static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data);
#ifdef CONFIG_HOTPLUG_CPU
/* This one keeps track of the previously set governor of a removed CPU */
-static DEFINE_PER_CPU(struct cpufreq_governor *, cpufreq_cpu_governor);
+static DEFINE_PER_CPU(char[CPUFREQ_NAME_LEN], cpufreq_cpu_governor);
#endif
static DEFINE_SPINLOCK(cpufreq_driver_lock);
* 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.
+ * - Lock should not be held across
+ * __cpufreq_governor(data, CPUFREQ_GOV_STOP);
*/
-static DEFINE_PER_CPU(int, policy_cpu);
+static DEFINE_PER_CPU(int, cpufreq_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); \
+ int policy_cpu = per_cpu(cpufreq_policy_cpu, cpu); \
BUG_ON(policy_cpu == -1); \
down_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu)); \
if (unlikely(!cpu_online(cpu))) { \
void unlock_policy_rwsem_read(int cpu)
{
- int policy_cpu = per_cpu(policy_cpu, cpu);
+ int policy_cpu = per_cpu(cpufreq_policy_cpu, cpu);
BUG_ON(policy_cpu == -1);
up_read(&per_cpu(cpu_policy_rwsem, policy_cpu));
}
void unlock_policy_rwsem_write(int cpu)
{
- int policy_cpu = per_cpu(policy_cpu, cpu);
+ int policy_cpu = per_cpu(cpufreq_policy_cpu, cpu);
BUG_ON(policy_cpu == -1);
up_write(&per_cpu(cpu_policy_rwsem, policy_cpu));
}
return policy->governor->show_setspeed(policy, buf);
}
+/**
+ * show_scaling_driver - show the current cpufreq HW/BIOS limitation
+ */
+static ssize_t show_bios_limit(struct cpufreq_policy *policy, char *buf)
+{
+ unsigned int limit;
+ int ret;
+ if (cpufreq_driver->bios_limit) {
+ ret = cpufreq_driver->bios_limit(policy->cpu, &limit);
+ if (!ret)
+ return sprintf(buf, "%u\n", limit);
+ }
+ return sprintf(buf, "%u\n", policy->cpuinfo.max_freq);
+}
+
#define define_one_ro(_name) \
static struct freq_attr _name = \
__ATTR(_name, 0444, show_##_name, NULL)
define_one_ro(scaling_available_governors);
define_one_ro(scaling_driver);
define_one_ro(scaling_cur_freq);
+define_one_ro(bios_limit);
define_one_ro(related_cpus);
define_one_ro(affected_cpus);
define_one_rw(scaling_min_freq);
NULL
};
+struct kobject *cpufreq_global_kobject;
+EXPORT_SYMBOL(cpufreq_global_kobject);
+
#define to_policy(k) container_of(k, struct cpufreq_policy, kobj)
#define to_attr(a) container_of(a, struct freq_attr, attr)
.release = cpufreq_sysfs_release,
};
-
-/**
- * cpufreq_add_dev - add a CPU device
- *
- * Adds the cpufreq interface for a CPU device.
+/*
+ * Returns:
+ * Negative: Failure
+ * 0: Success
+ * Positive: When we have a managed CPU and the sysfs got symlinked
*/
-static int cpufreq_add_dev(struct sys_device *sys_dev)
+static int cpufreq_add_dev_policy(unsigned int cpu,
+ struct cpufreq_policy *policy,
+ 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;
+#ifdef CONFIG_SMP
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);
-
-#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 = kzalloc(sizeof(struct cpufreq_policy), GFP_KERNEL);
- if (!policy) {
- ret = -ENOMEM;
- goto nomem_out;
- }
- if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL)) {
- kfree(policy);
- ret = -ENOMEM;
- goto nomem_out;
- }
- if (!alloc_cpumask_var(&policy->related_cpus, GFP_KERNEL)) {
- free_cpumask_var(policy->cpus);
- kfree(policy);
- ret = -ENOMEM;
- goto nomem_out;
- }
-
- policy->cpu = cpu;
- cpumask_copy(policy->cpus, cpumask_of(cpu));
-
- /* 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);
-
- /* 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
- */
- ret = cpufreq_driver->init(policy);
- if (ret) {
- dprintk("initialization failed\n");
- goto err_out;
- }
- policy->user_policy.min = policy->min;
- policy->user_policy.max = policy->max;
-
- blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
- CPUFREQ_START, policy);
-
-#ifdef CONFIG_SMP
-
#ifdef CONFIG_HOTPLUG_CPU
- if (per_cpu(cpufreq_cpu_governor, cpu)) {
- policy->governor = per_cpu(cpufreq_cpu_governor, cpu);
+ struct cpufreq_governor *gov;
+
+ gov = __find_governor(per_cpu(cpufreq_cpu_governor, cpu));
+ if (gov) {
+ policy->governor = gov;
dprintk("Restoring governor %s for cpu %d\n",
policy->governor->name, cpu);
}
#endif
for_each_cpu(j, policy->cpus) {
+ struct cpufreq_policy *managed_policy;
+
if (cpu == j)
continue;
/* Check for existing affected CPUs.
