[patch 00/10] NUMA: Memoryless Node support V1

[patch 00/10] NUMA: Memoryless Node support V1

[clameter]

This patch is addressing various issues with NUMA as a result of memory
less nodes being used. I think this is only a start fixing the most obvious
things, there may be more where this came from. I'd appreciate if someone
with a system with memoryless nodes could do systematic testing to see that
all the NUMA functionality works properly. Nishanth has done some testing
but he seems to be farily new to this.

The patchset is also part of my upload queue at
http://ftp.kernel.org/pub/linux/kernel/people/christoph/2.6.22-rc4-mm2

I know that some people are doing work based on this patchset. Will update
the patches in that location if more fixes are submitted. I hope Andrew
will get a new mm version out soon.

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 01/10] Memoryless nodes: Fix GFP_THISNODE behavior

[clameter]

[-- Attachment #1: memless_thisnode_fix --]
[-- Type: text/plain, Size: 5355 bytes --]

GFP_THISNODE checks that the zone selected is within the pgdat (node) of the
first zone of a nodelist. That only works if the node has memory. A
memoryless node will have its first node on another pgdat (node).

GFP_THISNODE currently will return simply memory on the first pgdat.
Thus it is returning memory on other nodes. GFP_THISNODE should fail
if there is no local memory on a node.


Add a new set of zonelists for each node that only contain the nodes
that belong to the zones itself so that no fallback is possible.

Then modify gfp_type to pickup the right zone based on the presence
of __GFP_THISNODE.

Drop the existing GFP_THISNODE checks from the page_allocators hot path.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/include/linux/gfp.h
===================================================================
--- linux-2.6.22-rc4-mm2.orig/include/linux/gfp.h	2007-06-18 11:46:24.000000000 -0700
+++ linux-2.6.22-rc4-mm2/include/linux/gfp.h	2007-06-18 11:47:11.000000000 -0700
@@ -116,22 +116,28 @@ static inline int allocflags_to_migratet
 
 static inline enum zone_type gfp_zone(gfp_t flags)
 {
+	int base = 0;
+
+#ifdef CONFIG_NUMA
+	if (flags & __GFP_THISNODE)
+		base = MAX_NR_ZONES;
+#endif
 #ifdef CONFIG_ZONE_DMA
 	if (flags & __GFP_DMA)
-		return ZONE_DMA;
+		return base + ZONE_DMA;
 #endif
 #ifdef CONFIG_ZONE_DMA32
 	if (flags & __GFP_DMA32)
-		return ZONE_DMA32;
+		return base + ZONE_DMA32;
 #endif
 	if ((flags & (__GFP_HIGHMEM | __GFP_MOVABLE)) ==
 			(__GFP_HIGHMEM | __GFP_MOVABLE))
-		return ZONE_MOVABLE;
+		return base + ZONE_MOVABLE;
 #ifdef CONFIG_HIGHMEM
 	if (flags & __GFP_HIGHMEM)
-		return ZONE_HIGHMEM;
+		return base + ZONE_HIGHMEM;
 #endif
-	return ZONE_NORMAL;
+	return base + ZONE_NORMAL;
 }
 
 static inline gfp_t set_migrateflags(gfp_t gfp, gfp_t migrate_flags)
Index: linux-2.6.22-rc4-mm2/mm/page_alloc.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/page_alloc.c	2007-06-18 11:46:26.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/page_alloc.c	2007-06-18 11:48:32.000000000 -0700
@@ -1430,9 +1430,6 @@ zonelist_scan:
 			!zlc_zone_worth_trying(zonelist, z, allowednodes))
 				continue;
 		zone = *z;
-		if (unlikely(NUMA_BUILD && (gfp_mask & __GFP_THISNODE) &&
-			zone->zone_pgdat != zonelist->zones[0]->zone_pgdat))
-				break;
 		if ((alloc_flags & ALLOC_CPUSET) &&
 			!cpuset_zone_allowed_softwall(zone, gfp_mask))
 				goto try_next_zone;
@@ -1553,7 +1550,10 @@ restart:
 	z = zonelist->zones;  /* the list of zones suitable for gfp_mask */
 
 	if (unlikely(*z == NULL)) {
-		/* Should this ever happen?? */
+		/*
+		 * Happens if we have an empty zonelist as a result of
+		 * GFP_THISNODE being used on a memoryless node
+		 */
 		return NULL;
 	}
 
@@ -2151,6 +2151,22 @@ static void build_zonelists_in_node_orde
 }
 
 /*
+ * Build gfp_thisnode zonelists
+ */
+static void build_thisnode_zonelists(pg_data_t *pgdat)
+{
+	enum zone_type i;
+	int j;
+	struct zonelist *zonelist;
+
+	for (i = 0; i < MAX_NR_ZONES; i++) {
+		zonelist = pgdat->node_zonelists + MAX_NR_ZONES + i;
+ 		j = build_zonelists_node(pgdat, zonelist, 0, i);
+		zonelist->zones[j] = NULL;
+	}
+}
+
+/*
  * Build zonelists ordered by zone and nodes within zones.
  * This results in conserving DMA zone[s] until all Normal memory is
  * exhausted, but results in overflowing to remote node while memory
@@ -2254,7 +2270,7 @@ static void build_zonelists(pg_data_t *p
 	int order = current_zonelist_order;
 
 	/* initialize zonelists */
-	for (i = 0; i < MAX_NR_ZONES; i++) {
+	for (i = 0; i < 2 * MAX_NR_ZONES; i++) {
 		zonelist = pgdat->node_zonelists + i;
 		zonelist->zones[0] = NULL;
 	}
@@ -2299,6 +2315,8 @@ static void build_zonelists(pg_data_t *p
 		/* calculate node order -- i.e., DMA last! */
 		build_zonelists_in_zone_order(pgdat, j);
 	}
+
+	build_thisnode_zonelists(pgdat);
 }
 
 /* Construct the zonelist performance cache - see further mmzone.h */
Index: linux-2.6.22-rc4-mm2/include/linux/mmzone.h
===================================================================
--- linux-2.6.22-rc4-mm2.orig/include/linux/mmzone.h	2007-06-18 11:46:24.000000000 -0700
+++ linux-2.6.22-rc4-mm2/include/linux/mmzone.h	2007-06-18 11:47:11.000000000 -0700
@@ -356,6 +356,7 @@ struct zone {
 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
 
 #ifdef CONFIG_NUMA
+#define MAX_ZONELISTS (2 * MAX_NR_ZONES)
 /*
  * We cache key information from each zonelist for smaller cache
  * footprint when scanning for free pages in get_page_from_freelist().
@@ -421,6 +422,7 @@ struct zonelist_cache {
 	unsigned long last_full_zap;		/* when last zap'd (jiffies) */
 };
 #else
+#define MAX_ZONELISTS MAX_NR_ZONES
 struct zonelist_cache;
 #endif
 
@@ -469,7 +471,7 @@ extern struct page *mem_map;
 struct bootmem_data;
 typedef struct pglist_data {
 	struct zone node_zones[MAX_NR_ZONES];
-	struct zonelist node_zonelists[MAX_NR_ZONES];
+	struct zonelist node_zonelists[MAX_ZONELISTS];
 	int nr_zones;
 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 	struct page *node_mem_map;

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 02/10] NUMA: Introduce node_memory_map

[clameter]

[-- Attachment #1: memless_memory_map --]
[-- Type: text/plain, Size: 4361 bytes --]

It is necessary to know if nodes have memory since we have recently
begun to add support for memoryless nodes. For that purpose we introduce
a new bitmap called

node_memory_map

A node has its bit in node_memory_map set if it has memory. If a node
has memory then it has at least one zone defined in its pgdat structure
that is located in the pgdat itself.

The node_memory_map can then be used in various places to insure that we
do the right thing when we encounter a memoryless node.

Signed-off-by: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.22-rc4-mm2/include/linux/nodemask.h
===================================================================
--- linux-2.6.22-rc4-mm2.orig/include/linux/nodemask.h	2007-06-18 11:46:26.000000000 -0700
+++ linux-2.6.22-rc4-mm2/include/linux/nodemask.h	2007-06-18 11:48:42.000000000 -0700
@@ -64,12 +64,16 @@
  *
  * int node_online(node)		Is some node online?
  * int node_possible(node)		Is some node possible?
+ * int node_memory(node)		Does a node have memory?
  *
  * int any_online_node(mask)		First online node in mask
  *
  * node_set_online(node)		set bit 'node' in node_online_map
  * node_set_offline(node)		clear bit 'node' in node_online_map
  *
+ * node_set_has_memory(node)		set bit 'node' in node_memory_map
+ * node_set_no_memory(node)		clear bit 'node' in node_memory_map
+ *
  * for_each_node(node)			for-loop node over node_possible_map
  * for_each_online_node(node)		for-loop node over node_online_map
  *
@@ -344,12 +348,14 @@ static inline void __nodes_remap(nodemas
 
 extern nodemask_t node_online_map;
 extern nodemask_t node_possible_map;
+extern nodemask_t node_memory_map;
 
