[CVE-2026-92513][MODERATE 7.0] RDMA/mana_ib: drain QP references after partial table insertion [ Upstream

[AL-KERNEL <alkernel@kernelcve.org>]

CVE: CVE-2026-92513
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: RDMA/mana_ib: drain QP references after partial table insertion [ Upstream
Commit: 4aaa2ab816c710be7385b31d08373edcdcd71656
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=4aaa2ab816c710be7385b31d08373edcdcd71656
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92513
Analysis date: Fri, 18 Sep 2026 07:06:24 -0400
ActionableScore: 6
ActionableScore lower bound: 5
Actionable bucket: Actionable Moderate / Strong Important candidate requiring manual review
Manual review required: YES

Summary:
A partial QP table insertion failure in RDMA mana_ib can leave transient completion-handler references alive while RDMA core frees the QP, creating a use-after-free that is locally reachable through RDMA uverbs access and requires manual review for possible exploitation beyond kernel DoS.

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

The original Linux kernel CVE announcement for CVE-2026-92513 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92513

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-2026-92513	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-92513 CHECK WITH IMPACT FROM ORIG NN IMPORTANT	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H';*CWE-416;CWE-362;*CWE-672;Other CVSS 'AV:L/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:H';BEST CVSS score: '7';DESCR 'An RDMA mana_ib QP can be exposed in the XArray after the send queue id is inserted but before the receive queue id is inserted. A concurrent completion handler can look up that QP and take a transient reference while the second insertion later fails and rollback removes only the send queue entry. This creates a reference lifetime mismatch where RDMA core may free the QP while lookup users still hold references, resulting in a use-after-free. For the CVSS the PR:L is used because a local process with access to RDMA uverbs or delegated RDMA resources can create QPs and hit the failing control path. The issue is not directly Internet reachable, but RDMA devices are often used in data center or cloud networks where a remote peer may help exercise completions once a local QP setup path is reachable. Impact is at least denial of service through kernel memory corruption and in the worst case may include confidentiality or integrity impact, so manual review is recommended.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate / Strong Important candidate requiring manual review (with actual score 6)	YES	UAF NETWORK KPANIC  DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFKP	YES	NO	checked

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

ActionableScore=6
ActionableScoreLower=5
PrimitiveClass=CLASS_B
RealUAF=YES
HostAdminRequired=NO
NamespaceScopedPrivilege=NO
ImportantCandidate=YES
ManualReview=REQUIRED
AutoDowngradeAllowed=NO

## 1. ActionableScore

* Raw conservative score: 5
* Conservative score: 5
* Raw paranoid score: 6
* Paranoid score: 6
* Final recommended bucket: **Actionable Moderate / Strong Important candidate requiring manual review**::

No floor or cap changed the conservative score. The REAL UAF FLOOR 5 is satisfied, but the raw conservative score already reaches 5. No restrictive deployment penalty was applied because host-admin-equivalent privilege is not positively established for the RDMA uverbs trigger model.

## 2. Signal breakdown

Conservative positive signals:

* +2 Real UAF / active lifetime violation: the QP can be published in the XArray, looked up by a concurrent completion handler, and later freed by RDMA core while references are still outstanding.
* +1 Reliable kernel crash / strong DoS: freeing a QP while completion-side users still hold references can realistically lead to kernel UAF faults or corruption.
* +1 Local below-host-admin trigger: a local process with access to RDMA uverbs or delegated RDMA resources can create QPs and exercise this driver path without being host root.
* +1 Concrete repeated release / refcount operation: the missing drain leaves a transient QP reference outstanding, and the completion-side user is expected to return that reference after RDMA core may already have freed the QP.

Paranoid additional signal:

* +1 Paranoid repeated-refcount leverage: the later reference return can perform a write-like lifetime/refcount mutation through storage whose original QP lifetime may already have ended. This is stronger than a shallow stale dereference, but still not enough for CLASS_A.

Non-awarded signals:

* +0 Controlled reclaim not demonstrated.
* +0 Arbitrary or directed kernel write not demonstrated.
* +0 Callback or function-pointer control not demonstrated.
* +0 Remote reachability not established. RDMA network activity may help exercise completions, but the vulnerable setup path is local RDMA object management.
* +0 Structured stale-object leverage not separately awarded because the strongest visible post-lifetime operation is already counted as concrete repeated refcount/lifetime mutation.
* +0 Weak LPE not separately awarded for the same repeated-refcount fact.
* +0 Extra availability not awarded because the same crash/UAF consequence is already counted once.
* +0 C/I impact not awarded directly because no disclosure, arbitrary write, controlled replacement, or demonstrated LPE is shown.
* +0 Race penalty not applied. The bug involves concurrency, but the supplied evidence does not prove a rare one-shot or externally uncontrollable timing condition.
* +0 Hardware rarity penalty not applied. MANA RDMA is device-specific, but driver-specific code alone is not enough to treat the path as rare or physical-only.

