[CVE-2026-98496][MODERATE 7.0] net: airoha: npu: cancel wdt_work after releasing the WDT IRQ

[CVE-2026-98496][MODERATE 7.0] net: airoha: npu: cancel wdt_work after releasing the WDT IRQ

[AL-KERNEL]

CVE: CVE-2026-98496
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: net: airoha: npu: cancel wdt_work after releasing the WDT IRQ
Commit: aca91d8fa0ecbc990ebf9a688c684dda78a43c00
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=aca91d8fa0ecbc990ebf9a688c684dda78a43c00
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98496
Analysis date: Fri, 23 Oct 2026 20:48:30 -0400
ActionableScore: 2
ActionableScore lower bound: unknown
Actionable bucket: Borderline / manual review recommended
Manual review required: YES

Summary:
A teardown race in the Airoha NPU driver can queue watchdog work after cancellation and execute it through a freed devm-managed context, creating a hardware-specific kernel use-after-free with a plausible crash impact but no established attacker-controlled reuse or privilege-escalation primitive.

======================================================================
ABOUT THIS REPORT
======================================================================

The original Linux kernel CVE announcement for CVE-2026-98496 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98496

The original announcement does not normally provide a security severity
estimate, CVSS assessment, or enough information to determine whether the
reported kernel bug represents a practically relevant security issue.

This report was generated by AL-KERNEL, an AI-assisted Linux kernel
vulnerability analysis system developed by Alexander Larkin. It combines
an autonomous classifier with LLM-assisted technical analysis and a
separate ActionableScore mechanism.

The purpose of this report is to prioritize Linux kernel CVEs before
manual review, identify cases that require prompt investigation, and
support automatic closure of issues that are unlikely to have meaningful
security impact.

Published priority for this report: MODERATE 7.0
Manual review required: YES

A detailed explanation of the methodology and priority rules is included
at the end of this message.

======================================================================
AL-KERNEL CLASSIFICATION RESULT
======================================================================

CVE-2026-98496	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-98496 CHECK WITH IMPACT FROM ORIG NN MODERATE | AEGIS IMPACT MODERATE_KPANIC AEGIS_AS=5 LEGACY=MODERATE_KPANIC RESULT=SAME V39_AEGIS_ONE_STEP_UP=MODERATE->MODERATE_KPANIC (final reconciliation agrees; DECISION: both MODERATE_KPANIC)	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:L/AC:H/PR:H/UI:N/S:U/C:H/I:H/A:H';*CWE-416;CWE-362;CWE-667;Other CVSS 'AV:L/AC:H/PR:H/UI:N/S:U/C:N/I:N/A:H';BEST CVSS score: '6.4';DESCR 'A race between the Airoha NPU watchdog IRQ and driver teardown can queue wdt_work after cancel_work_sync has returned. The containing struct airoha_npu is then released by devres, so the late work can dereference freed memory and create a use-after-free condition. A pending watchdog interrupt could also queue the work before it was initialized because INIT_WORK previously ran after devm_request_irq. Impact is at least a local denial of service through a kernel crash. In the worst case the use-after-free may permit confidentiality or integrity impact if the freed allocation is reused in a controllable way, although no reliable privilege escalation primitive is demonstrated. For the CVSS the PR:H reflects that reliable triggering normally requires administrative control to remove or unbind the platform driver. AC:H reflects the required timing between work cancellation, a watchdog interrupt, and managed resource teardown. The issue is not directly network reachable based on the available code and requires affected Airoha NPU hardware and driver activity. Automated impact decision path. The original SimpleNet classifier selected MODERATE_KPANIC. The primary LLM estimated paranoid CVSS 6.4 and the production ActionableScore returned 2 with recommendation Borderline / manual review recommended. The production rules then evaluated CVSS thresholds and ActionableScore thresholds. Recorded rule transitions were ActionableScore thresholds stage MODERATE_KPANIC to MODERATE because the ActionableScore threshold rule was applied. final reconciliation stage MODERATE to MODERATE_KPANIC because the cross model evidence reconciliation rule was applied. The legacy KP specialist has no review artifact for this run. The legacy FD or DH downgrade specialist has no review artifact for this run. Exact triggered rule markers were DECREASED_TO_MODERATEREG_BASED_ON_ACTIONABLESCORELESSTHAN4 and V39_AEGIS_ONE_STEP_UP and MANUAL_REVIEW_REQUIRED. The independent weaker Aegis shadow evaluated CVSS 2.2 and ActionableScore 5 with primitive class CLASS_B and manual review REQUIRED. Its independently reconciled impact was MODERATE_KPANIC. Cross model aggregation changed the production result under an enabled evidence gated merge rule. The final reported impact is MODERATE_KPANIC.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Borderline / manual review recommended (with actual score 2)	MAYBE	REMOTE INIT UAF NETWORK LINUS TEST  DECREASED_TO_MODERATEREG_BASED_ON_ACTIONABLESCORELESSTHAN4 KPANIC V39_AEGIS_ONE_STEP_UP MANUAL_REVIEW_REQUIRED	-	-	checked

======================================================================
ACTIONABLESCORE ANALYSIS
======================================================================

ActionableScore=2
PrimitiveClass=CLASS_B
RealUAF=YES
HostAdminRequired=UNKNOWN
NamespaceScopedPrivilege=UNKNOWN
ImportantCandidate=NO
ManualReview=RECOMMENDED
AutoDowngradeAllowed=YES
RequiresHostUID0=UNKNOWN
AuthorizationDecisionBypass=NO
ProtectedResourceImpact=NONE
B4PendingReview=NO
B4PendingConfirmed=NO

1. ActionableScore

* Raw conservative score: 2
* Conservative score: 2
* Raw paranoid score: 2
* Paranoid score: 2
* No floor or cap was applied.
* Final recommended bucket: **Borderline / manual review recommended**::

2. Signal breakdown

* +2 Real UAF. The watchdog IRQ can queue `wdt_work` after `cancel_work_sync()` returns, while devres subsequently frees the containing `struct airoha_npu`. The delayed callback then dereferences storage outside its valid lifetime.
* +1 Reliable kernel crash or strong DoS. Executing a work callback through a freed driver context can realistically produce an Oops or kernel panic.
* -1 Hard one-shot teardown race. The relevant interrupt must occur between work cancellation and managed IRQ release during device removal or probe unwind. The supplied evidence does not show that an attacker can repeatedly or directly control this timing.
* No host-admin penalty was applied. Administrative device removal is one possible path, but the patch also covers probe failure, and the complete attacker-controlled trigger is unresolved.
* No rare-hardware penalty was applied. Airoha NPU hardware is required, but the supplied evidence does not establish deployment rarity strongly enough for an additional numeric deduction.
* No structured stale-object leverage point was awarded. The visible evidence establishes a late callback using a fixed freed driver context, but does not show attacker-controlled reclaim, replacement contents, callback substitution, or security-sensitive field consumption.
* No repeated release or refcount point was awarded. There is no second `put`, `free`, `release`, or refcount mutation through the dangling reference.
* No confidentiality or integrity point was awarded. Reuse-based corruption is theoretically possible, but no disclosure, directed write, or protected-state modification is established.
* No additional availability point was awarded because the same possible kernel crash is already counted.
* NEW52 B5-R does not apply. There is no confirmed retry-accumulated heap OOB write.
* NEW53 B3-PENDING does not apply. There is no shared or file-backed storage write.
* V33, V32, and V34 do not apply. No attacker-influenced checksum placement, active packet-header corruption, privileged downstream metadata consumer, or cross-boundary networking chain is established.
* NEW54 B4-PENDING does not apply. The patch does not alter an authorization decision or restore a missing security guard.

Repeated-release checks:

* First lifetime operation capable of freeing the object: devres teardown frees the containing allocation, but this is not a release through the stale callback.
* Exact reference surviving into a second lifetime mutation: no.
* Concrete second lifetime or refcount mutation: no.
* Concrete repeated release or refcount +1: not awarded.
* Paranoid repeated-refcount leverage +1: withheld.
* FLOOR 5: not applicable.

3. Reachability analysis

The vulnerable ordering requires the Airoha NPU driver and corresponding platform hardware. The UAF can arise during driver removal or probe unwind when a watchdog interrupt queues work after cancellation but before the IRQ is released.

A deliberate driver unbind or device removal would ordinarily involve host device-management authority, but the supplied patch does not prove that this is the only complete trigger. Conversely, it does not establish that an ordinary user, namespace-scoped capability holder, network peer, or delegated service can initiate the complete sequence. `HostAdminRequired` and `NamespaceScopedPrivilege` therefore remain `UNKNOWN`.

