On Tue 18-12-18 20:27:43, Wei Yang wrote: [...] > BTW, would this eat lower zone's memory? For example, has less DMA32? Yes I think so. If the distribution should be even and some node(s) span only lower 32b address range then there is no other option than shrink the DMA32 zone. There is a note In the event, a node is too small to have both ZONE_NORMAL and ZONE_MOVABLE, kernelcore memory will take priority and other nodes will have a larger ZONE_MOVABLE. which explains that this might not be the case though. Btw. I have to say I quite do not like this interface not to mention the implementation. THere are users to rely on it though so we cannot remove it. There is a lot of room for cleanups there. -- Michal Hocko SUSE Labs
