--- Documentation/sched-design.txt | 234 ++++++ Documentation/sysctl/kernel.txt | 30 fs/pipe.c | 7 fs/proc/array.c | 2 include/linux/init_task.h | 4 include/linux/sched.h | 32 kernel/kthread.c | 1 kernel/sched.c | 1345 +++++++++++++++++++--------------------- kernel/softirq.c | 2 kernel/sysctl.c | 35 - kernel/workqueue.c | 2 11 files changed, 978 insertions(+), 716 deletions(-) Index: linux-2.6.22-rc2-ck1/include/linux/sched.h =================================================================== --- linux-2.6.22-rc2-ck1.orig/include/linux/sched.h 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/include/linux/sched.h 2007-05-19 20:11:57.000000000 +1000 @@ -150,8 +150,7 @@ extern unsigned long weighted_cpuload(co #define EXIT_ZOMBIE 16 #define EXIT_DEAD 32 /* in tsk->state again */ -#define TASK_NONINTERACTIVE 64 -#define TASK_DEAD 128 +#define TASK_DEAD 64 #define __set_task_state(tsk, state_value) \ do { (tsk)->state = (state_value); } while (0) @@ -537,8 +536,9 @@ struct signal_struct { #define MAX_USER_RT_PRIO 100 #define MAX_RT_PRIO MAX_USER_RT_PRIO +#define PRIO_RANGE (40) -#define MAX_PRIO (MAX_RT_PRIO + 40) +#define MAX_PRIO (MAX_RT_PRIO + PRIO_RANGE) #define rt_prio(prio) unlikely((prio) < MAX_RT_PRIO) #define rt_task(p) rt_prio((p)->prio) @@ -809,13 +809,6 @@ struct mempolicy; struct pipe_inode_info; struct uts_namespace; -enum sleep_type { - SLEEP_NORMAL, - SLEEP_NONINTERACTIVE, - SLEEP_INTERACTIVE, - SLEEP_INTERRUPTED, -}; - struct prio_array; struct task_struct { @@ -835,20 +828,33 @@ struct task_struct { int load_weight; /* for niceness load balancing purposes */ int prio, static_prio, normal_prio; struct list_head run_list; + /* + * This bitmap shows what priorities this task has received quota + * from for this major priority rotation on its current runqueue. + */ + DECLARE_BITMAP(bitmap, PRIO_RANGE + 1); struct prio_array *array; + /* Which major runqueue rotation did this task run */ + unsigned long rotation; unsigned short ioprio; #ifdef CONFIG_BLK_DEV_IO_TRACE unsigned int btrace_seq; #endif - unsigned long sleep_avg; unsigned long long timestamp, last_ran; unsigned long long sched_time; /* sched_clock time spent running */ - enum sleep_type sleep_type; unsigned int policy; cpumask_t cpus_allowed; - unsigned int time_slice, first_time_slice; + /* + * How much this task is entitled to run at the current priority + * before being requeued at a lower priority. + */ + int time_slice; + /* Is this the very first time_slice this task has ever run. */ + unsigned int first_time_slice; + /* How much this task receives at each priority level */ + int quota; #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) struct sched_info sched_info; Index: linux-2.6.22-rc2-ck1/kernel/sched.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/kernel/sched.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/kernel/sched.c 2007-05-19 20:12:08.000000000 +1000 @@ -16,6 +16,7 @@ * by Davide Libenzi, preemptible kernel bits by Robert Love. * 2003-09-03 Interactivity tuning by Con Kolivas. * 2004-04-02 Scheduler domains code by Nick Piggin + * 2007-03-02 Staircase deadline scheduling policy by Con Kolivas */ #include @@ -53,8 +54,9 @@ #include #include #include - +#include #include + #include /* @@ -84,147 +86,73 @@ unsigned long long __attribute__((weak)) #define USER_PRIO(p) ((p)-MAX_RT_PRIO) #define TASK_USER_PRIO(p) USER_PRIO((p)->static_prio) #define MAX_USER_PRIO (USER_PRIO(MAX_PRIO)) +#define SCHED_PRIO(p) ((p)+MAX_RT_PRIO) -/* - * Some helpers for converting nanosecond timing to jiffy resolution - */ -#define NS_TO_JIFFIES(TIME) ((TIME) / (1000000000 / HZ)) +/* Some helpers for converting to/from various scales.*/ #define JIFFIES_TO_NS(TIME) ((TIME) * (1000000000 / HZ)) - -/* - * These are the 'tuning knobs' of the scheduler: - * - * Minimum timeslice is 5 msecs (or 1 jiffy, whichever is larger), - * default timeslice is 100 msecs, maximum timeslice is 800 msecs. - * Timeslices get refilled after they expire. - */ -#define MIN_TIMESLICE max(5 * HZ / 1000, 1) -#define DEF_TIMESLICE (100 * HZ / 1000) -#define ON_RUNQUEUE_WEIGHT 30 -#define CHILD_PENALTY 95 -#define PARENT_PENALTY 100 -#define EXIT_WEIGHT 3 -#define PRIO_BONUS_RATIO 25 -#define MAX_BONUS (MAX_USER_PRIO * PRIO_BONUS_RATIO / 100) -#define INTERACTIVE_DELTA 2 -#define MAX_SLEEP_AVG (DEF_TIMESLICE * MAX_BONUS) -#define STARVATION_LIMIT (MAX_SLEEP_AVG) -#define NS_MAX_SLEEP_AVG (JIFFIES_TO_NS(MAX_SLEEP_AVG)) - -/* - * If a task is 'interactive' then we reinsert it in the active - * array after it has expired its current timeslice. (it will not - * continue to run immediately, it will still roundrobin with - * other interactive tasks.) - * - * This part scales the interactivity limit depending on niceness. - * - * We scale it linearly, offset by the INTERACTIVE_DELTA delta. - * Here are a few examples of different nice levels: - * - * TASK_INTERACTIVE(-20): [1,1,1,1,1,1,1,1,1,0,0] - * TASK_INTERACTIVE(-10): [1,1,1,1,1,1,1,0,0,0,0] - * TASK_INTERACTIVE( 0): [1,1,1,1,0,0,0,0,0,0,0] - * TASK_INTERACTIVE( 10): [1,1,0,0,0,0,0,0,0,0,0] - * TASK_INTERACTIVE( 19): [0,0,0,0,0,0,0,0,0,0,0] - * - * (the X axis represents the possible -5 ... 0 ... +5 dynamic - * priority range a task can explore, a value of '1' means the - * task is rated interactive.) - * - * Ie. nice +19 tasks can never get 'interactive' enough to be - * reinserted into the active array. And only heavily CPU-hog nice -20 - * tasks will be expired. Default nice 0 tasks are somewhere between, - * it takes some effort for them to get interactive, but it's not - * too hard. - */ - -#define CURRENT_BONUS(p) \ - (NS_TO_JIFFIES((p)->sleep_avg) * MAX_BONUS / \ - MAX_SLEEP_AVG) - -#define GRANULARITY (10 * HZ / 1000 ? : 1) - -#ifdef CONFIG_SMP -#define TIMESLICE_GRANULARITY(p) (GRANULARITY * \ - (1 << (((MAX_BONUS - CURRENT_BONUS(p)) ? : 1) - 1)) * \ - num_online_cpus()) -#else -#define TIMESLICE_GRANULARITY(p) (GRANULARITY * \ - (1 << (((MAX_BONUS - CURRENT_BONUS(p)) ? : 1) - 1))) -#endif - -#define SCALE(v1,v1_max,v2_max) \ - (v1) * (v2_max) / (v1_max) - -#define DELTA(p) \ - (SCALE(TASK_NICE(p) + 20, 40, MAX_BONUS) - 20 * MAX_BONUS / 40 + \ - INTERACTIVE_DELTA) - -#define TASK_INTERACTIVE(p) \ - ((p)->prio <= (p)->static_prio - DELTA(p)) - -#define INTERACTIVE_SLEEP(p) \ - (JIFFIES_TO_NS(MAX_SLEEP_AVG * \ - (MAX_BONUS / 2 + DELTA((p)) + 1) / MAX_BONUS - 1)) - -#define TASK_PREEMPTS_CURR(p, rq) \ - ((p)->prio < (rq)->curr->prio) - -#define SCALE_PRIO(x, prio) \ - max(x * (MAX_PRIO - prio) / (MAX_USER_PRIO / 2), MIN_TIMESLICE) - -static unsigned int static_prio_timeslice(int static_prio) -{ - if (static_prio < NICE_TO_PRIO(0)) - return SCALE_PRIO(DEF_TIMESLICE * 4, static_prio); - else - return SCALE_PRIO(DEF_TIMESLICE, static_prio); -} - -#ifdef CONFIG_SMP -/* - * Divide a load by a sched group cpu_power : (load / sg->__cpu_power) - * Since cpu_power is a 'constant', we can use a reciprocal divide. +#define MS_TO_NS(TIME) ((TIME) * 1000000) +#define MS_TO_US(TIME) ((TIME) * 1000) +#define US_TO_MS(TIME) ((TIME) / 1000) + +#define TASK_PREEMPTS_CURR(p, curr) ((p)->prio < (curr)->prio) + +/* + * This is the time all tasks within the same priority round robin. + * Value is in ms and set to a minimum of 10ms. Scales with number of cpus. + * Tunable via /proc interface. + */ +int rr_interval __read_mostly = 10; +int sched_interactive __read_mostly = 1; + +/* + * This contains a bitmap for each dynamic priority level with empty slots + * for the valid priorities each different nice level can have. It allows + * us to stagger the slots where differing priorities run in a way that + * keeps latency differences between different nice levels at a minimum. + * The purpose of a pre-generated matrix is for rapid lookup of next slot in + * O(1) time without having to recalculate every time priority gets demoted. + * All nice levels use priority slot 39 as this allows less niced tasks to + * get all priority slots better than that before expiration is forced. + * ie, where 0 means a slot for that priority, priority running from left to + * right is from prio 0 to prio 39: + * nice -20 0000000000000000000000000000000000000000 + * nice -10 1000100010001000100010001000100010010000 + * nice 0 1010101010101010101010101010101010101010 + * nice 5 1011010110110101101101011011010110110110 + * nice 10 1110111011101110111011101110111011101110 + * nice 15 1111111011111110111111101111111011111110 + * nice 19 1111111111111111111111111111111111111110 */ -static inline u32 sg_div_cpu_power(const struct sched_group *sg, u32 load) -{ - return reciprocal_divide(load, sg->reciprocal_cpu_power); -} +static unsigned long prio_matrix[PRIO_RANGE][BITS_TO_LONGS(PRIO_RANGE)] + __read_mostly; -/* - * Each time a sched group cpu_power is changed, - * we must compute its reciprocal value - */ -static inline void sg_inc_cpu_power(struct sched_group *sg, u32 val) -{ - sg->__cpu_power += val; - sg->reciprocal_cpu_power = reciprocal_value(sg->__cpu_power); -} -#endif - -/* - * task_timeslice() scales user-nice values [ -20 ... 0 ... 19 ] - * to time slice values: [800ms ... 100ms ... 