Merge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux...
authorLen Brown <len.brown@intel.com>
Tue, 9 Oct 2012 05:35:52 +0000 (01:35 -0400)
committerLen Brown <len.brown@intel.com>
Tue, 9 Oct 2012 05:35:52 +0000 (01:35 -0400)
Conflicts:
drivers/staging/omap-thermal/omap-thermal-common.
OMAP supplied dummy TC1 and TC2,
at the same time that the thermal tree removed them
from thermal_zone_device_register()

drivers/thermal/cpu_cooling.c b/drivers/thermal/cpu_cooling.c
propogate the upstream MAX_IDR_LEVEL re-name
to prevent a build failure

Previously-fixed-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Len Brown <len.brown@intel.com>
1  2 
drivers/acpi/thermal.c
drivers/hwmon/Kconfig
drivers/hwmon/Makefile
drivers/staging/omap-thermal/omap-thermal-common.c
drivers/thermal/cpu_cooling.c
drivers/thermal/thermal_sys.c

Simple merge
@@@ -334,19 -324,9 +334,9 @@@ config SENSORS_DA9052_AD
          This driver can also be built as module. If so, the module
          will be called da9052-hwmon.
  
- config SENSORS_EXYNOS4_TMU
-       tristate "Temperature sensor on Samsung EXYNOS4"
-       depends on ARCH_EXYNOS4
-       help
-         If you say yes here you get support for TMU (Thermal Management
-         Unit) on SAMSUNG EXYNOS4 series of SoC.
-         This driver can also be built as a module. If so, the module
-         will be called exynos4-tmu.
  config SENSORS_I5K_AMB
        tristate "FB-DIMM AMB temperature sensor on Intel 5000 series chipsets"
 -      depends on PCI && EXPERIMENTAL
 +      depends on PCI
        help
          If you say yes here you get support for FB-DIMM AMB temperature
          monitoring chips on systems with the Intel 5000 series chipset.
Simple merge
@@@ -268,9 -248,7 +270,8 @@@ int omap_thermal_expose_sensor(struct o
        /* Create thermal zone */
        data->omap_thermal = thermal_zone_device_register(domain,
                                OMAP_TRIP_NUMBER, 0, data, &omap_thermal_ops,
-                               1, 2, /*TODO: remove this when FW allows */
 -                              0, FAST_TEMP_MONITORING_RATE);
 +                              FAST_TEMP_MONITORING_RATE,
 +                              FAST_TEMP_MONITORING_RATE);
        if (IS_ERR_OR_NULL(data->omap_thermal)) {
                dev_err(bg_ptr->dev, "thermal zone device is NULL\n");
                return PTR_ERR(data->omap_thermal);
index 0000000,9050c1b..cc1c930
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,449 +1,449 @@@
 -      *id = *id & MAX_ID_MASK;
+ /*
+  *  linux/drivers/thermal/cpu_cooling.c
+  *
+  *  Copyright (C) 2012        Samsung Electronics Co., Ltd(http://www.samsung.com)
+  *  Copyright (C) 2012  Amit Daniel <amit.kachhap@linaro.org>
+  *
+  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  *  This program is free software; you can redistribute it and/or modify
+  *  it under the terms of the GNU General Public License as published by
+  *  the Free Software Foundation; version 2 of the License.
+  *
+  *  This program is distributed in the hope that it will be useful, but
+  *  WITHOUT ANY WARRANTY; without even the implied warranty of
+  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+  *  General Public License for more details.
+  *
+  *  You should have received a copy of the GNU General Public License along
+  *  with this program; if not, write to the Free Software Foundation, Inc.,
+  *  59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
+  *
+  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  */
+ #include <linux/kernel.h>
+ #include <linux/module.h>
+ #include <linux/thermal.h>
+ #include <linux/platform_device.h>
+ #include <linux/cpufreq.h>
+ #include <linux/err.h>
+ #include <linux/slab.h>
+ #include <linux/cpu.h>
+ #include <linux/cpu_cooling.h>
+ /**
+  * struct cpufreq_cooling_device
+  * @id: unique integer value corresponding to each cpufreq_cooling_device
+  *    registered.
+  * @cool_dev: thermal_cooling_device pointer to keep track of the the
+  *    egistered cooling device.
+  * @cpufreq_state: integer value representing the current state of cpufreq
+  *    cooling devices.
+  * @cpufreq_val: integer value representing the absolute value of the clipped
+  *    frequency.
+  * @allowed_cpus: all the cpus involved for this cpufreq_cooling_device.
+  * @node: list_head to link all cpufreq_cooling_device together.
