Re: ALSA via82xx fails since 2.6.2

From: Len Brown
Date: Fri Mar 12 2004 - 23:45:59 EST


On Fri, 2004-03-12 at 21:36, Jürgen Repolusk wrote:

> CPU0
> 0: 22589343 XT-PIC timer
> 1: 3716 XT-PIC i8042
> 2: 0 XT-PIC cascade
> 8: 2 XT-PIC rtc
> 9: 100000 XT-PIC acpi, uhci_hcd, uhci_hcd, yenta, eth1
> 10: 453565 XT-PIC yenta, eth0
> 11: 0 XT-PIC sonypi
> 12: 101345 XT-PIC i8042
> 14: 42157 XT-PIC ide0
> 15: 22 XT-PIC ide1
> NMI: 0
> LOC: 22589934
> ERR: 42084
> MIS: 0
>
>

curious that an audio driver does not show up on irq 5 where the device
claims to be...

> > how about when you boot with acpi=off or pci=noacpi?
> actually i gave it a try but it doesn't change anything. still the same error.

Thanks for confirming that this isn't an ACPI bug;-)

I think what's happening is that one of the devices on IRQ9 is pulling
on that line. ACPI is the 1st to register a handler on IRQ9, doesn't
know where the interrupts are coming from, and IRQ9 gets shut down
before the responsible device can register.

> > Are you sure you didn't see these messages before 2.6.2 -- was ACPI
> > enabled in the working release?
>
> Yes I'm sure that before 2.6.2 I did not see this message at all - sound was
> working real fine. with changin to 2.6.2 up to 2.6.4 now I've this problem.
>
> greets, jürgen

There are a couple of things you can do to debug.

boot with "noirqdebug" to treat the symptom. This will prevent IRQ9
from getting disabled. If a device registers and claims those
interrupts, then the system will function. If none does, then you'll
see IRQ9 in /proc/interrupts steadily climb and performance will be
poor.

remove your USB, yenta and eth1 hardware and see if one makes the issue
go away.

> ACPI: PCI Interrupt Link [LNKA] (IRQs 3 4 5 7 *9 10 11 12)

You can also use a feature in ACPI to distribute the interrupts and
perhaps isolate the offending device. Boot with...

acpi_irq_balance acpi_irq_pci=3,4,5,7 acpi_irq_isa=9

This should tell the code to balance interrupts across IRQs where
possible, tending towards 3,4,5,7 and avoiding 9; it already tends
towards 10.

cheers,
-Len


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/