diff options
Diffstat (limited to 'cacule-5.11.patch')
-rw-r--r-- | cacule-5.11.patch | 587 |
1 files changed, 355 insertions, 232 deletions
diff --git a/cacule-5.11.patch b/cacule-5.11.patch index 231ee8476dc4..7f6ccd2f9422 100644 --- a/cacule-5.11.patch +++ b/cacule-5.11.patch @@ -5,12 +5,12 @@ index 1d56a6b73a4e..4d55ff02310c 100644 @@ -1087,6 +1087,10 @@ Model available). If your platform happens to meet the requirements for EAS but you do not want to use it, change this value to 0. - + +sched_interactivity_factor (CacULE scheduler only) +================================================== +Sets the value *m* for interactivity score calculations. See +Figure 1 in https://web.cs.ucdavis.edu/~roper/ecs150/ULE.pdf - + sched_schedstats ================ diff --git a/Documentation/scheduler/sched-CacULE.rst b/Documentation/scheduler/sched-CacULE.rst @@ -96,13 +96,13 @@ index 000000000000..82b0847c468a + idle timer scheduler in order to avoid to get into priority + inversion problems which would deadlock the machine. diff --git a/include/linux/sched.h b/include/linux/sched.h -index 6e3a5eeec509..97103c8a91b2 100644 +index 6e3a5eeec509..f5a4fc49286f 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -450,10 +450,22 @@ struct sched_statistics { #endif }; - + +#ifdef CONFIG_CACULE_SCHED +struct cacule_node { + struct cacule_node* next; @@ -121,19 +121,34 @@ index 6e3a5eeec509..97103c8a91b2 100644 +#endif struct list_head group_node; unsigned int on_rq; - + +@@ -872,6 +884,12 @@ struct task_struct { + struct list_head sibling; + struct task_struct *group_leader; + ++#ifdef CONFIG_CACULE_SCHED ++ u64 fork_start_win_stamp; ++ unsigned int nr_forks_per_time; ++ int is_fake_interactive; ++#endif ++ + /* + * 'ptraced' is the list of tasks this task is using ptrace() on. + * diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h -index 3c31ba88aca5..cb819c3d86f3 100644 +index 3c31ba88aca5..20c85c808485 100644 --- a/include/linux/sched/sysctl.h +++ b/include/linux/sched/sysctl.h -@@ -31,6 +31,12 @@ extern unsigned int sysctl_sched_min_granularity; +@@ -31,6 +31,14 @@ extern unsigned int sysctl_sched_min_granularity; extern unsigned int sysctl_sched_wakeup_granularity; extern unsigned int sysctl_sched_child_runs_first; - + +#ifdef CONFIG_CACULE_SCHED -+extern int interactivity_factor; ++extern unsigned int interactivity_factor; +extern unsigned int interactivity_threshold; -+extern int cacule_max_lifetime; ++extern unsigned int cacule_max_lifetime; ++extern unsigned int fake_interactive_decay_time; ++extern unsigned int nr_fork_threshold; +#endif + enum sched_tunable_scaling { @@ -144,9 +159,9 @@ index a3d27421de8f..d0cfdf6e9bed 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -824,6 +824,17 @@ config UCLAMP_BUCKETS_COUNT - + endmenu - + +config CACULE_SCHED + bool "CacULE CPU scheduler" + default y @@ -176,43 +191,66 @@ index 38ef6d06888e..c8cf984c294e 100644 @@ -46,6 +46,9 @@ choice 1000 Hz is the preferred choice for desktop systems and other systems requiring fast interactive responses to events. - + + config HZ_2000 + bool "2000 HZ" + endchoice - + config HZ @@ -54,6 +57,7 @@ config HZ default 250 if HZ_250 default 300 if HZ_300 default 1000 if HZ_1000 + default 2000 if HZ_2000 - + config SCHED_HRTICK def_bool HIGH_RES_TIMERS +diff --git a/kernel/exit.c b/kernel/exit.c +index 04029e35e69a..9dfd515104db 100644 +--- a/kernel/exit.c ++++ b/kernel/exit.c +@@ -667,6 +667,17 @@ static void exit_notify(struct task_struct *tsk, int group_dead) + write_lock_irq(&tasklist_lock); + forget_original_parent(tsk, &dead); + ++#ifdef CONFIG_CACULE_SCHED ++ p = tsk->parent; ++ if (p) { ++ if (p->nr_forks_per_time) ++ p->nr_forks_per_time--; ++ ++ if (p->is_fake_interactive) ++ p->is_fake_interactive--; ++ } ++#endif ++ + if (group_dead) + kill_orphaned_pgrp(tsk->group_leader, NULL); + diff --git a/kernel/sched/core.c b/kernel/sched/core.c -index f0056507a373..4d8a3b232ae9 100644 +index f0056507a373..9ecce53ddcc1 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c -@@ -3554,7 +3554,13 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) - p->se.sum_exec_runtime = 0; +@@ -3555,6 +3555,14 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) p->se.prev_sum_exec_runtime = 0; p->se.nr_migrations = 0; + p->se.vruntime = 0; + +#ifdef CONFIG_CACULE_SCHED + p->se.cacule_node.vruntime = 0; -+#else - p->se.vruntime = 0; ++ p->fork_start_win_stamp = 0; ++ p->nr_forks_per_time = 0; ++ p->is_fake_interactive = 0; +#endif + INIT_LIST_HEAD(&p->se.group_node); - + #ifdef