[CVE-2026-97421][MODERATE REGULAR] RDMA/umem: Be careful about boundary conditions in ib_umem_find_best_pgsz() [ Upstream

[CVE-2026-97421][MODERATE REGULAR] RDMA/umem: Be careful about boundary conditions in ib_umem_find_best_pgsz() [ Upstream

[AL-KERNEL]

CVE: CVE-2026-97421
Priority: MODERATE REGULAR
AL-KERNEL base severity: MODERATE
KPANIC flag: NO
Patch: RDMA/umem: Be careful about boundary conditions in ib_umem_find_best_pgsz() [ Upstream
Commit: ae3cf7d2c4d506c5c91746321644e1a08d989e2a
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=ae3cf7d2c4d506c5c91746321644e1a08d989e2a
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97421
Analysis date: Thu, 24 Sep 2026 13:31:55 -0400
ActionableScore: 2
ActionableScore lower bound: 1
Actionable bucket: Ordinary Moderate
Manual review required: NO

Summary:
No one-sentence report phrase was found.

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

The original Linux kernel CVE announcement for CVE-2026-97421 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97421

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 REGULAR
Manual review required: NO

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

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

CVE-2026-97421	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-97421 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:L/I:L/A:H';CWE-190;CWE-681;*CWE-20;Other CVSS 'AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:L';BEST CVSS score: '5.8';DESCR 'An integer truncation and boundary calculation flaw in ib_umem_find_best_pgsz can process a 64 bit IOVA as an unsigned long value and can mishandle an IOVA range whose end overflows. This is especially relevant on 32 bit systems and for large RDMA memory regions. A malformed page size mask could select a page size that does not match the actual IOVA boundaries and may cause an affected RDMA operation to fail or behave incorrectly. For the CVSS the PR:L is used because a local process needs access to an RDMA verbs device and must be able to register or otherwise use a suitable memory region. The issue is not directly network reachable. The base impact is limited to availability because the patch does not demonstrate an out of bounds access or a memory corruption primitive. The paranoid assessment retains limited confidentiality and integrity impact because an incorrect DMA page table mapping is a plausible but unproven consequence that depends on the RDMA driver and hardware behavior.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Ordinary Moderate (with actual score 2)	YES	OOB SIMPLEFIX NETWORK MEMORY  DECREASED_TO_MODERATEREG_BASED_ON_ACTIONABLESCORELESSTHAN4	NO	NO	checked

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

ActionableScore=2
ActionableScoreLower=1
PrimitiveClass=NONE
RealUAF=NO
HostAdminRequired=UNKNOWN
NamespaceScopedPrivilege=UNKNOWN
ImportantCandidate=NO
ManualReview=RECOMMENDED
AutoDowngradeAllowed=YES

1. ActionableScore

* Raw conservative score: 1
* Conservative score: 1
* Raw paranoid score: 2
* Paranoid score: 2
* Final recommended bucket: **Ordinary Moderate**::

No floor or cap was applied.

2. Signal breakdown

* +1 Memory layout and address-range invariant restoration. The patch restores correct handling of the relationship between a 64 bit IOVA, the registered range length, and the page-size bitmap used for RDMA page-table construction.
* +1 Paranoid only. A truncated or overflowed IOVA range calculation can select an invalid hardware page size mask. This creates a limited unresolved concern that an affected RDMA driver could construct an incorrect mapping or fail an RDMA operation. The patch does not establish that such a mapping causes a DMA write to an attacker-selected target.

Negative signals:

* No host-admin penalty was applied. The complete permissions required to access the relevant RDMA verbs operation are not established by the supplied patch.
* No hard-race penalty was applied. This is deterministic arithmetic and boundary handling, not a race.
* No rare-hardware penalty was applied. RDMA hardware may be specialized, but the supplied evidence does not prove a physical-only or unusually inaccessible prerequisite.

Important non-awarded signals:

* Controlled reclaim is not demonstrated.
* Arbitrary or directed kernel write is not demonstrated.
* An out-of-bounds access is not demonstrated.
* A use-after-free, stale-object use, double release, or refcount mutation is not demonstrated.
* A kernel crash is not demonstrated.
* Remote reachability is not established. The affected logic is reached through local RDMA memory-registration and mapping paths.
* Confidentiality and integrity signals are not awarded. Incorrect page-size selection alone does not establish disclosure, unauthorized modification, or a useful DMA corruption primitive.
* Strong exploitation leverage is not awarded. The possible downstream behavior depends on specific RDMA drivers and HCA hardware handling of an invalid page-size choice.

