[CVE-2026-97945][MODERATE REGULAR] x86/mm: Fix user-space data loss with MADV_FREE and THP

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

CVE: CVE-2026-97945
Priority: MODERATE REGULAR
AL-KERNEL base severity: MODERATE
KPANIC flag: NO
Patch: x86/mm: Fix user-space data loss with MADV_FREE and THP
Commit: e4d569fe2f436a9b58c048fefabf56eab68c1b7e
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=e4d569fe2f436a9b58c048fefabf56eab68c1b7e
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97945
Analysis date: Sat, 26 Sep 2026 15:36:56 -0400
ActionableScore: 4
ActionableScore lower bound: 3
Actionable bucket: Borderline / manual review recommended
Manual review required: YES

Summary:
No one-sentence report phrase was found.

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

The original Linux kernel CVE announcement for CVE-2026-97945 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97945

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: YES

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

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

CVE-2026-97945	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-97945 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:N/I:H/A:H';CWE-682;CWE-703;Other CVSS 'AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:H/A:L';BEST CVSS score: '7.1';DESCR 'An x86 page table state tracking error causes pmd_modify() to discard the hardware dirty bit from writable PMD-mapped THP pages before it can be transferred into the saved dirty state. With MADV_FREE, a page rewritten after being marked lazyfree can later be reclaimed as clean and replaced with a zero page, causing confirmed application data loss. PMD-mapped file pages are also affected because modified data can lose its dirty state and may never reach storage through writeback. For the CVSS the PR:L value is selected because an unprivileged local process can use madvise() and mprotect() on its own mappings without administrative capabilities. The base AC:H reflects the required combination of PMD-mapped THP, dirty state transitions, and reclaim, NUMA hinting, or writeback activity. The paranoid AC:L reflects that the operation sequence can be reproduced once THP is available and memory pressure can be induced locally. The issue is not directly network reachable. Impact includes high integrity loss and possible service availability loss due to silent corruption or deletion of application data. No confidentiality impact or privilege escalation is supported because the bug loses dirty state and replaces discarded anonymous pages with zero pages rather than exposing unrelated memory.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Borderline / manual review recommended (with actual score 4)  SKIP CVE-2026-97945 UNKNOWN SKIP No affected files built, so skip this CVE NO  - - unknown 	MAYBE	HARDWARE MEMORY  DECREASED_TO_MODERATEREG_BASED_ON_FALSEPOSCHECKOFKP	-	-	checked

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

ActionableScore=4
ActionableScoreLower=3
PrimitiveClass=CLASS_C
RealUAF=NO
HostAdminRequired=NO
NamespaceScopedPrivilege=NO
ImportantCandidate=NO
ManualReview=RECOMMENDED
AutoDowngradeAllowed=YES

1. ActionableScore

* Raw conservative score: 3
* Conservative score: 3
* Raw paranoid score: 4
* Paranoid score: 4
* No floor or cap applied.
* Final recommended bucket: **Borderline / manual review recommended**::

2. Signal breakdown

Conservative signals:

* +1 Local unprivileged trigger. An ordinary process can exercise the affected state transitions through MADV_FREE, mprotect(), THP-backed mappings, and normal memory access.
* +1 Important filesystem/storage path. The defect affects VM dirty tracking, reclaim, page cache behavior, and writeback of PMD-mapped file THPs.
* +1 Persistent availability consequence. Anonymous data can be permanently replaced with zero pages, and file-backed modifications can be omitted from writeback. This is lasting data loss rather than transient resource pressure.

Paranoid-only signal:

* +1 Broad and realistically exercised exposure. x86, THP, mprotect(), NUMA hinting, and memory reclaim are broadly deployed, and the supplied report confirms production data loss. The exact combination is workload-dependent, so this is withheld from the conservative score.

Not awarded:

* No UAF, stale post-lifetime reference, OOB access, arbitrary write, controlled reclaim, or repeated release/refcount operation is established.
* No confidentiality point. Discarded anonymous pages return as fresh zero pages rather than exposing unrelated memory.
* No security integrity point. The evidence establishes loss of data belonging to affected mappings, but not unauthorized modification of another principal's protected data.
* No Dirty-Pipe-like ownership or COW violation. A write authorized for one object is not shown to modify independently permissioned storage.
* No strong semantic reinterpretation. The bug loses dirty-state metadata but does not reinterpret an active mapping as a different translation or ownership domain.
* No reliable kernel-crash point. The demonstrated result is silent data loss, not panic or Oops.
* No remote-reachability, privilege-escalation, authorization-bypass, or practical security exploit-chain point.

3. Reachability analysis

An ordinary local process can invoke madvise() and mprotect() on its own mappings without administrative capabilities. It can also create memory pressure within the resources available to it, although the final reclaim or NUMA interaction remains workload and system dependent.

No namespace capability is required, and namespace privilege is not part of the trigger. No prior privileged object provisioning, special device, firmware, or rare physical hardware is required. PMD-mapped THP must be present, and the data-loss sequence additionally requires reclaim, NUMA hinting, or file writeback behavior.

