{"id":"CVE-2026-99262","kind":"vulnerability","aliases":["CVE-2026-99262"],"related":[],"title":"Linux kernel: io_uring timeout cancel can be lost, stalling exiting tasks","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nio_uring/timeout: don't lose a cancel as io-wq arms a timeout\n\nAn IORING_OP_TIMEOUT run by io-wq, because of IOSQE_ASYNC, has the same\nproblem as poll, futex and waitid. io_wq_submit_work() looks at\nIO_WQ_WORK_CANCEL once, before it issues the request. A cancel in between\nflags the work, gets -EALREADY, and searches ->timeout_list in vain,\nwhile the worker then adds the request and starts the timer:\n\n  io-wq worker                          io_try_cancel()\n  io_wq_submit_work()\n    IO_WQ_WORK_CANCEL isn't set\n                                        io_async_cancel_one()\n                                          sets IO_WQ_WORK_CANCEL\n                                          = -EALREADY\n                                        io_timeout_cancel() = -ENOENT\n    io_timeout()\n      adds the request to ->timeout_list\n      hrtimer_start()\n    = IOU_ISSUE_SKIP_COMPLETE\n\nNobody looks at the flag again, and the timeout runs its course. At ring\nexit, io_uring_try_cancel_requests() cancels io-wq work and then kills\nthe timeouts the same way, so an exiting task can get stuck for the full\nduration of such a timeout.\n\nLook at the flag under ->timeout_lock, which the search takes as well,\nand fail the request if it is set. If the search came first, the flag is\nvisible. If it comes later, it finds the request on the list.","summary":"In the Linux kernel's io_uring, a cancel request that arrives while an io-wq worker is starting a timeout can be dropped, so the timeout keeps running. An exiting task can then stay stuck for the full timeout duration.","severity":"low","cvss":null,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["io_uring","kernel","linux","runtime"],"not_affected_if":["io-uring-disabled"],"how_to_check":null,"fix":null,"status":"open","published_at":"2026-11-03T12:17:33+00:00","first_seen_at":"2026-11-03T12:20:40+00:00","triaged_at":"2026-11-03T12:22:42+00:00","updated_at":"2026-11-03T12:22:42+00:00","category":"runtime","confidence":0.6,"triage":{"at":"2026-11-03T12:22:42Z","model":"claude-haiku-5-5","output":{"fix":"unknown","reach":"local","title":"Linux kernel: io_uring timeout cancel can be lost, stalling exiting tasks","impact":"low","summary":"In the Linux kernel's io_uring, a cancel request that arrives while an io-wq worker is starting a timeout can be dropped, so the timeout keeps running. An exiting task can then stay stuck for the full timeout duration.","audience":"businesses","category":"runtime","nicknames":[],"who_cares":["linux","kernel","io_uring"],"confidence":0.6,"fix_status":"unknown","needs_auth":"unknown","worst_case":"exiting task hangs for the timeout duration","analyst_note":"Check the kernel.org stable commit or NVD entry for affected version ranges and the fixed versions.","how_to_check":"unknown","plain_action":"","needs_analyst":true,"not_affected_if":["io-uring-disabled"],"who_is_affected":"Systems running a Linux kernel with io_uring, where applications submit IORING_OP_TIMEOUT requests with IOSQE_ASYNC. The sources do not give affected kernel version ranges or a fix commit. Systems without io_uring, or without this request pattern, are unlikely to be affected.","user_interaction":"no","needs_special_setup":"yes","severity_if_unknown":"low"},"cost_usd":0.000687,"input_tokens":4552,"output_tokens":463,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'-2026':14A,17A '-99262':15A,18A 'add':147C,188C 'arm':87C 'arriv':35B 'async':102C,173C 'came':272C 'cancel':6A,32B,82C,121C,129C,160C,168C,174C,180C,184C,218C,220C 'come':280C 'complet':199C 'cours':211C 'cve':13A,16A 'drop':48B 'durat':65B,243C 'ealreadi':136C,181C 'enoent':185C 'exit':11A,55B,214C,235C 'fail':262C 'find':283C 'first':273C 'flag':132C,204C,251C,275C 'follow':71C 'full':63B,242C 'futex':109C 'get':135C,238C 'hrtimer':194C 'io':3A,19B,29B,39B,76C,85C,97C,112C,118C,155C,158C,161C,165C,172C,177C,182C,186C,215C,222C 'io-wq':38B,84C,96C,154C,221C 'ior':91C 'iosq':101C 'iou':196C 'isn':169C 'issu':125C,197C 'keep':52B 'kernel':2A,21B,27B,69C 'kill':227C 'later':281C 'linux':1A,22B,26B,68C 'list':140C,193C,288C 'lock':254C 'look':116C,201C,248C 'lose':80C 'lost':9A 'nobodi':200C 'one':175C 'op':92C 'poll':108C 'problem':106C 'request':33B,127C,149C,190C,219C,264C,285C 'resolv':75C 'ring':213C 'run':53B,94C,209C 'runtim':23B 'search':138C,257C,271C 'set':171C,176C,268C 'skip':198C 'stall':10A 'start':43B,151C,195C 'stay':59B 'stuck':60B,239C 'submit':114C,163C 'take':258C 'task':12A,56B,236C 'timeout':5A,45B,51B,64B,89C,93C,139C,183C,187C,192C,208C,229C,247C,253C 'timer':153C 'tri':159C,217C 'ure':4A,20B,30B,216C 'uring/timeout':77C 'vain':142C 'visibl':277C 'vulner':72C 'waitid':111C 'way':232C 'well':260C 'work':115C,120C,134C,164C,167C,179C,224C 'worker':41B,145C,157C 'wq':40B,86C,98C,113C,119C,156C,162C,166C,178C,223C","nicknames":[],"mitigations":[],"who_affected":"Systems running a Linux kernel with io_uring, where applications submit IORING_OP_TIMEOUT requests with IOSQE_ASYNC. The sources do not give affected kernel version ranges or a fix commit. Systems without io_uring, or without this request pattern, are unlikely to be affected.","tier":"C","verified":{},"analyst_revisit_at":null,"fix_status":"unknown","audience":"businesses","plain_action":null,"pkeys":[],"ptokens":[],"labels":["iouring","kernel","linux","low","runtime"],"first_public_at":"2026-11-03T12:17:33+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-99262","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-99262","severity":null,"published_at":"2026-11-03T12:17:33+00:00","modified_at":"2026-11-03T12:17:33+00:00","first_seen_at":"2026-11-03T12:20:40+00:00","fetched_at":"2026-11-03T12:20:40+00:00","note":null}],"documents":[],"notice":"Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai"}