CONFIG_FAIR_GROUP_SCHED -@@ -3840,6 +3846,10 @@ void wake_up_new_task(struct task_struct *p) +@@ -3840,6 +3848,10 @@ void wake_up_new_task(struct task_struct *p) update_rq_clock(rq); post_init_entity_util_avg(p); - + +#ifdef CONFIG_CACULE_SCHED + p->se.cacule_node.cacule_start_time = sched_clock(); +#endif @@ -220,22 +258,36 @@ index f0056507a373..4d8a3b232ae9 100644 activate_task(rq, p, ENQUEUE_NOCLOCK); trace_sched_wakeup_new(p); check_preempt_curr(rq, p, WF_FORK); -@@ -7727,6 +7737,10 @@ void __init sched_init(void) +@@ -7727,6 +7739,10 @@ void __init sched_init(void) BUG_ON(&dl_sched_class + 1 != &stop_sched_class); #endif - + +#ifdef CONFIG_CACULE_SCHED -+ printk(KERN_INFO "CacULE CPU scheduler v5.11 by Hamad Al Marri."); ++ printk(KERN_INFO "CacULE CPU scheduler v5.11-r2 by Hamad Al Marri."); +#endif + wait_bit_init(); - + #ifdef CONFIG_FAIR_GROUP_SCHED diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c -index 2357921580f9..ac08a7ced508 100644 +index 2357921580f9..3adc9ee2bcfc 100644 --- a/kernel/sched/debug.c +++ b/kernel/sched/debug.c -@@ -557,21 +557,27 @@ void print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq) +@@ -535,8 +535,11 @@ static void print_rq(struct seq_file *m, struct rq *rq, int rq_cpu) + + void print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq) + { +- s64 MIN_vruntime = -1, min_vruntime, max_vruntime = -1, +- spread, rq0_min_vruntime, spread0; ++ s64 MIN_vruntime = -1, max_vruntime = -1, ++#if !defined(CONFIG_CACULE_SCHED) ++ min_vruntime, rq0_min_vruntime, spread0, ++#endif ++ spread; + struct rq *rq = cpu_rq(cpu); + struct sched_entity *last; + unsigned long flags; +@@ -557,21 +560,27 @@ void print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq) last = __pick_last_entity(cfs_rq); if (last) max_vruntime = last->vruntime; @@ -264,7 +316,7 @@ index 2357921580f9..ac08a7ced508 100644 cfs_rq->nr_spread_over); SEQ_printf(m, " .%-30s: %d\n", "nr_running", cfs_rq->nr_running); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c -index bbc78794224a..056b5673605f 100644 +index bbc78794224a..9e035ed1e746 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -19,6 +19,10 @@ @@ -277,23 +329,25 @@ index bbc78794224a..056b5673605f 100644 + * (C) 2020 Hamad Al Marri <hamad.s.almarri@gmail.com> */ #include "sched.h" - -@@ -113,6 +117,11 @@ int __weak arch_asym_cpu_priority(int cpu) + +@@ -113,6 +117,13 @@ int __weak arch_asym_cpu_priority(int cpu) */ #define fits_capacity(cap, max) ((cap) * 1280 < (max) * 1024) - + +#endif +#ifdef CONFIG_CACULE_SCHED -+int __read_mostly cacule_max_lifetime = 22000; // in ms -+int __read_mostly interactivity_factor = 32768; -+unsigned int __read_mostly interactivity_threshold = 20480; ++unsigned int __read_mostly cacule_max_lifetime = 22000; // in ms ++unsigned int __read_mostly interactivity_factor = 32768; ++unsigned int __read_mostly interactivity_threshold = 1000; ++unsigned int __read_mostly fake_interactive_decay_time = 1000; // in ms ++unsigned int __read_mostly nr_fork_threshold = 3; #endif - + #ifdef CONFIG_CFS_BANDWIDTH -@@ -253,6 +262,14 @@ static u64 __calc_delta(u64 delta_exec, unsigned long weight, struct load_weight - +@@ -253,6 +264,14 @@ static u64 __calc_delta(u64 delta_exec, unsigned long weight, struct load_weight + const struct sched_class fair_sched_class; - + + +#ifdef CONFIG_CACULE_SCHED +static inline struct sched_entity *se_of(struct cacule_node *cn) @@ -305,7 +359,7 @@ index bbc78794224a..056b5673605f 100644 /************************************************************** * CFS operations on generic schedulable entities: */ -@@ -512,7 +529,7 @@ void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec); +@@ -512,7 +531,7 @@ void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec); /************************************************************** * Scheduling class tree data structure manipulation methods: */ @@ -314,35 +368,75 @@ index bbc78794224a..056b5673605f 100644 static inline u64 max_vruntime(u64 max_vruntime, u64 vruntime) { s64 delta = (s64)(vruntime - max_vruntime); -@@ -568,7 +585,169 @@ static void update_min_vruntime(struct cfs_rq *cfs_rq) +@@ -568,7 +587,183 @@ static void update_min_vruntime(struct cfs_rq *cfs_rq) cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime; #endif } +#endif /* CONFIG_CACULE_SCHED */ + +#ifdef CONFIG_CACULE_SCHED ++static inline unsigned int is_fake_interactive(struct cacule_node *cn) ++{ ++ struct sched_entity *se = se_of(cn); ++ struct task_struct *parent = NULL; ++ struct cfs_rq *cfs_rq; ++ u64 win_time = fake_interactive_decay_time * 1000000ULL; ++ u64 now = sched_clock(); ++ ++ while (!parent) { ++ if (entity_is_task(se)) { ++ parent = task_of(se)->parent; ++ break; ++ } ++ ++ cfs_rq = group_cfs_rq(se); ++ ++ if (!cfs_rq->head && !cfs_rq->curr) ++ return 0; ++ ++ if (cfs_rq->head) ++ se = se_of(cfs_rq->head); ++ else if (cfs_rq->curr) ++ se = cfs_rq->curr; ++ } ++ ++ if (parent->is_fake_interactive ++ && (now - parent->fork_start_win_stamp > win_time)) ++ { ++ parent->fork_start_win_stamp = now; ++ parent->is_fake_interactive--; ++ } ++ ++ return parent->is_fake_interactive; ++} ++ +static unsigned int +calc_interactivity(u64 now, struct cacule_node *se) +{ -+ u64 l_se, vr_se, sleep_se = 1ULL, u64_factor; -+ unsigned int score_se; ++ u64 l_se, vr_se, sleep_se = 1ULL, u64_factor_m, _2m; ++ unsigned int score_se, fake_interactivity; + + /* + * in case of vruntime==0, logical OR with 1 would + * make sure that the least sig. bit is 1 + */ + l_se = now - se->cacule_start_time; -+ vr_se = se->vruntime | 1; -+ u64_factor = interactivity_factor; ++ vr_se = se->vruntime | 1; ++ u64_factor_m = interactivity_factor; ++ _2m = u64_factor_m << 1; + + /* safety check */ + if (likely(l_se > vr_se)) + sleep_se = (l_se - vr_se) | 1; + + if (sleep_se >= vr_se) -+ score_se = u64_factor / (sleep_se / vr_se); ++ score_se = u64_factor_m / (sleep_se / vr_se); + else -+ score_se = (u64_factor << 1) - (u64_factor / (vr_se / sleep_se)); ++ score_se = _2m - (u64_factor_m / (vr_se / sleep_se)); ++ ++ fake_interactivity = is_fake_interactive(se); ++ if (fake_interactivity) ++ score_se += (_2m * fake_interactivity) + 1; + + return score_se; +} @@ -352,6 +446,9 @@ index bbc78794224a..056b5673605f 100644 + if (se_of(cn)->vruntime == 0) + return 0; + ++ if (is_fake_interactive(cn)) ++ return 0; ++ + return calc_interactivity(sched_clock(), cn) < interactivity_threshold; +} + @@ -369,7 +466,7 @@ index bbc78794224a..056b5673605f 100644 + + return -1; +} -+ + +/* + * Does se have lower interactivity score value (i.e. interactive) than curr? If yes, return 1, + * otherwise return -1 @@ -399,47 +496,18 @@ index bbc78794224a..056b5673605f 100644 +static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *_se) +{ + struct cacule_node *se = &(_se->cacule_node); -+ struct cacule_node *iter, *next = NULL; -+ u64 now = sched_clock(); -+ unsigned int score_se = calc_interactivity(now, se); + + se->next = NULL; + se->prev = NULL; + + if (likely(cfs_rq->head)) { -+ -+ // start from tail -+ iter = cfs_rq->tail; -+ -+ // does se have higher IS than iter? -+ while (iter && entity_before_cached(now, score_se, iter) == -1) { -+ next = iter; -+ iter = iter->prev; -+ } -+ -+ // se in tail position -+ if (iter == cfs_rq->tail) { -+ cfs_rq->tail->next = se; -+ se->prev = cfs_rq->tail; -+ -+ cfs_rq->tail = se; -+ } -+ // else if not head no tail, insert se after iter -+ else if (iter) { -+ se->next = next; -+ se->prev = iter; -+ -+ iter->next = se; -+ next->prev = se; -+ } + // insert se at head -+ else { -+ se->next = cfs_rq->head; -+ cfs_rq->head->prev = se; ++ se->next = cfs_rq->head; ++ cfs_rq->head->prev = se; ++ ++ // lastly reset the head ++ cfs_rq->head = se; + -+ // lastly reset the head -+ cfs_rq->head = se; -+ } + } else { + // if empty rq + cfs_rq->head = se; @@ -470,7 +538,7 @@ index bbc78794224a..056b5673605f 100644 + struct cacule_node *next = se->next; + + prev->next = next; - ++ + if (next) + next->prev = prev; + } @@ -484,12 +552,12 @@ index bbc78794224a..056b5673605f 100644 /* * Enqueue an entity into the rb-tree: */ -@@ -626,16 +805,29 @@ static struct sched_entity *__pick_next_entity(struct sched_entity *se) - +@@ -626,16 +821,29 @@ static struct sched_entity *__pick_next_entity(struct sched_entity *se) + return rb_entry(next, struct sched_entity, run_node); } +#endif /* CONFIG_CACULE_SCHED */ - + #ifdef CONFIG_SCHED_DEBUG struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq) { @@ -505,35 +573,35 @@ index bbc78794224a..056b5673605f 100644 + return se_of(cn); +#else struct rb_node *last = rb_last(&cfs_rq->tasks_timeline.rb_root); - + if (!last) return NULL; - + return rb_entry(last, struct sched_entity, run_node); +#endif /* CONFIG_CACULE_SCHED */ } - + /************************************************************** -@@ -720,6 +912,7 @@ static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se) +@@ -720,6 +928,7 @@ static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se) return slice; } - + +#if !defined(CONFIG_CACULE_SCHED) /* * We calculate the vruntime slice of a