[CVE-2026-72129][MODERATE 7.0] nvmet-rdma: handle inline data with a nonzero offset

[CVE-2026-72129][MODERATE 7.0] nvmet-rdma: handle inline data with a nonzero offset

[AL-KERNEL]

CVE: CVE-2026-72129
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: nvmet-rdma: handle inline data with a nonzero offset
Commit: c2106ba1b14d644a5203bea1a50dbe25dcad713c
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=c2106ba1b14d644a5203bea1a50dbe25dcad713c
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-72129
Analysis date: Sat, 15 Aug 2026 10:06:03 -0400
ActionableScore: 8
ActionableScore lower bound: 7
Actionable bucket: Strong Important candidate / Actionable Moderate at minimum
Manual review required: YES

Summary:
A host-controlled inline data offset in `nvmet_rdma_use_inline_sg()` can underflow scatterlist length calculation when `inline_data_size > PAGE_SIZE`, causing the NVMe/RDMA target backend to read far beyond the intended inline buffer and potentially leading to kernel memory exposure, storage corruption, or target DoS.

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

The original Linux kernel CVE announcement for CVE-2026-72129 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-72129

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-2026-72129	MODERATE	CHECK WITH IMPACT FROM ORIG NN IMPORTANT	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H';CWE-191;CWE-125;*CWE-20;Other CVSS 'AV:A/AC:H/PR:N/UI:N/S:U/C:H/I:L/A:H';BEST CVSS score: '8.8';DESCR 'A host-controlled inline data offset in the NVMe/RDMA target path can be mapped incorrectly when inline_data_size is larger than PAGE_SIZE. An offset beyond the first page can make PAGE_SIZE minus off underflow, producing a very large scatterlist length and making the block backend read far past the first inline page. This is an out-of-bounds kernel memory read candidate and may also crash the target or corrupt backend data depending on how the oversized scatterlist is consumed. For the CVSS the PR:N is used because the trigger is sent by an NVMe/RDMA host over the RDMA fabric and does not require a local account on the target once the target port is reachable. The issue is adjacent-network reachable rather than general Internet reachable, because NVMe/RDMA is normally deployed on isolated storage or data-center fabrics. Impact is at least denial of service and possible confidentiality or integrity impact from kernel memory being read through an invalid scatterlist. For the paranoid score, choose the highest still-defensible interpretation supported by the bug class and patch context, with preference for manual-review sensitivity over autoclosed false negatives.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Strong Important candidate / Actionable Moderate at minimum (with actual score 8)	YES	OOB NVME IMPROVEONLY LINUS TEST KPANIC MEMORY  DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFKP	NO	NO	checked

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

ActionableScore=8
ActionableScoreLower=7

## 1. ActionableScore

* Conservative score: **7**
* Paranoid score: **8**
* Final recommended bucket: **Strong Important candidate / Actionable Moderate at minimum**::

## 2. Signal breakdown

* **Remote reachable / network-triggerable: +2**
  Trigger comes from an NVMe/RDMA host over the RDMA storage fabric. This is adjacent-network rather than public-Internet exposure, but it is still externally triggerable against the target.

* **Memory corruption, weak/indirect corruption candidate: +1**
  The primitive is not an OOB write or arbitrary write, but a host-controlled offset causes scatterlist length underflow and invalid scatterlist construction.

* **Memory layout invariant restoration: +1**
  The patch restores consistency between `off`, `page_off`, `page_idx`, `len`, `sg_count`, and the starting `inline_sg` entry.

* **Reliable kernel crash / strong DoS: +1**
  A huge `sg->length` around 4 GiB can make the backend read far beyond the intended inline buffer and plausibly crash or fault the target.

* **Confidentiality impact plausible: +1**
  The oversized scatterlist may cause kernel memory beyond the inline page to be read by the block backend. Depending on completion semantics, this can expose unintended kernel memory through storage data paths.

* **Integrity impact plausible: +1**
  The backend may consume unintended memory as write data, potentially corrupting target storage contents beyond the intended inline payload semantics.

* **Important filesystem/storage path: +1**
  The affected code is in the NVMe target RDMA storage path and can affect backend I/O processing.

* **Rare AND difficult-to-reach subsystem/configuration: -1**
  Practical exposure requires NVMe/RDMA target deployment and `inline_data_size > PAGE_SIZE`, so this is not a default general-purpose network path.

* **Paranoid additional weighting: +1**
  The bug is host-controlled, storage-facing, and creates an oversized scatterlist read primitive in kernel I/O code. This supports treating it as a Strong Important candidate for manual review.

