policy->cpus is cleared unconditionally now on hotplug-out of a CPU and it can be checked to know if a policy is active or not. Create helper routines to iterate over all active/inactive policies, based on policy->cpus field.

Replace all instances of for_each_policy() with for_each_active_policy() to make them iterate only for active policies. (We haven't made changes yet to keep inactive policies in the same list, but that will be followed in a later patch).

Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 66 ++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 59 insertions(+), 7 deletions(-)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 8cf0c0e7aea8..74d9fcbbe4f9 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -31,10 +31,62 @@ #include <linux/tick.h> #include <trace/events/power.h>

-/* Macros to iterate over lists */ -/* Iterate over online CPUs policies */ static LIST_HEAD(cpufreq_policy_list); -#define for_each_policy(__policy) \ + +static inline bool policy_is_inactive(struct cpufreq_policy *policy) +{ + return cpumask_empty(policy->cpus); +} + +static bool suitable_policy(struct cpufreq_policy *policy, bool active) +{ + return active == !policy_is_inactive(policy); +} + +/* Finds Next Acive/Inactive policy */ +static struct cpufreq_policy *next_policy(struct cpufreq_policy *policy, + bool active) +{ + do { + policy = list_next_entry(policy, policy_list); + + /* No more policies in the list */ + if (&policy->policy_list == &cpufreq_policy_list) + return NULL; + } while (!suitable_policy(policy, active)); + + return policy; +} + +static struct cpufreq_policy *first_policy(bool active) +{ + struct cpufreq_policy *policy; + + /* No policies in the list */ + if (list_empty(&cpufreq_policy_list)) + return NULL; + + policy = list_first_entry(&cpufreq_policy_list, typeof(*policy), + policy_list); + + if (!suitable_policy(policy, active)) + policy = next_policy(policy, active); + + return policy; +} + +/* Macros to iterate over CPU policies */ +#define for_each_suitable_policy(__policy, __active) \ + for (__policy = first_policy(__active); \ + __policy; \ + __policy = next_policy(__policy, __active)) + +#define for_each_active_policy(__policy) \ + for_each_suitable_policy(__policy, true) +#define for_each_inactive_policy(__policy) \ + for_each_suitable_policy(__policy, false) + +#define for_each_policy(__policy) \ list_for_each_entry(__policy, &cpufreq_policy_list, policy_list)

/* Iterate over governors */ @@ -1156,7 +1208,7 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif)

/* Check if this CPU already has a policy to manage it */ read_lock_irqsave(&cpufreq_driver_lock, flags); - for_each_policy(policy) { + for_each_active_policy(policy) { if (cpumask_test_cpu(cpu, policy->related_cpus)) { read_unlock_irqrestore(&cpufreq_driver_lock, flags); ret = cpufreq_add_policy_cpu(policy, cpu, dev); @@ -1674,7 +1726,7 @@ void cpufreq_suspend(void)

pr_debug("%s: Suspending Governors\n", __func__);

- for_each_policy(policy) { + for_each_active_policy(policy) { if (__cpufreq_governor(policy, CPUFREQ_GOV_STOP)) pr_err("%s: Failed to stop governor for policy: %p\n", __func__, policy); @@ -1708,7 +1760,7 @@ void cpufreq_resume(void)

pr_debug("%s: Resuming Governors\n", __func__);

- for_each_policy(policy) { + for_each_active_policy(policy) { if (cpufreq_driver->resume && cpufreq_driver->resume(policy)) pr_err("%s: Failed to resume driver: %p\n", __func__, policy); @@ -2354,7 +2406,7 @@ static int cpufreq_boost_set_sw(int state) struct cpufreq_policy *policy; int ret = -EINVAL;

- for_each_policy(policy) { + for_each_active_policy(policy) { freq_table = cpufreq_frequency_get_table(policy->cpu); if (freq_table) { ret = cpufreq_frequency_table_cpuinfo(policy,

We can extract the same information from cpufreq_cpu_data as it is also available for inactive policies now. And so don't need cpufreq_cpu_data_fallback anymore.

Also add a WARN_ON() for the case where we try to restore from an active policy.

