Received-SPF: pass (google.com: domain of linux-kernel+bounces-86931-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Message-ID: <a5aa10f3-e487-4e70-8010-1604bea3a936@gmail.com>
Date: Thu, 29 Feb 2024 17:35:01 +0200
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH v8 5/5] iio: light: Add support for APDS9306 Light Sensor
Content-Language: en-US, en-GB
To: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Subhajit Ghosh <subhajit.ghosh@tweaklogic.com>,
 Jonathan Cameron <jic23@kernel.org>, Lars-Peter Clausen <lars@metafoo.de>,
 Rob Herring <robh+dt@kernel.org>,
 Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
 Conor Dooley <conor+dt@kernel.org>, Marek Vasut <marex@denx.de>,
 Anshul Dalal <anshulusr@gmail.com>,
 Javier Carrasco <javier.carrasco.cruz@gmail.com>,
 Matt Ranostay <matt@ranostay.sg>,
 Stefan Windfeldt-Prytz <stefan.windfeldt-prytz@axis.com>,
 linux-iio@vger.kernel.org, devicetree@vger.kernel.org,
 linux-kernel@vger.kernel.org
References: <20240228122408.18619-1-subhajit.ghosh@tweaklogic.com>
 <20240228122408.18619-6-subhajit.ghosh@tweaklogic.com>
 <a828e77c-b3d4-49bb-b0bb-b9fd6cb7d114@gmail.com>
 <Zd9tApJClX7Frq20@smile.fi.intel.com>
 <45386f39-a034-4d70-a6d4-8804c27aadce@tweaklogic.com>
 <21ecfb62-30b7-4073-bad6-46a9e08e08b0@gmail.com>
 <ZeCJ3T8HVaQZC1Ps@smile.fi.intel.com>
From: Matti Vaittinen <mazziesaccount@gmail.com>
In-Reply-To: <ZeCJ3T8HVaQZC1Ps@smile.fi.intel.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit

On 2/29/24 15:42, Andy Shevchenko wrote:
> On Thu, Feb 29, 2024 at 02:58:52PM +0200, Matti Vaittinen wrote:
>> On 2/29/24 14:34, Subhajit Ghosh wrote:
>>> On 29/2/24 03:57, Andy Shevchenko wrote:
>>>> On Wed, Feb 28, 2024 at 03:08:56PM +0200, Matti Vaittinen wrote:
>>>>> On 2/28/24 14:24, Subhajit Ghosh wrote:
> 
> ...
> 
>>>>>> +    if (gain_new < 0) {
>>>>>> +        dev_err_ratelimited(dev, "Unsupported gain with time\n");
>>>>>> +        return gain_new;
>>>>>> +    }
>>>>
>>>> What is the difference between negative response from the function
>>>> itself and
>>>> similar in gain_new?
>>>>
>>> -ve response form the function is an error condition.
>>> -ve value in gain_new means - no valid gains could be computed.
>>> In case of error conditions from the function, the gain_new is also set
>>> to -1.
>>> My use case is valid hardware gain so I went for checking only gain_new.
>>> Matti will be the best person to answer on this.
>>
>> I now rely on the kerneldoc for the
>> iio_gts_find_new_gain_by_old_gain_time() as it seems reasonable to me:
>>
>> * Return: 0 if an exactly matching supported new gain was found. When a
>> * non-zero value is returned, the @new_gain will be set to a negative or
>> * positive value. The negative value means that no gain could be computed.
>> * Positive value will be the "best possible new gain there could be". There
>> * can be two reasons why finding the "best possible" new gain is not deemed
>> * successful. 1) This new value cannot be supported by the hardware. 2) The
>> new
>> * gain required to maintain the scale would not be an integer. In this case,
>> * the "best possible" new gain will be a floored optimal gain, which may or
>> * may not be supported by the hardware.
> 
>> Eg, if ret is zero, there is no need to check validity of the gain_new but
>> it is guaranteed to be one of the supported gains.
> 
> Right, but this kernel doc despite being so verbose does not fully answer my
> question. What is the semantic of that "negative value"? 

Current approach is to always investigate the function return value as 
error if the 'new_gain' is negative. Or, caller specific error if 
new_gain is unsuitable in some other way. When this is done, the 
absolute value of the negative 'new_gain' does not matter.

> I would expect to have
> the error code there (maybe different to what the function returns), but at
> least be able to return it to the upper layers if needed.

I am not sure I see the benefit of returning the new_gain over returning 
the error returned by the function. Well, maybe the benefit to be able 
to not evaluate the value returned by the 
iio_gts_find_new_gain_by_old_gain_time() - although I'm not sure I love it.

> Hence 2 ARs I see:
> 1) clarify the kernel documentation there;
> 2) update the semantic of the gain_new to simplify caller's code.

Yes, I agree. Patches welcome :) By the very least the kerneldoc can be 
improved. I'm undecided on benefits of having the error code in 'new_gain'.

The GTS API fixes shouldn't be required in the context of this driver 
series though.

Yours,
	--Matti

-- 
Matti Vaittinen
Linux kernel developer at ROHM Semiconductors
Oulu Finland

~~ When things go utterly wrong vim users can always type :help! ~~