On 4/3/19 11:40 AM, Oscar Salvador wrote: > On Wed, Apr 03, 2019 at 10:37:57AM +0200, Michal Hocko wrote: >> That being said it should be the caller of the hotplug code to tell >> the vmemmap allocation strategy. For starter, I would only pack vmemmaps >> for "regular" kernel zone memory. Movable zones should be more careful. >> We can always re-evaluate later when there is a strong demand for huge >> pages on movable zones but this is not the case now because those pages >> are not really movable in practice. > > I agree that makes sense to let the caller specify if it wants to allocate > vmemmaps per memblock or per memory-range, so we are more flexible when it > comes to granularity in hot-add/hot-remove operations.
Please also add possibility to not allocate from hotadded memory (i.e. allocate from node 0 as it is done now). I know about some MCDRAM users that want to expose as much as possible to userspace, and not even occupy those ~4% for memmap. > But the thing is that the zones are picked at onling stage, while > vmemmaps are created at hot-add stage, so I am not sure we can define > the strategy depending on the zone. >
