Skip to contentSkip to conversation

irq 15: nobody cared! with KT600 chipset and 2.6.0-test9

DISCUSSION Andi Kleen 4 participants 4 messages Nov 23, 2003 2 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Andi Kleen

author
#2 lore raw
Quoted 5 lines · Linus Torvalds
Quoted 3 lines · Marco d'Itri

This is an ASUS A7V600 motherboard, and the second IDE channel does not work at all. I verified that it works fine with a 2.4.x kernel.

Quoted 7 lines

Can you enable DEBUG in "arch/i386/pci/pci.h" and ACPI debugging and then report what the system says at bootup both with ACPI enabled, and with "pci=noacpi".

For some reason the system decides that your IDE controller is on irq3, which is clearly wrong. It's on irq15, but somebody told the PCI layer otherwise.

It's a long standing bug in how we handle VIA on board devices in ACPI. It was a big problem on x86-64 too until I cheated and used only PIC mode when there is a VIA southbridge.

Basically VIA has very specific rules on what interrupts can be assigned to the onboard devices and they use an unusual ACPI construct for that

See http://bugme.osdl.org/show_bug.cgi?id=1530 for some more details. (the same problem applies to 2.6 afaik)

Here's also some older description from VIA about the issues I saved:

Quoted 1 line
Quoted 1 line

1) The _SRS method can only cope with IRQs <= 15 if SB is before VT8233. After VT8235, _SRS method need not report IRQ for internal devices due to they are
+hardwired to fixed IRQ in APIC mode.

[AK: that is one problem with the ACPI code - it doesn't handle that the _SRS report doesn't report it. Even ignoring the problems with older SBs which would need some special handling code]

The BIOS returns fixed IRQs in the _PRS method when system is in APIC mode if SB is after
+VT8235.
(IDE-IRQ20, USB-IRQ21, Audio and Modem-IRQ22, LAN-IRQ23)

2) The _CRS method is hardcoded to return 0 ALWAYS for link devices in APIC mode. We feel strange for this situation. This control method will check internal device's
+interrupt line (Rx3C of PCI configuration) to return current IRQ setting.
Will it be possible in Linux kernel driver which will reset the Rx3C to zero ?
<<

Forcing PIC mode will at least make it work. Of course it would be better to fix the ACPI code to handle all this properly.

-Andi

#3 lore raw
Quoted 4 lines · Andi Kleen

It's a long standing bug in how we handle VIA on board devices in ACPI. It was a big problem on x86-64 too until I cheated and used only PIC mode when there is a VIA southbridge.

That doesn't explain the lack of autodetection, and the early irq15 registration.

That would explain why no interrupts make it at all, but why do we not even probe for irq15 here?

Linus

#4 lore raw
Quoted 7 lines · Linus Torvalds
Quoted 4 lines · Andi Kleen

It's a long standing bug in how we handle VIA on board devices in ACPI. It was a big problem on x86-64 too until I cheated and used only PIC mode when there is a VIA southbridge.

Quoted 5 lines

That doesn't explain the lack of autodetection, and the early irq15 registration.

That would explain why no interrupts make it at all, but why do we not even probe for irq15 here?

It's easy to test. does it work when booted with "noapic" ?

-Andi

#5 lore raw
Quoted 8 lines · Andi Kleen
Quoted 7 lines · Linus Torvalds
Quoted 4 lines · Andi Kleen

It's a long standing bug in how we handle VIA on board devices in ACPI. It was a big problem on x86-64 too until I cheated and used only PIC mode when there is a VIA southbridge.

Quoted 7 lines
Quoted 5 lines

That doesn't explain the lack of autodetection, and the early irq15 registration.

That would explain why no interrupts make it at all, but why do we not even probe for irq15 here?

It's easy to test. does it work when booted with "noapic" ?

Yes, I checked again.

Signature · 2 lines
ciao, |
Marco | [3232 ma6tjQg3wBT1w]