{"id":"CVE-2026-98464","state":"PUBLISHED","title":"RISC-V: KVM: Synchronize hrtimer callback during teardown","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nRISC-V: KVM: Synchronize hrtimer callback during teardown\n\nThe non-Sstc hrtimer callback clears next_set before its final uses of\nthe enclosing vCPU.  If teardown observes next_set as false while the\ncallback is still running, kvm_riscv_vcpu_timer_cancel() skips\nhrtimer_cancel() and kvm_destroy_vcpus() can free the vCPU before the\ncallback enters kvm_riscv_vcpu_set_interrupt().\n\nA guest can arm the timer with SBI TIME and request shutdown with SBI\nlegacy shutdown or SRST.  A VMM that honors KVM_EXIT_SYSTEM_EVENT and\ndestroys the VM supplies the teardown side of the race; no post-launch\nhost ioctl is needed to arm or request teardown.\n\nOn upstream master 62cc90241548, generic KASAN reported:\n\n  BUG: KASAN: slab-use-after-free in do_raw_spin_lock\n  Write of size 4 at addr ff60000005e58898\n\n  kvm_riscv_vcpu_set_interrupt\n  kvm_riscv_vcpu_hrtimer_expired\n  __hrtimer_run_queues\n  hrtimer_interrupt\n\nThe object was allocated by KVM_CREATE_VCPU and freed concurrently by:\n\n  kvm_destroy_vcpus\n  kvm_arch_destroy_vm\n  kvm_destroy_vm\n  __fput\n\nFor deterministic validation, I added mdelay(1000) immediately after\nthe existing next_set = false assignment.  This only widens the\nexisting post-clear callback window.  A no-delay trace build naturally\nreached the callback-after-teardown-start/before-deinit ordering in 12\nof 200 runs, but 1,500 stock-kernel stress iterations did not produce a\nKASAN report, so natural reproduction is timing-sensitive.\n\nAlways invoke hrtimer_cancel() for an initialized timer.  Preserve the\nexisting -EINVAL result when the timer is no longer set, but only after\nsynchronizing with a running callback.\n\nWith this patch, hrtimer_cancel() blocked for the full widened callback\nwindow before vCPU destruction.  KASAN reported no error in 100\nfixed-and-widened runs or 200 fix-only timing-sweep runs.","publishedAt":"2026-10-18T15:15:16.624Z","updatedAt":"2026-10-24T14:08:56.354Z","severity":{"score":9.3,"label":"Critical","version":"3.1","vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H","source":"Linux"},"assigner":"Linux","products":["Linux Linux"],"affected":[{"vendor":"Linux","product":"Linux","defaultStatus":"unaffected","versions":[{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"9bdbae64ae0dab93041ebc57ec4e93def793fc2d","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"2af03b606aadaaeb386658ef798c79bdcdef87c1","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"e66aced74f0122380603c9910a71d77618b28f0e","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"ea91f05a55b807ead3166546a2c75f4185ca645f","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"3cc6184bd5920c9685d87f4e260568feba08650d","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"aaad136d56d91252517272b68cd533e5714698d5","versionType":"git"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/riscv/kvm/vcpu_timer.c"]},{"vendor":"Linux","product":"Linux","defaultStatus":"affected","versions":[{"version":"5.16","status":"affected"},{"version":"0","status":"unaffected","lessThan":"5.16","versionType":"semver"},{"version":"6.1.189","status":"unaffected","lessThanOrEqual":"6.1.*","versionType":"semver"},{"version":"6.6.158","status":"unaffected","lessThanOrEqual":"6.6.*","versionType":"semver"},{"version":"6.12.112","status":"unaffected","lessThanOrEqual":"6.12.*","versionType":"semver"},{"version":"6.18.55","status":"unaffected","lessThanOrEqual":"6.18.*","versionType":"semver"},{"version":"7.2.9","status":"unaffected","lessThanOrEqual":"7.2.