AMD FX-6100 boot pauses with ACPI

[Ian Clark <mrrooster%gmail.com@localhost>]

Hi All,

I've been having some issues with boot hanging on a new AMD 6 core
bulldozer box. (6 core FX6100 - AMD 880G chipset, gigabyte
GA-880GMA-USB3 with F4 bios.)

http://mail-index.netbsd.org/netbsd-users/2011/11/14/msg009560.html

I was able to keep it fairly stable by booting a 5.99.59 kernel
without ACPI, although newer kernels seem less stable in this regard.

I've reformatted / (as I had bootloader issues preventing me loading a
newer kernel, and couldn't get the newest bootloader to work.)
Although I am running the same contents of / (root is ffs2, hd is 500g
or so)

So I'm now running 5.99.59 userland (from 22nd November IIRC) and
associated bootloader, with a kernel from a few days ago (5.99.62)

I still get the boot pauses if I enable ACPI, and the machine, even if
booted -2 won't stay up under heavy load (I can get 1 or 2 builds of
the source tree in maybe if I'm lucky.)

However, tonight I tried fiddling with the BIOS a bit more, and
disabled HPET, C1E support and C6 support. This seemed to alleviate
the issue somewhat, and I was able to re-enable all of these /except/
C1E without issue.

If I re-enable C1E I get to 'root file system type: ffs', then I have
to press a key to continue booting, and shutdown is unreliable, as
previously, however with C1E disabled I can boot and reboot cleanly.
(ALthough I have seen a couple of crashes at the final stage of
reboot.)

I have also taken the opportunity in the reformat to create a swap
partition, so hopefully I should be able to provide crashdumps if
required, although the hangs seem to be that, I find the machine with
a blank screen on the console and keyboard unresponsive. (No CAPSLOCK
for example.)

I'm currently running a full /src build to see if it's more stable
with C1E disabled.

Hopefully this info is useful to someone? Is it best to submit a bug
report? (I'm unsure of the etiquette for HEAD?)

Cheers,

Ian