Linaro Releases

The Linaro Qualcomm Landing Team is pleased to announce the new release of the Linaro Linux release for Qualcomm™ Snapdragon® 410 processor. The Linaro Linux release 17.04.1 is a Debian-based Linaro Build that provides developers with a desktop like environment using Debian and the LXQt desktop, as well as a console-only image.

What’s new in this release

This is a patch release with the following changes:

• Upgrade to Linux kernel 4.9.30
• After unplug/plug of the HDMI cable, the GUI does not come back
• synchronous external abort when display goes blank
• SMD channel are not closed on device removal
• Fixed release notes to include instructions about LK signing tool

Important note(s) about this release

A firmware/bootloader upgrade was required for this release. If you install using the SD card method, then the new bootloaders are flashed automatically, if you install with the fastboot images, then you need to install the latest bootloader package.

Known issues

For a list of all known issues, please check our Bugzilla

Features

The Linaro Linux version 17.04.1 for the Snapdragon 410 supports the following features:

• Provides a working Debian environment with access to Debian repositories (apt-get) and updates. It is based on Debian 9.0 (aka stretch).
• It is based on Linux kernel 4.9.30.
• It is based on proprietary firmware available on Qualcomm Developer Network or 96boards.org.
• The following images are released:
  • boot image that includes prebuilt kernel and initrd
  • developer image that includes core packages as well as typical development packages (headless)
  • alip image that includes a minimal desktop environment GUI using LXQt
• All images have a pre-configured user called linaro, and the password for this user is set to linaro
• The root file system should be flashed in the onboard eMMC.
• The following features are supported on the DragonBoard 410c:
  • Quad Core armv8 A53 CPU (@1.2GHz)
  • PSCI with support for cpuidle, SMP, hotplug and restart.
  • Adreno 306 GPU, powered by freedreno Mesa/Gallium GPU driver, version 13.0.6
    • OpenGL 3.1, OpenGLES 3.0, GLX, EGL
    • X11 -modesetting video driver
  • Cpufreq, using ondemand governor by default
  • CPU thermal sensors, and thermal management using the step wise governor
  • HDMI display and audio using the onboard ADV7533 MIPI/DSI Receiver with HDMI Transmitter from Analog Devices
  • UART, SD, eMMC
  • USB2.0 (Mouse, Keyboard, Storage, Ethernet)
  • Wifi and Bluetooth using integrated WCN3620, using proprietary firmware and wcn36xx driver
  • Hardware accelerated video codecs using dedicated Snapdragon coprocessor
  • Onboard GPS using Qualcomm GNSS subsystem
  • CSI interface with external ISP only, sample driver included for OV5645, with support for the following features:
    • CSI0 and/or CSI1 on the high speed expansion connector
    • streaming from the camera sensor or CSID test generator
    • raw dump to memory: 8bit packed UYVY format from OV5645 and resolutions of 2592×1944, 1920×1080 or 1280×960
    • format conversion from 8bit packed UYVY to NV12

Information about the DragonBoard 410c

For more information about the DragonBoard 410c, please check the following website and wiki:

• http://www.96boards.org/dragonboard410c
• http://www.96boards.org/dragonboard410c/documentation
• http://www.96boards.org/dragonboard410c/gettingstarted
• https://github.com/96boards/documentation/wiki
• https://developer.qualcomm.com/hardware/dragonboard-410c

How to install and use this release

To install this release please follow the instructions from the DragonBoard 410c Linux user guide.

Proprietary firmware

This release contains proprietary firmware. You can download the proprietary firmware separately, from here. All the required firmware files are pre-installed, and our images are bound to the Qualcomm Linux BSP license agreement. A copy of the LICENSE can be found in the image as /etc/QCOM-LINUX-BOARD-SUPPORT-LICENSE.

Running the ALIP Desktop image

The ALIP/LXQt image is expected to provide a desktop-like experience, as such it is recommended to use an HDMI monitor, as well as USB keyboard and mouse. The default image will directly boot to the login screen by default. However a root console login will also be started on the serial console.

Note: The default bootargs enable the kernel messages to be displayed on the serial console.

