[PATCH 6/6] net/phy: Expose 100base-T capability only

[Gavin Shan]

On Fri, May 06, 2016 at 11:00:02AM +1000, Benjamin Herrenschmidt wrote:
>On Thu, 2016-05-05 at 22:55 +0930, Joel Stanley wrote:
>> On Thu, May 5, 2016 at 10:57 AM, Cyril Bur <cyrilbur at gmail.com>
>> wrote:
>> > 
>> > From: Gavin Shan <gwshan at linux.vnet.ibm.com>
>> > 
>> > I don't know the specific PHY model deployed on Firestone BMC, but
>> > currently used one (PHY_ID_BCM54210E) isn't correct. BCM54210E
>> > supports 1000base-T but the PHY on Firestone BMC shouldn't support
>> > that mode as uboot uses 100base-T mode.
>> > 
>> > When exposing 1000base-T capability, the MAC is forced to have the
>> > same mode though the PHY is working in 100base-T mode in fact. It
>> > causes the mode mismatch between PHY/MAC and the ethernet network
>> > doesn't work.
>> > 
>> > It's a temporary fix. After the PHY model is confirmed, we need
>> > to enable the correct PHY driver instead current one for BCM54210E.
>> > 
>> > Signed-off-by: Gavin Shan <gwshan at linux.vnet.ibm.com>
>> > 
>> > [Cyril]
>> > The plot thickens, it does appear that this IS the correct PHY on
>> > firestones. This patch is definitely required for networking to be
>> > functional though.
>> We should get someone in Austin to boot a firestone without this
>> patch
>> and see what it does. You shouldn't need a pdu as long as you get it
>> right the first time :)
>
>The MAC in the ast2400 is a gigabit MAC. So it should all work normally
>at 1000bT. If not, then something is wrong that needs to be
>investigated:
>
>	- Either the interface between MAC and PHY isn't wired
>correctly (RGMII vs. RMII, RGMII uses more pins and a diff. clock)
>
>	- Or we don't set a setting or strap somewhere to switch
>between those modes (hardware strapping bit[7:6], SCU[A0], ...)
>
>	- Both ?
>
>If we are indeed limited to 100bT, then we should expose that one way
>or another in the device-tree in order to limit the advertized speed in
>the PHY driver.
>

Ben, thanks for review. Yeah, it's a temporary (tentative) fixup. The
BCM54210E driver isn't used by farady driver only, it's not reasonable
to limit the capability in the PHY driver as it's going to affect other
driver. As you suggested in another thread, the best option would be
removing the 1000BASE-T capability in farady driver after the PHY is
probed, before it's started (connected) when it's asked by device tree.

I will investigate it when getting a chance. It seems Pin multiplexing
isn't configured correctly for RGMII in uboot. I dumped the MAC control
register (offset: 0x50). We have 100Mb mode there in uboot.

Thanks,
Gavin

>Cheers,
>Ben.
>