Revert "PM: runtime: Update device status before letting suppliers suspend"
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>
Fri, 19 Mar 2021 14:47:25 +0000 (15:47 +0100)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Fri, 19 Mar 2021 15:35:47 +0000 (16:35 +0100)
Revert commit 44cc89f76464 ("PM: runtime: Update device status
before letting suppliers suspend") that introduced a race condition
into __rpm_callback() which allowed a concurrent rpm_resume() to
run and resume the device prematurely after its status had been
changed to RPM_SUSPENDED by __rpm_callback().

Fixes: 44cc89f76464 ("PM: runtime: Update device status before letting suppliers suspend")
Link: https://lore.kernel.org/linux-pm/24dfb6fc-5d54-6ee2-9195-26428b7ecf8a@intel.com/
Reported-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: 4.10+ <stable@vger.kernel.org> # 4.10+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
drivers/base/power/runtime.c

index 18b8242..a46a7e3 100644 (file)
@@ -325,22 +325,22 @@ static void rpm_put_suppliers(struct device *dev)
 static int __rpm_callback(int (*cb)(struct device *), struct device *dev)
        __releases(&dev->power.lock) __acquires(&dev->power.lock)
 {
-       bool use_links = dev->power.links_count > 0;
-       bool get = false;
        int retval, idx;
-       bool put;
+       bool use_links = dev->power.links_count > 0;
 
        if (dev->power.irq_safe) {
                spin_unlock(&dev->power.lock);
-       } else if (!use_links) {
-               spin_unlock_irq(&dev->power.lock);
        } else {
-               get = dev->power.runtime_status == RPM_RESUMING;
-
                spin_unlock_irq(&dev->power.lock);
 
-               /* Resume suppliers if necessary. */
-               if (get) {
+               /*
+                * Resume suppliers if necessary.
+                *
+                * The device's runtime PM status cannot change until this
+                * routine returns, so it is safe to read the status outside of
+                * the lock.
+                */
+               if (use_links && dev->power.runtime_status == RPM_RESUMING) {
                        idx = device_links_read_lock();
 
                        retval = rpm_get_suppliers(dev);
@@ -355,36 +355,24 @@ static int __rpm_callback(int (*cb)(struct device *), struct device *dev)
 
        if (dev->power.irq_safe) {
                spin_lock(&dev->power.lock);
-               return retval;
-       }
-
-       spin_lock_irq(&dev->power.lock);
-
-       if (!use_links)
-               return retval;
-
-       /*
-        * If the device is suspending and the callback has returned success,
-        * drop the usage counters of the suppliers that have been reference
-        * counted on its resume.
-        *
-        * Do that if the resume fails too.
-        */
-       put = dev->power.runtime_status == RPM_SUSPENDING && !retval;
-       if (put)
-               __update_runtime_status(dev, RPM_SUSPENDED);
-       else
-               put = get && retval;
-
-       if (put) {
-               spin_unlock_irq(&dev->power.lock);
-
-               idx = device_links_read_lock();
+       } else {
+               /*
+                * If the device is suspending and the callback has returned
+                * success, drop the usage counters of the suppliers that have
+                * been reference counted on its resume.
+                *
+                * Do that if resume fails too.
+                */
+               if (use_links
+                   && ((dev->power.runtime_status == RPM_SUSPENDING && !retval)
+                   || (dev->power.runtime_status == RPM_RESUMING && retval))) {
+                       idx = device_links_read_lock();
 
-fail:
-               rpm_put_suppliers(dev);
+ fail:
+                       rpm_put_suppliers(dev);
 
-               device_links_read_unlock(idx);
+                       device_links_read_unlock(idx);
+               }
 
                spin_lock_irq(&dev->power.lock);
        }