Running the Developer based image

If you have flashed the developer image, when booting the board you will end up in a root login on the serial console. If you have an HDMI monitor connected, you will also have login terminals on the display

How to get and customize the kernel source code

Building the Linux kernel from source

The Linux kernel used in this release is available via tags in the Linaro Qualcomm Landing Team git repository:

git: http://git.linaro.org/landing-teams/working/qualcomm/kernel.git
tag: debian-qcom-dragonboard410c-17.04.1
defconfig: arch/arm64/defconfig kernel/configs/distro.config

The kernel image (Image) is located in the boot image and partition and the kernel modules are installed in the root file system. It is possible for a user to rebuild the kernel and run a custom kernel image instead of the released kernel. You can build the kernel using any recent GCC release using the git tree, tag and defconfig mentioned above. This release only supports booting with device tree, as such both the device tree blobs need to be built as well.

The DragonBoard 410c is an armv8 platform, and the kernel is compiled for the Aarch64 target. Even though it is possible to build natively, on the target board, It is recommended to build the Linux kernel on a PC development host. In which case you need to install a cross compiler for the arm architecture. It is recommended to download the Linaro GCC cross compiler.

To build the Linux kernel, you can use the following instructions:

git clone -n http://git.linaro.org/landing-teams/working/qualcomm/kernel.git
cd kernel
git checkout -b kernel-17.04.1 debian-qcom-dragonboard410c-17.04.1
export ARCH=arm64
export CROSS_COMPILE=<path to your GCC cross compiler>/aarch64-linux-gnu-
make defconfig distro.config
make -j4 Image dtbs KERNELRELEASE=4.9.30-linaro-lt-qcom

Additionally, you might want or need to compile the kernel modules:

make -j4 modules KERNELRELEASE=4.9.30-linaro-lt-qcom

Building a boot image

You now need to create a valid boot image with your own kernel build.

On your host PC, we need to install the following tools:

sudo apt-get install device-tree-compiler
git clone git://codeaurora.org/quic/kernel/skales

The boot image consists of the table of device tree (dt.img), the kernel image (Image) and an init ramdisk image.

The dtbTool is a standalone application that will process the DTBs generated during the kernel build, to create the table of device tree image. This tool is included in the skales git tree cloned above.

./skales/dtbTool -o dt.img -s 2048 arch/arm64/boot/dts/qcom/

To create the boot image, you also need a ramdisk image, and you can use the one from the release:

wget http://builds.96boards.org/releases/dragonboard410c/linaro/debian/17.04.1/initrd.img-4.9.30-linaro-lt-qcom

The tool mkbootimg (also in the git tree previously cloned) is a standalone application that will process all files and create the boot image that can then be booted on the target board, or flash into the on-board eMMC. The boot image also contains the kernel bootargs, which can be changed as needed in the next command:

./skales/mkbootimg --kernel arch/arm64/boot/Image \
                   --ramdisk initrd.img-4.9.30-linaro-lt-qcom \
                   --output boot-db410c.img \
                   --dt dt.img \
                   --pagesize 2048 \
                   --base 0x80000000 \
                   --cmdline "root=/dev/disk/by-partlabel/rootfs rw rootwait console=ttyMSM0,115200n8"

Booting a custom boot image

Assuming you have now built a valid boot image called boot-db410c.img, you can run the following fastboot command to boot it on the board:

sudo fastboot boot boot-db410c.img

If you want to permanently use a custom kernel image, you can update the boot image and reflash it into the boot partition:

sudo fastboot flash boot boot-db410c.img

How to get and customize the bootloader

While the first stage bootloader is proprietary and released as firmware blob available on Qualcomm Developer Network or 96boards.org, the second stage bootloader is LK and is open source.

The original LK source code is available on CodeAurora.org, and the source code which is used in this release can be found in the Linaro Qualcomm Landing Team git repository:

git: http://git.linaro.org/landing-teams/working/qualcomm/lk.git
tag: dragonboard410c-LA.BR.1.2.7-03810-8x16.0-linaro1

