Mailinglist Archive: opensuse (5130 mails)

< Previous Next >
Re: [SLE] Changing hardware IRQ's
  • From: S Glasoe <srglasoe@xxxxxxxxxxx>
  • Date: Mon, 8 May 2006 14:26:23 -0500
  • Message-id: <200605081426.24045.srglasoe@xxxxxxxxxxx>
On Sunday 07 May 2006 7:03 am, Hylton Conacher (ZR1HPC) wrote:
> > I keep coming back to one of my first questions though. Think there has
> > been 3 different threads based on IRQ settings. Can't keep them
> > straight anymore. Does it matter if your devices share an IRQ? Or are
> > you just fixated on trying to get each PCI or onboard device _not_ to
> > share an IRQ? I know you may be getting an error message stating
> > something about unknown parameters (IRQ being one of them) when loading
> > a driver but my experience tells me that is just an old generic error
> > message that can't be more specific. Its giving you a hint as to what
> > to troubleshoot.
>
> Oooh, time for the deeep stuff :)
> I am more fixated on getting each device to have a single IRQ as I seem
> to remember from my past that having the same IRQ, on a Windows system
> anyway, would render either one or both of the devices inoperable ie
> they cannot both have the same request number to query the cpu.

Remember that there were never enough hardware based IRQs to go around to
uniquely identify all the hardware people wanted to put into a PC. One
manufacturer I worked at cascaded 3 Interrupt Controllers to get enough
unique hardware IRQs to support their SCSI controllers and CPU boards. Most
PCs used 1 or 2 Interrupt Controllers in the day when these were individual
ICs and needed the mainboard real estate for connections.

This was true in the deep, dark past of the early PCI days circa mid-90's to
late-90's. Once PCI became the dominant slot standard for mainboards, PCI
matured (got most of the nasty bugs out), PCI card manufacturers gained
more experience and got their bugs worked out, and the OS vendors got
things worked out then dedicated slots per IRQ weren't as necessary. Once
all these players got it straight that PCI bus, slot, IRQ, I/O port, memory
addresses, etc could all be used to uniquely identify devices things have
gotten much better.

What you are seeing now reminds you of those IRQ conflict days. I wager that
what is really happening is that there is a driver issue where your
particular mainboard and BIOS aren't helping with the unique identification
of all the devices. The driver(s) and/or Linux kernel may not know about
your hardware's idiosyncracies and therefore can't allow perfect operation.

> Whilst I realise it probably doesn't matter on Linux and that a newer
> version would help and that it seems that both the network card and
> sound are basically operating, although not fantastically, there must be
> something in this IRQ thing as cpu's have not changed their operating
> habits since Linux arrived, I don't think?

Using Konppix or Ubuntu/Kubuntu to troubleshoot may show that all your
hardware can work. Try the Live CD versions to see what happens. In regards
to hardware not much may have changed. Its the software though that is
changing more these days in regards to drivers, kernel drivers, BIOS and
firmware updates, etc. If the developers know of an issue they can
typically code for it.

> I know its thinking outside of the box but it is interesting.

Whatever solves it!

Stan

< Previous Next >