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;
MIME-Version: 1.0
References: <20190727085201.11743-1-christian@brauner.io> <20190727085201.11743-2-christian@brauner.io>
 <CAHk-=whrh5+aHmgqP9YhZ-yzCtUWT8fPi08ZSJdxusx7aHXOQQ@mail.gmail.com>
In-Reply-To: <CAHk-=whrh5+aHmgqP9YhZ-yzCtUWT8fPi08ZSJdxusx7aHXOQQ@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sat, 27 Jul 2019 09:41:25 -0700
Message-ID: <CAHk-=wgtGyQTLLb7m=jkMY5hOcj+TsOKSoeoJC7Jc1_h_XL+tw@mail.gmail.com>
Subject: Re: [PATCH v2 1/2] pidfd: add P_PIDFD to waitid()
To:     Christian Brauner <christian@brauner.io>
Cc:     Linux List Kernel Mailing <linux-kernel@vger.kernel.org>,
        Oleg Nesterov <oleg@redhat.com>, Arnd Bergmann <arnd@arndb.de>,
        "Eric W. Biederman" <ebiederm@xmission.com>,
        Kees Cook <keescook@chromium.org>,
        Joel Fernandes <joel@joelfernandes.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Tejun Heo <tj@kernel.org>, David Howells <dhowells@redhat.com>,
        Jann Horn <jannh@google.com>,
        Andrew Lutomirski <luto@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Aleksa Sarai <cyphar@cyphar.com>,
        Al Viro <viro@zeniv.linux.org.uk>,
        Android Kernel Team <kernel-team@android.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sat, Jul 27, 2019 at 9:28 AM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> Something like
>
>   struct pid *fd_to_pid(unsigned int fd)
>   {
>         struct fd f;
>         struct pid *pid;
...

I forgot to put my usual disclaimer about TOTALLY UNTESTED GARBAGE in
that email. I want to make that part clear: that code snippet was
meant as a rough guide of direction, not as a "this works".

Hopefully that was clear.

Also note again that one of the reasons I would prefer that
"fd_to_pid()" interface is that you _can_ do it cleverly with RCU
lookup, but that requires a lot of care.

In particular, I think all of our _existing_
"proc_pid(file_inode(file))" users are done while you actually hold a
reference to "struct file *", so they don't have to worry about races
with another thread doing the final ->release(). So the "clever" thing
is possible, but might need a _lot_ of care to make sure the 'struct
pid *' associated with the file still exists.

The example code sequence was not doing the clever thing, obviously.
So it was untested _and_ simple-stupid. But it has the interface that
I'd prefer.

              Linus