* [PATCH AUTOSEL 6.18-5.10] blk-cgroup: fix leaks and online flag on radix_tree_insert failure
[not found] <20260831133314.4125787-1-sashal@kernel.org>
@ 2026-08-31 13:25 ` Sasha Levin
2026-08-31 13:27 ` [PATCH AUTOSEL 6.18-6.12] iomap: don't make REQ_POLLED imply REQ_NOWAIT Sasha Levin
2026-08-31 13:28 ` [PATCH AUTOSEL 6.18-6.1] blk-cgroup: protect iterating blkgs with blkcg->lock in blkcg_print_stat() Sasha Levin
2 siblings, 0 replies; 3+ messages in thread
From: Sasha Levin @ 2026-08-31 13:25 UTC (permalink / raw)
To: patches, stable
Cc: Tao Cui, Tejun Heo, Jens Axboe, Sasha Levin, josef, cgroups,
linux-block, linux-kernel
From: Tao Cui <cuitao@kylinos.cn>
[ Upstream commit dbbca20764382b4d411ec2918f4e278ffe547acc ]
When radix_tree_insert() fails in blkg_create(), the error path has two
issues:
1. blkg->online is set to true unconditionally, even when the blkg was
never fully inserted. Move the assignment inside the success block.
2. The error path calls blkg_put() without first calling
percpu_ref_kill(). Because the refcount is still in percpu mode,
percpu_ref_put() only does this_cpu_sub() without checking for zero,
so blkg_release() is never triggered. This permanently leaks the
blkg memory, its percpu iostat, policy data, the parent blkg
reference, and the cgroup css reference — the latter preventing the
cgroup from ever being destroyed.
Fix by replacing blkg_put() with percpu_ref_kill(), matching the pattern
used in blkg_destroy().
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tao Cui <cuitao@kylinos.cn>
Link: https://patch.msgid.link/20260715132407.1469777-1-cui.tao@linux.dev
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
LLM Generated explanations, may be completely bogus:
# Stable Backport Analysis: `blk-cgroup: fix leaks and online flag on
radix_tree_insert failure`
**Local tree:** `v6.18.44` (Linux 6.18.44 stable)
---
## PHASE 1: COMMIT MESSAGE FORENSICS
**Step 1.1 — Subject line**
Record: `[blk-cgroup]` `[fix]` — fixes resource leaks and incorrect
`online` flag when `radix_tree_insert()` fails in `blkg_create()`.
**Step 1.2 — Tags**
Record:
- **Acked-by:** Tejun Heo `<tj@kernel.org>` (cgroup/block-cgroup
maintainer)
- **Signed-off-by:** Tao Cui `<cuitao@kylinos.cn>` (author)
- **Signed-off-by:** Jens Axboe `<axboe@kernel.dk>` (block layer
maintainer)
- **Link:**
https://patch.msgid.link/20260715132407.1469777-1-cui.tao@linux.dev
- No Fixes:, Reported-by:, Tested-by:, Reviewed-by:, or Cc: stable tags
- (Ignoring pipeline-added Signed-off-by: Sasha Levin per instructions)
**Step 1.3 — Body analysis**
Record:
- **Bug:** When `radix_tree_insert()` fails in `blkg_create()`, two
errors occur:
1. `blkg->online = true` is set even though the blkg was never
inserted into the tree.
2. Error path calls `blkg_put()` without `percpu_ref_kill()`. While
the refcount is still in percpu mode, `percpu_ref_put()` only
decrements a per-CPU counter and never checks for zero, so
`blkg_release()` is never called.
- **Symptom/failure mode:** Permanent leak of blkg memory, percpu
iostat, policy data, parent blkg reference, and cgroup css reference —
the css leak prevents the cgroup from ever being destroyed.
- **Root cause:** Wrong teardown primitive on the error path;
`blkg_destroy()` correctly uses `percpu_ref_kill()`.
**Step 1.4 — Hidden bug fix?**
Record: No — this is an explicit bug fix, not disguised cleanup.
---
## PHASE 2: DIFF ANALYSIS
**Step 2.1 — Inventory**
Record:
- **Files:** `block/blk-cgroup.c` only (+2 / −2 lines, 4 lines touched)
- **Function:** `blkg_create()`
- **Scope:** Single-file, surgical fix
**Step 2.2 — Code flow change**
Record:
- **Hunk 1:** `blkg->online = true` moved inside the `if (likely(!ret))`
success block.
- Before: online set unconditionally after insert attempt.
- After: online only set when insert succeeds.
- **Hunk 2:** Error path changed from `blkg_put(blkg)` to
`percpu_ref_kill(&blkg->refcnt)`.
- Before: percpu-mode put never triggers release callback.
- After: switches to atomic mode and triggers `blkg_release()` →
`__blkg_release()` → `css_put()` + `blkg_free()`.
**Step 2.3 — Bug mechanism**
Record: **Reference counting / resource leak fix.** Category (a) error-
path leak + (g) logic correctness (online flag). The percpu_ref
lifecycle requires `percpu_ref_kill()` before the final drop can trigger
the release function — documented in `include/linux/percpu-refcount.h`
lines 19–24.
**Step 2.4 — Fix quality**
Record: Obviously correct — mirrors `blkg_destroy()` at line 568.
Minimal change. Very low regression risk; only affects the rare
`radix_tree_insert()` failure path.
---
## PHASE 3: GIT HISTORY INVESTIGATION
**Step 3.1 — Blame**
Record: Buggy lines in this tree all from `5d324e5159d9e` (v6.18 merge,
Nov 2025). Same pattern present in `v6.12` and `v6.17` per `git show`.
**Step 3.2 — Fixes: tag**
Record: Not applicable — no Fixes: tag in commit message.
**Step 3.3 — Related file history**
Record:
- `93383b6681074` — "wait for blkcg cleanup before initializing new
disk" — reduces `-EEXIST` from `radix_tree_insert()` during disk
rebind, but does not fix the broken error path when insert still
fails.
- `5e5b7f2ef8549` — UAF fix in `__blkcg_rstat_flush()` (related
subsystem, separate issue).
- Fix commit on master: `dbbca20764382` (Jul 15, 2026); **not** an
ancestor of current HEAD (`merge-base` exit 1).
**Step 3.4 — Author context**
Record: Tao Cui; Acked-by Tejun Heo (blk-cgroup/cgroup maintainer). No
other Tao Cui commits in this tree's `block/blk-cgroup.c` history.
**Step 3.5 — Dependencies**
Record: Standalone — no series dependencies, no prerequisite commits
required. Self-contained 4-line change.
---
## PHASE 4: MAILING LIST AND EXTERNAL RESEARCH
**Step 4.1 — Original discussion**
Record:
- `b4 dig -c dbbca20764382`:
https://patch.msgid.link/20260715132407.1469777-1-cui.tao@linux.dev
- Series: v4 only (no v1–v3 in b4 results; v4 is the applied version)
- No NAKs found in saved mbox
- No explicit Cc: stable nomination in thread headers
**Step 4.2 — Reviewers**
Record: `b4 dig -w` CC'd: tj@kernel.org, axboe@kernel.dk,
josef@toxicpanda.com, cgroups@vger.kernel.org, linux-
block@vger.kernel.org. Tejun Heo Acked-by.