The trigger is not a race in the narrow timing sense. It is a deterministic dirty-bit state transition once the required mapping and VM operations occur. The issue is local and is not directly reachable through network traffic.

4. Primitive and severity interpretation

The exact primitive is loss of the hardware dirty bit when pmd_modify() transforms a writable, dirty PMD. Consequently, reclaim or writeback can treat modified data as clean.

This is neither active structured UAF nor fixed stale storage. It is a live page-table bookkeeping and state-transition defect. There is no first lifetime-ending operation, surviving dangling reference, or second refcount mutation.

The demonstrated consequences are silent anonymous-memory data loss and omission of modified file-backed THP data from writeback. Possible service disruption from corrupted or missing application data is plausible and operationally significant.

No evidence establishes unauthorized cross-principal corruption, kernel memory corruption, privilege escalation, disclosure, or a security-boundary bypass. The issue is therefore CLASS_C rather than CLASS_A or CLASS_B.

ImportantCandidate=NO because no Important-preserving security primitive is established. AutoDowngradeAllowed=YES because no unresolved UAF, ownership violation, controlled corruption, or authorization primitive would be lost by treating this as a Moderate-class security issue. Manual review remains recommended because the operational data-loss consequence is confirmed and potentially severe.

5. One-sentence report phrase

An x86 PMD dirty-state tracking error can cause rewritten MADV_FREE anonymous THPs to be discarded and modified file-backed THPs to miss writeback, producing silent local data loss without an established privilege escalation or cross-user corruption primitive.

6. Manual review recommendation

MANUAL CHECK RECOMMENDED

Confirmed production data loss warrants affected-version and workload review, but the supplied evidence does not establish an Important-grade kernel security primitive.

7. Call-site confidence

Call-site confidence: high

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

Patch: x86/mm: Fix user-space data loss with MADV_FREE and THP
Commit: e4d569fe2f436a9b58c048fefabf56eab68c1b7e
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=e4d569fe2f436a9b58c048fefabf56eab68c1b7e

Commit description:

Some of users of Polars (a data analytics library) have lost production
data from this bug. They seem to have just the right combination of
huge pages, MADV_FREE and heavy reclaim pressure.

pmd_modify() masks the old value with (_HPAGE_CHG_MASK & ~_PAGE_DIRTY),
silently discarding the hardware dirty bit.  The subsequent
pmd_mksaveddirty() call is supposed to transfer _PAGE_DIRTY into
_PAGE_SAVED_DIRTY when write-protecting, but the dirty bit was already
stripped from the value, so there is nothing left to transfer.

Contrast with pte_modify(), which keeps _PAGE_DIRTY_BITS in its mask,
and pud_modify(), which keeps _HPAGE_CHG_MASK untouched: pmd_modify()
is the odd one out.  Any pmd_modify() on a writable, dirty PMD loses
the dirty state.

One visible consequence is data loss with MADV_FREE on PMD-mapped THP:

    memset(buf, 0x5A, size);          // PMD-mapped THP, PMD dirty
    madvise(buf, size, MADV_FREE);    // PMD cleaned but left writable,
                                      // folio marked lazyfree
    memset(buf, 0x5A, size);          // hardware sets _PAGE_DIRTY again
    mprotect(buf, size, PROT_READ);   // pmd_modify() drops the dirty bit
    mprotect(buf, size, PROT_READ|PROT_WRITE);
    // ... memory pressure ...

Reclaim (e.g. under memcg pressure) then finds the lazyfree folio with
no dirty bit set anywhere and frees it in
__discard_anon_folio_pmd_locked(), even though the data was rewritten
after MADV_FREE; subsequent reads fault in fresh zero pages.  NUMA
hinting alone can trigger the same loss, as do_huge_pmd_numa_page()
restores the PMD through pmd_modify() as well.

PMD-mapped file THPs are affected too: mprotect()/NUMA hinting dropping
the dirty bit means rewritten data is never written back.

Fix it by keeping _PAGE_DIRTY in the preserved mask, exactly like
pte_modify() and pud_modify() do.  The existing
pmd_mksaveddirty()/pmd_clear_saveddirty() pair then performs the
hardware-dirty <-> saved-dirty transition based on the write bit,
preserving the shadow-stack encoding rules.

Fixes: bb3aadf ("x86/mm: Start actually marking _PAGE_SAVED_DIRTY")
Closes: https://lore.kernel.org/r/CAJxLxMUGu1-L+O_nAONOwOXnS=cNbNApCWqdthRjd76LThtSPg@mail.gmail.com/
Reported-by: Orson Peters <orsonpeters@gmail.com>
Signed-off-by: Vernon Yang <yanglincheng@kylinos.cn>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20260903031608.1194238-1-vernon2gm@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Changed files:
  arch/x86/include/asm/pgtable.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=e4d569fe2f436a9b58c048fefabf56eab68c1b7e

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

The original Linux kernel CVE announcement for CVE-2026-97945 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97945

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:H/A:L
  The Best / paranoid CVSS vector: AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:H/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: 3
  Paranoid ActionableScore: 4

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).