*","versionType":"semver"},{"version":"7.3","status":"unaffected","lessThanOrEqual":"*","versionType":"original_commit_for_fix"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/riscv/kvm/vcpu_timer.c"]}],"references":["https://git.kernel.org/stable/c/9bdbae64ae0dab93041ebc57ec4e93def793fc2d","https://git.kernel.org/stable/c/2af03b606aadaaeb386658ef798c79bdcdef87c1","https://git.kernel.org/stable/c/e66aced74f0122380603c9910a71d77618b28f0e","https://git.kernel.org/stable/c/ea91f05a55b807ead3166546a2c75f4185ca645f","https://git.kernel.org/stable/c/3cc6184bd5920c9685d87f4e260568feba08650d","https://git.kernel.org/stable/c/aaad136d56d91252517272b68cd533e5714698d5"],"weaknesses":[],"ingestedAt":"2026-10-29T20:13:43.249Z","source":{"repository":"CVEProject/cvelistV5","path":"cves/2026/98xxx/CVE-2026-98464.json","commit":"556c9d3a8cd17e218bb8e452ba7cd619a384ff47"},"raw":{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-98464","state":"PUBLISHED","datePublished":"2026-10-18T15:15:16.624Z","dateUpdated":"2026-10-24T14:08:56.354Z","assignerShortName":"Linux","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","dateReserved":"2026-09-25T10:25:14.350Z"},"containers":{"cna":{"title":"RISC-V: KVM: Synchronize hrtimer callback during teardown","descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nRISC-V: KVM: Synchronize hrtimer callback during teardown\n\nThe non-Sstc hrtimer callback clears next_set before its final uses of\nthe enclosing vCPU.  If teardown observes next_set as false while the\ncallback is still running, kvm_riscv_vcpu_timer_cancel() skips\nhrtimer_cancel() and kvm_destroy_vcpus() can free the vCPU before the\ncallback enters kvm_riscv_vcpu_set_interrupt().\n\nA guest can arm the timer with SBI TIME and request shutdown with SBI\nlegacy shutdown or SRST.  A VMM that honors KVM_EXIT_SYSTEM_EVENT and\ndestroys the VM supplies the teardown side of the race; no post-launch\nhost ioctl is needed to arm or request teardown.\n\nOn upstream master 62cc90241548, generic KASAN reported:\n\n  BUG: KASAN: slab-use-after-free in do_raw_spin_lock\n  Write of size 4 at addr ff60000005e58898\n\n  kvm_riscv_vcpu_set_interrupt\n  kvm_riscv_vcpu_hrtimer_expired\n  __hrtimer_run_queues\n  hrtimer_interrupt\n\nThe object was allocated by KVM_CREATE_VCPU and freed concurrently by:\n\n  kvm_destroy_vcpus\n  kvm_arch_destroy_vm\n  kvm_destroy_vm\n  __fput\n\nFor deterministic validation, I added mdelay(1000) immediately after\nthe existing next_set = false assignment.  This only widens the\nexisting post-clear callback window.  A no-delay trace build naturally\nreached the callback-after-teardown-start/before-deinit ordering in 12\nof 200 runs, but 1,500 stock-kernel stress iterations did not produce a\nKASAN report, so natural reproduction is timing-sensitive.\n\nAlways invoke hrtimer_cancel() for an initialized timer.  Preserve the\nexisting -EINVAL result when the timer is no longer set, but only after\nsynchronizing with a running callback.\n\nWith this patch, hrtimer_cancel() blocked for the full widened callback\nwindow before vCPU destruction.  KASAN reported no error in 100\nfixed-and-widened runs or 200 fix-only timing-sweep runs."}],"affected":[{"vendor":"Linux","product":"Linux","defaultStatus":"unaffected","versions":[{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"9bdbae64ae0dab93041ebc57ec4e93def793fc2d","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"2af03b606aadaaeb386658ef798c79bdcdef87c1","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"e66aced74f0122380603c9910a71d77618b28f0e","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"ea91f05a55b807ead3166546a2c75f4185ca645f","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"3cc6184bd5920c9685d87f4e260568feba08650d","versionType":"git"},{"version":"3a9f66cb25e18a3eeca36c08d9f823a35b3ddc22","status":"affected","lessThan":"aaad136d56d91252517272b68cd533e5714698d5","versionType":"git"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/riscv/kvm/vcpu_timer.c"]},{"vendor":"Linux","product":"Linux","defaultStatus":"affected","versions":[{"version":"5.16","status":"affected"},{"version":"0","status":"unaffected","lessThan":"5.16","versionType":"semver"},{"version":"6.1.189","status":"unaffected","lessThanOrEqual":"6.1.