 #if MAX_NUMNODES > 1
 #define num_online_nodes()	nodes_weight(node_online_map)
 #define num_possible_nodes()	nodes_weight(node_possible_map)
 #define node_online(node)	node_isset((node), node_online_map)
 #define node_possible(node)	node_isset((node), node_possible_map)
+#define node_memory(node)	node_isset((node), node_memory_map)
 #define first_online_node	first_node(node_online_map)
 #define next_online_node(nid)	next_node((nid), node_online_map)
 extern int nr_node_ids;
@@ -358,6 +364,8 @@ extern int nr_node_ids;
 #define num_possible_nodes()	1
 #define node_online(node)	((node) == 0)
 #define node_possible(node)	((node) == 0)
+#define node_memory(node)	((node) == 0)
+#define node_populated(node)	((node) == 0)
 #define first_online_node	0
 #define next_online_node(nid)	(MAX_NUMNODES)
 #define nr_node_ids		1
@@ -375,7 +383,11 @@ extern int nr_node_ids;
 #define node_set_online(node)	   set_bit((node), node_online_map.bits)
 #define node_set_offline(node)	   clear_bit((node), node_online_map.bits)
 
+#define node_set_has_memory(node)  set_bit((node), node_memory_map.bits)
+#define node_set_no_memory(node)   clear_bit((node), node_memory_map.bits)
+
 #define for_each_node(node)	   for_each_node_mask((node), node_possible_map)
 #define for_each_online_node(node) for_each_node_mask((node), node_online_map)
+#define for_each_memory_node(node) for_each_node_mask((node), node_memory_map)
 
 #endif /* __LINUX_NODEMASK_H */
Index: linux-2.6.22-rc4-mm2/mm/page_alloc.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/page_alloc.c	2007-06-18 11:48:32.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/page_alloc.c	2007-06-18 11:49:34.000000000 -0700
@@ -54,6 +54,9 @@ nodemask_t node_online_map __read_mostly
 EXPORT_SYMBOL(node_online_map);
 nodemask_t node_possible_map __read_mostly = NODE_MASK_ALL;
 EXPORT_SYMBOL(node_possible_map);
+nodemask_t node_memory_map __read_mostly = NODE_MASK_NONE;
+EXPORT_SYMBOL(node_memory_map);
+
 unsigned long totalram_pages __read_mostly;
 unsigned long totalreserve_pages __read_mostly;
 long nr_swap_pages;
@@ -2317,6 +2320,9 @@ static void build_zonelists(pg_data_t *p
 	}
 
 	build_thisnode_zonelists(pgdat);
+
+	if (pgdat->node_present_pages)
+		node_set_has_memory(local_node);
 }
 
 /* Construct the zonelist performance cache - see further mmzone.h */

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 03/10] Fix MPOL_INTERLEAVE behavior for memoryless nodes

[clameter]

[-- Attachment #1: memless_fix_interleave --]
[-- Type: text/plain, Size: 1353 bytes --]

MPOL_INTERLEAVE currently simply loops over all nodes. Allocations on
memoryless nodes will be redirected to nodes with memory. This results in
an imbalance because the neighboring nodes to memoryless nodes will get significantly
more interleave hits that the rest of the nodes on the system.

We can avoid this imbalance by clearing the nodes in the interleave node
set that have no memory.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/mm/mempolicy.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/mempolicy.c	2007-06-13 23:06:14.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/mempolicy.c	2007-06-14 00:49:43.000000000 -0700
@@ -185,7 +185,8 @@ static struct mempolicy *mpol_new(int mo
 	switch (mode) {
 	case MPOL_INTERLEAVE:
 		policy->v.nodes = *nodes;
-		if (nodes_weight(*nodes) == 0) {
+		nodes_and(policy->v.nodes, policy->v.nodes, node_memory_map);
+		if (nodes_weight(policy->v.nodes) == 0) {
 			kmem_cache_free(policy_cache, policy);
 			return ERR_PTR(-EINVAL);
 		}

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 04/10] OOM: use the node_memory_map instead of constructing one on the fly

[clameter]