**Step 4.3 — Bug report**
Record: No external bug report or syzbot link. Bug identified via code
review of percpu_ref lifecycle.
**Step 4.4 — Related patches**
Record: Complementary to `93383b6681074` (reduces trigger frequency) but
independently needed for correct error handling.
**Step 4.5 — Stable list**
Record: No stable@vger.kernel.org discussion found for this specific
fix.
---
## PHASE 5: CODE SEMANTIC ANALYSIS
**Step 5.1 — Key functions**
Record: `blkg_create()` modified; related: `blkg_destroy()`,
`blkg_release()`, `__blkg_release()`, `blkg_free()`.
**Step 5.2 — Callers**
Record: `blkg_create()` called from:
- `blkg_lookup_create()` — I/O hot path via `blkg_tryget_closest()` →
`bio_assoc_blkcg()` (line 2113)
- `blkg_conf_prep()` — cgroup sysfs configuration (uses
`radix_tree_preload`)
- `blkcg_init_disk()` — disk initialization (uses `radix_tree_preload`)
`blkg_lookup_create()` does **not** call `radix_tree_preload()`, so
`-ENOMEM` from `radix_tree_insert()` is reachable under memory pressure.
**Step 5.3 — Callees**
Record: On failure path after fix: `percpu_ref_kill()` →
`blkg_release()` → `__blkcg_rstat_flush()` + `call_rcu(__blkg_release)`
→ `css_put()` + `blkg_free()` → `blkg_free_workfn()` releases parent
ref, policy data, queue ref, percpu iostat.
**Step 5.4 — Reachability**
Record: Reachable from block I/O path when `CONFIG_BLK_CGROUP` is
enabled and a new blkg must be created for a cgroup/disk pair. Userspace
cgroup management can also trigger via `blkg_conf_prep()`. Unprivileged
users can trigger via I/O in their cgroup.
**Step 5.5 — Similar patterns**
Record: `blkg_destroy()` at line 568 already uses
`percpu_ref_kill(&blkg->refcnt)` — fix aligns error path with
established pattern. `include/linux/percpu-refcount.h` documents that
`percpu_ref_put()` does not check for zero before `percpu_ref_kill()`.
---
## PHASE 6: CROSS-REFERENCE WITH LOCAL TREE
**Step 6.1 — Buggy code present?**
Record: **YES.** Current tree at lines 436 and 443:
```436:444:block/blk-cgroup.c
blkg->online = true;
spin_unlock(&blkcg->lock);
if (!ret)
return blkg;
/* @blkg failed fully initialized, use the usual release path */
blkg_put(blkg);
return ERR_PTR(ret);
```
Bug present since at least v6.12 in this repository's history.
**Step 6.2 — Backport complications**
Record: Trivial change; `git apply --check` on upstream patch fails only
because stable has `err_put_css:` label that mainline parent lacks
(context line difference below the hunk). The three actual changed lines
apply without modification. Expected difficulty: **minor context
adjustment, not rework**.
**Step 6.3 — Related fixes already present?**
Record: `93383b6681074` is present (reduces `-EEXIST` trigger). This
specific leak fix is **not** present.
---
## PHASE 7: SUBSYSTEM AND MAINTAINER CONTEXT
**Step 7.1 — Subsystem**
Record: **block/blk-cgroup** — CORE/IMPORTANT subsystem. Affects all
systems using cgroup v1/v2 block controller (`CONFIG_BLK_CGROUP`).
**Step 7.2 — Activity**
Record: Active maintenance in 6.18.y — recent fixes include UAF
(`5e5b7f2ef8549`), disk reference leak (`b3e005f16cd98`), blkcg cleanup
wait (`93383b6681074`).
---
## PHASE 8: IMPACT AND RISK ASSESSMENT
**Step 8.1 — Who is affected**
Record: Systems with `CONFIG_BLK_CGROUP` enabled — container hosts
(Kubernetes, Docker, systemd cgroups), cloud VMs, any workload using
block I/O cgroup controller.
**Step 8.2 — Trigger conditions**
Record:
- `radix_tree_insert()` returns error (`-ENOMEM` most likely in
`blkg_lookup_create()` without preload; `-EEXIST` possible in races
despite `93383b6681074`)
- Requires blkg creation for a new cgroup/disk pair
- Unprivileged cgroup users can trigger via I/O; cgroup admin via sysfs
- Not every boot — requires memory pressure or specific race — but
consequences are permanent
**Step 8.3 — Failure mode severity**
Record:
- **Permanent memory/resource leak** (blkg, iostat, policy data)
- **Cgroup css reference leak → cgroup cannot be destroyed** —
functional breakage for container lifecycle
- **Incorrect online flag** — minor (e.g., `blkcg_print_one_stat()` at
line 1190 may process a non-inserted blkg)
- Severity: **HIGH** (resource leak with cgroup destruction blocked; not
a crash but serious operational impact)
**Step 8.4 — Risk-benefit**
Record:
- **Benefit:** HIGH — prevents unrecoverable resource leaks and stuck
cgroups
- **Risk:** VERY LOW — 4-line change, matches existing `blkg_destroy()`
pattern, only affects error path
- **Ratio:** Strongly favors backport
---
## PHASE 9: FINAL SYNTHESIS
**Step 9.1 — Evidence summary**
**FOR:**
- Real, verifiable resource leak on error path
- Cgroup css leak prevents cgroup destruction — serious for production
container workloads
- Small, surgical, maintainer-acked fix
- Buggy code confirmed present in v6.18.44
- Matches documented percpu_ref semantics and existing `blkg_destroy()`
pattern
- Reachable from common I/O and cgroup configuration paths
**AGAINST:**
- Rare trigger (radix_tree_insert failure)
- No syzbot/user crash report
- Upstream patch needs trivial context adjustment for `err_put_css:`
label
**UNRESOLVED:**
- No quantitative data on how often `radix_tree_insert()` fails in
production
**Step 9.2 — Stable rules checklist**
1. Obviously correct and tested? **PASS** — mechanism verified against
`percpu-refcount.h` and `blkg_destroy()`; Acked-by cgroup maintainer
2. Fixes real bug affecting users? **PASS** — permanent leak + cgroup
destruction blocked
3. Important issue? **PASS** — HIGH severity resource leak affecting
cgroup lifecycle
4. Small and contained? **PASS** — 4 lines, one function, one file
5. No new features or APIs? **PASS**
6. Can apply to local tree? **PASS** — trivial manual apply (minor
context difference only)
**Step 9.3 — Exception categories**
Record: Not applicable (standard bug fix, not device ID/quirk/build/doc
exception).
