Maybe this is my excuse to do some kernel-related development, then. :)
> >2. Once I got the APIC problem solved, I found that both my eepro100
> > and my scsi1 were sharing the same interrupt. / is on scsi1, so
> > that interrupt was causing lots of activity. A kernel complile
> > resulted in more "scsi: aborting command due to timeout..."
>
> This should _not_ matter, but the problem seems to be that not only were
> the interrupts shared, they also seem to be edge-triggered:
> > [...]
> and having shared edge-triggered interrupts is something that is very
> obviously not ever going to work reliably (Linux could try harder to
> make it work, but that would just make the problems happen less often
> rather than go away completely).
I'll take your word for it. It was working fine under NT, but NT didn't
have any drives under scsi1, just scsi0. Perhaps someone got lucky.
> I'd be interested to know if you can move the card back to a shared irq,
> if you can also get the BIOS to mark the interrupt as level-sensitive (I
> don't know if you even have such a BIOS option, but it's worth looking
> around in the "advanced chipset" settings etc for something like that).
Well, I ended up physically moving the PCI card to the second PCI bus
so that it would take something other than IRQ10. I searched for
*anything* interrupt-related in the BIOS that I might be able to change
before I moved the card, and didn't find anything. The SCSI BIOS had
more parameters to play with than the system BIOS, and it didn't have
anything interrupt-related, either. I'll double-check next time I
reboot it, though.
> I understand that it seems to work for you as-is, but I'd appreciate you
> trying out the above alternate setup, just to get some additional
> information..
That's a very diplomatic request, Linus, but I understand that no one
ever learned anything by taking something that works as-is. I would be
happy to break something (software, not hardware) to figure it out.
Noah
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu