Adding Hard Drives and IRQ Conflicts

Lennart Sorensen lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org
Tue Jan 2 20:48:26 UTC 2007


On Tue, Jan 02, 2007 at 03:02:27PM -0500, Paul King wrote:
> I have an ASUS AV7-133 motherboard using an Athlon K7 processor, and I have 
> found to my horror that I have umpteen devices all sharing IRQ 9. This 
> includes: 
>     ACPI (not sure what that is)

Advanced Configuration and Power management Interface (or something like
that).

It is an interface for managing device configuration and power
management.  It replaced APM among other things.  It is quite useful
(when not broken, which on many BIOS's it is).

>     Video card
>     My sound card
>     My ethernet card
>     On-board USB ports
>     PCI card USB ports
>     My PCI Ultra ATA card (connected to one drive)

Well normally interrupts 9 through 11 are available to be assigned for
PCI interrupts A through D.  PCI is designed to share interrupts so it
usually isn't a problem, although not sharing does have a very slight
performance benefit.  Often you can cause changes to the interrupt
assignments by changing which physical PCI slot a device is plugged
into.  Onboard devices you can't change of course.  Another way to
affect things is to disable onboard devices that aren't needed to free
up additional interrupts for use by PCI, for example parallel or serial
ports, or parhaps the PS/2 mouse port (irq 12) if you have joined this
millenium and moved to the far supperior USB mice.

I have seen indications that some video cards don't actually like
sharing their IRQ with certain types of devices.  Often moving the
conflicting device to another slot will resolve the issue with the video
card.

Interrupts used by standard devices:
0:timer (gotta have that one at least until HPET or APIC or something
else takes over that job.  Not likely to go away anytime soon though).

1:keyboard controller (if you have a usb keyboard, it might be possible
for this one to free up although I doubt it).

2:second interrupt controller cascade interrupt

3: second serial port (if you don't use serial, you can often free this
one by disabling the serial port in the bios).

4: first serial port (same as above).

5: sometimes used for a second isa parallel port, or older sound blaster
cards or older onboard audio (mainly anything isa based or emulating isa
hardware).  Can often be free for other uses.

6: Floppy controller.  Might be able to free it if you can disable the
floppy controller in the bios, although often you can't.  Not having any
floppy drives enabled does not mean that the floppy controller is not
still enabled.

7: first parallel port.  Can often be freed by disabling the parallel
port in the bios.

8: Real Time Clock.  Gotta have that one.

9: Usually free for use, often used for ACPI or other PCI related
hardware.  Sometimes used for video cards (although that was mainly back
in the EGA era).

10: Usually free for use by PCI

11: Usually free for use by PCI

12: PS/2 mouse port, if enabled.  Some systems can disable this port in
the bios freeing the interrupt for other uses.

13: FPU interrupt.  Legacy thing due to the fact the math processor used
to be a seperate chip and needed an interrupt to communicate with the
CPU to tell it when it needed attension.  Even though the FPU is now
integrated, the interrupt is still there for backwards compability with
old software I believe.

14: first IDE port.  If you don't have an IDE port, and you don't have
SATA pretending to be IDE (usually the case for many intel chipset
boards, especially if they need to be able to run windows), then you can
disable the port and free the IRQ.

15: second ide port.  Same as above.

I have run systems with ide disabled before that used PCI scsi for
all disks, and I used irq 15 for a parallel port on an isa card in
that machine.  It actually ran all PCI devices on a single interrupt
since I needed 4 serial ports and 2 parallel ports on that machine,
which was irq 14 since I had IDE disabled entirely at the time.

Machines which contain an APIC (advanced programmable interrupt
controller) can have interrupts above 15, and are much less likely to be
sharing interrupts in general.

> On my XP system, I found this while looking into the "System Information" 
> spawned by my Palm Desktop application. I find it rather odd that all these 
> devices are drawing on only one IRQ. 
> 
> In fact, I can't think offhand of anything I have that isn't going through IRQ 
> 9. Unless you count the on-board ATA controllers. I suspect that this is not 
> the operating system's fault. This is hardware-based.
> 
> I have a new hard drive that uses my PCI ATA card. When it is connected and 
> powered up, things work OK for a while, then I hear a "click" inside the box 
> (sounds like a hard drive "click"), then the operating system freezes 
> indefinitely (no mouse, no keyboard) until I reboot it. The obvious problem 
> here is with open files, and indeed many files got corrupted. Removing the hard 
> drive removed the freezing problem.
> 
> For anyone who knows more about hardware than me: does it actually sound like 
> an IRQ conflict (haven't heard of that problem in years). It appears that there 
> is an upper limit on the number of devices on the same IRQ until the system 
> starts to get confused.

Well a few ideas:

You overloaded your power supply by adding another drive
The driver for the IDE controller is buggy
The IDE controller is broken

--
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists





More information about the Legacy mailing list