This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [commit] Fix OpenBSD/i386 and OpenBSD/amd64 kernel trapframe unwinders
On Thu, Dec 22, 2005 at 04:38:14PM +0100, Mark Kettenis wrote:
> Yup. There's a seperate entry point for every interrupt vector (hence
> the strncmp() for "Xintr") and on top of that, interrupts can come
> from both userland and from within the kernel. I only want to
> terminate the backtrace when the interrupt came from userland.
That's no worse than for Linux. This is actually all solvable with
DWARF-2, and in fact, George Anzinger wrote bits to do it for Linux.
They're a bit disgusting, though (and I hope he'll agree if he's
reading this :-)
http://source.mvista.com/~ganzinger/common_kgdb_cfi_annotations.patch
He makes use of dwarf2 branches; I'm not sure if he ended up needing an
extra physical stack frame anywhere, but I'm convinced it's possible to
do without that.
--
Daniel Jacobowitz
CodeSourcery, LLC