**Step 9.4 — Decision rationale**
This commit fixes a genuine error-path bug in `blkg_create()` where
failed `radix_tree_insert()` leaves resources permanently leaked because
`blkg_put()` cannot trigger `blkg_release()` while the percpu refcount
is still in percpu mode. The css reference leak prevents cgroup
destruction — a significant operational issue for any system using block
cgroups. The fix is minimal, follows the established `blkg_destroy()`
pattern, is acked by the cgroup maintainer, and the buggy code is
present in this v6.18.44 tree without the fix.
---
## Verification
- [Phase 1] `git show dbbca20764382`: parsed subject, tags, body; Acked-
by Tejun Heo confirmed
- [Phase 2] Read `block/blk-cgroup.c` lines 371–452, 524–569, 1190; read
`include/linux/percpu-refcount.h` lines 19–24, 147–150, 326–352
- [Phase 2] Confirmed `blkg_put()` → `percpu_ref_put()` in `block/blk-
cgroup.h` line 336–338
- [Phase 3] `git describe HEAD`: v6.18.44; `git blame -L 419,452`: buggy
code from 5d324e5159d9e
- [Phase 3] `git merge-base --is-ancestor dbbca20764382 HEAD`: exit 1 —
fix NOT in tree
- [Phase 3] `git show v6.12/v6.17:v6.18:block/blk-cgroup.c`: same buggy
pattern back to v6.12
- [Phase 3] `93383b6681074` present in tree; complementary, not
substitute
- [Phase 4] `b4 dig -c dbbca20764382`: lore URL found; v4 series
- [Phase 4] `b4 dig -w`: maintainers CC'd; Tejun Heo Acked-by
- [Phase 4] `b4 dig -m /tmp/blkcg_fix.mbox`: no Cc: stable in thread
- [Phase 5] Grep callers: `blkg_create` at lines 510, 936, 1526;
`blkg_lookup_create` at 2113
- [Phase 5] Confirmed `blkg_lookup_create` does not use
`radix_tree_preload` (only `blkg_conf_prep` and `blkcg_init_disk` do)
- [Phase 6] Buggy code at lines 436, 443 confirmed in working tree
- [Phase 6] `git apply --check` fails on context below hunk
(`err_put_css:` vs `err_free_blkg:`); actual 3 changed lines are
identical — trivial backport
- [Phase 8] Failure mode: permanent leak + cgroup destruction blocked;
severity HIGH
**YES**
block/blk-cgroup.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index 812f60905a708..9aa3b6ae81ca8 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -432,15 +432,15 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg, struct gendisk *disk,
blkg->pd[i]->online = true;
}
}
+ blkg->online = true;
}
- blkg->online = true;
spin_unlock(&blkcg->lock);
if (!ret)
return blkg;
/* @blkg failed fully initialized, use the usual release path */
- blkg_put(blkg);
+ percpu_ref_kill(&blkg->refcnt);
return ERR_PTR(ret);
err_put_css:
--
2.53.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* [PATCH AUTOSEL 6.18-6.12] iomap: don't make REQ_POLLED imply REQ_NOWAIT
[not found] <20260831133314.4125787-1-sashal@kernel.org>
2026-08-31 13:25 ` [PATCH AUTOSEL 6.18-5.10] blk-cgroup: fix leaks and online flag on radix_tree_insert failure Sasha Levin
@ 2026-08-31 13:27 ` Sasha Levin
2026-08-31 13:28 ` [PATCH AUTOSEL 6.18-6.1] blk-cgroup: protect iterating blkgs with blkcg->lock in blkcg_print_stat() Sasha Levin
2 siblings, 0 replies; 3+ messages in thread
From: Sasha Levin @ 2026-08-31 13:27 UTC (permalink / raw)
To: patches, stable
Cc: Christoph Hellwig, Christian Brauner, Sasha Levin, axboe,
linux-xfs, linux-fsdevel, linux-kernel, linux-block
From: Christoph Hellwig <hch@lst.de>
[ Upstream commit 47f28b493daf9049307494689e515205f1c377af ]
As described in commit 2bc057692599 ("block: don't make REQ_POLLED imply
REQ_NOWAIT"), which fixed the same issue for the block device node, there
are valid cases to poll for I/O completion without REQ_NOWAIT.
Additionally, sing REQ_NOWAIT for file system writes is currently not
supported as file systems writes are not idempotent and would need a
retry of just the bio and not the entire operation to be fully supported.
Switch iomap to set REQ_POLLED and remove the now unused bio_set_polled
helper.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://patch.msgid.link/20260518062917.506483-1-hch@lst.de
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
LLM Generated explanations, may be completely bogus:
# Stable Backport Analysis: `iomap: don't make REQ_POLLED imply
REQ_NOWAIT`
**Local tree:** Linux **6.18.43** (`v6.18.43-1-gc7f0dac02d232`)
---
## PHASE 1: COMMIT MESSAGE FORENSICS
### Step 1.1: PARSE THE SUBJECT LINE
**Record:** `[iomap] [don't make / fix] [REQ_POLLED should not imply
REQ_NOWAIT on filesystem DIO bios]`
### Step 1.2: PARSE ALL COMMIT MESSAGE TAGS
**Record:**
- **Fixes:** — none
- **Reported-by:** — none
- **Tested-by:** — none
- **Reviewed-by:** — none
- **Acked-by:** — none
- **Link:** https://patch.msgid.link/20260518062917.506483-1-hch@lst.de
- **Cc: stable:** — none (expected for manual review)
- **Signed-off-by:** Christoph Hellwig `<hch@lst.de>`, Christian Brauner
`<brauner@kernel.org>` (merge commit)
- **Notable:** References upstream commit `2bc057692599` (block-layer
companion fix). No syzbot, no user bug reports.
### Step 1.3: ANALYZE THE COMMIT BODY TEXT
**Record:**
- **Bug:** `bio_set_polled()` propagates `REQ_NOWAIT` onto bios when
`IOCB_NOWAIT` is set. For iomap filesystem DIO this is incorrect —
filesystem writes are not idempotent at the bio level and cannot be
retried by re-submitting just the bio.
- **Symptom:** Polled filesystem DIO (e.g. io_uring
`IORING_SETUP_IOPOLL` on xfs/ext4 O_DIRECT) can hit spurious `-EAGAIN`
from the block layer, or fail to make progress — same class of bug
fixed for raw block devices in 2023.
- **Root cause:** iomap reused `bio_set_polled()` which couples
`REQ_POLLED` with conditional `REQ_NOWAIT`; block/fops.c was already
fixed to decouple them, but iomap was not.
- **Version info:** Commit dated 2026-05-18; not yet in this 6.18.43
tree.
### Step 1.4: DETECT HIDDEN BUG FIXES
**Record:** Not disguised — this is an explicit correctness fix, though
small. The removal of `bio_set_polled()` is cleanup after the last
caller is gone.
