Via PCI IRQ routing problem related? (was: PCI IRQ routing problem in 2.4.0)

[Pete Toscano <pete@research.netsol.com>]

[-- Attachment #1: Type: text/plain, Size: 4363 bytes --]

hmmm, would these sis-related pirq problems be related to the current
problems lots of people with a via chipset (at least the apollo pro 133a
chipset) and an smp-enabled kernel are seeing?  currently, people with
this chipset and an smp-enabled kernel have to disable apic if they wish
to use usb.  i've been told a few times that it's a problem with pci irq
routing, but have been able to find a fix.  reports of this problem pop
up every-so-often on the linux-usb list.

here's my dump_pirq output:

Interrupt routing table found at address 0xfdb50:
  Version 1.0, size 0x00a0
  Interrupt router is device 00:07.0
  PCI exclusive interrupt mask: 0x1c20 [5,10,11,12]
  Compatible router: vendor 0x1106 device 0x0596

Device 00:0f.0 (slot 1): FireWire (IEEE 1394)
  INTA: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:10.0 (slot 2): SCSI storage controller
  INTA: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:11.0 (slot 3): Multimedia audio controller
  INTA: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:12.0 (slot 4): Unknown mass storage controller
  INTA: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:13.0 (slot 5): Ethernet controller
  INTA: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:14.0 (slot 6): 
  INTA: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:01.0 (slot 0): PCI bridge
  INTA: link 0x01, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTB: link 0x02, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTC: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Device 00:07.0 (slot 0): ISA bridge
  INTC: link 0x03, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]
  INTD: link 0x05, irq mask 0xdeb8 [3,4,5,7,9,10,11,12,14,15]

Interrupt router at 00:07.0: VIA 82C596 PCI-to-ISA bridge
  PIRQA (link 0x01): irq 11
  PIRQB (link 0x02): irq 5
  PIRQC (link 0x03): irq 10
  PIRQD (link 0x05): irq 12

any ideas?  i also saved the lspci -vvvxxx and dmesg output if that'll
be helpful.

thanks,
pete



On Sun, 28 Jan 2001, Linus Torvalds wrote:

> On Sun, 28 Jan 2001, Tim Hockin wrote:
> > 
> > In reading the PIRQ specs, and making it work for our board, I thought
> > about this.  PIRQ states that link is chipset-dependant.  No chipset that I
> > have seen specifies what link should be.  So, as this case demonstrates, it
> > may be 'A' - the value the chipset expects, or 1, the logical index.
> > Either one makes sense, assuming the PIRQ routing code knows what link
> > means.  Here we see two BIOS vendors/versions that apparently do it
> > differently for the same chipset.    Grrr.
> 
> They _may_ do the same thing for the same chipset, it's just that we don't
> know exactly what that "same" thing is.
> 
> Ok, I want to see what people have. ANYBODY who has a SiS chipset, please
> take 5 seconds to do this as root (yes, you need to be root):
> 
> 	dump_pirq | mail -s "dump_pirq" torvalds@transmeta.com

-- 
Pete Toscano    p:sigsegv@psinet.com     w:pete@research.netsol.com
GPG fingerprint: D8F5 A087 9A4C 56BB 8F78  B29C 1FF0 1BA7 9008 2736

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]