Re: Tyan Tiger 100 (S1832) IO-APIC Problems

Randy.Dunlap said:
> On Wed, 25 Feb 2004 00:35:55 -0600 Russ W. Knize wrote:
>
> | Just in case anyne might be interested, I have been trying to debug this
> | issue.  It seems that pirq_find_routing_table() is not locating the
> | table in the range it is looking.  I turned on debug trace and added
> | some other debug code.  The PIRQ signature is not there in that range at
> | all.  Without the table, the router is not identified.  I also tried
> | forcing ACPI to be enabled, but that didn't seem to help.
>
> I don't know anything about the Tyan Tiger 100, but the PIRQ table
> was an M$ creation for Win 98 or thereabouts.  ACPI interrupt
> routing has replaced it on some systems, so I don't find it
> surprising that the PIRQ table isn't found.  Of course, I guess
> that it could be there and Linux just isn't finding it, wherever
> it's hidden.  Tough to say which case it is.
>
> I also don't recall the particular problems that you are having.
> Is it IO-APIC problems (as in $Subject), or is it just interrupt
> routing in general?  If the latter, perhaps asking your questions
> on the Linux ACPI mailing list would lead to some answers for you.
> ([email protected])

It's hard to say.  I have to force ACPI to get it to work at all
(acpi=force).  I imagine this is due to the buggy ACPI implementations of
the era.  Even with ACPI enabled, the result is the same (no interrupts)
and I get lots of complaining about various things from the ACPI driver. 
ACPI complains about the same things with "acpi=force noapic", so I don't
think the complaints are related, but who knows.  The BIOS has lots of
vague and poorly documented options, but a "binary search" of twiddling
through these didn't yield anything useful.

With "noapic", the board seems to work OK, which is what is leading me
down the IO-APIC path.  I am wondering at what point the IRQs become
misrouted.

> There are APIC and IO-APIC datasheets available from Intel and
> probably other processor and chipset manufacturers, but they are
> sorely lacking in anything similar to a theory of operation.

Thanks for the pointers.  I'll have a look through the Intel 440BX docs. 
My 440LX board finds the table and the router correctly.  I see that the
table itself is discarded, but the router is somehow used later on when
the IO-APIC is to be used.  These are the bits I am trying to understand
now.  I am wondering if there is another way to identify the router.

Thanks again,
Russ