Since release 17.04, the first stage bootloader expects the second stage bootloader to include a signature. The signature is typically used for secure boot. On Dragonboard 410c, since no private keys are available, the signature is not checked for integrity however the signature section, even if empty, must exist. To build the LK bootloader and add the signature section, you can use the following instructions:

git clone git://codeaurora.org/platform/prebuilts/gcc/linux-x86/arm/arm-eabi-4.8.git -b LA.BR.1.1.3.c4-01000-8x16.0
git clone http://git.linaro.org/landing-teams/working/qualcomm/lk.git -b dragonboard410c-LA.BR.1.2.7-03810-8x16.0-linaro1
git clone https://git.linaro.org/landing-teams/working/qualcomm/signlk.git
cd lk
make -j4 msm8916 EMMC_BOOT=1 TOOLCHAIN_PREFIX=<path to arm-eabi-4.8 tree>/bin/arm-eabi-
mv ./build-msm8916/emmc_appsboot.mbn ./build-msm8916/emmc_appsboot_unsigned.mbn
../signlk/signlk.sh -i=./build-msm8916/emmc_appsboot_unsigned.mbn -o=./build-msm8916/emmc_appsboot.mbn -d

The second stage bootloader is flashed on the aboot partition, you can now flash your board with:

sudo fastboot flash aboot ./build-msm8916/emmc_appsboot.mbn

How to get and customize Debian/Ubuntu packages source code

This release is based on Debian 9.0 (aka stretch), and it is not possible to use a different Debian release (e.g. it is not possible to downgrade to an older Debian release, nor is it possible to use a newer release, such as the one being currently developed).

Since all packages installed in Linaro Debian-based images are maintained either in Debian archives or in Linaro repositories, it is possible for users to update their environment with commands such as:

sudo apt-get update
sudo apt-get upgrade

All user space software is packaged using Ubuntu or Debian packaging process. As such you can find extensive information about using, patching and building packages in the Ubuntu packaging guide or The Debian New Maintainers Guide. If you quickly want to rebuild any package, you can run the following commands to fetch the package source code and install all build dependencies:

sudo apt-get update
sudo apt-get build-dep <pkg>
apt-get source <pkg>

Then you can rebuild the package locally with:

cd <pkg-version>
dpkg-buildpackage -b -us -uc

Notes:

• you can drop patches in debian/patches/ and update debian/patches/series before building with dpkg-buildpackage to customize the source code.
• all associated .deb files will be located in the root folder, and can be installed in the system with dpkg -i .deb.
• all these commands should be executed on the target directly, not on the development host. It is generally enough to build packages natively, on the target platform. For some packages, it is possible to cross compile Ubuntu/Debian packages however this goes beyond the scope of this wiki page.

Using the onboard GPS

The GPS software stack mostly runs on the DSP subsystem. The communication between the main CPU and the DSP is done with a specific IPC driver called QRTR (see ./net/qrtr/ in the kernel source tree). The DSP is not started automatically at boot. To start the GPS, the DSP needs to be started first. Once the DSP is started any gpsd client can be started and will be able to retrieve GPS data.

Please note that the sensitivity of the onboard antenna is quite low, so getting a FIX will take several minutes. Please refer to the dedicated application note to install an external antenna for better GPS performance.

To get started with GPS, first install the following packages:

sudo apt-get install gpsd-clients gnss-gpsd

The package gnss-gpsd will bring all the needed dependencies to use the onboard GPS. Then you need to start the DSP:

sudo systemctl start qdsp-start.service

From now on, you can use any gpsd client, such as gpsmon or xgps.

Using CSI camera

Introduction

Only basic use cases are supported at the moment. More features will be added, such as support for scaling, cropping.

The release is configured with no camera, and users with camera are expected to configure the DTS file accordingly. Please check commit “dts: Disable camera sensors in dtsi” in the kernel, it should mostly be a matter of reverting this patch. Note that this patch assumes that 2 OV5645 sensors are connected on each CSI port. The setup is validated using an upcoming STM based mezzanine (STM32F446 sensor board).

This release includes drivers for:

• OV5645 camera sensor;
• QC MSM camera sub-system (CSIPHY, CSID, ISPIF, VFE);
• QC Camera control interface.

The OV5645 camera sensor driver is in a final stage of implementation and is currently undergoing review in upstream lists.

The CAMSS (camera sub-system) driver is supports direct dump to memory and format conversion (UYVY to NV12). The version which supports direct dump to memory only was published on linux-media, the format conversion will follow. The CAMSS HW on 8016 consists of two CSIPHY modules, two CSID modules, ISPIF module and VFE module. The driver is implemented to configure each of these HW modules. It is a V4L2 driver which also utilizes the media controller framework to model the internal topology of the system. The driver registers V4L2 subdevice nodes for each of the CSIPHY and CSID modules, then registers two V4L2 subdevice nodes to represent the ISPIF module. Then it registers three V4L2 subdevice nodes to represent each of the RDI channels in the VFE and another V4L2 subdevice node for the processing part of the VFE.

