[CVE-2026-97496][IMPORTANT] drm/amdkfd: Fix OOB memory exposure in get_wave_state() [ Upstream

[CVE-2026-97496][IMPORTANT] drm/amdkfd: Fix OOB memory exposure in get_wave_state() [ Upstream

[AL-KERNEL]

CVE: CVE-2026-97496
Priority: IMPORTANT
AL-KERNEL base severity: IMPORTANT
KPANIC flag: YES
Patch: drm/amdkfd: Fix OOB memory exposure in get_wave_state() [ Upstream
Commit: d9183d974ddd5f09d029beaf359275ac20d4d5fe
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=d9183d974ddd5f09d029beaf359275ac20d4d5fe
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97496
Analysis date: Thu, 24 Sep 2026 18:13:25 -0400
ActionableScore: 7
ActionableScore lower bound: 6
Actionable bucket: Strong Important candidate
Manual review required: YES

Summary:
A local AMD KFD user can supply crafted MQD restore values that cause an unbounded out of bounds read through copy_to_user and disclose adjacent GTT or kernel memory including queue state and KASLR related pointers.

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

The original Linux kernel CVE announcement for CVE-2026-97496 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97496

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: IMPORTANT
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-2026-97496	IMPORTANT	https://www.kernelcve.org/list/?q=CVE-2026-97496 CHECK WITH IMPACT FROM ORIG NN LOW	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H';CWE-125;*CWE-200;CWE-191;*CWE-20;Other CVSS 'AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N';BEST CVSS score: '7.1';DESCR 'An out of bounds kernel memory disclosure occurs in AMD KFD get_wave_state because cp_hqd_cntl_stack_size and cp_hqd_cntl_stack_offset are trusted when calculating the source offset and copy length. A local attacker can provide crafted MQD values through the CRIU restore path and AMDKFD_IOC_RESTORE_PROCESS. An offset greater than the size can cause an integer underflow and produce a copy length close to 4 GiB. A size larger than the allocated control stack can expose adjacent GTT or kernel memory including queue MQDs, ring buffers, and KASLR related pointers. For the CVSS the PR:L value is selected because the attacker needs local code execution and access to the AMD KFD device and restore interface, but does not necessarily need full administrative control. The issue is not network reachable. The confirmed primitive is an out of bounds read through copy_to_user, so confidentiality impact is high while no direct kernel memory write primitive is present. The paranoid score includes high availability impact because a very large invalid copy may stall the system or trigger a kernel fault. This vulnerability requires manual review because the disclosed kernel data may assist further exploitation.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Strong Important candidate (with actual score 7)  SKIP CVE-2026-97496 UNKNOWN SKIP No affected files built, so skip this CVE NO  - - unknown 	MAYBE	KASAN OOB LEAK HARDWARE KPANIC MEMORY  KPANIC INCREASED_TO_HIGH_BASED_ON_GUESSCVSS	-	-	checked

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

ActionableScore=7
ActionableScoreLower=6
PrimitiveClass=CLASS_A
RealUAF=NO
HostAdminRequired=NO
NamespaceScopedPrivilege=UNKNOWN
ImportantCandidate=YES
ManualReview=REQUIRED
AutoDowngradeAllowed=NO

## 1. ActionableScore

* Raw conservative score: 6
* Conservative score: 6
* Raw paranoid score: 7
* Paranoid score: 7
* No floor or cap was applied.
* Final recommended bucket: **Strong Important candidate**::

## 2. Signal breakdown

Conservative signals:

* +2 — Substantial kernel disclosure primitive. Attacker-controlled MQD size and offset values produce a user-visible OOB read from adjacent GTT or kernel memory.
* +1 — Local unprivileged trigger. The supplied evidence establishes a local attacker using the KFD restore interface without requiring host-admin-equivalent control.
* +1 — High-control kernel API. The attacker controls the lengths and offsets that determine the source range supplied to `copy_to_user()`.
* +1 — Practical exploit chain. Crafted restore state directly leads to disclosure of adjacent MQDs, ring buffers, and KASLR-related pointers.
* +1 — Confidentiality impact. The disclosed bytes are returned to userspace and may contain security-sensitive kernel information.

