{"id":"CVE-2026-89908","published":"2026-09-16T11:16:59.807","lastModified":"2026-09-16T15:18:17.050","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nLoongArch: KVM: Preserve memslot arch flags on KVM_MR_FLAGS_ONLY\n\nkvm_arch_prepare_memory_region() computes new->arch.flags, i.e. whether\na memslot is KVM_MEM_HUGEPAGE_CAPABLE or KVM_MEM_HUGEPAGE_INCAPABLE,\nonly for KVM_MR_CREATE and KVM_MR_MOVE, and returns early for every\nother change. But the generic code allocates a zeroed memslot for every\nchange and never copies old->arch, so after a KVM_MR_FLAGS_ONLY update,\ne.g. toggling KVM_MEM_LOG_DIRTY_PAGES for live migration, the active\nmemslot has arch.flags == 0.\n\nWith both flags clear, fault_supports_huge_mapping() falls through to\nthe alignment check on the HVA range alone, which no longer verifies\nthat the GPA and HVA have the same offset within a PMD. A memslot that\nwas marked KVM_MEM_HUGEPAGE_INCAPABLE because of a GPA/HVA offset\nmismatch can then be mapped with PMD entries on read faults, and since\nkvm_map_page() aligns the gfn and the pfn independently, the guest ends\nup accessing the wrong host pages, exactly the \"d -> f, e -> g\" case\ndescribed in the comment above the check.\n\nCarry the arch flags over from the old memslot for KVM_MR_FLAGS_ONLY,\nas the GPA, HVA and size are guaranteed to be unchanged for that case.","cvssScore":8.8,"cvssSeverity":"HIGH","cvssVector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H","cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/27a9bfee3bbcb3cabb77797354f07e0e44e49831","tags":[]},{"url":"https://git.kernel.org/stable/c/4e4dbc341b1581dc512b85d98b768373b0398366","tags":[]},{"url":"https://git.kernel.org/stable/c/7c6df65b53846cd7a9a1c81c6ddb42fdc08603e8","tags":[]},{"url":"https://git.kernel.org/stable/c/bc7a6849b4395f45a602db91b0fb2fb2e9ed4bba","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}