Feedback from the Netdev 2.2 conference

The Netdev 2.2 conference took place in Seoul, South Korea. As we work on a diversity of networking topics at Free Electrons as part of our Linux kernel contributions, Free Electrons engineers Alexandre Belloni and Antoine Ténart went to Seoul to attend lots of interesting sessions and to meet with the Linux networking community. Below, they report on what they learned from this conference, by highlighting two talks they particularly liked.

Linux Networking Dietary Restrictions — slides — video

David S. Miller gave a keynote about reducing the size of core structures in the Linux kernel networking core. The idea behind his work is to use smaller structures which has many benefits in terms of performance as less cache misses will occur and less memory resources are needed. This is especially true in the networking core as small changes may have enormous impacts and improve performance a lot. Another argument from his maintainer hat perspective is the maintainability, where smaller structures usually means less complexity.

He presented five techniques he used to shrink the networking core data structures. The first one was to identify members of common base structures that are only used in sub-classes, as these members can easily be moved out and not impact all the data paths.

The second one makes use of what David calls “state compression”, aka. understanding the real width of the information stored in data structures and to pack flags together to save space. In his mind a boolean should take a single bit whereas in the kernel it requires way more space than that. While this is fine for many uses it makes sense to compress all these data in critical structures.

Then David S. Miller spoke about unused bits in pointers where in the kernel all pointers have 3 bits never used. He argued these bits are 3 boolean values that should be used to reduce core data structure sizes. This technique and the state compression one can be used by introducing helpers to safely access the data.

Another technique he used was to unionize members that aren’t used at the same time. This helps reducing even more the structure size by not having areas of memory never used during identified steps in the networking stack.

Finally he showed us the last technique he used, which was using lookup keys instead of pointers when the objects can be found cheaply based on their index. While this cannot be used for every object, it helped reducing some data structures.

While going through all these techniques he gave many examples to help understanding what can be saved and how it was effective. This was overall a great talk showing a critical aspect we do not always think of when writing drivers, which can lead to big performance improvements.

David S. Miller at Nedev 2.2

WireGuard: Next-generation Secure Kernel Network Tunnel — slides — video

Jason A. Donenfeld presented his new and shiny L3 network tunneling mechanism, in Linux. After two years of development this in-kernel formally proven cryptographic protocol is ready to be submitted upstream to get the first rounds of review.

The idea behind Wireguard is to provide, with a small code base, a simple interface to establish and maintain encrypted tunnels. Jason made a demo which was impressive by its simplicity when securely connecting two machines, while it can be a real pain when working with OpenVPN or IPsec. Under the hood this mechanism uses UDP packets on top of either IPv4 and IPv6 to transport encrypted packets using modern cryptographic principles. The authentication is similar to what SSH is using: static private/public key pairs. One particularly nice design choice is the fact that Wireguard is exposed as a stateless interface to the administrator whereas the protocol is stateful and timer based, which allow to put devices into sleep mode and not to care about it.

One of the difficulty to get Wireguard accepted upstream is its cryptographic needs, which do not match what can provide the kernel cryptographic framework. Jason knows this and plan to first send patches to rework the cryptographic framework so that his module nicely integrates with in-kernel APIs. First RFC patches for Wireguard should be sent at the end of 2017, or at the beginning of 2018.

We look forward to seeing Wireguard hit the mainline kernel, to allow everybody to establish secure tunnels in an easy way!

Jason A. Donenfeld at Netdev 2.2


Netdev 2.2 was again an excellent experience for us. It was an (almost) single track format, running alongside the workshops, allowing to not miss any session. The technical content let us dive deeply in the inner working of the network stack and stay up-to-date with the current developments.

Thanks for organizing this and for the impressive job, we had an amazing time!

Linux 4.14 released, Free Electrons contributions

Penguin from Mylène Josserand
Drawing from Mylène Josserand,
based on a picture from Samuel Blanc under CC-BY-SA
Linux 4.14, which is going to become the next Long Term Supported version, has been released a week ago by Linus Torvalds. As usual, did an interesting coverage of this release cycle merge window, highlighting the most important changes: The first half of the 4.14 merge window and The rest of the 4.14 merge window.

According to Linux Kernel Patch statistics, Free Electrons contributed 111 patches to this release, making it the 24th contributing company by number of commits: a somewhat lower than usual contribution level from our side. At least, Free Electrons cannot be blamed for trying to push more code into 4.14 because of its Long Term Support nature! 🙂

The main highlights of our contributions are:

  • On the RTC subsystem, Alexandre Belloni made as usual a number of fixes and improvements to various drivers, especially the ds1307 driver.
  • On the NAND subsystem, Boris Brezillon did a number of small improvements in various areas.
  • On the support for Marvell platforms
    • Antoine Ténart improved the ppv2 network driver used by the Marvell Armada 7K/8K SoCs: support for 10G speed and TSO support are the main highlights. In order to support 10G speed, Antoine added a driver in drivers/phy/ to configure the common PHYs in the Armada 7K/8K SoCs.
    • Thomas Petazzoni also improved the ppv2 network driver by adding support for TX interrupts and per-CPU RX interrupts.
    • Grégory Clement contributed some patches to enable NAND support on Armada 7K/8K, as well as a number of fixes in different areas (GPIO fix, clock handling fixes, etc.)
    • Miquèl Raynal contributed a fix for the Armada 3700 SPI controller driver.
  • On the support for Allwinner platforms
    • Maxime Ripard contributed the support for a new board, the BananaPI M2-Magic. Maxime also contributed a few fixes to the Allwinner DRM driver, and a few other misc fixes (clock, MMC, RTC, etc.).
    • Quentin Schulz contributed the support for the power button functionality of the AXP221 (PMIC used in several Allwinner platforms)
  • On the support for Atmel platforms, Quentin Schulz improved the clock drivers for this platform to properly support the Audio PLL, which allowed to fix the Atmel audio drivers. He also fixed suspend/resume support in the Atmel MMC driver to support the deep sleep mode of the SAMA5D2 processor.

