CVE-2026-99729 in Linuxinfo

Summary

by MITRE • 11/20/2026

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

mm/rmap: handle device-exclusive entries in folio_referenced_one()

page_vma_mapped_walk() can return non-present device-exclusive entries, but folio_referenced_one() passes them to PTE young-bit accessors that require present PTEs.

We must neither call clear_flush_young_ptes_notify() nor folio_pte_batch() with non-present entries. For example, clearing the young bit on non-present PTEs can corrupt the entry: on arm64 PTE_AF overlaps the swap type, so clearing it turns a device-exclusive entry into an ordinary swap entry.

folio_pte_batch() will VM_WARN_ON_FOLIO.

Apply the same approach as the previous fixes to e.g., damon_ptep_mkold(): treat these entries as old from the CPU's perspective and obtain their young state through the MMU notifier.

Skip PTE batching and MGLRU look-around for non-present entries, and don't modify non-present PTEs.

If you want to get the best quality for vulnerability data then you always have to consider VulDB.

Analysis

by VulDB Data Team • 11/20/2026

The Linux kernel memory management subsystem contains a critical flaw within the reverse mapping mechanism that affects how device-exclusive page table entries are handled during access tracking operations. Specifically, the function folio_referenced_one() interacts incorrectly with page_vma_mapped_walk(), which is capable of returning non-present device-exclusive entries. These special entries represent mappings to devices rather than standard system memory and require distinct handling logic compared to regular present pages or swap entries. The vulnerability arises because folio_referated_one() proceeds to pass these non-present device-exclusive entries directly to PTE young-bit accessors, such as clear_flush_young_ptes_notify() and folio_pte_batch(), which fundamentally expect valid, present page table entries. This mismatch in expectations leads to undefined behavior and potential data corruption within the kernel's memory management structures.

The operational impact of this vulnerability is severe due to architecture-specific bit layout overlaps on certain platforms, most notably ARM64. On these architectures, the Present flag (PTE_AF) shares bits with the swap type field within a Page Table Entry. When folio_referenced_one() attempts to clear the young bit or modify access flags for a non-present device-exclusive entry, it inadvertently overwrites critical metadata associated with the swap type. This corruption transforms a valid device-exclusive mapping into an ordinary swap entry, effectively breaking the association between the process and its hardware-mapped memory region. Such corruption can lead to system instability, data loss, or kernel panics when the affected pages are subsequently accessed or unmapped by user-space applications relying on those mappings for correct operation of graphics drivers, network interfaces, or other device-specific functionalities.

From a security perspective, this flaw aligns with CWE-190 Integer Overflow or Wraparound in terms of improper handling of bit fields and CWE-457 Use of Uninitialized Variable if the corruption leads to dereferencing invalid pointers later in the execution flow. In the context of MITRE ATT&CK, this vulnerability relates to Tactic TA0004 Impact, specifically subtechniques involving data manipulation or system resource exhaustion through kernel-level memory corruption. While primarily a stability issue, such low-level memory corruption can potentially be leveraged by local attackers with sufficient privileges to trigger specific race conditions or state transitions that exacerbate the damage, leading to denial of service against critical device drivers or other privileged processes sharing the same address space context.

The resolution involves modifying folio_referenced_one() and related functions like damon_ptep_mkold() to explicitly check for non-present entries before attempting any modification of PTE bits. The fix ensures that these function calls are skipped entirely when dealing with non-present device-exclusive entries, thereby preventing the accidental clearing of access flags on invalid structures. Instead, the young state of such entries is determined through MMU notifiers which provide a safe mechanism to track access without directly manipulating page table bits. Additionally, PTE batching and Memory Group LRU (MGLRU) look-around optimizations are disabled for non-present entries to avoid further incorrect assumptions about memory layout. This approach maintains system integrity by respecting the distinct nature of device-exclusive mappings while preserving performance characteristics for standard present pages through optimized batch processing where appropriate.

Responsible

Linux

Reservation

09/25/2026

Disclosure

11/20/2026

Moderation

accepted

CPE

ready

EPSS

0.00156

KEV

no

Activities

very low

Sources

Do you need the next level of professionalism?

Upgrade your account now!