{"id":"CVE-2026-98468","title":"KVM: arm64: Fix spurious warning for benign stage 2 teardown race","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\nKVM: arm64: Fix spurious warning for benign stage 2 teardown race\n\nkvmtool was used to establish an L1 guest with 8 CPUs and 8 GiB of RAM, an\nL2 guest with 4 CPUs and 4…","severity":"none","vendor":"Linux","product":"Linux","affected":["Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < f44a92633136b409ebb9b7036fceb7e11de60a20","Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < c23ed110d27ba99cf3a881986514f79aacedf476","Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < 330346dda3bf5f8a3b06723024f7d906326270cf","Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < 38b70fc453c3112f1a62583b89903ae41116cc27","Linux 6.11"],"published":"2026-10-18","updated":"2026-10-18","sourceUpdated":"2026-10-18T15:15:20.882Z","source":"CVEORG","sourceUrl":"https://www.cve.org/CVERecord?id=CVE-2026-98468","references":[{"url":"https://git.kernel.org/stable/c/f44a92633136b409ebb9b7036fceb7e11de60a20"},{"url":"https://git.kernel.org/stable/c/c23ed110d27ba99cf3a881986514f79aacedf476"},{"url":"https://git.kernel.org/stable/c/330346dda3bf5f8a3b06723024f7d906326270cf"},{"url":"https://git.kernel.org/stable/c/38b70fc453c3112f1a62583b89903ae41116cc27"}],"tags":["cve.org"],"ingestedAt":"2026-10-18T15:45:12.298Z","slug":"CVE-2026-98468","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\nKVM: arm64: Fix spurious warning for benign stage 2 teardown race\n\nkvmtool was used to establish an L1 guest with 8 CPUs and 8 GiB of RAM, an\nL2 guest with 4 CPUs and 4 GiB of RAM and an L3 guest with 2 CPUs and 2 GiB\nof RAM, all of which was then exited.\n\nUnder memory pressure in the L0 host warnings were observed due to\nmigration triggered by compaction:\n\nWARNING: arch/arm64/kvm/mmu.c:336 at __unmap_stage2_range+0x64/0x80,\nCPU#5: kcompactd0/66\n\nWhich was, in turn, triggered by an MMU notifier for the host invalidation:\n\nmmu_notifier_invalidate_range_start()\n  -> ... -> kvm_mmu_notifier_invalidate_range_start()\n    -> kvm_mmu_unmap_gfn_range()\n      -> kvm_unmap_gfn_range()\n        -> kvm_nested_s2_unmap()\n          -> kvm_stage2_unmap_range()\n            -> __unmap_stage2_range()\n               -> stage2_apply_range()\n       \t       <- -EINVAL, triggering a WARN_ON()\n\nRacing with L0's teardown of stage 2 page tables:\n\nexit_mm()\n  -> mmput()\n    -> __mmput()\n      -> exit_mmap()\n        -> mmu_notifier_release()\n\t  -> ... -> kvm_mmu_notifier_release()\n            -> kvm_flush_shadow_all()\n              -> kvm_arch_flush_shadow_all()\n\t        -> kvm_free_stage2_pgd()\n\t\t  -> [ acquire kvm->mmu_lock for write ]\n\t\t  -> mmu->pgt = NULL [ among other tasks ]\n\t\t  -> [ release kvm->mmu_lock for write ]\n\nIt turns out there is a benign race resulting in a spurious warning:\n\n\tThread A - notify: migration   | Thread B - notify: release\n\t-------------------------------|---------------------------------\n\t< kvm->mmu_lock held >         |\n\tstage2_apply_range()           |\n\t  get mmu->pgt, check !NULL    |\n\t  ...                          | kvm_arch_flush_shadow_all()\n\t  cond_resched_rwlock_write(); |   < contend, sleep kvm->mmu_lock >\n\t< drop kvm->mmu_lock >         |   < acquire kvm->mmu_lock>\n                                       |   ...\n\t\t\t\t       |   kvm_free_stage2_pgd()\n                                       |     mmu->pgt = NULL\n\t\t\t\t       |   < invalidate MMU >\n\t\t\t\t       |   ...\n\t\t\t\t       |   < release kvm->mmu_lock >\n\t[ scheduled ]\t\t       |\n\tstage2_apply_range()           |\n\t  < loop to next >             |\n\t  get, mmu->pgt, check !NULL   |\n\t  is NULL, return -EINVAL      |\n        __unmap_stage2_range()         |\n\t  WARN_ON(-EINVAL) <--- entirely spurious - the race was handled\n                                 correctly.\n\nFix the spurious warning by updating stage2_apply_range() to no longer\ntreat concurrent PGT teardown on lock release as an error - whether the\nwalker is tearing down page tables or doing something else this is a\nlegitimate reason to abort the operation without error.\n\nThis keeps the warning in place for all other circumstances.\n\nIn practice only __unmap_stage2_range() actually does anything with the\nerror so this only impacts that.\n\n## Affected\n\n- `Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < f44a92633136b409ebb9b7036fceb7e11de60a20`\n- `Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < c23ed110d27ba99cf3a881986514f79aacedf476`\n- `Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < 330346dda3bf5f8a3b06723024f7d906326270cf`\n- `Linux >= ec14c272408af43d392f65f55e66f3b94fc61921 < 38b70fc453c3112f1a62583b89903ae41116cc27`\n- `Linux 6.11`\n\n## Remediation\n\nRefer to the linked advisories for vendor-supplied fixes and affected version ranges.","depth":"sunlit","depthScore":3,"depthScoreParts":{"impact":2.8,"likelihood":0,"exploitation":0,"ransomware":0},"changes":[]}