cgroup: refactor cgroup_pidlist_find()
authorTejun Heo <tj@kernel.org>
Fri, 29 Nov 2013 15:42:59 +0000 (10:42 -0500)
committerTejun Heo <tj@kernel.org>
Fri, 29 Nov 2013 15:42:59 +0000 (10:42 -0500)
Rename cgroup_pidlist_find() to cgroup_pidlist_find_create() and
separate out finding proper to cgroup_pidlist_find().  Also, move
locking to the caller.

This patch is preparation for pidlist restructure and doesn't
introduce any behavior changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
kernel/cgroup.c

index ef019c075d5daecf0d76a560c9a8ddcc4042e14b..d58c30d3b82869ee7d492e299b9a3efb99caa359 100644 (file)
@@ -3575,48 +3575,50 @@ static int cmppid(const void *a, const void *b)
        return *(pid_t *)a - *(pid_t *)b;
 }
 
+static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
+                                                 enum cgroup_filetype type)
+{
+       struct cgroup_pidlist *l;
+       /* don't need task_nsproxy() if we're looking at ourself */
+       struct pid_namespace *ns = task_active_pid_ns(current);
+
+       lockdep_assert_held(&cgrp->pidlist_mutex);
+
+       list_for_each_entry(l, &cgrp->pidlists, links)
+               if (l->key.type == type && l->key.ns == ns)
+                       return l;
+       return NULL;
+}
+
 /*
  * find the appropriate pidlist for our purpose (given procs vs tasks)
  * returns with the lock on that pidlist already held, and takes care
  * of the use count, or returns NULL with no locks held if we're out of
  * memory.
  */
-static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
-                                                 enum cgroup_filetype type)
+static struct cgroup_pidlist *cgroup_pidlist_find_create(struct cgroup *cgrp,
+                                               enum cgroup_filetype type)
 {
        struct cgroup_pidlist *l;
-       /* don't need task_nsproxy() if we're looking at ourself */
-       struct pid_namespace *ns = task_active_pid_ns(current);
 
-       /*
-        * We can't drop the pidlist_mutex before taking the l->rwsem in case
-        * the last ref-holder is trying to remove l from the list at the same
-        * time. Holding the pidlist_mutex precludes somebody taking whichever
-        * list we find out from under us - compare release_pid_array().
-        */
-       mutex_lock(&cgrp->pidlist_mutex);
-       list_for_each_entry(l, &cgrp->pidlists, links) {
-               if (l->key.type == type && l->key.ns == ns) {
-                       /* make sure l doesn't vanish out from under us */
-                       down_write(&l->rwsem);
-                       mutex_unlock(&cgrp->pidlist_mutex);
-                       return l;
-               }
-       }
+       lockdep_assert_held(&cgrp->pidlist_mutex);
+
+       l = cgroup_pidlist_find(cgrp, type);
+       if (l)
+               return l;
+
        /* entry not found; create a new one */
        l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
-       if (!l) {
-               mutex_unlock(&cgrp->pidlist_mutex);
+       if (!l)
                return l;
-       }
+
        init_rwsem(&l->rwsem);
        INIT_DELAYED_WORK(&l->destroy_dwork, cgroup_pidlist_destroy_work_fn);
-       down_write(&l->rwsem);
        l->key.type = type;
-       l->key.ns = get_pid_ns(ns);
+       /* don't need task_nsproxy() if we're looking at ourself */
+       l->key.ns = get_pid_ns(task_active_pid_ns(current));
        l->owner = cgrp;
        list_add(&l->links, &cgrp->pidlists);
-       mutex_unlock(&cgrp->pidlist_mutex);
        return l;
 }
 
@@ -3662,17 +3664,26 @@ static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
        sort(array, length, sizeof(pid_t), cmppid, NULL);
        if (type == CGROUP_FILE_PROCS)
                length = pidlist_uniq(array, length);
-       l = cgroup_pidlist_find(cgrp, type);
+
+       mutex_lock(&cgrp->pidlist_mutex);
+
+       l = cgroup_pidlist_find_create(cgrp, type);
        if (!l) {
+               mutex_unlock(&cgrp->pidlist_mutex);
                pidlist_free(array);
                return -ENOMEM;
        }
-       /* store array, freeing old if necessary - lock already held */
+
+       /* store array, freeing old if necessary */
+       down_write(&l->rwsem);
        pidlist_free(l->list);
        l->list = array;
        l->length = length;
        l->use_count++;
        up_write(&l->rwsem);
+
+       mutex_unlock(&cgrp->pidlist_mutex);
+
        *lp = l;
        return 0;
 }