Re: Not enough IRQs for multiple PCMCIA cards

[Chris Howells <chrish-puGfsi27rH1aa/9Udqfwiw@public.gmane.org>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Nils,

On Saturday 10 May 2003 15:06, Nils Faerber wrote:
> For the record:
> 	I have exactly the same problem.

Whew, in some ways that's a relief :)

> > Regrettably, my BIOS (Award Asus L3800C BIOS) does not give me any
> > control how IRQs are designed or PnP OS settings.
>
> I have even the same machine :)

Yeah, I remember you helped answer a few questions of mine when I was thinking 
of buying the machine. I don't know if this particular machine is bad at 
controlling IRQ settings, or if all machines these days have such problems.

> OK.
> It is quite important to know a litle more about your setup.
> Why?
> I have tried with several kernels, pcmcia packages and configurations.
> Some worked, some not.

I'm currently using SuSE 8.2 with the stock SuSE 8.2 2.4.20-4GB kernel. I'm 
using the pcmcia-cs provided with SuSE 8.2 as well, and I've tried with both 
the orinoco and hostap drivers.

chris@galadriel:~> rpm -qa | grep pcmcia
pcmcia-3.2.3-47

> Do you use kernel pcmcia or the PCMCIA drivers from pcmcia-cs?

I'm using kernel PCMCIA, I've tried pcmcia-cs (just a matter of changing a 
setting in /etc/sysconfig/pcmcia on SuSE), but it didn't help.

> 	I currently use kernel pcmcia which currently does not work with twi
> cards.
>
> Which PCMCIA (cardbus) chipset driver do you use?
> 	I strongly recommend to use the kernel yenta socket driver! This gives
> best results with cardbus cards.

I must admit I haven't really tried any different kernels. I wonder if 
upgrading to 2.4.21-pre would help....

chris@galadriel:~> /sbin/lsmod | grep yenta
yenta_socket            9760   2
pcmcia_core            41824   0 [serial_cs hostap_cs ds yenta_socket]

> Some earlier kernel version with ACPI patch applied worked with
> pci=noacpi but this seems to have almost no effect on more recent
> kernels. I currently use 2.4.21-pre4 with ACPI 20030228.

Yeah, using ACPI for IRQ routing seems to work sometimes, though only very 
rarely. Is there any reason why it doesn't seem possible to control the way 
that ACPI allocates IRQs?

> Then there are some problems with certain PCMCIA drivers.
> In theory they should be able to share the one assigned interrupt. The
> serial_cs driver does so AFAIK. But many other drivers don't, like
> ide_cs. So some combinations may work, some not.

Ah, that's interesting. What kind of devices can serial_cs share IRQs with? 
PCI devices? How do I make it do that? Does it do it by default?

> This whole situation is, like you said, very unsatisfying :(

Yes, indeed. Previously it worked _sometimes_ (finding a valid IRQ to use) but 
it's stopped working now, I don't know why.

It's quite frustrating that Windows XP manages to handle IRQ allocation in a 
more satisfactory manner. (e.g. it gives the PCMCIA cards their own IRQ while 
making more of the PCI devices share their own IRQs).

> It is a real pain that there is no decent APIC in uniprocessor systems

I must admit I'm not too familiar with APICs. Does the Asus laptop have one at 
all?

> in order to have more than those shitty 15 IRQs... it's a shame... x86
> hardware is really crappy at times but there is not much of an
> alternative especially for notebooks :( (Oh oh ... no flame war start
> please!)

Yeah. I'm pretty sure my next laptop will be a Apple PowerBook :) Having a 
laptop which doesn't overheat if I use it on the bed will also be a novelty 
:)

- -- 
Cheers, Chris Howells -- chris-NvB7AskkBIqIudiWw5vspbVCufUGDwFn@public.gmane.org, howells@kde.org
Web: http://chrishowells.co.uk, PGP ID: 33795A2C
KDE: http://www.koffice.org, http://printing.kde.org, http://usability.kde.org

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2-rc1-SuSE (GNU/Linux)

iD4DBQE+vS93F8Iu1zN5WiwRAtRRAJinMjTl0KjtqYx7vEDgtoEnEx49AJ4misgZ
Py/oxp12CUb7r54meWBqnA==
=+C0Q
-----END PGP SIGNATURE-----


-------------------------------------------------------
Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara
The only event dedicated to issues related to Linux enterprise solutions
www.enterpriselinuxforum.com