+  *
+  * This structure is required for keeping information of each
+  * cpufreq_cooling_device registered as a list whose head is represented by
+  * cooling_cpufreq_list. In order to prevent corruption of this list a
+  * mutex lock cooling_cpufreq_lock is used.
+  */
+ struct cpufreq_cooling_device {
+       int id;
+       struct thermal_cooling_device *cool_dev;
+       unsigned int cpufreq_state;
+       unsigned int cpufreq_val;
+       struct cpumask allowed_cpus;
+       struct list_head node;
+ };
+ static LIST_HEAD(cooling_cpufreq_list);
+ static DEFINE_IDR(cpufreq_idr);
+ static struct mutex cooling_cpufreq_lock;
+ /* notify_table passes value to the CPUFREQ_ADJUST callback function. */
+ #define NOTIFY_INVALID NULL
+ struct cpufreq_cooling_device *notify_device;
+ /**
+  * get_idr - function to get a unique id.
+  * @idr: struct idr * handle used to create a id.
+  * @id: int * value generated by this function.
+  */
+ static int get_idr(struct idr *idr, int *id)
+ {
+       int err;
+ again:
+       if (unlikely(idr_pre_get(idr, GFP_KERNEL) == 0))
+               return -ENOMEM;
+       mutex_lock(&cooling_cpufreq_lock);
+       err = idr_get_new(idr, NULL, id);
+       mutex_unlock(&cooling_cpufreq_lock);
+       if (unlikely(err == -EAGAIN))
+               goto again;
+       else if (unlikely(err))
+               return err;
++      *id = *id & MAX_IDR_MASK;
+       return 0;
+ }
+ /**
+  * release_idr - function to free the unique id.
+  * @idr: struct idr * handle used for creating the id.
+  * @id: int value representing the unique id.
+  */
+ static void release_idr(struct idr *idr, int id)
+ {
+       mutex_lock(&cooling_cpufreq_lock);
+       idr_remove(idr, id);
+       mutex_unlock(&cooling_cpufreq_lock);
+ }
+ /* Below code defines functions to be used for cpufreq as cooling device */
+ /**
+  * is_cpufreq_valid - function to check if a cpu has frequency transition policy.
+  * @cpu: cpu for which check is needed.
+  */
+ static int is_cpufreq_valid(int cpu)
+ {
+       struct cpufreq_policy policy;
+       return !cpufreq_get_policy(&policy, cpu);
+ }
+ /**
+  * get_cpu_frequency - get the absolute value of frequency from level.
+  * @cpu: cpu for which frequency is fetched.
+  * @level: level of frequency of the CPU
+  *    e.g level=1 --> 1st MAX FREQ, LEVEL=2 ---> 2nd MAX FREQ, .... etc
+  */
+ static unsigned int get_cpu_frequency(unsigned int cpu, unsigned long level)
+ {
+       int ret = 0, i = 0;
+       unsigned long level_index;
+       bool descend = false;
+       struct cpufreq_frequency_table *table =
+                                       cpufreq_frequency_get_table(cpu);
+       if (!table)
+               return ret;
+       while (table[i].frequency != CPUFREQ_TABLE_END) {
+               if (table[i].frequency == CPUFREQ_ENTRY_INVALID)
+                       continue;
+               /*check if table in ascending or descending order*/
+               if ((table[i + 1].frequency != CPUFREQ_TABLE_END) &&
+                       (table[i + 1].frequency < table[i].frequency)
+                       && !descend) {
+                       descend = true;
+               }
+               /*return if level matched and table in descending order*/
+               if (descend && i == level)
+                       return table[i].frequency;
+               i++;
+       }
+       i--;
+       if (level > i || descend)
+               return ret;
+       level_index = i - level;
+       /*Scan the table in reverse order and match the level*/
+       while (i >= 0) {
+               if (table[i].frequency == CPUFREQ_ENTRY_INVALID)
+                       continue;
+               /*return if level matched*/
+               if (i == level_index)
+                       return table[i].frequency;
+               i--;
+       }
+       return ret;
+ }
+ /**
+  * cpufreq_apply_cooling - function to apply frequency clipping.
+  * @cpufreq_device: cpufreq_cooling_device pointer containing frequency
+  *    clipping data.
+  * @cooling_state: value of the cooling state.
