From: Nishanth Aravamudan <nacc@us.ibm.com>
To: Christoph Lameter <clameter@sgi.com>
Cc: lee.schermerhorn@hp.com, anton@samba.org,
akpm@linux-foundation.org, linux-mm@kvack.org
Subject: [PATCH v2] Add populated_map to account for memoryless nodes
Date: Mon, 11 Jun 2007 15:10:36 -0700 [thread overview]
Message-ID: <20070611221036.GA14458@us.ibm.com> (raw)
In-Reply-To: <Pine.LNX.4.64.0706111417540.20454@schroedinger.engr.sgi.com>
On 11.06.2007 [14:25:38 -0700], Christoph Lameter wrote:
> On Mon, 11 Jun 2007, Nishanth Aravamudan wrote:
>
> > @@ -2161,7 +2164,7 @@ static int node_order[MAX_NUMNODES];
> > static void build_zonelists_in_zone_order(pg_data_t *pgdat, int nr_nodes)
> > {
> > enum zone_type i;
> > - int pos, j, node;
> > + int pos, j;
> > int zone_type; /* needs to be signed */
> > struct zone *z;
> > struct zonelist *zonelist;
> > @@ -2171,7 +2174,7 @@ static void build_zonelists_in_zone_order(pg_data_t *pgdat, int nr_nodes)
> > pos = 0;
> > for (zone_type = i; zone_type >= 0; zone_type--) {
> > for (j = 0; j < nr_nodes; j++) {
> > - node = node_order[j];
> > + int node = node_order[j];
> > z = &NODE_DATA(node)->node_zones[zone_type];
> > if (populated_zone(z)) {
> > zonelist->zones[pos++] = z;
>
> Unrelated modifications.
Agreed -- sorry, I have just been refreshing/testing Lee and Anton's
original fixes.
> > @@ -2244,6 +2247,22 @@ static void set_zonelist_order(void)
> > current_zonelist_order = user_zonelist_order;
> > }
> >
> > +/*
> > + * setup_populate_map() - record nodes whose "policy_zone" is "on-node".
> > + */
> > +static void setup_populated_map(int nid)
> > +{
> > + pg_data_t *pgdat = NODE_DATA(nid);
> > + struct zonelist *zl = pgdat->node_zonelists + policy_zone;
> > + struct zone *z = zl->zones[0];
> > +
> > + VM_BUG_ON(!z);
> > + if (z->zone_pgdat == pgdat)
> > + node_set_populated(nid);
> > + else
> > + node_not_populated(nid);
> > +}
>
>
> A node is only populated if it has memory in the policy zone? I would
> say a node is populated if it has any memory in any zone.
>
> The above check may fail on x86_64 where only some nodes may have
> ZONE_NORMAL. Others only have ZONE_DMA32. Policy zone will be set to
> ZONE_NORMAL.
I agree here as well, updated below.
> > static void build_zonelists(pg_data_t *pgdat)
> > {
> > int j, node, load;
> > @@ -2327,6 +2346,15 @@ static void set_zonelist_order(void)
> > current_zonelist_order = ZONELIST_ORDER_ZONE;
> > }
> >
> > +/*
> > + * setup_populated_map - non-NUMA case
> > + * Only node 0 should be on-line, and it MUST be populated!
> > + */
> > +static void setup_populated_map(int nid)
> > +{
> > + node_set_populated(nid);
> > +}
>
> I'd say provide fallback functions so that node_populated() always
> returns true for !NUMA. That way it can be optimized out at compile
> time.
Already done in the original patch (node_populated() returns (node == 0)
if MAX_NUMODES <= 1), I think.
> > static void build_zonelists(pg_data_t *pgdat)
> > {
> > int node, local_node;
> > @@ -2381,6 +2409,7 @@ static int __build_all_zonelists(void *dummy)
> > for_each_online_node(nid) {
> > build_zonelists(NODE_DATA(nid));
> > build_zonelist_cache(NODE_DATA(nid));
> > + setup_populated_map(nid);
> > }
>
> Is it possible to move the set_populated_node into build_zonelists
> somehow?
>
> F.e. In build_zonelists_node you can check if nr_zones > 0 and then
> set it up?
I've tried to do this as well, please see below.
Split up Lee and Anton's original patch
(http://marc.info/?l=linux-mm&m=118133042025995&w=2), to allow for the
populated_map changes to go in on their own.
Add a populated_map nodemask to indicate a node has memory or not. We
have run into a number of issues (in practice and in code) with
assumptions about every node having memory. Having this nodemask allows
us to fix these issues; in particular, THISNODE allocations will come
from the node specified, only, and the INTERLEAVE policy will be able to
do the right thing with memoryless nodes.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
diff --git a/include/linux/nodemask.h b/include/linux/nodemask.h
index 52c54a5..751d3d7 100644
--- a/include/linux/nodemask.h
+++ b/include/linux/nodemask.h
@@ -64,12 +64,16 @@
*
* int node_online(node) Is some node online?
* int node_possible(node) Is some node possible?
+ * int node_populated(node) Is some node populated [at policy_zone]
*
* 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_populated(node) set bit 'node' in node_populated_map
+ * node_not_populated(node) clear bit 'node' in node_populated_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(nodemask_t *dstp, const nodemask_t *srcp,
extern nodemask_t node_online_map;
extern nodemask_t node_possible_map;
+extern nodemask_t node_populated_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_populated(node) node_isset((node), node_populated_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 +364,7 @@ extern int nr_node_ids;
#define num_possible_nodes() 1
#define node_online(node) ((node) == 0)
#define node_possible(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
@@ -375,6 +382,9 @@ 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_populated(node) set_bit((node), node_populated_map.bits)
+#define node_not_populated(node) clear_bit((node), node_populated_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)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 07cd5ae..1d20f8f 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -54,6 +54,9 @@ nodemask_t node_online_map __read_mostly = { { [0] = 1UL } };
EXPORT_SYMBOL(node_online_map);
nodemask_t node_possible_map __read_mostly = NODE_MASK_ALL;
EXPORT_SYMBOL(node_possible_map);
+nodemask_t node_populated_map __read_mostly = NODE_MASK_NONE;
+EXPORT_SYMBOL(node_populated_map);
+
unsigned long totalram_pages __read_mostly;
unsigned long totalreserve_pages __read_mostly;
long nr_swap_pages;
@@ -2251,6 +2254,7 @@ static void build_zonelists(pg_data_t *pgdat)
nodemask_t used_mask;
int local_node, prev_node;
struct zonelist *zonelist;
+ struct zone *z;
int order = current_zonelist_order;
/* initialize zonelists */
@@ -2299,6 +2303,18 @@ static void build_zonelists(pg_data_t *pgdat)
/* calculate node order -- i.e., DMA last! */
build_zonelists_in_zone_order(pgdat, j);
}
+
+ /*
+ * record nodes whose first fallback zone is "on-node" as
+ * populated
+ */
+ z = pgdat->node_zonelists->zones[0];
+
+ VM_BUG_ON(!z);
+ if (z->zone_pgdat == pgdat)
+ node_set_populated(local_node);
+ else
+ node_not_populated(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>
next prev parent reply other threads:[~2007-06-11 22:10 UTC|newest]
