[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: Native build...
On Thu, 13 Mar 2008, Michael L. Hitch wrote:
Tom Ivar Helbekkmo <tih%Hamartun.Priv.NO@localhost> writes:
I deliberatedly quit a long-running 'top', whose parent shell had been
swapped out. This now caused the VAX to drop to the chevron prompt,
claiming an 'ISP ERR'.
...and the PC tells me that it happened in /sys/arch/vax/vax/trap.c, at
line 125 of version 1.112 of that file:
l = curlwp;
KASSERT(l != NULL);
==> p = l->l_proc;
KASSERT(p != NULL);
I finally got my 4000/90A connected back up, and got a current kernel
built. I did a native "build.sh ... tools" sucessfully, and then tried
a native "build.sh ... kernel=GENERIC" and died with the "ISP ERR"
right after it tried (and failed) to link the kernel. I fixed the kernel
config file (commented out the nmi and xmi stuff), and ran a new build
and it died with "ISP ERR" this morning. I finally started looking at it
and found it died at the same place.
I was able to dump some of the interrupt stack and found the faulting
address multiple times.
What appears to be happening is that it traps, tries to access l->l_proc
and traps again until the interrupt stack overflows. I figured out that
curlwp is actually the SSP (which is supposed to be the address of the
current lwp), but I hung the machine before I could dump the SSP. I was
not able to get a crash just relinking the kernel, so now I'm starting
a full kernel build again and hope to see what SSP is at that point.
Matt and I may have this fixed now. A simple program I wrote to
allocate and write to a big chunk of memory and force swapping seems to
survive now. I'm starting a full native kernel build, but won't know if
it works for several hours. The fix is revision 1.159 of
Michael L. Hitch mhitch%montana.edu@localhost
Information Technology Center
Montana State University Bozeman, MT USA
Main Index |
Thread Index |