Re: [PATCH 3/3] slub: add hooks for kmemcheck

[Christoph Lameter <clameter@sgi.com>]

On Sat, 19 Apr 2008, Vegard Nossum wrote:

> We set it to zero to allow the compiler to optimize out code if
> CONFIG_KMEMCHECK=n. For instance
> 
>         if (kmemcheck_page_is_tracked(page) && !(s->flags & SLAB_NOTRACK)) {
> 
> Will be turned into if(0 && !(s->flags & 0)), which can be completely
> optimized away. On the other hand, if SLAB_NOTRACK is non-zero, this
> is not the case, since the flag must be tested.

The compiler will optize it again regardless of SLAB_NOTRACK. if (0 && 
whatever) is still 0.

> >  Should this not go into some kmemcheck.h file?
> 
> The implementations of these prototypes are in mm/slub_kmemcheck.c.
> They are only ever called from slub.c since they represent the
> interface between SLUB and kmemcheck.
> 
> Would you rather have it in include/linux/slub_kmemcheck.h?

Well lets see what Pekka says.

> Currently, with SLAB/SLUB debugging enabled, the page flag will be
> cleared, then accessed in check_object(), thus generating a few
> accesses to the page. It is safer to delay the clearing of the page
> flag until no more accesses can be generated for the page.

Hmmmm.. Okay that is already dealt with by a patch in Pekka's tree.