X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=drivers%2Fcpufreq%2Fcpufreq.c;h=a45cc89e387a3ee14ba2afcda2cb09ac5303b1f0;hb=refs%2Fheads%2Fvserver;hp=3e14a3a44415ee03bcbe13afd762a3ec85b55153;hpb=c7b5ebbddf7bcd3651947760f423e3783bbe6573;p=linux-2.6.git diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 3e14a3a44..a45cc89e3 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -4,13 +4,17 @@ * Copyright (C) 2001 Russell King * (C) 2002 - 2003 Dominik Brodowski * + * Oct 2005 - Ashok Raj + * Added handling for CPU hotplug + * Feb 2006 - Jacob Shin + * Fix handling for CPU hotplug -- affected CPUs + * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. * */ -#include #include #include #include @@ -23,37 +27,45 @@ #include #include #include +#include + +#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_CORE, \ + "cpufreq-core", msg) /** - * The "cpufreq driver" - the arch- or hardware-dependend low + * The "cpufreq driver" - the arch- or hardware-dependent low * level driver of CPUFreq support, and its spinlock. This lock * also protects the cpufreq_cpu_data array. */ -static struct cpufreq_driver *cpufreq_driver; -static struct cpufreq_policy *cpufreq_cpu_data[NR_CPUS]; -static spinlock_t cpufreq_driver_lock = SPIN_LOCK_UNLOCKED; +static struct cpufreq_driver *cpufreq_driver; +static struct cpufreq_policy *cpufreq_cpu_data[NR_CPUS]; +static DEFINE_SPINLOCK(cpufreq_driver_lock); /* internal prototypes */ static int __cpufreq_governor(struct cpufreq_policy *policy, unsigned int event); -static void handle_update(void *data); -static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci); +static void handle_update(struct work_struct *work); /** - * Two notifier lists: the "policy" list is involved in the - * validation process for a new CPU frequency policy; the + * Two notifier lists: the "policy" list is involved in the + * validation process for a new CPU frequency policy; the * "transition" list for kernel code that needs to handle * changes to devices when the CPU clock speed changes. * The mutex locks both lists. */ -static struct notifier_block *cpufreq_policy_notifier_list; -static struct notifier_block *cpufreq_transition_notifier_list; -static DECLARE_RWSEM (cpufreq_notifier_rwsem); +static BLOCKING_NOTIFIER_HEAD(cpufreq_policy_notifier_list); +static struct srcu_notifier_head cpufreq_transition_notifier_list; +static int __init init_cpufreq_transition_notifier_list(void) +{ + srcu_init_notifier_head(&cpufreq_transition_notifier_list); + return 0; +} +pure_initcall(init_cpufreq_transition_notifier_list); static LIST_HEAD(cpufreq_governor_list); -static DECLARE_MUTEX (cpufreq_governor_sem); +static DEFINE_MUTEX (cpufreq_governor_mutex); -static struct cpufreq_policy * cpufreq_cpu_get(unsigned int cpu) +struct cpufreq_policy *cpufreq_cpu_get(unsigned int cpu) { struct cpufreq_policy *data; unsigned long flags; @@ -80,24 +92,113 @@ static struct cpufreq_policy * cpufreq_cpu_get(unsigned int cpu) if (!kobject_get(&data->kobj)) goto err_out_put_module; - spin_unlock_irqrestore(&cpufreq_driver_lock, flags); - return data; - err_out_put_module: +err_out_put_module: module_put(cpufreq_driver->owner); - err_out_unlock: +err_out_unlock: spin_unlock_irqrestore(&cpufreq_driver_lock, flags); - err_out: +err_out: return NULL; } +EXPORT_SYMBOL_GPL(cpufreq_cpu_get); + -static void cpufreq_cpu_put(struct cpufreq_policy *data) +void cpufreq_cpu_put(struct cpufreq_policy *data) { kobject_put(&data->kobj); module_put(cpufreq_driver->owner); } +EXPORT_SYMBOL_GPL(cpufreq_cpu_put); + + +/********************************************************************* + * UNIFIED DEBUG HELPERS * + *********************************************************************/ +#ifdef CONFIG_CPU_FREQ_DEBUG + +/* what part(s) of the CPUfreq subsystem are debugged? */ +static unsigned int debug; + +/* is the debug output ratelimit'ed using printk_ratelimit? User can + * set or modify this value. + */ +static unsigned int debug_ratelimit = 1; + +/* is the printk_ratelimit'ing enabled? It's enabled after a successful + * loading of a cpufreq driver, temporarily disabled when a new policy + * is set, and disabled upon cpufreq driver removal + */ +static unsigned int disable_ratelimit = 1; +static DEFINE_SPINLOCK(disable_ratelimit_lock); + +static void cpufreq_debug_enable_ratelimit(void) +{ + unsigned long flags; + + spin_lock_irqsave(&disable_ratelimit_lock, flags); + if (disable_ratelimit) + disable_ratelimit--; + spin_unlock_irqrestore(&disable_ratelimit_lock, flags); +} + +static void cpufreq_debug_disable_ratelimit(void) +{ + unsigned long flags; + + spin_lock_irqsave(&disable_ratelimit_lock, flags); + disable_ratelimit++; + spin_unlock_irqrestore(&disable_ratelimit_lock, flags); +} + +void cpufreq_debug_printk(unsigned int type, const char *prefix, + const char *fmt, ...) +{ + char s[256]; + va_list args; + unsigned int len; + unsigned long flags; + + WARN_ON(!prefix); + if (type & debug) { + spin_lock_irqsave(&disable_ratelimit_lock, flags); + if (!disable_ratelimit && debug_ratelimit + && !printk_ratelimit()) { + spin_unlock_irqrestore(&disable_ratelimit_lock, flags); + return; + } + spin_unlock_irqrestore(&disable_ratelimit_lock, flags); + + len = snprintf(s, 256, KERN_DEBUG "%s: ", prefix); + + va_start(args, fmt); + len += vsnprintf(&s[len], (256 - len), fmt, args); + va_end(args); + + printk(s); + + WARN_ON(len < 5); + } +} +EXPORT_SYMBOL(cpufreq_debug_printk); + + +module_param(debug, uint, 0644); +MODULE_PARM_DESC(debug, "CPUfreq debugging: add 1 to debug core," + " 2 to debug drivers, and 4 to debug governors."); + +module_param(debug_ratelimit, uint, 0644); +MODULE_PARM_DESC(debug_ratelimit, "CPUfreq debugging:" + " set to 0 to disable ratelimiting."); + +#else /* !CONFIG_CPU_FREQ_DEBUG */ + +static inline void cpufreq_debug_enable_ratelimit(void) { return; } +static inline void cpufreq_debug_disable_ratelimit(void) { return; } + +#endif /* CONFIG_CPU_FREQ_DEBUG */ + /********************************************************************* * EXTERNALLY AFFECTING FREQUENCY CHANGES * @@ -108,14 +209,14 @@ static