Re: [PATCH v3 bpf-next 1/3] bpf: Tweak definition of KF_TRUSTED_ARGS

[Alexei Starovoitov <alexei.starovoitov@gmail.com>]

On Wed, Sep 21, 2022 at 9:49 AM Lorenzo Bianconi <lorenzo@kernel.org> wrote:
>
> +               /* These register types have special constraints wrt ref_obj_id
> +                * and offset checks. The rest of trusted args don't.
> +                */
> +               obj_ptr = reg->type == PTR_TO_CTX || reg->type == PTR_TO_BTF_ID ||
> +                         reg2btf_ids[base_type(reg->type)];
> +

..

>                 /* Check if argument must be a referenced pointer, args + i has
>                  * been verified to be a pointer (after skipping modifiers).
> +                * PTR_TO_CTX is ok without having non-zero ref_obj_id.
>                  */

Kumar,

Looking forward to your subsequent patch to split this function.
It's definitely getting unwieldy.

The comment above is double confusing.
1. I think you meant to say "PTR_TO_CTX is ok with zero ref_obj_id",
right? That double negate is not easy to parse.

2.
PTR_TO_CTX cannot have ref_obj_id != 0.
At least I don't think it's possible, but the comment implies
that such a case may exist.

I applied anyway, since big refactoring is coming shortly, right?