Re: [PATCH] mm: slub: fix conversion of freelist_corrupted()

[Eugeniu Rosca <erosca@de.adit-jv.com>]

Hello Dongli,

On Thu, Aug 13, 2020 at 11:57:51PM -0700, Dongli Zhang wrote:
> On 8/11/20 5:46 AM, Eugeniu Rosca wrote:
> > Commit 52f23478081ae0 ("mm/slub.c: fix corrupted freechain in
> > deactivate_slab()") suffered an update when picked up from LKML [1].
> > 
> > Specifically, relocating 'freelist = NULL' into 'freelist_corrupted()'
> > created a no-op statement. Fix it by sticking to the behavior intended
> > in the original patch [1]. Prefer the lowest-line-count solution.
> > 
> > [1] https://lore.kernel.org/linux-mm/20200331031450.12182-1-dongli.zhang@oracle.com/
> > 
> > Fixes: 52f23478081ae0 ("mm/slub.c: fix corrupted freechain in deactivate_slab()")
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Dongli Zhang <dongli.zhang@oracle.com>
> > Cc: <stable@vger.kernel.org>
> > Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
> > ---
> >  mm/slub.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/mm/slub.c b/mm/slub.c
> > index 68c02b2eecd9..9a3e963b02a3 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -677,7 +677,6 @@ static bool freelist_corrupted(struct kmem_cache *s, struct page *page,
> >  	if ((s->flags & SLAB_CONSISTENCY_CHECKS) &&
> >  	    !check_valid_pointer(s, page, nextfree)) {
> >  		object_err(s, page, freelist, "Freechain corrupt");
> > -		freelist = NULL;
> >  		slab_fix(s, "Isolate corrupted freechain");
> >  		return true;
> >  	}
> > @@ -2184,8 +2183,10 @@ static void deactivate_slab(struct kmem_cache *s, struct page *page,
> >  		 * 'freelist' is already corrupted.  So isolate all objects
> >  		 * starting at 'freelist'.
> >  		 */
> > -		if (freelist_corrupted(s, page, freelist, nextfree))
> > +		if (freelist_corrupted(s, page, freelist, nextfree)) {
> > +			freelist = NULL;
> 
> This is good to me.
> 
> However, this would confuse people when CONFIG_SLUB_DEBUG is not defined.
> 
> While reading the source code, people may be curious why to reset freelist when
> CONFIG_SLUB_DEBUG is even not defined.

This is a fair point. To address it, the `freelist = NULL` assignment
should be then moved into the body of freelist_corrupted(). If no
concerns on that, I will soon push a v2 implementing this proposal.

-- 
Best regards,
Eugeniu Rosca