[PATCH] net/phy: refactor RTL8211F initialization

[PATCH] net/phy: refactor RTL8211F initialization

[Shengzhou Liu]

RTL8211F needs to enalbe TXDLY for RGMII during
phy initialization, so move it to rtl8211f_config
for early initialization.

Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
cc: Joe Hershberger <joe.hershberger@gmail.com>
---
 drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
 1 file changed, 17 insertions(+), 8 deletions(-)

diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
index 3917c82..d48095b 100644
--- a/drivers/net/phy/realtek.c
+++ b/drivers/net/phy/realtek.c
@@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
 	return 0;
 }
 
+static int rtl8211f_config(struct phy_device *phydev)
+{
+	phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);
+
+	if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
+		/* enable TXDLY */
+		phy_write(phydev, MDIO_DEVAD_NONE,
+			  MIIM_RTL8211F_PAGE_SELECT, 0xd08);
+		phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
+	}
+
+	genphy_config_aneg(phydev);
+
+	return 0;
+}
+
 static int rtl8211x_parse_status(struct phy_device *phydev)
 {
 	unsigned int speed;
@@ -142,13 +158,6 @@ static int rtl8211f_parse_status(struct phy_device *phydev)
 		phydev->speed = SPEED_10;
 	}
 
-	if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
-		/* enable TXDLY */
-		phy_write(phydev, MDIO_DEVAD_NONE,
-			  MIIM_RTL8211F_PAGE_SELECT, 0xd08);
-		phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
-	}
-
 	return 0;
 }
 
@@ -209,7 +218,7 @@ static struct phy_driver RTL8211F_driver = {
 	.uid = 0x1cc916,
 	.mask = 0xffffff,
 	.features = PHY_GBIT_FEATURES,
-	.config = &rtl8211x_config,
+	.config = &rtl8211f_config,
 	.startup = &rtl8211f_startup,
 	.shutdown = &genphy_shutdown,
 };
-- 
2.1.0.27.g96db324

Re: [PATCH] net/phy: refactor RTL8211F initialization

[Florian Fainelli]

On 22/04/15 03:22, Shengzhou Liu wrote:
> RTL8211F needs to enalbe TXDLY for RGMII during
> phy initialization, so move it to rtl8211f_config
> for early initialization.
> 
> Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
> cc: Joe Hershberger <joe.hershberger@gmail.com>
> ---
>  drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
>  1 file changed, 17 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
> index 3917c82..d48095b 100644
> --- a/drivers/net/phy/realtek.c
> +++ b/drivers/net/phy/realtek.c
> @@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
>  	return 0;
>  }
>  
> +static int rtl8211f_config(struct phy_device *phydev)
> +{
> +	phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);

That part of the change is not documented but has to be, and you should
use genphy_soft_reset() which also takes care of waiting the required
amount of time before BMCR_RESET is cleared.

> +
> +	if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> +		/* enable TXDLY */
> +		phy_write(phydev, MDIO_DEVAD_NONE,
> +			  MIIM_RTL8211F_PAGE_SELECT, 0xd08);
> +		phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);

Do you need to reset the page selector somehow after this write?

> +	}
> +
> +	genphy_config_aneg(phydev);
> +
> +	return 0;
> +}
> +
>  static int rtl8211x_parse_status(struct phy_device *phydev)
>  {
>  	unsigned int speed;
> @@ -142,13 +158,6 @@ static int rtl8211f_parse_status(struct phy_device *phydev)
>  		phydev->speed = SPEED_10;
>  	}
>  
> -	if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> -		/* enable TXDLY */
> -		phy_write(phydev, MDIO_DEVAD_NONE,
> -			  MIIM_RTL8211F_PAGE_SELECT, 0xd08);
> -		phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
> -	}
> -
>  	return 0;
>  }
>  
> @@ -209,7 +218,7 @@ static struct phy_driver RTL8211F_driver = {
>  	.uid = 0x1cc916,
>  	.mask = 0xffffff,
>  	.features = PHY_GBIT_FEATURES,
> -	.config = &rtl8211x_config,
> +	.config = &rtl8211f_config,
>  	.startup = &rtl8211f_startup,
>  	.shutdown = &genphy_shutdown,
>  };
> 


-- 
Florian

RE: [PATCH] net/phy: refactor RTL8211F initialization

[Shengzhou.Liu]

