[1/2,v4] fs_enet/mii-fec.c: fix MII speed calculation

The MII speed calculation was based on the CPU clock (ppc_proc_freq),
but for MPC512x we must use the bus clock instead.

This patch makes it use the correct clock and makes sure we don't
clobber reserved bits in the MII_SPEED register.

Signed-off-by: Wolfgang Denk <wd@denx.de>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: <netdev@vger.kernel.org>

Signed-off-by: Wolfgang Denk <wd@denx.de>
---
 drivers/net/fs_enet/mii-fec.c |   35 +++++++++++++++++++++++++++++++----
 1 files changed, 31 insertions(+), 4 deletions(-)

On Thu, Jul 16, 2009 at 3:42 PM, Wolfgang Denk<wd@denx.de> wrote:
> The MII speed calculation was based on the CPU clock (ppc_proc_freq),
> but for MPC512x we must use the bus clock instead.
>
> This patch makes it use the correct clock and makes sure we don't
> clobber reserved bits in the MII_SPEED register.
>
> Signed-off-by: Wolfgang Denk <wd@denx.de>
> Cc: Grant Likely <grant.likely@secretlab.ca>
> Cc: Kumar Gala <galak@kernel.crashing.org>
> Cc: <netdev@vger.kernel.org>

Looks good to me.  Thanks for the work!

I assume this is tested.  I have not tested this on my board, and I've
got one question below, but otherwise I think I can say....
Acked-by: Grant Likely <grant.likely@secretlab.ca>

> -       fec->mii_speed = ((ppc_proc_freq + 4999999) / 5000000) << 1;
> +       if (get_bus_freq) {
> +               clock = get_bus_freq(ofdev->node);
> +
> +               if (!clock) {
> +                       dev_err(&ofdev->dev, "could not determine IPS/IPB clock\n");
> +                       goto out_unmap_regs;
> +               }
> +       } else
> +               clock = ppc_proc_freq;
> +
> +       /* scale for a MII clock <= 2.5 MHz */
> +       speed = (clock + 2499999) / 2500000;

The calculation has changed here for non mpc5121 users.  Shouldn't the
"clock = ppc_proc_freq;" line above be "clock = ppc_proc_freq / 2;"?
Or was this also a bug in the original driver?

g.

