Re: [irda-users] Re: Enhancements to smcinit

[<jfe...@fr...>]

jt...@hp... wrote:
> jfenalml wrote :
> 
>>I"m on the way to release on my home site 
>>(http://jfenal.free.fr/linux/nc6000.html) a strongly modified version of 
>>smcinit which works on HP/Compaq nc6000.
...
>>I also need some advice: would it be nice to try to deactivate the ttyS2 
>>(or other) UART via the right ioctl the same way setserial would do it, 
>>restoring values if no SuperIO is found, or should this be kept a pure 
>>and only setserial task ?
> 
> 	This is needed for all FIR devices (SMSC, NSC, VIA...),
> therefore I personally would rather add it to irattach. This way, when
> irattach exit, you would be able to restore the UART in its default
> config.

Right. I'll have a look to it.

I have some more questions:
- what is the target kernel: 2.4 or 2.6?
- if it would be irattach's job to unbind the serial driver from the 
ttySx, should'nt it be also its job to init hardware, as done in smcinit?

This last question is quite important since the order to make things 
work is the following:
- unbind (via setserial),
- smcinit,
- load smsc-ircc2.[k]o
- irattach

My thought was that this unbind it could be the job of either:
1. a super smcinit, with access to a database of pci-ids to smsc-ircc2 
parameters and ttySx (in order to avoid hardcoded pci-ids, which would 
permit to easily add new laptops)
2. the kernel module smsc-ircc2.

My preference goes to the first since:
- it can be done in user mode as it is already,
- the code is to be used only once at init time,
- doing 2. would bloat the kernel module with more code (init) and data 
(pci-ids to irq/dma/etc.)

> 	Thanks ;-)
you're welcome.

J.