[PATCH 1/2] cgroup: make serial_nr_cursor available throughout cgroup.c

[PATCH 1/2] cgroup: make serial_nr_cursor available throughout cgroup.c

[Li Zefan]

The next patch will use it to determine if a cgroup is newly created
while we're iterating the cgroup hierarchy.

Signed-off-by: Li Zefan <lizefan@huawei.com>
---

These two patches are based on the fixes I just sent.

---
 kernel/cgroup.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index e8a40e3..b1d98f2 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -198,6 +198,15 @@ static DEFINE_IDR(cgroup_hierarchy_idr);
 
 static struct cgroup_name root_cgroup_name = { .name = "/" };
 
+/*
+ * Assign a monotonically increasing serial number.  It guarantees
+ * cgroups with bigger numbers are newer than those with smaller
+ * numbers. Also with the list appending, it guarantees that sibling
+ * cgroups are always sorted in the ascending serial number order on
+ * the parent's ->children.
+ */
+static atomic64_t cgroup_serial_nr_cursor = ATOMIC64_INIT(0);
+
 /* This flag indicates whether tasks in the fork and exit paths should
  * check for fork/exit handlers to call. This avoids us having to do
  * extra work in the fork/exit path if none of the subsystems need to
@@ -4212,7 +4221,6 @@ static void offline_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
 static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 			     umode_t mode)
 {
-	static atomic64_t serial_nr_cursor = ATOMIC64_INIT(0);
 	struct cgroup *cgrp;
 	struct cgroup_name *name;
 	struct cgroupfs_root *root = parent->root;
@@ -4305,7 +4313,7 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 	 * sorted in the ascending serial number order on the parent's
 	 * ->children.
 	 */
-	cgrp->serial_nr = atomic64_inc_return(&serial_nr_cursor);
+	cgrp->serial_nr = atomic64_inc_return(&cgroup_serial_nr_cursor);
 
 	/* allocation complete, commit to creation */
 	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
-- 
1.8.0.2

[PATCH 2/2] cgroup: convert cgroup_cft_commit() to use cgroup_for_each_descendant_pre()

[Li Zefan]

We used root->allcg_list to iterate cgroup hierarchy because at that time
cgroup_for_each_descendant_pre() hasn't been invented.

Signed-off-by: Li Zefan <lizefan@huawei.com>
---
 include/linux/cgroup.h |  6 -----
 kernel/cgroup.c        | 62 +++++++++++++++++++++++++++-----------------------
 2 files changed, 33 insertions(+), 35 deletions(-)

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 835918c..f70b116 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -206,9 +206,6 @@ struct cgroup {
 	 */
 	struct list_head cset_links;
 
-	struct list_head allcg_node;	/* cgroupfs_root->allcg_list */
-	struct list_head cft_q_node;	/* used during cftype add/rm */
-
 	/*
 	 * Linked list running through all cgroups that can
 	 * potentially be reaped by the release agent. Protected by
@@ -317,9 +314,6 @@ struct cgroupfs_root {
 	/* A list running through the active hierarchies */
 	struct list_head root_list;
 
-	/* All cgroups on this root, cgroup_mutex protected */
-	struct list_head allcg_list;
-
 	/* Hierarchy-specific flags */
 	unsigned long flags;
 
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index b1d98f2..bc46c28 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -1399,7 +1399,6 @@ static void init_cgroup_housekeeping(struct cgroup *cgrp)
 	INIT_LIST_HEAD(&cgrp->children);
 	INIT_LIST_HEAD(&cgrp->files);
 	INIT_LIST_HEAD(&cgrp->cset_links);
-	INIT_LIST_HEAD(&cgrp->allcg_node);
 	INIT_LIST_HEAD(&cgrp->release_list);
 	INIT_LIST_HEAD(&cgrp->pidlists);
 	mutex_init(&cgrp->pidlist_mutex);
@@ -1414,12 +1413,10 @@ static void init_cgroup_root(struct cgroupfs_root *root)
 
