Re: init receiving SIGILL on XEN3_DOM0/amd64

[James Laird <abrasive%axdf.net@localhost>]

On Sun, May 18, 2008 at 02:38:16PM +0200, Manuel Bouyer wrote:
> On Sun, May 18, 2008 at 09:32:32PM +1000, James Laird wrote:
> > On Sun, May 18, 2008 at 01:14:34PM +0200, Manuel Bouyer wrote:
> > > On Sun, May 18, 2008 at 01:09:54PM +0200, Christoph Egger wrote:
> > > > What are the normal conditions where the kernels sends SIGILL to 
> > > > userland?
> > > 
> > > When the CPU sends an illegal intruction trap, I guess.
> > 
> > Turning on debugging in arch/amd64/amd64/trap.c (TRAP_SIGDEBUG) would 
> > confirm that:
> > 
> > pid 1 (init): ILL at rip 7f7ffde03670 addr 7f7fffffffd8
> > rip 0x7f7ffde03670  rsp 0x7f7fffffde98  rfl 0x202
> 
> Could you set a breakpoint just after this printf (or add a call to
> Debugger()), and see what is at the rip ?
> The output of both
> x/i 7f7ffde03670
> and
> x/x 7f7ffde03670 (and subsequent addresses, just press enter to get the
> next value) could be usefull.
> 
> If you have a working amd64 kernel, you can also compare this with what
> gdb /sbin/init says for this address.

Okay, now some interesting things manifest.

db> x/i 7f7ffde03670
0x7f7ffde03670: invalid address
db> x/x 7f7ffde03670
0x7f7ffde03670: 10ec8348
db> x/i 7f7ffde03670
0x7f7ffde03670: subq    $0x10,%rsp

Is that... normal?