5ms] - * - * The higher a thread's priority, the bigger timeslices - * it gets during one round of execution. But even the lowest - * priority thread gets MIN_TIMESLICE worth of execution time. - */ - -static inline unsigned int task_timeslice(struct task_struct *p) -{ - return static_prio_timeslice(p->static_prio); -} +struct rq; /* * These are the runqueue data structures: */ - struct prio_array { - unsigned int nr_active; - DECLARE_BITMAP(bitmap, MAX_PRIO+1); /* include 1 bit for delimiter */ + /* Tasks queued at each priority */ struct list_head queue[MAX_PRIO]; + + /* + * The bitmap of priorities queued for this array. While the expired + * array will never have realtime tasks on it, it is simpler to have + * equal sized bitmaps for a cheap array swap. Include 1 bit for + * delimiter. + */ + DECLARE_BITMAP(prio_bitmap, MAX_PRIO + 1); + + /* + * The best static priority (of the dynamic priority tasks) queued + * this array. + */ + int best_static_prio; + +#ifdef CONFIG_SMP + /* For convenience looks back at rq */ + struct rq *rq; +#endif }; /* @@ -260,14 +188,24 @@ struct rq { */ unsigned long nr_uninterruptible; - unsigned long expired_timestamp; /* Cached timestamp set by update_cpu_clock() */ unsigned long long most_recent_timestamp; struct task_struct *curr, *idle; unsigned long next_balance; struct mm_struct *prev_mm; + struct prio_array *active, *expired, arrays[2]; - int best_expired_prio; + unsigned long *dyn_bitmap, *exp_bitmap; + + /* + * The current dynamic priority level this runqueue is at per static + * priority level. + */ + int prio_level[PRIO_RANGE]; + + /* How many times we have rotated the priority queue */ + unsigned long prio_rotation; + atomic_t nr_iowait; #ifdef CONFIG_SMP @@ -606,12 +544,9 @@ static inline struct rq *this_rq_lock(vo #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) /* * Called when a process is dequeued from the active array and given - * the cpu. We should note that with the exception of interactive - * tasks, the expired queue will become the active queue after the active - * queue is empty, without explicitly dequeuing and requeuing tasks in the - * expired queue. (Interactive tasks may be requeued directly to the - * active queue, thus delaying tasks in the expired queue from running; - * see scheduler_tick()). + * the cpu. We should note that the expired queue will become the active + * queue after the active queue is empty, without explicitly dequeuing and + * requeuing tasks in the expired queue. * * This function is only called from sched_info_arrive(), rather than * dequeue_task(). Even though a task may be queued and dequeued multiple @@ -709,71 +644,239 @@ sched_info_switch(struct task_struct *pr #define sched_info_switch(t, next) do { } while (0) #endif /* CONFIG_SCHEDSTATS || CONFIG_TASK_DELAY_ACCT */ +static inline int task_queued(struct task_struct *task) +{ + return !list_empty(&task->run_list); +} + +static inline void set_dynamic_bit(struct task_struct *p, struct rq *rq) +{ + __set_bit(p->prio, p->array->prio_bitmap); +} + /* - * Adding/removing a task to/from a priority array: + * Removing from a runqueue. */ -static void dequeue_task(struct task_struct *p, struct prio_array *array) +static void dequeue_task(struct task_struct *p, struct rq *rq) { - array->nr_active--; - list_del(&p->run_list); - if (list_empty(array->queue + p->prio)) - __clear_bit(p->prio, array->bitmap); + list_del_init(&p->run_list); + if (list_empty(p->array->queue + p->prio)) + __clear_bit(p->prio, p->array->prio_bitmap); } -static void enqueue_task(struct task_struct *p, struct prio_array *array) +static void reset_first_time_slice(struct task_struct *p) { - sched_info_queued(p); - list_add_tail(&p->run_list, array->queue + p->prio); - __set_bit(p->prio, array->bitmap); - array->nr_active++; + if (unlikely(p->first_time_slice)) + p->first_time_slice = 0; +} + +/* + * The task is being queued on a fresh array so it has its entitlement + * bitmap cleared. + */ +static void task_new_array(struct task_struct *p, struct rq *rq, + struct prio_array *array) +{ + bitmap_zero(p->bitmap, PRIO_RANGE); + p->rotation = rq->prio_rotation; + p->time_slice = p->quota; p->array = array; + reset_first_time_slice(p); +} + +/* Find the first slot from the relevant prio_matrix entry */ +static int first_prio_slot(struct task_struct *p) +{ + if (unlikely(p->policy == SCHED_BATCH)) + return p->static_prio; + return SCHED_PRIO(find_first_zero_bit( + prio_matrix[USER_PRIO(p->static_prio)], PRIO_RANGE)); } /* - * Put task to the end of the run list without the overhead of dequeue - * followed by enqueue. + * In sched_interactive mode priority allocation occurs per process per rq + * array swap. In !sched_interactive mode all waking tasks must obey the + * current prio level of all other tasks running per array swap. */ -static void requeue_task(struct task_struct *p, struct prio_array *array) +static int minprio(struct rq *rq, int uprio) { - list_move_tail(&p->run_list, array->queue + p->prio); + if (sched_interactive) + return MAX_RT_PRIO; + return rq->prio_level[uprio]; } -static inline void -enqueue_task_head(struct task_struct *p, struct prio_array *array) +/* + * Find the first unused slot by this task that is also in its prio_matrix + * level. SCHED_BATCH tasks do not use the priority matrix. They only take + * priority slots from their static_prio and above. + */ +static int next_entitled_slot(struct task_struct *p, struct rq *rq) { - list_add(&p->run_list, array->queue + p->prio); - __set_bit(p->prio, array->bitmap); - array->nr_active++; - p->array = array; + int search_prio = MAX_RT_PRIO, uprio = USER_PRIO(p->static_prio); + struct prio_array *array = rq->active; + DECLARE_BITMAP(tmp, PRIO_RANGE); + + /* + * Go straight to expiration if there are higher priority tasks + * already expired. + */ + if (p->static_prio > rq->expired->best_static_prio) + return MAX_PRIO; + if (!rq->prio_level[uprio]) + rq->prio_level[uprio] = MAX_RT_PRIO; + /* + * Only priorities equal to the prio_level and above for their + * static_prio are acceptable, and only if it's not better than + * a queued better static_prio's prio_level. + */ + if (p->static_prio < array->best_static_prio) { + if (likely(p->policy != SCHED_BATCH)) + array->best_static_prio = p->static_prio; + } else if (p->static_prio == array->best_static_prio) { + search_prio = minprio(rq, uprio); + } else { + int i; + + search_prio = minprio(rq, uprio); + /* A bound O(n) function, worst case n is 40 */ + for (i = array->best_static_prio; i <= p->static_prio ; i++) { + if (!rq->prio_level[USER_PRIO(i)]) + rq->prio_level[USER_PRIO(i)] = MAX_RT_PRIO; + search_prio = max(search_prio, + rq->prio_level[USER_PRIO(i)]); + } + } + if (unlikely(p->policy == SCHED_BATCH)) { + search_prio = max(search_prio, p->static_prio); + return SCHED_PRIO(find_next_zero_bit(p->bitmap, PRIO_RANGE, + USER_PRIO(search_prio))); + } + bitmap_or(tmp, p->bitmap, prio_matrix[uprio], PRIO_RANGE); + return SCHED_PRIO(find_next_zero_bit(tmp, PRIO_RANGE, + USER_PRIO(search_prio))); +} + +static void queue_expired(struct task_struct *p, struct rq *rq) +{ + task_new_array(p, rq, rq->expired); + p->prio = p->normal_prio = first_prio_slot(p); + if (p->static_prio < rq->expired->best_static_prio) + rq->expired->best_static_prio = p->static_prio; + reset_first_time_slice(p); } +#ifdef CONFIG_SMP /* - * __normal_prio - return the priority that is based on the static - * priority but is modified by bonuses/penalties. - * - * We scale the actual sleep average [0 .... MAX_SLEEP_AVG] - * into the -5 ... 0 ... +5 bonus/penalty range. - * - * We use 25% of the full 0...39 priority range so that: - * - * 1) nice +19 interactive tasks do not preempt nice 0 CPU hogs. - * 2) nice -20 CPU hogs do not get preempted by nice 0 tasks. - * - * Both properties are important to certain workloads. + * If we're waking up a task that was previously on a different runqueue, + * update its data appropriately. Note we may be reading data from src_rq-> + * outside of lock, but the occasional inaccurate result should be harmless. + */ + static void update_if_moved(struct task_struct *p, struct rq *rq) +{ + struct rq *src_rq = p->array->rq; + + if (src_rq == rq) + return; + /* + * Only need to set p->array when p->rotation == rq->prio_rotation as + * they will be set in recalc_task_prio when != rq->prio_rotation. + */ + if (p->rotation == src_rq->prio_rotation) { + p->rotation = rq->prio_rotation; + if (p->array == src_rq->expired) + p->array = rq->expired; + else + p->array = rq->active; + } else + p->rotation = 0; +} +#else +static inline void update_if_moved(struct task_struct *p, struct rq *rq) +{ +} +#endif + +/* + * recalc_task_prio determines what priority a non rt_task will be + * queued at. If the task has already been running during this runqueue's + * major rotation (rq->prio_rotation) then it continues at the same + * priority if it has tick entitlement left. If it does not have entitlement + * left, it finds the next priority slot according to its nice value that it + * has not extracted quota from. If it has not run during this major + * rotation, it starts at the next_entitled_slot and has its bitmap quota + * cleared. If it does not have any slots left it has all its slots reset and + * is queued on the expired at its first_prio_slot. */ +static void recalc_task_prio(struct task_struct *p, struct rq *rq) +{ + struct prio_array *array = rq->active; + int queue_prio; + + update_if_moved(p, rq); + if (p->rotation == rq->prio_rotation) { + if (p->array == array) { + if (p->time_slice > 0) + return; + p->time_slice = p->quota; + } else if (p->array == rq->expired) { + queue_expired(p, rq); + return; + } else + task_new_array(p, rq, array); + } else + task_new_array(p, rq, array); + + queue_prio = next_entitled_slot(p, rq); + if (queue_prio >= MAX_PRIO) { + queue_expired(p, rq); + return; + } + p->prio = p->normal_prio = queue_prio; + __set_bit(USER_PRIO(p->prio), p->bitmap); +} -static inline int __normal_prio(struct task_struct *p) +/* + * Adding to a runqueue. The dynamic priority queue that it is added to is + * determined by recalc_task_prio() above. + */ +static inline void __enqueue_task(struct task_struct *p, struct rq *rq) { - int bonus, prio; + if (rt_task(p)) + p->array = rq->active; + else + recalc_task_prio(p, rq); - bonus = CURRENT_BONUS(p) - MAX_BONUS / 2; + sched_info_queued(p); + set_dynamic_bit(p, rq); +} - prio = p->static_prio - bonus; - if (prio < MAX_RT_PRIO) - prio = MAX_RT_PRIO; - if (prio > MAX_PRIO-1) - prio = MAX_PRIO-1; - return prio; +static void enqueue_task(struct task_struct *p, struct rq *rq) +{ + __enqueue_task(p, rq); + list_add_tail(&p->run_list, p->array->queue + p->prio); +} + +static inline void enqueue_task_head(struct task_struct *p, struct rq *rq) +{ + __enqueue_task(p, rq); + list_add(&p->run_list, p->array->queue + p->prio); +} + +/* + * requeue_task