Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1531084 > unrolled thread

Re: [patch] mm, slab: faster active and free stats

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [patch] mm, slab: faster active and free stats Joonsoo Kim <iamjoonsoo.kim@lge.com> - 2016-11-28 08:40 +0100
    Re: [patch] mm, slab: faster active and free stats David Rientjes <rientjes@google.com> - 2016-11-30 02:00 +0100
      RE: [patch] mm, slab: faster active and free stats 김준수/선임연구원/SW Platform(연)AOT팀(iamjoonsoo.kim@lge.com)          <iamjoonsoo.kim@lge.com> - 2016-12-02 09:10 +0100
        [patch -mm] mm, slab: maintain total slab count instead of active  count David Rientjes <rientjes@google.com> - 2016-12-05 05:30 +0100

#1531084 — Re: [patch] mm, slab: faster active and free stats

On Fri, Nov 11, 2016 at 02:30:39AM -0800, David Rientjes wrote:
> On Fri, 11 Nov 2016, Joonsoo Kim wrote:
> 
> > Hello, David.
> > 
> > Maintaining acitve/free_slab counters looks so complex. And, I think
> > that we don't need to maintain these counters for faster slabinfo.
> > Key point is to remove iterating n->slabs_partial list.
> > 
> > We can calculate active slab/object by following equation as you did in
> > this patch.
> > 
> > active_slab(n) = n->num_slab - the number of free_slab
> > active_object(n) = n->num_slab * cachep->num - n->free_objects
> > 
> > To get the number of free_slab, we need to iterate n->slabs_free list
> > but I guess it would be small enough.
> > 
> > If you don't like to iterate n->slabs_free list in slabinfo, just
> > maintaining the number of slabs_free would be enough.
> > 
> 
> Hi Joonsoo,
> 
> It's a good point, although I don't think the patch has overly complex 
> logic to keep track of slab state.
> 
> We don't prefer to do any iteration in get_slabinfo() since users can 
> read /proc/slabinfo constantly; it's better to just settle the stats when 
> slab state changes instead of repeating an expensive operation over and 
> over if someone is running slabtop(1) or /proc/slabinfo is scraped 
> regularly for stats.
> 
> That said, I imagine there are more clever ways to arrive at the same 
> answer, and you bring up a good point about maintaining a n->num_slabs and 
> n->free_slabs rather than n->active_slabs and n->free_slabs.
> 
> I don't feel strongly about either approach, but I think some improvement, 
> such as what this patch provides, is needed to prevent how expensive 
> simply reading /proc/slabinfo can be.

Hello,

Sorry for long delay.
I agree that this improvement is needed. Could you try the approach
that maintains n->num_slabs and n->free_slabs? I guess that it would be
simpler than this patch so more maintainable.

Thanks.

[toc] | [next] | [standalone]

#1532873

On Mon, 28 Nov 2016, Joonsoo Kim wrote:

> Hello,
> 
> Sorry for long delay.
> I agree that this improvement is needed. Could you try the approach
> that maintains n->num_slabs and n->free_slabs? I guess that it would be
> simpler than this patch so more maintainable.
> 

Ok, what do you think about the following?  I'm not sure it's that much 
more simpler.


mm, slab: track total number of slabs instead of active slabs

Rather than tracking the number of active slabs for each node, track the
total number of slabs.  This is a minor improvement that avoids active
slab tracking when a slab goes from free to partial or partial to free.

