diff mbox

ARM: dts: add phy-reset property for rk3066a-rayeager emac

Message ID 1510022959-19440-1-git-send-email-zyw@rock-chips.com
State New, archived
Headers show

Commit Message

Chris Zhong Nov. 7, 2017, 2:49 a.m. UTC
The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
GPIO1_D6, this pin should be pull down then pull up to reset the phy.
Add a phy-reset property in emac, make the phy can be reset when emac
power on.

Signed-off-by: Chris Zhong <zyw@rock-chips.com>
---

 arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
 1 file changed, 2 insertions(+)

Comments

Vladimir Zapolskiy Nov. 7, 2017, 7:54 a.m. UTC | #1
Hello Chris,

On 11/07/2017 04:49 AM, Chris Zhong wrote:
> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
> Add a phy-reset property in emac, make the phy can be reset when emac
> power on.

for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
please reference to Documentation/devicetree/bindings/net/mdio.txt

Can you try to reuse them instead of adding new custom properties?

As a side question, which is mainly addressed to Sergei and Roger,
I don't quite understand why PHY properties were initially added to
MAC/MDIO bus device tree nodes, in my opinion they must be moved under
PHY device tree nodes.

--
With best wishes,
Vladimir

> 
> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
> ---
> 
>  arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts b/arch/arm/boot/dts/rk3066a-rayeager.dts
> index 570157f..6064a0a 100644
> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
> @@ -173,6 +173,8 @@
>  	pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>  	phy = <&phy0>;
>  	phy-supply = <&vcc_rmii>;
> +	phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
> +	phy-reset-duration = <10>; /* millisecond */
>  	status = "okay";
>  
>  	phy0: ethernet-phy@0 {
>
Chris Zhong Nov. 7, 2017, 9:51 a.m. UTC | #2
On 2017年11月07日 15:54, Vladimir Zapolskiy wrote:
> Hello Chris,
>
> On 11/07/2017 04:49 AM, Chris Zhong wrote:
>> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
>> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
>> Add a phy-reset property in emac, make the phy can be reset when emac
>> power on.
> for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
> please reference to Documentation/devicetree/bindings/net/mdio.txt
>
> Can you try to reuse them instead of adding new custom properties?
This phy-reset is from Documentation/devicetree/bindings/net/arc_emac.txt.
And copy from arch/arm/boot/dts/rk3036-kylin.dts.
Can we just use these properties, they are not new.
> As a side question, which is mainly addressed to Sergei and Roger,
> I don't quite understand why PHY properties were initially added to
> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
> PHY device tree nodes.
>
> --
> With best wishes,
> Vladimir
>
>> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>> ---
>>
>>   arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts b/arch/arm/boot/dts/rk3066a-rayeager.dts
>> index 570157f..6064a0a 100644
>> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
>> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
>> @@ -173,6 +173,8 @@
>>   	pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>>   	phy = <&phy0>;
>>   	phy-supply = <&vcc_rmii>;
>> +	phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
>> +	phy-reset-duration = <10>; /* millisecond */
>>   	status = "okay";
>>   
>>   	phy0: ethernet-phy@0 {
>>
>
>
Andrew Lunn Nov. 7, 2017, 3:59 p.m. UTC | #3
> As a side question, which is mainly addressed to Sergei and Roger,
> I don't quite understand why PHY properties were initially added to
> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
> PHY device tree nodes.

If you look at the implementation, they are not actually PHY
properties. They are MDIO properties. It is the MDIO driver using them
to perform an 'MDIO bus reset'. The property names are wrong.

The device tree structure could be better. It mixes the MAC and the
MDIO properties in the same node. It would of been better to have
separate nodes. 

The MDIO core now has equivalent functionality. So yes, it would be
better to use the generic MDIO properties, not these driver
properties.

	 Andrew
Florian Fainelli Nov. 7, 2017, 6:26 p.m. UTC | #4
On 11/07/2017 01:51 AM, Chris Zhong wrote:
> 
> 
> On 2017年11月07日 15:54, Vladimir Zapolskiy wrote:
>> Hello Chris,
>>
>> On 11/07/2017 04:49 AM, Chris Zhong wrote:
>>> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
>>> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
>>> Add a phy-reset property in emac, make the phy can be reset when emac
>>> power on.
>> for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
>> please reference to Documentation/devicetree/bindings/net/mdio.txt
>>
>> Can you try to reuse them instead of adding new custom properties?
> This phy-reset is from Documentation/devicetree/bindings/net/arc_emac.txt.
> And copy from arch/arm/boot/dts/rk3036-kylin.dts.
> Can we just use these properties, they are not new.