is only called when p->static_prio does not change. p->prio + * can change with dynamic tasks. + */ +static void requeue_task(struct task_struct *p, struct rq *rq, + struct prio_array *old_array, int old_prio) +{ + if (p->array == rq->expired) + queue_expired(p, rq); + list_move_tail(&p->run_list, p->array->queue + p->prio); + if (!rt_task(p)) { + if (list_empty(old_array->queue + old_prio)) + __clear_bit(old_prio, old_array->prio_bitmap); + set_dynamic_bit(p, rq); + } } /* @@ -786,17 +889,24 @@ static inline int __normal_prio(struct t */ /* - * Assume: static_prio_timeslice(NICE_TO_PRIO(0)) == DEF_TIMESLICE - * If static_prio_timeslice() is ever changed to break this assumption then - * this code will need modification - */ -#define TIME_SLICE_NICE_ZERO DEF_TIMESLICE -#define LOAD_WEIGHT(lp) \ - (((lp) * SCHED_LOAD_SCALE) / TIME_SLICE_NICE_ZERO) -#define PRIO_TO_LOAD_WEIGHT(prio) \ - LOAD_WEIGHT(static_prio_timeslice(prio)) -#define RTPRIO_TO_LOAD_WEIGHT(rp) \ - (PRIO_TO_LOAD_WEIGHT(MAX_RT_PRIO) + LOAD_WEIGHT(rp)) + * task_timeslice - the total duration a task can run during one major + * rotation. Returns value in milliseconds as the smallest value can be 1. + */ +static int task_timeslice(struct task_struct *p) +{ + int slice = p->quota; /* quota is in us */ + + if (!rt_task(p)) + slice += (PRIO_RANGE - 1 - TASK_USER_PRIO(p)) * slice; + return US_TO_MS(slice); +} + +/* + * The load weight is basically the task_timeslice in ms. Realtime tasks are + * special cased to be proportionately larger than nice -20 by their + * rt_priority. The weight for rt tasks can only be arbitrary at best. + */ +#define RTPRIO_TO_LOAD_WEIGHT(rp) (rr_interval * 20 * (40 + rp)) static void set_load_weight(struct task_struct *p) { @@ -813,7 +923,7 @@ static void set_load_weight(struct task_ #endif p->load_weight = RTPRIO_TO_LOAD_WEIGHT(p->rt_priority); } else - p->load_weight = PRIO_TO_LOAD_WEIGHT(p->static_prio); + p->load_weight = task_timeslice(p); } static inline void @@ -841,28 +951,38 @@ static inline void dec_nr_running(struct } /* - * Calculate the expected normal priority: i.e. priority - * without taking RT-inheritance into account. Might be - * boosted by interactivity modifiers. Changes upon fork, - * setprio syscalls, and whenever the interactivity - * estimator recalculates. + * __activate_task - move a task to the runqueue. */ -static inline int normal_prio(struct task_struct *p) +static inline void __activate_task(struct task_struct *p, struct rq *rq) +{ + enqueue_task(p, rq); + inc_nr_running(p, rq); +} + +/* + * __activate_idle_task - move idle task to the _front_ of runqueue. + */ +static inline void __activate_idle_task(struct task_struct *p, struct rq *rq) { - int prio; + enqueue_task_head(p, rq); + inc_nr_running(p, rq); +} +static inline int normal_prio(struct task_struct *p) +{ if (has_rt_policy(p)) - prio = MAX_RT_PRIO-1 - p->rt_priority; + return MAX_RT_PRIO-1 - p->rt_priority; + /* Other tasks all have normal_prio set in recalc_task_prio */ + if (likely(p->prio >= MAX_RT_PRIO && p->prio < MAX_PRIO)) + return p->prio; else - prio = __normal_prio(p); - return prio; + return p->static_prio; } /* * Calculate the current priority, i.e. the priority * taken into account by the scheduler. This value might - * be boosted by RT tasks, or might be boosted by - * interactivity modifiers. Will be RT if the task got + * be boosted by RT tasks as it will be RT if the task got * RT-boosted. If not then it returns p->normal_prio. */ static int effective_prio(struct task_struct *p) @@ -879,111 +999,41 @@ static int effective_prio(struct task_st } /* - * __activate_task - move a task to the runqueue. + * All tasks have quotas based on rr_interval. RT tasks all get rr_interval. + * From nice 1 to 19 they are smaller than it only if they are at least one + * tick still. Below nice 0 they get progressively larger. + * ie nice -6..0 = rr_interval. nice -10 = 2.5 * rr_interval + * nice -20 = 10 * rr_interval. nice 1-19 = rr_interval / 2. + * Value returned is in microseconds. */ -static void __activate_task(struct task_struct *p, struct rq *rq) +static inline unsigned int rr_quota(struct task_struct *p) { - struct prio_array *target = rq->active; - - if (batch_task(p)) - target = rq->expired; - enqueue_task(p, target); - inc_nr_running(p, rq); -} + int nice = TASK_NICE(p), rr = rr_interval; -/* - * __activate_idle_task - move idle task to the _front_ of runqueue. - */ -static inline void __activate_idle_task(struct task_struct *p, struct rq *rq) -{ - enqueue_task_head(p, rq->active); - inc_nr_running(p, rq); + if (!rt_task(p)) { + if (nice < -6) { + rr *= nice * nice; + rr /= 40; + } else if (nice > 0) + rr = rr / 2 ? : 1; + } + return MS_TO_US(rr); } -/* - * Recalculate p->normal_prio and p->prio after having slept, - * updating the sleep-average too: - */ -static int recalc_task_prio(struct task_struct *p, unsigned long long now) +/* Every time we set the quota we need to set the load weight */ +static void set_quota(struct task_struct *p) { - /* Caller must always ensure 'now >= p->timestamp' */ - unsigned long sleep_time = now - p->timestamp; - - if (batch_task(p)) - sleep_time = 0; - - if (likely(sleep_time > 0)) { - /* - * This ceiling is set to the lowest priority that would allow - * a task to be reinserted into the active array on timeslice - * completion. - */ - unsigned long ceiling = INTERACTIVE_SLEEP(p); - - if (p->mm && sleep_time > ceiling && p->sleep_avg < ceiling) { - /* - * Prevents user tasks from achieving best priority - * with one single large enough sleep. - */ - p->sleep_avg = ceiling; - /* - * Using INTERACTIVE_SLEEP() as a ceiling places a - * nice(0) task 1ms sleep away from promotion, and - * gives it 700ms to round-robin with no chance of - * being demoted. This is more than generous, so - * mark this sleep as non-interactive to prevent the - * on-runqueue bonus logic from intervening should - * this task not receive cpu immediately. - */ - p->sleep_type = SLEEP_NONINTERACTIVE; - } else { - /* - * Tasks waking from uninterruptible sleep are - * limited in their sleep_avg rise as they - * are likely to be waiting on I/O - */ - if (p->sleep_type == SLEEP_NONINTERACTIVE && p->mm) { - if (p->sleep_avg >= ceiling) - sleep_time = 0; - else if (p->sleep_avg + sleep_time >= - ceiling) { - p->sleep_avg = ceiling; - sleep_time = 0; - } - } - - /* - * This code gives a bonus to interactive tasks. - * - * The boost works by updating the 'average sleep time' - * value here, based on ->timestamp. The more time a - * task spends sleeping, the higher the average gets - - * and the higher the priority boost gets as well. - */ - p->sleep_avg += sleep_time; - - } - if (p->sleep_avg > NS_MAX_SLEEP_AVG) - p->sleep_avg = NS_MAX_SLEEP_AVG; - } - - return effective_prio(p); + p->quota = rr_quota(p); + set_load_weight(p); } /* * activate_task - move a task to the runqueue and do priority recalculation - * - * Update all the scheduling statistics stuff. (sleep average - * calculation, priority modifiers, etc.) */ static void activate_task(struct task_struct *p, struct rq *rq, int local) { - unsigned long long now; - - if (rt_task(p)) - goto out; + unsigned long long now = sched_clock(); - now = sched_clock(); #ifdef CONFIG_SMP if (!local) { /* Compensate for drifting sched_clock */ @@ -1004,32 +1054,9 @@ static void activate_task(struct task_st (now - p->timestamp) >> 20); } - p->prio = recalc_task_prio(p, now); - - /* - * This checks to make sure it's not an uninterruptible task - * that is now waking up. - */ - if (p->sleep_type == SLEEP_NORMAL) { - /* - * Tasks which were woken up by interrupts (ie. hw events) - * are most likely of interactive nature. So we give them - * the credit of extending their sleep time to the period - * of time they spend on the runqueue, waiting for execution - * on a CPU, first time around: - */ - if (in_interrupt()) - p->sleep_type = SLEEP_INTERRUPTED; - else { - /* - * Normal first-time wakeups get a credit too for - * on-runqueue time, but it will be weighted down: - */ - p->sleep_type = SLEEP_INTERACTIVE; - } - } + set_quota(p); + p->prio = effective_prio(p); p->timestamp = now; -out: __activate_task(p, rq); } @@ -1039,8 +1066,7 @@ out: static void deactivate_task(struct task_struct *p, struct rq *rq) { dec_nr_running(p, rq); - dequeue_task(p, p->array); - p->array = NULL; + dequeue_task(p, rq); } /* @@ -1133,7 +1159,7 @@ migrate_task(struct task_struct *p, int * If the task is not on a runqueue (and not running), then * it is sufficient to simply update the task's cpu field. */ - if (!p->array && !task_running(rq, p)) { + if (!task_queued(p) && !task_running(rq, p)) { set_task_cpu(p, dest_cpu); return 0; } @@ -1164,7 +1190,7 @@ void wait_task_inactive(struct task_stru repeat: rq = task_rq_lock(p, &flags); /* Must be off runqueue entirely, not preempted. */ - if (unlikely(p->array || task_running(rq, p))) { + if (unlikely(task_queued(p) || task_running(rq, p))) { /* If it's preempted, we yield. It could be a while. */ preempted = !task_running(rq, p); task_rq_unlock(rq, &flags); @@ -1243,6 +1269,25 @@ static inline unsigned long cpu_avg_load } /* + * Divide a load by a sched group cpu_power : (load / sg->__cpu_power) + * Since cpu_power is a 'constant', we can use a reciprocal divide. + */ +static inline u32 sg_div_cpu_power(const struct sched_group *sg, u32 load) +{ + return reciprocal_divide(load, sg->reciprocal_cpu_power); +} + +/* + * Each time a sched group cpu_power is changed, + * we must compute its reciprocal value + */ +static inline void sg_inc_cpu_power(struct sched_group *sg, u32 val) +{ + sg->__cpu_power += val; + sg->reciprocal_cpu_power = reciprocal_value(sg->__cpu_power); +} + +/* * find_idlest_group finds and returns the least busy CPU group within the * domain. */ @@ -1439,6 +1484,31 @@ static inline int wake_idle(int cpu, str } #endif +/* + * We need to have a special definition for an idle runqueue when testing + * for preemption on CONFIG_HOTPLUG_CPU as the idle task may be scheduled as + * a realtime task in sched_idle_next. + */ +#ifdef CONFIG_HOTPLUG_CPU +#define rq_idle(rq) ((rq)->curr == (rq)->idle && !rt_task((rq)->curr)) +#else +#define rq_idle(rq) ((rq)->curr == (rq)->idle) +#endif + +static inline int task_preempts_curr(struct task_struct *p, struct rq *rq) +{ + struct task_struct *curr = rq->curr; + + return ((p->array == task_rq(p)->active && + TASK_PREEMPTS_CURR(p, curr)) || rq_idle(rq)); +} + +static inline void try_preempt(struct task_struct *p, struct rq *rq) +{ + if (task_preempts_curr(p, rq)) + resched_task(rq->curr); +} + /*** * try_to_wake_up - wake up a thread * @p: the to-be-woken-up thread @@ -1470,7 +1540,7 @@ static int try_to_wake_up(struct task_st if (!