]> git.hungrycats.org Git - linux/log
linux
7 days agoallocator-hints: claim only what the space tests measured: unfsynced files lost upstream
Zygo Blaxell [Thu, 12 Nov 2026 23:46:59 +0000 (18:46 -0500)]
allocator-hints: claim only what the space tests measured: unfsynced files lost

"Not yet committed" was inferred, not recorded; the cover and TESTING.md
now say the lost files had not been fsynced, pending a commit-timing
measurement.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7 days agoallocator-hints: DATA_ONLY and METADATA_ONLY space results (hs1112); cover says which...
Zygo Blaxell [Thu, 12 Nov 2026 23:20:34 +0000 (18:20 -0500)]
allocator-hints: DATA_ONLY and METADATA_ONLY space results (hs1112); cover says which hints lose data

TESTING.md: four rows and the arm descriptions for dataonly/metaonly on
misc-next and 6.18 with negative controls; the closed space-estimate
entry notes the two hints confirmed fixed; a new open item for a hint
change not clearing space_info->full.  Cover: patch 12 paragraph splits
the effect by hint (DATA_ONLY/NONE_ONLY lose acked data, METADATA_ONLY
only df Avail) and the reproducer covers both.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7 days agoallocator-hints: TESTING.md: the space-test loss is loss of acked data
Zygo Blaxell [Thu, 12 Nov 2026 21:19:44 +0000 (16:19 -0500)]
allocator-hints: TESTING.md: the space-test loss is loss of acked data

A plain write() that returned success is acked; the 10-18 text called
the lost un-fsynced files "not loss of acked data".  Both the hint-arm
classification and the 6.18 finding are corrected.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7 days agoallocator-hints: summarize the space reproducer in the cover; TESTING.md closed list
Zygo Blaxell [Thu, 12 Nov 2026 19:40:16 +0000 (14:40 -0500)]
allocator-hints: summarize the space reproducer in the cover; TESTING.md closed list

The cover described patch 12's reproducer only by a harness script name,
which reviewers cannot see; it now says what the test does.  TESTING.md
names the script and records it as the fstests candidate, counts four
follow-ups (9-12), and moves the fixed and done items out of "Not
covered / open" into a new "Closed" list (Zygo 2026-11-12).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7 days agoallocator-hints: patch 9 keeps the hint values in the UAPI header; patches 10-12...
Zygo Blaxell [Thu, 12 Nov 2026 17:19:46 +0000 (12:19 -0500)]
allocator-hints: patch 9 keeps the hint values in the UAPI header; patches 10-12 reworded (uh1112)

Re-exported from misc-next topics/allocator-hints 6a3533244111 (testing
9379f731c8e9, zb64 9e0e654aa3f3); export verified, fixup unchanged.
Still 12 patches, Cheng Yongkang's kobject fix excluded (now in
btrfs/for-next 4ab0b18d8a57).

Zygo 11-12: patch 9 drops the CONFIG_ guard instead of moving the values
to volumes.h; patch 10 no longer refuses the replace source (option A);
patch 11 and 12 messages updated (12: statfs changes f_bavail only).

Cover: Cheng paragraph dropped; patch 9 paragraph rewritten; patch 10,
11 and 12 paragraphs extended to match; diffstat now lists
include/uapi/linux/btrfs_tree.h.  Other covers' Date: churn reverted.
TESTING.md: 2026-11-12 round (oracles, object comparison, buildchecks,
headers_install with a control).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agobalance-resume: fold in remap-balance-leak (8 patches); cover DRAFT for Zygo (rm1111)
Zygo Blaxell [Thu, 12 Nov 2026 02:01:24 +0000 (21:01 -0500)]
balance-resume: fold in remap-balance-leak (8 patches); cover DRAFT for Zygo (rm1111)

Zygo 11-11: remap-balance-leak is merged into balance-resume, fixes first
(1 limit_min range, 2-4 the METADATA_REMAP fixes, 5 vrange, 6 limit
write-back, 7 queued remap chunks not written as processed, 8 legacy
filters).  No Cc stable (remap tree is experimental), Zygo 11-11.

Re-exported from misc-next topics/balance-resume 02f11b2bbcbc (testing
03d5cbfcdf57, zb64 cbed368de48b); export verified.  topics/remap-balance-leak
dropped; README table and ORDER updated.  Cover: 0/8, patches 2-4 and 7
described, Fixes named as hash ("title", vX.Y); DRAFT for Zygo.  TESTING.md:
2026-11-11 round added, remap-balance-leak's TESTING.md folded in.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agoscrub-commit-pause: TESTING: 7.3 buildcheck, promoted SHAs (pc1111)
Zygo Blaxell [Wed, 11 Nov 2026 06:20:32 +0000 (01:20 -0500)]
scrub-commit-pause: TESTING: 7.3 buildcheck, promoted SHAs (pc1111)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agoscrub-commit-pause: new series; replace-finish-deadlock patch 2 reworded; re-export...
Zygo Blaxell [Wed, 11 Nov 2026 06:07:19 +0000 (01:07 -0500)]
scrub-commit-pause: new series; replace-finish-deadlock patch 2 reworded; re-export after the pc1111 misc-next promotion

misc-next testing 6f8be7703b41, zb64 a6ceb4e01937 (base 83cd81353093;
kdave and btrfs fetched first, no public commit with the base's tree,
so no base-commit line).

- New series topics/scrub-commit-pause (1 patch, Zygo, e74d5a2d0a4c):
  "btrfs: scrub: do not let a transaction commit pass under queued
  stripes" (Fixes: 1dc4888e725d, v6.6; no Cc stable, Zygo 11-11).  The
  0/1 cover is a DRAFT for Zygo's review: it states the discard=sync-only
  prerequisite, the per-pause cost (drain of up to 128 queued stripes,
  paths dropped) and the harmless impact, and leaves the value to the
  reviewers.  TESTING.md has rd1110's numbers and the pc1111 gate.
- README: series table row, ORDER entry (last), "What the fixup
  contains" item 4, and a "Before sending" note.
- fixups/post-merge.patch: + scrub.c (scrub_pause_point(), which -X ours
  drops at the raid56-scrub-replace conflict); 667 lines added, 2
  removed, 7 files; verified against testing by export.sh.
- replace-finish-deadlock patch 2 (now 0c648f0b012b): changelog reworded
  on the lanes (Zygo 11-11: "affects every profile" before RAID5);
  tree unchanged.  Its cover's From line follows.
- The other covers changed only in their Date: headers (reverted).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agocovers: name each Fixes commit with its title and release (Zygo 11-11)
Zygo Blaxell [Wed, 11 Nov 2026 05:29:04 +0000 (00:29 -0500)]
covers: name each Fixes commit with its title and release (Zygo 11-11)

Same rule as the replace-fixes cover: deadlock-fixes, raid56-fixes,
raid56-scrub-replace and reflink-inline-fixes quoted bare hashes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agoreplace-fixes: cover: drop Josef's trim patch, name the Fixes commits with title...
Zygo Blaxell [Wed, 11 Nov 2026 05:28:11 +0000 (00:28 -0500)]
replace-fixes: cover: drop Josef's trim patch, name the Fixes commits with title and release (Zygo 11-11)

Josef's "btrfs: don't trim the target of a device replace" is (or soon
will be) upstream already, so the cover no longer mentions it.  Each
Fixes reference now quotes the commit title and the release that
introduced it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agoreplace-finish-deadlock: cover: patch 2 says it affects every profile first (Zygo...
Zygo Blaxell [Wed, 11 Nov 2026 05:15:05 +0000 (00:15 -0500)]
replace-finish-deadlock: cover: patch 2 says it affects every profile first (Zygo 11-11)

The patch 2 paragraph now opens with the profile-independence and drops
the trailing sentence.  The patch 2 changelog gets the same reword on the
lanes (pc1111); the exported patch follows with the next re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agobalance-resume: patch 1 has a Fixes tag and no Cc stable (Zygo 11-11)
Zygo Blaxell [Wed, 11 Nov 2026 05:14:51 +0000 (00:14 -0500)]
balance-resume: patch 1 has a Fixes tag and no Cc stable (Zygo 11-11)

README row and TESTING.md: the limit_min fix is "not a high severity
issue"; it keeps Fixes: 12907fc79818 and is not for stable.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8 days agoreplace-fixes: review edits
Zygo Blaxell [Wed, 11 Nov 2026 05:39:57 +0000 (00:39 -0500)]
replace-fixes: review edits

 * Clarify the reason for comparing replace with dev/add.
 * Drop the "unfixed" tag from scrub testing--the behavior doesn't need to change.

