Re: [PATCH RFC net-next v2 03/12] net: mdio: mdiobus_register: update validation test

[Michael Walle]

Hi Russell,

Am 2023-01-03 23:19, schrieb Russell King (Oracle):
> On Tue, Jan 03, 2023 at 11:21:08AM +0100, Michael Walle wrote:
> > Am 2023-01-03 11:13, schrieb Russell King (Oracle):
> > > On Wed, Dec 28, 2022 at 12:07:19AM +0100, Michael Walle wrote:
> > > > +	if (!bus || !bus->name)
> > > > +		return -EINVAL;
> > > > +
> > > > +	/* An access method always needs both read and write operations */
> > > > +	if ((bus->read && !bus->write) ||
> > > > +	    (!bus->read && bus->write) ||
> > > > +	    (bus->read_c45 && !bus->write_c45) ||
> > > > +	    (!bus->read_c45 && bus->write_c45))
> > >
> > > I wonder whether the following would be even more readable:
> > >
> > > 	if (!bus->read != !bus->write || !bus->read_c45 != !bus->write_c45)
> >
> > That's what Andrew had originally. But there was a comment from Sergey 
> > [1]
> > which I agree with. I had a hard time wrapping my head around that, so 
> > I
> > just listed all the possible bad cases.
>
> The only reason I suggested it was because when looked at your code,
> it also took several reads to work out what it was trying to do!
>
> Would using !!bus->read != !!bus->write would help or make it worse,
> !!ptr being the more normal way to convert something to a boolean?

IMHO that makes it even harder. But I doubt we will find an expression
that will work for everyone. I'll go with your suggestion/Andrew's first
version in the next iteration.

-michael