Re: [PATCH v5 6/7] btf: Add lazy sorting validation for binary search

[Eduard Zingerman <eddyz87@gmail.com>]

On Fri, 2025-11-07 at 11:01 -0800, Alexei Starovoitov wrote:
> On Fri, Nov 7, 2025 at 10:58 AM Eduard Zingerman <eddyz87@gmail.com>
> wrote:
> > 
> > On Fri, 2025-11-07 at 10:54 -0800, Alexei Starovoitov wrote:
> > 
> > [...]
> > 
> > > > > > > @@ -610,7 +674,7 @@ s32 btf_find_by_name_kind(const
> > > > > > > struct
> > > > > > > btf
> > > > > > > *btf, const char *name, u8 kind)
> > > > > > >                       goto out;
> > > > > > >       }
> > > > > > > 
> > > > > > > -     if (btf->nr_sorted_types != BTF_NEED_SORT_CHECK) {
> > > > > > > +     if (btf_check_sorted((struct btf *)btf)) {
> > > > > >                                   ^
> > > > > > 
> > > > > > The const cast here enables the concurrent writes discussed
> > > > > > above.
> > > > > > Is
> > > > > > there a reason to mark the btf parameter as const if we're
> > > > > > modifying it?
> > > > > 
> > > > > Hi team, is casting away const an acceptable approach for our
> > > > > codebase?
> > > > 
> > > > Casting away const is undefined behaviour, e.g. see paragraph
> > > > 6.7.3.6
> > > > N1570 ISO/IEC 9899:201x Programming languages — C.
> > > > 
> > > > Both of the problems above can be avoided if kernel will do
> > > > sorted
> > > > check non-lazily. But Andrii and Alexei seem to like that
> > > > property.
> > > 
> > > Ihor is going to move BTF manipulations into resolve_btfid.
> > > Sorting of BTF should be in resolve_btfid as well.
> > > This way the build process will guarantee that BTF is sorted
> > > to the kernel liking. So the kernel doesn't even need to check
> > > that BTF is sorted.
> > 
> > This would be great.
> > Does this imply that module BTFs are sorted too?
> 
> Yes. The module build is supposed to use the kernel build tree where
> kernel BTF expectations will match resolve_btfid actions.
> Just like compiler and config flags should be the same.

There is also program BTF. E.g. btf_find_by_name_kind() is called for
program BTF in bpf_check_attach_target(). I think it would be fine to
check program BTF for being sorted at the BTF load time.