On Mon, Apr 30, 2018 at 12:42:16PM +0000, Matthew Wilcox wrote: > On Mon, Apr 30, 2018 at 12:47:04PM +0300, Kirill A. Shutemov wrote: > > On Wed, Apr 18, 2018 at 11:49:09AM -0700, Matthew Wilcox wrote: > > > From: Matthew Wilcox <mawilcox@xxxxxxxxxxxxx> > > > > > > This gives us five words of space in a single union in struct page. > > > The compound_mapcount moves position (from offset 24 to offset 20) > > > on 64-bit systems, but that does not seem likely to cause any trouble. > > > > Yeah, it should be fine. > > > > Acked-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx> > > I was wondering if it might make sense to make compound_mapcount an > atomic_long_t. It'd guarantee no overflow, and prevent the location > from moving. It would only make sense if we change mapcount too. I mean, what the point if the first split_huge_pmd() will cause the overflow. -- Kirill A. Shutemov
