On Thu, 2007-03-08 at 23:33 +0100, Rafael J. Wysocki wrote:

> > Unfortunately I won't be able to actually try this on hardware until the
> > 20th or so.
> 
> OK, it's not an urgent thing. ;-)

True :)

> Well, I don't think so.  If I understand the definition of system_state
> correctly, it is initially equal to SYSTEM_BOOTING.  Then, it's changed to
> SYSTEM_RUNNING in init/main.c after the bootmem has been freed.

No, I think you're confusing bootmem with initmem right now. If you
actually look at the code then free_all_bootmem is called as part of
mem_init() on powerpc, which is called from start_kernel() a long time
before initcalls are done and system state is set.

Put it this way. By the time initcalls are done, I can no longer use
bootmem. I tested this and it panics. But if you look at the code in
init/main.c, system_state is only changed after initcalls are done.

> Anyway, the patch works on x86_64. :-)

Yeah but it worked before too ;)

johannes