Q: Kconfig: 'If unsure, say N'

Q: Kconfig: 'If unsure, say N'

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 1905 bytes --]

Hi,

In quite a number of Kconfig help text entries I see this:
"If unsure, say N."

But that raises the question: How can I be sure?

To me, this comes across as that the person who implemented the feature
recommends *against* using it, unless you think you know better then the
person who implemented it. Which is quite a high bar.

IIRC I did come across an entry which paraphrased said:
"This module can be useful in situation Y, but you run a real risk of
physically damaging your board when you use it.
So normally you REALLY should not enable this, but if you still need it
then the functionality is implemented in this module.

If unsure, say N."

Which is an excellent reason not to enable it ;-)
Moreover, it specifies when you can/should go against the advise and
tells you what the risk is if you do.

But the vast majority just says "If unsure, say N."

The problem is that I'd need a better justification to enable (as 'y' or
'm') a module then "Based on (the rest of) the Help text, this looks
really useful".

Not to discuss these specifically, but just for illustration:
``drivers/firmware/arm_scmi/transports/Kconfig`` has this
option: ``ARM_SCMI_TRANSPORT_SMC_ATOMIC_ENABLE``
which IIUC enables an *optional* feature for an atomic transaction.

Which sounds useful and harmless, yet ... "If unsure, say N."

And the trigger which caused me to actually write this email was
"scsi: ufs: rockchip: initial support for UFS"
which I interpret as: if you want to use UFS on Rockchip based devices,
you should enable this. But ... "If unsure, say N."

So it would be really helpful if the Kconfig help text:
1) did not say "If unsure, say N."
2) If the recommendation is indeed to NOT enable it ('normally'),
   specify why and under what situation/condition you can go against the
   maintainer/implementers recommendation

Cheers,
  Diederik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: Q: Kconfig: 'If unsure, say N'

[Harald Arnesen]

Diederik de Haas [2024-11-08 15:39:30]:

> In quite a number of Kconfig help text entries I see this:
> "If unsure, say N."
> 
> But that raises the question: How can I be sure?

I usually interpret this as "If you don't know that you need it, you 
probably don't".
-- 
Hilsen Harald

Re: Q: Kconfig: 'If unsure, say N'

[Sudeep Holla]

On Fri, Nov 08, 2024 at 03:39:30PM +0100, Diederik de Haas wrote:
> Hi,
>
> In quite a number of Kconfig help text entries I see this:
> "If unsure, say N."
>
> But that raises the question: How can I be sure?
>
> To me, this comes across as that the person who implemented the feature
> recommends *against* using it, unless you think you know better then the
> person who implemented it. Which is quite a high bar.
>

If you don't know about the feature and it is not having any user-interface
why do you want to enable it. You must understand the feature to enable it
and use it IMO.

> IIRC I did come across an entry which paraphrased said:
> "This module can be useful in situation Y, but you run a real risk of
> physically damaging your board when you use it.
> So normally you REALLY should not enable this, but if you still need it
> then the functionality is implemented in this module.
>
> If unsure, say N."
>
> Which is an excellent reason not to enable it ;-)
> Moreover, it specifies when you can/should go against the advise and
> tells you what the risk is if you do.
>
> But the vast majority just says "If unsure, say N."
>
> The problem is that I'd need a better justification to enable (as 'y' or
> 'm') a module then "Based on (the rest of) the Help text, this looks
> really useful".
>

You seem to contradict yourself here. If you have understood the help
text and think it is useful, it seems to me as an indication that you
are not unsure really. So you can enable it if you want TBH.

> Not to discuss these specifically, but just for illustration:
> ``drivers/firmware/arm_scmi/transports/Kconfig`` has this
> option: ``ARM_SCMI_TRANSPORT_SMC_ATOMIC_ENABLE``
> which IIUC enables an *optional* feature for an atomic transaction.
>

"If you want the SCMI SMC based transport to operate in atomic
mode, avoiding any kind of sleeping behaviour for selected
transactions on the TX path, answer Y.

Enabling atomic mode operations allows any SCMI driver using this
transport to optionally ask for atomic SCMI transactions and operate
in atomic context too, at the price of using a number of busy-waiting
primitives all over instead."

So you read the above text, understood and find it useful. You must be
not unsure of this feature then, so what does that text bother you.
It is just a caution to users who are just build and not looked at the
code, or have no idea about the feature or doesn't understand the help
text.

--
Regards,
Sudeep

Re: Q: Kconfig: 'If unsure, say N'

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 2278 bytes --]

Thanks for your response.

On Tue Nov 12, 2024 at 6:46 AM CET, Sudeep Holla wrote:
> On Fri, Nov 08, 2024 at 03:39:30PM +0100, Diederik de Haas wrote:
> > In quite a number of Kconfig help text entries I see this:
> > "If unsure, say N."
> >
> > But that raises the question: How can I be sure?
>
> If you don't know about the feature and it is not having any user-interface
> why do you want to enable it. You must understand the feature to enable it
> and use it IMO.

Ok, understood.

> You seem to contradict yourself here. If you have understood the help
> text and think it is useful, it seems to me as an indication that you
> are not unsure really. So you can enable it if you want TBH.
>
> > Not to discuss these specifically, but just for illustration:
> > ``drivers/firmware/arm_scmi/transports/Kconfig`` has this
> > option: ``ARM_SCMI_TRANSPORT_SMC_ATOMIC_ENABLE``
> > which IIUC enables an *optional* feature for an atomic transaction.
> >
>
> "If you want the SCMI SMC based transport to operate in atomic
> mode, avoiding any kind of sleeping behaviour for selected
> transactions on the TX path, answer Y.
>
> Enabling atomic mode operations allows any SCMI driver using this
> transport to optionally ask for atomic SCMI transactions and operate
> in atomic context too, at the price of using a number of busy-waiting
> primitives all over instead."
>
> So you read the above text, understood and find it useful. You must be
> not unsure of this feature then, so what does that text bother you.

In this case I was wondering whether this was related to Table 1 in
"Power and Performance Management using Arm SCMI Specification"
whitepaper.
And I do think that I understand most of the Help text, but I'm not/less
sure about the part starting with "at the price of ...".

> It is just a caution to users who are just build and not looked at the
> code, or have no idea about the feature or doesn't understand the help
> text.

And my question was exactly about "just a caution", which raises the
question with me "caution for what?" ... which is an unspecified risk.
So to be on the safe side, I will not enable modules with "If unsure,
say N" unless I fully understand the source code.

Cheers,
  Diederik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: Q: Kconfig: 'If unsure, say N'

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 440 bytes --]

Hi,

On Fri Nov 8, 2024 at 4:05 PM CET, Harald Arnesen wrote:
> Diederik de Haas [2024-11-08 15:39:30]:
>
> > In quite a number of Kconfig help text entries I see this:
> > "If unsure, say N."
> > 
> > But that raises the question: How can I be sure?
>
> I usually interpret this as "If you don't know that you need it, you 
> probably don't".

Yeah, that does sound like a sensible strategy. Thanks :)

Cheers,
  Diederik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]