Re: [PATCH] regulator: core: avoid regulator_resolve_supply() race condition

[David Collins <collinsd@codeaurora.org>]

Hello Mark,

On 1/21/21 12:30 PM, Marek Szyprowski wrote:
> Hi Mark,
> 
> On 21.01.2021 16:44, Mark Brown wrote:
>> On Thu, Jan 21, 2021 at 10:41:59AM +0100, Marek Szyprowski wrote:
>>> On 18.01.2021 21:49, Mark Brown wrote:
>>>> Does this help (completely untested):
>>> Sadly nope. I get same warning:
>> Try this instead:
>>
>> diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
>> index 3ae5ccd9277d..31503776dbd7 100644
>> --- a/drivers/regulator/core.c
>> +++ b/drivers/regulator/core.c
>> @@ -1823,17 +1823,6 @@ static int regulator_resolve_supply(struct regulator_dev *rdev)
>>    	if (rdev->supply)
>>    		return 0;
>>    
>> -	/*
>> -	 * Recheck rdev->supply with rdev->mutex lock held to avoid a race
>> -	 * between rdev->supply null check and setting rdev->supply in
>> -	 * set_supply() from concurrent tasks.
>> -	 */
>> -	regulator_lock(rdev);
>> -
>> -	/* Supply just resolved by a concurrent task? */
>> -	if (rdev->supply)
>> -		goto out;
>> -
>>    	r = regulator_dev_lookup(dev, rdev->supply_name);
>>    	if (IS_ERR(r)) {
>>    		ret = PTR_ERR(r);
>> @@ -1885,12 +1874,29 @@ static int regulator_resolve_supply(struct regulator_dev *rdev)
>>    		goto out;
>>    	}
>>    
>> +	/*
>> +	 * Recheck rdev->supply with rdev->mutex lock held to avoid a race
>> +	 * between rdev->supply null check and setting rdev->supply in
>> +	 * set_supply() from concurrent tasks.
>> +	 */
>> +	regulator_lock(rdev);
>> +
>> +	/* Supply just resolved by a concurrent task? */
>> +	if (rdev->supply) {
>> +		regulator_unlock(rdev);
>> +		put_device(&r->dev);
>> +		return ret;
>> +	}
>> +
>>    	ret = set_supply(rdev, r);
>>    	if (ret < 0) {
>> +		regulator_unlock(rdev);
>>    		put_device(&r->dev);
>> -		goto out;
>> +		return ret;
>>    	}
>>    
>> +	regulator_unlock(rdev);
>> +
>>    	/*
>>    	 * In set_machine_constraints() we may have turned this regulator on
>>    	 * but we couldn't propagate to the supply if it hadn't been resolved
>> @@ -1901,12 +1907,11 @@ static int regulator_resolve_supply(struct regulator_dev *rdev)
>>    		if (ret < 0) {
>>    			_regulator_put(rdev->supply);
>>    			rdev->supply = NULL;
>> -			goto out;
>> +			goto out_rdev_lock;
> 
> drivers/regulator/core.c:1910:4: error: label ‘out_rdev_lock’ used but
> not defined
> 
>>    		}
>>    	}
>>    
>>    out:
>> -	regulator_unlock(rdev);
>>    	return ret;
>>    }
>>    
> 
> It looks that it finally fixes the locking issue, with the above goto
> removed completely to fix build. Feel free to add:
> 
> Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
> 
> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>

Thank you for making this fix.  I'm sorry that I missed the potential 
deadlock issue resulting from the regulator_enable() call inside 
regulator_resolve_supply() with rdev->mutex locked.  Your fix avoids 
deadlock while still ensuring that the there isn't a set supply race 
condition.

Take care,
David

-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project