[Devel] Re: [RFC][PATCH 1/3] CGroups: Add a per-subsystem hierarchy_mutex

[Paul Menage]

On Wed, Dec 10, 2008 at 4:37 PM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu at jp.fujitsu.com> wrote:
>> +static void cgroup_unlock_hierarchy(struct cgroupfs_root *root)
>> +{
>> +     int i;
>> +
>> +     for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
>> +             struct cgroup_subsys *ss = subsys[i];
>> +             if (ss->root == root)
>> +                     mutex_unlock(&ss->hierarchy_mutex);
>> +     }
>> +}
>> +
> Maybe no problem..but I don't like releasing lock in the order of acquiring lock.
>
>        for (i = CGROUP_SUBSYS_COUNT - 1; i >=0; i--)  ?

The order that you release the locks is irrelevant for correctness. In
this case, since the only callers of cgroup_lock_hierarchy() also hold
cgroup_mutex and hence can't race with one another, the order of
locking is irrelevant for correctness too - right now the locking
order is just designed to keep lockdep happy.

I think that the reverse-ordered loop is less readable for no gain.

Paul
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers