Re: [PATCH 02/15] thermal: sysfs: lock tz while on access to mode properties

On 二, 2016-06-07 at 14:52 +0530, Keerthy wrote:

On Tuesday 07 June 2016 02:38 PM, Keerthy wrote:

quoted

Hi Eduardo,

On Tuesday 31 May 2016 12:01 PM, Eduardo Valentin wrote:

quoted

Serialized calls to tz.ops in user facing
sysfs handler mode_show()  and mode_store().

This seems to be causing a deadlock at boot time during the ending
stages of boot:

http://pastebin.ubuntu.com/17085291/

It took a while to git bisect on linux-next.

Seems like you introduced new locking at the sysfs layer which
causes
this deadlock as the underlying code again tries to acquire the
same
tz->lock.

I confirm reverting this patch helps me boot on linux-next. This
patch 
can be dropped as the lower layer functions are already acquiring tz-

quoted

lock.

Thanks Vignesh for reporting the deadlock.

the root cause of the deadlock is that some platform driver invokes
thermal_zone_device_update() in .set_mode(), after mode changed.

If we wants to lock tz->ops->set_mode(), we should cleanup all the
platform thermal drivers, and deliver a thermal_zone_device_update()
in mode_store(), at the same time.

thanks,
rui

Regards,
Keerthy

quoted

Regards,
Keerthy

quoted

Cc: Zhang Rui <rui.zhang@intel.com>
Cc: linux-pm@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Eduardo Valentin <edubezval@gmail.com>
---
  drivers/thermal/thermal_sysfs.c | 13 ++++++++++---
  1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/drivers/thermal/thermal_sysfs.c

b/drivers/thermal/thermal_sysfs.c
index ee983ca..1db2406 100644

--- a/drivers/thermal/thermal_sysfs.c
+++ b/drivers/thermal/thermal_sysfs.c

@@ -62,7 +62,9 @@ mode_show(struct device *dev, struct

device_attribute *attr, char *buf)
      if (!tz->ops->get_mode)
          return -EPERM;

+    mutex_lock(&tz->lock);
      result = tz->ops->get_mode(tz, &mode);
+    mutex_unlock(&tz->lock);
      if (result)
          return result;

@@ -75,17 +77,22 @@ mode_store(struct device *dev, struct

device_attribute *attr,
         const char *buf, size_t count)
  {
      struct thermal_zone_device *tz = to_thermal_zone(dev);
+    enum thermal_device_mode mode = THERMAL_DEVICE_DISABLED;
      int result;

      if (!tz->ops->set_mode)
          return -EPERM;

      if (!strncmp(buf, "enabled", sizeof("enabled") - 1))
-        result = tz->ops->set_mode(tz, THERMAL_DEVICE_ENABLED);
+        mode = THERMAL_DEVICE_ENABLED;
      else if (!strncmp(buf, "disabled", sizeof("disabled") - 1))
-        result = tz->ops->set_mode(tz, THERMAL_DEVICE_DISABLED);
+        mode = THERMAL_DEVICE_DISABLED;
      else
-        result = -EINVAL;
+        return -EINVAL;
+
+    mutex_lock(&tz->lock);
+    result = tz->ops->set_mode(tz, mode);
+    mutex_unlock(&tz->lock);

      if (result)
          return result;

--
To unsubscribe from this list: send the line "unsubscribe linux-pm" 
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html