On Wed, Dec 16, 2020 at 08:47:25AM +0100, Christian König wrote: > Am 15.12.20 um 20:08 schrieb Andy Shevchenko: > > On Tue, Dec 15, 2020 at 03:30:44PM +0100, Christian König wrote: > > > Am 15.12.20 um 04:03 schrieb Andrew Morton: > > ... > > > > > Question: Can the fact that __GFP_ZERO is effectively ignored cause an > > > information leak if new size is larger than old size and the array is > > > somehow copied to user space? > > > > > > I think the answer is no, but just wanted to double check. Maybe we should > > > note that here. > > kmalloc()/kmalloc_array()/etc has the same. Should it be mentioned there as well? > > No, they don't. If kmalloc()/kmalloc_array() would ignore __GFP_ZERO we > would have quite a problem. > > It is only krealloc()/krealloc_array() which ignore __GFP_ZERO when they > don't reallocate memory because newsize is smaller than oldsize. In other > words the freed up space is not cleared in any way. Yes, true. So, you meant that comment now a bit misleading. I agree. -- With Best Regards, Andy Shevchenko