> -----Original Message-----
> From: Florian Fainelli [mailto:f.fainelli@gmail.com]
> Sent: Thursday, April 23, 2015 12:57 AM
> To: Liu Shengzhou-B36685; netdev@vger.kernel.org; joe.hershberger@gmail.com
> Subject: Re: [PATCH] net/phy: refactor RTL8211F initialization
> 
> On 22/04/15 03:22, Shengzhou Liu wrote:
> > RTL8211F needs to enalbe TXDLY for RGMII during phy initialization, so
> > move it to rtl8211f_config for early initialization.
> >
> > Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
> > cc: Joe Hershberger <joe.hershberger@gmail.com>
> > ---
> >  drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
> >  1 file changed, 17 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
> > index 3917c82..d48095b 100644
> > --- a/drivers/net/phy/realtek.c
> > +++ b/drivers/net/phy/realtek.c
> > @@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
> >  	return 0;
> >  }
> >
> > +static int rtl8211f_config(struct phy_device *phydev) {
> > +	phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);
> 
> That part of the change is not documented but has to be, and you should use
> genphy_soft_reset() which also takes care of waiting the required amount of
> time before BMCR_RESET is cleared.

There is no genphy_soft_reset() in current u-boot tree.  

> > +
> > +	if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> > +		/* enable TXDLY */
> > +		phy_write(phydev, MDIO_DEVAD_NONE,
> > +			  MIIM_RTL8211F_PAGE_SELECT, 0xd08);
> > +		phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
> 
> Do you need to reset the page selector somehow after this write?

Actually no need to reset the page selector.

> > +	}
> > +
> > +	genphy_config_aneg(phydev);
> > +
> > +	return 0;
> > +}
> > +
> >  static int rtl8211x_parse_status(struct phy_device *phydev)  {
> >  	unsigned int speed;
> > @@ -142,13 +158,6 @@ static int rtl8211f_parse_status(struct phy_device
> *phydev)
> >  		phydev->speed = SPEED_10;
> >  	}
> >
> > -	if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> > -		/* enable TXDLY */
> > -		phy_write(phydev, MDIO_DEVAD_NONE,
> > -			  MIIM_RTL8211F_PAGE_SELECT, 0xd08);
> > -		phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
> > -	}
> > -
> >  	return 0;
> >  }
> >
> > @@ -209,7 +218,7 @@ static struct phy_driver RTL8211F_driver = {
> >  	.uid = 0x1cc916,
> >  	.mask = 0xffffff,
> >  	.features = PHY_GBIT_FEATURES,
> > -	.config = &rtl8211x_config,
> > +	.config = &rtl8211f_config,
> >  	.startup = &rtl8211f_startup,
> >  	.shutdown = &genphy_shutdown,
> >  };
> >
> 
> 
> --
> Florian

Re: [PATCH] net/phy: refactor RTL8211F initialization

[Florian Fainelli]

Le 22/04/2015 20:53, Shengzhou.Liu@freescale.com a écrit :
>> -----Original Message-----
>> From: Florian Fainelli [mailto:f.fainelli@gmail.com]
>> Sent: Thursday, April 23, 2015 12:57 AM
>> To: Liu Shengzhou-B36685; netdev@vger.kernel.org; joe.hershberger@gmail.com
>> Subject: Re: [PATCH] net/phy: refactor RTL8211F initialization
>>
>> On 22/04/15 03:22, Shengzhou Liu wrote:
>>> RTL8211F needs to enalbe TXDLY for RGMII during phy initialization, so
>>> move it to rtl8211f_config for early initialization.
>>>
>>> Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
>>> cc: Joe Hershberger <joe.hershberger@gmail.com>
>>> ---
>>>  drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
>>>  1 file changed, 17 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
>>> index 3917c82..d48095b 100644
>>> --- a/drivers/net/phy/realtek.c
>>> +++ b/drivers/net/phy/realtek.c
>>> @@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
>>>  	return 0;
>>>  }
>>>
>>> +static int rtl8211f_config(struct phy_device *phydev) {
>>> +	phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);
>>
>> That part of the change is not documented but has to be, and you should use
>> genphy_soft_reset() which also takes care of waiting the required amount of
>> time before BMCR_RESET is cleared.
> 
> There is no genphy_soft_reset() in current u-boot tree. 

