slab: defer slab_destroy in free_block()
In free_block(), if freeing object makes new free slab and number of
free_objects exceeds free_limit, we start to destroy this new free slab
with holding the kmem_cache node lock. Holding the lock is useless and,
generally, holding a lock as least as possible is good thing. I never
measure performance effect of this, but we'd be better not to hold the
lock as much as possible.
Commented by Christoph:
This is also good because kmem_cache_free is no longer called while
holding the node lock. So we avoid one case of recursion.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
diff --git a/mm/slab.c b/mm/slab.c
index 205632c..f6ad8d3 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -242,7 +242,8 @@
static int drain_freelist(struct kmem_cache *cache,
struct kmem_cache_node *n, int tofree);
static void free_block(struct kmem_cache *cachep, void **objpp, int len,
- int node);
+ int node, struct list_head *list);
+static void slabs_destroy(struct kmem_cache *cachep, struct list_head *list);
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp);
static void cache_reap(struct work_struct *unused);
@@ -1030,6 +1031,7 @@
struct array_cache *ac, int node)
{
struct kmem_cache_node *n = get_node(cachep, node);
+ LIST_HEAD(list);
if (ac->avail) {
spin_lock(&n->list_lock);
@@ -1041,9 +1043,10 @@
if (n->shared)
transfer_objects(n->shared, ac, ac->limit);
- free_block(cachep, ac->entry, ac->avail, node);
+ free_block(cachep, ac->entry, ac->avail, node, &list);
ac->avail = 0;
spin_unlock(&n->list_lock);
+ slabs_destroy(cachep, &list);
}
}
@@ -1087,6 +1090,7 @@
struct kmem_cache_node *n;
struct array_cache *alien = NULL;
int node;
+ LIST_HEAD(list);
node = numa_mem_id();
@@ -1111,8 +1115,9 @@
} else {
n = get_node(cachep, nodeid);
spin_lock(&n->list_lock);
- free_block(cachep, &objp, 1, nodeid);
+ free_block(cachep, &objp, 1, nodeid, &list);
spin_unlock(&n->list_lock);
+ slabs_destroy(cachep, &list);
}
return 1;
}
@@ -1182,6 +1187,7 @@
struct array_cache *nc;
struct array_cache *shared;
struct array_cache **alien;
+ LIST_HEAD(list);
/* cpu is dead; no one can alloc from it. */
nc = cachep->array[cpu];
@@ -1196,7 +1202,7 @@
/* Free limit for this kmem_cache_node */
n->free_limit -= cachep->batchcount;
if (nc)
- free_block(cachep, nc->entry, nc->avail, node);
+ free_block(cachep, nc->entry, nc->avail, node, &list);
if (!cpumask_empty(mask)) {
spin_unlock_irq(&n->list_lock);
@@ -1206,7 +1212,7 @@
shared = n->shared;
if (shared) {
free_block(cachep, shared->entry,
- shared->avail, node);
+ shared->avail, node, &list);
n->shared = NULL;
}
@@ -1221,6 +1227,7 @@
free_alien_cache(alien);
}
free_array_cache:
+ slabs_destroy(cachep, &list);
kfree(nc);
}
/*
@@ -2056,6 +2063,16 @@
kmem_cache_free(cachep->freelist_cache, freelist);
}
+static void slabs_destroy(struct kmem_cache *cachep, struct list_head *list)
+{
+ struct page *page, *n;
+
+ list_for_each_entry_safe(page, n, list, lru) {
+ list_del(&page->lru);
+ slab_destroy(cachep, page);
+ }
+}
+
/**
* calculate_slab_order - calculate size (page order) of slabs
* @cachep: pointer to the cache that is being created
@@ -2459,13 +2476,15 @@
struct array_cache *ac;
int node = numa_mem_id();
struct kmem_cache_node *n;
+ LIST_HEAD(list);
check_irq_off();
ac = cpu_cache_get(cachep);
n = get_node(cachep, node);
spin_lock(&n->list_lock);
- free_block(cachep, ac->entry, ac->avail, node);
+ free_block(cachep, ac->entry, ac->avail, node, &list);
spin_unlock(&n->list_lock);
+ slabs_destroy(cachep, &list);
ac->avail = 0;
}
@@ -3393,9 +3412,10 @@
/*
* Caller needs to acquire correct kmem_cache_node's list_lock
+ * @list: List of detached free slabs should be freed by caller
*/
-static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
- int node)
+static void free_block(struct kmem_cache *cachep, void **objpp,
+ int nr_objects, int node, struct list_head *list)
{
int i;
struct kmem_cache_node *n = get_node(cachep, node);
@@ -3418,13 +3438,7 @@
if (page->active == 0) {
if (n->free_objects > n->free_limit) {
n->free_objects -= cachep->num;
- /* No need to drop any previously held
- * lock here, even if we have a off-slab slab
- * descriptor it is guaranteed to come from
- * a different cache, refer to comments before
- * alloc_slabmgmt.
- */
- slab_destroy(cachep, page);
+ list_add_tail(&page->lru, list);
} else {
list_add(&page->lru, &n->slabs_free);
}
@@ -3443,6 +3457,7 @@
int batchcount;
struct kmem_cache_node *n;
int node = numa_mem_id();
+ LIST_HEAD(list);
batchcount = ac->batchcount;
#if DEBUG
@@ -3464,7 +3479,7 @@
}
}
- free_block(cachep, ac->entry, batchcount, node);
+ free_block(cachep, ac->entry, batchcount, node, &list);
free_done:
#if STATS
{
@@ -3485,6 +3500,7 @@
}
#endif
spin_unlock(&n->list_lock);
+ slabs_destroy(cachep, &list);
ac->avail -= batchcount;
memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
}
@@ -3765,12 +3781,13 @@
n = get_node(cachep, node);
if (n) {
struct array_cache *shared = n->shared;
+ LIST_HEAD(list);
spin_lock_irq(&n->list_lock);
if (shared)
free_block(cachep, shared->entry,
- shared->avail, node);
+ shared->avail, node, &list);
n->shared = new_shared;
if (!n->alien) {
@@ -3780,6 +3797,7 @@
n->free_limit = (1 + nr_cpus_node(node)) *
cachep->batchcount + cachep->num;
spin_unlock_irq(&n->list_lock);
+ slabs_destroy(cachep, &list);
kfree(shared);
free_alien_cache(new_alien);
continue;
@@ -3869,6 +3887,7 @@
cachep->shared = shared;
for_each_online_cpu(i) {
+ LIST_HEAD(list);
struct array_cache *ccold = new->new[i];
int node;
struct kmem_cache_node *n;
@@ -3879,8 +3898,9 @@
node = cpu_to_mem(i);
n = get_node(cachep, node);
spin_lock_irq(&n->list_lock);
- free_block(cachep, ccold->entry, ccold->avail, node);
+ free_block(cachep, ccold->entry, ccold->avail, node, &list);
spin_unlock_irq(&n->list_lock);
+ slabs_destroy(cachep, &list);
kfree(ccold);
}
kfree(new);
@@ -3988,6 +4008,7 @@
static void drain_array(struct kmem_cache *cachep, struct kmem_cache_node *n,
struct array_cache *ac, int force, int node)
{
+ LIST_HEAD(list);
int tofree;
if (!ac || !ac->avail)
@@ -4000,12 +4021,13 @@
tofree = force ? ac->avail : (ac->limit + 4) / 5;
if (tofree > ac->avail)
tofree = (ac->avail + 1) / 2;
- free_block(cachep, ac->entry, tofree, node);
+ free_block(cachep, ac->entry, tofree, node, &list);
ac->avail -= tofree;
memmove(ac->entry, &(ac->entry[tofree]),
sizeof(void *) * ac->avail);
}
spin_unlock_irq(&n->list_lock);
+ slabs_destroy(cachep, &list);
}
}