Re: [PATCH bpf-next v3 01/10] bpf: Add dynptr type for skb metadata

[Martin KaFai Lau <martin.lau@linux.dev>]

On 7/21/25 3:52 AM, Jakub Sitnicki wrote:
> @@ -21788,12 +21798,17 @@ static void specialize_kfunc(struct bpf_verifier_env *env,
>   	if (offset)
>   		return;
>   
> -	if (func_id == special_kfunc_list[KF_bpf_dynptr_from_skb]) {
> +	if (func_id == special_kfunc_list[KF_bpf_dynptr_from_skb] ||
> +	    func_id == special_kfunc_list[KF_bpf_dynptr_from_skb_meta]) {

I don't think this check is needed. The skb_meta is writable to tc.

>   		seen_direct_write = env->seen_direct_write;
>   		is_rdonly = !may_access_direct_pkt_data(env, NULL, BPF_WRITE);

is_rdonly is always false here.

>   
> -		if (is_rdonly)
> -			*addr = (unsigned long)bpf_dynptr_from_skb_rdonly;
> +		if (is_rdonly) {
> +			if (func_id == special_kfunc_list[KF_bpf_dynptr_from_skb])
> +				*addr = (unsigned long)bpf_dynptr_from_skb_rdonly;
> +			else if (func_id == special_kfunc_list[KF_bpf_dynptr_from_skb_meta])
> +				*addr = (unsigned long)bpf_dynptr_from_skb_meta_rdonly;
> +		}

[ ... ]

> +int bpf_dynptr_from_skb_meta_rdonly(struct __sk_buff *skb, u64 flags,

so I suspect this is never used and not needed now. Please check.
It can be revisited in the future when other hooks are supported. It will be a 
useful comment in the commit message.

> +				    struct bpf_dynptr *ptr__uninit)
> +{
> +	return dynptr_from_skb_meta(skb, flags, ptr__uninit, true);
> +}