The CCI (camera control interface) driver is a version which originates from QC Android driver and is now separated from the CAMSS driver and compiled on our linux. For this another V4L2 driver and media device are created – this is only a temporary work to enable control on the camera sensor. Proper implementation will follow.

Basic usage

Make sure that you have the following package installed:

sudo apt-get install v4l-utils

To ensure that your sensor is properly connected, you can inspect the output of the following command:

sudo media-ctl -d /dev/media1 -p

If everything is ok, you should see something like this:

entity 13: ov5645 1-0076 (1 pad, 1 link)
type V4L2 subdev subtype Unknown flags 0
device node name /dev/v4l-subdev9
pad0: Source
[fmt:UYVY8_2X8/1920x1080 field:none
crop:(0,0)/0x0]
-> "msm_csiphy0":0 [ENABLED,IMMUTABLE]

At this point you need to configure the pipeline: link CSIPHY to CSID, CSID to ISPIF, ISPIF to VFE. Then configure formats on all entities in the pipeline. For direct dump to memory (RDI channels) this looks like this:

sudo media-ctl -d /dev/media1 -l '"msm_csiphy0":1->"msm_csid0":0[1],"msm_csid0":1->"msm_ispif0":0[1],"msm_ispif0":1->"msm_vfe0_rdi0":0[1]'
sudo media-ctl -d /dev/media1 -V '"ov5645 1-0076":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_csiphy0":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_csid0":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_ispif0":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_vfe0_rdi0":0[fmt:UYVY8_2X8/1920x1080 field:none]'

At this point, the pipeline should be configured and ready to be used by any application that can use v4l2. For example, you can use Gstreamer to take a JPEG picture:

gst-launch-1.0 v4l2src device=/dev/video0 num-buffers=1 ! 'video/x-raw,format=UYVY,width=1920,height=1080,framerate=30/1' ! jpegenc ! filesink location=image01.jpg

Or you can use Gstreamer to show a live preview from the camera:

gst-launch-1.0 v4l2src ! glimagesink

Pipeline configuration for the format conversion looks like this:

sudo media-ctl -d /dev/media1 -l '"msm_csiphy0":1->"msm_csid0":0[1],"msm_csid0":1->"msm_ispif0":0[1],"msm_ispif0":1->"msm_vfe0_pix":0[1]'
sudo media-ctl -d /dev/media1 -V '"ov5645 1-0076":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_csiphy0":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_csid0":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_ispif0":0[fmt:UYVY8_2X8/1920x1080 field:none],"msm_vfe0_pix":0[fmt:UYVY8_2X8/1920x1080 field:none]'

And similar Gstreamer pipeline for a JPEG picture:

gst-launch-1.0 v4l2src device=/dev/video3 num-buffers=1 ! 'video/x-raw,format=NV12,width=1920,height=1080,framerate=30/1' ! jpegenc ! filesink location=image02.jpg

If you have a second camera sensor and intend to use it concurrently then link and configure another pipeline which includes the second camera and the unused entities. Use a v4l2 application the same way only pointing the correct video device node used in this pipeline.

Video record pipeline

Starting with this release a basic video recording GStreamer pipeline is supported involving the camera and video encoder. Currently this has the following limitations:

• The frame data must be vertically aligned to 32 lines. For the OV5645 camera this means that only 1280×960 resolution is supported.
• A GStreamer video encoder plugin is needed which is not a part of the release. To enable this the user must patch and rebuild gstreamer1.0-plugins-good package – this is recommended for advanced users only.

To enable the video encoder plugin follow these steps. Install needed packages and dependencies:

sudo apt-get install build-essential fakeroot devscripts quilt
sudo apt-get build-dep gstreamer1.0-plugins-good

Get gstreamer1.0-plugins-good source:

apt-get source gstreamer1.0-plugins-good
cd gst-plugins-good1.0-1.10.4/

Get the video encoder plugin patches. These are the 12 last commits from here. Download them as patch files and to apply them do the following for each patch:

quilt import path_to_patch.patch
quilt push

Edit debian/rules and add --enable-v4l2-probe and --without-libv4l2 options to DEB_CONFIGURE_EXTRA_FLAGS.

