Subject: Re: Boot blocks croak on kernels with full debug table
To: Craig Metz <cmetz@sundance.itd.nrl.navy.mil>
From: Chris G Demetriou <Chris_G_Demetriou@UX2.SP.CS.CMU.EDU>
List: port-i386
Date: 05/22/1996 12:06:58
> 	Even if it loads the debug symbols into memory and eats memory (in
> my case, using ~10MB of 16MB), it should boot and run the kernel (or at least
> give you an error message) instead of hard locking when it's supposed to be
> jumping to the kernel entry point. The kernel should then immediately reclaim
> the symbol table as usable memory. BSD/OS and 4.4-Lite on a SPARC can do
> this just fine.

As far as I know, 4.4-Lite didn't actually load symbols.  I doubt that
BSD/OS loads them, as well.

NetBSD actually does load the symbols into memory, and even keeps them
there if you have DDB enabled.

The 'kernelname' vs. 'kernelname.gdb' distinction came from CSRG.
Even debugging 4.4BSD systems, you shouldn't copy 'kernelname.gdb' to
/, you'd copy 'kernelname' to / and then use 'kernelname.gdb' later
with gdb when analyzing the running system or crash dumps.  (I know
this because i spent plenty of time  working with 'real' 4.4BSD on
development machines at Berkeley...  8-)


In any case, I agree with you that the boot block shouldn't
arbitrarily lose.  It should either work properly or determine that it
cannot, and if it makes that determination should do _something_ sane...



cgd