Re: [PATCH bpf-next v8 1/4] bpf: add btf pointer to struct bpf_ctx_arg_aux.

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

On 2/8/24 6:37 PM, thinker.li@gmail.com wrote:
> From: Kui-Feng Lee <thinker.li@gmail.com>
> 
> Enable the providers to use types defined in a module instead of in the
> kernel (btf_vmlinux).
> 
> Signed-off-by: Kui-Feng Lee <thinker.li@gmail.com>
> ---
>   include/linux/bpf.h | 1 +
>   kernel/bpf/btf.c    | 2 +-
>   2 files changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> index 1ebbee1d648e..9a2ee9456989 100644
> --- a/include/linux/bpf.h
> +++ b/include/linux/bpf.h
> @@ -1416,6 +1416,7 @@ struct bpf_ctx_arg_aux {
>   	u32 offset;
>   	enum bpf_reg_type reg_type;
>   	u32 btf_id;
> +	struct btf *btf;

It will leave a 4 bytes hole. Not a big deal considering the sizeof won't change 
regardless but still would be nice to avoid it.

Moving "struct btf *btf" to the top will need code churns in all the existing 
bpf_iter_* because they do not use the ".offset = " and ".reg_type = " style to 
initialize. I am going to move this pointer addition before the "u32 btf_id;" 
instead. No need to resend.

>   };
>   
>   struct btf_mod_pair {
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index 8e06d29961f1..7c6c9fefdbd6 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -6266,7 +6266,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
>   			}
>   
>   			info->reg_type = ctx_arg_info->reg_type;
> -			info->btf = btf_vmlinux;
> +			info->btf = ctx_arg_info->btf ? ctx_arg_info->btf : btf_vmlinux;
>   			info->btf_id = ctx_arg_info->btf_id;
>   			return true;
>   		}