No direct packet-driven trigger is shown. Networking activity might influence NPU watchdog behavior, but the patch does not establish attacker control over the watchdog interrupt or over the required simultaneous teardown. The race is a narrow, asynchronous, largely externally uncontrolled teardown window. Prior hardware and driver availability are required, but no special boot option or namespace configuration is identified.

4. Primitive and severity interpretation

The exact primitive is an asynchronous post-teardown use-after-free. The stale storage is a fixed devm-managed driver context containing `wdt_work`, and the stale operation is execution of the normal watchdog work callback after the containing object has been freed.

This is `CLASS_B` because a real lifetime violation is directly established. It is not `CLASS_A` because controlled reclaim, attacker-shaped replacement data, arbitrary read or write, callback hijacking, refcount takeover, and privilege escalation are not demonstrated.

The demonstrated consequence is a plausible kernel crash or host DoS. A stronger consequence would require the freed allocation to be reused in a way that influences fields consumed by the watchdog callback, but that mechanism is unresolved and unsupported by the supplied code. The separate pre-fix ordering where an IRQ could queue an uninitialized work item also supports invalid-state or crash concern, but does not independently establish a stronger primitive.

`ImportantCandidate=NO` because this is presently a constrained, one-shot driver teardown UAF without meaningful attacker-controlled reuse or an established trust-boundary crossing. `AutoDowngradeAllowed=YES` because no independent exploitation primitive beyond the fixed-context stale callback remains unresolved. Operational severity is borderline Moderate and hardware-specific.

5. One-sentence report phrase

A teardown race in the Airoha NPU driver can queue watchdog work after cancellation and execute it through a freed devm-managed context, creating a hardware-specific kernel use-after-free with a plausible crash impact but no established attacker-controlled reuse or privilege-escalation primitive.

6. Manual review recommendation

MANUAL CHECK RECOMMENDED

The UAF is concrete, but review should focus on the body of `airoha_npu_wdt_work`, possible allocation reuse before callback execution, and whether unprivileged traffic can reliably provoke watchdog interrupts during an independently reachable teardown path.

7. Call-site confidence

Call-site confidence: medium

======================================================================
UPSTREAM PATCH SUMMARY
======================================================================

Patch: net: airoha: npu: cancel wdt_work after releasing the WDT IRQ
Commit: aca91d8fa0ecbc990ebf9a688c684dda78a43c00
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=aca91d8fa0ecbc990ebf9a688c684dda78a43c00

Commit description:

airoha_npu_remove() calls cancel_work_sync() on each core's wdt_work,
but the watchdog IRQ that queues it is requested with devm_request_irq()
and is freed only after .remove() returns.  airoha_npu_wdt_handler() can
therefore schedule_work() again once the cancel has returned.  struct
airoha_npu, which contains the work, is devm_kzalloc()'d and is freed in
that same unwind, so the late work dereferences freed memory.

Register the work with devm_work_autocancel() before devm_request_irq()
and drop .remove().  Devres runs in reverse order, so the IRQ is freed
before cancel_work_sync(), including when probe fails.  A cancel left in
.remove() cannot get that order.  Initializing the work first also stops
a pending watchdog interrupt from queuing an uninitialized work item.
Probe currently calls INIT_WORK() only after devm_request_irq().

This issue was identified during our ongoing static-analysis research
while reviewing kernel code.

Fixes: 23290c7 ("net: airoha: Introduce Airoha NPU support")
Cc: stable@vger.kernel.org # 6.15+
Co-developed-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Myeonghun Pak <mhun512@gmail.com>
Acked-by: Lorenzo Bianconi <lorenzo.bianconi@oss.qualcomm.com>
Link: https://patch.msgid.link/20260922000914.542068-1-mhun512@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Changed files:
  drivers/net/ethernet/airoha/airoha_npu.c

Diff excerpt:

Not included in this email. See the upstream URL for the full patch.

Full patch:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=aca91d8fa0ecbc990ebf9a688c684dda78a43c00

======================================================================
DETAILED REPORT METHODOLOGY
======================================================================

The original Linux kernel CVE announcement for CVE-2026-98496 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98496

The original CVE announcement normally does not include a security-level
estimate. In particular, it may not contain a CVSS assessment, an impact
level, or enough information to determine whether the reported bug is a
practically relevant security issue. One purpose of this parallel CVE list
is to provide that missing technical and prioritization information.

The original goal of the AL-KERNEL project was to prioritize Linux kernel
CVE analysis automatically before manual review. The system can also help
identify non-security issues that may be suitable for automatic closure.

This report was generated by AL-KERNEL, an AI-assisted Linux kernel
vulnerability analysis system developed by Alexander Larkin.

The first analysis stage combines an autonomous classifier with additional
LLM-based analysis. The autonomous classifier runs locally on a CPU and is
based on a backpropagation neural network. Together, these mechanisms
produce a technical vulnerability description, identify likely weakness
types, estimate CVSS severity, and provide input for ActionableScore.

Two CVSS estimates are retained because incomplete kernel vulnerability
information often permits more than one defensible interpretation:

  Conservative CVSS vector: AV:L/AC:H/PR:H/UI:N/S:U/C:N/I:N/A:H
  The Best / paranoid CVSS vector: AV:L/AC:H/PR:H/UI:N/S:U/C:H/I:H/A:H
  The Best / paranoid CVSS score: 6.4

The conservative vector represents a lower-impact interpretation.
The Best/paranoid vector intentionally represents a plausible upper-bound
interpretation and should not automatically be treated as demonstrated
real-world impact.

CVSS may also need to be adjusted for a particular Linux deployment,
because actual reachability, privileges, enabled kernel configuration,
hardware, namespaces, exposed device nodes, and other environmental
conditions can differ significantly between systems.

A separate ActionableScore mechanism evaluates practical remediation
urgency. Its analysis may include reachability, attack prerequisites,
subsystem exposure, memory-corruption characteristics, denial-of-service
reliability, and possible confidentiality, integrity, or
privilege-escalation impact.

  Conservative ActionableScore: unknown
  Paranoid ActionableScore: 2

The final base severity is taken directly from the second tab-separated
field of the AL-KERNEL classification result. ActionableScore does not
replace or independently override that final AL-KERNEL decision, and
if ActionableScore adjusted impact level of ALKERNEL, then you would see
self-readable flags above like INCREASED_TO_HIGH_BASED_ON_ACTIONABLESCOREHIGHEREQTHAN7.

For an AL-KERNEL result of MODERATE, this report uses the following
additional presentation split:

  ActionableScore below 5   -> MODERATE REGULAR
  ActionableScore 5 or more -> MODERATE 7.0

The distinction between MODERATE REGULAR and MODERATE 7.0 makes it
possible to identify Moderate issues that should receive manual analysis
and fixes before lower-priority MODERATE REGULAR issues. In many cases,
MODERATE REGULAR fixes may wait for a later rebase or routine update.

There is one override in which MODERATE REGULAR becomes MODERATE 7.0
even when the ActionableScore is below 5. When the AL-KERNEL result
contains the KPANIC flag, a MODERATE result is always presented as
MODERATE 7.0. The KPANIC flag selected with few regexps without
usage of AI at all, so it helps to detect cases when Kernel Crash happens
and similar (to filter False-Negative results from the LLM usage).

KPANIC indicates that a reliable kernel crash, kernel panic, or similarly
serious kernel availability impact was identified by the classification
workflow.

AL-KERNEL base severity for this report: MODERATE
KPANIC detected for this report: YES
Published priority for this report (same as in Subject): MODERATE 7.0

These results are intended to support engineering triage. They are
machine-generated estimates, and cases marked for manual review should
be validated by a human security engineer before final disposition.
For more info read docs linked from here: https://kernelcve.org/
(and you can submit you own patch there to generate such a report
for non-existant CVE-id yet).

Note that in many cases this AI tool selects higher severity, than
real is (means you can expect Importants instead of Moderate 7.0 or
Moderates 7.0 instead of regular Moderates). If you see such cases,
please use reply email interface to add additional manual analyses
info to this particular CVE.
And please, please, let me know when you see Lows instead of Importants
or Important instead of Low (because particular for such cases I
need to tune this AI tool to make it better for this one and next similar).
My contact email for such notifications is alexanjelausa@gmail.com (and both
send reply to CVE record itself too and see "reply" button below for howto reply).