EROFS With Linux 7.2 Better Handles Large Sparse AI Datasets, More Efficient I/O

Linux 7.2 brings two notable improvements to the EROFS read-only file system. First, chunk-based inode mapping has been optimized for more efficient I/O via an updated erofs_map_chunks() implementation. Second, sparse file support has been added to the pcluster layout, specifically to better handle large sparse AI datasets — without this, overlayfs would allocate excess space on copy-up and SEEK_HOLE would fail to report holes. Additionally, the EROFS FSCACHE back-end, previously deprecated due to NETFS becoming a hard dependency, has been fully removed in Linux 7.2, with equivalent functionality now covered by file-backed mounts and fanotify pre-content hooks.