`btrfs_finish_one_ordered` (add_pending_csums -> -ENOSPC, metadata
space_info 26.6 MiB over-reserved) during the metadata fill and again at
`__btrfs_update_delayed_inode` after the remount, while the control filled
-metadata to a clean ENOSPC. Classification: **ENOSPC transaction abort with loss of un-fsynced data
-only, not loss of acked data**: the 24125 lost files were written and
-closed but never fsynced and not yet committed when the first abort hit;
+metadata to a clean ENOSPC. Classification: **ENOSPC transaction abort with loss of acked data**
+(corrected 2026-11-12; the 10-18 text said "not loss of acked data").
+The 24125 lost files were created, written and closed with success
+returns but never fsynced, and were not yet committed when the first
+abort hit. A plain write() that returned success is acked, so this is
+silent loss of acked data, the outcome patch 12 prevents. What survived:
every data file synced before the metadata phase (43/43) verifies, and no
file that had been fsynced or covered by a completed sync was lost. Mechanism, from the code: the per-profile available
space simulation (`btrfs_update_per_profile_avail()` ->
raid1 chunk. Upstream fixed this in v7.1 (Qu Wenruo's per-profile
estimate, "This leads to various ENOSPC related transaction abort and
flips the fs read-only"); the 6.18 candidate carries those three commits
- and passes both arms. Loss class: ENOSPC transaction abort, un-fsynced
- data only (the abort hit while the writers were still writing; data
- synced before survived).
+ and passes both arms. Loss class: ENOSPC transaction abort with loss of
+ acked data (files whose writes had succeeded but were not yet
+ committed; corrected 2026-11-12); data synced before survived.
- e2 (patch 8) negative control: done 2026-11-06 (ah1106, above).
- `make headers_install` on the series (patch 9): done 2026-11-12, above.