---
## PHASE 2: DIFF ANALYSIS
### Step 2.1: INVENTORY THE CHANGES
**Record:**
- `fs/iomap/direct-io.c`: 1 line changed (`bio_set_polled` →
`bio->bi_opf |= REQ_POLLED`)
- `include/linux/bio.h`: 14 lines removed (`bio_set_polled()` helper +
comment)
- **Functions modified:** `iomap_dio_submit_bio()`; `bio_set_polled()`
removed
- **Scope:** Single-subsystem, 2 files, ~16 lines total — surgical fix
### Step 2.2: UNDERSTAND THE CODE FLOW CHANGE
**Record:**
- **Hunk 1 (`iomap_dio_submit_bio`):** Before: for async HIPRI DIO, call
`bio_set_polled(bio, iocb)` which sets `REQ_POLLED` and also
`REQ_NOWAIT` when `IOCB_NOWAIT` is set. After: only `REQ_POLLED` is
set; `IOCB_NOWAIT` is handled separately at the iomap layer via
`IOMAP_NOWAIT` (line 654–655).
- **Hunk 2 (`bio.h`):** Remove now-dead `bio_set_polled()` helper (only
caller was iomap).
### Step 2.3: IDENTIFY THE BUG MECHANISM
**Record:**
- **Category:** Logic / correctness fix (incorrect flag propagation)
- **Mechanism:** `REQ_NOWAIT` on a bio causes the block layer to return
`-EAGAIN` instead of blocking on resource contention
(`__bio_queue_enter`, tag allocation in `blk-mq`). For filesystem DIO
through iomap, `IOCB_NOWAIT` is already translated to `IOMAP_NOWAIT`
for filesystem-level handling; passing `REQ_NOWAIT` to the block layer
is both unnecessary and harmful for writes.
### Step 2.4: ASSESS THE FIX QUALITY
**Record:**
- Obviously correct: mirrors the already-accepted block-layer fix
pattern in `block/fops.c`.
- Minimal: one-line functional change plus dead-code removal.
- **Regression risk:** Very low. Block device path already uses the same
pattern. `IOMAP_NOWAIT` continues to handle filesystem-level non-
blocking semantics.
---
## PHASE 3: GIT HISTORY INVESTIGATION
### Step 3.1: BLAME THE CHANGED LINES
**Record:** Shallow repository limits blame — all lines attribute to
`a112b91dd6349` (unrelated sunrpc backport). Verified current buggy code
exists at `fs/iomap/direct-io.c:77` and `include/linux/bio.h:688-693`.
Kernel.org history (via curl) shows iomap polled-IO support added in
`daa99c5a3319` (2023-08-01, Jens Axboe: "iomap: only set iocb->private
for polled bio"); block fix `2bc057692599` (2023-08-08) updated
`bio_set_polled()` but left iomap calling it.
### Step 3.2: FOLLOW THE FIXES: TAG
**Record:** No `Fixes:` tag. Referenced commit `2bc057692599` ("block:
don't make REQ_POLLED imply REQ_NOWAIT") exists as a git object in this
tree; `block/fops.c` already uses the decoupled pattern (`IOCB_NOWAIT`
and `REQ_POLLED` set independently). iomap was the remaining caller of
`bio_set_polled()`.
### Step 3.3: CHECK FILE HISTORY FOR RELATED CHANGES
**Record:** Shallow repo prevents meaningful `git log` on these files.
External kernel.org log confirms this is a standalone 1-patch fix (not
part of a series). Related prior fix: `2bc057692599` (block layer,
2023).
### Step 3.4: CHECK THE AUTHOR'S OTHER COMMITS
**Record:** Christoph Hellwig is the iomap maintainer. Christian Brauner
is VFS maintainer who applied the patch. Strong subsystem authority.
### Step 3.5: CHECK FOR DEPENDENT/PREREQUISITE COMMITS
**Record:** No prerequisites. Self-contained. Depends only on existing
`IOCB_HIPRI`/polled-IO infrastructure already present in 6.18.43. Commit
`47f28b493daf` is NOT in this tree (object not found via `git cat-
file`).
---
## PHASE 4: MAILING LIST AND EXTERNAL RESEARCH
### Step 4.1: FIND THE ORIGINAL PATCH DISCUSSION
**Record:** `b4 dig -c` could not run — commit not in local repo.
Fetched via spinics.net:
- URL: https://www.spinics.net/lists/linux-fsdevel/msg338671.html
- Single patch, no series revisions found
- CC'd: `axboe`, `linux-block`, `linux-fsdevel`, `linux-xfs`, `djwong`,
`brauner`
### Step 4.2: CHECK WHO REVIEWED THE PATCH
**Record:** CC list includes block maintainer (Axboe), XFS, fsdevel,
block lists. Brauner applied to `vfs-7.2.iomap` branch. No explicit
Reviewed-by in commit; no NAKs found.
### Step 4.3: SEARCH FOR THE BUG REPORT
**Record:** No bug report, syzbot, or crash trace. Bug identified by
code analysis and parity with the 2023 block-layer fix. Failure mode
inferred from block commit message: "repeated -EAGAIN submissions and
not make any progress."
### Step 4.4: CHECK FOR RELATED PATCHES AND SERIES
**Record:** Standalone 1/1 patch. Companion to `2bc057692599` (already
in stable block path).
### Step 4.5: CHECK STABLE MAILING LIST HISTORY
**Record:** Not searched (no stable nomination found in thread). Absence
of `Cc: stable` is not a negative signal per review guidelines.
---
## PHASE 5: CODE SEMANTIC ANALYSIS
### Step 5.1: IDENTIFY KEY FUNCTIONS IN THE DIFF
**Record:** `iomap_dio_submit_bio()`, `bio_set_polled()` (removed)
### Step 5.2: TRACE CALLERS
**Record:** `iomap_dio_submit_bio()` called from iomap DIO write/read
paths in `fs/iomap/direct-io.c`. Reachable via `iomap_dio_rw()` →
filesystem `read_iter`/`write_iter` on xfs, ext4, f2fs, gfs2, zonefs,
btrfs (partial). io_uring sets `IOCB_HIPRI` for `IORING_SETUP_IOPOLL`
(`io_uring/rw.c:891-895`) and may set `IOCB_NOWAIT` for nonblock issue
(`io_uring/rw.c:950-954`).
### Step 5.3: TRACE CALLEES
**Record:** After fix: `bio->bi_opf |= REQ_POLLED`, then `submit_bio()`
(or filesystem `submit_io` hook). Block layer checks `REQ_NOWAIT` in
`__bio_queue_enter()` → `bio_wouldblock_error()` → `-EAGAIN`.
### Step 5.4: FOLLOW THE CALL CHAIN
**Record:** Userspace io_uring IOPOLL → `IOCB_HIPRI` + possibly
`IOCB_NOWAIT` → `xfs_file_read_iter`/`ext4_file_write_iter` →
`iomap_dio_rw` → `iomap_dio_submit_bio` → block layer. **Reachable from
userspace** on common filesystems with `.iopoll = iocb_bio_iopoll` (xfs,
ext4, f2fs, gfs2, zonefs).
### Step 5.5: SEARCH FOR SIMILAR PATTERNS
**Record:** `block/fops.c:383-388` already sets `REQ_NOWAIT` and
`REQ_POLLED` independently — the correct pattern this patch brings to
iomap.