void cpufreq_cpu_put(struct cpufreq_policy *data) * * This function alters the system "loops_per_jiffy" for the clock * speed change. Note that loops_per_jiffy cannot be updated on SMP - * systems as each CPU might be scaled differently. So, use the arch + * systems as each CPU might be scaled differently. So, use the arch * per-CPU loops_per_jiffy value wherever possible. */ #ifndef CONFIG_SMP static unsigned long l_p_j_ref; static unsigned int l_p_j_ref_freq; -static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci) +static void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci) { if (ci->flags & CPUFREQ_CONST_LOOPS) return; @@ -123,56 +224,74 @@ static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci) if (!l_p_j_ref_freq) { l_p_j_ref = loops_per_jiffy; l_p_j_ref_freq = ci->old; + dprintk("saving %lu as reference value for loops_per_jiffy;" + "freq is %u kHz\n", l_p_j_ref, l_p_j_ref_freq); } if ((val == CPUFREQ_PRECHANGE && ci->old < ci->new) || (val == CPUFREQ_POSTCHANGE && ci->old > ci->new) || - (val == CPUFREQ_RESUMECHANGE)) - loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq, ci->new); + (val == CPUFREQ_RESUMECHANGE || val == CPUFREQ_SUSPENDCHANGE)) { + loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq, + ci->new); + dprintk("scaling loops_per_jiffy to %lu" + "for frequency %u kHz\n", loops_per_jiffy, ci->new); + } } #else -static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci) { return; } +static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci) +{ + return; +} #endif /** - * cpufreq_notify_transition - call notifier chain and adjust_jiffies on frequency transition + * cpufreq_notify_transition - call notifier chain and adjust_jiffies + * on frequency transition. * - * This function calls the transition notifiers and the "adjust_jiffies" function. It is called - * twice on all CPU frequency changes that have external effects. + * This function calls the transition notifiers and the "adjust_jiffies" + * function. It is called twice on all CPU frequency changes that have + * external effects. */ void cpufreq_notify_transition(struct cpufreq_freqs *freqs, unsigned int state) { + struct cpufreq_policy *policy; + BUG_ON(irqs_disabled()); freqs->flags = cpufreq_driver->flags; + dprintk("notification %u of frequency transition to %u kHz\n", + state, freqs->new); - down_read(&cpufreq_notifier_rwsem); + policy = cpufreq_cpu_data[freqs->cpu]; switch (state) { + case CPUFREQ_PRECHANGE: - /* detect if the driver reported a value as "old frequency" which - * is not equal to what the cpufreq core thinks is "old frequency". + /* detect if the driver reported a value as "old frequency" + * which is not equal to what the cpufreq core thinks is + * "old frequency". */ if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) { - if ((likely(cpufreq_cpu_data[freqs->cpu])) && - (likely(cpufreq_cpu_data[freqs->cpu]->cur)) && - (unlikely(freqs->old != cpufreq_cpu_data[freqs->cpu]->cur))) - { - printk(KERN_WARNING "Warning: CPU frequency is %u, " - "cpufreq assumed %u kHz.\n", freqs->old, cpufreq_cpu_data[freqs->cpu]->cur); - freqs->old = cpufreq_cpu_data[freqs->cpu]->cur; + if ((policy) && (policy->cpu == freqs->cpu) && + (policy->cur) && (policy->cur != freqs->old)) { + dprintk("Warning: CPU frequency is" + " %u, cpufreq assumed %u kHz.\n", + freqs->old, policy->cur); + freqs->old = policy->cur; } } - notifier_call_chain(&cpufreq_transition_notifier_list, CPUFREQ_PRECHANGE, freqs); + srcu_notifier_call_chain(&cpufreq_transition_notifier_list, + CPUFREQ_PRECHANGE, freqs); adjust_jiffies(CPUFREQ_PRECHANGE, freqs); break; + case CPUFREQ_POSTCHANGE: adjust_jiffies(CPUFREQ_POSTCHANGE, freqs); - notifier_call_chain(&cpufreq_transition_notifier_list, CPUFREQ_POSTCHANGE, freqs); - if (likely(cpufreq_cpu_data[freqs->cpu])) - cpufreq_cpu_data[freqs->cpu]->cur = freqs->new; + srcu_notifier_call_chain(&cpufreq_transition_notifier_list, + CPUFREQ_POSTCHANGE, freqs); + if (likely(policy) && likely(policy->cpu == freqs->cpu)) + policy->cur = freqs->new; break; } - up_read(&cpufreq_notifier_rwsem); } EXPORT_SYMBOL_GPL(cpufreq_notify_transition); @@ -182,41 +301,72 @@ EXPORT_SYMBOL_GPL(cpufreq_notify_transition); * SYSFS INTERFACE * *********************************************************************/ +static struct cpufreq_governor *__find_governor(const char *str_governor) +{ + struct cpufreq_governor *t; + + list_for_each_entry(t, &cpufreq_governor_list, governor_list) + if (!strnicmp(str_governor,t->name,CPUFREQ_NAME_LEN)) + return t; + + return NULL; +} + /** * cpufreq_parse_governor - parse a governor string */ -int cpufreq_parse_governor (char *str_governor, unsigned int *policy, +static int cpufreq_parse_governor (char *str_governor, unsigned int *policy, struct cpufreq_governor **governor) { + int err = -EINVAL; + if (!cpufreq_driver) - return -EINVAL; + goto out; + if (cpufreq_driver->setpolicy) { if (!strnicmp(str_governor, "performance", CPUFREQ_NAME_LEN)) { *policy = CPUFREQ_POLICY_PERFORMANCE; - return 0; - } else if (!strnicmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) { + err = 0; + } else if (!strnicmp(str_governor, "powersave", + CPUFREQ_NAME_LEN)) { *policy = CPUFREQ_POLICY_POWERSAVE; - return 0; + err = 0; } - return -EINVAL; - } else { + } else if (cpufreq_driver->target) { struct cpufreq_governor *t; - down(&cpufreq_governor_sem); - if (!cpufreq_driver || !cpufreq_driver->target) - goto out; - list_for_each_entry(t, &cpufreq_governor_list, governor_list) { - if (!strnicmp(str_governor,t->name,CPUFREQ_NAME_LEN)) { - *governor = t; - up(&cpufreq_governor_sem); - return 0; + + mutex_lock(&cpufreq_governor_mutex); + + t = __find_governor(str_governor); + + if (t == NULL) { + char *name = kasprintf(GFP_KERNEL, "cpufreq_%s", + str_governor); + + if (name) { + int ret; + + mutex_unlock(&cpufreq_governor_mutex); + ret = request_module(name); + mutex_lock(&cpufreq_governor_mutex); + + if (ret == 0) + t = __find_governor(str_governor); } + + kfree(name); + } + + if (t != NULL) { + *governor = t; + err = 0; } - out: - up(&cpufreq_governor_sem); + + mutex_unlock(&cpufreq_governor_mutex); } - return -EINVAL; + out: + return err; } -EXPORT_SYMBOL_GPL(cpufreq_parse_governor); /* drivers/base/cpu.c */ @@ -224,17 +374,18 @@ extern struct sysdev_class cpu_sysdev_class; /** - * cpufreq_per_cpu_attr_read() / show_##file_name() - print out cpufreq information + * cpufreq_per_cpu_attr_read() / show_##file_name() - + * print out cpufreq information * * Write out information from cpufreq_driver->policy[cpu]; object must be * "unsigned int". */ -#define show_one(file_name, object) \ -static ssize_t show_##file_name \ -(struct cpufreq_policy * policy, char *buf) \ -{ \ - return sprintf (buf, "%u\n", policy->object); \ +#define show_one(file_name, object) \ +static ssize_t show_##file_name \ +(struct cpufreq_policy * policy, char *buf) \ +{ \ + return sprintf (buf, "%u\n", policy->object); \ } show_one(cpuinfo_min_freq, cpuinfo.min_freq); @@ -243,6 +394,9 @@ show_one(scaling_min_freq, min); show_one(scaling_max_freq, max); show_one(scaling_cur_freq, cur); +static int __cpufreq_set_policy(struct cpufreq_policy *data, + struct cpufreq_policy *policy); + /** * cpufreq_per_cpu_attr_write() / store_##file_name() - sysfs write access */ @@ -261,7 +415,12 @@ static ssize_t store_##file_name \ if (ret != 1) \ return -EINVAL; \ \ - ret = cpufreq_set_policy(&new_policy); \ + lock_cpu_hotplug(); \ + mutex_lock(&policy->lock); \ + ret = __cpufreq_set_policy(policy, &new_policy); \ + policy->user_policy.object = policy->object; \ + mutex_unlock(&policy->lock); \ + unlock_cpu_hotplug(); \ \ return ret ? ret : count; \ } @@ -272,7 +431,8 @@ store_one(scaling_max_freq,max); /** * show_cpuinfo_cur_freq - current CPU frequency as detected by hardware */ -static ssize_t show_cpuinfo_cur_freq (struct cpufreq_policy * policy, char *buf) +static ssize_t show_cpuinfo_cur_freq (struct cpufreq_policy * policy, + char *buf) { unsigned int cur_freq = cpufreq_get(policy->cpu); if (!cur_freq) @@ -284,7 +444,8 @@ static ssize_t show_cpuinfo_cur_freq (struct cpufreq_policy * policy, char *buf) /** * show_scaling_governor - show the current policy for the specified CPU */ -static ssize_t show_scaling_governor (struct cpufreq_policy * policy, char *buf) +static ssize_t show_scaling_governor (struct cpufreq_policy * policy, + char *buf) { if(policy->policy == CPUFREQ_POLICY_POWERSAVE) return sprintf(buf, "powersave\n"); @@ -299,8 +460,8 @@ static ssize_t show_scaling_governor (struct cpufreq_policy * policy, char *buf) /** * store_scaling_governor - store policy for the specified CPU */ -static ssize_t store_scaling_governor (struct cpufreq_policy * policy, - const char *buf, size_t count) +static ssize_t store_scaling_governor (struct cpufreq_policy * policy, + const char *buf, size_t count) { unsigned int ret = -EINVAL; char str_governor[16]; @@ -314,12 +475,27 @@ static ssize_t store_scaling_governor (struct cpufreq_policy * policy, if (ret != 1) return -EINVAL; - if (cpufreq_parse_governor(str_governor, &new_policy.policy, &new_policy.governor)) + if (cpufreq_parse_governor(str_governor, &new_policy.policy, + &new_policy.governor)) return -EINVAL; - ret = cpufreq_set_policy(&new_policy); + lock_cpu_hotplug(); + + /* Do not use cpufreq_set_policy here or the user_policy.max + will be wrongly overridden */ + mutex_lock(&policy->lock); + ret = __cpufreq_set_policy(policy, &new_policy); + + policy->user_policy.policy = policy->policy; + policy->user_policy.governor = policy->governor; + mutex_unlock(&policy->lock); - return ret ? ret : count; + unlock_cpu_hotplug(); + + if (ret) + return ret; + else + return count; } /** @@ -333,7 +509,7 @@ static ssize_t show_scaling_driver (struct cpufreq_policy * policy, char *buf) /** * show_scaling_available_governors - show the available CPUfreq governors */ -static ssize_t show_scaling_available_governors (struct cpufreq_policy * policy, +static ssize_t show_scaling_available_governors (struct cpufreq_policy *policy, char *buf) { ssize_t i = 0; @@ -349,30 +525,41 @@ static ssize_t show_scaling_available_governors (struct cpufreq_policy * policy, goto out; i += scnprintf(&buf[i], CPUFREQ_NAME_LEN, "%s ", t->name); } - out: +out: + i += sprintf(&buf[i], "\n"); + return i; +} +/** + * show_affected_cpus - show the CPUs affected by each transition + */ +static ssize_t show_affected_cpus (struct cpufreq_policy * policy, char *buf) +{ + ssize_t i = 0; + unsigned int cpu; + + for_each_cpu_mask(cpu, policy->cpus) { + if (i) + i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), " "); + i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), "%u", cpu); + if (i >= (PAGE_SIZE - 5)) + break; + } i += sprintf(&buf[i], "\n"); return i; } #define define_one_ro(_name) \ -struct freq_attr _name = { \ - .attr = { .name = __stringify(_name), .mode = 0444 }, \ - .show = show_##_name, \ -} +static struct freq_attr _name = \ +__ATTR(_name, 0444, show_##_name, NULL) #define define_one_ro0400(_name) \ -struct freq_attr _name = { \ - .attr = { .name = __stringify(_name), .mode = 0400 }, \ - .show = show_##_name, \ -} +static struct freq_attr _name = \ +__ATTR(_name, 0400, show_##_name, NULL) #define define_one_rw(_name) \ -struct freq_attr _name = { \ - .attr = { .name = __stringify(_name), .mode = 0644 }, \ - .show = show_##_name, \ - .store = store_##_name, \ -} +static struct freq_attr _name = \ +__ATTR(_name, 0644, show_##_name, store_##_name) define_one_ro0400(cpuinfo_cur_freq); define_one_ro(cpuinfo_min_freq); @@ -380,6 +567,7 @@ define_one_ro(cpuinfo_max_freq); define_one_ro(scaling_available_governors); define_one_ro(scaling_driver); define_one_ro(scaling_cur_freq); +define_one_ro(affected_cpus); define_one_rw(scaling_min_freq); define_one_rw(scaling_max_freq); define_one_rw(scaling_governor); @@ -389,6 +577,7 @@ static struct attribute * default_attrs[] = { &cpuinfo_max_freq.attr, &scaling_min_freq.attr, &scaling_max_freq.attr, + &affected_cpus.attr, &scaling_governor.attr, &scaling_driver.attr, &scaling_available_governors.attr, @@ -406,12 +595,16 @@ static ssize_t show(struct kobject * kobj, struct attribute * attr ,char * buf) policy = cpufreq_cpu_get(policy->cpu); if (!policy) return -EINVAL; - ret = fattr->show ? fattr->show(policy,buf) : 0; + if (fattr->show) + ret = fattr->show(policy, buf); + else + ret = -EIO; + cpufreq_cpu_put(policy); return ret; } -static ssize_t store(struct kobject * kobj, struct attribute * attr, +static ssize_t store(struct kobject * kobj, struct attribute * attr, const char * buf, size_t count) { struct cpufreq_policy * policy = to_policy(kobj); @@ -420,7 +613,11 @@ static ssize_t store(struct kobject * kobj, struct attribute * attr, policy = cpufreq_cpu_get(policy->cpu); if (!policy) return -EINVAL; - ret = fattr->store ? fattr->store(policy,buf,count) : 0; + if (fattr->store) + ret = fattr->store(policy, buf, count); + else + ret = -EIO; + cpufreq_cpu_put(policy); return ret; } @@ -428,6 +625,7 @@ static ssize_t store(struct kobject * kobj, struct attribute * attr, static void cpufreq_sysfs_release(struct kobject * kobj) { struct cpufreq_policy * policy = to_policy(kobj); + dprintk("last reference is dropped\n"); complete(&policy->kobj_unregister); } @@ -446,7 +644,7 @@ static struct kobj_type ktype_cpufreq = { /** * cpufreq_add_dev - add a CPU device * - * Adds the cpufreq interface for a CPU device. + * Adds the cpufreq interface for a CPU device. */ static int cpufreq_add_dev (struct sys_device * sys_dev) { @@ -455,30 +653,90 @@ static int cpufreq_add_dev (struct sys_device * sys_dev) struct cpufreq_policy new_policy; struct cpufreq_policy *policy; struct freq_attr **drv_attr; + struct sys_device *cpu_sys_dev; unsigned long flags; + unsigned int j; +#ifdef CONFIG_SMP + struct cpufreq_policy *managed_policy; +#endif - if (!try_module_get(cpufreq_driver->owner)) - return -EINVAL; + if (cpu_is_offline(cpu)) + return 0; + + cpufreq_debug_disable_ratelimit(); + dprintk("adding CPU %u\n", cpu); + +#ifdef CONFIG_SMP + /* check whether a different CPU already registered this + * CPU because it is in the same boat. */ + policy = cpufreq_cpu_get(cpu); + if (unlikely(policy)) { + cpufreq_cpu_put(policy); + cpufreq_debug_enable_ratelimit(); + return 0; + } +#endif + + if (!try_module_get(cpufreq_driver->owner)) { + ret = -EINVAL; + goto module_out; + } - policy = kmalloc(sizeof(struct cpufreq_policy), GFP_KERNEL); + policy = kzalloc(sizeof(struct cpufreq_policy), GFP_KERNEL); if (!policy) { ret = -ENOMEM; goto nomem_out; } - memset(policy, 0, sizeof(struct cpufreq_policy)); policy->cpu = cpu; - init_MUTEX_LOCKED(&policy->lock); + policy->cpus = cpumask_of_cpu(cpu); + + mutex_init(&policy->lock); + mutex_lock(&policy->lock); init_completion(&policy->kobj_unregister); - INIT_WORK(&policy->update, handle_update, (void *)(long)cpu); + INIT_WORK(&policy->update, handle_update); /* call driver. From then on the cpufreq must be able * to accept all calls to ->verify and ->setpolicy for this CPU */ ret = cpufreq_driver->init(policy); - if (ret) + if (ret) { + dprintk("initialization failed\n"); + mutex_unlock(&policy->lock); goto err_out; + } + +#ifdef CONFIG_SMP + for_each_cpu_mask(j, policy->cpus) { + if (cpu == j) + continue; + /* check for existing affected CPUs. They may not be aware + * of it due to CPU Hotplug. + */ + managed_policy = cpufreq_cpu_get(j); + if (unlikely(managed_policy)) { + spin_lock_irqsave(&cpufreq_driver_lock, flags); + managed_policy->cpus = policy->cpus; + cpufreq_cpu_data[cpu] = managed_policy; + spin_unlock_irqrestore(&cpufreq_driver_lock, flags); + + dprintk("CPU already managed, adding link\n"); + ret = sysfs_create_link(&sys_dev->kobj, + &managed_policy->kobj, + "cpufreq"); + if (ret) { + mutex_unlock(&policy->lock); + goto err_out_driver_exit; + } + + cpufreq_debug_enable_ratelimit(); + mutex_unlock(&policy->lock); + ret = 0; + goto err_out_driver_exit; /* call driver->exit() */ + } + } +#endif memcpy(&new_policy, policy, sizeof(struct cpufreq_policy)); /* prepare interface data */ @@ -487,9 +745,10 @@ static int cpufreq_add_dev (struct sys_device * sys_dev) strlcpy(policy->kobj.name, "cpufreq", KOBJ_NAME_LEN); ret = kobject_register(&policy->kobj); - if (ret) - goto err_out; - + if (ret) { + mutex_unlock(&policy->lock); + goto err_out_driver_exit; + } /* set up files for this cpu device */ drv_attr = cpufreq_driver->attr; while ((drv_attr) && (*drv_attr)) { @@ -502,35 +761,66 @@ static int cpufreq_add_dev (struct sys_device * sys_dev) sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr); spin_lock_irqsave(&cpufreq_driver_lock, flags); - cpufreq_cpu_data[cpu] = policy; + for_each_cpu_mask(j, policy->cpus) + cpufreq_cpu_data[j] = policy; spin_unlock_irqrestore(&cpufreq_driver_lock, flags); + + /* symlink affected CPUs */ + for_each_cpu_mask(j, policy->cpus) { + if (j == cpu) + continue; + if (!cpu_online(j)) + continue; + + dprintk("CPU %u already managed, adding link\n", j); + cpufreq_cpu_get(cpu); + cpu_sys_dev = get_cpu_sysdev(j); + ret = sysfs_create_link(&cpu_sys_dev->kobj, &policy->kobj, + "cpufreq"); + if (ret) { + mutex_unlock(&policy->lock); + goto err_out_unregister; + } + } + policy->governor = NULL; /* to assure that the starting sequence is * run in cpufreq_set_policy */ - up(&policy->lock); - + mutex_unlock(&policy->lock); + /* set default policy */ - ret = cpufreq_set_policy(&new_policy); - if (ret) + if (ret) { + dprintk("setting policy failed\n"); goto err_out_unregister; + } module_put(cpufreq_driver->owner); + dprintk("initialization complete\n"); + cpufreq_debug_enable_ratelimit(); + return 0; err_out_unregister: spin_lock_irqsave(&cpufreq_driver_lock, flags); - cpufreq_cpu_data[cpu] = NULL; + for_each_cpu_mask(j, policy->cpus) + cpufreq_cpu_data[j] = NULL; spin_unlock_irqrestore(&cpufreq_driver_lock, flags); kobject_unregister(&policy->kobj); wait_for_completion(&policy->kobj_unregister); +err_out_driver_exit: + if (cpufreq_driver->exit) + cpufreq_driver->exit(policy); + err_out: kfree(policy); nomem_out: module_put(cpufreq_driver->owner); +module_out: + cpufreq_debug_enable_ratelimit(); return ret; } @@ -545,45 +835,110 @@ static int cpufreq_remove_dev (struct sys_device * sys_dev) unsigned int cpu = sys_dev->id; unsigned long flags; struct cpufreq_policy *data; +#ifdef CONFIG_SMP + struct