From 332fbdbca3f7716c5620970755ae054d213bcc4e Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:45 -0400 Subject: lockup_detector: Touch_softlockup cleanups and softlockup_tick removal Just some code cleanup to make touch_softlockup clearer and remove the softlockup_tick function as it is no longer needed. Also remove the /proc softlockup_thres call as it has been changed to watchdog_thres. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-3-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/timer.c | 1 - 1 file changed, 1 deletion(-) (limited to 'kernel/timer.c') diff --git a/kernel/timer.c b/kernel/timer.c index aeb6a54f277..e8de5eb07a0 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -1225,7 +1225,6 @@ void run_local_timers(void) { hrtimer_run_queues(); raise_softirq(TIMER_SOFTIRQ); - softlockup_tick(); } /* -- cgit v1.2.3-70-g09d2 From 83cd4fe27ad8446619b2e030b171b858501de87d Mon Sep 17 00:00:00 2001 From: Venkatesh Pallipadi Date: Fri, 21 May 2010 17:09:41 -0700 Subject: sched: Change nohz idle load balancing logic to push model In the new push model, all idle CPUs indeed go into nohz mode. There is still the concept of idle load balancer (performing the load balancing on behalf of all the idle cpu's in the system). Busy CPU kicks the nohz balancer when any of the nohz CPUs need idle load balancing. The kickee CPU does the idle load balancing on behalf of all idle CPUs instead of the normal idle balance. This addresses the below two problems with the current nohz ilb logic: * the idle load balancer continued to have periodic ticks during idle and wokeup frequently, even though it did not have any rebalancing to do on behalf of any of the idle CPUs. * On x86 and CPUs that have APIC timer stoppage on idle CPUs, this periodic wakeup can result in a periodic additional interrupt on a CPU doing the timer broadcast. Also currently we are migrating the unpinned timers from an idle to the cpu doing idle load balancing (when all the cpus in the system are idle, there is no idle load balancing cpu and timers get added to the same idle cpu where the request was made. So the existing optimization works only on semi idle system). And In semi idle system, we no longer have periodic ticks on the idle load balancer CPU. Using that cpu will add more delays to the timers than intended (as that cpu's timer base may not be uptodate wrt jiffies etc). This was causing mysterious slowdowns during boot etc. For now, in the semi idle case, use the nearest busy cpu for migrating timers from an idle cpu. This is good for power-savings anyway. Signed-off-by: Venkatesh Pallipadi Signed-off-by: Suresh Siddha Signed-off-by: Peter Zijlstra Cc: Thomas Gleixner LKML-Reference: <1274486981.2840.46.camel@sbs-t61.sc.intel.com> Signed-off-by: Ingo Molnar --- include/linux/sched.h | 9 +- kernel/hrtimer.c | 8 +- kernel/sched.c | 34 ++++- kernel/sched_fair.c | 329 ++++++++++++++++++++++++++++------------------- kernel/time/tick-sched.c | 8 +- kernel/timer.c | 8 +- 6 files changed, 237 insertions(+), 159 deletions(-) (limited to 'kernel/timer.c') diff --git a/include/linux/sched.h b/include/linux/sched.h index c2d4316a04b..a3e5b1cd043 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -271,14 +271,11 @@ extern int runqueue_is_locked(int cpu); extern cpumask_var_t nohz_cpu_mask; #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ) -extern int select_nohz_load_balancer(int cpu); -extern int get_nohz_load_balancer(void); +extern void select_nohz_load_balancer(int stop_tick); +extern int get_nohz_timer_target(void); extern int nohz_ratelimit(int cpu); #else -static inline int select_nohz_load_balancer(int cpu) -{ - return 0; -} +static inline void select_nohz_load_balancer(int stop_tick) { } static inline int nohz_ratelimit(int cpu) { diff --git a/kernel/hrtimer.c b/kernel/hrtimer.c index 5c69e996bd0..e934339fbbe 100644 --- a/kernel/hrtimer.c +++ b/kernel/hrtimer.c @@ -144,12 +144,8 @@ struct hrtimer_clock_base *lock_hrtimer_base(const struct hrtimer *timer, static int hrtimer_get_target(int this_cpu, int pinned) { #ifdef CONFIG_NO_HZ - if (!pinned && get_sysctl_timer_migration() && idle_cpu(this_cpu)) { - int preferred_cpu = get_nohz_load_balancer(); - - if (preferred_cpu >= 0) - return preferred_cpu; - } + if (!pinned && get_sysctl_timer_migration() && idle_cpu(this_cpu)) + return get_nohz_timer_target(); #endif return this_cpu; } diff --git a/kernel/sched.c b/kernel/sched.c index a757f6b11cb..132950b33dd 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -460,7 +460,7 @@ struct rq { unsigned long last_load_update_tick; #ifdef CONFIG_NO_HZ u64 nohz_stamp; - unsigned char in_nohz_recently; + unsigned char nohz_balance_kick; #endif unsigned int skip_clock_update; @@ -1194,6 +1194,27 @@ static void resched_cpu(int cpu) } #ifdef CONFIG_NO_HZ +/* + * In the semi idle case, use the nearest busy cpu for migrating timers + * from an idle cpu. This is good for power-savings. + * + * We don't do similar optimization for completely idle system, as + * selecting an idle cpu will add more delays to the timers than intended + * (as that cpu's timer base may not be uptodate wrt jiffies etc). + */ +int get_nohz_timer_target(void) +{ + int cpu = smp_processor_id(); + int i; + struct sched_domain *sd; + + for_each_domain(cpu, sd) { + for_each_cpu(i, sched_domain_span(sd)) + if (!idle_cpu(i)) + return i; + } + return cpu; +} /* * When add_timer_on() enqueues a timer into the timer wheel of an * idle CPU then this timer might expire before the next timer event @@ -7791,6 +7812,10 @@ void __init sched_init(void) rq->idle_stamp = 0; rq->avg_idle = 2*sysctl_sched_migration_cost; rq_attach_root(rq, &def_root_domain); +#ifdef CONFIG_NO_HZ + rq->nohz_balance_kick = 0; + init_sched_softirq_csd(&per_cpu(remote_sched_softirq_cb, i)); +#endif #endif init_rq_hrtick(rq); atomic_set(&rq->nr_iowait, 0); @@ -7835,8 +7860,11 @@ void __init sched_init(void) zalloc_cpumask_var(&nohz_cpu_mask, GFP_NOWAIT); #ifdef CONFIG_SMP #ifdef CONFIG_NO_HZ - zalloc_cpumask_var(&nohz.cpu_mask, GFP_NOWAIT); - alloc_cpumask_var(&nohz.ilb_grp_nohz_mask, GFP_NOWAIT); + zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT); + alloc_cpumask_var(&nohz.grp_idle_mask, GFP_NOWAIT); + atomic_set(&nohz.load_balancer, nr_cpu_ids); + atomic_set(&nohz.first_pick_cpu, nr_cpu_ids); + atomic_set(&nohz.second_pick_cpu, nr_cpu_ids); #endif /* May be allocated at isolcpus cmdline parse time */ if (cpu_isolated_map == NULL) diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index 22b8b4f2b61..6ee2e0af665 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -3091,13 +3091,40 @@ out_unlock: } #ifdef CONFIG_NO_HZ + +static DEFINE_PER_CPU(struct call_single_data, remote_sched_softirq_cb); + +static void trigger_sched_softirq(void *data) +{ + raise_softirq_irqoff(SCHED_SOFTIRQ); +} + +static inline void init_sched_softirq_csd(struct call_single_data *csd) +{ + csd->func = trigger_sched_softirq; + csd->info = NULL; + csd->flags = 0; + csd->priv = 0; +} + +/* + * idle load balancing details + * - One of the idle CPUs nominates itself as idle load_balancer, while + * entering idle. + * - This idle load balancer CPU will also go into tickless mode when + * it is idle, just like all other idle CPUs + * - When one of the busy CPUs notice that there may be an idle rebalancing + * needed, they will kick the idle load balancer, which then does idle + * load balancing for all the idle CPUs. + */ static struct { atomic_t load_balancer; - cpumask_var_t cpu_mask; - cpumask_var_t ilb_grp_nohz_mask; -} nohz ____cacheline_aligned = { - .load_balancer = ATOMIC_INIT(-1), -}; + atomic_t first_pick_cpu; + atomic_t second_pick_cpu; + cpumask_var_t idle_cpus_mask; + cpumask_var_t grp_idle_mask; + unsigned long next_balance; /* in jiffy units */ +} nohz ____cacheline_aligned; int get_nohz_load_balancer(void) { @@ -3151,17 +3178,17 @@ static inline struct sched_domain *lowest_flag_domain(int cpu, int flag) */ static inline int is_semi_idle_group(struct sched_group *ilb_group) { - cpumask_and(nohz.ilb_grp_nohz_mask, nohz.cpu_mask, + cpumask_and(nohz.grp_idle_mask, nohz.idle_cpus_mask, sched_group_cpus(ilb_group)); /* * A sched_group is semi-idle when it has atleast one busy cpu * and atleast one idle cpu. */ - if (cpumask_empty(nohz.ilb_grp_nohz_mask)) + if (cpumask_empty(nohz.grp_idle_mask)) return 0; - if (cpumask_equal(nohz.ilb_grp_nohz_mask, sched_group_cpus(ilb_group))) + if (cpumask_equal(nohz.grp_idle_mask, sched_group_cpus(ilb_group))) return 0; return 1; @@ -3194,7 +3221,7 @@ static int find_new_ilb(int cpu) * Optimize for the case when we have no idle CPUs or only one * idle CPU. Don't walk the sched_domain hierarchy in such cases */ - if (cpumask_weight(nohz.cpu_mask) < 2) + if (cpumask_weight(nohz.idle_cpus_mask) < 2) goto