bpf: Return -ENOTSUPP when attaching to non-kernel BTF
authorAndrii Nakryiko <andrii@kernel.org>
Tue, 8 Dec 2020 06:43:26 +0000 (22:43 -0800)
committerDaniel Borkmann <daniel@iogearbox.net>
Tue, 8 Dec 2020 16:14:27 +0000 (17:14 +0100)
Return -ENOTSUPP if tracing BPF program is attempted to be attached with
specified attach_btf_obj_fd pointing to non-kernel (neither vmlinux nor
module) BTF object. This scenario might be supported in the future and isn't
outright invalid, so -EINVAL isn't the most appropriate error code.

Suggested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20201208064326.667389-1-andrii@kernel.org
kernel/bpf/syscall.c

index 0cd3cc2..287be33 100644 (file)
@@ -2121,8 +2121,11 @@ static int bpf_prog_load(union bpf_attr *attr, union bpf_attr __user *uattr)
                        if (IS_ERR(attach_btf))
                                return -EINVAL;
                        if (!btf_is_kernel(attach_btf)) {
+                               /* attaching through specifying bpf_prog's BTF
+                                * objects directly might be supported eventually
+                                */
                                btf_put(attach_btf);
-                               return -EINVAL;
+                               return -ENOTSUPP;
                        }
                }
        } else if (attr->attach_btf_id) {