9 days agoreplace-finish-deadlock: cover: replace counts include the 7.3 round
Zygo Blaxell [Wed, 11 Nov 2026 01:10:04 +0000 (20:10 -0500)]
replace-finish-deadlock: cover: replace counts include the 7.3 round

72 replaces per profile (48 with thread_pool=64) with the series, none
hung: KASAN and non-KASAN kernels, 12 per arm and kernel.  Still a draft
for Zygo's review.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9 days agoreplace-finish-deadlock: TESTING: the 7.3 round
Zygo Blaxell [Wed, 11 Nov 2026 00:58:34 +0000 (19:58 -0500)]
replace-finish-deadlock: TESTING: the 7.3 round

7.3 promoted (topic 0b22c35e0e7b, testing 7ab729bb2f60, zb64
ace590553e2b): stress r5/r5tp64/r6p/r1 12/12 on KASAN+FI #327 and
non-KASAN #328; gate column added (all PASS or the known non-verdicts:
rr-scrubabort INVALID, scrub-fserror-stop scrub arm INVALID with the
replace arm PASS on rerun, unwritten-readback retry artifact).  No
change to the series or the cover.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9 days agoreplace-finish-deadlock: new series, re-export after the rf1109 misc-next promotion
Zygo Blaxell [Tue, 10 Nov 2026 16:47:26 +0000 (11:47 -0500)]
replace-finish-deadlock: new series, re-export after the rf1109 misc-next promotion

misc-next testing d249ec595c9c, zb64 efaeb5436501 (base 83cd81353093;
kdave and btrfs fetched first, no public commit with the base's tree,
so no base-commit line).

- New series topics/replace-finish-deadlock (2 patches, Zygo), the
  dev-replace bio counter deadlocks found by rh1108: "btrfs: do not wait
  for device replace in the bio completion workers" (Fixes: 7609afac6775)
  and "btrfs: do not hold the bio counter across the data csum lookup"
  (Fixes: 1c2b3ee3b0ec).  Cover letter is a DRAFT for Zygo's review
  (no Cc stable, no tool-use note, like the other covers); TESTING.md
  has the misc-next and 6.18 rounds, 7.3 follows after its gate.
- README: series table row and the integration ORDER.
- fixups/post-merge.patch unchanged (the series merges cleanly last).
- The zb64-only "'btrfs replace' hangs at end of replacing a device"
  commit dropped from misc-next zb64 was never exported; no change.
- The other covers changed only in their Date: headers (reverted).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9 days agobalance-resume: TESTING: 2026-11-10 round (the limit_min fix as patch 1, lr1109)
Zygo Blaxell [Tue, 10 Nov 2026 08:48:56 +0000 (03:48 -0500)]
balance-resume: TESTING: 2026-11-10 round (the limit_min fix as patch 1, lr1109)

The intro and the numbering note follow the 4-patch series (rounds
before 2026-11-10: add one to their patch numbers).  New round: the bug
re-derived from the source, the fix, why it goes first, the five new
limit-test arms with their unfixed outcomes, and the gate on 6.18 #36,
misc-next #71 and 7.3 #5 (all PASS) with the 7.3 #3 control (lm2, l2m,
l2m1 and lres FAIL as predicted).  The 2026-11-08 round's alias item is
marked closed.  remap-balance-leak: the misc-next #71 result row.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9 days agobalance-resume: re-export after the lr1109 misc-next promotion
Zygo Blaxell [Tue, 10 Nov 2026 08:09:48 +0000 (03:09 -0500)]
balance-resume: re-export after the lr1109 misc-next promotion

misc-next testing 3f4e1b6e5913, zb64 d4cc0ed0de64 (base 83cd81353093;
kdave and btrfs fetched first, no public commit with the base's tree,
so no base-commit line).

- balance-resume is now 4 patches: "btrfs: apply the balance limit_min
  filter only to a limit range" (Fixes: 12907fc79818) first, then the
  vrange bound (2), the limit write-back (3) and the legacy filters (4),
  replayed unchanged.  The cover's headers, shortlog and diffstat are
  regenerated; README's series table and "Before sending" note follow
  the renumbering.
- fixups/post-merge.patch: one hunk offset in volumes.c (+13 lines
  above it), same content.
- The other covers changed only in their Date: headers (reverted).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9 days agobalance-resume: cover: the limit_min fix as patch 1 (draft for Zygo's review)
Zygo Blaxell [Tue, 10 Nov 2026 08:07:39 +0000 (03:07 -0500)]
balance-resume: cover: the limit_min fix as patch 1 (draft for Zygo's review)

"btrfs: apply the balance limit_min filter only to a limit range" goes
first in the series (Zygo 2026-11-09: "add that to the balance-resume
series"): a new paragraph for it, the later patches renumbered (vrange
2, limit write-back 3, legacy filters 4), and the limit test entry
gains the single-limit arms.  The headers, shortlog and diffstat follow
at the re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
10 days agoremap-balance-leak: cover letter (draft for Zygo's review)
Zygo Blaxell [Tue, 10 Nov 2026 04:28:19 +0000 (23:28 -0500)]
remap-balance-leak: cover letter (draft for Zygo's review)

The series is one patch.  The README convention (every series has a
cover, single-patch ones included) gives it a 0/1 cover; export.sh now
keeps it, so the patch becomes [PATCH 1/1].  Headers, shortlog and
diffstat are from an export.sh dry run against misc-next testing
5fa1c4f82174 (verified, no base-commit), so the next re-export changes
only the Date: lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
10 days agonocmtime, remap-balance-leak: re-export after the nc1108 misc-next promotion
Zygo Blaxell [Mon, 9 Nov 2026 13:42:32 +0000 (08:42 -0500)]
nocmtime, remap-balance-leak: re-export after the nc1108 misc-next promotion

misc-next testing 5fa1c4f82174, zb64 fb7546be6023 (base 83cd81353093).