[-- Attachment #1: memless_oom_kill --]
[-- Type: text/plain, Size: 1150 bytes --]

constrained_alloc() builds its own memory map for nodes with memory.
We have that available in node_memory_map now. So simplify the code.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/mm/oom_kill.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/oom_kill.c	2007-06-13 23:11:32.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/oom_kill.c	2007-06-13 23:12:39.000000000 -0700
@@ -176,14 +176,7 @@ static inline int constrained_alloc(stru
 {
 #ifdef CONFIG_NUMA
 	struct zone **z;
-	nodemask_t nodes;
-	int node;
-
-	nodes_clear(nodes);
-	/* node has memory ? */
-	for_each_online_node(node)
-		if (NODE_DATA(node)->node_present_pages)
-			node_set(node, nodes);
+	nodemask_t nodes = node_memory_map;
 
 	for (z = zonelist->zones; *z; z++)
 		if (cpuset_zone_allowed_softwall(*z, gfp_mask))

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 05/10] Memoryless Nodes: No need for kswapd

[clameter]

[-- Attachment #1: memless_no_kswapd --]
[-- Type: text/plain, Size: 920 bytes --]

A node without memory does not need a kswapd. So use the memory map instead
of the online map when starting kswapd.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/mm/vmscan.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/vmscan.c	2007-06-18 11:46:25.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/vmscan.c	2007-06-18 11:49:47.000000000 -0700
@@ -1735,7 +1735,7 @@ static int __init kswapd_init(void)
 	int nid;
 
 	swap_setup();
-	for_each_online_node(nid)
+	for_each_memory_node(nid)
  		kswapd_run(nid);
 	hotcpu_notifier(cpu_callback, 0);
 	return 0;

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 06/10] Memoryless Node: Slab support

[clameter]

[-- Attachment #1: memless_slab --]
[-- Type: text/plain, Size: 2151 bytes --]

Slab should not allocate control structures for nodes without memory. This may seem
to work right now but its unreliable since not all allocations can fall back due
to the use of GFP_THISNODE.

Switching a few for_each_online_node's to for_each_memory_node will allow us to
only allocate for nodes that actually have memory.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/mm/slab.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/slab.c	2007-06-18 11:46:25.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/slab.c	2007-06-18 11:49:53.000000000 -0700
@@ -1564,7 +1564,7 @@ void __init kmem_cache_init(void)
 		/* Replace the static kmem_list3 structures for the boot cpu */
 		init_list(&cache_cache, &initkmem_list3[CACHE_CACHE], node);
 
-		for_each_online_node(nid) {
+		for_each_memory_node(nid) {
 			init_list(malloc_sizes[INDEX_AC].cs_cachep,
 				  &initkmem_list3[SIZE_AC + nid], nid);
 
@@ -1942,7 +1942,7 @@ static void __init set_up_list3s(struct 
 {
 	int node;
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		cachep->nodelists[node] = &initkmem_list3[index + node];
 		cachep->nodelists[node]->next_reap = jiffies +
 		    REAPTIMEOUT_LIST3 +
@@ -2073,7 +2073,7 @@ static int __init_refok setup_cpu_cache(
 			g_cpucache_up = PARTIAL_L3;
 		} else {
 			int node;
-			for_each_online_node(node) {
+			for_each_memory_node(node) {
 				cachep->nodelists[node] =
 				    kmalloc_node(sizeof(struct kmem_list3),
 						GFP_KERNEL, node);
@@ -3787,7 +3787,7 @@ static int alloc_kmemlist(struct kmem_ca
 	struct array_cache *new_shared;
 	struct array_cache **new_alien = NULL;
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 
                 if (use_alien_caches) {
                         new_alien = alloc_alien_cache(node, cachep->limit);

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 07/10] Memoryless nodes: SLUB support

[clameter]

[-- Attachment #1: memless_slub --]
[-- Type: text/plain, Size: 3389 bytes --]

Simply switch all for_each_online_node to for_each_memory_node. That way
SLUB only operates on nodes with memory. Any allocation attempt on a
memoryless node will fall whereupon SLUB will fetch memory from a nearby
node (depending on how memory policies and cpuset describe fallback).

Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.22-rc4-mm2/mm/slub.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/slub.c	2007-06-18 11:16:15.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/slub.c	2007-06-18 11:28:50.000000000 -0700
@@ -2086,7 +2086,7 @@ static void free_kmem_cache_nodes(struct
 {
 	int node;
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n = s->node[node];
 		if (n && n != &s->local_node)
 			kmem_cache_free(kmalloc_caches, n);
@@ -2104,7 +2104,7 @@ static int init_kmem_cache_nodes(struct 
 	else
 		local_node = 0;
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n;
 
 		if (local_node == node)
@@ -2366,7 +2366,7 @@ static inline int kmem_cache_close(struc
 	/* Attempt to free all objects */
 	free_kmem_cache_cpus(s);
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n = get_node(s, node);
 
 		n->nr_partial -= free_list(s, n, &n->partial);
@@ -2937,7 +2937,7 @@ int kmem_cache_shrink(struct kmem_cache 
 	if (!scratch)
 		return -ENOMEM;
 
-	for_each_online_node(node)
+	for_each_memory_node(node)
 		__kmem_cache_shrink(s, get_node(s, node), scratch);
 
 	kfree(scratch);
@@ -3008,7 +3008,7 @@ int kmem_cache_defrag(int percent, int n
 		scratch = kmalloc(sizeof(struct list_head) * s->objects,
 								GFP_KERNEL);
 		if (node == -1) {
-			for_each_online_node(node)
+			for_each_memory_node(node)
 				pages += __kmem_cache_defrag(s, percent,
 							node, scratch);
 		} else
@@ -3392,7 +3392,7 @@ static unsigned long validate_slab_cache
 	unsigned long count = 0;
 
 	flush_all(s);
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n = get_node(s, node);
 
 		count += validate_slab_node(s, n);
@@ -3611,7 +3611,7 @@ static int list_locations(struct kmem_ca
 	/* Push back cpu slabs */
 	flush_all(s);
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n = get_node(s, node);
 		unsigned long flags;
 		struct page *page;
@@ -3723,7 +3723,7 @@ static unsigned long slab_objects(struct
 		}
 	}
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n = get_node(s, node);
 
 		if (flags & SO_PARTIAL) {
@@ -3751,7 +3751,7 @@ static unsigned long slab_objects(struct
 
 	x = sprintf(buf, "%lu", total);
 #ifdef CONFIG_NUMA
-	for_each_online_node(node)
+	for_each_memory_node(node)
 		if (nodes[node])
 			x += sprintf(buf + x, " N%d=%lu",
 					node, nodes[node]);
@@ -3772,7 +3772,7 @@ static int any_slab_objects(struct kmem_
 			return 1;
 	}
 
-	for_each_online_node(node) {
+	for_each_memory_node(node) {
 		struct kmem_cache_node *n = get_node(s, node);
 
 		if (n && (n->nr_partial || atomic_read(&n->nr_slabs)))

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [patch 07/10] Memoryless nodes: SLUB support

[Lee Schermerhorn]

On Mon, 2007-06-18 at 12:20 -0700, clameter@sgi.com wrote:
> plain text document attachment (memless_slub)
> Simply switch all for_each_online_node to for_each_memory_node. That way
> SLUB only operates on nodes with memory. Any allocation attempt on a
> memoryless node will fall whereupon SLUB will fetch memory from a nearby
> node (depending on how memory policies and cpuset describe fallback).
> 
> Signed-off-by: Christoph Lameter <clameter@sgi.com>
> 
> Index: linux-2.6.22-rc4-mm2/mm/slub.c
> ===================================================================
> --- linux-2.6.22-rc4-mm2.orig/mm/slub.c	2007-06-18 11:16:15.000000000 -0700
> +++ linux-2.6.22-rc4-mm2/mm/slub.c	2007-06-18 11:28:50.000000000 -0700

<snip>

Christoph:

This patch didn't apply to 22-rc4-mm2.  Does it assume some other SLUB
patches?

I resolved the conflicts by just doing what the description says:
replacing all 'for_each_online_node" with "for_each_memory_node", but I
was surprised that this one patch out of 10 didn't apply.  I'm probably
missing some other patch.

Lee

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [patch 07/10] Memoryless nodes: SLUB support

[Christoph Lameter]

On Wed, 20 Jun 2007, Lee Schermerhorn wrote:

> This patch didn't apply to 22-rc4-mm2.  Does it assume some other SLUB
> patches?

Yes sorry this was based on SLUB with patches already accepted by Andrew 
for SLUB.
 
> I resolved the conflicts by just doing what the description says:
> replacing all 'for_each_online_node" with "for_each_memory_node", but I
> was surprised that this one patch out of 10 didn't apply.  I'm probably
> missing some other patch.