Did you mean to submit that against u-boot or Linux? If the latter,
there is absolutely no need to have the same file compile under u-boot
and Linux without changes, that is too restrictive. There is a
genphy_soft_reset() which takes care of waiting for BMCR_RESET to clear,
please use it, there is no guarantee otherwise that a PHY has completed
a reset.
--
Florian

RE: [PATCH] net/phy: refactor RTL8211F initialization

[Shengzhou.Liu]

> -----Original Message-----
> From: Florian Fainelli [mailto:f.fainelli@gmail.com]
> Sent: Thursday, April 23, 2015 12:39 PM
> To: Liu Shengzhou-B36685; netdev@vger.kernel.org; joe.hershberger@gmail.com
> Subject: Re: [PATCH] net/phy: refactor RTL8211F initialization
> 
> Le 22/04/2015 20:53, Shengzhou.Liu@freescale.com a écrit :
> >> -----Original Message-----
> >> From: Florian Fainelli [mailto:f.fainelli@gmail.com]
> >> Sent: Thursday, April 23, 2015 12:57 AM
> >> To: Liu Shengzhou-B36685; netdev@vger.kernel.org;
> >> joe.hershberger@gmail.com
> >> Subject: Re: [PATCH] net/phy: refactor RTL8211F initialization
> >>
> >> On 22/04/15 03:22, Shengzhou Liu wrote:
> >>> RTL8211F needs to enalbe TXDLY for RGMII during phy initialization,
> >>> so move it to rtl8211f_config for early initialization.
> >>>
> >>> Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
> >>> cc: Joe Hershberger <joe.hershberger@gmail.com>
> >>> ---
> >>>  drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
> >>>  1 file changed, 17 insertions(+), 8 deletions(-)
> >>>
> >>> diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
> >>> index 3917c82..d48095b 100644
> >>> --- a/drivers/net/phy/realtek.c
> >>> +++ b/drivers/net/phy/realtek.c
> >>> @@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
> >>>  	return 0;
> >>>  }
> >>>
> >>> +static int rtl8211f_config(struct phy_device *phydev) {
> >>> +	phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);
> >>
> >> That part of the change is not documented but has to be, and you
> >> should use
> >> genphy_soft_reset() which also takes care of waiting the required
> >> amount of time before BMCR_RESET is cleared.
> >
> > There is no genphy_soft_reset() in current u-boot tree.
> 
> Did you mean to submit that against u-boot or Linux? If the latter, there is
> absolutely no need to have the same file compile under u-boot and Linux
> without changes, that is too restrictive. There is a
> genphy_soft_reset() which takes care of waiting for BMCR_RESET to clear,
> please use it, there is no guarantee otherwise that a PHY has completed a
> reset.
> --
> Florian
This patch is for u-boot, not for kernel. There is no genphy_soft_reset() in u-boot.

[U-Boot] [PATCH] net/phy: refactor RTL8211F initialization

[Joe Hershberger]

On Thu, Apr 23, 2015 at 12:58 AM, Shengzhou.Liu at freescale.com
<Shengzhou.Liu@freescale.com> wrote:
>> -----Original Message-----
>> From: Florian Fainelli [mailto:f.fainelli at gmail.com]
>> Sent: Thursday, April 23, 2015 12:39 PM
>> To: Liu Shengzhou-B36685; netdev at vger.kernel.org; joe.hershberger at gmail.com
>> Subject: Re: [PATCH] net/phy: refactor RTL8211F initialization
>>
>> Did you mean to submit that against u-boot or Linux? If the latter, there is
>> absolutely no need to have the same file compile under u-boot and Linux
>> without changes, that is too restrictive. There is a
>> genphy_soft_reset() which takes care of waiting for BMCR_RESET to clear,
>> please use it, there is no guarantee otherwise that a PHY has completed a
>> reset.
>> --
>> Florian
> This patch is for u-boot, not for kernel. There is no genphy_soft_reset() in u-boot.

You sent this patch to the netdev list and not the u-boot list.  I've
added the u-boot list here, but you should send v2 to the u-boot list.

Thanks,
-Joe

[U-Boot] [PATCH] net/phy: refactor RTL8211F initialization

[Joe Hershberger]

Hi Shengzhou Liu,

On Wed, Apr 22, 2015 at 5:22 AM, Shengzhou Liu
<Shengzhou.Liu@freescale.com> wrote:
> RTL8211F needs to enalbe TXDLY for RGMII during
> phy initialization, so move it to rtl8211f_config
> for early initialization.
>
> Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
> cc: Joe Hershberger <joe.hershberger@gmail.com>
> ---
>  drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
>  1 file changed, 17 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
> index 3917c82..d48095b 100644
> --- a/drivers/net/phy/realtek.c
> +++ b/drivers/net/phy/realtek.c
> @@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
>         return 0;
>  }
>
> +static int rtl8211f_config(struct phy_device *phydev)
> +{
> +       phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);

Do you not need to disable the phy interrupt here?

