CVE-2026-99729: mm/rmap: handle device-exclusive entries in folio_referenced_one()
In the Linux kernel, the following vulnerability has been resolved:
mm/rmap: handle device-exclusive entries in folioreferencedone()
pagevmamappedwalk() can return non-present device-exclusive entries, but folioreferencedone() passes them to PTE young-bit accessors that require present PTEs.
We must neither call clearflushyoungptesnotify() nor folioptebatch() with non-present entries. For example, clearing the young bit on non-present PTEs can corrupt the entry: on arm64 PTEAF overlaps the swap type, so clearing it turns a device-exclusive entry into an ordinary swap entry.
folioptebatch() will VMWARNONFOLIO.
Apply the same approach as the previous fixes to e.g., damonptepmkold(): 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.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 7.3.5
Frequently Asked Questions
What systems are exposed to this issue?
Systems running Linux kernel code that encounters non-present device-exclusive entries during folio reference tracking are exposed. The issue is specifically relevant where page_vma_mapped_walk() returns such entries.
What could happen if the vulnerable code processes a device-exclusive entry?
Clearing the young bit on a non-present entry can corrupt it. On arm64, PTE_AF overlaps the swap type, so clearing it can convert a device-exclusive entry into an ordinary swap entry; the batching path can also trigger VM_WARN_ON_FOLIO.
Is attacker interaction or a particular configuration required?
The provided information does not describe an attacker-controlled trigger or any configuration prerequisite. It identifies the necessary runtime condition as processing non-present device-exclusive page-table entries.
What is the mitigation when a patched kernel cannot be deployed immediately?
No operational workaround is provided. The fix avoids modifying non-present PTEs, treats them as old from the CPU perspective, obtains young state through the MMU notifier, and skips PTE batching and MGLRU look-around for those entries.