cvs commit: src/sys/alpha/alpha support.s src/sys/i386/i386 swtch.s src/sys/kern kern_shutdown.c src/sys/sys systm.h

[Nate Lawson]

On Tue, 20 Jan 2004, Mike Silbersack wrote:
> On Tue, 20 Jan 2004, Bruce Evans wrote:
> <much deleted>
> > This can be obtained by running a debugger on the panic dump.  Line
> > numbers and source file names can also be printed by the debugger
> > if the executable has at least line numbers in its debugging info.
>
> True, if you can convince the reporter to enable crashdumps and recreate
> the panic... having even roughly accurate line numbers would increase the
> usefulness of PRs quite a bit.

Except when crashdumps don't work or are 1 GB+.  There are a few things we
need:

* Fix crash dumps.  I can't get a dump on my laptop, ATA.

* Sparse crash dumps.  We don't need non-anonymous pages, for instance.

* Fix GDB backtrace output (patch posted by gallatin@ and obrien@ may be
working on integrating it.)

* Fix backtrace() so the output goes to the console log.  Right now, we
get output that needs to be hand-transcribed or use a serial console.

* Clean up the overly verbose page fault code.  Maybe put the full output
under some VERBOSE option.  All you need is the PC, trap type, faulting
address, and curproc.  Everything else is better gotten through DDB or a
dump.

* Fix printf on SMP.  This combined with overly verbose page fault code
often creates scrambled console output.

There are some things that would be nice:

* Dumps over Ethernet

* GDB over Ethernet

We don't need file/line added to the panic message since panic messages
are unambiguous.

-Nate