+  */
+ static int cpufreq_apply_cooling(struct cpufreq_cooling_device *cpufreq_device,
+                               unsigned long cooling_state)
+ {
+       unsigned int cpuid, clip_freq;
+       struct cpumask *maskPtr = &cpufreq_device->allowed_cpus;
+       unsigned int cpu = cpumask_any(maskPtr);
+       /* Check if the old cooling action is same as new cooling action */
+       if (cpufreq_device->cpufreq_state == cooling_state)
+               return 0;
+       clip_freq = get_cpu_frequency(cpu, cooling_state);
+       if (!clip_freq)
+               return -EINVAL;
+       cpufreq_device->cpufreq_state = cooling_state;
+       cpufreq_device->cpufreq_val = clip_freq;
+       notify_device = cpufreq_device;
+       for_each_cpu(cpuid, maskPtr) {
+               if (is_cpufreq_valid(cpuid))
+                       cpufreq_update_policy(cpuid);
+       }
+       notify_device = NOTIFY_INVALID;
+       return 0;
+ }
+ /**
+  * cpufreq_thermal_notifier - notifier callback for cpufreq policy change.
+  * @nb:       struct notifier_block * with callback info.
+  * @event: value showing cpufreq event for which this function invoked.
+  * @data: callback-specific data
+  */
+ static int cpufreq_thermal_notifier(struct notifier_block *nb,
+                                       unsigned long event, void *data)
+ {
+       struct cpufreq_policy *policy = data;
+       unsigned long max_freq = 0;
+       if (event != CPUFREQ_ADJUST || notify_device == NOTIFY_INVALID)
+               return 0;
+       if (cpumask_test_cpu(policy->cpu, &notify_device->allowed_cpus))
+               max_freq = notify_device->cpufreq_val;
+       /* Never exceed user_policy.max*/
+       if (max_freq > policy->user_policy.max)
+               max_freq = policy->user_policy.max;
+       if (policy->max != max_freq)
+               cpufreq_verify_within_limits(policy, 0, max_freq);
+       return 0;
+ }
+ /*
+  * cpufreq cooling device callback functions are defined below
+  */
+ /**
+  * cpufreq_get_max_state - callback function to get the max cooling state.
+  * @cdev: thermal cooling device pointer.
+  * @state: fill this variable with the max cooling state.
+  */
+ static int cpufreq_get_max_state(struct thermal_cooling_device *cdev,
+                                unsigned long *state)
+ {
+       int ret = -EINVAL, i = 0;
+       struct cpufreq_cooling_device *cpufreq_device;
+       struct cpumask *maskPtr;
+       unsigned int cpu;
+       struct cpufreq_frequency_table *table;
+       mutex_lock(&cooling_cpufreq_lock);
+       list_for_each_entry(cpufreq_device, &cooling_cpufreq_list, node) {
+               if (cpufreq_device && cpufreq_device->cool_dev == cdev)
+                       break;
+       }
+       if (cpufreq_device == NULL)
+               goto return_get_max_state;
+       maskPtr = &cpufreq_device->allowed_cpus;
+       cpu = cpumask_any(maskPtr);
+       table = cpufreq_frequency_get_table(cpu);
+       if (!table) {
+               *state = 0;
+               ret = 0;
+               goto return_get_max_state;
+       }
+       while (table[i].frequency != CPUFREQ_TABLE_END) {
+               if (table[i].frequency == CPUFREQ_ENTRY_INVALID)
+                       continue;
+               i++;
+       }
+       if (i > 0) {
+               *state = --i;
+               ret = 0;
+       }
+ return_get_max_state:
+       mutex_unlock(&cooling_cpufreq_lock);
+       return ret;
+ }
+ /**
+  * cpufreq_get_cur_state - callback function to get the current cooling state.
+  * @cdev: thermal cooling device pointer.
+  * @state: fill this variable with the current cooling state.
+  */
+ static int cpufreq_get_cur_state(struct thermal_cooling_device *cdev,
+                                unsigned long *state)
+ {
+       int ret = -EINVAL;
+       struct cpufreq_cooling_device *cpufreq_device;
+       mutex_lock(&cooling_cpufreq_lock);
+       list_for_each_entry(cpufreq_device, &cooling_cpufreq_list, node) {
+               if (cpufreq_device && cpufreq_device->cool_dev == cdev) {
+                       *state = cpufreq_device->cpufreq_state;
+                       ret = 0;
+                       break;
+               }
+       }
+       mutex_unlock(&cooling_cpufreq_lock);
+       return ret;
+ }
+ /**
+  * cpufreq_set_cur_state - callback function to set the current cooling state.
+  * @cdev: thermal cooling device pointer.
+  * @state: set this variable to the current cooling state.
