{"cve":"CVE-2026-53034","nvd":{"published":"2026-06-24T17:17:14.843","modified":"2026-07-15T14:38:08.400","status":"Analyzed","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nbpf, sockmap: Fix af_unix null-ptr-deref in proto update\n\nunix_stream_connect() sets sk_state (`WRITE_ONCE(sk->sk_state,\nTCP_ESTABLISHED)`) _before_ it assigns a peer (`unix_peer(sk) = newsk`).\nsk_state == TCP_ESTABLISHED makes sock_map_sk_state_allowed() believe that\nsocket is properly set up, which would include having a defined peer. IOW,\nthere's a window when unix_stream_bpf_update_proto() can be called on\nsocket which still has unix_peer(sk) == NULL.\n\n         CPU0 bpf                            CPU1 connect\n         --------                            ------------\n\n                                WRITE_ONCE(sk->sk_state, TCP_ESTABLISHED)\nsock_map_sk_state_allowed(sk)\n...\nsk_pair = unix_peer(sk)\nsock_hold(sk_pair)\n                                sock_hold(newsk)\n                                smp_mb__after_atomic()\n                                unix_peer(sk) = newsk\n\nBUG: kernel NULL pointer dereference, address: 0000000000000080\nRIP: 0010:unix_stream_bpf_update_proto+0xa0/0x1b0\nCall Trace:\n  sock_map_link+0x564/0x8b0\n  sock_map_update_common+0x6e/0x340\n  sock_map_update_elem_sys+0x17d/0x240\n  __sys_bpf+0x26db/0x3250\n  __x64_sys_bpf+0x21/0x30\n  do_syscall_64+0x6b/0x3a0\n  entry_SYSCALL_64_after_hwframe+0x76/0x7e\n\nInitial idea was to move peer assignment _before_ the sk_state update[1],\nbut that involved an additional memory barrier, and changing the hot path\nwas rejected.\nThen a NULL check during proto update in unix_stream_bpf_update_proto() was\nconsidered[2], but the follow-up discussion[3] focused on the root cause,\ni.e. sockmap update taking a wrong lock. Or, more specifically, missing\nunix_state_lock()[4].\nIn the end it was concluded that teaching sockmap about the af_unix locking\nwould be unnecessarily complex[5].\nComplexity aside, since BPF_PROG_TYPE_SCHED_CLS and BPF_PROG_TYPE_SCHED_ACT\nare allowed to update sockmaps, sock_map_update_elem() taking the unix\nlock, as it is currently implemented in unix_state_lock():\nspin_lock(&unix_sk(s)->lock), would be problematic. unix_state_lock() taken\nin a process context, followed by a softirq-context TC BPF program\nattempting to take the same spinlock -- deadlock[6].\nThis way we circled back to the peer check idea[2].\n\n[1]: https://lore.kernel.org/netdev/ba5c50aa-1df4-40c2-ab33-a72022c5a32e@rbox.co/\n[2]: https://lore.kernel.org/netdev/20240610174906.32921-1-kuniyu@amazon.com/\n[3]: https://lore.kernel.org/netdev/7603c0e6-cd5b-452b-b710-73b64bd9de26@linux.dev/\n[4]: https://lore.kernel.org/netdev/CAAVpQUA+8GL_j63CaKb8hbxoL21izD58yr1NvhOhU=j+35+3og@mail.gmail.com/\n[5]: https://lore.kernel.org/bpf/CAAVpQUAHijOMext28Gi10dSLuMzGYh+jK61Ujn+fZ-wvcODR2A@mail.gmail.com/\n[6]: https://lore.kernel.org/bpf/dd043c69-4d03-46fe-8325-8f97101435cf@linux.dev/\n\nSummary of scenarios where af_unix/stream connect() may race a sockmap\nupdate:\n\n1. connect() vs. bpf(BPF_MAP_UPDATE_ELEM), i.e. sock_map_update_elem_sys()\n\n   Implemented NULL check is sufficient. Once assigned, socket peer won't\n   be released until socket fd is released. And that's not an issue because\n   sock_map_update_elem_sys() bumps fd refcnf.\n\n2. connect() vs BPF program doing update\n\n   Update restricted per verifier.c:may_update_sockmap() to\n\n      BPF_PROG_TYPE_TRACING/BPF_TRACE_ITER\n      BPF_PROG_TYPE_SOCK_OPS (bpf_sock_map_update() only)\n      BPF_PROG_TYPE_SOCKET_FILTER\n      BPF_PROG_TYPE_SCHED_CLS\n      BPF_PROG_TYPE_SCHED_ACT\n      BPF_PROG_TYPE_XDP\n      BPF_PROG_TYPE_SK_REUSEPORT\n      BPF_PROG_TYPE_FLOW_DISSECTOR\n      BPF_PROG_TYPE_SK_LOOKUP\n\n   Plus one more race to consider:\n\n            CPU0 bpf                            CPU1 connect\n            --------                            ------------\n\n                                   WRITE_ONCE(sk->sk_state, TCP_ESTABLISHED)\n   sock_map_sk_state_allowed(sk)\n                                   sock_hold(newsk)\n                                   smp_mb__after_atomic()\n                \n---truncated---","cvss_v4":null,"cvss_v3":{"version":"3.1","vector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H","score":5.5},"cwes":["CWE-476"],"references":["https://git.kernel.org/stable/c/041eb6348d73ee5e15fc8161f1eac5a6e8289ca0","https://git.kernel.org/stable/c/37bfcd164161b47d00b1c3bd20adc816a6977ce0","https://git.kernel.org/stable/c/4913c94a3adcdbb64c552110c0c243cb1fdbb317","https://git.kernel.org/stable/c/75b7d3b3f8bd4e59eb3af1b11a43c64c0c2db6f4","https://git.kernel.org/stable/c/a94d3dd78ee8b63e6b8ad629081c952c93ee5a10"],"products":["linux:linux_kernel"],"source":"nvd","synced_at":"2026-10-19T22:41:17.107060303+00:00"},"epss":{"cve":"CVE-2026-53034","score":0.00128,"percentile":0.02436,"date":"2026-10-31"},"kev":null,"osv":[],"attribution":["This product uses the NVD API but is not endorsed or certified by the NVD.","EPSS scores: FIRST Exploit Prediction Scoring System, https://www.first.org/epss","Known exploited vulnerabilities: CISA Known Exploited Vulnerabilities Catalog, https://www.cisa.gov/known-exploited-vulnerabilities-catalog","Package advisories: OSV (https://osv.dev), from the GitHub Advisory Database, PyPA, Go, RustSec and other ecosystem databases, under CC-BY 4.0 or CC0","CVE records: the CVE Program's CVE List (https://www.cve.org), under its terms of use (https://www.cve.org/Legal/TermsOfUse). CVE is a registered trademark of The MITRE Corporation; this product is not affiliated with the CVE Program."]}