For some time I've been chasing a problem on powerpc
with gdb hanging (both gdb5.0 and the new gdb53).
The machine ends up with a bad kernel stack
and dsitraps itself to death and remains difficult
to debug even with an ICE. The problem occurs after
gdb/ptrace() sets the new process to PT_STEP and the
new process starts to execute. (I'll try and hook
up a LA to get nearer the problem)

The problem occurred between 8/1 and 8/29 and since
I have to use my own platform code I wasn't sure if the
problem with my code or -current. By pulling out
changes by date I have finally eliminated my platform
changes (but you never know) and it seems confined to arch/powerpc/powerpc.
The mostly likely candidates are

   oea/pmap.c 1.11 -> 1.14
   powerpc/locore_subr.S 1.14 -> 1.18
   powerpc/trap_subr.S 1.42 -> 1.49

are other 32bit powerpc/oea people seeing this?  Any help
would be appreciated.

-Derek