Acked-by: Saravana Kannan skannan@codeaurora.org Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 25 ++++++------------------- 1 file changed, 6 insertions(+), 19 deletions(-)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index e899a5446d0e..eb0c3a802b14 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -101,7 +101,6 @@ static LIST_HEAD(cpufreq_governor_list); */ static struct cpufreq_driver *cpufreq_driver; static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data); -static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data_fallback); static DEFINE_RWLOCK(cpufreq_driver_lock); DEFINE_MUTEX(cpufreq_governor_lock);

@@ -1081,13 +1080,14 @@ static struct cpufreq_policy *cpufreq_policy_restore(unsigned int cpu) unsigned long flags;

read_lock_irqsave(&cpufreq_driver_lock, flags); - - policy = per_cpu(cpufreq_cpu_data_fallback, cpu); - + policy = per_cpu(cpufreq_cpu_data, cpu); read_unlock_irqrestore(&cpufreq_driver_lock, flags);

- if (policy) + if (likely(policy)) { + /* Policy should be inactive here */ + WARN_ON(!policy_is_inactive(policy)); policy->governor = NULL; + }

return policy; } @@ -1383,11 +1383,8 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) if (cpufreq_driver->exit) cpufreq_driver->exit(policy); err_set_policy_cpu: - if (recover_policy) { - /* Do not leave stale fallback data behind. */ - per_cpu(cpufreq_cpu_data_fallback, cpu) = NULL; + if (recover_policy) cpufreq_policy_put_kobj(policy); - } cpufreq_policy_free(policy);

nomem_out: @@ -1401,21 +1398,11 @@ static int __cpufreq_remove_dev_prepare(struct device *dev, { unsigned int cpu = dev->id, cpus; int ret; - unsigned long flags; struct cpufreq_policy *policy;

pr_debug("%s: unregistering CPU %u\n", __func__, cpu);

- write_lock_irqsave(&cpufreq_driver_lock, flags); - policy = cpufreq_cpu_get_raw(cpu); - - /* Save the policy somewhere when doing a light-weight tear-down */ - if (cpufreq_suspended) - per_cpu(cpufreq_cpu_data_fallback, cpu) = policy; - - write_unlock_irqrestore(&cpufreq_driver_lock, flags); - if (!policy) { pr_debug("%s: No cpu_data found\n", __func__); return -EINVAL;

History of which governor was used last is common to all CPUs within a policy and maintaining it per-cpu isn't the best approach for sure.

Apart from wasting memory, this also increases the complexity of managing this data structure as it has to be updated for all CPUs.

To make that somewhat simpler, lets store this information in a new field 'last_governor' in struct cpufreq_policy and update it on removal of last cpu of a policy.

As a side-effect it also solves an old problem, consider a system with two clusters 0 & 1. And there is one policy per cluster.

Cluster 0: CPU0 and 1. Cluster 1: CPU2 and 3.

- CPU2 is first brought online, and governor is set to performance (default as cpufreq_cpu_governor wasn't set). - Governor is changed to ondemand. - CPU2 is taken offline and cpufreq_cpu_governor is updated for CPU2. - CPU3 is brought online. - Because cpufreq_cpu_governor wasn't set for CPU3, the default governor performance is picked for CPU3.

This patch fixes the bug as we now have a single variable to update for policy.

Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 30 +++++++++++++++--------------- include/linux/cpufreq.h | 1 + 2 files changed, 16 insertions(+), 15 deletions(-)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index e6a63d6ba6f1..16275ba6428e 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -104,9 +104,6 @@ static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data); static DEFINE_RWLOCK(cpufreq_driver_lock); DEFINE_MUTEX(cpufreq_governor_lock);

-/* This one keeps track of the previously set governor of a removed CPU */ -static DEFINE_PER_CPU(char[CPUFREQ_NAME_LEN], cpufreq_cpu_governor); - /* Flag to suspend/resume CPUFreq governors */ static bool cpufreq_suspended;

@@ -1017,7 +1014,7 @@ static void cpufreq_init_policy(struct cpufreq_policy *policy) memcpy(&new_policy, policy, sizeof(*policy));

/* Update governor of new_policy to the governor used before hotplug */ - gov = find_governor(per_cpu(cpufreq_cpu_governor, policy->cpu)); + gov = find_governor(policy->last_governor); if (gov) pr_debug("Restoring governor %s for cpu %d\n", policy->governor->name, policy->cpu); @@ -1411,14 +1408,15 @@ static int __cpufreq_remove_dev_prepare(struct device *dev, pr_err("%s: Failed to stop governor\n", __func__); return ret; } - - strncpy(per_cpu(cpufreq_cpu_governor, cpu), - policy->governor->name, CPUFREQ_NAME_LEN); }

