At Mon, 11 Nov 2002 14:26:46 +0000 (UTC), "Ben Harris" wrote:
> It's quite useful when you've got a kernel that's crashing after init's been
> started (so printfs in main() are no use) but before you can get a
> single-user shell (so ktrace isn't useful either).  It'd be nice if
> something analogous to SYSCALL_DEBUG could be preserved (though more things
> like UVM histories would be even better).

Indeed.  I just used it for that purpose like a week ago, maybe two...

A better mechanism would be nice, but "it got the job done well enough
at the time," and it really was meant to be pretty much (only) a "oh
boy, why is init croaking on me now" debugging mechanism.  8-)

It would be nice if we had a general facility (for debugging) to log
traps and syscalls and other "important system events" (e.g. forks,
switches, and for mips maybe cache flushes, ASID uses/rollover, etc.)


chris