Suggested-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: David Rientjes <rientjes@google.com>
---
 mm/slab.c | 48 +++++++++++++++++++++---------------------------
 mm/slab.h |  4 ++--
 2 files changed, 23 insertions(+), 29 deletions(-)

diff --git a/mm/slab.c b/mm/slab.c
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -227,7 +227,7 @@ static void kmem_cache_node_init(struct kmem_cache_node *parent)
 	INIT_LIST_HEAD(&parent->slabs_full);
 	INIT_LIST_HEAD(&parent->slabs_partial);
 	INIT_LIST_HEAD(&parent->slabs_free);
-	parent->active_slabs = 0;
+	parent->total_slabs = 0;
 	parent->free_slabs = 0;
 	parent->shared = NULL;
 	parent->alien = NULL;
@@ -1381,20 +1381,18 @@ slab_out_of_memory(struct kmem_cache *cachep, gfp_t gfpflags, int nodeid)
 		cachep->name, cachep->size, cachep->gfporder);
 
 	for_each_kmem_cache_node(cachep, node, n) {
-		unsigned long active_objs = 0, free_objs = 0;
-		unsigned long active_slabs, num_slabs;
+		unsigned long total_slabs, free_slabs, free_objs;
 
 		spin_lock_irqsave(&n->list_lock, flags);
-		active_slabs = n->active_slabs;
-		num_slabs = active_slabs + n->free_slabs;
-
-		active_objs += (num_slabs * cachep->num) - n->free_objects;
-		free_objs += n->free_objects;
+		total_slabs = n->total_slabs;
+		free_slabs = n->free_slabs;
+		free_objs = n->free_objects;
 		spin_unlock_irqrestore(&n->list_lock, flags);
 
-		pr_warn("  node %d: slabs: %ld/%ld, objs: %ld/%ld, free: %ld\n",
-			node, active_slabs, num_slabs, active_objs,
-			num_slabs * cachep->num, free_objs);
+		pr_warn("  node %d: slabs: %ld/%ld, objs: %ld/%ld\n",
+			node, total_slabs - free_slabs, total_slabs,
+			(total_slabs * cachep->num) - free_objs,
+			total_slabs * cachep->num);
 	}
 #endif
 }
@@ -2307,6 +2305,7 @@ static int drain_freelist(struct kmem_cache *cache,
 		page = list_entry(p, struct page, lru);
 		list_del(&page->lru);
 		n->free_slabs--;
+		n->total_slabs--;
 		/*
 		 * Safe to drop the lock. The slab is no longer linked
 		 * to the cache.
@@ -2741,13 +2740,12 @@ static void cache_grow_end(struct kmem_cache *cachep, struct page *page)
 	n = get_node(cachep, page_to_nid(page));
 
 	spin_lock(&n->list_lock);
+	n->total_slabs++;
 	if (!page->active) {
 		list_add_tail(&page->lru, &(n->slabs_free));
 		n->free_slabs++;
-	} else {
+	} else
 		fixup_slab_list(cachep, n, page, &list);
-		n->active_slabs++;
-	}
 
 	STATS_INC_GROWN(cachep);
 	n->free_objects += cachep->num - page->active;
@@ -2935,10 +2933,8 @@ static struct page *get_first_slab(struct kmem_cache_node *n, bool pfmemalloc)
 	if (sk_memalloc_socks())
 		page = get_valid_first_slab(n, page, &page_is_free, pfmemalloc);
 
-	if (page && page_is_free) {
-		n->active_slabs++;
+	if (page && page_is_free)
 		n->free_slabs--;
-	}
 
 	return page;
 }
@@ -3441,7 +3437,6 @@ static void free_block(struct kmem_cache *cachep, void **objpp,
 		if (page->active == 0) {
 			list_add(&page->lru, &n->slabs_free);
 			n->free_slabs++;
-			n->active_slabs--;
 		} else {
 			/* Unconditionally move a slab to the end of the
 			 * partial list on free - maximum time for the
@@ -3457,6 +3452,7 @@ static void free_block(struct kmem_cache *cachep, void **objpp,
 		page = list_last_entry(&n->slabs_free, struct page, lru);
 		list_move(&page->lru, list);
 		n->free_slabs--;
+		n->total_slabs--;
 	}
 }
 
