To Alan who is perhaps reading this: Who is an expert on this type of stuff?

Davide Libenzi wrote:
> On Mon, 26 May 2003, Thomas Winischhofer wrote:
> 
> 
>>and I had (and have) no problems with irqs or USB (or anything) on any
>>of these machines.

First, let me say that I know NIL about irq routing. But fact is, I had 
my machines running with webcams, floppy drives and mice (all via USB, 
that is) - and had no problem.

But you got me puzzled: As a matter of fact, it seems that ALL (!) my 
650 variants show different routing tables, mostly like yours.

dmesg with pci-debug enabled and lspci -vxxx printouts attached.

650 = "ISA bridge" revision 0
M650 = revision 4
651 = revision 0x25

Interestingly, as soon as pci-debugging was enabled, the log is full 
with error messages, and I suddenly actually _had_ problems with my 
network card and my wireless card (and assumingly the USB stuff, too, 
conclusing from the "failed" statements in the log)....

>>Are you sure that checking the revision number of the device is enough?
> 
> 
> It seems reasonble, at least without having the spec for the chipset. All
> my searches failed about docs. Previous cases are correctly handled like
> before, as you can see from the patch.

I myself doubt this now. If I am reading the dmesg output correctly, 
even the machine with revision 0 (plain 650) is routing _some_ of the 
interrupts with 0x6x requests...

> You happen to have the spec for the SIS650 ?

That's a good one :) I have been bugging SiS since 2001 for docs, all I 
got was nothing.

 > The reality is that the
> chipset issues 0x60..0x63 router request and the current kernel will not
> correctly initialize device under that spot. So either we have a too
> strict implementation in the current router (we should by default pass
> thru unknown requests) or we need to fix it according to the new requests.
> The "stdroute" thing is a way to keep the strict behaviour by default, and
> yet have the ability to pass thru w/out the need of patching the kernel.
> 
> 
> Below is reported the boot with PCI debug enabled and the content of my
> PCI devices configuration space. If you have a better idea on how to fix
> it, please submit a patch.

As said, I know nothing about this stuff. The last time I dealt with 
irqs was on the Amiga...



Thomas

-- 
Thomas Winischhofer
Vienna/Austria
thomas AT winischhofer DOT net          *** http://www.winischhofer.net/
twini AT xfree86 DOT org