(old_state & state)) goto out; - if (p->array) + if (task_queued(p)) goto out_running; cpu = task_cpu(p); @@ -1563,7 +1633,7 @@ out_set_cpu: old_state = p->state; if (!(old_state & state)) goto out; - if (p->array) + if (task_queued(p)) goto out_running; this_cpu = smp_processor_id(); @@ -1572,25 +1642,9 @@ out_set_cpu: out_activate: #endif /* CONFIG_SMP */ - if (old_state == TASK_UNINTERRUPTIBLE) { + if (old_state == TASK_UNINTERRUPTIBLE) rq->nr_uninterruptible--; - /* - * Tasks on involuntary sleep don't earn - * sleep_avg beyond just interactive state. - */ - p->sleep_type = SLEEP_NONINTERACTIVE; - } else - - /* - * Tasks that have marked their sleep as noninteractive get - * woken up with their sleep average not weighted in an - * interactive way. - */ - if (old_state & TASK_NONINTERACTIVE) - p->sleep_type = SLEEP_NONINTERACTIVE; - - activate_task(p, rq, cpu == this_cpu); /* * Sync wakeups (i.e. those types of wakeups where the waker * has indicated that it will leave the CPU in short order) @@ -1599,10 +1653,9 @@ out_activate: * the waker guarantees that the freshly woken up task is going * to be considered on this CPU.) */ - if (!sync || cpu != this_cpu) { - if (TASK_PREEMPTS_CURR(p, rq)) - resched_task(rq->curr); - } + activate_task(p, rq, cpu == this_cpu); + if (!sync || cpu != this_cpu) + try_preempt(p, rq); success = 1; out_running: @@ -1625,7 +1678,6 @@ int fastcall wake_up_state(struct task_s return try_to_wake_up(p, state, 0); } -static void task_running_tick(struct rq *rq, struct task_struct *p); /* * Perform scheduler related setup for a newly forked process p. * p is forked by current. @@ -1653,7 +1705,6 @@ void fastcall sched_fork(struct task_str p->prio = current->normal_prio; INIT_LIST_HEAD(&p->run_list); - p->array = NULL; #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) if (unlikely(sched_info_on())) memset(&p->sched_info, 0, sizeof(p->sched_info)); @@ -1665,30 +1716,31 @@ void fastcall sched_fork(struct task_str /* Want to start with kernel preemption disabled. */ task_thread_info(p)->preempt_count = 1; #endif + if (unlikely(p->policy == SCHED_FIFO)) + goto out; /* * Share the timeslice between parent and child, thus the * total amount of pending timeslices in the system doesn't change, * resulting in more scheduling fairness. */ local_irq_disable(); - p->time_slice = (current->time_slice + 1) >> 1; - /* - * The remainder of the first timeslice might be recovered by - * the parent if the child exits early enough. - */ - p->first_time_slice = 1; - current->time_slice >>= 1; - p->timestamp = sched_clock(); - if (unlikely(!current->time_slice)) { + if (current->time_slice > 0) { + current->time_slice /= 2; + if (current->time_slice) + p->time_slice = current->time_slice; + else + p->time_slice = 1; /* - * This case is rare, it happens when the parent has only - * a single jiffy left from its timeslice. Taking the - * runqueue lock is not a problem. + * The remainder of the first timeslice might be recovered by + * the parent if the child exits early enough. */ - current->time_slice = 1; - task_running_tick(cpu_rq(cpu), current); - } + p->first_time_slice = 1; + } else + p->time_slice = 0; + + p->timestamp = sched_clock(); local_irq_enable(); +out: put_cpu(); } @@ -1710,38 +1762,16 @@ void fastcall wake_up_new_task(struct ta this_cpu = smp_processor_id(); cpu = task_cpu(p); - /* - * We decrease the sleep average of forking parents - * and children as well, to keep max-interactive tasks - * from forking tasks that are max-interactive. The parent - * (current) is done further down, under its lock. - */ - p->sleep_avg = JIFFIES_TO_NS(CURRENT_BONUS(p) * - CHILD_PENALTY / 100 * MAX_SLEEP_AVG / MAX_BONUS); - - p->prio = effective_prio(p); - if (likely(cpu == this_cpu)) { + activate_task(p, rq, 1); if (!(clone_flags & CLONE_VM)) { /* * The VM isn't cloned, so we're in a good position to * do child-runs-first in anticipation of an exec. This * usually avoids a lot of COW overhead. */ - if (unlikely(!current->array)) - __activate_task(p, rq); - else { - p->prio = current->prio; - p->normal_prio = current->normal_prio; - list_add_tail(&p->run_list, ¤t->run_list); - p->array = current->array; - p->array->nr_active++; - inc_nr_running(p, rq); - } set_need_resched(); - } else - /* Run child last */ - __activate_task(p, rq); + } /* * We skip the following code due to cpu == this_cpu * @@ -1758,19 +1788,16 @@ void fastcall wake_up_new_task(struct ta */ p->timestamp = (p->timestamp - this_rq->most_recent_timestamp) + rq->most_recent_timestamp; - __activate_task(p, rq); - if (TASK_PREEMPTS_CURR(p, rq)) - resched_task(rq->curr); + activate_task(p, rq, 0); + try_preempt(p, rq); /* * Parent and child are on different CPUs, now get the - * parent runqueue to update the parent's ->sleep_avg: + * parent runqueue to update the parent's ->flags: */ task_rq_unlock(rq, &flags); this_rq = task_rq_lock(current, &flags); } - current->sleep_avg = JIFFIES_TO_NS(CURRENT_BONUS(current) * - PARENT_PENALTY / 100 * MAX_SLEEP_AVG / MAX_BONUS); task_rq_unlock(this_rq, &flags); } @@ -1785,23 +1812,17 @@ void fastcall wake_up_new_task(struct ta */ void fastcall sched_exit(struct task_struct *p) { + struct task_struct *parent; unsigned long flags; struct rq *rq; - /* - * If the child was a (relative-) CPU hog then decrease - * the sleep_avg of the parent as well. - */ - rq = task_rq_lock(p->parent, &flags); - if (p->first_time_slice && task_cpu(p) == task_cpu(p->parent)) { - p->parent->time_slice += p->time_slice; - if (unlikely(p->parent->time_slice > task_timeslice(p))) - p->parent->time_slice = task_timeslice(p); - } - if (p->sleep_avg < p->parent->sleep_avg) - p->parent->sleep_avg = p->parent->sleep_avg / - (EXIT_WEIGHT + 1) * EXIT_WEIGHT + p->sleep_avg / - (EXIT_WEIGHT + 1); + parent = p->parent; + rq = task_rq_lock(parent, &flags); + if (p->first_time_slice > 0 && task_cpu(p) == task_cpu(parent)) { + parent->time_slice += p->time_slice; + if (unlikely(parent->time_slice > parent->quota)) + parent->time_slice = parent->quota; + } task_rq_unlock(rq, &flags); } @@ -2133,23 +2154,17 @@ void sched_exec(void) * pull_task - move a task from a remote runqueue to the local runqueue. * Both runqueues must be locked. */ -static void pull_task(struct rq *src_rq, struct prio_array *src_array, - struct task_struct *p, struct rq *this_rq, - struct prio_array *this_array, int this_cpu) +static void pull_task(struct rq *src_rq, struct task_struct *p, + struct rq *this_rq, int this_cpu) { - dequeue_task(p, src_array); + dequeue_task(p, src_rq); dec_nr_running(p, src_rq); set_task_cpu(p, this_cpu); inc_nr_running(p, this_rq); - enqueue_task(p, this_array); + enqueue_task(p, this_rq); p->timestamp = (p->timestamp - src_rq->most_recent_timestamp) + this_rq->most_recent_timestamp; - /* - * Note that idle threads have a prio of MAX_PRIO, for this test - * to be always true for them. - */ - if (TASK_PREEMPTS_CURR(p, this_rq)) - resched_task(this_rq->curr); + try_preempt(p, this_rq); } /* @@ -2192,7 +2207,16 @@ int can_migrate_task(struct task_struct return 1; } -#define rq_best_prio(rq) min((rq)->curr->prio, (rq)->best_expired_prio) +static inline int rq_best_prio(struct rq *rq) +{ + int best_prio, exp_prio; + + best_prio = sched_find_first_bit(rq->dyn_bitmap); + exp_prio = find_next_bit(rq->exp_bitmap, MAX_PRIO, MAX_RT_PRIO); + if (unlikely(best_prio > exp_prio)) + best_prio = exp_prio; + return best_prio; +} /* * move_tasks tries to move up to max_nr_move tasks and max_load_move weighted @@ -2208,7 +2232,7 @@ static int move_tasks(struct rq *this_rq { int idx, pulled = 0, pinned = 0, this_best_prio, best_prio, best_prio_seen, skip_for_load; - struct prio_array *array, *dst_array; + struct prio_array *array; struct list_head *head, *curr; struct task_struct *tmp; long rem_load_move; @@ -2235,26 +2259,21 @@ static int move_tasks(struct rq *this_rq * be cache-cold, thus switching CPUs has the least effect * on them. */ - if (busiest->expired->nr_active) { - array = busiest->expired; - dst_array = this_rq->expired; - } else { - array = busiest->active; - dst_array = this_rq->active; - } - + array = busiest->expired; new_array: - /* Start searching at priority 0: */ - idx = 0; + /* Expired arrays don't have RT tasks so they're always MAX_RT_PRIO+ */ + if (array == busiest->expired) + idx = MAX_RT_PRIO; + else + idx = 0; skip_bitmap: if (!idx) - idx = sched_find_first_bit(array->bitmap); + idx = sched_find_first_bit(array->prio_bitmap); else - idx = find_next_bit(array->bitmap, MAX_PRIO, idx); + idx = find_next_bit(array->prio_bitmap, MAX_PRIO, idx); if (idx >= MAX_PRIO) { - if (array == busiest->expired && busiest->active->nr_active) { + if (array == busiest->expired) { array = busiest->active; - dst_array = this_rq->active; goto new_array; } goto out; @@ -2285,7 +2304,7 @@ skip_queue: goto skip_bitmap; } - pull_task(busiest, array, tmp, this_rq, dst_array, this_cpu); + pull_task(busiest, tmp, this_rq, this_cpu); pulled++; rem_load_move -= tmp->load_weight; @@ -3242,11 +3261,36 @@ EXPORT_PER_CPU_SYMBOL(kstat); /* * This is called on clock ticks and on context switches. * Bank in p->sched_time the ns elapsed since the last tick or switch. + * CPU scheduler quota accounting is also performed here in microseconds. + * The value returned