Build gstreamer1.0-plugins-good package:

debuild -b -uc -us

Install the newly built gstreamer1.0-plugins-good package:

dpkg -i ../gstreamer1.0-plugins-good_1.10.4-1_arm64.deb

Now you are ready to do video recording! Configure the pipeline:

sudo media-ctl -d /dev/media1 -l '"msm_csiphy0":1->"msm_csid0":0[1],"msm_csid0":1->"msm_ispif0":0[1],"msm_ispif0":1->"msm_vfe0_pix":0[1]'
sudo media-ctl -d /dev/media1 -V '"ov5645 1-0076":0[fmt:UYVY8_2X8/1280x960 field:none],"msm_csiphy0":0[fmt:UYVY8_2X8/1280x960 field:none],"msm_csid0":0[fmt:UYVY8_2X8/1280x960 field:none],"msm_ispif0":0[fmt:UYVY8_2X8/1280x960 field:none],"msm_vfe0_pix":0[fmt:UYVY8_2X8/1280x960 field:none]'

And do video recording:

gst-launch-1.0 -e v4l2src device=/dev/video3 ! video/x-raw,format=NV12,width=1280,height=960,framerate=30/1 ! v4l2video5h264enc extra-controls="controls,h264_profile=4,video_bitrate=2000000;" ! h264parse ! mp4mux ! filesink location=video.mp4

When using the above command make sure that video device node numbers are correct or check and use the correct ones:

• examine the output of media-ctl -d /dev/media1 -p and check that /dev/video3 is connected to msm_vfe0_pix
• examine the output of gst-inspect-1.0 video4linux2 and check the name of the V4L2 H.264 Encoder plugin – e.g. v4l2video5h264enc

Using CSID Test Generator

If you do not have any sensor, it is possible to use the internal CSID Test Generator (TG). To enable it download and compile the yavta tool from here:

git clone git://git.ideasonboard.org/yavta.git
cd yavta
make

Then, enable the test generator:

sudo ./yavta --no-query -w '0x009f0903 1' /dev/v4l-subdev2

Then, configure the pipeline:

sudo media-ctl -d /dev/media1 -l '"msm_csid0":1->"msm_ispif0":0[1],"msm_ispif0":1->"msm_vfe0_rdi0":0[1]'
sudo media-ctl -d /dev/media1 -V '"msm_csid0":1[fmt:UYVY2X8/1920x1080],"msm_ispif0":0[fmt:UYVY2X8/1920x1080],"msm_vfe0_rdi0":0[fmt:UYVY2X8/1920x1080]'

Finally, you can use any v4l2 application, such as Gstreamer.

Optimized video pipeline

In order to exercise a fully optimized video pipeline some specific settings need to be applied throughout the pipeline elements. With the fully optimized video use case there is no extra buffer copy, and the various drivers involved are sharing all video buffers using dmabuf.

To exercise the opimized video pipeline, you can use for example the following gstreamer command:

GST_GL_PLATFORM=egl gst-launch-1.0 filesrc location=<path to file> ! qtdemux name=m m.video_0 ! h264parse ! v4l2video0dec capture-io-mode=dmabuf ! glimagesink

Audio configuration settings

The release has support for both analog and digital audio (HDMI). When using the ALIP desktop image, to switch back and forth you can use PulseAudio volume control application, and in the configuration tab, you will be able to chose which profile to use. Note that by default, PulseAudio will use the analog output, and you need to switch to HDMI at first boot if your HDMI monitor supports HDMI audio.

Feedback and Support

For general question or support request, please go to 96boards.org Community forum.

For any bug related to this release, please submit issues to the 96Board.org Bug tracking system. To submit a bug, follow this link.

Bugs will be reviewed and prioritized by the team. For any bug report it is recommended to provide as much information as possible, and at the very list please include the name of the release you are using, the output of dpkg -l to list all packages installed, as well, as the boot log (output of dmesg).

How to contribute

We very much encourage developers to use and contribute to our releases, using the following instructions.

---

Qualcomm Snapdragon is product of Qualcomm Technologies, Inc.