Hi Florian,
On 2018/12/17 18:03, Florian Fainelli wrote:
> On 12/17/18 2:55 PM, John Rama wrote:
>> Hi, Andrew, Florian and Viven
>>
>> thank you your feedback !!
>>
>>>> I would recommend checking a newer kernel anyway which would have
>>>> support for the mv88e6xxx internal MDIO bus as it might allow you to
>>>> solve that specific problem. A kernel that contains
>>>> a3c53be55c955b7150cda17874c3fcb4eeb97a89 ("net: dsa: mv88e6xxx: Support
>>>> multiple MDIO busses") might work better and actually help here, though
>>>> I don't know enough about that specific switch model, Andrew or Vivien
>>>> would.
>>>>
>>>>>
>>
>> I tried with kernel 4.9.0, and confirmed that same problem happens.
>> (I wanted to try more latest one, but newer version is using phylink,
>> which I need couple of time to understand how it works)
>> I think this is not the driver side issue, maybe dsa side (but not sure for
>> newer kernel)
>> or I misunderstanding some concept.
>>
>>> Since you have the switch in single address mode, i don't see why this
>>> should not work.
>>
>> Why conflicts happens is because of the code at slave.c (@kernel 4.9.0)
>>
>> /* pseudo code */
>> dsa_slave_phy_setup() {
>> phy_dn = of_parse_phandle(port_dn, "phy-handle", 0);
>> if (phy_dn) {
>> phy_id = of_mdio_parse_addr(&slave_dev->dev, phy_dn);
>> dsa_slave_phy_connect(p, slave_dev, phy_id); <= register with phy_id
>> }
>> if (!p->phy) {
>> dsa_slave_phy_connect(p, slave_dev, p->port); <= register with port
>> number
>> }
>> }
>
> That was later fixed with this commit:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=399ba77a94e1ee01f747b168c429a121164ac962
>
> can you try to backport that change and see if that helps?
>
Yes, that's the one !!
I confirmed that the problem is solved by backporting this patch to both kernel
4.9.0 and 4.1.15.
Thanks a lot.
John
