Re: [PATCH net-next] net: phy: constify phydev->drv

[Christian Marangi <ansuelsmth@gmail.com>]

On Sat, Feb 03, 2024 at 05:56:19PM +0100, Andrew Lunn wrote:
> On Fri, Feb 02, 2024 at 05:41:45PM +0000, Russell King (Oracle) wrote:
> > Device driver structures are shared between all devices that they
> > match, and thus nothing should never write to the device driver
> 
> nothing should never ???
> 
> I guess the never should be ever?
> 
> > diff --git a/drivers/net/phy/xilinx_gmii2rgmii.c b/drivers/net/phy/xilinx_gmii2rgmii.c
> > index 7fd9fe6a602b..7b1bc5fcef9b 100644
> > --- a/drivers/net/phy/xilinx_gmii2rgmii.c
> > +++ b/drivers/net/phy/xilinx_gmii2rgmii.c
> > @@ -22,7 +22,7 @@
> >  
> >  struct gmii2rgmii {
> >  	struct phy_device *phy_dev;
> > -	struct phy_driver *phy_drv;
> > +	const struct phy_driver *phy_drv;
> >  	struct phy_driver conv_phy_drv;
> >  	struct mdio_device *mdio;
> >  };
> 
> Did you build testing include xilinx_gmii2rgmii.c ? It does funky
> things with phy_driver structures.
>

Looking at the probe function it seems they only swap phy_drv with
conv_phy_drv but it doesn't seems they touch stuff in the phy_dev
struct. Looks like the thing while hackish, seems clean enough to follow
the rule of not touching the OPs and causing side effects.

-- 
	Ansuel