* They may not be aware of it due to CPU Hotplug.
+ * cpufreq_cpu_put is called when the device is removed
+ * in __cpufreq_remove_dev()
*/
- managed_policy = cpufreq_cpu_get(j); /* FIXME: Where is this released? What about error paths? */
+ managed_policy = cpufreq_cpu_get(j);
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;
+ per_cpu(cpufreq_policy_cpu, cpu) = managed_policy->cpu;
+
+ if (lock_policy_rwsem_write(cpu) < 0) {
+ /* Should not go through policy unlock path */
+ if (cpufreq_driver->exit)
+ cpufreq_driver->exit(policy);
+ cpufreq_cpu_put(managed_policy);
+ return -EBUSY;
+ }
spin_lock_irqsave(&cpufreq_driver_lock, flags);
cpumask_copy(managed_policy->cpus, policy->cpus);
&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() */
+ cpufreq_cpu_put(managed_policy);
+ /*
+ * Success. We only needed to be added to the mask.
+ * Call driver->exit() because only the cpu parent of
+ * the kobj needed to call init().
+ */
+ if (cpufreq_driver->exit)
+ cpufreq_driver->exit(policy);
+
+ if (!ret)
+ return 1;
+ else
+ return ret;
}
}
#endif
- memcpy(&new_policy, policy, sizeof(struct cpufreq_policy));
+ return ret;
+}
+
+
+/* symlink affected CPUs */
+static int cpufreq_add_dev_symlink(unsigned int cpu,
+ struct cpufreq_policy *policy)
+{
+ unsigned int j;
+ int ret = 0;
+
+ for_each_cpu(j, policy->cpus) {
+ struct cpufreq_policy *managed_policy;
+ struct sys_device *cpu_sys_dev;
+
+ if (j == cpu)
+ continue;
+ if (!cpu_online(j))
+ continue;
+
+ dprintk("CPU %u already managed, adding link\n", j);
+ managed_policy = cpufreq_cpu_get(cpu);
+ cpu_sys_dev = get_cpu_sysdev(j);
+ ret = sysfs_create_link(&cpu_sys_dev->kobj, &policy->kobj,
+ "cpufreq");
+ if (ret) {
+ cpufreq_cpu_put(managed_policy);
+ return ret;
+ }
+ }
+ return ret;
+}
+
+static int cpufreq_add_dev_interface(unsigned int cpu,
+ struct cpufreq_policy *policy,
+ struct sys_device *sys_dev)
+{
+ struct cpufreq_policy new_policy;
+ struct freq_attr **drv_attr;
+ unsigned long flags;
+ int ret = 0;
+ unsigned int j;
/* prepare interface data */
- ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq, &sys_dev->kobj,
- "cpufreq");
+ ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq,
+ &sys_dev->kobj, "cpufreq");
if (ret)
- goto err_out_driver_exit;
+ return ret;
/* set up files for this cpu device */
drv_attr = cpufreq_driver->attr;
while ((drv_attr) && (*drv_attr)) {
ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr));
if (ret)
- goto err_out_driver_exit;
+ goto err_out_kobj_put;
drv_attr++;
}
if (cpufreq_driver->get) {
ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr);
if (ret)
- goto err_out_driver_exit;
+ goto err_out_kobj_put;
}
if (cpufreq_driver->target) {
ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
if (ret)
- goto err_out_driver_exit;
+ goto err_out_kobj_put;
+ }
+ if (cpufreq_driver->bios_limit) {
+ ret = sysfs_create_file(&policy->kobj, &bios_limit.attr);
+ if (ret)
+ goto err_out_kobj_put;
}
spin_lock_irqsave(&cpufreq_driver_lock, flags);
for_each_cpu(j, policy->cpus) {
+ if (!cpu_online(j))
+ continue;
per_cpu(cpufreq_cpu_data, j) = policy;
- per_cpu(policy_cpu, j) = policy->cpu;
+ per_cpu(cpufreq_policy_cpu, j) = policy->cpu;
}
spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
- /* symlink affected CPUs */
- for_each_cpu(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;
- }
+ ret = cpufreq_add_dev_symlink(cpu, policy);
+ if (ret)
+ goto err_out_kobj_put;
- policy->governor = NULL; /* to assure that the starting sequence is
- * run in cpufreq_set_policy */
+ memcpy(&new_policy, policy, sizeof(struct cpufreq_policy));
+ /* assure that the starting sequence is run in __cpufreq_set_policy */
+ policy->governor = NULL;
/* set default policy */
ret = __cpufreq_set_policy(policy, &new_policy);
if (ret) {
dprintk("setting policy failed\n");
- goto err_out_unregister;
+ if (cpufreq_driver->exit)
+ cpufreq_driver->exit(policy);
}
+ return ret;
+
+err_out_kobj_put:
+ kobject_put(&policy->kobj);
+ wait_for_completion(&policy->kobj_unregister);
+ return ret;
+}
+
+
+/**
+ * cpufreq_add_dev - add a CPU device
+ *
+ * Adds the cpufreq interface for a CPU device.