from sched_clock() occasionally gives bogus values so + * some sanity checking is required. */ -static inline void -update_cpu_clock(struct task_struct *p, struct rq *rq, unsigned long long now) +static void +update_cpu_clock(struct task_struct *p, struct rq *rq, unsigned long long now, + int tick) { - p->sched_time += now - p->last_ran; + long time_diff = now - p->last_ran; + + if (tick) { + /* + * Called from scheduler_tick() there should be less than two + * jiffies worth, and not negative/overflow. + */ + if (time_diff > JIFFIES_TO_NS(2) || time_diff < 0) + time_diff = JIFFIES_TO_NS(1); + } else { + /* + * Called from context_switch there should be less than one + * jiffy worth, and not negative/overflow. There should be + * some time banked here so use a nominal 1us. + */ + if (time_diff > JIFFIES_TO_NS(1) || time_diff < 1) + time_diff = 1000; + } + /* time_slice accounting is done in usecs to avoid overflow on 32bit */ + if (p != rq->idle && p->policy != SCHED_FIFO) + p->time_slice -= time_diff / 1000; + p->sched_time += time_diff; p->last_ran = rq->most_recent_timestamp = now; } @@ -3267,27 +3311,6 @@ unsigned long long current_sched_time(co } /* - * We place interactive tasks back into the active array, if possible. - * - * To guarantee that this does not starve expired tasks we ignore the - * interactivity of a task if the first expired task had to wait more - * than a 'reasonable' amount of time. This deadline timeout is - * load-dependent, as the frequency of array switched decreases with - * increasing number of running tasks. We also ignore the interactivity - * if a better static_prio task has expired: - */ -static inline int expired_starving(struct rq *rq) -{ - if (rq->curr->static_prio > rq->best_expired_prio) - return 1; - if (!STARVATION_LIMIT || !rq->expired_timestamp) - return 0; - if (jiffies - rq->expired_timestamp > STARVATION_LIMIT * rq->nr_running) - return 1; - return 0; -} - -/* * Account user cpu time to a process. * @p: the process that the cpu time gets accounted to * @hardirq_offset: the offset to subtract from hardirq_count() @@ -3360,87 +3383,47 @@ void account_steal_time(struct task_stru cpustat->steal = cputime64_add(cpustat->steal, tmp); } -static void task_running_tick(struct rq *rq, struct task_struct *p) +/* + * The task has used up its quota of running in this prio_level so it must be + * dropped a priority level, all managed by recalc_task_prio(). + */ +static void task_expired_entitlement(struct rq *rq, struct task_struct *p) { - if (p->array != rq->active) { - /* Task has expired but was not scheduled yet */ - set_tsk_need_resched(p); + int overrun; + + reset_first_time_slice(p); + if (rt_task(p)) { + p->time_slice += p->quota; + list_move_tail(&p->run_list, p->array->queue + p->prio); return; } - spin_lock(&rq->lock); + overrun = p->time_slice; + dequeue_task(p, rq); + enqueue_task(p, rq); /* - * The task was running during this tick - update the - * time slice counter. Note: we do not update a thread's - * priority until it either goes to sleep or uses up its - * timeslice. This makes it possible for interactive tasks - * to use up their timeslices at their highest priority levels. + * Subtract any extra time this task ran over its time_slice; ie + * overrun will either be 0 or negative. */ - if (rt_task(p)) { - /* - * RR tasks need a special form of timeslice management. - * FIFO tasks have no timeslices. - */ - if ((p->policy == SCHED_RR) && !--p->time_slice) { - p->time_slice = task_timeslice(p); - p->first_time_slice = 0; - set_tsk_need_resched(p); - - /* put it at the end of the queue: */ - requeue_task(p, rq->active); - } - goto out_unlock; - } - if (!--p->time_slice) { - dequeue_task(p, rq->active); - set_tsk_need_resched(p); - p->prio = effective_prio(p); - p->time_slice = task_timeslice(p); - p->first_time_slice = 0; - - if (!rq->expired_timestamp) - rq->expired_timestamp = jiffies; - if (!TASK_INTERACTIVE(p) || expired_starving(rq)) { - enqueue_task(p, rq->expired); - if (p->static_prio < rq->best_expired_prio) - rq->best_expired_prio = p->static_prio; - } else - enqueue_task(p, rq->active); - } else { - /* - * Prevent a too long timeslice allowing a task to monopolize - * the CPU. We do this by splitting up the timeslice into - * smaller pieces. - * - * Note: this does not mean the task's timeslices expire or - * get lost in any way, they just might be preempted by - * another task of equal priority. (one with higher - * priority would have preempted this task already.) We - * requeue this task to the end of the list on this priority - * level, which is in essence a round-robin of tasks with - * equal priority. - * - * This only applies to tasks in the interactive - * delta range with at least TIMESLICE_GRANULARITY to requeue. - */ - if (TASK_INTERACTIVE(p) && !((task_timeslice(p) - - p->time_slice) % TIMESLICE_GRANULARITY(p)) && - (p->time_slice >= TIMESLICE_GRANULARITY(p)) && - (p->array == rq->active)) { + p->time_slice += overrun; +} - requeue_task(p, rq->active); - set_tsk_need_resched(p); - } - } -out_unlock: +/* This manages tasks that have run out of timeslice during a scheduler_tick */ +static void task_running_tick(struct rq *rq, struct task_struct *p) +{ + /* SCHED_FIFO tasks never run out of timeslice. */ + if (p->time_slice > 0 || p->policy == SCHED_FIFO) + return; + /* p->time_slice <= 0 */ + spin_lock(&rq->lock); + if (likely(task_queued(p))) + task_expired_entitlement(rq, p); + set_tsk_need_resched(p); spin_unlock(&rq->lock); } /* * This function gets called by the timer code, with HZ frequency. * We call it with interrupts disabled. - * - * It also gets called by the fork code, when changing the parent's - * timeslices. */ void scheduler_tick(void) { @@ -3450,7 +3433,7 @@ void scheduler_tick(void) int idle_at_tick = idle_cpu(cpu); struct rq *rq = cpu_rq(cpu); - update_cpu_clock(p, rq, now); + update_cpu_clock(p, rq, now, 1); if (!idle_at_tick) task_running_tick(rq, p); @@ -3499,10 +3482,55 @@ EXPORT_SYMBOL(sub_preempt_count); #endif -static inline int interactive_sleep(enum sleep_type sleep_type) +static void reset_prio_levels(struct rq *rq) { - return (sleep_type == SLEEP_INTERACTIVE || - sleep_type == SLEEP_INTERRUPTED); + rq->active->best_static_prio = MAX_PRIO - 1; + rq->expired->best_static_prio = MAX_PRIO - 1; + memset(rq->prio_level, 0, sizeof(int) * PRIO_RANGE); +} + +/* + * next_dynamic_task finds the next suitable dynamic task. + */ +static inline struct task_struct *next_dynamic_task(struct rq *rq, int idx) +{ + struct prio_array *array = rq->active; + struct task_struct *next; + struct list_head *queue; + int nstatic; + +retry: + if (idx >= MAX_PRIO) { + /* There are no more tasks in the active array. Swap arrays */ + array = rq->expired; + rq->expired = rq->active; + rq->active = array; + rq->exp_bitmap = rq->expired->prio_bitmap; + rq->dyn_bitmap = rq->active->prio_bitmap; + rq->prio_rotation++; + idx = find_next_bit(rq->dyn_bitmap, MAX_PRIO, MAX_RT_PRIO); + reset_prio_levels(rq); + } + queue = array->queue + idx; + next = list_entry(queue->next, struct task_struct, run_list); + if (unlikely(next->time_slice <= 0)) { + /* + * Unlucky enough that this task ran out of time_slice + * before it hit a scheduler_tick so it should have its + * priority reassessed and choose another task (possibly + * the same one) + */ + task_expired_entitlement(rq, next); + idx = find_next_bit(rq->dyn_bitmap, MAX_PRIO, MAX_RT_PRIO); + goto retry; + } + next->rotation = rq->prio_rotation; + nstatic = next->static_prio; + if (nstatic < array->best_static_prio) + array->best_static_prio = nstatic; + if (idx > rq->prio_level[USER_PRIO(nstatic)]) + rq->prio_level[USER_PRIO(nstatic)] = idx; + return next; } /* @@ -3511,13 +3539,11 @@ static inline int interactive_sleep(enum asmlinkage void __sched schedule(void) { struct task_struct *prev, *next; - struct prio_array *array; struct list_head *queue; unsigned long long now; - unsigned long run_time; - int cpu, idx, new_prio; long *switch_count; struct rq *rq; + int cpu, idx; /* * Test if we are atomic. Since do_exit() needs to call into @@ -3553,18 +3579,6 @@ need_resched_nonpreemptible: schedstat_inc(rq, sched_cnt); now = sched_clock(); - if (likely((long long)(now - prev->timestamp) < NS_MAX_SLEEP_AVG)) { - run_time = now - prev->timestamp; - if (unlikely((long long)(now - prev->timestamp) < 0)) - run_time = 0; - } else - run_time = NS_MAX_SLEEP_AVG; - - /* - * Tasks charged proportionately less run_time at high sleep_avg to - * delay them losing their interactive status - */ - run_time /= (CURRENT_BONUS(prev) ? : 1); spin_lock_irq(&rq->lock); @@ -3586,59 +3600,29 @@ need_resched_nonpreemptible: idle_balance(cpu, rq); if (!rq->nr_running) { next = rq->idle; - rq->expired_timestamp = 0; goto switch_tasks; } } - array = rq->active; - if (unlikely(!array->nr_active)) { - /* - * Switch the active and expired arrays. - */ - schedstat_inc(rq, sched_switch); - rq->active = rq->expired; - rq->expired = array; - array = rq->active; - rq->expired_timestamp = 0; - rq->best_expired_prio = MAX_PRIO; + idx = sched_find_first_bit(rq->dyn_bitmap); + if (!rt_prio(idx)) + next = next_dynamic_task(rq, idx); + else { + queue = rq->active->queue + idx; + next = list_entry(queue->next, struct task_struct, run_list); } - - idx = sched_find_first_bit(array->bitmap); - queue = array->queue + idx; - next = list_entry(queue->next, struct task_struct, run_list); - - if (!rt_task(next) && interactive_sleep(next->sleep_type)) { - unsigned long long delta = now - next->timestamp; - if (unlikely((long long)(now - next->timestamp) < 0)) - delta = 0; - - if (next->sleep_type == SLEEP_INTERACTIVE) - delta = delta * (ON_RUNQUEUE_WEIGHT * 128 / 100) / 128; - - array = next->array; - new_prio = recalc_task_prio(next, next->timestamp + delta); - - if (unlikely(next->prio != new_prio)) { - dequeue_task(next, array); - next->prio = new_prio; - enqueue_task(next, array); - } - } - next->sleep_type = SLEEP_NORMAL; switch_tasks: - if (next == rq->idle) + if (next == rq->idle) { + reset_prio_levels(rq); + rq->prio_rotation++; schedstat_inc(rq, sched_goidle); + } prefetch(next); prefetch_stack(next); clear_tsk_need_resched(prev); rcu_qsctr_inc(task_cpu(prev)); - update_cpu_clock(prev, rq, now); - - prev->sleep_avg -= run_time; - if ((long)prev->sleep_avg <= 0) - prev->sleep_avg = 0; + update_cpu_clock(prev, rq, now, 0); prev->timestamp = prev->last_ran = now; sched_info_switch(prev, next); @@ -4074,29 +4058,22 @@ EXPORT_SYMBOL(sleep_on_timeout); */ void rt_mutex_setprio(struct task_struct *p, int prio) { - struct prio_array *array; unsigned long flags; + int queued, oldprio; struct rq *rq; - int oldprio; BUG_ON(prio < 0 || prio > MAX_PRIO); rq = task_rq_lock(p, &flags); oldprio = p->prio; - array = p->array; - if (array) - dequeue_task(p, array); + queued = task_queued(p); + if (queued) + dequeue_task(p, rq); p->prio = prio; - if (array) { - /* - * If changing to an RT priority then queue it - * in the active array! - */ - if (rt_task(p)) - array = rq->active; - enqueue_task(p, array); + if (queued) { + enqueue_task(p, rq); /* * Reschedule if we are currently running on this runqueue and * our priority decreased, or if we are not currently running on @@ -4105,8 +4082,8 @@ void rt_mutex_setprio(struct task_struct if (task_running(rq, p)) { if (p->prio > oldprio) resched_task(rq->curr); - } else if (TASK_PREEMPTS_CURR(p, rq)) - resched_task(rq->curr); + } else + try_preempt(p, rq); } task_rq_unlock(rq, &flags); } @@ -4115,8 +4092,7 @@ void rt_mutex_setprio(struct task_struct void set_user_nice(struct task_struct *p, long nice) { - struct prio_array *array; - int old_prio, delta; + int queued, old_prio,delta; unsigned long flags; struct rq *rq; @@ -4137,20 +4113,20 @@ void set_user_nice(struct task_struct *p p->static_prio = NICE_TO_PRIO(nice); goto out_unlock; } - array = p->array; - if (array) { - dequeue_task(p, array); + queued = task_queued(p); + if (queued) { + dequeue_task(p, rq); dec_raw_weighted_load(rq, p); } p->static_prio = NICE_TO_PRIO(nice); - set_load_weight(p); old_prio = p->prio; p->prio = effective_prio(p); + set_quota(p); delta = p->prio - old_prio; - if (array) { - enqueue_task(p, array); + if (queued) { + enqueue_task(p, rq); inc_raw_weighted_load(rq, p); /* * If the task increased its priority or is running and @@ -4226,11 +4202,23 @@ asmlinkage long sys_nice(int increment) * * This is the priority value as seen by users in /proc. * RT tasks are offset by -200. Normal tasks are centered - * around 0, value goes from -16 to +15. + * around 1, value goes from 0 to +79. Values higher than + * 39 indicate task is on the expired array. This is done + * lockless and may rarely return an active instead of + * expired value. */ int task_prio(const struct task_struct *p) { - return p->prio - MAX_RT_PRIO; + int prio = p->prio - MAX_RT_PRIO; + + if (task_queued(p)) { + struct rq *rq = task_rq(p); + struct prio_array *array = p->array; + + if (rq && rq->expired == array) + prio += PRIO_RANGE; + } + return prio; } /** @@ -4273,19 +4261,14 @@ static inline struct task_struct *find_p /* Actually do priority change: must hold rq lock. */ static void __setscheduler(struct task_struct *p, int policy, int prio) { - BUG_ON(p->array); + BUG_ON(task_queued(p)); p->policy = policy; p->rt_priority = prio; p->normal_prio = normal_prio(p); /* we are holding p->pi_lock already */ p->prio = rt_mutex_getprio(p); - /* - * SCHED_BATCH tasks are treated as perpetual CPU hogs: - */ - if (policy == SCHED_BATCH) - p->sleep_avg = 0; - set_load_weight(p); + set_quota(p); } /** @@ -4299,8 +4282,7 @@ static void __setscheduler(struct task_s int sched_setscheduler(struct task_struct *p, int policy, struct sched_param *param) { - int retval, oldprio, oldpolicy = -1; - struct prio_array *array; + int queued, retval, oldprio, oldpolicy = -1; unsigned long flags; struct rq *rq; @@ -4374,12 +4356,12 @@ recheck: spin_unlock_irqrestore(&p->pi_lock, flags); goto recheck; } - array = p->array; - if (array) + queued = task_queued(p); + if (queued) deactivate_task(p, rq); oldprio = p->prio; __setscheduler(p, policy, param->sched_priority); - if (array) { + if (queued) { __activate_task(p, rq); /* * Reschedule if we are currently running on this runqueue and @@ -4389,8 +4371,8 @@ recheck: if (task_running(rq, p)) { if (p->prio > oldprio) resched_task(rq->curr); - } else if (TASK_PREEMPTS_CURR(p, rq)) - resched_task(rq->curr); + } else + try_preempt(p, rq); } __task_rq_unlock(rq); spin_unlock_irqrestore(&p->pi_lock, flags); @@ -4663,40 +4645,27 @@ asmlinkage long sys_sched_getaffinity(pi * sys_sched_yield - yield the current processor to other threads. * * This function yields the current CPU by moving the calling thread - * to the expired array. If there are no other threads running on this - * CPU then this function will return. + * to the expired array if SCHED_NORMAL or the end of its current priority + * queue if a realtime task. If there are no other threads running on this + * cpu this function will return. */ asmlinkage long sys_sched_yield(void) { struct rq *rq = this_rq_lock(); - struct prio_array *array = current->array, *target = rq->expired; + struct task_struct *p = current; schedstat_inc(rq, yld_cnt); - /* - * We implement yielding by moving the task into the expired - * queue. - * - * (special rule: RT tasks will just roundrobin in the active - * array.) - */ - if (rt_task(current)) - target = rq->active; + if (rq->nr_running == 1) + schedstat_inc(rq, yld_both_empty); + else { + struct prio_array *old_array = p->array; + int old_prio = p->prio; - if (array->nr_active == 1) { - schedstat_inc(rq, yld_act_empty); - if (!rq->expired->nr_active) - schedstat_inc(rq, yld_both_empty); - } else if (!rq->expired->nr_active) - schedstat_inc(rq, yld_exp_empty); - - if (array != target) { - dequeue_task(current, array); - enqueue_task(current, target); - } else - /* - * requeue_task is cheaper so perform that if possible. - */ - requeue_task(current, array); + /* p->prio will be updated in requeue_task via queue_expired */ + if (!rt_task(p)) + p->array = rq->expired; + requeue_task(p, rq, old_array, old_prio); + } /* * Since we are going to call schedule() anyway, there's @@ -4906,8 +4875,8 @@ long sys_sched_rr_get_interval(pid_t pid if (retval) goto out_unlock; - jiffies_to_timespec(p->policy == SCHED_FIFO ? - 0 : task_timeslice(p), &t); + t = ns_to_timespec(p->policy == SCHED_FIFO ? 0 : + MS_TO_NS(task_timeslice(p))); read_unlock(&tasklist_lock); retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0; out_nounlock: @@ -5003,10 +4972,10 @@ void __cpuinit init_idle(struct task_str struct rq *rq = cpu_rq(cpu); unsigned long flags; - idle->timestamp = sched_clock(); - idle->sleep_avg = 0; - idle->array = NULL; - idle->prio = idle->normal_prio = MAX_PRIO; + bitmap_zero(idle->bitmap, PRIO_RANGE); + idle->timestamp = idle->last_ran = sched_clock(); + idle->array = rq->active; + idle->prio = idle->normal_prio = NICE_TO_PRIO(0); idle->state = TASK_RUNNING; idle->cpus_allowed = cpumask_of_cpu(cpu); set_task_cpu(idle, cpu); @@ -5125,7 +5094,7 @@ static int __migrate_task(struct task_st goto out; set_task_cpu(p, dest_cpu); - if (p->array) { + if (task_queued(p)) { /* * Sync timestamp with rq_dest's before activating. * The same thing could be achieved by doing this step @@ -5136,8 +5105,7 @@ static int __migrate_task(struct task_st + rq_dest->most_recent_timestamp; deactivate_task(p, rq_src); __activate_task(p, rq_dest); - if (TASK_PREEMPTS_CURR(p, rq_dest)) - resched_task(rq_dest->curr); + try_preempt(p, rq_dest); } ret = 1; out: @@ -5434,7 +5402,7 @@ migration_call(struct notifier_block *nf /* Idle task back to normal (off runqueue, low prio) */ rq = task_rq_lock(rq->idle, &flags); deactivate_task(rq->idle, rq); - rq->idle->static_prio = MAX_PRIO; + rq->idle->static_prio = NICE_TO_PRIO(0); __setscheduler(rq->idle, SCHED_NORMAL, 0); migrate_dead_tasks(cpu); task_rq_unlock(rq, &flags); @@ -6960,6 +6928,13 @@ void __init sched_init_smp(void) /* Move init over to a non-isolated CPU */ if (set_cpus_allowed(current, non_isolated_cpus) < 0) BUG(); + + /* + * Assume that every added cpu gives us slightly less overall latency + * allowing us to increase the base rr_interval, but in a non linear + * fashion. + */ + rr_interval *= 1 + ilog2(num_online_cpus()); } #else void __init sched_init_smp(void) @@ -6982,6 +6957,16 @@ void __init sched_init(void) int i, j, k; int highest_cpu = 0; + /* Generate the priority matrix */ + for (i = 0; i < PRIO_RANGE; i++) { + bitmap_fill(prio_matrix[i], PRIO_RANGE); + j = PRIO_RANGE * PRIO_RANGE / (PRIO_RANGE - i); + for (k = 0; k <= PRIO_RANGE * (PRIO_RANGE - 1); k += j) { + __clear_bit(PRIO_RANGE - 1 - (k / PRIO_RANGE), + prio_matrix[i]); + } + } + for_each_possible_cpu(i) { struct prio_array *array; struct rq *rq; @@ -6990,11 +6975,16 @@ void __init sched_init(void) spin_lock_init(&rq->lock); lockdep_set_class(&rq->lock, &rq->rq_lock_key); rq->nr_running = 0; + rq->prio_rotation = 0; rq->active = rq->arrays; rq->expired = rq->arrays + 1; - rq->best_expired_prio = MAX_PRIO; + reset_prio_levels(rq); + rq->dyn_bitmap = rq->active->prio_bitmap; + rq->exp_bitmap = rq->expired->prio_bitmap; #ifdef CONFIG_SMP + rq->active->rq = rq; + rq->expired->rq = rq; rq->sd = NULL; for (j = 1; j < 3; j++) rq->cpu_load[j] = 0; @@ -7007,17 +6997,16 @@ void __init sched_init(void) atomic_set(&rq->nr_iowait, 0); for (j = 0; j < 2; j++) { + array = rq->arrays + j; - for (k = 0; k < MAX_PRIO; k++) { + for (k = 0; k < MAX_PRIO; k++) INIT_LIST_HEAD(array->queue + k); - __clear_bit(k, array->bitmap); - } - // delimiter for bitsearch - __set_bit(MAX_PRIO, array->bitmap); + bitmap_zero(array->prio_bitmap, MAX_PRIO); + /* delimiter for bitsearch */ + __set_bit(MAX_PRIO, array->prio_bitmap); } highest_cpu = i; } - set_load_weight(&init_task); #ifdef CONFIG_SMP @@ -7072,10 +7061,10 @@ EXPORT_SYMBOL(__might_sleep); #ifdef CONFIG_MAGIC_SYSRQ void normalize_rt_tasks(void) { - struct prio_array *array; struct task_struct *p; unsigned long flags; struct rq *rq; + int queued; read_lock_irq(&tasklist_lock); for_each_process(p) { @@ -7085,11 +7074,11 @@ void normalize_rt_tasks(void) spin_lock_irqsave(&p->pi_lock, flags); rq = __task_rq_lock(p); - array = p->array; - if (array) + queued = task_queued(p); + if (queued) deactivate_task(p, task_rq(p)); __setscheduler(p, SCHED_NORMAL, 0); - if (array) { + if (queued) { __activate_task(p, task_rq(p)); resched_task(rq->curr); } Index: linux-2.6.22-rc2-ck1/kernel/sysctl.