Subject: Sparc-LX running 1.6 crashing ("lockmgr: no context")
To: None <port-sparc@netbsd.org>
From: Volker Borchert <v.borchert@vistecprivat.de>
List: port-sparc
Date: 04/19/2003 08:32:34
Hi all,

I am having serious trouble with a SS-LX I bought for use as DSL router
for my home network, with filtering and NAT/PAT. It is equipped with 24 MB
memory, an IBM DPES-31080 disk, a 501-2015 FSBE/S as outside interface.
It is running plain 1.6.

I set it up by copying over stuff from the SS-10 that temporarily did the
job. (Why change a working system? The SS-10 is overkill for the job, and
I want it for other things that can make use of its horsepower.)

Things seemed to work - bind, sendmail/fetchmail, newsx. But whenever I
open Netscape and go ahead brwosing, it goes down almost immediately and
bites the dust, sometimes so bad that it corrupts the NVRAM. A typical
console log (cut'n'paste from xterm serial console) begins with

cpu0: NMI: system interrupts: 40000000<VME=0,SBUS=0,ME>
panic: lockmgr: no context
syncing disks... 9 9 9 9 8 8 8 8 8 8 8 8 8 8 8 8 8 7 7 7 7 7 giving up
Frame pointer is at 0xf01e45a0

It then tries to dump core without success:

dumping to dev 7,1 offset 288247
dump data fault: pc=0xf018a32c addr=0x4 sfsr=126<PERR=0,LVL=1,AT=1,FT=1,FAV,OW>
panic: kernel fault
Frame pointer is at 0xf01e41d8

I went almost crazy over this. Removed first memory bank, shifted up the
other two. Re-seated everything. Checked kernel config for SS-10 specifics
left in and SS-LX specifics left out. Tried with a GENERIC kernel. No
change at all.

I worked around the problem by (ab)using that SS-10 for the job again.
I moved the disk, edited /etc/ifconfig.* to reflect use of hme0/le0 as
inside/outside interfaces rather than le0/le1, and voila. No trouble
with this box. I have no other SS-LX to check whether this one has some
subtle hardware problem.

Has anyone seen the above symptoms on any sun4m machine? On an LX? (From
what I have read on this list, the "no context" panic would indicate
problems with multiprocessing. But the LX definitively has only one CPU.)

Is it possible that the GENERIC kernel and the one I built myself on the
SS-10 both use some SuperSPARC{,-II}-only features? Or that the sun4m
support is incomplete as far as the microSPARC-II / Sunergy is concerned?

Thanks in advance for any hints,

	Volker

PS: Complete console traces, kernel config, etc. available upon request.