Subject: Re: boot crash with GENERIC
To: Matt Thomas <matt@3am-software.com>
From: Monroe Williams <monroe@pobox.com>
List: port-macppc
Date: 02/12/2003 11:42:32
on 2/12/03 8:38 AM, Matt Thomas at matt@3am-software.com wrote:
> At 01:59 AM 2/12/2003, Monroe Williams wrote:
>> on 2/11/03 1:34 PM, Emmanuel Dreyfus at manu@netbsd.org wrote:
>>
>>> Anyone has seen that one, or is it only at mine? I get it with latest
>>> GENERIC kernel from releng.netbsd.org
>>>
>>> total memory = 65536 KB
>>> avail memory = 51684 KB
>>> using 844 buffers containing 3376 KB of memory
>>> no active packagepanic: no cpu node
>>> Stopped in pid 0.1 (swapper) at cpu_Debugger+0x10: lwz r0, r1,
>>> 0x14
>>> db> trace
>>> 0x005e75b0: at panic+18c
>>> 0x005e7670: at calc_delayconst+f4
>>> 0x005e76b0: at cpu_configure+10
>>> 0x005e76c0: at configure+54
>>> 0x005e76d0: at main+4d8
>>> 0x005e7730: at __start+98
>>
>> I get exactly the same thing from a GENERIC kernel built with source
>> retrieved from anon CVS the last several days.
>>
>> It might be relevant that this is on an OF 1.0.5 machine.
>
> Did it generic boot before? Can I get a copy of your of your OFW tree?
Yes and yes. (See below.)
>> Kernels with my usual -current config are even worse -- they get part-way
>> through the second-stage boot loader counts and die with
>>
>> DEFAULT CATCH!, code=FFF00700
>>
>> I haven't yet narrowed down what the difference between these two is, but my
>> config isn't that far from GENERIC. Perhaps the size of the kernel -- mine
>> are a bit bigger. Actually, I just noticed that the bad ones are over 5MB.
>> Hmmm...
>
> Note that ofwboot starts at 6MB and given the 1MB starting addressing,
> anything over 5MB is bad news. You can rebuild ofwboot use a larger address.
I thought that might be the case. I can probably get rid of some drivers
and get the size down below 5M. I think adding raidframe was the last
straw.
>> I still have a copy of the source tree from sometime before the big CVS
>> repository rearrangement sitting around, and I can boot kernels built from
>> it just fine. (They come out labelled "NetBSD 1.6K" if that means
>> anything...)
>
> Oh. If your kernels have pseudo-device openfirm, grab
> ftp://ftp.netbsd.org/pub/incoming/matt/ofdump.c and then compile
> and run it as ofdump -l and send me the output.
-----
[monroe@macppc temp]$ ./ofdump -l
ofdump: OFIOCGET(3, "driver,AAPL,MacOS,PowerPC"): Bad address
ofdump: OFIOCGET(3, "driver,AAPL,MacOSX,PowerPC"): Bad address
[Caching 60 nodes and 456 properties]
ff828700: /device-tree
ff828f80: /PowerPC,60?@0
ff829878: /chosen
ff8299a8: /memory@0
ff829af0: /openprom
ff829bb0: /AAPL,ROM@ffc00000
ff829dc8: /options
ff82a9d8: /aliases
ff82ac18: /packages
ff82aca0: /deblocker
ff82b4a0: /disk-label
ff82b9e0: /obp-tftp
ff82de20: /mac-files
ff82e618: /mac-parts
ff82ed78: /aix-boot
ff82f1f0: /fat-files
ff8307c0: /iso-9660-files
ff831108: /xcoff-loader
ff831ac8: /terminal-emulator
ff831b60: /bandit@f2000000
ff832d50: /gc@10
ff833188: /53c94@10000
ff834a10: /sd
ff835640: /st
ff8362b8: /mace@11000
ff837130: /escc@13000
ff837288: /ch-a@13020
ff837938: /ch-b@13000
ff837fe8: /awacs@14000
ff8380d0: /swim3@15000
ff8391d8: /via-cuda@16000
ff839d68: /adb
ff839e58: /keyboard@0
ff83a5a8: /mouse@1
ff83a658: /pram@0
ff83a708: /rtc@0
ff83abd0: /power-mgt@0
ff83acf0: /mesh@18000
ff83c858: /sd
ff83d488: /st
ff83e190: /nvram@1d000
ff8401c0: /pci106b,1@b
ff840398: /pci-bridge@d
ff841000: /pci16b8,12@0
ff841270: /pci16b8,12@0,1
ff8414e0: /pci16b8,21@0,2
ff841750: /pci16b8,1@1
ff8419e8: /UltraTek100+@2
ff86bbc8: /FrmTk-0@10,18182c0
ff86c028: /sd
ff86c698: /FrmTk-1@11,18182c0
ff86caf8: /sd
ff882250: /pci-bridge@f
ff882eb8: /pci1186,1112@4
ff883180: /pci1186,1112@5
ff883448: /pci1186,1112@6
ff883710: /pci1186,1112@7
ff83e2c8: /chaos@f0000000
ff883ca8: /control@35800,0,0
ff885060: /planb@36800,0,0
ff83f258: /hammerhead@f8000000
[monroe@macppc temp]$
-----
For those keeping score, that's a Power Mac 7500 with a G3 upgrade, a Sonnet
Tempo Trio (ATA/FireWire/USB2) in slot D, and a D-Link DFE-570TX (4-port
Ethernet) in slot F.
-- monroe
------------------------------------------------------------------------
Monroe Williams monroe@pobox.com