On 09/15/2015 02:52 PM, Andrew Morton wrote:
>> /*
>> * If movable zone has already been setup, newly added memory should be check.
>> * If its address is higher than movable zone, it should be added as movable.
>> + * And if system boots up with movable_node and config CONFIG_MOVABLE_NOD and
>> + * added memory does not overlap the zone before MOVABLE_ZONE,
>> + * the memory is added as movable
>> * Without this check, movable zone may overlap with other zone.
>> */
>> static int should_add_memory_movable(int nid, u64 start, u64 size)
>> @@ -1208,6 +1211,11 @@ static int should_add_memory_movable(int
>> unsigned long start_pfn = start >> PAGE_SHIFT;
>> pg_data_t *pgdat = NODE_DATA(nid);
>> struct zone *movable_zone = pgdat->node_zones + ZONE_MOVABLE;
>> + struct zone *pre_zone = pgdat->node_zones + (ZONE_MOVABLE - 1);
>> +
>> + if (movable_node_is_enabled()
>> + && zone_end_pfn(pre_zone) <= start_pfn)
>> + return 1;
This check seems goofy to me. According to the description, we're
looking at a node here which has all of its zones empty. So it
definitely has zone->spanned_pages=0. zone_end_pfn() is looking at
zone->zone_start_pfn too, which is also 0, presumably.
So why is it bothering to look at the pfns if they're potentially
"garbage"? It seems like we really want something like this:
if (all_node_zones_empty(pgdat)) {
/*
* We usually want a ZONE_NORMAL before we add a
* ZONE_MOVABLE since ZONE_MOVABLE is mildly crippled.
* We only want ZONE_MOVABLE first when 'movable_node'
* mode is on.
*/
return movable_node_is_enabled();
}
Either way, this is a behavior change. It's one that is triggered by a
config option plus a boot option, but it might surprise some users. Is
this new behavior documented?
--
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>