Skip to contentSkip to conversation

fix output of numactl --hardware with >BITS_PER_LONG cpus

PATCH New Arnd Bergmann 1 participant 1 message Jan 28, 2009

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fix output of numactl --hardware with >BITS_PER_LONG cpus

There was evidently a glitch in the conversion of numactl to libnuma2:

The print_node_cpus function tries to find the size of the bitmask, which it no longer needs to. This results in printing only the first word of the CPU mask.

This fixes the function to print all available CPUs again.

Found by Bill Buros on a 128-way 64 bit system.

numactl.c +6 -18
index 2de33d8..f1a7166 100644
@@ -199,28 +199,16 @@ static void print_distances(int maxnode)
 
 void print_node_cpus(int node)
 {
-	int len = 1;
 	int conf_cpus = numa_num_configured_cpus();
+	int i, err;
+	struct bitmask *cpus;
 
-	for (;;) { 
-		int i, err;
-		struct bitmask *cpus;
-
-		cpus = numa_bitmask_alloc(conf_cpus);
-		errno = 0;
-		err = numa_node_to_cpus(node, cpus);
-		if (err < 0) {
-			if (errno == ERANGE) {
-				len *= 2; 
-				continue;
-			}
-			break; 
-		}
-		for (i = 0; i < len*BITS_PER_LONG; i++) 
+	cpus = numa_bitmask_alloc(conf_cpus);
+	err = numa_node_to_cpus(node, cpus);
+	if (err >= 0) 
+		for (i = 0; i < conf_cpus; i++) 
 			if (numa_bitmask_isbitset(cpus, i))
 				printf(" %d", i);
-		break;
-	}
 	putchar('\n');
 }