Re: [PATCH bpf-next v2] bpf: arm64: Fix panic due to missing BTI at indirect jump targets

[Anton Protopopov]

On 25/12/25 07:46PM, Xu Kuohai wrote:
> On 12/25/2025 7:04 PM, Anton Protopopov wrote:
> 
> [...]
> 
> > > > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > > > index da6a00dd313f..a3a89d4b4dae 100644
> > > > --- a/include/linux/bpf.h
> > > > +++ b/include/linux/bpf.h
> > > > @@ -3875,13 +3875,32 @@ void bpf_insn_array_release(struct bpf_map *map);
> > > >    void bpf_insn_array_adjust(struct bpf_map *map, u32 off, u32 len);
> > > >    void bpf_insn_array_adjust_after_remove(struct bpf_map *map, u32 off, u32 len);
> > > > +enum bpf_insn_array_type {
> > > > +	BPF_INSN_ARRAY_VOID,
> > > 
> > > What is the purpose for BPF_INSN_ARRAY_VOID? Do we really need it?
> > 
> > There seems to be no need for a name for the default case,
> > but BPF_INSN_ARRAY_JUMP_TABLE should be != 0, so can be just
> > 
> > enum bpf_insn_array_type {
> > 	BPF_INSN_ARRAY_JUMP_TABLE = 1,
> > };
> > 
> 
> Having only BPF_INSN_ARRAY_JUMP_TABLE feels incomplete, since there
> would be no enum value to indicate an instruction array without a
> specific purpose, like the insn_arrays created in selftests [1].

Yes, but it is also never used explicitly, right?
The only usage is in "x != BPF_INSN_ARRAY_JUMP_TABLE".

> [1] https://lore.kernel.org/bpf/20251105090410.1250500-5-a.s.protopopov@xxxxxxxxx/
> 
> [...]
>