The patch titled
Subject: ksm: allocate roots when needed
has been added to the -mm tree. Its filename is
ksm-allocate-roots-when-needed.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/SubmitChecklist when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Hugh Dickins <hughd@xxxxxxxxxx>
Subject: ksm: allocate roots when needed
It is a pity to have MAX_NUMNODES+MAX_NUMNODES tree roots statically
allocated, particularly when very few users will ever actually tune
merge_across_nodes 0 to use more than 1+1 of those trees. Not a big deal
(only 16kB wasted on each machine with CONFIG_MAXSMP), but a pity.
Start off with 1+1 statically allocated, then if merge_across_nodes is
ever tuned, allocate for nr_node_ids+nr_node_ids. Do not attempt to free
up the extra if it's tuned back, that would be a waste of effort.
Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
Cc: Mel Gorman <mgorman@xxxxxxx>
Cc: Petr Holasek <pholasek@xxxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Cc: Izik Eidus <izik.eidus@xxxxxxxxxxxxxxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/ksm.c | 72 ++++++++++++++++++++++++++++++++++++-----------------
1 file changed, 49 insertions(+), 23 deletions(-)
diff -puN mm/ksm.c~ksm-allocate-roots-when-needed mm/ksm.c
--- a/mm/ksm.c~ksm-allocate-roots-when-needed
+++ a/mm/ksm.c
@@ -183,8 +183,10 @@ struct rmap_item {
#define STABLE_FLAG 0x200 /* is listed from the stable tree */
/* The stable and unstable tree heads */
-static struct rb_root root_unstable_tree[MAX_NUMNODES];
-static struct rb_root root_stable_tree[MAX_NUMNODES];
+static struct rb_root one_stable_tree[1] = { RB_ROOT };
+static struct rb_root one_unstable_tree[1] = { RB_ROOT };
+static struct rb_root *root_stable_tree = one_stable_tree;
+static struct rb_root *root_unstable_tree = one_unstable_tree;
/* Recently migrated nodes of stable tree, pending proper placement */
static LIST_HEAD(migrate_nodes);
@@ -224,8 +226,10 @@ static unsigned int ksm_thread_sleep_mil
#ifdef CONFIG_NUMA
/* Zeroed when merging across nodes is not allowed */
static unsigned int ksm_merge_across_nodes = 1;
+static int ksm_nr_node_ids = 1;
#else
#define ksm_merge_across_nodes 1U
+#define ksm_nr_node_ids 1
#endif
#define KSM_RUN_STOP 0
@@ -508,7 +512,7 @@ static void remove_node_from_stable_tree
list_del(&stable_node->list);
else
rb_erase(&stable_node->node,
- &root_stable_tree[NUMA(stable_node->nid)]);
+ root_stable_tree + NUMA(stable_node->nid));
free_stable_node(stable_node);
}
@@ -644,7 +648,7 @@ static void remove_rmap_item_from_tree(s
BUG_ON(age > 1);
if (!age)
rb_erase(&rmap_item->node,
- &root_unstable_tree[NUMA(rmap_item->nid)]);
+ root_unstable_tree + NUMA(rmap_item->nid));
ksm_pages_unshared--;
rmap_item->address &= PAGE_MASK;
}
@@ -742,7 +746,7 @@ static int remove_all_stable_nodes(void)
int nid;
int err = 0;
- for (nid = 0; nid < nr_node_ids; nid++) {
+ for (nid = 0; nid < ksm_nr_node_ids; nid++) {
while (root_stable_tree[nid].rb_node) {
stable_node = rb_entry(root_stable_tree[nid].rb_node,
struct stable_node, node);
@@ -1150,6 +1154,7 @@ static struct page *try_to_merge_two_pag
static struct page *stable_tree_search(struct page *page)
{
int nid;
+ struct rb_root *root;
struct rb_node **new;
struct rb_node *parent;
struct stable_node *stable_node;
@@ -1163,8 +1168,9 @@ static struct page *stable_tree_search(s
}
nid = get_kpfn_nid(page_to_pfn(page));
+ root = root_stable_tree + nid;
again:
- new = &root_stable_tree[nid].rb_node;