In addition to making direct contributions, Free Electrons is also involved in the Linux kernel development by having a number of its engineers act as Linux kernel maintainers. As part of this effort, Free Electrons engineers have reviewed, merged and sent pull requests for a large number of contributions from other developers:

  • Boris Brezillon, as the NAND subsystem maintainer and MTD subsystem co-maintainer, merged 68 patches from other developers.
  • Alexandre Belloni, as the RTC subsystem maintainer and Atmel ARM platform co-maintainer, merged 32 patches from other developers.
  • Grégory Clement, as the Marvell ARM platform co-maintainer, merged 29 patches from other developers.
  • Maxime Ripard, as the Allwinner ARM platform co-maintainer, merged 18 patches from other developers.

This flow of patches from kernel maintainers to other kernel maintainers is also nicely described for the 4.14 release by the Patch flow into the mainline for 4.14 article.

The detailed list of our contributions:

Mender: How to integrate an OTA updater

Recently, our customer Senic asked us to integrate an Over-The-Air (OTA) mechanism in their embedded Linux system, and after some discussion, they ended up chosing Mender. This article will detail an example of Mender’s integration and how to use it.

What is Mender?

Mender is an open source remote updater for embedded devices. It is composed of a client installed on the embedded device, and a management server installed on a remote server. However, the server is not mandatory as Mender can be used standalone, with updates triggered directly on the embedded device.

Image taken from Mender’s website

In order to offer a fallback in case of failure, Mender uses the double partition layout: the device will have at least 2 rootfs partitions, one active and one inactive. Mender will deploy an update on the inactive partition, so that in case of an error during the update process, it will still have the active partition intact. If the update succeeds, it will switch to the updated partition: the active partition becomes inactive and the inactive one becomes the new active. As the kernel and the device tree are stored in the /boot folder of the root filesystem, it is possible to easily update an entire system. Note that Mender needs at least 4 partitions:

  • bootloader partition
  • data persistent partition
  • rootfs + kernel active partition
  • rootfs + kernel inactive partition

It is, of course, customizable if you need more partitions.

Two reference devices are supported: the BeagleBone Black and a virtual device. In our case, the board was a Nanopi-Neo, which is based on an Allwinner H3.

Mender provides a Yocto Project layer containing all the necessary classes and recipes to make it work. The most important thing to know is that it will produce an image ready to be written to an SD card to flash empty boards. It will also produce “artifacts” (files with .mender extension) that will be used to update an existing system.

Installation and setup

In this section, we will see how to setup the Mender client and server for your project. Most of the instructions are taken from the Mender documentation that we found well detailed and really pleasant to read. We’ll simply summarize the most important steps.

Server side

The Mender server will allow you to remotely update devices. The server can be installed in two modes:

  • demo mode: Used to test a demo server. It can be nice to test it if you just want to quickly deploy a Mender solution, for testing purpose only. It includes a demo layer that simplify and configure for you a default Mender server on localhost of your workstation.
  • production mode: Used for production. We will focus on this mode as we wanted to use Mender in a production context. This mode allows to customize the server configuration: IP address, certificates, etc. Because of that, some configuration will be necessary (which is not the case in the demo mode).

In order to install the Mender server, you should first install Docker CE and Docker Compose. Have a look at the corresponding Docker instructions.


  • Download the integration repository from Mender:
  • $ git clone mender-server
  • Checkout 1.1.0 tag (latest version at the moment of the test)
  • $ cd mender-server
    $ git checkout 1.1.0 -b my-production-setup
  • Copy the template folder and update all the references to “template”
  • $ cp -a template production
    $ cd production
    $ sed -i -e 's#/template/#/production/#g' prod.yml
  • Download Docker images
  • $ ./run pull
  • Use the keygen script to create certificates for domain names (e.g. and
  • $ ../keygen
  • Some persistent storage will be needed by Mender so create a few Docker volumes:
  • $ docker volume create --name=mender-artifacts
    $ docker volume create --name=mender-deployments-db
    $ docker volume create --name=mender-useradm-db
    $ docker volume create --name=mender-inventory-db
    $ docker volume create --name=mender-deviceadm-db
    $ docker volume create --name=mender-deviceauth-db

Final configuration

This final configuration will link the generated keys with the Mender server. All the modifications will be in the prod.yml file.

  • Locate the storage-proxy service in prod.yml and set it to your domain name. In our case under the networks.mender.aliases
  • Locate the minio service. Set MINIO_ACCESS_KEY to “mender-deployments” and the MINIO_SECRET_KEY to a generated password (with e.g.: $ apg -n1 -a0 -m32)
  • Locate the mender-deployments service. Set DEPLOYMENTS_AWS_AUTH_KEY and DEPLOYMENTS_AWS_AUTH_SECRET to respectively the value of MINIO_ACCESS_KEY and MINIO_SECRET_KEY. Set DEPLOYMENTS_AWS_URI to point to your domain such as

Start the server

Make sure that the domain names you have defined ( and are accessible, potentially by adding them to /etc/hosts if you’re just testing.

  • Start the server
  • $ ./run up -d
  • If it is a new installation, request initial user login:
  • $ curl -X POST  -D - --cacert keys-generated/certs/api-gateway/cert.crt
  • Check that you can create a user and login to mender UI:
  •  $ firefox 

Client side – Yocto Project

Mender has a Yocto Project layer to easily interface with your own layer.
We will see how to customize your layer and image components (U-Boot, Linux kernel) to correctly configure it for Mender use.

In this section, we will assume that you have your own U-Boot and your own kernel repositories (and thus, recipes) and that you retrieved the correct branch of this layer.

