{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-99672","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-25T10:25:14.428Z","datePublished":"2026-11-20T05:50:21.150Z","dateUpdated":"2026-11-22T11:09:35.468Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-11-22T11:09:35.468Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\ncan: protect the receive lists with a per-device lock\n\nThe receive lists of a CAN device are located in the device's\ncan_ml_priv and travel with the device when it is moved to another\nnetwork namespace. But they are protected by the per-netns\nrcvlists_lock. So can_rx_unregister() needs dev_net(dev) to pick the\nright lock and silently returns when the passed namespace does not\nmatch dev_net(dev) anymore.\n\nisotp_release() and bcm_release() leave their notifier list before they\ntake any lock that serializes against the netdev notifier, which runs\nunder RTNL. When the bound device is moved to another namespace in this\nwindow, the notifier skips the socket and the release path has to remove\nthe filters itself. isotp_release() passes the socket's namespace, which\ndoes not match dev_net(dev) anymore. bcm_rx_unreg() passes dev_net(dev),\nwhich can_rx_unregister() reads again for its check, so a namespace\nchange between both reads makes the check fail as well. In both cases\ncan_rx_unregister() silently returns. As can_rx_register() stores\nrcv->sk without a reference, the remaining receivers point to the freed\nsocket (isotp) or the freed op (bcm):\n\n  BUG: KASAN: use-after-free in isotp_rcv+0x1570/0x24d0\n  Read of size 1 at addr ffff888118130552 by task isotp_ns_uaf/578\n   can_rcv_filter+0x4af/0x8c0\n   can_receive+0x28d/0x3c0\n   can_rcv+0x2a9/0x310\n   __netif_receive_skb_one_core+0x21a/0x260\n   process_backlog+0x210/0x760\n\nBesides this, the lists of a device are protected by the rcvlists_lock\nof the namespace that is passed in. Callers which pass dev_net(dev) read\nit at different times, so the same lists may be modified under the locks\nof different namespaces while the device is moved.\n\nTaking RTNL in these release paths is no option: raw_bind(),\nraw_setsockopt() and raw_release() already hold RTNL while adding and\nremoving up to CAN_RAW_FILTER_MAX filters per socket, which unprivileged\nusers can trigger repeatedly and which makes RTNL a global bottleneck.\n\nInstead fix the root cause and add a spinlock to struct\ncan_dev_rcv_lists, which protects the receive lists of a CAN device or\nthe 'any' CAN device lists of a namespace. It is initialized in\ncan_set_ml_priv(), which is used by all CAN network devices. With this,\ncan_rx_unregister() finds the subscription independently of the\ncurrent namespace of the device. The protocols keep the device valid\nwith their tracked references. The per-netns rcvlists_lock only protects\nthe receive list statistics anymore, and CAN_RAW, CAN_BCM and CAN_ISOTP\npass the namespace that has been used for can_rx_register().\n\nSo isotp_release() and bcm_release() now remove their filters correctly\nwithout RTNL, and raw_release(), raw_setsockopt(), raw_bind() and\nisotp_bind() do not need RTNL anymore and add and remove their filters\nunder lock_sock(). To serialize bind() with NETDEV_UNREGISTER, bind()\npublishes the device inside the RCU section of its lookup, and\nraw_notify() and isotp_notify() recheck the bound device under lock_sock().\nAs RTNL does not keep NETDEV_DOWN out of bind() anymore, bind() checks\nIFF_UP after publishing the bound device, with a barrier pair against\nraw_notify() and isotp_notify(), so a down transition during bind() is\nstill reported with ENETDOWN."}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","baseScore":7.8,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - isotp_release() list_del()s the socket from isotp_notifier_list before rtnl_lock(), so rtnl_setlink()'s __dev_change_net_namespace() runs NETDEV_UNREGISTER without calling isotp_notify(). can_rx_unregister(sock_net(sk), dev) then returns on the dev_net(dev) mismatch and leaves receiver.data set. A later can_send() loopback on that vcan reaches the stale entry from can_rcv().