[PATCH] net: phy: shift phy_init_hw after device_add during phy device registration

[PATCH] net: phy: shift phy_init_hw after device_add during phy device registration

[Grygorii Strashko]

Now the call of phy_init_hw() in phy_device_register() is converted to
be just a NOP, because phy_init_hw() is called before PHY devices probing
and phydev->drv == NULL at this moment.

phy_device_register()
 |- phy_init_hw() -> NOP
 |- device_add()
    |- phy_probe() -> may be if driver is ready
       |- phydev->drv = <&phy_driver>

Hence, shift call of phy_init_hw() after device_add() and in such way
increase possibility of its execution.

This patch is RFC because:
- this patch will change PHY's callbacks call order: .config_init() will
be called after .probe()
- i'm not sure that phy_init_hw() need to be called from phy_device_register() and
maybe it can be moved in other place, like phy_probe()

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
---
 drivers/net/phy/phy_device.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
index 2f6989b..496940be 100644
--- a/drivers/net/phy/phy_device.c
+++ b/drivers/net/phy/phy_device.c
@@ -352,16 +352,16 @@ int phy_device_register(struct phy_device *phydev)
 		return -EINVAL;
 	phydev->bus->phy_map[phydev->addr] = phydev;
 
-	/* Run all of the fixups for this PHY */
-	err = phy_init_hw(phydev);
+	err = device_add(&phydev->dev);
 	if (err) {
-		pr_err("PHY %d failed to initialize\n", phydev->addr);
+		pr_err("PHY %d failed to add\n", phydev->addr);
 		goto out;
 	}
 
-	err = device_add(&phydev->dev);
+	/* Run all of the fixups for this PHY */
+	err = phy_init_hw(phydev);
 	if (err) {
-		pr_err("PHY %d failed to add\n", phydev->addr);
+		pr_err("PHY %d failed to initialize\n", phydev->addr);
 		goto out;
 	}
 
-- 
1.7.9.5

Re: [PATCH] net: phy: shift phy_init_hw after device_add during phy device registration

[Grygorii Strashko]

Sorry, this is  [RFC PATCH].

On 04/04/2014 04:38 PM, Grygorii Strashko wrote:
> Now the call of phy_init_hw() in phy_device_register() is converted to
> be just a NOP, because phy_init_hw() is called before PHY devices probing
> and phydev->drv == NULL at this moment.
>
> phy_device_register()
>   |- phy_init_hw() -> NOP
>   |- device_add()
>      |- phy_probe() -> may be if driver is ready
>         |- phydev->drv = <&phy_driver>
>
> Hence, shift call of phy_init_hw() after device_add() and in such way
> increase possibility of its execution.
>
> This patch is RFC because:
> - this patch will change PHY's callbacks call order: .config_init() will
> be called after .probe()
> - i'm not sure that phy_init_hw() need to be called from phy_device_register() and
> maybe it can be moved in other place, like phy_probe()
>
> Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
> ---
>   drivers/net/phy/phy_device.c |   10 +++++-----
>   1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
> index 2f6989b..496940be 100644
> --- a/drivers/net/phy/phy_device.c
> +++ b/drivers/net/phy/phy_device.c
> @@ -352,16 +352,16 @@ int phy_device_register(struct phy_device *phydev)
>   		return -EINVAL;
>   	phydev->bus->phy_map[phydev->addr] = phydev;
>
> -	/* Run all of the fixups for this PHY */
> -	err = phy_init_hw(phydev);
> +	err = device_add(&phydev->dev);
>   	if (err) {
> -		pr_err("PHY %d failed to initialize\n", phydev->addr);
> +		pr_err("PHY %d failed to add\n", phydev->addr);
>   		goto out;
>   	}
>
> -	err = device_add(&phydev->dev);
> +	/* Run all of the fixups for this PHY */
> +	err = phy_init_hw(phydev);
>   	if (err) {
> -		pr_err("PHY %d failed to add\n", phydev->addr);
> +		pr_err("PHY %d failed to initialize\n", phydev->addr);
>   		goto out;
>   	}
>
>