9. Frequently Asked Questions

9.1. Why doesn’t the installed system use the whole partition?

The filesystem image installed via RAUC was probably created for a size smaller than the partition on the target device.

Especially in cases where the same bundle will be installed on devices which use different partition sizes, tar archives are preferable to filesystem images. When RAUC installs from a tar archive, it will first create a new filesystem on the target partition, allowing use of the full size.

9.2. Is it possible to use RAUC without D-Bus (Client/Server mode)?

Yes. If you compile RAUC using the -Dservice=false configure option, you will be able to compile RAUC without service mode and without D-Bus support:

meson setup -Dservice=false build

Then every call of the command line tool will be executed directly rather than being forwarded to the RAUC service process running on your machine.

9.3. Why does RAUC not have an ext2 / ext3 file type?

ext4 is the successor of ext3. There is no advantage in using ext3 over ext4.

Some people still tend to select ext2 when they want a file system without journaling. This is not necessary, as one can turn off journaling in ext4, either during creation:

mkfs.ext4 -O ^has_journal

or later with:

tune2fs -O ^has_journal

Note that even if there is only an ext4 slot type available, potentially each file system mountable as ext4 should work (with the filename suffix adapted).

9.4. Is the RAUC bundle format forwards/backwards compatible?

While RAUC now supports two bundle formats (verity and crypt) in addition to the original format (plain), all are still installable by default. Support for the old format can be disabled via the configuration. Going forward, any issue with installing bundles using old formats or features would be considered a bug, except after an explicit deprecation period of several years.

Newer RAUC versions have added features (such as casync), slot types (eMMC/MBR/GPT bootloader partitions) and bundle formats (verity and crypt). Only if you use those features by enabling them in the bundle manifest, older versions of RAUC that cannot handle them will refuse to install the bundle. As long as you don’t enable new features, our intention is that bundles created by newer versions will be installable by older versions and any such issues would be considered a bug.

Some background is described in the Forward and Backward Compatibility section.

If there are ever reasons that require an incompatible change, you can use a two step migration using an intermediate version.

9.5. Can I use RAUC with a dm-verity-protected partition?

Yes you can, as the offline-generated dm-verity hash tree is simply part of the image that RAUC writes to the partition. To ensure RAUC does not corrupt the partition by executing hooks or writing slot status information, use type=raw in the respective slot config and use a global (see slot status file) on a separate non-redundant partition with setting statusfile=</path/to/global.status>.

9.6. Can I use RAUC with a dm-crypt-protected partition?

Yes you can, by using the /dev/mapper/<devicename> as the device for the slot (with the type of the filesystem of your choice). This way, RAUC interacts only with the unencrypted device/content.

For example, with an encrypted root filesystem slot (perhaps unlocked by an initramfs loaded from a different partition):


Remember to unlock the inactive slots as well so that RAUC can write to them.

9.7. What causes a payload size that is not a multiple of 4kiB?

RAUC versions up to 1.4 had an issue in the casync bundle signature generation, which caused two signatures to be appended. While the squashfs payload size is a multiple of 4kiB, the end of the first signature was not aligned. As RAUC uses the second (“outer”) signature during verification, this didn’t cause problems. RAUC 1.5 fixed the casync bundle generation and added stricter checks, which rejected the older bundles. In RAUC 1.5.1, this was reduced to a notification message.

To avoid the message, you can recreate the bundle with RAUC 1.5 and newer.

9.8. How can I refer to devices if the numbering is not fixed?

There are many reasons why device numbering might change from one kernel version to the next, across boots or even between hardware variants. In the context of RAUC, this is mainly relevant for block, MTD and UBI devices.

In almost all cases, the proper way to configure this is to use udev rules.

For block devices, udev ships with rules which create symlinks in /dev/disk/by-path/. These are not affected by changes in the probe order or by other devices that are not always connected. For example, on an emulated ARM machine, this results in:

root@qemuarm:~# ls -l /dev/disk/by-path
lrwxrwxrwx    1 root     root             9 Nov 18 12:46 platform-a003c00.virtio_mmio -> ../../vda

By using /dev/disk/by-path/platform-a003c00.virtio_mmio in your configuration, you ensure that you always refer to the same block device.

For UBI volumes, no equivalent rules are currently shipped by udev, so custom rules can be used. Depending on how the symlinks should be named, different rules could be used:

# Use the volume name instead of the number
SUBSYSTEM=="ubi", KERNEL=="ubi*_*", ATTRS{mtd_num}=="*", SYMLINK+="$parent_%s{name}"
# Use the MTD device number instead of the UBI device number
SUBSYSTEM=="ubi", KERNEL=="ubi*_*", ATTRS{mtd_num}=="*", SYMLINK+="ubi_mtd%s{mtd_num}_%s{name}"
# Use the MTD device name instead of the UBI device number
SUBSYSTEM=="ubi", KERNEL=="ubi*_*", ATTRS{mtd_num}=="*", IMPORT{program}="/bin/sh -ec 'echo MTD_NAME=$(cat /sys/class/mtd/mtd%s{mtd_num}/name)'" SYMLINK+="ubi_%E{MTD_NAME}_%s{name}"

When enabling all of these rules (which you should not do), you will get something like:

crw------- 1 root root 249,  0 Nov 18 13:46 /dev/ubi0
crw------- 1 root root 249,  1 Nov 18 13:46 /dev/ubi0_0
lrwxrwxrwx 1 root root       6 Nov 18 13:46 /dev/ubi0_rauc-test -> ubi0_0
lrwxrwxrwx 1 root root       6 Nov 18 13:46 /dev/ubi_nandsim_rauc-test -> ubi0_0
crw------- 1 root root  10, 59 Nov 18 13:46 /dev/ubi_ctrl
lrwxrwxrwx 1 root root       6 Nov 18 13:46 /dev/ubi_mtd3_rauc-test -> ubi0_0

Custom udev rules can also be very useful when you want to refer to the active data partition (in a scenario with redundant data partitions) with a fixed name.

9.9. Why does the installation fail with a resize2fs error?

When installing a ext4 image with resize=true configured for that slot, it may happen that the ext4 image has features enabled which are not supported by the currently used version of resize2fs.

For example, e2fstools 1.47 enabled the orphan_file feature by default and is included in Yocto mickledore. When an image generated by Yocto mickledore is installed on an older release (perhaps built using kirkstone or older), that version of resize2fs will refuse to modify the filesystem and the installation will be aborted:

LastError: Installation error: Failed updating slot rootfs.1: Failed to run resize2fs: Child process exited with code 1

In the log output from RAUC, you’ll find more details:

rauc[409]: opening slot device /dev/mmcblk0p2
rauc[409]: writing data to device /dev/mmcblk0p2
rauc[409]: Resizing /dev/mmcblk0p2
rauc[551]: resize2fs 1.46.5 (30-Dec-2021)
rauc[551]: resize2fs: Filesystem has unsupported feature(s) (/dev/mmcblk0p2)
rauc[409]: Installation error: Failed updating slot rootfs.1: Failed to run resize2fs: Child process exited with code 1

The solution for this is to disable the unsupported filesystem features during the image generation. When using Yocto, in case of the orphan_file file feature, you could use:

EXTRA_IMAGECMD:ext4:append = " -O ^orphan_file"

in the image recipe or an appropriate conf file to disable the feature until all systems have been updated with versions of resize2fs which support this feature.