[CVE-2026-98687][MODERATE REGULAR] thunderbolt: Reject oversized XDomain properties responses

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

CVE: CVE-2026-98687
Priority: MODERATE REGULAR
AL-KERNEL base severity: MODERATE
KPANIC flag: NO
Patch: thunderbolt: Reject oversized XDomain properties responses
Commit: a766a86ff48acee76aeb8e2575771040749e509d
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=a766a86ff48acee76aeb8e2575771040749e509d
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98687
Analysis date: Mon, 19 Oct 2026 01:10:44 -0400
ActionableScore: 1
ActionableScore lower bound: unknown
Actionable bucket: Ordinary Moderate
Manual review required: NO

Summary:
A malicious Thunderbolt XDomain peer can advertise an oversized properties response that causes a 12-byte heap out-of-bounds read, but no disclosure, kernel write, reliable panic, or privilege-escalation primitive is established.

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

The original Linux kernel CVE announcement for CVE-2026-98687 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98687

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-2026-98687	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-98687 CHECK WITH IMPACT FROM ORIG NN IMPORTANT | AEGIS IMPACT MODERATE_KPANIC AEGIS_AS=1 LEGACY=MODERATE RESULT=DIFF (final Legacy and Aegis differ; FINAL: MODERATE, Aegis: MODERATE_KPANIC)	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:A/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:H';CWE-125;CWE-126;*CWE-20;Other CVSS 'AV:A/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L';BEST CVSS score: '7.6';DESCR 'A heap out of bounds read occurs in tb_xdp_properties_request because a malicious XDomain peer can advertise 48 data dwords while the fixed response structure contains space for only 45. The resulting memcpy reads 192 bytes from the 180 byte res data array and therefore reads 12 bytes beyond the source heap object. The issue is adjacent rather than generally Internet reachable because triggering requires access to a Thunderbolt XDomain connection or fabric. For the CVSS the PR:N is selected because the malicious peer does not need a local account or privileges on the target system. The demonstrated primitive is an out of bounds read and not an out of bounds write or UAF. Base impact is limited to availability because the supplied code does not establish that adjacent heap data is returned to the peer. In the paranoid interpretation the copied bytes may enter downstream property parsing and could cause limited confidentiality or integrity impact and a host wide crash. There is no supported arbitrary write or privilege escalation primitive. Automated impact decision path. The original SimpleNet classifier selected IMPORTANT. The primary LLM estimated paranoid CVSS 7.6 and the production ActionableScore returned 1 with recommendation Ordinary Moderate. The production rules then evaluated CVSS thresholds and ActionableScore thresholds. Recorded rule transitions were ActionableScore thresholds stage IMPORTANT to MODERATE because the ActionableScore threshold rule was applied. The legacy KP specialist has no review artifact for this run. The legacy FD or DH downgrade specialist has no review artifact for this run. Exact triggered rule markers were DECREASED_TO_MODERATEREG_BASED_ON_ACTIONABLESCORELESSTHAN4 and V41_INHERITED_KPANIC_PROMOTION_BLOCKED and MANUAL_REVIEW_REQUIRED. The independent weaker Aegis shadow evaluated CVSS 4.3 and ActionableScore 1 with primitive class CLASS_C and manual review REQUIRED. Its independently reconciled impact was MODERATE_KPANIC. The models disagreed but the aggregation rules retained the production result. The final reported impact is MODERATE.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Ordinary Moderate (with actual score 1)	MAYBE	OOB SIMPLEFIX IMPROVEONLY LINUS  DECREASED_TO_MODERATEREG_BASED_ON_ACTIONABLESCORELESSTHAN4 V41_INHERITED_KPANIC_PROMOTION_BLOCKED MANUAL_REVIEW_REQUIRED	-	-	checked

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

ActionableScore=1
PrimitiveClass=CLASS_C
RealUAF=NO
HostAdminRequired=NO
NamespaceScopedPrivilege=NO
ImportantCandidate=NO
ManualReview=RECOMMENDED
AutoDowngradeAllowed=YES
RequiresHostUID0=NO
AuthorizationDecisionBypass=NO
ProtectedResourceImpact=NONE
B4PendingReview=NO
B4PendingConfirmed=NO

1. ActionableScore

* Raw conservative score: 1
* Conservative score: 1
* Raw paranoid score: 1
* Paranoid score: 1
* No floor or cap was applied.
* Final recommended bucket: **Ordinary Moderate**::

The conservative and paranoid scores are identical because the patch establishes only a bounded 12-byte heap out-of-bounds read. It does not establish disclosure, a write, corrupted active state, reliable host panic, or privilege escalation.

2. Signal breakdown

Positive signals:

* +1 Firmware-mediated external influence. A malicious Thunderbolt XDomain peer controls the six-bit response length consumed through the Thunderbolt controller path.
* +0 Limited OOB read. `memcpy()` reads 192 bytes from a 180-byte `res->data` array, but this is not a write primitive and has no demonstrated disclosure path.
* +0 Weak corruption concern. The three extra dwords are copied into the assembled property block, but no security-sensitive downstream use of those bytes is established.

Negative signals:

* No negative points were applied.
* No physical-only penalty was applied because the supplied evidence establishes an adjacent XDomain peer but does not prove that attacker presence must be limited to direct physical attachment in every affected topology.
* No rare-subsystem penalty was applied merely because Thunderbolt XDomain is hardware-specific.
* No constrained-corruption penalty was applied because the limited-read classification already excludes corruption bonuses. Applying another reduction would double-count the same limitation.

