Re: [PATCH v4] can: Convert to runtime_pm

[Marc Kleine-Budde <mkl@pengutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 1290 bytes --]

On 01/07/2015 04:58 PM, Sören Brinkmann wrote:
>> I think you have to convert the _remove() function, too. Have a look at
>> the gpio-zynq.c driver:
>>
>>> static int zynq_gpio_remove(struct platform_device *pdev)
>>> {
>>> 	struct zynq_gpio *gpio = platform_get_drvdata(pdev);
>>>
>>> 	pm_runtime_get_sync(&pdev->dev);
>>
>> However I don't understand why the get_sync() is here. Maybe Sören can help?
> 
> IIRC, the concern was that the remove function may be called while the device is
> runtime suspended. Hence the remove function needs to resume the device since the
> remove function may access the HW.

What about the corresponding runtime_put()? Would some counter be
unbalanced upon device removal?

Without having tested it, unloading and loading, i.e. reloading a CAN
driver is easier than reloading the gpio driver on an average embedded
system. Kedareswara please test your driver with something like:
modprobe; ifconfig up; cansend; ifconfig down; rmmod in a loop.

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   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]