Re: [PATCH] User namespaces: set of cleanups (eventually for linux-next?)

[David Howells <dhowells@redhat.com>]

Serge E. Hallyn <serue@us.ibm.com> wrote:

> David, since this consumes your patch, I wasn't sure whether it was
> appropriate to put your signed-off-by on here or not.  I decided
> doing so was the worse of the potential offenses...

The way I think I'd've done it is to include my patch description and
signed-off-by then list your additional changes (just keep on incrementing the
point numbers) and your signed-off-by.

> -	new_user = alloc_uid(current->nsproxy->user_ns, new->uid);
> +	new_user = alloc_uid(current_user()->user_ns, new->uid);

That should be current_user_ns() rather than current_user()->user_ns.  I made
this change before adding the macro.

> +	key_put(new->thread_keyring);
> +	new->thread_keyring = NULL;

Superfluous.  copy_creds() does this immediately upon return.

Also, in copy_creds(), should the session and process keyrings be discarded if
CLONE_NEWUSER is set?  Actually, I think that should be dealt with by a patch
to deal with namespacing keyrings as the user-default keyrings need to be
namespaced rather than here.

David