X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fcpufreq%2Fcpufreq.c;h=1d41496ed2f83b648e1311ee8fecccbd56ad741d;hb=357585892e56f7c7bec4a9c8dfaf90257c8756c6;hp=7a7859dd0d9859869f0e851aef8ebf8017a16a89;hpb=e00d9967e3addea86dded46deefc5daec5d52e5a;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 7a7859d..1d41496 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 @@ -23,47 +27,110 @@ #include #include #include +#include -#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_CORE, "cpufreq-core", msg) +#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 struct cpufreq_driver *cpufreq_driver; +static struct cpufreq_policy *cpufreq_cpu_data[NR_CPUS]; +#ifdef CONFIG_HOTPLUG_CPU +/* This one keeps track of the previously set governor of a removed CPU */ +static struct cpufreq_governor *cpufreq_cpu_governor[NR_CPUS]; +#endif static DEFINE_SPINLOCK(cpufreq_driver_lock); - -/* we keep a copy of all ->add'ed CPU's struct sys_device here; - * as it is only accessed in ->add and ->remove, no lock or reference - * count is necessary. +/* + * cpu_policy_rwsem is a per CPU reader-writer semaphore designed to cure + * all cpufreq/hotplug/workqueue/etc related lock issues. + * + * The rules for this semaphore: + * - Any routine that wants to read from the policy structure will + * do a down_read on this semaphore. + * - Any routine that will write to the policy structure and/or may take away + * the policy altogether (eg. CPU hotplug), will hold this lock in write + * mode before doing so. + * + * Additional rules: + * - All holders of the lock should check to make sure that the CPU they + * are concerned with are online after they get the lock. + * - Governor routines that can be called in cpufreq hotplug path should not + * take this sem as top level hotplug notifier handler takes this. */ -static struct sys_device *cpu_sys_devices[NR_CPUS]; +static DEFINE_PER_CPU(int, policy_cpu); +static DEFINE_PER_CPU(struct rw_semaphore, cpu_policy_rwsem); + +#define lock_policy_rwsem(mode, cpu) \ +int lock_policy_rwsem_##mode \ +(int cpu) \ +{ \ + int policy_cpu = per_cpu(policy_cpu, cpu); \ + BUG_ON(policy_cpu == -1); \ + down_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu)); \ + if (unlikely(!cpu_online(cpu))) { \ + up_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu)); \ + return -1; \ + } \ + \ + return 0; \ +} + +lock_policy_rwsem(read, cpu); +EXPORT_SYMBOL_GPL(lock_policy_rwsem_read); + +lock_policy_rwsem(write, cpu); +EXPORT_SYMBOL_GPL(lock_policy_rwsem_write); + +void unlock_policy_rwsem_read(int cpu) +{ + int policy_cpu = per_cpu(policy_cpu, cpu); + BUG_ON(policy_cpu == -1); + up_read(&per_cpu(cpu_policy_rwsem, policy_cpu)); +} +EXPORT_SYMBOL_GPL(unlock_policy_rwsem_read); + +void unlock_policy_rwsem_write(int cpu) +{ + int policy_cpu = per_cpu(policy_cpu, cpu); + BUG_ON(policy_cpu == -1); + up_write(&per_cpu(cpu_policy_rwsem, policy_cpu)); +} +EXPORT_SYMBOL_GPL(unlock_policy_rwsem_write); /* 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 unsigned int __cpufreq_get(unsigned int cpu); +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 bool init_cpufreq_transition_notifier_list_called; +static int __init init_cpufreq_transition_notifier_list(void) +{ + srcu_init_notifier_head(&cpufreq_transition_notifier_list); + init_cpufreq_transition_notifier_list_called = true; + 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); -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; @@ -90,20 +157,19 @@ 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); + void cpufreq_cpu_put(struct cpufreq_policy *data) { kobject_put(&data->kobj); @@ -132,7 +198,7 @@ static unsigned int debug_ratelimit = 1; static unsigned int disable_ratelimit = 1; static DEFINE_SPINLOCK(disable_ratelimit_lock); -static inline void cpufreq_debug_enable_ratelimit(void) +static void cpufreq_debug_enable_ratelimit(void) { unsigned long flags; @@ -142,7 +208,7 @@ static inline void cpufreq_debug_enable_ratelimit(void) spin_unlock_irqrestore(&disable_ratelimit_lock, flags); } -static inline void cpufreq_debug_disable_ratelimit(void) +static void cpufreq_debug_disable_ratelimit(void) { unsigned long flags; @@ -151,17 +217,19 @@ static inline void cpufreq_debug_disable_ratelimit(void) spin_unlock_irqrestore(&disable_ratelimit_lock, flags); } -void cpufreq_debug_printk(unsigned int type, const char *prefix, const char *fmt, ...) +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()) { + if (!disable_ratelimit && debug_ratelimit + && !printk_ratelimit()) { spin_unlock_irqrestore(&disable_ratelimit_lock, flags); return; } @@ -182,10 +250,12 @@ 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_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."); +MODULE_PARM_DESC(debug_ratelimit, "CPUfreq debugging:" + " set to 0 to disable ratelimiting."); #else /* !CONFIG_CPU_FREQ_DEBUG */ @@ -204,14 +274,14 @@ static inline void cpufreq_debug_disable_ratelimit(void) { return; } * * 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; @@ -219,62 +289,74 @@ 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); + 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 || 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); + 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); + 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]->cpu == freqs->cpu)) && - (likely(cpufreq_cpu_data[freqs->cpu]->cur)) && - (unlikely(freqs->old != cpufreq_cpu_data[freqs->cpu]->cur))) - { - dprintk(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])) && - (likely(cpufreq_cpu_data[freqs->cpu]->cpu == 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); @@ -284,41 +366,72 @@ 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 */ -static 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("%s", 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 */ @@ -326,17 +439,18 @@ 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); @@ -345,12 +459,15 @@ 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 */ #define store_one(file_name, object) \ static ssize_t store_##file_name \ -(struct cpufreq_policy * policy, const char *buf, size_t count) \ +(struct cpufreq_policy *policy, const char *buf, size_t count) \ { \ unsigned int ret = -EINVAL; \ struct cpufreq_policy new_policy; \ @@ -363,7 +480,8 @@ static ssize_t store_##file_name \ if (ret != 1) \ return -EINVAL; \ \ - ret = cpufreq_set_policy(&new_policy); \ + ret = __cpufreq_set_policy(policy, &new_policy); \ + policy->user_policy.object = policy->object; \ \ return ret ? ret : count; \ } @@ -374,9 +492,10 @@ 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); + unsigned int cur_freq = __cpufreq_get(policy->cpu); if (!cur_freq) return sprintf(buf, ""); return sprintf(buf, "%u\n", cur_freq); @@ -386,7 +505,7 @@ 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"); @@ -401,8 +520,8 @@ 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]; @@ -416,18 +535,27 @@ 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); + /* Do not use cpufreq_set_policy here or the user_policy.max + will be wrongly overridden */ + ret = __cpufreq_set_policy(policy, &new_policy); - return ret ? ret : count; + policy->user_policy.policy = policy->policy; + policy->user_policy.governor = policy->governor; + + if (ret) + return ret; + else + return count; } /** * show_scaling_driver - show the cpufreq driver currently loaded */ -static ssize_t show_scaling_driver (struct cpufreq_policy * policy, char *buf) +static ssize_t show_scaling_driver(struct cpufreq_policy *policy, char *buf) { return scnprintf(buf, CPUFREQ_NAME_LEN, "%s\n", cpufreq_driver->name); } @@ -435,8 +563,8 @@ 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, - char *buf) +static ssize_t show_scaling_available_governors(struct cpufreq_policy *policy, + char *buf) { ssize_t i = 