- down_read(&policy->rwsem); + down_write(&policy->rwsem); cpus = cpumask_weight(policy->cpus); - up_read(&policy->rwsem); + + if (has_target() && cpus == 1) + strncpy(policy->last_governor, policy->governor->name, + CPUFREQ_NAME_LEN); + up_write(&policy->rwsem);

if (cpu != policy->cpu) { sysfs_remove_link(&dev->kobj, "cpufreq"); @@ -2135,7 +2133,8 @@ EXPORT_SYMBOL_GPL(cpufreq_register_governor);

void cpufreq_unregister_governor(struct cpufreq_governor *governor) { - int cpu; + struct cpufreq_policy *policy; + unsigned long flags;

if (!governor) return; @@ -2143,12 +2142,13 @@ void cpufreq_unregister_governor(struct cpufreq_governor *governor) if (cpufreq_disabled()) return;

- 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"); + /* clear last_governor for all inactive policies */ + read_lock_irqsave(&cpufreq_driver_lock, flags); + for_each_inactive_policy(policy) { + if (!strcmp(policy->last_governor, governor->name)) + strcpy(policy->last_governor, "\0"); } + read_unlock_irqrestore(&cpufreq_driver_lock, flags);

mutex_lock(&cpufreq_governor_mutex); list_del(&governor->governor_list); diff --git a/include/linux/cpufreq.h b/include/linux/cpufreq.h index 2ee4888c1f47..48e37c07eb84 100644 --- a/include/linux/cpufreq.h +++ b/include/linux/cpufreq.h @@ -80,6 +80,7 @@ struct cpufreq_policy { struct cpufreq_governor *governor; /* see below */ void *governor_data; bool governor_enabled; /* governor start/stop flag */ + char last_governor[CPUFREQ_NAME_LEN]; /* last governor used */

struct work_struct update; /* if update_policy() needs to be * called, but you're in IRQ context */

Later commits would change the way policies are managed today. Policies wouldn't be freed on cpu hotplug (currently they aren't freed only for suspend), and while the CPU is offline, the sysfs cpufreq files would still be present.

User may accidentally try to update the sysfs files in following directory: '/sys/devices/system/cpu/cpuX/cpufreq/'. And that would result in undefined behavior as policy wouldn't be active then.

Apart from updating the store() routine, we also update __cpufreq_get() which can call cpufreq_out_of_sync(). The later routine tries to update policy->cur and starts notifying kernel about it.

Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 11 +++++++++++ 1 file changed, 11 insertions(+)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index c08de5e985c7..983961ba2f72 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -875,11 +875,18 @@ static ssize_t store(struct kobject *kobj, struct attribute *attr,

down_write(&policy->rwsem);

+ /* Updating inactive policies is invalid, so avoid doing that. */ + if (unlikely(policy_is_inactive(policy))) { + ret = -EPERM; + goto unlock_policy_rwsem; + } + if (fattr->store) ret = fattr->store(policy, buf, count); else ret = -EIO;

+unlock_policy_rwsem: up_write(&policy->rwsem);

up_read(&cpufreq_rwsem); @@ -1607,6 +1614,10 @@ static unsigned int __cpufreq_get(struct cpufreq_policy *policy)

ret_freq = cpufreq_driver->get(policy->cpu);

+ /* Updating inactive policies is invalid, so avoid doing that. */ + if (unlikely(policy_is_inactive(policy))) + return ret_freq; + if (ret_freq && policy->cur && !(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) { /* verify no discrepancy between actual and

When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu.

There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume

To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier.

- Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed.

- Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver).

Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence.

Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface..

Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online'

We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks.