+  */
+ static int cpufreq_set_cur_state(struct thermal_cooling_device *cdev,
+                                unsigned long state)
+ {
+       int ret = -EINVAL;
+       struct cpufreq_cooling_device *cpufreq_device;
+       mutex_lock(&cooling_cpufreq_lock);
+       list_for_each_entry(cpufreq_device, &cooling_cpufreq_list, node) {
+               if (cpufreq_device && cpufreq_device->cool_dev == cdev) {
+                       ret = 0;
+                       break;
+               }
+       }
+       if (!ret)
+               ret = cpufreq_apply_cooling(cpufreq_device, state);
+       mutex_unlock(&cooling_cpufreq_lock);
+       return ret;
+ }
+ /* Bind cpufreq callbacks to thermal cooling device ops */
+ static struct thermal_cooling_device_ops const cpufreq_cooling_ops = {
+       .get_max_state = cpufreq_get_max_state,
+       .get_cur_state = cpufreq_get_cur_state,
+       .set_cur_state = cpufreq_set_cur_state,
+ };
+ /* Notifier for cpufreq policy change */
+ static struct notifier_block thermal_cpufreq_notifier_block = {
+       .notifier_call = cpufreq_thermal_notifier,
+ };
+ /**
+  * cpufreq_cooling_register - function to create cpufreq cooling device.
+  * @clip_cpus: cpumask of cpus where the frequency constraints will happen.
+  */
+ struct thermal_cooling_device *cpufreq_cooling_register(
+       struct cpumask *clip_cpus)
+ {
+       struct thermal_cooling_device *cool_dev;
+       struct cpufreq_cooling_device *cpufreq_dev = NULL;
+       unsigned int cpufreq_dev_count = 0, min = 0, max = 0;
+       char dev_name[THERMAL_NAME_LENGTH];
+       int ret = 0, i;
+       struct cpufreq_policy policy;
+       list_for_each_entry(cpufreq_dev, &cooling_cpufreq_list, node)
+               cpufreq_dev_count++;
+       /*Verify that all the clip cpus have same freq_min, freq_max limit*/
+       for_each_cpu(i, clip_cpus) {
+               /*continue if cpufreq policy not found and not return error*/
+               if (!cpufreq_get_policy(&policy, i))
+                       continue;
+               if (min == 0 && max == 0) {
+                       min = policy.cpuinfo.min_freq;
+                       max = policy.cpuinfo.max_freq;
+               } else {
+                       if (min != policy.cpuinfo.min_freq ||
+                               max != policy.cpuinfo.max_freq)
+                               return ERR_PTR(-EINVAL);
+ }
+       }
+       cpufreq_dev = kzalloc(sizeof(struct cpufreq_cooling_device),
+                       GFP_KERNEL);
+       if (!cpufreq_dev)
+               return ERR_PTR(-ENOMEM);
+       cpumask_copy(&cpufreq_dev->allowed_cpus, clip_cpus);
+       if (cpufreq_dev_count == 0)
+               mutex_init(&cooling_cpufreq_lock);
+       ret = get_idr(&cpufreq_idr, &cpufreq_dev->id);
+       if (ret) {
+               kfree(cpufreq_dev);
+               return ERR_PTR(-EINVAL);
+       }
+       sprintf(dev_name, "thermal-cpufreq-%d", cpufreq_dev->id);
+       cool_dev = thermal_cooling_device_register(dev_name, cpufreq_dev,
+                                               &cpufreq_cooling_ops);
+       if (!cool_dev) {
+               release_idr(&cpufreq_idr, cpufreq_dev->id);
+               kfree(cpufreq_dev);
+               return ERR_PTR(-EINVAL);
+       }
+       cpufreq_dev->cool_dev = cool_dev;
+       cpufreq_dev->cpufreq_state = 0;
+       mutex_lock(&cooling_cpufreq_lock);
+       list_add_tail(&cpufreq_dev->node, &cooling_cpufreq_list);
+       /* Register the notifier for first cpufreq cooling device */
+       if (cpufreq_dev_count == 0)
+               cpufreq_register_notifier(&thermal_cpufreq_notifier_block,
+                                               CPUFREQ_POLICY_NOTIFIER);
+       mutex_unlock(&cooling_cpufreq_lock);
+       return cool_dev;
+ }
+ EXPORT_SYMBOL(cpufreq_cooling_register);
+ /**
+  * cpufreq_cooling_unregister - function to remove cpufreq cooling device.
+  * @cdev: thermal cooling device pointer.
