From d88c16919793a9f5dc93e7956da5bb089c7600b4 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 18 Jul 2008 08:59:24 +0200 Subject: Revert parts of "ftrace: do not trace scheduler functions" the removal of -mno-spe in the !ftrace case was not intended. Signed-off-by: Ingo Molnar --- kernel/Makefile | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/Makefile b/kernel/Makefile index 985ddb7da4d..15ab63ffe64 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -11,6 +11,8 @@ obj-y = sched.o fork.o exec_domain.o panic.o printk.o profile.o \ hrtimer.o rwsem.o nsproxy.o srcu.o semaphore.o \ notifier.o ksysfs.o pm_qos_params.o sched_clock.o +CFLAGS_REMOVE_sched.o = -mno-spe + ifdef CONFIG_FTRACE # Do not trace debug files and internal ftrace files CFLAGS_REMOVE_lockdep.o = -pg -- cgit v1.2.3-18-g5258 From 13b40c1e40f3261e83ee514a08b77dbecb93021b Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Tue, 1 Jul 2008 10:32:50 -0700 Subject: sched: reduce stack size in isolated_cpu_setup() * Remove 16k stack requirements in isolated_cpu_setup when NR_CPUS=4096. Signed-off-by: Mike Travis Cc: Andrew Morton Signed-off-by: Ingo Molnar --- kernel/sched.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/sched.c b/kernel/sched.c index 99e6d850eca..1ee18dbb451 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6768,7 +6768,8 @@ static cpumask_t cpu_isolated_map = CPU_MASK_NONE; /* Setup the mask of cpus configured for isolated domains */ static int __init isolated_cpu_setup(char *str) { - int ints[NR_CPUS], i; + static int __initdata ints[NR_CPUS]; + int i; str = get_options(str, ARRAY_SIZE(ints), ints); cpus_clear(cpu_isolated_map); -- cgit v1.2.3-18-g5258 From 7ebefa8ceefed44cc321be70afc54a585a68ac0b Mon Sep 17 00:00:00 2001 From: Dmitry Adamushko Date: Tue, 1 Jul 2008 23:32:15 +0200 Subject: sched: rework of "prioritize non-migratable tasks over migratable ones" (1) handle in a generic way all cases when a newly woken-up task is not migratable (not just a corner case when "rt_se->nr_cpus_allowed == 1") (2) if current is to be preempted, then make sure "p" will be picked up by pick_next_task_rt(). i.e. move task's group at the head of its list as well. currently, it's not a case for the group-scheduling case as described here: http://www.ussg.iu.edu/hypermail/linux/kernel/0807.0/0134.html Signed-off-by: Dmitry Adamushko Cc: Steven Rostedt Cc: Gregory Haskins Signed-off-by: Ingo Molnar --- kernel/sched_rt.c | 68 ++++++++++++++++++++++++++++++++----------------------- 1 file changed, 40 insertions(+), 28 deletions(-) (limited to 'kernel') diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 47ceac9e855..d3d1cccb3d7 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -599,11 +599,7 @@ static void __enqueue_rt_entity(struct sched_rt_entity *rt_se) if (group_rq && (rt_rq_throttled(group_rq) || !group_rq->rt_nr_running)) return; - if (rt_se->nr_cpus_allowed == 1) - list_add(&rt_se->run_list, queue); - else - list_add_tail(&rt_se->run_list, queue); - + list_add_tail(&rt_se->run_list, queue); __set_bit(rt_se_prio(rt_se), array->bitmap); inc_rt_tasks(rt_se, rt_rq); @@ -688,32 +684,34 @@ static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep) * Put task to the end of the run list without the overhead of dequeue * followed by enqueue. */ -static -void requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se) +static void +requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head) { - struct rt_prio_array *array = &rt_rq->active; - if (on_rt_rq(rt_se)) { - list_del_init(&rt_se->run_list); - list_add_tail(&rt_se->run_list, - array->queue + rt_se_prio(rt_se)); + struct rt_prio_array *array = &rt_rq->active; + struct list_head *queue = array->queue + rt_se_prio(rt_se); + + if (head) + list_move(&rt_se->run_list, queue); + else + list_move_tail(&rt_se->run_list, queue); } } -static