Skip to content
Fencer Intel
no cvss —

CVE-2026-97945 json

x86/mm: Fix user-space data loss with MADV_FREE and THP

published last mo. modified 24d ago also EUVD-2026-86799

cvss
—
not scored
cisa kev
Not listed
not in CISA's catalog
epss
0.2%
EPSS 0.2% chance of exploitation in 30 days, percentile 10%
public exploits
0
none known

Summary

In the Linux kernel, the following vulnerability has been resolved:

x86/mm: Fix user-space data loss with MADV_FREE and THP

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.

Patch / fix

  • Linux Linux 6.6.158, 6.12.111, 6.18.53, 7.2.7, 7.3
  • linux Debian 13 6.12.111-1
  • linux Debian 14 7.2.7-1
  • linux-6.12 Debian 12 6.12.111-1~deb12u1

debian advisories DEBIAN-CVE-2026-97945

  • linux Ubuntu 26.04 LTS 7.0.0-42.42
  • linux-aws Ubuntu 26.04 LTS 7.0.0-1016.17
  • linux-gcp Ubuntu 26.04 LTS 7.0.0-1016.18
  • linux-ibm Ubuntu 26.04 LTS 7.0.0-1017.17
  • linux-riscv Ubuntu 26.04 LTS 7.0.0-42.42.1
  • linux-oracle Ubuntu 26.04 LTS 7.0.0-1015.16
  • linux-hwe-7.0 Ubuntu 24.04 LTS 7.0.0-42.42~24.04.1
  • linux-oem-7.0 Ubuntu 26.04 LTS 7.0.0-1017.17

ubuntu advisories UBUNTU-CVE-2026-97945

Ubuntu security data © Canonical Ltd., from https://github.com/canonical/ubuntu-security-notices via OSV; changed: normalised to fixed package versions per release. CC BY-SA 4.0

  • kernel-rt Real Time Module 15 SP7 6.4.0-150700.7.81.1
  • kernel-syms-rt Real Time Module 15 SP7 6.4.0-150700.7.81.1
  • kernel-source-rt Real Time Module 15 SP7 6.4.0-150700.7.81.1
  • kernel-64kb SLE Module for Basesystem 15 SP7 6.4.0-150700.53.87.3
  • kernel-syms SLE Module for Development Tools 15 SP7 6.4.0-150700.53.87.1
  • kernel-azure SLE Module for Public Cloud 15 SP7 6.4.0-150700.53.87.3
  • kernel-source SLE Module for Development Tools 15 SP7, SLE Module for Basesystem 15 SP7 6.4.0-150700.53.87.3
  • kernel-default SLE Workstation Extension 15 SP7, SLE Module for Legacy 15 SP7, SLE Module for Basesystem 15 SP7 +2 more 6.4.0-150700.53.87.3

suse advisories SUSE-SU-2026:4940-1SUSE-SU-2026:4935-1SUSE-SU-2026:4934-1SUSE-SU-2026:24473-1SUSE-SU-2026:24553-1SUSE-SU-2026:24422-1

  • kernel-rt openSUSE Leap 16.0 6.12.0-160000.40.1
  • dtb-aarch64 openSUSE Leap 16.0 6.12.0-160000.40.1
  • kernel-64kb openSUSE Leap 16.0 6.12.0-160000.40.1
  • kernel-syms openSUSE Leap 16.0 6.12.0-160000.40.1
  • kernel-azure openSUSE Leap 16.0 6.12.0-160000.40.1
  • kernel-obs-qa openSUSE Leap 16.0 6.12.0-160000.40.1
  • kernel-source openSUSE Leap 16.0 6.12.0-160000.40.1
  • kernel-default openSUSE Leap 16.0 6.12.0-160000.40.1

opensuse advisories openSUSE-SU-2026:22167-1openSUSE-SU-2026:12074-1

  • kernel Azure Linux 3.0 6.6.158.1-1

azure linux advisories AZL-104843

Azure Linux vulnerability data, Copyright (c) Microsoft Corporation, from https://github.com/microsoft/AzureLinuxVulnerabilityData via OSV, under the MIT licence; changed: normalised to fixed package versions per release. MIT

Is CVE-2026-97945 in your stack? Fencer shows which of your repositories are affected as soon as an advisory lands, and opens the fix as a pull request.

Check with Fencer

Vendor advisories

What the vendors themselves say about their products: affected, fixed or not affected, and what to do.

red hat product security vex CVE-2026-97945 updated cvss 5.5

kernel: x86/mm: Fix user-space data loss with MADV_FREE and THP

  • Red Hat Enterprise Linux 6 perf, kernel, python-perf, kernel-kdump, kernel-firmware, kernel-bootwrapper, +1 more
    affected: workaround, no fix planned “Out of support scope”
  • Red Hat Enterprise Linux 9 rv, perf, rtla, kernel, bpftool, libperf, +82 more Red Hat Enterprise Linux 10 kernel Red Hat OpenShift Container Platform 4 openshift/ose-rhel-coreos-9 Red Hat OpenShift Container Platform 5 openshift/ose-rhel-coreos-9
    affected: workaround, no fix available “Fix deferred”
  • Red Hat Enterprise Linux 7 perf, kernel, bpftool, kernel-rt, python-perf, kernel-kdump, +8 more Red Hat Enterprise Linux 8 perf, kernel, bpftool, kernel-rt, kernel-core, kernel-tools, +29 more Red Hat Enterprise Linux 9 kernel-rt, kernel-rt-selftests-internal Red Hat OpenShift Container Platform 4 openshift/ose-rhel-coreos-8
    not affected: vulnerable code not present

Red Hat CSAF VEX © Red Hat, Inc., from https://security.access.redhat.com/data/csaf/v2/vex/; changed: reduced to product statuses and remediations. CC BY 4.0 · csaf

Exploit prediction

0.21% chance of exploitation in the next 30 days, higher than 10% of all scored CVEs (FIRST EPSS, 2026-11-11).

Affected products

  • Linux Linux

Classification

Technology feeds

References

Sources

Fencer Intel keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.