{"id":"CVE-2026-98979","kind":"vulnerability","aliases":["CVE-2026-98979"],"related":[],"title":"Linux kernel: automount under a detached mount crashes the kernel","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nnamespace: refuse an automount below a mount that is in no namespace\n\nIt's possible to add autmounts even when the parent mount isn't in the\nmount namespace of the caller. The only requirement we have is that the\nparent mount namespace must not be NULL, i.e., unmounted.\n\nProblem is that clone_private_mount() has MNT_NS_INTERNAL which makes\nthat trivially true. So that passes the test and attach_recursive_mnt()\naccepts that as a mount point and funny enough, count_mounts()\ndereferences MNT_NS_INTERNAL. The problem is it is an error pointer...\n\nSo we can reach this in userspace via fanotify. A filesystem mark on the\nlower filesystem of an overlay reports paths on the layer clone and\nreading the event hands out a descriptor on it. For example with debugfs\nas the lower layer it goes kaboom:\n\n  openat(evfd, \"tracing\", O_DIRECTORY)\n\n  Oops: general protection fault\n  KASAN: null-ptr-deref in range [0x1d0-0x1d7]\n  RIP: 0010:count_mounts+0x35/0x200\n   attach_recursive_mnt\n   finish_automount\n\nAnd since that sleeping beauty happens under namespace_sem held for\nwriting every mount operation on the system blocks from then on.\nCongrats.\n\nUse is_mounted() instead which rejects unmounted and internal mounts\nalike. The open fails with EINVAL just as it did before\nclone_private_mount() used MNT_NS_INTERNAL.","summary":"A flaw in the Linux kernel's mount handling lets an automount be attached below a mount that is in no namespace, which dereferences an error pointer and crashes the kernel. The reported trigger is opening a path reported by fanotify on an overlay's lower filesystem, after which all mount operations on the system block.","severity":"high","cvss":null,"epss":0.00209,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["containers","infrastructure","linux-kernel","overlayfs","self-hosted"],"not_affected_if":["default-config"],"how_to_check":"- Run `uname -r` and compare against the fixed kernel versions listed in the NVD entry (unknown in the sources).\n- Check whether the kernel build includes the commit that changes `attach_recursive_mnt` / `is_mounted()` in fs/namespace.c (unknown in the sources).","fix":"- Update the Linux kernel to a release that includes the upstream fix for CVE-2026-98979 (exact version unknown in the sources).\n- Reboot into the patched kernel; the fix is in the kernel, not in userspace packages.","status":"open","published_at":"2026-10-23T13:17:19+00:00","first_seen_at":"2026-10-23T21:57:33+00:00","triaged_at":"2026-10-25T04:56:19+00:00","updated_at":"2026-10-25T12:38:50+00:00","category":"infrastructure","confidence":0.6,"triage":{"at":"2026-10-25T04:56:19Z","model":"claude-haiku-5-5","output":{"fix":"- Update the Linux kernel to a release that includes the upstream fix for CVE-2026-98979 (exact version unknown in the sources).\n- Reboot into the patched kernel; the fix is in the kernel, not in userspace packages.","reach":"local","title":"Linux kernel: automount under a detached mount crashes the kernel","impact":"high","summary":"A flaw in the Linux kernel's mount handling lets an automount be attached below a mount that is in no namespace, which dereferences an error pointer and crashes the kernel. The reported trigger is opening a path reported by fanotify on an overlay's lower filesystem, after which all mount operations on the system block.","category":"infrastructure","nicknames":[],"who_cares":["linux-kernel","self-hosted","containers","overlayfs"],"confidence":0.6,"needs_auth":"unknown","worst_case":"Kernel crash and a system-wide hang of all mount operations","analyst_note":"Confirm affected and fixed kernel version ranges and the fix commit from the NVD entry and the upstream stable tree.","how_to_check":"- Run `uname -r` and compare against the fixed kernel versions listed in the NVD entry (unknown in the sources).