I think you should be fine with that approach. There is a later patch that 
does more for_each_memory stuff for the policy layer. I'd appreciate it if 
you could check if that proposed change in semantics for memoryless 
nodes makes sense to you.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [patch 07/10] Memoryless nodes: SLUB support

[Lee Schermerhorn]

On Wed, 2007-06-20 at 09:53 -0700, Christoph Lameter wrote:
> On Wed, 20 Jun 2007, Lee Schermerhorn wrote:
> 
> > This patch didn't apply to 22-rc4-mm2.  Does it assume some other SLUB
> > patches?
> 
> Yes sorry this was based on SLUB with patches already accepted by Andrew 
> for SLUB.

No problem.  That's what I thought happened, as I did see a big SLUB
series of yours go by.  Just wanted to be sure.

>  
> > I resolved the conflicts by just doing what the description says:
> > replacing all 'for_each_online_node" with "for_each_memory_node", but I
> > was surprised that this one patch out of 10 didn't apply.  I'm probably
> > missing some other patch.
> 
> I think you should be fine with that approach. 

It appears to be running OK.

> There is a later patch that 
> does more for_each_memory stuff for the policy layer. I'd appreciate it if 
> you could check if that proposed change in semantics for memoryless 
> nodes makes sense to you.

So far, so good.  We moved all of our equipment between buildings over
the weekend and I still don't have all of the test machines back
on-line.  I have tested your patches with Nish's on my ia64 NUMA
platform, and it seems to be "doing the right thing" for this particular
configuration because the DMA zone on the hardware-interleaved
pseudo-node is too small to allocate even one 256MB huge page.  On a
larger system, I think that it will be allocating some huge pages out of
the DMA zone--which I don't want.  I have an idea for fixing that that
we can discuss at some point.

I'll continue to look at/test your memoryless-node patches for other
than hugepage allocations and on other platforms, as they come on line.

Lee

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 08/10] Uncached allocator: Handle memoryless nodes

[clameter]

[-- Attachment #1: memless_mspec --]
[-- Type: text/plain, Size: 1792 bytes --]

The checks for node_online in the uncached allocator are made to make sure
that memory is available on these nodes. Thus switch all the checks to use
the node_memory and for_each_memory_node functions.

Cc: jes@sgi.com
Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.22-rc4-mm2/arch/ia64/kernel/uncached.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/arch/ia64/kernel/uncached.c	2007-06-13 23:29:58.000000000 -0700
+++ linux-2.6.22-rc4-mm2/arch/ia64/kernel/uncached.c	2007-06-13 23:32:35.000000000 -0700
@@ -196,7 +196,7 @@ unsigned long uncached_alloc_page(int st
 	nid = starting_nid;
 
 	do {
-		if (!node_online(nid))
+		if (!node_memory(nid))
 			continue;
 		uc_pool = &uncached_pools[nid];
 		if (uc_pool->pool == NULL)
@@ -268,7 +268,7 @@ static int __init uncached_init(void)
 {
 	int nid;
 
-	for_each_online_node(nid) {
+	for_each_memory_node(nid) {
 		uncached_pools[nid].pool = gen_pool_create(PAGE_SHIFT, nid);
 		mutex_init(&uncached_pools[nid].add_chunk_mutex);
 	}
Index: linux-2.6.22-rc4-mm2/drivers/char/mspec.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/drivers/char/mspec.c	2007-06-13 23:28:15.000000000 -0700
+++ linux-2.6.22-rc4-mm2/drivers/char/mspec.c	2007-06-13 23:29:35.000000000 -0700
@@ -353,7 +353,7 @@ mspec_init(void)
 		is_sn2 = 1;
 		if (is_shub2()) {
 			ret = -ENOMEM;
-			for_each_online_node(nid) {
+			for_each_memory_node(nid) {
 				int actual_nid;
 				int nasid;
 				unsigned long phys;

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [patch 08/10] Uncached allocator: Handle memoryless nodes

[Jes Sorensen]

clameter@sgi.com wrote:
> The checks for node_online in the uncached allocator are made to make sure
> that memory is available on these nodes. Thus switch all the checks to use
> the node_memory and for_each_memory_node functions.
> 
> Cc: jes@sgi.com
> Signed-off-by: Christoph Lameter <clameter@sgi.com>

Seems fair to me:

