[CVE-2026-92507][MODERATE 7.0] RDMA/core: Fix potential use after free in ib_dealloc_pd_user() [ Upstream

[CVE-2026-92507][MODERATE 7.0] RDMA/core: Fix potential use after free in ib_dealloc_pd_user() [ Upstream

[AL-KERNEL]

CVE: CVE-2026-92507
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: RDMA/core: Fix potential use after free in ib_dealloc_pd_user() [ Upstream
Commit: df0e9efe61ca706a07b1c9bf93fcb9032455efc4
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=df0e9efe61ca706a07b1c9bf93fcb9032455efc4
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92507
Analysis date: Thu, 17 Sep 2026 22:03:42 -0400
ActionableScore: 5
ActionableScore lower bound: unknown
Actionable bucket: Actionable Moderate / Strong Important candidate requiring manual review
Manual review required: YES

Summary:
RDMA core can leave an ib_pd visible through restrack while ib_dealloc_pd_user() destroys internal resources, allowing a local RDMA capable user to race netlink resource inspection against deallocation and potentially trigger a kernel use after free.

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

The original Linux kernel CVE announcement for CVE-2026-92507 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92507

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-2026-92507	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-92507 CHECK WITH IMPACT FROM ORIG NN MODERATE	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:N/I:N/A:H';BEST CVSS score: '7';DESCR 'A race condition in RDMA core can leave an ib_pd visible through restrack while ib_dealloc_pd_user() is already destroying internal and vendor specific resources. A local user with access to RDMA device interfaces or RDMA restrack netlink paths may race resource inspection against PD deallocation and reach a use after free condition. For the CVSS the PR:L is used because reliable triggering can be possible with local RDMA access or delegated device access rather than full administrative privileges. The issue is not directly network reachable through normal RDMA traffic and requires local control plane interaction. Impact is at least local denial of service via kernel crash. In the worst case it may allow confidentiality or integrity impact due to kernel use after free and should be reviewed manually.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate / Strong Important candidate requiring manual review (with actual score 5)	YES	INIT UAF NETWORK LINUS KPANIC 	NO	NO	checked

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

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

## 1. ActionableScore

* Raw conservative score: 4
* Conservative score: 5
* Raw paranoid score: 5
* Paranoid score: 5
* Applied conservative floor/cap: Real UAF FLOOR 5 applied because this is an active structured lifetime bug reachable below confirmed host-admin and the stale object remains externally discoverable through restrack while destruction is in progress.
* Final recommended bucket: **Actionable Moderate / Strong Important candidate requiring manual review**::

## 2. Signal breakdown

Positive signals:

* +2 Real UAF / lifetime violation: the commit directly states that the PD remains accessible through restrack while vendor-specific resources associated with the PD may already be freed, creating a potential use-after-free.
* +1 Reliable kernel crash / strong DoS: stale access to partially destroyed RDMA core/vendor resources can realistically produce an Oops or kernel crash.
* +1 Attacker-influenced active-object lifetime: a local actor with RDMA/uverbs and restrack access can plausibly create/use/deallocate PD objects and race resource inspection against destruction.
* +1 Paranoid structured stale-object leverage: the stale object is an active RDMA protection-domain object reachable through restrack/netlink, and the stale path is not just a passive diagnostic pointer. It may consume structured resource state after associated resources are freed.

Non-awarded / withheld signals:

* +2 Strong exploitation leverage not awarded: controlled reclaim, attacker-shaped replacement object, arbitrary write, callback control, or refcount takeover is not demonstrated.
* Concrete repeated release/refcount +1 not awarded: the patch does not show an exact same dangling reference receiving a second lifetime-mutating put/release/free after a first release may have ended lifetime.
* Paranoid repeated-refcount +1 not awarded: no concrete second refcount/lifetime mutation through the same stale reference is shown.
* Weak LPE +1 not separately awarded: any LPE concern is already represented by the unresolved active-object UAF and structured stale-object review need.
* Extra Availability +1 not awarded: the same potential kernel crash is already counted once.
* Confidentiality and Integrity points not awarded: no disclosure path, arbitrary read/write, protected object modification, or demonstrated privilege escalation is established.
* Remote reachable +2 not awarded: this is not directly reachable through normal RDMA network traffic. It requires local control-plane interaction.
* Race/timing penalty not applied: there is a short race window, but the attacker may be able to repeatedly exercise create/deallocate and restrack inspection paths. The evidence does not prove a one-shot or externally uncontrollable race.
* Host-admin penalty not applied: full host-admin-equivalent privilege is not positively established.

