{"id":"CVE-2026-98061","published":"2026-09-25T11:17:35.110","lastModified":"2026-09-25T11:17:35.110","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nbpf: Reject tail calls directly from callback frames\n\nA tail call from a non-zero frame is modeled as a return from that frame.\nThe verifier makes R0 unknown and calls prepare_func_exit() for the taken\nbranch.\n\nWhen the current frame is a synchronous callback, prepare_func_exit()\nenforces the callback return-value contract and marks R0 precise. Since the\ntail-call path synthesized R0 rather than deriving it from an instruction,\nprecision backtracking reaches the callback-calling instruction with R0\nstill requested and triggers the \"callback unexpected regs\" verifier bug.\nA CAP_BPF task can therefore cause a WARN and an -EFAULT BPF_PROG_LOAD.\n\nTail calls reachable from callbacks are already rejected later by\ncheck_max_stack_depth(). Reject a tail call made directly by a callback\nbefore constructing the inconsistent return state, using the existing\ndiagnostic. Tail calls from ordinary subprograms keep their current\nbehavior.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/266aa4ad0b2e82397cd9045752c9bff03d98eddd","tags":[]},{"url":"https://git.kernel.org/stable/c/617c8266e49f45747f71e74177646a63f72b7025","tags":[]},{"url":"https://git.kernel.org/stable/c/96d31b28263c429a56e4de85bcf744a5c320b3a3","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}