You are browsing documentation for a version other than the latest stable release. Switch to the latest stable release, 1.2.
In order to ensure that the necessary components for Mender have been integrated correctly, use this checklist to verify each of them in turn. Use this when all components have been built successfully, and the device has booted correctly.
This checklist will verify some key functionality aspects of the Mender integration. It will verify that:
As part of the test, we will need two different Linux kernels, in order to verify that both are booted correctly when they should. Therefore, before building the images you will test with, run the commands:
bitbake -c clean linux-yocto bitbake -c cleansstate linux-yocto
linux-yocto may need to be switched with the particular kernel being used with your device.
Then build the image normally.
Copy the resulting
.sdimg file to a safe location, for example
base-image.sdimg and then repeat step 1.
Run the following commands:
sudo losetup /dev/loop0 base-image.sdimg sudo kpartx -a /dev/loop0 sudo dd if=tmp/deploy/images/<machine>/<image>.ext4 of=/dev/mapper/loop0p3 sync sudo kpartx -d /dev/loop0 sudo losetup -d /dev/loop0
The last number in
loop0p3 should correspond to the partition number of the second rootfs partition, and may need to be adjusted if
MENDER_ROOTFS_PART_B has been changed.
kpartx tool may require installation before it can be used.
Some Linux distributions may try to auto-mount the devices that
kpartx maps. If so you need to unmount them again manually before calling
The reason for doing this manual image patching is to have two kernels in the image with differing build dates, which will be useful for verification later.
base-image.sdimg image to the device and boot it.
Verify that the two commands
fw_setenv are in the path and are executable. Calling them with no arguments will should give a variable listing and an error about missing variable name, respectively. This verifies that u-boot-fw-utils tools are present on the device.
Now we will test that storing and reading values in the U-Boot environment works. Run:
fw_setenv mender_linux_test 1
Reboot, and interrupt U-Boot so you get the U-Boot prompt.
The output should be:
in other words the same value that was set earlier.
While still at the U-Boot prompt, run:
setenv mender_uboot_test 1 saveenv run bootcmd
You should be back inside Linux now, so run:
The output should be:
in other words the same value that was set earlier. This verifies that U-Boot the bootloader, and u-boot-fw-utils agree on format for the U-Boot environment.
Now we will verify that Mender is running. Run the following:
If Mender has been enabled as a daemon, either through inheriting
mender-systemd, it should return a PID. If not, it should return nothing. This verifies that Mender is started as a service if applicable.
Now we verify that the booted kernel is from the rootfs. The easiest way to check this is to check the build date, which can be seen by running:
Make sure it matches the build date and time of the first kernel you just built, and not an older one. Make a note of this time, as you will need it later. This verifies that the correct kernel is loaded from partition A.
Now we will look at which rootfs is mounted. Run
mount with no arguments. The file system mounted as root (signified by the
<device> on / entry) should be
If the device listed is an ambiguous device, such as
/dev/root, you can use an alternative method for verifying it. If you call the following series of commands:
stat -c %D / stat -c %t%02T /dev/mmcblk0p2
The output of the two commands should be identical. This verifies that the correct rootfs is mounted when partition A is active.
If you have selected a different device using either
MENDER_STORAGE_DEVICE_BASE in the Yocto configuration, the
/dev/mmcblk0p2 entry may be different, but it should always correspond to the value in
Everything we have tested so far has been for partition A; we will now verify both kernel and rootfs for partition B. Run the following:
fw_setenv mender_boot_part 3
The number is the number of the second rootfs partition, and corresponds to the last component of the
MENDER_ROOTFS_PART_B variable. If you've changed this variable in the Yocto configuration, you may need to use a different number.
Repeat step 12, but this time verify that the build date is later than the first time you did this step. This is because we replaced the kernel of the second rootfs partition with a newer one. This verifies that the correct kernel is loaded from partition B.
Repeat step 13, but this time using
/dev/mmcblk0p3. This verifies that the correct rootfs is mounted when partition B is active.
As in the previous step,
/dev/mmcblk0p3 may be different if
MENDER_STORAGE_DEVICE_BASE have been changed.
Run the following commands:
fw_setenv mender_boot_part 2 fw_setenv upgrade_available 1 fw_setenv bootcount 0
mender_boot_part number is the number of the first rootfs partition, and corresponds to the last component of the
MENDER_ROOTFS_PART_A variable. If you've changed this variable in the Yocto configuration, you may need to use a different number.
Reboot, but pull the power plug before the system has had time to finish booting. Mender will auto-commit the update if it is enabled as a service, which will defeat the purpose of this test, so it's important that the power is cut after the kernel has started booting, but before Mender has started.
/dev/mmcblk0p3. The detected device should not be
/dev/mmcblk0p2, this indicates that the rollback has not worked. Otherwise this verifies that rollback, indeed, has worked.
That's it! You have now verified compatibility with Mender!