Skip to contentSkip to conversation

Btrfs progs release 6.3

DISCUSSION David Sterba 3 participants 3 messages Apr 27, 2023 Btrfs file system

lore
Share & tools
lore

David Sterba

author maintainer
#0 lore raw

Hi,

btrfs-progs version 6.3 have been released.

There are some fixes and tweaks, the experimental status of block-group-tree has been removed, namely the conversion but please still use it with care.

Otherwise there's nothing interesting from users' POV, more documentation is being moved from archived wiki, updated tests.

The CI on github has been extended to cover devel and pre-release branches, further enhancements are planned to do e.g. quick build tests for pull requests or use self-hosted runners for heavier tests that may need to run against a recent kernel. (https://github.com/kdave/btrfs-progs/actions)

Release process now uses more of the github features, so besides the tag itself there's also a release with changelog (https://github.com/kdave/btrfs-progs/releases).
This allows to add "artifacts" i.e. other files besides the sources.

There are now static binaries published along with the release. They're built inside the CI workflow "Static binaries" (also available there for download). The upload is still done from an intermediate host so if you don't trust github CI or me copying the files as-is, please build and use your own.

Please let me know if you have suggestions for improvements or if you find problems with the CI/release things. It's new to me and I'm sure the are already best practices to follow.

Changelog:
* mkfs: option -R deprecated, options unified in -O (-R still works)
* mkfs: fix potential race with udev leading to EBUSY due to repeatedly opened file descriptors
* block-group-tree is out of experimental mode
* available as 'mkfs.btrfs -O block-group-tree'
* btrfstune can do in-place conversion to/from (use with care)
* balance: fix recognizing old and new syntax
* subvol snapshot: specific error if a failure is caused by an active swapfile
* tree-stats: rephrase warning when run on a mounted filesystem
* completion: 'filesystem du' also completes files
* check: fix docs, help text and warning that --force + --repair works on a mounted filesystem
* build: fix static build when static libudev is available
* documentation:
* more updates from wiki, developer docs, changelogs
* reformatting
* updates and fixes
* other:
* test updates and fixes
* CI cleanups and old files removed
* integration with Github actions

Tarballs: https://www.kernel.org/pub/linux/kernel/people/kdave/btrfs-progs/
Git: git://git.kernel.org/pub/scm/linux/kernel/git/kdave/btrfs-progs.git

#1 lore raw
Quoted 52 lines · David Sterba

Hi,

btrfs-progs version 6.3 have been released.

There are some fixes and tweaks, the experimental status of block-group-tree has been removed, namely the conversion but please still use it with care.

Otherwise there's nothing interesting from users' POV, more documentation is being moved from archived wiki, updated tests.

The CI on github has been extended to cover devel and pre-release branches, further enhancements are planned to do e.g. quick build tests for pull requests or use self-hosted runners for heavier tests that may need to run against a recent kernel. (https://github.com/kdave/btrfs-progs/actions)

Release process now uses more of the github features, so besides the tag itself there's also a release with changelog (https://github.com/kdave/btrfs-progs/releases).
This allows to add "artifacts" i.e. other files besides the sources.

There are now static binaries published along with the release. They're built inside the CI workflow "Static binaries" (also available there for download). The upload is still done from an intermediate host so if you don't trust github CI or me copying the files as-is, please build and use your own.

Please let me know if you have suggestions for improvements or if you find problems with the CI/release things. It's new to me and I'm sure the are already best practices to follow.

Changelog:
* mkfs: option -R deprecated, options unified in -O (-R still works)
* mkfs: fix potential race with udev leading to EBUSY due to repeatedly opened file descriptors
* block-group-tree is out of experimental mode
* available as 'mkfs.btrfs -O block-group-tree'
* btrfstune can do in-place conversion to/from (use with care)
* balance: fix recognizing old and new syntax
* subvol snapshot: specific error if a failure is caused by an active swapfile
* tree-stats: rephrase warning when run on a mounted filesystem
* completion: 'filesystem du' also completes files
* check: fix docs, help text and warning that --force + --repair works on a mounted filesystem
* build: fix static build when static libudev is available
* documentation:
* more updates from wiki, developer docs, changelogs
* reformatting
* updates and fixes
* other:
* test updates and fixes
* CI cleanups and old files removed
* integration with Github actions

Quoted 2 lines

Tarballs: https://www.kernel.org/pub/linux/kernel/people/kdave/btrfs-progs/
Git: git://git.kernel.org/pub/scm/linux/kernel/git/kdave/btrfs-progs.git

It seems like my patch to default to 4k pages[1] was missed for this release, despite having three reviewed-by tags.

Can we get this pulled in for the next release then? And I guess the patch will need to be updated to say 6.4 instead of 6.3...

[1]: https://lore.kernel.org/linux-btrfs/[email protected]/

--
真実はいつも一つ!/ Always, there's only one truth!

#2 lore raw
Quoted 6 lines · Neal Gompa
Quoted 2 lines · David Sterba

Tarballs: https://www.kernel.org/pub/linux/kernel/people/kdave/btrfs-progs/
Git: git://git.kernel.org/pub/scm/linux/kernel/git/kdave/btrfs-progs.git

It seems like my patch to default to 4k pages[1] was missed for this release, despite having three reviewed-by tags.

I'm aware of the patch, changes like that need more than just three reviews. Switching to sectorsize different than page size affects kernels/arches that are configured like that, which is now namely arm64 but power8/9 is still around and I've checked only a few main distros.

The testing coverage of the subpage code is very low on arm64 due to slow hardware some of us have and upstream support of the boards is lagging behind. First we need enough testing coverage for functionality and then at least some basic performance results. Anand sent some benchmarks which is good, but still not sufficient.

The situation with hardware may change soon so this is a start of the evaluation.

Quoted 2 lines

Can we get this pulled in for the next release then? And I guess the patch will need to be updated to say 6.4 instead of 6.3...

Changing defaults can take time and I can't promise any specific version for that. We had a change in 5.15, it's better to pack more changes in defaults to limit number of configuration so if there's next round the 4K will be among considered options. Until then we need to raise the confidence that this will not cause problems.

There's a tracking issue for that with the points above
https://github.com/kdave/btrfs-progs/issues/604 .

What we can do now is to add a warning or reminder that on a host with 64K pages the portability may be limited and that this should be reconsidered, with examples where this applies etc.

And once the change to 4k is done, backward compatibility would be possibly kept by adding 'page' as a special keyword for sectorsize.