out_done; for_each_flag_domain(cpu, sd, SD_POWERSAVINGS_BALANCE) { @@ -3202,7 +3229,7 @@ static int find_new_ilb(int cpu) do { if (is_semi_idle_group(ilb_group)) - return cpumask_first(nohz.ilb_grp_nohz_mask); + return cpumask_first(nohz.grp_idle_mask); ilb_group = ilb_group->next; @@ -3210,98 +3237,116 @@ static int find_new_ilb(int cpu) } out_done: - return cpumask_first(nohz.cpu_mask); + return nr_cpu_ids; } #else /* (CONFIG_SCHED_MC || CONFIG_SCHED_SMT) */ static inline int find_new_ilb(int call_cpu) { - return cpumask_first(nohz.cpu_mask); + return nr_cpu_ids; } #endif +/* + * Kick a CPU to do the nohz balancing, if it is time for it. We pick the + * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle + * CPU (if there is one). + */ +static void nohz_balancer_kick(int cpu) +{ + int ilb_cpu; + + nohz.next_balance++; + + ilb_cpu = get_nohz_load_balancer(); + + if (ilb_cpu >= nr_cpu_ids) { + ilb_cpu = cpumask_first(nohz.idle_cpus_mask); + if (ilb_cpu >= nr_cpu_ids) + return; + } + + if (!cpu_rq(ilb_cpu)->nohz_balance_kick) { + struct call_single_data *cp; + + cpu_rq(ilb_cpu)->nohz_balance_kick = 1; + cp = &per_cpu(remote_sched_softirq_cb, cpu); + __smp_call_function_single(ilb_cpu, cp, 0); + } + return; +} + /* * This routine will try to nominate the ilb (idle load balancing) * owner among the cpus whose ticks are stopped. ilb owner will do the idle - * load balancing on behalf of all those cpus. If all the cpus in the system - * go into this tickless mode, then there will be no ilb owner (as there is - * no need for one) and all the cpus will sleep till the next wakeup event - * arrives... - * - * For the ilb owner, tick is not stopped. And this tick will be used - * for idle load balancing. ilb owner will still be part of - * nohz.cpu_mask.. + * load balancing on behalf of all those cpus. * - * While stopping the tick, this cpu will become the ilb owner if there - * is no other owner. And will be the owner till that cpu becomes busy - * or if all cpus in the system stop their ticks at which point - * there is no need for ilb owner. + * When the ilb owner becomes busy, we will not have new ilb owner until some + * idle CPU wakes up and goes back to idle or some busy CPU tries to kick + * idle load balancing by kicking one of the idle CPUs. * - * When the ilb owner becomes busy, it nominates another owner, during the - * next busy scheduler_tick() + * Ticks are stopped for the ilb owner as well, with busy CPU kicking this + * ilb owner CPU in future (when there is a need for idle load balancing on + * behalf of all idle CPUs). */ -int select_nohz_load_balancer(int stop_tick) +void select_nohz_load_balancer(int stop_tick) { int cpu = smp_processor_id(); if (stop_tick) { - cpu_rq(cpu)->in_nohz_recently = 1; - if (!cpu_active(cpu)) { if (atomic_read(&nohz.load_balancer) != cpu) - return 0; + return; /* * If we are going offline and still the leader, * give up! */ - if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) + if (atomic_cmpxchg(&nohz.load_balancer, cpu, + nr_cpu_ids) != cpu) BUG(); - return 0; + return; } - cpumask_set_cpu(cpu, nohz.cpu_mask); + cpumask_set_cpu(cpu, nohz.idle_cpus_mask); - /* time for ilb owner also to sleep */ - if (cpumask_weight(nohz.cpu_mask) == num_active_cpus()) { - if (atomic_read(&nohz.load_balancer) == cpu) - atomic_set(&nohz.load_balancer, -1); - return 0; - } + if (atomic_read(&nohz.first_pick_cpu) == cpu) + atomic_cmpxchg(&nohz.first_pick_cpu, cpu, nr_cpu_ids); + if (atomic_read(&nohz.second_pick_cpu) == cpu) + atomic_cmpxchg(&nohz.second_pick_cpu, cpu, nr_cpu_ids); - if (atomic_read(&nohz.load_balancer) == -1) { - /* make me the ilb owner */ - if (atomic_cmpxchg(&nohz.load_balancer, -1, cpu) == -1) - return 1; - } else if (atomic_read(&nohz.load_balancer) == cpu) { + if (atomic_read(&nohz.load_balancer) >= nr_cpu_ids) { int new_ilb; - if (!(sched_smt_power_savings || - sched_mc_power_savings)) - return 1; + /* make me the ilb owner */ + if (atomic_cmpxchg(&nohz.load_balancer, nr_cpu_ids, + cpu) != nr_cpu_ids) + return; + /* * Check to see if there is a more power-efficient * ilb. */ new_ilb = find_new_ilb(cpu); if (new_ilb < nr_cpu_ids && new_ilb != cpu) { - atomic_set(&nohz.load_balancer, -1); + atomic_set(&nohz.load_balancer, nr_cpu_ids); resched_cpu(new_ilb); - return 0; + return; } - return 1; + return; } } else { - if (!cpumask_test_cpu(cpu, nohz.cpu_mask)) - return 0; + if (!cpumask_test_cpu(cpu, nohz.idle_cpus_mask)) + return; - cpumask_clear_cpu(cpu, nohz.cpu_mask); + cpumask_clear_cpu(cpu, nohz.idle_cpus_mask); if (atomic_read(&nohz.load_balancer) == cpu) - if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) + if (atomic_cmpxchg(&nohz.load_balancer, cpu, + nr_cpu_ids) != cpu) BUG(); } - return 0; + return; } #endif @@ -3383,11 +3428,101 @@ out: rq->next_balance = next_balance; } +#ifdef CONFIG_NO_HZ /* - * run_rebalance_domains is triggered when needed from the scheduler tick. - * In CONFIG_NO_HZ case, the idle load balance owner will do the + * In CONFIG_NO_HZ case, the idle balance kickee will do the * rebalancing for all the cpus for whom scheduler ticks are stopped. */ +static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) +{ + struct rq *this_rq = cpu_rq(this_cpu); + struct rq *rq; + int balance_cpu; + + if (idle != CPU_IDLE || !this_rq->nohz_balance_kick) + return; + + for_each_cpu(balance_cpu, nohz.idle_cpus_mask) { + if (balance_cpu == this_cpu) + continue; + + /* + * If this cpu gets work to do, stop the load balancing + * work being done for other cpus. Next load + * balancing owner will pick it up. + */ + if (need_resched()) { + this_rq->nohz_balance_kick = 0; + break; + } + + raw_spin_lock_irq(&this_rq->lock); + update_cpu_load(this_rq); + raw_spin_unlock_irq(&this_rq->lock); + + rebalance_domains(balance_cpu, CPU_IDLE); + + rq = cpu_rq(balance_cpu); + if (time_after(this_rq->next_balance, rq->next_balance)) + this_rq->next_balance = rq->next_balance; + } + nohz.next_balance = this_rq->next_balance; + this_rq->nohz_balance_kick = 0; +} + +/* + * Current heuristic for kicking the idle load balancer + * - first_pick_cpu is the one of the busy CPUs. It will kick + * idle load balancer when it has more than one process active. This + * eliminates the need for idle load balancing altogether when we have + * only one running process in the system (common case). + * - If there are more than one busy CPU, idle load balancer may have + * to run for active_load_balance to happen (i.e., two busy CPUs are + * SMT or core siblings and can run better if they move to different + * physical CPUs). So, second_pick_cpu is the second of the busy CPUs + * which will kick idle load balancer as soon as it has any load. + */ +static inline int nohz_kick_needed(struct rq *rq, int cpu) +{ + unsigned long now = jiffies; + int ret; + int first_pick_cpu, second_pick_cpu; + + if (time_before(now, nohz.next_balance)) + return 0; + + if (!rq->nr_running) + return 0; + + first_pick_cpu = atomic_read(&nohz.first_pick_cpu); + second_pick_cpu = atomic_read(&nohz.second_pick_cpu); + + if (first_pick_cpu < nr_cpu_ids && first_pick_cpu != cpu && + second_pick_cpu < nr_cpu_ids && second_pick_cpu != cpu) + return 0; + + ret = atomic_cmpxchg(&nohz.first_pick_cpu, nr_cpu_ids, cpu); + if (ret == nr_cpu_ids || ret == cpu) { + atomic_cmpxchg(&nohz.second_pick_cpu, cpu, nr_cpu_ids); + if (rq->nr_running > 1) + return 1; + } else { + ret = atomic_cmpxchg(&nohz.second_pick_cpu, nr_cpu_ids, cpu); + if (ret == nr_cpu_ids || ret == cpu) { + if (rq->nr_running) + return 1; + } + } + return 0; +} +#else +static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) { } +#endif + +/* + * run_rebalance_domains is triggered when needed from the scheduler tick. + * Also triggered for nohz idle balancing (with nohz_balancing_kick set). + */ static void run_rebalance_domains(struct softirq_action *h) { int this_cpu = smp_processor_id(); @@ -3397,40 +3532,12 @@ static void run_rebalance_domains(struct softirq_action *h) rebalance_domains(this_cpu, idle); -#ifdef CONFIG_NO_HZ /* - * If this cpu is the owner for idle load balancing, then do the + * If this cpu has a pending nohz_balance_kick, then do the * balancing on behalf of the other idle cpus whose ticks are * stopped. */ - if (this_rq->idle_at_tick && - atomic_read(&nohz.load_balancer) == this_cpu) { - struct rq *rq; - int balance_cpu; - - for_each_cpu(balance_cpu, nohz.cpu_mask) { - if (balance_cpu == this_cpu) - continue; - - /* - * If this cpu gets work to do, stop the load balancing - * work being done for other cpus. Next load - * balancing owner will pick it up. - */ - if (need_resched()) - break; - - rq = cpu_rq(balance_cpu); - raw_spin_lock_irq(&rq->lock); - update_cpu_load(rq); - raw_spin_unlock_irq(&rq->lock); - rebalance_domains(balance_cpu, CPU_IDLE); - - if (time_after(this_rq->next_balance, rq->next_balance)) - this_rq->next_balance = rq->next_balance; - } - } -#endif + nohz_idle_balance(this_cpu, idle); } static inline int on_null_domain(int cpu) @@ -3440,57 +3547,17 @@ static inline int on_null_domain(int cpu) /* * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. - * - * In case of CONFIG_NO_HZ, this is the place where we nominate a new - * idle load balancing owner or decide to stop the periodic load balancing, - * if the whole system is idle. */ static inline void trigger_load_balance(struct rq *rq, int cpu) { -#ifdef CONFIG_NO_HZ - /* - * If we were in the nohz mode recently and busy at the current - * scheduler tick, then check if we need to nominate new idle - * load balancer. - */ - if (rq->in_nohz_recently && !rq->idle_at_tick) { - rq->in_nohz_recently = 0; - - if (atomic_read(&nohz.load_balancer) == cpu) { - cpumask_clear_cpu(cpu, nohz.cpu_mask); - atomic_set(&nohz.load_balancer, -1); - } - - if (atomic_read(&nohz.load_balancer) == -1) { - int ilb = find_new_ilb(cpu); - - if (ilb < nr_cpu_ids) - resched_cpu(ilb); - } - } - - /* - * If this cpu is idle and doing idle load balancing for all the - * cpus with ticks stopped, is it time for that to stop? - */ - if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) == cpu && - cpumask_weight(nohz.cpu_mask) == num_online_cpus()) { - resched_cpu(cpu); - return; - } - - /* - * If this cpu is idle and the idle load balancing is done by - * someone else, then no need raise the SCHED_SOFTIRQ - */ - if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) != cpu && - cpumask_test_cpu(cpu, nohz.cpu_mask)) - return; -#endif /* Don't need to rebalance while attached to NULL domain */ if (time_after_eq(jiffies, rq->next_balance) && likely(!on_null_domain(cpu))) raise_softirq(SCHED_SOFTIRQ); +#ifdef CONFIG_NO_HZ + else if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu))) + nohz_balancer_kick(cpu); +#endif } static void rq_online_fair(struct rq *rq) diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 1d7b9bc1c03..5f171f04ab0 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -408,13 +408,7 @@ void tick_nohz_stop_sched_tick(int inidle) * the scheduler tick in nohz_restart_sched_tick. */ if (!ts->tick_stopped) { - if (select_nohz_load_balancer(1)) { - /* - * sched tick not stopped! - */ - cpumask_clear_cpu(cpu, nohz_cpu_mask); - goto out; - } + select_nohz_load_balancer(1); ts->idle_tick = hrtimer_get_expires(&ts->sched_timer); ts->tick_stopped = 1; diff --git a/kernel/timer.c b/kernel/timer.c index ee305c8d4e1..48d6aec0789 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -679,12 +679,8 @@ __mod_timer(struct timer_list *timer, unsigned long expires, cpu = smp_processor_id(); #if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) - if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) { - int preferred_cpu = get_nohz_load_balancer(); - - if (preferred_cpu >= 0) - cpu = preferred_cpu; - } + if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) + cpu = get_nohz_timer_target(); #endif new_base = per_cpu(tvec_bases, cpu); -- cgit v1.2.3-70-g09d2 From 866e26115cba6b59cec669b6307599e3e4440491 Mon Sep 17 00:00:00 2001 From: "J. Bruce Fields" Date: Tue, 20 Jul 2010 15:23:15 -0700 Subject: timers: Document meaning of deferrable timer Steal some text from 6e453a67510 "Add support for deferrable timers". A reader shouldn't have to dig through the git logs for the basic description of a deferrable timer. Signed-off-by: J. Bruce Fields Cc: johnstul@us.ibm.com Signed-off-by: Andrew Morton Signed-off-by: Thomas Gleixner --- kernel/timer.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'kernel/timer.c') diff --git a/kernel/timer.c b/kernel/timer.c index ee305c8d4e1..ce98685cd1c 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -90,8 +90,13 @@ static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; /* * Note that all tvec_bases are 2 byte aligned and lower bit of - * base in timer_list is guaranteed to be zero. Use the LSB for - * the new flag to indicate whether the timer is deferrable + * base in timer_list is guaranteed to be zero. Use the LSB to + * indicate whether the timer is deferrable. + * + * A deferrable timer will work normally when the system is busy, but + * will not cause a CPU to come out of idle just to service it; instead, + * the timer will be serviced when the CPU eventually wakes up with a + * subsequent non-deferrable timer. */ #define TBASE_DEFERRABLE_FLAG (0x1) -- cgit v1.2.3-70-g09d2 From 22b8f15c2f7130bb0386f548428df2ffd4e81903 Mon Sep 17 00:00:00 2001 From: Patrick Pannuto Date: Mon, 19 Jul 2010 15:09:26 -0700 Subject: timer: Added usleep[_range] timer usleep[_range] are finer precision implementations of msleep and are designed to be drop-in replacements for udelay where a precise sleep / busy-wait is unnecessary. They also allow an easy interface to specify slack when a precise (ish) wakeup is unnecessary to help minimize wakeups Signed-off-by: Patrick Pannuto Cc: akinobu.mita@gmail.com Cc: sboyd@codeaurora.org Acked-by: Arjan van de Ven LKML-Reference: <4C44CDD2.1070708@codeaurora.org> Signed-off-by: Thomas Gleixner --- include/linux/delay.h | 6 ++++++ kernel/timer.c | 22 ++++++++++++++++++++++ 2 files changed, 28 insertions(+) (limited to 'kernel/timer.c') diff --git a/include/linux/delay.h b/include/linux/delay.h index fd832c6d419..0e303d1aacd 100644 --- a/include/linux/delay.h +++ b/include/linux/delay.h @@ -45,6 +45,12 @@ extern unsigned long lpj_fine; void calibrate_delay(void); void msleep(unsigned int msecs); unsigned long msleep_interruptible(unsigned int msecs); +void usleep_range(unsigned long min, unsigned long max); + +static inline void usleep(unsigned long usecs) +{ + usleep_range(usecs, usecs); +} static inline void ssleep(unsigned int seconds) { diff --git a/kernel/timer.c b/kernel/timer.c index ce98685cd1c..f110f241ab6 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -1755,3 +1755,25 @@ unsigned long msleep_interruptible(unsigned int msecs) } EXPORT_SYMBOL(msleep_interruptible); + +static int __sched do_usleep_range(unsigned long min, unsigned long max) +{ + ktime_t kmin; + unsigned long delta; + + kmin = ktime_set(0, min * NSEC_PER_USEC); + delta = max - min; + return schedule_hrtimeout_range(&kmin, delta, HRTIMER_MODE_REL); +} + +/** + * usleep_range - Drop in replacement for udelay where wakeup is flexible + * @min: Minimum time in usecs to sleep + * @max: Maximum time in usecs to sleep + */ +void usleep_range(unsigned long min, unsigned long max) +{ + __set_current_state(TASK_UNINTERRUPTIBLE); + do_usleep_range(min, max); +} +EXPORT_SYMBOL(usleep_range); -- cgit v1.2.3-70-g09d2 From 8cadd2831bf3abc94f4530e7fdbab7bb39b6b27d Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Mon, 10 May 2010 14:26:20 -0700 Subject: timer: add on-stack deferrable timer interfaces In some cases (for instance with kernel threads) it may be desireable to use on-stack deferrable timers to get their power saving benefits. Add interfaces to support this for the IPS driver. Signed-off-by: Jesse Barnes Signed-off-by: Matthew Garrett --- include/linux/timer.h | 15 +++++++++++++++ kernel/timer.c | 13 +++++++++++++ 2 files changed, 28 insertions(+) (limited to 'kernel/timer.c') diff --git a/include/linux/timer.h b/include/linux/timer.h index ea965b857a5..38cf093ef62 100644 --- a/include/linux/timer.h +++ b/include/linux/timer.h @@ -100,6 +100,13 @@ void init_timer_deferrable_key(struct timer_list *timer, setup_timer_on_stack_key((timer), #timer, &__key, \ (fn), (data)); \ } while (0) +#define setup_deferrable_timer_on_stack(timer, fn, data) \ + do { \ + static struct lock_class_key __key; \ + setup_deferrable_timer_on_stack_key((timer), #timer, \ + &__key, (fn), \ + (data)); \ + } while (0) #else #define init_timer(timer)\ init_timer_key((timer), NULL, NULL) @@ -111,6 +118,8 @@ void init_timer_deferrable_key(struct timer_list *timer, setup_timer_key((timer), NULL, NULL, (fn), (data)) #define setup_timer_on_stack(timer, fn, data)\ setup_timer_on_stack_key((timer), NULL, NULL, (fn), (data)) +#define setup_deferrable_timer_on_stack(timer, fn, data)\ + setup_deferrable_timer_on_stack_key((timer), NULL, NULL, (fn), (data)) #endif #ifdef CONFIG_DEBUG_OBJECTS_TIMERS @@ -150,6 +159,12 @@ static inline void setup_timer_on_stack_key(struct timer_list *timer, init_timer_on_stack_key(timer, name, key); } +extern void setup_deferrable_timer_on_stack_key(struct timer_list *timer, + const char *name, + struct lock_class_key *key, + void (*function)(unsigned long), + unsigned long data); + /** * timer_pending - is a timer pending? * @timer: the timer in question diff --git a/kernel/timer.c b/kernel/timer.c index ee305c8d4e1..efde11e197c 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -577,6 +577,19 @@ static void __init_timer(struct timer_list *timer, lockdep_init_map(&timer->lockdep_map, name, key, 0); } +void setup_deferrable_timer_on_stack_key(struct timer_list *timer, + const char *name, + struct lock_class_key *key, + void (*function)(unsigned long), + unsigned long data) +{ + timer->function = function; + timer->data = data; + init_timer_on_stack_key(timer, name, key); + timer_set_deferrable(timer); +} +EXPORT_SYMBOL_GPL(setup_deferrable_timer_on_stack_key); + /** * init_timer_key - initialize a timer * @timer: the timer to be initialized -- cgit v1.2.3-70-g09d2 From e1b004c3ef9c59db5f013528628b51c8653155ec Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 4 Aug 2010 10:53:00 +0200 Subject: Revert "timer: Added usleep[_range] timer" This reverts commit 22b8f15c2f7130bb0386f548428df2ffd4e81903 to merge an advanced version. Signed-off-by: Thomas Gleixner --- include/linux/delay.h | 6 ------ kernel/timer.c | 22 ---------------------- 2 files changed, 28 deletions(-) (limited to 'kernel/timer.c') diff --git a/include/linux/delay.h b/include/linux/delay.h index 0e303d1aacd..fd832c6d419 100644 --- a/include/linux/delay.h +++ b/include/linux/delay.h @@ -45,12 +45,6 @@ extern unsigned long lpj_fine; void calibrate_delay(void); void msleep(unsigned int msecs); unsigned long msleep_interruptible(unsigned int msecs); -void usleep_range(unsigned long min, unsigned long max); - -static inline void usleep(unsigned long usecs) -{ - usleep_range(usecs, usecs); -} static inline void ssleep(unsigned int seconds) { diff --git a/kernel/timer.c b/kernel/timer.c index f110f241ab6..ce98685cd1c 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -1755,25 +1755,3 @@ unsigned long msleep_interruptible(unsigned int msecs) } EXPORT_SYMBOL(msleep_interruptible); - -static int __sched do_usleep_range(unsigned long min, unsigned long max) -{ - ktime_t kmin; - unsigned long delta; - - kmin = ktime_set(0, min * NSEC_PER_USEC); - delta = max - min; - return schedule_hrtimeout_range(&kmin, delta, HRTIMER_MODE_REL); -} - -/** - * usleep_range - Drop in replacement for udelay where wakeup is flexible - * @min: Minimum time in usecs to sleep - * @max: Maximum time in usecs to sleep - */ -void usleep_range(unsigned long min, unsigned long max) -{ - __set_current_state(TASK_UNINTERRUPTIBLE); - do_usleep_range(min, max); -} -EXPORT_SYMBOL(usleep_range); -- cgit v1.2.3-70-g09d2 From 5e7f5a178bba45c5aca3448fddecabd4e28f1f6b Mon Sep 17 00:00:00 2001 From: Patrick Pannuto Date: Mon, 2 Aug 2010 15:01:04 -0700 Subject: timer: Added usleep_range timer usleep_range is a finer precision implementations of msleep and is designed to be a drop-in replacement for udelay where a precise sleep / busy-wait is unnecessary. Since an easy interface to hrtimers could lead to an undesired proliferation of interrupts, we provide only a "range" API, forcing the caller to think about an acceptable tolerance on both ends and hopefully avoiding introducing another interrupt. INTRO As discussed here ( http://lkml.org/lkml/2007/8/3/250 ), msleep(1) is not precise enough for many drivers (yes, sleep precision is an unfair notion, but consistently sleeping for ~an order of magnitude greater than requested is worth fixing). This patch adds a usleep API so that udelay does not have to be used. Obviously not every udelay can be replaced (those in atomic contexts or being used for simple bitbanging come to mind), but there are many, many examples of mydriver_write(...) /* Wait for hardware to latch */ udelay(100) in various drivers where a busy-wait loop is neither beneficial nor necessary, but msleep simply does not provide enough precision and people are using a busy-wait loop instead. CONCERNS FROM THE RFC Why is udelay a problem / necessary? Most callers of udelay are in device/ driver initialization code, which is serial... As I see it, there is only benefit to sleeping over a delay; the notion of "refactoring" areas that use udelay was presented, but I see usleep as the refactoring. Consider i2c, if the bus is busy, you need to wait a bit (say 100us) before trying again, your current options are: * udelay(100) * msleep(1) <-- As noted above, actually as high as ~20ms on some platforms, so not really an option * Manually set up an hrtimer to try again in 100us (which is what usleep does anyway...) People choose the udelay route because it is EASY; we need to provide a better easy route. Device / driver / boot code is *currently* serial, but every few months someone makes noise about parallelizing boot, and IMHO, a little forward-thinking now is one less thing to worry about if/when that ever happens udelay's could be preempted Sure, but if udelay plans on looping 1000 times, and it gets preempted on loop 200, whenever it's scheduled again, it is going to do the next 800 loops. Is the interruptible case needed? Probably not, but I see usleep as a very logical parallel to msleep, so it made sense to include the "full" API. Processors are getting faster (albeit not as quickly as they are becoming more parallel), so if someone wanted to be interruptible for a few usecs, why not let them? If this is a contentious point, I'm happy to remove it. OTHER THOUGHTS I believe there is also value in exposing the usleep_range option; it gives the scheduler a lot more flexibility and allows the programmer to express his intent much more clearly; it's something I would hope future driver writers will take advantage of. To get the results in the NUMBERS section below, I literally s/udelay/usleep the kernel tree; I had to go in and undo the changes to the USB drivers, but everything else booted successfully; I find that extremely telling in and of itself -- many people are using a delay API where a sleep will suit them just fine. SOME ATTEMPTS AT NUMBERS It turns out that calculating quantifiable benefit on this is challenging, so instead I will simply present the current state of things, and I hope this to be sufficient: How many udelay calls are there in 2.6.35-rc5? udealy(ARG) >= | COUNT 1000 | 319 500 | 414 100 | 1146 20 | 1832 I am working on Android, so that is my focus for this. The following table is a modified usleep that simply printk's the amount of time requested to sleep; these tests were run on a kernel with udelay >= 20 --> usleep "boot" is power-on to lock screen "power collapse" is when the power button is pushed and the device suspends "resume" is when the power button is pushed and the lock screen is displayed (no touchscreen events or anything, just turning on the display) "use device" is from the unlock swipe to clicking around a bit; there is no sd card in this phone, so fail loading music, video, camera ACTION | TOTAL NUMBER OF USLEEP CALLS | NET TIME (us) boot | 22 | 1250 power-collapse | 9 | 1200 resume | 5 | 500 use device | 59 | 7700 The most interesting category to me is the "use device" field; 7700us of busy-wait time that could be put towards better responsiveness, or at the least less power usage. Signed-off-by: Patrick Pannuto Cc: apw@canonical.com Cc: corbet@lwn.net Cc: arjan@linux.intel.com Cc: Randy Dunlap Cc: Andrew Morton Signed-off-by: Thomas Gleixner --- include/linux/delay.h | 1 + kernel/timer.c | 22 ++++++++++++++++++++++ 2 files changed, 23 insertions(+) (limited to 'kernel/timer.c') diff --git a/include/linux/delay.h b/include/linux/delay.h index fd832c6d419..a6ecb34cf54 100644 --- a/include/linux/delay.h +++ b/include/linux/delay.h @@ -45,6 +45,7 @@ extern unsigned long lpj_fine; void calibrate_delay(void); void msleep(unsigned int msecs); unsigned long msleep_interruptible(unsigned int msecs); +void usleep_range(unsigned long min, unsigned long max); static inline void ssleep(unsigned int seconds) { diff --git a/kernel/timer.c b/kernel/timer.c index ce98685cd1c..723a62e86dc 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -1755,3 +1755,25 @@ unsigned long msleep_interruptible(unsigned int msecs) } EXPORT_SYMBOL(msleep_interruptible); + +static int __sched do_usleep_range(unsigned long min, unsigned long max) +{ + ktime_t kmin; + unsigned long delta; + + kmin = ktime_set(0, min * NSEC_PER_USEC); + delta = (max - min) * NSEC_PER_USEC; + return schedule_hrtimeout_range(&kmin, delta, HRTIMER_MODE_REL); +} + +/** + * usleep_range - Drop in replacement for udelay where wakeup is flexible + * @min: Minimum time in usecs to sleep + * @max: Maximum time in usecs to sleep + */ +void usleep_range(unsigned long min, unsigned long max) +{ + __set_current_state(TASK_UNINTERRUPTIBLE); + do_usleep_range(min, max); +} +EXPORT_SYMBOL(usleep_range); -- cgit v1.2.3-70-g09d2 From 0caa621065b2cc05d4e53655a34fd989f500b040 Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Mon, 9 Aug 2010 16:32:50 -0700 Subject: kernel/timer.c: fix kernel-doc function parameter warning Fix kernel-doc warning, add @timer description: Warning(kernel/timer.c:335): No description found for parameter 'timer' Signed-off-by: Randy Dunlap Cc: Thomas Gleixner Signed-off-by: Linus Torvalds --- kernel/timer.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel/timer.c') diff --git a/kernel/timer.c b/kernel/timer.c index f1b8afe1ad8..97bf05baade 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -326,6 +326,7 @@ EXPORT_SYMBOL_GPL(round_jiffies_up_relative); /** * set_timer_slack - set the allowed slack for a timer + * @timer: the timer to be modified * @slack_hz: the amount of time (in jiffies) allowed for rounding * * Set the amount of time, in jiffies, that a certain timer has -- cgit v1.2.3-70-g09d2 From e360adbe29241a0194e10e20595360dd7b98a2b3 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 14 Oct 2010 14:01:34 +0800 Subject: irq_work: Add generic hardirq context callbacks Provide a mechanism that allows running code in IRQ context. It is most useful for NMI code that needs to interact with the rest of the system -- like wakeup a task to drain buffers. Perf currently has such a mechanism, so extract that and provide it as a generic feature, independent of perf so that others may also benefit. The IRQ context callback is generated through self-IPIs where possible, or on architectures like powerpc the decrementer (the built-in timer facility) is set to generate an interrupt immediately. Architectures that don't have anything like this get to do with a callback from the timer tick. These architectures can call irq_work_run() at the tail of any IRQ handlers that might enqueue such work (like the perf IRQ handler) to avoid undue latencies in processing the work. Signed-off-by: Peter Zijlstra Acked-by: Kyle McMartin Acked-by: Martin Schwidefsky [ various fixes ] Signed-off-by: Huang Ying LKML-Reference: <1287036094.7768.291.camel@yhuang-dev> Signed-off-by: Ingo Molnar --- arch/alpha/Kconfig | 1 + arch/alpha/include/asm/perf_event.h | 5 -- arch/alpha/kernel/time.c | 30 +++---- arch/arm/Kconfig | 1 + arch/arm/include/asm/perf_event.h | 12 --- arch/arm/kernel/perf_event.c | 8 +- arch/frv/Kconfig | 1 + arch/frv/lib/Makefile | 2 +- arch/frv/lib/perf_event.c | 19 ---- arch/parisc/Kconfig | 1 + arch/parisc/include/asm/perf_event.h | 3 +- arch/powerpc/Kconfig | 1 + arch/powerpc/include/asm/paca.h | 2 +- arch/powerpc/kernel/time.c | 42 ++++----- arch/s390/Kconfig | 1 + arch/s390/include/asm/perf_event.h | 3 +- arch/sh/Kconfig | 1 + arch/sh/include/asm/perf_event.h | 7 -- arch/sparc/Kconfig | 2 + arch/sparc/include/asm/perf_event.h | 4 - arch/sparc/kernel/pcr.c | 8 +- arch/x86/Kconfig | 1 + arch/x86/include/asm/entry_arch.h | 4 +- arch/x86/include/asm/hardirq.h | 2 +- arch/x86/include/asm/hw_irq.h | 2 +- arch/x86/include/asm/irq_vectors.h | 4 +- arch/x86/kernel/Makefile | 1 + arch/x86/kernel/cpu/perf_event.c | 19 ---- arch/x86/kernel/entry_64.S | 6 +- arch/x86/kernel/irq.c | 8 +- arch/x86/kernel/irq_work.c | 30 +++++++ arch/x86/kernel/irqinit.c | 6 +- include/linux/irq_work.h | 20 +++++ include/linux/perf_event.h | 11 +-- init/Kconfig | 8 ++ kernel/Makefile | 2 + kernel/irq_work.c | 164 +++++++++++++++++++++++++++++++++++ kernel/perf_event.c | 104 ++-------------------- kernel/timer.c | 7 +- 39 files changed, 311 insertions(+), 242 deletions(-) delete mode 100644 arch/frv/lib/perf_event.c create mode 100644 arch/x86/kernel/irq_work.c create mode 100644 include/linux/irq_work.h create mode 100644 kernel/irq_work.c (limited to 'kernel/timer.c') diff --git a/arch/alpha/Kconfig b/arch/alpha/Kconfig index b9647bb66d1..d04ccd73af4 100644 --- a/arch/alpha/Kconfig +++ b/arch/alpha/Kconfig @@ -9,6 +9,7 @@ config ALPHA select HAVE_IDE select HAVE_OPROFILE select HAVE_SYSCALL_WRAPPERS + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select HAVE_DMA_ATTRS help diff --git a/arch/alpha/include/asm/perf_event.h b/arch/alpha/include/asm/perf_event.h index 4157cd3c44a..fe792ca818f 100644 --- a/arch/alpha/include/asm/perf_event.h +++ b/arch/alpha/include/asm/perf_event.h @@ -1,11 +1,6 @@ #ifndef __ASM_ALPHA_PERF_EVENT_H #define __ASM_ALPHA_PERF_EVENT_H -/* Alpha only supports software events through this interface. */ -extern void set_perf_event_pending(void); - -#define PERF_EVENT_INDEX_OFFSET 0 - #ifdef CONFIG_PERF_EVENTS extern void init_hw_perf_events(void); #else diff --git a/arch/alpha/kernel/time.c b/arch/alpha/kernel/time.c index 396af1799ea..0f1d8493cfc 100644 --- a/arch/alpha/kernel/time.c +++ b/arch/alpha/kernel/time.c @@ -41,7 +41,7 @@ #include #include #include -#include +#include #include #include @@ -83,25 +83,25 @@ static struct { unsigned long est_cycle_freq; -#ifdef CONFIG_PERF_EVENTS +#ifdef CONFIG_IRQ_WORK -DEFINE_PER_CPU(u8, perf_event_pending); +DEFINE_PER_CPU(u8, irq_work_pending); -#define set_perf_event_pending_flag() __get_cpu_var(perf_event_pending) = 1 -#define test_perf_event_pending() __get_cpu_var(perf_event_pending) -#define clear_perf_event_pending() __get_cpu_var(perf_event_pending) = 0 +#define set_irq_work_pending_flag() __get_cpu_var(irq_work_pending) = 1 +#define test_irq_work_pending() __get_cpu_var(irq_work_pending) +#define clear_irq_work_pending() __get_cpu_var(irq_work_pending) = 0 -void set_perf_event_pending(void) +void set_irq_work_pending(void) { - set_perf_event_pending_flag(); + set_irq_work_pending_flag(); } -#else /* CONFIG_PERF_EVENTS */ +#else /* CONFIG_IRQ_WORK */ -#define test_perf_event_pending() 0 -#define clear_perf_event_pending() +#define test_irq_work_pending() 0 +#define clear_irq_work_pending() -#endif /* CONFIG_PERF_EVENTS */ +#endif /* CONFIG_IRQ_WORK */ static inline __u32 rpcc(void) @@ -191,9 +191,9 @@ irqreturn_t timer_interrupt(int irq, void *dev) write_sequnlock(&xtime_lock); - if (test_perf_event_pending()) { - clear_perf_event_pending(); - perf_event_do_pending(); + if (test_irq_work_pending()) { + clear_irq_work_pending(); + irq_work_run(); } #ifndef CONFIG_SMP diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 88c97bc7a6f..7c0dfccd05b 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -23,6 +23,7 @@ config ARM select HAVE_KERNEL_GZIP select HAVE_KERNEL_LZO select HAVE_KERNEL_LZMA + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select PERF_USE_VMALLOC select HAVE_REGS_AND_STACK_ACCESS_API diff --git a/arch/arm/include/asm/perf_event.h b/arch/arm/include/asm/perf_event.h index b5799a3b711..c4aa4e8c6af 100644 --- a/arch/arm/include/asm/perf_event.h +++ b/arch/arm/include/asm/perf_event.h @@ -12,18 +12,6 @@ #ifndef __ARM_PERF_EVENT_H__ #define __ARM_PERF_EVENT_H__ -/* - * NOP: on *most* (read: all supported) ARM platforms, the performance - * counter interrupts are regular interrupts and not an NMI. This - * means that when we receive the interrupt we can call - * perf_event_do_pending() that handles all of the work with - * interrupts disabled. - */ -static inline void -set_perf_event_pending(void) -{ -} - /* ARM performance counters start from 1 (in the cp15 accesses) so use the * same indexes here for consistency. */ #define PERF_EVENT_INDEX_OFFSET 1 diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index 6cc6521881a..49643b1467e 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c @@ -1092,7 +1092,7 @@ armv6pmu_handle_irq(int irq_num, * platforms that can have the PMU interrupts raised as an NMI, this * will not work. */ - perf_event_do_pending(); + irq_work_run(); return IRQ_HANDLED; } @@ -2068,7 +2068,7 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev) * platforms that can have the PMU interrupts raised as an NMI, this * will not work. */ - perf_event_do_pending(); + irq_work_run(); return IRQ_HANDLED; } @@ -2436,7 +2436,7 @@ xscale1pmu_handle_irq(int irq_num, void *dev) armpmu->disable(hwc, idx); } - perf_event_do_pending(); + irq_work_run(); /* * Re-enable the PMU. @@ -2763,7 +2763,7 @@ xscale2pmu_handle_irq(int irq_num, void *dev) armpmu->disable(hwc, idx); } - perf_event_do_pending(); + irq_work_run(); /* * Re-enable the PMU. diff --git a/arch/frv/Kconfig b/arch/frv/Kconfig index 16399bd2499..0f2417df632 100644 --- a/arch/frv/Kconfig +++ b/arch/frv/Kconfig @@ -7,6 +7,7 @@ config FRV default y select HAVE_IDE select HAVE_ARCH_TRACEHOOK + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS config ZONE_DMA diff --git a/arch/frv/lib/Makefile b/arch/frv/lib/Makefile index f4709756d0d..4ff2fb1e6b1 100644 --- a/arch/frv/lib/Makefile +++ b/arch/frv/lib/Makefile @@ -5,4 +5,4 @@ lib-y := \ __ashldi3.o __lshrdi3.o __muldi3.o __ashrdi3.o __negdi2.o __ucmpdi2.o \ checksum.o memcpy.o memset.o atomic-ops.o atomic64-ops.o \ - outsl_ns.o outsl_sw.o insl_ns.o insl_sw.o cache.o perf_event.o + outsl_ns.o outsl_sw.o insl_ns.o insl_sw.o cache.o diff --git a/arch/frv/lib/perf_event.c b/arch/frv/lib/perf_event.c deleted file mode 100644 index 9ac5acfd2e9..00000000000 --- a/arch/frv/lib/perf_event.c +++ /dev/null @@ -1,19 +0,0 @@ -/* Performance event