*","versionType":"semver"},{"version":"6.6.158","status":"unaffected","lessThanOrEqual":"6.6.*","versionType":"semver"},{"version":"6.12.112","status":"unaffected","lessThanOrEqual":"6.12.*","versionType":"semver"},{"version":"6.18.55","status":"unaffected","lessThanOrEqual":"6.18.*","versionType":"semver"},{"version":"7.2.9","status":"unaffected","lessThanOrEqual":"7.2.*","versionType":"semver"},{"version":"7.3","status":"unaffected","lessThanOrEqual":"*","versionType":"original_commit_for_fix"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/riscv/kvm/vcpu_timer.c"]}],"references":[{"url":"https://git.kernel.org/stable/c/9bdbae64ae0dab93041ebc57ec4e93def793fc2d"},{"url":"https://git.kernel.org/stable/c/2af03b606aadaaeb386658ef798c79bdcdef87c1"},{"url":"https://git.kernel.org/stable/c/e66aced74f0122380603c9910a71d77618b28f0e"},{"url":"https://git.kernel.org/stable/c/ea91f05a55b807ead3166546a2c75f4185ca645f"},{"url":"https://git.kernel.org/stable/c/3cc6184bd5920c9685d87f4e260568feba08650d"},{"url":"https://git.kernel.org/stable/c/aaad136d56d91252517272b68cd533e5714698d5"}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H","baseScore":9.3,"baseSeverity":"CRITICAL"},"scenarios":[{"lang":"en","value":"AV:L - Under KVM_RUN, a guest supervisor ECALL follows kvm_riscv_vcpu_exit()->kvm_riscv_vcpu_sbi_ecall(); SBI TIME arms kvm_riscv_vcpu_update_hrtimer(), while shutdown/SRST drives teardown to the patched kvm_riscv_vcpu_timer_cancel(). This is local guest/hypervisor input, not a network protocol.\nAC:L - On a non-Sstc host, the guest chooses the timer deadline through SBI TIME and immediately issues shutdown/SRST, causing VMM teardown while the callback runs. The attacker creates both sides of the race, so the short scheduling window does not impose an attacker-uncontrollable condition.\nPR:N - kvm_riscv_vcpu_sbi_ecall() dispatches TIME and SRST or legacy shutdown from an already-running guest without any host credential or capability check; guest S-mode is within the untrusted VM authority, so no privilege in host KVM is required.\nUI:N - kvm_riscv_vcpu_sbi_system_reset() sets KVM_EXIT_SYSTEM_EVENT, after which a VMM that honors the guest's request tears down the VM automatically; no person must perform an additional action.\nS:C - The guest-triggered kvm_riscv_vcpu_hrtimer_expired() callback accesses host kernel memory after the host struct kvm_vcpu is freed, crossing the KVM guest-to-host isolation boundary.\nC:H - kvm_riscv_vcpu_hrtimer_expired() reconstructs and uses a freed vCPU, and kvm_riscv_vcpu_set_interrupt()/kvm_vcpu_kick() read reclaimed lock, wait, mode, and CPU state. Reuse of this UAF object can provide host-kernel memory disclosure primitives.\nI:H - kvm_riscv_vcpu_set_interrupt() acquires a raw spinlock and performs two __set_bit() writes through the freed vCPU before kicking it; if the slab object is reclaimed, these operations corrupt the new host object and provide exploitable UAF write primitives.\nA:H - KASAN recorded a four-byte slab-use-after-free write in do_raw_spin_lock after kvm_destroy_vcpus() freed the vCPU. Corrupting or following the reclaimed lock and wakeup state can oops, panic, or hang the host and its VMs."}]}],"providerMetadata":{"shortName":"Linux","orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","dateUpdated":"2026-10-24T14:08:56.354Z"},"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.16","versionEndExcluding":"6.1.189"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.16","versionEndExcluding":"6.6.158"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.16","versionEndExcluding":"6.12.112"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.16","versionEndExcluding":"6.18.55"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.16","versionEndExcluding":"7.2.9"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"5.16","versionEndExcluding":"7.3"}]}]}],"x_generator":{"engine":"bippy-1.2.0"}}}}}