regulator: core: fix a possible race in disable_work handling

A race condition between queueing and processing the disable_work instances results in having a work instance in the queue and the deferred_disables variable of regulator device structure having a value '0'. If no new regulator_disable_deferred() call later from clients, the deferred_disables variable value remains '0' and hits BUG() in regulator_disable_work() when the queued instance scheduled for processing the work.

The race occurs as below:

	Core-0					     Core-1
	.....	       /* deferred_disables = 2 */   .....
	.....	       /* disable_work is queued */  .....
	.....					     .....

regulator_disable_deferred: regulator_disable_work:
mutex_lock(&rdev->mutex); .....
rdev->deferred_disables++; mutex_lock(&rdev->mutex);
mutex_unlock(&rdev->mutex); .....

   queue_delayed_work(...)		    count =rdev->deferred_disables;
	.....				    rdev->deferred_disables = 0;
	.....					     .....
	.....				    mutex_unlock(&rdev->mutex);
	.....					     .....
	.....				    return;
	.....					     .....
	/* No new regulator_disable_deferred() calls from clients */
	/* The newly queued instance is scheduled for processing */
	.....					     .....
regulator_disable_work:
	.....

mutex_lock(&rdev->mutex); BUG_ON(!rdev->deferred_disables); /* deferred_disables = 0 */

The race is fixed by removing the work instance that is queued while processing the previous queued instance. Cancel the newly queued instance from disable_work() handler just after reset the deferred_disables variable to value '0'. Also move the work queueing step before mutex_unlock in regulator_disable_deferred().

Also use mod_delayed_work() in the pace of queue_delayed_work() as queue_delayed_work() always uses the delay requested in the first call when multiple consumers call regulator_disable_deferred() close in time and does not guarantee the semantics of regulator_disable_deferred().

---
 drivers/regulator/core.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

index e567fa5..9f4d484 100644
@@ -2396,6 +2396,14 @@ static void regulator_disable_work(struct work_struct *work)
 	count = rdev->deferred_disables;
 	rdev->deferred_disables = 0;
 
+	/*
+	 * Workqueue functions queue the new work instance while the previous
+	 * work instance is being processed. Cancel the queued work instance
+	 * as the work instance under processing does the job of the queued
+	 * work instance.
+	 */
+	cancel_delayed_work(&rdev->disable_work);
+
 	for (i = 0; i < count; i++) {
 		ret = _regulator_disable(rdev);
 		if (ret != 0)

@@ -2439,10 +2447,10 @@ int regulator_disable_deferred(struct regulator *regulator, int ms)
 
 	mutex_lock(&rdev->mutex);
 	rdev->deferred_disables++;
+	mod_delayed_work(system_power_efficient_wq, &rdev->disable_work,
+			 msecs_to_jiffies(ms));
 	mutex_unlock(&rdev->mutex);
 
-	queue_delayed_work(system_power_efficient_wq, &rdev->disable_work,
-			   msecs_to_jiffies(ms));
 	return 0;
 }
 EXPORT_SYMBOL_GPL(regulator_disable_deferred);

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

This shows two things simultaneously holding rdev->mutex...

Fixed this in the new patch version(V2).
https://patchwork.kernel.org/patch/9836529/

The patch

regulator: core: fix a possible race in disable_work handling

has been applied to the regulator tree at

git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator.git

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark

From: Tirupathi Reddy <[email protected]>
Date: Wed, 12 Jul 2017 17:08:13 +0530
Subject: [PATCH] regulator: core: fix a possible race in disable_work handling

A race condition between queueing and processing the disable_work instances results in having a work instance in the queue and the deferred_disables variable of regulator device structure having a value '0'. If no new regulator_disable_deferred() call later from clients, the deferred_disables variable value remains '0' and hits BUG() in regulator_disable_work() when the queued instance scheduled for processing the work.

The race occurs as below:

	Core-0					     Core-1
	.....	       /* deferred_disables = 2 */   .....
	.....	       /* disable_work is queued */  .....
	.....					     .....

regulator_disable_deferred: regulator_disable_work:
mutex_lock(&rdev->mutex); .....
rdev->deferred_disables++; .....
mutex_unlock(&rdev->mutex); .....

   queue_delayed_work(...)		    mutex_lock(&rdev->mutex);
	.....				    count =rdev->deferred_disables;
	.....				    rdev->deferred_disables = 0;
	.....					     .....
	.....				    mutex_unlock(&rdev->mutex);
	.....					     .....
	.....				    return;
	.....					     .....
	/* No new regulator_disable_deferred() calls from clients */
	/* The newly queued instance is scheduled for processing */
	.....					     .....
regulator_disable_work:
	.....

mutex_lock(&rdev->mutex); BUG_ON(!rdev->deferred_disables); /* deferred_disables = 0 */

The race is fixed by removing the work instance that is queued while processing the previous queued instance. Cancel the newly queued instance from disable_work() handler just after reset the deferred_disables variable to value '0'. Also move the work queueing step before mutex_unlock in regulator_disable_deferred().

Also use mod_delayed_work() in the pace of queue_delayed_work() as queue_delayed_work() always uses the delay requested in the first call when multiple consumers call regulator_disable_deferred() close in time and does not guarantee the semantics of regulator_disable_deferred().

---
 drivers/regulator/core.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

index e567fa54980b..9f4d484eb25d 100644
@@ -2396,6 +2396,14 @@ static void regulator_disable_work(struct work_struct *work)
 	count = rdev->deferred_disables;
 	rdev->deferred_disables = 0;
 
+	/*
+	 * Workqueue functions queue the new work instance while the previous
+	 * work instance is being processed. Cancel the queued work instance
+	 * as the work instance under processing does the job of the queued
+	 * work instance.
+	 */
+	cancel_delayed_work(&rdev->disable_work);
+
 	for (i = 0; i < count; i++) {
 		ret = _regulator_disable(rdev);
 		if (ret != 0)

@@ -2439,10 +2447,10 @@ int regulator_disable_deferred(struct regulator *regulator, int ms)
 
 	mutex_lock(&rdev->mutex);
 	rdev->deferred_disables++;
+	mod_delayed_work(system_power_efficient_wq, &rdev->disable_work,
+			 msecs_to_jiffies(ms));
 	mutex_unlock(&rdev->mutex);
 
-	queue_delayed_work(system_power_efficient_wq, &rdev->disable_work,
-			   msecs_to_jiffies(ms));
 	return 0;
 }
 EXPORT_SYMBOL_GPL(regulator_disable_deferred);

2.13.2