- Jeff Layton's FMODE_NOCMTIME series replaces dedupe-mtime in the lane
  (topics/nocmtime).  It is Jeff's, so exclude.txt drops its four
  subjects and topics/dedupe-mtime is removed.  Two of its reverts from
  testing need stored resolutions (scripts/revert-resolutions/, README);
  export.sh's verify step now reverts the dropped commits newest first.
- "btrfs: free queued remap chunks when balance bails out" leaves
  balance-resume for its own series, topics/remap-balance-leak (1 patch,
  no cover yet), merged right after balance-resume: README series table
  and ORDER.  Its leak test rows move to its TESTING.md, plus the
  2026-11-09 misc-next result.
- balance-resume is now vrange (1), limit write-back (2), legacy filters
  (3); TESTING.md's rounds before 2026-11-08 are renumbered to match.
  The cover changed only in its From/Date headers (patch count 3 -> 3,
  shortlog and diffstat already current).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
10 days agobalance-resume: TESTING: 6.18 gate results for the limit commit (bl1108)
Zygo Blaxell [Mon, 9 Nov 2026 07:44:04 +0000 (02:44 -0500)]
balance-resume: TESTING: 6.18 gate results for the limit commit (bl1108)

Limit test rows filled from ahive #10 (PASS x2) and the #9 control
(FAIL as predicted, 3a PASS); a 6.18 gate subsection with the rest of
the balance-resume suite, the legacy variants, the no-unmount check and
nocmtime-test.sh.  The alias note is corrected: the limit_min check sees
the already decremented limit, so limit=L with M matches relocates
2M+1-L, not none (measured: M=7, limit=9 relocated 6 on both kernels).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: failed item writes undo no work, but the rest needs a new...
Zygo Blaxell [Mon, 9 Nov 2026 04:02:04 +0000 (23:02 -0500)]
balance-resume: cover: failed item writes undo no work, but the rest needs a new balance

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: record the intent for failed item writes
Zygo Blaxell [Mon, 9 Nov 2026 04:01:53 +0000 (23:01 -0500)]
balance-resume: cover: record the intent for failed item writes

Zygo (2026-11-08): keep the error handling consistent with the existing
code (a failed write-back at resume or limit update ends the balance and
deletes its item), but record the intent in the cover.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: TESTING: failure-path audit of the new item writes, injection plan...
Zygo Blaxell [Mon, 9 Nov 2026 03:26:48 +0000 (22:26 -0500)]
balance-resume: TESTING: failure-path audit of the new item writes, injection plan (bl1108)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: diffstat after the resume write-back failure message (bl1108)
Zygo Blaxell [Mon, 9 Nov 2026 03:26:09 +0000 (22:26 -0500)]
balance-resume: cover: diffstat after the resume write-back failure message (bl1108)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: TESTING: 2026-11-08 round (limit writeback, bl1108), not run yet
Zygo Blaxell [Mon, 9 Nov 2026 02:33:58 +0000 (21:33 -0500)]
balance-resume: TESTING: 2026-11-08 round (limit writeback, bl1108), not run yet

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: rewrap the on-disk state paragraph
Zygo Blaxell [Mon, 9 Nov 2026 02:33:28 +0000 (21:33 -0500)]
balance-resume: cover: rewrap the on-disk state paragraph

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: limit filter (Zygo's two sentences, patch 2 paragraph, test)
Zygo Blaxell [Mon, 9 Nov 2026 02:33:20 +0000 (21:33 -0500)]
balance-resume: cover: limit filter (Zygo's two sentences, patch 2 paragraph, test)

Zygo's sentences verbatim: "Adding vrange to any balance filter makes it
idempotent, except for the limit filter which is handled in the second
patch." in place of the claim that all filters can be made idempotent
(the preceding clause now says only what a vrange can be added without),
and "There are no balance item updates during the balance unless the
limit filter is used." in place of the "no balance item updates" half
(the "no new on-disk state" half stays: the patch reuses the item's
fields).  New paragraph on patch 2's limit writeback, and a Testing
bullet for balance-resume-limit-test.sh.  For Zygo's review.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: new series shape (remap fix leaves, limit patch added)
Zygo Blaxell [Mon, 9 Nov 2026 02:32:41 +0000 (21:32 -0500)]
balance-resume: cover: new series shape (remap fix leaves, limit patch added)

The remap fix becomes its own one-patch series: drop its paragraph.  The
series is now vrange (1), limit writeback (2), legacy heuristics (3):
renumber the patch references, update the shortlog and the diffstat
(misc-next candidate refs/cand/bl1108/misc-next/balance-resume).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: cover: "for most of them", comma, tie the guarantee to the floor
Zygo Blaxell [Mon, 9 Nov 2026 02:00:44 +0000 (21:00 -0500)]
balance-resume: cover: "for most of them", comma, tie the guarantee to the floor

Zygo's review rulings (2026-11-08) on 84ced0b97bc03: the usage-range
filter did get a heuristic exception, so "not updated for most of them";
a comma after the parenthetical; and the guarantee depends on the
find_next_chunk() floor described later, so say so where it is claimed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
11 days agobalance-resume: mention bitrot in heuristics, not keeping up with new balance filters
Zygo Blaxell [Mon, 9 Nov 2026 01:34:01 +0000 (20:34 -0500)]
balance-resume: mention bitrot in heuristics, not keeping up with new balance filters

12 days agoexport: exclude: give the REQ_META pick its own lane-only block
Zygo Blaxell [Sun, 8 Nov 2026 02:37:29 +0000 (21:37 -0500)]
export: exclude: give the REQ_META pick its own lane-only block

Mark Korenberg's raid56 REQ_META patch was NACKed upstream; it is
carried only while stripe_meta is, and resent only after stripe_meta is
accepted.  Separate it from the 11-06 "posted separately" block, whose
patches are on the list and expected to land.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
12 days agostripe-alloc: TESTING: Mark Korenberg's REQ_META pick at the head of stripe_meta...
Zygo Blaxell [Sun, 8 Nov 2026 02:34:20 +0000 (21:34 -0500)]
stripe-alloc: TESTING: Mark Korenberg's REQ_META pick at the head of stripe_meta (km1107)

Also list its subject in scripts/exclude.txt: Mark posts it himself.  Covers
unchanged, no re-export (the lanes are not promoted yet).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agoallocator-hints: TESTING: fixed-width mirror arms and their negative control (ah1106)
Zygo Blaxell [Sat, 7 Nov 2026 00:30:49 +0000 (19:30 -0500)]
allocator-hints: TESTING: fixed-width mirror arms and their negative control (ah1106)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agoexport: re-export from misc-next testing 3b5b4d424790 (zb64 dcf19c743b68)
Zygo Blaxell [Fri, 6 Nov 2026 20:44:36 +0000 (15:44 -0500)]
export: re-export from misc-next testing 3b5b4d424790 (zb64 dcf19c743b68)

After xf1106: the stripe_alloc extract-hook drop is folded into the
patches that added the hooks (4 and 40), so stripe-alloc is 85 patches
again, matching its cover.  Trees unchanged; no base-commit line (no
public commit has the lane base tree).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agostripe-alloc, compress-fixes: TESTING: extract-hook drop folded into patches 4 and...
Zygo Blaxell [Fri, 6 Nov 2026 20:41:21 +0000 (15:41 -0500)]
stripe-alloc, compress-fixes: TESTING: extract-hook drop folded into patches 4 and 40 (xf1106)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agoexport: re-export from misc-next testing 86f7f846934c (zb64 d411e967283c)
Zygo Blaxell [Fri, 6 Nov 2026 10:49:12 +0000 (05:49 -0500)]
export: re-export from misc-next testing 86f7f846934c (zb64 d411e967283c)

After the br1105, dl1106 and fl1106 promotions.  Base 83cd81353093; no
public commit on kdave/btrfs/linus has its tree, so the mails carry no
base-commit line.

exclude.txt: the patches by others that the covers name as posted
separately (Josef Bacik's data integrity fixes v2 2/5-5/5, Cheng
Yongkang's devid kobject fix) are left out of the mails; the topics keep
them.  Series: stripe-alloc 86, replace-fixes 10, allocator-hints 12,
dev-extent-defrag 2.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agobalance-resume: TESTING: write-back is conditional; no-unmount resume check
Zygo Blaxell [Fri, 6 Nov 2026 10:42:58 +0000 (05:42 -0500)]
balance-resume: TESTING: write-back is conditional; no-unmount resume check

