[irda-users] Re: [linux-usb-devel] stir4200

[Jean T. <jt...@bo...>]

On Tue, Jun 24, 2003 at 08:50:23PM +0200, Martin Diehl wrote:
> 
> On Tue, 24 Jun 2003, Tomek Grzejszczyk wrote:
> 
> > 	I'am trying to write driver for STIR4200 usb-irda bridge controller. I'am
> > not writing it from scratch but rather modify current irda-usb driver
> > from 2.4.20 kernel. Right now I'am testing what I've written but I've
> 
> I'm not sure whether this would be the best approach. From looking across 
> the stir4200 datasheet I'm pretty sure it's completely different from 
> irda-usb class. So I think all you can keep from irda-usb is some network 
> and usb skeleton which might even be a sub-optimal thing to start from for 
> this device - but it's your decision of course.

	I already pointed to the alpha driver on my web page ;-)

> The second thing I'd suggest is using the very latest 2.4 kernel, probably 
> 2.4.22-pre2, in order to get all the recent important irda and usb fixes.
> Or maybe even better 2.5.

	I will merge new drivers in 2.5.X first, and reject drivers
that I can't integrate in 2.5.X. I'm starting to get fed up that for
each driver I have to spend a few days just to integrate it properly
in 2.5.X because the author just didn't care.

> Have you seen the experimental driver for the stir4200 at Jeans Homepage?
> Maybe you could start with this one or cooperate with the other author to 
> avoid having to start all over?

	The main issues with the driver on my web page :
	o sequencial transmission of frames is busted, the driver
would need a scheme to properly wait for the end of the transmission
of the previous frame, and the HW doesn't seem to offer an easy way to
do that.
	o Rx of frames at 4Mb/s doesn't timely push the frame to the
host and bundle multiple frames in a descriptor without any way to
separate them (you will get multiple RR bundled together).
	To be polite, I was not impressed by the hardware design. This
is going to be difficult to workaround those "features".

> Martin
> 
> PS: I've added Jean and the irda-users to CC because you send another 
> request there and I think the thread would be more appropriate there.

	Correct ;-)

	Jean