On 08/01/2012 12:30 PM, Fabio Baltieri wrote:
[...]
>>> +void can_led_init(struct net_device *netdev)
>>> +{
>>> +	struct can_priv *priv = netdev_priv(netdev);
>>> +	void *res;
>>> +
>>> +	res = devres_alloc(can_led_release, 0, GFP_KERNEL);
>>                                             ^
>> I'm not really sure if this is working. For example, pinctrl [1]
>> allocates a double pointer here. The res pointer here and in
>> can_led_release simply points to invalid memory. But as long as you
>> don't dereference it, it should work.
>>
>> [1] http://lxr.free-electrons.com/source/drivers/pinctrl/core.c#L862
> 
> Actually that's also used by libata-core (not in lxr yet) and dma-mapping:
> 
> http://lxr.free-electrons.com/source/drivers/base/dma-mapping.c#L193
> 
> actually, res should point at the end of some internal devres structure,
> and is only used as return value in this case.
> 
> Of course, in this case the release function can only use the struct
> device pointer.
> 
> I've run some fail test and ftraced the whole thing and it seems to work
> pretty good!  I'll send a v5 with the rename later.

Thanks for checking. I just noticed, the pinctrl usecase is a bit different.

Marc
-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |