ACPI regression on IBM x336, dual Xeon EM64T, FreeBSD 6.0-BETA5

[John Baldwin]

On Thursday 06 October 2005 09:02 am, Hroi Sigurdsson wrote:
> Hi.
>
> The new ACPI code in FreeBSD 6.0/amd64 is causing hard lock-ups on IBM
> x336 servers during boot, right after printing:
>
> Timecounter *mumble* frequency *mumble* Hz quality *mumble*
> Timecounters tick every 1.000 msec
>
> I haven't been able to gather more than that from 6.0, unfortunately due
> to KVM limitations.
>
> Setting hint.acpi.0.disabled=1 enables booting, but then the second CPU
> isn't found. BIOS version is 1.11 (newest).
> FreeBSD 5.4-STABLE does not have this problem (We went with 6.0 for the
> improved mpt driver, but decided to run 5.4 with the mpt driver from 6.0
> instead).
>
> Various debug output gathered from 5.4 is at
> http://quentin.asdf.dk/~hroi/freebsd/
>
> How can I help debug this further? I have another spare (single-CPU)
> x336 available to test with. I can hook it to serial console server, IP
> KVM and remote power switch if needed.

Can you try doing 'hint.apic.0.disabled=1' instead and leaving ACPI enabled 
and seeing if it works?  Also, are you able to capture the full dmesg from 
the broken boot using a serial console or some such?

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org