Re: [PATCH v2 1/2] hwmon: ina2xx: convert driver to using regmap

[Marc Titinger <mtitinger@baylibre.com>]

On 28/10/2015 03:47, Guenter Roeck wrote:
> On Tue, Oct 27, 2015 at 10:51:07AM +0100, Marc Titinger wrote:
>> Any sysfs "show" read access from the client app will result in reading
>> all registers (8 with ina226). Depending on the host this can limit the
>> best achievable read rate.
>>
>> This changeset allows for individual register accesses through regmap.
>>
>> Tested with BeagleBone Black (Baylibre-ACME) and ina226.
>>
>> Signed-off-by: Marc Titinger <mtitinger@baylibre.com>
>> ---
>>
>>    v2:
>> 	- rename 'rv' to 'regval' for clarity
>> 	- fix missed smbus_xxx api change to regmap
>> 	- rename ina2xx_do_update to ina2xx_read_reg
>> 	- fix indentation
>>
>>   drivers/hwmon/ina2xx.c | 211 +++++++++++++++++++------------------------------
>>   1 file changed, 82 insertions(+), 129 deletions(-)
>>
>> diff --git a/drivers/hwmon/ina2xx.c b/drivers/hwmon/ina2xx.c
>> index 4d28150..5e7ada8 100644
>> --- a/drivers/hwmon/ina2xx.c
>> +++ b/drivers/hwmon/ina2xx.c
>
> [ ... ]
>>
>> -	mutex_lock(&data->update_lock);
>>   	data->rshunt = val;
>>   	status = ina2xx_calibrate(data);
>> -	mutex_unlock(&data->update_lock);
>
> I think this can result in a race conditon if multiple processes
> try to update the shunt resistor value at the same time in a
> multi-core system. There is no guarantee that the value programmed
> into the chip matches the value that is written into 'rshunt'.
> So I think we still need the mutex, unless you have a better
> idea how to avoid the race.
>
Right, I had that same afterthought when removing the mutex at that 
position. Fix incoming!

M.


> Thanks,
> Guenter
>