[CVE-2026-92512][MODERATE 7.0] RDMA/core: Fix use after free in ib_query_qp() [ Upstream

[CVE-2026-92512][MODERATE 7.0] RDMA/core: Fix use after free in ib_query_qp() [ Upstream

[AL-KERNEL]

CVE: CVE-2026-92512
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: RDMA/core: Fix use after free in ib_query_qp() [ Upstream
Commit: 2ef2ea52ae41bd9efa83139e97ae839de7e17156
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=2ef2ea52ae41bd9efa83139e97ae839de7e17156
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92512
Analysis date: Fri, 18 Sep 2026 00:33:21 -0400
ActionableScore: 6
ActionableScore lower bound: 5
Actionable bucket: Actionable Moderate / Strong Important candidate requiring manual review
Manual review required: YES

Summary:
A local RDMA user can race RDMA netlink QP dumping against QP destruction so that `ib_query_qp()` accesses vendor-specific QP resources after teardown has started, causing a real UAF with demonstrated kernel crash and unresolved exploitation leverage.

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

The original Linux kernel CVE announcement for CVE-2026-92512 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92512

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-2026-92512	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-92512 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-667;Other CVSS 'AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H';BEST CVSS score: '7';DESCR 'A race in RDMA QP resource tracking can allow a QP to remain visible to the netlink query path while ib_destroy_qp_user is already destroying vendor specific QP resources. A local process can race RDMA netlink QP dumping against QP destruction, causing ib_query_qp to access stale or partially destroyed kernel driver state and triggering a use-after-free condition. For the CVSS the PR:L is used because reliable triggering requires local user access to RDMA userspace or RDMA netlink interfaces, not full administrative privileges in the common practical model. The issue is not directly network reachable because the vulnerable path is the local RDMA control plane rather than remote packet processing. Impact is at least local denial of service via kernel crash. In the paranoid interpretation, the UAF bug class and vendor object lifetime mismatch may allow confidentiality or integrity impact and possible privilege escalation, so this case should be reviewed manually. YES REQUIRES MANUAL CHECK';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate / Strong Important candidate requiring manual review (with actual score 6)	YES	INIT UAF NETWORK KERNEL_PANIC_PLUS_UAF LINUS KPANIC  DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFDH	NO	NO	checked

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

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

## 1. ActionableScore

* Raw conservative score: 5
* Conservative score: 5
* Raw paranoid score: 6
* Paranoid score: 6
* Floor/cap changes: none
* Final recommended bucket: **Actionable Moderate / Strong Important candidate requiring manual review**::

## 2. Signal breakdown

Conservative signals:

* +2 Real UAF / active lifetime violation. The QP remains visible through RDMA restrack while `ib_destroy_qp_user()` has already destroyed vendor-specific QP resources, allowing `ib_query_qp()` through the netlink dump path to access stale or partially destroyed state.
* +1 Reliable kernel crash / strong DoS. The supplied trace shows a kernel NULL dereference in `ib_query_qp()` via `mlx5_ib_query_qp()` during RDMA netlink QP dumping.
* +1 Local below-host-admin trigger. The vulnerable path is reachable from local RDMA userspace / RDMA netlink activity and the supplied context supports `PR:L` rather than host-admin-only triggering.
* +1 Attacker-influenced repeatable lifetime manipulation. A local RDMA user can plausibly create/destroy QPs and concurrently request QP resource dumps, repeatedly driving the destroy/query lifetime window.

Paranoid additional signal:

* +1 Structured stale-object exploitation leverage. The stale object is an active structured RDMA QP whose vendor-specific state is consumed by the query path after teardown has started. This is more than a passive cleanup pointer, but controlled reclaim or arbitrary read/write is not demonstrated.