to-be-inserted task. * -@@ -729,6 +922,7 @@ static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se) +@@ -729,6 +938,7 @@ static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se) { return calc_delta_fair(sched_slice(cfs_rq, se), se); } +#endif /* CONFIG_CACULE_SCHED */ - + #include "pelt.h" #ifdef CONFIG_SMP -@@ -836,14 +1030,46 @@ static void update_tg_load_avg(struct cfs_rq *cfs_rq) +@@ -836,14 +1046,46 @@ static void update_tg_load_avg(struct cfs_rq *cfs_rq) } #endif /* CONFIG_SMP */ - + +#ifdef CONFIG_CACULE_SCHED +static void normalize_lifetime(u64 now, struct sched_entity *se) +{ @@ -576,13 +644,13 @@ index bbc78794224a..056b5673605f 100644 - u64 delta_exec; + u64 now = sched_clock(); + u64 delta_exec, delta_fair; - + if (unlikely(!curr)) return; -@@ -860,8 +1086,15 @@ static void update_curr(struct cfs_rq *cfs_rq) +@@ -860,8 +1102,15 @@ static void update_curr(struct cfs_rq *cfs_rq) curr->sum_exec_runtime += delta_exec; schedstat_add(cfs_rq->exec_clock, delta_exec); - + +#ifdef CONFIG_CACULE_SCHED + delta_fair = calc_delta_fair(delta_exec, curr); + curr->vruntime += delta_fair; @@ -592,52 +660,52 @@ index bbc78794224a..056b5673605f 100644 curr->vruntime += calc_delta_fair(delta_exec, curr); update_min_vruntime(cfs_rq); +#endif - + if (entity_is_task(curr)) { struct task_struct *curtask = task_of(curr); -@@ -1029,7 +1262,6 @@ update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) +@@ -1029,7 +1278,6 @@ update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) static inline void update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { - if (!schedstat_enabled()) return; - -@@ -1061,7 +1293,7 @@ update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se) + +@@ -1061,7 +1309,7 @@ update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se) /* * We are starting a new run period: */ - se->exec_start = rq_clock_task(rq_of(cfs_rq)); + se->exec_start = sched_clock(); } - + /************************************************** -@@ -4115,7 +4347,7 @@ static inline void update_misfit_status(struct task_struct *p, struct rq *rq) {} - +@@ -4115,7 +4363,7 @@ static inline void update_misfit_status(struct task_struct *p, struct rq *rq) {} + static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se) { -#ifdef CONFIG_SCHED_DEBUG +#if defined(CONFIG_SCHED_DEBUG) && !defined(CONFIG_CACULE_SCHED) s64 d = se->vruntime - cfs_rq->min_vruntime; - + if (d < 0) -@@ -4126,6 +4358,7 @@ static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se) +@@ -4126,6 +4374,7 @@ static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se) #endif } - + +#if !defined(CONFIG_CACULE_SCHED) static void place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) { -@@ -4157,6 +4390,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) +@@ -4157,6 +4406,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) /* ensure we never gain time by being placed backwards. */ se->vruntime = max_vruntime(se->vruntime, vruntime); } +#endif /* CONFIG_CACULE_SCHED */ - + static void check_enqueue_throttle(struct cfs_rq *cfs_rq); - -@@ -4215,18 +4449,23 @@ static inline bool cfs_bandwidth_used(void); + +@@ -4215,18 +4465,23 @@ static inline bool cfs_bandwidth_used(void); static void enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { @@ -645,7 +713,7 @@ index bbc78794224a..056b5673605f 100644 bool renorm = !(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATED); +#endif bool curr = cfs_rq->curr == se; - + +#if !defined(CONFIG_CACULE_SCHED) /* * If we're the current task, we must renormalise before calling @@ -654,126 +722,141 @@ index bbc78794224a..056b5673605f 100644 if (renorm && curr) se->vruntime += cfs_rq->min_vruntime; +#endif - + update_curr(cfs_rq); - + +#if !defined(CONFIG_CACULE_SCHED) /* * Otherwise, renormalise after, such that we're placed at the current * moment in time, instead of some random moment in the past. Being -@@ -4235,6 +4474,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) +@@ -4235,6 +4490,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) */ if (renorm && !curr) se->vruntime += cfs_rq->min_vruntime; +#endif - + /* * When enqueuing a sched_entity, we must: -@@ -4249,8 +4489,10 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) +@@ -4249,8 +4505,10 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) update_cfs_group(se); account_entity_enqueue(cfs_rq, se); - + +#if !defined(CONFIG_CACULE_SCHED) if (flags & ENQUEUE_WAKEUP) place_entity(cfs_rq, se, 0); +#endif - + check_schedstat_required(); update_stats_enqueue(cfs_rq, se, flags); -@@ -4271,6 +4513,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) +@@ -4271,6 +4529,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) check_enqueue_throttle(cfs_rq); } - + +#if !defined(CONFIG_CACULE_SCHED) static void __clear_buddies_last(struct sched_entity *se) { for_each_sched_entity(se) { -@@ -4315,6 +4558,7 @@ static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se) +@@ -4315,6 +4574,7 @@ static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se) if (cfs_rq->skip == se) __clear_buddies_skip(se); } +#endif /* !CONFIG_CACULE_SCHED */ - + static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq); - -@@ -4339,13 +4583,16 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) - + +@@ -4339,13 +4599,16 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) + update_stats_dequeue(cfs_rq, se, flags); - + +#if !defined(CONFIG_CACULE_SCHED) clear_buddies(cfs_rq, se); +#endif - + if (se != cfs_rq->curr) __dequeue_entity(cfs_rq, se); se->on_rq = 0; account_entity_dequeue(cfs_rq, se); - + +#if !defined(CONFIG_CACULE_SCHED) /* * Normalize after update_curr(); which will also have moved * min_vruntime if @se is the one holding it back. But before doing -@@ -4354,12 +4601,14 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) +@@ -4354,12 +4617,14 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) */ if (!(flags & DEQUEUE_SLEEP)) se->vruntime -= cfs_rq->min_vruntime; +#endif - + /* return excess runtime on last dequeue */ return_cfs_rq_runtime(cfs_rq); - + update_cfs_group(se); - + +#if !defined(CONFIG_CACULE_SCHED) /* * Now advance min_vruntime if @se was the entity holding it back, * except when: DEQUEUE_SAVE && !DEQUEUE_MOVE, in this case we'll be -@@ -4368,8 +4617,23 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) +@@ -4368,8 +4633,24 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) */ if ((flags & (DEQUEUE_SAVE | DEQUEUE_MOVE)) != DEQUEUE_SAVE) update_min_vruntime(cfs_rq); +#endif } - + +#ifdef CONFIG_CACULE_SCHED ++static struct sched_entity * ++pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr); ++ +/* + * Preempt the current task with a newly woken task if needed: + */ +static void +check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr) +{ -+ u64 now = sched_clock(); -+ + // does head have higher IS than curr -+ if (entity_before(now, &curr->cacule_node, cfs_rq->head) == 1) ++ if (pick_next_entity(cfs_rq, curr) != curr) + resched_curr(rq_of(cfs_rq)); +} +#else /* * Preempt the current task with a newly woken task if needed: */ -@@ -4409,6 +4673,7 @@ check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr) +@@ -4409,6 +4690,7 @@ check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr) if (delta > ideal_runtime) resched_curr(rq_of(cfs_rq)); } +#endif /* CONFIG_CACULE_SCHED */ - + static void set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) -@@ -4443,6 +4708,21 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) +@@ -4443,6 +4725,35 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) se->prev_sum_exec_runtime = se->sum_exec_runtime; } - + +#ifdef CONFIG_CACULE_SCHED +static struct sched_entity * +pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr) +{ -+ struct cacule_node *se = cfs_rq->head; ++ struct cacule_node *next, *se = cfs_rq->head; ++ u64 now = sched_clock(); ++ unsigned int score_se; + + if (unlikely(!se)) -+ se = &curr->cacule_node; -+ else if (unlikely(curr -+ && entity_before(sched_clock(), se, &curr->cacule_node) == 1)) ++ return curr; ++ ++ score_se = calc_interactivity(now, se); ++ ++ next = se->next; ++ while (next) { ++ if (entity_before_cached(now, score_se, next) == 1) { ++ se = next; ++ score_se = calc_interactivity(now, se); ++ } ++ ++ next = next->next; ++ } ++ ++ if (unlikely(curr && entity_before_cached(now, score_se, &curr->cacule_node) == 1)) + se = &curr->cacule_node; + + return se_of(se); @@ -781,26 +864,44 @@ index bbc78794224a..056b5673605f 100644 +#else static int wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se); - -@@ -4503,6 +4783,7 @@ pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr) - + +@@ -4503,6 +4814,7 @@ pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr) + return se; } +#endif /* CONFIG_CACULE_SCHED */ - + static bool check_cfs_rq_runtime(struct cfs_rq *cfs_rq); - -@@ -5605,7 +5886,9 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) + +@@ -5605,7 +5917,9 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) hrtick_update(rq); } - + +#if !defined(CONFIG_CACULE_SCHED) static void set_next_buddy(struct sched_entity *se); +#endif - + /* * The dequeue_task method is called before nr_running is -@@ -5637,12 +5920,14 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) +@@ -5619,6 +5933,17 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) + int task_sleep = flags & DEQUEUE_SLEEP; + int idle_h_nr_running = task_has_idle_policy(p); + bool was_sched_idle = sched_idle_rq(rq); ++#ifdef CONFIG_CACULE_SCHED ++ struct task_struct *parent = p->parent; ++ ++ if (task_sleep && parent) { ++ if (parent->nr_forks_per_time) ++ parent->nr_forks_per_time--; ++ ++ if (parent->is_fake_interactive) ++ parent->is_fake_interactive--; ++ } ++#endif + + util_est_dequeue(&rq->cfs, p); + +@@ -5637,12 +5962,14 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) if (cfs_rq->load.weight) { /* Avoid re-evaluating load for this entity: */ se = parent_entity(se); @@ -815,32 +916,32 @@ index bbc78794224a..056b5673605f 100644 break; } flags |= DEQUEUE_SLEEP; -@@ -5758,6 +6043,7 @@ static unsigned long capacity_of(int cpu) +@@ -5758,6 +6085,7 @@ static unsigned long capacity_of(int cpu) return cpu_rq(cpu)->cpu_capacity; } - + +#if !defined(CONFIG_CACULE_SCHED) static void record_wakee(struct task_struct *p) { /* -@@ -5804,6 +6090,7 @@ static int wake_wide(struct task_struct *p) +@@ -5804,6 +6132,7 @@ static int wake_wide(struct task_struct *p) return 0; return 1; } -+#endif - ++#endif /* CONFIG_CACULE_SCHED */ + /* * The purpose of wake_affine() is to quickly determine on which CPU we can run -@@ -6484,6 +6771,7 @@ static unsigned long cpu_util_without(int cpu, struct task_struct *p) +@@ -6484,6 +6813,7 @@ static unsigned long cpu_util_without(int cpu, struct task_struct *p) return min_t(unsigned long, util, capacity_orig_of(cpu)); } - + +#if !defined(CONFIG_CACULE_SCHED) /* * Predicts what cpu_util(@cpu) would return if @p was migrated (and enqueued) * to @dst_cpu. -@@ -6717,6 +7005,57 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu) - +@@ -6717,6 +7047,57 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu) + return -1; } +#endif /* CONFIG_CACULE_SCHED */ @@ -894,13 +995,13 @@ index bbc78794224a..056b5673605f 100644 + return new_cpu; +} +#endif - + /* * select_task_rq_fair: Select target runqueue for the waking task in domains -@@ -6741,6 +7080,26 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) +@@ -6741,6 +7122,26 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) /* SD_flags and WF_flags share the first nibble */ int sd_flag = wake_flags & 0xF; - + +#ifdef CONFIG_CACULE_SCHED + struct sched_entity *se = &p->se; + @@ -915,7 +1016,7 @@ index bbc78794224a..056b5673605f 100644 + + new_cpu = find_least_IS_cpu(p); + -+ if (likely(new_cpu != -1)) ++ if (new_cpu != -1) + return new_cpu; + + new_cpu = prev_cpu; @@ -923,16 +1024,16 @@ index bbc78794224a..056b5673605f 100644 +#else if (wake_flags & WF_TTWU) { record_wakee(p); - -@@ -6753,6 +7112,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) - + +@@ -6753,6 +7154,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) + want_affine = !wake_wide(p) && cpumask_test_cpu(cpu, p->cpus_ptr); } +#endif /* CONFIG_CACULE_SCHED */ - + rcu_read_lock(); for_each_domain(cpu, tmp) { -@@ -6799,6 +7159,7 @@ static void detach_entity_cfs_rq(struct sched_entity *se); +@@ -6799,6 +7201,7 @@ static void detach_entity_cfs_rq(struct sched_entity *se); */ static void migrate_task_rq_fair(struct task_struct *p, int new_cpu) { @@ -940,31 +1041,31 @@ index bbc78794224a..056b5673605f 100644 /* * As blocked tasks retain absolute vruntime the migration needs to * deal with this by subtracting the old and adding the new -@@ -6824,6 +7185,7 @@ static void migrate_task_rq_fair(struct task_struct *p, int new_cpu) - +@@ -6824,6 +7227,7 @@ static void migrate_task_rq_fair(struct task_struct *p, int new_cpu) + se->vruntime -= min_vruntime; } +#endif /* CONFIG_CACULE_SCHED */ - + if (p->on_rq == TASK_ON_RQ_MIGRATING) { /* -@@ -6869,6 +7231,7 @@ balance_fair(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) +@@ -6869,6 +7273,7 @@ balance_fair(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) } #endif /* CONFIG_SMP */ - + +#if !defined(CONFIG_CACULE_SCHED) static unsigned long wakeup_gran(struct sched_entity *se) { unsigned long gran = sysctl_sched_wakeup_granularity; -@@ -6947,6 +7310,7 @@ static void set_skip_buddy(struct sched_entity *se) +@@ -6947,6 +7352,7 @@ static void set_skip_buddy(struct sched_entity *se) for_each_sched_entity(se) cfs_rq_of(se)->skip = se; } +#endif /* CONFIG_CACULE_SCHED */ - + /* * Preempt the