *__task_rq_lock(struct task_struct *p) + __acquires(rq->lock); +struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) __acquires(p->pi_lock) - __acquires(rq->lock) -{ - struct rq *rq; - - for (;;) { - raw_spin_lock_irqsave(&p->pi_lock, *flags); - rq = task_rq(p); - raw_spin_lock(&rq->lock); - /* - * move_queued_task() task_rq_lock() - * - * ACQUIRE (rq->lock) - * [S] ->on_rq = MIGRATING [L] rq = task_rq() - * WMB (__set_task_cpu()) ACQUIRE (rq->lock); - * [S] ->cpu = new_cpu [L] task_rq() - * [L] ->on_rq - * RELEASE (rq->lock) - * - * If we observe the old cpu in task_rq_lock, the acquire of - * the old rq->lock will fully serialize against the stores. - * - * If we observe the new cpu in task_rq_lock, the acquire will - * pair with the WMB to ensure we must then also see migrating. - */ - if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { - lockdep_pin_lock(&rq->lock); - return rq; - } - raw_spin_unlock(&rq->lock); - raw_spin_unlock_irqrestore(&p->pi_lock, *flags); - - while (unlikely(task_on_rq_migrating(p))) - cpu_relax(); - } -} + __acquires(rq->lock); static inline void __task_rq_unlock(struct rq *rq) __releases(rq->lock) -- cgit v1.2.3 From eb58075149b7f0300ff19142e6245fe75db2a081 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 31 Jul 2015 21:28:18 +0200 Subject: sched/core: Introduce 'struct rq_flags' In order to be able to pass around more than just the IRQ flags in the future, add a rq_flags structure. No difference in code generation for the x86_64-defconfig build I tested. Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 98 +++++++++++++++++++++++++------------------------ kernel/sched/deadline.c | 6 +-- kernel/sched/sched.h | 14 ++++--- 3 files changed, 62 insertions(+), 56 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 1b609a886795..7c7db60115b4 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -173,7 +173,7 @@ static struct rq *this_rq_lock(void) /* * __task_rq_lock - lock the rq @p resides on. */ -struct rq *__task_rq_lock(struct task_struct *p) +struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf) __acquires(rq->lock) { struct rq *rq; @@ -197,14 +197,14 @@ struct rq *__task_rq_lock(struct task_struct *p) /* * task_rq_lock - lock p->pi_lock and lock the rq @p resides on. */ -struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) +struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf) __acquires(p->pi_lock) __acquires(rq->lock) { struct rq *rq; for (;;) { - raw_spin_lock_irqsave(&p->pi_lock, *flags); + raw_spin_lock_irqsave(&p->pi_lock, rf->flags); rq = task_rq(p); raw_spin_lock(&rq->lock); /* @@ -228,7 +228,7 @@ struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) return rq; } raw_spin_unlock(&rq->lock); - raw_spin_unlock_irqrestore(&p->pi_lock, *flags); + raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags); while (unlikely(task_on_rq_migrating(p))) cpu_relax(); @@ -1150,12 +1150,12 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) static int __set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask, bool check) { - unsigned long flags; - struct rq *rq; unsigned int dest_cpu; + struct rq_flags rf; + struct rq *rq; int ret = 0; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * Must re-check here, to close a race against __kthread_bind(), @@ -1184,7 +1184,7 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, if (task_running(rq, p) || p->state == TASK_WAKING) { struct migration_arg arg = { p, dest_cpu }; /* Need help from migration thread: drop lock and wait. */ - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg); tlb_migrate_finish(p->mm); return 0; @@ -1198,7 +1198,7 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, lockdep_pin_lock(&rq->lock); } out: - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return ret; } @@ -1382,8 +1382,8 @@ out: */ unsigned long wait_task_inactive(struct task_struct *p, long match_state) { - unsigned long flags; int running, queued; + struct rq_flags rf; unsigned long ncsw; struct rq *rq; @@ -1418,14 +1418,14 @@ unsigned long wait_task_inactive(struct task_struct *p, long match_state) * lock now, to be *sure*. If we're wrong, we'll * just go back and repeat. */ - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); trace_sched_wait_task(p); running = task_running(rq, p); queued = task_on_rq_queued(p); ncsw = 0; if (!match_state || p->state == match_state) ncsw = p->nvcsw | LONG_MIN; /* sets MSB */ - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); /* * If it changed from the expected state, bail out now. @@ -1723,17 +1723,18 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags) */ static int ttwu_remote(struct task_struct *p, int wake_flags) { + struct rq_flags rf; struct rq *rq; int ret = 0; - rq = __task_rq_lock(p); + rq = __task_rq_lock(p, &rf); if (task_on_rq_queued(p)) { /* check_preempt_curr() may use rq clock */ update_rq_clock(rq); ttwu_do_wakeup(rq, p, wake_flags); ret = 1; } - __task_rq_unlock(rq); + __task_rq_unlock(rq, &rf); return ret; } @@ -2486,12 +2487,12 @@ extern void init_dl_bw(struct dl_bw *dl_b); */ void wake_up_new_task(struct task_struct *p) { - unsigned long flags; + struct rq_flags rf; struct rq *rq; - raw_spin_lock_irqsave(&p->pi_lock, flags); /* Initialize new task's runnable average */ init_entity_runnable_average(&p->se); + raw_spin_lock_irqsave(&p->pi_lock, rf.flags); #ifdef CONFIG_SMP /* * Fork balancing, do it here and not earlier because: @@ -2503,7 +2504,7 @@ void wake_up_new_task(struct task_struct *p) /* Post initialize new task's util average when its cfs_rq is set */ post_init_entity_util_avg(&p->se); - rq = __task_rq_lock(p); + rq = __task_rq_lock(p, &rf); activate_task(rq, p, 0); p->on_rq = TASK_ON_RQ_QUEUED; trace_sched_wakeup_new(p); @@ -2519,7 +2520,7 @@ void wake_up_new_task(struct task_struct *p) lockdep_pin_lock(&rq->lock); } #endif - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); } #ifdef CONFIG_PREEMPT_NOTIFIERS @@ -2935,7 +2936,7 @@ EXPORT_PER_CPU_SYMBOL(kernel_cpustat); */ unsigned long long task_sched_runtime(struct task_struct *p) { - unsigned long flags; + struct rq_flags rf; struct rq *rq; u64 ns; @@ -2955,7 +2956,7 @@ unsigned long long task_sched_runtime(struct task_struct *p) return p->se.sum_exec_runtime; #endif - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * Must be ->curr _and_ ->on_rq. If dequeued, we would * project cycles that may never be accounted to this @@ -2966,7 +2967,7 @@ unsigned long long task_sched_runtime(struct task_struct *p) p->sched_class->update_curr(rq); } ns = p->se.sum_exec_runtime; - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return ns; } @@ -3524,12 +3525,13 @@ EXPORT_SYMBOL(default_wake_function); void rt_mutex_setprio(struct task_struct *p, int prio) { int oldprio, queued, running, queue_flag = DEQUEUE_SAVE | DEQUEUE_MOVE; - struct rq *rq; const struct sched_class *prev_class; + struct rq_flags rf; + struct rq *rq; BUG_ON(prio > MAX_PRIO); - rq = __task_rq_lock(p); + rq = __task_rq_lock(p, &rf); /* * Idle task boosting is a nono in general. There is one @@ -3605,7 +3607,7 @@ void rt_mutex_setprio(struct task_struct *p, int prio) check_class_changed(rq, p, prev_class, oldprio); out_unlock: preempt_disable(); /* avoid rq from going away on us */ - __task_rq_unlock(rq); + __task_rq_unlock(rq, &rf); balance_callback(rq); preempt_enable(); @@ -3615,7 +3617,7 @@ out_unlock: void set_user_nice(struct task_struct *p, long nice) { int old_prio, delta, queued; - unsigned long flags; + struct rq_flags rf; struct rq *rq; if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE) @@ -3624,7 +3626,7 @@ void set_user_nice(struct task_struct *p, long nice) * We have to be careful, if called from sys_setpriority(), * the task might be in the middle of scheduling on another CPU. */ - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * The RT priorities are set via sched_setscheduler(), but we still * allow the 'normal' nice value to be set - but as expected @@ -3655,7 +3657,7 @@ void set_user_nice(struct task_struct *p, long nice) resched_curr(rq); } out_unlock: - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); } EXPORT_SYMBOL(set_user_nice); @@ -3952,11 +3954,11 @@ static int __sched_setscheduler(struct task_struct *p, MAX_RT_PRIO - 1 - attr->sched_priority; int retval, oldprio, oldpolicy = -1, queued, running; int new_effective_prio, policy = attr->sched_policy; - unsigned long flags; const struct sched_class *prev_class; - struct rq *rq; + struct rq_flags rf; int reset_on_fork; int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE; + struct rq *rq; /* may grab non-irq protected spin_locks */ BUG_ON(in_interrupt()); @@ -4051,13 +4053,13 @@ recheck: * To be able to change p->policy safely, the appropriate * runqueue lock must be held. */ - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * Changing the policy of the stop threads its a very bad idea */ if (p == rq->stop) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EINVAL; } @@ -4074,7 +4076,7 @@ recheck: goto change; p->sched_reset_on_fork = reset_on_fork; - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return 0; } change: @@ -4088,7 +4090,7 @@ change: if (rt_bandwidth_enabled() && rt_policy(policy) && task_group(p)->rt_bandwidth.rt_runtime == 0 && !task_group_is_autogroup(task_group(p))) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EPERM; } #endif @@ -4103,7 +4105,7 @@ change: */ if (!cpumask_subset(span, &p->cpus_allowed) || rq->rd->dl_bw.bw == 0) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EPERM; } } @@ -4113,7 +4115,7 @@ change: /* recheck policy now with rq lock held */ if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) { policy = oldpolicy = -1; - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); goto recheck; } @@ -4123,7 +4125,7 @@ change: * is available. */ if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EBUSY; } @@ -4168,7 +4170,7 @@ change: check_class_changed(rq, p, prev_class, oldprio); preempt_disable(); /* avoid rq from going away on us */ - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); if (pi) rt_mutex_adjust_pi(p); @@ -5021,10 +5023,10 @@ SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid, { struct task_struct *p; unsigned int time_slice; - unsigned long flags; + struct rq_flags rf; + struct timespec t; struct rq *rq; int retval; - struct timespec t; if (pid < 0) return -EINVAL; @@ -5039,11 +5041,11 @@ SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid, if (retval) goto out_unlock; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); time_slice = 0; if (p->sched_class->get_rr_interval) time_slice = p->sched_class->get_rr_interval(rq, p); - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); rcu_read_unlock(); jiffies_to_timespec(time_slice, &t); @@ -5307,11 +5309,11 @@ int migrate_task_to(struct task_struct *p, int target_cpu) */ void sched_setnuma(struct task_struct *p, int nid) { - struct rq *rq; - unsigned long flags; bool queued, running; + struct rq_flags rf; + struct rq *rq; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); queued = task_on_rq_queued(p); running = task_current(rq, p); @@ -5326,7 +5328,7 @@ void sched_setnuma(struct task_struct *p, int nid) p->sched_class->set_curr_task(rq); if (queued) enqueue_task(rq, p, ENQUEUE_RESTORE); - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); } #endif /* CONFIG_NUMA_BALANCING */ @@ -7757,10 +7759,10 @@ void sched_move_task(struct task_struct *tsk) { struct task_group *tg; int queued, running; - unsigned long flags; + struct rq_flags rf; struct rq *rq; - rq = task_rq_lock(tsk, &flags); + rq = task_rq_lock(tsk, &rf); running = task_current(rq, tsk); queued = task_on_rq_queued(tsk); @@ -7792,7 +7794,7 @@ void sched_move_task(struct task_struct *tsk) if (queued) enqueue_task(rq, tsk, ENQUEUE_RESTORE | ENQUEUE_MOVE); - task_rq_unlock(rq, tsk, &flags); + task_rq_unlock(rq, tsk, &rf); } #endif /* CONFIG_CGROUP_SCHED */ diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 8f9b5af4e857..738e3c84dfe1 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -591,10 +591,10 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer) struct sched_dl_entity, dl_timer); struct task_struct *p = dl_task_of(dl_se); - unsigned long flags; + struct rq_flags rf; struct rq *rq; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * The task might have changed its scheduling policy to something @@ -677,7 +677,7 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer) #endif unlock: - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); /* * This can free the task_struct, including this hrtimer, do not touch diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index aab4cf05d48a..a5eecb1e5e4b 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1451,13 +1451,17 @@ static inline void sched_rt_avg_update(struct rq *rq, u64 rt_delta) { } static inline void sched_avg_update(struct rq *rq) { } #endif -struct rq *__task_rq_lock(struct task_struct *p) +struct rq_flags { + unsigned long flags; +}; + +struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf) __acquires(rq->lock); -struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) +struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf) __acquires(p->pi_lock) __acquires(rq->lock); -static inline void __task_rq_unlock(struct rq *rq) +static inline void __task_rq_unlock(struct rq *rq, struct rq_flags *rf) __releases(rq->lock) { lockdep_unpin_lock(&rq->lock); @@ -1465,13 +1469,13 @@ static inline void __task_rq_unlock(struct rq *rq) } static inline void -task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags) +task_rq_unlock(struct rq *rq, struct task_struct *p, struct rq_flags *rf) __releases(rq->lock) __releases(p->pi_lock) { lockdep_unpin_lock(&rq->lock); raw_spin_unlock(&rq->lock); - raw_spin_unlock_irqrestore(&p->pi_lock, *flags); + raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags); } #ifdef CONFIG_SMP -- cgit v1.2.3 From e7904a28f5331c21d17af638cb477c83662e3cb6 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 1 Aug 2015 19:25:08 +0200 Subject: locking/lockdep, sched/core: Implement a better lock pinning scheme The problem with the existing lock pinning is that each pin is of value 1; this mean you can simply unpin if you know its pinned, without having any extra information. This scheme generates a random (16 bit) cookie for each pin and requires this same cookie to unpin. This means you have to keep the cookie in context. No objsize difference for !LOCKDEP kernels. Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Linus Torvalds Cc: Mike Galbraith Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- include/linux/lockdep.h | 23 ++++++++++---- kernel/locking/lockdep.c | 71 +++++++++++++++++++++++++++++++++++++------ kernel/sched/core.c | 79 +++++++++++++++++++++++++----------------------- kernel/sched/deadline.c | 11 ++++--- kernel/sched/fair.c | 6 ++-- kernel/sched/idle_task.c | 2 +- kernel/sched/rt.c | 6 ++-- kernel/sched/sched.h | 8 +++-- kernel/sched/stop_task.c | 2 +- 9 files changed, 140 insertions(+), 68 deletions(-) diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h index d026b190c530..cf9bf9612702 100644 --- a/include/linux/lockdep.h +++ b/include/linux/lockdep.h @@ -354,8 +354,13 @@ extern void lockdep_set_current_reclaim_state(gfp_t gfp_mask); extern void lockdep_clear_current_reclaim_state(void); extern void lockdep_trace_alloc(gfp_t mask); -extern void lock_pin_lock(struct lockdep_map *lock); -extern void lock_unpin_lock(struct lockdep_map *lock); +struct pin_cookie { unsigned int val; }; + +#define NIL_COOKIE (struct pin_cookie){ .val = 0U, } + +extern struct pin_cookie lock_pin_lock(struct lockdep_map *lock); +extern void lock_repin_lock(struct lockdep_map *lock, struct pin_cookie); +extern void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie); # define INIT_LOCKDEP .lockdep_recursion = 0, .lockdep_reclaim_gfp = 0, @@ -371,8 +376,9 @@ extern void lock_unpin_lock(struct lockdep_map *lock); #define lockdep_recursing(tsk) ((tsk)->lockdep_recursion) -#define lockdep_pin_lock(l) lock_pin_lock(&(l)->dep_map) -#define lockdep_unpin_lock(l) lock_unpin_lock(&(l)->dep_map) +#define lockdep_pin_lock(l) lock_pin_lock(&(l)->dep_map) +#define lockdep_repin_lock(l,c) lock_repin_lock(&(l)->dep_map, (c)) +#define lockdep_unpin_lock(l,c) lock_unpin_lock(&(l)->dep_map, (c)) #else /* !CONFIG_LOCKDEP */ @@ -425,8 +431,13 @@ struct lock_class_key { }; #define lockdep_recursing(tsk) (0) -#define lockdep_pin_lock(l) do { (void)(l); } while (0) -#define lockdep_unpin_lock(l) do { (void)(l); } while (0) +struct pin_cookie { }; + +#define NIL_COOKIE (struct pin_cookie){ } + +#define lockdep_pin_lock(l) ({ struct pin_cookie cookie; cookie; }) +#define lockdep_repin_lock(l, c) do { (void)(l); (void)(c); } while (0) +#define lockdep_unpin_lock(l, c) do { (void)(l); (void)(c); } while (0) #endif /* !LOCKDEP */ diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c index ed9410936a22..d7f94f4c811d 100644 --- a/kernel/locking/lockdep.c +++ b/kernel/locking/lockdep.c @@ -45,6 +45,7 @@ #include #include #include +#include #include @@ -3554,7 +3555,35 @@ static int __lock_is_held(struct lockdep_map *lock) return 0; } -static void __lock_pin_lock(struct lockdep_map *lock) +static struct pin_cookie __lock_pin_lock(struct lockdep_map *lock) +{ + struct pin_cookie cookie = NIL_COOKIE; + struct task_struct *curr = current; + int i; + + if (unlikely(!debug_locks)) + return cookie; + + for (i = 0; i < curr->lockdep_depth; i++) { + struct held_lock *hlock = curr->held_locks + i; + + if (match_held_lock(hlock, lock)) { + /* + * Grab 16bits of randomness; this is sufficient to not + * be guessable and still allows some pin nesting in + * our u32 pin_count. + */ + cookie.val = 1 + (prandom_u32() >> 16); + hlock->pin_count += cookie.val; + return cookie; + } + } + + WARN(1, "pinning an unheld lock\n"); + return cookie; +} + +static void __lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie) { struct task_struct *curr = current; int i; @@ -3566,7 +3595,7 @@ static void __lock_pin_lock(struct lockdep_map *lock) struct held_lock *hlock = curr->held_locks + i; if (match_held_lock(hlock, lock)) { - hlock->pin_count++; + hlock->pin_count += cookie.val; return; } } @@ -3574,7 +3603,7 @@ static void __lock_pin_lock(struct lockdep_map *lock) WARN(1, "pinning an unheld lock\n"); } -static void __lock_unpin_lock(struct lockdep_map *lock) +static void __lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie) { struct task_struct *curr = current; int i; @@ -3589,7 +3618,11 @@ static void __lock_unpin_lock(struct lockdep_map *lock) if (WARN(!hlock->pin_count, "unpinning an unpinned lock\n")) return; - hlock->pin_count--; + hlock->pin_count -= cookie.val; + + if (WARN((int)hlock->pin_count < 0, "pin count corrupted\n")) + hlock->pin_count = 0; + return; } } @@ -3720,24 +3753,44 @@ int lock_is_held(struct lockdep_map *lock) } EXPORT_SYMBOL_GPL(lock_is_held); -void lock_pin_lock(struct lockdep_map *lock) +struct pin_cookie lock_pin_lock(struct lockdep_map *lock) { + struct pin_cookie cookie = NIL_COOKIE; unsigned long flags; if (unlikely(current->lockdep_recursion)) - return; + return cookie; raw_local_irq_save(flags); check_flags(flags); current->lockdep_recursion = 1; - __lock_pin_lock(lock); + cookie = __lock_pin_lock(lock); current->lockdep_recursion = 0; raw_local_irq_restore(flags); + + return cookie; } EXPORT_SYMBOL_GPL(lock_pin_lock); -void lock_unpin_lock(struct lockdep_map *lock) +void lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie) +{ + unsigned long flags; + + if (unlikely(current->lockdep_recursion)) + return; + + raw_local_irq_save(flags); + check_flags(flags); + + current->lockdep_recursion = 1; + __lock_repin_lock(lock, cookie); + current->lockdep_recursion = 0; + raw_local_irq_restore(flags); +} +EXPORT_SYMBOL_GPL(lock_repin_lock); + +void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie) { unsigned long flags; @@ -3748,7 +3801,7 @@ void lock_unpin_lock(struct lockdep_map *lock) check_flags(flags); current->lockdep_recursion = 1; - __lock_unpin_lock(lock); + __lock_unpin_lock(lock, cookie); current->lockdep_recursion = 0; raw_local_irq_restore(flags); } diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 7c7db60115b4..71c5a753e6e9 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -184,7 +184,7 @@ struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf) rq = task_rq(p); raw_spin_lock(&rq->lock); if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { - lockdep_pin_lock(&rq->lock); + rf->cookie = lockdep_pin_lock(&rq->lock); return rq; } raw_spin_unlock(&rq->lock); @@ -224,7 +224,7 @@ struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf) * pair with the WMB to ensure we must then also see migrating. */ if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { - lockdep_pin_lock(&rq->lock); + rf->cookie = lockdep_pin_lock(&rq->lock); return rq; } raw_spin_unlock(&rq->lock); @@ -1193,9 +1193,9 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, * OK, since we're going to drop the lock immediately * afterwards anyway. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf.cookie); rq = move_queued_task(rq, p, dest_cpu); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, rf.cookie); } out: task_rq_unlock(rq, p, &rf); @@ -1669,8 +1669,8 @@ static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_fl /* * Mark the task runnable and perform wakeup-preemption. */ -static void -ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) +static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags, + struct pin_cookie cookie) { check_preempt_curr(rq, p, wake_flags); p->state = TASK_RUNNING; @@ -1682,9 +1682,9 @@ ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) * Our task @p is fully woken up and running; so its safe to * drop the rq->lock, hereafter rq is only used for statistics. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); p->sched_class->task_woken(rq, p); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); } if (rq->idle_stamp) { @@ -1702,7 +1702,8 @@ ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) } static void -ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags) +ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags, + struct pin_cookie cookie) { lockdep_assert_held(&rq->lock); @@ -1712,7 +1713,7 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags) #endif ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_WAKING); - ttwu_do_wakeup(rq, p, wake_flags); + ttwu_do_wakeup(rq, p, wake_flags, cookie); } /* @@ -1731,7 +1732,7 @@ static int ttwu_remote(struct task_struct *p, int wake_flags) if (task_on_rq_queued(p)) { /* check_preempt_curr() may use rq clock */ update_rq_clock(rq); - ttwu_do_wakeup(rq, p, wake_flags); + ttwu_do_wakeup(rq, p, wake_flags, rf.cookie); ret = 1; } __task_rq_unlock(rq, &rf); @@ -1744,6 +1745,7 @@ void sched_ttwu_pending(void) { struct rq *rq = this_rq(); struct llist_node *llist = llist_del_all(&rq->wake_list); + struct pin_cookie cookie; struct task_struct *p; unsigned long flags; @@ -1751,15 +1753,15 @@ void sched_ttwu_pending(void) return; raw_spin_lock_irqsave(&rq->lock, flags); - lockdep_pin_lock(&rq->lock); + cookie = lockdep_pin_lock(&rq->lock); while (llist) { p = llist_entry(llist, struct task_struct, wake_entry); llist = llist_next(llist); - ttwu_do_activate(rq, p, 0); + ttwu_do_activate(rq, p, 0, cookie); } - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock_irqrestore(&rq->lock, flags); } @@ -1846,6 +1848,7 @@ bool cpus_share_cache(int this_cpu, int that_cpu) static void ttwu_queue(struct task_struct *p, int cpu) { struct rq *rq = cpu_rq(cpu); + struct pin_cookie cookie; #if defined(CONFIG_SMP) if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) { @@ -1856,9 +1859,9 @@ static void ttwu_queue(struct task_struct *p, int cpu) #endif raw_spin_lock(&rq->lock); - lockdep_pin_lock(&rq->lock); - ttwu_do_activate(rq, p, 0); - lockdep_unpin_lock(&rq->lock); + cookie = lockdep_pin_lock(&rq->lock); + ttwu_do_activate(rq, p, 0, cookie); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); } @@ -2055,7 +2058,7 @@ out: * ensure that this_rq() is locked, @p is bound to this_rq() and not * the current task. */ -static void try_to_wake_up_local(struct task_struct *p) +static void try_to_wake_up_local(struct task_struct *p, struct pin_cookie cookie) { struct rq *rq = task_rq(p); @@ -2072,11 +2075,11 @@ static void try_to_wake_up_local(struct task_struct *p) * disabled avoiding further scheduler activity on it and we've * not yet picked a replacement task. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); raw_spin_lock(&p->pi_lock); raw_spin_lock(&rq->lock); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); } if (!(p->state & TASK_NORMAL)) @@ -2087,7 +2090,7 @@ static void try_to_wake_up_local(struct task_struct *p) if (!task_on_rq_queued(p)) ttwu_activate(rq, p, ENQUEUE_WAKEUP); - ttwu_do_wakeup(rq, p, 0); + ttwu_do_wakeup(rq, p, 0, cookie); if (schedstat_enabled()) ttwu_stat(p, smp_processor_id(), 0); out: @@ -2515,9 +2518,9 @@ void wake_up_new_task(struct task_struct *p) * Nothing relies on rq->lock after this, so its fine to * drop it. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf.cookie); p->sched_class->task_woken(rq, p); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, rf.cookie); } #endif task_rq_unlock(rq, p, &rf); @@ -2782,7 +2785,7 @@ asmlinkage __visible void schedule_tail(struct task_struct *prev) */ static __always_inline struct rq * context_switch(struct rq *rq, struct task_struct *prev, - struct task_struct *next) + struct task_struct *next, struct pin_cookie cookie) { struct mm_struct *mm, *oldmm; @@ -2814,7 +2817,7 @@ context_switch(struct rq *rq, struct task_struct *prev, * of the scheduler it's an obvious special-case), so we * do an early lockdep release here: */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); spin_release(&rq->lock.dep_map, 1, _THIS_IP_); /* Here we just switch the register state and the stack. */ @@ -3154,7 +3157,7 @@ static inline void schedule_debug(struct task_struct *prev) * Pick up the highest-prio task: */ static inline struct task_struct * -pick_next_task(struct rq *rq, struct task_struct *prev) +pick_next_task(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { const struct sched_class *class = &fair_sched_class; struct task_struct *p; @@ -3165,20 +3168,20 @@ pick_next_task(struct rq *rq, struct task_struct *prev) */ if (likely(prev->sched_class == class && rq->nr_running == rq->cfs.h_nr_running)) { - p = fair_sched_class.pick_next_task(rq, prev); + p = fair_sched_class.pick_next_task(rq, prev, cookie); if (unlikely(p == RETRY_TASK)) goto again; /* assumes fair_sched_class->next == idle_sched_class */ if (unlikely(!p)) - p = idle_sched_class.pick_next_task(rq, prev); + p = idle_sched_class.pick_next_task(rq, prev, cookie); return p; } again: for_each_class(class) { - p = class->pick_next_task(rq, prev); + p = class->pick_next_task(rq, prev, cookie); if (p) { if (unlikely(p == RETRY_TASK)) goto again; @@ -3232,6 +3235,7 @@ static void __sched notrace __schedule(bool preempt) { struct task_struct *prev, *next; unsigned long *switch_count; + struct pin_cookie cookie; struct rq *rq; int cpu; @@ -3265,7 +3269,7 @@ static void __sched notrace __schedule(bool preempt) */ smp_mb__before_spinlock(); raw_spin_lock(&rq->lock); - lockdep_pin_lock(&rq->lock); + cookie = lockdep_pin_lock(&rq->lock); rq->clock_skip_update <<= 1; /* promote REQ to ACT */ @@ -3287,7 +3291,7 @@ static void __sched notrace __schedule(bool preempt) to_wakeup = wq_worker_sleeping(prev); if (to_wakeup) - try_to_wake_up_local(to_wakeup); + try_to_wake_up_local(to_wakeup, cookie); } } switch_count = &prev->nvcsw; @@ -3296,7 +3300,7 @@ static void __sched notrace __schedule(bool preempt) if (task_on_rq_queued(prev)) update_rq_clock(rq); - next = pick_next_task(rq, prev); + next = pick_next_task(rq, prev, cookie); clear_tsk_need_resched(prev); clear_preempt_need_resched(); rq->clock_skip_update = 0; @@ -3307,9 +3311,9 @@ static void __sched notrace __schedule(bool preempt) ++*switch_count; trace_sched_switch(preempt, prev, next); - rq = context_switch(rq, prev, next); /* unlocks the rq */ + rq = context_switch(rq, prev, next, cookie); /* unlocks the rq */ } else { - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock_irq(&rq->lock); } @@ -5392,6 +5396,7 @@ static void migrate_tasks(struct rq *dead_rq) { struct rq *rq = dead_rq; struct task_struct *next, *stop = rq->stop; + struct pin_cookie cookie; int dest_cpu; /* @@ -5423,8 +5428,8 @@ static void migrate_tasks(struct rq *dead_rq) /* * pick_next_task assumes pinned rq->lock. */ - lockdep_pin_lock(&rq->lock); - next = pick_next_task(rq, &fake_task); + cookie = lockdep_pin_lock(&rq->lock); + next = pick_next_task(rq, &fake_task, cookie); BUG_ON(!next); next->sched_class->put_prev_task(rq, next); @@ -5437,7 +5442,7 @@ static void migrate_tasks(struct rq *dead_rq) * because !cpu_active at this point, which means load-balance * will not interfere. Also, stop-machine. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); raw_spin_lock(&next->pi_lock); raw_spin_lock(&rq->lock); diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 738e3c84dfe1..ba53a87bb978 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -670,9 +670,9 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer) * Nothing relies on rq->lock after this, so its safe to drop * rq->lock. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf.cookie); push_dl_task(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, rf.cookie); } #endif @@ -1125,7 +1125,8 @@ static struct sched_dl_entity *pick_next_dl_entity(struct rq *rq, return rb_entry(left, struct sched_dl_entity, rb_node); } -struct task_struct *pick_next_task_dl(struct rq *rq, struct task_struct *prev) +struct task_struct * +pick_next_task_dl(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct sched_dl_entity *dl_se; struct task_struct *p; @@ -1140,9 +1141,9 @@ struct task_struct *pick_next_task_dl(struct rq *rq, struct task_struct *prev) * disabled avoiding further scheduler activity on it and we're * being very careful to re-start the picking loop. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); pull_dl_task(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); /* * pull_rt_task() can drop (and re-acquire) rq->lock; this * means a stop task can slip in, in which case we need to diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index b8a33abce650..91395e1552ae 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -5542,7 +5542,7 @@ preempt: } static struct task_struct * -pick_next_task_fair(struct rq *rq, struct task_struct *prev) +pick_next_task_fair(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct cfs_rq *cfs_rq = &rq->cfs; struct sched_entity *se; @@ -5655,9 +5655,9 @@ idle: * further scheduler activity on it and we're being very careful to * re-start the picking loop. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); new_tasks = idle_balance(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); /* * Because idle_balance() releases (and re-acquires) rq->lock, it is * possible for any higher priority task to appear. In that case we diff --git a/kernel/sched/idle_task.c b/kernel/sched/idle_task.c index 47ce94931f1b..2ce5458bbe1d 100644 --- a/kernel/sched/idle_task.c +++ b/kernel/sched/idle_task.c @@ -24,7 +24,7 @@ static void check_preempt_curr_idle(struct rq *rq, struct task_struct *p, int fl } static struct task_struct * -pick_next_task_idle(struct rq *rq, struct task_struct *prev) +pick_next_task_idle(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { put_prev_task(rq, prev); diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index 19e13060fcd5..68deaf901a12 100644 --- a/kernel/sched/rt.c +++ b/kernel/sched/rt.c @@ -1524,7 +1524,7 @@ static struct task_struct *_pick_next_task_rt(struct rq *rq) } static struct task_struct * -pick_next_task_rt(struct rq *rq, struct task_struct *prev) +pick_next_task_rt(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct task_struct *p; struct rt_rq *rt_rq = &rq->rt; @@ -1536,9 +1536,9 @@ pick_next_task_rt(struct rq *rq, struct task_struct *prev) * disabled avoiding further scheduler activity on it and we're * being very careful to re-start the picking loop. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); pull_rt_task(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); /* * pull_rt_task() can drop (and re-acquire) rq->lock; this * means a dl or stop task can slip in, in which case we need diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index a5eecb1e5e4b..0b6a838e9e73 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1202,7 +1202,8 @@ struct sched_class { * tasks. */ struct task_struct * (*pick_next_task) (struct rq *rq, - struct task_struct *prev); + struct task_struct *prev, + struct pin_cookie cookie); void (*put_prev_task) (struct rq *rq, struct task_struct *p); #ifdef CONFIG_SMP @@ -1453,6 +1454,7 @@ static inline void sched_avg_update(struct rq *rq) { } struct rq_flags { unsigned long flags; + struct pin_cookie cookie; }; struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf) @@ -1464,7 +1466,7 @@ struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf) static inline void __task_rq_unlock(struct rq *rq, struct rq_flags *rf) __releases(rq->lock) { - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf->cookie); raw_spin_unlock(&rq->lock); } @@ -1473,7 +1475,7 @@ task_rq_unlock(struct rq *rq, struct task_struct *p, struct rq_flags *rf) __releases(rq->lock) __releases(p->pi_lock) { - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf->cookie); raw_spin_unlock(&rq->lock); raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags); } diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c index cbc67da10954..604297a08b3a 100644 --- a/kernel/sched/stop_task.c +++ b/kernel/sched/stop_task.c @@ -24,7 +24,7 @@ check_preempt_curr_stop(struct rq *rq, struct task_struct *p, int flags) } static struct task_struct * -pick_next_task_stop(struct rq *rq, struct task_struct *prev) +pick_next_task_stop(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct task_struct *stop = rq->stop; -- cgit v1.2.3 From 2159197d66770ec01f75c93fb11dc66df81fd45b Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 28 Apr 2016 12:49:38 +0200 Subject: sched/core: Enable increased load resolution on 64-bit kernels Mike ran into the low load resolution limitation on his big machine. So reenable these bits; nobody could ever reproduce/analyze the reported power usage claim and Google has been running with this for years as well. Reported-by: Mike Galbraith Tested-by: Mike Galbraith Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- kernel/sched/sched.h | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 0b6a838e9e73..066a4c2d2695 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -49,11 +49,13 @@ static inline void cpu_load_update_active(struct rq *this_rq) { } * and does not change the user-interface for setting shares/weights. * * We increase resolution only if we have enough bits to allow this increased - * resolution (i.e. BITS_PER_LONG > 32). The costs for increasing resolution - * when BITS_PER_LONG <= 32 are pretty high and the returns do not justify the - * increased costs. + * resolution (i.e. 64bit). The costs for increasing resolution when 32bit are + * pretty high and the returns do not justify the increased costs. + * + * Really only required when CONFIG_FAIR_GROUP_SCHED is also set, but to + * increase coverage and consistency always enable it on 64bit platforms. */ -#if 0 /* BITS_PER_LONG > 32 -- currently broken: it increases power usage under light load */ +#ifdef CONFIG_64BIT # define SCHED_LOAD_RESOLUTION 10 # define scale_load(w) ((w) << SCHED_LOAD_RESOLUTION) # define scale_load_down(w) ((w) >> SCHED_LOAD_RESOLUTION) -- cgit v1.2.3 From 6ecdd74962f246dfe8750b7bea481a1c0816315d Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Tue, 5 Apr 2016 12:12:26 +0800 Subject: sched/fair: Generalize the load/util averages resolution definition Integer metric needs fixed point arithmetic. In sched/fair, a few metrics, e.g., weight, load, load_avg, util_avg, freq, and capacity, may have different fixed point ranges, which makes their update and usage error-prone. In order to avoid the errors relating to the fixed point range, we definie a basic fixed point range, and then formalize all metrics to base on the basic range. The basic range is 1024 or (1 << 10). Further, one can recursively apply the basic range to have larger range. Pointed out by Ben Segall, weight (visible to user, e.g., NICE-0 has 1024) and load (e.g., NICE_0_LOAD) have independent ranges, but they must be well calibrated. Signed-off-by: Yuyang Du Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: bsegall@google.com Cc: dietmar.eggemann@arm.com Cc: lizefan@huawei.com Cc: morten.rasmussen@arm.com Cc: pjt@google.com Cc: umgwanakikbuti@gmail.com Cc: vincent.guittot@linaro.org Link: http://lkml.kernel.org/r/1459829551-21625-2-git-send-email-yuyang.du@intel.com Signed-off-by: Ingo Molnar --- include/linux/sched.h | 16 +++++++++++++--- kernel/sched/fair.c | 4 ---- kernel/sched/sched.h | 15 ++++++++++----- 3 files changed, 23 insertions(+), 12 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index d894f2d61388..7d779d70a3a5 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -936,10 +936,20 @@ enum cpu_idle_type { CPU_MAX_IDLE_TYPES }; +/* + * Integer metrics need fixed point arithmetic, e.g., sched/fair + * has a few: load, load_avg, util_avg, freq, and capacity. + * + * We define a basic fixed point arithmetic range, and then formalize + * all these metrics based on that basic range. + */ +# define SCHED_FIXEDPOINT_SHIFT 10 +# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) + /* * Increase resolution of cpu_capacity calculations */ -#define SCHED_CAPACITY_SHIFT 10 +#define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) /* @@ -1205,8 +1215,8 @@ struct load_weight { * 1) load_avg factors frequency scaling into the amount of time that a * sched_entity is runnable on a rq into its weight. For cfs_rq, it is the * aggregated such weights of all runnable and blocked sched_entities. - * 2) util_avg factors frequency and cpu scaling into the amount of time - * that a sched_entity is running on a CPU, in the range [0..SCHED_LOAD_SCALE]. + * 2) util_avg factors frequency and cpu capacity scaling into the amount of time + * that a sched_entity is running on a CPU, in the range [0..SCHED_CAPACITY_SCALE]. * For cfs_rq, it is the aggregated such times of all runnable and * blocked sched_entities. * The 64 bit load_sum can: diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 91395e1552ae..76ca86e9fc20 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2662,10 +2662,6 @@ static u32 __compute_runnable_contrib(u64 n) return contrib + runnable_avg_yN_sum[n]; } -#if (SCHED_LOAD_SHIFT - SCHED_LOAD_RESOLUTION) != 10 || SCHED_CAPACITY_SHIFT != 10 -#error "load tracking assumes 2^10 as unit" -#endif - #define cap_scale(v, s) ((v)*(s) >> SCHED_CAPACITY_SHIFT) /* diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 066a4c2d2695..ad83361f9e67 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -56,18 +56,23 @@ static inline void cpu_load_update_active(struct rq *this_rq) { } * increase coverage and consistency always enable it on 64bit platforms. */ #ifdef CONFIG_64BIT -# define SCHED_LOAD_RESOLUTION 10 -# define scale_load(w) ((w) << SCHED_LOAD_RESOLUTION) -# define scale_load_down(w) ((w) >> SCHED_LOAD_RESOLUTION) +# define SCHED_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT + SCHED_FIXEDPOINT_SHIFT) +# define scale_load(w) ((w) << SCHED_FIXEDPOINT_SHIFT) +# define scale_load_down(w) ((w) >> SCHED_FIXEDPOINT_SHIFT) #else -# define SCHED_LOAD_RESOLUTION 0 +# define SCHED_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT) # define scale_load(w) (w) # define scale_load_down(w) (w) #endif -#define SCHED_LOAD_SHIFT (10 + SCHED_LOAD_RESOLUTION) #define SCHED_LOAD_SCALE (1L << SCHED_LOAD_SHIFT) +/* + * NICE_0's weight (visible to users) and its load (invisible to users) have + * independent ranges, but they should be well calibrated. We use scale_load() + * and scale_load_down(w) to convert between them, and the following must be true: + * scale_load(sched_prio_to_weight[20]) == NICE_0_LOAD + */ #define NICE_0_LOAD SCHED_LOAD_SCALE #define NICE_0_SHIFT SCHED_LOAD_SHIFT -- cgit v1.2.3 From 172895e6b5216eba3e0880460829a8baeefd55f3 Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Tue, 5 Apr 2016 12:12:27 +0800 Subject: sched/fair: Rename SCHED_LOAD_SHIFT to NICE_0_LOAD_SHIFT and remove SCHED_LOAD_SCALE After cleaning up the sched metrics, there are two definitions that are ambiguous and confusing: SCHED_LOAD_SHIFT and SCHED_LOAD_SHIFT. Resolve this: - Rename SCHED_LOAD_SHIFT to NICE_0_LOAD_SHIFT, which better reflects what it is. - Replace SCHED_LOAD_SCALE use with SCHED_CAPACITY_SCALE and remove SCHED_LOAD_SCALE. Suggested-by: Ben Segall Signed-off-by: Yuyang Du Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: dietmar.eggemann@arm.com Cc: lizefan@huawei.com Cc: morten.rasmussen@arm.com Cc: pjt@google.com Cc: umgwanakikbuti@gmail.com Cc: vincent.guittot@linaro.org Link: http://lkml.kernel.org/r/1459829551-21625-3-git-send-email-yuyang.du@intel.com [ Rewrote the changelog and fixed the build on 32-bit kernels. ] Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 4 ++-- kernel/sched/sched.h | 22 +++++++++++----------- 2 files changed, 13 insertions(+), 13 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 76ca86e9fc20..e1485710d1ec 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -719,7 +719,7 @@ void post_init_entity_util_avg(struct sched_entity *se) { struct cfs_rq *cfs_rq = cfs_rq_of(se); struct sched_avg *sa = &se->avg; - long cap = (long)(scale_load_down(SCHED_LOAD_SCALE) - cfs_rq->avg.util_avg) / 2; + long cap = (long)(SCHED_CAPACITY_SCALE - cfs_rq->avg.util_avg) / 2; if (cap > 0) { if (cfs_rq->avg.util_avg != 0) { @@ -7010,7 +7010,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s if (busiest->group_type == group_overloaded && local->group_type == group_overloaded) { load_above_capacity = busiest->sum_nr_running * - SCHED_LOAD_SCALE; + scale_load_down(NICE_0_LOAD); if (load_above_capacity > busiest->group_capacity) load_above_capacity -= busiest->group_capacity; else diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index ad83361f9e67..d24e91b0a722 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -56,25 +56,25 @@ static inline void cpu_load_update_active(struct rq *this_rq) { } * increase coverage and consistency always enable it on 64bit platforms. */ #ifdef CONFIG_64BIT -# define SCHED_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT + SCHED_FIXEDPOINT_SHIFT) +# define NICE_0_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT + SCHED_FIXEDPOINT_SHIFT) # define scale_load(w) ((w) << SCHED_FIXEDPOINT_SHIFT) # define scale_load_down(w) ((w) >> SCHED_FIXEDPOINT_SHIFT) #else -# define SCHED_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT) +# define NICE_0_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT) # define scale_load(w) (w) # define scale_load_down(w) (w) #endif -#define SCHED_LOAD_SCALE (1L << SCHED_LOAD_SHIFT) - /* - * NICE_0's weight (visible to users) and its load (invisible to users) have - * independent ranges, but they should be well calibrated. We use scale_load() - * and scale_load_down(w) to convert between them, and the following must be true: - * scale_load(sched_prio_to_weight[20]) == NICE_0_LOAD + * Task weight (visible to users) and its load (invisible to users) have + * independent resolution, but they should be well calibrated. We use + * scale_load() and scale_load_down(w) to convert between them. The + * following must be true: + * + * scale_load(sched_prio_to_weight[USER_PRIO(NICE_TO_PRIO(0))]) == NICE_0_LOAD + * */ -#define NICE_0_LOAD SCHED_LOAD_SCALE -#define NICE_0_SHIFT SCHED_LOAD_SHIFT +#define NICE_0_LOAD (1L << NICE_0_LOAD_SHIFT) /* * Single value that decides SCHED_DEADLINE internal math precision. @@ -863,7 +863,7 @@ DECLARE_PER_CPU(struct sched_domain *, sd_asym); struct sched_group_capacity { atomic_t ref; /* - * CPU capacity of this group, SCHED_LOAD_SCALE being max capacity + * CPU capacity of this group, SCHED_CAPACITY_SCALE being max capacity * for a single CPU. */ unsigned int capacity; -- cgit v1.2.3 From 7b5953345efe4f226bb52cbea04558d16ec7ebfa Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Tue, 5 Apr 2016 12:12:28 +0800 Subject: sched/fair: Add detailed description to the sched load avg metrics These sched metrics have become complex enough, so describe them in detail at their definition. Signed-off-by: Yuyang Du Signed-off-by: Peter Zijlstra (Intel) [ Fixed the text to improve its spelling and typography. ] Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: bsegall@google.com Cc: dietmar.eggemann@arm.com Cc: lizefan@huawei.com Cc: morten.rasmussen@arm.com Cc: pjt@google.com Cc: umgwanakikbuti@gmail.com Cc: vincent.guittot@linaro.org Link: http://lkml.kernel.org/r/1459829551-21625-4-git-send-email-yuyang.du@intel.com Signed-off-by: Ingo Molnar --- include/linux/sched.h | 60 +++++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 49 insertions(+), 11 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index 7d779d70a3a5..57faf789c88f 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1211,18 +1211,56 @@ struct load_weight { }; /* - * The load_avg/util_avg accumulates an infinite geometric series. - * 1) load_avg factors frequency scaling into the amount of time that a - * sched_entity is runnable on a rq into its weight. For cfs_rq, it is the - * aggregated such weights of all runnable and blocked sched_entities. - * 2) util_avg factors frequency and cpu capacity scaling into the amount of time - * that a sched_entity is running on a CPU, in the range [0..SCHED_CAPACITY_SCALE]. - * For cfs_rq, it is the aggregated such times of all runnable and + * The load_avg/util_avg accumulates an infinite geometric series + * (see __update_load_avg() in kernel/sched/fair.c). + * + * [load_avg definition] + * + * load_avg = runnable% * scale_load_down(load) + * + * where runnable% is the time ratio that a sched_entity is runnable. + * For cfs_rq, it is the aggregated load_avg of all runnable and * blocked sched_entities. - * The 64 bit load_sum can: - * 1) for cfs_rq, afford 4353082796 (=2^64/47742/88761) entities with - * the highest weight (=88761) always runnable, we should not overflow - * 2) for entity, support any load.weight always runnable + * + * load_avg may also take frequency scaling into account: + * + * load_avg = runnable% * scale_load_down(load) * freq% + * + * where freq% is the CPU frequency normalized to the highest frequency. + * + * [util_avg definition] + * + * util_avg = running% * SCHED_CAPACITY_SCALE + * + * where running% is the time ratio that a sched_entity is running on + * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable + * and blocked sched_entities. + * + * util_avg may also factor frequency scaling and CPU capacity scaling: + * + * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% + * + * where freq% is the same as above, and capacity% is the CPU capacity + * normalized to the greatest capacity (due to uarch differences, etc). + * + * N.B., the above ratios (runnable%, running%, freq%, and capacity%) + * themselves are in the range of [0, 1]. To do fixed point arithmetics, + * we therefore scale them to as large a range as necessary. This is for + * example reflected by util_avg's SCHED_CAPACITY_SCALE. + * + * [Overflow issue] + * + * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities + * with the highest load (=88761), always runnable on a single cfs_rq, + * and should not overflow as the number already hits PID_MAX_LIMIT. + * + * For all other cases (including 32-bit kernels), struct load_weight's + * weight will overflow first before we do, because: + * + * Max(load_avg) <= Max(load.weight) + * + * Then it is the load_weight's responsibility to consider overflow + * issues. */ struct sched_avg { u64 last_update_time, load_sum; -- cgit v1.2.3 From 7b20b916e953cabef569541f991a0a583bc344cb Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Tue, 3 May 2016 05:54:27 +0800 Subject: sched/fair: Optimize sum computation with a lookup table __compute_runnable_contrib() uses a loop to compute sum, whereas a table lookup can do it faster in a constant amount of time. The program to generate the constants is located at: Documentation/scheduler/sched-avg.txt Signed-off-by: Yuyang Du Signed-off-by: Peter Zijlstra (Intel) Reviewed-by: Morten Rasmussen Acked-by: Vincent Guittot Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: bsegall@google.com Cc: dietmar.eggemann@arm.com Cc: juri.lelli@arm.com Cc: pjt@google.com Link: http://lkml.kernel.org/r/1462226078-31904-2-git-send-email-yuyang.du@intel.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index e1485710d1ec..8c381a6d56e3 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2602,6 +2602,16 @@ static const u32 runnable_avg_yN_sum[] = { 17718,18340,18949,19545,20128,20698,21256,21802,22336,22859,23371, }; +/* + * Precomputed \Sum y^k { 1<=k<=n, where n%32=0). Values are rolled down to + * lower integers. See Documentation/scheduler/sched-avg.txt how these + * were generated: + */ +static const u32 __accumulated_sum_N32[] = { + 0, 23371, 35056, 40899, 43820, 45281, + 46011, 46376, 46559, 46650, 46696, 46719, +}; + /* * Approximate: * val * y^n, where y^32 ~= 0.5 (~1 scheduling period) @@ -2650,14 +2660,9 @@ static u32 __compute_runnable_contrib(u64 n) else if (unlikely(n >= LOAD_AVG_MAX_N)) return LOAD_AVG_MAX; - /* Compute \Sum k^n combining precomputed values for k^i, \Sum k^j */ - do { - contrib /= 2; /* y^LOAD_AVG_PERIOD = 1/2 */ - contrib += runnable_avg_yN_sum[LOAD_AVG_PERIOD]; - - n -= LOAD_AVG_PERIOD; - } while (n > LOAD_AVG_PERIOD); - + /* Since n < LOAD_AVG_MAX_N, n/LOAD_AVG_PERIOD < 11 */ + contrib = __accumulated_sum_N32[n/LOAD_AVG_PERIOD]; + n %= LOAD_AVG_PERIOD; contrib = decay_load(contrib, n); return contrib + runnable_avg_yN_sum[n]; } -- cgit v1.2.3 From db6ea2fb094fb3a6afc36d3e4229bc162638ad24 Mon Sep 17 00:00:00 2001 From: Wanpeng Li Date: Tue, 3 May 2016 12:38:25 +0800 Subject: sched/debug: Print out idle balance values even on !CONFIG_SCHEDSTATS kernels The max_idle_balance_cost and avg_idle values which are tracked and ar used to capture short idle incidents, are not associated with schedstats, however the information of these two values isn't printed out on !CONFIG_SCHEDSTATS kernels. Fix this by moving the value printout out of the CONFIG_SCHEDSTATS section. Signed-off-by: Wanpeng Li Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1462250305-4523-1-git-send-email-wanpeng.li@hotmail.com Signed-off-by: Ingo Molnar --- kernel/sched/debug.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c index 4fbc3bd5ff60..cf905f655ba1 100644 --- a/kernel/sched/debug.c +++ b/kernel/sched/debug.c @@ -626,15 +626,16 @@ do { \ #undef P #undef PN -#ifdef CONFIG_SCHEDSTATS -#define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n); -#define P64(n) SEQ_printf(m, " .%-30s: %Ld\n", #n, rq->n); - #ifdef CONFIG_SMP +#define P64(n) SEQ_printf(m, " .%-30s: %Ld\n", #n, rq->n); P64(avg_idle); P64(max_idle_balance_cost); +#undef P64 #endif +#ifdef CONFIG_SCHEDSTATS +#define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n); + if (schedstat_enabled()) { P(yld_count); P(sched_count); @@ -644,7 +645,6 @@ do { \ } #undef P -#undef P64 #endif spin_lock_irqsave(&sched_debug_lock, flags); print_cfs_stats(m, cpu); -- cgit v1.2.3 From b52fad2db5d792d89975cebf2fe1646a7af28ed0 Mon Sep 17 00:00:00 2001 From: Matt Fleming Date: Tue, 3 May 2016 20:46:54 +0100 Subject: sched/fair: Update rq clock before updating nohz CPU load If we're accessing rq_clock() (e.g. in sched_avg_update()) we should update the rq clock before calling cpu_load_update(), otherwise any time calculations will be stale. All other paths currently call update_rq_clock(). Signed-off-by: Matt Fleming Signed-off-by: Peter Zijlstra (Intel) Reviewed-by: Wanpeng Li Cc: Frederic Weisbecker Cc: Linus Torvalds Cc: Mel Gorman Cc: Mike Galbraith Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1462304814-11715-1-git-send-email-matt@codeblueprint.co.uk Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 1 + 1 file changed, 1 insertion(+) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 8c381a6d56e3..7a00c7c2dad0 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4724,6 +4724,7 @@ void cpu_load_update_nohz_stop(void) load = weighted_cpuload(cpu_of(this_rq)); raw_spin_lock(&this_rq->lock); + update_rq_clock(this_rq); cpu_load_update_nohz(this_rq, curr_jiffies, load); raw_spin_unlock(&this_rq->lock); } -- cgit v1.2.3 From 0a9b23ce46cd5d3a360fbefca8ffce441c55046e Mon Sep 17 00:00:00 2001 From: Dietmar Eggemann Date: Fri, 29 Apr 2016 20:32:38 +0100 Subject: sched/fair: Remove stale power aware scheduling comments Commit 8e7fbcbc22c1 ("sched: Remove stale power aware scheduling remnants and dysfunctional knobs") deleted the power aware scheduling support. This patch gets rid of the remaining power aware scheduling related comments in the code as well. Signed-off-by: Dietmar Eggemann Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1461958364-675-2-git-send-email-dietmar.eggemann@arm.