using gdb's 'kvm proc' to debug LWPs

[Richard Hansen]

Hi all,

I'd like some help examining LWP backtraces on a running NetBSD system 
in a qemu-kvm virtual machine.

I've successfully done the following to examine LWP backtraces in crash 
dumps:

   $ gdb /path/to/netbsd.gdb
   (gdb) target kvm /path/to/netbsd.0.core
   (gdb) dir /path/to/kernel/source/code
   (gdb) kvm proc 0xdeadbeef
   (gdb) bt full

I'm now trying to do something similar on a *running* NetBSD kernel in a 
qemu-kvm virtual machine (Linux host).  qemu has remote gdb stubs, which 
makes it great for stepping the kernel without having to deal with 
serial ports or rebuilding the kernel with KGDB enabled.

I'm able to successfully connect and do normal gdb things:

   $ kvm <args> -S -s
   $ gdb /path/to/netbsd.gdb
   (gdb) target remote localhost:1234
   (gdb) dir /path/to/kernel/source/code
   (gdb) bt full

So I can see the current LWP backtrace easily.  Unfortunately, the 
following fails:

   (gdb) kvm proc 0xdeadbeef
   No kernel memory image.

Does anyone know how I can examine the backtraces of other LWPs from a 
live kernel?  Does 'kvm proc' work with a KGDB setup?

Thanks,
Richard