Paranoid-only signal:

* +1 — Strong DoS is technically plausible because an underflow can request a copy approaching 4 GiB from an invalid or excessively large kernel mapping, potentially causing a kernel fault or severe stall.

Not awarded:

* No integrity point. The established operation reads kernel memory and copies it to userspace. It does not write to the source or establish a kernel write primitive.
* No UAF or lifetime-corruption points.
* No remote-reachability point.
* No controlled-reclaim, callback, credential, or privilege-escalation point.
* No additional availability point beyond the paranoid crash or stall signal.
* No rare-hardware penalty. AMD GPU support is hardware-specific, but the supplied evidence does not establish genuinely rare or physical-only hardware.

## 3. Reachability analysis

A local process with access to the AMD KFD device and `AMDKFD_IOC_RESTORE_PROCESS` can supply restored MQD state containing malicious control-stack size and offset values. The supplied analysis positively places this below host-admin-equivalent privilege, although device-node access or membership in an authorized graphics group may still be required.

The operation is local and is not reachable through network traffic. No evidence establishes that the complete path is user-namespace scoped, so `NamespaceScopedPrivilege=UNKNOWN`.

The system must have a supported AMD GPU, the AMDGPU/KFD stack enabled, and the relevant device exposed to the attacker. These are feature prerequisites rather than evidence of rare hardware. There is no race or timing requirement. A crafted restore request directly establishes the invalid arithmetic and OOB copy.

## 4. Primitive and severity interpretation

The exact primitive is an attacker-directed OOB kernel read with direct disclosure through `copy_to_user()`. If the supplied offset exceeds the supplied size, unsigned subtraction can create a length near 4 GiB. Independently, a size such as 1 MiB against a 4 KiB control-stack allocation causes the copy to read adjacent memory.

This is not a passive internal invalid read. The data leave the kernel and can include other queues' MQDs, ring buffers, and KASLR-related pointers. That makes it a substantial disclosure and therefore `CLASS_A`.

There is no real UAF, repeated release, stale-object use, or demonstrated write primitive. Integrity impact and privilege escalation remain speculative and are not scored. The paranoid score only adds the defensible possibility of host DoS from the extremely large invalid copy.

`ImportantCandidate=YES` because the established disclosure can expose substantial adjacent kernel or GTT contents and exploitation-relevant pointers. `AutoDowngradeAllowed=NO` because automatically treating this as an ordinary moderate invalid read would discard a concrete attacker-controlled kernel disclosure primitive.

Likely operational severity is Important where untrusted local GPU-compute users have access to KFD and the CRIU restore path.

## 5. One-sentence report phrase

A local AMD KFD user can supply crafted MQD restore values that cause an unbounded out of bounds read through copy_to_user and disclose adjacent GTT or kernel memory including queue state and KASLR related pointers.

## 6. Manual review recommendation

MANUAL CHECK REQUIRED

The issue provides a concrete, attacker-controlled kernel disclosure primitive, and manual review should confirm device permissions, restore-path availability, affected GPU generations, and the practical extent of adjacent memory exposure.

## 7. Call-site confidence

Call-site confidence: high

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

Patch: drm/amdkfd: Fix OOB memory exposure in get_wave_state() [ Upstream
Commit: d9183d974ddd5f09d029beaf359275ac20d4d5fe
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=d9183d974ddd5f09d029beaf359275ac20d4d5fe

Changed files:
  drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.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=d9183d974ddd5f09d029beaf359275ac20d4d5fe

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

The original Linux kernel CVE announcement for CVE-2026-97496 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97496

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:L/PR:L/UI:N/S:U/C:H/I:N/A:N
  The Best / paranoid CVSS vector: AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H
  The Best / paranoid CVSS score: 7.1

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: 6
  Paranoid ActionableScore: 7

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: IMPORTANT
KPANIC detected for this report: YES
Published priority for this report (same as in Subject): IMPORTANT

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).