The kernel is a kernel.org release with Debian's configuration of that branch and the synodeb drivers for the supported models. One kernel package serves every model.
Until November 2026 the packages were named after the first supported model:
linux-image-ds923* and synology-ds923, kernels <version>-ds923. Those names are now
transitional packages: sudo apt full-upgrade moves an existing system to the synodeb names
by itself, and the transitional packages can be removed afterwards.
The platform driver was renamed at the same time, from synology_ds923 to synology_nas:
kernel parameters in /etc/default/synodeb (e.g. synology_nas.spinup_delay=8) take the new
prefix.
There are two, named as on kernel.org:
| metapackage | series | |
|---|---|---|
linux-image-synodeb-longterm |
longterm (6.18.y), projected end of life December 2028 | default, also installed as linux-image-synodeb |
linux-image-synodeb-stable |
the newest stable branch (7.3.y, then 7.4.y…) | newer features, each branch maintained for a few months only |
To follow the stable series: sudo apt install linux-image-synodeb-stable (plus
linux-headers-synodeb-stable if you build modules). With both installed, the DOM boots the
newest stable kernel; remove linux-image-synodeb-stable to go back to longterm.
New releases of both series are built and published automatically, usually within a day of
their release on kernel.org. Install them with sudo apt update && sudo apt full-upgrade, then
reboot. A kernel update is never installed automatically, because it writes the DOM.
The DOM keeps two kernels: if the new one does not boot, the next boot starts the previous one and you get a notification. A normal boot never writes the DOM. A kernel update writes about 25 MB to the DOM; Synology does not publish its endurance, but even MLC flash should last thousands of updates.
The kernels on the DOM come from /boot of the running root. A btrfs snapshot taken before a
kernel update holds the older kernels and their /lib/modules only: booting it with the newer
kernel on the DOM leaves the kernel without its modules (no network, no disk or GPU drivers).
Before rebooting into a snapshot, mount it and copy its kernels to the DOM:
sudo synodeb-dom sync --root /mnt/snapshotIt copies the two kernels that root had chosen (its /boot/vmlinuz and vmlinuz.old links),
only if their modules are in it, makes DEBIAN the default boot entry and arms a kernel trial
when DEBIAN_OLD is there. The DOM is not touched when the snapshot lacks the modules of its
kernel, e.g. with /boot on a separate partition. To boot, the snapshot must then take the
place of the subvolume the boot entries name (rootflags=subvol=@rootfs, see
sudo synodeb-dom status).
For DKMS modules install linux-headers-synodeb (or linux-headers-synodeb-stable): it follows the
kernel of its series. Module signatures are checked but not enforced, so locally built modules
load. ZFS (zfs-dkms) works with the longterm series only, until OpenZFS supports the newest
stable branch.
RAID arrays created by a 7.x kernel record their logical block size in the md superblock, and a stock 6.18 kernel refuses to assemble them. The synodeb longterm kernel carries a small patch that reads and keeps that field, so these arrays work with both series. Moving such disks to another machine needs a kernel newer than 6.18 there.