void requeue_task_rt(struct rq *rq, struct task_struct *p) +static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head) { struct sched_rt_entity *rt_se = &p->rt; struct rt_rq *rt_rq; for_each_sched_rt_entity(rt_se) { rt_rq = rt_rq_of_se(rt_se); - requeue_rt_entity(rt_rq, rt_se); + requeue_rt_entity(rt_rq, rt_se, head); } } static void yield_task_rt(struct rq *rq) { - requeue_task_rt(rq, rq->curr); + requeue_task_rt(rq, rq->curr, 0); } #ifdef CONFIG_SMP @@ -753,6 +751,30 @@ static int select_task_rq_rt(struct task_struct *p, int sync) */ return task_cpu(p); } + +static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p) +{ + cpumask_t mask; + + if (rq->curr->rt.nr_cpus_allowed == 1) + return; + + if (p->rt.nr_cpus_allowed != 1 + && cpupri_find(&rq->rd->cpupri, p, &mask)) + return; + + if (!cpupri_find(&rq->rd->cpupri, rq->curr, &mask)) + return; + + /* + * There appears to be other cpus that can accept + * current and none to run 'p', so lets reschedule + * to try and push current away: + */ + requeue_task_rt(rq, p, 1); + resched_task(rq->curr); +} + #endif /* CONFIG_SMP */ /* @@ -778,18 +800,8 @@ static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p) * to move current somewhere else, making room for our non-migratable * task. */ - if((p->prio == rq->curr->prio) - && p->rt.nr_cpus_allowed == 1 - && rq->curr->rt.nr_cpus_allowed != 1) { - cpumask_t mask; - - if (cpupri_find(&rq->rd->cpupri, rq->curr, &mask)) - /* - * There appears to be other cpus that can accept - * current, so lets reschedule to try and push it away - */ - resched_task(rq->curr); - } + if (p->prio == rq->curr->prio && !need_resched()) + check_preempt_equal_prio(rq, p); #endif } @@ -1415,7 +1427,7 @@ static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued) * on the queue: */ if (p->rt.run_list.prev != p->rt.run_list.next) { - requeue_task_rt(rq, p); + requeue_task_rt(rq, p, 0); set_tsk_need_resched(p); } } -- cgit v1.2.3-18-g5258 From e761b7725234276a802322549cee5255305a0930 Mon Sep 17 00:00:00 2001 From: Max Krasnyansky Date: Tue, 15 Jul 2008 04:43:49 -0700 Subject: cpu hotplug, sched: Introduce cpu_active_map and redo sched domain managment (take 2) This is based on Linus' idea of creating cpu_active_map that prevents scheduler load balancer from migrating tasks to the cpu that is going down. It allows us to simplify domain management code and avoid unecessary domain rebuilds during cpu hotplug event handling. Please ignore the cpusets part for now. It needs some more work in order to avoid crazy lock nesting. Although I did simplfy and unify domain reinitialization logic. We now simply call partition_sched_domains() in all the cases. This means that we're using exact same code paths as in cpusets case and hence the test below cover cpusets too. Cpuset changes to make rebuild_sched_domains() callable from various contexts are in the separate patch (right next after this one). This not only boots but also easily handles while true; do make clean; make -j 8; done and while true; do on-off-cpu 1; done at the same time. (on-off-cpu 1 simple does echo 0/1 > /sys/.../cpu1/online thing). Suprisingly the box (dual-core Core2) is quite usable. In fact I'm typing this on right now in gnome-terminal and things are moving just fine. Also this is running with most of the debug features enabled (lockdep, mutex, etc) no BUG_ONs or lockdep complaints so far. I believe I addressed all of the Dmitry's comments for original Linus' version. I changed both fair and rt balancer to mask out non-active cpus. And replaced cpu_is_offline() with !cpu_active() in the main scheduler code where it made sense (to me). Signed-off-by: Max Krasnyanskiy Acked-by: Linus Torvalds Acked-by: Peter Zijlstra Acked-by: Gregory Haskins Cc: dmitry.adamushko@gmail.com Cc: pj@sgi.com Signed-off-by: Ingo Molnar --- kernel/cpu.c | 30 ++++++++++++--- kernel/cpuset.c | 2 +- kernel/sched.c | 108 ++++++++++++++++++++++------------------------------ kernel/sched_fair.c | 3 ++ kernel/sched_rt.c | 7 ++++ 5 files changed, 80 insertions(+), 70 deletions(-) (limited to 'kernel') diff --git a/kernel/cpu.c b/kernel/cpu.c index cfb1d43ab80..a1ac7ea245d 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -64,6 +64,8 @@ void __init cpu_hotplug_init(void) cpu_hotplug.refcount = 0; } +cpumask_t cpu_active_map; + #ifdef CONFIG_HOTPLUG_CPU void get_online_cpus(void) @@ -291,11 +293,20 @@ int __ref cpu_down(unsigned int cpu) int err = 0; cpu_maps_update_begin(); - if (cpu_hotplug_disabled) + + if (cpu_hotplug_disabled) { err = -EBUSY; - else - err = _cpu_down(cpu, 0); + goto out; + } + + cpu_clear(cpu, cpu_active_map); + + err = _cpu_down(cpu, 0); + + if (cpu_online(cpu)) + cpu_set(cpu, cpu_active_map); +out: cpu_maps_update_done(); return err; } @@ -355,11 +366,18 @@ int __cpuinit cpu_up(unsigned int cpu) } cpu_maps_update_begin(); - if (cpu_hotplug_disabled) + + if (cpu_hotplug_disabled) { err = -EBUSY; - else - err = _cpu_up(cpu, 0); + goto out; + } + err = _cpu_up(cpu, 0); + + if (cpu_online(cpu)) + cpu_set(cpu, cpu_active_map); + +out: cpu_maps_update_done(); return err; } diff --git a/kernel/cpuset.c b/kernel/cpuset.c index 459d601947a..3c3ef02f65f 100644 --- a/kernel/cpuset.c +++ b/kernel/cpuset.c @@ -564,7 +564,7 @@ update_domain_attr(struct sched_domain_attr *dattr, struct cpuset *c) * partition_sched_domains(). */ -static void rebuild_sched_domains(void) +void rebuild_sched_domains(void) { struct kfifo *q; /* queue of cpusets to be scanned */ struct cpuset *cp; /* scans q */ diff --git a/kernel/sched.c b/kernel/sched.c index 1ee18dbb451..c237624a8a0 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2881,7 +2881,7 @@ static void sched_migrate_task(struct task_struct *p, int dest_cpu) rq = task_rq_lock(p, &flags); if (!cpu_isset(dest_cpu, p->cpus_allowed) - || unlikely(cpu_is_offline(dest_cpu))) + || unlikely(!cpu_active(dest_cpu))) goto out; /* force the process onto the specified CPU */ @@ -3849,7 +3849,7 @@ int select_nohz_load_balancer(int stop_tick) /* * If we are going offline and still the leader, give up! */ - if (cpu_is_offline(cpu) && + if (!cpu_active(cpu) && atomic_read(&nohz.load_balancer) == cpu) { if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) BUG(); @@ -5876,7 +5876,7 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu) struct rq *rq_dest, *rq_src; int ret = 0, on_rq; - if (unlikely(cpu_is_offline(dest_cpu))) + if (unlikely(!cpu_active(dest_cpu))) return ret; rq_src = cpu_rq(src_cpu); @@ -7553,18 +7553,6 @@ void __attribute__((weak)) arch_update_cpu_topology(void) { } -/* - * Free current domain masks. - * Called after all cpus are attached to NULL domain. - */ -static void free_sched_domains(void) -{ - ndoms_cur = 0; - if (doms_cur != &fallback_doms) - kfree(doms_cur); - doms_cur = &fallback_doms; -} - /* * Set up scheduler domains and groups. Callers must hold the hotplug lock. * For now this just excludes isolated cpus, but could be used to @@ -7643,7 +7631,7 @@ static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur, * ownership of it and will kfree it when done with it. If the caller * failed the kmalloc call, then it can pass in doms_new == NULL, * and partition_sched_domains() will fallback to the single partition - * 'fallback_doms'. + * 'fallback_doms', it also forces the domains to be rebuilt. * * Call with hotplug lock held */ @@ -7657,12 +7645,8 @@ void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, /* always unregister in case we don't destroy any domains */ unregister_sched_domain_sysctl(); - if (doms_new == NULL) { - ndoms_new = 1; - doms_new = &fallback_doms; - cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map); - dattr_new = NULL; - } + if (doms_new == NULL) + ndoms_new = 0; /* Destroy deleted domains */ for (i = 0; i < ndoms_cur; i++) { @@ -7677,6 +7661,14 @@ match1: ; } + if (doms_new == NULL) { + ndoms_cur = 0; + ndoms_new = 1; + doms_new = &fallback_doms; + cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map); + dattr_new = NULL; + } + /* Build new domains */ for (i = 0; i < ndoms_new; i++) { for (j = 0; j < ndoms_cur; j++) { @@ -7707,17 +7699,10 @@ match2: #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) int arch_reinit_sched_domains(void) { - int err; - get_online_cpus(); - mutex_lock(&sched_domains_mutex); - detach_destroy_domains(&cpu_online_map); - free_sched_domains(); - err = arch_init_sched_domains(&cpu_online_map); - mutex_unlock(&sched_domains_mutex); + rebuild_sched_domains(); put_online_cpus(); - - return err; + return 0; } static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt) @@ -7783,14 +7768,30 @@ int sched_create_sysfs_power_savings_entries(struct sysdev_class *cls) } #endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */ +#ifndef CONFIG_CPUSETS /* - * Force a reinitialization of the sched domains hierarchy. The domains - * and groups cannot be updated in place without racing with the balancing - * code, so we temporarily attach all running cpus to the NULL domain - * which will prevent rebalancing while the sched domains are recalculated. + * Add online and remove offline CPUs from the scheduler domains. + * When cpusets are enabled they take over this function. */ static int update_sched_domains(struct notifier_block *nfb, unsigned long action, void *hcpu) +{ + switch (action) { + case CPU_ONLINE: + case CPU_ONLINE_FROZEN: + case CPU_DEAD: + case CPU_DEAD_FROZEN: + partition_sched_domains(0, NULL, NULL); + return NOTIFY_OK; + + default: + return NOTIFY_DONE; + } +} +#endif + +static int update_runtime(struct notifier_block *nfb, + unsigned long action, void *hcpu) { int cpu = (int)(long)hcpu; @@ -7798,44 +7799,18 @@ static int update_sched_domains(struct notifier_block *nfb, case CPU_DOWN_PREPARE: case CPU_DOWN_PREPARE_FROZEN: disable_runtime(cpu_rq(cpu)); - /* fall-through */ - case CPU_UP_PREPARE: - case CPU_UP_PREPARE_FROZEN: - detach_destroy_domains(&cpu_online_map); - free_sched_domains(); return NOTIFY_OK; - case CPU_DOWN_FAILED: case CPU_DOWN_FAILED_FROZEN: case CPU_ONLINE: case CPU_ONLINE_FROZEN: enable_runtime(cpu_rq(cpu)); - /* fall-through */ - case CPU_UP_CANCELED: - case CPU_UP_CANCELED_FROZEN: - case CPU_DEAD: - case CPU_DEAD_FROZEN: - /* - * Fall through and re-initialise the domains. - */ - break; + return NOTIFY_OK; + default: return NOTIFY_DONE; } - -#ifndef CONFIG_CPUSETS - /* - * Create default domain partitioning if cpusets are disabled. - * Otherwise we let cpusets rebuild the domains based on the - * current setup. - */ - - /* The hotplug lock is already held by cpu_up/cpu_down */ - arch_init_sched_domains(&cpu_online_map); -#endif - - return NOTIFY_OK; } void __init sched_init_smp(void) @@ -7855,8 +7830,15 @@ void __init sched_init_smp(void) cpu_set(smp_processor_id(), non_isolated_cpus); mutex_unlock(&sched_domains_mutex); put_online_cpus(); + +#ifndef CONFIG_CPUSETS /* XXX: Theoretical race here - CPU may be hotplugged now */ hotcpu_notifier(update_sched_domains, 0); +#endif + + /* RT runtime code needs to handle some hotplug events */ + hotcpu_notifier(update_runtime, 0); + init_hrtick(); /* Move init over to a non-isolated CPU */ diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index f2aa987027d..d924c679dfa 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -1004,6 +1004,8 @@ static void yield_task_fair(struct rq *rq) * not idle and an idle cpu is available. The span of cpus to * search starts with cpus closest then further out as needed, * so we always favor a closer, idle cpu. + * Domains may include CPUs that are not usable for migration, + * hence