Machine and distro configurations

  • Make sure that the kernel image and Device Tree files are installed in the root filesystem image
  • RDEPENDS_kernel-base += "kernel-image kernel-devicetree"
  • Update the distro to inherit the mender-full class and add systemd as the init manager (we only tested Mender’s integration with systemd)
  • # Enable systemd for Mender
    DISTRO_FEATURES_append = " systemd"
    VIRTUAL-RUNTIME_init_manager = "systemd"
    VIRTUAL-RUNTIME_initscripts = ""
    INHERIT += "mender-full"
  • By default, Mender assumes that your storage device is /dev/mmcblk0, that mmcblk0p1 is your boot partition (containing the bootloader), that mmcblk0p2 and mmcblk0p3 are your two root filesystem partitions, and that mmcblk0p5 is your data partition. If that’s the case for you, then everything is fine! However, if you need a different layout, you need to update your machine configuration. Mender’s client will retrieve which storage device to use by using the MENDER_STORAGE_DEVICE variable (which defaults to mmcblk0). The partitions themselves should be specified using MENDER_BOOT_PART, MENDER_ROOTFS_PART_A, MENDER_ROOTFS_PART_B and ROOTFS_DATA_PART. If you need to change the default storage or the partitions’ layout, edit in your machine configuration the different variables according to your need. Here is an example for /dev/sda:
  • MENDER_STORAGE_DEVICE = "/dev/sda"
  • Do not forget to update the artifact name in your local.conf, for example:

    MENDER_ARTIFACT_NAME = "release-1"

As described in Mender’s documentation, Mender will store the artifact name in its artifact image. It must be unique which is what we expect because an artifact will represent a release tag or a delivery. Note that if you forgot to update it and upload an artifact with the same name as an existing in the web UI, it will not be taken into account.

U-Boot configuration tuning

Some modifications in U-Boot are necessary to be able to perform the rollback (use a different partition after an unsuccessful update)

  • Mender needs BOOTCOUNT support in U-Boot. It creates a bootcount variable that will be incremented each time a reboot appears (or reset to 1 after a power-on reset). Mender will use this variable in its rollback mechanism.
    Make sure to enable it in your U-Boot configuration. This will most likely require a patch to your board .h configuration file, enabling:
  • Remove environment variables that will be redefined by Mender. They are defined in Mender’s documentation.
  • Update your U-Boot recipe to inherit Mender’s one and make sure to provide U-Boot virtual package (using PROVIDES)
  • # Mender integration
    require recipes-bsp/u-boot/
    PROVIDES += "u-boot"
    RPROVIDES_${PN} += "u-boot"
    BOOTENV_SIZE = "0x20000"

    The BOOTENV_SIZE must be set the same content as the U-Boot CONFIG_ENV_SIZE variable. It will be used by the u-boot-fw-utils tool to retrieve the U-Boot environment variables.

    Mender is using u-boot-fw-utils so make sure that you have a recipe for it and that Mender include’s file is included. To do that, you can create a bbappend file on the default recipe or create your own recipe if you need a specific version. Have a look at Mender’s documentation example.

  • Tune your U-Boot environment to use Mender’s variables. Here are some examples of the modifications to be done. Set the root= kernel argument to use ${mender_kernel_root}, set the bootcmd to load the kernel image and Device Tree from ${mender_uboot_root} and to run mender_setup. Make sure that you are loading the Linux kernel image and Device Tree file from the root filesystem /boot directory.
    setenv bootargs 'console=${console} root=${mender_kernel_root} rootwait'
    setenv mmcboot 'load ${mender_uboot_root} ${fdt_addr_r} boot/my-device-tree.dtb; load ${mender_uboot_root} ${kernel_addr_r} boot/zImage; bootz ${kernel_addr_r} - ${fdt_addr_r}'
    setenv bootcmd 'run mender_setup; run mmcboot'

Mender’s client recipe

As stated in the introduction, Mender has a client, in the form of a userspace application, that will be used on the target. Mender’s layer has a Yocto recipe for it but it does not have our server certificates. To establish a connection between the client and the server, the certificates have to be installed in the image. For that, a bbappend recipe will be created. It will also allow to perform additional Mender configuration, such as defining the server URL.

  • Create a bbappend for the Mender recipe
  • FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
    SRC_URI_append = " file://server.crt"
  • Copy your server certificates in the bbappend recipe folder

Recompile an image and now, we should have everything we need to be able to update an image. Do not hesitate to run the integration checklist, it is really a convenient way to know if everything is correctly configured (or not).

If you want to be more robust and secure, you can sign your artifacts to be sure that they come from a trusted source. If you want this feature, have a look at this documentation.


Standalone mode

To update an artifact using the standalone mode (i.e. without server), here are the commands to use. You will need to update them according to your needs.

  • On your work station, create a simple HTTP server in your Yocto deploy folder:
  • $ python -m SimpleHTTPServer
  • On the target, start mender in standalone mode
  • $ mender -log-level info -rootfs

    You can also use the mender command to start an update from a local .mender file, provided by a USB key or SD card.

  • Once finished, you will have to reboot the target manually
  • $ reboot

    After the first reboot, you will be on the the new active partition (if the previous one was /dev/mmcblk0p2, you should be on /dev/mmcblk0p3). Check the kernel version, artifact name or command line:

    $ uname -a
    $ cat /etc/mender/artifact_info
    $ cat /proc/cmdline

    If you are okay with this update, you will have to commit the modification otherwise the update will not be persistent and once you will reboot the board, Mender will rollback to the previous partition:

    $ mender -commit

Using Mender’s server UI

The Mender server UI provides a management interface to deploy updates on all your devices. It knows about all your devices, their current software version, and you can plan deployments on all or a subset of your devices. Here are the basic steps to trigger a deployment:

  • Login (or create an account) into the mender server UI:
  • Power-up your device
  • The first time, you will have to authorize the device. You will find it in your “dashboard” or in the “devices” section.
  • After authorizing it, it will retrieve device information such as current software version, MAC address, network interface, and so on
  • To update a partition, you will have to create a deployment using an artifact.
  • Upload the new artifact in the server UI using the “Artifacts” section
  • Deploy the new artifact using the “deployment” or the “devices” section. You will retrieve the status of the deployment in the “status” field. It will be in “installing”, “rebooting”, etc. The board will reboot and the partition should be updated.


