Re: [patch 1/4] kref: warn kref_put() with unreferenced kref

[Greg KH <greg@kroah.com>]

On Tue, Apr 25, 2006 at 12:34:10PM +0800, Akinobu Mita wrote:
> On Mon, Apr 24, 2006 at 08:51:28PM -0700, Greg KH wrote:
> > > @@ -49,6 +49,7 @@ void kref_get(struct kref *kref)
> > >   */
> > >  int kref_put(struct kref *kref, void (*release)(struct kref *kref))
> > >  {
> > > +	WARN_ON(atomic_read(&kref->refcount) < 1);
> > 
> > How can this ever be true?  If the refcount _ever_ goes below 1, the
> > object is freed.
> 
> The idea of detection kref_put() with unreferenced object was stolen
> from BUG_ON()es in blocks/ll_rw_blk.c and fs/bio.c
> 
> ll_rw_blk.c:    BUG_ON(atomic_read(&ioc->refcount) == 0);
> 
> bio.c:          BIO_BUG_ON(!atomic_read(&bio->bi_cnt));
> 
> But the kref counter usually does not become zero. Because kref is
> trying to reduce the number of atomic_dec_and_test()
> 
> So this patch also set kref counter to zero here:
> 
> > > +	if (atomic_read(&kref->refcount) == 1)
> > > +		atomic_set(&kref->refcount, 0);

But again, if this happens the memory will be freed.  So again,
CONFIG_DEBUG_SLAB will catch this kind of thing.

Is there some reason you created these patches?  Were you trying to
debug something that was tracked down to a kref issue?

thanks,

greg k-h