\nAC:L - The attacker creates the vcan (vcan_link_ops does not set netns_immutable), binds a CAN_ISOTP or CAN_BCM socket, and races close() with the RTM_SETLINK move, then sets IFF_UP and transmits the registered CAN id. Both threads and the device state are under the attacker's control.\nPR:L - can_create() does not check capabilities, so socket(AF_CAN, SOCK_DGRAM, CAN_ISOTP) is unprivileged. The move needs CAP_NET_ADMIN only in the netns user namespace, via netlink_net_capable() in rtnetlink_rcv_msg() and rtnl_link_get_net_capable(), which the root of an unprivileged user namespace holds.\nUI:N - The attacker alone performs socket(), isotp_bind(), close, RTM_SETLINK, and can_send() on a vcan they created.\nS:U - The dangling receiver.data is an isotp_sock or bcm_op in the host kernel; can_rcv_filter() corrupts that kernel memory. The fault stays inside the kernel's own security authority.\nC:H - can_rx_register() stores the sock or bcm_op in receiver.data with no reference. After the failed unregister, can_rcv_filter() calls isotp_rcv() on the freed isotp_sock (KASAN read of so->opt.flags) and bcm_rx_handler() memcpy()s through op->frames into an skb queued by bcm_send_to_user() to op->sk.\nI:H - isotp_rcv_ff() stores the received CAN bytes through so->rx.buf (initialized to the inline rx.sbuf) of the freed isotp_sock, and bcm_rx_update_and_send() memcpy()s rxframe into op->last_frames for op->cfsiz bytes. Reusing those freed slots turns the callbacks into kernel memory writes.\nA:H - The published KASAN trace is a use-after-free read in isotp_rcv() from can_rcv_filter() during process_backlog(). The same callback spin_lock()s so->rx_lock and cancels hrtimers on the freed object, which oopses or hangs the kernel."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["include/linux/can/can-ml.h","include/net/netns/can.h","net/can/af_can.c","net/can/bcm.c","net/can/isotp.c","net/can/raw.c"],"versions":[{"version":"8e8cda6d737d356054c9eeef642aec0e8ae7e6bc","lessThan":"59d91f555ada337458c48316987b8ef46487f885","status":"affected","versionType":"git"},{"version":"8e8cda6d737d356054c9eeef642aec0e8ae7e6bc","lessThan":"4b625dde57c0a40a933bf017ff93a64a7306aa43","status":"affected","versionType":"git"},{"version":"8e8cda6d737d356054c9eeef642aec0e8ae7e6bc","lessThan":"228e795ec650c33ebbce8ba81ce983c84d64d889","status":"affected","versionType":"git"},{"version":"8e8cda6d737d356054c9eeef642aec0e8ae7e6bc","lessThan":"0f3ad9ab40e3dad06d5c0693806e07935ceedbc3","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["include/linux/can/can-ml.h","include/net/netns/can.h","net/can/af_can.c","net/can/bcm.c","net/can/isotp.c","net/can/raw.c"],"versions":[{"version":"4.12","status":"affected"},{"version":"0","lessThan":"4.12","status":"unaffected","versionType":"semver"},{"version":"6.12.118","lessThanOrEqual":"6.12.*","status":"unaffected","versionType":"semver"},{"version":"6.18.64","lessThanOrEqual":"6.18.*","status":"unaffected","versionType":"semver"},{"version":"7.3.5","lessThanOrEqual":"7.3.*","status":"unaffected","versionType":"semver"},{"version":"7.4-rc2","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":"4.12","versionEndExcluding":"6.12.118"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"4.12","versionEndExcluding":"6.18.64"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"4.12","versionEndExcluding":"7.3.5"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"4.12","versionEndExcluding":"7.4-rc2"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/59d91f555ada337458c48316987b8ef46487f885"},{"url":"https://git.kernel.org/stable/c/4b625dde57c0a40a933bf017ff93a64a7306aa43"},{"url":"https://git.kernel.org/stable/c/228e795ec650c33ebbce8ba81ce983c84d64d889"},{"url":"https://git.kernel.org/stable/c/0f3ad9ab40e3dad06d5c0693806e07935ceedbc3"}],"title":"can: protect the receive lists with a per-device lock","x_generator":{"engine":"bippy-1.2.0"}}}}