CVE-2026-99727: mm/huge_memory: properly handle refcounted PUD mappings on fork
In the Linux kernel, the following vulnerability has been resolved:
mm/hugememory: properly handle refcounted PUD mappings on fork
copyhugepud() copies device-DAX entries without taking a folio reference or updating mapcounts and the child's RSS.
For example, MADVGUARDINSTALL sets VMMAYBEGUARD on a device-DAX VMA, forcing copypagerange() to traverse the whole VMA on fork. Any remaining huge PUDs are then copied without accounting for the new mappings. Unmapping them in the child decrements accounting that was never added.
Use vmnormalpagepud() to identify refcounted mappings. Take a folio reference and add the child's PUD rmap and RSS accounting before installing the copied entry. Pass src and dst VMAs consistently instead of src and dst MM. Move the TODO.
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 - Upgrade
Upgrade
debian/linuxto a version that resolves this vulnerability.Fixed in 6.1.176-1Fixed in 6.1.192-2Fixed in 6.12.107-1Fixed in 6.12.117-1Fixed in 7.3.5-1
Event History
Frequently Asked Questions
Which workloads are exposed to the affected path?
The affected path involves processes that fork while using device-DAX VMAs containing huge PUD mappings. The description specifically notes that MADV_GUARD_INSTALL can set VM_MAYBE_GUARD on such a VMA, causing fork processing to traverse the VMA and copy remaining huge PUD entries.
What happens after an affected mapping is copied into a child process?
The copied device-DAX huge PUD entry may be installed without taking a folio reference or updating the child mapping counts and RSS. If the child later unmaps it, accounting is decremented even though it was never added.