@@ -4109,8 +4105,8 @@ static void cache_reap(struct work_struct *w)
 void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 {
 	unsigned long active_objs, num_objs, active_slabs;
-	unsigned long num_slabs = 0, free_objs = 0, shared_avail = 0;
-	unsigned long num_slabs_free = 0;
+	unsigned long total_slabs = 0, free_objs = 0, shared_avail = 0;
+	unsigned long free_slabs = 0;
 	int node;
 	struct kmem_cache_node *n;
 
@@ -4118,9 +4114,8 @@ void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 		check_irq_on();
 		spin_lock_irq(&n->list_lock);
 
-		num_slabs += n->active_slabs + n->free_slabs;
-		num_slabs_free += n->free_slabs;
-
+		total_slabs += n->total_slabs;
+		free_slabs += n->free_slabs;
 		free_objs += n->free_objects;
 
 		if (n->shared)
@@ -4128,15 +4123,14 @@ void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 
 		spin_unlock_irq(&n->list_lock);
 	}
-	num_objs = num_slabs * cachep->num;
-	active_slabs = num_slabs - num_slabs_free;
-
+	num_objs = total_slabs * cachep->num;
+	active_slabs = total_slabs - free_slabs;
 	active_objs = num_objs - free_objs;
 
 	sinfo->active_objs = active_objs;
 	sinfo->num_objs = num_objs;
 	sinfo->active_slabs = active_slabs;
-	sinfo->num_slabs = num_slabs;
+	sinfo->num_slabs = total_slabs;
 	sinfo->shared_avail = shared_avail;
 	sinfo->limit = cachep->limit;
 	sinfo->batchcount = cachep->batchcount;
diff --git a/mm/slab.h b/mm/slab.h
--- a/mm/slab.h
+++ b/mm/slab.h
@@ -432,8 +432,8 @@ struct kmem_cache_node {
 	struct list_head slabs_partial;	/* partial list first, better asm code */
 	struct list_head slabs_full;
 	struct list_head slabs_free;
-	unsigned long active_slabs;	/* length of slabs_partial+slabs_full */
-	unsigned long free_slabs;	/* length of slabs_free */
+	unsigned long total_slabs;	/* length of all slab lists */
+	unsigned long free_slabs;	/* length of free slab list only */
 	unsigned long free_objects;
 	unsigned int free_limit;
 	unsigned int colour_next;	/* Per-node cache coloring */

[toc] | [prev] | [next] | [standalone]

#1534724

Hello, David.

There is some problem on my e-mail client so I have to use another one.
Please understand broken reply style.

Yeah, I like this version much. Can we do account slabs_free directly in get_first_slab()
and get_valid_first_slab()? Passing page_is_free isn't needed if we do it directly in
those functions.

One nitpick is that if we don't replace variable name, num_slabs with total_slabs, we will
get less churn the code. However, total_slabs looks better than num_slabs.

Thanks.

-----Original Message-----
From: David Rientjes [mailto:rientjes@google.com] 
Sent: Wednesday, November 30, 2016 9:57 AM
To: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>; Greg Thelen <gthelen@google.com>; Aruna Ramakrishna <aruna.ramakrishna@oracle.com>; Christoph Lameter <cl@linux.com>; linux-kernel@vger.kernel.org; linux-mm@kvack.org
Subject: Re: [patch] mm, slab: faster active and free stats

On Mon, 28 Nov 2016, Joonsoo Kim wrote:

> Hello,
> 
> Sorry for long delay.
> I agree that this improvement is needed. Could you try the approach 
> that maintains n->num_slabs and n->free_slabs? I guess that it would 
> be simpler than this patch so more maintainable.
> 

Ok, what do you think about the following?  I'm not sure it's that much more simpler.


mm, slab: track total number of slabs instead of active slabs

Rather than tracking the number of active slabs for each node, track the
total number of slabs.  This is a minor improvement that avoids active
slab tracking when a slab goes from free to partial or partial to free.