> +
> +       if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> +               /* enable TXDLY */
> +               phy_write(phydev, MDIO_DEVAD_NONE,
> +                         MIIM_RTL8211F_PAGE_SELECT, 0xd08);

Why do you not need to change the page back to default? Does it only
apply to one following command or something? I haven't read the data
sheet for this phy to understand its behavior, but want to make sure
it's clear here.  Please at least add a comment describing why the
page need not be changed back.

> +               phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);

Is this TX delay board specific? Seems like it would be. Should it be
parameterized to come from a board CONFIG_? If not, at least add a
comment describing these magic numbers.

> +       }
> +
> +       genphy_config_aneg(phydev);
> +
> +       return 0;
> +}
> +
>  static int rtl8211x_parse_status(struct phy_device *phydev)
>  {
>         unsigned int speed;
> @@ -142,13 +158,6 @@ static int rtl8211f_parse_status(struct phy_device *phydev)
>                 phydev->speed = SPEED_10;
>         }
>
> -       if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> -               /* enable TXDLY */
> -               phy_write(phydev, MDIO_DEVAD_NONE,
> -                         MIIM_RTL8211F_PAGE_SELECT, 0xd08);
> -               phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
> -       }
> -
>         return 0;
>  }
>
> @@ -209,7 +218,7 @@ static struct phy_driver RTL8211F_driver = {
>         .uid = 0x1cc916,
>         .mask = 0xffffff,
>         .features = PHY_GBIT_FEATURES,
> -       .config = &rtl8211x_config,
> +       .config = &rtl8211f_config,
>         .startup = &rtl8211f_startup,
>         .shutdown = &genphy_shutdown,
>  };
> --
> 2.1.0.27.g96db324
>

[U-Boot] [PATCH] net/phy: refactor RTL8211F initialization

[Shengzhou.Liu at freescale.com]

> -----Original Message-----
> From: Joe Hershberger [mailto:joe.hershberger at gmail.com]
> Sent: Thursday, April 23, 2015 10:42 PM
> To: Liu Shengzhou-B36685
> Cc: u-boot
> Subject: Re: [PATCH] net/phy: refactor RTL8211F initialization
> 
> Hi Shengzhou Liu,
> 
> On Wed, Apr 22, 2015 at 5:22 AM, Shengzhou Liu <Shengzhou.Liu@freescale.com>
> wrote:
> > RTL8211F needs to enalbe TXDLY for RGMII during phy initialization, so
> > move it to rtl8211f_config for early initialization.
> >
> > Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
> > cc: Joe Hershberger <joe.hershberger@gmail.com>
> > ---
> >  drivers/net/phy/realtek.c | 25 +++++++++++++++++--------
> >  1 file changed, 17 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
> > index 3917c82..d48095b 100644
> > --- a/drivers/net/phy/realtek.c
> > +++ b/drivers/net/phy/realtek.c
> > @@ -43,6 +43,22 @@ static int rtl8211x_config(struct phy_device *phydev)
> >         return 0;
> >  }
> >
> > +static int rtl8211f_config(struct phy_device *phydev) {
> > +       phy_write(phydev, MDIO_DEVAD_NONE, MII_BMCR, BMCR_RESET);
> 
> Do you not need to disable the phy interrupt here?
No need, It's disabled by default.

> > +
> > +       if (phydev->interface == PHY_INTERFACE_MODE_RGMII) {
> > +               /* enable TXDLY */
> > +               phy_write(phydev, MDIO_DEVAD_NONE,
> > +                         MIIM_RTL8211F_PAGE_SELECT, 0xd08);
> 
> Why do you not need to change the page back to default? Does it only apply to
> one following command or something? I haven't read the data sheet for this
> phy to understand its behavior, but want to make sure it's clear here.
> Please at least add a comment describing why the page need not be changed
> back.
There is no other command, so it's working without back to default. 
To avoid potential problem if one not set expected page, will have the page back to default 0 in v2.
 
> > +               phy_write(phydev, MDIO_DEVAD_NONE, 0x11, 0x109);
> 
> Is this TX delay board specific? Seems like it would be. Should it be
> parameterized to come from a board CONFIG_? If not, at least add a comment
> describing these magic numbers.

It is not board specific. Will replace the magic number.