Re: [PATCH] mm, uprobes: fix multiple free of ->uprobes_state.xol_area

[Eric Biggers]

Hi Oleg,

On Wed, Aug 30, 2017 at 06:37:14PM +0200, Oleg Nesterov wrote:
> On 08/29, Eric Biggers wrote:
> >
> > --- a/kernel/events/uprobes.c
> > +++ b/kernel/events/uprobes.c
> > @@ -1262,8 +1262,6 @@ void uprobe_end_dup_mmap(void)
> >  
> >  void uprobe_dup_mmap(struct mm_struct *oldmm, struct mm_struct *newmm)
> >  {
> > -	newmm->uprobes_state.xol_area = NULL;
> > -
> >  	if (test_bit(MMF_HAS_UPROBES, &oldmm->flags)) {
> >  		set_bit(MMF_HAS_UPROBES, &newmm->flags);
> >  		/* unconditionally, dup_mmap() skips VM_DONTCOPY vmas */
> > diff --git a/kernel/fork.c b/kernel/fork.c
> > index cbbea277b3fb..b7e9e57b71ea 100644
> > --- a/kernel/fork.c
> > +++ b/kernel/fork.c
> > @@ -785,6 +785,13 @@ static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
> >  #endif
> >  }
> >  
> > +static void mm_init_uprobes_state(struct mm_struct *mm)
> > +{
> > +#ifdef CONFIG_UPROBES
> > +	mm->uprobes_state.xol_area = NULL;
> > +#endif
> > +}
> > +
> >  static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
> >  	struct user_namespace *user_ns)
> >  {
> > @@ -812,6 +819,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
> >  #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
> >  	mm->pmd_huge_pte = NULL;
> >  #endif
> > +	mm_init_uprobes_state(mm);
> 
> ACK, but I have cosmetic nit, this doesn't match other uprobe helpers.
> 
> I'd suggest to add uprobe_init_state() into kernel/events/uprobes.c and
> the dummy !CONFIG_UPROBES version into include/linux/uprobes.h.
> 
> Not that I think this will be more clean, personally I would simply add a
> ifdef(CONFIG_UPROBES) line into mm_init(), but this will be more consistent.
> 
> Oleg.
> 

I was going to do that originally, but I didn't like that it wouldn't be inlined
despite being just a single pointer assignment.  And mm_init_uprobes_state() is
at least consistent with the other mm_init_*() helpers in kernel/fork.c.  I
don't feel too strongly though, and it's true that nothing outside of uprobes.h
and uprobes.c references xol_area currently, so maybe it would be clearer to
have it there.

Eric