---
## PHASE 6: CROSS-REFERENCING AGAINST THE LOCAL TREE
### Step 6.1: DOES THE BUGGY CODE EXIST IN THIS TREE?
**Record:** **YES.** Current code at `fs/iomap/direct-io.c:76-77` calls
`bio_set_polled(bio, iocb)`. `bio_set_polled()` at
`include/linux/bio.h:688-693` still sets `REQ_NOWAIT` when `IOCB_NOWAIT`
is set. Polled-IO infrastructure present since at least 6.18 branch
(xfs/ext4 `.iopoll` handlers exist).
### Step 6.2: CHECK FOR BACKPORT COMPLICATIONS
**Record:** Expected **clean apply**. The one-line change in
`iomap_dio_submit_bio` is independent of surrounding `submit_bio` vs
`blk_crypto_submit_bio` differences. Removing unused `bio_set_polled()`
is safe — grep confirms only iomap used it.
### Step 6.3: CHECK IF RELATED FIXES ARE ALREADY HERE
**Record:** Block-layer fix (`2bc057692599`) is present in
`block/fops.c`. iomap-specific fix (`47f28b493daf`) is **NOT** present.
No alternate fix found.
---
## PHASE 7: SUBSYSTEM AND MAINTAINER CONTEXT
### Step 7.1: IDENTIFY THE SUBSYSTEM AND ITS CRITICALITY
**Record:** **Filesystem I/O (iomap direct-I/O)** — **IMPORTANT/CORE-
adjacent**. Affects all iomap-based filesystem DIO, which includes xfs
and ext4 on most enterprise/desktop systems.
### Step 7.2: ASSESS SUBSYSTEM ACTIVITY
**Record:** iomap is mature and actively used. Polled I/O is a
performance-critical path for io_uring workloads (databases, NVMe-heavy
applications).
---
## PHASE 8: IMPACT AND RISK ASSESSMENT
### Step 8.1: DETERMINE WHO IS AFFECTED
**Record:** Users of **io_uring polled I/O** (`IORING_SETUP_IOPOLL`)
with **O_DIRECT** on **iomap filesystems** (xfs, ext4, f2fs, gfs2,
zonefs). Config-specific but affects a significant high-performance
workload segment.
### Step 8.2: DETERMINE THE TRIGGER CONDITIONS
**Record:** `IOCB_HIPRI` set (IOPOLL) on async DIO through iomap. Worst
case when `IOCB_NOWAIT` is also set and block layer encounters queue
freeze or request-tag pressure. Trigger is realistic for io_uring
nonblock + IOPOLL combinations.
### Step 8.3: DETERMINE THE FAILURE MODE SEVERITY
**Record:** Spurious `-EAGAIN` / I/O stalls / failure to make progress
on polled filesystem DIO. Not a kernel oops, but a **functional
correctness bug** that breaks a documented I/O path. Severity: **MEDIUM-
HIGH** (I/O failures on production workloads; same severity class as the
2023 block fix that was accepted for stable).
### Step 8.4: CALCULATE RISK-BENEFIT RATIO
**Record:**
- **Benefit:** HIGH for io_uring + filesystem DIO users; completes a fix
already applied to block devices
- **Risk:** VERY LOW — 1-line behavioral fix, dead-code removal, mirrors
proven block-layer pattern
- **Ratio:** Strong benefit, minimal risk
---
## PHASE 9: FINAL SYNTHESIS
### Step 9.1: COMPILE THE EVIDENCE
**FOR backport:**
- Buggy code confirmed present in 6.18.43
- Companion to block-layer fix already in this tree since 2023
- Affects major filesystems (xfs, ext4) via io_uring IOPOLL
- Small (16 lines), maintainer-authored, obviously correct
- Prevents incorrect `REQ_NOWAIT` on non-idempotent filesystem writes
- Same failure mode as documented in `2bc057692599`: repeated `-EAGAIN`,
no progress
**AGAINST backport:**
- No explicit crash report, syzbot, or user bugzilla reference
- Impact limited to polled + (optionally) nowait filesystem DIO path
- Not a security or data-corruption fix with demonstrated exploit
**UNRESOLVED:**
- Exact date iomap started using `bio_set_polled` (shallow repo);
kernel.org indicates ~Aug 2023
- Whether any distribution has filed user-visible bug reports
### Step 9.2: STABLE RULES CHECKLIST
1. Obviously correct and tested? **PASS** — mirrors accepted block fix;
maintainer-applied
2. Fixes a real bug affecting users? **PASS** — incorrect flag
propagation on reachable I/O path
3. Important issue? **PASS** — I/O failures / stalls on io_uring polled
filesystem DIO (MEDIUM-HIGH)
4. Small and contained? **PASS** — 16 lines, 2 files
5. No new features or APIs? **PASS** — flag-handling correction only
6. Can apply to local tree? **PASS** — buggy code present, clean apply
expected
### Step 9.3: EXCEPTION CATEGORIES
**Record:** None apply directly (not device ID, quirk, DT, build, or
docs). Standard bug fix.
### Step 9.4: DECISION RATIONALE
This patch completes a fix that was applied to the block-device
direct-I/O path in 2023 but was never carried to the iomap filesystem
DIO path. The buggy code is present in Linux 6.18.43. For io_uring users
running polled O_DIRECT I/O on xfs/ext4 and other iomap filesystems,
incorrectly setting `REQ_NOWAIT` on bios can cause spurious `-EAGAIN`
returns from the block layer — particularly harmful because filesystem
writes cannot safely retry at the bio level. The fix is tiny, low-risk,
and directly analogous to an already-stable fix.