Here are some issues we faced when we integrated Mender for our device. The Mender documentation also has a troubleshooting section so have a look at it if you are facing issues. Otherwise, the community seems to be active even if we did not need to interact with it as it worked like a charm when we tried it.

Update systemd’s service starting

By default, the Mender systemd service will start after the service “resolved” the domain name. On our target device, the network was available only via WiFi. We had to wait for the wlan0 interface to be up and configured to automatically connect a network before starting Mender’s service. Otherwise, it leads to an error due to the network being unreachable. To solve this issue which is specific to our platform, we set the systemd dependencies to “” to be sure that a network is available:


It now matches our use case because the Mender service will start only if the wlan0 connection is available and working.

Certificate expired

The certificates generated and used by Mender have a validity period. In case your board does not have a RTC set, Mender can fail with the error:

systemctl status mender
... level=error msg="authorize failed: transient error: authorization request failed: failed to execute authorization request:
Post https:///api/devices/v1/authentication/auth_requests: x509: certificate has expired or is not yet valid" module=state

To solve this issue, update the date on your board and make sure your RTC is correctly set.

Device deletion

While testing Mender’s server (version 1.0), we always used the same board and got into the issue that the board was already registered in the Server UI but had a different Device ID (which is used by Mender to identify devices). Because of that, the server was always rejecting the authentication. The next release of the Mender server offers the possibility to remove a device so we updated the Mender’s server to the last version.

Deployments not taken into account

Note that the Mender’s client is checking by default every 30 minutes if a deployment is available for this device. During testing, you may want to reduce this period, which you can in the Mender’s configuration file using its UpdatePollIntervalSeconds variable.


Mender is an OTA updater for Embedded devices. It has a great documentation in the form of tutorials which makes the integration easy. While testing it, the only issues we got were related to our custom platform or were already indicated in the documentation. Deploying it on a board was not difficult, only some U-Boot/kernel and Yocto Project modifications were necessary. All in all, Mender worked perfectly fine for our project!

Back from ELCE: selection of talks from the Free Electrons team

As discussed in our previous blog post, Free Electrons had a strong presence at the Embedded Linux Conference Europe, with 7 attendees, 4 talks, one BoF and one poster during the technical show case.

In this blog post, we would like to highlight a number of talks from the conference that we found interesting. Each Free Electrons engineer who attended the conference has selected one talk, and gives his/her feedback about this talk.

uClibc Today: Still Makes Sense – Alexey Brodkin

Talk selected by Michael Opdenacker

uClibc Today: Still Makes Sense – Alexey BrodkinAlexey Brodkin, an active contributor to the uClibc library, shared recent updates about this C library, trying to show people that the project was still active and making progress, after a few years during which it appeared to be stalled. Alexey works for Synopsys, the makers of the ARC architecture, which uClibc supports.

If you look at the repository for uClibc releases, you will see that since version 1.0.0 released in 2015, the project has made 26 releases according to a predictable schedule. The project also runs runtime regression tests on all its releases, which weren’t done before. The developers have also added support for 4 new architectures (arm64 in particular), and uClibc remains the default C library that Buildroot proposes.

Alexey highlighted that in spite of the competition from the musl library, which causes several projects to switch from uClibc to musl, uClibc still makes a lot of sense today. As a matter of fact, it supports more hardware architectures than glibc and musl do, as it’s the only one to support platforms without an MMU (such as noMMU ARM, Blackfin, m68k, Xtensa) and as the library size is still smaller that what you get with musl (though a static hello_world program is much smaller with musl if you have a close look at the library comparison tests he mentioned).

Alexey noted that the uClibc++ project is still alive too, and used in OpenWRT/Lede by default.

Read the slides and watch the video of his talk.

Identifying and Supporting “X-Compatible” hardware blocks – Chen-Yu Tsai

Talk selected by Quentin Schulz

Identifying and Supporting “X-Compatible” hardware blocks – Chen-Yu TsaiAn SoC is made of multiple IP blocks from different vendors. In some cases the source or model of the hardware blocks are neither documented nor marketed by the SoC vendor. However, since there are only very few vendors of a given IP block, stakes are high that your SoC vendor’s undocumented IP block is compatible with a known one.

With his experience in developing drivers for multiple IP blocks present in Allwinner SoCs and as a maintainer of those same SoCs, Chen-Yu first explained that SoC vendors often either embed some vendors’ licensed IP blocks in their SoCs and add the glue around it for platform- or SoC-specific hardware (clocks, resets and control signals), or they clone IP blocks with the same logic but some twists (missing, obfuscated or rearranged registers).

To identify the IP block, we can dig into the datasheet or the vendor BSP and compare those with well documented datasheets such as the one for NXP i.MX6, TI KeyStone II or the Zynq UltraScale+ MPSoC, or with mainline drivers. Asking the community is also a good idea as someone might have encountered an IP with the same behaviour before and can help us identify it quicker.

Some good identifiers for IPs could be register layouts or names along with DMA logic and descriptor format. For the unlucky ones that have been provided only a blob, they can look for the symbols in it and that may slightly help in the process.

He also mentioned that identifying an IP block is often the result of the developer’s experience in identifying IPs and other time just pure luck. Unfortunately, there are times when someone couldn’t identify the IP and wrote a complete driver just to be told by someone else in the community that this IP reminds him of that IP in which case the work done can be just thrown away. That’s where the community plays a strong role, to help us in our quest to identify an IP.