 	INIT_LIST_HEAD(&root->subsys_list);
 	INIT_LIST_HEAD(&root->root_list);
-	INIT_LIST_HEAD(&root->allcg_list);
 	root->number_of_cgroups = 1;
 	cgrp->root = root;
 	cgrp->name = &root_cgroup_name;
 	init_cgroup_housekeeping(cgrp);
-	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
 }
 
 static int cgroup_init_root_id(struct cgroupfs_root *root)
@@ -2775,65 +2772,74 @@ static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
 	return ret;
 }
 
-static DEFINE_MUTEX(cgroup_cft_mutex);
-
 static void cgroup_cfts_prepare(void)
-	__acquires(&cgroup_cft_mutex) __acquires(&cgroup_mutex)
+	__acquires(&cgroup_mutex)
 {
 	/*
 	 * Thanks to the entanglement with vfs inode locking, we can't walk
 	 * the existing cgroups under cgroup_mutex and create files.
-	 * Instead, we increment reference on all cgroups and build list of
-	 * them using @cgrp->cft_q_node.  Grab cgroup_cft_mutex to ensure
-	 * exclusive access to the field.
+	 * Instead, we use cgroup_for_each_descendant_pre() and drop RCU
+	 * read lock before calling cgroup_addrm_files().
 	 */
-	mutex_lock(&cgroup_cft_mutex);
 	mutex_lock(&cgroup_mutex);
 }
 
 static void cgroup_cfts_commit(struct cgroup_subsys *ss,
 			       struct cftype *cfts, bool is_add)
-	__releases(&cgroup_mutex) __releases(&cgroup_cft_mutex)
+	__releases(&cgroup_mutex)
 {
 	LIST_HEAD(pending);
-	struct cgroup *cgrp, *n;
+	struct cgroup *cgrp, *root = &ss->root->top_cgroup;
 	struct super_block *sb = ss->root->sb;
+	struct inode *inode;
+	struct dentry *prev = NULL;
+	u64 serial_nr = atomic64_read(&cgroup_serial_nr_cursor);
 
 	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
-	if (cfts && ss->root != &rootnode &&
-	    atomic_inc_not_zero(&sb->s_active)) {
-		list_for_each_entry(cgrp, &ss->root->allcg_list, allcg_node) {
-			dget(cgrp->dentry);
-			list_add_tail(&cgrp->cft_q_node, &pending);
-		}
-	} else {
-		sb = NULL;
+	if (!cfts || ss->root == &rootnode ||
+	    !atomic_inc_not_zero(&sb->s_active)) {
+		mutex_unlock(&cgroup_mutex);
+		return;
 	}
+	mutex_unlock(&cgroup_mutex);
 
+	inode = root->dentry->d_inode;
+	mutex_lock(&inode->i_mutex);
+	mutex_lock(&cgroup_mutex);
+	cgroup_addrm_files(root, ss, cfts, is_add);
 	mutex_unlock(&cgroup_mutex);
+	mutex_unlock(&inode->i_mutex);
 
 	/*
 	 * All new cgroups will see @cfts update on @ss->cftsets.  Add/rm
 	 * files for all cgroups which were created before.
 	 */
-	list_for_each_entry_safe(cgrp, n, &pending, cft_q_node) {
-		struct inode *inode = cgrp->dentry->d_inode;
+	rcu_read_lock();
+	cgroup_for_each_descendant_pre(cgrp, root) {
+		if (cgroup_is_dead(cgrp))
+			continue;
+
+		inode = cgrp->dentry->d_inode;
+		dget(cgrp->dentry);
+		rcu_read_unlock();
+
+		dput(prev);
+		prev = cgrp->dentry;
 
 		mutex_lock(&inode->i_mutex);
 		mutex_lock(&cgroup_mutex);
-		if (!cgroup_is_dead(cgrp))
+		if (cgrp->serial_nr <= serial_nr && !cgroup_is_dead(cgrp))
 			cgroup_addrm_files(cgrp, ss, cfts, is_add);
 		mutex_unlock(&cgroup_mutex);
 		mutex_unlock(&inode->i_mutex);
 