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agobalance-resume, dev-extent-defrag: recovered items bounded at mount; aligned-holes...
Zygo Blaxell [Fri, 6 Nov 2026 10:28:30 +0000 (05:28 -0500)]
balance-resume, dev-extent-defrag: recovered items bounded at mount; aligned-holes patch joins the series

balance-resume cover: one sentence on recovered items (vrange added or
lowered, written back once at resume) and the legacy-test bullet.
dev-extent-defrag cover: name patch 2 (aligned holes) in the intro and
the testing paragraph.  Both TESTING.md: the 2026-11-06 round.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agodeadlock-fixes: TESTING: 2026-11-06 changed only comments
Zygo Blaxell [Fri, 6 Nov 2026 07:28:59 +0000 (02:28 -0500)]
deadlock-fixes: TESTING: 2026-11-06 changed only comments

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agoscripts/export.sh: name a public base-commit, or none
Zygo Blaxell [Fri, 6 Nov 2026 07:27:11 +0000 (02:27 -0500)]
scripts/export.sh: name a public base-commit, or none

The lane base is a local copy of kdave's "=== misc-next on b-for-next ==="
marker, so the base-commit line in the mails named a commit no reviewer
can fetch.  Name instead a commit on a public remote-tracking branch
(kdave, btrfs, linus; fetched first) with the same tree as the lane base,
and send without a base-commit line when there is none.  PUBBASE=
overrides the search.  The verify step checks every base-commit line.

Dry runs against misc-next/testing 9457797a25d7: no tree match (kdave's
misc-next has moved on), 21 series without base-commit, verified;
PUBBASE=kdave/misc-next: every line names ee901676f63d, verified.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agocovers: drop TESTING.md pointers, lane wording and harness paths
Zygo Blaxell [Fri, 6 Nov 2026 06:33:01 +0000 (01:33 -0500)]
covers: drop TESTING.md pointers, lane wording and harness paths

Zygo 2026-11-06: TESTING.md and the lanes do not belong in cover
letters; an upstream reviewer gets only the cover letter and patches.

In the 18 covers that had them: no "details in TESTING.md" pointers;
Subject lines of allocator-hints and read-policies lose "as carried on
misc-next"; "misc-next lane" / "misc-next and 6.18" test-kernel lists
become the kernel config facts (KASAN, lockdep); comparisons against
"the lanes" / "previous lane kernels" become "without the series" (or
name the earlier level patch); "Port notes for misc-next" become
"Changes for the current base"; the lane-only paragraphs about patches
carried ahead of the series (Cheng's devid kobject fix, Josef's data
integrity fixes) shrink to what a reviewer needs (posted separately,
why it matters); the stripe-alloc lane-only extract-hook paragraph
(folds into patches 4 and 40 at the next export) is dropped; harness,
forensics, ck110x tags and README pointers are reworded or dropped.
Only touched paragraphs are reflowed; shortlog/diffstat/base-commit
untouched.  TESTING.md files keep the lane detail.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agodedupe-fixes: cover testing paragraph as a report, with kernel and TESTING.md pointer
Zygo Blaxell [Fri, 6 Nov 2026 06:05:50 +0000 (01:05 -0500)]
dedupe-fixes: cover testing paragraph as a report, with kernel and TESTING.md pointer

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agoTESTING: the hang kit is in raid56-wh-harness now
Zygo Blaxell [Fri, 6 Nov 2026 06:01:02 +0000 (01:01 -0500)]
TESTING: the hang kit is in raid56-wh-harness now

deadlock-fixes, dedupe-fixes and stripe-alloc TESTING.md pointed at
~/share/raid56-hang-debug for the reproducers and captures.  The kit was
merged into raid56-wh-harness on 2026-11-06: scripts at its top level,
cited captures in forensics/hang-kit-captures/.  Drops deadlock-fixes'
"not in the harness".  No cover letter named the kit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agodedupe-fixes: trim the cover letter
Zygo Blaxell [Fri, 6 Nov 2026 05:51:46 +0000 (00:51 -0500)]
dedupe-fixes: trim the cover letter

13 days agobalance-resume: testing for the no-change-on-resume rule (br1105)
Zygo Blaxell [Fri, 6 Nov 2026 05:20:35 +0000 (00:20 -0500)]
balance-resume: testing for the no-change-on-resume rule (br1105)

Cover: Testing entries for the new balance-resume-convert-test.sh and
the extended legacy test (convert -> soft, user vrange -> unchanged).
TESTING.md: the soft change in patches 2/3 and the 2026-11-05 results,
including the negative controls on the previous lane kernels.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agobalance-resume: cover letter review fixes
Zygo Blaxell [Fri, 6 Nov 2026 05:02:00 +0000 (00:02 -0500)]
balance-resume: cover letter review fixes

find_next_chunk() returns an address rather than allocating; say how a
user-supplied vrange end is lowered; patch 3 applies per chunk type;
"no completed block group is repeated" (Zygo's wording).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 days agobalance-resume: human-edited cover letter
Zygo Blaxell [Fri, 6 Nov 2026 04:45:48 +0000 (23:45 -0500)]
balance-resume: human-edited cover letter

2 weeks agoscripts/export.sh: strip Assisted-by only from Zygo's patches
Zygo Blaxell [Fri, 6 Nov 2026 01:52:12 +0000 (20:52 -0500)]
scripts/export.sh: strip Assisted-by only from Zygo's patches

Zygo: removing every author's Assisted-by is too much.  Strip the tag
only from exported patches whose From: is Zygo Blaxell
<ce3g8jdj@umail.furryterror.org>; other authors' patches keep their
tags.  The verify step checks only Zygo's patches, and skips cover
letters as the stripping does.

Dry run against misc-next/testing 2365f2f67131: verified; 11 tags left,
all in other authors' patches.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agoscripts/export.sh: strip Assisted-by from the exported patches
Zygo Blaxell [Fri, 6 Nov 2026 01:42:26 +0000 (20:42 -0500)]
scripts/export.sh: strip Assisted-by from the exported patches

Upstream dropped the Assisted-by rule in 1e035297ffda ("docs: remove
Assisted-by references"); tool use is described in the cover letter
instead.  The lane commits keep the tag, so Zygo can see which model
wrote or edited a patch, but the send-email copies lose it: every
"Assisted-by:" line in the changelog part of each exported patch
(headers to the first "---") is removed, whoever added it.  Cover
letters are left alone.  The verify step now fails if any is left.

