[net.unix-wizards] Debugging a driver

olmstead@sri-unix (10/28/82)

I'm looking for advice on interpreting a 4.1 kernel core dump (the kind
that ends up in places like /usr/crash/vmcore.#).  Most important to me
is a stack trace that will tell me how my driver got to the function
that called "panic."  I can get at all (most of ?) my global data with
adb(1), but getting the stack trace has thus far proved impossible.
I've looked at the conflicting advice given in the manual pages for
crash(8) and analyze(8), and a look at the code in locore.s tells me I
should ignore the mention of "dumpstack" in the manual page for crash.
Analyze also mentions doing something weird to adb's map registers, but
adb seems to do the right thing if left alone.  So, I guess my question
is:
	Given the stack pointer, frame pointer and other goodies
	stashed by doadump(), what, if anything, do I need to tell adb
	about virtual memory, frame registers, etc., to get the trace?
	And how do I do it?

			Not afraid to say "Thanks in advance,"

			Patrick Olmstead
			(...ucbvax!menlo70!sytek!olmstead)