Re: [PATCH v2] net: ethernet: sun4i-emac: replace magic number with macro

[Conley Lee]

On 01/10/22 at 02:31下午, Andrew Lunn wrote:
> Date: Mon, 10 Jan 2022 14:31:28 +0100
> From: Andrew Lunn <andrew@xxxxxxx>
> To: Conley Lee <conleylee@xxxxxxxxxxx>
> Cc: davem@xxxxxxxxxxxxx, kuba@xxxxxxxxxx, mripard@xxxxxxxxxx,
>  wens@xxxxxxxx, clabbe.montjoie@xxxxxxxxx, netdev@xxxxxxxxxxxxxxx,
>  linux-arm-kernel@xxxxxxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v2] net: ethernet: sun4i-emac: replace magic number
>  with macro
> 
> > @@ -637,7 +637,9 @@ static void emac_rx(struct net_device *dev)
> >  		if (!rxcount) {
> >  			db->emacrx_completed_flag = 1;
> >  			reg_val = readl(db->membase + EMAC_INT_CTL_REG);
> > -			reg_val |= (0xf << 0) | (0x01 << 8);
> > +			reg_val |=
> > +				(EMAC_INT_CTL_TX_EN | EMAC_INT_CTL_TX_ABRT_EN |
> > +				 EMAC_INT_CTL_RX_EN);
> 
> Putting the first value on the next line is a bit odd. This would be
> preferred:
> 
> +			reg_val |= (EMAC_INT_CTL_TX_EN |
> +                                   EMAC_INT_CTL_TX_ABRT_EN |
> +				    EMAC_INT_CTL_RX_EN);
> 
> I also have to wonder why two | have become three? (0x01 << 8) is
> clearly a single value. (0xf << 0) should either be a single macro, or
> 4 macros since 0xf is four bits. Without looking into the details, i
> cannot say this is wrong, but it does look strange.
> 
>        Andrew
> 
Thanks for your suggestion. The (0xf << 0) mask enable tx finish and tx abort
interrupts at hardware level. And the reason this mask has 4 bits is that
sun4i emac has 2 tx channels. I reduce it into two macros EMAC_INT_CTL_TX_EN 
and EMAC_INT_CTL_TX_ABRT_EN, this may be more readable, since we always
enable both tx channels in the driver.