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/kernel/sysctl.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/kernel/sysctl.c 2007-05-19 20:12:08.000000000 +1000 @@ -78,6 +78,8 @@ extern int percpu_pagelist_fraction; extern int compat_log; extern int maps_protect; extern int sysctl_stat_interval; +extern int rr_interval; +extern int sched_interactive; /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */ static int maxolduid = 65535; @@ -161,6 +163,14 @@ int sysctl_legacy_va_layout; #endif +/* Constants for minimum and maximum testing. + We use these as one-element integer vectors. */ +static int __read_mostly zero; +static int __read_mostly one = 1; +static int __read_mostly one_hundred = 100; +static int __read_mostly five_thousand = 5000; + + /* The default sysctl tables: */ static ctl_table root_table[] = { @@ -501,6 +511,25 @@ static ctl_table kern_table[] = { .mode = 0444, .proc_handler = &proc_dointvec, }, + { + .ctl_name = CTL_UNNUMBERED, + .procname = "rr_interval", + .data = &rr_interval, + .maxlen = sizeof (int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .strategy = &sysctl_intvec, + .extra1 = &one, + .extra2 = &five_thousand, + }, + { + .ctl_name = CTL_UNNUMBERED, + .procname = "interactive", + .data = &sched_interactive, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec, + }, #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) { .ctl_name = KERN_UNKNOWN_NMI_PANIC, @@ -619,12 +648,6 @@ static ctl_table kern_table[] = { { .ctl_name = 0 } }; -/* Constants for minimum and maximum testing in vm_table. - We use these as one-element integer vectors. */ -static int zero; -static int one_hundred = 100; - - static ctl_table vm_table[] = { { .ctl_name = VM_OVERCOMMIT_MEMORY, Index: linux-2.6.22-rc2-ck1/Documentation/sched-design.txt =================================================================== --- linux-2.6.22-rc2-ck1.orig/Documentation/sched-design.txt 2006-11-30 11:30:31.000000000 +1100 +++ linux-2.6.22-rc2-ck1/Documentation/sched-design.txt 2007-05-19 20:11:57.000000000 +1000 @@ -1,11 +1,14 @@ - Goals, Design and Implementation of the - new ultra-scalable O(1) scheduler + Goals, Design and Implementation of the ultra-scalable O(1) scheduler by + Ingo Molnar and theStaircase Deadline cpu scheduler policy designed by + Con Kolivas. - This is an edited version of an email Ingo Molnar sent to - lkml on 4 Jan 2002. It describes the goals, design, and - implementation of Ingo's new ultra-scalable O(1) scheduler. - Last Updated: 18 April 2002. + This was originally an edited version of an email Ingo Molnar sent to + lkml on 4 Jan 2002. It describes the goals, design, and implementation + of Ingo's ultra-scalable O(1) scheduler. It now contains a description + of the Staircase Deadline priority scheduler that was built on this + design. + Last Updated: Fri, 4 May 2007 Goal @@ -163,3 +166,222 @@ certain code paths and data constructs. code is smaller than the old one. Ingo + + +Staircase Deadline cpu scheduler policy +================================================ + +Design summary +============== + +A novel design which incorporates a foreground-background descending priority +system (the staircase) via a bandwidth allocation matrix according to nice +level. + + +Features +======== + +A starvation free, strict fairness O(1) scalable design with interactivity +as good as the above restrictions can provide. There is no interactivity +estimator, no sleep/run measurements and only simple fixed accounting. +The design has strict enough a design and accounting that task behaviour +can be modelled and maximum scheduling latencies can be predicted by +the virtual deadline mechanism that manages runqueues. The prime concern +in this design is to maintain fairness at all costs determined by nice level, +yet to maintain as good interactivity as can be allowed within the +constraints of strict fairness. + + +Design description +================== + +SD works off the principle of providing each task a quota of runtime that it is +allowed to run at a number of priority levels determined by its static priority +(ie. its nice level). If the task uses up its quota it has its priority +decremented to the next level determined by a priority matrix. Once every +runtime quota has been consumed of every priority level, a task is queued on the +"expired" array. When no other tasks exist with quota, the expired array is +activated and fresh quotas are handed out. This is all done in O(1). + +Design details +============== + +Each task keeps a record of its own entitlement of cpu time. Most of the rest of +these details apply to non-realtime tasks as rt task management is straight +forward. + +Each runqueue keeps a record of what major epoch it is up to in the +rq->prio_rotation field which is incremented on each major epoch. It also +keeps a record of the current prio_level for each static priority task. + +Each task keeps a record of what major runqueue epoch it was last running +on in p->rotation. It also keeps a record of what priority levels it has +already been allocated quota from during this epoch in a bitmap p->bitmap. + +The only tunable that determines all other details is the RR_INTERVAL. This +is set to 8ms, and is scaled gently upwards with more cpus. This value is +tunable via a /proc interface. + +All tasks are initially given a quota based on RR_INTERVAL. This is equal to +RR_INTERVAL between nice values of -6 and 0, half that size above nice 0, and +progressively larger for nice values from -1 to -20. This is assigned to +p->quota and only changes with changes in nice level. + +As a task is first queued, it checks in recalc_task_prio to see if it has run at +this runqueue's current priority rotation. If it has not, it will have its +p->prio level set according to the first slot in a "priority matrix" and will be +given a p->time_slice equal to the p->quota, and has its allocation bitmap bit +set in p->bitmap for this prio level. It is then queued on the current active +priority array. + +If a task has already been running during this major epoch, and it has +p->time_slice left and the rq->prio_quota for the task's p->prio still +has quota, it will be placed back on the active array, but no more quota +will be added. + +If a task has been running during this major epoch, but does not have +p->time_slice left, it will find the next lowest priority in its bitmap that it +has not been allocated quota from. It then gets the a full quota in +p->time_slice. It is then queued on the current active priority array at the +newly determined lower priority. + +If a task has been running during this major epoch, and does not have +any entitlement left in p->bitmap and no time_slice left, it will have its +bitmap cleared, and be queued at its best prio again, but on the expired +priority array. + +When a task is queued, it has its relevant bit set in the array->prio_bitmap. + +p->time_slice is stored in nanosconds and is updated via update_cpu_clock on +schedule() and scheduler_tick. If p->time_slice is below zero then the +recalc_task_prio is readjusted and the task rescheduled. + + +Priority Matrix +=============== + +In order to minimise the latencies between tasks of different nice levels +running concurrently, the dynamic priority slots where different nice levels +are queued are dithered instead of being sequential. What this means is that +there are 40 priority slots where a task may run during one major rotation, +and the allocation of slots is dependant on nice level. In the +following table, a zero represents a slot where the task may run. + +PRIORITY:0..................20.................39 +nice -20 0000000000000000000000000000000000000000 +nice -10 1000100010001000100010001000100010010000 +nice 0 1010101010101010101010101010101010101010 +nice 5 1011010110110101101101011011010110110110 +nice 10 1110111011101110111011101110111011101110 +nice 15 1111111011111110111111101111111011111110 +nice 19 1111111111111111111111111111111111111110 + +As can be seen, a nice -20 task runs in every priority slot whereas a nice 19 +task only runs one slot per major rotation. This dithered table allows for the +smallest possible maximum latencies between tasks of varying nice levels, thus +allowing vastly different nice levels to be used. + +SCHED_BATCH tasks are managed slightly differently, receiving only the top +slots from its priority bitmap giving it equal cpu as SCHED_NORMAL, but +slightly higher latencies. + + +Modelling deadline behaviour +============================ + +As the accounting in this design is hard and not modified by sleep average +calculations or interactivity modifiers, it is possible to accurately +predict the maximum latency that a task may experience under different +conditions. This is a virtual deadline mechanism enforced by mandatory +timeslice expiration and not outside bandwidth measurement. + +The maximum duration a task can run during one major epoch is determined by its +nice value. Nice 0 tasks can run at 19 different priority levels for RR_INTERVAL +duration during each epoch. Nice 10 tasks can run at 9 priority levels for each +epoch, and so on. The table in the priority matrix above demonstrates how this +is enforced. + +Therefore the maximum duration a runqueue epoch can take is determined by +the number of tasks running, and their nice level. After that, the maximum +duration it can take before a task can wait before it get scheduled is +determined by the position of its first