[PATCH] btree: Fix the bug to release whole btree nodes

[PATCH] btree: Fix the bug to release whole btree nodes

[Minfei Huang]

I use the btree which pickes up the latest version of kernel source
3.14-rc2 in my own module. When the btree module is removed, a warning arised:

kmem_cache_destroy btree_node: Slab cache still has objects
CPU: 13 PID: 9150 Comm: rmmod Tainted: GF          O 3.14.0-rc2 #1
Hardware name: Inspur NF5270M3/NF5270M3, BIOS CHEETAH_2.1.3 09/10/2013
ffff881ff8643b18 ffff881ffdc23ea8 ffffffff815a4ecc 0000000000000000
ffff881ff8643ac0 ffff881ffdc23ec8 ffffffff811610df 0000000000000880
ffffffffa057da60 ffff881ffdc23ed8 ffffffffa057d57c ffff881ffdc23f78
Call Trace:
[<ffffffff815a4ecc>] dump_stack+0x49/0x5d
[<ffffffff811610df>] kmem_cache_destroy+0xcf/0xe0
[<ffffffffa057d57c>] btree_module_exit+0x10/0x12 [btree]
[<ffffffff810d7948>] SyS_delete_module+0x198/0x1f0
[<ffffffff815aac89>] ? retint_swapgs+0xe/0x13
[<ffffffff810a561d>] ? trace_hardirqs_on_caller+0xfd/0x1c0
[<ffffffff812addde>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[<ffffffff815b3652>] system_call_fastpath+0x16/0x1b

The cause is that it doesn't release the last btree node,
when height = 1 and fill = 1.

Signed-off-by: Minfei Huang <huangminfei@ucloud.cn>
CC: Joern Engel <joern@logfs.org>
CC: Johannes Berg <johannes@sipsolutions.net>
---
 lib/btree.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/lib/btree.c b/lib/btree.c
index f9a4846..725bf8b 100644
--- a/lib/btree.c
+++ b/lib/btree.c
@@ -198,6 +198,8 @@ EXPORT_SYMBOL_GPL(btree_init);
 
 void btree_destroy(struct btree_head *head)
 {
+	if (head->node)
+		mempool_free(head->node, head->mempool);
 	mempool_destroy(head->mempool);
 	head->mempool = NULL;
 }
-- 
1.7.1

Re: [PATCH] btree: Fix the bug to release whole btree nodes

[Andrew Morton]

On Thu,  8 May 2014 12:21:12 +0800 Minfei Huang <huangminfei@ucloud.cn> wrote:

> I use the btree which pickes up the latest version of kernel source
> 3.14-rc2 in my own module. When the btree module is removed, a warning arised:
> 
> kmem_cache_destroy btree_node: Slab cache still has objects
> CPU: 13 PID: 9150 Comm: rmmod Tainted: GF          O 3.14.0-rc2 #1
> Hardware name: Inspur NF5270M3/NF5270M3, BIOS CHEETAH_2.1.3 09/10/2013
> ffff881ff8643b18 ffff881ffdc23ea8 ffffffff815a4ecc 0000000000000000
> ffff881ff8643ac0 ffff881ffdc23ec8 ffffffff811610df 0000000000000880
> ffffffffa057da60 ffff881ffdc23ed8 ffffffffa057d57c ffff881ffdc23f78
> Call Trace:
> [<ffffffff815a4ecc>] dump_stack+0x49/0x5d
> [<ffffffff811610df>] kmem_cache_destroy+0xcf/0xe0
> [<ffffffffa057d57c>] btree_module_exit+0x10/0x12 [btree]
> [<ffffffff810d7948>] SyS_delete_module+0x198/0x1f0
> [<ffffffff815aac89>] ? retint_swapgs+0xe/0x13
> [<ffffffff810a561d>] ? trace_hardirqs_on_caller+0xfd/0x1c0
> [<ffffffff812addde>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> [<ffffffff815b3652>] system_call_fastpath+0x16/0x1b
> 
> The cause is that it doesn't release the last btree node,
> when height = 1 and fill = 1.