Dry run against misc-next/testing 2365f2f67131: 226 lines removed, all
patches still apply to the topic trees, fixup verified.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agocovers: sb1104: 6.18 read-repair-fixes starts with Filipe's three scrub fixes
Zygo Blaxell [Thu, 5 Nov 2026 08:59:50 +0000 (03:59 -0500)]
covers: sb1104: 6.18 read-repair-fixes starts with Filipe's three scrub fixes

TESTING.md of read-repair-fixes and raid56-scrub-replace: the 6.18 backport, its placement,
the progress-ioctl UAF reproducer (#316 KASAN hit, #317 none) and the gate.  The series sent
upstream is unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agocovers: ck1104: stripe-alloc drops its extract hooks; compress-fixes promoted
Zygo Blaxell [Wed, 4 Nov 2026 22:54:38 +0000 (17:54 -0500)]
covers: ck1104: stripe-alloc drops its extract hooks; compress-fixes promoted

stripe-alloc: a new tip commit on both lanes drops the dead extract hooks
of the stripe_alloc and stripe_alloc_allow_rmw property handlers (not
inheritable) and moves prop_compression_extract() back next to
prop_compression_ignore().  The testing octopus then merges with
koraynilay's level series without a resolution.  Cover paragraph plus a
TESTING.md section: caller audit, new mount-state non-inheritance case,
gate on #316 / #66.  compress-fixes: the integration note now says the
merge is clean; TESTING.md records the ck1104 promotion and the
Assisted-by tag on the edited 6.18 koraynilay 3/6 backport.  Not a
re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agocompress-fixes TESTING.md: oldvalues mechanism as observed
Zygo Blaxell [Wed, 4 Nov 2026 02:39:22 +0000 (21:39 -0500)]
compress-fixes TESTING.md: oldvalues mechanism as observed

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agocompress-fixes: cover and TESTING.md for the rebuild on koraynilay v6 (ck1103)
Zygo Blaxell [Wed, 4 Nov 2026 02:35:31 +0000 (21:35 -0500)]
compress-fixes: cover and TESTING.md for the rebuild on koraynilay v6 (ck1103)

Candidates only (refs/cand/ck1103/*): the testing octopus needs an
extract() prototype resolution against stripe-alloc, which waits for
Zygo. Not re-exported.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agoTESTING.md: rb1103 on the 6.18 lane (raid56-fixes, stripe-alloc, replace-fixes, compr...
Zygo Blaxell [Tue, 3 Nov 2026 20:46:00 +0000 (15:46 -0500)]
TESTING.md: rb1103 on the 6.18 lane (raid56-fixes, stripe-alloc, replace-fixes, compress-fixes)

6.18 subsections next to the misc-next rb1103 ones: Josef v2 1/5, 2/5,
5/5 message-only; v2 3/5 and 4/5 replace v1 in place, with the 6.18 port
of 3/5 (spin_lock_irq() on ordered_tree_lock, which 6.18 takes from bio
completion with IRQs off); Filipe's patch replaces ours in compress-fixes
(btrfs/for-next 75d0e4dd64f2a since today).  Gate on 6.18 #313 (bhive,
ahive) equal to rb1101; compress-props PASS, negative control on #311.
3/5's IOERR branch is BLIND on 6.18 too.  Not a re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agocovers: rb1103 misc-next: Josef data-integrity v2, Filipe's nodatacow/compress patch...
Zygo Blaxell [Tue, 3 Nov 2026 20:20:55 +0000 (15:20 -0500)]
covers: rb1103 misc-next: Josef data-integrity v2, Filipe's nodatacow/compress patch replaces ours

raid56-fixes, replace-fixes: Josef 1/5 and 5/5 are v2 (code unchanged,
new tags).  stripe-alloc: Josef 2/5-4/5 v2 at the head; 3/5 reworked
("wait for ordered extent IO in fast fsync"), which answers the v1 race
report; 4/5 DEBUG_WARN.  compress-fixes: Filipe Manana's "dont propagate
nodatacow mount option on files with inherited compression" (Qu R-b,
already in btrfs/for-next 75d0e4dd64f2a) replaces our patch 2.  Our
legacy-parent btrfs_inherit_iflags() hunk is dropped with no replacement,
so upstream behaviour applies.  TESTING.md for all four topics has the
rb1103 gate results; 3/5's crash window is BLIND in every harness test.
Not a re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agoreplace-fixes: TESTING: drop the dasync-raid5 case
Zygo Blaxell [Mon, 2 Nov 2026 23:35:02 +0000 (18:35 -0500)]
replace-fixes: TESTING: drop the dasync-raid5 case

The case could only be BLIND (no discard reaches a raid5 member) and is
gone from replace-discard-test.sh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agocovers: rb1101 misc-next (for-next a919d4a3e831): Josef 1/5 replaces our split-bio...
Zygo Blaxell [Mon, 2 Nov 2026 02:29:08 +0000 (21:29 -0500)]
covers: rb1101 misc-next (for-next a919d4a3e831): Josef 1/5 replaces our split-bio fix, Qu rmw_rbio pair upstream, Josef 2-4/5 + 5/5 and Cheng v2 carried at topic heads

raid56-fixes cover: 10 patches (Qu pair in for-next and the base; Josef 1/5
as patch 3, ours dropped), renumbered, new base.  stripe-alloc cover: Josef
2/5-4/5 at the head, 3/5 v1 with the open Sashiko race report.
replace-fixes cover + TESTING.md: Josef 5/5 carried at the head until
upstream takes it.  allocator-hints cover: Cheng v2 at the head.
raid56-scrub-replace cover: Filipe prerequisite hash = current for-next
0bade584e95b9.  Not a re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agoTESTING.md: replace-fixes: replace with discards (fstrim, discard=sync/async, no -K)
Zygo Blaxell [Sun, 1 Nov 2026 02:29:17 +0000 (22:29 -0400)]
TESTING.md: replace-fixes: replace with discards (fstrim, discard=sync/async, no -K)

New rows for replace-discard-test.sh on both lanes, base zb64 and the
josef-trim candidate (lane zb64 + Josef Bacik's "btrfs: don't trim the
target of a device replace"), plus discard=async reruns of
replace-status and replace-cancel-persist.  FITRIM during a replace
loses the target's copy on both lane bases (single: the fs no longer
mounts; raid1: target alone does not mount; raid5: silent, 169296 csum
errors on the target); the candidate PASSes every row.  The bug is
upstream, not fixed by this series.  Lists the checks that are invalid
in discard mode (the poison strip checks) and what is used instead.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agoexport: misc-next after knobmv1030; no stored resolution needed
Zygo Blaxell [Sat, 31 Oct 2026 05:54:59 +0000 (01:54 -0400)]
export: misc-next after knobmv1030; no stored resolution needed

Re-exported from misc-next: base bf921b86bdf2, testing 0dd345dcdb09,
zb64 9291348d2cbe (knobmv1030 promoted 2026-10-31: the rbio cache size
knob, test-knobs 78dfea36c175, declares its function below
btrfs_rbio_unpin_stale() in raid56.h).

The knob now reverts cleanly from testing, so the stored resolution for
the old knob commit 3213446fec93 is removed and README.md says none is
in use.  The patches, the fixup (6 files, +636/-2) and the fixup target
(tree 911f5da3a97a) are unchanged apart from headers and base SHAs.

Build check of the fixup target on bhive: W=1 fs/btrfs and bzImage,
BTRFS_DEBUG/BTRFS_ASSERT on and off, 0 warnings, both link.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 weeks agoexport: misc-next topics after restructure1030, without the test-only commits
Zygo Blaxell [Sat, 31 Oct 2026 02:39:50 +0000 (22:39 -0400)]
export: misc-next topics after restructure1030, without the test-only commits

Re-exported from misc-next: base bf921b86bdf2, testing d88841f5cf02,
zb64 db4cce38a238 (lanes promoted 2026-10-30: review1029, replace-resume,
restructure1030).

scripts/export.sh now leaves out every commit marked DO-NOT-UPSTREAM
(subject or body, any spelling, or "dropped before upstream") and what
scripts/exclude.txt lists (topic dm-stripfault, topic test-knobs).  A
topic that loses commits is replayed without them with the original
idents and messages, and the export stops instead of hand-resolving if a
later commit no longer applies.  Stacked topics are exported on their
parent with prerequisite-patch-id lines.  The fixup target is testing
with the dropped commits reverted; a revert that conflicts uses a stored
resolution from scripts/revert-resolutions/<sha>.patch, or stops.  One
is in use: reverting the rbio cache size knob (test-knobs 3213446fec93)
conflicts in raid56.h next to raid56-fixes' uncache_range declaration.

