[PATCH v5 2/2] i2c: aspeed: support ast2600 i2cv2 new register mode driver

[Krzysztof Kozlowski]

On 23/02/2023 01:58, Ryan Chen wrote:
> Hello Krzysztof,
> 
>> -----Original Message-----
>> From: Krzysztof Kozlowski <krzysztof.kozlowski at linaro.org>
>> Sent: Wednesday, February 22, 2023 4:28 PM
>> To: Ryan Chen <ryan_chen at aspeedtech.com>; Rob Herring
>> <robh+dt at kernel.org>; Krzysztof Kozlowski
>> <krzysztof.kozlowski+dt at linaro.org>; Joel Stanley <joel at jms.id.au>; Andrew
>> Jeffery <andrew at aj.id.au>; Philipp Zabel <p.zabel at pengutronix.de>;
>> openbmc at lists.ozlabs.org; linux-arm-kernel at lists.infradead.org;
>> linux-aspeed at lists.ozlabs.org; linux-kernel at vger.kernel.org
>> Subject: Re: [PATCH v5 2/2] i2c: aspeed: support ast2600 i2cv2 new register
>> mode driver
>>
>> On 22/02/2023 04:36, Ryan Chen wrote:
>>
>>>>> +
>>>>> +	return 0;
>>>>> +
>>>>> +free_irq:
>>>>> +	devm_free_irq(&pdev->dev, i2c_bus->irq, i2c_bus);
>>>>
>>>> Why?
>>>>
>>>>> +unmap:
>>>>> +	devm_iounmap(&pdev->dev, i2c_bus->reg_base);
>>>>
>>>> Why?
>>>>
>>>>> +free_mem:
>>>>> +	devm_kfree(&pdev->dev, i2c_bus);
>>>>
>>>> Why?
>>>>
>>>
>>> Sorry, those are probe following, if any error, will goto this label.
>>> To release mem/unmap/free_irq. Is this unnecessary?
>>
>> Releasing managed resources is usualyl unnecessary. Therefore I am asking
>> why do you think it is necessary here?
>>
>>> I saw many driver submit is remove all probe fail goto label, is just return
>> ERR.
>>> Do you mean I direct go for this way?
>>
>> Why would you do it differently?
> 
> Thanks, I will remove those labels, and modify to dev_err_probe in previous probe return.
> 
>>>
>>>>> +
>>>>> +	return ret;
>>>>> +}
>>>>> +
>>>>> +static int ast2600_i2c_remove(struct platform_device *pdev) {
>>>>> +	struct ast2600_i2c_bus *i2c_bus = platform_get_drvdata(pdev);
>>>>> +
>>>>> +	/* Disable everything. */
>>>>> +	writel(0, i2c_bus->reg_base + AST2600_I2CC_FUN_CTRL);
>>>>> +	writel(0, i2c_bus->reg_base + AST2600_I2CM_IER);
>>>>> +
>>>>> +	devm_free_irq(&pdev->dev, i2c_bus->irq, i2c_bus);
>>>>> +
>>>>> +	i2c_del_adapter(&i2c_bus->adap);
>>>>
>>>> Wrong order of cleanup. It should be reversed to the probe, unless
>>>> you have some reason, but then please explain.
>>>
>>> Sorry, this in remove function, it should do i2c_del_adapter(&i2c_bus->adap)
>> in the end.
>>> Why this should revered to probe?
>>
>> Because it's logical, you do the same with error paths of probe, it it usually the
>> only way to make sure some of the resources are not used by some other piece
>> (e.g. interrupt handler is called while i2c adapter is unregistered).
> 
> Sorry, I can't catch your point.
> Do you mean remove devm_free_irq(&pdev->dev, i2c_bus->irq, i2c_bus);
> Keep i2c_del_adapter(&i2c_bus->adap) here, because interrupt is called while i2c adapter is unregistered ?

Again, maybe clearer, actions should be in reversed order comparing to
the probe actions. Why would you remove pieces of code from here if I
asked to do them in different order?

Best regards,
Krzysztof