Suggested-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: David Rientjes <rientjes@google.com>
---
 mm/slab.c | 48 +++++++++++++++++++++---------------------------
 mm/slab.h |  4 ++--
 2 files changed, 23 insertions(+), 29 deletions(-)

diff --git a/mm/slab.c b/mm/slab.c
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -227,7 +227,7 @@ static void kmem_cache_node_init(struct kmem_cache_node *parent)
 	INIT_LIST_HEAD(&parent->slabs_full);
 	INIT_LIST_HEAD(&parent->slabs_partial);
 	INIT_LIST_HEAD(&parent->slabs_free);
-	parent->active_slabs = 0;
+	parent->total_slabs = 0;
 	parent->free_slabs = 0;
 	parent->shared = NULL;
 	parent->alien = NULL;
@@ -1381,20 +1381,18 @@ slab_out_of_memory(struct kmem_cache *cachep, gfp_t gfpflags, int nodeid)
 		cachep->name, cachep->size, cachep->gfporder);
 
 	for_each_kmem_cache_node(cachep, node, n) {
-		unsigned long active_objs = 0, free_objs = 0;
-		unsigned long active_slabs, num_slabs;
+		unsigned long total_slabs, free_slabs, free_objs;
 
 		spin_lock_irqsave(&n->list_lock, flags);
-		active_slabs = n->active_slabs;
-		num_slabs = active_slabs + n->free_slabs;
-
-		active_objs += (num_slabs * cachep->num) - n->free_objects;
-		free_objs += n->free_objects;
+		total_slabs = n->total_slabs;
+		free_slabs = n->free_slabs;
+		free_objs = n->free_objects;
 		spin_unlock_irqrestore(&n->list_lock, flags);
 
-		pr_warn("  node %d: slabs: %ld/%ld, objs: %ld/%ld, free: %ld\n",
-			node, active_slabs, num_slabs, active_objs,
-			num_slabs * cachep->num, free_objs);
+		pr_warn("  node %d: slabs: %ld/%ld, objs: %ld/%ld\n",
+			node, total_slabs - free_slabs, total_slabs,
+			(total_slabs * cachep->num) - free_objs,
+			total_slabs * cachep->num);
 	}
 #endif
 }
@@ -2307,6 +2305,7 @@ static int drain_freelist(struct kmem_cache *cache,
 		page = list_entry(p, struct page, lru);
 		list_del(&page->lru);
 		n->free_slabs--;
+		n->total_slabs--;
 		/*
 		 * Safe to drop the lock. The slab is no longer linked
 		 * to the cache.
@@ -2741,13 +2740,12 @@ static void cache_grow_end(struct kmem_cache *cachep, struct page *page)
 	n = get_node(cachep, page_to_nid(page));
 
 	spin_lock(&n->list_lock);
+	n->total_slabs++;
 	if (!page->active) {
 		list_add_tail(&page->lru, &(n->slabs_free));
 		n->free_slabs++;
-	} else {
+	} else
 		fixup_slab_list(cachep, n, page, &list);
-		n->active_slabs++;
-	}
 
 	STATS_INC_GROWN(cachep);
 	n->free_objects += cachep->num - page->active;
@@ -2935,10 +2933,8 @@ static struct page *get_first_slab(struct kmem_cache_node *n, bool pfmemalloc)
 	if (sk_memalloc_socks())
 		page = get_valid_first_slab(n, page, &page_is_free, pfmemalloc);
 
-	if (page && page_is_free) {
-		n->active_slabs++;
+	if (page && page_is_free)
 		n->free_slabs--;
-	}
 
 	return page;
 }
@@ -3441,7 +3437,6 @@ static void free_block(struct kmem_cache *cachep, void **objpp,
 		if (page->active == 0) {
 			list_add(&page->lru, &n->slabs_free);
 			n->free_slabs++;
-			n->active_slabs--;
 		} else {
 			/* Unconditionally move a slab to the end of the
 			 * partial list on free - maximum time for the
@@ -3457,6 +3452,7 @@ static void free_block(struct kmem_cache *cachep, void **objpp,
 		page = list_last_entry(&n->slabs_free, struct page, lru);
 		list_move(&page->lru, list);
 		n->free_slabs--;
+		n->total_slabs--;
 	}
 }
 