Chen-Yu then went on with the presentation of the different ways to handle the multiple variants of an IP blocks in drivers. He said that the core logic of all IP drivers is usually presented as a library and that the different variants have a driver with their own resources, extra setup and use this library. Also, a good practice is to use booleans to select features of IP blocks instead of using the ID of each variant.
For IPs whose registers have been altered, the way to go is usually to write a table for the register offsets, or use regmaps when bitfields are also modified. When the IP block differs a bit too much, custom callbacks should be used.

He ended his talk with his return from experience on multiple IP blocks (UART, USB OTG, GMAC, EMAC and HDMI) present in Allwinner SoCs and the differences developers had to handle when adding support for them.

Read the slides and watch the video of his talk.

printk(): The Most Useful Tool is Now Showing its Age – Steven Rostedt & Sergey Senozhatsky

Talks selected by Boris Brezillon. Boris also covered the related talk “printk: It’s Old, What Can We Do to Make It Young Again?” from the same speakers.

printk(): The Most Useful Tool is Now Showing its Age – Steven Rostedt & Sergey SenozhatskyMaybe I should be ashamed of saying that but printk() is one of the basic tool I’m using to debug kernel code, and I must say it did the job so far, so when I saw these presentations talking about how to improve printk() I was a bit curious. What could be wrong in printk() implementation?
Before attending the talks, I never digged into printk()’s code, because it just worked for me, but what I thought was a simple piece of code appeared to be a complex infrastructure with locking scheme that makes you realize how hard it is when several CPUs are involved.

At its core, printk() is supposed to store logs into a circular buffer and push new entries to one or several consoles. In his first talk Steven Rostedt walked through the history of printk() and explained why it became more complex when support for multi CPU appeared. He also detailed why printk() is not re-entrant and the problem it causes when called from an NMI handler. He finally went through some fixes that made the situation a bit better and advertised the 2nd half of the talk driven by Sergey Senozhatsky.

Note that between these two presentations, the printk() rework has been discussed at Kernel Summit, so Sergey already had some feedback on his proposals. While Steven presentation focused mainly on the main printk() function, Sergey gave a bit more details on why printk() can deadlock, and one of the reasons why preventing deadlocks is so complicated is that printk() delegates the ‘print to console’ aspect to console drivers which have their own locking scheme. To address that, it is proposed to move away from the callback approach and let console drivers poll for new entries in the console buffer instead, which would remove part of the locking issues. The problem with this approach is that it brings even more uncertainty on when the logs are printed on the consoles, and one of the nice things about printk() in its current form is that you are likely to have the log printed on the output before printk() returns (which helps a lot when you debug things).

He also mentioned other solutions to address other possible deadlocks, but I must admit I got lost at some point, so if you’re interested in this topic I recommend that you watch the video (printk(): The Most Useful Tool is Now Showing its Age, sadly no video is available for the second talk) and read the slides (printk(): The Most Useful Tool is Now Showing its Age and printk: It’s Old, What Can We Do to Make It Young Again?).

More robust I2C designs with a new fault-injection driver – Wolfram Sang

Talk selected by Miquèl Raynal

More robust I2C designs with a new fault-injection driver – Wolfram SangAlthough Wolfram had a lot of troubles starting its presentation lacking a proper HDMI adaptater, he gave an illuminating talk about how, as an I2C subsystem maintainer, he would like to strengthen the robustness of I2C drivers.

He first explained some basics of the I2C bus like START and STOP conditions and introduced us to a few errors he regularly spots in drivers. For instance, some badly written drivers used a START and STOP sequence while a “repeated START” was needed. This is very bad because another master on the bus could, in this little spare idle delay, decide to grab the medium and send its message. Then the message right after the repeated START would not have the expected effect. Of course plenty other errors can happen: stalled bus (SDA or SCL stuck low), lost arbitration, faulty bits… All these situations are usually due to incorrect sequences sent by the driver.

To avoid so much pain debugging obscure situations where this happens, he decided to use an extended I2C-gpio interface to access SDA and SCL from two external GPIOs and this way forces faulty situations by simply pinning high or low one line (or both) and see how the driver reacts. The I2C specification and framework provide everything to get out of a faulty situation, it is just a matter of using it (sending a STOP condition, clocking 9 times, operate a reset, etc).

Wolfram is aware of his quite conservative approach but he is really scared about breaking users by using random quirks so he tried with this talk to explain his point of view and the solutions he wants to promote.

Two questions that you might have a hard time hearing were also interesting. The first person asked if he ever considered using a “default faulty chip” designed to do by itself this kind of fault injection and see how the host reacts and behaves. Wolfram said buying hardware is too much effort for debugging, so he was more motivated to get something very easy and straightforward to use. Someone else asked if he thought about multiple clients situation, but from Wolfram’s point of view, all clients are in the same state whether the bus is busy or free and should not badly behave if we clock 9 times.

Watch the video and grab the slides.

HDMI 4k Video: Lessons Learned – Hans Verkuil

Talk selected by Maxime Ripard

HDMI 4k Video: Lessons Learned – Hans VerkuilHaving worked recently on a number of display related drivers, it was quite natural to go see what I was probably going to work on in a quite close future.

Hans started by talking about HDMI in general, and the various signals that could go through it. He then went on with what was basically a war story about all the mistakes, gotchas and misconceptions that he encountered while working on a video-conference box for Cisco. He covered the hardware itself, but also more low-level oriented aspects, such as the clocks frequencies needed to operate properly, the various signals you could look at for debugging or the issues that might come with the associated encoding and / or color spaces, especially when you want to support as many displays as you can. He also pointed out the flaws in the specifications that might lead to implementation inconsistencies. He concluded with the flaws of various HDMI adapters, the issues that might arise using them on various OSes, and how to work around them when doable.

Watch the video and the slides.

