IO-APIC IRQs strange? (was Re: linux 2.1.98 sound timeouts)

Michael Poole (poole@graviton.res.cmu.edu)
27 Apr 1998 23:01:22 -0400


Michael Poole <poole@graviton.res.cmu.edu> writes:

> I'm seeing the same thing, and have been for a while, with my sound
> card as well -- however, it gives the time out error every time an app
> tries to use the wave output functions, and all I hear are quiet clicks.

After a patch from Artur Skawina that "should" have better supported
my sound card didn't give any better results, I did a bit more digging.
It turned up a very surprising result -- /proc/interrupts doesn't list
all the interactions between IRQs and interrupt handlers that I was
expecting, and this meant that I had been looking in entirely the wrong
place for problems.

I discovered the discrepancy by booting back to 2.0.33, where I fully
expected sound to work -- after all, it had worked before. However, it
did not -- the ad1848 driver failed the IRQ10 test, which is where my
(non-PNP) sound card lives. So I looked in /proc/interrupts, and it
revealed said that my aic7xxx and Intel EEPro100 were on IRQ 10. Under
2.1.x, these were listed under IRQ 19 -- so I had assumed there was no
conflict. I was wrong. After going into the BIOS and listing IRQ 10
(and a number of others, after trial and error) as being used by ISA
peripherals, 2.0.33 reported the SCSI and NIC being on IRQ 14, and sound
worked again -- both in 2.0.33 and 2.1.97.

So this raises a question in my mind: Why were the SCSI and network
cards causing interference, apparently slurping away the interrupts
which "should" have been received by the sound card driver? I understand
that it was a misconfiguration on my part in the first place, but I
didn't guess it until today because guessing was virtually all I could
do under the 2.1 kernels. Would it be somehow possible to discover, if
not rectify, this kind of occurence? I am willing and able to test
patches to diagnose or to try to fix this, if any experienced kernel
hackers are interested.

- Michael

P.S.: the Sound Galaxy driver from Artur Skawina is in good working
order, now that the other problems have been fixed.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu