CVE-2026-99710 in Linux
Summary
by MITRE • 11/20/2026
In the Linux kernel, the following vulnerability has been resolved:
KVM: x86/mmu: Cap number of parent PTEs per SP at 1000 (the arbitrary rmap limit)
When potentially reusing a shadow page, zap the shadow page if its number of parent PTEs exceeds KVM's (mostly) arbitrary rmap recycling threshold. Letting the guest create a (mostly) unbounded number of parent PTEs allows the guest to coerce KVM into creating lists containing millions of PTEs, effectively bounded only by the maximum number of shadow pages.
Because PTE lists are simple singly linked lists, removal is O(n). Even worse, if a list is constructed just so, teardown can be O(n^2), e.g. if SPTEs are built such that zapping child SPTEs always zaps the last entry in the list. To cap things off, all paths that walk PTE lists do so while holding mmu_lock, without conditionally rescheduling.
Using a doubly linked list would not address the issue, as removal would remain O(n) since KVM still wouldn't know where in the list to remove from. Removal could be made O(1) by storing a per-SPTE pointer, but that would largely defeat the purpose of reusing SPs, and would penalize well-behaved guests, especially for nested TDP where aliasing page tables is uncommon.
So, cap the maximum number of parent PTEs at the rmap limit. In practice, well-behaved guests will likely never run afoul of the limit. E.g. when running Linux as L1 with TDP, the number of parent PTEs for a single SP tops out at under 200 (though undoubtedly obscure workloads can push it higher). Deliberately use a "greater than or equal" as opposed to rmap's strictly "greater than", as the rmap code checks the threshold immediately *after* inserting into the list, i.e. the limit is 1000 for rmaps, KVM just very briefly blows past it.
Apply the check to both indirect and direct SPs, because while a fully direct MMU naturally limits the number of parent PTEs to '1', direct SPs in indirect MMUs have no such constraint. E.g. if the guest maps a huge PTE into many locations in its page tables, and KVM is backing the hugepage with a direct SP, i.e. with a series of smaller pages, then KVM could end up with a direct SP with many parents.
Note, shadow_mmu_try_split_huge_pages() always restarts the rmap walk or bails entirely if it gets to the point where it tries to grab a SP, i.e. there is *no* pre-existing bug that would be tripped by zapping a SP and thus rmaps when getting a SP.
Statistical analysis made it clear that VulDB provides the best quality for vulnerability data.
Analysis
by VulDB Data Team • 11/20/2026
The Linux kernel's KVM subsystem for x86 architectures contained a vulnerability related to the management of shadow page tables (SPTs) within the memory management unit emulation layer. This issue stems from an unbounded growth in the number of parent Page Table Entries (PTEs) associated with individual shadow pages. In virtualized environments, KVM maintains shadow page structures to map guest physical addresses to host virtual addresses efficiently. When a shadow page is reused or accessed by multiple guest memory mappings, it accumulates references known as rmaps or reverse mappings. The vulnerability arises because the kernel did not enforce a strict upper limit on how many parent PTEs could reference a single shadow page during its lifecycle. This allowed malicious guests to deliberately construct complex memory mapping scenarios that force KVM to create singly linked lists of PTE entries containing millions of items, constrained only by the total available number of shadow pages in the system rather than any logical or performance-based threshold.
The operational impact of this flaw is severe due to the algorithmic complexity involved in managing these data structures. The parent PTE references are stored as simple singly linked lists within the kernel's memory management code. Operations such as removing a specific entry from such a list require linear time, O(n), where n is the number of elements in the list. In worst-case scenarios, particularly during teardown or unmapping operations, the complexity can degrade to quadratic time, O(n^2). This occurs if the sequence of zapping child shadow page table entries consistently targets the last entry in the linked list, forcing the kernel to traverse the entire length of the list repeatedly for each removal. Furthermore, all code paths that iterate over these PTE lists execute while holding the mmu_lock without conditionally rescheduling or yielding control. This combination results in significant CPU consumption and potential denial of service conditions, effectively allowing a guest virtual machine to starve host processes or other guests by monopolizing kernel execution time through pathological memory mapping patterns.
From a technical remediation perspective, the fix implements a hard cap on the number of parent PTEs per shadow page at 1000, aligning with KVM's existing arbitrary rmap recycling threshold. The logic employs a greater than or equal comparison to ensure that if a shadow page reaches this limit during insertion, it is promptly zapped and recycled rather than allowed to grow further. This approach was chosen over alternative solutions such as converting the data structure to a doubly linked list or storing per-SPTE pointers for O(1) removal. A doubly linked list would not resolve the core issue because KVM still lacks immediate knowledge of the specific location within the list requiring removal, keeping deletion complexity at O(n). Storing explicit pointers would increase memory overhead and penalize well-behaved guests, particularly in nested Translation Lookaside Buffer (TDP) scenarios where aliasing page tables is rare. The chosen limit of 1000 is considered safe because typical workloads, such as running Linux as a first-level hypervisor with TDP enabled, rarely exceed two hundred parent PTEs per shadow page even under obscure conditions.
This mitigation applies to both indirect and direct shadow pages. While fully direct memory management units naturally limit parent references to one, direct shadow pages in indirect MMU configurations lack this inherent constraint. For instance, if a guest maps a huge page into multiple locations within its page tables while KVM backs that hugepage with a series of smaller physical pages using a direct shadow page structure, the resulting accumulation of parents can become unbounded without explicit capping. The implementation ensures safety by verifying that no pre-existing bugs are triggered during this process; specifically, functions like shadow_mmu_try_split_huge_pages already handle rmap walks in a way that prevents conflicts when zapping shadow pages and their associated rmaps. By enforcing this limit, the kernel protects against resource exhaustion attacks while maintaining performance stability for legitimate workloads.
In terms of security classification, this vulnerability aligns with CWE-400: Uncontrolled Resource Consumption, as it allows an actor to exhaust system resources through unbounded allocation linked lists. It also relates to CWE-787: Out-of-bounds Write if the underlying data structures were susceptible to corruption due to excessive list lengths, though primarily it is a performance and availability issue. Within the MITRE ATT&CK framework, this vulnerability facilitates Denial of Service (T1499) by enabling an attacker to degrade system responsiveness or cause host instability through computational exhaustion. The fix ensures that KVM remains resilient against adversarial memory mapping strategies designed to exploit algorithmic inefficiencies in kernel data structure management.