Thread overview: 140+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-06-11 20:27 [PATCH] Add populated_map to account for memoryless nodes Nishanth Aravamudan, Lee Schermerhorn
2007-06-11 21:25 ` Christoph Lameter
2007-06-11 22:10 ` Nishanth Aravamudan [this message]
2007-06-11 22:42 ` [PATCH v2] " Christoph Lameter
2007-06-11 22:52 ` [PATCH v3] " Nishanth Aravamudan
2007-06-11 23:00 ` Christoph Lameter
2007-06-11 23:41 ` [PATCH v4] " Nishanth Aravamudan
2007-06-11 23:45 ` Christoph Lameter
2007-06-12 0:07 ` [PATCH] populated_map: fix !NUMA case, remove comment Nishanth Aravamudan
2007-06-12 0:41 ` Christoph Lameter
2007-06-12 1:43 ` Nishanth Aravamudan
2007-06-12 1:45 ` Christoph Lameter
2007-06-12 1:52 ` Nishanth Aravamudan
2007-06-12 2:39 ` Nishanth Aravamudan
2007-06-12 2:02 ` Nishanth Aravamudan
2007-06-12 2:20 ` Christoph Lameter
2007-06-12 2:32 ` Nishanth Aravamudan
2007-06-12 2:54 ` Christoph Lameter
2007-06-12 3:20 ` Nishanth Aravamudan
2007-06-12 3:21 ` Christoph Lameter
2007-06-12 3:31 ` Nishanth Aravamudan
2007-06-12 15:06 ` Lee Schermerhorn
2007-06-12 17:28 ` Nishanth Aravamudan
2007-06-12 18:43 ` Christoph Lameter
2007-06-12 18:48 ` Lee Schermerhorn
2007-06-12 18:51 ` Christoph Lameter
2007-06-12 19:44 ` Lee Schermerhorn
2007-06-12 19:48 ` Christoph Lameter
2007-06-12 19:58 ` Christoph Lameter
2007-06-12 20:01 ` Nishanth Aravamudan
2007-06-13 15:30 ` Lee Schermerhorn
2007-06-13 17:58 ` Nishanth Aravamudan
2007-06-13 18:21 ` Lee Schermerhorn
2007-06-13 19:01 ` Nishanth Aravamudan
2007-06-13 22:51 ` Christoph Lameter
2007-06-14 15:50 ` Lee Schermerhorn
2007-06-14 15:57 ` Christoph Lameter
2007-06-14 16:54 ` Lee Schermerhorn
2007-06-14 16:09 ` Nishanth Aravamudan
2007-06-14 16:15 ` Christoph Lameter
2007-06-14 17:07 ` Lee Schermerhorn
2007-06-14 17:16 ` Christoph Lameter
2007-06-14 18:04 ` Lee Schermerhorn
2007-06-14 22:35 ` Nishanth Aravamudan
2007-06-13 22:50 ` Christoph Lameter
2007-06-13 23:09 ` Nishanth Aravamudan
2007-06-13 23:12 ` Christoph Lameter
2007-06-13 23:18 ` Nishanth Aravamudan
2007-06-13 23:26 ` Christoph Lameter
2007-06-13 23:56 ` Nishanth Aravamudan
2007-06-14 14:23 ` Lee Schermerhorn
2007-06-13 22:49 ` Christoph Lameter
2007-06-12 19:55 ` Nishanth Aravamudan
2007-06-12 18:41 ` Christoph Lameter
2007-06-12 19:07 ` Lee Schermerhorn
2007-06-12 19:13 ` Christoph Lameter
2007-06-11 23:08 ` [PATCH][RFC] Fix INTERLEAVE with memoryless nodes Nishanth Aravamudan
2007-06-11 23:10 ` [PATCH v6][RFC] Fix hugetlb pool allocation with empty nodes Nishanth Aravamudan
2007-06-11 23:11 ` [PATCH][RFC] hugetlb: numafy several functions Nishanth Aravamudan
2007-06-11 23:13 ` [PATCH][RFC] hugetlb: add per-node nr_hugepages sysfs attribute Nishanth Aravamudan
2007-06-11 23:40 ` Christoph Lameter
2007-06-11 23:42 ` Christoph Lameter
2007-06-12 0:19 ` Nishanth Aravamudan
2007-06-12 0:43 ` Christoph Lameter
2007-06-12 2:19 ` Nishanth Aravamudan
2007-06-12 2:22 ` Christoph Lameter
2007-06-12 2:34 ` Nishanth Aravamudan
2007-06-11 23:38 ` [PATCH][RFC] hugetlb: numafy several functions Christoph Lameter
2007-06-11 23:17 ` [PATCH v6][RFC] Fix hugetlb pool allocation with empty nodes Christoph Lameter
2007-06-12 0:15 ` Nishanth Aravamudan
2007-06-12 0:47 ` Christoph Lameter
2007-06-12 2:12 ` Nishanth Aravamudan
