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:   Tue, 6 Nov 2018 16:07:01 -0800
From:   Andrew Morton <akpm@linux-foundation.org>
To:     =?UTF-8?B?Y2hvdXJ5emhvdQ==?= (=?UTF-8?B?5ZGo5aiB?=) 
        <chouryzhou@tencent.com>
Cc:     "gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
        "arve@android.com" <arve@android.com>,
        "tkjos@android.com" <tkjos@android.com>,
        "dave@stgolabs.net" <dave@stgolabs.net>,
        "devel@driverdev.osuosl.org" <devel@driverdev.osuosl.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH V2] binder: ipc namespace support for android binder
Message-Id: <20181106160701.5e73de3056dee7aa560e43a3@linux-foundation.org>
In-Reply-To: <5FBCBE569E134E4CA167B91C0A77FD610198F6BB7D@EXMBX-SZMAIL022.tencent.com>
References: <5FBCBE569E134E4CA167B91C0A77FD610198F6BB7D@EXMBX-SZMAIL022.tencent.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, 29 Oct 2018 06:18:11 +0000 chouryzhou(周威) <chouryzhou@tencent.com> wrote:

>   We are working for running android in container, but we found that binder is
> not isolated by ipc namespace. Since binder is a form of IPC and therefore should
> be tied to ipc namespace. With this patch, we can run more than one android
> container on one host.
>   This patch move "binder_procs" and "binder_context" into ipc_namespace,
> driver will find the context from it when opening. Althought statistics in debugfs
> remain global.
> 
> ...
>
> --- a/ipc/namespace.c
> +++ b/ipc/namespace.c
> @@ -56,6 +56,9 @@ static struct ipc_namespace *create_ipc_ns(struct user_namespace *user_ns,
>         ns->ucounts = ucounts;
>  
>         err = mq_init_ns(ns);
> +       if (err)
> +               goto fail_put;
> +       err = binder_init_ns(ns);
>         if (err)
>                 goto fail_put;
>  

Don't we need an mq_put_mnt() if binder_init_ns() fails?

free_ipc_ns() seems to have forgotten about that too.  In which case it
must be madly leaking mounts so probably I'm wrong.  Confused.