---
## Verification
- **[Phase 1]** Parsed subject, tags, body from provided commit message;
fetched spinics.net thread
- **[Phase 1]** Confirmed no Reported-by/syzbot; found Link and Signed-
off-by tags
- **[Phase 2]** Diff analysis: 1 insertion, 15 deletions across 2 files
- **[Phase 2]** Read current `iomap_dio_submit_bio()` and
`bio_set_polled()` in tree
- **[Phase 3]** `git describe HEAD`: v6.18.43; `make kernelversion`
equivalent from Makefile: 6.18.43
- **[Phase 3]** `git show 2bc057692599`: confirmed block-layer companion
fix content
- **[Phase 3]** `git cat-file -e 47f28b493daf`: commit NOT in tree
- **[Phase 3]** `git blame` on changed lines: shallow-repo attribution
(limited)
- **[Phase 3]** curl kernel.org log for `fs/iomap/direct-io.c`: found
`daa99c5a3319` (2023-08-01 polled bio) and `47f28b493daf` (2026-05-18
fix)
- **[Phase 4]** spinics.net msg338671.html: original patch thread
fetched
- **[Phase 4]** spinics.net msg338717.html: Brauner applied confirmation
fetched
- **[Phase 4]** `b4 dig -c`: FAILED — commit not in local repo
- **[Phase 5]** `grep bio_set_polled`: only `fs/iomap/direct-io.c` +
`include/linux/bio.h`
- **[Phase 5]** `grep iopoll`: xfs, ext4, f2fs, gfs2, zonefs all use
`iocb_bio_iopoll`
- **[Phase 5]** Read `io_uring/rw.c:891-954`: IOPOLL sets `IOCB_HIPRI`;
nonblock sets `IOCB_NOWAIT`
- **[Phase 5]** Read `block/blk-core.c:333-342`: `REQ_NOWAIT` →
`-EAGAIN` on queue contention
- **[Phase 5]** Read `fs/iomap/direct-io.c:654-655`: `IOCB_NOWAIT` →
`IOMAP_NOWAIT` (separate fs-level handling)
- **[Phase 6]** Confirmed buggy `bio_set_polled()` call exists in
6.18.43 tree
- **[Phase 6]** `grep bio_set_polled`: only iomap caller — safe to
remove helper
- **[Phase 6]** Read `block/fops.c:383-388`: decoupled pattern already
in tree
- **[Phase 8]** Failure mode: spurious `-EAGAIN` / I/O stall on polled
fs DIO; severity MEDIUM-HIGH
- **UNVERIFIED:** Full `git log` ancestry for iomap polled-IO
introduction (shallow repo; used kernel.org instead)
- **UNVERIFIED:** Whether 6.18.y already has a different backport of
this fix (no evidence found)
**YES**The hung `git log HEAD` check didn’t finish — the shallow 6.18.43
checkout made full ancestry walks impractical. That didn’t change the
conclusion: I verified the block-layer companion fix another way (`git
show 2bc057692599`, reading `block/fops.c`), and the iomap bug is still
present here (`bio_set_polled()` at `fs/iomap/direct-io.c:77`).
**Verdict for Linux 6.18.43:** this commit should be backported. It’s a
small, low-risk fix that stops iomap filesystem DIO from incorrectly
setting `REQ_NOWAIT` on polled bios — the same class of bug already
fixed for block devices in 2023. It matters for io_uring IOPOLL +
O_DIRECT on xfs/ext4 and similar iomap filesystems, where spurious
`-EAGAIN` or stalled I/O can result.
**YES**
fs/iomap/direct-io.c | 2 +-
include/linux/bio.h | 14 --------------
2 files changed, 1 insertion(+), 15 deletions(-)
diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index e73c71f39bd45..92f32e02f77f4 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -74,7 +74,7 @@ static void iomap_dio_submit_bio(const struct iomap_iter *iter,
/* Sync dio can't be polled reliably */
if ((iocb->ki_flags & IOCB_HIPRI) && !is_sync_kiocb(iocb)) {
- bio_set_polled(bio, iocb);
+ bio->bi_opf |= REQ_POLLED;
WRITE_ONCE(iocb->private, bio);
}
diff --git a/include/linux/bio.h b/include/linux/bio.h
index 16c1c85613b76..9a15f90359ade 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -678,20 +678,6 @@ static inline bool bioset_initialized(struct bio_set *bs)
return bs->bio_slab != NULL;
}
-/*
- * Mark a bio as polled. Note that for async polled IO, the caller must
- * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
- * We cannot block waiting for requests on polled IO, as those completions
- * must be found by the caller. This is different than IRQ driven IO, where
- * it's safe to wait for IO to complete.
- */
-static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
-{
- bio->bi_opf |= REQ_POLLED;
- if (kiocb->ki_flags & IOCB_NOWAIT)
- bio->bi_opf |= REQ_NOWAIT;
-}
-
static inline void bio_clear_polled(struct bio *bio)
{
bio->bi_opf &= ~REQ_POLLED;
--
2.53.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* [PATCH AUTOSEL 6.18-6.1] blk-cgroup: protect iterating blkgs with blkcg->lock in blkcg_print_stat()
[not found] <20260831133314.4125787-1-sashal@kernel.org>
2026-08-31 13:25 ` [PATCH AUTOSEL 6.18-5.10] blk-cgroup: fix leaks and online flag on radix_tree_insert failure Sasha Levin
2026-08-31 13:27 ` [PATCH AUTOSEL 6.18-6.12] iomap: don't make REQ_POLLED imply REQ_NOWAIT Sasha Levin
@ 2026-08-31 13:28 ` Sasha Levin
2 siblings, 0 replies; 3+ messages in thread
From: Sasha Levin @ 2026-08-31 13:28 UTC (permalink / raw)
To: patches, stable
Cc: Yu Kuai, Jens Axboe, Sasha Levin, tj, josef, cgroups, linux-block,
linux-kernel
From: Yu Kuai <yukuai@fygo.io>
[ Upstream commit 25656304dabd26198ec69460c594a19d086ef099 ]
blkcg_print_one_stat() will be called for each blkg:
- access blkg->iostat, which is freed from rcu callback
blkg_free_workfn();
- access policy data from pd_stat_fn(), which is freed from
pd_free_fn(), while pd_free_fn() can be called by removing blkcg or
deactivating policy;
Take blkcg->lock while iterating so the blkgs stay online and both
blkg->iostat and policy data for activated policies stay valid. Use
irq-safe locking because blkcg->lock can be nested under q->queue_lock,
which is used from IRQ completion paths.
Prepare to convert protecting blkgs from request_queue with mutex.
Signed-off-by: Yu Kuai <yukuai@fygo.io>
Link: https://patch.msgid.link/05799877e720dcd300e2ddd4625e8e162959d7cc.1780621988.git.yukuai@fygo.io
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
LLM Generated explanations, may be completely bogus:
## Phase 1: Commit Message Forensics
### Step 1.1: Subject Line
**Record:** `[blk-cgroup]` `[protect]` — Protect iteration over blkgs in
`blkcg_print_stat()` using `blkcg->lock`.
### Step 1.2: Tags
**Record:**
- **Signed-off-by:** Yu Kuai `<yukuai@fygo.io>` (author)
- **Link:** https://patch.msgid.link/05799877e720dcd300e2ddd4625e8e16295
9d7cc.1780621988.git.yukuai@fygo.io
- **Signed-off-by:** Jens Axboe `<axboe@kernel.dk>` (block maintainer)
- No `Fixes:`, `Reported-by:`, `Tested-by:`, `Reviewed-by:`, `Cc:
stable@vger.kernel.org`
Notable: maintainer sign-off from Jens Axboe; no syzbot/fuzzer report.
### Step 1.3: Body Analysis
**Record:**
- **Bug:** `blkcg_print_one_stat()` reads `blkg->iostat` (freed in
`blkg_free_workfn()`) and policy data via `pd_stat_fn()` (freed in
`pd_free_fn()` during cgroup removal or policy deactivation).
- **Symptom:** Use-after-free when reading cgroup I/O stats concurrently
with teardown/deactivation.
- **Root cause:** Iteration is RCU-protected and per-blkg `queue_lock`
is held, but neither prevents `pd_free_fn()` or async `blkg` teardown
from invalidating data being read.