Thanks.

> --- a/lib/btree.c
> +++ b/lib/btree.c
> @@ -198,6 +198,8 @@ EXPORT_SYMBOL_GPL(btree_init);
>  
>  void btree_destroy(struct btree_head *head)
>  {
> +	if (head->node)
> +		mempool_free(head->node, head->mempool);
>  	mempool_destroy(head->mempool);
>  	head->mempool = NULL;

We don't really need the test - mempool_free(NULL, ...) is an OK thing
to do.  We conventionally will rely upon this for optimisation reasons,
although not having the test can sometimes make the code less clear.

--- a/lib/btree.c~lib-btreec-fix-leak-of-whole-btree-nodes-fix
+++ a/lib/btree.c
@@ -198,8 +198,7 @@ EXPORT_SYMBOL_GPL(btree_init);
 
 void btree_destroy(struct btree_head *head)
 {
-	if (head->node)
-		mempool_free(head->node, head->mempool);
+	mempool_free(head->node, head->mempool);
 	mempool_destroy(head->mempool);
 	head->mempool = NULL;
 }
_

[PATCH] btree: Fix the bug to release whole btree nodes

[Minfei Huang]

I use the btree which pickes up the latest version of kernel source
3.14-rc2 in my own module. When the btree module is removed, a warning arised:

kmem_cache_destroy btree_node: Slab cache still has objects
CPU: 13 PID: 9150 Comm: rmmod Tainted: GF          O 3.14.0-rc2 #1
Hardware name: Inspur NF5270M3/NF5270M3, BIOS CHEETAH_2.1.3 09/10/2013
ffff881ff8643b18 ffff881ffdc23ea8 ffffffff815a4ecc 0000000000000000
ffff881ff8643ac0 ffff881ffdc23ec8 ffffffff811610df 0000000000000880
ffffffffa057da60 ffff881ffdc23ed8 ffffffffa057d57c ffff881ffdc23f78
Call Trace:
[<ffffffff815a4ecc>] dump_stack+0x49/0x5d
[<ffffffff811610df>] kmem_cache_destroy+0xcf/0xe0
[<ffffffffa057d57c>] btree_module_exit+0x10/0x12 [btree]
[<ffffffff810d7948>] SyS_delete_module+0x198/0x1f0
[<ffffffff815aac89>] ? retint_swapgs+0xe/0x13
[<ffffffff810a561d>] ? trace_hardirqs_on_caller+0xfd/0x1c0
[<ffffffff812addde>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[<ffffffff815b3652>] system_call_fastpath+0x16/0x1b

The cause is that it doesn't release the last btree node,
when height = 1 and fill = 1.

Signed-off-by: Minfei Huang <huangminfei@ucloud.cn>
CC: Joern Engel <joern@logfs.org>
CC: Johannes Berg <johannes@sipsolutions.net>
---
 lib/btree.c |   12 ++++++++++++
 1 files changed, 12 insertions(+), 0 deletions(-)

diff --git a/lib/btree.c b/lib/btree.c
index f9a4846..fd17d26 100644
--- a/lib/btree.c
+++ b/lib/btree.c
@@ -444,6 +444,16 @@ static void btree_shrink(struct btree_head *head, struct btree_geo *geo)
 	mempool_free(node, head->mempool);
 }
 
+static void btree_shrink_head(struct btree_head *head)
+{
+	if (head->height != 1)
+		return;
+
+	mempool_free(head->node, head->mempool);
+	head->node = NULL;
+	head->height--;
+}
+
 static int btree_insert_level(struct btree_head *head, struct btree_geo *geo,
 			      unsigned long *key, void *val, int level,
 			      gfp_t gfp)
@@ -620,6 +630,8 @@ static void *btree_remove_level(struct btree_head *head, struct btree_geo *geo,
 			rebalance(head, geo, key, level, node, fill - 1);
 		else if (fill - 1 == 1)
 			btree_shrink(head, geo);
+		else if (fill - 1 == 0)
+			btree_shrink_head(head);
 	}
 
 	return ret;
-- 
1.7.1