The Serial Device Bus – Johan Hovold

Talk selected by Thomas Petazzoni

The Serial Device Bus – Johan HovoldJohan started his talk at ELCE by exposing the problem with how serial ports (UARTs) are currently handled in the Linux kernel. Serial ports are handled by the TTY layer, which allows user-space applications to send and receive data with what is connected on the other side of the UART. However, the kernel doesn’t provide a good mechanism to model the device that is connected at the other side of the UART, such as a Bluetooth chip. Due to this, people have resorted to either writing user-space drivers for such devices (which falls short when those devices need additional resources such as regulators, GPIOs, etc.) or to developing specific TTY line-discipline in the kernel. The latter also doesn’t work very well because a line discipline needs to be explicitly attached to a UART to operate, which requires a user-space program such as hciattach used in Bluetooth applications.

In order to address this problem, Johan picked up the work initially started by Rob Herring (Linaro), which consisted in writing a serial device bus (serdev in short), which consists in turning UART into a proper bus, with bus controllers (UART controllers) and devices connected to this bus, very much like other busses in the Linux kernel (I2C, SPI, etc.). serdev was initially merged in Linux 4.11, with some improvements being merged in follow-up versions. Johan then described in details how serdev works. First, there is a TTY port controller, which instead of registering the traditional TTY character device will register the serdev controller and its slave devices. Thanks to this, one can described in its Device Tree child nodes of the UART controller node, which will be probed as serdev slaves. There is then a serdev API to allow the implementation of serdev slave drivers, that can send and receive data of the UART. Already a few drivers are using this API: hci_serdev, hci_bcm, hci_ll, hci_nokia (Bluetooth) and qca_uart (Ethernet).

We found this talk very interesting, as it clearly explained what is the use case for serdev and how it works, and it should become a very useful subsystem for many embedded applications that use UART-connected devices.

Watch the video and the slides.

GStreamer for Tiny Devices – Olivier Crête

Talk selected by Grégory Clement

GStreamer for Tiny Devices – Olivier CrêteThe purpose of this talk was to show how to shrink Gstreamer to make it fit in an embedded Linux device. First, Olivier Crête introduced what GStreamer is, it was very high level but well done. Then after presenting the issue, he showed step by step how he managed to reduce the footprint of a GStreamer application to fit in his device.

In a first part it was a focus on features specific to GStreamer such as how to generate only the needed plugins. Then most of the tricks showed could be used for any C or C++ application. The talk was pretty short so there was no useless or boring part Moreover, the speaker himself was good and dynamic.

To conclude it was a very pleasant talk teaching step by step how to reduce the footprint of an application being GStreamer or not.

Watch the video and the slides.

Free Electrons at NetDev 2.2

NetDev 2.2Back in April 2017, Free Electrons engineer Antoine Ténart participated to NetDev 2.1, the most important conference discussing Linux networking support. After the conference, Antoine published a summary of it, reporting on the most interesting talks and topics that have been discussed.

Next week, NetDev 2.2 takes place in Seoul, South Korea, and this time around, two Free Electrons engineers will be attending the event: Alexandre Belloni and Antoine Ténart. We are getting more and more projects with networking related topics, and therefore the wide range of talks proposed at NetDev 2.2 will definitely help grow our expertise in this field.

Do not hesitate to get in touch with Alexandre or Antoine if you are also attending this event!

Buildroot Long Term Support releases: from 2017.02 to 2017.02.6 and counting

Buildroot LogoBuildroot is a widely used embedded Linux build systems. A large number of companies and projects use Buildroot to produce customized embedded Linux systems for a wide range of embedded devices. Most of those devices are now connected to the Internet, and therefore subject to attacks if the software they run is not regularly updated to address security vulnerabilities.

The Buildroot project publishes a new release every three months, with each release providing a mix of new features, new packages, package updates, build infrastructure improvements… and security fixes. However, until earlier this year, as soon as a new version was published, the maintenance of the previous version stopped. This means that in order to stay up to date in terms of security fixes, users essentially had two options:

  1. Update their Buildroot version regularly. The big drawback is that they get not only security updates, but also many other package updates, which may be problematic when a system is in production.
  2. Stick with their original Buildroot version, and carefully monitor CVEs and security vulnerabilities in the packages they use, and update the corresponding packages, which obvisouly is a time-consuming process.

Starting with 2017.02, the Buildroot community has decided to offer one long term supported release every year: 2017.02 will be supported one year in terms of security updates and bug fixes, until 2018.02 is released. The usual three-month release cycle still applies, with 2017.05 and 2017.08 already being released, but users interested in a stable Buildroot version that is kept updated for security issues can stay on 2017.02.

Since 2017.02 was released on February 28th, 2017, six minor versions were published on a fairly regularly basis, almost every month, except in August:

With about 60 to 130 commits between each minor version, it is relatively easy for users to check what has been changed, and evaluate the impact of upgrading to the latest minor version to benefit from the security updates. The commits integrated in those minor versions are carefully chosen with the idea that users should be able to easily update existing systems.

In total, those six minor versions include 526 commits, of which 183 commits were security updates, representing roughly one third of the total number of commits. The other commits have been:

  • 140 commits to fix build issues
  • 57 commits to bump versions of packages for bug fixes. These almost exclusively include updates to the Linux kernel, using its LTS versions. For other packages, we are more conservative and generally don’t upgrade them.
  • 17 commits to address issues in the licensing description of the packages
  • 186 commits to fix miscellaneous issues, ranging from runtime issues affecting packages to bugs in the build infrastructure

The Buildroot community has already received a number of bug reports, patches or suggestions specifically targetting the 2017.02 LTS version, which indicates that developers and companies have started to adopt this LTS version.

Therefore, if you are interested in using Buildroot for a product, you should probably consider using the LTS version! We very much welcome feedback on this version, and help in monitoring the security vulnerabilities affecting software packages in Buildroot.

Mali OpenGL support on Allwinner platforms with mainline Linux

As most people know, getting GPU-based 3D acceleration to work on ARM platforms has always been difficult, due to the closed nature of the support for such GPUs. Most vendors provide closed-source binary-only OpenGL implementations in the form of binary blobs, whose quality depend on the vendor.