Repeated-release/refcount checklist:

* First release can end lifetime: not established for the exact same restrack-visible reference.
* Exact same reference survives: stale PD visibility through restrack is established, but not a second release of the exact same reference.
* Second operation is a real lifetime/refcount mutation: not established.
* Dedicated +1 Concrete repeated release/refcount awarded: no.
* Paranoid repeated-refcount leverage awarded: no.
* FLOOR 5 applied: yes, due to active structured UAF surface, not due to repeated release/refcount.

## 3. Reachability analysis

The likely attacker is a local user or process with access to RDMA device interfaces and RDMA restrack/netlink inspection paths. The vulnerable condition arises when PD destruction in `ib_dealloc_pd_user()` overlaps with netlink/restrack access that can still obtain the PD through `rdma_restrack_get()`.

This is not a direct network packet-processing vulnerability. Normal remote RDMA traffic is not shown to reach the vulnerable operation by itself.

Full host-admin-equivalent privilege is not established. RDMA device access may be delegated through device permissions, service accounts, or RDMA-capable workloads, so treating this as strictly host-root-only would be an unsupported restrictive assumption.

Namespace-scoped privilege is unresolved. The input does not establish that the complete create/deallocate plus restrack inspection sequence is available inside attacker-created namespaces.

Prior setup may include RDMA-capable hardware, configured RDMA devices, and accessible uverbs/restrack interfaces. These are deployment prerequisites, but not enough by themselves to apply a rare-hardware or host-admin penalty.

Race repeatability is plausible because PD lifecycle operations and restrack inspection can be attempted repeatedly, but the exact timing window remains relevant.

## 4. Primitive and severity interpretation

The exact primitive is a lifetime ordering bug: `rdma_restrack_del()` happened too late in `ib_dealloc_pd_user()`, leaving the PD visible through restrack after internal/vendor-specific resources could already be freed.

This is an active structured object case, not fixed stale storage. The stale object is an RDMA protection domain and associated resource state reachable through a kernel resource-tracking interface.

The demonstrated consequence is potential kernel use-after-free and likely local DoS through crash. The speculative consequence is broader exploitation from stale structured RDMA resource access, but no controlled reclaim, arbitrary read/write, or demonstrated privilege escalation is established.

PrimitiveClass is CLASS_B because a real kernel lifetime/UAF primitive is established, but strong Class A exploitation primitives are not.

ImportantCandidate is YES because this is an active structured UAF in RDMA core with unresolved exploitation leverage beyond a simple NULL dereference.

AutoDowngradeAllowed is NO because automatic closure would discard a concrete kernel lifetime primitive where stale externally visible resource access remains meaningfully unresolved.

Operational severity is at least Actionable Moderate. It is not proven Important, but it requires manual review for possible stronger impact.

## 5. One-sentence report phrase

RDMA core can leave an ib_pd visible through restrack while ib_dealloc_pd_user() destroys internal resources, allowing a local RDMA capable user to race netlink resource inspection against deallocation and potentially trigger a kernel use after free.

## 6. Manual review recommendation

MANUAL CHECK REQUIRED

Reason: real active-object UAF is established, reachable through RDMA control-plane paths below confirmed host-admin, and exploitation leverage beyond crash is unresolved.

## 7. Call-site confidence

Call-site confidence: medium

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

Patch: RDMA/core: Fix potential use after free in ib_dealloc_pd_user() [ Upstream
Commit: df0e9efe61ca706a07b1c9bf93fcb9032455efc4
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=df0e9efe61ca706a07b1c9bf93fcb9032455efc4

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=df0e9efe61ca706a07b1c9bf93fcb9032455efc4

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

The original Linux kernel CVE announcement for CVE-2026-92507 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-92507

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: unknown
  Paranoid ActionableScore: 5

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).