current task with a newly woken task if needed: -@@ -6955,9 +7319,12 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ +@@ -6955,9 +7361,12 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ { struct task_struct *curr = rq->curr; struct sched_entity *se = &curr->se, *pse = &p->se; @@ -974,23 +1075,23 @@ index bbc78794224a..056b5673605f 100644 int scale = cfs_rq->nr_running >= sched_nr_latency; int next_buddy_marked = 0; +#endif /* CONFIG_CACULE_SCHED */ - + if (unlikely(se == pse)) return; -@@ -6971,10 +7338,12 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ +@@ -6971,10 +7380,12 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ if (unlikely(throttled_hierarchy(cfs_rq_of(pse)))) return; - + +#if !defined(CONFIG_CACULE_SCHED) if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) { set_next_buddy(pse); next_buddy_marked = 1; } +#endif /* CONFIG_CACULE_SCHED */ - + /* * We can come here with TIF_NEED_RESCHED already set from new task -@@ -7004,6 +7373,11 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ +@@ -7004,6 +7415,11 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ find_matching_se(&se, &pse); update_curr(cfs_rq_of(se)); BUG_ON(!pse); @@ -1002,14 +1103,14 @@ index bbc78794224a..056b5673605f 100644 if (wakeup_preempt_entity(se, pse) == 1) { /* * Bias pick_next to pick the sched entity that is -@@ -7013,11 +7387,14 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ +@@ -7013,11 +7429,14 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ set_next_buddy(pse); goto preempt; } +#endif /* CONFIG_CACULE_SCHED */ - + return; - + preempt: resched_curr(rq); + @@ -1017,15 +1118,15 @@ index bbc78794224a..056b5673605f 100644 /* * Only set the backward buddy when the current task is still * on the rq. This can happen when a wakeup gets interleaved -@@ -7032,6 +7409,7 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ - +@@ -7032,6 +7451,7 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ + if (sched_feat(LAST_BUDDY) && scale && entity_is_task(se)) set_last_buddy(se); +#endif /* CONFIG_CACULE_SCHED */ } - + struct task_struct * -@@ -7206,7 +7584,10 @@ static void yield_task_fair(struct rq *rq) +@@ -7206,7 +7626,10 @@ static void yield_task_fair(struct rq *rq) { struct task_struct *curr = rq->curr; struct cfs_rq *cfs_rq = task_cfs_rq(curr); @@ -1033,60 +1134,60 @@ index bbc78794224a..056b5673605f 100644 +#if !defined(CONFIG_CACULE_SCHED) struct sched_entity *se = &curr->se; +#endif - + /* * Are we the only task in the tree? -@@ -7214,7 +7595,9 @@ static void yield_task_fair(struct rq *rq) +@@ -7214,7 +7637,9 @@ static void yield_task_fair(struct rq *rq) if (unlikely(rq->nr_running == 1)) return; - + +#if !defined(CONFIG_CACULE_SCHED) clear_buddies(cfs_rq, se); +#endif - + if (curr->policy != SCHED_BATCH) { update_rq_clock(rq); -@@ -7230,7 +7613,9 @@ static void yield_task_fair(struct rq *rq) +@@ -7230,7 +7655,9 @@ static void yield_task_fair(struct rq *rq) rq_clock_skip_update(rq); } - + +#if !defined(CONFIG_CACULE_SCHED) set_skip_buddy(se); +#endif } - + static bool yield_to_task_fair(struct rq *rq, struct task_struct *p) -@@ -7241,8 +7626,10 @@ static bool yield_to_task_fair(struct rq *rq, struct task_struct *p) +@@ -7241,8 +7668,10 @@ static bool yield_to_task_fair(struct rq *rq, struct task_struct *p) if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se))) return false; - + +#if !defined(CONFIG_CACULE_SCHED) /* Tell the scheduler that we'd really like pse to run next. */ set_next_buddy(se); +#endif - + yield_task_fair(rq); - -@@ -7470,6 +7857,7 @@ static int task_hot(struct task_struct *p, struct lb_env *env) + +@@ -7470,6 +7899,7 @@ static int task_hot(struct task_struct *p, struct lb_env *env) if (env->sd->flags & SD_SHARE_CPUCAPACITY) return 0; - + +#if !defined(CONFIG_CACULE_SCHED) /* * Buddy candidates are cache hot: */ -@@ -7477,6 +7865,7 @@ static int task_hot(struct task_struct *p, struct lb_env *env) +@@ -7477,6 +7907,7 @@ static int task_hot(struct task_struct *p, struct lb_env *env) (&p->se == cfs_rq_of(&p->se)->next || &p->se == cfs_rq_of(&p->se)->last)) return 1; +#endif - + if (sysctl_sched_migration_cost == -1) return 1; -@@ -10765,11 +11154,30 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued) +@@ -10765,11 +11196,38 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued) update_overutilized_status(task_rq(curr)); } - + +#ifdef CONFIG_CACULE_SCHED /* * called on fork with the child task as argument from the parent's context @@ -1099,6 +1200,8 @@ index bbc78794224a..056b5673605f 100644 + struct sched_entity *curr; + struct rq *rq = this_rq(); + struct rq_flags rf; ++ struct task_struct *parent = p->parent; ++ u64 now = sched_clock(); + + rq_lock(rq, &rf); + update_rq_clock(rq); @@ -1109,37 +1212,43 @@ index bbc78794224a..056b5673605f 100644 + update_curr(cfs_rq); + + rq_unlock(rq, &rf); ++ ++ parent->fork_start_win_stamp = now; ++ if (parent->nr_forks_per_time >= nr_fork_threshold) ++ parent->is_fake_interactive++; ++ ++ parent->nr_forks_per_time++; +} +#else static void task_fork_fair(struct task_struct *p) { struct cfs_rq *cfs_rq; -@@ -10800,6 +11208,7 @@ static void task_fork_fair(struct task_struct *p) +@@ -10800,6 +11258,7 @@ static void task_fork_fair(struct task_struct *p) se->vruntime -= cfs_rq->min_vruntime; rq_unlock(rq, &rf); } +#endif /* CONFIG_CACULE_SCHED */ - + /* * Priority of the task has changed. Check to see if we preempt -@@ -10912,6 +11321,8 @@ static void attach_entity_cfs_rq(struct sched_entity *se) +@@ -10912,6 +11371,8 @@ static void attach_entity_cfs_rq(struct sched_entity *se) static void detach_task_cfs_rq(struct task_struct *p) { struct sched_entity *se = &p->se; + +#if !defined(CONFIG_CACULE_SCHED) struct cfs_rq *cfs_rq = cfs_rq_of(se); - + if (!vruntime_normalized(p)) { -@@ -10922,6 +11333,7 @@ static void detach_task_cfs_rq(struct task_struct *p) +@@ -10922,6 +11383,7 @@ static void detach_task_cfs_rq(struct task_struct *p) place_entity(cfs_rq, se, 0); se->vruntime -= cfs_rq->min_vruntime; } +#endif - + detach_entity_cfs_rq(se); } -@@ -10929,12 +11341,17 @@ static void detach_task_cfs_rq(struct task_struct *p) +@@ -10929,12 +11391,17 @@ static void detach_task_cfs_rq(struct task_struct *p) static void attach_task_cfs_rq(struct task_struct *p) { struct sched_entity *se = &p->se; @@ -1147,17 +1256,17 @@ index bbc78794224a..056b5673605f 100644 +#if !defined(CONFIG_CACULE_SCHED) struct cfs_rq *cfs_rq = cfs_rq_of(se); +#endif - + attach_entity_cfs_rq(se); - + +#if !defined(CONFIG_CACULE_SCHED) if (!vruntime_normalized(p)) se->vruntime += cfs_rq->min_vruntime; +#endif } - + static void switched_from_fair(struct rq *rq, struct task_struct *p) -@@ -10990,13 +11407,22 @@ static void set_next_task_fair(struct rq *rq, struct task_struct *p, bool first) +@@ -10990,13 +11457,22 @@ static void set_next_task_fair(struct rq *rq, struct task_struct *p, bool first) void init_cfs_rq(struct cfs_rq *cfs_rq) { cfs_rq->tasks_timeline = RB_ROOT_CACHED; @@ -1178,7 +1287,7 @@ index bbc78794224a..056b5673605f 100644 + cfs_rq->tail = NULL; +#endif } - + #ifdef CONFIG_FAIR_GROUP_SCHED diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 282a6bbaacd7..a3b7316dd537 100644 @@ -1186,7 +1295,7 @@ index 282a6bbaacd7..a3b7316dd537 100644 +++ b/kernel/sched/sched.h @@ -516,10 +516,13 @@ struct cfs_rq { unsigned int idle_h_nr_running; /* SCHED_IDLE */ - + u64 exec_clock; + +#if !defined(CONFIG_CACULE_SCHED) @@ -1195,9 +1304,9 @@ index 282a6bbaacd7..a3b7316dd537 100644 u64 min_vruntime_copy; #endif +#endif /* CONFIG_CACULE_SCHED */ - + struct rb_root_cached tasks_timeline; - + @@ -528,9 +531,15 @@ struct cfs_rq { * It is set to NULL otherwise (i.e when none are currently running). */ @@ -1211,14 +1320,14 @@ index 282a6bbaacd7..a3b7316dd537 100644 struct sched_entity *last; struct sched_entity *skip; +#endif // CONFIG_CACULE_SCHED - + #ifdef CONFIG_SCHED_DEBUG unsigned int nr_spread_over; diff --git a/kernel/sysctl.c b/kernel/sysctl.c -index 62fbd09b5dc1..a0bf55bbb3a7 100644 +index 62fbd09b5dc1..5f10c51fa7df 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c -@@ -1659,6 +1659,29 @@ static struct ctl_table kern_table[] = { +@@ -1659,6 +1659,43 @@ static struct ctl_table kern_table[] = { .mode = 0644, .proc_handler = proc_dointvec, }, @@ -1226,7 +1335,7 @@ index 62fbd09b5dc1..a0bf55bbb3a7 100644 + { + .procname = "sched_interactivity_factor", + .data = &interactivity_factor, -+ .maxlen = sizeof(int), ++ .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, @@ -1240,7 +1349,21 @@ index 62fbd09b5dc1..a0bf55bbb3a7 100644 + { + .procname = "sched_max_lifetime_ms", + .data = &cacule_max_lifetime, -+ .maxlen = sizeof(int), ++ .maxlen = sizeof(unsigned int), ++ .mode = 0644, ++ .proc_handler = proc_dointvec, ++ }, ++ { ++ .procname = "sched_fake_interactive_decay_time_ms", ++ .data = &fake_interactive_decay_time, ++ .maxlen = sizeof(unsigned int), ++ .mode = 0644, ++ .proc_handler = proc_dointvec, ++ }, ++ { ++ .procname = "sched_nr_fork_threshold", ++ .data = &nr_fork_threshold, ++ .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, |