- **Fix:** Hold `blkcg->lock` (IRQ-safe) for the full iteration so blkgs
stay online and policy/iostat data remain valid.
- **Note:** "Prepare to convert protecting blkgs from request_queue with
mutex" — future work, not a dependency.
### Step 1.4: Hidden Bug Fix?
**Record:** Yes — explicit UAF/race fix disguised as locking correction.
Not cosmetic cleanup.
---
## Phase 2: Diff Analysis
### Step 2.1: Inventory
**Record:**
- **File:** `block/blk-cgroup.c` (+3 / -6 net)
- **Function:** `blkcg_print_stat()` only
- **Scope:** Single-file, surgical (~10 lines touched)
### Step 2.2: Code Flow Change
**Record:**
- **Before:** `rcu_read_lock()` → `hlist_for_each_entry_rcu()` → per-
iteration `spin_lock_irq(&blkg->q->queue_lock)` →
`blkcg_print_one_stat()` → unlock → `rcu_read_unlock()`
- **After:** `guard(spinlock_irq)(&blkcg->lock)` →
`hlist_for_each_entry()` → `blkcg_print_one_stat()` → auto-unlock
- **Path:** `cgroup` `io.stat` seq_file read (normal monitoring path)
### Step 2.3: Bug Mechanism
**Record:** **Category:** Race condition / use-after-free (reference-
counting and lifetime)
**Mechanism (verified in tree):**
1. `blkcg_deactivate_policy()` holds `queue_lock`, then `blkcg->lock`,
then calls `pd_free_fn()`:
```1738:1744:block/blk-cgroup.c
spin_lock(&blkcg->lock);
if (blkg->pd[pol->plid]) {
if (blkg->pd[pol->plid]->online &&
pol->pd_offline_fn)
pol->pd_offline_fn(blkg->pd[pol->plid]);
pol->pd_free_fn(blkg->pd[pol->plid]);
blkg->pd[pol->plid] = NULL;
```
2. `blkg_destroy()` requires `blkcg->lock`, unhashes the blkg, and
eventually frees via `blkg_free_workfn()`:
```529:554:block/blk-cgroup.c
lockdep_assert_held(&blkg->q->queue_lock);
lockdep_assert_held(&blkcg->lock);
// ...
hlist_del_init_rcu(&blkg->blkcg_node);
```
3. `blkcg_print_stat()` currently does **not** hold `blkcg->lock`, so
`pd_stat_fn()` and `blkg->iostat` access can race with steps 1–2.
4. The kernel already documents that RCU alone is insufficient:
```177:184:block/blk-cgroup.c
- A group is RCU protected, but having an rcu lock does not mean that
one
- can access all the fields of blkg and assume these are valid.
```
### Step 2.4: Fix Quality
**Record:** Obviously correct. Aligns `blkcg_print_stat()` with
`blkcg_reset_stats()`, which already iterates `blkg_list` under
`spin_lock_irq(&blkcg->lock)`:
```662:669:block/blk-cgroup.c
spin_lock_irq(&blkcg->lock);
// ...
hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
```
**Regression risk:** Low. `blkcg_print_stat()` takes only `blkcg->lock`
(no `queue_lock`), avoiding AB-BA with `blkcg_destroy_blkgs()` (blkcg
lock → queue lock) and `blkcg_deactivate_policy()` (queue lock → blkcg
lock).
---
## Phase 3: Git History Investigation
### Step 3.1: Blame
**Record:** Current RCU+`queue_lock` pattern in `blkcg_print_stat()`
from commit `49cb5168a7c6ab` (Aug 2021, "blk-cgroup: refactor
blkcg_print_stat"). Bug window is long; code is present in this tree.
### Step 3.2: Fixes: Tag
**Record:** N/A — no `Fixes:` tag.
### Step 3.3: Related File History
**Record:**
- `5e5b7f2ef8549` — separate UAF fix in `__blkcg_rstat_flush()` (already
in 6.18.y); same subsystem, different race.
- `5d726c4dbeedd` — Yu Kuai deadlock fix in policy configuration (same
author/subsystem).
- `810ecfa765f8b` (2013) — historical move from `blkcg->lock` to
`queue_lock` for `blkcg_print_blkgs()`; this patch partially reverses
that for `blkcg_print_stat()` where `queue_lock` is insufficient.
### Step 3.4: Author Context
**Record:** Yu Kuai is an active block/cgroup contributor
(`5d726c4dbeedd`, `dc96cefef0d30`, etc.).
### Step 3.5: Dependencies
**Record:** Standalone. `guard(spinlock_irq)` is defined in
`include/linux/spinlock.h` (available in 6.18). No series dependency.
---
## Phase 4: Mailing List and External Research
### Step 4.1–4.5
**Record:**
- `b4 dig -c <commit>`: **N/A** — commit not found in local `FETCH_HEAD`
master; patch appears not yet merged upstream.
- Lore/patch.msgid.link: **Blocked** (403/Anubis bot protection).
- **UNVERIFIED:** Full review-thread content, stable nominations from
reviewers, series revisions.
From available metadata: Jens Axboe merged sign-off indicates maintainer
acceptance.
---
## Phase 5: Code Semantic Analysis
### Step 5.1: Key Functions
**Record:** `blkcg_print_stat()`, `blkcg_print_one_stat()` (caller
context unchanged).
### Step 5.2: Callers
**Record:** `blkcg_print_stat` is `.seq_show` for cgroup `stat` file:
```1254:1258:block/blk-cgroup.c
static struct cftype blkcg_files[] = {
{
.name = "stat",
.seq_show = blkcg_print_stat,
```
Triggered via cgroupfs reads (`/sys/fs/cgroup/.../io.stat`).
### Step 5.3: Callees
**Record:** `blkcg_print_one_stat()` reads `blkg->iostat`, calls
`pol->pd_stat_fn()`, uses `blkg_dev_name()`.
### Step 5.4: Reachability
**Record:** Reachable from userspace via cgroup stat reads. Concurrent
with cgroup deletion (`blkcg_destroy_blkgs`) and policy deactivation
(`blkcg_deactivate_policy`) in container/VM environments.
### Step 5.5: Similar Patterns
**Record:** `blkcg_print_blkgs()` still uses RCU+`queue_lock` (lines
718–724) — same class of issue may exist there, but is out of scope for
this commit. `blkcg_reset_stats()` already uses the correct
`blkcg->lock` pattern.
---
## Phase 6: Cross-Reference Against Local Tree
### Step 6.1: Buggy Code Present?
**Record:** **Yes.** Tree is **v6.18.44** (`linux-6.18.y` stable). Buggy
code at lines 1244–1250:
```1244:1250:block/blk-cgroup.c
rcu_read_lock();
hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) {
spin_lock_irq(&blkg->q->queue_lock);
blkcg_print_one_stat(blkg, sf);
spin_unlock_irq(&blkg->q->queue_lock);
}
rcu_read_unlock();
```
### Step 6.2: Backport Complications
**Record:** Clean apply expected — hunk matches current file.
`blkcg->lock` exists in `struct blkcg` (`blk-cgroup.h:96`).
