{"id":"CVE-2026-98644","kind":"vulnerability","aliases":["CVE-2026-98644"],"related":[],"title":"Linux kernel: BPF batch lookup can freeze a CPU on large hash maps","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nbpf: Avoid soft lockup in __htab_map_lookup_and_delete_batch()\n\n__htab_map_lookup_and_delete_batch() has no rescheduling point. The\nbatch count bounds how many entries are copied out, not how many\nbuckets are visited, so one BPF_MAP_LOOKUP_BATCH call can walk the\nmap end to end. The empty-bucket fast path is worse: it stays inside\na single rcu_read_lock() / bpf_disable_instrumentation() section for\nany run of consecutive empty buckets.\n\nThat holds up on small maps, but it falls apart at scale. On a\n144-CPU arm64 host running a CONFIG_PREEMPT_NONE kernel, periodic\nBPF_MAP_LOOKUP_BATCH calls against an LRU hash map with 16,777,216\nbuckets held a CPU inside the batch op for 77+ seconds and triggered\nthe soft lockup watchdog.\n\nCommit 75134f16e7dd (\"bpf: Add schedule points in batch ops\") fixed this\nsame problem in the generic batch ops, but not in this htab-native path,\nwhich every htab-based hash map variant uses for its lookup[_and_delete]\nbatch ops.\n\nComplete that fix here. Leave the critical section after 64 consecutive\nempty buckets, call cond_resched_tasks_rcu_qs(), and resume at the saved\nbucket cursor. No locks are held at that point, and resuming from the\ncursor is already the function's behavior for non-empty buckets. Add the\nsame call to the per-bucket loop after copy_to_user(), where every lock\nhas been dropped. cond_resched_rcu() is not enough here: sleeping with\nbpf_prog_active elevated makes tracing programs on that CPU silently\nskip their invocations.\n\nPlain cond_resched() is not enough either. It is a no-op under PREEMPT\nand PREEMPT_LAZY, the only models arm64 and x86 have offered since\ncommit 7dadeaa6e851 (\"sched: Further restrict the preemption modes\").\nIt is also never a Tasks RCU quiescent state, in any model: the\nreschedule counts as a preemption. The walking task stays a holdout and\nstalls every synchronize_rcu_tasks() caller, ftrace and BPF trampoline\nteardown included, until the syscall returns [1].\ncond_resched_tasks_rcu_qs() is the usual tool for that [2]. It reports\nthe quiescent state at each yield and still reschedules as\ncond_resched() does on PREEMPT_NONE and PREEMPT_VOLUNTARY kernels.","summary":"A BPF_MAP_LOOKUP_BATCH call on a large hash map could walk the whole map without yielding the CPU, which triggered soft lockup warnings and stalled Tasks RCU waiters. The fix adds periodic rescheduling points to the htab batch lookup path, so a local caller can cause a denial of service but not gain code execution from this flaw as described.","severity":"unknown","cvss":null,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["bpf","kernel","linux","package","self-hosted"],"not_affected_if":["no-bpf-hash-map-batch-use"],"how_to_check":"- `uname -r` and compare against the kernel version that carries the upstream fix commit (unknown from sources)\n- Check whether the host runs BPF hash map batch operations (`bpftool map` usage) on very large LRU hash maps, unknown from sources","fix":"- Update the Linux kernel to a release that includes the upstream commit resolving this issue (exact version unknown from sources)","status":"open","published_at":"2026-10-18T16:17:17+00:00","first_seen_at":"2026-10-23T21:57:16+00:00","triaged_at":"2026-10-25T05:20:52+00:00","updated_at":"2026-10-25T05:20:52+00:00","category":"package","confidence":0.5,"triage":{"at":"2026-10-25T05:20:52Z","model":"claude-haiku-5-5","output":{"fix":"- Update the Linux kernel to a release that includes the upstream commit resolving this issue (exact version unknown from sources)","reach":"local","title":"Linux kernel: BPF batch lookup can freeze a CPU on large hash maps","impact":"low","summary":"A BPF_MAP_LOOKUP_BATCH call on a large hash map could walk the whole map without yielding the CPU, which triggered soft lockup warnings and stalled Tasks RCU waiters. The fix adds periodic rescheduling points to the htab batch lookup path, so a local caller can cause a denial of service but not gain code execution from this flaw as described.","category":"package","nicknames":[],"who_cares":["linux","kernel","bpf","self-hosted"],"confidence":0.5,"needs_auth":"unknown","worst_case":"local caller stalls a CPU and delays RCU for seconds","analyst_note":"Confirm the fixed kernel versions and stable backport list from the NVD entry or