This situation is getting better and better through vendor-funded initiatives like for the Broadcom VC4 and VC5, or through reverse engineering projects like Nouveau on Tegra SoCs, Etnaviv on Vivante GPUs, Freedreno on Qualcomm’s. However there are still GPUs where you do not have the option to use a free software stack: PowerVR from Imagination Technologies and Mali from ARM (even though there is some progress on the reverse engineering effort).

Allwinner SoCs are using either a Mali GPU from ARM or a PowerVR from Imagination Technologies, and therefore, support for OpenGL on those platforms using a mainline Linux kernel has always been a problem. This is also further complicated by the fact that Allwinner is mostly interested in Android, which uses a different C library that avoids its use in traditional glibc-based systems (or through the use of libhybris).

However, we are happy to announce that Allwinner gave us clearance to publish the userspace binary blobs that allows to get OpenGL supported on Allwinner platforms that use a Mali GPU from ARM, using a recent mainline Linux kernel. Of course, those are closed source binary blobs and not a nice fully open-source solution, but it nonetheless allows everyone to have OpenGL support working, while taking advantage of all the benefits of a recent mainline Linux kernel. We have successfully used those binary blobs on customer projects involving the Allwinner A33 SoCs, and they should work on all Allwinner SoCs using the Mali GPU.

In order to get GPU support to work on your Allwinner platform, you will need:

  • The kernel-side driver, available on Maxime Ripard’s Github repository. This is essentially the Mali kernel-side driver from ARM, plus a number of build and bug fixes to make it work with recent mainline Linux kernels.
  • The Device Tree description of the GPU. We introduced Device Tree bindings for Mali GPUs in the mainline kernel a while ago, so that Device Trees can describe such GPUs. Such description has been added for the Allwinner A23 and A33 SoCs as part of this commit.
  • The userspace blob, which is available on Free Electrons GitHub repository. It currently provides the r6p2 version of the driver, with support for both fbdev and X11 systems. Hopefully, we’ll gain access to newer versions in the future, with additional features (such as GBM support).

If you want to use it in your system, the first step is to have the GPU definition in your device tree if it’s not already there. Then, you need to compile the kernel module:

git clone
cd sunxi-mali
./ -r r6p2 -b
./ -r r6p2 -i

It should install the mali.ko Linux kernel module into the target filesystem.

Now, you can copy the OpenGL userspace blobs that match your setup, most likely the fbdev or X11-dma-buf variant. For example, for fbdev:

git clone
cd mali-blobs
cp -a r6p2/fbdev/lib/lib_fb_dev/lib* $TARGET_DIR/usr/lib

You should be all set. Of course, you will have to link your OpenGL applications or libraries against those user-space blobs. You can check that everything works using OpenGL test programs such as es2_gears for example.

Linux 4.13 released, Free Electrons contributions

Linux 4.13 was released last Sunday by Linus Torvalds, and the major new features of this release were described in details by LWN in a set of articles: part 1 and part 2.

This release gathers 13006 non-merge commits, amongst which 239 were made by Free Electrons engineers. According to the LWN article on 4.13 statistics, this makes Free Electrons the 13th contributing company by number of commits, the 10th by lines changed.

The most important contributions from Free Electrons for this release have been:

  • In the RTC subsystem
    • Alexandre Belloni introduced a new method for registering RTC devices, with one step for the allocation, and one step for the registration itself, which allows to solve race conditions in a number of drivers.
    • Alexandre Belloni added support for exposing the non-volatile memory found in some RTC devices through the Linux kernel nvmem framework, making them usable from userspace. A few drivers were changed to use this new mechanism.
  • In the MTD/NAND subsystem
    • Boris Brezillon did a large number of fixes and minor improvements in the NAND subsystem, both in the core and in a few drivers.
    • Thomas Petazzoni contributed the support for on-die ECC, specifically with Micron NANDs. This allows to use the ECC calculation capabilities of the NAND chip itself, as opposed to using software ECC (calculated by the CPU) or ECC done by the NAND controller.
    • Thomas Petazzoni contributed a few improvements to the FSMC NAND driver, used on ST Spear platforms. The main improvement is to support the ->setup_data_interface() callback, which allows to configure optimal timings in the NAND controller.
  • Support for Allwinner ARM platforms
    • Alexandre Belloni improved the sun4i PWM driver to use the so-called atomic API and support hardware read out.
    • Antoine Ténart improved the sun4i-ss cryptographic engine driver to support the Allwinner A13 processor, in addition to the already supported A10.
    • Maxime Ripard contributed HDMI support for the Allwinner A10 processor (in the DRM subsystem) and a number of related changes to the Allwinner clock support.
    • Quentin Schulz improved the support for battery charging through the AXP20x PMIC, used on Allwinner platforms.
  • Support for Atmel ARM platforms
    • Alexandre Belloni added suspend/resume support for the Atmel SAMA5D2 clock driver. This is part of a larger effort to implement the backup mode for the SAMA5D2 processor.
    • Alexandre Belloni added suspend/resume support in the tcb_clksrc driver, used as for clocksource and clockevents on Atmel SAMA5D2.
    • Alexandre Belloni cleaned up a number of drivers, removing support for non-DT probing, which is possible now that the AVR32 architecture has been dropped. Indeed, the AVR32 processors used to share the same drivers as the Atmel ARM processors.
    • Alexandre Belloni added the core support for the backup mode on Atmel SAMA5D2, a suspend/resume state with significant power savings.
    • Boris Brezillon switched Atmel platforms to use the new binding for the EBI and NAND controllers.
    • Boris Brezillon added support for timing configuration in the Atmel NAND driver.
    • Quentin Schulz added suspend/resume support to the Bosch m_can driver, used on Atmel platforms.
  • Support for Marvell ARM platforms
    • Antoine Ténart contributed a completely new driver (3200+ lines of code) for the Inside Secure EIP197 cryptographic engine, used in the Marvell Armada 7K and 8K processors. He also subsequently contributed a number of fixes and improvements for this driver.
    • Antoine Ténart improved the existing mvmdio driver, used to communicate with Ethernet PHYs over MDIO on Marvell platforms to support the XSMI variant found on Marvell Armada 7K/8K, used to communicate with 10G capable PHYs.
    • Antoine Ténart contributed minimal support for 10G Ethernet in the mvpp2 driver, used on Marvell Armada 7K/8K. For now, the driver still relies on low-level initialization done by the bootloader, but additional changes in 4.14 and 4.15 will remove this limitation.
    • Grégory Clement added a new pinctrl driver to configure the pin-muxing on the Marvell Armada 37xx processors.
    • Grégory Clement did a large number of changes to the clock drivers used on the Marvell Armada 7K/8K processors to prepare the addition of pinctrl support.
    • Grégory Clement added support for Marvell Armada 7K/8K to the existing mvebu-gpio driver.
    • Thomas Petazzoni added support for the ICU, a specialized interrupt controller used on the Marvell Armada 7K/8K, for all devices located in the CP110 part of the processor.
    • Thomas Petazzoni removed a work-around to properly resume per-CPU interrupts on the older Marvell Armada 370/XP platforms.
  • Support for RaspberryPi platforms
    • Boris Brezillon added runtime PM support to the HDMI encoder driver used on RaspberryPi platforms, and contributed a few other fixes to the VC4 DRM driver.

