Re: [RFC bpf-next 7/8] bpftool: add BTF dump "format meta" to dump header/metadata

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Alan Maguire <alan.maguire@oracle.com>
Cc: ast@kernel.org, daniel@iogearbox.net, andrii@kernel.org,
	acme@kernel.org,  martin.lau@linux.dev, song@kernel.org,
	yhs@fb.com, john.fastabend@gmail.com,  kpsingh@kernel.org,
	sdf@google.com, haoluo@google.com, jolsa@kernel.org,
	 quentin@isovalent.com, mykolal@fb.com, bpf@vger.kernel.org
Subject: Re: [RFC bpf-next 7/8] bpftool: add BTF dump "format meta" to dump header/metadata
Date: Fri, 2 Jun 2023 09:57:45 -0700	[thread overview]
Message-ID: <CAEf4BzahDAkuaxOQf10zMv8rKPA0Hno1y6m8R1w=g+X2ryO6Kw@mail.gmail.com> (raw)
In-Reply-To: <20230531201936.1992188-8-alan.maguire@oracle.com>

On Wed, May 31, 2023 at 1:21 PM Alan Maguire <alan.maguire@oracle.com> wrote:
>
> Provide a way to dump BTF header and metadata info via
> bpftool; for example
>
> $ bpftool btf dump file vmliux format meta
> BTF: data size 4963656
> Header: magic 0xeb9f, version 1, flags 0x0, hdr_len 32
> Types: len 2927556, offset 0
> Strings: len 2035881, offset 2927556
> Metadata header found: len 184, offset 4963440, flags 0x1
> Description: 'generated by dwarves v1.25'
> CRC 0x6da2a930 ; base CRC 0x0
> Kind metadata for 20 kinds:
>        BTF_KIND_UNKN[ 0] flags 0x0    info_sz  0 elem_sz  0
>         BTF_KIND_INT[ 1] flags 0x0    info_sz  4 elem_sz  0
>         BTF_KIND_PTR[ 2] flags 0x0    info_sz  0 elem_sz  0
>       BTF_KIND_ARRAY[ 3] flags 0x0    info_sz 12 elem_sz  0
>      BTF_KIND_STRUCT[ 4] flags 0x0    info_sz  0 elem_sz 12
>       BTF_KIND_UNION[ 5] flags 0x0    info_sz  0 elem_sz 12
>        BTF_KIND_ENUM[ 6] flags 0x0    info_sz  0 elem_sz  8
>         BTF_KIND_FWD[ 7] flags 0x0    info_sz  0 elem_sz  0
>     BTF_KIND_TYPEDEF[ 8] flags 0x0    info_sz  0 elem_sz  0
>    BTF_KIND_VOLATILE[ 9] flags 0x0    info_sz  0 elem_sz  0
>       BTF_KIND_CONST[10] flags 0x0    info_sz  0 elem_sz  0
>    BTF_KIND_RESTRICT[11] flags 0x0    info_sz  0 elem_sz  0
>        BTF_KIND_FUNC[12] flags 0x0    info_sz  0 elem_sz  0
>  BTF_KIND_FUNC_PROTO[13] flags 0x0    info_sz  0 elem_sz  8
>         BTF_KIND_VAR[14] flags 0x0    info_sz  4 elem_sz  0
>     BTF_KIND_DATASEC[15] flags 0x0    info_sz  0 elem_sz 12
>       BTF_KIND_FLOAT[16] flags 0x0    info_sz  0 elem_sz  0
>    BTF_KIND_DECL_TAG[17] flags 0x1    info_sz  4 elem_sz  0
>    BTF_KIND_TYPE_TAG[18] flags 0x1    info_sz  0 elem_sz  0
>      BTF_KIND_ENUM64[19] flags 0x0    info_sz  0 elem_sz 12

nit: looks very weird to be right aligned for the "BTF_KIND_xxx"
column, let's align it left here?

Also, btfdump ([0]) emits stats on per-kind basis, and I found it
quite useful on multiple occasions, do you think it would be
worthwhile to add that to bpftool as well. It looks like this in
btfdump's case:

BTF types
=======================================
Total        2293996 bytes (90517 types)
FuncProto:    731420 bytes (21455 types)
Struct:       625944 bytes (7575 types)
Func:         480876 bytes (40073 types)
Enum:         137652 bytes (1721 types)
Ptr:          132180 bytes (11015 types)
Array:         67440 bytes (2810 types)
Union:         57744 bytes (1348 types)
Const:         28140 bytes (2345 types)
Typedef:       20964 bytes (1747 types)
Var:            5024 bytes (314 types)
Datasec:        3780 bytes (1 types)
Enum64:         1500 bytes (7 types)
Fwd:             828 bytes (69 types)
Int:             240 bytes (15 types)
Volatile:        228 bytes (19 types)
Restrict:         24 bytes (2 types)
Float:            12 bytes (1 types)

  [0] https://github.com/anakryiko/btfdump

>
> Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
> ---
>  tools/bpf/bpftool/btf.c | 46 +++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 46 insertions(+)
>
> diff --git a/tools/bpf/bpftool/btf.c b/tools/bpf/bpftool/btf.c
> index 91fcb75babe3..da4257e00ba8 100644
> --- a/tools/bpf/bpftool/btf.c
> +++ b/tools/bpf/bpftool/btf.c
> @@ -504,6 +504,47 @@ static int dump_btf_c(const struct btf *btf,
>         return err;
>  }
>

[...]