Acked-by: Jes Sorensen <jes@sgi.com>


> Index: linux-2.6.22-rc4-mm2/arch/ia64/kernel/uncached.c
> ===================================================================
> --- linux-2.6.22-rc4-mm2.orig/arch/ia64/kernel/uncached.c	2007-06-13 23:29:58.000000000 -0700
> +++ linux-2.6.22-rc4-mm2/arch/ia64/kernel/uncached.c	2007-06-13 23:32:35.000000000 -0700
> @@ -196,7 +196,7 @@ unsigned long uncached_alloc_page(int st
>  	nid = starting_nid;
>  
>  	do {
> -		if (!node_online(nid))
> +		if (!node_memory(nid))
>  			continue;
>  		uc_pool = &uncached_pools[nid];
>  		if (uc_pool->pool == NULL)
> @@ -268,7 +268,7 @@ static int __init uncached_init(void)
>  {
>  	int nid;
>  
> -	for_each_online_node(nid) {
> +	for_each_memory_node(nid) {
>  		uncached_pools[nid].pool = gen_pool_create(PAGE_SHIFT, nid);
>  		mutex_init(&uncached_pools[nid].add_chunk_mutex);
>  	}
> Index: linux-2.6.22-rc4-mm2/drivers/char/mspec.c
> ===================================================================
> --- linux-2.6.22-rc4-mm2.orig/drivers/char/mspec.c	2007-06-13 23:28:15.000000000 -0700
> +++ linux-2.6.22-rc4-mm2/drivers/char/mspec.c	2007-06-13 23:29:35.000000000 -0700
> @@ -353,7 +353,7 @@ mspec_init(void)
>  		is_sn2 = 1;
>  		if (is_shub2()) {
>  			ret = -ENOMEM;
> -			for_each_online_node(nid) {
> +			for_each_memory_node(nid) {
>  				int actual_nid;
>  				int nasid;
>  				unsigned long phys;
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 09/10] Memoryless node: Allow profiling data to fall back to other nodes

[clameter]

[-- Attachment #1: memless_profile --]
[-- Type: text/plain, Size: 1382 bytes --]

Processors on memoryless nodes must be able to fall back to remote nodes
in order to get a profiling buffer. This may lead to excessive NUMA traffic
but I think we should allow this rather than failing.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/kernel/profile.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/kernel/profile.c	2007-06-13 23:36:42.000000000 -0700
+++ linux-2.6.22-rc4-mm2/kernel/profile.c	2007-06-13 23:36:55.000000000 -0700
@@ -346,7 +346,7 @@ static int __devinit profile_cpu_callbac
 		per_cpu(cpu_profile_flip, cpu) = 0;
 		if (!per_cpu(cpu_profile_hits, cpu)[1]) {
 			page = alloc_pages_node(node,
-					GFP_KERNEL | __GFP_ZERO | GFP_THISNODE,
+					GFP_KERNEL | __GFP_ZERO,
 					0);
 			if (!page)
 				return NOTIFY_BAD;
@@ -354,7 +354,7 @@ static int __devinit profile_cpu_callbac
 		}
 		if (!per_cpu(cpu_profile_hits, cpu)[0]) {
 			page = alloc_pages_node(node,
-					GFP_KERNEL | __GFP_ZERO | GFP_THISNODE,
+					GFP_KERNEL | __GFP_ZERO,
 					0);
 			if (!page)
 				goto out_free;

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[patch 10/10] Memoryless nodes: Update memory policy and page migration

[clameter]

[-- Attachment #1: memless_migrate --]
[-- Type: text/plain, Size: 4109 bytes --]

Online nodes now may have no memory. The checks and initialization must therefore
be changed to no longer use the online functions.

This will correctly initialize the interleave on bootup to only target
nodes with memory and will make sys_move_pages return an error when a page
is to be moved to a memoryless node. Similarly we will get an error if
MPOL_BIND and MPOL_INTERLEAVE is used on a memoryless node.

These are somewhat new semantics. So far one could specify memoryless nodes
and we would maybe do the right thing and just ignore the node (or we'd do
something strange like with MPOL_INTERLEAVE). If we want to allow the
specification of memoryless nodes via memory policies then we need to keep
checking for online nodes.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>

Index: linux-2.6.22-rc4-mm2/mm/migrate.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/migrate.c	2007-06-14 00:49:43.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/migrate.c	2007-06-18 11:30:53.000000000 -0700
@@ -963,7 +963,7 @@ asmlinkage long sys_move_pages(pid_t pid
 				goto out;
 
 			err = -ENODEV;
-			if (!node_online(node))
+			if (!node_memory(node))
 				goto out;
 
 			err = -EACCES;
Index: linux-2.6.22-rc4-mm2/mm/mempolicy.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/mm/mempolicy.c	2007-06-18 11:24:51.000000000 -0700
+++ linux-2.6.22-rc4-mm2/mm/mempolicy.c	2007-06-18 11:30:53.000000000 -0700
@@ -130,7 +130,7 @@ static int mpol_check_policy(int mode, n
 			return -EINVAL;
 		break;
 	}
-	return nodes_subset(*nodes, node_online_map) ? 0 : -EINVAL;
+	return nodes_subset(*nodes, node_memory_map) ? 0 : -EINVAL;
 }
 
 /* Generate a custom zonelist for the BIND policy. */
@@ -495,9 +495,9 @@ static void get_zonemask(struct mempolic
 		*nodes = p->v.nodes;
 		break;
 	case MPOL_PREFERRED:
-		/* or use current node instead of online map? */
+		/* or use current node instead of memory_map? */
 		if (p->v.preferred_node < 0)
-			*nodes = node_online_map;
+			*nodes = node_memory_map;
 		else
 			node_set(p->v.preferred_node, *nodes);
 		break;
@@ -1606,7 +1606,7 @@ int mpol_parse_options(char *value, int 
 		*nodelist++ = '\0';
 		if (nodelist_parse(nodelist, *policy_nodes))
 			goto out;
-		if (!nodes_subset(*policy_nodes, node_online_map))
+		if (!nodes_subset(*policy_nodes, node_memory_map))
 			goto out;
 	}
 	if (!strcmp(value, "default")) {
@@ -1631,9 +1631,9 @@ int mpol_parse_options(char *value, int 
 			err = 0;
 	} else if (!strcmp(value, "interleave")) {
 		*policy = MPOL_INTERLEAVE;
-		/* Default to nodes online if no nodelist */
+		/* Default to nodes memory map if no nodelist */
 		if (!nodelist)
-			*policy_nodes = node_online_map;
+			*policy_nodes = node_memory_map;
 		err = 0;
 	}
 out:
@@ -1674,14 +1674,14 @@ void __init numa_policy_init(void)
 
 	/*
 	 * Use the specified nodemask for init, or fall back to
-	 * node_online_map.
+	 * node_memory_map.
 	 */
 	if (policy_sysinit == MPOL_DEFAULT)
 		nmask = NULL;
 	else if (!nodes_empty(nmask_sysinit))
 		nmask = &nmask_sysinit;
 	else
-		nmask = &node_online_map;
+		nmask = &node_memory_map;
 
 	if (do_set_mempolicy(policy_sysinit, nmask))
 		printk("numa_policy_init: setting init policy failed\n");
@@ -1945,7 +1945,7 @@ int show_numa_map(struct seq_file *m, vo
 		seq_printf(m, " huge");
 	} else {
 		check_pgd_range(vma, vma->vm_start, vma->vm_end,
-				&node_online_map, MPOL_MF_STATS, md);
+				&node_memory_map, MPOL_MF_STATS, md);
 	}
 
 	if (!md->pages)
@@ -1972,7 +1972,7 @@ int show_numa_map(struct seq_file *m, vo
 	if (md->writeback)
 		seq_printf(m," writeback=%lu", md->writeback);
 
-	for_each_online_node(n)
+	for_each_memory_node(n)
 		if (md->node[n])
 			seq_printf(m, " N%d=%lu", n, md->node[n]);
 out:

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [patch 00/10] NUMA: Memoryless Node support V1

[Andrew Morton]

On Mon, 18 Jun 2007 12:19:56 -0700
clameter@sgi.com wrote:

> This patch is addressing various issues with NUMA as a result of memory
> less nodes being used. I think this is only a start fixing the most obvious
> things, there may be more where this came from. I'd appreciate if someone
> with a system with memoryless nodes could do systematic testing to see that
> all the NUMA functionality works properly. Nishanth has done some testing
> but he seems to be farily new to this.
> 
> The patchset is also part of my upload queue at
> http://ftp.kernel.org/pub/linux/kernel/people/christoph/2.6.22-rc4-mm2
> 
> I know that some people are doing work based on this patchset. Will update
> the patches in that location if more fixes are submitted.

OK, I'll duck version 1 for now.  Am hugely backlogged at present and I'm
mainly looking for simple-and-safe fixes.

Plus I'm generally going more slowly and deliberately in the vague hope
that others will follow suit.

> I hope Andrew
> will get a new mm version out soon.

umm, maybe, it depends on how much crap I merge today.  I'll be spending
the rest of the week overseas, so I guess I'd better try.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>