sys_device *cpu_sys_dev; + unsigned int j; +#endif + + cpufreq_debug_disable_ratelimit(); + dprintk("unregistering CPU %u\n", cpu); spin_lock_irqsave(&cpufreq_driver_lock, flags); data = cpufreq_cpu_data[cpu]; if (!data) { spin_unlock_irqrestore(&cpufreq_driver_lock, flags); + cpufreq_debug_enable_ratelimit(); return -EINVAL; } cpufreq_cpu_data[cpu] = NULL; - spin_unlock_irqrestore(&cpufreq_driver_lock, flags); - if (!kobject_get(&data->kobj)) + +#ifdef CONFIG_SMP + /* if this isn't the CPU which is the parent of the kobj, we + * only need to unlink, put and exit + */ + if (unlikely(cpu != data->cpu)) { + dprintk("removing link\n"); + cpu_clear(cpu, data->cpus); + spin_unlock_irqrestore(&cpufreq_driver_lock, flags); + sysfs_remove_link(&sys_dev->kobj, "cpufreq"); + cpufreq_cpu_put(data); + cpufreq_debug_enable_ratelimit(); + return 0; + } +#endif + + + if (!kobject_get(&data->kobj)) { + spin_unlock_irqrestore(&cpufreq_driver_lock, flags); + cpufreq_debug_enable_ratelimit(); return -EFAULT; + } + +#ifdef CONFIG_SMP + /* if we have other CPUs still registered, we need to unlink them, + * or else wait_for_completion below will lock up. Clean the + * cpufreq_cpu_data[] while holding the lock, and remove the sysfs + * links afterwards. + */ + if (unlikely(cpus_weight(data->cpus) > 1)) { + for_each_cpu_mask(j, data->cpus) { + if (j == cpu) + continue; + cpufreq_cpu_data[j] = NULL; + } + } + spin_unlock_irqrestore(&cpufreq_driver_lock, flags); + + if (unlikely(cpus_weight(data->cpus) > 1)) { + for_each_cpu_mask(j, data->cpus) { + if (j == cpu) + continue; + dprintk("removing link for cpu %u\n", j); + cpu_sys_dev = get_cpu_sysdev(j); + sysfs_remove_link(&cpu_sys_dev->kobj, "cpufreq"); + cpufreq_cpu_put(data); + } + } +#else + spin_unlock_irqrestore(&cpufreq_driver_lock, flags); +#endif + + mutex_lock(&data->lock); if (cpufreq_driver->target) __cpufreq_governor(data, CPUFREQ_GOV_STOP); + mutex_unlock(&data->lock); kobject_unregister(&data->kobj); kobject_put(&data->kobj); /* we need to make sure that the underlying kobj is actually - * not referenced anymore by anybody before we proceed with + * not referenced anymore by anybody before we proceed with * unloading. */ + dprintk("waiting for dropping of refcount\n"); wait_for_completion(&data->kobj_unregister); + dprintk("wait complete\n"); if (cpufreq_driver->exit) cpufreq_driver->exit(data); kfree(data); + cpufreq_debug_enable_ratelimit(); return 0; } -static void handle_update(void *data) +static void handle_update(struct work_struct *work) { - unsigned int cpu = (unsigned int)(long)data; + struct cpufreq_policy *policy = + container_of(work, struct cpufreq_policy, update); + unsigned int cpu = policy->cpu; + dprintk("handle_update for cpu %u called\n", cpu); cpufreq_update_policy(cpu); } @@ -596,11 +951,12 @@ static void handle_update(void *data) * We adjust to current frequency first, and need to clean up later. So either call * to cpufreq_update_policy() or schedule handle_update()). */ -static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq, unsigned int new_freq) +static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq, + unsigned int new_freq) { struct cpufreq_freqs freqs; - printk(KERN_WARNING "Warning: CPU frequency out of sync: cpufreq and timing " + dprintk("Warning: CPU frequency out of sync: cpufreq and timing " "core thinks of %u, is %u kHz.\n", old_freq, new_freq); freqs.cpu = cpu; @@ -611,7 +967,31 @@ static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq, unsigne } -/** +/** + * cpufreq_quick_get - get the CPU frequency (in kHz) from policy->cur + * @cpu: CPU number + * + * This is the last known freq, without actually getting it from the driver. + * Return value will be same as what is shown in scaling_cur_freq in sysfs. + */ +unsigned int cpufreq_quick_get(unsigned int cpu) +{ + struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); + unsigned int ret_freq = 0; + + if (policy) { + mutex_lock(&policy->lock); + ret_freq = policy->cur; + mutex_unlock(&policy->lock); + cpufreq_cpu_put(policy); + } + + return (ret_freq); +} +EXPORT_SYMBOL(cpufreq_quick_get); + + +/** * cpufreq_get - get the current CPU frequency (in kHz) * @cpu: CPU number * @@ -620,7 +1000,7 @@ static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq, unsigne unsigned int cpufreq_get(unsigned int cpu) { struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); - unsigned int ret = 0; + unsigned int ret_freq = 0; if (!policy) return 0; @@ -628,42 +1008,124 @@ unsigned int cpufreq_get(unsigned int cpu) if (!cpufreq_driver->get) goto out; - down(&policy->lock); + mutex_lock(&policy->lock); - ret = cpufreq_driver->get(cpu); + ret_freq = cpufreq_driver->get(cpu); - if (ret && policy->cur && !(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) - { - /* verify no discrepancy between actual and saved value exists */ - if (unlikely(ret != policy->cur)) { - cpufreq_out_of_sync(cpu, policy->cur, ret); + if (ret_freq && policy->cur && + !(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) { + /* verify no discrepancy between actual and + saved value exists */ + if (unlikely(ret_freq != policy->cur)) { + cpufreq_out_of_sync(cpu, policy->cur, ret_freq); schedule_work(&policy->update); } } - up(&policy->lock); + mutex_unlock(&policy->lock); - out: +out: cpufreq_cpu_put(policy); - return (ret); + return (ret_freq); } EXPORT_SYMBOL(cpufreq_get); +/** + * cpufreq_suspend - let the low level driver prepare for suspend + */ + +static int cpufreq_suspend(struct sys_device * sysdev, pm_message_t pmsg) +{ + int cpu = sysdev->id; + int ret = 0; + unsigned int cur_freq = 0; + struct cpufreq_policy *cpu_policy; + + dprintk("suspending cpu %u\n", cpu); + + if (!cpu_online(cpu)) + return 0; + + /* we may be lax here as interrupts are off. Nonetheless + * we need to grab the correct cpu policy, as to check + * whether we really run on this CPU. + */ + + cpu_policy = cpufreq_cpu_get(cpu); + if (!cpu_policy) + return -EINVAL; + + /* only handle each CPU group once */ + if (unlikely(cpu_policy->cpu != cpu)) { + cpufreq_cpu_put(cpu_policy); + return 0; + } + + if (cpufreq_driver->suspend) { + ret = cpufreq_driver->suspend(cpu_policy, pmsg); + if (ret) { + printk(KERN_ERR "cpufreq: suspend failed in ->suspend " + "step on CPU %u\n", cpu_policy->cpu); + cpufreq_cpu_put(cpu_policy); + return ret; + } + } + + + if (cpufreq_driver->flags & CPUFREQ_CONST_LOOPS) + goto out; + + if (cpufreq_driver->get) + cur_freq = cpufreq_driver->get(cpu_policy->cpu); + + if (!cur_freq || !cpu_policy->cur) { + printk(KERN_ERR "cpufreq: suspend failed to assert current " + "frequency is what timing core thinks it is.\n"); + goto out; + } + + if (unlikely(cur_freq != cpu_policy->cur)) { + struct cpufreq_freqs freqs; + + if (!(cpufreq_driver->flags & CPUFREQ_PM_NO_WARN)) + dprintk("Warning: CPU frequency is %u, " + "cpufreq assumed %u kHz.\n", + cur_freq, cpu_policy->cur); + + freqs.cpu = cpu; + freqs.old = cpu_policy->cur; + freqs.new = cur_freq; + + srcu_notifier_call_chain(&cpufreq_transition_notifier_list, + CPUFREQ_SUSPENDCHANGE, &freqs); + adjust_jiffies(CPUFREQ_SUSPENDCHANGE, &freqs); + + cpu_policy->cur = cur_freq; + } + +out: + cpufreq_cpu_put(cpu_policy); + return 0; +} + /** * cpufreq_resume - restore proper CPU frequency handling after resume * * 1.) resume CPUfreq hardware support (cpufreq_driver->resume()) * 2.) if ->target and !CPUFREQ_CONST_LOOPS: verify we're in sync - * 3.) schedule call cpufreq_update_policy() ASAP as interrupts are restored. + * 3.) schedule call cpufreq_update_policy() ASAP as interrupts are + * restored. */ static int cpufreq_resume(struct sys_device * sysdev) { int cpu = sysdev->id; - unsigned int ret = 0; + int ret = 0; struct cpufreq_policy *cpu_policy; + dprintk("resuming cpu %u\n", cpu); + if (!cpu_online(cpu)) return 0; @@ -676,6 +1138,22 @@ static int cpufreq_resume(struct sys_device * sysdev) if (!cpu_policy) return -EINVAL; + /* only handle each CPU group once */ + if (unlikely(cpu_policy->cpu != cpu)) { + cpufreq_cpu_put(cpu_policy); + return 0; + } + + if (cpufreq_driver->resume) { + ret = cpufreq_driver->resume(cpu_policy); + if (ret) { + printk(KERN_ERR "cpufreq: resume failed in ->resume " + "step on CPU %u\n", cpu_policy->cpu); + cpufreq_cpu_put(cpu_policy); + return ret; + } + } + if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) { unsigned int cur_freq = 0; @@ -683,21 +1161,27 @@ static int cpufreq_resume(struct sys_device * sysdev) cur_freq = cpufreq_driver->get(cpu_policy->cpu); if (!cur_freq || !cpu_policy->cur) { - printk(KERN_ERR "cpufreq: resume failed to assert current frequency is what timing core thinks it is.\n"); + printk(KERN_ERR "cpufreq: resume failed to assert " + "current frequency is what timing core " + "thinks it is.\n"); goto out; } if (unlikely(cur_freq != cpu_policy->cur)) { struct cpufreq_freqs freqs; - printk(KERN_WARNING "Warning: CPU frequency is %u, " - "cpufreq assumed %u kHz.\n", cur_freq, cpu_policy->cur); + if (!(cpufreq_driver->flags & CPUFREQ_PM_NO_WARN)) + dprintk("Warning: CPU frequency" + "is %u, cpufreq assumed %u kHz.\n", + cur_freq, cpu_policy->cur); freqs.cpu = cpu; freqs.old = cpu_policy->cur; freqs.new = cur_freq; - notifier_call_chain(&cpufreq_transition_notifier_list, CPUFREQ_RESUMECHANGE, &freqs); + srcu_notifier_call_chain( + &cpufreq_transition_notifier_list, + CPUFREQ_RESUMECHANGE, &freqs); adjust_jiffies(CPUFREQ_RESUMECHANGE, &freqs); cpu_policy->cur = cur_freq; @@ -713,6 +1197,7 @@ out: static struct sysdev_driver cpufreq_sysdev_driver = { .add = cpufreq_add_dev, .remove = cpufreq_remove_dev, + .suspend = cpufreq_suspend, .resume = cpufreq_resume, }; @@ -726,30 +1211,30 @@ static struct sysdev_driver cpufreq_sysdev_driver = { * @nb: notifier function to register * @list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER * - * Add a driver to one of two lists: either a list of drivers that + * Add a driver to one of two lists: either a list of drivers that * are notified about clock rate changes (once before and once after * the transition), or a list of drivers that are notified about * changes in cpufreq policy. * * This function may sleep, and has the same return conditions as - * notifier_chain_register. + * blocking_notifier_chain_register. */ int cpufreq_register_notifier(struct notifier_block *nb, unsigned int list) { int ret; - down_write(&cpufreq_notifier_rwsem); switch (list) { case CPUFREQ_TRANSITION_NOTIFIER: - ret = notifier_chain_register(&cpufreq_transition_notifier_list, nb); + ret = srcu_notifier_chain_register( + &cpufreq_transition_notifier_list, nb); break; case CPUFREQ_POLICY_NOTIFIER: - ret = notifier_chain_register(&cpufreq_policy_notifier_list, nb); + ret = blocking_notifier_chain_register( + &cpufreq_policy_notifier_list, nb); break; default: ret = -EINVAL; } - up_write(&cpufreq_notifier_rwsem); return ret; } @@ -764,24 +1249,24 @@ EXPORT_SYMBOL(cpufreq_register_notifier); * Remove a driver from the CPU frequency notifier list. * * This function may sleep, and has the same return conditions as - * notifier_chain_unregister. + * blocking_notifier_chain_unregister. */ int cpufreq_unregister_notifier(struct notifier_block *nb, unsigned int list) { int ret; - down_write(&cpufreq_notifier_rwsem); switch (list) { case CPUFREQ_TRANSITION_NOTIFIER: - ret = notifier_chain_unregister(&cpufreq_transition_notifier_list, nb); + ret = srcu_notifier_chain_unregister( + &cpufreq_transition_notifier_list, nb); break; case CPUFREQ_POLICY_NOTIFIER: - ret = notifier_chain_unregister(&cpufreq_policy_notifier_list, nb); + ret = blocking_notifier_chain_unregister( + &cpufreq_policy_notifier_list, nb); break; default: ret = -EINVAL; } - up_write(&cpufreq_notifier_rwsem); return ret; } @@ -793,101 +1278,110 @@ EXPORT_SYMBOL(cpufreq_unregister_notifier); *********************************************************************/ +/* Must be called with lock_cpu_hotplug held */ int __cpufreq_driver_target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation) { int retval = -EINVAL; - lock_cpu_hotplug(); - if (cpu_online(policy->cpu)) + + dprintk("target for CPU %u: %u kHz, relation %u\n", policy->cpu, + target_freq, relation); + if (cpu_online(policy->cpu) && cpufreq_driver->target) retval = cpufreq_driver->target(policy, target_freq, relation); - unlock_cpu_hotplug(); + return retval; } EXPORT_SYMBOL_GPL(__cpufreq_driver_target); - int cpufreq_driver_target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation) { - unsigned int ret; + int ret; policy = cpufreq_cpu_get(policy->cpu); if (!policy) return -EINVAL; - down(&policy->lock); + lock_cpu_hotplug(); + mutex_lock(&policy->lock); ret = __cpufreq_driver_target(policy, target_freq, relation); - up(&policy->lock); + mutex_unlock(&policy->lock); + unlock_cpu_hotplug(); cpufreq_cpu_put(policy); - return ret; } EXPORT_SYMBOL_GPL(cpufreq_driver_target); - -static int __cpufreq_governor(struct cpufreq_policy *policy, unsigned int event) +int cpufreq_driver_getavg(struct cpufreq_policy *policy) { - int ret = -EINVAL; + int ret = 0; - if (!try_module_get(policy->governor->owner)) + policy = cpufreq_cpu_get(policy->cpu); + if (!policy) return -EINVAL; - ret = policy->governor->governor(policy, event); + mutex_lock(&policy->lock); - /* we keep one module reference alive for each CPU governed by this CPU */ - if ((event != CPUFREQ_GOV_START) || ret) - module_put(policy->governor->owner); - if ((event == CPUFREQ_GOV_STOP) && !ret) - module_put(policy->governor->owner); + if (cpu_online(policy->cpu) && cpufreq_driver->getavg) + ret = cpufreq_driver->getavg(policy->cpu); + + mutex_unlock(&policy->lock); + cpufreq_cpu_put(policy); return ret; } +EXPORT_SYMBOL_GPL(cpufreq_driver_getavg); +/* + * Locking: Must be called with the lock_cpu_hotplug() lock held + * when "event" is CPUFREQ_GOV_LIMITS + */ -int cpufreq_governor(unsigned int cpu, unsigned int event) +static int __cpufreq_governor(struct cpufreq_policy *policy, + unsigned int event) { - int ret = 0; - struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); + int ret; - if (!policy) + if (!try_module_get(policy->governor->owner)) return -EINVAL; - down(&policy->lock); - ret = __cpufreq_governor(policy, event); - up(&policy->lock); + dprintk("__cpufreq_governor for CPU %u, event %u\n", + policy->cpu, event); + ret = policy->governor->governor(policy, event); - cpufreq_cpu_put(policy); + /* we keep one module reference alive for + each CPU governed by this CPU */ + if ((event != CPUFREQ_GOV_START) || ret) + module_put(policy->governor->owner); + if ((event == CPUFREQ_GOV_STOP) && !ret) + module_put(policy->governor->owner); return ret; } -EXPORT_SYMBOL_GPL(cpufreq_governor); int cpufreq_register_governor(struct cpufreq_governor *governor) { - struct cpufreq_governor *t; + int err; if (!governor) return -EINVAL; - down(&cpufreq_governor_sem); - - list_for_each_entry(t, &cpufreq_governor_list, governor_list) { - if (!strnicmp(governor->name,t->name,CPUFREQ_NAME_LEN)) { - up(&cpufreq_governor_sem); - return -EBUSY; - } - } - list_add(&governor->governor_list, &cpufreq_governor_list); + mutex_lock(&cpufreq_governor_mutex); - up(&cpufreq_governor_sem); + err = -EBUSY; + if (__find_governor(governor->name) == NULL) { + err = 0; + list_add(&governor->governor_list, &cpufreq_governor_list); + } - return 0; + mutex_unlock(&cpufreq_governor_mutex); + return err; } EXPORT_SYMBOL_GPL(cpufreq_register_governor); @@ -897,9 +1391,9 @@ void cpufreq_unregister_governor(struct cpufreq_governor *governor) if (!governor) return; - down(&cpufreq_governor_sem); + mutex_lock(&cpufreq_governor_mutex); list_del(&governor->governor_list); - up(&cpufreq_governor_sem); + mutex_unlock(&cpufreq_governor_mutex); return; } EXPORT_SYMBOL_GPL(cpufreq_unregister_governor); @@ -926,65 +1420,78 @@ int cpufreq_get_policy(struct cpufreq_policy *policy, unsigned int cpu) if (!cpu_policy) return -EINVAL; - down(&cpu_policy->lock); + mutex_lock(&cpu_policy->lock); memcpy(policy, cpu_policy, sizeof(struct cpufreq_policy)); - up(&cpu_policy->lock); + mutex_unlock(&cpu_policy->lock); cpufreq_cpu_put(cpu_policy); - return 0; } EXPORT_SYMBOL(cpufreq_get_policy); -static int __cpufreq_set_policy(struct cpufreq_policy *data, struct cpufreq_policy *policy) +/* + * data : current policy. + * policy : policy to be set. + * Locking: Must be called with the lock_cpu_hotplug() lock held + */ +static int __cpufreq_set_policy(struct cpufreq_policy *data, + struct cpufreq_policy *policy) { int ret = 0; - memcpy(&policy->cpuinfo, - &data->cpuinfo, - sizeof(struct cpufreq_cpuinfo)); + cpufreq_debug_disable_ratelimit(); + dprintk("setting new policy for CPU %u: %u - %u kHz\n", policy->cpu, + policy->min, policy->max); + + memcpy(&policy->cpuinfo, &data->cpuinfo, + sizeof(struct cpufreq_cpuinfo)); + + if (policy->min > data->min && policy->min > policy->max) { + ret = -EINVAL; + goto error_out; + } /* verify the cpu speed can be set within this limit */ ret = cpufreq_driver->verify(policy); if (ret) goto error_out; - down_read(&cpufreq_notifier_rwsem); - /* adjust if necessary - all reasons */ - notifier_call_chain(&cpufreq_policy_notifier_list, CPUFREQ_ADJUST, - policy); + blocking_notifier_call_chain(&cpufreq_policy_notifier_list, + CPUFREQ_ADJUST, policy); /* adjust if necessary - hardware incompatibility*/ - notifier_call_chain(&cpufreq_policy_notifier_list, CPUFREQ_INCOMPATIBLE, - policy); + blocking_notifier_call_chain(&cpufreq_policy_notifier_list, + CPUFREQ_INCOMPATIBLE, policy); /* verify the cpu speed can be set within this limit, which might be different to the first one */ ret = cpufreq_driver->verify(policy); - if (ret) { - up_read(&cpufreq_notifier_rwsem); + if (ret) goto error_out; - } /* notification of the new policy */ - notifier_call_chain(&cpufreq_policy_notifier_list, CPUFREQ_NOTIFY, - policy); + blocking_notifier_call_chain(&cpufreq_policy_notifier_list, + CPUFREQ_NOTIFY, policy); - up_read(&cpufreq_notifier_rwsem); + data->min = policy->min; + data->max = policy->max; - data->min = policy->min; - data->max = policy->max; + dprintk("new min and max freqs are %u - %u kHz\n", + data->min, data->max); if (cpufreq_driver->setpolicy) { data->policy = policy->policy; + dprintk("setting range\n"); ret = cpufreq_driver->setpolicy(policy); } else { if (policy->governor != data->governor) { /* save old, working values */ struct cpufreq_governor *old_gov = data->governor; + dprintk("governor switch\n"); + /* end old governor */ if (data->governor) __cpufreq_governor(data, CPUFREQ_GOV_STOP); @@ -993,19 +1500,24 @@ static int __cpufreq_set_policy(struct cpufreq_policy *data, struct cpufreq_poli data->governor = policy->governor; if (__cpufreq_governor(data, CPUFREQ_GOV_START)) { /* new governor failed, so re-start old one */ + dprintk("starting governor %s failed\n", + data->governor->name); if (old_gov) { data->governor = old_gov; - __cpufreq_governor(data, CPUFREQ_GOV_START); + __cpufreq_governor(data, + CPUFREQ_GOV_START); } ret = -EINVAL; goto error_out; } /* might be a policy change, too, so fall through */ } + dprintk("governor: change or update limits\n"); __cpufreq_governor(data, CPUFREQ_GOV_LIMITS); } - error_out: +error_out: + cpufreq_debug_enable_ratelimit(); return ret; } @@ -1027,8 +1539,10 @@ int cpufreq_set_policy(struct cpufreq_policy *policy) if (!data) return -EINVAL; + lock_cpu_hotplug(); + /* lock this CPU */ - down(&data->lock); + mutex_lock(&data->lock); ret = __cpufreq_set_policy(data, policy); data->user_policy.min = data->min; @@ -1036,7 +1550,9 @@ int cpufreq_set_policy(struct cpufreq_policy *policy) data->user_policy.policy = data->policy; data->user_policy.governor = data->governor; - up(&data->lock); + mutex_unlock(&data->lock); + + unlock_cpu_hotplug(); cpufreq_cpu_put(data); return ret; @@ -1060,25 +1576,78 @@ int cpufreq_update_policy(unsigned int cpu) if (!data) return -ENODEV; - down(&data->lock); + lock_cpu_hotplug(); + mutex_lock(&data->lock); - memcpy(&policy, - data, - sizeof(struct cpufreq_policy)); + dprintk("updating policy for CPU %u\n", cpu); + memcpy(&policy, data, sizeof(struct cpufreq_policy)); policy.min = data->user_policy.min; policy.max = data->user_policy.max; policy.policy = data->user_policy.policy; policy.governor = data->user_policy.governor; - ret = __cpufreq_set_policy(data, &policy); + /* BIOS might change freq behind our back + -> ask driver for current freq and notify governors about a change */ + if (cpufreq_driver->get) { + policy.cur = cpufreq_driver->get(cpu); + if (!data->cur) { + dprintk("Driver did not initialize current freq"); + data->cur = policy.cur; + } else { + if (data->cur != policy.cur) + cpufreq_out_of_sync(cpu, data->cur, + policy.cur); + } + } - up(&data->lock); + ret = __cpufreq_set_policy(data, &policy); + mutex_unlock(&data->lock); + unlock_cpu_hotplug(); cpufreq_cpu_put(data); return ret; } EXPORT_SYMBOL(cpufreq_update_policy); +static int cpufreq_cpu_callback(struct notifier_block *nfb, + unsigned long action, void *hcpu) +{ + unsigned int cpu = (unsigned long)hcpu; + struct cpufreq_policy *policy; + struct sys_device *sys_dev; + + sys_dev = get_cpu_sysdev(cpu); + + if (sys_dev) { + switch (action) { + case CPU_ONLINE: + cpufreq_add_dev(sys_dev); + break; + case CPU_DOWN_PREPARE: + /* + * We attempt to put this cpu in lowest frequency + * possible before going down. This will permit + * hardware-managed P-State to switch other related + * threads to min or higher speeds if possible. + */ + policy = cpufreq_cpu_data[cpu]; + if (policy) { + cpufreq_driver_target(policy, policy->min, + CPUFREQ_RELATION_H); + } + break; + case CPU_DEAD: + cpufreq_remove_dev(sys_dev); + break; + } + } + return NOTIFY_OK; +} + +static struct notifier_block __cpuinitdata cpufreq_cpu_notifier = +{ + .notifier_call = cpufreq_cpu_callback, +}; /********************************************************************* * REGISTER / UNREGISTER CPUFREQ DRIVER * @@ -1089,9 +1658,9 @@ EXPORT_SYMBOL(cpufreq_update_policy); * @driver_data: A struct cpufreq_driver containing the values# * submitted by the CPU Frequency driver. * - * Registers a CPU Frequency driver to this core code. This code + * Registers a CPU Frequency driver to this core code. This code * returns zero on success, -EBUSY when another driver got here first - * (and isn't unregistered in the meantime). + * (and isn't unregistered in the meantime). * */ int cpufreq_register_driver(struct cpufreq_driver *driver_data) @@ -1103,6 +1672,8 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data) ((!driver_data->setpolicy) && (!driver_data->target))) return -EINVAL; + dprintk("trying to register driver %s\n", driver_data->name); + if (driver_data->setpolicy) driver_data->flags |= CPUFREQ_CONST_LOOPS; @@ -1127,7 +1698,10 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data) /* if all ->init() calls failed, unregister */ if (ret) { - sysdev_driver_unregister(&cpu_sysdev_class, &cpufreq_sysdev_driver); + dprintk("no CPU initialized for driver %s\n", + driver_data->name); + sysdev_driver_unregister(&cpu_sysdev_class, + &cpufreq_sysdev_driver); spin_lock_irqsave(&cpufreq_driver_lock, flags); cpufreq_driver = NULL; @@ -1135,6 +1709,12 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data) } } + if (!ret) { + register_hotcpu_notifier(&cpufreq_cpu_notifier); + dprintk("driver %s up and running\n", driver_data->name); + cpufreq_debug_enable_ratelimit(); + } + return (ret); } EXPORT_SYMBOL_GPL(cpufreq_register_driver); @@ -1143,7 +1723,7 @@ EXPORT_SYMBOL_GPL(cpufreq_register_driver); /** * cpufreq_unregister_driver - unregister the current CPUFreq driver * - * Unregister the current CPUFreq driver. Only call this if you have + * Unregister the current CPUFreq driver. Only call this if you have * the right to do so, i.e. if you have succeeded in initialising before! * Returns zero if successful, and -EINVAL if the cpufreq_driver is * currently not initialised. @@ -1152,10 +1732,17 @@ int cpufreq_unregister_driver(struct cpufreq_driver *driver) { unsigned long flags; - if (!cpufreq_driver || (driver != cpufreq_driver)) + cpufreq_debug_disable_ratelimit(); + + if (!cpufreq_driver || (driver != cpufreq_driver)) { + cpufreq_debug_enable_ratelimit(); return -EINVAL; + } + + dprintk("unregistering driver %s\n", driver->name); sysdev_driver_unregister(&cpu_sysdev_class, &cpufreq_sysdev_driver); + unregister_hotcpu_notifier(&cpufreq_cpu_notifier); spin_lock_irqsave(&cpufreq_driver_lock, flags); cpufreq_driver = NULL;