Call-site confidence: **medium-high**. The changed helper and release path are visible, and the commit explicitly states that the block backend reads far past the first inline page, but the exact backend readout and user-visible disclosure path is not fully shown in the supplied diff.

## 3. Reachability analysis

An NVMe/RDMA host that can reach the target port can influence the inline data offset. No local account on the target is required once the attacker has access to the RDMA storage fabric, so the practical privilege model is closer to remote-adjacent `PR:N` than local `PR:L`.

This is not normally exposed to the public Internet. NVMe/RDMA targets are typically deployed in storage or data-center networks with restricted fabric access. Namespaces and containers on the target do not meaningfully reduce privileges needed for the network trigger, but fabric-level authentication, zoning, and target ACLs may limit who can send the malformed request.

The path is configuration-dependent because `inline_data_size` must be larger than `PAGE_SIZE`. That reduces broad exposure but does not make the bug low priority because the affected path is a privileged storage I/O path and the malformed scatterlist can drive backend reads outside the intended buffer.

## 4. Severity interpretation

This is stronger than an ordinary Moderate. The realistic evidence supports at least **Actionable Moderate** because the bug is externally triggerable over an RDMA storage fabric and can produce a large out-of-bounds kernel read through a scatterlist length underflow.

The theoretical worst case is **Strong Important candidate** because the oversized scatterlist may expose kernel memory or corrupt storage data through backend I/O processing. However, arbitrary code execution or a direct local privilege escalation primitive is not demonstrated by the patch, so the score should not be treated as Critical without additional evidence.

## 5. One-sentence report phrase

A host-controlled inline data offset in `nvmet_rdma_use_inline_sg()` can underflow scatterlist length calculation when `inline_data_size > PAGE_SIZE`, causing the NVMe/RDMA target backend to read far beyond the intended inline buffer and potentially leading to kernel memory exposure, storage corruption, or target DoS.

## 6. Manual review recommendation

**MANUAL CHECK REQUIRED**

Manual review is required because this is a network-adjacent storage target bug with host-controlled scatterlist construction and an explicit out-of-bounds backend read. The likely minimum impact is DoS, but confidentiality and integrity impact are plausible enough that it should not be auto-closed as an ordinary Moderate.

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

Patch: nvmet-rdma: handle inline data with a nonzero offset
Commit: c2106ba1b14d644a5203bea1a50dbe25dcad713c
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=c2106ba1b14d644a5203bea1a50dbe25dcad713c

Commit description:

nvmet_rdma_use_inline_sg() maps the host-controlled inline data offset
into the per-command inline scatterlist.  The bounds check admits any
offset with off + len <= inline_data_size, but the mapping still assumes
the data begins in the first inline page:

	sg->offset = off;
	sg->length = min_t(int, len, PAGE_SIZE - off);

When a port is configured with inline_data_size > PAGE_SIZE (settable up
to max(SZ_16K, PAGE_SIZE)), an offset in (PAGE_SIZE, inline_data_size]
makes "PAGE_SIZE - off" underflow, so sg->length is set to ~4 GiB and
the block backend reads far past the first inline page.  num_pages(len)
also ignores the offset, so an in-bounds offset whose [off, off+len)
span crosses a page boundary under-counts the scatterlist.

Map the offset properly: split it into a page index and an in-page
offset, start the scatterlist at that page, and size the page count from
page_off + len.  Because the request scatterlist may now start at
inline_sg[page_idx] rather than inline_sg[0], generalize the inline-SGL
identity test in nvmet_rdma_release_rsp() to a range test; otherwise the
persistent inline scatterlist is mistaken for an allocated one and
nvmet_req_free_sgls() frees an inline page (and warns in
free_large_kmalloc()).

Fixes: 0d5ee2b ("nvmet-rdma: support max(16KB, PAGE_SIZE) inline data")
Cc: stable@vger.kernel.org
Suggested-by: Keith Busch <kbusch@kernel.org>
Reported-by: Bryam Vargas <hexlabsecurity@proton.me>
Signed-off-by: Bryam Vargas <hexlabsecurity@proton.me>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Changed files:
  drivers/nvme/target/rdma.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=c2106ba1b14d644a5203bea1a50dbe25dcad713c

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

The original Linux kernel CVE announcement for CVE-2026-72129 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-72129

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:H/PR:N/UI:N/S:U/C:H/I:L/A:H
  The Best / paranoid CVSS vector: AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
  The Best / paranoid CVSS score: 8.8

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: 7
  Paranoid ActionableScore: 8

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).