handling - * - * Copyright (C) 2009 Red Hat, Inc. All Rights Reserved. - * Written by David Howells (dhowells@redhat.com) - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public Licence - * as published by the Free Software Foundation; either version - * 2 of the Licence, or (at your option) any later version. - */ - -#include - -/* - * mark the performance event as pending - */ -void set_perf_event_pending(void) -{ -} diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig index 907417d187e..79a04a9394d 100644 --- a/arch/parisc/Kconfig +++ b/arch/parisc/Kconfig @@ -16,6 +16,7 @@ config PARISC select RTC_DRV_GENERIC select INIT_ALL_POSSIBLE select BUG + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select GENERIC_ATOMIC64 if !64BIT help diff --git a/arch/parisc/include/asm/perf_event.h b/arch/parisc/include/asm/perf_event.h index cc146427d8f..1e0fd8ba6c0 100644 --- a/arch/parisc/include/asm/perf_event.h +++ b/arch/parisc/include/asm/perf_event.h @@ -1,7 +1,6 @@ #ifndef __ASM_PARISC_PERF_EVENT_H #define __ASM_PARISC_PERF_EVENT_H -/* parisc only supports software events through this interface. */ -static inline void set_perf_event_pending(void) { } +/* Empty, just to avoid compiling error */ #endif /* __ASM_PARISC_PERF_EVENT_H */ diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig index 631e5a0fb6a..4b1e521d966 100644 --- a/arch/powerpc/Kconfig +++ b/arch/powerpc/Kconfig @@ -138,6 +138,7 @@ config PPC select HAVE_OPROFILE select HAVE_SYSCALL_WRAPPERS if PPC64 select GENERIC_ATOMIC64 if PPC32 + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select HAVE_REGS_AND_STACK_ACCESS_API select HAVE_HW_BREAKPOINT if PERF_EVENTS && PPC_BOOK3S_64 diff --git a/arch/powerpc/include/asm/paca.h b/arch/powerpc/include/asm/paca.h index 1ff6662f7fa..9b287fdd8ea 100644 --- a/arch/powerpc/include/asm/paca.h +++ b/arch/powerpc/include/asm/paca.h @@ -129,7 +129,7 @@ struct paca_struct { u8 soft_enabled; /* irq soft-enable flag */ u8 hard_enabled; /* set if irqs are enabled in MSR */ u8 io_sync; /* writel() needs spin_unlock sync */ - u8 perf_event_pending; /* PM interrupt while soft-disabled */ + u8 irq_work_pending; /* IRQ_WORK interrupt while soft-disable */ /* Stuff for accurate time accounting */ u64 user_time; /* accumulated usermode TB ticks */ diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c index 8533b3b83f5..54888eb10c3 100644 --- a/arch/powerpc/kernel/time.c +++ b/arch/powerpc/kernel/time.c @@ -53,7 +53,7 @@ #include #include #include -#include +#include #include #include @@ -493,60 +493,60 @@ void __init iSeries_time_init_early(void) } #endif /* CONFIG_PPC_ISERIES */ -#ifdef CONFIG_PERF_EVENTS +#ifdef CONFIG_IRQ_WORK /* * 64-bit uses a byte in the PACA, 32-bit uses a per-cpu variable... */ #ifdef CONFIG_PPC64 -static inline unsigned long test_perf_event_pending(void) +static inline unsigned long test_irq_work_pending(void) { unsigned long x; asm volatile("lbz %0,%1(13)" : "=r" (x) - : "i" (offsetof(struct paca_struct, perf_event_pending))); + : "i" (offsetof(struct paca_struct, irq_work_pending))); return x; } -static inline void set_perf_event_pending_flag(void) +static inline void set_irq_work_pending_flag(void) { asm volatile("stb %0,%1(13)" : : "r" (1), - "i" (offsetof(struct paca_struct, perf_event_pending))); + "i" (offsetof(struct paca_struct, irq_work_pending))); } -static inline void clear_perf_event_pending(void) +static inline void clear_irq_work_pending(void) { asm volatile("stb %0,%1(13)" : : "r" (0), - "i" (offsetof(struct paca_struct, perf_event_pending))); + "i" (offsetof(struct paca_struct, irq_work_pending))); } #else /* 32-bit */ -DEFINE_PER_CPU(u8, perf_event_pending); +DEFINE_PER_CPU(u8, irq_work_pending); -#define set_perf_event_pending_flag() __get_cpu_var(perf_event_pending) = 1 -#define test_perf_event_pending() __get_cpu_var(perf_event_pending) -#define clear_perf_event_pending() __get_cpu_var(perf_event_pending) = 0 +#define set_irq_work_pending_flag() __get_cpu_var(irq_work_pending) = 1 +#define test_irq_work_pending() __get_cpu_var(irq_work_pending) +#define clear_irq_work_pending() __get_cpu_var(irq_work_pending) = 0 #endif /* 32 vs 64 bit */ -void set_perf_event_pending(void) +void set_irq_work_pending(void) { preempt_disable(); - set_perf_event_pending_flag(); + set_irq_work_pending_flag(); set_dec(1); preempt_enable(); } -#else /* CONFIG_PERF_EVENTS */ +#else /* CONFIG_IRQ_WORK */ -#define test_perf_event_pending() 0 -#define clear_perf_event_pending() +#define test_irq_work_pending() 0 +#define clear_irq_work_pending() -#endif /* CONFIG_PERF_EVENTS */ +#endif /* CONFIG_IRQ_WORK */ /* * For iSeries shared processors, we have to let the hypervisor @@ -587,9 +587,9 @@ void timer_interrupt(struct pt_regs * regs) calculate_steal_time(); - if (test_perf_event_pending()) { - clear_perf_event_pending(); - perf_event_do_pending(); + if (test_irq_work_pending()) { + clear_irq_work_pending(); + irq_work_run(); } #ifdef CONFIG_PPC_ISERIES diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index f0777a47e3a..958f0dadead 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -95,6 +95,7 @@ config S390 select HAVE_KVM if 64BIT select HAVE_ARCH_TRACEHOOK select INIT_ALL_POSSIBLE + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select HAVE_KERNEL_GZIP select HAVE_KERNEL_BZIP2 diff --git a/arch/s390/include/asm/perf_event.h b/arch/s390/include/asm/perf_event.h index 3840cbe7763..a75f168d271 100644 --- a/arch/s390/include/asm/perf_event.h +++ b/arch/s390/include/asm/perf_event.h @@ -4,7 +4,6 @@ * Copyright 2009 Martin Schwidefsky, IBM Corporation. */ -static inline void set_perf_event_pending(void) {} -static inline void clear_perf_event_pending(void) {} +/* Empty, just to avoid compiling error */ #define PERF_EVENT_INDEX_OFFSET 0 diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig index 35b6c3f8517..35b6879628a 100644 --- a/arch/sh/Kconfig +++ b/arch/sh/Kconfig @@ -16,6 +16,7 @@ config SUPERH select HAVE_ARCH_TRACEHOOK select HAVE_DMA_API_DEBUG select HAVE_DMA_ATTRS + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select PERF_USE_VMALLOC select HAVE_KERNEL_GZIP diff --git a/arch/sh/include/asm/perf_event.h b/arch/sh/include/asm/perf_event.h index 3d0c9f36d15..14308bed7ea 100644 --- a/arch/sh/include/asm/perf_event.h +++ b/arch/sh/include/asm/perf_event.h @@ -26,11 +26,4 @@ extern int register_sh_pmu(struct sh_pmu *); extern int reserve_pmc_hardware(void); extern void release_pmc_hardware(void); -static inline void set_perf_event_pending(void) -{ - /* Nothing to see here, move along. */ -} - -#define PERF_EVENT_INDEX_OFFSET 0 - #endif /* __ASM_SH_PERF_EVENT_H */ diff --git a/arch/sparc/Kconfig b/arch/sparc/Kconfig index 9212cd42a83..3e9d31401fb 100644 --- a/arch/sparc/Kconfig +++ b/arch/sparc/Kconfig @@ -26,6 +26,7 @@ config SPARC select ARCH_WANT_OPTIONAL_GPIOLIB select RTC_CLASS select RTC_DRV_M48T59 + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select PERF_USE_VMALLOC select HAVE_DMA_ATTRS @@ -54,6 +55,7 @@ config SPARC64 select RTC_DRV_BQ4802 select RTC_DRV_SUN4V select RTC_DRV_STARFIRE + select HAVE_IRQ_WORK select HAVE_PERF_EVENTS select PERF_USE_VMALLOC diff --git a/arch/sparc/include/asm/perf_event.h b/arch/sparc/include/asm/perf_event.h index 727af70646c..6e8bfa1786d 100644 --- a/arch/sparc/include/asm/perf_event.h +++ b/arch/sparc/include/asm/perf_event.h @@ -1,10 +1,6 @@ #ifndef __ASM_SPARC_PERF_EVENT_H #define __ASM_SPARC_PERF_EVENT_H -extern void set_perf_event_pending(void); - -#define PERF_EVENT_INDEX_OFFSET 0 - #ifdef CONFIG_PERF_EVENTS #include diff --git a/arch/sparc/kernel/pcr.c b/arch/sparc/kernel/pcr.c index c4a6a50b484..b87873c0e8e 100644 --- a/arch/sparc/kernel/pcr.c +++ b/arch/sparc/kernel/pcr.c @@ -7,7 +7,7 @@ #include #include -#include +#include #include #include @@ -43,14 +43,14 @@ void __irq_entry deferred_pcr_work_irq(int irq, struct pt_regs *regs) old_regs = set_irq_regs(regs); irq_enter(); -#ifdef CONFIG_PERF_EVENTS - perf_event_do_pending(); +#ifdef CONFIG_IRQ_WORK + irq_work_run(); #endif irq_exit(); set_irq_regs(old_regs); } -void set_perf_event_pending(void) +void arch_irq_work_raise(void) { set_softint(1 << PIL_DEFERRED_PCR_WORK); } diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 9815221976a..fd227d6b8d9 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -25,6 +25,7 @@ config X86 select HAVE_IDE select HAVE_OPROFILE select HAVE_PERF_EVENTS if (!M386 && !M486) + select HAVE_IRQ_WORK select HAVE_IOREMAP_PROT select HAVE_KPROBES select ARCH_WANT_OPTIONAL_GPIOLIB diff --git a/arch/x86/include/asm/entry_arch.h b/arch/x86/include/asm/entry_arch.h index 8e8ec663a98..b8e96a18676 100644 --- a/arch/x86/include/asm/entry_arch.h +++ b/arch/x86/include/asm/entry_arch.h @@ -49,8 +49,8 @@ BUILD_INTERRUPT(apic_timer_interrupt,LOCAL_TIMER_VECTOR) BUILD_INTERRUPT(error_interrupt,ERROR_APIC_VECTOR) BUILD_INTERRUPT(spurious_interrupt,SPURIOUS_APIC_VECTOR) -#ifdef CONFIG_PERF_EVENTS -BUILD_INTERRUPT(perf_pending_interrupt, LOCAL_PENDING_VECTOR) +#ifdef CONFIG_IRQ_WORK +BUILD_INTERRUPT(irq_work_interrupt, IRQ_WORK_VECTOR) #endif #ifdef CONFIG_X86_THERMAL_VECTOR diff --git a/arch/x86/include/asm/hardirq.h b/arch/x86/include/asm/hardirq.h index aeab29aee61..55e4de613f0 100644 --- a/arch/x86/include/asm/hardirq.h +++ b/arch/x86/include/asm/hardirq.h @@ -14,7 +14,7 @@ typedef struct { #endif unsigned int x86_platform_ipis; /* arch dependent */ unsigned int apic_perf_irqs; - unsigned int apic_pending_irqs; + unsigned int apic_irq_work_irqs; #ifdef CONFIG_SMP unsigned int irq_resched_count; unsigned int irq_call_count; diff --git a/arch/x86/include/asm/hw_irq.h b/arch/x86/include/asm/hw_irq.h index 46c0fe05f23..3a54a1ca1a0 100644 --- a/arch/x86/include/asm/hw_irq.h +++ b/arch/x86/include/asm/hw_irq.h @@ -29,7 +29,7 @@ extern void apic_timer_interrupt(void); extern void x86_platform_ipi(void); extern void error_interrupt(void); -extern void perf_pending_interrupt(void); +extern void irq_work_interrupt(void); extern void spurious_interrupt(void); extern void thermal_interrupt(void); diff --git a/arch/x86/include/asm/irq_vectors.h b/arch/x86/include/asm/irq_vectors.h index e2ca3009255..6af0894dafb 100644 --- a/arch/x86/include/asm/irq_vectors.h +++ b/arch/x86/include/asm/irq_vectors.h @@ -114,9 +114,9 @@ #define X86_PLATFORM_IPI_VECTOR 0xed /* - * Performance monitoring pending work vector: + * IRQ work vector: */ -#define LOCAL_PENDING_VECTOR 0xec +#define IRQ_WORK_VECTOR 0xec #define UV_BAU_MESSAGE 0xea diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 9d3f485e5dd..7490bf8d145 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -35,6 +35,7 @@ obj-y := process_$(BITS).o signal.o entry_$(BITS).o obj-y += traps.o irq.o irq_$(BITS).o dumpstack_$(BITS).o obj-y += time.o ioport.o ldt.o dumpstack.o obj-y += setup.o x86_init.o i8259.o irqinit.o jump_label.o +obj-$(CONFIG_IRQ_WORK) += irq_work.o obj-$(CONFIG_X86_VISWS) += visws_quirks.o obj-$(CONFIG_X86_32) += probe_roms_32.o obj-$(CONFIG_X86_32) += sys_i386_32.o i386_ksyms_32.o diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index e2513f26ba8..fe73c1844a9 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1196,25 +1196,6 @@ static int x86_pmu_handle_irq(struct pt_regs *regs) return handled; } -void smp_perf_pending_interrupt(struct pt_regs *regs) -{ - irq_enter(); - ack_APIC_irq(); - inc_irq_stat(apic_pending_irqs); - perf_event_do_pending(); - irq_exit(); -} - -void set_perf_event_pending(void) -{ -#ifdef CONFIG_X86_LOCAL_APIC - if (!x86_pmu.apic || !x86_pmu_initialized()) - return; - - apic->send_IPI_self(LOCAL_PENDING_VECTOR); -#endif -} - void perf_events_lapic_init(void) { if (!x86_pmu.apic || !x86_pmu_initialized()) diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 17be5ec7cbb..c375c79065f 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -1023,9 +1023,9 @@ apicinterrupt ERROR_APIC_VECTOR \ apicinterrupt SPURIOUS_APIC_VECTOR \ spurious_interrupt smp_spurious_interrupt -#ifdef CONFIG_PERF_EVENTS -apicinterrupt LOCAL_PENDING_VECTOR \ - perf_pending_interrupt smp_perf_pending_interrupt +#ifdef CONFIG_IRQ_WORK +apicinterrupt IRQ_WORK_VECTOR \ + irq_work_interrupt smp_irq_work_interrupt #endif /* diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c index 91fd0c70a18..44edb03fc9e 100644 --- a/arch/x86/kernel/irq.c +++ b/arch/x86/kernel/irq.c @@ -67,10 +67,10 @@ static int show_other_interrupts(struct seq_file *p, int prec) for_each_online_cpu(j) seq_printf(p, "%10u ", irq_stats(j)->apic_perf_irqs); seq_printf(p, " Performance monitoring interrupts\n"); - seq_printf(p, "%*s: ", prec, "PND"); + seq_printf(p, "%*s: ", prec, "IWI"); for_each_online_cpu(j) - seq_printf(p, "%10u ", irq_stats(j)->apic_pending_irqs); - seq_printf(p, " Performance pending work\n"); + seq_printf(p, "%10u ", irq_stats(j)->apic_irq_work_irqs); + seq_printf(p, " IRQ work interrupts\n"); #endif if (x86_platform_ipi_callback) { seq_printf(p, "%*s: ", prec, "PLT"); @@ -185,7 +185,7 @@ u64 arch_irq_stat_cpu(unsigned int cpu) sum += irq_stats(cpu)->apic_timer_irqs; sum += irq_stats(cpu)->irq_spurious_count; sum += irq_stats(cpu)->apic_perf_irqs; - sum += irq_stats(cpu)->apic_pending_irqs; + sum += irq_stats(cpu)->apic_irq_work_irqs; #endif if (x86_platform_ipi_callback) sum += irq_stats(cpu)->x86_platform_ipis; diff --git a/arch/x86/kernel/irq_work.c b/arch/x86/kernel/irq_work.c new file mode 100644 index 00000000000..ca8f703a1e7 --- /dev/null +++ b/arch/x86/kernel/irq_work.c @@ -0,0 +1,30 @@ +/* + * x86 specific code for irq_work + * + * Copyright (C) 2010 Red Hat, Inc., Peter Zijlstra + */ + +#include +#include +#include +#include + +void smp_irq_work_interrupt(struct pt_regs *regs) +{ + irq_enter(); + ack_APIC_irq(); + inc_irq_stat(apic_irq_work_irqs); + irq_work_run(); + irq_exit(); +} + +void arch_irq_work_raise(void) +{ +#ifdef CONFIG_X86_LOCAL_APIC + if (!cpu_has_apic) + return; + + apic->send_IPI_self(IRQ_WORK_VECTOR); + apic_wait_icr_idle(); +#endif +} diff --git a/arch/x86/kernel/irqinit.c b/arch/x86/kernel/irqinit.c index 990ae7cfc57..713969b9266 100644 --- a/arch/x86/kernel/irqinit.c +++ b/arch/x86/kernel/irqinit.c @@ -224,9 +224,9 @@ static void __init apic_intr_init(void) alloc_intr_gate(SPURIOUS_APIC_VECTOR, spurious_interrupt); alloc_intr_gate(ERROR_APIC_VECTOR, error_interrupt); - /* Performance monitoring interrupts: */ -# ifdef CONFIG_PERF_EVENTS - alloc_intr_gate(LOCAL_PENDING_VECTOR, perf_pending_interrupt); + /* IRQ work interrupts: */ +# ifdef CONFIG_IRQ_WORK + alloc_intr_gate(IRQ_WORK_VECTOR, irq_work_interrupt); # endif #endif diff --git a/include/linux/irq_work.h b/include/linux/irq_work.h new file mode 100644 index 00000000000..4fa09d4d0b7 --- /dev/null +++ b/include/linux/irq_work.h @@ -0,0 +1,20 @@ +#ifndef _LINUX_IRQ_WORK_H +#define _LINUX_IRQ_WORK_H + +struct irq_work { + struct irq_work *next; + void (*func)(struct irq_work *); +}; + +static inline +void init_irq_work(struct irq_work *entry, void (*func)(struct irq_work *)) +{ + entry->next = NULL; + entry->func = func; +} + +bool irq_work_queue(struct irq_work *entry); +void irq_work_run(void); +void irq_work_sync(struct irq_work *entry); + +#endif /* _LINUX_IRQ_WORK_H */ diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index a9227e98520..2ebfc9ae475 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -486,6 +486,7 @@ struct perf_guest_info_callbacks { #include #include #include +#include #include #include @@ -672,11 +673,6 @@ struct perf_buffer { void *data_pages[0]; }; -struct perf_pending_entry { - struct perf_pending_entry *next; - void (*func)(struct perf_pending_entry *); -}; - struct perf_sample_data; typedef void (*perf_overflow_handler_t)(struct perf_event *, int, @@ -784,7 +780,7 @@ struct perf_event { int pending_wakeup; int pending_kill; int pending_disable; - struct perf_pending_entry pending; + struct irq_work pending; atomic_t event_limit; @@ -898,8 +894,6 @@ extern int perf_event_init_task(struct task_struct *child); extern void perf_event_exit_task(struct task_struct *child); extern void perf_event_free_task(struct task_struct *task); extern void perf_event_delayed_put(struct task_struct *task); -extern void set_perf_event_pending(void); -extern void perf_event_do_pending(void); extern void perf_event_print_debug(void); extern void perf_pmu_disable(struct pmu *pmu); extern void perf_pmu_enable(struct pmu *pmu); @@ -1078,7 +1072,6 @@ static inline int perf_event_init_task(struct task_struct *child) { return 0; } static inline void perf_event_exit_task(struct task_struct *child) { } static inline void perf_event_free_task(struct task_struct *task) { } static inline void perf_event_delayed_put(struct task_struct *task) { } -static inline void perf_event_do_pending(void) { } static inline void perf_event_print_debug(void) { } static inline int perf_event_task_disable(void) { return -EINVAL; } static inline int perf_event_task_enable(void) { return -EINVAL; } diff --git a/init/Kconfig b/init/Kconfig index 2de5b1cbadd..1ef0b439908 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -21,6 +21,13 @@ config CONSTRUCTORS depends on !UML default y +config HAVE_IRQ_WORK + bool + +config IRQ_WORK + bool + depends on HAVE_IRQ_WORK + menu "General setup" config EXPERIMENTAL @@ -987,6 +994,7 @@ config PERF_EVENTS default y if (PROFILING || PERF_COUNTERS) depends on HAVE_PERF_EVENTS select ANON_INODES + select IRQ_WORK help Enable kernel support for various performance events provided by software and hardware. diff --git a/kernel/Makefile b/kernel/Makefile index d52b473c99a..4d9bf5f8531 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -23,6 +23,7 @@ CFLAGS_REMOVE_rtmutex-debug.o = -pg