mdev

[Vladimir Dronnikov]

>> > However, I suspect that having a hotplug wrapper script in examples might
>> > be more useful.
>>
>> This all seems to be the step to /etc/hotplug.d/ ages.
>
> So if we don't implement this functionality ourselves in the busybox binary,
> it could be implemented badly by someone else?  That's usually true, yes, but
> not a very strong argument for adding code to busybox.
>

Exactly. I just recall that times and it makes me happy people finally
have invented better tools.
A simple, understandable _declarative syntax_ for configuration,
instead of procedural approach, I treat here a very good improvement.

>>
>> To split mdev and tune obtained parts separately is a good option then.
>
> What are "tune obtained parts"?
>

Sorry for the language. I meant when we split (A+B) to the separate
parts A and B we can improve/change/freeze-changes each of the two
parts more or less independently. The whole discussion seems to
discard this idea, though.

>> To be brief: if you want no sound from beeper, you unload the
>> pcspkr.ko module. If you then want the sound, you load the module.
>> When it is compiled in, you just have no choice.
>
> "if you want the sound, you load the module".  So you're saying that this
> option must be manually user selectable, but defaulting to off instead of on
> is unacceptable, and having boot scripts insmod it isn't good enough?  I'm
> not sure I'm following here.

I just said _iff_ there were such an option, _I_'d make in on. It is
up to others what value will they set.

>
> I'm also not sure why some variant of "modprobe /lib/modules/*" isn't an
> option if you want to load everything that's loadable at boot time.

I assure you it would take much longer. Moreover, it will result in
unnecessary kernel memory consumptions. (Moreover, it would resemble a
lady, turning her ridicule upside down and make all the thingies
spread over the table, just to find her favorite tiny lipstick. Sorry
for such an irrelevant passage...)

Nevertheless, it is quite valid.

> They're all supposed to have probe code to see if there's any device for them to
> attach to.  Why are we designing code to iterate through available devices in
> sysfs when we could just iterate through the _modules_ instead?  (There may
> be a legitimate reason it won't work, I just wanted to make sure we'd looked
> into it first.)

The key words are "iterate through _available_ devices". That is the
point, I presume.

Best regards,
--
Vladimir