3. Reachability analysis

The evaluated attacker is a local process using an RDMA verbs device and a memory-registration or related RDMA mapping operation that reaches ib_umem_find_best_pgsz.

The patch establishes that the vulnerable computation accepts an IOVA argument and uses umem length and a page-size bitmap. It does not show the complete caller chain, device-node permissions, RDMA cgroup policy, device assignment policy, or whether an ordinary unprivileged user can create the required RDMA object on the evaluated system.

HostAdminRequired is therefore UNKNOWN. It would be incorrect to assume host-admin access merely because RDMA deployment is often administratively managed. It would also be incorrect to claim below-host-admin reachability without evidence that the complete RDMA registration and mapping sequence is available to an ordinary local user.

NamespaceScopedPrivilege is UNKNOWN. The supplied material does not establish that all required RDMA device, verbs, memory-registration, and DMA-mapping operations can be performed using only attacker-created namespaces or namespace-scoped capabilities.

No direct network trigger is shown. RDMA peer traffic alone is not shown to control the local virt argument, umem length, or page-size-selection call path. No prior privileged setup is proven by the patch, although practical RDMA device availability may depend on local policy and hardware provisioning.

The issue is especially relevant on 32 bit kernels because unsigned long can truncate the u64 IOVA. Large IOVA ranges and lengths above 4 GB are also relevant to the overflow and GENMASK boundary conditions. These are input and platform constraints, but not a narrow timing race.

4. Primitive and severity interpretation

The established primitive is an integer truncation and boundary-validation defect in ib_umem_find_best_pgsz. The old code accepted virt as unsigned long even though umem iova is u64. On 32 bit systems this can discard high IOVA bits. The old calculation can also overflow while computing the final virtual address and can derive an invalid GENMASK boundary for large lengths.

This is not an established UAF, OOB read, OOB write, stale-reference, double-free, ownership violation, or authorization bypass. The patch does not show a concrete malformed page table being accepted by hardware, a DMA transfer to an incorrect target, a kernel crash, or attacker-controlled post-validation memory corruption.

The direct practical consequence supported by the patch is incorrect selection of a page size or rejection behavior for an RDMA memory mapping operation. In the conservative interpretation this is an invalid-state or operation-failure issue. The paranoid interpretation recognizes that page-size selection feeds RDMA mapping construction, so driver- and device-specific review is appropriate to verify whether a bad selected page size could result in an incorrect DMA translation instead of a clean operation failure.

PrimitiveClass is NONE because the patch establishes arithmetic validation defects but no meaningful corruption, lifetime, ownership, authorization, or security-boundary primitive. RealUAF is NO.

ImportantCandidate is NO because there is no demonstrated exploitation leverage beyond an invalid page-size calculation. AutoDowngradeAllowed is YES because no unresolved concrete UAF, write primitive, ownership violation, authorization bypass, or active-object lifetime corruption is established. Manual review remains recommended because RDMA page-table and HCA behavior are driver-specific, and the supplied patch does not include relevant call sites or hardware error-handling behavior.

Likely operational severity is Low-like or ordinary Moderate. The issue should be patched in affected RDMA-enabled environments, especially for 32 bit kernels or deployments using large IOVA ranges, but the supplied evidence does not support Important classification.

5. One-sentence report phrase

An integer truncation and overflow validation flaw in RDMA umem page-size selection can derive an incorrect IOVA page-size mask for large mappings, especially on 32 bit kernels, but no concrete memory-corruption or privilege-escalation primitive is established.

6. Manual review recommendation

MANUAL CHECK RECOMMENDED

Review callers of ib_umem_find_best_pgsz and affected RDMA driver page-table programming paths to determine whether an invalid selected page size is rejected safely or can produce a device-visible DMA mapping inconsistency.

7. Call-site confidence

Call-site confidence: medium

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

Patch: RDMA/umem: Be careful about boundary conditions in ib_umem_find_best_pgsz() [ Upstream
Commit: ae3cf7d2c4d506c5c91746321644e1a08d989e2a
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=ae3cf7d2c4d506c5c91746321644e1a08d989e2a

Changed files:
  drivers/infiniband/core/umem.c
  include/rdma/ib_umem.h

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=ae3cf7d2c4d506c5c91746321644e1a08d989e2a

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

The original Linux kernel CVE announcement for CVE-2026-97421 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97421

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:L
  The Best / paranoid CVSS vector: AV:L/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:H
  The Best / paranoid CVSS score: 5.8

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: 1
  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: NO
Published priority for this report (same as in Subject): MODERATE REGULAR

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).