0; struct cpufreq_governor *t; @@ -451,19 +579,17 @@ 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) + +static ssize_t show_cpus(cpumask_t mask, char *buf) { ssize_t i = 0; unsigned int cpu; - for_each_cpu_mask(cpu, policy->cpus) { + for_each_cpu_mask(cpu, mask) { if (i) i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), " "); i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), "%u", cpu); @@ -474,6 +600,50 @@ static ssize_t show_affected_cpus (struct cpufreq_policy * policy, char *buf) return i; } +/** + * show_related_cpus - show the CPUs affected by each transition even if + * hw coordination is in use + */ +static ssize_t show_related_cpus(struct cpufreq_policy *policy, char *buf) +{ + if (cpus_empty(policy->related_cpus)) + return show_cpus(policy->cpus, buf); + return show_cpus(policy->related_cpus, buf); +} + +/** + * show_affected_cpus - show the CPUs affected by each transition + */ +static ssize_t show_affected_cpus(struct cpufreq_policy *policy, char *buf) +{ + return show_cpus(policy->cpus, buf); +} + +static ssize_t store_scaling_setspeed(struct cpufreq_policy *policy, + const char *buf, size_t count) +{ + unsigned int freq = 0; + unsigned int ret; + + if (!policy->governor || !policy->governor->store_setspeed) + return -EINVAL; + + ret = sscanf(buf, "%u", &freq); + if (ret != 1) + return -EINVAL; + + policy->governor->store_setspeed(policy, freq); + + return count; +} + +static ssize_t show_scaling_setspeed(struct cpufreq_policy *policy, char *buf) +{ + if (!policy->governor || !policy->governor->show_setspeed) + return sprintf(buf, "\n"); + + return policy->governor->show_setspeed(policy, buf); +} #define define_one_ro(_name) \ static struct freq_attr _name = \ @@ -493,56 +663,82 @@ 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(related_cpus); define_one_ro(affected_cpus); define_one_rw(scaling_min_freq); define_one_rw(scaling_max_freq); define_one_rw(scaling_governor); +define_one_rw(scaling_setspeed); -static struct attribute * default_attrs[] = { +static struct attribute *default_attrs[] = { &cpuinfo_min_freq.attr, &cpuinfo_max_freq.attr, &scaling_min_freq.attr, &scaling_max_freq.attr, &affected_cpus.attr, + &related_cpus.attr, &scaling_governor.attr, &scaling_driver.attr, &scaling_available_governors.attr, + &scaling_setspeed.attr, NULL }; #define to_policy(k) container_of(k,struct cpufreq_policy,kobj) #define to_attr(a) container_of(a,struct freq_attr,attr) -static ssize_t show(struct kobject * kobj, struct attribute * attr ,char * buf) +static ssize_t show(struct kobject *kobj, struct attribute *attr ,char *buf) { - struct cpufreq_policy * policy = to_policy(kobj); - struct freq_attr * fattr = to_attr(attr); - ssize_t ret; + struct cpufreq_policy *policy = to_policy(kobj); + struct freq_attr *fattr = to_attr(attr); + ssize_t ret = -EINVAL; policy = cpufreq_cpu_get(policy->cpu); if (!policy) - return -EINVAL; - ret = fattr->show ? fattr->show(policy,buf) : -EIO; + goto no_policy; + + if (lock_policy_rwsem_read(policy->cpu) < 0) + goto fail; + + if (fattr->show) + ret = fattr->show(policy, buf); + else + ret = -EIO; + + unlock_policy_rwsem_read(policy->cpu); +fail: cpufreq_cpu_put(policy); +no_policy: return ret; } -static ssize_t store(struct kobject * kobj, struct attribute * attr, - const char * buf, size_t count) +static ssize_t store(struct kobject *kobj, struct attribute *attr, + const char *buf, size_t count) { - struct cpufreq_policy * policy = to_policy(kobj); - struct freq_attr * fattr = to_attr(attr); - ssize_t ret; + struct cpufreq_policy *policy = to_policy(kobj); + struct freq_attr *fattr = to_attr(attr); + ssize_t ret = -EINVAL; policy = cpufreq_cpu_get(policy->cpu); if (!policy) - return -EINVAL; - ret = fattr->store ? fattr->store(policy,buf,count) : -EIO; + goto no_policy; + + if (lock_policy_rwsem_write(policy->cpu) < 0) + goto fail; + + if (fattr->store) + ret = fattr->store(policy, buf, count); + else + ret = -EIO; + + unlock_policy_rwsem_write(policy->cpu); +fail: cpufreq_cpu_put(policy); +no_policy: return ret; } -static void cpufreq_sysfs_release(struct kobject * kobj) +static void cpufreq_sysfs_release(struct kobject *kobj) { - struct cpufreq_policy * policy = to_policy(kobj); + struct cpufreq_policy *policy = to_policy(kobj); dprintk("last reference is dropped\n"); complete(&policy->kobj_unregister); } @@ -562,17 +758,24 @@ 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) +static int cpufreq_add_dev(struct sys_device *sys_dev) { unsigned int cpu = sys_dev->id; int ret = 0; 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 (cpu_is_offline(cpu)) + return 0; cpufreq_debug_disable_ratelimit(); dprintk("adding CPU %u\n", cpu); @@ -582,9 +785,7 @@ static int cpufreq_add_dev (struct sys_device * sys_dev) * CPU because it is in the same boat. */ policy = cpufreq_cpu_get(cpu); if (unlikely(policy)) { - cpu_sys_devices[cpu] = sys_dev; - dprintk("CPU already managed, adding link\n"); - sysfs_create_link(&sys_dev->kobj, &policy->kobj, "cpufreq"); + cpufreq_cpu_put(policy); cpufreq_debug_enable_ratelimit(); return 0; } @@ -595,20 +796,24 @@ static int cpufreq_add_dev (struct sys_device * sys_dev) 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; policy->cpus = cpumask_of_cpu(cpu); - init_MUTEX_LOCKED(&policy->lock); + /* Initially set CPU itself as the policy_cpu */ + per_cpu(policy_cpu, cpu) = cpu; + lock_policy_rwsem_write(cpu); + init_completion(&policy->kobj_unregister); - INIT_WORK(&policy->update, handle_update, (void *)(long)cpu); + INIT_WORK(&policy->update, handle_update); + /* Set governor before ->init, so that driver could check it */ + policy->governor = CPUFREQ_DEFAULT_GOVERNOR; /* call driver. From then on the cpufreq must be able * to accept all calls to ->verify and ->setpolicy for this CPU */ @@ -617,50 +822,124 @@ static int cpufreq_add_dev (struct sys_device * sys_dev) dprintk("initialization failed\n"); goto err_out; } + policy->user_policy.min = policy->cpuinfo.min_freq; + policy->user_policy.max = policy->cpuinfo.max_freq; + +#ifdef CONFIG_SMP + +#ifdef CONFIG_HOTPLUG_CPU + if (cpufreq_cpu_governor[cpu]){ + policy->governor = cpufreq_cpu_governor[cpu]; + dprintk("Restoring governor %s for cpu %d\n", + policy->governor->name, cpu); + } +#endif + + 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); // FIXME: Where is this released? What about error paths? + if (unlikely(managed_policy)) { + + /* Set proper policy_cpu */ + unlock_policy_rwsem_write(cpu); + per_cpu(policy_cpu, cpu) = managed_policy->cpu; + + if (lock_policy_rwsem_write(cpu) < 0) + goto err_out_driver_exit; + + 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) + goto err_out_driver_exit; + + cpufreq_debug_enable_ratelimit(); + ret = 0; + goto err_out_driver_exit; /* call driver->exit() */ + } + } +#endif memcpy(&new_policy, policy, sizeof(struct cpufreq_policy)); /* prepare interface data */ - policy->kobj.parent = &sys_dev->kobj; - policy->kobj.ktype = &ktype_cpufreq; - strlcpy(policy->kobj.name, "cpufreq", KOBJ_NAME_LEN); - - ret = kobject_register(&policy->kobj); + ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq, &sys_dev->kobj, + "cpufreq"); if (ret) - goto err_out; + goto err_out_driver_exit; /* set up files for this cpu device */ drv_attr = cpufreq_driver->attr; while ((drv_attr) && (*drv_attr)) { - sysfs_create_file(&policy->kobj, &((*drv_attr)->attr)); + ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr)); + if (ret) + goto err_out_driver_exit; drv_attr++; } - if (cpufreq_driver->get) - sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr); - if (cpufreq_driver->target) - sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr); + if (cpufreq_driver->get) { + ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr); + if (ret) + goto err_out_driver_exit; + } + if (cpufreq_driver->target) { + ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr); + if (ret) + goto err_out_driver_exit; + } spin_lock_irqsave(&cpufreq_driver_lock, flags); - for_each_cpu_mask(j, policy->cpus) + for_each_cpu_mask(j, policy->cpus) { cpufreq_cpu_data[j] = policy; + per_cpu(policy_cpu, j) = policy->cpu; + } 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) + goto err_out_unregister; + } + policy->governor = NULL; /* to assure that the starting sequence is * run in cpufreq_set_policy */ - up(&policy->lock); - + /* set default policy */ - - ret = cpufreq_set_policy(&new_policy); + ret = __cpufreq_set_policy(policy, &new_policy); + policy->user_policy.policy = policy->policy; + policy->user_policy.governor = policy->governor; + if (ret) { dprintk("setting policy failed\n"); goto err_out_unregister; } + unlock_policy_rwsem_write(cpu); + + kobject_uevent(&policy->kobj, KOBJ_ADD); module_put(cpufreq_driver->owner); - cpu_sys_devices[cpu] = sys_dev; dprintk("initialization complete\n"); cpufreq_debug_enable_ratelimit(); - + return 0; @@ -670,31 +949,39 @@ err_out_unregister: cpufreq_cpu_data[j] = NULL; spin_unlock_irqrestore(&cpufreq_driver_lock, flags); - kobject_unregister(&policy->kobj); + kobject_put(&policy->kobj); wait_for_completion(&policy->kobj_unregister); +err_out_driver_exit: + if (cpufreq_driver->exit) + cpufreq_driver->exit(policy); + err_out: + unlock_policy_rwsem_write(cpu); kfree(policy); nomem_out: module_put(cpufreq_driver->owner); - module_out: +module_out: cpufreq_debug_enable_ratelimit(); return ret; } /** - * cpufreq_remove_dev - remove a CPU device + * __cpufreq_remove_dev - remove a CPU device * * Removes the cpufreq interface for a CPU device. + * Caller should already have policy_rwsem in write mode for this CPU. + * This routine frees the rwsem before returning. */ -static int cpufreq_remove_dev (struct sys_device * sys_dev) +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 @@ -706,8 +993,8 @@ static int cpufreq_remove_dev (struct sys_device * sys_dev) if (!data) { spin_unlock_irqrestore(&cpufreq_driver_lock, flags); - cpu_sys_devices[cpu] = NULL; cpufreq_debug_enable_ratelimit(); + unlock_policy_rwsem_write(cpu); return -EINVAL; } cpufreq_cpu_data[cpu] = NULL; @@ -715,28 +1002,26 @@ static int cpufreq_remove_dev (struct sys_device * sys_dev) #ifdef CONFIG_SMP /* if this isn't the CPU which is the parent of the kobj, we - * only need to unlink, put and exit + * 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"); - cpu_sys_devices[cpu] = NULL; cpufreq_cpu_put(data); cpufreq_debug_enable_ratelimit(); + unlock_policy_rwsem_write(cpu); return 0; } #endif - cpu_sys_devices[cpu] = NULL; +#ifdef CONFIG_SMP - if (!kobject_get(&data->kobj)) { - spin_unlock_irqrestore(&cpufreq_driver_lock, flags); - cpufreq_debug_enable_ratelimit(); - return -EFAULT; - } +#ifdef CONFIG_HOTPLUG_CPU + cpufreq_cpu_governor[cpu] = data->governor; +#endif -#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 @@ -757,7 +1042,11 @@ static int cpufreq_remove_dev (struct sys_device * sys_dev) if (j == cpu) continue; dprintk("removing link for cpu %u\n", j); - sysfs_remove_link(&cpu_sys_devices[j]->kobj, "cpufreq"); +#ifdef CONFIG_HOTPLUG_CPU + cpufreq_cpu_governor[j] = data->governor; +#endif + cpu_sys_dev = get_cpu_sysdev(j); + sysfs_remove_link(&cpu_sys_dev->kobj, "cpufreq"); cpufreq_cpu_put(data); } } @@ -765,18 +1054,15 @@ static int cpufreq_remove_dev (struct sys_device * sys_dev) spin_unlock_irqrestore(&cpufreq_driver_lock, flags); #endif - down(&data->lock); if (cpufreq_driver->target) __cpufreq_governor(data, CPUFREQ_GOV_STOP); - cpufreq_driver->target = NULL; - up(&data->lock); - kobject_unregister(&data->kobj); + unlock_policy_rwsem_write(cpu); 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"); @@ -789,14 +1075,31 @@ static int cpufreq_remove_dev (struct sys_device * sys_dev) kfree(data); cpufreq_debug_enable_ratelimit(); - return 0; } -static void handle_update(void *data) +static int cpufreq_remove_dev(struct sys_device *sys_dev) { - unsigned int cpu = (unsigned int)(long)data; + unsigned int cpu = sys_dev->id; + int retval; + + if (cpu_is_offline(cpu)) + return 0; + + if (unlikely(lock_policy_rwsem_write(cpu))) + BUG(); + + retval = __cpufreq_remove_dev(sys_dev); + return retval; +} + + +static void handle_update(struct work_struct *work) +{ + 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); } @@ -810,11 +1113,12 @@ 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; - dprintk(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; @@ -825,42 +1129,76 @@ static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq, unsigne } -/** - * cpufreq_get - get the current CPU frequency (in kHz) +/** + * cpufreq_quick_get - get the CPU frequency (in kHz) from policy->cur * @cpu: CPU number * - * Get the CPU current (static) CPU frequency + * 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_get(unsigned int cpu) +unsigned int cpufreq_quick_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; + if (policy) { + ret_freq = policy->cur; + cpufreq_cpu_put(policy); + } + + return ret_freq; +} +EXPORT_SYMBOL(cpufreq_quick_get); - if (!cpufreq_driver->get) - goto out; - down(&policy->lock); +static unsigned int __cpufreq_get(unsigned int cpu) +{ + struct cpufreq_policy *policy = cpufreq_cpu_data[cpu]; + unsigned int ret_freq = 0; + + if (!cpufreq_driver->get) + return ret_freq; - 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); + return ret_freq; +} - out: - cpufreq_cpu_put(policy); +/** + * cpufreq_get - get the current CPU frequency (in kHz) + * @cpu: CPU number + * + * Get the CPU current (static) CPU frequency + */ +unsigned int cpufreq_get(unsigned int cpu) +{ + unsigned int ret_freq = 0; + struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); + + if (!policy) + goto out; + + if (unlikely(lock_policy_rwsem_read(cpu))) + goto out_policy; + + ret_freq = __cpufreq_get(cpu); - return (ret); + unlock_policy_rwsem_read(cpu); + +out_policy: + cpufreq_cpu_put(policy); +out: + return ret_freq; } EXPORT_SYMBOL(cpufreq_get); @@ -869,14 +1207,14 @@ 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) +static int cpufreq_suspend(struct sys_device *sysdev, pm_message_t pmsg) { int cpu = sysdev->id; - unsigned int ret = 0; + int ret = 0; unsigned int cur_freq = 0; struct cpufreq_policy *cpu_policy; - dprintk("resuming cpu %u\n", cpu); + dprintk("suspending cpu %u\n", cpu); if (!cpu_online(cpu)) return 0; @@ -891,22 +1229,18 @@ static int cpufreq_suspend(struct sys_device * sysdev, pm_message_t pmsg) return -EINVAL; /* only handle each CPU group once */ - if (unlikely(cpu_policy->cpu != cpu)) { - cpufreq_cpu_put(cpu_policy); - return 0; - } + if (unlikely(cpu_policy->cpu != cpu)) + goto out; 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; + goto out; } } - if (cpufreq_driver->flags & CPUFREQ_CONST_LOOPS) goto out; @@ -923,7 +1257,7 @@ static int cpufreq_suspend(struct sys_device * sysdev, pm_message_t pmsg) struct cpufreq_freqs freqs; if (!(cpufreq_driver->flags & CPUFREQ_PM_NO_WARN)) - dprintk(KERN_DEBUG "Warning: CPU frequency is %u, " + dprintk("Warning: CPU frequency is %u, " "cpufreq assumed %u kHz.\n", cur_freq, cpu_policy->cur); @@ -931,16 +1265,16 @@ static int cpufreq_suspend(struct sys_device * sysdev, pm_message_t pmsg) freqs.old = cpu_policy->cur; freqs.new = cur_freq; - notifier_call_chain(&cpufreq_transition_notifier_list, + srcu_notifier_call_chain(&cpufreq_transition_notifier_list, CPUFREQ_SUSPENDCHANGE, &freqs); adjust_jiffies(CPUFREQ_SUSPENDCHANGE, &freqs); cpu_policy->cur = cur_freq; } - out: +out: cpufreq_cpu_put(cpu_policy); - return 0; + return ret; } /** @@ -951,10 +1285,10 @@ static int cpufreq_suspend(struct sys_device * sysdev, pm_message_t pmsg) * 3.) schedule call cpufreq_update_policy() ASAP