Subject: Re: kern/10228: ktrace -c panics and freezes the system
To: John Hawkinson <jhawk@MIT.EDU>
From: Andreas Wrede <email@example.com>
Date: 05/29/2000 17:18:50
On Mon, 29 May 2000, John Hawkinson wrote:
> | [....]
> | uvm_fault(....) -> 1
> | fatal page fault in supervisor mode
> | trape 6 code 2 eip .....
> | oanix: trap
> | Begin traceback...
> | trap() at tarp+0x22d
> "oanix"? "tarp"?
> are these consistent typos or parity bozons for "panic" and "trap"?
Sorry, they are typos. The same couple of lines repeat over and over
> There appear to be three independant problems:
> a) your original page fault failure; no clue on this. New ktrace code
> from Bill Sommerfeld?
I don't know what the original trap is is - it scrolled of the
screen. If need be, I can set up with serial console and repeat..
> b) Recursive traceback printing. This is my fault, and I'll scratch
> my head and decide how best to deal with it. Perhaps we should
> not try to print tracebacks recursively.
> c) System freezes after 5 traps. No clue here.
Could be more - that's how many fit on the 80x24 screen.
> I presume this is reproducable.
Yes. On two different machines. You don't even to turn ktracing on
first, ie. a 'ktrace -c -p xxx' will do it.