Re: [PATCH v3 2/9] i2c: mlxbf: Use dev_err_probe in probe function

[Andi Shyti]

On Tue, Aug 08, 2023 at 05:55:29PM +0200, Krzysztof Kozlowski wrote:
> On 08/08/2023 13:47, Andi Shyti wrote:
> > Hi Krzysztof,
> > 
> > On Tue, Aug 08, 2023 at 01:31:31PM +0200, Krzysztof Kozlowski wrote:
> >> On 08/08/2023 13:29, Andi Shyti wrote:
> >>> Hi Krzysztof,
> >>>
> >>> On Tue, Aug 08, 2023 at 10:36:40AM +0200, Krzysztof Kozlowski wrote:
> >>>> On 08/08/2023 03:29, Liao Chang wrote:
> >>>>> Use the dev_err_probe function instead of dev_err in the probe function
> >>>>> so that the printed messge includes the return value and also handles
> >>>>> -EPROBE_DEFER nicely.
> >>>>>
> >>>>> Reviewed-by: Andi Shyti <andi.shyti@xxxxxxxxxx>
> >>>>> Signed-off-by: Liao Chang <liaochang1@xxxxxxxxxx>
> >>>>
> >>>> ...
> >>>>
> >>>>> @@ -2413,10 +2399,8 @@ static int mlxbf_i2c_probe(struct platform_device *pdev)
> >>>>>  	ret = devm_request_irq(dev, irq, mlxbf_i2c_irq,
> >>>>>  			       IRQF_SHARED | IRQF_PROBE_SHARED,
> >>>>>  			       dev_name(dev), priv);
> >>>>> -	if (ret < 0) {
> >>>>> -		dev_err(dev, "Cannot get irq %d\n", irq);
> >>>>> -		return ret;
> >>>>> -	}
> >>>>> +	if (ret < 0)
> >>>>> +		return dev_err_probe(dev, ret, "Cannot get irq %d\n", irq);
> >>>>
> >>>> I don't think this is needed:
> >>>> https://lore.kernel.org/all/20230721094641.77189-1-frank.li@xxxxxxxx/
> >>>
> >>> Hmm, that's a bit borderline, I'd say. The change to
> >>
> >> What's borderline exactly? devm_request_threaded_irq_probe() is coming,
> >> right? If it is accepted this hunk is useless and soon should be
> >> replaced with proper one.
> > 
> > Such change is out of the scope of this series, there are two
> > options that I'd prefer (in the listed order):
> > 
> >  1. accept the patch as it is, this patch is not sent today the
> >     first time and at the current state it's correct.
> >  2. not accept a change on this line
> 
> The 2 is what I commented here. This change should not be made and
> instead we should just switch all such users to new API, because this is
> preferred for all error messages, when applicable and does not result in
> lost context. If there was no such API, sure, but we have this API coming.

To me the patch is correct... I am OK also with 2, but I find 1
more complete... let's say that it's a matter of taste?

> > Replacing devm_request_irq belongs to another series and,
> > besides, I don't want to ask Liao to hold on this series for such
> > trivialities.
> 
> So the comment about this redundant and unneeded change, thus switching
> to new API you call 'triviality' but a comment of yours of changing the
> tone of error message to 'please' is appropriate.
> https://lore.kernel.org/all/20230807231320.svssge6uymw3jiho@intel.intel/
> 
> That's double standards.

That was a joke, the review was somewhere else in my comment.

Thanks for your inputs,
Andi