On Tue 30-01-18 13:11:06, Pavel Tatashin wrote: > Hi Michal, > > Thank you for taking care of the problem. The patch may introduce a > small performance regression during normal boot, as we add a branch > into a hot initialization path. But, it fixes a current problem, so: > > Reviewed-by: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx> Thanks! > However, I think we should change the hotplug code to also not to > touch the map area until struct pages are initialized. > > Currently, we loop through "struct page"s several times during memory hotplug: > > 1. memset(0) in sparse_add_one_section() > 2. loop in __add_section() to set do: set_page_node(page, nid); and > SetPageReserved(page); > 3. loop in pages_correctly_reserved() to check that SetPageReserved is set. > 4. loop in memmap_init_zone() to call __init_single_pfn() You might be very well correct but the hotplug code is quite subtle and we do depend on PageReserved at some unexpected places so it is not that easy I am afraid. My TODO list in the hotplug is quite long. If you feel like you want to work on that I would be more than happy. -- Michal Hocko SUSE Labs -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