slot on the matrix. + +In the following examples, these are _worst case scenarios_ and would rarely +occur, but can be modelled nonetheless to determine the maximum possible +latency. + +So for example, if two nice 0 tasks are running, and one has just expired as +another is activated for the first time receiving a full quota for this +runqueue rotation, the first task will wait: + +nr_tasks * max_duration + nice_difference * rr_interval +1 * 19 * RR_INTERVAL + 0 = 152ms + +In the presence of a nice 10 task, a nice 0 task would wait a maximum of +1 * 10 * RR_INTERVAL + 0 = 80ms + +In the presence of a nice 0 task, a nice 10 task would wait a maximum of +1 * 19 * RR_INTERVAL + 1 * RR_INTERVAL = 160ms + +More useful than these values, though, are the average latencies which are +a matter of determining the average distance between priority slots of +different nice values and multiplying them by the tasks' quota. For example +in the presence of a nice -10 task, a nice 0 task will wait either one or +two slots. Given that nice -10 tasks have a quota 2.5 times the RR_INTERVAL, +this means the latencies will alternate between 2.5 and 5 RR_INTERVALs or +20 and 40ms respectively (on uniprocessor at 1000HZ). + + +Achieving interactivity +======================= + +A requirement of this scheduler design was to achieve good interactivity +despite being a completely fair deadline based design. The disadvantage of +designs that try to achieve interactivity is that they usually do so at +the expense of maintaining fairness. As cpu speeds increase, the requirement +for some sort of metered unfairness towards interactive tasks becomes a less +desirable phenomenon, but low latency and fairness remains mandatory to +good interactive performance. + +This design relies on the fact that interactive tasks, by their nature, +sleep often. Most fair scheduling designs end up penalising such tasks +indirectly giving them less than their fair possible share because of the +sleep, and have to use a mechanism of bonusing their priority to offset +this based on the duration they sleep. This becomes increasingly inaccurate +as the number of running tasks rises and more tasks spend time waiting on +runqueues rather than sleeping, and it is impossible to tell whether the +task that's waiting on a runqueue only intends to run for a short period and +then sleep again after than runqueue wait. Furthermore, all such designs rely +on a period of time to pass to accumulate some form of statistic on the task +before deciding on how much to give them preference. The shorter this period, +the more rapidly bursts of cpu ruin the interactive tasks behaviour. The +longer this period, the longer it takes for interactive tasks to get low +scheduling latencies and fair cpu. + +This design does not measure sleep time at all. Interactive tasks that sleep +often will wake up having consumed very little if any of their quota for +the current major priority rotation. The longer they have slept, the less +likely they are to even be on the current major priority rotation. Once +woken up, though, they get to use up a their full quota for that epoch, +whether part of a quota remains or a full quota. Overall, however, they +can still only run as much cpu time for that epoch as any other task of the +same nice level. This means that two tasks behaving completely differently +from fully cpu bound to waking/sleeping extremely frequently will still +get the same quota of cpu, but the latter will be using its quota for that +epoch in bursts rather than continuously. This guarantees that interactive +tasks get the same amount of cpu as cpu bound ones. + +The other requirement of interactive tasks is also to obtain low latencies +for when they are scheduled. Unlike fully cpu bound tasks and the maximum +latencies possible described in the modelling deadline behaviour section +above, tasks that sleep will wake up with quota available usually at the +current runqueue's priority_level or better. This means that the most latency +they are likely to see is one RR_INTERVAL, and often they will preempt the +current task if it is not of a sleeping nature. This then guarantees very +low latency for interactive tasks, and the lowest latencies for the least +cpu bound tasks. + + +Fri, 4 May 2007 +Con Kolivas Index: linux-2.6.22-rc2-ck1/Documentation/sysctl/kernel.txt =================================================================== --- linux-2.6.22-rc2-ck1.orig/Documentation/sysctl/kernel.txt 2007-05-19 20:10:46.000000000 +1000 +++ linux-2.6.22-rc2-ck1/Documentation/sysctl/kernel.txt 2007-05-19 20:12:08.000000000 +1000 @@ -25,6 +25,7 @@ show up in /proc/sys/kernel: - domainname - hostname - hotplug +- interactive - java-appletviewer [ binfmt_java, obsolete ] - java-interpreter [ binfmt_java, obsolete ] - kstack_depth_to_print [ X86 only ] @@ -43,6 +44,7 @@ show up in /proc/sys/kernel: - printk - real-root-dev ==> Documentation/initrd.txt - reboot-cmd [ SPARC only ] +- rr_interval - rtsig-max - rtsig-nr - sem @@ -164,6 +166,21 @@ Default value is "/sbin/hotplug". ============================================================== +interactive: + +The staircase-deadline cpu scheduler can be set in either purely +forward-looking mode for absolutely rigid fairness and cpu distribution +according to nice level, or it can allow a small per-process history +to smooth out cpu usage perturbations common in interactive tasks by +enabling this sysctl. While small fairness issues can arise with this +enabled, overall fairness is usually still strongly maintained and +starvation is never possible. Enabling this can significantly smooth +out 3d graphics and games. + +Default value is 1 (enabled). + +============================================================== + l2cr: (PPC only) This flag controls the L2 cache of G3 processor boards. If @@ -288,6 +305,19 @@ rebooting. ??? ============================================================== +rr_interval: + +This is the smallest duration that any cpu process scheduling unit +will run for. Increasing this value can increase throughput of cpu +bound tasks substantially but at the expense of increased latencies +overall. This value is in milliseconds and the default value chosen +depends on the number of cpus available at scheduler initialisation +with a minimum of 8. + +Valid values are from 1-5000. + +============================================================== + rtsig-max & rtsig-nr: The file rtsig-max can be used to tune the maximum number Index: linux-2.6.22-rc2-ck1/fs/pipe.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/fs/pipe.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/fs/pipe.c 2007-05-19 20:11:57.000000000 +1000 @@ -41,12 +41,7 @@ void pipe_wait(struct pipe_inode_info *p { DEFINE_WAIT(wait); - /* - * Pipes are system-local resources, so sleeping on them - * is considered a noninteractive wait: - */ - prepare_to_wait(&pipe->wait, &wait, - TASK_INTERRUPTIBLE | TASK_NONINTERACTIVE); + prepare_to_wait(&pipe->wait, &wait, TASK_INTERRUPTIBLE); if (pipe->inode) mutex_unlock(&pipe->inode->i_mutex); schedule(); Index: linux-2.6.22-rc2-ck1/fs/proc/array.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/fs/proc/array.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/fs/proc/array.c 2007-05-19 20:11:57.000000000 +1000 @@ -165,7 +165,6 @@ static inline char * task_state(struct t rcu_read_lock(); buffer += sprintf(buffer, "State:\t%s\n" - "SleepAVG:\t%lu%%\n" "Tgid:\t%d\n" "Pid:\t%d\n" "PPid:\t%d\n" @@ -173,7 +172,6 @@ static inline char * task_state(struct t "Uid:\t%d\t%d\t%d\t%d\n" "Gid:\t%d\t%d\t%d\t%d\n", get_task_state(p), - (p->sleep_avg/1024)*100/(1020000000/1024), p->tgid, p->pid, pid_alive(p) ? rcu_dereference(p->real_parent)->tgid : 0, pid_alive(p) && p->ptrace ? rcu_dereference(p->parent)->pid : 0, Index: linux-2.6.22-rc2-ck1/include/linux/init_task.h =================================================================== --- linux-2.6.22-rc2-ck1.orig/include/linux/init_task.h 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/include/linux/init_task.h 2007-05-19 20:11:57.000000000 +1000 @@ -125,13 +125,15 @@ extern struct group_info init_groups; .prio = MAX_PRIO-20, \ .static_prio = MAX_PRIO-20, \ .normal_prio = MAX_PRIO-20, \ + .rotation = 0, \ .policy = SCHED_NORMAL, \ .cpus_allowed = CPU_MASK_ALL, \ .mm = NULL, \ .active_mm = &init_mm, \ .run_list = LIST_HEAD_INIT(tsk.run_list), \ .ioprio = 0, \ - .time_slice = HZ, \ + .time_slice = 1000000000, \ + .quota = 1000000000, \ .tasks = LIST_HEAD_INIT(tsk.tasks), \ .ptrace_children= LIST_HEAD_INIT(tsk.ptrace_children), \ .ptrace_list = LIST_HEAD_INIT(tsk.ptrace_list), \ Index: linux-2.6.22-rc2-ck1/kernel/softirq.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/kernel/softirq.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/kernel/softirq.c 2007-05-19 20:11:57.000000000 +1000 @@ -488,7 +488,7 @@ void __init softirq_init(void) static int ksoftirqd(void * __bind_cpu) { - set_user_nice(current, 19); + set_user_nice(current, 15); current->flags |= PF_NOFREEZE; set_current_state(TASK_INTERRUPTIBLE); Index: linux-2.6.22-rc2-ck1/kernel/workqueue.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/kernel/workqueue.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/kernel/workqueue.c 2007-05-19 20:11:57.000000000 +1000 @@ -298,8 +298,6 @@ static int worker_thread(void *__cwq) if (!cwq->wq->freezeable) current->flags |= PF_NOFREEZE; - set_user_nice(current, -5); - for (;;) { prepare_to_wait(&cwq->more_work, &wait, TASK_INTERRUPTIBLE); if (!freezing(current) && !cwq->should_stop Index: linux-2.6.22-rc2-ck1/kernel/kthread.c =================================================================== --- linux-2.6.22-rc2-ck1.orig/kernel/kthread.c 2007-05-19 20:10:51.000000000 +1000 +++ linux-2.6.22-rc2-ck1/kernel/kthread.c 2007-05-19 20:11:57.000000000 +1000 @@ -220,7 +220,6 @@ static __init void kthreadd_setup(void) ignore_signals(tsk); - set_user_nice(tsk, -5); set_cpus_allowed(tsk, CPU_MASK_ALL); }