covers: the 6.18 lane no longer carries the defrag NO_FLUSH patch

The 2026-09-16 rebase onto v6.18.52 met Filipe's broader fix for the
defrag reservation deadlock (upstream ba02eab28041f, backported to 6.18
as a9701dd06e82c: the reservation moves up-front into defrag_one_range()
and defrag_one_locked_target() no longer reserves).  Our 6.18-only
"btrfs: defrag: don't flush on reservation while holding the cluster
locked" was dropped in its favour, and the probe-margin hold that used
to hook defrag_one_locked_target() now reaches defrag through the
up-front btrfs_delalloc_reserve_space() -- the commit message of "hold
the probe margin until the metadata reservation charges it" already
lists that path, so nothing there changes.  Only carry_one_range() (the
log-tail carry) keeps a defrag.c hook.

Update the README (one 6.18-only patch now, not two) and the cover
letter paragraph that said both upstream-fixed bugs were carried only on
the 6.18 backport.

Assisted-by: Claude:claude-fable-5-1

diff --git a/covers/README.md b/covers/README.md
index 77a4b1733adb3a41e43240d479ed32c6861fb9ca..499f19c9e34811cecae644cacd5b29561833e62c 100644 (file)
--- a/covers/README.md
+++ b/covers/README.md
@@ -3,7 +3,7 @@
 This directory lives on the misc-next lane (`zygo-misc-next-zb64`), the
 submission lane, and describes the topics as they are on
 `misc-next/topics/*`.  The 6.18 lane carries the same topics as backports
-plus two 6.18-only patches (below); the linus/master lane is a formality
+plus one 6.18-only patch (below); the linus/master lane is a formality
 and may drift.  Reverified against the lane state on 2026-09-16 (both lanes rebased: 6.18 -> v6.18.52, misc-next -> kdave 7c6b5acf; new topic balance-resume added to both); round-1 review results included.
 
 | file | series | topic branch | state |
@@ -28,10 +28,14 @@ self-contained; no cover letter yet).  The other topics (backref-hang,
 dedupe-fixes, dev-extent-defrag, prealloc-size) are single patches with
 self-contained messages.
 
-6.18-only patches (not for submission): `btrfs: stripe_alloc: refuse
-space_cache=v1` (v1 no longer exists upstream) and `btrfs: defrag: don't
-flush on reservation while holding the cluster locked` (upstream fixed it
-differently in ba02eab28041f, 7.3); the 6.18 raid56-fixes topic also keeps
+6.18-only patch (not for submission): `btrfs: stripe_alloc: refuse
+space_cache=v1` (v1 no longer exists upstream).  The earlier 6.18-only
+`btrfs: defrag: don't flush on reservation while holding the cluster
+locked` was dropped at the 2026-09-16 rebase: upstream's broader fix
+ba02eab28041f (7.3) is backported to 6.18 as a9701dd06e82c, and the
+probe-margin hold reaches defrag through its up-front
+btrfs_delalloc_reserve_space() instead of a defrag.c hook.  The 6.18
+raid56-fixes topic also keeps
 `btrfs: do not write repaired read sectors back on a read-only mount`
 (upstream 8ceaad6cd6e7, v7.0 -- a stable backport request, not a
 submission).

diff --git a/covers/stripe-alloc.txt b/covers/stripe-alloc.txt
index 9b96fdabb605af0ebfb7512dac0abab846088832..0793c1a7af69570a60bcf7291317325e715b1c2c 100644 (file)
--- a/covers/stripe-alloc.txt
+++ b/covers/stripe-alloc.txt
@@ -239,12 +239,12 @@ lib/raid6 (sent to the md maintainers, Cc btrfs and stable):
       picked an x1 variant.
 
 Two more bugs found the same way were fixed upstream independently
-before this posting and are carried only on the 6.18 backport:
-btrfs_end_repair_bio() writing repaired sectors back on a read-only
-mount (upstream 8ceaad6cd6e7, v7.0; a stable backport is the right
-action for 6.18) and defrag reserving with a flush while holding a
-locked cluster (upstream ba02eab28041f, v7.3, fixed more broadly than
-our patch).
+before this posting: btrfs_end_repair_bio() writing repaired sectors
+back on a read-only mount (upstream 8ceaad6cd6e7, v7.0; our guard is
+carried only on the 6.18 backport, where a stable backport is the right
+action) and defrag reserving with a flush while holding a locked cluster
+(upstream ba02eab28041f, v7.3, fixed more broadly than our patch, which
+the 6.18 lane dropped once that fix was backported to it).
 
 The write hole
 ==============