-		list_del_init(&cgrp->cft_q_node);
-		dput(cgrp->dentry);
+		rcu_read_lock();
 	}
+	rcu_read_unlock();
+	dput(prev);
 
 	if (sb)
 		deactivate_super(sb);
-
-	mutex_unlock(&cgroup_cft_mutex);
 }
 
 /**
@@ -4316,7 +4322,6 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 	cgrp->serial_nr = atomic64_inc_return(&cgroup_serial_nr_cursor);
 
 	/* allocation complete, commit to creation */
-	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
 	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
 	root->number_of_cgroups++;
 
@@ -4555,7 +4560,6 @@ static void cgroup_offline_fn(struct work_struct *work)
 
 	/* delete this cgroup from parent->children */
 	list_del_rcu(&cgrp->sibling);
-	list_del_init(&cgrp->allcg_node);
 
 	dput(d);
 
-- 
1.8.0.2

[PATCH 1/2][UPDATED] cgroup: make serial_nr_cursor available throughout cgroup.c

[Li Zefan]

The next patch will use it to determine if a cgroup is newly created
while we're iterating the cgroup hierarchy.

Signed-off-by: Li Zefan <lizefan@huawei.com>
---

forgot to remove the old comment in cgroup_create().

---
 kernel/cgroup.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index e8a40e3..60383e8 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -198,6 +198,15 @@ static DEFINE_IDR(cgroup_hierarchy_idr);
 
 static struct cgroup_name root_cgroup_name = { .name = "/" };
 
+/*
+ * Assign a monotonically increasing serial number.  It guarantees
+ * cgroups with bigger numbers are newer than those with smaller
+ * numbers. Also with the list appending, it guarantees that sibling
+ * cgroups are always sorted in the ascending serial number order on
+ * the parent's ->children.
+ */
+static atomic64_t cgroup_serial_nr_cursor = ATOMIC64_INIT(0);
+
 /* This flag indicates whether tasks in the fork and exit paths should
  * check for fork/exit handlers to call. This avoids us having to do
  * extra work in the fork/exit path if none of the subsystems need to
@@ -4212,7 +4221,6 @@ static void offline_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
 static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 			     umode_t mode)
 {
-	static atomic64_t serial_nr_cursor = ATOMIC64_INIT(0);
 	struct cgroup *cgrp;
 	struct cgroup_name *name;
 	struct cgroupfs_root *root = parent->root;
@@ -4299,13 +4307,7 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 		goto err_free_all;
 	lockdep_assert_held(&dentry->d_inode->i_mutex);
 
-	/*
-	 * Assign a monotonically increasing serial number.  With the list
-	 * appending below, it guarantees that sibling cgroups are always
-	 * sorted in the ascending serial number order on the parent's
-	 * ->children.
-	 */
-	cgrp->serial_nr = atomic64_inc_return(&serial_nr_cursor);
+	cgrp->serial_nr = atomic64_inc_return(&cgroup_serial_nr_cursor);
 
 	/* allocation complete, commit to creation */
 	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
-- 
1.8.0.2

Re: [PATCH 1/2][UPDATED] cgroup: make serial_nr_cursor available throughout cgroup.c

[Tejun Heo]

>From 794611a1dfcb055d7d41ce133378dd8197d73e38 Mon Sep 17 00:00:00 2001
From: Li Zefan <lizefan@huawei.com>
Date: Tue, 18 Jun 2013 18:53:53 +0800

The next patch will use it to determine if a cgroup is newly created
while we're iterating the cgroup hierarchy.

tj: Rephrased the comment on top of cgroup_serial_nr_cursor.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
Applied to cgroup/for-3.11 with minor update.

Thanks.

 kernel/cgroup.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 1d4f471..e6571ca 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -198,6 +198,15 @@ static DEFINE_IDR(cgroup_hierarchy_idr);
 
 static struct cgroup_name root_cgroup_name = { .name = "/" };
 
+/*
+ * Assign a monotonically increasing serial number to cgroups.  It
+ * guarantees cgroups with bigger numbers are newer than those with smaller
+ * numbers.  Also, as cgroups are always appended to the parent's
+ * ->children list, it guarantees that sibling cgroups are always sorted in
+ * the ascending serial number order on the list.
+ */
+static atomic64_t cgroup_serial_nr_cursor = ATOMIC64_INIT(0);
+
 /* This flag indicates whether tasks in the fork and exit paths should
  * check for fork/exit handlers to call. This avoids us having to do
  * extra work in the fork/exit path if none of the subsystems need to
@@ -4222,7 +4231,6 @@ static void offline_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
 static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 			     umode_t mode)
 {
-	static atomic64_t serial_nr_cursor = ATOMIC64_INIT(0);
 	struct cgroup *cgrp;
 	struct cgroup_name *name;
 	struct cgroupfs_root *root = parent->root;
@@ -4309,13 +4317,7 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 		goto err_free_all;
 	lockdep_assert_held(&dentry->d_inode->i_mutex);
 
-	/*
-	 * Assign a monotonically increasing serial number.  With the list
-	 * appending below, it guarantees that sibling cgroups are always
-	 * sorted in the ascending serial number order on the parent's
-	 * ->children.
-	 */
-	cgrp->serial_nr = atomic64_inc_return(&serial_nr_cursor);
+	cgrp->serial_nr = atomic64_inc_return(&cgroup_serial_nr_cursor);
 
 	/* allocation complete, commit to creation */
 	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
-- 
1.8.2.1

[PATCH 2/2] cgroup: convert cgroup_cft_commit() to use cgroup_for_each_descendant_pre()

[Tejun Heo]

>From e8c82d20a9f729cf4b9f73043f7fd4e0872bebfd Mon Sep 17 00:00:00 2001
From: Li Zefan <lizefan@huawei.com>
Date: Tue, 18 Jun 2013 18:48:37 +0800

We used root->allcg_list to iterate cgroup hierarchy because at that time
cgroup_for_each_descendant_pre() hasn't been invented.

tj: In cgroup_cfts_commit(), s/@serial_nr/@update_upto/, move the
    assignment right above releasing cgroup_mutex and explain what's
    going on there.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
Applied to cgroup/for-3.11 with some minor updates.

Thanks.

 include/linux/cgroup.h |  6 ----
 kernel/cgroup.c        | 80 +++++++++++++++++++++++++++-----------------------
 2 files changed, 44 insertions(+), 42 deletions(-)

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index f975227..b283658 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -206,9 +206,6 @@ struct cgroup {
 	 */
 	struct list_head cset_links;
 
-	struct list_head allcg_node;	/* cgroupfs_root->allcg_list */
-	struct list_head cft_q_node;	/* used during cftype add/rm */
-
 	/*
 	 * Linked list running through all cgroups that can
 	 * potentially be reaped by the release agent. Protected by
@@ -313,9 +310,6 @@ struct cgroupfs_root {
 	/* A list running through the active hierarchies */
 	struct list_head root_list;
 
-	/* All cgroups on this root, cgroup_mutex protected */
-	struct list_head allcg_list;
-
 	/* Hierarchy-specific flags */
 	unsigned long flags;
 
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index e6571ca..0ed7d8d 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -1399,7 +1399,6 @@ static void init_cgroup_housekeeping(struct cgroup *cgrp)
 	INIT_LIST_HEAD(&cgrp->children);
 	INIT_LIST_HEAD(&cgrp->files);
 	INIT_LIST_HEAD(&cgrp->cset_links);
-	INIT_LIST_HEAD(&cgrp->allcg_node);
 	INIT_LIST_HEAD(&cgrp->release_list);
 	INIT_LIST_HEAD(&cgrp->pidlists);
 	mutex_init(&cgrp->pidlist_mutex);
@@ -1414,12 +1413,10 @@ static void init_cgroup_root(struct cgroupfs_root *root)
 
 	INIT_LIST_HEAD(&root->subsys_list);
 	INIT_LIST_HEAD(&root->root_list);
-	INIT_LIST_HEAD(&root->allcg_list);
 	root->number_of_cgroups = 1;
 	cgrp->root = root;
 	cgrp->name = &root_cgroup_name;
 	init_cgroup_housekeeping(cgrp);
-	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
 }
 
 static int cgroup_init_root_id(struct cgroupfs_root *root)
@@ -2785,65 +2782,78 @@ static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
 	return ret;
 }
 
