Ext4Kit 0.1.1

0.1.1 — 2026-06-14

Fixes a bug where freshly formatted volumes mounted read-only.

• Formatted volumes are writable again. The checksum-seed safety check
  now only forces read-only when a volume actually carries metadata
  checksums (metadata_csum). lwext4's mkfs sets the csum_seed flag
  without metadata_csum, which tripped the old check — so every
  newfs_fskit-formatted volume (and the benchmark images) wrongly mounted
  read-only. Volumes whose UUID was changed after format are still protected.
• Benchmark images now get a real UUID, and the test harness gained a
  --csum check that guards both directions of this policy.

Ext4Kit 0.1.0

Changelog

0.1.0 — 2026-06-14

First public beta. Mount Linux ext4 drives on macOS & read & write them like
any other volume — no kernel extensions, built on Apple's FSKit. Runs on
macOS 15.4+; you don't need a developer account to use it.

Beta — back up your data first. Test on a scratch volume before you
trust Ext4Kit with anything important. The read/write path is verified
against live mounts. Formatting (newfs_fskit), checking (fsck_fskit), &
read-only mounts pass the build & test harness but haven't been run against
a live mount yet.

What you can do

• Mount & unmount ext4 volumes: mount -F -t ext4 diskN /Volumes/ext4.
• Read files, directories, & symlinks, with real permissions & timestamps.
• Write for real: create, delete, rename (including renaming over an
  existing file), hard-link, symlink, & edit files.
• Change metadata with chmod, chown, touch, & truncate.
• Set extended attributes — macOS xattrs round-trip through Linux's
  user. namespace.
• Rename a volume with diskutil rename.
• Format & check: newfs_fskit -t ext4 -L LABEL makes a fresh volume;
  fsck_fskit -t ext4 runs a read-only structural check.

Safe by default

• Journaled writes. Ext4Kit replays the ext4 journal on mount & refuses
  to mount writable if replay fails, so an interrupted write can't silently
  corrupt the volume.
• Linux stays happy. Metadata checksums stay up to date — e2fsck comes
  back clean after a macOS write session.
• Read-only when it should be. mount -o ro, write-protected media, &
  volumes with an unusual checksum seed all mount read-only automatically.
• Open files survive deletion. Deleting an open file keeps it readable
  until the last handle closes, the way POSIX expects.

Faster

• Sequential writes run about 3× faster, with far fewer device writes.
• Listing a 20,000-entry directory pages instantly instead of re-scanning
  from the start (about 150× faster).
• Repeated stat of deep paths hits the cache & skips the disk entirely.

Limitations

• inline_data volumes don't mount — format with mkfs.ext4 -O ^inline_data.
• No sparse files: writing past the end of a file fills the gap with zeros.
• Extended-attribute values are capped at one filesystem block (about 4 KiB).
• Filenames that aren't valid UTF-8 stay hidden & can't be deleted on macOS.
• File data flows through the extension process — no kernel-offloaded I/O yet.

See the README for the full list.

Tooling

• Benchmarks/ — an lwext4 test harness with data-integrity checks,
  corrupt-image fuzzing, randomized soak runs, & crash/journal-replay tests.
• CI builds Debug & Release & runs the full harness on every push.
• Scripts/release.sh — archive, Developer ID sign, notarize, & staple.

License

Ext4Kit's own code is MIT. The shipped app statically links lwext4, so the
binary is a combined work under GPL-2.0-or-later — the complete source is
this tag plus the pinned Vendor/lwext4 submodule. See
THIRD_PARTY_LICENSES.md.