On 28.07.2025 17:14, Kent Overstreet wrote:
Schedule notes for users:
I've been digging through the bug tracker and polling users to see what
bugs are still outstanding, and - it's not much.
So, the experimental label is coming off in 6.18.
As always, if you do hit a bug, please report it.
I can now confirm that bcachefs is getting stable and the test cases
with intentionally data corruption (simulation of a real world case I
had) gets bcachefs back to a consistent state (after 2 runs of: bcachefs
fsck -f -y ${DEV}). That's a base requirement for a stable filesystem.
Version of bcachefs-tools is git
530e8ade4e6af7d152f4f79bf9f2b9dec6441f2b and kernel is
6.16.0-200.fc42.x86_64.
See for details, I made data corruption even worser with running the
destroy script 5x:
https://lore.kernel.org/linux-bcachefs/aa613c37-153c-43e4-b68e-9d50744be7de@xxxxxxxxxxxxx/
Great work Kent and the other contributors.
Unfortunately btrfs can't be repaired to a consistent state with the
same testcase. I'd like to be that testcase fixed also for BTRFS as a
stable filesystem (versions: 6.16.0-200.fc42.x86_64, btrfs-progs v6.15,
-EXPERIMENTAL -INJECT -STATIC +LZO +ZSTD +UDEV +FSVERITY +ZONED
CRYPTO=libgcrypt).
(I reported that already far in the past on the mailing list, see here:
https://lore.kernel.org/linux-btrfs/63f8866f-ceda-4228-b595-e37b016e7b1f@xxxxxxxxxxxxx/).
Thnx.
Ciao,
Gerhard
