Re: [PATCH RFC bpf-next 1/2] bpf: propagate nullness information for reg to reg comparisons

[John Fastabend <john.fastabend@gmail.com>]

Eduard Zingerman wrote:
> > On Tue, 2022-08-23 at 16:15 -0700, John Fastabend wrote:
> 
> Hi John,
> 
> Thank you for commenting!
> 
> > > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > > index 2c1f8069f7b7..c48d34625bfd 100644
> > > --- a/kernel/bpf/verifier.c
> > > +++ b/kernel/bpf/verifier.c
> > > @@ -472,6 +472,11 @@ static bool type_may_be_null(u32 type)
> > >  	return type & PTR_MAYBE_NULL;
> > >  }
> > >  
> > > +static bool type_is_pointer(enum bpf_reg_type type)
> > > +{
> > > +	return type != NOT_INIT && type != SCALAR_VALUE;
> > > +}
> > > +
> > 
> > Instead of having another helper is_pointer_value() could work here?
> > Checking if we need NOT_INIT in that helper now.
> 
> Do you mean modifying the `__is_pointer_value` by adding a check
> `reg->type != NOT_INIT`?
> 
> I tried this out and tests are passing, but __is_pointer_value /
> is_pointer_value are used in a lot of places, seem to be a scary
> change, to be honest.

Agree it looks scary I wanted to play around with it more. I agree
its not the same and off to investigate a few places we use
__is_pointer_value now. Might add a few more tests while I'm at it.

> 
> Thanks,
> Eduard