{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-89521","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-11T19:38:34.718Z","datePublished":"2026-09-11T19:44:03.257Z","dateUpdated":"2026-09-13T06:30:19.786Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-09-13T06:30:19.786Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nsched/core: Handle pick_task() releasing the rq lock\n\nCore scheduling's pick_next_task() breaks when a ->pick_task()\nimplementation can release the rq lock. The selection state derived on entry\nis only valid while the lock is held continuously. Once a pick can drop the\nlock, an interleaving selection can invalidate all of it: the single-CPU\nfast path can commit an uncookied pick although the core went cookied during\nthe release, and forceidle committed by the interleaving selection skews the\nrestarted pass's accounting.\n\nFix it by restarting the whole selection when a pick returns RETRY_TASK\nafter releasing the lock: a single restart point above the state derivation\nreplaces the per-loop restart labels, so a retry picks up state committed by\ninterleaving selections and accounts and resets forceidle like a fresh\nselection would.\n\nneed_sync and fi_before latch across retries. Clock validity can't be\nre-derived - there is no program-ordered way to tell whether the own and\ncore rq clocks are still updated after the lock was released, as other\nlockers' pin cycles may or may not have invalidated them. When restarting,\nclear core_clock_updated so that the sibling loop re-updates the core rq,\nand update the own rq clock if invalidated."}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:H","baseScore":7.3,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - The flaw is in pick_next_task() in kernel/sched/core.c, reached only from the local __schedule() hotpath. An attacker triggers it with prctl(PR_SCHED_CORE) and their own threads; it is not reachable from network packet processing, ksmbd/nfsd, or other remote handlers.\nAC:L - The attacker controls both sides of the race: they create cookied and uncookied tasks, pin them on both SMT siblings, and generate concurrent scheduling so pick_task_scx() drops the rq lock during core-wide selection. Success does not depend on an uncontrollable victim state or memory layout.\nPR:L - prctl(PR_SCHED_CORE) is available to an unprivileged process (only ptrace_may_access on its own tasks). The lock-dropping pick_task() is sched_ext, which an attacker can load with CAP_BPF/CAP_PERFMON via a user namespace or BPF token, or which may already be loaded on the host.\nUI:N - The attacker enables core scheduling, runs their own threads on SMT siblings, and drives context switches themselves. No victim action such as mounting a filesystem or opening a file is required.\nS:U - Impact is confined to the host kernel scheduler (SMT cookie selection and forceidle/vruntime accounting). This is a same-host isolation failure, not a VM/guest-to-host escape or IOMMU/DMA boundary bypass.\nC:H - After pick_task() drops the rq lock, the single-CPU fast path can commit an uncookied pick while an interleaving selection made the core cookied, so SMT siblings run mismatched cookies and bypass core scheduling's L1TF/MDS isolation, enabling cross-domain memory disclosure via SMT side channels.\nI:L - The misplaced RETRY_TASK restart skips forceidle reset and can apply invalidated rq clocks to vruntime and forceidle accounting, which is limited corruption of scheduler metadata rather than an arbitrary kernel write or control-flow hijack.\nA:H - Invalidated rq clocks and skewed forceidle accounting after a lock-dropping pick can apply a huge bogus vruntime delta or incorrectly force-idle SMT siblings, stranding runnable tasks in a hang-class scheduling DoS on the core."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["kernel/sched/core.c"],"versions":[{"version":"4c95380701f58b8112f0b891de8d160e4199e19d","lessThan":"88ed5a66467ca2a5148b9997af9c71d8c43060ad","status":"affected","versionType":"git"},{"version":"4c95380701f58b8112f0b891de8d160e4199e19d","lessThan":"c10b216a072ff5c57bc880a05f87eb519aecc529","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["kernel/sched/core.c"],"versions":[{"version":"6.19","status":"affected"},{"version":"0","lessThan":"6.19","status":"unaffected","versionType":"semver"},{"version":"7.2.4","lessThanOrEqual":"7.2.*","status":"unaffected","versionType":"semver"},{"version":"7.3-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":"6.19","versionEndExcluding":"7.2.4"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.19","versionEndExcluding":"7.3-rc1"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/88ed5a66467ca2a5148b9997af9c71d8c43060ad"},{"url":"https://git.kernel.org/stable/c/c10b216a072ff5c57bc880a05f87eb519aecc529"}],"title":"sched/core: Handle pick_task() releasing the rq lock","x_generator":{"engine":"bippy-1.2.0"}}}}