devcgroup: avoid using cgroup_lock
Li Zefan [Thu, 2 Apr 2009 23:57:32 +0000 (16:57 -0700)]
There is nothing special that has to be protected by cgroup_lock,
so introduce devcgroup_mtuex for it's own use.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

security/device_cgroup.c

index 3aacd0f..5fda7df 100644 (file)
@@ -11,6 +11,7 @@
 #include <linux/uaccess.h>
 #include <linux/seq_file.h>
 #include <linux/rcupdate.h>
+#include <linux/mutex.h>
 
 #define ACC_MKNOD 1
 #define ACC_READ  2
 #define DEV_CHAR  2
 #define DEV_ALL   4  /* this represents all devices */
 
+static DEFINE_MUTEX(devcgroup_mutex);
+
 /*
  * whitelist locking rules:
- * hold cgroup_lock() for update/read.
+ * hold devcgroup_mutex for update/read.
  * hold rcu_read_lock() for read.
  */
 
@@ -67,7 +70,7 @@ static int devcgroup_can_attach(struct cgroup_subsys *ss,
 }
 
 /*
- * called under cgroup_lock()
+ * called under devcgroup_mutex
  */
 static int dev_whitelist_copy(struct list_head *dest, struct list_head *orig)
 {
@@ -92,7 +95,7 @@ free_and_exit:
 
 /* Stupid prototype - don't bother combining existing entries */
 /*
- * called under cgroup_lock()
+ * called under devcgroup_mutex
  */
 static int dev_whitelist_add(struct dev_cgroup *dev_cgroup,
                        struct dev_whitelist_item *wh)
@@ -130,7 +133,7 @@ static void whitelist_item_free(struct rcu_head *rcu)
 }
 
 /*
- * called under cgroup_lock()
+ * called under devcgroup_mutex
  */
 static void dev_whitelist_rm(struct dev_cgroup *dev_cgroup,
                        struct dev_whitelist_item *wh)
@@ -185,8 +188,10 @@ static struct cgroup_subsys_state *devcgroup_create(struct cgroup_subsys *ss,
                list_add(&wh->list, &dev_cgroup->whitelist);
        } else {
                parent_dev_cgroup = cgroup_to_devcgroup(parent_cgroup);
+               mutex_lock(&devcgroup_mutex);
                ret = dev_whitelist_copy(&dev_cgroup->whitelist,
                                &parent_dev_cgroup->whitelist);
+               mutex_unlock(&devcgroup_mutex);
                if (ret) {
                        kfree(dev_cgroup);
                        return ERR_PTR(ret);
@@ -273,7 +278,7 @@ static int devcgroup_seq_read(struct cgroup *cgroup, struct cftype *cft,
  * does the access granted to dev_cgroup c contain the access
  * requested in whitelist item refwh.
  * return 1 if yes, 0 if no.
- * call with c->lock held
+ * call with devcgroup_mutex held
  */
 static int may_access_whitelist(struct dev_cgroup *c,
                                       struct dev_whitelist_item *refwh)
@@ -426,11 +431,11 @@ static int devcgroup_access_write(struct cgroup *cgrp, struct cftype *cft,
                                  const char *buffer)
 {
        int retval;
-       if (!cgroup_lock_live_group(cgrp))
-               return -ENODEV;
+
+       mutex_lock(&devcgroup_mutex);
        retval = devcgroup_update_access(cgroup_to_devcgroup(cgrp),
                                         cft->private, buffer);
-       cgroup_unlock();
+       mutex_unlock(&devcgroup_mutex);
        return retval;
 }