Important non-awarded signals:

* No strong or constrained overwrite exists. The defect is a source-side OOB read.
* Confidentiality was not awarded because the extra heap bytes are not shown to be returned to the peer, userspace, or another observable endpoint.
* Integrity was not awarded because no protected kernel state or active descriptor is shown to be modified.
* Reliable kernel crash was not awarded. A fault is possible at an unfavorable allocation boundary or under memory debugging, but a host-wide panic is not established.
* Arbitrary read, arbitrary write, controlled target selection, controlled reclaim, UAF, callback influence, and LPE are not demonstrated.
* Remote IP reachability is not established. The attacker must be an adjacent Thunderbolt XDomain peer.
* NEW52 B5-R does not apply. Condition 1 fails because this is an OOB read rather than an OOB write. Conditions 2 and 4 also fail because there is no retry-accumulated offset and no corrupted written state later consumed.
* NEW53 B3-PENDING does not apply because no shared-backing write occurs.
* V33 does not apply because attacker-influenced corruption of active structured state is not established.
* V32 and V34 do not apply because no skb checksum metadata or checksum-completion write is involved.
* B4-PENDING does not apply because there is no authorization decision or DENY-to-ALLOW transition.

3. Reachability analysis

The triggering actor is a malicious Thunderbolt XDomain peer that supplies an oversized response length. It does not need a local account, host UID 0, a capability, or host-administrator authority, so `HostAdminRequired=NO` and `RequiresHostUID0=NO`.

This is adjacent hardware-fabric reachability rather than general network reachability. The vulnerable host must have Thunderbolt XDomain support active and communicate with the peer. Existing platform authorization or device-enrollment policy may restrict practical exposure, but no such mandatory prerequisite is established by the supplied patch.

User and network namespaces are not part of the trigger. No namespace-scoped capability is required or relevant to the malicious peer, so `NamespaceScopedPrivilege=NO`.

There is no race, retry timing, or repeated lifetime transition. A crafted response with a sufficiently large header-derived length is enough to reach the invalid read once the XDomain exchange occurs.

4. Primitive and severity interpretation

The exact primitive is a 12-byte heap out-of-bounds source read:

* advertised data length after header adjustment: 48 dwords;
* available `res->data` capacity: 45 dwords;
* attempted read: 192 bytes;
* valid source region: 180 bytes;
* excess read: 12 bytes.

This is neither active structured-object lifetime corruption nor fixed stale storage. There is no release, free, stale reference, second refcount operation, or post-lifetime access. `RealUAF=NO`.

The copied extra bytes enter an assembled property block, but the supplied call path does not show that they are disclosed or used to modify protected state. The demonstrated consequence is therefore an invalid kernel heap read. A diagnostic report or fault is possible, but a reliable production panic remains unresolved.

`PrimitiveClass=CLASS_C` is appropriate because the established behavior is a bounded invalid read without an independently useful disclosure, write, lifetime, or authorization primitive.

`ImportantCandidate=NO` because no Class A primitive and no strong unresolved Class B primitive exists. `AutoDowngradeAllowed=YES` because the evidence supports ordinary Moderate or Low-like handling after review rather than mandatory Important preservation.

Likely operational severity is Low to Moderate. Exposure is limited to systems using Thunderbolt XDomain communication with an attacker-controlled or compromised peer.

5. One-sentence report phrase

A malicious Thunderbolt XDomain peer can advertise an oversized properties response that causes a 12-byte heap out-of-bounds read, but no disclosure, kernel write, reliable panic, or privilege-escalation primitive is established.

6. Manual review recommendation

MANUAL CHECK RECOMMENDED

Confirm whether the assembled property block can reach any peer-visible output or security-sensitive downstream property consumer. The current evidence does not justify mandatory Important preservation.

7. Call-site confidence

Call-site confidence: medium

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

Patch: thunderbolt: Reject oversized XDomain properties responses
Commit: a766a86ff48acee76aeb8e2575771040749e509d
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=a766a86ff48acee76aeb8e2575771040749e509d

Commit description:

tb_xdp_properties_request() allocates room for 45 data dwords in its
252-byte response buffer.  The XDomain length field is six bits wide,
however, and a malicious peer can set it to 63.  After the fixed response
fields are subtracted, the driver treats this as 48 data dwords.

Commit 322e934 ("thunderbolt: Clamp XDomain response data copy to
allocation size") only bounds the copy against data_len.  If data_len is
at least 48, memcpy() reads 192 bytes from the 180-byte res->data array,
causing a 12-byte heap out-of-bounds read.  Commit 4db2bd2
("thunderbolt: Limit XDomain response copy to actual frame size") limits
the earlier copy but does not constrain this header-derived length.

Reject response data lengths that exceed the allocated source buffer
before copying them into the assembled property block.

Fixes: d1ff702 ("thunderbolt: Add support for XDomain discovery protocol")
Cc: stable@vger.kernel.org
Assisted-by: LLM
Signed-off-by: Daehyeon Ko <4ncienth@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Changed files:
  drivers/thunderbolt/xdomain.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=a766a86ff48acee76aeb8e2575771040749e509d

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

The original Linux kernel CVE announcement for CVE-2026-98687 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98687

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:A/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
  The Best / paranoid CVSS vector: AV:A/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:H
  The Best / paranoid CVSS score: 7.6

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: 1

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).