[PATCH v3 2/7] userns: Simpilify MNT_NODEV handling.

[Eric W. Biederman]

Andy Lutomirski <luto at amacapital.net> writes:

> On Wed, Sep 16, 2015 at 1:02 PM, Seth Forshee
> <seth.forshee at canonical.com> wrote:
>> From: "Eric W. Biederman" <ebiederm at xmission.com>
>>
>> - Consolidate the testing if a device node may be opened in a new
>>   function may_open_dev.
>>
>> - Move the check for allowing access to device nodes on filesystems
>>   not mounted in the initial user namespace from mount time to open
>>   time and include it in may_open_dev.
>>
>> This set of changes removes the implicit adding of MNT_NODEV which
>> simplifies the logic in fs/namespace.c and removes a potentially
>> problematic user visible difference in how normal and unprivileged
>> mount namespaces work.
>>
>> Signed-off-by: "Eric W. Biederman" <ebiederm at xmission.com>
>
>> -               /* Only in special cases allow devices from mounts
>> -                * created outside the initial user namespace.
>> -                */
>> -               if (!(type->fs_flags & FS_USERNS_DEV_MOUNT)) {
>> -                       flags |= MS_NODEV;
>> -                       mnt_flags |= MNT_NODEV | MNT_LOCK_NODEV;
>> -               }
>
> This is an ABI change.  It's probably okay, but I think the commit
> message should make it clear what's happening.

You mean it should include in big flashing neon letters
            ***REGRESSION FIX***
?

It is longer in coming than I had hoped.  But that is part of the reason
I did not fix the security hole this way.  Getting the s_user_ns stuff
just so has been non-trivial.

I do agree that because this is a user visible change we do need to keep
our eyes peeled for pieces of userspace software that may depend on the
exact details of the current behavior.

Eric