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 13 +++---------- 1 file changed, 3 insertions(+), 10 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 7a00c7c2dad0..537d71ec18ab 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7027,9 +7027,8 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s * We're trying to get all the cpus to the average_load, so we don't * want to push ourselves above the average load, nor do we wish to * reduce the max loaded cpu below the average load. At the same time, - * we also don't want to reduce the group load below the group capacity - * (so that we can implement power-savings policies etc). Thus we look - * for the minimum possible imbalance. + * we also don't want to reduce the group load below the group + * capacity. Thus we look for the minimum possible imbalance. */ max_pull = min(busiest->avg_load - sds->avg_load, load_above_capacity); @@ -7053,10 +7052,7 @@ 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 - * if there is an imbalance. If there isn't an imbalance, and - * the user has opted for power-savings, it returns a group whose - * CPUs can be put to idle by rebalancing those tasks elsewhere, if - * such a group exists. + * if there is an imbalance. * * Also calculates the amount of weighted load which should be moved * to restore balance. @@ -7064,9 +7060,6 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s * @env: The load balancing environment. * * Return: - The busiest group if imbalance exists. - * - If no imbalance and user has opted for power-savings balance, - * return the least loaded group whose CPUs can be - * put to idle by rebalancing its tasks onto our group. */ static struct sched_group *find_busiest_group(struct lb_env *env) { -- cgit v1.2.3 From 885e542ce827f5f102fe9628d63c6430c8b7ab2c Mon Sep 17 00:00:00 2001 From: Dietmar Eggemann Date: Fri, 29 Apr 2016 20:32:39 +0100 Subject: sched/fair: Fix comment in calculate_imbalance() The comment in calculate_imbalance() was introduced in commit: 2dd73a4f09be ("[PATCH] sched: implement smpnice") which described the logic as it was then, but a later commit: b18855500fc4 ("sched/balancing: Fix 'local->avg_load > sds->avg_load' case in calculate_imbalance()") .. complicated this logic some more so that the comment does not match anymore. Update the comment to match the code. Signed-off-by: Dietmar Eggemann Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1461958364-675-3-git-send-email-dietmar.eggemann@arm.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 537d71ec18ab..51f7a4b62985 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7000,9 +7000,10 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s } /* - * In the presence of smp nice balancing, certain scenarios can have - * max load less than avg load(as we skip the groups at or below - * its cpu_capacity, while calculating max_load..) + * Avg load of busiest sg can be less and avg load of local sg can + * be greater than avg load across all sgs of sd because avg load + * factors in sg capacity and sgs with smaller group_type are + * skipped when updating the busiest sg: */ if (busiest->avg_load <= sds->avg_load || local->avg_load >= sds->avg_load) { -- cgit v1.2.3 From e9d867a67fd03ccc07248ca4e9c2f74fed494d5b Mon Sep 17 00:00:00 2001 From: Peter Zijlstra (Intel) Date: Thu, 10 Mar 2016 12:54:08 +0100 Subject: sched: Allow per-cpu kernel threads to run on online && !active In order to enable symmetric hotplug, we must mirror the online && !active state of cpu-down on the cpu-up side. However, to retain sanity, limit this state to per-cpu kthreads. Aside from the change to set_cpus_allowed_ptr(), which allow moving the per-cpu kthreads on, the other critical piece is the cpu selection for pinned tasks in select_task_rq(). This avoids dropping into select_fallback_rq(). select_fallback_rq() cannot be allowed to select !active cpus because its used to migrate user tasks away. And we do not want to move user tasks onto cpus that are in transition. Requested-by: Thomas Gleixner Signed-off-by: Peter Zijlstra (Intel) Tested-by: Thomas Gleixner Cc: Lai Jiangshan Cc: Jan H. Schönherr Cc: Oleg Nesterov Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160301152303.GV6356@twins.programming.kicks-ass.net Signed-off-by: Thomas Gleixner --- arch/powerpc/kernel/smp.c | 2 +- arch/s390/kernel/smp.c | 2 +- include/linux/cpumask.h | 6 ++---- kernel/sched/core.c | 49 ++++++++++++++++++++++++++++++++++++++++------- 4 files changed, 46 insertions(+), 13 deletions(-) diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c index 8cac1eb41466..55c924b65f71 100644 --- a/arch/powerpc/kernel/smp.c +++ b/arch/powerpc/kernel/smp.c @@ -565,7 +565,7 @@ int __cpu_up(unsigned int cpu, struct task_struct *tidle) smp_ops->give_timebase(); /* Wait until cpu puts itself in the online & active maps */ - while (!cpu_online(cpu) || !cpu_active(cpu)) + while (!cpu_online(cpu)) cpu_relax(); return 0; diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index 40a6b4f9c36c..7b89a7572100 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -832,7 +832,7 @@ int __cpu_up(unsigned int cpu, struct task_struct *tidle) pcpu_attach_task(pcpu, tidle); pcpu_start_fn(pcpu, smp_start_secondary, NULL); /* Wait until cpu puts itself in the online & active maps */ - while (!cpu_online(cpu) || !cpu_active(cpu)) + while (!cpu_online(cpu)) cpu_relax(); return 0; } diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h index 40cee6b77a93..e828cf65d7df 100644 --- a/include/linux/cpumask.h +++ b/include/linux/cpumask.h @@ -743,12 +743,10 @@ set_cpu_present(unsigned int cpu, bool present) static inline void set_cpu_online(unsigned int cpu, bool online) { - if (online) { + if (online) cpumask_set_cpu(cpu, &__cpu_online_mask); - cpumask_set_cpu(cpu, &__cpu_active_mask); - } else { + else cpumask_clear_cpu(cpu, &__cpu_online_mask); - } } static inline void diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8b489fcac37b..8bfd7d4f1c21 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1082,13 +1082,21 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) static int __set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask, bool check) { + const struct cpumask *cpu_valid_mask = cpu_active_mask; + unsigned int dest_cpu; unsigned long flags; struct rq *rq; - unsigned int dest_cpu; int ret = 0; rq = task_rq_lock(p, &flags); + if (p->flags & PF_KTHREAD) { + /* + * Kernel threads are allowed on online && !active CPUs + */ + cpu_valid_mask = cpu_online_mask; + } + /* * Must re-check here, to close a race against __kthread_bind(), * sched_setaffinity() is not guaranteed to observe the flag. @@ -1101,18 +1109,28 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, if (cpumask_equal(&p->cpus_allowed, new_mask)) goto out; - if (!cpumask_intersects(new_mask, cpu_active_mask)) { + if (!cpumask_intersects(new_mask, cpu_valid_mask)) { ret = -EINVAL; goto out; } do_set_cpus_allowed(p, new_mask); + if (p->flags & PF_KTHREAD) { + /* + * For kernel threads that do indeed end up on online && + * !active we want to ensure they are strict per-cpu threads. + */ + WARN_ON(cpumask_intersects(new_mask, cpu_online_mask) && + !cpumask_intersects(new_mask, cpu_active_mask) && + p->nr_cpus_allowed != 1); + } + /* Can the task run on the task's current CPU? If so, we're done */ if (cpumask_test_cpu(task_cpu(p), new_mask)) goto out; - dest_cpu = cpumask_any_and(cpu_active_mask, new_mask); + dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask); if (task_running(rq, p) || p->state == TASK_WAKING) { struct migration_arg arg = { p, dest_cpu }; /* Need help from migration thread: drop lock and wait. */ @@ -1431,6 +1449,25 @@ EXPORT_SYMBOL_GPL(kick_process); /* * ->cpus_allowed is protected by both rq->lock and p->pi_lock + * + * A few notes on cpu_active vs cpu_online: + * + * - cpu_active must be a subset of cpu_online + * + * - on cpu-up we allow per-cpu kthreads on the online && !active cpu, + * see __set_cpus_allowed_ptr(). At this point the newly online + * cpu isn't yet part of the sched domains, and balancing will not + * see it. + * + * - on cpu-down we clear cpu_active() to mask the sched domains and + * avoid the load balancer to place new tasks on the to be removed + * cpu. Existing tasks will remain running there and will be taken + * off. + * + * This means that fallback selection must not select !active CPUs. + * And can assume that any active CPU must be online. Conversely + * select_task_rq() below may allow selection of !active CPUs in order + * to satisfy the above rules. */ static int select_fallback_rq(int cpu, struct task_struct *p) { @@ -1449,8 +1486,6 @@ static int select_fallback_rq(int cpu, struct task_struct *p) /* Look for allowed, online CPU in same node. */ for_each_cpu(dest_cpu, nodemask) { - if (!cpu_online(dest_cpu)) - continue; if (!cpu_active(dest_cpu)) continue; if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p))) @@ -1461,8 +1496,6 @@ static int select_fallback_rq(int cpu, struct task_struct *p) for (;;) { /* Any allowed, online CPU? */ for_each_cpu(dest_cpu, tsk_cpus_allowed(p)) { - if (!cpu_online(dest_cpu)) - continue; if (!cpu_active(dest_cpu)) continue; goto out; @@ -1514,6 +1547,8 @@ int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags) if (p->nr_cpus_allowed > 1) cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags); + else + cpu = cpumask_any(tsk_cpus_allowed(p)); /* * In order not to call set_task_cpu() on a blocking task we need -- cgit v1.2.3 From 9cf7243d5d83d27aca47f842107bfa02b5f11d16 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:09 +0100 Subject: sched: Make set_cpu_rq_start_time() a built in hotplug state Start distangling the maze of hotplug notifiers in the scheduler. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Signed-off-by: Thomas Gleixner --- include/linux/cpuhotplug.h | 1 + include/linux/sched.h | 1 + kernel/cpu.c | 6 ++++++ kernel/sched/core.c | 16 +++++++++------- 4 files changed, 17 insertions(+), 7 deletions(-) diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h index 5d68e15e46b7..99fd1d2f76fe 100644 --- a/include/linux/cpuhotplug.h +++ b/include/linux/cpuhotplug.h @@ -8,6 +8,7 @@ enum cpuhp_state { CPUHP_BRINGUP_CPU, CPUHP_AP_IDLE_DEAD, CPUHP_AP_OFFLINE, + CPUHP_AP_SCHED_STARTING, CPUHP_AP_NOTIFY_STARTING, CPUHP_AP_ONLINE, CPUHP_TEARDOWN_CPU, diff --git a/include/linux/sched.h b/include/linux/sched.h index 52c4847b05e2..39597d0a005e 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -372,6 +372,7 @@ extern void cpu_init (void); extern void trap_init(void); extern void update_process_times(int user); extern void scheduler_tick(void); +extern int sched_cpu_starting(unsigned int cpu); extern void sched_show_task(struct task_struct *p); diff --git a/kernel/cpu.c b/kernel/cpu.c index 3e3f6e49eabb..f46d02b966bf 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -1236,6 +1236,12 @@ static struct cpuhp_step cpuhp_ap_states[] = { .name = "ap:offline", .cant_stop = true, }, + /* First state is scheduler control. Interrupts are disabled */ + [CPUHP_AP_SCHED_STARTING] = { + .name = "sched:starting", + .startup = sched_cpu_starting, + .teardown = NULL, + }, /* * Low level startup/teardown notifiers. Run with interrupts * disabled. Will be removed once the notifiers are converted to diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8bfd7d4f1c21..4df9aaae27a2 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5473,10 +5473,10 @@ static struct notifier_block migration_notifier = { .priority = CPU_PRI_MIGRATION, }; -static void set_cpu_rq_start_time(void) +static void set_cpu_rq_start_time(unsigned int cpu) { - int cpu = smp_processor_id(); struct rq *rq = cpu_rq(cpu); + rq->age_stamp = sched_clock_cpu(cpu); } @@ -5486,10 +5486,6 @@ static int sched_cpu_active(struct notifier_block *nfb, int cpu = (long)hcpu; switch (action & ~CPU_TASKS_FROZEN) { - case CPU_STARTING: - set_cpu_rq_start_time(); - return NOTIFY_OK; - case CPU_DOWN_FAILED: set_cpu_active(cpu, true); return NOTIFY_OK; @@ -5511,6 +5507,12 @@ static int sched_cpu_inactive(struct notifier_block *nfb, } } +int sched_cpu_starting(unsigned int cpu) +{ + set_cpu_rq_start_time(cpu); + return 0; +} + static int __init migration_init(void) { void *cpu = (void *)(long)smp_processor_id(); @@ -7426,7 +7428,7 @@ void __init sched_init(void) if (cpu_isolated_map == NULL) zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT); idle_thread_set_boot_cpu(); - set_cpu_rq_start_time(); + set_cpu_rq_start_time(smp_processor_id()); #endif init_sched_fair_class(); -- cgit v1.2.3 From e26fbffd32c28107d9d268b432706ccf84fb6411 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:10 +0100 Subject: sched: Allow hotplug notifiers to be setup early Prevent the SMP scheduler related notifiers to be executed before the smp scheduler is initialized and install them early. This is a preparatory change for further consolidation of the hotplug notifier maze. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 59 ++++++++++++++++++++++++++++++++--------------------- 1 file changed, 36 insertions(+), 23 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 4df9aaae27a2..328502c9af00 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5195,6 +5195,8 @@ out: #ifdef CONFIG_SMP +static bool sched_smp_initialized __read_mostly; + #ifdef CONFIG_NUMA_BALANCING /* Migrate current task p to target_cpu */ int migrate_task_to(struct task_struct *p, int target_cpu) @@ -5513,25 +5515,6 @@ int sched_cpu_starting(unsigned int cpu) return 0; } -static int __init migration_init(void) -{ - void *cpu = (void *)(long)smp_processor_id(); - int err; - - /* Initialize migration for the boot CPU */ - err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu); - BUG_ON(err == NOTIFY_BAD); - migration_call(&migration_notifier, CPU_ONLINE, cpu); - register_cpu_notifier(&migration_notifier); - - /* Register cpu active notifiers */ - cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE); - cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE); - - return 0; -} -early_initcall(migration_init); - static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */ #ifdef CONFIG_SCHED_DEBUG @@ -6711,6 +6694,9 @@ static int sched_domains_numa_masks_update(struct notifier_block *nfb, { int cpu = (long)hcpu; + if (!sched_smp_initialized) + return NOTIFY_DONE; + switch (action & ~CPU_TASKS_FROZEN) { case CPU_ONLINE: sched_domains_numa_masks_set(cpu); @@ -7129,6 +7115,9 @@ static int num_cpus_frozen; /* used to mark begin/end of suspend/resume */ static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action, void *hcpu) { + if (!sched_smp_initialized) + return NOTIFY_DONE; + switch (action) { case CPU_ONLINE_FROZEN: case CPU_DOWN_FAILED_FROZEN: @@ -7169,6 +7158,9 @@ static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action, bool overflow; int cpus; + if (!sched_smp_initialized) + return NOTIFY_DONE; + switch (action) { case CPU_DOWN_PREPARE: rcu_read_lock_sched(); @@ -7216,10 +7208,6 @@ void __init sched_init_smp(void) cpumask_set_cpu(smp_processor_id(), non_isolated_cpus); mutex_unlock(&sched_domains_mutex); - hotcpu_notifier(sched_domains_numa_masks_update, CPU_PRI_SCHED_ACTIVE); - hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE); - hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE); - init_hrtick(); /* Move init over to a non-isolated CPU */ @@ -7230,7 +7218,32 @@ void __init sched_init_smp(void) init_sched_rt_class(); init_sched_dl_class(); + sched_smp_initialized = true; } + +static int __init migration_init(void) +{ + void *cpu = (void *)(long)smp_processor_id(); + int err; + + /* Initialize migration for the boot CPU */ + err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu); + BUG_ON(err == NOTIFY_BAD); + migration_call(&migration_notifier, CPU_ONLINE, cpu); + register_cpu_notifier(&migration_notifier); + + /* Register cpu active notifiers */ + cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE); + cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE); + + hotcpu_notifier(sched_domains_numa_masks_update, CPU_PRI_SCHED_ACTIVE); + hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE); + hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE); + + return 0; +} +early_initcall(migration_init); + #else void __init sched_init_smp(void) { -- cgit v1.2.3 From 135fb3e19773e66f56b60e3b9fdda6166e77c55d Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:11 +0100 Subject: sched: Consolidate the notifier maze We can maintain the ordering of the scheduler cpu hotplug functionality nicely in one notifer. Get rid of the maze. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Signed-off-by: Thomas Gleixner --- include/linux/cpu.h | 12 ++-- kernel/sched/core.c | 174 +++++++++++++++++++++------------------------------- 2 files changed, 73 insertions(+), 113 deletions(-) diff --git a/include/linux/cpu.h b/include/linux/cpu.h index f9b1fab4388a..17017051bfb1 100644 --- a/include/linux/cpu.h +++ b/include/linux/cpu.h @@ -61,19 +61,15 @@ struct notifier_block; enum { /* * SCHED_ACTIVE marks a cpu which is coming up active during - * CPU_ONLINE and CPU_DOWN_FAILED and must be the first - * notifier. CPUSET_ACTIVE adjusts cpuset according to - * cpu_active mask right after SCHED_ACTIVE. During - * CPU_DOWN_PREPARE, SCHED_INACTIVE and CPUSET_INACTIVE are - * ordered in the similar way. + * CPU_ONLINE and CPU_DOWN_FAILED and must be the first notifier. Is + * also cpuset according to cpu_active mask right after activating the + * cpu. During CPU_DOWN_PREPARE, SCHED_INACTIVE reversed the operation. * * This ordering guarantees consistent cpu_active mask and * migration behavior to all cpu notifiers. */ CPU_PRI_SCHED_ACTIVE = INT_MAX, - CPU_PRI_CPUSET_ACTIVE = INT_MAX - 1, - CPU_PRI_SCHED_INACTIVE = INT_MIN + 1, - CPU_PRI_CPUSET_INACTIVE = INT_MIN, + CPU_PRI_SCHED_INACTIVE = INT_MIN, /* migration should happen before other stuff but after perf */ CPU_PRI_PERF = 20, diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 328502c9af00..8a03a04f5c3a 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5482,39 +5482,6 @@ static void set_cpu_rq_start_time(unsigned int cpu) rq->age_stamp = sched_clock_cpu(cpu); } -static int sched_cpu_active(struct notifier_block *nfb, - unsigned long action, void *hcpu) -{ - int cpu = (long)hcpu; - - switch (action & ~CPU_TASKS_FROZEN) { - case CPU_DOWN_FAILED: - set_cpu_active(cpu, true); - return NOTIFY_OK; - - default: - return NOTIFY_DONE; - } -} - -static int sched_cpu_inactive(struct notifier_block *nfb, - unsigned long action, void *hcpu) -{ - switch (action & ~CPU_TASKS_FROZEN) { - case CPU_DOWN_PREPARE: - set_cpu_active((long)hcpu, false); - return NOTIFY_OK; - default: - return NOTIFY_DONE; - } -} - -int sched_cpu_starting(unsigned int cpu) -{ - set_cpu_rq_start_time(cpu); - return 0; -} - static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */ #ifdef CONFIG_SCHED_DEBUG @@ -6662,10 +6629,13 @@ static void sched_init_numa(void) init_numa_topology_type(); } -static void sched_domains_numa_masks_set(int cpu) +static void sched_domains_numa_masks_set(unsigned int cpu) { - int i, j; int node = cpu_to_node(cpu); + int i, j; + + if (!sched_smp_initialized) + return; for (i = 0; i < sched_domains_numa_levels; i++) { for (j = 0; j < nr_node_ids; j++) { @@ -6675,54 +6645,23 @@ static void sched_domains_numa_masks_set(int cpu) } } -static void sched_domains_numa_masks_clear(int cpu) +static void sched_domains_numa_masks_clear(unsigned int cpu) { int i, j; + + if (!sched_smp_initialized) + return; + for (i = 0; i < sched_domains_numa_levels; i++) { for (j = 0; j < nr_node_ids; j++) cpumask_clear_cpu(cpu, sched_domains_numa_masks[i][j]); } } -/* - * Update sched_domains_numa_masks[level][node] array when new cpus - * are onlined. - */ -static int sched_domains_numa_masks_update(struct notifier_block *nfb, - unsigned long action, - void *hcpu) -{ - int cpu = (long)hcpu; - - if (!sched_smp_initialized) - return NOTIFY_DONE; - - switch (action & ~CPU_TASKS_FROZEN) { - case CPU_ONLINE: - sched_domains_numa_masks_set(cpu); - break; - - case CPU_DEAD: - sched_domains_numa_masks_clear(cpu); - break; - - default: - return NOTIFY_DONE; - } - - return NOTIFY_OK; -} #else -static inline void sched_init_numa(void) -{ -} - -static int sched_domains_numa_masks_update(struct notifier_block *nfb, - unsigned long action, - void *hcpu) -{ - return 0; -} +static inline void sched_init_numa(void) { } +static void sched_domains_numa_masks_set(unsigned int cpu) { } +static void sched_domains_numa_masks_clear(unsigned int cpu) { } #endif /* CONFIG_NUMA */ static int __sdt_alloc(const struct cpumask *cpu_map) @@ -7112,16 +7051,12 @@ static int num_cpus_frozen; /* used to mark begin/end of suspend/resume */ * If we come here as part of a suspend/resume, don't touch cpusets because we * want to restore it back to its original state upon resume anyway. */ -static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action, - void *hcpu) +static void cpuset_cpu_active(bool frozen) { if (!sched_smp_initialized) - return NOTIFY_DONE; - - switch (action) { - case CPU_ONLINE_FROZEN: - case CPU_DOWN_FAILED_FROZEN: + return; + if (frozen) { /* * num_cpus_frozen tracks how many CPUs are involved in suspend * resume sequence. As long as this is not the last online @@ -7131,38 +7066,28 @@ static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action, num_cpus_frozen--; if (likely(num_cpus_frozen)) { partition_sched_domains(1, NULL, NULL); - break; + return; } - /* * This is the last CPU online operation. So fall through and * restore the original sched domains by considering the * cpuset configurations. */ - - case CPU_ONLINE: - cpuset_update_active_cpus(true); - break; - default: - return NOTIFY_DONE; } - return NOTIFY_OK; + cpuset_update_active_cpus(true); } -static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action, - void *hcpu) +static int cpuset_cpu_inactive(unsigned int cpu, bool frozen) { unsigned long flags; - long cpu = (long)hcpu; struct dl_bw *dl_b; bool overflow; int cpus; if (!sched_smp_initialized) - return NOTIFY_DONE; + return 0; - switch (action) { - case CPU_DOWN_PREPARE: + if (!frozen) { rcu_read_lock_sched(); dl_b = dl_bw_of(cpu); @@ -7174,17 +7099,60 @@ static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action, rcu_read_unlock_sched(); if (overflow) - return notifier_from_errno(-EBUSY); + return -EBUSY; cpuset_update_active_cpus(false); - break; - case CPU_DOWN_PREPARE_FROZEN: + } else { num_cpus_frozen++; partition_sched_domains(1, NULL, NULL); - break; + } + return 0; +} + +static int sched_cpu_active(struct notifier_block *nfb, unsigned long action, + void *hcpu) +{ + unsigned int cpu = (unsigned long)hcpu; + + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_DOWN_FAILED: + case CPU_ONLINE: + set_cpu_active(cpu, true); + sched_domains_numa_masks_set(cpu); + cpuset_cpu_active(action & CPU_TASKS_FROZEN); + return NOTIFY_OK; default: return NOTIFY_DONE; } - return NOTIFY_OK; +} + +static int sched_cpu_inactive(struct notifier_block *nfb, + unsigned long action, void *hcpu) +{ + unsigned int cpu = (unsigned long)hcpu; + int ret; + + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_DOWN_PREPARE: + set_cpu_active(cpu, false); + ret = cpuset_cpu_inactive(cpu, action & CPU_TASKS_FROZEN); + if (ret) { + set_cpu_active(cpu, true); + return notifier_from_errno(ret); + } + return NOTIFY_OK; + + case CPU_DEAD: + sched_domains_numa_masks_clear(cpu); + return NOTIFY_OK; + default: + return NOTIFY_DONE; + } +} + +int sched_cpu_starting(unsigned int cpu) +{ + set_cpu_rq_start_time(cpu); + return 0; } void __init sched_init_smp(void) @@ -7236,10 +7204,6 @@ static int __init migration_init(void) cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE); cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE); - hotcpu_notifier(sched_domains_numa_masks_update, CPU_PRI_SCHED_ACTIVE); - hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE); - hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE); - return 0; } early_initcall(migration_init); -- cgit v1.2.3 From c6d2c7475ced868cfc636cd5a55a428da94537c3 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:12 +0100 Subject: sched: Move sched_domains_numa_masks_clear() to DOWN_PREPARE This is the last operation on the cpu before vanishing. No point in calling that on CPU_DEAD. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8a03a04f5c3a..541f9ab8ce4f 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -7139,9 +7139,6 @@ static int sched_cpu_inactive(struct notifier_block *nfb, set_cpu_active(cpu, true); return notifier_from_errno(ret); } - return NOTIFY_OK; - - case CPU_DEAD: sched_domains_numa_masks_clear(cpu); return NOTIFY_OK; default: -- cgit v1.2.3 From 40190a78f85fec29f0fdd21f6b4415712085711e Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:13 +0100 Subject: sched/hotplug: Convert cpu_[in]active notifiers to state machine Now that we reduced everything into single notifiers, it's simple to move them into the hotplug state machine space. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Signed-off-by: Thomas Gleixner --- include/linux/cpu.h | 12 --------- include/linux/cpuhotplug.h | 1 + include/linux/sched.h | 2 ++ kernel/cpu.c | 8 ++++-- kernel/sched/core.c | 67 +++++++++++++++------------------------------- 5 files changed, 30 insertions(+), 60 deletions(-) diff --git a/include/linux/cpu.h b/include/linux/cpu.h index 17017051bfb1..b22b000cf6ee 100644 --- a/include/linux/cpu.h +++ b/include/linux/cpu.h @@ -59,18 +59,6 @@ struct notifier_block; * CPU notifier priorities. */ enum { - /* - * SCHED_ACTIVE marks a cpu which is coming up active during - * CPU_ONLINE and CPU_DOWN_FAILED and must be the first notifier. Is - * also cpuset according to cpu_active mask right after activating the - * cpu. During CPU_DOWN_PREPARE, SCHED_INACTIVE reversed the operation. - * - * This ordering guarantees consistent cpu_active mask and - * migration behavior to all cpu notifiers. - */ - CPU_PRI_SCHED_ACTIVE = INT_MAX, - CPU_PRI_SCHED_INACTIVE = INT_MIN, - /* migration should happen before other stuff but after perf */ CPU_PRI_PERF = 20, CPU_PRI_MIGRATION = 10, diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h index 99fd1d2f76fe..9e07468bf1c5 100644 --- a/include/linux/cpuhotplug.h +++ b/include/linux/cpuhotplug.h @@ -13,6 +13,7 @@ enum cpuhp_state { CPUHP_AP_ONLINE, CPUHP_TEARDOWN_CPU, CPUHP_AP_ONLINE_IDLE, + CPUHP_AP_ACTIVE, CPUHP_AP_SMPBOOT_THREADS, CPUHP_AP_NOTIFY_ONLINE, CPUHP_AP_ONLINE_DYN, diff --git a/include/linux/sched.h b/include/linux/sched.h index 39597d0a005e..1e5f961b1a74 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -373,6 +373,8 @@ extern void trap_init(void); extern void update_process_times(int user); extern void scheduler_tick(void); extern int sched_cpu_starting(unsigned int cpu); +extern int sched_cpu_activate(unsigned int cpu); +extern int sched_cpu_deactivate(unsigned int cpu); extern void sched_show_task(struct task_struct *p); diff --git a/kernel/cpu.c b/kernel/cpu.c index f46d02b966bf..15402b72fbc7 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -923,8 +923,6 @@ void cpuhp_online_idle(enum cpuhp_state state) st->state = CPUHP_AP_ONLINE_IDLE; - /* The cpu is marked online, set it active now */ - set_cpu_active(cpu, true); /* Unpark the stopper thread and the hotplug thread of this cpu */ stop_machine_unpark(cpu); kthread_unpark(st->thread); @@ -1259,6 +1257,12 @@ static struct cpuhp_step cpuhp_ap_states[] = { [CPUHP_AP_ONLINE] = { .name = "ap:online", }, + /* First state is scheduler control. Interrupts are enabled */ + [CPUHP_AP_ACTIVE] = { + .name = "sched:active", + .startup = sched_cpu_activate, + .teardown = sched_cpu_deactivate, + }, /* Handle smpboot threads park/unpark */ [CPUHP_AP_SMPBOOT_THREADS] = { .name = "smpboot:threads", diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 541f9ab8ce4f..73bcd937d436 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -6634,9 +6634,6 @@ static void sched_domains_numa_masks_set(unsigned int cpu) int node = cpu_to_node(cpu); int i, j; - if (!sched_smp_initialized) - return; - for (i = 0; i < sched_domains_numa_levels; i++) { for (j = 0; j < nr_node_ids; j++) { if (node_distance(j, node) <= sched_domains_numa_distance[i]) @@ -6649,9 +6646,6 @@ static void sched_domains_numa_masks_clear(unsigned int cpu) { int i, j; - if (!sched_smp_initialized) - return; - for (i = 0; i < sched_domains_numa_levels; i++) { for (j = 0; j < nr_node_ids; j++) cpumask_clear_cpu(cpu, sched_domains_numa_masks[i][j]); @@ -7051,12 +7045,9 @@ static int num_cpus_frozen; /* used to mark begin/end of suspend/resume */ * If we come here as part of a suspend/resume, don't touch cpusets because we * want to restore it back to its original state upon resume anyway. */ -static void cpuset_cpu_active(bool frozen) +static void cpuset_cpu_active(void) { - if (!sched_smp_initialized) - return; - - if (frozen) { + if (cpuhp_tasks_frozen) { /* * num_cpus_frozen tracks how many CPUs are involved in suspend * resume sequence. As long as this is not the last online @@ -7077,17 +7068,14 @@ static void cpuset_cpu_active(bool frozen) cpuset_update_active_cpus(true); } -static int cpuset_cpu_inactive(unsigned int cpu, bool frozen) +static int cpuset_cpu_inactive(unsigned int cpu) { unsigned long flags; struct dl_bw *dl_b; bool overflow; int cpus; - if (!sched_smp_initialized) - return 0; - - if (!frozen) { + if (!cpuhp_tasks_frozen) { rcu_read_lock_sched(); dl_b = dl_bw_of(cpu); @@ -7108,42 +7096,33 @@ static int cpuset_cpu_inactive(unsigned int cpu, bool frozen) return 0; } -static int sched_cpu_active(struct notifier_block *nfb, unsigned long action, - void *hcpu) +int sched_cpu_activate(unsigned int cpu) { - unsigned int cpu = (unsigned long)hcpu; + set_cpu_active(cpu, true); - switch (action & ~CPU_TASKS_FROZEN) { - case CPU_DOWN_FAILED: - case CPU_ONLINE: - set_cpu_active(cpu, true); + if (sched_smp_initialized) { sched_domains_numa_masks_set(cpu); - cpuset_cpu_active(action & CPU_TASKS_FROZEN); - return NOTIFY_OK; - default: - return NOTIFY_DONE; + cpuset_cpu_active(); } + return 0; } -static int sched_cpu_inactive(struct notifier_block *nfb, - unsigned long action, void *hcpu) +int sched_cpu_deactivate(unsigned int cpu) { - unsigned int cpu = (unsigned long)hcpu; int ret; - switch (action & ~CPU_TASKS_FROZEN) { - case CPU_DOWN_PREPARE: - set_cpu_active(cpu, false); - ret = cpuset_cpu_inactive(cpu, action & CPU_TASKS_FROZEN); - if (ret) { - set_cpu_active(cpu, true); - return notifier_from_errno(ret); - } - sched_domains_numa_masks_clear(cpu); - return NOTIFY_OK; - default: - return NOTIFY_DONE; + set_cpu_active(cpu, false); + + if (!sched_smp_initialized) + return 0; + + ret = cpuset_cpu_inactive(cpu); + if (ret) { + set_cpu_active(cpu, true); + return ret; } + sched_domains_numa_masks_clear(cpu); + return 0; } int sched_cpu_starting(unsigned int cpu) @@ -7197,10 +7176,6 @@ static int __init migration_init(void) migration_call(&migration_notifier, CPU_ONLINE, cpu); register_cpu_notifier(&migration_notifier); - /* Register cpu active notifiers */ - cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE); - cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE); - return 0; } early_initcall(migration_init); -- cgit v1.2.3 From b2454caa8977ade27292a71f2def5e403e24b4d5 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 10 Mar 2016 12:54:14 +0100 Subject: sched/hotplug: Move sync_rcu to be with set_cpu_active(false) The sync_rcu stuff is specificically for clearing bits in the active mask, such that everybody will observe the bit cleared and will not consider the cleared CPU for load-balancing etc. Signed-off-by: Peter Zijlstra (Intel) Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.169219710@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/cpu.c | 15 --------------- kernel/sched/core.c | 14 ++++++++++++++ 2 files changed, 14 insertions(+), 15 deletions(-) diff --git a/kernel/cpu.c b/kernel/cpu.c index 15402b72fbc7..c134a35374a1 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -703,21 +703,6 @@ static int takedown_cpu(unsigned int cpu) struct cpuhp_cpu_state *st = per_cpu_ptr(&cpuhp_state, cpu); int err; - /* - * By now we've cleared cpu_active_mask, wait for all preempt-disabled - * and RCU users of this state to go away such that all new such users - * will observe it. - * - * For CONFIG_PREEMPT we have preemptible RCU and its sync_rcu() might - * not imply sync_sched(), so wait for both. - * - * Do sync before park smpboot threads to take care the rcu boost case. - */ - if (IS_ENABLED(CONFIG_PREEMPT)) - synchronize_rcu_mult(call_rcu, call_rcu_sched); - else - synchronize_rcu(); - /* Park the smpboot threads */ kthread_park(per_cpu_ptr(&cpuhp_state, cpu)->thread); smpboot_park_threads(cpu); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 73bcd937d436..0a3107809972 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -7112,6 +7112,20 @@ int sched_cpu_deactivate(unsigned int cpu) int ret; set_cpu_active(cpu, false); + /* + * We've cleared cpu_active_mask, wait for all preempt-disabled and RCU + * users of this state to go away such that all new such users will + * observe it. + * + * For CONFIG_PREEMPT we have preemptible RCU and its sync_rcu() might + * not imply sync_sched(), so wait for both. + * + * Do sync before park smpboot threads to take care the rcu boost case. + */ + if (IS_ENABLED(CONFIG_PREEMPT)) + synchronize_rcu_mult(call_rcu, call_rcu_sched); + else + synchronize_rcu(); if (!sched_smp_initialized) return 0; -- cgit v1.2.3 From 94baf7a5d882cde0b4d591f4ab89cc32ee39ac6a Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:15 +0100 Subject: sched/migration: Move prepare transition to SCHED_STARTING state We can piggy pack that on the SCHED_STARTING state. It's not required before the cpu actually comes online. Name the function proper as it has nothing to do with migration. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.248226511@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 0a3107809972..bafc308e6d45 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5424,11 +5424,6 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) switch (action & ~CPU_TASKS_FROZEN) { - case CPU_UP_PREPARE: - rq->calc_load_update = calc_load_update; - account_reset_rq(rq); - break; - case CPU_ONLINE: /* Update our root-domain */ raw_spin_lock_irqsave(&rq->lock, flags); @@ -7139,9 +7134,19 @@ int sched_cpu_deactivate(unsigned int cpu) return 0; } +static void sched_rq_cpu_starting(unsigned int cpu) +{ + struct rq *rq = cpu_rq(cpu); + + rq->calc_load_update = calc_load_update; + account_reset_rq(rq); + update_max_interval(); +} + int sched_cpu_starting(unsigned int cpu) { set_cpu_rq_start_time(cpu); + sched_rq_cpu_starting(cpu); return 0; } @@ -7182,11 +7187,8 @@ void __init sched_init_smp(void) static int __init migration_init(void) { void *cpu = (void *)(long)smp_processor_id(); - int err; - /* Initialize migration for the boot CPU */ - err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu); - BUG_ON(err == NOTIFY_BAD); + sched_rq_cpu_starting(smp_processor_id()); migration_call(&migration_notifier, CPU_ONLINE, cpu); register_cpu_notifier(&migration_notifier); -- cgit v1.2.3 From e9cd8fa4fcfd67c95db9b87c0fff88fa23cb00e5 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:16 +0100 Subject: sched/migration: Move calc_load_migrate() into CPU_DYING It really does not matter when we fold the load for the outgoing cpu. It's almost dead anyway, so there is no harm if we fail to fold the few microseconds which are required for going fully away. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.328739226@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index bafc308e6d45..688e8a83208c 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5447,9 +5447,6 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) migrate_tasks(rq); BUG_ON(rq->nr_running != 1); /* the migration thread */ raw_spin_unlock_irqrestore(&rq->lock, flags); - break; - - case CPU_DEAD: calc_load_migrate(rq); break; #endif -- cgit v1.2.3 From 7d97669933eb94245ec9b715753753ec5ca8f646 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:17 +0100 Subject: sched/migration: Move CPU_ONLINE into scheduler state The alleged requirement that the migration notifier has a lower priority than perf is completely undocumented and there is no indication at all that this is true. perf does not even handle the CPU_ONLINE notification and perf really has nothing to do with migration. Move the CPU_ONLINE code into the sched_activate_cpu() state callback. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.421743581@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 33 ++++++++++++++++++++++----------- 1 file changed, 22 insertions(+), 11 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 688e8a83208c..8d8d9034edff 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5424,17 +5424,6 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) switch (action & ~CPU_TASKS_FROZEN) { - case CPU_ONLINE: - /* Update our root-domain */ - raw_spin_lock_irqsave(&rq->lock, flags); - if (rq->rd) { - BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); - - set_rq_online(rq); - } - raw_spin_unlock_irqrestore(&rq->lock, flags); - break; - #ifdef CONFIG_HOTPLUG_CPU case CPU_DYING: sched_ttwu_pending(); @@ -7090,12 +7079,34 @@ static int cpuset_cpu_inactive(unsigned int cpu) int sched_cpu_activate(unsigned int cpu) { + struct rq *rq = cpu_rq(cpu); + unsigned long flags; + set_cpu_active(cpu, true); if (sched_smp_initialized) { sched_domains_numa_masks_set(cpu); cpuset_cpu_active(); } + + /* + * Put the rq online, if not already. This happens: + * + * 1) In the early boot process, because we build the real domains + * after all cpus have been brought up. + * + * 2) At runtime, if cpuset_cpu_active() fails to rebuild the + * domains. + */ + raw_spin_lock_irqsave(&rq->lock, flags); + if (rq->rd) { + BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); + set_rq_online(rq); + } + raw_spin_unlock_irqrestore(&rq->lock, flags); + + update_max_interval(); + return 0; } -- cgit v1.2.3 From f2785ddb5367e217365099294b89d6a84668069e Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:18 +0100 Subject: sched/hotplug: Move migration CPU_DYING to sched_cpu_dying() Remove the hotplug notifier and make it an explicit state. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.502222097@linutronix.de Signed-off-by: Thomas Gleixner --- include/linux/cpu.h | 2 -- include/linux/sched.h | 6 +++++ kernel/cpu.c | 2 +- kernel/sched/core.c | 72 ++++++++++++++++----------------------------------- 4 files changed, 29 insertions(+), 53 deletions(-) diff --git a/include/linux/cpu.h b/include/linux/cpu.h index b22b000cf6ee..21597dcac0e2 100644 --- a/include/linux/cpu.h +++ b/include/linux/cpu.h @@ -59,9 +59,7 @@ struct notifier_block; * CPU notifier priorities. */ enum { - /* migration should happen before other stuff but after perf */ CPU_PRI_PERF = 20, - CPU_PRI_MIGRATION = 10, /* bring up workqueues before normal notifiers and down after */ CPU_PRI_WORKQUEUE_UP = 5, diff --git a/include/linux/sched.h b/include/linux/sched.h index 1e5f961b1a74..47835cf8aefa 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -376,6 +376,12 @@ extern int sched_cpu_starting(unsigned int cpu); extern int sched_cpu_activate(unsigned int cpu); extern int sched_cpu_deactivate(unsigned int cpu); +#ifdef CONFIG_HOTPLUG_CPU +extern int sched_cpu_dying(unsigned int cpu); +#else +# define sched_cpu_dying NULL +#endif + extern void sched_show_task(struct task_struct *p); #ifdef CONFIG_LOCKUP_DETECTOR diff --git a/kernel/cpu.c b/kernel/cpu.c index c134a35374a1..d6eeb8c5ef88 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -1223,7 +1223,7 @@ static struct cpuhp_step cpuhp_ap_states[] = { [CPUHP_AP_SCHED_STARTING] = { .name = "sched:starting", .startup = sched_cpu_starting, - .teardown = NULL, + .teardown = sched_cpu_dying, }, /* * Low level startup/teardown notifiers. Run with interrupts diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8d8d9034edff..a9a65ed772e3 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5411,51 +5411,6 @@ static void set_rq_offline(struct rq *rq) } } -/* - * migration_call - callback that gets triggered when a CPU is added. - * Here we can start up the necessary migration thread for the new CPU. - */ -static int -migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) -{ - int cpu = (long)hcpu; - unsigned long flags; - struct rq *rq = cpu_rq(cpu); - - switch (action & ~CPU_TASKS_FROZEN) { - -#ifdef CONFIG_HOTPLUG_CPU - case CPU_DYING: - sched_ttwu_pending(); - /* Update our root-domain */ - raw_spin_lock_irqsave(&rq->lock, flags); - if (rq->rd) { - BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); - set_rq_offline(rq); - } - migrate_tasks(rq); - BUG_ON(rq->nr_running != 1); /* the migration thread */ - raw_spin_unlock_irqrestore(&rq->lock, flags); - calc_load_migrate(rq); - break; -#endif - } - - update_max_interval(); - - return NOTIFY_OK; -} - -/* - * Register at high priority so that task migration (migrate_all_tasks) - * happens before everything else. This has to be lower priority than - * the notifier in the perf_event subsystem, though. - */ -static struct notifier_block migration_notifier = { - .notifier_call = migration_call, - .priority = CPU_PRI_MIGRATION, -}; - static void set_cpu_rq_start_time(unsigned int cpu) { struct rq *rq = cpu_rq(cpu); @@ -7158,6 +7113,28 @@ int sched_cpu_starting(unsigned int cpu) return 0; } +#ifdef CONFIG_HOTPLUG_CPU +int sched_cpu_dying(unsigned int cpu) +{ + struct rq *rq = cpu_rq(cpu); + unsigned long flags; + + /* Handle pending wakeups and then migrate everything off */ + sched_ttwu_pending(); + raw_spin_lock_irqsave(&rq->lock, flags); + if (rq->rd) { + BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); + set_rq_offline(rq); + } + migrate_tasks(rq); + BUG_ON(rq->nr_running != 1); + raw_spin_unlock_irqrestore(&rq->lock, flags); + calc_load_migrate(rq); + update_max_interval(); + return 0; +} +#endif + void __init sched_init_smp(void) { cpumask_var_t non_isolated_cpus; @@ -7194,12 +7171,7 @@ void __init sched_init_smp(void) static int __init migration_init(void) { - void *cpu = (void *)(long)smp_processor_id(); - sched_rq_cpu_starting(smp_processor_id()); - migration_call(&migration_notifier, CPU_ONLINE, cpu); - register_cpu_notifier(&migration_notifier); - return 0; } early_initcall(migration_init); -- cgit v1.2.3 From aaddd7d1c740ab3c5efaad7a34650b6dc680c21c Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:19 +0100 Subject: sched/hotplug: Make activate() the last hotplug step The scheduler can handle per cpu threads before the cpu is set to active and it does not allow user space threads on the cpu before active is set. Attaching to the scheduling domains is also not required before user space threads can be handled. Move the activation to the end of the hotplug state space. That also means that deactivation is the first action when a cpu is shut down. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.597477199@linutronix.de Signed-off-by: Thomas Gleixner --- include/linux/cpuhotplug.h | 2 +- kernel/cpu.c | 15 +++++++++------ 2 files changed, 10 insertions(+), 7 deletions(-) diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h index 9e07468bf1c5..386374d19987 100644 --- a/include/linux/cpuhotplug.h +++ b/include/linux/cpuhotplug.h @@ -13,11 +13,11 @@ enum cpuhp_state { CPUHP_AP_ONLINE, CPUHP_TEARDOWN_CPU, CPUHP_AP_ONLINE_IDLE, - CPUHP_AP_ACTIVE, CPUHP_AP_SMPBOOT_THREADS, CPUHP_AP_NOTIFY_ONLINE, CPUHP_AP_ONLINE_DYN, CPUHP_AP_ONLINE_DYN_END = CPUHP_AP_ONLINE_DYN + 30, + CPUHP_AP_ACTIVE, CPUHP_ONLINE, }; diff --git a/kernel/cpu.c b/kernel/cpu.c index d6eeb8c5ef88..d948e44c471e 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -1242,12 +1242,6 @@ static struct cpuhp_step cpuhp_ap_states[] = { [CPUHP_AP_ONLINE] = { .name = "ap:online", }, - /* First state is scheduler control. Interrupts are enabled */ - [CPUHP_AP_ACTIVE] = { - .name = "sched:active", - .startup = sched_cpu_activate, - .teardown = sched_cpu_deactivate, - }, /* Handle smpboot threads park/unpark */ [CPUHP_AP_SMPBOOT_THREADS] = { .name = "smpboot:threads", @@ -1269,6 +1263,15 @@ static struct cpuhp_step cpuhp_ap_states[] = { * The dynamically registered state space is here */ +#ifdef CONFIG_SMP + /* Last state is scheduler control setting the cpu active */ + [CPUHP_AP_ACTIVE] = { + .name = "sched:active", + .startup = sched_cpu_activate, + .teardown = sched_cpu_deactivate, + }, +#endif + /* CPU is fully up and running. */ [CPUHP_ONLINE] = { .name = "online", -- cgit v1.2.3 From 20a5c8cc74ade5027c2b0e2bc724278afd6054f3 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:20 +0100 Subject: sched/fair: Make ilb_notifier an explicit call No need for an extra notifier. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.693720241@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 1 + kernel/sched/fair.c | 15 +-------------- kernel/sched/sched.h | 4 ++++ 3 files changed, 6 insertions(+), 14 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index a9a65ed772e3..28ffd6882ea6 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -7131,6 +7131,7 @@ int sched_cpu_dying(unsigned int cpu) raw_spin_unlock_irqrestore(&rq->lock, flags); calc_load_migrate(rq); update_max_interval(); + nohz_balance_exit_idle(cpu); return 0; } #endif diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 0fe30e66aff1..8b6db3626566 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7711,7 +7711,7 @@ static void nohz_balancer_kick(void) return; } -static inline void nohz_balance_exit_idle(int cpu) +void nohz_balance_exit_idle(unsigned int cpu) { if (unlikely(test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))) { /* @@ -7784,18 +7784,6 @@ void nohz_balance_enter_idle(int cpu) atomic_inc(&nohz.nr_cpus); set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)); } - -static int sched_ilb_notifier(struct notifier_block *nfb, - unsigned long action, void *hcpu) -{ - switch (action & ~CPU_TASKS_FROZEN) { - case CPU_DYING: - nohz_balance_exit_idle(smp_processor_id()); - return NOTIFY_OK; - default: - return NOTIFY_DONE; - } -} #endif static DEFINE_SPINLOCK(balancing); @@ -8600,7 +8588,6 @@ __init void init_sched_fair_class(void) #ifdef CONFIG_NO_HZ_COMMON nohz.next_balance = jiffies; zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT); - cpu_notifier(sched_ilb_notifier, 0); #endif #endif /* SMP */ diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index ec2e8d23527e..16a27b624ee5 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1743,6 +1743,10 @@ enum rq_nohz_flag_bits { }; #define nohz_flags(cpu) (&cpu_rq(cpu)->nohz_flags) + +extern void nohz_balance_exit_idle(unsigned int cpu); +#else +static inline void nohz_balance_exit_idle(unsigned int cpu) { } #endif #ifdef CONFIG_IRQ_TIME_ACCOUNTING -- cgit v1.2.3 From e5ef27d0f5acf9f1db2882d7546a41c021f66820 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 10 Mar 2016 12:54:21 +0100 Subject: sched: Make hrtick_notifier an explicit call No need for an extra notifier. We don't need to handle all these states. It's sufficient to kill the timer when the cpu dies. Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160310120025.770528462@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/sched/core.c | 34 +--------------------------------- 1 file changed, 1 insertion(+), 33 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 28ffd6882ea6..9c710ad0ac22 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -249,29 +249,6 @@ void hrtick_start(struct rq *rq, u64 delay) } } -static int -hotplug_hrtick(struct notifier_block *nfb, unsigned long action, void *hcpu) -{ - int cpu = (int)(long)hcpu; - - switch (action) { - case CPU_UP_CANCELED: - case CPU_UP_CANCELED_FROZEN: - case CPU_DOWN_PREPARE: - case CPU_DOWN_PREPARE_FROZEN: - case CPU_DEAD: - case CPU_DEAD_FROZEN: - hrtick_clear(cpu_rq(cpu)); - return NOTIFY_OK; - } - - return NOTIFY_DONE; -} - -static __init void init_hrtick(void) -{ - hotcpu_notifier(hotplug_hrtick, 0); -} #else /* * Called to set the hrtick timer state. @@ -288,10 +265,6 @@ void hrtick_start(struct rq *rq, u64 delay) hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay), HRTIMER_MODE_REL_PINNED); } - -static inline void init_hrtick(void) -{ -} #endif /* CONFIG_SMP */ static void init_rq_hrtick(struct rq *rq) @@ -315,10 +288,6 @@ static inline void hrtick_clear(struct rq *rq) static inline void init_rq_hrtick(struct rq *rq) { } - -static inline void init_hrtick(void) -{ -} #endif /* CONFIG_SCHED_HRTICK */ /* @@ -7132,6 +7101,7 @@ int sched_cpu_dying(unsigned int cpu) calc_load_migrate(rq); update_max_interval(); nohz_balance_exit_idle(cpu); + hrtick_clear(rq); return 0; } #endif @@ -7157,8 +7127,6 @@ void __init sched_init_smp(void) cpumask_set_cpu(smp_processor_id(), non_isolated_cpus); mutex_unlock(&sched_domains_mutex); - init_hrtick(); - /* Move init over to a non-isolated CPU */ if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0) BUG(); -- cgit v1.2.3 From 8c5e95548d41a48b1eb2be741107a259251ebd86 Mon Sep 17 00:00:00 2001 From: Muhammad Falak R Wani Date: Thu, 5 May 2016 15:21:19 +0530 Subject: sched/core: Remove unused variable Remove unused variable 'ret', and directly return 0. Signed-off-by: Muhammad Falak R Wani Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1462441879-10092-1-git-send-email-falakreyaz@gmail.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 71c5a753e6e9..a6d3e7a9c0a9 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -8019,7 +8019,7 @@ static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk) static int sched_rt_global_constraints(void) { unsigned long flags; - int i, ret = 0; + int i; raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags); for_each_possible_cpu(i) { @@ -8031,7 +8031,7 @@ static int sched_rt_global_constraints(void) } raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags); - return ret; + return 0; } #endif /* CONFIG_RT_GROUP_SCHED */ -- cgit v1.2.3 From 58fe9c4621b7219e724c0b7af053112f974a08c3 Mon Sep 17 00:00:00 2001 From: Davidlohr Bueso Date: Sun, 8 May 2016 20:58:10 -0700 Subject: sched/core: Fix comment typo in wake_q_add() ... the comment clearly refers to wake_up_q(), and not wake_up_list(). Signed-off-by: Davidlohr Bueso Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: dave@stgolabs.net Link: http://lkml.kernel.org/r/1462766290-28664-1-git-send-email-dave@stgolabs.net Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index a6d3e7a9c0a9..e09f92c3a096 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -465,7 +465,7 @@ void wake_q_add(struct wake_q_head *head, struct task_struct *task) * wakeup due to that. * * This cmpxchg() implies a full barrier, which pairs with the write - * barrier implied by the wakeup in wake_up_list(). + * barrier implied by the wakeup in wake_up_q(). */ if (cmpxchg(&node->next, NULL, WAKE_Q_TAIL)) return; -- cgit v1.2.3 From c58d25f371f5e4b2dfbec3a7bd6f3c24dd79095b Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 12 May 2016 09:19:59 +0200 Subject: sched/fair: Move record_wakee() Since I want to make ->task_woken() conditional on the task getting migrated, we cannot use it to call record_wakee(). Move it to select_task_rq_fair(), which gets called in almost all the same conditions. The only exception is if the woken task (@p) is CPU-bound (as per the nr_cpus_allowed test in select_task_rq()). Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Hunter Cc: Ben Segall Cc: Linus Torvalds Cc: Matt Fleming Cc: Mike Galbraith Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Paul Turner Cc: Pavan Kondeti Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: byungchul.park@lge.com Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 61 +++++++++++++++++++++++++++++------------------------ 1 file changed, 33 insertions(+), 28 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 84e465ae7c63..48633a1c3b46 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4810,24 +4810,6 @@ static unsigned long cpu_avg_load_per_task(int cpu) return 0; } -static void record_wakee(struct task_struct *p) -{ - /* - * Rough decay (wiping) for cost saving, don't worry - * about the boundary, really active task won't care - * about the loss. - */ - if (time_after(jiffies, current->wakee_flip_decay_ts + HZ)) { - current->wakee_flips >>= 1; - current->wakee_flip_decay_ts = jiffies; - } - - if (current->last_wakee != p) { - current->last_wakee = p; - current->wakee_flips++; - } -} - static void task_waking_fair(struct task_struct *p) { struct sched_entity *se = &p->se; @@ -4847,7 +4829,6 @@ static void task_waking_fair(struct task_struct *p) #endif se->vruntime -= min_vruntime; - record_wakee(p); } #ifdef CONFIG_FAIR_GROUP_SCHED @@ -4965,17 +4946,39 @@ static long effective_load(struct task_group *tg, int cpu, long wl, long wg) #endif +static void record_wakee(struct task_struct *p) +{ + /* + * Only decay a single time; tasks that have less then 1 wakeup per + * jiffy will not have built up many flips. + */ + if (time_after(jiffies, current->wakee_flip_decay_ts + HZ)) { + current->wakee_flips >>= 1; + current->wakee_flip_decay_ts = jiffies; + } + + if (current->last_wakee != p) { + current->last_wakee = p; + current->wakee_flips++; + } +} + /* * Detect M:N waker/wakee relationships via a switching-frequency heuristic. + * * A waker of many should wake a different task than the one last awakened - * at a frequency roughly N times higher than one of its wakees. In order - * to determine whether we should let the load spread vs consolodating to - * shared cache, we look for a minimum 'flip' frequency of llc_size in one - * partner, and a factor of lls_size higher frequency in the other. With - * both conditions met, we can be relatively sure that the relationship is - * non-monogamous, with partner count exceeding socket size. Waker/wakee - * being client/server, worker/dispatcher, interrupt source or whatever is - * irrelevant, spread criteria is apparent partner count exceeds socket size. + * at a frequency roughly N times higher than one of its wakees. + * + * In order to determine whether we should let the load spread vs consolidating + * to shared cache, we look for a minimum 'flip' frequency of llc_size in one + * partner, and a factor of lls_size higher frequency in the other. + * + * With both conditions met, we can be relatively sure that the relationship is + * non-monogamous, with partner count exceeding socket size. + * + * Waker/wakee being client/server, worker/dispatcher, interrupt source or + * whatever is irrelevant, spread criteria is apparent partner count exceeds + * socket size. */ static int wake_wide(struct task_struct *p) { @@ -5280,8 +5283,10 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f int want_affine = 0; int sync = wake_flags & WF_SYNC; - if (sd_flag & SD_BALANCE_WAKE) + if (sd_flag & SD_BALANCE_WAKE) { + record_wakee(p); want_affine = !wake_wide(p) && cpumask_test_cpu(cpu, tsk_cpus_allowed(p)); + } rcu_read_lock(); for_each_domain(cpu, tmp) { -- cgit v1.2.3 From b5179ac70de85ef477cedf8b026a57913754cf1e Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 11 May 2016 16:10:34 +0200 Subject: sched/fair: Prepare to fix fairness problems on migration Mike reported that our recent attempt to fix migration problems: 3a47d5124a95 ("sched/fair: Fix fairness issue on migration") broke interactivity and the signal starve test. We reverted that commit and now let's try it again more carefully, with some other underlying problems fixed first. One problem is that I assumed ENQUEUE_WAKING was only set when we do a cross-cpu wakeup (migration), which isn't true. This means we now destroy the vruntime history of tasks and wakeup-preemption suffers. Cure this by making my assumption true, only call sched_class::task_waking() when we do a cross-cpu wakeup. This avoids the indirect call in the case we do a local wakeup. Reported-by: Mike Galbraith Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Hunter Cc: Ben Segall Cc: Linus Torvalds Cc: Matt Fleming Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Paul Turner Cc: Pavan Kondeti Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: byungchul.park@lge.com Cc: linux-kernel@vger.kernel.org Fixes: 3a47d5124a95 ("sched/fair: Fix fairness issue on migration") Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 29 +++++++++++++++++++++-------- kernel/sched/fair.c | 37 +++++++++++++++++++++++++++++++++++++ 2 files changed, 58 insertions(+), 8 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 1e622f254df4..1f73e2554bc1 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1709,14 +1709,22 @@ static void ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags, struct pin_cookie cookie) { + int en_flags = ENQUEUE_WAKEUP; + lockdep_assert_held(&rq->lock); #ifdef CONFIG_SMP if (p->sched_contributes_to_load) rq->nr_uninterruptible--; + + /* + * If we migrated; we must have called sched_class::task_waking(). + */ + if (wake_flags & WF_MIGRATED) + en_flags |= ENQUEUE_WAKING; #endif - ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_WAKING); + ttwu_activate(rq, p, en_flags); ttwu_do_wakeup(rq, p, wake_flags, cookie); } @@ -1762,7 +1770,11 @@ void sched_ttwu_pending(void) while (llist) { p = llist_entry(llist, struct task_struct, wake_entry); llist = llist_next(llist); - ttwu_do_activate(rq, p, 0, cookie); + /* + * See ttwu_queue(); we only call ttwu_queue_remote() when + * its a x-cpu wakeup. + */ + ttwu_do_activate(rq, p, WF_MIGRATED, cookie); } lockdep_unpin_lock(&rq->lock, cookie); @@ -1849,7 +1861,7 @@ bool cpus_share_cache(int this_cpu, int that_cpu) } #endif /* CONFIG_SMP */ -static void ttwu_queue(struct task_struct *p, int cpu) +static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags) { struct rq *rq = cpu_rq(cpu); struct pin_cookie cookie; @@ -1864,7 +1876,7 @@ static void ttwu_queue(struct task_struct *p, int cpu) raw_spin_lock(&rq->lock); cookie = lockdep_pin_lock(&rq->lock); - ttwu_do_activate(rq, p, 0, cookie); + ttwu_do_activate(rq, p, wake_flags, cookie); lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); } @@ -2034,17 +2046,18 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) p->sched_contributes_to_load = !!task_contributes_to_load(p); p->state = TASK_WAKING; - if (p->sched_class->task_waking) - p->sched_class->task_waking(p); - cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags); if (task_cpu(p) != cpu) { wake_flags |= WF_MIGRATED; + + if (p->sched_class->task_waking) + p->sched_class->task_waking(p); + set_task_cpu(p, cpu); } #endif /* CONFIG_SMP */ - ttwu_queue(p, cpu); + ttwu_queue(p, cpu, wake_flags); stat: if (schedstat_enabled()) ttwu_stat(p, cpu, wake_flags); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 48633a1c3b46..445bcd2d7ee1 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -3254,6 +3254,37 @@ static inline void check_schedstat_required(void) #endif } + +/* + * MIGRATION + * + * dequeue + * update_curr() + * update_min_vruntime() + * vruntime -= min_vruntime + * + * enqueue + * update_curr() + * update_min_vruntime() + * vruntime += min_vruntime + * + * this way the vruntime transition between RQs is done when both + * min_vruntime are up-to-date. + * + * WAKEUP (remote) + * + * ->task_waking_fair() + * vruntime -= min_vruntime + * + * enqueue + * update_curr() + * update_min_vruntime() + * vruntime += min_vruntime + * + * this way we don't have the most up-to-date min_vruntime on the originating + * CPU and an up-to-date min_vruntime on the destination CPU. + */ + static void enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { @@ -4810,6 +4841,12 @@ static unsigned long cpu_avg_load_per_task(int cpu) return 0; } +/* + * Called to migrate a waking task; as blocked tasks retain absolute vruntime + * the migration needs to deal with this by subtracting the old and adding the + * new min_vruntime -- the latter is done by enqueue_entity() when placing + * the task on the new runqueue. + */ static void task_waking_fair(struct task_struct *p) { struct sched_entity *se = &p->se; -- cgit v1.2.3 From 59efa0bac9cf8b2ef8d08f7632826c6d90f6a9bb Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 10 May 2016 18:24:37 +0200 Subject: sched/core: Kill sched_class::task_waking to clean up the migration logic With sched_class::task_waking being called only when we do set_task_cpu(), we can make sched_class::migrate_task_rq() do the work and eliminate sched_class::task_waking entirely. Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Hunter Cc: Ben Segall Cc: Linus Torvalds Cc: Matt Fleming Cc: Mike Galbraith Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Paul Turner Cc: Pavan Kondeti Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: byungchul.park@lge.com Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 9 +------- kernel/sched/fair.c | 58 +++++++++++++++++++++++++--------------------------- kernel/sched/sched.h | 7 +++---- 3 files changed, 32 insertions(+), 42 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 1f73e2554bc1..636c4b9cac38 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1717,11 +1717,8 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags, if (p->sched_contributes_to_load) rq->nr_uninterruptible--; - /* - * If we migrated; we must have called sched_class::task_waking(). - */ if (wake_flags & WF_MIGRATED) - en_flags |= ENQUEUE_WAKING; + en_flags |= ENQUEUE_MIGRATED; #endif ttwu_activate(rq, p, en_flags); @@ -2049,10 +2046,6 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags); if (task_cpu(p) != cpu) { wake_flags |= WF_MIGRATED; - - if (p->sched_class->task_waking) - p->sched_class->task_waking(p); - set_task_cpu(p, cpu); } #endif /* CONFIG_SMP */ diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 445bcd2d7ee1..24ce01b73906 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -3273,7 +3273,7 @@ static inline void check_schedstat_required(void) * * WAKEUP (remote) * - * ->task_waking_fair() + * ->migrate_task_rq_fair() (p->state == TASK_WAKING) * vruntime -= min_vruntime * * enqueue @@ -3292,7 +3292,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) * Update the normalized vruntime before updating min_vruntime * through calling update_curr(). */ - if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING)) + if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATED)) se->vruntime += cfs_rq->min_vruntime; /* @@ -4841,33 +4841,6 @@ static unsigned long cpu_avg_load_per_task(int cpu) return 0; } -/* - * Called to migrate a waking task; as blocked tasks retain absolute vruntime - * the migration needs to deal with this by subtracting the old and adding the - * new min_vruntime -- the latter is done by enqueue_entity() when placing - * the task on the new runqueue. - */ -static void task_waking_fair(struct task_struct *p) -{ - struct sched_entity *se = &p->se; - struct cfs_rq *cfs_rq = cfs_rq_of(se); - u64 min_vruntime; - -#ifndef CONFIG_64BIT - u64 min_vruntime_copy; - - do { - min_vruntime_copy = cfs_rq->min_vruntime_copy; - smp_rmb(); - min_vruntime = cfs_rq->min_vruntime; - } while (min_vruntime != min_vruntime_copy); -#else - min_vruntime = cfs_rq->min_vruntime; -#endif - - se->vruntime -= min_vruntime; -} - #ifdef CONFIG_FAIR_GROUP_SCHED /* * effective_load() calculates the load change as seen from the root_task_group @@ -5402,6 +5375,32 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f */ static void migrate_task_rq_fair(struct task_struct *p) { + /* + * As blocked tasks retain absolute vruntime the migration needs to + * deal with this by subtracting the old and adding the new + * min_vruntime -- the latter is done by enqueue_entity() when placing + * the task on the new runqueue. + */ + if (p->state == TASK_WAKING) { + struct sched_entity *se = &p->se; + struct cfs_rq *cfs_rq = cfs_rq_of(se); + u64 min_vruntime; + +#ifndef CONFIG_64BIT + u64 min_vruntime_copy; + + do { + min_vruntime_copy = cfs_rq->min_vruntime_copy; + smp_rmb(); + min_vruntime = cfs_rq->min_vruntime; + } while (min_vruntime != min_vruntime_copy); +#else + min_vruntime = cfs_rq->min_vruntime; +#endif + + se->vruntime -= min_vruntime; + } + /* * We are supposed to update the task to "current" time, then its up to date * and ready to go to new CPU/cfs_rq. But we have difficulty in getting @@ -8672,7 +8671,6 @@ const struct sched_class fair_sched_class = { .rq_online = rq_online_fair, .rq_offline = rq_offline_fair, - .task_waking = task_waking_fair, .task_dead = task_dead_fair, .set_cpus_allowed = set_cpus_allowed_common, #endif diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index ab6adb159e23..e51145e76807 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1168,7 +1168,7 @@ extern const u32 sched_prio_to_wmult[40]; * * ENQUEUE_HEAD - place at front of runqueue (tail if not specified) * ENQUEUE_REPLENISH - CBS (replenish runtime and postpone deadline) - * ENQUEUE_WAKING - sched_class::task_waking was called + * ENQUEUE_MIGRATED - the task was migrated during wakeup * */ @@ -1183,9 +1183,9 @@ extern const u32 sched_prio_to_wmult[40]; #define ENQUEUE_HEAD 0x08 #define ENQUEUE_REPLENISH 0x10 #ifdef CONFIG_SMP -#define ENQUEUE_WAKING 0x20 +#define ENQUEUE_MIGRATED 0x20 #else -#define ENQUEUE_WAKING 0x00 +#define ENQUEUE_MIGRATED 0x00 #endif #define RETRY_TASK ((void *)-1UL) @@ -1217,7 +1217,6 @@ struct sched_class { int (*select_task_rq)(struct task_struct *p, int task_cpu, int sd_flag, int flags); void (*migrate_task_rq)(struct task_struct *p); - void (*task_waking) (struct task_struct *task); void (*task_woken) (struct rq *this_rq, struct task_struct *task); void (*set_cpus_allowed)(struct task_struct *p, -- cgit v1.2.3 From 2f950354e6d535b892f133d20bd6a8b09430424c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 11 May 2016 19:27:56 +0200 Subject: sched/fair: Fix fairness issue on migration Pavan reported that in the presence of very light tasks (or cgroups) the placement of migrated tasks can cause severe fairness issues. The problem is that enqueue_entity() places the task before it updates time, thereby it can place the task far in the past (remember that light tasks will shoot virtual time forward at a high speed, so in relation to the pre-existing light task, we can land far in the past). This is done because update_curr() needs the current task, and we might be placing the current task. The obvious solution is to differentiate between the current and any other task; placing the current before we update time, and placing any other task after, such that !curr tasks end up at the current moment in time, and not in the past. This commit re-introduces the previously reverted commit: 3a47d5124a95 ("sched/fair: Fix fairness issue on migration") ... which is now safe to do, after we've also fixed another underlying bug first, in: sched/fair: Prepare to fix fairness problems on migration and cleaned up other details in the migration code: sched/core: Kill sched_class::task_waking Reported-by: Pavan Kondeti Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 24ce01b73906..d28d89d774aa 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -3288,17 +3288,27 @@ static inline void check_schedstat_required(void) static void enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { + bool renorm = !(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATED); + bool curr = cfs_rq->curr == se; + /* - * Update the normalized vruntime before updating min_vruntime - * through calling update_curr(). + * If we're the current task, we must renormalise before calling + * update_curr(). */ - if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATED)) + if (renorm && curr) se->vruntime += cfs_rq->min_vruntime; + update_curr(cfs_rq); + /* - * Update run-time statistics of the 'current'. + * Otherwise, renormalise after, such that we're placed at the current + * moment in time, instead of some random moment in the past. Being + * placed in the past could significantly boost this task to the + * fairness detriment of existing tasks. */ - update_curr(cfs_rq); + if (renorm && !curr) + se->vruntime += cfs_rq->min_vruntime; + enqueue_entity_load_avg(cfs_rq, se); account_entity_enqueue(cfs_rq, se); update_cfs_shares(cfs_rq); @@ -3314,7 +3324,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) update_stats_enqueue(cfs_rq, se); check_spread(cfs_rq, se); } - if (se != cfs_rq->curr) + if (!curr) __enqueue_entity(cfs_rq, se); se->on_rq = 1; -- cgit v1.2.3 From 444969223c81c7d0a95136b7b4cfdcfbc96ac5bd Mon Sep 17 00:00:00 2001 From: Wanpeng Li Date: Wed, 4 May 2016 14:45:34 +0800 Subject: sched/nohz: Fix affine unpinned timers mess The following commit: 9642d18eee2c ("nohz: Affine unpinned timers to housekeepers")' intended to affine unpinned timers to housekeepers: unpinned timers(full dynaticks, idle) => nearest busy housekeepers(otherwise, fallback to any housekeepers) unpinned timers(full dynaticks, busy) => nearest busy housekeepers(otherwise, fallback to any housekeepers) unpinned timers(houserkeepers, idle) => nearest busy housekeepers(otherwise, fallback to itself) However, the !idle_cpu(i) && is_housekeeping_cpu(cpu) check modified the intention to: unpinned timers(full dynaticks, idle) => any housekeepers(no mattter cpu topology) unpinned timers(full dynaticks, busy) => any housekeepers(no mattter cpu topology) unpinned timers(housekeepers, idle) => any busy cpus(otherwise, fallback to any housekeepers) This patch fixes it by checking if there are busy housekeepers nearby, otherwise falls to any housekeepers/itself. After the patch: unpinned timers(full dynaticks, idle) => nearest busy housekeepers(otherwise, fallback to any housekeepers) unpinned timers(full dynaticks, busy) => nearest busy housekeepers(otherwise, fallback to any housekeepers) unpinned timers(housekeepers, idle) => nearest busy housekeepers(otherwise, fallback to itself) Signed-off-by: Wanpeng Li Signed-off-by: Peter Zijlstra (Intel) [ Fixed the changelog. ] Cc: Frederic Weisbecker Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Fixes: 'commit 9642d18eee2c ("nohz: Affine unpinned timers to housekeepers")' Link: http://lkml.kernel.org/r/1462344334-8303-1-git-send-email-wanpeng.li@hotmail.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 636c4b9cac38..6f6962a356b4 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -533,7 +533,10 @@ int get_nohz_timer_target(void) rcu_read_lock(); for_each_domain(cpu, sd) { for_each_cpu(i, sched_domain_span(sd)) { - if (!idle_cpu(i) && is_housekeeping_cpu(cpu)) { + if (cpu == i) + continue; + + if (!idle_cpu(i) && is_housekeeping_cpu(i)) { cpu = i; goto unlock; } -- cgit v1.2.3 From 1be0eb2a97d756fb7dd8c9baf372d81fa9699c09 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 6 May 2016 12:21:23 +0200 Subject: sched/fair: Clean up scale confusion Wanpeng noted that the scale_load_down() in calculate_imbalance() was weird. I agree, it should be SCHED_CAPACITY_SCALE, since we're going to compare against busiest->group_capacity, which is in [capacity] units. Reported-by: Wanpeng Li Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Yuyang Du Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index d28d89d774aa..23381056b4b5 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7066,8 +7066,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s */ if (busiest->group_type == group_overloaded && local->group_type == group_overloaded) { - load_above_capacity = busiest->sum_nr_running * - scale_load_down(NICE_0_LOAD); + load_above_capacity = busiest->sum_nr_running * SCHED_CAPACITY_SCALE; if (load_above_capacity > busiest->group_capacity) load_above_capacity -= busiest->group_capacity; else -- cgit v1.2.3 From cfa10334318d8212d007da8c771187643c9cef35 Mon Sep 17 00:00:00 2001 From: Morten Rasmussen Date: Fri, 29 Apr 2016 20:32:40 +0100 Subject: sched/fair: Correct unit of load_above_capacity In calculate_imbalance() load_above_capacity currently has the unit [capacity] while it is used as being [load/capacity]. Not only is it wrong it also makes it unlikely that load_above_capacity is ever used as the subsequent code picks the smaller of load_above_capacity and the avg_load This patch ensures that load_above_capacity has the right unit [load/capacity]. Signed-off-by: Morten Rasmussen [ Changed changelog to note it was in capacity unit; +rebase. ] Signed-off-by: Peter Zijlstra (Intel) Cc: Dietmar Eggemann Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Link: http://lkml.kernel.org/r/1461958364-675-4-git-send-email-dietmar.eggemann@arm.