Commit cfebe563 authored Feb 11, 2009 by Li Zefan Committed by Linus Torvalds Feb 11, 2009

cgroups: fix lockdep subclasses overflow



I enabled all cgroup subsystems when compiling kernel, and then:
 # mount -t cgroup -o net_cls xxx /mnt
 # mkdir /mnt/0

This showed up immediately:
 BUG: MAX_LOCKDEP_SUBCLASSES too low!
 turning off the locking correctness validator.

It's caused by the cgroup hierarchy lock:
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
		if (ss->root == root)
			mutex_lock_nested(&ss->hierarchy_mutex, i);
	}

Now we have 9 cgroup subsystems, and the above 'i' for net_cls is 8, but
MAX_LOCKDEP_SUBCLASSES is 8.

This patch uses different lockdep keys for different subsystems.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

parent 01c4a428

include/linux/cgroup.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -378,6 +378,7 @@ struct cgroup_subsys {
		* - initiating hotplug events
		*/
		struct mutex hierarchy_mutex;
		struct lock_class_key subsys_key;

		/*
		* Link to parent, and list entry in parent's children.

kernel/cgroup.c

+2 −1

Original line number	Diff line number	Diff line
		@@ -2351,7 +2351,7 @@ static void cgroup_lock_hierarchy(struct cgroupfs_root *root)
		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
		if (ss->root == root)
		mutex_lock_nested(&ss->hierarchy_mutex, i);
		mutex_lock(&ss->hierarchy_mutex);
		}
		}

		@@ -2637,6 +2637,7 @@ static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
		BUG_ON(!list_empty(&init_task.tasks));

		mutex_init(&ss->hierarchy_mutex);
		lockdep_set_class(&ss->hierarchy_mutex, &ss->subsys_key);
		ss->active = 1;
		}