On 05/01/2013 03:30 PM, David Rientjes wrote:
On Wed, 1 May 2013, Cody P Schafer wrote:diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c index a221fac..0bdca10 100644 --- a/mm/memory_hotplug.c +++ b/mm/memory_hotplug.c @@ -915,6 +915,7 @@ static void node_states_set_node(int node, struct memory_notify *arg) int __ref online_pages(unsigned long pfn, unsigned long nr_pages, int online_type) { + unsigned long flags; unsigned long onlined_pages = 0; struct zone *zone; int need_zonelists_rebuild = 0; @@ -993,7 +994,11 @@ int __ref online_pages(unsigned long pfn, unsigned long nr_pages, int online_typ zone->managed_pages += onlined_pages; zone->present_pages += onlined_pages; + + pgdat_resize_lock(zone->zone_pgdat, &flags); zone->zone_pgdat->node_present_pages += onlined_pages; + pgdat_resize_unlock(zone->zone_pgdat, &flags); + if (onlined_pages) { node_states_set_node(zone_to_nid(zone), &arg); if (need_zonelists_rebuild)Why? You can't get a partial read of a word-sized data structure.
Guaranteed to be stable means that if I'm a reader and pgdat_resize_lock(), node_present_pages had better not change at all until I pgdat_resize_unlock().
If nothing needs this guarantee, we should change the rules of pgdat_resize_lock(). I played it safe and went with following the existing rules.
-- 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>