Because it already exists does not mean it's correct, in fact, it is not
at all because it places the reset property for a PHY into the MAC node,
which is just not what this is, what we should be using instead is the
following patch series:

http://patchwork.ozlabs.org/project/netdev/list/?series=9267

http://patchwork.ozlabs.org/patch/828499/
http://patchwork.ozlabs.org/patch/828505/
http://patchwork.ozlabs.org/patch/828501/
http://patchwork.ozlabs.org/patch/828502/

>> As a side question, which is mainly addressed to Sergei and Roger,
>> I don't quite understand why PHY properties were initially added to
>> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
>> PHY device tree nodes.
>>
>> -- 
>> With best wishes,
>> Vladimir
>>
>>> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>>> ---
>>>
>>>   arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>>>   1 file changed, 2 insertions(+)
>>>
>>> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>> b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>> index 570157f..6064a0a 100644
>>> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>> @@ -173,6 +173,8 @@
>>>       pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>>>       phy = <&phy0>;
>>>       phy-supply = <&vcc_rmii>;
>>> +    phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
>>> +    phy-reset-duration = <10>; /* millisecond */
>>>       status = "okay";
>>>         phy0: ethernet-phy@0 {
>>>
>>
>>
>
Chris Zhong Nov. 8, 2017, 9:49 a.m. UTC | #5
Hi Florian Fainelli


On 2017年11月08日 02:26, Florian Fainelli wrote:
> On 11/07/2017 01:51 AM, Chris Zhong wrote:
>>
>> On 2017年11月07日 15:54, Vladimir Zapolskiy wrote:
>>> Hello Chris,
>>>
>>> On 11/07/2017 04:49 AM, Chris Zhong wrote:
>>>> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
>>>> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
>>>> Add a phy-reset property in emac, make the phy can be reset when emac
>>>> power on.
>>> for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
>>> please reference to Documentation/devicetree/bindings/net/mdio.txt
>>>
>>> Can you try to reuse them instead of adding new custom properties?
>> This phy-reset is from Documentation/devicetree/bindings/net/arc_emac.txt.
>> And copy from arch/arm/boot/dts/rk3036-kylin.dts.
>> Can we just use these properties, they are not new.
> Because it already exists does not mean it's correct, in fact, it is not
> at all because it places the reset property for a PHY into the MAC node,
> which is just not what this is, what we should be using instead is the
> following patch series:
>
> http://patchwork.ozlabs.org/project/netdev/list/?series=9267
>
> http://patchwork.ozlabs.org/patch/828499/
> http://patchwork.ozlabs.org/patch/828505/
> http://patchwork.ozlabs.org/patch/828501/
> http://patchwork.ozlabs.org/patch/828502/
Okay, this series works for me, and I will push a new version patch 
following it,
with a reset-gpios property under phy node.
And hope this series can be applied soon.

Thanks
Chris Zhong


>>> As a side question, which is mainly addressed to Sergei and Roger,
>>> I don't quite understand why PHY properties were initially added to
>>> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
>>> PHY device tree nodes.
>>>
>>> -- 
>>> With best wishes,
>>> Vladimir
>>>
>>>> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>>>> ---
>>>>
>>>>    arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>>>>    1 file changed, 2 insertions(+)
>>>>
>>>> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>> b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>> index 570157f..6064a0a 100644
>>>> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>> @@ -173,6 +173,8 @@
>>>>        pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>>>>        phy = <&phy0>;
>>>>        phy-supply = <&vcc_rmii>;
>>>> +    phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
>>>> +    phy-reset-duration = <10>; /* millisecond */
>>>>        status = "okay";
>>>>          phy0: ethernet-phy@0 {
>>>>
>>>
>
Roger Quadros Nov. 9, 2017, 8:20 a.m. UTC | #6
Hi,

On 07/11/17 09:54, Vladimir Zapolskiy wrote:
> Hello Chris,
> 
> On 11/07/2017 04:49 AM, Chris Zhong wrote:
>> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
>> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
>> Add a phy-reset property in emac, make the phy can be reset when emac
>> power on.
> 
> for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
> please reference to Documentation/devicetree/bindings/net/mdio.txt
> 
> Can you try to reuse them instead of adding new custom properties?
> 
> As a side question, which is mainly addressed to Sergei and Roger,
> I don't quite understand why PHY properties were initially added to
> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
> PHY device tree nodes.