@@ -4109,8 +4105,8 @@ static void cache_reap(struct work_struct *w)
 void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 {
 	unsigned long active_objs, num_objs, active_slabs;
-	unsigned long num_slabs = 0, free_objs = 0, shared_avail = 0;
-	unsigned long num_slabs_free = 0;
+	unsigned long total_slabs = 0, free_objs = 0, shared_avail = 0;
+	unsigned long free_slabs = 0;
 	int node;
 	struct kmem_cache_node *n;
 
@@ -4118,9 +4114,8 @@ void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 		check_irq_on();
 		spin_lock_irq(&n->list_lock);
 
-		num_slabs += n->active_slabs + n->free_slabs;
-		num_slabs_free += n->free_slabs;
-
+		total_slabs += n->total_slabs;
+		free_slabs += n->free_slabs;
 		free_objs += n->free_objects;
 
 		if (n->shared)
@@ -4128,15 +4123,14 @@ void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 
 		spin_unlock_irq(&n->list_lock);
 	}
-	num_objs = num_slabs * cachep->num;
-	active_slabs = num_slabs - num_slabs_free;
-
+	num_objs = total_slabs * cachep->num;
+	active_slabs = total_slabs - free_slabs;
 	active_objs = num_objs - free_objs;
 
 	sinfo->active_objs = active_objs;
 	sinfo->num_objs = num_objs;
 	sinfo->active_slabs = active_slabs;
-	sinfo->num_slabs = num_slabs;
+	sinfo->num_slabs = total_slabs;
 	sinfo->shared_avail = shared_avail;
 	sinfo->limit = cachep->limit;
 	sinfo->batchcount = cachep->batchcount;
diff --git a/mm/slab.h b/mm/slab.h
--- a/mm/slab.h
+++ b/mm/slab.h
@@ -432,8 +432,8 @@ struct kmem_cache_node {
 	struct list_head slabs_partial;	/* partial list first, better asm code */
 	struct list_head slabs_full;
 	struct list_head slabs_free;
-	unsigned long active_slabs;	/* length of slabs_partial+slabs_full */
-	unsigned long free_slabs;	/* length of slabs_free */
+	unsigned long total_slabs;	/* length of all slab lists */
+	unsigned long free_slabs;	/* length of free slab list only */
 	unsigned long free_objects;
 	unsigned int free_limit;
 	unsigned int colour_next;	/* Per-node cache coloring */

[toc] | [prev] | [next] | [standalone]

#1535803 — [patch -mm] mm, slab: maintain total slab count instead of active count

Rather than tracking the number of active slabs for each node, track the
total number of slabs.  This is a minor improvement that avoids active
slab tracking when a slab goes from free to partial or partial to free.

For slab debugging, this also removes an explicit free count since it can 
easily be inferred by the difference in number of total objects and number 
of active objects.

Suggested-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: David Rientjes <rientjes@google.com>
---
 For -mm because this depends on 
 mm-slab-faster-active-and-free-stats.patch

 mm/slab.c | 70 ++++++++++++++++++++++++++-------------------------------------
 mm/slab.h |  4 ++--
 2 files changed, 31 insertions(+), 43 deletions(-)

diff --git a/mm/slab.c b/mm/slab.c
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -227,7 +227,7 @@ static void kmem_cache_node_init(struct kmem_cache_node *parent)
 	INIT_LIST_HEAD(&parent->slabs_full);
 	INIT_LIST_HEAD(&parent->slabs_partial);
 	INIT_LIST_HEAD(&parent->slabs_free);
-	parent->active_slabs = 0;
+	parent->total_slabs = 0;
 	parent->free_slabs = 0;
 	parent->shared = NULL;
 	parent->alien = NULL;
@@ -1376,20 +1376,18 @@ slab_out_of_memory(struct kmem_cache *cachep, gfp_t gfpflags, int nodeid)
 		cachep->name, cachep->size, cachep->gfporder);
 
