// the take
kdbus was an attempt to move D-Bus message passing into the Linux kernel instead of running it through a userspace daemon, pairing with systemd's sd-bus on the userspace side. It's aimed at kernel developers and distro maintainers who wanted lower-latency, zero-copy IPC for init systems, not application developers looking for a D-Bus library.
The design includes real primitives for a production IPC layer: a pool-based memory model for zero-copy message transfer, a namespace/domain concept for container isolation, and a policy system for connection-level access control. The test suite (test-benchmark, test-policy-ns, test-metadata-ns) shows the authors were actually exercising multi-namespace and security-boundary behavior, not just happy-path IPC.
This never got merged into mainline Linux — Linus and other kernel maintainers rejected moving D-Bus semantics into kernel space, and the last commit is from 2017, so there is no path to using this on any supported kernel today. Being out-of-tree kernel code, it demands matching kernel headers and ABI versions exactly, and any staleness means it won't build against modern kernels at all. systemd itself dropped kdbus support years ago in favor of a pure userspace kdbus-free sd-bus, so even the one consumer this was built for no longer depends on it. Kernel-resident IPC also means bugs here are kernel panics or local privilege escalations, not recoverable userspace crashes — a much higher severity class for an idea that didn't ship.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →