Re: [PATCH 1/4] kernel: add a helper to get an owning user namespace for a namespace

[Serge E. Hallyn]

On Wed, Aug 31, 2016 at 01:38:35PM -0700, Andrey Vagin wrote:
> On Tue, Aug 30, 2016 at 7:56 PM, Serge E. Hallyn <serge@xxxxxxxxxx> wrote:
> > On Fri, Aug 26, 2016 at 04:08:08PM -0700, Andrei Vagin wrote:
> >> +struct ns_common *ns_get_owner(struct ns_common *ns)
> >> +{
> >> +     struct user_namespace *my_user_ns = current_user_ns();
> >> +     struct user_namespace *owner, *p;
> >> +
> >> +     /* See if the owner is in the current user namespace */
> >> +     owner = p = ns->ops->get_owner(ns);
> >> +     for (;;) {
> >> +             if (!p)
> >> +                     return ERR_PTR(-EPERM);
> >> +             if (p == my_user_ns)
> >> +                     break;
> >> +             p = p->parent;
> >> +     }
> >> +
> >> +     return &get_user_ns(owner)->ns;
> >
> > get_user_ns() bumps the owner's refcount.  I don't see where
> > this is being dropped, especially when ns_ioctl() uses it in
> > the next patch.
> 
> It is dropped in __ns_get_path if a namespace has a dentry, otherwise
> it is dropped from nsfs_evict.
> 
> static void *__ns_get_path(struct path *path, struct ns_common *ns)
>         |                return -EPERM;
> ...
>         ns->ops->put(ns);                                                      |
> got_it:
>         |        /* See if the owner is in the current user namespace
> */
>         path->mnt = mnt;
>         |        owner = p = ns->ops->get_owner(ns);
>         path->dentry = dentry;
>         |        for (;;) {
>         return NULL;
> ...
> 
> static void nsfs_evict(struct inode *inode)                                    |
> {
>         |        if (!ns_capable(user_ns, CAP_SYS_ADMIN))
>         struct ns_common *ns = inode->i_private;
>         |                return -EPERM;
>         clear_inode(inode);                                                    |
>         ns->ops->put(ns);
>         |        cred = prepare_creds();
> }

Gotcha, thanks.