-static DEFINE_MUTEX(cgroup_cft_mutex);
-
 static void cgroup_cfts_prepare(void)
-	__acquires(&cgroup_cft_mutex) __acquires(&cgroup_mutex)
+	__acquires(&cgroup_mutex)
 {
 	/*
 	 * Thanks to the entanglement with vfs inode locking, we can't walk
 	 * the existing cgroups under cgroup_mutex and create files.
-	 * Instead, we increment reference on all cgroups and build list of
-	 * them using @cgrp->cft_q_node.  Grab cgroup_cft_mutex to ensure
-	 * exclusive access to the field.
+	 * Instead, we use cgroup_for_each_descendant_pre() and drop RCU
+	 * read lock before calling cgroup_addrm_files().
 	 */
-	mutex_lock(&cgroup_cft_mutex);
 	mutex_lock(&cgroup_mutex);
 }
 
 static void cgroup_cfts_commit(struct cgroup_subsys *ss,
 			       struct cftype *cfts, bool is_add)
-	__releases(&cgroup_mutex) __releases(&cgroup_cft_mutex)
+	__releases(&cgroup_mutex)
 {
 	LIST_HEAD(pending);
-	struct cgroup *cgrp, *n;
+	struct cgroup *cgrp, *root = &ss->root->top_cgroup;
 	struct super_block *sb = ss->root->sb;
+	struct dentry *prev = NULL;
+	struct inode *inode;
+	u64 update_upto;
 
 	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
-	if (cfts && ss->root != &rootnode &&
-	    atomic_inc_not_zero(sb->s_active)) {
-		list_for_each_entry(cgrp, &ss->root->allcg_list, allcg_node) {
-			dget(cgrp->dentry);
-			list_add_tail(&cgrp->cft_q_node, &pending);
-		}
-	} else {
-		sb = NULL;
+	if (!cfts || ss->root == &rootnode ||
+	    !atomic_inc_not_zero(&sb->s_active)) {
+		mutex_unlock(&cgroup_mutex);
+		return;
 	}
 
-	mutex_unlock(&cgroup_mutex);
-
 	/*
-	 * All new cgroups will see @cfts update on @ss->cftsets.  Add/rm
-	 * files for all cgroups which were created before.
+	 * All cgroups which are created after we drop cgroup_mutex will
+	 * have the updated set of files, so we only need to update the
+	 * cgroups created before the current @cgroup_serial_nr_cursor.
 	 */
-	list_for_each_entry_safe(cgrp, n, &pending, cft_q_node) {
-		struct inode *inode = cgrp->dentry->d_inode;
+	update_upto = atomic64_read(&cgroup_serial_nr_cursor);
+
+	mutex_unlock(&cgroup_mutex);
+
+	/* @root always needs to be updated */
+	inode = root->dentry->d_inode;
+	mutex_lock(&inode->i_mutex);
+	mutex_lock(&cgroup_mutex);
+	cgroup_addrm_files(root, ss, cfts, is_add);
+	mutex_unlock(&cgroup_mutex);
+	mutex_unlock(&inode->i_mutex);
+
+	/* add/rm files for all cgroups created before */
+	rcu_read_lock();
+	cgroup_for_each_descendant_pre(cgrp, root) {
+		if (cgroup_is_dead(cgrp))
+			continue;
+
+		inode = cgrp->dentry->d_inode;
+		dget(cgrp->dentry);
+		rcu_read_unlock();
+
+		dput(prev);
+		prev = cgrp->dentry;
 
 		mutex_lock(&inode->i_mutex);
 		mutex_lock(&cgroup_mutex);
-		if (!cgroup_is_dead(cgrp))
+		if (cgrp->serial_nr <= update_upto && !cgroup_is_dead(cgrp))
 			cgroup_addrm_files(cgrp, ss, cfts, is_add);
 		mutex_unlock(&cgroup_mutex);
 		mutex_unlock(&inode->i_mutex);
 
-		list_del_init(&cgrp->cft_q_node);
-		dput(cgrp->dentry);
+		rcu_read_lock();
 	}
-
-	if (sb)
-		deactivate_super(sb);
-
-	mutex_unlock(&cgroup_cft_mutex);
+	rcu_read_unlock();
+	dput(prev);
+	deactivate_super(sb);
 }
 
