systemd 260: The End of System V and a New Era for Linux Init

From First Candidate to Stable

System V Compatibility: Fully and Finally Gone

The most disruptive change in systemd 260 is the permanent, unconditional removal of System V init script support. After more than a decade of carrying the compatibility layer as a courtesy, the following components are deleted entirely:

• REMOVEDsystemd-sysv-generator — the tool that auto-translated /etc/init.d scripts into transient unit files at boot.
• REMOVEDsystemd-rc-local-generator and rc-local.service — the legacy rc.local execution path.
• REMOVEDsystemd-sysv-install — the hook used by systemctl enable/disable/is-enabled for SysV scripts.

Any service that has not been migrated to a native .service unit file will silently fail or refuse to start after upgrading. Administrators running third-party software — older database installers, legacy monitoring agents, or enterprise middleware — should audit all running services with systemctl list-units --type=service before upgrading.

Kernel & Dependency Floor Raised

The official release notes confirm the previous kernel baseline was Linux 5.4; it has been raised to 5.10, with 5.14 or the 6.6 LTS series recommended for full functionality. Code supporting kernels older than 5.10 has been removed entirely.

Beyond the kernel, several core library dependencies have been bumped:

• RAISEDglibc 2.31 → 2.34
• RAISEDOpenSSL 2.x → 3.0
• RAISEDPython minimum → 3.9
• RAISEDlibseccomp 2.3.1 → 2.4.0
• REMOVEDlibidn support dropped — IDN now requires libidn2

What systemd 260 Adds

• NEW mstack & systemd-mstack — A new mechanism for composing OverlayFS and bind-mount setups using a structured .mstack/ directory layout. A companion CLI tool, systemd-mstack, allows interactive management. Particularly useful for containerized and sandboxed service deployments.
• NEW CPUSchedulingPolicy=ext — Services can now opt in to the SCHED_EXT scheduler, enabling extensible, BPF-programmable CPU scheduling policies per service unit.
• NEW MemoryTHP= — A new unit setting provides per-service control over Transparent Huge Pages, useful for fine-tuning memory performance in high-throughput workloads.
• NEW FANCY_NAME= field — Added to os-release, similar to PRETTY_NAME but explicitly allowing ANSI sequences and non-ASCII Unicode glyphs. Shown by systemd-hostnamed and hostnamectl.
• NEW tpm2_id udev built-in — Extracts vendor and model identification from connected TPM2 devices as they are probed, simplifying TPM management.
• NEW systemctl enqueue-marked — A new verb that calls the EnqueueMarkedJobs() D-Bus method directly from the command line.
• NEW xaccess device delegation — systemd-logind and systemd-udevd gain the xaccess mechanism for delegating access to specific devices to users with specially marked sessions.
• NEW systemd-vmspawn –ephemeral — Virtual machines can now be marked ephemeral, created on demand and discarded after use — ideal for testing and automation pipelines.
• NEW AI tooling documentation — The project now ships AGENTS.md and CLAUDE.md files, providing AI coding assistants with structured guidance on the codebase’s conventions and development workflow.

systemd-networkd Improvements

Several confirmed improvements land in the networking stack:

• NEWModemManager integration via the “simple connect” protocol, enabling easier mobile broadband management through systemd-networkd.
• IMPROVEDThe Varlink and JSON interfaces now report IP addresses as human-readable strings in addition to the existing integer array format — making monitoring and automation tooling more ergonomic.
• NEW.link files gain additional parameters: ScatterGather=, TCPECNSegmentationOffload=, GenericReceiveOffloadList=, and more, for fine-grained Ethernet device tuning.

Integrity & Access Control

• NEWsystemd-repart adds basic integrity checks for encrypted volumes.
• NEWThe xaccess mechanism in systemd-logind and udev enables more granular, session-scoped device access delegation.
• IMPROVEDsystemd-portabled now runs as a user-level service, allowing unprivileged users to start portable services on supported kernel versions.

When Will It Reach Your System?

Rolling-release distributions — Arch Linux, Fedora, openSUSE Tumbleweed — are expected to absorb systemd 260 quickly, as their users typically track current kernels already. The release is intended for first-half 2026 Linux distributions.

Long-term support enterprise distributions and LTS Ubuntu builds will likely delay, as the kernel 5.10 requirement itself may be a significant ask for some server fleets. RHEL derivatives are expected to take additional time before shipping this update in their stable channels.

The most critical action item for all administrators: audit and migrate any remaining System V init scripts before upgrading. There is no automatic fallback in systemd 260.