{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-99346","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-25T10:25:14.409Z","datePublished":"2026-11-19T14:22:01.304Z","dateUpdated":"2026-11-22T11:07:48.175Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-11-22T11:07:48.175Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nKVM: SVM: Trigger new ASID allocation in both VMCBs on pCPU switch\n\nWhen the pCPU where the VMCB was most recently used is switched, reset\nthe ASID generation in both VMCBs, triggering new ASID allocation for\nthe immediate VMRUN as well as the next VMRUN on the other VMCB.\n\nThe ASID is shared between vmcb01 and vmcb02, and gets flushed on every\nnested transition. However, since pCPU tracking is done per VMCB, it is\npossible for one VMCB to allocate a new ASID when migrated to a new\npCPU, and then the other VMCB reuses that ASID on the old pCPU. This can\nresult in the same ASID being used by multiple vCPUs on the old pCPU.\n\nExample scenario:\n- vCPU runs on pCPU A, vmcb01 is active, asid=1.\n- vCPU migrates to pCPU B, vmcb01 pCPU changes, new asid=2.\n- Another vCPU runs on pCPU A and allocates asid=2 as well.\n- vCPU migrates back to pCPU A, and then switches to vmcb02 before it\n  runs again with vmcb01.\n- No pCPU switch is detected for vmcb02, so VMRUN is done with asid=2.\n- Two vCPUs end up using asid=2 on the same pCPU.\n\nKeep the VMCB dirtying to the active VMCB only. Clean bits are tracked\nby a pCPU for each VMCB, so do not unnecessarily dirty a VMCB if its\npCPU does not change.\n\nAdditionally, initialize the tracked pCPU for vmcb02 to -1 on nested\nenablement as hardening, so that a new ASID allocation is always\ntriggered when nested is disabled and re-enabled.\n\nNo performance regression was noticed when overcommitting L1 vCPUs in L0\n(to force rescheduling), pinning L1 <-> L2 vCPUs, and running CPUID in a\ntight loop bouncing between 2 vCPUs in L2.\n\nAn alternative (and perhaps more proper) fix would be tracking the ASID\nper-VMCB instead (e.g. [1]). However, that's a more involved change, and\nit would result in having different ASIDs for L1 and L2 without actually\nproperly maintaining them. It would probably work because all TLB\nflushes target the current VMCB, and the other VMCB is always flushed on\nnested transitions, but the code ends up in an arguably more fragile\nstate. Punt a proper clean fix to an incoming (and overdue) overhaul of\nSVM's ASID usage [2].\n\n[1] https://lore.kernel.org/lkml/20250205182402.2147495-2-yosry.ahmed@linux.dev\n[2] https://lore.kernel.org/kvm/20260728003557.1136583-1-yosry@kernel.org"}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H","baseScore":8.8,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - pre_svm_run() runs only from svm_vcpu_run() on KVM_RUN (kvm_vcpu_ioctl -> kvm_arch_vcpu_ioctl_run -> vcpu_enter_guest). Nothing in a remote protocol is parsed; the attacker makes current_vmcb vmcb02 through KVM_SET_NESTED_STATE (svm_set_nested_state) or vmrun_interception() -> nested_svm_vmrun().\nAC:L - The attacker sched_setaffinity()s the KVM_RUN thread so kvm_arch_vcpu_load() changes vcpu->cpu, and svm_set_nested_state() selects vmcb02 whose cpu still equals the old pCPU, so pre_svm_run() skips new_asid() and svm_vcpu_run() writes that foreign svm->asid to vmcb->control.asid. Repeated new_asid() calls walk sd->next_asid, including the wrap to min_asid, until the reused value is one another vCPU already holds on the target pCPU.