Merge branch 'for-3.11' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup

Pull cgroup changes from Tejun Heo:
 "This pull request contains the following changes.

   - cgroup_subsys_state (css) reference counting has been converted to
     percpu-ref.  css is what each resource controller embeds into its
     own control structure and perform reference count against.  It may
     be used in hot paths of various subsystems and is similar to module
     refcnt in that aspect.  For example, block-cgroup's css refcnting
     was showing up a lot in Mikulaus's device-mapper scalability work
     and this should alleviate it.

   - cgroup subtree iterator has been updated so that RCU read lock can
     be released after grabbing reference.  This allows simplifying its
     users which requires blocking which used to build iteration list
     under RCU read lock and then traverse it outside.  This pull
     request contains simplification of cgroup core and device-cgroup.
     A separate pull request will update cpuset.

   - Fixes for various bugs including corner race conditions and RCU
     usage bugs.

   - A lot of cleanups and some prepartory work for the planned unified
     hierarchy support."

* 'for-3.11' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: (48 commits)
  cgroup: CGRP_ROOT_SUBSYS_BOUND should also be ignored when mounting an existing hierarchy
  cgroup: CGRP_ROOT_SUBSYS_BOUND should be ignored when comparing mount options
  cgroup: fix deadlock on cgroup_mutex via drop_parsed_module_refcounts()
  cgroup: always use RCU accessors for protected accesses
  cgroup: fix RCU accesses around task->cgroups
  cgroup: fix RCU accesses to task->cgroups
  cgroup: grab cgroup_mutex in drop_parsed_module_refcounts()
  cgroup: fix cgroupfs_root early destruction path
  cgroup: reserve ID 0 for dummy_root and 1 for unified hierarchy
  cgroup: implement for_each_[builtin_]subsys()
  cgroup: move init_css_set initialization inside cgroup_mutex
  cgroup: s/for_each_subsys()/for_each_root_subsys()/
  cgroup: clean up find_css_set() and friends
  cgroup: remove cgroup->actual_subsys_mask
  cgroup: prefix global variables with "cgroup_"
  cgroup: convert CFTYPE_* flags to enums
  cgroup: rename cont to cgrp
  cgroup: clean up cgroup_serial_nr_cursor
  cgroup: convert cgroup_cft_commit() to use cgroup_for_each_descendant_pre()
  cgroup: make serial_nr_cursor available throughout cgroup.c
  ...

1 files changed, 18 insertions, 38 deletions

diff --git a/security/device_cgroup.c b/security/device_cgroup.c
index dd0dc57..e8aad69 100644
--- a/security/device_cgroup.c
+++ b/security/device_cgroup.c
@@ -49,8 +49,6 @@ struct dev_cgroup {
 	struct cgroup_subsys_state css;
 	struct list_head exceptions;
 	enum devcg_behavior behavior;
-	/* temporary list for pending propagation operations */
-	struct list_head propagate_pending;
 };
 
 static inline struct dev_cgroup *css_to_devcgroup(struct cgroup_subsys_state *s)
@@ -241,7 +239,6 @@ static struct cgroup_subsys_state *devcgroup_css_alloc(struct cgroup *cgroup)
 	if (!dev_cgroup)
 		return ERR_PTR(-ENOMEM);
 	INIT_LIST_HEAD(&dev_cgroup->exceptions);
-	INIT_LIST_HEAD(&dev_cgroup->propagate_pending);
 	dev_cgroup->behavior = DEVCG_DEFAULT_NONE;
 
 	return &dev_cgroup->css;
@@ -445,34 +442,6 @@ static void revalidate_active_exceptions(struct dev_cgroup *devcg)
 }
 
 /**
- * get_online_devcg - walks the cgroup tree and fills a list with the online
- * 		      groups
- * @root: cgroup used as starting point
- * @online: list that will be filled with online groups
- *
- * Must be called with devcgroup_mutex held. Grabs RCU lock.
- * Because devcgroup_mutex is held, no devcg will become online or offline
- * during the tree walk (see devcgroup_online, devcgroup_offline)
- * A separated list is needed because propagate_behavior() and
- * propagate_exception() need to allocate memory and can block.
- */
-static void get_online_devcg(struct cgroup *root, struct list_head *online)
-{
-	struct cgroup *pos;
-	struct dev_cgroup *devcg;
-
-	lockdep_assert_held(&devcgroup_mutex);
-
-	rcu_read_lock();
-	cgroup_for_each_descendant_pre(pos, root) {
-		devcg = cgroup_to_devcgroup(pos);
-		if (is_devcg_online(devcg))
-			list_add_tail(&devcg->propagate_pending, online);
-	}
-	rcu_read_unlock();
-}
-
-/**
  * propagate_exception - propagates a new exception to the children
  * @devcg_root: device cgroup that added a new exception
  * @ex: new exception to be propagated
@@ -482,15 +451,24 @@ static void get_online_devcg(struct cgroup *root, struct list_head *online)
 static int propagate_exception(struct dev_cgroup *devcg_root,
 			       struct dev_exception_item *ex)
 {
-	struct cgroup *root = devcg_root->css.cgroup;
-	struct dev_cgroup *devcg, *parent, *tmp;
+	struct cgroup *root = devcg_root->css.cgroup, *pos;
 	int rc = 0;
-	LIST_HEAD(pending);
 
-	get_online_devcg(root, &pending);
+	rcu_read_lock();
 
-	list_for_each_entry_safe(devcg, tmp, &pending, propagate_pending) {
-		parent = cgroup_to_devcgroup(devcg->css.cgroup->parent);
+	cgroup_for_each_descendant_pre(pos, root) {
+		struct dev_cgroup *devcg = cgroup_to_devcgroup(pos);
+
+		/*
+		 * Because devcgroup_mutex is held, no devcg will become
+		 * online or offline during the tree walk (see on/offline
+		 * methods), and online ones are safe to access outside RCU
+		 * read lock without bumping refcnt.
+		 */
+		if (!is_devcg_online(devcg))
+			continue;
+
+		rcu_read_unlock();
 
 		/*
 		 * in case both root's behavior and devcg is allow, a new
@@ -512,8 +490,10 @@ static int propagate_exception(struct dev_cgroup *devcg_root,
 		}
 		revalidate_active_exceptions(devcg);
 
-		list_del_init(&devcg->propagate_pending);
+		rcu_read_lock();
 	}
+
+	rcu_read_unlock();
 	return rc;
 }