"Eric S. Raymond" <[EMAIL PROTECTED]> writes:
> Eric W. Biederman <[EMAIL PROTECTED]>:
> > In this case the dual athlon motherboard the Tyan Thunder K7/ Tyan
> > Guiness/ what ever other names it has, has 4 pci interrupt lines that
> > feed into the AMD766 southbridge. There is a 2 byte routing register in
> > subfuction 3, offset 0x56, that has a set of mappings from 4 irq pins
> > to isa interrupts. The stock BIOS mapping is:
> >
> > PIRQA -> 10
> > PIRQB -> 5
> > PIRQC -> 3
> > PIRQD -> 11
> >
> > When I looked nothing else uses those ISA interrupts, so having a
> > conflict would suprise me.
This teaches to trust to my memory. IRQ 3 should have conflicts...
Except that there is only one external serial port so if the other
internal serial port is disabled we are probably o.k.
> Here's what we saw with the magic setting "Use PCI Interupt Entries in MP
> table" off, SB Live! in slot 4:
>
> 3 Serial port 1
> 4 Serial port 0
> 5 SCSI controller 1
> 7 Parallel port
> 10 SCSI controller 2, Radeon
> 11 USB controller, SB Live!, NIC #2.
> 15 NIC #1
>
> Moving the SB! Live to slot 5 changed its IRQ from 11 to 5. We had
> lockups with this configuration whether the card was in slot 4 or 5.
>
> > As far as I can tell all the "Use PCI Interupt Entries in MP table"
> > does when enable is:
> >
> > In the ioapic and probably the i82559 disable interrupts 3,5,10,11.
> > Assign irqs with higher numbers. (I.e. the pci irq field)
> > In the ioapic enable pins 0x10 - 0x13 and use the higher numbered
> > interrupts.
> > List these interupts in the MP table.
> >
> > And when "Use PCI Interrupt Entries in MP table" is disabled:
> >
> > In the ioapic and the i82559 enable interrupts 3,5,10,11.
> > Assign irqs to the devices with the low numbers.
> > In the ioapic disable pins 0x10 - 0x13.
> > Don't list the higher alias in the MP table.
> >
> > So I don't see how that can be your problem, unless the i82559
> > acknowledgement logic is locking you up or you moved the problematic
> > card to a different slot.
>
> When we set the "ISA interrupts in MP table" option in the BIOS, with
> the SB Live! in slot 5, we saw this.
>
> PIRQA -> 16 Adaptec #1, SB Live!
> PIRQB -> 17 Adaptec #2, Radeon
> PIRQC -> 18 NIC #1
> PIRQD -> 19 NIC #2, USB controller
>
> No lockups with this configuration.
O.k. So without "PCI interrupts in MP table" option you had:
> 3 Serial port 1
> 4 Serial port 0
> PIRQA -> 5 SCSI controller 1, SB Live!
> 7 Parallel port
> PIRQB -> 10 SCSI controller 2, Radeon
> PIRQD -> 11 USB controller, NIC #2.
> PIRQC -> 15 NIC #1
And there are only the four lines on the board. Hmm. I would have
to see a dump from lspci -s 7.3 and see bytes 0x56 & 0x57 but from
what I see here it looks like all the world like PIRQC that NIC #1 uses
is misreported in the table. And since the hang happened when
you logged in with ssh that make some sense.
Do you really have a second serial port? There is hardware that can
implement one but I haven't seen that hardware configuration. Hmm. And you
don't have irqs 14 & 15 mapped to ide. That is different.
O.k. So for the moment I at least have a reasonable hunch why you were
having lockups. I wouldn't suprise me if sharing an irq with an onboard
device would cause problems.
Eric
