{"id":"CVE-2026-98942","title":"sched_ext: Build the cid tables privately and publish them with RCU","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\nsched_ext: Build the cid tables privately and publish them with RCU\n\nThe cid tables are visible to the cid kfuncs while being modified: the\nfirst enable publishes the g…","severity":"none","vendor":"Linux","product":"Linux","affected":["Linux >= 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 < 0e1a74369bdbc357d3d9713734418d914a981553","Linux >= 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 < 3a773220d39ba993dfe5d135f8b610be10794d5d","Linux < 7.2.10","Linux (all versions)"],"published":"2026-10-23","updated":"2026-10-23","sourceUpdated":"2026-10-23T12:12:14.180Z","source":"CVEORG","sourceUrl":"https://www.cve.org/CVERecord?id=CVE-2026-98942","references":[{"url":"https://git.kernel.org/stable/c/0e1a74369bdbc357d3d9713734418d914a981553"},{"url":"https://git.kernel.org/stable/c/3a773220d39ba993dfe5d135f8b610be10794d5d"}],"tags":["cve.org"],"ingestedAt":"2026-10-23T13:24:55.561Z","slug":"CVE-2026-98942","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\nsched_ext: Build the cid tables privately and publish them with RCU\n\nThe cid tables are visible to the cid kfuncs while being modified: the\nfirst enable publishes the global pointers before filling them,\nops.init_cids() overrides rewrite them in place, and re-enables rebuild\nthem in place. A racing TRACING or SYSCALL program can read unfilled\nentries, including uninitialized memory in the kmalloc'd tables, or torn\ntopo updates.\n\nTie the tables' lifetimes to the root sched instead: each root enable\nbuilds a fresh set privately and publishes the per-table __rcu globals once\nthe layout is final, and root disable unpublishes and RCU-frees the set. A\nnon-NULL global is now always a fully built table which stays valid for the\nreader's RCU read section, and lookups stay two loads. Kfuncs treat NULL as\nno-mapping, also after the scheduler exits instead of reporting the stale\nlast mapping.\n\nThe cid kfuncs are available whether the root scheduler is cid-form or\ncpu-form, the latter to allow gradual migration to cids. Every root\ntherefore builds and publishes a default mapping.\n\nEvery reader must either be gated on scheduler liveness or NULL-check\ninside an RCU read section. Fix the two kfuncs that were neither:\nscx_bpf_this_cid() read the table with no RCU or preemption protection and\nscx_bpf_task_cid() relied on KF_RCU, which doesn't put a sleepable program\nin an RCU read section. The hotplug callbacks are instead serialized by\nretiring the tables inside the cpus_read_lock() section that clears\nscx_root.\n\nv2: Document why every root builds the tables (desc + cid.c comment).\n\n## Affected\n\n- `Linux >= 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 < 0e1a74369bdbc357d3d9713734418d914a981553`\n- `Linux >= 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 < 3a773220d39ba993dfe5d135f8b610be10794d5d`\n- `Linux < 7.2.10`\n- `Linux (all versions)`\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":[]}