Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 17 Feb 2020 15:59:08 +0100
From:   Christian Brauner <christian.brauner@ubuntu.com>
To:     Michael Kerrisk <mtk.manpages@gmail.com>
Cc:     Dmitry Safonov <dima@arista.com>, Andrei Vagin <avagin@openvz.org>,
        Linux Kernel <linux-kernel@vger.kernel.org>,
        Dmitry Safonov <0x7f454c46@gmail.com>,
        Adrian Reber <adrian@lisas.de>,
        Andy Lutomirski <luto@kernel.org>,
        Arnd Bergmann <arnd@arndb.de>,
        Cyrill Gorcunov <gorcunov@openvz.org>,
        "Eric W. Biederman" <ebiederm@xmission.com>,
        "H. Peter Anvin" <hpa@zytor.com>, Ingo Molnar <mingo@redhat.com>,
        Jann Horn <jannh@google.com>, Jeff Dike <jdike@addtoit.com>,
        Oleg Nesterov <oleg@redhat.com>,
        Pavel Emelyanov <xemul@virtuozzo.com>,
        Shuah Khan <shuah@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Vincenzo Frascino <vincenzo.frascino@arm.com>,
        containers <containers@lists.linux-foundation.org>,
        criu@openvz.org, Linux API <linux-api@vger.kernel.org>,
        x86@kernel.org, Andrei Vagin <avagin@gmail.com>
Subject: Re: Time Namespaces: CLONE_NEWTIME and clone3()?
Message-ID: <20200217145908.7epzz5nescccwvzv@wittgenstein>
References: <20191112012724.250792-1-dima@arista.com>
 <20191112012724.250792-4-dima@arista.com>
 <CAHO5Pa2_7BzZPCXjFj4f=YoX28M4q2Au=h6GrzN-EjRffMo1iw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <CAHO5Pa2_7BzZPCXjFj4f=YoX28M4q2Au=h6GrzN-EjRffMo1iw@mail.gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Feb 17, 2020 at 03:20:55PM +0100, Michael Kerrisk wrote:
> Hello Dmitry, Andrei,
> 
> Is the CLONE_NEWTIME flag intended to be usable with clone3()? The
> mail quoted below implies (in my reading) that this should be possible
> once clone3() is available, which it is by now. (See also [1].)
> 
> If the answer is yes, CLONE_NEWTIME  should be usable with clone3(),
> then I have a bug report and a question.
> 
> I successfully used CLONE_NEWTIME with unshare(). But if I try to use
> CLONE_NEWSIGNAL with clone3(), it errors out with EINVAL, because of

s/CLONE_NEWSIGNAL/CLONE_NEWTIME/

> the following check in clone3_args_valid():
> 
>         /*
>          * - make the CLONE_DETACHED bit reuseable for clone3
>          * - make the CSIGNAL bits reuseable for clone3
>          */
>         if (kargs->flags & (CLONE_DETACHED | CSIGNAL))
>                 return false;
> 
> The problem is that CLONE_NEWTIME matches one of the bits in the
> CSIGNAL mask. If the intention is to allow CLONE_NEWTIME with
> clone3(), then either the bit needs to be redefined, or the error
> checking in clone3_args_valid() needs to be reworked.

If this is intended to be useable with clone3() the check should be
adapted to allow for CLONE_NEWTIME. (I asked about this a while ago I
think.)
But below rather sounds like it should simply be an unshare() flag. The
code seems to set frozen_offsets to true right after copy_namespaces()
in timens_on_fork(new_ns, tsk) and so the offsets can't be changed
anymore unless I'm reading this wrong.
Alternatives seem to either make timens_offsets writable once after fork
and before exec, I guess - though that's probably not going to work
with the vdso judging from timens_on_fork().

The other alternative is that Andrei and Dmitry send me a patch to
enable CLONE_NEWTIME with clone3() by exposing struct timens_offsets (or
a version of it) in the uapi and extend struct clone_args to include a
pointer to a struct timens_offset that is _only_ set when CLONE_NEWTIME
is set.
Though the unshare() way sounds way less invasive simpler.

Christian