On Mon, 2 Dec 2013, Greg KH wrote: > On Mon, Dec 02, 2013 at 04:33:20PM +0000, Christoph Lameter wrote: > > On Wed, 27 Nov 2013, Greg KH wrote: > > > > > Just make the kobject "dynamic" instead of embedded in struct kmem_cache > > > and all will be fine. I can't believe this code has been broken for > > > this long. > > > > The slub code is was designed to use an embedded structure since we > > only get the kobj pointer passed to us from sysfs. If kobj is not > > embedded then how can we get from the sysfs object to the kmem_cache > > structure from the sysfs callbacks? Sysfs was designed to have embedded > > objects as far as I can recall. > > Yes, it's designed to have embedded objects, so then use it that way and > clean up the structure when the kobject goes away. Don't use a > different reference count for your structure than the one in the kobject > and think that all will be fine. We need our own reference count. So we just have to defer the release of the kmem_cache struct until the ->release callback is triggered. The put of the embedded kobject must be the last action on the kmem_cache structure which will then trigger release and that will trigger the kmem_cache_free(). -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>