as interrupts are * restored. */ -static int cpufreq_resume(struct sys_device * sysdev) +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); @@ -972,18 +1306,15 @@ static int cpufreq_resume(struct sys_device * sysdev) return -EINVAL; /* only handle each CPU group once */ - if (unlikely(cpu_policy->cpu != cpu)) { - cpufreq_cpu_put(cpu_policy); - return 0; - } + if (unlikely(cpu_policy->cpu != cpu)) + goto fail; 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; + goto fail; } } @@ -1004,7 +1335,7 @@ static int cpufreq_resume(struct sys_device * sysdev) struct cpufreq_freqs freqs; if (!(cpufreq_driver->flags & CPUFREQ_PM_NO_WARN)) - dprintk(KERN_WARNING "Warning: CPU frequency" + dprintk("Warning: CPU frequency " "is %u, cpufreq assumed %u kHz.\n", cur_freq, cpu_policy->cur); @@ -1012,7 +1343,8 @@ static int cpufreq_resume(struct sys_device * sysdev) freqs.old = cpu_policy->cur; freqs.new = cur_freq; - notifier_call_chain(&cpufreq_transition_notifier_list, + srcu_notifier_call_chain( + &cpufreq_transition_notifier_list, CPUFREQ_RESUMECHANGE, &freqs); adjust_jiffies(CPUFREQ_RESUMECHANGE, &freqs); @@ -1022,6 +1354,7 @@ static int cpufreq_resume(struct sys_device * sysdev) out: schedule_work(&cpu_policy->update); +fail: cpufreq_cpu_put(cpu_policy); return ret; } @@ -1043,30 +1376,32 @@ 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); + WARN_ON(!init_cpufreq_transition_notifier_list_called); + 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; } @@ -1081,24 +1416,24 @@ 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; } @@ -1115,99 +1450,123 @@ int __cpufreq_driver_target(struct cpufreq_policy *policy, unsigned int relation) { int retval = -EINVAL; - lock_cpu_hotplug(); + 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); + if (unlikely(lock_policy_rwsem_write(policy->cpu))) + return -EINVAL; ret = __cpufreq_driver_target(policy, target_freq, relation); - up(&policy->lock); + unlock_policy_rwsem_write(policy->cpu); 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; - dprintk("__cpufreq_governor for CPU %u, event %u\n", policy->cpu, event); - ret = policy->governor->governor(policy, event); - - /* 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); + cpufreq_cpu_put(policy); return ret; } +EXPORT_SYMBOL_GPL(__cpufreq_driver_getavg); +/* + * 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) + /* Only must be defined when default governor is known to have latency + restrictions, like e.g. conservative or ondemand. + That this is the case is already ensured in Kconfig + */ +#ifdef CONFIG_CPU_FREQ_GOV_PERFORMANCE + struct cpufreq_governor *gov = &cpufreq_gov_performance; +#else + struct cpufreq_governor *gov = NULL; +#endif + + if (policy->governor->max_transition_latency && + policy->cpuinfo.transition_latency > + policy->governor->max_transition_latency) { + if (!gov) + return -EINVAL; + else { + printk(KERN_WARNING "%s governor failed, too long" + " transition latency of HW, fallback" + " to %s governor\n", + policy->governor->name, + gov->name); + policy->governor = gov; + } + } + + 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); @@ -1217,9 +1576,9 @@ 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); @@ -1246,18 +1605,20 @@ int cpufreq_get_policy(struct cpufreq_policy *policy, unsigned int cpu) if (!cpu_policy) return -EINVAL; - down(&cpu_policy->lock); memcpy(policy, cpu_policy, sizeof(struct cpufreq_policy)); - up(&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. + */ +static int __cpufreq_set_policy(struct cpufreq_policy *data, + struct cpufreq_policy *policy) { int ret = 0; @@ -1265,43 +1626,42 @@ static int __cpufreq_set_policy(struct cpufreq_policy *data, struct cpufreq_poli 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)); + memcpy(&policy->cpuinfo, &data->cpuinfo, + sizeof(struct cpufreq_cpuinfo)); + + if (policy->min > data->max || policy->max < data->min) { + 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); + dprintk("new min and max freqs are %u - %u kHz\n", + data->min, data->max); if (cpufreq_driver->setpolicy) { data->policy = policy->policy; @@ -1322,10 +1682,12 @@ 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); + 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; @@ -1336,47 +1698,12 @@ static int __cpufreq_set_policy(struct cpufreq_policy *data, struct cpufreq_poli __cpufreq_governor(data, CPUFREQ_GOV_LIMITS); } - error_out: +error_out: cpufreq_debug_enable_ratelimit(); return ret; } /** - * cpufreq_set_policy - set a new CPUFreq policy - * @policy: policy to be set. - * - * Sets a new CPU frequency and voltage scaling policy. - */ -int cpufreq_set_policy(struct cpufreq_policy *policy) -{ - int ret = 0; - struct cpufreq_policy *data; - - if (!policy) - return -EINVAL; - - data = cpufreq_cpu_get(policy->cpu); - if (!data) - return -EINVAL; - - /* lock this CPU */ - down(&data->lock); - - ret = __cpufreq_set_policy(data, policy); - data->user_policy.min = data->min; - data->user_policy.max = data->max; - data->user_policy.policy = data->policy; - data->user_policy.governor = data->governor; - - up(&data->lock); - cpufreq_cpu_put(data); - - return ret; -} -EXPORT_SYMBOL(cpufreq_set_policy); - - -/** * cpufreq_update_policy - re-evaluate an existing cpufreq policy * @cpu: CPU which shall be re-evaluated * @@ -1392,26 +1719,72 @@ int cpufreq_update_policy(unsigned int cpu) if (!data) return -ENODEV; - down(&data->lock); + if (unlikely(lock_policy_rwsem_write(cpu))) + return -EINVAL; dprintk("updating policy for CPU %u\n", cpu); - memcpy(&policy, - data, - sizeof(struct cpufreq_policy)); + 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; + /* 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); + } + } + ret = __cpufreq_set_policy(data, &policy); - up(&data->lock); + unlock_policy_rwsem_write(cpu); cpufreq_cpu_put(data); return ret; } EXPORT_SYMBOL(cpufreq_update_policy); +static int __cpuinit cpufreq_cpu_callback(struct notifier_block *nfb, + unsigned long action, void *hcpu) +{ + unsigned int cpu = (unsigned long)hcpu; + struct sys_device *sys_dev; + + sys_dev = get_cpu_sysdev(cpu); + if (sys_dev) { + switch (action) { + case CPU_ONLINE: + case CPU_ONLINE_FROZEN: + cpufreq_add_dev(sys_dev); + break; + case CPU_DOWN_PREPARE: + case CPU_DOWN_PREPARE_FROZEN: + if (unlikely(lock_policy_rwsem_write(cpu))) + BUG(); + + __cpufreq_remove_dev(sys_dev); + break; + case CPU_DOWN_FAILED: + case CPU_DOWN_FAILED_FROZEN: + cpufreq_add_dev(sys_dev); + break; + } + } + return NOTIFY_OK; +} + +static struct notifier_block __refdata cpufreq_cpu_notifier = +{ + .notifier_call = cpufreq_cpu_callback, +}; /********************************************************************* * REGISTER / UNREGISTER CPUFREQ DRIVER * @@ -1422,9 +1795,9 @@ 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) @@ -1462,8 +1835,10 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data) /* if all ->init() calls failed, unregister */ if (ret) { - dprintk("no CPU initialized for driver %s\n", driver_data->name); - 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; @@ -1472,11 +1847,12 @@ 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); + return ret; } EXPORT_SYMBOL_GPL(cpufreq_register_driver); @@ -1484,7 +1860,7 @@ 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. @@ -1503,6 +1879,7 @@ int cpufreq_unregister_driver(struct cpufreq_driver *driver) 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; @@ -1511,3 +1888,16 @@ int cpufreq_unregister_driver(struct cpufreq_driver *driver) return 0; } EXPORT_SYMBOL_GPL(cpufreq_unregister_driver); + +static int __init cpufreq_core_init(void) +{ + int cpu; + + for_each_possible_cpu(cpu) { + per_cpu(policy_cpu, cpu) = -1; + init_rwsem(&per_cpu(cpu_policy_rwsem, cpu)); + } + return 0; +} + +core_initcall(cpufreq_core_init);