Kernel panic with 2.6.15.1 on IBM 9076-N81

[Patrick Finnegan]

Ok.  I've figured some things out, working from 2.6.15.1:

First, the console doesn't get set up properly.  The call to 
disable_early_printk() fron console_init() turns off the console, and 
none of the other console routines init the serial console.

Second, there's a panic (I commented out the disable_early_printk call
so that I could see what was happening):

requesting IPIs ...
IPIs requested...
Processor 1 found.
Unable to handle kernel paging request for data at address 0x00000080
Faulting instruction address: 0xc0000000000394d4
Oops: Kernel access of bad area, sig: 11 [#1]
SMP NR_CPUS=32 NUM PSERIES
Modules linked in:
NC0000000000394D4 LR: C000000000039348 CTR: C00000000002A750
REGS: c00000000ff87b80 TRAP: 0300  Not tainted  (2.6.15.1-power3-smp)
MSR:  A000000000001032 <ME,IR,DR>  CR: 22000002  XER: 20000000
DAR: 0000000000000080, DSISR: 0000000002200000
TASK = c00000003ffbf0a0[0 'swapper' THREAD: c00000000ff841
GPR00: 0000000000000080 C00000000FF87E00 C00000000042E4A8 A000000000001032
GPR04: A000000000001032 0000000000000004 0000000000000F 0000000000000001
GPR08: 0000020 0000000000000000 D000080080204C00 0000000000000000
GPR12: C00000000FFEF000 C00000000036F400 0000000000000000 0000000000000000
GPR16: 0000000000000000 000000000000000 0000000000000000 0000000000000
GPR20: 0000000000000000 0000000000C00000 0000000000004000 0000000000000001
GPR24: 00000000000001 0000000000000000 0000000000000000 0000000000000000
GPR28: C000000000431CB8 0000000000000002 C00000000038F1C8 C000000002579000
NIP [C0000000000394D4] .mpic_setupthis_cpu+0x1d4/0x210
LR [C000000000039348] .mpic_setup_this_cpu+0x48/0x2
Call Trace:
[C00000000FF87E90] [C00000000002A760] .smp_mpic_setup_cpu+0x10/0x30
[C00000000FF87F00] [C00000000002AD1C] .start_se0x9c/0x150
[C00000000FF87F90] [C00000000000827C] .start_secondary_prolog+0xc/0x10
Instruction dump:
7c00eb78 7d6bfa14 79290020 e94b01e8 7d2b4b78 4082ff68 7c0907b4 7c0a5a14
7d20052c 7c0004ac 4bffff60 38090080 <7d60052c> 7c0004ac 60000000 60000000

(I had to reconstruct that from multiple retries of the same panic, since
the shared SP console tends to "eat" characters, and setting the Open Firmware
input-device and output-device variables doesn't actually redirect the
console.  Yes, I know it's still missing some characters.)

Pat
-- 
Purdue University Research Computing ---  http://www.rcac.purdue.edu/
The Computer Refuge                  ---  http://computer-refuge.org