CFLAGS_REMOVE_cgroup-debug.o = -pg CFLAGS_REMOVE_sched_clock.o = -pg CFLAGS_REMOVE_perf_event.o = -pg +CFLAGS_REMOVE_irq_work.o = -pg endif obj-$(CONFIG_FREEZER) += freezer.o @@ -100,6 +101,7 @@ obj-$(CONFIG_TRACING) += trace/ obj-$(CONFIG_X86_DS) += trace/ obj-$(CONFIG_RING_BUFFER) += trace/ obj-$(CONFIG_SMP) += sched_cpupri.o +obj-$(CONFIG_IRQ_WORK) += irq_work.o obj-$(CONFIG_PERF_EVENTS) += perf_event.o obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o obj-$(CONFIG_USER_RETURN_NOTIFIER) += user-return-notifier.o diff --git a/kernel/irq_work.c b/kernel/irq_work.c new file mode 100644 index 00000000000..f16763ff848 --- /dev/null +++ b/kernel/irq_work.c @@ -0,0 +1,164 @@ +/* + * Copyright (C) 2010 Red Hat, Inc., Peter Zijlstra + * + * Provides a framework for enqueueing and running callbacks from hardirq + * context. The enqueueing is NMI-safe. + */ + +#include +#include +#include +#include + +/* + * An entry can be in one of four states: + * + * free NULL, 0 -> {claimed} : free to be used + * claimed NULL, 3 -> {pending} : claimed to be enqueued + * pending next, 3 -> {busy} : queued, pending callback + * busy NULL, 2 -> {free, claimed} : callback in progress, can be claimed + * + * We use the lower two bits of the next pointer to keep PENDING and BUSY + * flags. + */ + +#define IRQ_WORK_PENDING 1UL +#define IRQ_WORK_BUSY 2UL +#define IRQ_WORK_FLAGS 3UL + +static inline bool irq_work_is_set(struct irq_work *entry, int flags) +{ + return (unsigned long)entry->next & flags; +} + +static inline struct irq_work *irq_work_next(struct irq_work *entry) +{ + unsigned long next = (unsigned long)entry->next; + next &= ~IRQ_WORK_FLAGS; + return (struct irq_work *)next; +} + +static inline struct irq_work *next_flags(struct irq_work *entry, int flags) +{ + unsigned long next = (unsigned long)entry; + next |= flags; + return (struct irq_work *)next; +} + +static DEFINE_PER_CPU(struct irq_work *, irq_work_list); + +/* + * Claim the entry so that no one else will poke at it. + */ +static bool irq_work_claim(struct irq_work *entry) +{ + struct irq_work *next, *nflags; + + do { + next = entry->next; + if ((unsigned long)next & IRQ_WORK_PENDING) + return false; + nflags = next_flags(next, IRQ_WORK_FLAGS); + } while (cmpxchg(&entry->next, next, nflags) != next); + + return true; +} + + +void __weak arch_irq_work_raise(void) +{ + /* + * Lame architectures will get the timer tick callback + */ +} + +/* + * Queue the entry and raise the IPI if needed. + */ +static void __irq_work_queue(struct irq_work *entry) +{ + struct irq_work **head, *next; + + head = &get_cpu_var(irq_work_list); + + do { + next = *head; + /* Can assign non-atomic because we keep the flags set. */ + entry->next = next_flags(next, IRQ_WORK_FLAGS); + } while (cmpxchg(head, next, entry) != next); + + /* The list was empty, raise self-interrupt to start processing. */ + if (!irq_work_next(entry)) + arch_irq_work_raise(); + + put_cpu_var(irq_work_list); +} + +/* + * Enqueue the irq_work @entry, returns true on success, failure when the + * @entry was already enqueued by someone else. + * + * Can be re-enqueued while the callback is still in progress. + */ +bool irq_work_queue(struct irq_work *entry) +{ + if (!irq_work_claim(entry)) { + /* + * Already enqueued, can't do! + */ + return false; + } + + __irq_work_queue(entry); + return true; +} +EXPORT_SYMBOL_GPL(irq_work_queue); + +/* + * Run the irq_work entries on this cpu. Requires to be ran from hardirq + * context with local IRQs disabled. + */ +void irq_work_run(void) +{ + struct irq_work *list, **head; + + head = &__get_cpu_var(irq_work_list); + if (*head == NULL) + return; + + BUG_ON(!in_irq()); + BUG_ON(!irqs_disabled()); + + list = xchg(head, NULL); + while (list != NULL) { + struct irq_work *entry = list; + + list = irq_work_next(list); + + /* + * Clear the PENDING bit, after this point the @entry + * can be re-used. + */ + entry->next = next_flags(NULL, IRQ_WORK_BUSY); + entry->func(entry); + /* + * Clear the BUSY bit and return to the free state if + * no-one else claimed it meanwhile. + */ + cmpxchg(&entry->next, next_flags(NULL, IRQ_WORK_BUSY), NULL); + } +} +EXPORT_SYMBOL_GPL(irq_work_run); + +/* + * Synchronize against the irq_work @entry, ensures the entry is not + * currently in use. + */ +void irq_work_sync(struct irq_work *entry) +{ + WARN_ON_ONCE(irqs_disabled()); + + while (irq_work_is_set(entry, IRQ_WORK_BUSY)) + cpu_relax(); +} +EXPORT_SYMBOL_GPL(irq_work_sync); diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 634f86a4b2f..99b9700e74d 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -2206,12 +2206,11 @@ static void free_event_rcu(struct rcu_head *head) kfree(event); } -static void perf_pending_sync(struct perf_event *event); static void perf_buffer_put(struct perf_buffer *buffer); static void free_event(struct perf_event *event) { - perf_pending_sync(event); + irq_work_sync(&event->pending); if (!event->parent) { atomic_dec(&nr_events); @@ -3162,16 +3161,7 @@ void perf_event_wakeup(struct perf_event *event) } } -/* - * Pending wakeups - * - * Handle the case where we need to wakeup up from NMI (or rq->lock) context. - * - * The NMI bit means we cannot possibly take locks. Therefore, maintain a - * single linked list and use cmpxchg() to add entries lockless. - */ - -static void perf_pending_event(struct perf_pending_entry *entry) +static void perf_pending_event(struct irq_work *entry) { struct perf_event *event = container_of(entry, struct perf_event, pending); @@ -3187,89 +3177,6 @@ static void perf_pending_event(struct perf_pending_entry *entry) } } -#define PENDING_TAIL ((struct perf_pending_entry *)-1UL) - -static DEFINE_PER_CPU(struct perf_pending_entry *, perf_pending_head) = { - PENDING_TAIL, -}; - -static void perf_pending_queue(struct perf_pending_entry *entry, - void (*func)(struct perf_pending_entry *)) -{ - struct perf_pending_entry **head; - - if (cmpxchg(&entry->next, NULL, PENDING_TAIL) != NULL) - return; - - entry->func = func; - - head = &get_cpu_var(perf_pending_head); - - do { - entry->next = *head; - } while (cmpxchg(head, entry->next, entry) != entry->next); - - set_perf_event_pending(); - - put_cpu_var(perf_pending_head); -} - -static int __perf_pending_run(void) -{ - struct perf_pending_entry *list; - int nr = 0; - - list = xchg(&__get_cpu_var(perf_pending_head), PENDING_TAIL); - while (list != PENDING_TAIL) { - void (*func)(struct perf_pending_entry *); - struct perf_pending_entry *entry = list; - - list = list->next; - - func = entry->func; - entry->next = NULL; - /* - * Ensure we observe the unqueue before we issue the wakeup, - * so that we won't be waiting forever. - * -- see perf_not_pending(). - */ - smp_wmb(); - - func(entry); - nr++; - } - - return nr; -} - -static inline int perf_not_pending(struct perf_event *event) -{ - /* - * If we flush on whatever cpu we run, there is a chance we don't - * need to wait. - */ - get_cpu(); - __perf_pending_run(); - put_cpu(); - - /* - * Ensure we see the proper queue state before going to sleep - * so that we do not miss the wakeup. -- see perf_pending_handle() - */ - smp_rmb(); - return event->pending.next == NULL; -} - -static void perf_pending_sync(struct perf_event *event) -{ - wait_event(event->waitq, perf_not_pending(event)); -} - -void perf_event_do_pending(void) -{ - __perf_pending_run(); -} - /* * We assume there is only KVM supporting the callbacks. * Later on, we might change it to a list if there is @@ -3319,8 +3226,7 @@ static void perf_output_wakeup(struct perf_output_handle *handle) if (handle->nmi) { handle->event->pending_wakeup = 1; - perf_pending_queue(&handle->event->pending, - perf_pending_event); + irq_work_queue(&handle->event->pending); } else perf_event_wakeup(handle->event); } @@ -4356,8 +4262,7 @@ static int __perf_event_overflow(struct perf_event *event, int nmi, event->pending_kill = POLL_HUP; if (nmi) { event->pending_disable = 1; - perf_pending_queue(&event->pending, - perf_pending_event); + irq_work_queue(&event->pending); } else perf_event_disable(event); } @@ -5374,6 +5279,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, INIT_LIST_HEAD(&event->event_entry); INIT_LIST_HEAD(&event->sibling_list); init_waitqueue_head(&event->waitq); + init_irq_work(&event->pending, perf_pending_event); mutex_init(&event->mmap_mutex); diff --git a/kernel/timer.c b/kernel/timer.c index 97bf05baade..68a9ae7679b 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -37,7 +37,7 @@ #include #include #include -#include +#include #include #include @@ -1279,7 +1279,10 @@ void update_process_times(int user_tick) run_local_timers(); rcu_check_callbacks(cpu, user_tick); printk_tick(); - perf_event_do_pending(); +#ifdef CONFIG_IRQ_WORK + if (in_irq()) + irq_work_run(); +#endif scheduler_tick(); run_posix_cpu_timers(p); } -- cgit v1.2.3-70-g09d2