Left out: dm-stripfault (493cea52681f, 8968252ad50b), test-knobs
(9ab48be32773 reclaim exclusive-operation gap, 3213446fec93 rbio cache
size knob).  Every exported topic tip is the lane tip itself.

21 series: logical-ino-costs 3, balance-resume 3, compress-fixes 5,
deadlock-fixes 3, dedupe-fixes 1, delalloc-fixes 1, dev-extent-defrag 1,
prealloc-size 1, raid56-fixes 12, read-repair-fixes 4,
raid56-scrub-replace 12 (on read-repair-fixes), reclaim-fixes 3,
stripe-alloc 85 (on reclaim-fixes), low-space 3 (on stripe-alloc),
dedupe-mtime 1, allocator-hints 12, read-policies 9, replace-fixes 10,
dev-stats 1, resize-missing 1, reflink-inline-fixes 2 (new).

Verified by export.sh: every series' patches give its branch's tree,
which equals the lane branch; the leaf branches merged with -X ours plus
fixups/post-merge.patch (6 files, +636/-2) give the fixup target, tree
911f5da3a97a.

Covers: stripe-alloc rewritten for the 85-patch subtopic order
(prerequisites, renumbered citations, debug subtopics as not for
merging, review and BTRFS_DEBUG=n build notes); raid56-fixes (Qu's two
patches, no knob), deadlock-fixes (relock folded), reclaim-fixes (gap
not posted), low-space (prerequisites, review1029 results),
raid56-scrub-replace (stacked on read-repair-fixes, abort gate),
replace-fixes (10 patches: resume chunk skip, single device mount, stop
on fs error), read-repair-fixes, allocator-hints, from the review1029
message changes.  TESTING.md updated for stripe-alloc (116 -> 85 map),
replace-fixes, raid56-fixes, raid56-scrub-replace, deadlock-fixes,
low-space and reclaim-fixes.  README.md documents the exclusions, the
stacking and the new fixup.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3 weeks agoreclaim-fixes: re-export from the final candidate, TESTING.md results
Zygo Blaxell [Wed, 28 Oct 2026 13:49:41 +0000 (09:49 -0400)]
reclaim-fixes: re-export from the final candidate, TESTING.md results

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3 weeks agoreclaim-fixes: new topic (candidate: cover, patches, TESTING.md)
Zygo Blaxell [Wed, 28 Oct 2026 07:32:05 +0000 (03:32 -0400)]
reclaim-fixes: new topic (candidate: cover, patches, TESTING.md)

The generic block group reclaim fixes split out of topics/stripe-alloc by
the sa-fold rework, exported from refs/cand/sa-fold/mn/reclaim.  Not
promoted; export.sh not run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3 weeks agodeadlock-fixes TESTING.md: -EUCLEAN candidate test results
Zygo Blaxell [Tue, 27 Oct 2026 05:55:36 +0000 (01:55 -0400)]
deadlock-fixes TESTING.md: -EUCLEAN candidate test results

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3 weeks agodeadlock-fixes TESTING.md: -EUCLEAN for leaked delalloc in the clone lock loop (candi...
Zygo Blaxell [Tue, 27 Oct 2026 01:02:06 +0000 (21:02 -0400)]
deadlock-fixes TESTING.md: -EUCLEAN for leaked delalloc in the clone lock loop (candidate)

Zygo chose -EUCLEAN with a warning over -EAGAIN for delalloc that survives
the flush-and-retry rounds of patch 2.  The candidate folds it into patch 2;
the patch files stay the lane's until promotion.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3 weeks agoreflink-inline-fixes: new topic (cover, patches, TESTING.md); deadlock-fixes: relock...
Zygo Blaxell [Mon, 26 Oct 2026 23:41:57 +0000 (19:41 -0400)]
reflink-inline-fixes: new topic (cover, patches, TESTING.md); deadlock-fixes: relock candidate

reflink-inline-fixes holds two upstream fixes to copy_inline_to_page():
the folio tail zeroed at the wrong offset (reproduced with large folios
on 6.18) and the delalloc bits left set when decompression fails.  The
patches are exported from the candidate refs/cand/reflink/mn/reflink-inline-fixes
(be33f3b2272b), not from a lane branch: the topic is not promoted, so a
re-export with scripts/export.sh before promotion drops this directory.

deadlock-fixes/TESTING.md: the candidate relock commit placed after patch
2 (it fixes patch 2's double unlock), the v1 hang it went through on 6.18,
and the -EAGAIN analysis.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3 weeks agoRe-export onto the rebased misc-next lane (base 99bc6552614f)
Zygo Blaxell [Sun, 25 Oct 2026 07:16:15 +0000 (03:16 -0400)]
Re-export onto the rebased misc-next lane (base 99bc6552614f)

misc-next was rebased onto btrfs/for-next 0713fa6a4356 plus the marker
commit. The new lane tips are testing e8519e93c9a2 and zb64 315863b5a610.
Every series is re-exported against that base.

- raid56-scrub-replace is 0/12. Filipe Manana's -EIO fix (ec0795c045c3)
  is in the base now. The cover drops the paragraph about it and
  renumbers its patch references; TESTING.md gets a numbering note.
- README: the base, the testing and zb64 tips, every branch tip, and the
  scrub-replace row. The "git am --skip Filipe's patch" note is dropped.
- stripe-alloc cover: the base hash in its prerequisites section.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

3 weeks agostripe-alloc: REORDER-PLAN rev 3 and the reorder engine
Zygo Blaxell [Fri, 23 Oct 2026 17:38:57 +0000 (13:38 -0400)]
stripe-alloc: REORDER-PLAN rev 3 and the reorder engine

o-sord's revision 3 of the subtopic reorder: data-only reclaim in S3,
extended to metadata block groups by a new S5 commit, order S1 S2 S3 S4
S6a S5 S6b.  Candidates refs/tmp/o-sord/p16 (misc-next, 164 commits) and
q16 (6.18, 180) build W=1-clean at every boundary and at every commit
from S3 on; final trees equal the lane tips plus the documented
residual.  scripts/sa-reorder/ holds the engine, inputs, logs and
checks that regenerate them.  Both are on the pre-rebase bases and get
replayed after the 2026-10-23 lane rebase.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes: no stable tag on the after-abort write guards
Zygo Blaxell [Fri, 23 Oct 2026 01:57:25 +0000 (21:57 -0400)]
read-repair-fixes: no stable tag on the after-abort write guards

Zygo: keep patches 3 and 4 (read repair and scrub write-back after an
abort) but without Cc stable while upstream revises the rule.  Message
change only; the trees are unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes 0/4, raid56-fixes 0/12: leave raid56 as upstream after an abort
Zygo Blaxell [Fri, 23 Oct 2026 01:42:01 +0000 (21:42 -0400)]
read-repair-fixes 0/4, raid56-fixes 0/12: leave raid56 as upstream after an abort