\n- Check whether the kernel build includes the commit that changes `attach_recursive_mnt` / `is_mounted()` in fs/namespace.c (unknown in the sources).","needs_analyst":true,"not_affected_if":["default-config"],"who_is_affected":"Systems running a Linux kernel without this fix that use fanotify filesystem marks on a lower filesystem beneath an overlay mount, or that otherwise allow unprivileged users to trigger automounts. Hosts without overlay layers or fanotify watchers on such filesystems are less likely to be exposed, but the sources do not say which configurations are required.","user_interaction":"no","needs_special_setup":"yes","severity_if_unknown":"high"},"cost_usd":0.000703,"input_tokens":3738,"output_tokens":659,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'+0':257C '-2026':12A,15A '-98979':13A,16A '0010':254C '0x1d0':251C '0x1d0-0x1d7':250C '0x1d7':252C 'accept':165C 'add':108C 'alik':297C 'attach':39B,162C,259C 'autmount':109C 'automount':3A,37B,95C,263C 'beauti':268C 'block':81B,282C 'caller':123C 'clone':144C,212C,308C 'congrat':286C 'contain':17B 'count':174C,255C 'crash':8A,54B 'cve':11A,14A 'debugf':226C 'deref':247C 'derefer':49B,176C 'descriptor':220C 'detach':6A 'directori':238C 'einval':302C 'enough':173C 'error':51B,186C 'even':110C 'event':216C 'everi':276C 'evfd':235C 'exampl':224C 'fail':300C 'fanotifi':66B,196C 'fault':242C 'filesystem':72B,198C,203C 'finish':262C 'flaw':27B 'follow':87C 'funni':172C 'general':240C 'goe':232C 'hand':217C 'handl':34B 'happen':269C 'held':273C 'host':25B 'i.e':139C 'infrastructur':18B 'instead':290C 'intern':150C,179C,295C,314C 'isn':115C 'kaboom':233C 'kasan':243C 'kernel':2A,10A,21B,31B,56B,85C 'layer':211C,230C 'let':35B 'linux':1A,20B,30B,84C 'linux-kernel':19B 'lower':71B,202C,229C 'make':152C 'mark':199C 'mnt':148C,164C,177C,261C,312C 'mount':7A,33B,42B,76B,98C,114C,119C,133C,146C,169C,175C,256C,277C,289C,296C,310C 'must':135C 'namespac':47B,92C,103C,120C,134C,271C 'ns':149C,178C,313C 'null':138C,245C 'null-ptr-deref':244C 'o':237C 'oop':239C 'open':61B,299C 'openat':234C 'oper':77B,278C 'overlay':69B,206C 'overlayf':22B 'parent':113C,132C 'pass':158C 'path':63B,208C 'point':170C 'pointer':52B,187C 'possibl':106C 'privat':145C,309C 'problem':141C,181C 'protect':241C 'ptr':246C 'rang':249C 'reach':191C 'read':214C 'recurs':163C,260C 'refus':93C 'reject':292C 'report':58B,64B,207C 'requir':126C 'resolv':91C 'rip':253C 'self':24B 'self-host':23B 'sem':272C 'sinc':265C 'sleep':267C 'system':80B,281C 'test':160C 'trace':236C 'trigger':59B 'trivial':154C 'true':155C 'unmount':140C,293C 'use':287C,311C 'userspac':194C 'via':195C 'vulner':88C 'write':275C 'x35/0x200':258C","nicknames":[],"mitigations":[],"who_affected":"Systems running a Linux kernel without this fix that use fanotify filesystem marks on a lower filesystem beneath an overlay mount, or that otherwise allow unprivileged users to trigger automounts. Hosts without overlay layers or fanotify watchers on such filesystems are less likely to be exposed, but the sources do not say which configurations are required.","tier":"B","verified":{},"analyst_revisit_at":null,"first_public_at":"2026-10-23T13:17:19+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-98979","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-98979","severity":null,"published_at":"2026-10-23T13:17:19+00:00","modified_at":"2026-10-23T13:17:19+00:00","first_seen_at":"2026-10-23T21:57:33+00:00","fetched_at":"2026-10-23T22:29:03+00:00","note":null},{"source":"epss","external_id":"CVE-2026-98979","url":"https://api.first.org/data/v1/epss?cve=CVE-2026-98979","severity":null,"published_at":null,"modified_at":null,"first_seen_at":"2026-10-25T12:38:50+00:00","fetched_at":"2026-10-25T12:38:50+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"}