From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753277AbbG1VCA (ORCPT ); Tue, 28 Jul 2015 17:02:00 -0400 Received: from mail-io0-f180.google.com ([209.85.223.180]:34970 "EHLO mail-io0-f180.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753112AbbG1VB7 (ORCPT ); Tue, 28 Jul 2015 17:01:59 -0400 MIME-Version: 1.0 In-Reply-To: References: <20150728171500.GA2871@www.outflux.net> <87d1zcm9iz.fsf@x220.int.ebiederm.org> Date: Tue, 28 Jul 2015 14:01:58 -0700 X-Google-Sender-Auth: UOIzh1JT_Bs2FEuOcuc2NJ8XuJ4 Message-ID: Subject: Re: [PATCH] user_ns: use correct check for single-threadedness From: Kees Cook To: Ricky Zhou Cc: "Eric W. Biederman" , Oleg Nesterov , David Howells , LKML , Andrew Morton , Peter Zijlstra , Ingo Molnar , "Kirill A. Shutemov" , Rik van Riel , Vladimir Davydov , Julien Tinnes Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jul 28, 2015 at 1:55 PM, Ricky Zhou wrote: > On Tue, Jul 28, 2015 at 11:17 AM, Eric W. Biederman > wrote: >> Kees Cook writes: >> >>> From: Ricky Zhou >>> >>> Checking mm_users > 1 does not mean a process is multithreaded. For >>> example, reading /proc/PID/maps temporarily increments mm_users, allowing >>> other processes to (accidentally) interfere with unshare() calls. >>> >>> This fixes observed failures of unshare(CLONE_NEWUSER) incorrectly >>> returning EINVAL if another processes happened to be simultaneously >>> reading the maps file. >>> >>> Signed-off-by: Ricky Zhou >>> Signed-off-by: Kees Cook >>> Cc: stable@vger.kernel.org >> >> This looks like a good fix. Any chance you can drudge up the commit where >> this hack came in so that Greg & Company know how far to back port this? > > userns_install in user_namespace.c (affects setns of a user > namespace): cde1975bc242f3e1072bde623ef378e547b73f91. > > The check in check_unshare_flags is a little more complex. The > incorrect check was added in > cf2e340f4249b781b3d2beb41e891d08581f0e10 but I don't think it would > have triggered under any supported combination of flags at that point. > > From 50804fe3737ca6a5942fdc2057a18a8141d00141 until > 6e556ce209b09528dbf1931cbfd5d323e1345926, the bug affected > unshare(CLONE_NEWPID). That's back to v3.8, so this goes quite a way, it seems. -Kees > > From b2e0d98705e60e45bbb3c0032c48824ad7ae0704 onward, > unshare(CLONE_NEWUSER) is affected. > > Thanks, > Ricky >>> >>> --- >>> kernel/fork.c | 4 ++-- >>> kernel/user_namespace.c | 3 ++- >>> 2 files changed, 4 insertions(+), 3 deletions(-) >>> >>> diff --git a/kernel/fork.c b/kernel/fork.c >>> index dbd9b8d7b7cc..7d138f152dcd 100644 >>> --- a/kernel/fork.c >>> +++ b/kernel/fork.c >>> @@ -75,6 +75,7 @@ >>> #include >>> #include >>> #include >>> +#include >>> >>> #include >>> #include >>> @@ -1876,8 +1877,7 @@ static int check_unshare_flags(unsigned long unshare_flags) >>> * needs to unshare vm. >>> */ >>> if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { >>> - /* FIXME: get_task_mm() increments ->mm_users */ >>> - if (atomic_read(¤t->mm->mm_users) > 1) >>> + if (!current_is_single_threaded()) >>> return -EINVAL; >>> } >>> >>> diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c >>> index 4109f8320684..7e0e021e4304 100644 >>> --- a/kernel/user_namespace.c >>> +++ b/kernel/user_namespace.c >>> @@ -22,6 +22,7 @@ >>> #include >>> #include >>> #include >>> +#include >>> >>> static struct kmem_cache *user_ns_cachep __read_mostly; >>> static DEFINE_MUTEX(userns_state_mutex); >>> @@ -977,7 +978,7 @@ static int userns_install(struct nsproxy *nsproxy, struct ns_common *ns) >>> return -EINVAL; >>> >>> /* Threaded processes may not enter a different user namespace */ >>> - if (atomic_read(¤t->mm->mm_users) > 1) >>> + if (!current_is_single_threaded()) >>> return -EINVAL; >>> >>> if (current->fs->users != 1) >>> -- >>> 1.9.1 -- Kees Cook Chrome OS Security