Zygo: the "no writes after abort" rule is less strict than assumed and
is under revision upstream, so raid56 keeps upstream behavior.  The
parity-scrub check (read-repair-fixes 5) and the RMW write-back check
(raid56-fixes 12) are dropped from both lanes; the covers, README and
TESTING.md follow.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes 0/5, raid56-fixes 0/13: export the after-abort fixes
Zygo Blaxell [Thu, 22 Oct 2026 22:48:45 +0000 (18:48 -0400)]
read-repair-fixes 0/5, raid56-fixes 0/13: export the after-abort fixes

s-repair's follow-up is promoted on both lanes.  read-repair-fixes gains
patches 2-5: the repair bio searches the same csum root as the read it
repairs, and neither read repair nor scrub writes after a transaction
abort.  Its cover is rewritten for the series.  raid56-fixes gains the
RMW write-back guard as patch 12; the DO-NOT-UPSTREAM knob stays last,
as 13.  TESTING.md marks the rows promoted, and the README has the new
rows and heads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes: TESTING: s-repair P4 split into P4a (scrub) and P4b (raid56)
Zygo Blaxell [Thu, 22 Oct 2026 22:41:50 +0000 (18:41 -0400)]
read-repair-fixes: TESTING: s-repair P4 split into P4a (scrub) and P4b (raid56)

raid56 patches carry no Cc stable, so the scrub patch is split: P4a
scrub.c (Cc 6.6+), P4b raid56.c parity scrub (no Cc).  New candidate
SHAs; tip trees identical to the tested ones; P4a alone W=1-clean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes: TESTING: s-repair follow-up rows (candidate, not promoted)
Zygo Blaxell [Thu, 22 Oct 2026 22:34:48 +0000 (18:34 -0400)]
read-repair-fixes: TESTING: s-repair follow-up rows (candidate, not promoted)

Read repair after an abort (repair bbio inherits csum_search_commit_root)
and no repair or scrub writes after an abort or shutdown:
refs/cand/s-repair/{mn,618}/{topic,zb64}, plus R56 on raid56-fixes.
New arms abortgood, shutdown, scrubabort (lane FAIL, candidate PASS on
both lanes), dirtyabort (FAIL everywhere, open for Zygo), and the
candidates' regression runs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes: add the stable markup
Zygo Blaxell [Thu, 22 Oct 2026 18:06:57 +0000 (14:06 -0400)]
read-repair-fixes: add the stable markup

The patch now carries "Cc: stable@vger.kernel.org # 6.6+" (the bug is
in every kernel since v6.3).  Re-exported; TESTING.md and the README
have the new lane heads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoread-repair-fixes: new series (repair bio end_io for submit failures)
Zygo Blaxell [Thu, 22 Oct 2026 12:15:25 +0000 (08:15 -0400)]
read-repair-fixes: new series (repair bio end_io for submit failures)

The read-repair NULL end_io fix is added to both lanes as a new topic.
It is exported here with a 0/1 cover.  TESTING.md (m-reloc's abort arm,
q-repair's fail_function, repair and corrupt-recovery rows) now records
the lane SHAs, and the README has the new row, the ORDER entry and the
new heads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md read-repair-fixes: inject arm, repair and corrupt-recovery regression...
Zygo Blaxell [Thu, 22 Oct 2026 12:11:37 +0000 (08:11 -0400)]
TESTING.md read-repair-fixes: inject arm, repair and corrupt-recovery regression (q-repair)

Still a candidate (refs/cand/m-reloc-repair), not promoted.  fail_function
inject arm on ehive: misc-next and 6.18 lanes FAIL with the NULL end_io
oops, both candidates PASS 3/3 with every injection inside a repair
lookup.  Repair control and corrupt-recovery-test.sh PASS on all four
kernels; W=1 adds no warnings.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agostripe-alloc: export patches 115-116 (reclaim backoff) and describe them
Zygo Blaxell [Thu, 22 Oct 2026 11:31:13 +0000 (07:31 -0400)]
stripe-alloc: export patches 115-116 (reclaim backoff) and describe them

f-oct's reclaim backoff is promoted on both lanes: 115 backs off reclaim
after repeated ENOSPC, and 116 lets queued exclusive operations in
between reclaimed groups.  The cover's subtopic 3 describes both, with
the traced wait times.  TESTING.md marks them promoted, REORDER-PLAN.md
places them in subtopic 3, and the README and post-merge fixup follow the
new heads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md stripe-alloc: reclaim-futile-test row, 6.18 candidate result (f-oct)
Zygo Blaxell [Thu, 22 Oct 2026 10:39:50 +0000 (06:39 -0400)]
TESTING.md stripe-alloc: reclaim-futile-test row, 6.18 candidate result (f-oct)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agolow-space: export patch 3 (merge-reservation flush) and describe it in the cover
Zygo Blaxell [Thu, 22 Oct 2026 09:13:48 +0000 (05:13 -0400)]
low-space: export patch 3 (merge-reservation flush) and describe it in the cover

m-reloc's "btrfs: relocation: flush fully before failing a merge
reservation" is promoted on both lanes.  The cover becomes 0/3 with a
paragraph on the fix, TESTING.md marks it promoted with the lane SHAs,
and the README has the new heads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md read-repair-fixes: new candidate topic (NULL end_io on a failed repair...
Zygo Blaxell [Thu, 22 Oct 2026 09:11:01 +0000 (05:11 -0400)]
TESTING.md read-repair-fixes: new candidate topic (NULL end_io on a failed repair submit)

The candidate, refs/cand/m-reloc-repair, is not promoted.
read-repair-submit-fail-test.sh abort arm: FAIL (oops) on the misc-next
and 6.18 lanes, PASS on both candidates.  The repair control and W=1 are
clean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md low-space: candidate patch 3 flush-mode rationale, refs on the new lanes...
Zygo Blaxell [Thu, 22 Oct 2026 09:11:01 +0000 (05:11 -0400)]
TESTING.md low-space: candidate patch 3 flush-mode rationale, refs on the new lanes, abort 2 open task

The candidate keeps FLUSH_EVICT.  44d354abf33e chose FLUSH_LIMIT only to
avoid -EINTR.  No existing mode runs delayed refs and commits without
delalloc or steal.  The deadlock analysis is now in the commit message.

The refs are rebuilt with the new message: mn topic 9971f3238ceb, zb64
49c28497f96e on 2550544f9c6e; 618 topic 19e3ca085180, zb64 829ecda8abee
on 7626a4a10198.  The code is unchanged apart from its comment.

Abort 2 (the commit-time delayed-ref ENOSPC after a relocation's burst of
frees) is written up as an open task with the numbers from combo-compact
and cand-compact.  It is not fixed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md stripe-alloc: reclaim-futile-test row (reclaimbackoff candidate, f-oct)
Zygo Blaxell [Thu, 22 Oct 2026 08:04:12 +0000 (04:04 -0400)]
TESTING.md stripe-alloc: reclaim-futile-test row (reclaimbackoff candidate, f-oct)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoREADME: stripe-alloc cover rewritten
Zygo Blaxell [Thu, 22 Oct 2026 07:56:18 +0000 (03:56 -0400)]
README: stripe-alloc cover rewritten

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agostripe-alloc: rewrite the cover letter by subtopic
Zygo Blaxell [Thu, 22 Oct 2026 07:56:11 +0000 (03:56 -0400)]
stripe-alloc: rewrite the cover letter by subtopic

The old cover was written for 65 patches and marked stale.  The new one
describes the series in the order the branch will be rebuilt: activation
and the no-RMW core, the allow-rmw property, stripe-only reclaim, the two
fsync implementations, stripe_meta, and debug code not for merging.
Patch numbers stay the current export numbers until the rebuild.