 	for_each_kmem_cache_node(cachep, node, n) {
-		unsigned long active_objs = 0, free_objs = 0;
-		unsigned long active_slabs, num_slabs;
+		unsigned long total_slabs, free_slabs, free_objs;
 
 		spin_lock_irqsave(&n->list_lock, flags);
-		active_slabs = n->active_slabs;
-		num_slabs = active_slabs + n->free_slabs;
-
-		active_objs += (num_slabs * cachep->num) - n->free_objects;
-		free_objs += n->free_objects;
+		total_slabs = n->total_slabs;
+		free_slabs = n->free_slabs;
+		free_objs = n->free_objects;
 		spin_unlock_irqrestore(&n->list_lock, flags);
 
-		pr_warn("  node %d: slabs: %ld/%ld, objs: %ld/%ld, free: %ld\n",
-			node, active_slabs, num_slabs, active_objs,
-			num_slabs * cachep->num, free_objs);
+		pr_warn("  node %d: slabs: %ld/%ld, objs: %ld/%ld\n",
+			node, total_slabs - free_slabs, total_slabs,
+			(total_slabs * cachep->num) - free_objs,
+			total_slabs * cachep->num);
 	}
 #endif
 }
@@ -2302,6 +2300,7 @@ static int drain_freelist(struct kmem_cache *cache,
 		page = list_entry(p, struct page, lru);
 		list_del(&page->lru);
 		n->free_slabs--;
+		n->total_slabs--;
 		/*
 		 * Safe to drop the lock. The slab is no longer linked
 		 * to the cache.
@@ -2736,13 +2735,12 @@ static void cache_grow_end(struct kmem_cache *cachep, struct page *page)
 	n = get_node(cachep, page_to_nid(page));
 
 	spin_lock(&n->list_lock);
+	n->total_slabs++;
 	if (!page->active) {
 		list_add_tail(&page->lru, &(n->slabs_free));
 		n->free_slabs++;
-	} else {
+	} else
 		fixup_slab_list(cachep, n, page, &list);
-		n->active_slabs++;
-	}
 
 	STATS_INC_GROWN(cachep);
 	n->free_objects += cachep->num - page->active;
@@ -2869,7 +2867,7 @@ static inline void fixup_slab_list(struct kmem_cache *cachep,
 
 /* Try to find non-pfmemalloc slab if needed */
 static noinline struct page *get_valid_first_slab(struct kmem_cache_node *n,
-			struct page *page, bool *page_is_free, bool pfmemalloc)
+					struct page *page, bool pfmemalloc)
 {
 	if (!page)
 		return NULL;
@@ -2888,10 +2886,9 @@ static noinline struct page *get_valid_first_slab(struct kmem_cache_node *n,
 
 	/* Move pfmemalloc slab to the end of list to speed up next search */
 	list_del(&page->lru);
-	if (*page_is_free) {
-		WARN_ON(page->active);
+	if (!page->active) {
 		list_add_tail(&page->lru, &n->slabs_free);
-		*page_is_free = false;
+		n->free_slabs++;
 	} else
 		list_add_tail(&page->lru, &n->slabs_partial);
 
@@ -2903,7 +2900,7 @@ static noinline struct page *get_valid_first_slab(struct kmem_cache_node *n,
 	n->free_touched = 1;
 	list_for_each_entry(page, &n->slabs_free, lru) {
 		if (!PageSlabPfmemalloc(page)) {
-			*page_is_free = true;
+			n->free_slabs--;
 			return page;
 		}
 	}
@@ -2914,26 +2911,19 @@ static noinline struct page *get_valid_first_slab(struct kmem_cache_node *n,
 static struct page *get_first_slab(struct kmem_cache_node *n, bool pfmemalloc)
 {
 	struct page *page;
-	bool page_is_free = false;
 