[ Something similar attempted by Saravana earlier ] Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 174 +++++++++++++++++++++++++--------------------- 1 file changed, 95 insertions(+), 79 deletions(-)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index b5e883f497dd..b07a3387f2bc 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -977,14 +977,15 @@ static inline int add_remove_cpu_dev_symlink(struct cpufreq_policy *policy, return 0; }

-/* Add/remove symlinks for all affected CPUs */ +/* Add/remove symlinks for all related CPUs */ static int cpufreq_add_remove_dev_symlink(struct cpufreq_policy *policy, bool add) { unsigned int j; int ret = 0;

- for_each_cpu(j, policy->cpus) { + /* Some related CPUs might not be present (physically hotplugged) */ + for_each_cpu_and(j, policy->related_cpus, cpu_present_mask) { if (j == policy->kobj_cpu) continue;

@@ -1092,7 +1093,7 @@ static int cpufreq_add_policy_cpu(struct cpufreq_policy *policy, } }

- return sysfs_create_link(&dev->kobj, &policy->kobj, "cpufreq"); + return 0; }

static struct cpufreq_policy *cpufreq_policy_restore(unsigned int cpu) @@ -1112,7 +1113,7 @@ static struct cpufreq_policy *cpufreq_policy_restore(unsigned int cpu) return policy; }

-static struct cpufreq_policy *cpufreq_policy_alloc(void) +static struct cpufreq_policy *cpufreq_policy_alloc(int cpu) { struct cpufreq_policy *policy;

@@ -1133,6 +1134,11 @@ static struct cpufreq_policy *cpufreq_policy_alloc(void) init_completion(&policy->kobj_unregister); INIT_WORK(&policy->update, handle_update);

+ policy->cpu = cpu; + + /* Set this once on allocation */ + policy->kobj_cpu = cpu; + return policy;

err_free_cpumask: @@ -1151,10 +1157,11 @@ static void cpufreq_policy_put_kobj(struct cpufreq_policy *policy) blocking_notifier_call_chain(&cpufreq_policy_notifier_list, CPUFREQ_REMOVE_POLICY, policy);

- down_read(&policy->rwsem); + down_write(&policy->rwsem); + cpufreq_add_remove_dev_symlink(policy, false); kobj = &policy->kobj; cmp = &policy->kobj_unregister; - up_read(&policy->rwsem); + up_write(&policy->rwsem); kobject_put(kobj);

/* @@ -1185,27 +1192,14 @@ static void cpufreq_policy_free(struct cpufreq_policy *policy) kfree(policy); }

-static int update_policy_cpu(struct cpufreq_policy *policy, unsigned int cpu, - struct device *cpu_dev) +static void update_policy_cpu(struct cpufreq_policy *policy, unsigned int cpu) { - int ret; - if (WARN_ON(cpu == policy->cpu)) - return 0; - - /* Move kobject to the new policy->cpu */ - ret = kobject_move(&policy->kobj, &cpu_dev->kobj); - if (ret) { - pr_err("%s: Failed to move kobj: %d\n", __func__, ret); - return ret; - } + return;

down_write(&policy->rwsem); policy->cpu = cpu; - policy->kobj_cpu = cpu; up_write(&policy->rwsem); - - return 0; }

/** @@ -1223,13 +1217,25 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) int ret = -ENOMEM; struct cpufreq_policy *policy; unsigned long flags; - bool recover_policy = cpufreq_suspended; - - if (cpu_is_offline(cpu)) - return 0; + bool recover_policy = !sif;

pr_debug("adding CPU %u\n", cpu);

+ /* + * Only possible if 'cpu' wasn't physically present earlier and we are + * here from subsys_interface add callback. A hotplug notifier will + * follow and we will handle it like logical CPU hotplug then. For now, + * just create the sysfs link. + */ + if (cpu_is_offline(cpu)) { + policy = per_cpu(cpufreq_cpu_data, cpu); + /* No need to create link of the first cpu of a policy */ + if (!policy) + return 0; + + return add_remove_cpu_dev_symlink(policy, cpu, true); + } + if (!down_read_trylock(&cpufreq_rwsem)) return 0;

@@ -1249,7 +1255,7 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) policy = recover_policy ? cpufreq_policy_restore(cpu) : NULL; if (!policy) { recover_policy = false; - policy = cpufreq_policy_alloc(); + policy = cpufreq_policy_alloc(cpu); if (!policy) goto nomem_out; } @@ -1260,12 +1266,8 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) * the creation of a brand new one. So we need to perform this update * by invoking update_policy_cpu(). */ - if (recover_policy && cpu != policy->cpu) { - WARN_ON(update_policy_cpu(policy, cpu, dev)); - } else { - policy->cpu = cpu; - policy->kobj_cpu = cpu; - } + if (recover_policy && cpu != policy->cpu) + update_policy_cpu(policy, cpu);

cpumask_copy(policy->cpus, cpumask_of(cpu));

