On (04/26/17 14:59), Joonsoo Kim wrote: > On Wed, Apr 26, 2017 at 11:37:18AM +0900, Sergey Senozhatsky wrote: > > On (04/26/17 09:52), js1...@gmail.com wrote: > > [..] > > > struct zram_entry { > > > + struct rb_node rb_node; > > > + u32 len; > > > + u32 checksum; > > > + unsigned long refcount; > > > > use refcount_t? what do you think? > > We don't need atomic operation for refcount but refcount_t does. > API of refcount_t provides additional guarantee that refcount will not > overflow but I'm not sure that this overhead is needed here.
refcount_t has some additional debugging features which probably could be helpful one day, but not necessarily needed in current zram_dedup implementation, agree. asked just in case. -ss