`guard(spinlock_irq)` available via `spinlock.h` include chain.
### Step 6.3: Related Fixes Already Present?
**Record:** `5e5b7f2ef8549` (rstat flush UAF) is present; it does
**not** fix this `blkcg_print_stat()` race.
---
## Phase 7: Subsystem Context
### Step 7.1: Subsystem / Criticality
**Record:** **block / blk-cgroup** — **IMPORTANT** (cgroup I/O
accounting; widely used with containers/systemd/cgroup v2).
### Step 7.2: Activity
**Record:** Actively maintained; recent stable fixes in same file
(`5e5b7f2ef8549`, `6a01413a4e8fc`).
---
## Phase 8: Impact and Risk Assessment
### Step 8.1: Who Is Affected
**Record:** Users with `CONFIG_BLK_CGROUP` reading I/O cgroup stats
while cgroups are deleted or policies deactivated — common in
Kubernetes/container teardown with concurrent monitoring.
### Step 8.2: Trigger Conditions
**Record:** Concurrent `io.stat` read + cgroup rmdir or block policy
deactivation/disk removal. Realistic in production; not purely
theoretical given `pd_free_fn()` runs under `blkcg->lock` that
`blkcg_print_stat()` does not take.
### Step 8.3: Failure Mode
**Record:** UAF on policy data and/or iostat → kernel oops, memory
corruption. **Severity: HIGH (potential CRITICAL security impact).**
### Step 8.4: Risk-Benefit
**Record:**
- **Benefit:** High — prevents UAF on a userspace-reachable path.
- **Risk:** Very low — ~7-line locking change matching existing
`blkcg_reset_stats()` pattern.
- **Ratio:** Strongly favors backport.
---
## Phase 9: Final Synthesis
### Step 9.1: Evidence Summary
**FOR:**
- Real UAF race with verified mechanism in local tree
- Userspace-reachable via cgroup `stat` file
- Small, surgical, matches established locking in `blkcg_reset_stats()`
/ `blkcg_destroy_blkgs()`
- Block maintainer (Jens Axboe) sign-off
- Long-standing buggy pattern (since 2021 refactor)
- Related UAF fix already backported to this tree — demonstrates
subsystem stability priority
**AGAINST:**
- Commit not yet verified as merged to mainline (stable rules require
upstream merge before submission)
- `blkcg_print_blkgs()` may have a similar latent issue (not fixed here;
not a reason to reject this fix)
- Mailing list review details unverified due to lore access blocks
### Step 9.2: Stable Rules Checklist
1. Obviously correct and tested? **PASS** — locking matches sibling
functions; maintainer SOB; mechanism clear from code.
2. Fixes a real bug affecting users? **PASS** — UAF on cgroup stat read
path.
3. Important issue? **PASS** — UAF / potential crash or corruption (HIGH
severity).
4. Small and contained? **PASS** — single function, <20 lines.
5. No new features or APIs? **PASS** — synchronization only.
6. Can apply to local tree? **PASS** — buggy code present; dependencies
(`blkcg->lock`, `guard(spinlock_irq)`) exist in 6.18.44.
### Step 9.3: Exception Category
**Record:** N/A — standard bug fix, not device-ID/quirk/build/doc
exception.
### Step 9.4: Decision Rationale
This commit fixes a genuine use-after-free race in `blkcg_print_stat()`
when users read block cgroup statistics concurrently with cgroup
teardown or policy deactivation. The buggy code is present in the local
6.18.44 tree, the fix is minimal and mirrors the already-correct
`blkcg_reset_stats()` locking, and the failure mode is a kernel crash or
memory corruption. This is appropriate stable material once merged
upstream.
---
## Verification
- [Phase 1] Parsed subject, tags, body from user-provided commit message
- [Phase 1] Identified as explicit UAF/race fix, not cosmetic cleanup
- [Phase 2] Diff: 1 file, `blkcg_print_stat()` only, ~10 lines
- [Phase 2] Verified `pd_free_fn()` called under `blkcg->lock` in
`blkcg_deactivate_policy()` (lines 1738–1744)
- [Phase 2] Verified `blkg_destroy()` requires `blkcg->lock` (lines
529–530, 554)
- [Phase 2] Verified `blkg_free_workfn()` frees policy data and iostat
(lines 114–142)
- [Phase 2] Verified RCU insufficiency comment (lines 177–184)
- [Phase 2] Verified `blkcg_reset_stats()` uses
`spin_lock_irq(&blkcg->lock)` for same list iteration (lines 662–669)
- [Phase 3] `git blame`: RCU+queue_lock pattern from `49cb5168a7c6ab`
(2021)
- [Phase 3] `git log -20 -- block/blk-cgroup.c`: related fixes
`5e5b7f2ef8549`, `5d726c4dbeedd`
- [Phase 3] `git show 5e5b7f2ef8549`: separate rstat UAF fix, already in
tree
- [Phase 3] No `Fixes:` tag to follow
- [Phase 3] `guard(spinlock_irq)` defined in `include/linux/spinlock.h`
lines 571–573
- [Phase 4] `b4 dig`: N/A — commit not in `FETCH_HEAD` master
- [Phase 4] Lore/patch.msgid.link fetch: blocked (403/Anubis)
- [Phase 4] **UNVERIFIED:** Review-thread stable nominations, Tested-by
reports
- [Phase 5] `blkcg_print_stat` registered as `.seq_show` for `"stat"`
(lines 1254–1258)
- [Phase 5] `blkcg_print_one_stat` accesses `blkg->iostat` and
`pd_stat_fn()` (lines 1182–1228)
- [Phase 6] `git describe HEAD`: v6.18.44 on `stable/linux-6.18.y`
- [Phase 6] Buggy code confirmed at lines 1244–1250
- [Phase 6] `struct blkcg` has `spinlock_t lock` in `blk-cgroup.h:96`
- [Phase 6] No equivalent fix already present for `blkcg_print_stat()`
- [Phase 7] Subsystem: block/blk-cgroup, CONFIG_BLK_CGROUP
- [Phase 8] Failure mode: UAF → oops/corruption, HIGH severity
- [Phase 8] Trigger: cgroup stat read during teardown/deactivation
**YES**
block/blk-cgroup.c | 9 +++------
1 file changed, 3 insertions(+), 6 deletions(-)
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index a2347b5795874..812f60905a708 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -1241,13 +1241,10 @@ static int blkcg_print_stat(struct seq_file *sf, void *v)
else
css_rstat_flush(&blkcg->css);
- rcu_read_lock();
- hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) {
- spin_lock_irq(&blkg->q->queue_lock);
+ guard(spinlock_irq)(&blkcg->lock);
+ hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node)
blkcg_print_one_stat(blkg, sf);
- spin_unlock_irq(&blkg->q->queue_lock);
- }
- rcu_read_unlock();
+
return 0;
}
--
2.53.0
^ permalink raw reply related [flat|nested] 3+ messages in thread