Re: UBSAN: Undefined behaviour in mm/page_alloc.c

[Michal Hocko <mhocko@kernel.org>]

On Fri 09-11-18 18:41:53, Tetsuo Handa wrote:
> On 2018/11/09 17:43, Michal Hocko wrote:
> > @@ -4364,6 +4353,17 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, int preferred_nid,
> >  	gfp_t alloc_mask; /* The gfp_t that was actually used for allocation */
> >  	struct alloc_context ac = { };
> >  
> > +	/*
> > +	 * In the slowpath, we sanity check order to avoid ever trying to
> 
> Please keep the comment up to dated.

Does this following look better?

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 9fc10a1029cf..bf9aecba4222 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -4354,10 +4354,8 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, int preferred_nid,
 	struct alloc_context ac = { };
 
 	/*
-	 * In the slowpath, we sanity check order to avoid ever trying to
-	 * reclaim >= MAX_ORDER areas which will never succeed. Callers may
-	 * be using allocators in order of preference for an area that is
-	 * too large.
+	 * There are several places where we assume that the order value is sane
+	 * so bail out early if the request is out of bound.
 	 */
 	if (order >= MAX_ORDER) {
 		WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN));

> I don't like that comments in OOM code is outdated.
> 
> > +	 * reclaim >= MAX_ORDER areas which will never succeed. Callers may
> > +	 * be using allocators in order of preference for an area that is
> > +	 * too large.
> > +	 */
> > +	if (order >= MAX_ORDER) {
> 
> Also, why not to add BUG_ON(gfp_mask & __GFP_NOFAIL); here?

Because we do not want to blow up the kernel just because of a stupid
usage of the allocator. Can you think of an example where it would
actually make any sense?

I would argue that such a theoretical abuse would blow up on an
unchecked NULL ptr access. Isn't that enough?
-- 
Michal Hocko
SUSE Labs