Re: [RFC] Allow user namespace inside chroot

[Eric W. Biederman]

Jann Horn <jannh@xxxxxxxxxx> writes:

> On Mon, Oct 15, 2018 at 7:10 PM <nagarathnam.muthusamy@xxxxxxxxxx> wrote:
>> @@ -1281,7 +1285,7 @@ static int userns_install(struct nsproxy *nsproxy, struct ns_common *ns)
>>                 return -ENOMEM;
>>
>>         put_user_ns(cred->user_ns);
>> -       set_cred_user_ns(cred, get_user_ns(user_ns));
>> +       set_cred_user_ns(cred, get_user_ns(user_ns), 0);
>
> This looks bogus. With this, I think your restriction can be bypassed
> if process A forks a child B, B creates a new user namespace, then A
> enters the user namespace with setns() and has full capabilities. Am I
> missing something?

Nope.  I feel silly for missing that angle.

Even without the full capabilities the userns_install angle will place
you at the root of the mount namespace outside of the chroot.

At which point I have visions of the special cases multiplying like
bunnies make this work.  Without a very strong case I don't like this at all.

Eric