 /**
@@ -4320,7 +4330,6 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 	cgrp->serial_nr = atomic64_inc_return(&cgroup_serial_nr_cursor);
 
 	/* allocation complete, commit to creation */
-	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
 	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
 	root->number_of_cgroups++;
 
@@ -4559,7 +4568,6 @@ static void cgroup_offline_fn(struct work_struct *work)
 
 	/* delete this cgroup from parent->children */
 	list_del_rcu(&cgrp->sibling);
-	list_del_init(&cgrp->allcg_node);
 
 	dput(d);
 
-- 
1.8.2.1

[PATCH 3/3] cgroup: clean up cgroup_serial_nr_cursor

[Tejun Heo]

>From 00356bd5f0f5e04183fb15805eb29e97c2fc20ac Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Tue, 18 Jun 2013 11:14:22 -0700

cgroup_serial_nr_cursor was created atomic64_t because I thought it
was never gonna used for anything other than assigning unique numbers
to cgroups and didn't want to worry about synchronization; however,
now we're using it as an event-stamp to distinguish cgroups created
before and after certain point which assumes that it's protected by
cgroup_mutex.

Let's make it clear by making it a u64.  Also, rename it to
cgroup_serial_nr_next and make it point to the next nr to allocate so
that where it's pointing to is clear and more conventional.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
---
Applied to cgroup/for-3.11 as a follow-up.

Thanks.

 kernel/cgroup.c | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 0ed7d8d..65f333e 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -203,9 +203,10 @@ static struct cgroup_name root_cgroup_name = { .name = "/" };
  * guarantees cgroups with bigger numbers are newer than those with smaller
  * numbers.  Also, as cgroups are always appended to the parent's
  * ->children list, it guarantees that sibling cgroups are always sorted in
- * the ascending serial number order on the list.
+ * the ascending serial number order on the list.  Protected by
+ * cgroup_mutex.
  */
-static atomic64_t cgroup_serial_nr_cursor = ATOMIC64_INIT(0);
+static u64 cgroup_serial_nr_next = 1;
 
 /* This flag indicates whether tasks in the fork and exit paths should
  * check for fork/exit handlers to call. This avoids us having to do
@@ -2803,7 +2804,7 @@ static void cgroup_cfts_commit(struct cgroup_subsys *ss,
 	struct super_block *sb = ss->root->sb;
 	struct dentry *prev = NULL;
 	struct inode *inode;
-	u64 update_upto;
+	u64 update_before;
 
 	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
 	if (!cfts || ss->root == &rootnode ||
@@ -2815,9 +2816,9 @@ static void cgroup_cfts_commit(struct cgroup_subsys *ss,
 	/*
 	 * All cgroups which are created after we drop cgroup_mutex will
 	 * have the updated set of files, so we only need to update the
-	 * cgroups created before the current @cgroup_serial_nr_cursor.
+	 * cgroups created before the current @cgroup_serial_nr_next.
 	 */
-	update_upto = atomic64_read(&cgroup_serial_nr_cursor);
+	update_before = cgroup_serial_nr_next;
 
 	mutex_unlock(&cgroup_mutex);
 
@@ -2844,7 +2845,7 @@ static void cgroup_cfts_commit(struct cgroup_subsys *ss,
 
 		mutex_lock(&inode->i_mutex);
 		mutex_lock(&cgroup_mutex);
-		if (cgrp->serial_nr <= update_upto && !cgroup_is_dead(cgrp))
+		if (cgrp->serial_nr < update_before && !cgroup_is_dead(cgrp))
 			cgroup_addrm_files(cgrp, ss, cfts, is_add);
 		mutex_unlock(&cgroup_mutex);
 		mutex_unlock(&inode->i_mutex);
@@ -4327,7 +4328,7 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 		goto err_free_all;
 	lockdep_assert_held(&dentry->d_inode->i_mutex);
 
-	cgrp->serial_nr = atomic64_inc_return(&cgroup_serial_nr_cursor);
+	cgrp->serial_nr = cgroup_serial_nr_next++;
 
 	/* allocation complete, commit to creation */
 	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
-- 
1.8.2.1