It is worth mentioning that Miquèl Raynal, recently hired by Free Electrons, sees his first kernel patch merged: nand: fix wrong default oob layout for small pages using soft ecc.

Free Electrons engineers are not only contributors, but also maintainers of various subsystems in the Linux kernel, which means they are involved in the process of reviewing, discussing and merging patches contributed to those subsystems:

  • Maxime Ripard, as the Allwinner platform co-maintainer, merged 113 patches from other contributors
  • Boris Brezillon, as the MTD/NAND maintainer, merged 62 patches from other contributors
  • Alexandre Belloni, as the RTC maintainer and Atmel platform co-maintainer, merged 57 patches from other contributors
  • Grégory Clement, as the Marvell EBU co-maintainer, merged 47 patches from other contributors

Here is the commit by commit detail of our contributors to 4.13:

Updated bleeding edge toolchains on

Two months ago, we announced a new service from Free Electrons: free and ready-to-use Linux cross-compilation toolchains, for a large number of architectures and C libraries, available at

Bleeding edge toolchain updates

All our bleeding edge toolchains have been updated, with the latest version of the toolchain components:

  • gcc 7.2.0, which was released 2 days ago
  • glibc 2.26, which was released 2 weeks ago
  • binutils 2.29
  • gdb 8.0

Those bleeding edge toolchains are now based on Buildroot 2017.08-rc2, which brings a nice improvement: the host tools (gcc, binutils, etc.) are no longer linked statically against gmp, mpfr and other host libraries. They are dynamically linked against them with an appropriate rpath encoded into the gcc and binutils binaries to find those shared libraries regardless of the installation location of the toolchain.

However, due to gdb 8.0 requiring a C++11 compiler on the host machine (at least gcc 4.8), our bleeding edge toolchains are now built in a Debian Jessie system instead of Debian Squeeze, which means that at least glibc 2.14 is needed on the host system to use them.

The only toolchains for which the tests are not successful are the MIPS64R6 toolchains, due to the Linux kernel not building properly for this architecture with gcc 7.x. This issue has already been reported upstream.

Stable toolchain updates

We haven’t changed the component versions of our stable toolchains, but we made a number of fixes to them:

  • The armv7m and m68k-coldfire toolchains have been rebuilt with a fixed version of elf2flt that makes the toolchain linker directly usable. This fixes building the Linux kernel using those toolchains.
  • The mips32r5 toolchain has been rebuilt with NaN 2008 encoding (instead of NaN legacy), which makes the resulting userspace binaries actually executable by the Linux kernel, which expects NaN 2008 encoding on mips32r5 by default.
  • Most mips toolchains for musl have been rebuilt, with Buildroot fixes for the creation of the dynamic linker symbolic link. This has no effect on the toolchain itself, but also the tests under Qemu to work properly and validate the toolchains.

Other improvements

We made a number of small improvements to the site:

  • Each architecture now has a page that lists all toolchain versions available. This allows to easily find a toolchain that matches your requirements (in terms of gcc version, kernel headers version, etc.). See All aarch64 toolchains for an example.
  • We added a FAQ as well as a news page.

As usual, we welcome feedback about our toolchains, either on our bug tracker or by mail at

Free Electrons proposes an I3C subsystem for the Linux kernel

MIPI I3C fact sheet, from the MIPI I3C white paper
MIPI I3C fact sheet, from the MIPI I3C white paper
At the end of 2016, the MIPI consortium has finalized the first version of its I3C specification, a new communication bus that aims at replacing older busses like I2C or SPI. According to the specification, I3C gets closer to SPI data rate while requiring less pins and adding interesting mechanisms like in-band interrupts, hotplug capability or automatic discovery of devices connected on the bus. In addition, I3C provides backward compatibility with I2C: I3C and legacy I2C devices can be connected on a common bus controlled by an I3C master.

For more details about I3C, we suggest reading the MIPI I3C Whitepaper, as unfortunately MIPI has not publicly released the specifications for this protocol.

For the last few months, Free Electrons engineer Boris Brezillon has been working with Cadence to develop a Linux kernel subsystem to support this new bus, as well as Cadence’s I3C master controller IP. We have now posted the first version of our patch series to the Linux kernel mailing list for review, and we already received a large number of very useful comments from the kernel community.

Free Electrons is proud to be pioneering the support for this new bus in the Linux kernel, and hopes to see other developers contribute to this subsystem in the near future!