Advanced Configuration

Target Specific Configuration

Different target boards have different layouts for GPIO, LED’s and more. Often this leads to a necessity for configurations to be specified per target. In this example we will be looking at how to configure the LED’s for two different targets. First, lets start by modifying our config.exs to include configs for each target.

use Mix.Config

config :blinky, led_list: [ :red, :green ]

import_config "#{Mix.Project.config[:target]}.exs"

This will load separate Mix Configs for each target passed. Lets say we have targets rpi and bbb each of those files would look like this.

rpi

config :nerves_io_led, names: [ red: "led0", green: "led1" ]

bbb

config :nerves_io_led, names: [
  led0: "beaglebone:green:usr0",
  led1: "beaglebone:green:usr1",
  led2: "beaglebone:green:usr2",
  led3: "beaglebone:green:usr3"
]

Root Filesystem Additions

Sometimes you want to ship additional files and configurations with your firmware. You can do this by providing your own directory of root file system additions. This is done by configuring the firmware assembler and telling it where to find the folder it should use as an overlay.

# config/config.exs

config :nerves, :firmware,
  rootfs_additions: "config/rootfs-additions"

This declares that the contents of the folder at config/rootfs-additions will be merges into the root file system when mix firmware is called. You can also specify different rootfs additions per target as illustrated above.

Overwriting Files in the Root File System

Any files in the rootfs_additions will overwrite those present in the underlying system. This can be useful if you want to change the value of included files in the underlying Nerves system. Lets say for example that you want to change the behaviour of erlinit. You can include your own erlinit.config

config/rootfs-additions/etc/erlinit.config

# Uncomment to hang the board rather than rebooting when Erlang exits
#--hang-on-exit

# Enable UTF-8 filename handling in Erlang and custom inet configuration
-e LANG=en_US.UTF-8;LANGUAGE=en;ERL_INETRC=/etc/erl_inetrc

# Mount the configdata partition
# See http://www.linuxfromscratch.org/lfs/view/6.3/chapter08/fstab.html about
# ignoring warning the Linux kernel warning about using UTF8 with vfat.
-m /dev/mmcblk0p3:/root:vfat::

# Erlang release search path
-r /srv/erlang

# Hostname
-d "/usr/bin/boardid -b bbb -n 4"
-n nerves-%.4s

It is important to note that if you replace a config file, that you should first obtain and modify the original file. A trick for doing this is to expand the rootfs.squashfs. You can do this using unsquashfs.

$ unsquashfs path/to/rootfs.squashfs

This file is typically found in _build/(Target)/(Mix.env)/nerves/system/images/rootfs.squashfs. It will be expanded into the current directory under squashfs-root

Overwriting Files in the Boot Partition

Different targets have different boot partition contents. To overwrite files in the boot partition, you will need to do this in your own fwup.conf file.

# config/config.exs

config :nerves, :firmware,
  fwup_conf: "config/fwup.conf"

In your included fwup.conf file, you can use absolute paths, or Env variables to point to the location of included files.

Lets say you have a Raspberry Pi and you want to change the contents of the cmdline.txt file. You can do this by editing the fwup.conf as follows

# fwup.conf

file-resource cmdline.txt {
    host-path = "${NERVES_APP}/config/cmdline.txt"
}

You can use NERVES_APP environment veriable to point to the root of your elixir app. This variable is automatically managed for you by nerves_bootstrap

Partitions

Nerves firmware uses Master Boot Record partition layout which states that you can define 4 partitions. By default, the root filesystem partition is mounted as read only. This is to prevent corruption of the rootfs due to “improper shutdowns”. With embedded systems, it is expected that the power can be pulled from the device at any time. This could be problematic if you are performing a write operation on the filesystem. Because of this layout, we add a read write partition called app_data. This is mounted at /root and is dictated in etc/erlinit.config

+----------------------------+
| MBR                        |
+----------------------------+
| Boot partition (FAT32)     |
+----------------------------+
| p1*: Rootfs A (squashfs)   |
+----------------------------+
| p1*: Rootfs B (squashfs)   |
+----------------------------+
| p2: App Data  (FAT32)      |
+----------------------------+

You can enable and mount an additional read write partition by modifying the fwup.conf file. This strategy is typically used to define two locations where data can be written. Lets say you want to persist some infrequently written configuration data and some frequently written log data. It would best be handled by separate partitions so that the important infrequently written configuration data is not corrupted due to the frequent log data.

First, start by defining a new space on the disk for the partition.

# The boot partition
define(BOOT_PART_OFFSET, 63)
define(BOOT_PART_COUNT, 16321)

# Let the rootfs have room to grow up to 128 MiB and align
# it to the nearest 1 MB boundary
define(ROOTFS_A_PART_OFFSET, 16384)
define(ROOTFS_A_PART_COUNT, 289044)
define(ROOTFS_B_PART_OFFSET, 305428)
define(ROOTFS_B_PART_COUNT, 289044)

# Config partition
define(CONFIG_PART_OFFSET, 594472)
define(CONFIG_PART_COUNT, 1048576)

# Log partition
define(LOG_PART_OFFSET, 1643048)
define(LOG_PART_COUNT, 1048576)

In this example, we are changing the app data partition to CONFIG_PART and adding LOG_PART.

Next, we change the mapping to include these two new partitions.

mbr mbr-a {
    partition 0 {
        block-offset = ${BOOT_PART_OFFSET}
        block-count = ${BOOT_PART_COUNT}
        type = 0xc # FAT32
        boot = true
    }
    partition 1 {
        block-offset = ${ROOTFS_A_PART_OFFSET}
        block-count = ${ROOTFS_A_PART_COUNT}
        type = 0x83 # Linux
    }
    partition 2 {
        block-offset = ${CONFIG_PART_OFFSET}
        block-count = ${CONFIG_PART_COUNT}
        type = 0xc # FAT32
    }
    partition 3 {
        block-offset = ${LOG_PART_OFFSET}
        block-count = ${LOG_PART_COUNT}
        type = 0x83 # Linux
    }
}

mbr mbr-b {
    partition 0 {
        block-offset = ${BOOT_PART_OFFSET}
        block-count = ${BOOT_PART_COUNT}
        type = 0xc # FAT32
        boot = true
    }
    partition 1 {
        block-offset = ${ROOTFS_B_PART_OFFSET}
        block-count = ${ROOTFS_B_PART_COUNT}
        type = 0x83 # Linux
    }
    partition 2 {
        block-offset = ${CONFIG_PART_OFFSET}
        block-count = ${CONFIG_PART_COUNT}
        type = 0xc # FAT32
    }
    partition 3 {
        block-offset = ${LOG_PART_OFFSET}
        block-count = ${LOG_PART_COUNT}
        type = 0x83 # Linux
    }
}

This layout defines our system as follows.

+----------------------------+
| MBR                        |
+----------------------------+
| Boot partition (FAT32)     |
+----------------------------+
| p1*: Rootfs A (squashfs)   |
+----------------------------+
| p1*: Rootfs B (squashfs)   |
+----------------------------+
| p2: Config      (FAT32)    |
| p3: Log         (EXT4)     |
+----------------------------+

Mounting the Partition

Mounting your new partition can be handled by either erlinit or by your Elixir application. To have erlinit mount the partition for you, you will need to supply your own erlinit.config file and edit the -m option

# Mount the configdata partition
# See http://www.linuxfromscratch.org/lfs/view/6.3/chapter08/fstab.html about
# ignoring warning the Linux kernel warning about using UTF8 with vfat.
-m /dev/mmcblk0p3:/root:vfat::;/dev/mmcblk0p4:/mnt/log:ext4::

The other option is to handle it in your Elixir code. This can be useful if you want to scan the disk for corruption and reformat / seed it. Erlinit can only attempt to mount the partition.

First we can initialize it

System.cmd("mke2fs", ["-t", "ext4", "-L", "LOGDATA", "/dev/mmcblk0p4"])

Then we can mount the partition

System.cmd("mount", ["-t", "ext4", "/dev/mmcblk0p4", "/mnt/log"])