@@ -1444,29 +1446,14 @@ static int __cpufreq_remove_dev_prepare(struct device *dev, CPUFREQ_NAME_LEN); up_write(&policy->rwsem);

- if (cpu != policy->kobj_cpu) { - sysfs_remove_link(&dev->kobj, "cpufreq"); - } else if (cpus > 1) { - /* Nominate new CPU */ - int new_cpu = cpumask_any_but(policy->cpus, cpu); - struct device *cpu_dev = get_cpu_device(new_cpu); - - sysfs_remove_link(&cpu_dev->kobj, "cpufreq"); - ret = update_policy_cpu(policy, new_cpu, cpu_dev); - if (ret) { - if (sysfs_create_link(&cpu_dev->kobj, &policy->kobj, - "cpufreq")) - pr_err("%s: Failed to restore kobj link to cpu:%d\n", - __func__, cpu_dev->id); - return ret; - } + if (cpu != policy->cpu) + return 0;

- if (!cpufreq_suspended) - pr_debug("%s: policy Kobject moved to cpu: %d from: %d\n", - __func__, new_cpu, cpu); - } else if (cpufreq_driver->stop_cpu) { + if (cpus > 1) + /* Nominate new CPU */ + update_policy_cpu(policy, cpumask_any_but(policy->cpus, cpu)); + else if (cpufreq_driver->stop_cpu) cpufreq_driver->stop_cpu(policy); - }

return 0; } @@ -1487,32 +1474,11 @@ static int __cpufreq_remove_dev_finish(struct device *dev, cpumask_clear_cpu(cpu, policy->cpus); up_write(&policy->rwsem);

- /* If cpu is last user of policy, free policy */ - if (policy_is_inactive(policy)) { - if (has_target()) { - ret = __cpufreq_governor(policy, - CPUFREQ_GOV_POLICY_EXIT); - if (ret) { - pr_err("%s: Failed to exit governor\n", - __func__); - return ret; - } - } - - if (!cpufreq_suspended) - cpufreq_policy_put_kobj(policy); - - /* - * Perform the ->exit() even during light-weight tear-down, - * since this is a core component, and is essential for the - * subsequent light-weight ->init() to succeed. - */ - if (cpufreq_driver->exit) - cpufreq_driver->exit(policy); + /* Not the last cpu of policy, start governor again ? */ + if (!policy_is_inactive(policy)) { + if (!has_target()) + return 0;

- if (!cpufreq_suspended) - cpufreq_policy_free(policy); - } else if (has_target()) { ret = __cpufreq_governor(policy, CPUFREQ_GOV_START); if (!ret) ret = __cpufreq_governor(policy, CPUFREQ_GOV_LIMITS); @@ -1521,8 +1487,34 @@ static int __cpufreq_remove_dev_finish(struct device *dev, pr_err("%s: Failed to start governor\n", __func__); return ret; } + + return 0; + } + + /* If cpu is last user of policy, free policy */ + if (has_target()) { + ret = __cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT); + if (ret) { + pr_err("%s: Failed to exit governor\n", __func__); + return ret; + } }

+ /* Free the policy kobjects only if the driver is getting removed. */ + if (sif) + cpufreq_policy_put_kobj(policy); + + /* + * Perform the ->exit() even during light-weight tear-down, + * since this is a core component, and is essential for the + * subsequent light-weight ->init() to succeed. + */ + if (cpufreq_driver->exit) + cpufreq_driver->exit(policy); + + if (sif) + cpufreq_policy_free(policy); + return 0; }

@@ -1536,8 +1528,32 @@ static int cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif) unsigned int cpu = dev->id; int ret;

- if (cpu_is_offline(cpu)) + /* + * Only possible if 'cpu' is getting physically removed now. A hotplug + * notifier should have already been called and we just need to remove + * link or free policy here. + */ + if (cpu_is_offline(cpu)) { + struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu); + struct cpumask mask; + + if (!policy) + return 0; + + cpumask_copy(&mask, policy->related_cpus); + cpumask_clear_cpu(cpu, &mask); + + /* + * Free policy only if all policy->related_cpus are removed + * physically. + */ + if (cpumask_intersects(&mask, cpu_present_mask)) + return add_remove_cpu_dev_symlink(policy, cpu, false); + + cpufreq_policy_put_kobj(policy); + cpufreq_policy_free(policy); return 0; + }

ret = __cpufreq_remove_dev_prepare(dev, sif);

policy->kobj is required to be initialized once in the lifetime of a policy. Currently we are initializing it from __cpufreq_add_dev() and that doesn't look to be the best place for doing so as we have to do this on special cases (like: !recover_policy).

