selinux: no recursive read_lock of policy_rwlock in security_genfs_sid()
authorWaiman Long <Waiman.Long@hp.com>
Mon, 23 Jun 2014 15:28:51 +0000 (11:28 -0400)
committerPaul Moore <pmoore@redhat.com>
Mon, 23 Jun 2014 20:52:55 +0000 (16:52 -0400)
commitf31e799459659ae88c341aeac16a8a5efb1271d4
tree912cc7f6dd3350c32501a344cf438f7ec2269c72
parent6e51f9cbfa04a92b40e7f9c1e76c8ecbff534a22
selinux: no recursive read_lock of policy_rwlock in security_genfs_sid()

With the introduction of fair queued rwlock, recursive read_lock()
may hang the offending process if there is a write_lock() somewhere
in between.

With recursive read_lock checking enabled, the following error was
reported:

=============================================
[ INFO: possible recursive locking detected ]
3.16.0-rc1 #2 Tainted: G            E
---------------------------------------------
load_policy/708 is trying to acquire lock:
 (policy_rwlock){.+.+..}, at: [<ffffffff8125b32a>]
security_genfs_sid+0x3a/0x170

but task is already holding lock:
 (policy_rwlock){.+.+..}, at: [<ffffffff8125b48c>]
security_fs_use+0x2c/0x110

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(policy_rwlock);
  lock(policy_rwlock);

This patch fixes the occurrence of recursive read_lock() of
policy_rwlock by adding a helper function __security_genfs_sid()
which requires caller to take the lock before calling it. The
security_fs_use() was then modified to call the new helper function.

Signed-off-by: Waiman Long <Waiman.Long@hp.com>
Acked-by: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: Paul Moore <pmoore@redhat.com>
security/selinux/ss/services.c