 	assert_spin_locked(&n->list_lock);
-	page = list_first_entry_or_null(&n->slabs_partial,
-			struct page, lru);
+	page = list_first_entry_or_null(&n->slabs_partial, struct page, lru);
 	if (!page) {
 		n->free_touched = 1;
-		page = list_first_entry_or_null(&n->slabs_free,
-				struct page, lru);
+		page = list_first_entry_or_null(&n->slabs_free, struct page,
+						lru);
 		if (page)
-			page_is_free = true;
+			n->free_slabs--;
 	}
 
 	if (sk_memalloc_socks())
-		page = get_valid_first_slab(n, page, &page_is_free, pfmemalloc);
-
-	if (page && page_is_free) {
-		n->active_slabs++;
-		n->free_slabs--;
-	}
+		page = get_valid_first_slab(n, page, pfmemalloc);
 
 	return page;
 }
@@ -3436,7 +3426,6 @@ static void free_block(struct kmem_cache *cachep, void **objpp,
 		if (page->active == 0) {
 			list_add(&page->lru, &n->slabs_free);
 			n->free_slabs++;
-			n->active_slabs--;
 		} else {
 			/* Unconditionally move a slab to the end of the
 			 * partial list on free - maximum time for the
@@ -3452,6 +3441,7 @@ static void free_block(struct kmem_cache *cachep, void **objpp,
 		page = list_last_entry(&n->slabs_free, struct page, lru);
 		list_move(&page->lru, list);
 		n->free_slabs--;
+		n->total_slabs--;
 	}
 }
 
@@ -4104,8 +4094,8 @@ static void cache_reap(struct work_struct *w)
 void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 {
 	unsigned long active_objs, num_objs, active_slabs;
-	unsigned long num_slabs = 0, free_objs = 0, shared_avail = 0;
-	unsigned long num_slabs_free = 0;
+	unsigned long total_slabs = 0, free_objs = 0, shared_avail = 0;
+	unsigned long free_slabs = 0;
 	int node;
 	struct kmem_cache_node *n;
 
@@ -4113,9 +4103,8 @@ void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 		check_irq_on();
 		spin_lock_irq(&n->list_lock);
 
-		num_slabs += n->active_slabs + n->free_slabs;
-		num_slabs_free += n->free_slabs;
-
+		total_slabs += n->total_slabs;
+		free_slabs += n->free_slabs;
 		free_objs += n->free_objects;
 
 		if (n->shared)
@@ -4123,15 +4112,14 @@ void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)
 
 		spin_unlock_irq(&n->list_lock);
 	}
-	num_objs = num_slabs * cachep->num;
-	active_slabs = num_slabs - num_slabs_free;
-
+	num_objs = total_slabs * cachep->num;
+	active_slabs = total_slabs - free_slabs;
 	active_objs = num_objs - free_objs;
 
 	sinfo->active_objs = active_objs;
 	sinfo->num_objs = num_objs;
 	sinfo->active_slabs = active_slabs;
-	sinfo->num_slabs = num_slabs;
+	sinfo->num_slabs = total_slabs;
 	sinfo->shared_avail = shared_avail;
 	sinfo->limit = cachep->limit;
 	sinfo->batchcount = cachep->batchcount;
diff --git a/mm/slab.h b/mm/slab.h
--- a/mm/slab.h
+++ b/mm/slab.h
@@ -447,8 +447,8 @@ struct kmem_cache_node {
 	struct list_head slabs_partial;	/* partial list first, better asm code */
 	struct list_head slabs_full;
 	struct list_head slabs_free;
-	unsigned long active_slabs;	/* length of slabs_partial+slabs_full */
-	unsigned long free_slabs;	/* length of slabs_free */
+	unsigned long total_slabs;	/* length of all slab lists */
+	unsigned long free_slabs;	/* length of free slab list only */
 	unsigned long free_objects;
 	unsigned int free_limit;
 	unsigned int colour_next;	/* Per-node cache coloring */

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web