Applying with these fixes on my side:
- 3/5: added Fixes: 3681ad22775c ("xfs: Teach the fs where the balloon
inode is"), and mount and remount added to the list of leaking calls
- 4/5, 5/5: "opened only through ..._OPEN_BALLOON, which takes
CAP_SYS_ADMIN" reworded: a file handle opens the balloon too
- 4/5, 5/5: "Physically adjacent extents are merged" -> "Extents that
follow each other both in the file and on disk are merged"
On Wed, 11 Nov 2026 12:31:30 +0100, Denis V. Lunev wrote:
> Daejun Park (2):
> ms/xfs: don't skip free extents in later batches of a whole-AG FITRIM
> ms/xfs: don't skip free extents of the longest length in a whole-AG
> FITRIM
>
> Denis V. Lunev (3):
> xfs: release the balloon inode when xfs_balloon_get() refuses it
> ext4: discard the balloon on FITRIM
> xfs: discard the balloon on FITRIM
Separate patches, not needed for this series; I created a task for
each:
1. A balloon discard error fails FITRIM, while both file systems
ignore discard errors in the free space pass. ploop_trim() then
stops and skips its smaller minlen rounds, which worked before this
series. Please say whether that difference is wanted; if not, treat
a balloon discard error the way the free space pass does.
https://virtuozzo.atlassian.net/browse/VSTOR-154840
2. The chunks end on multiples of the discard granularity counted
from the partition start, but the block layer aligns discards to
the start of the whole disk. On a partition that does not start on
a cluster boundary every chunk boundary splits a cluster, and ploop
keeps it mapped. Partitions made by ploop start at one cluster, so
this is minor: add the partition start to the remainder, or narrow
the comment.
https://virtuozzo.atlassian.net/browse/VSTOR-154841
--
Konstantin Khorenko,
Virtuozzo Linux Kernel Team
_______________________________________________
Devel mailing list
[email protected]
https://lists.openvz.org/mailman/listinfo/devel
