[RFC PATCH net-next v3 03/15] net: dsa: mt7530: use regmap to access switch register space

[Andrew Lunn]

On Wed, Mar 29, 2023 at 07:33:17PM +0100, Daniel Golle wrote:
> On Wed, Mar 29, 2023 at 06:24:21PM +0200, Andrew Lunn wrote:
> > Thanks for splitting this patchset up. This is much easier to review.
> > 
> > > +static u32
> > > +mt7530_mii_read(struct mt7530_priv *priv, u32 reg)
> > > +{
> > > +	int ret;
> > > +	u32 val;
> > > +
> > > +	ret = regmap_read(priv->regmap, reg, &val);
> > > +	if (ret) {
> > > +		dev_err(priv->dev,
> > > +			"failed to read mt7530 register\n");
> > > +		return ret;
> > 
> > This is a u32 function. ret should be negative on error, which is
> > going to be turned positive in order to return a u32. So you probably
> > want to make this an int function.
> 
> This is a pre-existing flaw in the code. As we are accessing 32-bit
> registers there has just never been any meaningful error handling.

O.K. At least i would not return the negative error code. Return 0, or
0xdeadbeef or something. And consider adding a WARN_ON_ONCE() so it is very loud
when it goes wrong.

> I guess the correct solution would be to not use the return value only
> to indicate success or error, and use an additional u32* parameter for
> the read value.

Yes, that is what mv88e6xxx does.

> However, I was hestitating to convert all the calls (they are many) to
> follow that improved paradigm.

Yes, leave that for another time.

     Andrew