+ *
+ * The Oracle says: try running cpufreq registration/unregistration concurrently
+ * with with cpu hotplugging and all hell will break loose. Tried to clean this
+ * mess up, but more thorough testing is needed. - Mathieu
+ */
+static int cpufreq_add_dev(struct sys_device *sys_dev)
+{
+ unsigned int cpu = sys_dev->id;
+ int ret = 0, found = 0;
+ struct cpufreq_policy *policy;
+ unsigned long flags;
+ unsigned int j;
+#ifdef CONFIG_HOTPLUG_CPU
+ int sibling;
+#endif
+
+ 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;
+ }
+
+ ret = -ENOMEM;
+ policy = kzalloc(sizeof(struct cpufreq_policy), GFP_KERNEL);
+ if (!policy)
+ goto nomem_out;
+
+ if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL))
+ goto err_free_policy;
+
+ if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL))
+ goto err_free_cpumask;
+
+ policy->cpu = cpu;
+ cpumask_copy(policy->cpus, cpumask_of(cpu));
+
+ /* Initially set CPU itself as the policy_cpu */
+ per_cpu(cpufreq_policy_cpu, cpu) = cpu;
+ ret = (lock_policy_rwsem_write(cpu) < 0);
+ WARN_ON(ret);
+
+ init_completion(&policy->kobj_unregister);
+ INIT_WORK(&policy->update, handle_update);
+
+ /* Set governor before ->init, so that driver could check it */
+#ifdef CONFIG_HOTPLUG_CPU
+ for_each_online_cpu(sibling) {
+ struct cpufreq_policy *cp = per_cpu(cpufreq_cpu_data, sibling);
+ if (cp && cp->governor &&
+ (cpumask_test_cpu(cpu, cp->related_cpus))) {
+ policy->governor = cp->governor;
+ found = 1;
+ break;
+ }
+ }
+#endif
+ if (!found)
+ 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
+ */
+ ret = cpufreq_driver->init(policy);
+ if (ret) {
+ dprintk("initialization failed\n");
+ goto err_unlock_policy;
+ }
+ policy->user_policy.min = policy->min;
+ policy->user_policy.max = policy->max;
+
+ blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
+ CPUFREQ_START, policy);
+
+ ret = cpufreq_add_dev_policy(cpu, policy, sys_dev);
+ if (ret) {
+ if (ret > 0)
+ /* This is a managed cpu, symlink created,
+ exit with 0 */
+ ret = 0;
+ goto err_unlock_policy;
+ }
+
+ ret = cpufreq_add_dev_interface(cpu, policy, sys_dev);
+ if (ret)
+ goto err_out_unregister;
unlock_policy_rwsem_write(cpu);
kobject_put(&policy->kobj);
wait_for_completion(&policy->kobj_unregister);
-err_out_driver_exit:
- if (cpufreq_driver->exit)
- cpufreq_driver->exit(policy);
-
-err_out:
+err_unlock_policy:
unlock_policy_rwsem_write(cpu);
+err_free_cpumask:
+ free_cpumask_var(policy->cpus);
+err_free_policy:
kfree(policy);
-
nomem_out:
module_put(cpufreq_driver->owner);
module_out:
#ifdef CONFIG_SMP
#ifdef CONFIG_HOTPLUG_CPU
- per_cpu(cpufreq_cpu_governor, cpu) = data->governor;
+ strncpy(per_cpu(cpufreq_cpu_governor, cpu), data->governor->name,
+ CPUFREQ_NAME_LEN);
#endif
/* if we have other CPUs still registered, we need to unlink them,
continue;
dprintk("removing link for cpu %u\n", j);
#ifdef CONFIG_HOTPLUG_CPU
- per_cpu(cpufreq_cpu_governor, j) = data->governor;
+ strncpy(per_cpu(cpufreq_cpu_governor, j),
+ data->governor->name, CPUFREQ_NAME_LEN);
#endif
cpu_sys_dev = get_cpu_sysdev(j);
sysfs_remove_link(&cpu_sys_dev->kobj, "cpufreq");
if (cpufreq_driver->target)