The PHY reset *has* to be performed at the MDIO bus driver level so that they
can be pin-strap configured correctly and be detected during MDIO bus scan.

- We have boards where PHYs won't be detected if RESET is not done before the scan. (due to invalid h/w pin-strapping config)
- We have boards where a single GPIO line controls reset of all PHYs on the bus.

Due to these reasons the RESET control lies with the MDIO bus driver.

> 
> --
> With best wishes,
> Vladimir
> 
>>
>> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>> ---
>>
>>  arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>>  1 file changed, 2 insertions(+)
>>
>> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts b/arch/arm/boot/dts/rk3066a-rayeager.dts
>> index 570157f..6064a0a 100644
>> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
>> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
>> @@ -173,6 +173,8 @@
>>  	pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>>  	phy = <&phy0>;
>>  	phy-supply = <&vcc_rmii>;
>> +	phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
>> +	phy-reset-duration = <10>; /* millisecond */
>>  	status = "okay";
>>  
>>  	phy0: ethernet-phy@0 {
>>
Roger Quadros Nov. 9, 2017, 8:28 a.m. UTC | #7
On 08/11/17 11:49, Chris Zhong wrote:
> Hi Florian Fainelli
> 
> 
> On 2017年11月08日 02:26, Florian Fainelli wrote:
>> On 11/07/2017 01:51 AM, Chris Zhong wrote:
>>>
>>> On 2017年11月07日 15:54, Vladimir Zapolskiy wrote:
>>>> Hello Chris,
>>>>
>>>> On 11/07/2017 04:49 AM, Chris Zhong wrote:
>>>>> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
>>>>> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
>>>>> Add a phy-reset property in emac, make the phy can be reset when emac
>>>>> power on.
>>>> for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
>>>> please reference to Documentation/devicetree/bindings/net/mdio.txt
>>>>
>>>> Can you try to reuse them instead of adding new custom properties?
>>> This phy-reset is from Documentation/devicetree/bindings/net/arc_emac.txt.
>>> And copy from arch/arm/boot/dts/rk3036-kylin.dts.
>>> Can we just use these properties, they are not new.
>> Because it already exists does not mean it's correct, in fact, it is not
>> at all because it places the reset property for a PHY into the MAC node,
>> which is just not what this is, what we should be using instead is the
>> following patch series:
>>
>> http://patchwork.ozlabs.org/project/netdev/list/?series=9267
>>
>> http://patchwork.ozlabs.org/patch/828499/
>> http://patchwork.ozlabs.org/patch/828505/
>> http://patchwork.ozlabs.org/patch/828501/
>> http://patchwork.ozlabs.org/patch/828502/
> Okay, this series works for me, and I will push a new version patch following it,
> with a reset-gpios property under phy node.
> And hope this series can be applied soon.
> 

If 'reset-gpios' and 'reset-delay-us' as per [1] works for you
then why not use that?

