Re: [PATCH] mm: fix lazy vmap purging (use-after-free error)

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Mon, Feb 23, 2009 at 10:27:44AM +0100, Ingo Molnar wrote:
> 
> * Andrew Morton <akpm@linux-foundation.org> wrote:
> 
> > On Mon, 23 Feb 2009 10:07:35 +0100 Ingo Molnar <mingo@elte.hu> wrote:
> > 
> > > > ---
> > > > 
> > > >  init/main.c         |    3 +++
> > > >  kernel/rcuclassic.c |    4 +++-
> > > >  kernel/rcutree.c    |    4 +++-
> > > >  3 files changed, 9 insertions(+), 2 deletions(-)
> > > > 
> > > > diff --git a/init/main.c b/init/main.c
> > > > index 8442094..51f4b71 100644
> > > > --- a/init/main.c
> > > > +++ b/init/main.c
> > > > @@ -121,6 +121,8 @@ static char *static_command_line;
> > > >  static char *execute_command;
> > > >  static char *ramdisk_execute_command;
> > > >  
> > > > +int idle_task_is_really_idle;	/* set to 1 late in boot. */
> > > > +
> > > >  #ifdef CONFIG_SMP
> > > >  /* Setup configured maximum number of CPUs to activate */
> > > >  unsigned int __initdata setup_max_cpus = NR_CPUS;
> > > > @@ -463,6 +465,7 @@ static noinline void __init_refok rest_init(void)
> > > >  	 * at least once to get things moving:
> > > >  	 */
> > > >  	init_idle_bootup_task(current);
> > > > +	idle_task_is_really_idle = 1;
> > > >  	preempt_enable_no_resched();
> > > >  	schedule();
> > > >  	preempt_disable();
> > > 
> > > Could you please use system_state instead? We could insert a new 
> > > stage - or just use SYSTEM_RUNNING as the trigger.
> > 
> > I think the standalone flag is better (once those 
> > extern-decls-in-C get fixed).
> > 
> > system_state's semantics have, err, evolved over time.  If 
> > this happens again (and the patch sneaks past my attention) 
> > then there's a risk that code which depends upon system_state 
> > will break - this has happened in the past.  Plus piling more 
> > dependencies on system_state of course makes any evolution of 
> > its semantics harder to do...
> 
> All we need is a SYSTEM_BOOTING_EARLY boundary - prior which 
> there's no real scheduling yet. I used SYSTEM_SCHEDULER_BOOTING 
> state before and it worked well and wasnt fragile.
> 
> Our system_state semantics problems were more rooted in the fact 
> that the SYSTEM_BOOTING stage wasnt well defined. But if we did 
> a SYSTEM_SCHEDULER_BOOTING stage that would be pretty 
> bit-rot-safe.

Actually, I should be able to simply use (system_state != SYSTEM_BOOTING),
because the system state transitions to SYSTEM_RUNNING at essentially
the same place that I set my new global variable.  Especially given that
this approach saves others from any bad guesses I might make about the
existing uses of SYSTEM_BOOTING.  ;-)

I will also credit kmemcheck on the resulting patch.

							Thanx, Paul