Re: [PATCH net-next] net: mdio: mscc-miim: Use devm_platform_get_and_ioremap_resource()

[Andrew Lunn <andrew@lunn.ch>]

On Fri, Jun 11, 2021 at 09:35:04AM +0800, Yang Yingliang wrote:
> Hi,
> 
> On 2021/6/11 0:01, Andrew Lunn wrote:
> > > -	dev->regs = devm_ioremap_resource(&pdev->dev, res);
> > > +	dev->regs = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
> > >   	if (IS_ERR(dev->regs)) {
> > Here, only dev->regs is considered.
> > 
> > >   		dev_err(&pdev->dev, "Unable to map MIIM registers\n");
> > >   		return PTR_ERR(dev->regs);
> > >   	}
> > 
> > 
> > > +	dev->phy_regs = devm_platform_get_and_ioremap_resource(pdev, 1, &res);
> > > +	if (res && IS_ERR(dev->phy_regs)) {
> > Here you look at both res and dev->phy_regs.
> > 
> > This seems inconsistent. Can devm_platform_get_and_ioremap_resource()
> > return success despite res being NULL?
> No, if res is NULL, devm_platform_get_and_ioremap_resource() returns failed.
> But, before this patch, if the internal phy res is NULL, it doesn't return
> error
> code, so I checked the res to make sure it doesn't change the origin code
> logic.

O.K, so IORESOURCE_MEM, 1 is optional. By making this change, i think
you have made this less clear. So i would say it is O.K. to change the
first platform_get_resource(pdev, IORESOURCE_MEM, 0) and
devm_ioremap_resource(&pdev->dev, res) to one call, but i would leave
the second pair alone.

    Andrew