+  */
+ void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev)
+ {
+       struct cpufreq_cooling_device *cpufreq_dev = NULL;
+       unsigned int cpufreq_dev_count = 0;
+       mutex_lock(&cooling_cpufreq_lock);
+       list_for_each_entry(cpufreq_dev, &cooling_cpufreq_list, node) {
+               if (cpufreq_dev && cpufreq_dev->cool_dev == cdev)
+                       break;
+               cpufreq_dev_count++;
+       }
+       if (!cpufreq_dev || cpufreq_dev->cool_dev != cdev) {
+               mutex_unlock(&cooling_cpufreq_lock);
+               return;
+       }
+       list_del(&cpufreq_dev->node);
+       /* Unregister the notifier for the last cpufreq cooling device */
+       if (cpufreq_dev_count == 1) {
+               cpufreq_unregister_notifier(&thermal_cpufreq_notifier_block,
+                                       CPUFREQ_POLICY_NOTIFIER);
+       }
+       mutex_unlock(&cooling_cpufreq_lock);
+       thermal_cooling_device_unregister(cpufreq_dev->cool_dev);
+       release_idr(&cpufreq_idr, cpufreq_dev->id);
+       if (cpufreq_dev_count == 1)
+               mutex_destroy(&cooling_cpufreq_lock);
+       kfree(cpufreq_dev);
+ }
+ EXPORT_SYMBOL(cpufreq_cooling_unregister);
@@@ -694,84 -703,19 +703,16 @@@ thermal_remove_hwmon_sysfs(struct therm
  static void thermal_zone_device_set_polling(struct thermal_zone_device *tz,
                                            int delay)
  {
 -      cancel_delayed_work(&(tz->poll_queue));
 -
 -      if (!delay)
 -              return;
 -
        if (delay > 1000)
 -              queue_delayed_work(system_freezable_wq, &(tz->poll_queue),
 -                                    round_jiffies(msecs_to_jiffies(delay)));
 +              mod_delayed_work(system_freezable_wq, &tz->poll_queue,
 +                               round_jiffies(msecs_to_jiffies(delay)));
 +      else if (delay)
 +              mod_delayed_work(system_freezable_wq, &tz->poll_queue,
 +                               msecs_to_jiffies(delay));
        else
 -              queue_delayed_work(system_freezable_wq, &(tz->poll_queue),
 -                                    msecs_to_jiffies(delay));
 +              cancel_delayed_work(&tz->poll_queue);
  }
  
- static void thermal_zone_device_passive(struct thermal_zone_device *tz,
-                                       int temp, int trip_temp, int trip)
- {
-       int trend = 0;
-       struct thermal_cooling_device_instance *instance;
-       struct thermal_cooling_device *cdev;
-       long state, max_state;
-       /*
-        * Above Trip?
-        * -----------
-        * Calculate the thermal trend (using the passive cooling equation)
-        * and modify the performance limit for all passive cooling devices
-        * accordingly.  Note that we assume symmetry.
-        */
-       if (temp >= trip_temp) {
-               tz->passive = true;
-               trend = (tz->tc1 * (temp - tz->last_temperature)) +
-                       (tz->tc2 * (temp - trip_temp));
-               /* Heating up? */
-               if (trend > 0) {
-                       list_for_each_entry(instance, &tz->cooling_devices,
-                                           node) {
-                               if (instance->trip != trip)
-                                       continue;
-                               cdev = instance->cdev;
-                               cdev->ops->get_cur_state(cdev, &state);
-                               cdev->ops->get_max_state(cdev, &max_state);
-                               if (state++ < max_state)
-                                       cdev->ops->set_cur_state(cdev, state);
-                       }
-               } else if (trend < 0) { /* Cooling off? */
-                       list_for_each_entry(instance, &tz->cooling_devices,
-                                           node) {
-                               if (instance->trip != trip)
-                                       continue;
-                               cdev = instance->cdev;
-                               cdev->ops->get_cur_state(cdev, &state);
-                               cdev->ops->get_max_state(cdev, &max_state);
-                               if (state > 0)
-                                       cdev->ops->set_cur_state(cdev, --state);
-                       }
-               }
-               return;
-       }
-       /*
-        * Below Trip?
-        * -----------
-        * Implement passive cooling hysteresis to slowly increase performance
-        * and avoid thrashing around the passive trip point.  Note that we
-        * assume symmetry.
-        */
-       list_for_each_entry(instance, &tz->cooling_devices, node) {
-               if (instance->trip != trip)
-                       continue;
-               cdev = instance->cdev;
-               cdev->ops->get_cur_state(cdev, &state);
-               cdev->ops->get_max_state(cdev, &max_state);
-               if (state > 0)
-                       cdev->ops->set_cur_state(cdev, --state);
-               if (state == 0)
-                       tz->passive = false;
-       }
- }
  static void thermal_zone_device_check(struct work_struct *work)
  {
        struct thermal_zone_device *tz = container_of(work, struct