Skip to contentSkip to conversation

Fix test/prefered v2

PATCH Cliff Wickman 2 participants 2 messages Jan 15, 2009 – Jan 16, 2009

lore
Share & tools
lore
#0 lore raw

Hi Andi,

Quoted 6 lines

Date: Sat, 10 Jan 2009 07:18:09 +0100
From: Andi Kleen <[email protected]>
To: Andi Kleen <[email protected]>
Cc: [email protected], [email protected]
Subject: [PATCH] Fix test/prefered v2

Quoted 8 lines

Fix test/prefered v2

Couple of bugs here:

- First the kernel errors out of the nodemask is < MAXNUMNODES, so have to discover that at runtime. There's no function in libnuma for this (perhaps there should be one?), so i wrote one here.

There is numa_num_possible_nodes(). It determines the size of the kernel nodemask_t (MAX_NUMNODES) by counting the size of Mems_allowed:
in /proc/self/status. (And if that is not available, falls back to a procedure very similar to your max_numnode().)
The patch, but using numa_num_possible_nodes() is below.
It returns the same results, in my tests.

Is there some subtle difference between numa_num_possible_nodes() and your function that I'm missing?

-Cliff

Quoted 50 lines

- The loop set the node in the wrong mask, breaking the test

v2: Correct max_numnode detection

test/prefered.c +20 -2
@@ -6,20 +6,38 @@
 #include <assert.h>
 #include <unistd.h>
 #include <stdlib.h>
+#include <errno.h>
 
 #define err(x) perror(x),exit(1)
 
+/* Discover MAXNUMNODE of the kernel */
+int max_numnode(void)
+{	
+	int v;
+	unsigned size = 64;
+	struct bitmask *mask = numa_bitmask_alloc(size);
+	while (get_mempolicy(NULL, mask->maskp, mask->size, &v, MPOL_F_ADDR) < 0 &&
+			errno == EINVAL) {
+		numa_bitmask_free(mask);
+		size <<= 1;
+		mask = numa_bitmask_alloc(size);
+	}
+	numa_bitmask_free(mask);
+	return size;
+}
+
 int main(void)
 {
 	int max = numa_max_node();
+	int maxmask = max_numnode();
 	struct bitmask *nodes, *mask;
 	int pagesize = getpagesize();
 	int i;
 	int pol;
 	int node;
 	int err = 0;
-	nodes = numa_bitmask_alloc(max+1);
-	mask = numa_bitmask_alloc(max+1);
+	nodes = numa_bitmask_alloc(maxmask);
+	mask = numa_bitmask_alloc(maxmask);
 
 	for (i = max; i >= 0; --i) { 
 		char *mem = mmap(NULL, pagesize*(max+1), PROT_READ|PROT_WRITE, 

test/prefered.c line 58

Quoted 9 lines
test/prefered.c +1 -1
@@ -33,8 +51,8 @@
 
 		numa_bitmask_clearall(nodes);
 		numa_bitmask_clearall(mask);
-		numa_bitmask_setbit(mask, i);
+		numa_bitmask_setbit(nodes, i);
 
 		if (mbind(adr,  pagesize, MPOL_PREFERRED, nodes->maskp,
 							nodes->size, 0) < 0)
 			err("mbind");
---
 test/prefered.c |    8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
test/prefered.c +5 -3
Index: numactl-dev/test/prefered.c
===================================================================
@@ -6,20 +6,22 @@
 #include <assert.h>
 #include <unistd.h>
 #include <stdlib.h>
+#include <errno.h>
 
 #define err(x) perror(x),exit(1)
 
 int main(void)
 {
 	int max = numa_max_node();
+	int maxmask = numa_num_possible_nodes();
 	struct bitmask *nodes, *mask;
 	int pagesize = getpagesize();
 	int i;
 	int pol;
 	int node;
 	int err = 0;
-	nodes = numa_bitmask_alloc(max+1);
-	mask = numa_bitmask_alloc(max+1);
+	nodes = numa_bitmask_alloc(maxmask);
+	mask = numa_bitmask_alloc(maxmask);
 
 	for (i = max; i >= 0; --i) { 
 		char *mem = mmap(NULL, pagesize*(max+1), PROT_READ|PROT_WRITE, 
@@ -33,7 +35,7 @@ int main(void)
 
 		numa_bitmask_clearall(nodes);
 		numa_bitmask_clearall(mask);
-		numa_bitmask_setbit(mask, i);
+		numa_bitmask_setbit(nodes, i);
 
 		if (mbind(adr,  pagesize, MPOL_PREFERRED, nodes->maskp,
 							nodes->size, 0) < 0)
#1 lore raw
Quoted 1 line

There is numa_num_possible_nodes(). It determines the size of

Ah that's better.

Quoted 8 lines

the kernel nodemask_t (MAX_NUMNODES) by counting the size of Mems_allowed:
in /proc/self/status. (And if that is not available, falls back to a procedure very similar to your max_numnode().)
The patch, but using numa_num_possible_nodes() is below.
It returns the same results, in my tests.

Is there some subtle difference between numa_num_possible_nodes() and your function that I'm missing?

No your patch should be ok.

-Andi