Hi,

On Thu, Jul 16, 2009 at 11:42:25PM +0200, Wolfgang Denk wrote:
> The MII speed calculation was based on the CPU clock (ppc_proc_freq),
> but for MPC512x we must use the bus clock instead.
> 
> This patch makes it use the correct clock and makes sure we don't
> clobber reserved bits in the MII_SPEED register.
> 
> Signed-off-by: Wolfgang Denk <wd@denx.de>
> Cc: Grant Likely <grant.likely@secretlab.ca>
> Cc: Kumar Gala <galak@kernel.crashing.org>
> Cc: <netdev@vger.kernel.org>
> 
> Signed-off-by: Wolfgang Denk <wd@denx.de>
> ---
>  drivers/net/fs_enet/mii-fec.c |   35 +++++++++++++++++++++++++++++++----
>  1 files changed, 31 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/fs_enet/mii-fec.c b/drivers/net/fs_enet/mii-fec.c
> index 75a0999..62b2d7a 100644
> --- a/drivers/net/fs_enet/mii-fec.c
> +++ b/drivers/net/fs_enet/mii-fec.c
> @@ -103,11 +103,11 @@ static int fs_enet_fec_mii_reset(struct mii_bus *bus)
>  static int __devinit fs_enet_mdio_probe(struct of_device *ofdev,
>                                          const struct of_device_id *match)
>  {
> -	struct device_node *np = NULL;
>  	struct resource res;
>  	struct mii_bus *new_bus;
>  	struct fec_info *fec;
> -	int ret = -ENOMEM, i;
> +	int (*get_bus_freq)(struct device_node *) = match->data;
> +	int ret = -ENOMEM, clock, speed;
>  
>  	new_bus = mdiobus_alloc();
>  	if (!new_bus)
> @@ -133,13 +133,34 @@ static int __devinit fs_enet_mdio_probe(struct of_device *ofdev,
>  	if (!fec->fecp)
>  		goto out_fec;
>  
> -	fec->mii_speed = ((ppc_proc_freq + 4999999) / 5000000) << 1;
> +	if (get_bus_freq) {
> +		clock = get_bus_freq(ofdev->node);
> +
> +		if (!clock) {
> +			dev_err(&ofdev->dev, "could not determine IPS/IPB clock\n");
> +			goto out_unmap_regs;
> +		}
> +	} else
> +		clock = ppc_proc_freq;
> +
> +	/* scale for a MII clock <= 2.5 MHz */
> +	speed = (clock + 2499999) / 2500000;
> +
> +	/* only 6 bits (25:30) available for MII speed */
> +	if (speed > 0x3F) {
> +		speed = 0x3F;
> +		dev_err(&ofdev->dev,
> +			"MII clock (%d Hz) exceeds max (2.5 MHz)\n",
> +			clock / speed);
> +	}
> +
> +	fec->mii_speed = speed << 1;
>  
>  	setbits32(&fec->fecp->fec_r_cntrl, FEC_RCNTRL_MII_MODE);
>  	setbits32(&fec->fecp->fec_ecntrl, FEC_ECNTRL_PINMUX |
>  	                                  FEC_ECNTRL_ETHER_EN);
>  	out_be32(&fec->fecp->fec_ievent, FEC_ENET_MII);
> -	out_be32(&fec->fecp->fec_mii_speed, fec->mii_speed);
> +	clrsetbits_be32(&fec->fecp->fec_mii_speed, 0x7E, fec->mii_speed);
>  
>  	new_bus->phy_mask = ~0;
>  	new_bus->irq = kmalloc(sizeof(int) * PHY_MAX_ADDR, GFP_KERNEL);
> @@ -188,6 +209,12 @@ static struct of_device_id fs_enet_mdio_fec_match[] = {
>  	{
>  		.compatible = "fsl,pq1-fec-mdio",
>  	},
> +#if defined(CONFIG_PPC_MPC512x)
> +	{
> +		.compatible = "fsl,mpc5121-fec-mdio",
> +		.data = mpc5xxx_get_bus_frequency,
> +	},
> +#endif

Grepping through 'drivers/*' I see that #ifdefing compatible-entries is highly
uncommon (just 3 hits). I think a guideline would be useful. Most people like
to avoid #ifdefs at any cost, while I personally think it doesn't spoil
readability too much here. Other opinions?

Regards,

   Wolfram

Dear Grant Likely,

In message <fa686aa40907161544je92317dy7abea6819746847@mail.gmail.com> you wrote:
>
> I assume this is tested.  I have not tested this on my board, and I've

It was tested, but as it turns out not quite well :-(

> got one question below, but otherwise I think I can say....
...
> The calculation has changed here for non mpc5121 users.  Shouldn't the
> "clock = ppc_proc_freq;" line above be "clock = ppc_proc_freq / 2;"?
> Or was this also a bug in the original driver?

You are right. That's a bug in the new code. Fixed in next version of
the patch.

Best regards,

Wolfgang Denk

Dear Wolfram Sang,

In message <20090717093307.GB3150@pengutronix.de> you wrote:
> 
...
> > @@ -188,6 +209,12 @@ static struct of_device_id fs_enet_mdio_fec_match[] = {
> >  	{
> >  		.compatible = "fsl,pq1-fec-mdio",
> >  	},
> > +#if defined(CONFIG_PPC_MPC512x)
> > +	{
> > +		.compatible = "fsl,mpc5121-fec-mdio",
> > +		.data = mpc5xxx_get_bus_frequency,
> > +	},
> > +#endif
>
> Grepping through 'drivers/*' I see that #ifdefing compatible-entries is highly
> uncommon (just 3 hits). I think a guideline would be useful. Most people like
> to avoid #ifdefs at any cost, while I personally think it doesn't spoil
> readability too much here. Other opinions?

An older version of the patch tried to "hide" the ifdef in a 512x
specific header, so at least common code would remain clean, but I
agree with Grant that this current version looks cleaner globally.

Best regards,

Wolfgang Denk