Non-awarded signals:

* +0 Strong LPE / strong exploitation leverage. Controlled reclaim, object replacement, callback control, arbitrary write, or demonstrated privilege escalation is not established.
* +0 Concrete repeated release/refcount. The patch does not show the same exact dangling reference receiving a second lifetime/refcount mutation.
* +0 Paranoid repeated-refcount leverage. No second stale `put` / `release` / refcount mutation is established.
* +0 Remote reachability. The vulnerable operation is RDMA control-plane netlink/query handling, not direct remote packet processing.
* +0 Extra availability. The crash is already counted once.
* +0 Confidentiality / integrity as direct signals. C/I impact is possible in a paranoid CVSS sense for a UAF, but no concrete disclosure, arbitrary write, or privilege-escalation primitive is shown.
* No hard-race penalty. The issue is a race, but the attacker can plausibly repeat QP creation/destruction and netlink dumping. The evidence does not prove a one-shot or externally uncontrollable timing window.
* No rare-hardware penalty. RDMA hardware/device access is a deployment condition, but the evidence does not justify treating it as physical-only or too rare for triage purposes.

Repeated-release/refcount checklist:

* First release can end lifetime: not applicable as a repeated-release primitive.
* Exact same reference survives to a second release: not established.
* Second operation is lifetime/refcount mutation: not established.
* Concrete repeated release/refcount +1 awarded: no.
* Paranoid repeated-refcount +1 awarded: no.
* FLOOR 5 applied: no. Conservative raw score already reaches 5.

## 3. Reachability analysis

The evaluated attacker is a local user or process with access to RDMA userspace objects and the RDMA netlink query path. The trigger requires racing QP destruction against a netlink QP resource dump so that restrack lookup can expose a QP whose vendor-specific resources are already being torn down.

Host-admin-equivalent privilege is not positively established. The practical model is below host admin, consistent with `PR:L`, although exact deployment policy may depend on RDMA device permissions and distribution configuration.

Namespace-scoped privilege is unresolved. The input does not establish that the complete QP creation, destruction, and RDMA netlink query sequence is fully available inside attacker-created namespaces.

Prior host setup may exist in the sense that RDMA hardware and drivers must be present and usable, but no separate host-admin setup step is proven as part of the vulnerability trigger. The race appears repeatable through repeated QP lifecycle operations and dump requests.

## 4. Primitive and severity interpretation

The exact primitive is a real use-after-free / stale lifetime use of an RDMA QP and its vendor-specific resources. This is an active structured object, not fixed stale storage.

The stale use is not merely a destructor-only cleanup fault. The stale QP is consumed by the normal RDMA netlink query path through `fill_res_qp_entry_query()`, `ib_query_qp()`, and the mlx5 provider query implementation after teardown has already invalidated provider resources.

This is `CLASS_B`: a real kernel lifetime/corruption primitive requiring exploitation-leverage analysis. It is not `CLASS_A` because controlled reclaim, directed corruption, arbitrary kernel read/write, callback control, or demonstrated LPE is not supplied.

Demonstrated consequence is local kernel crash / DoS. Speculative consequence is broader memory corruption or privilege escalation due to UAF in structured RDMA provider state, but that is not proven. ImportantCandidate is `YES` because the bug is a below-host-admin active-object UAF with unresolved exploitation leverage. AutoDowngradeAllowed is `NO` because automatically closing it as ordinary crash-only Moderate would discard a concrete active lifetime primitive in kernel RDMA object handling.

Likely operational severity: actionable Moderate at minimum, with Important review warranted.

## 5. One-sentence report phrase

A local RDMA user can race RDMA netlink QP dumping against QP destruction so that `ib_query_qp()` accesses vendor-specific QP resources after teardown has started, causing a real UAF with demonstrated kernel crash and unresolved exploitation leverage.

## 6. Manual review recommendation

MANUAL CHECK REQUIRED

Reason: this is a real active-object UAF in RDMA QP lifetime handling, reachable below host-admin in the practical model, with demonstrated crash and unresolved post-lifetime exploitation potential.

## 7. Call-site confidence

Call-site confidence: high

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

Patch: RDMA/core: Fix use after free in ib_query_qp() [ Upstream
Commit: 2ef2ea52ae41bd9efa83139e97ae839de7e17156
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=2ef2ea52ae41bd9efa83139e97ae839de7e17156

Changed files:
  drivers/infiniband/core/verbs.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=2ef2ea52ae41bd9efa83139e97ae839de7e17156

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

The original Linux kernel CVE announcement for CVE-2026-92512 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92512

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:N/I:N/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).