Re: [PATCH bpf-next v4 1/3] libbpf: Introduce 'btf_custom_path' to 'bpf_obj_open_opts'

[Shuyi Cheng]

On 7/16/21 12:51 PM, Andrii Nakryiko wrote:
> On Tue, Jul 13, 2021 at 5:43 AM Shuyi Cheng
> <chengshuyi@xxxxxxxxxxxxxxxxx> wrote:
> > btf_custom_path allows developers to load custom BTF, and subsequent
> > CO-RE will use custom BTF for relocation.
> >
> > Learn from Andrii's comments in [0], add the btf_custom_path parameter
> > to bpf_obj_open_opts, you can directly use the skeleton's
> > <objname>_bpf__open_opts function to pass in the btf_custom_path
> > parameter.
> >
> > Prior to this, there was also a developer who provided a patch with
> > similar functions. It is a pity that the follow-up did not continue to
> > advance. See [1].
> >
> >          [0]https://lore.kernel.org/bpf/CAEf4BzbJZLjNoiK8_VfeVg_Vrg=9iYFv+po-38SMe=UzwDKJ=Q@xxxxxxxxxxxxxx/#t
> >          [1]https://yhbt.net/lore/all/CAEf4Bzbgw49w2PtowsrzKQNcxD4fZRE6AKByX-5-dMo-+oWHHA@xxxxxxxxxxxxxx/
> >
> > Signed-off-by: Shuyi Cheng <chengshuyi@xxxxxxxxxxxxxxxxx>
> > ---
> >   tools/lib/bpf/libbpf.c | 36 ++++++++++++++++++++++++++++++------
> >   tools/lib/bpf/libbpf.h |  9 ++++++++-
> >   2 files changed, 38 insertions(+), 7 deletions(-)
> >
> > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> > index 1e04ce7..6e11a7b 100644
> > --- a/tools/lib/bpf/libbpf.c
> > +++ b/tools/lib/bpf/libbpf.c
> > @@ -498,6 +498,13 @@ struct bpf_object {
> >           * it at load time.
> >           */
> >          struct btf *btf_vmlinux;
> > +       /* Path to the custom BTF to be used for BPF CO-RE relocations.
> > +        * This custom BTF completely replaces the use of vmlinux BTF
> > +        * for the purpose of CO-RE relocations.
> > +        * NOTE: any other BPF feature (e.g., fentry/fexit programs,
> > +        * struct_ops, etc) will need actual kernel BTF at /sys/kernel/btf/vmlinux.
> > +        */
>
> this comment completely duplicates the one from bpf_object_open_opts,
> I'll remove or shorten it
>
> > +       char *btf_custom_path;
> >          /* vmlinux BTF override for CO-RE relocations */
> >          struct btf *btf_vmlinux_override;
> >          /* Lazily initialized kernel module BTFs */
> > @@ -2645,10 +2652,6 @@ static bool obj_needs_vmlinux_btf(const struct bpf_object *obj)
> >          struct bpf_program *prog;
> >          int i;
> >
> > -       /* CO-RE relocations need kernel BTF */
> > -       if (obj->btf_ext && obj->btf_ext->core_relo_info.len)
> > -               return true;
> > -
> >          /* Support for typed ksyms needs kernel BTF */
> >          for (i = 0; i < obj->nr_extern; i++) {
> >                  const struct extern_desc *ext;
> > @@ -2665,6 +2668,13 @@ static bool obj_needs_vmlinux_btf(const struct bpf_object *obj)
> >                          return true;
> >          }
> >
> > +       /* CO-RE relocations need kernel BTF, only when btf_custom_path
> > +        * is not specified
> > +        */
> > +       if (obj->btf_ext && obj->btf_ext->core_relo_info.len
> > +               && !obj->btf_custom_path)
> > +               return true;
>
> not sure why you moved it, I'll move it back to minimize code churn

You're right. 👍

regards,
Shuyi
> > +
> >          return false;
> >   }
>
> [...]