cvs commit: src/sys/dev/acpica acpi_quirks acpi_timer.c acpivar.h

[Nate Lawson]

John Baldwin wrote:
> On Friday 08 October 2004 01:56 pm, Nate Lawson wrote:
> 
>>njl         2004-10-08 17:56:47 UTC
>>
>>  FreeBSD src repository
>>
>>  Modified files:
>>    sys/dev/acpica       acpi_quirks acpi_timer.c acpivar.h
>>  Log:
>>  Update a quirk for the ASUS P5A to disable the timer.  It appears to work
>>fine with acpi but the timer runs twice as fast.  Note that the main
>>problem (system doesn't work properly with acpi disabled) should be fixed
>>separately.
> 
> 
> Actually, it's not really a problem that can be fixed.  $PIR and the actual 
> link devices return different capabilities as far as the list of possible 
> IRQs for each link device.  There's not much we can do to fix the fact that 
> according to ACPI, links 3 and 4 use IRQs 5 and 6 when $PIR says that neither 
> IRQ is valid for either link.

I've been analyzing how Windows and Linux handle IRQ routing.  There are 
some interesting parts that I've mentioned before but thought I'd 
summarize publically:

* Some systems, notably laptops, require all PCI irqs to be routed to 
the SCI (irq 9 almost always).  Sony VAIOs are one example.

* Both $PIR and _PRT are used for ACPI irq routing.  It merges them via 
some unknown algorithm (prefer $PIR?)

As for this particular system, perhaps it would be helped by a PCI 
quirk.  Also, was Kevin running the latest version of his BIOS?  Also, 
the slot the ethernet card was in might change things.

http://www.pcphotovideo.com/richa/page33.html

-- 
Nate