Re: [PATCH v3 2/3] i2c: iproc: Add Broadcom iProc I2C Driver

[Ray Jui]

On 1/13/2015 11:51 PM, Uwe Kleine-König wrote:
> Hello,
> 
> On Tue, Jan 13, 2015 at 06:14:17PM -0800, Ray Jui wrote:
>>>> +	irq = platform_get_irq(pdev, 0);
>>>> +	if (irq < 0) {
>>> irq == 0 should be handled as error, too.
>>>
>> Ah. I thought zero is a valid global interrupt number, and I see other
>> drivers checking against < 0 as well. Is my understanding incorrect?
> These are wrong, too. 0 should never be a valid interrupt number. There
> are some exceptions but mostly for historic reasons. The right handling
> is used for example in drivers/i2c/busses/i2c-efm32.c.
> 
Okay. Will check against <= 0. Thanks.

>>>> +		dev_err(dev->device, "no irq resource\n");
>>>> +		return irq;
>>>> +	}
>> [...]
>>>> +static int bcm_iproc_i2c_remove(struct platform_device *pdev)
>>>> +{
>>>> +	struct bcm_iproc_i2c_dev *dev = platform_get_drvdata(pdev);
>>>> +
>>>> +	i2c_del_adapter(&dev->adapter);
>>>> +	bcm_iproc_i2c_disable(dev);
>>> I think you have a problem here if bcm_iproc_i2c_remove is called while
>>> an irq is still being serviced. I'm not sure how to prevent this
>>> properly for a shared interrupt.
>>>
>> Can I grab i2c_lock_adapter to ensure the bus is locked (so there's no
>> outstanding transactions or IRQs by the time we remove the adapter)? But
>> I see no I2C bus driver does this in their remove function...
> The problem I pointed out is the reason for some driver authors not to
> use devm_request_irq. If you use plain request_irq and the matching
> free_irq in the .remove callback you can be sure that the irq isn't
> running any more as soon as free_irq returns.
> 
Okay. Will change to use request_irq and make sure that it's freed in
the remove function. Also, the interrupt is dedicated to the I2C
controller, so I'll remove the IRQF_SHARED flag.

> BTW, if you use vim, you can add
> 
> 	set cinoptions=(,:
> 	if has("autocmd")
> 		filetype plugin indent on
> 	endif
> 
> to your .vimrc. Then while typing vim does the indention right and
> consistent, and with the = command you can reindent.
> 
Wow this is excellent! Just tried and it works perfectly. Thanks a lot!!!

> Best regards
> Uwe
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/