+ new = &root->rb_node;
parent = NULL;
while (*new) {
@@ -1219,7 +1225,7 @@ again:
list_del(&page_node->list);
DO_NUMA(page_node->nid = nid);
rb_link_node(&page_node->node, parent, new);
- rb_insert_color(&page_node->node, &root_stable_tree[nid]);
+ rb_insert_color(&page_node->node, root);
get_page(page);
return page;
@@ -1227,11 +1233,10 @@ replace:
if (page_node) {
list_del(&page_node->list);
DO_NUMA(page_node->nid = nid);
- rb_replace_node(&stable_node->node,
- &page_node->node, &root_stable_tree[nid]);
+ rb_replace_node(&stable_node->node, &page_node->node, root);
get_page(page);
} else {
- rb_erase(&stable_node->node, &root_stable_tree[nid]);
+ rb_erase(&stable_node->node, root);
page = NULL;
}
stable_node->head = &migrate_nodes;
@@ -1250,13 +1255,15 @@ static struct stable_node *stable_tree_i
{
int nid;
unsigned long kpfn;
+ struct rb_root *root;
struct rb_node **new;
struct rb_node *parent = NULL;
struct stable_node *stable_node;
kpfn = page_to_pfn(kpage);
nid = get_kpfn_nid(kpfn);
- new = &root_stable_tree[nid].rb_node;
+ root = root_stable_tree + nid;
+ new = &root->rb_node;
while (*new) {
struct page *tree_page;
@@ -1295,7 +1302,7 @@ static struct stable_node *stable_tree_i
set_page_stable_node(kpage, stable_node);
DO_NUMA(stable_node->nid = nid);
rb_link_node(&stable_node->node, parent, new);
- rb_insert_color(&stable_node->node, &root_stable_tree[nid]);
+ rb_insert_color(&stable_node->node, root);
return stable_node;
}
@@ -1325,7 +1332,7 @@ struct rmap_item *unstable_tree_search_i
int nid;
nid = get_kpfn_nid(page_to_pfn(page));
- root = &root_unstable_tree[nid];
+ root = root_unstable_tree + nid;
new = &root->rb_node;
while (*new) {
@@ -1422,7 +1429,7 @@ static void cmp_and_merge_page(struct pa
if (stable_node->head != &migrate_nodes &&
get_kpfn_nid(stable_node->kpfn) != NUMA(stable_node->nid)) {
rb_erase(&stable_node->node,
- &root_stable_tree[NUMA(stable_node->nid)]);
+ root_stable_tree + NUMA(stable_node->nid));
stable_node->head = &migrate_nodes;
list_add(&stable_node->list, stable_node->head);
}
@@ -1574,7 +1581,7 @@ static struct rmap_item *scan_get_next_r
}
}
- for (nid = 0; nid < nr_node_ids; nid++)
+ for (nid = 0; nid < ksm_nr_node_ids; nid++)
root_unstable_tree[nid] = RB_ROOT;
spin_lock(&ksm_mmlist_lock);
@@ -2094,8 +2101,8 @@ static void ksm_check_stable_tree(unsign
struct rb_node *node;
int nid;
- for (nid = 0; nid < nr_node_ids; nid++) {
- node = rb_first(&root_stable_tree[nid]);
+ for (nid = 0; nid < ksm_nr_node_ids; nid++) {
+ node = rb_first(root_stable_tree + nid);
while (node) {
stable_node = rb_entry(node, struct stable_node, node);
if (stable_node->kpfn >= start_pfn &&
@@ -2105,7 +2112,7 @@ static void ksm_check_stable_tree(unsign
* which is why we keep kpfn instead of page*
*/
remove_node_from_stable_tree(stable_node);
- node = rb_first(&root_stable_tree[nid]);
+ node = rb_first(root_stable_tree + nid);
} else
node = rb_next(node);
cond_resched();
@@ -2298,8 +2305,31 @@ static ssize_t merge_across_nodes_store(
if (ksm_merge_across_nodes != knob) {
if (ksm_pages_shared || remove_all_stable_nodes())
err = -EBUSY;
- else
+ else if (root_stable_tree == one_stable_tree) {
+ struct rb_root *buf;
+ /*
+ * This is the first time that we switch away from the
+ * default of merging across nodes: must now allocate
+ * a buffer to hold as many roots as may be needed.
+ * Allocate stable and unstable together:
+ * MAXSMP NODES_SHIFT 10 will use 16kB.