We can initialize it from a more obvious place cpufreq_policy_alloc() and that will make code look cleaner, specially the error handling part.

The error handling part of __cpufreq_add_dev() was doing almost the same thing while recover_policy is true or false. Fix that as well by always calling cpufreq_policy_put_kobj() with an additional parameter to skip notification part of it.

Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 46 +++++++++++++++++++++------------------------- 1 file changed, 21 insertions(+), 25 deletions(-)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index a16dd26b39f1..1137e7942278 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1117,9 +1117,10 @@ static struct cpufreq_policy *cpufreq_policy_restore(unsigned int cpu) return policy; }

-static struct cpufreq_policy *cpufreq_policy_alloc(int cpu) +static struct cpufreq_policy *cpufreq_policy_alloc(struct device *dev) { struct cpufreq_policy *policy; + int ret;

policy = kzalloc(sizeof(*policy), GFP_KERNEL); if (!policy) @@ -1131,6 +1132,13 @@ static struct cpufreq_policy *cpufreq_policy_alloc(int cpu) if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL)) goto err_free_cpumask;

+ ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq, &dev->kobj, + "cpufreq"); + if (ret) { + pr_err("%s: failed to init policy->kobj: %d\n", __func__, ret); + goto err_free_rcpumask; + } + INIT_LIST_HEAD(&policy->policy_list); init_rwsem(&policy->rwsem); spin_lock_init(&policy->transition_lock); @@ -1138,13 +1146,15 @@ static struct cpufreq_policy *cpufreq_policy_alloc(int cpu) init_completion(&policy->kobj_unregister); INIT_WORK(&policy->update, handle_update);

- policy->cpu = cpu; + policy->cpu = dev->id;

/* Set this once on allocation */ - policy->kobj_cpu = cpu; + policy->kobj_cpu = dev->id;

return policy;

+err_free_rcpumask: + free_cpumask_var(policy->related_cpus); err_free_cpumask: free_cpumask_var(policy->cpus); err_free_policy: @@ -1153,13 +1163,14 @@ static struct cpufreq_policy *cpufreq_policy_alloc(int cpu) return NULL; }

-static void cpufreq_policy_put_kobj(struct cpufreq_policy *policy) +static void cpufreq_policy_put_kobj(struct cpufreq_policy *policy, bool notify) { struct kobject *kobj; struct completion *cmp;

- blocking_notifier_call_chain(&cpufreq_policy_notifier_list, - CPUFREQ_REMOVE_POLICY, policy); + if (notify) + blocking_notifier_call_chain(&cpufreq_policy_notifier_list, + CPUFREQ_REMOVE_POLICY, policy);

down_write(&policy->rwsem); cpufreq_add_remove_dev_symlink(policy, false); @@ -1249,7 +1260,7 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) policy = recover_policy ? cpufreq_policy_restore(cpu) : NULL; if (!policy) { recover_policy = false; - policy = cpufreq_policy_alloc(cpu); + policy = cpufreq_policy_alloc(dev); if (!policy) goto nomem_out; } @@ -1280,15 +1291,6 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) policy->user_policy.min = policy->min; policy->user_policy.max = policy->max;

- /* prepare interface data */ - ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq, - &dev->kobj, "cpufreq"); - if (ret) { - pr_err("%s: failed to init policy->kobj: %d\n", - __func__, ret); - goto err_init_policy_kobj; - } - write_lock_irqsave(&cpufreq_driver_lock, flags); for_each_cpu(j, policy->related_cpus) per_cpu(cpufreq_cpu_data, j) = policy; @@ -1380,18 +1382,12 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif)

err_out_unregister: err_get_freq: - if (!recover_policy) { - kobject_put(&policy->kobj); - wait_for_completion(&policy->kobj_unregister); - } -err_init_policy_kobj: up_write(&policy->rwsem);

if (cpufreq_driver->exit) cpufreq_driver->exit(policy); err_set_policy_cpu: - if (recover_policy) - cpufreq_policy_put_kobj(policy); + cpufreq_policy_put_kobj(policy, recover_policy); cpufreq_policy_free(policy);

nomem_out: @@ -1490,7 +1486,7 @@ static int __cpufreq_remove_dev_finish(struct device *dev,

