Re: [BUG] Use of probe_kernel_address() in task_rcu_dereference() without checking return value

[Oleg Nesterov <oleg@redhat.com>]

On 09/02, Peter Zijlstra wrote:
>
> On Mon, Sep 02, 2019 at 03:40:03PM +0200, Oleg Nesterov wrote:
> > diff --git a/include/linux/sched.h b/include/linux/sched.h
> > index 8dc1811..1f9b021 100644
> > --- a/include/linux/sched.h
> > +++ b/include/linux/sched.h
> > @@ -1134,7 +1134,10 @@ struct task_struct {
> >  
> >  	struct tlbflush_unmap_batch	tlb_ubc;
> >  
> > -	struct rcu_head			rcu;
> > +	union {
> > +		bool			xxx;
> > +		struct rcu_head		rcu;
> > +	};
> >  
> >  	/* Cache last used pipe for splice(): */
> >  	struct pipe_inode_info		*splice_pipe;
> > diff --git a/kernel/exit.c b/kernel/exit.c
> > index a75b6a7..baacfce 100644
> > --- a/kernel/exit.c
> > +++ b/kernel/exit.c
> > @@ -182,6 +182,11 @@ static void delayed_put_task_struct(struct rcu_head *rhp)
> >  	put_task_struct(tsk);
> >  }
> >  
> > +void call_delayed_put_task_struct(struct task_struct *p)
> > +{
> > +	if (xchg(&p->xxx, 1))
> > +		call_rcu(&p->rcu, delayed_put_task_struct);
> > +}
> 
> I think this is the first usage of xchg() on _Bool, also not all archs
> implement xchg8()

I didn't even notice I used "bool" ;)


speaking of the users of task_rcu_dereference(), membarrier_global_expedited()
does

		rcu_read_lock();
		p = task_rcu_dereference(&cpu_rq(cpu)->curr);
		if (p && p->mm && (atomic_read(&p->mm->membarrier_state) &
				   MEMBARRIER_STATE_GLOBAL_EXPEDITED)) {

This asks for READ_ONCE, but this is minor. Why can't p->mm be freed?

I guess it is fine to read the garbage from &p->mm->membarrier_state if we race
with the exiting task, but in theory this looks unsafe if CONFIG_DEBUG_PAGEALLOC.

Another possible user of probe_slab_address() or I am totally confused?

Oleg.