\nPR:L - kvm_dev_ioctl(KVM_CREATE_VM) and kvm_vcpu_ioctl(KVM_RUN, KVM_SET_NESTED_STATE) never call capable(). A process that can open /dev/kvm can svm_set_efer(EFER_SVME) -> svm_allocate_nested() and run the vCPU; host CAP_SYS_ADMIN is not on this path.\nUI:N - The attacker issues KVM_SET_NESTED_STATE or a guest VMRUN and moves their own vCPU thread. The other VM takes no action; its vCPU keeps running with the ASID new_asid() already saved in svm->asid.\nS:C - svm->asid is shared by vmcb01 and vmcb02 and svm_vcpu_run() copies it to vmcb->control.asid, while new_asid() allocates from per-pCPU svm_cpu_data common to every VM. Reusing that ASID on the old pCPU aliases NPT TLB entries with another VM, outside the attacking guest's memslots.\nC:H - Once the stolen ASID is installed, later svm_vcpu_run() entries leave tlb_ctl at TLB_CONTROL_DO_NOTHING, so a GPA load uses the other VM's cached nested translation and returns that guest's page. The hit covers the other vCPU's TLB-resident pages, not a few leaked bytes.\nI:H - Those NPT TLB entries carry the writable bit KVM sets for ordinary RAM memslots, so a store with the stolen ASID writes the other VM's HPA, and an instruction fetch can execute attacker bytes at an overlapping GPA. svm_flush_tlb_asid() flushes only that ASID and does not separate the two guests.\nA:H - Fetches and stores that hit the other VM's HPA corrupt that guest and can triple-fault or panic it. The alias survives later VMRUNs because pre_svm_run() does not allocate a new ASID and svm_flush_tlb_asid() cannot tell the two vCPUs apart."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/x86/kvm/svm/nested.c","arch/x86/kvm/svm/svm.c"],"versions":[{"version":"193015adf40d0465c240d4e9a7b6e4b84b531f8b","lessThan":"5636d57b61068918085360aaad18ab57872e5800","status":"affected","versionType":"git"},{"version":"193015adf40d0465c240d4e9a7b6e4b84b531f8b","lessThan":"058173bce103b6ac405b2bbf61e29a7f3767348f","status":"affected","versionType":"git"},{"version":"193015adf40d0465c240d4e9a7b6e4b84b531f8b","lessThan":"7d787fb34e825218702496fb0d5892529bea9366","status":"affected","versionType":"git"},{"version":"193015adf40d0465c240d4e9a7b6e4b84b531f8b","lessThan":"bf260dc9e7b8bfcd06c1a8e739797b1cf0917c72","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/x86/kvm/svm/nested.c","arch/x86/kvm/svm/svm.c"],"versions":[{"version":"5.13","status":"affected"},{"version":"0","lessThan":"5.13","status":"unaffected","versionType":"semver"},{"version":"6.12.117","lessThanOrEqual":"6.12.*","status":"unaffected","versionType":"semver"},{"version":"6.18.61","lessThanOrEqual":"6.18.*","status":"unaffected","versionType":"semver"},{"version":"7.3.2","lessThanOrEqual":"7.3.*","status":"unaffected","versionType":"semver"},{"version":"7.4-rc1","lessThanOrEqual":"*","status":"unaffected","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.13","versionEndExcluding":"6.12.117"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.13","versionEndExcluding":"6.18.61"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.13","versionEndExcluding":"7.3.2"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.13","versionEndExcluding":"7.4-rc1"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/5636d57b61068918085360aaad18ab57872e5800"},{"url":"https://git.kernel.org/stable/c/058173bce103b6ac405b2bbf61e29a7f3767348f"},{"url":"https://git.kernel.org/stable/c/7d787fb34e825218702496fb0d5892529bea9366"},{"url":"https://git.kernel.org/stable/c/bf260dc9e7b8bfcd06c1a8e739797b1cf0917c72"}],"title":"KVM: SVM: Trigger new ASID allocation in both VMCBs on pCPU switch","x_generator":{"engine":"bippy-1.2.0"}}}}