we need to mask them out (cpu_active_map) * * Returns the CPU we should wake onto. */ @@ -1031,6 +1033,7 @@ static int wake_idle(int cpu, struct task_struct *p) || ((sd->flags & SD_WAKE_IDLE_FAR) && !task_hot(p, task_rq(p)->clock, sd))) { cpus_and(tmp, sd->span, p->cpus_allowed); + cpus_and(tmp, tmp, cpu_active_map); for_each_cpu_mask(i, tmp) { if (idle_cpu(i)) { if (i != task_cpu(p)) { diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index d3d1cccb3d7..50735bb9614 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -933,6 +933,13 @@ static int find_lowest_rq(struct task_struct *task) if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask)) return -1; /* No targets found */ + /* + * Only consider CPUs that are usable for migration. + * I guess we might want to change cpupri_find() to ignore those + * in the first place. + */ + cpus_and(*lowest_mask, *lowest_mask, cpu_active_map); + /* * At this point we have built a mask of cpus representing the * lowest priority tasks in the system. Now we want to elect -- cgit v1.2.3-18-g5258 From 39b0fad7121eace85770e7a4c6dc35dfd2879768 Mon Sep 17 00:00:00 2001 From: Max Krasnyansky Date: Tue, 15 Jul 2008 20:56:26 -0700 Subject: cpu hotplug: Make cpu_active_map synchronization dependency clear This goes on top of the cpu_active_map (take 2) patch. Currently we depend on the stop_machine to provide nescessesary synchronization for the cpu_active_map updates. As Dmitry Adamushko pointed this is fragile and is not much clearer than the previous scheme. In other words we do not want to depend on the internal stop machine operation here. So make the synchronization rules clear by doing synchronize_sched() after clearing out cpu active bit. Tested on quad-Core2 with: while true; do for i in 1 2 3; do echo 0 > /sys/devices/system/cpu/cpu$i/online done for i in 1 2 3; do echo 1 > /sys/devices/system/cpu/cpu$i/online done done and stress -c 200 No lockdep, preempt or other complaints. Signed-off-by: Max Krasnyansky Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/cpu.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'kernel') diff --git a/kernel/cpu.c b/kernel/cpu.c index a1ac7ea245d..033603c1d7c 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -301,6 +301,16 @@ int __ref cpu_down(unsigned int cpu) cpu_clear(cpu, cpu_active_map); + /* + * Make sure the all cpus did the reschedule and are not + * using stale version of the cpu_active_map. + * This is not strictly necessary becuase stop_machine() + * that we run down the line already provides the required + * synchronization. But it's really a side effect and we do not + * want to depend on the innards of the stop_machine here. + */ + synchronize_sched(); + err = _cpu_down(cpu, 0); if (cpu_online(cpu)) -- cgit v1.2.3-18-g5258 From 577b4a58d2e74a4d48050eeea3e3f952ce04eb86 Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 11 Jul 2008 13:34:54 +0100 Subject: sched: fix warning in inc_rt_tasks() to not declare variable 'rq' if it's not needed Fix inc_rt_tasks() to not declare variable 'rq' if it's not needed. It is declared if CONFIG_SMP or CONFIG_RT_GROUP_SCHED, but only used if CONFIG_SMP. This is a consequence of patch 1f11eb6a8bc92536d9e93ead48fa3ffbd1478571 plus patch 1100ac91b6af02d8639d518fad5b434b1bf44ed6. Signed-off-by: David Howells Signed-off-by: Ingo Molnar --- kernel/sched_rt.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 47ceac9e855..147004c651c 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -505,7 +505,9 @@ void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) rt_rq->rt_nr_running++; #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED if (rt_se_prio(rt_se) < rt_rq->highest_prio) { +#ifdef CONFIG_SMP struct rq *rq = rq_of_rt_rq(rt_rq); +#endif rt_rq->highest_prio = rt_se_prio(rt_se); #ifdef CONFIG_SMP -- cgit v1.2.3-18-g5258 From 31656519e132f6612584815f128c83976a9aaaef Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 18 Jul 2008 18:01:23 +0200 Subject: sched, x86: clean up hrtick implementation random uvesafb failures were reported against Gentoo: http://bugs.gentoo.org/show_bug.cgi?id=222799 and Mihai Moldovan bisected it back to: > 8f4d37ec073c17e2d4aa8851df5837d798606d6f is first bad commit > commit 8f4d37ec073c17e2d4aa8851df5837d798606d6f > Author: Peter Zijlstra > Date: Fri Jan 25 21:08:29 2008 +0100 > > sched: high-res preemption tick Linus suspected it to be hrtick + vm86 interaction and observed: > Btw, Peter, Ingo: I think that commit is doing bad things. They aren't > _incorrect_ per se, but they are definitely bad. > > Why? > > Using random _TIF_WORK_MASK flags is really impolite for doing > "scheduling" work. There's a reason that arch/x86/kernel/entry_32.S > special-cases the _TIF_NEED_RESCHED flag: we don't want to exit out of > vm86 mode unnecessarily. > > See the "work_notifysig_v86" label, and how it does that > "save_v86_state()" thing etc etc. Right, I never liked having to fiddle with those TIF flags. Initially I needed it because the hrtimer base lock could not nest in the rq lock. That however is fixed these days. Currently the only reason left to fiddle with the TIF flags is remote wakeups. We cannot program a remote cpu's hrtimer. I've been thinking about using the new and improved IPI function call stuff to implement hrtimer_start_on(). However that does require that smp_call_function_single(.wait=0) works from interrupt context - /me looks at the latest series from Jens - Yes that does seem to be supported, good. Here's a stab at cleaning this stuff up ... Mihai reported test success as well. Signed-off-by: Peter Zijlstra Tested-by: Mihai Moldovan Cc: Michal Januszewski Cc: Antonino Daplas Signed-off-by: Ingo Molnar --- kernel/Kconfig.hz | 2 +- kernel/sched.c | 202 ++++++++++++++++------------------------------------ kernel/sched_fair.c | 5 +- 3 files changed, 63 insertions(+), 146 deletions(-) (limited to 'kernel') diff --git a/kernel/Kconfig.hz b/kernel/Kconfig.hz index 526128a2e62..2a202a84675 100644 --- a/kernel/Kconfig.hz +++ b/kernel/Kconfig.hz @@ -55,4 +55,4 @@ config HZ default 1000 if HZ_1000 config SCHED_HRTICK - def_bool HIGH_RES_TIMERS && X86 + def_bool HIGH_RES_TIMERS diff --git a/kernel/sched.c b/kernel/sched.c index 1ee18dbb451..c13c75e9f9f 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -571,8 +571,10 @@ struct rq { #endif #ifdef CONFIG_SCHED_HRTICK - unsigned long hrtick_flags; - ktime_t hrtick_expire; +#ifdef CONFIG_SMP + int hrtick_csd_pending; + struct call_single_data hrtick_csd; +#endif struct hrtimer hrtick_timer; #endif @@ -983,13 +985,6 @@ static struct rq *this_rq_lock(void) return rq; } -static void __resched_task(struct task_struct *p, int tif_bit); - -static inline void resched_task(struct task_struct *p) -{ - __resched_task(p, TIF_NEED_RESCHED); -} - #ifdef CONFIG_SCHED_HRTICK /* * Use HR-timers to deliver accurate preemption points. @@ -1001,25 +996,6 @@ static inline void resched_task(struct task_struct *p) * When we get rescheduled we reprogram the hrtick_timer outside of the * rq->lock. */ -static inline void resched_hrt(struct task_struct *p) -{ - __resched_task(p, TIF_HRTICK_RESCHED); -} - -static inline void resched_rq(struct rq *rq) -{ - unsigned long flags; - - spin_lock_irqsave(&rq->lock, flags); - resched_task(rq->curr); - spin_unlock_irqrestore(&rq->lock, flags); -} - -enum { - HRTICK_SET, /* re-programm hrtick_timer */ - HRTICK_RESET, /* not a new slice */ - HRTICK_BLOCK, /* stop hrtick operations */ -}; /* * Use hrtick when: @@ -1030,72 +1006,17 @@ static inline int hrtick_enabled(struct rq *rq) { if (!sched_feat(HRTICK)) return 0; - if (unlikely(test_bit(HRTICK_BLOCK, &rq->hrtick_flags))) + if (!cpu_online(cpu_of(rq))) return 0; return