[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAEf4BzYPfMvNt57oP7YH1Subi6vE7ptcjkdBkdRCVcs=hw5LSQ@mail.gmail.com>
Date: Wed, 7 Aug 2019 13:10:34 -0700
From: Andrii Nakryiko <andrii.nakryiko@...il.com>
To: Alexei Starovoitov <ast@...com>
Cc: Alexei Starovoitov <alexei.starovoitov@...il.com>,
Andrii Nakryiko <andriin@...com>, bpf <bpf@...r.kernel.org>,
Networking <netdev@...r.kernel.org>,
Daniel Borkmann <daniel@...earbox.net>,
Yonghong Song <yhs@...com>, Kernel Team <Kernel-team@...com>
Subject: Re: [PATCH v4 bpf-next 02/14] libbpf: convert libbpf code to use new
btf helpers
On Wed, Aug 7, 2019 at 1:01 PM Alexei Starovoitov <ast@...com> wrote:
>
> On 8/7/19 12:59 PM, Andrii Nakryiko wrote:
> > On Wed, Aug 7, 2019 at 12:30 PM Alexei Starovoitov
> > <alexei.starovoitov@...il.com> wrote:
> >>
> >> On Tue, Aug 06, 2019 at 10:37:54PM -0700, Andrii Nakryiko wrote:
> >>> Simplify code by relying on newly added BTF helper functions.
> >>>
> >>> Signed-off-by: Andrii Nakryiko <andriin@...com>
> >> ..
> >>>
> >>> - for (i = 0, vsi = (struct btf_var_secinfo *)(t + 1);
> >>> - i < vars; i++, vsi++) {
> >>> + for (i = 0, vsi = (void *)btf_var_secinfos(t); i < vars; i++, vsi++) {
> >>
> >>> + struct btf_member *m = (void *)btf_members(t);
> >> ...
> >>> case BTF_KIND_ENUM: {
> >>> - struct btf_enum *m = (struct btf_enum *)(t + 1);
> >>> - __u16 vlen = BTF_INFO_VLEN(t->info);
> >>> + struct btf_enum *m = (void *)btf_enum(t);
> >>> + __u16 vlen = btf_vlen(t);
> >> ...
> >>> case BTF_KIND_FUNC_PROTO: {
> >>> - struct btf_param *m = (struct btf_param *)(t + 1);
> >>> - __u16 vlen = BTF_INFO_VLEN(t->info);
> >>> + struct btf_param *m = (void *)btf_params(t);
> >>> + __u16 vlen = btf_vlen(t);
> >>
> >> So all of these 'void *' type hacks are only to drop const-ness ?
> >
> > Yes.
> >
> >> May be the helpers shouldn't be taking const then?
> >>
> >
> > Probably not, because then we'll have much wider-spread problem of
> > casting const pointers into non-const when passing btf_type into
> > helpers.
> > I think const as a default is the right choice, because normally BTF
> > is immutable and btf__type_by_id is returning const pointer, etc.
> > That's typical and expected use-case. btf_dedup and BTF sanitization +
> > datasec size setting pieces are an exception that have to modify BTF
> > types in place before passing it to user.
> >
> > So realistically I think we can just leave it as (void *), or I can do
> > explicit non-const type casts, or we can just not use helpers for
> > mutable cases. Do you have a preference?
>
> Hmm. Take a const into the helper and drop it there?
> I'd like to avoid all these 'void *'.
Well, I guess it's a way to do this as well. Given it's C, I guess
it's acceptable to be so free about constness. I'll update patches.
Powered by blists - more mailing lists