Skip to contentSkip to conversation

regulator: lock supply in regulator enable

PATCH Applied Mattias Wallin 3 participants 7 messages Nov 4, 2010 – Nov 6, 2010 Voltage and current regulator framework

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 regulator: lock supply in regulator enable

This patch add locks around regulator supply enable.

---
The previous patch I sent solves a problem seen in our system. This patch does not solve a problem I have seen but I still think it should be there. Or at least some locking of the supply in regulator enable. What do you guys think?

---
 drivers/regulator/core.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
drivers/regulator/core.c +2 -0
index c625633..d10ad4a 100644
@@ -1270,7 +1270,9 @@ static int _regulator_enable(struct regulator_dev *rdev)
 
 	/* do we need to enable the supply regulator first */
 	if (rdev->supply) {
+		mutex_lock(&rdev->supply->mutex);
 		ret = _regulator_enable(rdev->supply);
+		mutex_unlock(&rdev->supply->mutex);
 		if (ret < 0) {
 			printk(KERN_ERR "%s: failed to enable %s: %d\n",
 			       __func__, rdev_get_name(rdev), ret);
Signature · 1 line
1.6.3.3
#1 lore raw
Quoted 8 lines · Mattias Wallin

This patch add locks around regulator supply enable.

---
The previous patch I sent solves a problem seen in our system. This patch does not solve a problem I have seen but I still think it should be there. Or at least some locking of the supply in regulator enable. What do you guys think?

This sounds like guesswork. What exactly is the problem in your system ?

Liam

Signature · 3 lines
Freelance Developer, SlimLogic Ltd
ASoC and Voltage Regulator Maintainer.
http://www.slimlogic.co.uk
#2 lore raw

As I wrote, the problem that I had is solved with my previous patch. Right now I have no visible problem but I still think there is locks missing and would like your opinion on it.

/Wallin

Quoted 13 lines · Liam Girdwood
Quoted 8 lines · Mattias Wallin

This patch add locks around regulator supply enable.

---
The previous patch I sent solves a problem seen in our system. This patch does not solve a problem I have seen but I still think it should be there. Or at least some locking of the supply in regulator enable. What do you guys think?

This sounds like guesswork. What exactly is the problem in your system ?

Liam

#3 lore raw
Quoted 4 lines · Mattias Wallin

As I wrote, the problem that I had is solved with my previous patch. Right now I have no visible problem but I still think there is locks missing and would like your opinion on it.

Quoted 15 lines

/Wallin

Quoted 12 lines · Liam Girdwood
Quoted 8 lines · Mattias Wallin

This patch add locks around regulator supply enable.

---
The previous patch I sent solves a problem seen in our system. This patch does not solve a problem I have seen but I still think it should be there. Or at least some locking of the supply in regulator enable. What do you guys think?

This sounds like guesswork. What exactly is the problem in your system ?

Sorry, got a busy schedule atm. Can you give us your reasoning behind why you think we need a lock here ?

Thanks

Liam

Signature · 3 lines
Freelance Developer, SlimLogic Ltd
ASoC and Voltage Regulator Maintainer.
http://www.slimlogic.co.uk
#4 lore raw
Quoted 3 lines

Sorry, got a busy schedule atm. Can you give us your reasoning behind why you think we need a lock here ?

Yes, when we enter regulator_enable() we take the lock only for that specific regulator rdev->mutex
and calls the locked function _regulator_enable().
This locked function then checks if we have a supply and recursively calls the locked _regulator_enable() again but with the supply rdev as argument. The supply rdev regulator is however not locked at this stage, only the "original" supplied regulator is locked.
So if we have two regulators with the same supply regulator trying to enable at approx. the same time they will both
enter the _regulator_enable without locks and we could get a race condition.
This would probably result in reference counting error and unbalanced regulator warnings.

In our system we make use the regulator hierarchy quite heavily.

Quoted 4 lines

Thanks

Liam

/Mattias Wallin

#5 lore raw
Quoted 10 lines · Mattias Wallin
Quoted 2 lines

Sorry, got a busy schedule atm. Can you give us your reasoning behind why you think we need a lock here ?

Yes, when we enter regulator_enable() we take the lock only for that specific regulator rdev->mutex
and calls the locked function _regulator_enable().
This locked function then checks if we have a supply and recursively calls the locked _regulator_enable() again but with the supply rdev as argument. The supply rdev regulator is however not locked at this stage, only the "original" supplied regulator is locked.
So if we have two regulators with the same supply regulator trying to enable at approx. the same time they will both
enter the _regulator_enable without locks and we could get a race condition.
This would probably result in reference counting error and unbalanced regulator warnings.

Quoted 2 lines

In our system we make use the regulator hierarchy quite heavily.

Quoted 1 line

Applied.

Thanks

Liam

Signature · 3 lines
Freelance Developer, SlimLogic Ltd
ASoC and Voltage Regulator Maintainer.
http://www.slimlogic.co.uk
Acked-by on v1 1/1
#6 lore raw
In reply to Mattias Wallin
Quoted 3 lines · Mattias Wallin

This patch add locks around regulator supply enable.

Quoted 4 lines

The previous patch I sent solves a problem seen in our system. This patch does not solve a problem I have seen but I still think it should be there. Or at least some locking of the supply in regulator enable. What do you guys think?

If we're locking the regulator when it's directly enabled we need to do the same when enabling it via other ways.