OpenVMX

OpenVMS-compatible environment for Linux. OpenVMX implements VMS system services, the DCL command language, Record Management Services, and VMS-style kernel extensions — packaged as a bootable Linux distribution.

OpenVMX is two named layers, GNU/Linux-style: OpenVMX is the VMS-compatible product (login banner, SHOW SYSTEM, DCL — what a user touches), running on OVMX/Linux, the Linux base layer underneath (kernel, boot, distro tooling) — the rough equivalent of the VAX/Alpha hardware OpenVMS itself ran on. See Architecture for where the split falls.

Features

• System Services — sys$assign, sys$qio, sys$crmpsc, sys$enq, sys$setimr, sys$creprc, sys$asctim, sys$gettim, mailboxes, event flags, ASTs, logical names — kernel-resident in the executive (vms.ko), reached through /dev/vms
• Runtime Library — lib$, str$, mth$, ots$ routines for memory, strings, math, signals, date/time
• DCL Shell — the DCL command language: 50+ built-in verbs, symbol management, lexical functions, script execution, VMS-style file specifications (see the command reference)
• Record Management Services — Sequential, relative, and indexed file organizations with the FAB/RAB/NAM/XAB interface; file-share and record locking arbitrated by the distributed lock manager
• Files-11 ODS-2 — Genuine ODS-2 volumes read and written over the executive ACP — INITIALIZE writes a real volume, MOUNT parses the home block/SCB, and RMS resolves through FID → file header → extents — not a host-filesystem passthrough
• Authentication — Binary SYSUAF with the Purdy password hash; DCL login (vms_login) and OpenSSH-based multi-user access
• Cluster participation — A real distributed lock manager (6-mode $ENQ/$DEQ, blocking ASTs, deadlock detection, lock value blocks) running cross-node over the SCS wire; cluster membership resident in the executive, so SHOW CLUSTER and $GETSYI read the real member block; RMS file/record locking behind the DLM
• TCP/IP Services — VMS-faithful IP layered product: the TCPIP$CONFIG configuration plane, TCPIP$ logicals, BGn: socket transport, PING, and an inetd-style auxiliary server
• Logical Name Manager — Hierarchical name tables; LNM$SYSTEM is executive-resident and visible node-wide
• Kernel Module — vms.ko (the VMS executive: access control, ASTs, event flags, the distributed lock manager, mailboxes, and the Files-11 ODS-2 ACP)
• Freestanding Syscall Layer — libvmssys runs without glibc (x86_64, Alpha, VAX, aarch64)
• Bootable Distro — Static musl binaries, initramfs, QEMU boot

Quick Start

Bootable QEMU VM — the OpenVMX runtime

This is OpenVMX proper: it boots OVMX/Linux (kernel modules, init), which hands off to the VMS-native toolchain (the IMGACT.EXE image activator and LINK.EXE, on a musl userland).

# Build the kernel + initramfs
docker build -f distro/Dockerfile.bootable -o dist .

# Boot it
./distro/boot/run-qemu.sh dist/vmlinuz dist/initramfs-ovmx.cpio.gz

Local Build — libraries, DCL, tests

# Prerequisites (Debian/Ubuntu)
sudo apt install cmake gcc make libc6-dev libreadline-dev flex bison

# Build
cmake -B build -DCMAKE_BUILD_TYPE=Debug -DBUILD_TESTS=ON -DBUILD_TOOLS=ON
cmake --build build -j$(nproc)

# Run tests
cd build && ctest --output-on-failure

Documentation

Start at the Documentation Index for the full map. Key entries:

• Architecture — component layers, boot sequence, data flow
• Building — all build modes, CMake options, kernel modules
• Install Guide — PRODUCT INSTALL a kit onto a target volume (steps checked in CI against the real e2e gate)
• Upgrade Guide — upgrade an installed system in place, preserving site config and user data (steps checked in CI against the real e2e gate)
• DCL Command Reference — the built-in verb set
• System Services API — the sys$ reference
• Compatibility Surface — per-surface status register
• Release Roadmap to 1.0 — milestone ladder and roadmap of record

Project Status

OpenVMX ships on a milestone ladder toward 1.0. The current release is V0.6 (cluster correctness): a genuine VMScluster participant — the distributed lock manager runs the full ladder over the SCS wire, RMS file/record locking reaches the real arbitrator, and cluster membership lives in the executive. Earlier milestones delivered the executive / DCL / RMS / kernel foundation (0.3), faithful install-and-boot (0.4), and the authenticity flip to genuine Files-11 ODS-2 over the executive ACP with binary SYSUAF/Purdy login (0.5). See the release roadmap to 1.0 for the full milestone ladder and current state.

First-class release architectures are x86_64, Alpha (LP64), and VAX (on the NetBSD substrate); aarch64 is also supported. Every release is co-released across these architectures through a shared build/release gate.

OpenVMX runs as a bootable QEMU VM — the VMS-native runtime (image activation via IMGACT.EXE, linking via LINK.EXE, musl userland) — where vms.ko provides the VMS executive and userspace reaches it through /dev/vms. The QEMU / real-kernel path is the only OpenVMX runtime; there is no Docker-based way to run OpenVMX (Rule 9).

License

GPL-2.0 — see LICENSE.