Re: [PATCH 01/13] net: phy: Support Marvell 88E1318S PHY

[Ezequiel Garcia <ezequiel@xxxxxxxxxxxxxxxxxxxx>]

Hi Sebastian,

On 05/03/2015 07:47 PM, Sebastian Hesselbarth wrote:
> On 03.05.2015 11:50, Sebastian Hesselbarth wrote:
>> On 02.05.2015 22:28, Ezequiel Garcia wrote:
>>> On 05/02/2015 01:53 PM, Ezequiel Garcia wrote:
>>>> On 04/09/2015 10:03 PM, Sebastian Hesselbarth wrote:
>>>>> This adds support for the Marvell 88E1318S Gigabit Ethernet PHY.
>>>>>
>>>>> Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@xxxxxxxxx>
>> [...]
>>>>
>>>> With this commit, the Openblocks A6 PHY is not brought up, and so
>>>> network is broken.
>>>>
>>>> I'll look into it (unless you have an idea),
>>>>
>>>
>>> Hm.. the devicetree needs to specify the phy-mode:
>>>
>>> diff --git a/dts/src/arm/kirkwood-openblocks_a6.dts
>>> b/dts/src/arm/kirkwood-openblocks_a6.dts
>>> index fb9dc22..c53e37f 100644
>>> --- a/dts/src/arm/kirkwood-openblocks_a6.dts
>>> +++ b/dts/src/arm/kirkwood-openblocks_a6.dts
>>> @@ -171,6 +171,7 @@
>>>   &eth0 {
>>>          status = "okay";
>>>          ethernet0-port@0 {
>>> +               phy-mode = "rgmii-id";
>>>                  phy-handle = <&ethphy0>;
>>>          };
>>>   };
>>>
>>> However, network works really crappy even after this change (either in
>>> barebox or in Linux).
>>
>> Ezequiel,
>>
>> RGMII phy-mode would have been my first guess.
>>
>> Can you define "crappy"  with respect to ethernet?
>>

On barebox, I'm seeing tftp making slow progress but with lots and lots of
timeouts. And then on the Linux side, there are many NFSroot disconnects,
and userspace is unresponsive.

I haven't really done any serious measurements.

>> The patch doesn't do anything different than the linux marvell phy
>> code would do.

Yeah, I don't think the issue in on the PHY init.

>> Do you have access to A6's u-boot source and can match
>> up the PHY init sequence?

Yes. Although U-Boot is barely readable, can't see anything interesting
being done with the PHY. (from ftp://ftp.plathome.co.jp/pub/OBSA6/uboot/)

MV_VOID mvEthE1310PhyBasicInit(MV_U32 ethPortNum)
{
        MV_U16 reg;

        /* Set phy address */
        MV_REG_WRITE(ETH_PHY_ADDR_REG(ethPortNum), mvBoardPhyAddrGet(ethPortNum));

        /* reset the phy */
        mvEthPhyRegRead(mvBoardPhyAddrGet(ethPortNum),0,&reg);
        reg |= BIT15;
        mvEthPhyRegWrite(mvBoardPhyAddrGet(ethPortNum),0,reg);
        mvOsDelay(10);

        /* Leds link and activity*/
        mvEthPhyRegWrite(mvBoardPhyAddrGet(ethPortNum),22,0x3);
        mvEthPhyRegRead(mvBoardPhyAddrGet(ethPortNum),16,&reg);
        reg &= ~0xfff;
        reg |= 0x671;
        mvEthPhyRegWrite(mvBoardPhyAddrGet(ethPortNum),16,reg);
        mvEthPhyRegWrite(mvBoardPhyAddrGet(ethPortNum),22,0x0);
}

> 
> Maybe we need some 1V8 RGMII setup for A6 too?
> 
> https://lkml.org/lkml/2015/5/3/159
> 
> IO Configuration 0 register 0xf10100e0, bit 7:

Hm, it seems it's off in U-Boot:

openblocks>> md f10100e0
f10100e0: 1b1b1b1b 011b1b1b 00000000 00000000    ................

-- 
Ezequiel Garcia, VanguardiaSur
www.vanguardiasur.com.ar

_______________________________________________
barebox mailing list
barebox@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/barebox