Linux Wireless Woes

fractal

OK..

Im trying to get Wireless networking running in Linux (A thing I've done many times before)

But this time im running on a fairly minimal system so it doesnt fall to cardmgr to config everything.

So i do this:
insmod pcmcia_core (This works grand, loading linux card services)
insmod yenta_socket (this is where ther are problems)

Spouts the following to me:
Using /lib/modules/2.4.22/pcmcia/yenta_socket.o
PCI: No IRQ known for interrupt pin A of device 00:12.0. Please try using pci=biosirq
Yenta IRQ list 0000, PCI irq0
Socket status: 10000011
ti113x: Routing card interrupts to PCI

So i try passing pci=biosirq to the kernel at boot like has worked for me before and i get the same message without the whole helpful "Please try using pci=biosirq" bit..

Anyways I think maybe this all is a vain attempt at sadistic kernel developers to make me think it isnt working when in fact it is.. So I insmod ds, hermes, orinoco, orinoco_cs and start up cardmgr (All without problem) And then i plug in my Orinoco-silver wirless
card.. all that work only to be greeted by a: orinoco_cs: RequestIRQ: Resource in use

I know the PCMCIA Adaptor is working coz if I try the same thing on the same hardware
booting into RedHat 9.1 (kernel 2.4.20-8) and pass the pci=biosirq param all is sunny and
works.. BEEP,BEEP it goes and allis well..

Any help GREATLY appreciated..

PS: This could maybe have gone in the Nets/Comms->Wireless group too but I thought here might be better as its more of a driver problem.

Snowbat

So your hardware works in "RedHat 9.1 (kernel 2.4.20-8)" but which kernel does "fairly minimal system" use?

I had similar fun trying to get an Enterasys wifi card in PCI adapter working in RH8.0 (2.4.18-4) but my case it was due to 430HX boards (tried two). It seems the 430HX interrupt router was not fully supported in 2.4.18-4. A kernel recompile with the then latest pcmcia-cs gave the same error and booting with pci=biosirq produced a kernel panic. Identical setup on a 430VX board, a VIA-based board, and Asus CUSL2-C worked fine (pci=biosirq not required).

http://sourceforge.net/forum/forum.php?thread_id=857418&forum_id=33428


I didn't get any further with it so I put the 430VX board into use.

voxpop

I also had many problems with setting up my wireless card on gentoo, got the same sort of problems aswell. It turned out that the correct module wasnt being loaded for my card

Most pcmcia HOW-TOs recommend you use the pcmcia-cs package http://pcmcia-cs.sourceforge.net/
instead of the built in kernel pcmcia drivers. If you havent tried these then that would be my first step although my own smc wireless card would only work with the built in kernel pcmcia drivers

Second check out that your pcmcia hardware is listed in /etc/pcmcia/config and associated with the correct module.

If you want to manually set pci address (you might need to set the pci address in /etc/pcmcia/config file) then you can look it up using lspci

Hope the above helps

fractal

Thanks to the replies but I've found a solution!

The problem was with the 2.4.22 (and pre 2.4.22 i think) kernel implementation of the yenta_socket driver for pci -> pcmcia bridges..

In certain cases pertaining to certain hardware it seems that an IRQ cannot be established for use with the bridge device..

2.4.20-8 works coz I presume RedHat nicely added their own patches for the problem..

The minimal system i was talking about was a self-built router/firewall system with a ~450kb custom compiled kernel.. (2.4.22)

Anyways after looking at quite a few PCMCIA setup tutorials i would have thought a nice workaround would have been to allow manually specifed IRQ's to be passed to yenta_socket (but this isnt implemented)

2.4.24 fixes this problem anyways and the latest prepatch towards 2.4.25 adds a few more fixes for PCMCIA...

I think I've learned quite a lot over the past few days about how PCMCIA subsystem works so fingers crossed I wont run into another problem like this again..

Thanks everyone for replies.