Re: [PATCH v9 bpf-next 08/14] bpf: Add btf_struct_ids_match function

[Jiri Olsa <jolsa@redhat.com>]

On Tue, Aug 04, 2020 at 11:27:55PM -0700, Andrii Nakryiko wrote:

SNIP

> > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> > index 7bacc2f56061..ba05b15ad599 100644
> > --- a/kernel/bpf/btf.c
> > +++ b/kernel/bpf/btf.c
> > @@ -4160,6 +4160,37 @@ int btf_struct_access(struct bpf_verifier_log *log,
> >         return -EINVAL;
> >  }
> >
> > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > +                         int off, u32 id, u32 need_type_id)
> > +{
> > +       const struct btf_type *type;
> > +       int err;
> > +
> > +       /* Are we already done? */
> > +       if (need_type_id == id && off == 0)
> > +               return true;
> > +
> > +again:
> > +       type = btf_type_by_id(btf_vmlinux, id);
> > +       if (!type)
> > +               return false;
> > +       err = btf_struct_walk(log, type, off, 1, &id);
> 
> nit: this size=1 looks a bit artificial, seems like btf_struct_walk()
> will work with size==0 just as well, no?

right, it will work the same for 0 ... not sure why I put
originaly 1 byte for size.. probably got mixed up by some
condition in btf_struct_walk that I thought 0 wouldn't pass,
but it should work, I'll change it, it's less tricky

> 
> > +       if (err != WALK_STRUCT)
> > +               return false;
> > +
> > +       /* We found nested struct object. If it matches
> > +        * the requested ID, we're done. Otherwise let's
> > +        * continue the search with offset 0 in the new
> > +        * type.
> > +        */
> > +       if (need_type_id != id) {
> > +               off = 0;
> > +               goto again;
> > +       }
> > +
> > +       return true;
> > +}
> > +
> >  int btf_resolve_helper_id(struct bpf_verifier_log *log,
> >                           const struct bpf_func_proto *fn, int arg)
> >  {
> > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > index b6ccfce3bf4c..bb6ca19f282d 100644
> > --- a/kernel/bpf/verifier.c
> > +++ b/kernel/bpf/verifier.c
> > @@ -3960,16 +3960,21 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
> >                                 goto err_type;
> >                 }
> >         } else if (arg_type == ARG_PTR_TO_BTF_ID) {
> > +               bool ids_match = false;
> > +
> >                 expected_type = PTR_TO_BTF_ID;
> >                 if (type != expected_type)
> >                         goto err_type;
> >                 if (!fn->check_btf_id) {
> > -                       if (reg->btf_id != meta->btf_id) {
> > -                               verbose(env, "Helper has type %s got %s in R%d\n",
> > -                                       kernel_type_name(meta->btf_id),
> > -                                       kernel_type_name(reg->btf_id), regno);
> > -
> > -                               return -EACCES;
> > +                       if (reg->btf_id != meta->btf_id || reg->off) {
> 
> Will it ever succeed if reg->btf_id == meta->btf_id, but reg->off > 0?
> That would require recursively nested type, which is not possible,
> right? Or what am I missing? Is it just a simplification of the error
> handling path?

ok, I wanted to cover all possible cases, but did not realized this
one is not possible ;-) will revert it to previous version

thanks,
jirka