MeshXFS: A clustered filesystem built on XFS, written by AI

MeshXFS is a POSIX clustered filesystem built on XFS that lets multiple nodes mount the same shared block device read-write. It supports shared iSCSI SANs and dual-primary DRBD, using either a network lock manager or on-disk SCSI compare-and-write (CAW) locking, and implements fencing and survivor-led journal recovery to protect consistency. The clustering, distributed-locking, fencing and on-disk envelope logic was produced by AI agents that started from upstream XFS and substantially modified it. MeshXFS ships as a kernel module (mxfs.ko) with userland tools (mkfs.mxfs, chk_mxfs/fsck.mxfs, resize_mxfs, mxfs_admin), DKMS packages for Proxmox VE 9, RHEL 9, Ubuntu 24.04/26.04 and Debian 12/13, plus a Proxmox storage plugin.

Release 0.90.130 focuses on stronger recovery and broader platform support: Ubuntu 26.04 and Debian 12 builds, fixes for pre-6.5/6.6 kernel write/freezing edge cases, and a network lock manager that operates correctly over multipath at 2, 4, 8 and 16 nodes (and single-path at larger node counts). A node that loses all storage paths is fenced and its journal is replayed while remaining nodes continue; the others now refuse any late lock releases. Validation includes many configurations (2-16 nodes, direct and multipath iSCSI, on-disk CAW, DRBD dual-primary) run against a 31-row test suite plus path-fault scenarios on an Ubuntu 24.04 test rig. The prior 0.90.129 work added robust Proxmox VE shared-SAN operation, quicker DRBD whole-disk setup, QDevice-assisted DRBD recovery and multiple locking and recovery fixes.