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 23381056b4b5..218f8e83db73 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7067,9 +7067,11 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s if (busiest->group_type == group_overloaded && local->group_type == group_overloaded) { load_above_capacity = busiest->sum_nr_running * SCHED_CAPACITY_SCALE; - if (load_above_capacity > busiest->group_capacity) + if (load_above_capacity > busiest->group_capacity) { load_above_capacity -= busiest->group_capacity; - else + load_above_capacity *= NICE_0_LOAD; + load_above_capacity /= busiest->group_capacity; + } else load_above_capacity = ~0UL; } -- cgit v1.2.3 From 20878232c52329f92423d27a60e48b6a6389e0dd Mon Sep 17 00:00:00 2001 From: Vik Heyndrickx Date: Thu, 28 Apr 2016 20:46:28 +0200 Subject: sched/loadavg: Fix loadavg artifacts on fully idle and on fully loaded systems Systems show a minimal load average of 0.00, 0.01, 0.05 even when they have no load at all. Uptime and /proc/loadavg on all systems with kernels released during the last five years up until kernel version 4.6-rc5, show a 5- and 15-minute minimum loadavg of 0.01 and 0.05 respectively. This should be 0.00 on idle systems, but the way the kernel calculates this value prevents it from getting lower than the mentioned values. Likewise but not as obviously noticeable, a fully loaded system with no processes waiting, shows a maximum 1/5/15 loadavg of 1.00, 0.99, 0.95 (multiplied by number of cores). Once the (old) load becomes 93 or higher, it mathematically can never get lower than 93, even when the active (load) remains 0 forever. This results in the strange 0.00, 0.01, 0.05 uptime values on idle systems. Note: 93/2048 = 0.0454..., which rounds up to 0.05. It is not correct to add a 0.5 rounding (=1024/2048) here, since the result from this function is fed back into the next iteration again, so the result of that +0.5 rounding value then gets multiplied by (2048-2037), and then rounded again, so there is a virtual "ghost" load created, next to the old and active load terms. By changing the way the internally kept value is rounded, that internal value equivalent now can reach 0.00 on idle, and 1.00 on full load. Upon increasing load, the internally kept load value is rounded up, when the load is decreasing, the load value is rounded down. The modified code was tested on nohz=off and nohz kernels. It was tested on vanilla kernel 4.6-rc5 and on centos 7.1 kernel 3.10.0-327. It was tested on single, dual, and octal cores system. It was tested on virtual hosts and bare hardware. No unwanted effects have been observed, and the problems that the patch intended to fix were indeed gone. Tested-by: Damien Wyart Signed-off-by: Vik Heyndrickx Signed-off-by: Peter Zijlstra (Intel) Cc: Cc: Doug Smythies Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Fixes: 0f004f5a696a ("sched: Cure more NO_HZ load average woes") Link: http://lkml.kernel.org/r/e8d32bff-d544-7748-72b5-3c86cc71f09f@veribox.net Signed-off-by: Ingo Molnar --- kernel/sched/loadavg.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/kernel/sched/loadavg.c b/kernel/sched/loadavg.c index ef7159012cf3..b0b93fd33af9 100644 --- a/kernel/sched/loadavg.c +++ b/kernel/sched/loadavg.c @@ -99,10 +99,13 @@ long calc_load_fold_active(struct rq *this_rq) static unsigned long calc_load(unsigned long load, unsigned long exp, unsigned long active) { - load *= exp; - load += active * (FIXED_1 - exp); - load += 1UL << (FSHIFT - 1); - return load >> FSHIFT; + unsigned long newload; + + newload = load * exp + active * (FIXED_1 - exp); + if (active >= load) + newload += FIXED_1-1; + + return newload / FIXED_1; } #ifdef CONFIG_NO_HZ_COMMON -- cgit v1.2.3 From ade42e092b5d1fb9a77b026f019b9953d66f1573 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 11 May 2016 14:23:30 +0200 Subject: sched/core: Use tsk_cpus_allowed() instead of accessing ->cpus_allowed Use the future-safe accessor for struct task_struct's. Signed-off-by: Thomas Gleixner Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: linux-kernel@vger.kernel.org Link: http://lkml.kernel.org/r/1462969411-17735-1-git-send-email-bigeasy@linutronix.de Signed-off-by: Ingo Molnar --- kernel/sched/cpudeadline.c | 4 ++-- kernel/sched/cpupri.c | 4 ++-- kernel/sched/deadline.c | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) diff --git a/kernel/sched/cpudeadline.c b/kernel/sched/cpudeadline.c index 5a75b08cfd85..5be58820465c 100644 --- a/kernel/sched/cpudeadline.c +++ b/kernel/sched/cpudeadline.c @@ -103,10 +103,10 @@ int cpudl_find(struct cpudl *cp, struct task_struct *p, const struct sched_dl_entity *dl_se = &p->dl; if (later_mask && - cpumask_and(later_mask, cp->free_cpus, &p->cpus_allowed)) { + cpumask_and(later_mask, cp->free_cpus, tsk_cpus_allowed(p))) { best_cpu = cpumask_any(later_mask); goto out; - } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && + } else if (cpumask_test_cpu(cpudl_maximum(cp), tsk_cpus_allowed(p)) && dl_time_before(dl_se->deadline, cp->elements[0].dl)) { best_cpu = cpudl_maximum(cp); if (later_mask) diff --git a/kernel/sched/cpupri.c b/kernel/sched/cpupri.c index 981fcd7dc394..11e9705bf937 100644 --- a/kernel/sched/cpupri.c +++ b/kernel/sched/cpupri.c @@ -103,11 +103,11 @@ int cpupri_find(struct cpupri *cp, struct task_struct *p, if (skip) continue; - if (cpumask_any_and(&p->cpus_allowed, vec->mask) >= nr_cpu_ids) + if (cpumask_any_and(tsk_cpus_allowed(p), vec->mask) >= nr_cpu_ids) continue; if (lowest_mask) { - cpumask_and(lowest_mask, &p->cpus_allowed, vec->mask); + cpumask_and(lowest_mask, tsk_cpus_allowed(p), vec->mask); /* * We have to ensure that we have at least one bit diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 0ac6c84f3371..4c9b4eee3f1c 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -1393,7 +1393,7 @@ static struct rq *find_lock_later_rq(struct task_struct *task, struct rq *rq) if (double_lock_balance(rq, later_rq)) { if (unlikely(task_rq(task) != rq || !cpumask_test_cpu(later_rq->cpu, - &task->cpus_allowed) || + tsk_cpus_allowed(task)) || task_running(rq, task) || !dl_task(task) || !task_on_rq_queued(task))) { -- cgit v1.2.3 From 50605ffbdaf6d7ccab70d4631fd8347fc78af14f Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 11 May 2016 14:23:31 +0200 Subject: sched/core: Provide a tsk_nr_cpus_allowed() helper tsk_nr_cpus_allowed() is an accessor for task->nr_cpus_allowed which allows us to change the representation of ->nr_cpus_allowed if required. Signed-off-by: Thomas Gleixner Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: linux-kernel@vger.kernel.org Link: http://lkml.kernel.org/r/1462969411-17735-2-git-send-email-bigeasy@linutronix.de Signed-off-by: Ingo Molnar --- include/linux/sched.h | 5 +++++ kernel/sched/core.c | 2 +- kernel/sched/deadline.c | 28 ++++++++++++++-------------- kernel/sched/rt.c | 24 ++++++++++++------------ 4 files changed, 32 insertions(+), 27 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index f8fea8a5b5ab..38526b67e787 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1930,6 +1930,11 @@ extern int arch_task_struct_size __read_mostly; /* Future-safe accessor for struct task_struct's cpus_allowed. */ #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed) +static inline int tsk_nr_cpus_allowed(struct task_struct *p) +{ + return p->nr_cpus_allowed; +} + #define TNF_MIGRATED 0x01 #define TNF_NO_GROUP 0x02 #define TNF_SHARED 0x04 diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 6f6962a356b4..404c0784b1fc 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1585,7 +1585,7 @@ int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags) { lockdep_assert_held(&p->pi_lock); - if (p->nr_cpus_allowed > 1) + if (tsk_nr_cpus_allowed(p) > 1) cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags); else cpu = cpumask_any(tsk_cpus_allowed(p)); diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 4c9b4eee3f1c..fcb7f0217ff4 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -134,7 +134,7 @@ static void inc_dl_migration(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq) { struct task_struct *p = dl_task_of(dl_se); - if (p->nr_cpus_allowed > 1) + if (tsk_nr_cpus_allowed(p) > 1) dl_rq->dl_nr_migratory++; update_dl_migration(dl_rq); @@ -144,7 +144,7 @@ static void dec_dl_migration(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq) { struct task_struct *p = dl_task_of(dl_se); - if (p->nr_cpus_allowed > 1) + if (tsk_nr_cpus_allowed(p) > 1) dl_rq->dl_nr_migratory--; update_dl_migration(dl_rq); @@ -966,7 +966,7 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags) enqueue_dl_entity(&p->dl, pi_se, flags); - if (!task_current(rq, p) && p->nr_cpus_allowed > 1) + if (!task_current(rq, p) && tsk_nr_cpus_allowed(p) > 1) enqueue_pushable_dl_task(rq, p); } @@ -1040,9 +1040,9 @@ select_task_rq_dl(struct task_struct *p, int cpu, int sd_flag, int flags) * try to make it stay here, it might be important. */ if (unlikely(dl_task(curr)) && - (curr->nr_cpus_allowed < 2 || + (tsk_nr_cpus_allowed(curr) < 2 || !dl_entity_preempt(&p->dl, &curr->dl)) && - (p->nr_cpus_allowed > 1)) { + (tsk_nr_cpus_allowed(p) > 1)) { int target = find_later_rq(p); if (target != -1 && @@ -1063,7 +1063,7 @@ static void check_preempt_equal_dl(struct rq *rq, struct task_struct *p) * Current can't be migrated, useless to reschedule, * let's hope p can move out. */ - if (rq->curr->nr_cpus_allowed == 1 || + if (tsk_nr_cpus_allowed(rq->curr) == 1 || cpudl_find(&rq->rd->cpudl, rq->curr, NULL) == -1) return; @@ -1071,7 +1071,7 @@ static void check_preempt_equal_dl(struct rq *rq, struct task_struct *p) * p is migratable, so let's not schedule it and * see if it is pushed or pulled somewhere else. */ - if (p->nr_cpus_allowed != 1 && + if (tsk_nr_cpus_allowed(p) != 1 && cpudl_find(&rq->rd->cpudl, p, NULL) != -1) return; @@ -1186,7 +1186,7 @@ static void put_prev_task_dl(struct rq *rq, struct task_struct *p) { update_curr_dl(rq); - if (on_dl_rq(&p->dl) && p->nr_cpus_allowed > 1) + if (on_dl_rq(&p->dl) && tsk_nr_cpus_allowed(p) > 1) enqueue_pushable_dl_task(rq, p); } @@ -1287,7 +1287,7 @@ static int find_later_rq(struct task_struct *task) if (unlikely(!later_mask)) return -1; - if (task->nr_cpus_allowed == 1) + if (tsk_nr_cpus_allowed(task) == 1) return -1; /* @@ -1433,7 +1433,7 @@ static struct task_struct *pick_next_pushable_dl_task(struct rq *rq) BUG_ON(rq->cpu != task_cpu(p)); BUG_ON(task_current(rq, p)); - BUG_ON(p->nr_cpus_allowed <= 1); + BUG_ON(tsk_nr_cpus_allowed(p) <= 1); BUG_ON(!task_on_rq_queued(p)); BUG_ON(!dl_task(p)); @@ -1472,7 +1472,7 @@ retry: */ if (dl_task(rq->curr) && dl_time_before(next_task->dl.deadline, rq->curr->dl.deadline) && - rq->curr->nr_cpus_allowed > 1) { + tsk_nr_cpus_allowed(rq->curr) > 1) { resched_curr(rq); return 0; } @@ -1619,9 +1619,9 @@ static void task_woken_dl(struct rq *rq, struct task_struct *p) { if (!task_running(rq, p) && !test_tsk_need_resched(rq->curr) && - p->nr_cpus_allowed > 1 && + tsk_nr_cpus_allowed(p) > 1 && dl_task(rq->curr) && - (rq->curr->nr_cpus_allowed < 2 || + (tsk_nr_cpus_allowed(rq->curr) < 2 || !dl_entity_preempt(&p->dl, &rq->curr->dl))) { push_dl_tasks(rq); } @@ -1725,7 +1725,7 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p) if (task_on_rq_queued(p) && rq->curr != p) { #ifdef CONFIG_SMP - if (p->nr_cpus_allowed > 1 && rq->dl.overloaded) + if (tsk_nr_cpus_allowed(p) > 1 && rq->dl.overloaded) queue_push_tasks(rq); #else if (dl_task(rq->curr)) diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index 67afa06cc8bc..d5690b722691 100644 --- a/kernel/sched/rt.c +++ b/kernel/sched/rt.c @@ -334,7 +334,7 @@ static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) rt_rq = &rq_of_rt_rq(rt_rq)->rt; rt_rq->rt_nr_total++; - if (p->nr_cpus_allowed > 1) + if (tsk_nr_cpus_allowed(p) > 1) rt_rq->rt_nr_migratory++; update_rt_migration(rt_rq); @@ -351,7 +351,7 @@ static void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) rt_rq = &rq_of_rt_rq(rt_rq)->rt; rt_rq->rt_nr_total--; - if (p->nr_cpus_allowed > 1) + if (tsk_nr_cpus_allowed(p) > 1) rt_rq->rt_nr_migratory--; update_rt_migration(rt_rq); @@ -1324,7 +1324,7 @@ enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags) enqueue_rt_entity(rt_se, flags); - if (!task_current(rq, p) && p->nr_cpus_allowed > 1) + if (!task_current(rq, p) && tsk_nr_cpus_allowed(p) > 1) enqueue_pushable_task(rq, p); } @@ -1413,7 +1413,7 @@ select_task_rq_rt(struct task_struct *p, int cpu, int sd_flag, int flags) * will have to sort it out. */ if (curr && unlikely(rt_task(curr)) && - (curr->nr_cpus_allowed < 2 || + (tsk_nr_cpus_allowed(curr) < 2 || curr->prio <= p->prio)) { int target = find_lowest_rq(p); @@ -1437,7 +1437,7 @@ static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p) * Current can't be migrated, useless to reschedule, * let's hope p can move out. */ - if (rq->curr->nr_cpus_allowed == 1 || + if (tsk_nr_cpus_allowed(rq->curr) == 1 || !cpupri_find(&rq->rd->cpupri, rq->curr, NULL)) return; @@ -1445,7 +1445,7 @@ static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p) * p is migratable, so let's not schedule it and * see if it is pushed or pulled somewhere else. */ - if (p->nr_cpus_allowed != 1 + if (tsk_nr_cpus_allowed(p) != 1 && cpupri_find(&rq->rd->cpupri, p, NULL)) return; @@ -1579,7 +1579,7 @@ static void put_prev_task_rt(struct rq *rq, struct task_struct *p) * The previous task needs to be made eligible for pushing * if it is still active */ - if (on_rt_rq(&p->rt) && p->nr_cpus_allowed > 1) + if (on_rt_rq(&p->rt) && tsk_nr_cpus_allowed(p) > 1) enqueue_pushable_task(rq, p); } @@ -1629,7 +1629,7 @@ static int find_lowest_rq(struct task_struct *task) if (unlikely(!lowest_mask)) return -1; - if (task->nr_cpus_allowed == 1) + if (tsk_nr_cpus_allowed(task) == 1) return -1; /* No other targets possible */ if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask)) @@ -1762,7 +1762,7 @@ static struct task_struct *pick_next_pushable_task(struct rq *rq) BUG_ON(rq->cpu != task_cpu(p)); BUG_ON(task_current(rq, p)); - BUG_ON(p->nr_cpus_allowed <= 1); + BUG_ON(tsk_nr_cpus_allowed(p) <= 1); BUG_ON(!task_on_rq_queued(p)); BUG_ON(!rt_task(p)); @@ -2122,9 +2122,9 @@ static void task_woken_rt(struct rq *rq, struct task_struct *p) { if (!task_running(rq, p) && !test_tsk_need_resched(rq->curr) && - p->nr_cpus_allowed > 1 && + tsk_nr_cpus_allowed(p) > 1 && (dl_task(rq->curr) || rt_task(rq->curr)) && - (rq->curr->nr_cpus_allowed < 2 || + (tsk_nr_cpus_allowed(rq->curr) < 2 || rq->curr->prio <= p->prio)) push_rt_tasks(rq); } @@ -2197,7 +2197,7 @@ static void switched_to_rt(struct rq *rq, struct task_struct *p) */ if (task_on_rq_queued(p) && rq->curr != p) { #ifdef CONFIG_SMP - if (p->nr_cpus_allowed > 1 && rq->rt.overloaded) + if (tsk_nr_cpus_allowed(p) > 1 && rq->rt.overloaded) queue_push_tasks(rq); #else if (p->prio < rq->curr->prio) -- cgit v1.2.3 From ef0491ea17f8019821c7e9c8e801184ecf17f85a Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Fri, 13 May 2016 15:30:13 +0200 Subject: ARM: Hide finish_arch_post_lock_switch() from modules The introduction of switch_mm_irqs_off() brought back an old bug regarding the use of preempt_enable_no_resched: As part of: 62b94a08da1b ("sched/preempt: Take away preempt_enable_no_resched() from modules") the definition of preempt_enable_no_resched() is only available in built-in code, not in loadable modules, so we can't generally use it from header files. However, the ARM version of finish_arch_post_lock_switch() calls preempt_enable_no_resched() and is defined as a static inline function in asm/mmu_context.h. This in turn means we cannot include asm/mmu_context.h from modules. With today's tip tree, asm/mmu_context.h gets included from linux/mmu_context.h, which is normally the exact pattern one would expect, but unfortunately, linux/mmu_context.h can be included from the vhost driver that is a loadable module, now causing this compile time error with modular configs: In file included from ../include/linux/mmu_context.h:4:0, from ../drivers/vhost/vhost.c:18: ../arch/arm/include/asm/mmu_context.h: In function 'finish_arch_post_lock_switch': ../arch/arm/include/asm/mmu_context.h:88:3: error: implicit declaration of function 'preempt_enable_no_resched' [-Werror=implicit-function-declaration] preempt_enable_no_resched(); Andy already tried to fix the bug by including linux/preempt.h from asm/mmu_context.h, but that didn't help. Arnd suggested reordering the header files, which wasn't popular, so let's use this workaround instead: The finish_arch_post_lock_switch() definition is now also hidden inside of #ifdef MODULE, so we don't see anything referencing preempt_enable_no_resched() from a header file. I've built a few hundred randconfig kernels with this, and did not see any new problems. Tested-by: Guenter Roeck Signed-off-by: Steven Rostedt Signed-off-by: Arnd Bergmann Acked-by: Russell King Cc: Alexander Shishkin Cc: Andy Lutomirski Cc: Andy Lutomirski Cc: Ard Biesheuvel Cc: Arnaldo Carvalho de Melo Cc: Borislav Petkov Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Linus Torvalds Cc: Mel Gorman Cc: Peter Zijlstra Cc: Russell King - ARM Linux Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: linux-arm-kernel@lists.infradead.org Fixes: f98db6013c55 ("sched/core: Add switch_mm_irqs_off() and use it in the scheduler") Link: http://lkml.kernel.org/r/1463146234-161304-1-git-send-email-arnd@arndb.de Signed-off-by: Ingo Molnar --- arch/arm/include/asm/mmu_context.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm/include/asm/mmu_context.h b/arch/arm/include/asm/mmu_context.h index ed73babc0dc9..3cc14dd8587c 100644 --- a/arch/arm/include/asm/mmu_context.h +++ b/arch/arm/include/asm/mmu_context.h @@ -67,6 +67,7 @@ static inline void check_and_switch_context(struct mm_struct *mm, cpu_switch_mm(mm->pgd, mm); } +#ifndef MODULE #define finish_arch_post_lock_switch \ finish_arch_post_lock_switch static inline void finish_arch_post_lock_switch(void) @@ -88,6 +89,7 @@ static inline void finish_arch_post_lock_switch(void) preempt_enable_no_resched(); } } +#endif /* !MODULE */ #endif /* CONFIG_MMU */ -- cgit v1.2.3