kernel.org commit referenced in the sources.","how_to_check":"- `uname -r` and compare against the kernel version that carries the upstream fix commit (unknown from sources)\n- Check whether the host runs BPF hash map batch operations (`bpftool map` usage) on very large LRU hash maps, unknown from sources","needs_analyst":true,"not_affected_if":["no-bpf-hash-map-batch-use"],"who_is_affected":"Hosts running Linux kernels without this fix that let users or services call BPF_MAP_LOOKUP_BATCH on large hash maps, such as multi-CPU arm64 servers using PREEMPT_NONE kernels. The sources do not say which kernel versions are fixed, so the exact range is unknown.","user_interaction":"no","needs_special_setup":"unknown","severity_if_unknown":"low"},"cost_usd":0.000725,"input_tokens":4104,"output_tokens":630,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'-2026':15A,18A '-98644':16A,19A '1':443C '144':191C '16':213C '2':455C '216':215C '64':284C '75134f16e7dd':234C '77':225C '777':214C '7dadeaa6e851':395C 'activ':355C 'add':59B,236C,324C 'alreadi':314C 'also':404C 'apart':186C 'arm64':193C,388C 'avoid':100C 'base':263C 'batch':4A,31B,66B,109C,115C,121C,141C,205C,222C,240C,249C,273C 'behavior':318C 'bound':123C 'bpf':3A,20B,28B,99C,138C,166C,202C,235C,353C,435C 'bucket':133C,153C,176C,216C,287C,299C,323C,332C 'call':32B,142C,206C,288C,327C 'caller':72B,432C 'caus':74B 'code':82B 'commit':233C,394C 'complet':275C 'cond':289C,344C,368C,444C,468C 'config':197C 'consecut':174C,285C 'copi':128C,335C 'could':38B 'count':122C,416C 'cpu':9A,46B,192C,219C,362C 'critic':281C 'cursor':300C,312C 'cve':14A,17A 'delet':108C,114C,272C 'denial':76B 'describ':88B 'disabl':167C 'drop':343C 'either':373C 'elev':356C 'empti':152C,175C,286C,322C 'empty-bucket':151C 'end':147C,149C 'enough':349C,372C 'entri':126C 'everi':260C,339C,428C 'execut':83B 'fall':185C 'fast':154C 'fix':58B,242C,277C 'flaw':86B 'follow':94C 'freez':7A 'ftrace':433C 'function':316C 'gain':81B 'generic':248C 'hash':12A,36B,210C,264C 'held':217C,304C 'hold':178C 'holdout':425C 'host':26B,194C 'htab':65B,104C,110C,256C,262C 'htab-bas':261C 'htab-nat':255C 'includ':438C 'insid':160C,220C 'instrument':168C 'invoc':366C 'kernel':2A,21B,92C,200C,477C 'larg':11A,35B 'lazi':384C 'leav':279C 'linux':1A,22B,91C 'local':71B 'lock':165C,302C,340C 'lockup':50B,102C,231C 'lookup':5A,30B,67B,106C,112C,140C,204C,270C 'loop':333C 'lru':209C 'make':357C 'mani':125C,132C 'map':13A,29B,37B,42B,105C,111C,139C,146C,182C,203C,211C,265C 'mode':401C 'model':387C,413C 'nativ':257C 'never':405C 'no-op':377C 'non':321C 'non-empti':320C 'none':199C,473C 'offer':392C 'one':137C 'op':223C,241C,250C,274C,379C 'packag':23B 'path':68B,155C,258C 'per':331C 'per-bucket':330C 'period':60B,201C 'plain':367C 'point':62B,119C,238C,307C 'preempt':198C,381C,383C,472C,475C 'preemption':400C,419C 'problem':245C 'prog':354C 'program':359C 'qs':293C,448C 'quiescent':409C,459C 'rcu':55B,163C,292C,346C,408C,430C,447C 'read':164C 'report':457C 'resch':290C,345C,369C,445C,469C 'reschedul':61B,118C,415C,466C 'resolv':98C 'restrict':398C 'resum':295C,309C 'return':442C 'run':172C,195C 'save':298C 'scale':188C 'sched':396C 'schedul':237C 'second':226C 'section':169C,282C 'self':25B 'self-host':24B 'servic':78B 'silent':363C 'sinc':393C 'singl':162C 'skip':364C 'sleep':351C 'small':181C 'soft':49B,101C,230C 'stall':53B,427C 'state':410C,460C 'stay':159C,423C 'still':465C 'synchron':429C 'syscal':441C 'task':54B,291C,407C,422C,431C,446C 'teardown':437C 'tool':452C 'trace':358C 'trampolin':436C 'trigger':48B,228C 'use':267C 'user':337C 'usual':451C 'variant':266C 'visit':135C 'voluntari':476C 'vulner':95C 'waiter':56B 'walk':39B,144C,421C 'warn':51B 'watchdog':232C 'whole':41B 'without':43B 'wors':157C 'x86':390C 'yield':44B,463C","nicknames":[],"mitigations":[],"who_affected":"Hosts running Linux kernels without this fix that let users or services call BPF_MAP_LOOKUP_BATCH on large hash maps, such as multi-CPU arm64 servers using PREEMPT_NONE kernels. The sources do not say which kernel versions are fixed, so the exact range is unknown.","tier":"C","verified":{},"analyst_revisit_at":null,"first_public_at":"2026-10-18T16:17:17+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-98644","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-98644","severity":null,"published_at":"2026-10-18T16:17:17+00:00","modified_at":"2026-10-19T10:19:28+00:00","first_seen_at":"2026-10-23T21:57:16+00:00","fetched_at":"2026-10-23T22:28:56+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"}