+ */
+ buf = kcalloc(nr_node_ids + nr_node_ids,
+ sizeof(*buf), GFP_KERNEL | __GFP_ZERO);
+ /* Let us assume that RB_ROOT is NULL is zero */
+ if (!buf)
+ err = -ENOMEM;
+ else {
+ root_stable_tree = buf;
+ root_unstable_tree = buf + nr_node_ids;
+ /* Stable tree is empty but not the unstable */
+ root_unstable_tree[0] = one_unstable_tree[0];
+ }
+ }
+ if (!err) {
ksm_merge_across_nodes = knob;
+ ksm_nr_node_ids = knob ? 1 : nr_node_ids;
+ }
}
mutex_unlock(&ksm_thread_mutex);
@@ -2378,15 +2408,11 @@ static int __init ksm_init(void)
{
struct task_struct *ksm_thread;
int err;
- int nid;
err = ksm_slab_init();
if (err)
goto out;
- for (nid = 0; nid < nr_node_ids; nid++)
- root_stable_tree[nid] = RB_ROOT;
-
ksm_thread = kthread_run(ksm_scan_thread, NULL, "ksmd");
if (IS_ERR(ksm_thread)) {
printk(KERN_ERR "ksm: creating kthread failed\n");
_
Patches currently in -mm which might be from hughd@xxxxxxxxxx are
origin.patch
linux-next.patch
revert-x86-mm-make-spurious_fault-check-explicitly-check-the-present-bit.patch
pageattr-prevent-pse-and-gloabl-leftovers-to-confuse-pmd-pte_present-and-pmd_huge.patch
mm-memcg-only-evict-file-pages-when-we-have-plenty.patch
mm-vmscan-save-work-scanning-almost-empty-lru-lists.patch
mm-vmscan-clarify-how-swappiness-highest-priority-memcg-interact.patch
mm-vmscan-improve-comment-on-low-page-cache-handling.patch
mm-vmscan-clean-up-get_scan_count.patch
mm-vmscan-clean-up-get_scan_count-fix.patch
mm-vmscan-compaction-works-against-zones-not-lruvecs.patch
mm-vmscan-compaction-works-against-zones-not-lruvecs-fix.patch
mm-reduce-rmap-overhead-for-ex-ksm-page-copies-created-on-swap-faults.patch
mm-page_allocc-__setup_per_zone_wmarks-make-min_pages-unsigned-long.patch
mm-vmscanc-__zone_reclaim-replace-max_t-with-max.patch
mmksm-use-new-hashtable-implementation.patch
mm-make-madvisemadv_willneed-support-swap-file-prefetch.patch
mm-make-madvisemadv_willneed-support-swap-file-prefetch-fix.patch
mm-make-madvisemadv_willneed-support-swap-file-prefetch-fix-fix.patch
mm-avoid-calling-pgdat_balanced-needlessly.patch
mm-numa-fix-minor-typo-in-numa_next_scan.patch
mm-numa-take-thp-into-account-when-migrating-pages-for-numa-balancing.patch
mm-numa-handle-side-effects-in-count_vm_numa_events-for-config_numa_balancing.patch
mm-move-page-flags-layout-to-separate-header.patch
mm-fold-page-_last_nid-into-page-flags-where-possible.patch
mm-numa-cleanup-flow-of-transhuge-page-migration.patch
mm-dont-inline-page_mapping.patch
swap-make-each-swap-partition-have-one-address_space.patch
swap-make-each-swap-partition-have-one-address_space-fix.patch
swap-make-each-swap-partition-have-one-address_space-fix-fix.patch
swap-add-per-partition-lock-for-swapfile.patch
swap-add-per-partition-lock-for-swapfile-fix-fix-fix.patch
memcg-reduce-the-size-of-struct-memcg-244-fold.patch
memcg-reduce-the-size-of-struct-memcg-244-fold-fix.patch
ksm-allow-trees-per-numa-node.patch
ksm-add-sysfs-abi-documentation.patch
ksm-trivial-tidyups.patch
ksm-trivial-tidyups-fix.patch
ksm-reorganize-ksm_check_stable_tree.patch
ksm-get_ksm_page-locked.patch
ksm-remove-old-stable-nodes-more-thoroughly.patch
ksm-make-ksm-page-migration-possible.patch
ksm-make-merge_across_nodes-migration-safe.patch
ksm-enable-ksm-page-migration.patch
mm-remove-offlining-arg-to-migrate_pages.patch
ksm-stop-hotremove-lockdep-warning.patch
mm-shmem-use-new-radix-tree-iterator.patch
mm-mlockc-document-scary-looking-stack-expansion-mlock-chain.patch
mmu_notifier_unregister-null-pointer-deref-and-multiple-release-callouts.patch
mm-use-up-free-swap-space-before-reaching-oom-kill.patch
memcg-stop-warning-on-memcg_propagate_kmem.patch
mm-use-long-type-for-page-counts-in-mm_populate-and-get_user_pages.patch
mm-accelerate-mm_populate-treatment-of-thp-pages.patch
mm-accelerate-munlock-treatment-of-thp-pages.patch
tmpfs-fix-use-after-free-of-mempolicy-object.patch
tmpfs-fix-mempolicy-object-leaks.patch
tmpfs-fix-mempolicy-object-leaks-fix.patch
ksm-add-some-comments.patch
ksm-treat-unstable-nid-like-in-stable-tree.patch
ksm-shrink-32-bit-rmap_item-back-to-32-bytes.patch
mmksm-foll_migration-do-migration_entry_wait.patch
mmksm-swapoff-might-need-to-copy.patch
mm-cleanup-swapcache-in-do_swap_page.patch
ksm-allocate-roots-when-needed.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
