mergerfs v2.34.1

Changelog

• Fix regression testing for implemented functions by @trapexit in #1109

Full Changelog: 2.34.0...2.34.1

Documentation

https://github.com/trapexit/mergerfs/blob/2.34.1/README.md

Donations

https://github.com/trapexit/support

mergerfs v2.34.0

What's Changed

• Rework support section in readme by @trapexit in #1032
• Add details about usage of FUSE to docs by @trapexit in #1039
• Fix typos by @Gelma in #1045
• Error when given invalid policy names by @trapexit in #1057
• Support doc update by @trapexit in #1058
• Fix setting of stat vars for symlinkify by @trapexit in #1085
• Update fuse_kernel.h by @trapexit in #1087
• Tweaks for 32bit systems by @trapexit in #1089
• Remove unnecessary libfuse abstractions by @trapexit in #1091
• Remove write_buf, simplify FUSE msg dispatching by @trapexit in #1092
• Add CodeQL workflow for GitHub code scanning by @lgtm-com in #1093
• Remove libfuse abstraction in prep for adding request data by @trapexit in #1104
• Misc cleanup by @trapexit in #1105
• Fix reading of setxattr name by @trapexit in #1106

Full Changelog: 2.33.5...2.34.0

Support

Consider supporting mergerfs' development: https://github.com/trapexit/support

mergerfs-2.33.5

What's Changed

• Return ENOENT when dotdot for root node requested by @trapexit in #1031

Support

If you like mergerfs please consider supporting the project: https://github.com/sponsors/trapexit

mergerfs-2.33.4

What's Changed

• Update bug_report.md by @trapexit in #1007
• Rework node slab garbage collection to limit blocking work threads by @trapexit in #1020
• Fix query of attr during symlink by @trapexit in #1023
• Change ENOENT to ESTALE for looking up paths to handle rename race conditions by @trapexit in #1025
• Lock less often during logging by @trapexit in #1026

Full Changelog: 2.33.3...2.33.4

mergerfs-2.33.3

What's Changed

• Call FUSE getattr rather than syscall for extra logic

Full Changelog: 2.33.2...2.33.3

mergerfs-2.33.2

What's Changed

• Get attrs for link, not target

Full Changelog: 2.33.1...2.33.2

mergerfs-2.33.1

Fixes a regression with file locks not being properly setup.

Commits:

• ee8532c Ensure server handling of locks is disabled

mergerfs-2.33.0

Sorry for taking so long to get this version out. Been working on 3DO related projects and busy with work. I had hoped with COVID lockdowns I'd have spent more time working on mergerfs but the opposite occurred for a number of reasons. Regardless, now that this release is out, besides bug fixes, I intend to make the trek towards version 3.0 which will have major changes.

There was a lot of under-the-hood changes to sensitive code. I've stress tested it quite a bit but I'd not be surprised if I missed something. If anything seems odd please file a ticket or reach out directly. Enjoy.

Summary:

Improvements

• Major cleanup of libfuse.
• Reduced memory overhead of file metadata held in memory significantly.
• Reduce memory fragmentation by reusing certain objects via mmap'ed memory pools.
• Removed ioctl based runtime control. Wasn't used by anyone and will be completely changed in v3.

Bug Fixes

• Fix bug in mfs policy when used in non-create situations (rare)
• Fix bug in pfrd policy

Features

• follow-symlinks: allows a number of modes where mergerfs will resolve symlinks rather than return them as symlinks.
• rename-exdev, link-exdev: configurable ways to handle when rename or link return exdev. Instead of failing with EXDEV it can instead create a symlink to the file.

Changes

• Removed most of the debug logging. The intent is to change how the logging is done eventually. I just didn't get around to it for this release. strace is typically more useful anyway. If you run with -d it will spit out INIT flag details though. It will also write certain memory details to /tmp/mergerfs.<pid>.info

Commits:

• 68b3026 Remove embedded name in node struct
• 18dead4 Add new debug printing routines
• 68719da Remove ioctl runtime section, use getfattr instead of xattr
• c384509 Remove subscribestar from sponsor platforms
• 5f737cb Add option to log node memory usage metrics
• 930dad3 Reduce struct node size
• 8150957 Fix regression from remember_node refactor
• c929781 Add malloc_trim configure test
• 313aa2a Update ghc::filesystem and nonstd::optional
• 4ea0de3 Rework dirents buffer management
• 6b5c484 Major rework of memory allocation using fixed mem pools
• f598d3b Add details on dropping caches before benchmarks
• 122f9c1 Add new cirrus-ci targets
• 5263a65 Remove usage printing regression in mount_bsd
• 8def16a Fix infinite loop in mfs action policy
• 43a6d66 Major cleanup of libfuse to remove unneeded features
• df721eb Update docs with Chia wallet address
• 54e8500 Add more clarity in what to provide when submitting bug reports or questions
• 1aaf742 Update 'features' section of readme
• b4c031d tweak bug report template language
• 61c2187 Remove config ioctl calls
• 7013ff9 Update nonstd::optional to v3.4.0
• 3ce0dc5 Update ghc::filesystem to v1.5.8
• 098f353 pfrd: fix mod by zero error when all branches are filtered
• c06db9c pfrd: fix mod by zero error when all branches are filtered
• 80f3099 properly initialize fuse_buf in worker loop
• bcd5fde properly initialize fuse_buf in worker loop
• c40f42e mention bind-propagation and add link to hotio
• 44efbaa Update README.md
• 9573a05 Update bug_report.md
• 495afef README: update strace args and add question on snapraid
• 58b4eb9 Fix typo
• 9a90609 remove travis-ci
• 8adebc9 new features: follow-symlinks, rename-exdev, link-exdev
• ed0c1db fix parsing of relative branch paths
• d337574 fix parsing of relative branch paths
• 6a9f7e6 readme: update support section
• 3bc189c README: update support section
• dfb544f README: add warning/clearity about caching script and hardlinks and cow
• 538467b config: rework global config, remove rwlock, make branches RCU like
• 3808e26 Update FUNDING.yml
• e9e17ba README: misc updates and tweaks

mergerfs-2.32.6

8e47bf4 pfrd: fix mod by zero error when all branches are filtered

mergerfs-2.32.5

Fix an uninitialized variable that could cause invalid behavior when reading from the FUSE device. Only likely to be hit if using splice_read option.

• bcd5fde: properly initialize fuse_buf in worker loop