Re: 3.15 regression: wrong cgroup magic

[Linus Torvalds <torvalds@linux-foundation.org>]

On Mon, Jun 2, 2014 at 5:56 PM, Andy Lutomirski <luto@amacapital.net> wrote:
>
> In particular, this piece:
>
> -       sb->s_magic = CGROUP_SUPER_MAGIC;
>
> The result is that cgroup shows up with the wrong magic number, so my
> code goes "oh crap, cgroupfs isn't mounted" and fails.
>
> I can change my code to hack around this, but I can imagine other
> things getting tripped up.  Is there still time to fix this?

Sure. Send me a tested patch. I'm assuming it's going to look something like

  --- a/kernel/cgroup.c
  +++ b/kernel/cgroup.c
  @@ -54,6 +54,7 @@
   #include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
   #include <linux/kthread.h>
   #include <linux/delay.h>
  +#include <linux/magic.h>

   #include <linux/atomic.h>

  @@ -1607,6 +1608,8 @@ out_unlock:
          dentry = kernfs_mount(fs_type, flags, root->kf_root, &new_sb);
          if (IS_ERR(dentry) || !new_sb)
                  cgroup_put(&root->cgrp);
  +       else
  +               dentry->d_sb->s_magic = CGROUP_SUPER_MAGIC;
          return dentry;
   }

but somebody definitely needs to test it.

               Linus