Re: [PATCH bpf-next 3/6] bpf: Introduce function-by-function verification

[Stanislav Fomichev <sdf@fomichev.me>]

On 01/07, Alexei Starovoitov wrote:
> New llvm and old llvm with libbpf help produce BTF that distinguish global and
> static functions. Unlike arguments of static function the arguments of global
> functions cannot be removed or optimized away by llvm. The compiler has to use
> exactly the arguments specified in a function prototype. The argument type
> information allows the verifier validate each global function independently.
> For now only supported argument types are pointer to context and scalars. In
> the future pointers to structures, sizes, pointer to packet data can be
> supported as well. Consider the following example:
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -2621,8 +2621,8 @@ static s32 btf_func_check_meta(struct btf_verifier_env *env,
>  		return -EINVAL;
>  	}
>  
> -	if (btf_type_vlen(t)) {
> -		btf_verifier_log_type(env, t, "vlen != 0");
> +	if (btf_type_vlen(t) > BTF_FUNC_EXTERN) {
> +		btf_verifier_log_type(env, t, "invalid func linkage");
>  		return -EINVAL;
Sorry for bringing it up after the review:

This effectively teaches kernel about BTF_KIND_FUNC scope argument,
right? Which means, if I take clang from the tip
(https://github.com/llvm/llvm-project/commit/fbb64aa69835c8e3e9efe0afc8a73058b5a0fb3c#diff-f191c05d1eb0a6ca0e89d7e7938d73d4)
and take 5.4 kernel, it will reject BTF because it now has a
BTF_KIND_FUNC with global scope (any 'main' function is global and has
non-zero vlen).

What's the general guidance on the situation where clang starts
spitting out some BTF and released kernels reject it? Is there some list of
flags I can pass to clang to not emit some of the BTF features?
Or am I missing something?