Re: [RFC PATCH v1 08/10] bpf: verifier: refactor kfunc specialization

[Eduard Zingerman <eddyz87@gmail.com>]

On Wed, 2025-10-08 at 19:27 +0100, Mykyta Yatsenko wrote:
> On 10/3/25 23:08, Eduard Zingerman wrote:
> > On Fri, 2025-10-03 at 17:04 +0100, Mykyta Yatsenko wrote:
> > 
> > [...]
> > 
> > > @@ -3354,18 +3344,29 @@ static int add_kfunc_call(struct bpf_verifier_env *env, u32 func_id, s16 offset)
> > >   			return err;
> > >   	}
> > >   
> > > +	err = btf_distill_func_proto(&env->log, desc_btf,
> > > +				     func_proto, func_name,
> > > +				     &func_model);
> > > +	if (err)
> > > +		return err;
> > > +
> > > +	call_imm = kfunc_call_imm(addr, func_id);
> > > +	/* Check whether the relative offset overflows desc->imm */
> > > +	if ((unsigned long)(s32)call_imm != call_imm) {
> >
> > This error was previously reported only when !bpf_jit_supports_far_kfunc_call().
>
> Sorry for the delayed response.
> This condition can only be true if call_imm is a 64 bit address. But if
> bpf_jit_supports_far_kfunc_call() is true, call_imm holds func_id which 
> is u32, anyway, so we can't hit this error.

Oh, fair enough.
Would be a bit more obvious, of this check was pushed into
kfunc_call_imm(), but that would change code structure for error
return, so I don't insist.

> > > +		verbose(env, "address of kernel function %s is out of range\n",
> > > +			func_name);
> > > +		return -EINVAL;
> > > +	}
> > > +

[...]