Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
To:     Bjorn Andersson <bjorn.andersson@linaro.org>,
        Ohad Ben-Cohen <ohad@wizery.com>,
        Mathieu Poirier <mathieu.poirier@linaro.org>
CC:     <linux-remoteproc@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        <linux-stm32@st-md-mailman.stormreply.com>,
        <arnaud.pouliquen@foss.st.com>
Subject: [PATCH 0/4] rpmsg: ctrl: Add ability to instantiate rpmsg device locally
Date:   Fri, 4 Jun 2021 11:14:02 +0200
Message-ID: <20210604091406.15901-1-arnaud.pouliquen@foss.st.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Purpose:
  Allow the user space application to create and release an rpmsg device by adding
  rpmsg ioctrl to the /dev/rpmsg_ctrl interface
    
Aim:
  The current implementation is based on the enumeration of services by the 
  remote processor to create a new channel and instantiate associated rpmsg device.
  There is no solution to create a rpmsg channel on user application request.
  If the rpmsg char driver allows adding a new endpoint over an existing channel, it
  does not offer the ability to create a new one.
  Adding the IOCTRL to instantiate rpmsg channels from the user application will allow
  to dynamically create and destroy rpmsg devices. Some examples of use are:
  - activate the service at the initiative of the application,
  - remove the communication on a specific channel before entering the suspend mode,
  - creating a temporary channel for debugging purposes.

Concerns:
  This implementation is very simple but allows the user application to create rpmsg devices
  without any limitations.
  - A device can be created even if there is no match with the driver (and it is difficult to
    check a match with module drivers).
  - It is not really possible to add a counter to limit the number of devices, because a device
    can be released by the remote side.
  
  The question is: should we add protection? Notice that there is no protection for RPMSG_CREATE_EPT_IOCTL.

  If this is not considered safe, another approach could be to create an rpmsg_ctrl API to add controls.
  This API would be used by a rpmsg driver to add is own user controls, such as creating/deleting devices.

How to test it:
  - This series can be applied on git/andersson/remoteproc.git for-next branch (dc0e14fa833b)
    + the "Restructure the rpmsg char to decorrelate the control part" series[1]
  - to test the ioctrl, a rpmsgexportdev tool is available here: https://github.com/arnopo/rpmsgexport

[1]https://patchwork.kernel.org/project/linux-remoteproc/list/?series=483793

Arnaud Pouliquen (4):
  rpmsg: ctrl: Introduce RPMSG_CREATE_DEV_IOCTL
  rpmsg: ctrl: Introduce RPMSG_RELEASE_DEV_IOCTL
  rpmsg: ctrl: Add check on rpmsg device removability from user space
  rpmsg: Add a removable attribute to the rpmsg device

 drivers/rpmsg/rpmsg_core.c |  2 ++
 drivers/rpmsg/rpmsg_ctrl.c | 52 +++++++++++++++++++++++++++++++++++---
 include/linux/rpmsg.h      |  2 ++
 include/uapi/linux/rpmsg.h | 10 ++++++++
 4 files changed, 62 insertions(+), 4 deletions(-)

-- 
2.17.1