hrtimer_is_hres_active(&rq->hrtick_timer); } -/* - * Called to set the hrtick timer state. - * - * called with rq->lock held and irqs disabled - */ -static void hrtick_start(struct rq *rq, u64 delay, int reset) -{ - assert_spin_locked(&rq->lock); - - /* - * preempt at: now + delay - */ - rq->hrtick_expire = - ktime_add_ns(rq->hrtick_timer.base->get_time(), delay); - /* - * indicate we need to program the timer - */ - __set_bit(HRTICK_SET, &rq->hrtick_flags); - if (reset) - __set_bit(HRTICK_RESET, &rq->hrtick_flags); - - /* - * New slices are called from the schedule path and don't need a - * forced reschedule. - */ - if (reset) - resched_hrt(rq->curr); -} - static void hrtick_clear(struct rq *rq) { if (hrtimer_active(&rq->hrtick_timer)) hrtimer_cancel(&rq->hrtick_timer); } -/* - * Update the timer from the possible pending state. - */ -static void hrtick_set(struct rq *rq) -{ - ktime_t time; - int set, reset; - unsigned long flags; - - WARN_ON_ONCE(cpu_of(rq) != smp_processor_id()); - - spin_lock_irqsave(&rq->lock, flags); - set = __test_and_clear_bit(HRTICK_SET, &rq->hrtick_flags); - reset = __test_and_clear_bit(HRTICK_RESET, &rq->hrtick_flags); - time = rq->hrtick_expire; - clear_thread_flag(TIF_HRTICK_RESCHED); - spin_unlock_irqrestore(&rq->lock, flags); - - if (set) { - hrtimer_start(&rq->hrtick_timer, time, HRTIMER_MODE_ABS); - if (reset && !hrtimer_active(&rq->hrtick_timer)) - resched_rq(rq); - } else - hrtick_clear(rq); -} - /* * High-resolution timer tick. * Runs from hardirq context with interrupts disabled. @@ -1115,27 +1036,37 @@ static enum hrtimer_restart hrtick(struct hrtimer *timer) } #ifdef CONFIG_SMP -static void hotplug_hrtick_disable(int cpu) +/* + * called from hardirq (IPI) context + */ +static void __hrtick_start(void *arg) { - struct rq *rq = cpu_rq(cpu); - unsigned long flags; - - spin_lock_irqsave(&rq->lock, flags); - rq->hrtick_flags = 0; - __set_bit(HRTICK_BLOCK, &rq->hrtick_flags); - spin_unlock_irqrestore(&rq->lock, flags); + struct rq *rq = arg; - hrtick_clear(rq); + spin_lock(&rq->lock); + hrtimer_restart(&rq->hrtick_timer); + rq->hrtick_csd_pending = 0; + spin_unlock(&rq->lock); } -static void hotplug_hrtick_enable(int cpu) +/* + * Called to set the hrtick timer state. + * + * called with rq->lock held and irqs disabled + */ +static void hrtick_start(struct rq *rq, u64 delay) { - struct rq *rq = cpu_rq(cpu); - unsigned long flags; + struct hrtimer *timer = &rq->hrtick_timer; + ktime_t time = ktime_add_ns(timer->base->get_time(), delay); - spin_lock_irqsave(&rq->lock, flags); - __clear_bit(HRTICK_BLOCK, &rq->hrtick_flags); - spin_unlock_irqrestore(&rq->lock, flags); + timer->expires = time; + + if (rq == this_rq()) { + hrtimer_restart(timer); + } else if (!rq->hrtick_csd_pending) { + __smp_call_function_single(cpu_of(rq), &rq->hrtick_csd); + rq->hrtick_csd_pending = 1; + } } static int @@ -1150,16 +1081,7 @@ hotplug_hrtick(struct notifier_block *nfb, unsigned long action, void *hcpu) case CPU_DOWN_PREPARE_FROZEN: case CPU_DEAD: case CPU_DEAD_FROZEN: - hotplug_hrtick_disable(cpu); - return NOTIFY_OK; - - case CPU_UP_PREPARE: - case CPU_UP_PREPARE_FROZEN: - case CPU_DOWN_FAILED: - case CPU_DOWN_FAILED_FROZEN: - case CPU_ONLINE: - case CPU_ONLINE_FROZEN: - hotplug_hrtick_enable(cpu); + hrtick_clear(cpu_rq(cpu)); return NOTIFY_OK; } @@ -1170,46 +1092,45 @@ static void init_hrtick(void) { hotcpu_notifier(hotplug_hrtick, 0); } -#endif /* CONFIG_SMP */ +#else +/* + * Called to set the hrtick timer state. + * + * called with rq->lock held and irqs disabled + */ +static void hrtick_start(struct rq *rq, u64 delay) +{ + hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay), HRTIMER_MODE_REL); +} -static void init_rq_hrtick(struct rq *rq) +static void init_hrtick(void) { - rq->hrtick_flags = 0; - hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); - rq->hrtick_timer.function = hrtick; - rq->hrtick_timer.cb_mode = HRTIMER_CB_IRQSAFE_NO_SOFTIRQ; } +#endif /* CONFIG_SMP */ -void hrtick_resched(void) +static void init_rq_hrtick(struct rq *rq) { - struct rq *rq; - unsigned long flags; +#ifdef CONFIG_SMP + rq->hrtick_csd_pending = 0; - if (!test_thread_flag(TIF_HRTICK_RESCHED)) - return; + rq->hrtick_csd.flags = 0; + rq->hrtick_csd.func = __hrtick_start; + rq->hrtick_csd.info = rq; +#endif - local_irq_save(flags); - rq = cpu_rq(smp_processor_id()); - hrtick_set(rq); - local_irq_restore(flags); + hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + rq->hrtick_timer.function = hrtick; + rq->hrtick_timer.cb_mode = HRTIMER_CB_IRQSAFE_NO_SOFTIRQ; } #else static inline void hrtick_clear(struct rq *rq) { } -static inline void hrtick_set(struct rq *rq) -{ -} - static inline void init_rq_hrtick(struct rq *rq) { } -void hrtick_resched(void) -{ -} - static inline void init_hrtick(void) { } @@ -1228,16 +1149,16 @@ static inline void init_hrtick(void) #define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG) #endif -static void __resched_task(struct task_struct *p, int tif_bit) +static void resched_task(struct task_struct *p) { int cpu; assert_spin_locked(&task_rq(p)->lock); - if (unlikely(test_tsk_thread_flag(p, tif_bit))) + if (unlikely(test_tsk_thread_flag(p, TIF_NEED_RESCHED))) return; - set_tsk_thread_flag(p, tif_bit); + set_tsk_thread_flag(p, TIF_NEED_RESCHED); cpu = task_cpu(p); if (cpu == smp_processor_id()) @@ -1303,10 +1224,10 @@ void wake_up_idle_cpu(int cpu) #endif /* CONFIG_NO_HZ */ #else /* !CONFIG_SMP */ -static void __resched_task(struct task_struct *p, int tif_bit) +static void resched_task(struct task_struct *p) { assert_spin_locked(&task_rq(p)->lock); - set_tsk_thread_flag(p, tif_bit); + set_tsk_need_resched(p); } #endif /* CONFIG_SMP */ @@ -4395,7 +4316,7 @@ asmlinkage void __sched schedule(void) struct task_struct *prev, *next; unsigned long *switch_count; struct rq *rq; - int cpu, hrtick = sched_feat(HRTICK); + int cpu; need_resched: preempt_disable(); @@ -4410,7 +4331,7 @@ need_resched_nonpreemptible: schedule_debug(prev); - if (hrtick) + if (sched_feat(HRTICK)) hrtick_clear(rq); /* @@ -4457,9 +4378,6 @@ need_resched_nonpreemptible: } else spin_unlock_irq(&rq->lock); - if (hrtick) - hrtick_set(rq); - if (unlikely(reacquire_kernel_lock(current) < 0)) goto need_resched_nonpreemptible; diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index f2aa987027d..6893b3ed65f 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -878,7 +878,6 @@ entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued) #ifdef CONFIG_SCHED_HRTICK static void hrtick_start_fair(struct rq *rq, struct task_struct *p) { - int requeue = rq->curr == p; struct sched_entity *se = &p->se; struct cfs_rq *cfs_rq = cfs_rq_of(se); @@ -899,10 +898,10 @@ static void hrtick_start_fair(struct rq *rq, struct task_struct *p) * Don't schedule slices shorter than 10000ns, that just * doesn't make sense. Rely on vruntime for fairness. */ - if (!requeue) + if (rq->curr != p) delta = max(10000LL, delta); - hrtick_start(rq, delta, requeue); + hrtick_start(rq, delta); } } #else /* !CONFIG_SCHED_HRTICK */ -- cgit v1.2.3-18-g5258 From ba42059fbd0aa1ac91b582412b5fedb1258f241f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 20 Jul 2008 11:02:06 +0200 Subject: sched: hrtick_enabled() should use cpu_active() Peter pointed out that hrtick_enabled() should use cpu_active(). Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/sched.c b/kernel/sched.c index 85cf246cfdf..62b1b8ecb5c 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1006,7 +1006,7 @@ static inline int hrtick_enabled(struct rq *rq) { if (!sched_feat(HRTICK)) return 0; - if (!cpu_online(cpu_of(rq))) + if (!cpu_active(cpu_of(rq))) return 0; return hrtimer_is_hres_active(&rq->hrtick_timer); } -- cgit v1.2.3-18-g5258