2007-06-12 2:21 ` Christoph Lameter
2007-06-12 2:25 ` Christoph Lameter
2007-06-12 2:34 ` Nishanth Aravamudan
2007-06-12 2:55 ` Christoph Lameter
2007-06-12 3:17 ` Nishanth Aravamudan
2007-06-12 3:19 ` Christoph Lameter
2007-06-12 3:30 ` Nishanth Aravamudan
2007-06-12 3:48 ` Christoph Lameter
2007-06-12 5:07 ` Nishanth Aravamudan
2007-06-12 18:47 ` Christoph Lameter
2007-06-12 17:43 ` Nishanth Aravamudan
2007-06-12 18:49 ` Christoph Lameter
2007-06-12 2:33 ` Nishanth Aravamudan
2007-06-12 3:44 ` William Lee Irwin III
2007-06-12 3:50 ` Christoph Lameter
2007-06-12 3:53 ` William Lee Irwin III
2007-06-12 3:53 ` Christoph Lameter
2007-06-12 4:14 ` William Lee Irwin III
2007-06-12 5:09 ` Nishanth Aravamudan
2007-06-12 5:15 ` William Lee Irwin III
2007-06-12 17:36 ` Nishanth Aravamudan
2007-06-12 18:50 ` Christoph Lameter
2007-06-12 17:45 ` Nishanth Aravamudan
2007-06-12 19:13 ` William Lee Irwin III
2007-06-13 0:04 ` [PATCH v7][RFC] " Nishanth Aravamudan
2007-06-13 15:26 ` [PATCH v3][RFC] hugetlb: numafy several functions Nishanth Aravamudan
2007-06-13 15:28 ` [PATCH v3][RFC] hugetlb: add per-node nr_hugepages sysfs attribute Nishanth Aravamudan
2007-06-13 18:23 ` Lee Schermerhorn
2007-06-13 19:19 ` [PATCH v4][RFC] " Nishanth Aravamudan
2007-06-13 20:05 ` Lee Schermerhorn
2007-06-13 20:29 ` Nishanth Aravamudan
2007-06-13 21:02 ` Lee Schermerhorn
2007-07-23 19:23 ` Christoph Lameter
2007-07-23 20:14 ` Lee Schermerhorn
2007-06-13 21:04 ` [PATCH v7][RFC] Fix hugetlb pool allocation with empty nodes Lee Schermerhorn
2007-06-13 21:50 ` [PATCH v7][UPDATE][RFC] " Nishanth Aravamudan
2007-06-12 14:28 ` [PATCH v6][RFC] " Lee Schermerhorn
2007-06-11 23:15 ` [PATCH][RFC] Fix INTERLEAVE with memoryless nodes Christoph Lameter
2007-06-12 0:14 ` [PATCH v2][RFC] " Nishanth Aravamudan
2007-06-12 0:42 ` Christoph Lameter
2007-06-12 0:57 ` Andrew Morton
2007-06-12 1:12 ` Christoph Lameter
2007-06-12 1:41 ` Nishanth Aravamudan
2007-06-12 1:52 ` Andrew Morton
2007-06-12 2:03 ` Nishanth Aravamudan
2007-06-12 14:19 ` [PATCH v2] Add populated_map to account for " Lee Schermerhorn
2007-06-12 17:32 ` Nishanth Aravamudan
2007-06-12 18:45 ` Christoph Lameter
2007-06-12 19:17 ` Lee Schermerhorn
2007-06-12 19:22 ` Christoph Lameter
2007-06-12 19:49 ` Nishanth Aravamudan
2007-06-12 19:51 ` Christoph Lameter
2007-06-12 20:00 ` Nishanth Aravamudan
2007-06-12 20:03 ` Christoph Lameter
2007-06-12 20:10 ` Christoph Lameter
2007-06-12 19:52 ` Christoph Lameter
2007-06-12 19:58 ` Christoph Lameter
2007-06-12 20:00 ` Nishanth Aravamudan
2007-06-12 20:06 ` Christoph Lameter
2007-06-12 14:10 ` [PATCH] " Lee Schermerhorn
2007-06-12 17:35 ` Nishanth Aravamudan
2007-06-12 18:39 ` Christoph Lameter
2007-06-12 18:54 ` Lee Schermerhorn
2007-06-12 19:00 ` Christoph Lameter
2007-06-12 2:27 ` KAMEZAWA Hiroyuki
2007-06-12 2:46 ` Nishanth Aravamudan
2007-06-12 2:53 ` Christoph Lameter
2007-06-12 3:04 ` KAMEZAWA Hiroyuki
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20070611221036.GA14458@us.ibm.com \
--to=nacc@us.ibm.com \
--cc=akpm@linux-foundation.org \
--cc=anton@samba.org \
--cc=clameter@sgi.com \
--cc=lee.schermerhorn@hp.com \
--cc=linux-mm@kvack.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.