The stripe_meta section says why it may be dropped (the four verdicts
from STRIPE-META-CLAIMS.md, with what was not measured), what dropping
costs users and the raid1/raid1c3 metadata alternative.  Statements that
are no longer true of the code are corrected (system chunks and log
blocks are covered, NOCOW runs live only under allow_rmw, metadata run
waits are unbounded).  Numbers that could not be traced to TESTING.md,
a commit message or the forensics were removed.  Body 1549 -> 818 lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agostripe-alloc: REORDER-PLAN.md, plan to rebuild the series by subtopic
Zygo Blaxell [Thu, 22 Oct 2026 07:45:38 +0000 (03:45 -0400)]
stripe-alloc: REORDER-PLAN.md, plan to rebuild the series by subtopic

Classifies all 114 patches into Zygo's six subtopics (activation and
no-RMW core, allow-rmw, stripe-only reclaim, fsync implementations,
stripe_meta, debug).  A pure reorder does not work: the trial replays
the first 13 patches and stops at old 24, which edits code from later
subtopics.  Lists the eleven dependencies (D1-D11), the splits that
remove them, the two decisions needed first (D1: split old 14's
private-run machinery; D4: whether park/run stats are observability or
debug), the trial refs and the per-subtopic summaries.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agostripe-alloc: STRIPE-META-CLAIMS.md, evidence for dropping stripe_meta
Zygo Blaxell [Thu, 22 Oct 2026 06:17:21 +0000 (02:17 -0400)]
stripe-alloc: STRIPE-META-CLAIMS.md, evidence for dropping stripe_meta

Verdicts on Zygo's four claims for dropping stripe_meta from the series,
from existing benchmarks, the series history and the mailing list (no
new tests): not faster than raid1 (holds in practice), trapping eats
the space saving (holds on aged and full filesystems), high complexity
and stall/deadlock risk (holds), overlap with Johannes Thumshirn's
zoned ENOSPC work (same problems, different mechanisms, no code
collision).  Ends with the cost of dropping it: 14 patches, about
+1211/-130 lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md low-space: candidate patch 3 (relocation merge refill), reloc-merge-enospc...
Zygo Blaxell [Thu, 22 Oct 2026 05:32:28 +0000 (01:32 -0400)]
TESTING.md low-space: candidate patch 3 (relocation merge refill), reloc-merge-enospc-test results

Candidate refs/cand/m-reloc (not promoted): merge_reloc_root() refills
with FLUSH_EVICT instead of FLUSH_LIMIT.  Lane FAIL 3 of 3 reclaim arms
(merge_reloc_roots -28); candidate: no merge failure in 6 arms, 5 merge
refills served only by FLUSH_DELAYED_REFS.  A second fill-edge abort
(commit-time delayed refs after a relocation's metadata burst) remains
and is listed as open.  Cover letter: candidate note.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md stripe-alloc: question 7, btrfs-progs --enqueue loses briefly freed exclus...
Zygo Blaxell [Wed, 21 Oct 2026 23:27:39 +0000 (19:27 -0400)]
TESTING.md stripe-alloc: question 7, btrfs-progs --enqueue loses briefly freed exclusive operations (f-oct)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoresize-missing: cover letter and TESTING.md
Zygo Blaxell [Wed, 21 Oct 2026 23:16:11 +0000 (19:16 -0400)]
resize-missing: cover letter and TESTING.md

The cover explains the problem (a smaller replacement can't be used for
a missing device), what the patch allows and refuses, the two bugs fixed
in Qu's 2019 version and the test results.  TESTING.md records n-rsz's
runs on both lanes (forensics/n-rsz-20261021, harness 9a2358c).  The
patch is re-exported as 1/1 now that the series has a cover.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoREADME: 19 series
Zygo Blaxell [Wed, 21 Oct 2026 21:47:43 +0000 (17:47 -0400)]
README: 19 series

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoresize-missing: export the new series and add it to the README
Zygo Blaxell [Wed, 21 Oct 2026 21:47:30 +0000 (17:47 -0400)]
resize-missing: export the new series and add it to the README

topics/resize-missing is one patch, Qu Wenruo's 2019 "shrink a missing
device" resize change, rewritten: growth of a missing device is now
rejected for +size and absolute sizes too, and every rejection returns.
It merges into misc-next/testing 723245adb5c1 with no conflict, so
fixups/post-merge.patch is unchanged.  No cover letter yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoraid56-scrub-replace: read-only scrub fix, scrub error device names, parity-scrub...
Zygo Blaxell [Wed, 21 Oct 2026 11:10:20 +0000 (07:10 -0400)]
raid56-scrub-replace: read-only scrub fix, scrub error device names, parity-scrub repairs counted (patches 9, 12, 13)

Promoted on both lanes 2026-10-21.  The read-only fix goes before the
raid6 rebuild, so the former patches 9-10 are now 10-11.  misc-next
raid56-scrub-replace 9970d7e44abf, testing 318dec80d29b, zb64
9c343098c7ea.  Cover describes patches 9-13.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md raid56-scrub-replace: candidates R (read-only scrub writes nothing), L...
Zygo Blaxell [Wed, 21 Oct 2026 11:07:16 +0000 (07:07 -0400)]
TESTING.md raid56-scrub-replace: candidates R (read-only scrub writes nothing), L (log lines name the read device), U (count the parity scrub's data repairs)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md stripe-alloc: rows for patches 113 (reclaim debug lines) and 114 (stripe_r...
Zygo Blaxell [Wed, 21 Oct 2026 10:33:25 +0000 (06:33 -0400)]
TESTING.md stripe-alloc: rows for patches 113 (reclaim debug lines) and 114 (stripe_reclaim_now) (f-oct)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agostripe-alloc: export patches 113-114 (reclaim debug log, stripe_reclaim_now)
Zygo Blaxell [Wed, 21 Oct 2026 07:29:19 +0000 (03:29 -0400)]
stripe-alloc: export patches 113-114 (reclaim debug log, stripe_reclaim_now)

Promoted on both lanes 2026-10-21: misc-next stripe-alloc bcadbfc8ac02,
testing 12c700ee1440, zb64 aa5e2b4c73c0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md dev-stats: devstats-persist results, #5 negative control, W=1, regressions
Zygo Blaxell [Wed, 21 Oct 2026 05:01:37 +0000 (01:01 -0400)]
TESTING.md dev-stats: devstats-persist results, #5 negative control, W=1, regressions

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agodev-stats: new one-patch series (write dirty device stats when no transaction is...
Zygo Blaxell [Wed, 21 Oct 2026 05:00:37 +0000 (01:00 -0400)]
dev-stats: new one-patch series (write dirty device stats when no transaction is running)

Promoted on both lanes 2026-10-21 as a new topic, droppable on its own:
misc-next topic 55ade1cf195c, testing 5eed033e8449, zb64 e2c9218febf9.
Cover letter states the behaviour change and that reviewers may reject it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoraid56-fixes: export patch 11 (RMW writes back repaired sectors); knob is now patch 12
Zygo Blaxell [Wed, 21 Oct 2026 02:28:51 +0000 (22:28 -0400)]
raid56-fixes: export patch 11 (RMW writes back repaired sectors); knob is now patch 12

Promoted on both lanes 2026-10-21: misc-next raid56-fixes f7e693d10be2,
testing bbc86ab46e84, zb64 3e1390a9de03.  The integration merge with
stripe-alloc conflicts where both insert helpers before rmw_rbio(); the
post-merge fixup carries the resolution (README item 1).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec

4 weeks agoTESTING.md raid56-fixes: the detector kernel ran 5WP and 6PW only
Zygo Blaxell [Wed, 21 Oct 2026 02:25:13 +0000 (22:25 -0400)]
TESTING.md raid56-fixes: the detector kernel ran 5WP and 6PW only

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6wxsJn1aY7PwnhgXdz6Ec