[1] Documentation/devicetree/bindings/net/mdio.txt

> Thanks
> Chris Zhong
> 
> 
>>>> As a side question, which is mainly addressed to Sergei and Roger,
>>>> I don't quite understand why PHY properties were initially added to
>>>> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
>>>> PHY device tree nodes.
>>>>
>>>> -- 
>>>> With best wishes,
>>>> Vladimir
>>>>
>>>>> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>>>>> ---
>>>>>
>>>>>    arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>>>>>    1 file changed, 2 insertions(+)
>>>>>
>>>>> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>> b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>> index 570157f..6064a0a 100644
>>>>> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>> @@ -173,6 +173,8 @@
>>>>>        pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>>>>>        phy = <&phy0>;
>>>>>        phy-supply = <&vcc_rmii>;
>>>>> +    phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
>>>>> +    phy-reset-duration = <10>; /* millisecond */
>>>>>        status = "okay";
>>>>>          phy0: ethernet-phy@0 {
>>>>>
>>>>
>>
>
Roger Quadros Nov. 9, 2017, 8:40 a.m. UTC | #8
On 09/11/17 10:28, Roger Quadros wrote:
> On 08/11/17 11:49, Chris Zhong wrote:
>> Hi Florian Fainelli
>>
>>
>> On 2017年11月08日 02:26, Florian Fainelli wrote:
>>> On 11/07/2017 01:51 AM, Chris Zhong wrote:
>>>>
>>>> On 2017年11月07日 15:54, Vladimir Zapolskiy wrote:
>>>>> Hello Chris,
>>>>>
>>>>> On 11/07/2017 04:49 AM, Chris Zhong wrote:
>>>>>> The ethernet phy of rk3066a-rayeager has a reset pin, it controlled by
>>>>>> GPIO1_D6, this pin should be pull down then pull up to reset the phy.
>>>>>> Add a phy-reset property in emac, make the phy can be reset when emac
>>>>>> power on.
>>>>> for PHY reset there are properties 'reset-gpios' and 'reset-delay-us',
>>>>> please reference to Documentation/devicetree/bindings/net/mdio.txt
>>>>>
>>>>> Can you try to reuse them instead of adding new custom properties?
>>>> This phy-reset is from Documentation/devicetree/bindings/net/arc_emac.txt.
>>>> And copy from arch/arm/boot/dts/rk3036-kylin.dts.
>>>> Can we just use these properties, they are not new.
>>> Because it already exists does not mean it's correct, in fact, it is not
>>> at all because it places the reset property for a PHY into the MAC node,
>>> which is just not what this is, what we should be using instead is the
>>> following patch series:
>>>
>>> http://patchwork.ozlabs.org/project/netdev/list/?series=9267
>>>
>>> http://patchwork.ozlabs.org/patch/828499/
>>> http://patchwork.ozlabs.org/patch/828505/
>>> http://patchwork.ozlabs.org/patch/828501/
>>> http://patchwork.ozlabs.org/patch/828502/
>> Okay, this series works for me, and I will push a new version patch following it,
>> with a reset-gpios property under phy node.
>> And hope this series can be applied soon.
>>
> 
> If 'reset-gpios' and 'reset-delay-us' as per [1] works for you
> then why not use that?
> 
> [1] Documentation/devicetree/bindings/net/mdio.txt

I take back my statement.

You should use the above only if you have a single reset line connected to multiple PHYs
on the same MDIO bus.
In all other cases the reset properties should be part of PHY device node.

However, we need to still ensure that the mdio_bus driver issues the reset. This piece
is still missing in the code.

> 
>> Thanks
>> Chris Zhong
>>
>>
>>>>> As a side question, which is mainly addressed to Sergei and Roger,
>>>>> I don't quite understand why PHY properties were initially added to
>>>>> MAC/MDIO bus device tree nodes, in my opinion they must be moved under
>>>>> PHY device tree nodes.
>>>>>
>>>>> -- 
>>>>> With best wishes,
>>>>> Vladimir
>>>>>
>>>>>> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>>>>>> ---
>>>>>>
>>>>>>    arch/arm/boot/dts/rk3066a-rayeager.dts | 2 ++
>>>>>>    1 file changed, 2 insertions(+)
>>>>>>
>>>>>> diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>>> b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>>> index 570157f..6064a0a 100644
>>>>>> --- a/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>>> +++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
>>>>>> @@ -173,6 +173,8 @@
>>>>>>        pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
>>>>>>        phy = <&phy0>;
>>>>>>        phy-supply = <&vcc_rmii>;
>>>>>> +    phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
>>>>>> +    phy-reset-duration = <10>; /* millisecond */
>>>>>>        status = "okay";
>>>>>>          phy0: ethernet-phy@0 {
>>>>>>
>>>>>
>>>
>>
>
diff mbox

Patch

diff --git a/arch/arm/boot/dts/rk3066a-rayeager.dts b/arch/arm/boot/dts/rk3066a-rayeager.dts
index 570157f..6064a0a 100644
--- a/arch/arm/boot/dts/rk3066a-rayeager.dts
+++ b/arch/arm/boot/dts/rk3066a-rayeager.dts
@@ -173,6 +173,8 @@ 
 	pinctrl-0 = <&emac_xfer>, <&emac_mdio>, <&rmii_rst>;
 	phy = <&phy0>;
 	phy-supply = <&vcc_rmii>;
+	phy-reset-gpios = <&gpio1 RK_PD6 GPIO_ACTIVE_LOW>; /* PHY_RST */
+	phy-reset-duration = <10>; /* millisecond */
 	status = "okay";
 
 	phy0: ethernet-phy@0 {