/* Free the policy kobjects only if the driver is getting removed. */ if (sif) - cpufreq_policy_put_kobj(policy); + cpufreq_policy_put_kobj(policy, true);

/* * Perform the ->exit() even during light-weight tear-down, @@ -1538,7 +1534,7 @@ static int cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif) if (cpumask_intersects(&mask, cpu_present_mask)) return add_remove_cpu_dev_symlink(policy, cpu, false);

- cpufreq_policy_put_kobj(policy); + cpufreq_policy_put_kobj(policy, true); cpufreq_policy_free(policy); return 0; }

On cpu hot-unplug, we don't need to wait for POST_DEAD notification to restart the governor if the policy has atleast one online cpu left. We can restart the governor right from the DOWN_PREPARE notification instead.

[ Something similar attempted by Saravana earlier ]

Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq.c | 58 ++++++++++++++++++++--------------------------- 1 file changed, 24 insertions(+), 34 deletions(-)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index b030432cbfad..8d64b5e74ab2 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1398,8 +1398,8 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif) static int __cpufreq_remove_dev_prepare(struct device *dev, struct subsys_interface *sif) { - unsigned int cpu = dev->id, cpus; - int ret; + unsigned int cpu = dev->id; + int ret = 0; struct cpufreq_policy *policy;

pr_debug("%s: unregistering CPU %u\n", __func__, cpu); @@ -1419,26 +1419,33 @@ static int __cpufreq_remove_dev_prepare(struct device *dev, }

down_write(&policy->rwsem); - cpus = cpumask_weight(policy->cpus); + cpumask_clear_cpu(cpu, policy->cpus);

- if (has_target() && cpus == 1) - strncpy(policy->last_governor, policy->governor->name, - CPUFREQ_NAME_LEN); + if (policy_is_inactive(policy)) { + if (has_target()) + strncpy(policy->last_governor, policy->governor->name, + CPUFREQ_NAME_LEN); + } else if (cpu == policy->cpu) { + /* Nominate new CPU */ + policy->cpu = cpumask_any(policy->cpus); + } up_write(&policy->rwsem);

- if (cpu != policy->cpu) - return 0; + /* Start governor again for active policy */ + if (!policy_is_inactive(policy)) { + if (has_target()) { + ret = __cpufreq_governor(policy, CPUFREQ_GOV_START); + if (!ret) + ret = __cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);

- if (cpus > 1) { - /* Nominate new CPU */ - down_write(&policy->rwsem); - policy->cpu = cpumask_any_but(policy->cpus, cpu); - up_write(&policy->rwsem); + if (ret) + pr_err("%s: Failed to start governor\n", __func__); + } } else if (cpufreq_driver->stop_cpu) { cpufreq_driver->stop_cpu(policy); }

- return 0; + return ret; }

static int __cpufreq_remove_dev_finish(struct device *dev, @@ -1446,33 +1453,16 @@ static int __cpufreq_remove_dev_finish(struct device *dev, { unsigned int cpu = dev->id; int ret; - struct cpufreq_policy *policy = cpufreq_cpu_get_raw(cpu); + struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);

if (!policy) { pr_debug("%s: No cpu_data found\n", __func__); return -EINVAL; }

- down_write(&policy->rwsem); - cpumask_clear_cpu(cpu, policy->cpus); - up_write(&policy->rwsem); - - /* Not the last cpu of policy, start governor again ? */ - if (!policy_is_inactive(policy)) { - if (!has_target()) - return 0; - - ret = __cpufreq_governor(policy, CPUFREQ_GOV_START); - if (!ret) - ret = __cpufreq_governor(policy, CPUFREQ_GOV_LIMITS); - - if (ret) { - pr_err("%s: Failed to start governor\n", __func__); - return ret; - } - + /* Only proceed for inactive policies */ + if (!policy_is_inactive(policy)) return 0; - }

/* If cpu is last user of policy, free policy */ if (has_target()) {