On Tue 14-09-21 10:01:26, Vlastimil Babka wrote:

On 9/14/21 05:40, Feng Tang wrote:

[...]

quoted

+/* Whether the 'nodes' are all movable nodes */
+static inline bool movable_only_nodes(nodemask_t *nodes)
+{
+	struct zonelist *zonelist;
+	struct zoneref *z;
+
+	if (nodes_empty(*nodes))
+		return false;
+
+	zonelist =
+	    &NODE_DATA(first_node(*nodes))->node_zonelists[ZONELIST_FALLBACK];
+	z = first_zones_zonelist(zonelist, ZONE_NORMAL,	nodes);
+	return (!z->zone) ? true : false;
+}

Hmm, could all that become just this?

!nodes_intersects(&node_states[N_NORMAL_MEMORY], nodes)

Maybe yes but I find the zonelist approach much easier to follow even
though the code looks more complex at first sight. It talks about an
empty zone list for ZONE_NORMAL request which is quite clear from the
scribble. I always have to re-learn how the N*MEMORY works TBH. Maybe
this is just me though.
-- 
Michal Hocko
SUSE Labs