__cpufreq_governor(data, CPUFREQ_GOV_STOP);
- unlock_policy_rwsem_write(cpu);
-
kobject_put(&data->kobj);
/* we need to make sure that the underlying kobj is actually
if (cpufreq_driver->exit)
cpufreq_driver->exit(data);
+ unlock_policy_rwsem_write(cpu);
+
free_cpumask_var(data->related_cpus);
free_cpumask_var(data->cpus);
kfree(data);
static int cpufreq_suspend(struct sys_device *sysdev, pm_message_t pmsg)
{
- int cpu = sysdev->id;
int ret = 0;
- unsigned int cur_freq = 0;
+
+ int cpu = sysdev->id;
struct cpufreq_policy *cpu_policy;
dprintk("suspending cpu %u\n", cpu);
if (cpufreq_driver->suspend) {
ret = cpufreq_driver->suspend(cpu_policy, pmsg);
- if (ret) {
+ if (ret)
printk(KERN_ERR "cpufreq: suspend failed in ->suspend "
"step on CPU %u\n", cpu_policy->cpu);
- goto out;
- }
- }
-
- 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_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.
+ * 2.) schedule call cpufreq_update_policy() ASAP as interrupts are
+ * restored. It will verify that the current freq is in sync with
+ * what we believe it to be. This is a bit later than when it
+ * should be, but nonethteless it's better than calling
+ * cpufreq_driver->get() here which might re-enable interrupts...
*/
static int cpufreq_resume(struct sys_device *sysdev)
{
- int cpu = sysdev->id;
int ret = 0;
+
+ int cpu = sysdev->id;
struct cpufreq_policy *cpu_policy;
dprintk("resuming cpu %u\n", cpu);
}
}
- if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
- unsigned int cur_freq = 0;
-
- if (cpufreq_driver->get)
- 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");
- 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_RESUMECHANGE, &freqs);
- adjust_jiffies(CPUFREQ_RESUMECHANGE, &freqs);
-
- cpu_policy->cur = cur_freq;
- }
- }
-
-out:
schedule_work(&cpu_policy->update);
+
fail:
cpufreq_cpu_put(cpu_policy);
return ret;
void cpufreq_unregister_governor(struct cpufreq_governor *governor)
{
+#ifdef CONFIG_HOTPLUG_CPU
+ int cpu;
+#endif
+
if (!governor)
return;
+#ifdef CONFIG_HOTPLUG_CPU
+ for_each_present_cpu(cpu) {
+ if (cpu_online(cpu))
+ continue;
+ if (!strcmp(per_cpu(cpufreq_cpu_governor, cpu), governor->name))
+ strcpy(per_cpu(cpufreq_cpu_governor, cpu), "\0");
+ }
+#endif
+
mutex_lock(&cpufreq_governor_mutex);
list_del(&governor->governor_list);
mutex_unlock(&cpufreq_governor_mutex);
dprintk("governor switch\n");
/* end old governor */
- if (data->governor)
+ if (data->governor) {
+ /*
+ * Need to release the rwsem around governor
+ * stop due to lock dependency between
+ * cancel_delayed_work_sync and the read lock
+ * taken in the delayed work handler.
+ */
+ unlock_policy_rwsem_write(data->cpu);
__cpufreq_governor(data, CPUFREQ_GOV_STOP);
+ lock_policy_rwsem_write(data->cpu);
+ }
/* start new governor */
data->governor = policy->governor;
int cpu;
for_each_possible_cpu(cpu) {
- per_cpu(policy_cpu, cpu) = -1;
+ per_cpu(cpufreq_policy_cpu, cpu) = -1;
init_rwsem(&per_cpu(cpu_policy_rwsem, cpu));
}
+
+ cpufreq_global_kobject = kobject_create_and_add("cpufreq",
+ &cpu_sysdev_class.kset.kobj);
+ BUG_ON(!cpufreq_global_kobject);
+
return 0;
}
-
core_initcall(cpufreq_core_init);