Re: 2.6.20 "IRQ handler type mismatch for IRQ 4"

[Alan <alan@lxorguk.ukuu.org.uk>]

> [   23.783913] Serial: 8250/16550 driver $Revision: 1.90 $ 4 ports, IRQ 
> sharing enabled
> [   23.787063] pnp: Device 00:0c activated.
> [   23.787420] 00:0c: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A

So the PnP layers put a device on IRQ 4, which is sensible

> [   37.516000] eth1: orinoco_cs at 0.0, irq 4, io 0x0100-0x013f

The PCMCIA layer then ignores this and takes IRQ 4 simply because
although PnP has used it the IRQ isnt currently live

> [   53.580000] IRQ handler type mismatch for IRQ 4
> [   53.580000] current handler: pcmcia0.0

The serial IRQ setup fails

> [   53.580000]  [<c0141a25>] setup_irq+0x135/0x1e0
> [   53.580000]  [<c0230930>] serial8250_interrupt+0x0/0xf0
> [   53.580000]  [<c0141b73>] request_irq+0xa3/0xc0
> [   53.580000]  [<c023086e>] serial8250_startup+0x47e/0x4b0
> [   53.580000]  [<c022bf58>] uart_startup+0x48/0x160
> [   53.580000]  [<c022cb9f>] uart_open+0xbf/0x460

This is all scary and generally not needed spew but handy for debug.

Looks like a PCMCIA layer bug. It should be respecting IRQ assignment
(not just allocation) by other resource configuration layers.

Alan