On 02.04.19 10:28, Oscar Salvador wrote: > On Mon, Apr 01, 2019 at 01:53:06PM +0200, Michal Hocko wrote: >> On Mon 01-04-19 09:59:36, Oscar Salvador wrote: >>> On Fri, Mar 29, 2019 at 02:42:43PM +0100, Michal Hocko wrote: >>>> Having a larger contiguous area is definitely nice to have but you also >>>> have to consider the other side of the thing. If we have a movable >>>> memblock with unmovable memory then we are breaking the movable >>>> property. So there should be some flexibility for caller to tell whether >>>> to allocate on per device or per memblock. Or we need something to move >>>> memmaps during the hotremove. >>> >>> By movable memblock you mean a memblock whose pages can be migrated over when >>> this memblock is offlined, right? >> >> I am mostly thinking about movable_node kernel parameter which makes >> newly hotpluged memory go into ZONE_MOVABLE and people do use that to >> make sure such a memory can be later hotremoved. > > Uhm, I might be missing your point, but hot-added memory that makes use of > vmemmap pages can be hot-removed as any other memory. > > Vmemmap pages do not account as unmovable memory, they just stick around > until all sections they referred to have been removed, and then, we proceed > with removing them. > So, to put it in another way: vmemmap pages are left in the system until the > whole memory device (DIMM, virt mem-device or whatever) is completely > hot-removed. Indeed, separate memblocks can be offlined, but vmemmap is removed along with remove_memory(). -- Thanks, David / dhildenb