Repeated-release/refcount checks:

* First lifetime-ending operation can occur: yes, RDMA core may free the QP while driver-side table and transient references still exist.
* Exact same object reference survives: yes, the completion handler can retain a transient reference to the same QP.
* Second operation mutates lifetime/refcount state: yes, normal destruction waits for concurrent lookup users to return references, implying a real refcount/lifetime completion path.
* Concrete repeated release/refcount +1 awarded: yes.
* Paranoid repeated-refcount leverage +1 awarded: yes, because the stale operation may mutate lifetime/refcount state after object lifetime ended.
* FLOOR 5 applied: satisfied but did not change score.

## 3. Reachability analysis

The evaluated attacker is a local userspace RDMA consumer with access to the MANA RDMA device through uverbs or delegated RDMA resources. This is below host-admin-equivalent privilege in the normal RDMA application model, so HostAdminRequired is set to NO.

The complete trigger requires the ability to create a UD QP and hit the partial insertion failure path where the send-queue id is inserted but receive-queue insertion fails. A concurrent completion handler must also look up the partially published QP and take a transient reference. This is a local control-plane plus device/completion concurrency issue, not a direct Internet or packet-only network trigger.

Namespace-scoped privilege is set to NO because this is not primarily a namespace capability issue. Prior host setup is normally required only in the sense that the MANA RDMA device and driver must exist and be exposed to the user or service, but that is a deployment prerequisite rather than proof that the attacker must be host administrator.

Special hardware is involved because the bug is in the MANA RDMA driver. No hardware penalty is applied because the patch establishes a real kernel lifetime bug and because device-specific driver code is not automatically low priority.

Race repeatability is unresolved. The interleaving is real, but the evidence does not show that the timing is so narrow or uncontrollable that an ActionableScore race penalty is justified.

## 4. Primitive and severity interpretation

The exact primitive is an active-object kernel lifetime violation in `struct mana_ib_qp`. The QP is inserted into a lookup table before the full table insertion succeeds. A completion handler can legally find it and take a transient reference. If the second insertion fails, the old rollback removes only the send-queue entry and returns without dropping the initial table reference and without waiting for transient lookup users.

This is active structured object lifetime corruption, not fixed stale storage. It is also not merely an ordinary cleanup stale dereference. The important distinction is that a QP reference can remain outstanding while RDMA core frees the QP, and later completion-side reference return can mutate lifetime/refcount state through a potentially freed object.

PrimitiveClass is CLASS_B. It is not CLASS_A because controlled reclaim, attacker-shaped replacement contents, arbitrary read/write, callback control, refcount takeover, or demonstrated LPE are not established. However, it is stronger than CLASS_C crash-only behavior because the patch-supported primitive includes real post-lifetime object use and a concrete unresolved reference-lifetime mismatch.

Demonstrated consequence is kernel use-after-free with realistic DoS. Speculative consequence is privilege escalation or broader memory corruption if the stale QP storage can be usefully reused or if the post-free refcount operation can be steered, but that is not proven by the supplied patch.

ImportantCandidate is YES because this is a real active-object UAF with a concrete reference lifetime mismatch and unresolved exploitation leverage beyond an ordinary stale cleanup fault. AutoDowngradeAllowed is NO because automatic Moderate closure would discard a meaningful unresolved Class B lifetime/refcount corruption primitive.

Likely operational severity is Actionable Moderate at minimum, with Strong Important review warranted.

## 5. One-sentence report phrase

A partial QP table insertion failure in RDMA mana_ib can leave transient completion-handler references alive while RDMA core frees the QP, creating a use-after-free that is locally reachable through RDMA uverbs access and requires manual review for possible exploitation beyond kernel DoS.

## 6. Manual review recommendation

MANUAL CHECK REQUIRED

Reason: the patch establishes a real active-object UAF and a concrete unresolved reference-lifetime mismatch where a later reference return may operate on already freed QP storage. Controlled exploitation is not proven, but automatic downgrade is not appropriate.

## 7. Call-site confidence

Call-site confidence: medium

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

Patch: RDMA/mana_ib: drain QP references after partial table insertion [ Upstream
Commit: 4aaa2ab816c710be7385b31d08373edcdcd71656
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=4aaa2ab816c710be7385b31d08373edcdcd71656

Changed files:
  drivers/infiniband/hw/mana/qp.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=4aaa2ab816c710be7385b31d08373edcdcd71656

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

The original Linux kernel CVE announcement for CVE-2026-92513 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92513

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:L/UI:N/S:U/C:L/I:L/A:H
  The Best / paranoid CVSS vector: AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H
  The Best / paranoid CVSS score: 7

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: 5
  Paranoid ActionableScore: 6

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).