Re: Debugging kernel with gdb

Quoth Paul Lalonde <[email protected]>:
> I'm debugging Nix's reboot code.  The good news is that a straight reboot
> now works.  The bad news is that jumping to a new kernel does not.  It
> looks like it's dying in the reboot trampoline, which points to some
> mis-managed expectation about machine state (page table, gdt?).
> 
> I'm trying to debug this with qemu and gdb, but though the elf generated by
> 'mk 9pc64.elf' is enough to run gdb and connect to the qemu-hosted kernel,
> it has no debug symbols.  Operating with addresses gleaned from a .lst file
> is less than pleasant.
> 
> Is there a way to generate symbols for functions in the .elf?
> 
> Thanks,
>   Paul
> 

I don't believe there is.

It may be interesting to implement a /proc-compatible interface
that talks to gdb stubs for this kind of thing; then we could
point acid at any gdb remote server and go.