RE: [PATCH V3 net-next 3/4] net: phy: realtek: add dt property to enable ALDPS mode

Hi Jisheng,

-----Original Message-----
From: Jisheng Zhang <redacted>
Sent: 2021年6月9日 9:57
To: Joakim Zhang <redacted>
Cc: davem@davemloft.net; kuba@kernel.org; robh+dt@kernel.org;
andrew@lunn.ch; hkallweit1@gmail.com; linux@armlinux.org.uk;
f.fainelli@gmail.com; dl-linux-imx [off-list ref];
netdev@vger.kernel.org; devicetree@vger.kernel.org;
linux-kernel@vger.kernel.org
Subject: Re: [PATCH V3 net-next 3/4] net: phy: realtek: add dt property to
enable ALDPS mode

On Tue, 8 Jun 2021 10:14:40 +0000
Joakim Zhang [off-list ref] wrote:

quoted

Hi Jisheng,

Hi,

quoted

quoted

-----Original Message-----
From: Jisheng Zhang <redacted>
Sent: 2021年6月8日 17:51
To: Joakim Zhang <redacted>
Cc: davem@davemloft.net; kuba@kernel.org; robh+dt@kernel.org;
andrew@lunn.ch; hkallweit1@gmail.com; linux@armlinux.org.uk;
f.fainelli@gmail.com; dl-linux-imx [off-list ref];
netdev@vger.kernel.org; devicetree@vger.kernel.org;
linux-kernel@vger.kernel.org
Subject: Re: [PATCH V3 net-next 3/4] net: phy: realtek: add dt
property to enable ALDPS mode

On Tue,  8 Jun 2021 11:15:34 +0800
Joakim Zhang [off-list ref] wrote:

quoted

If enable Advance Link Down Power Saving (ALDPS) mode, it will
change crystal/clock behavior, which cause RXC clock stop for
dozens to hundreds of miliseconds. This is comfirmed by Realtek
engineer. For some MACs, it needs RXC clock to support RX logic,
after this patch, PHY can generate continuous RXC clock during

auto-negotiation.

quoted

quoted

quoted

ALDPS default is disabled after hardware reset, it's more
reasonable to add a property to enable this feature, since ALDPS
would introduce side

effect.

quoted

This patch adds dt property "realtek,aldps-enable" to enable ALDPS
mode per users' requirement.

Jisheng Zhang enables this feature, changes the default behavior.
Since mine patch breaks the rule that new implementation should
not break existing design, so Cc'ed let him know to see if it can be

accepted.

quoted

quoted

quoted

Cc: Jisheng Zhang <redacted>
Signed-off-by: Joakim Zhang <redacted>
---
 drivers/net/phy/realtek.c | 20 +++++++++++++++++---
 1 file changed, 17 insertions(+), 3 deletions(-)

diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
index ca258f2a9613..79dc55bb4091 100644
--- a/drivers/net/phy/realtek.c
+++ b/drivers/net/phy/realtek.c

@@ -76,6 +76,7 @@ MODULE_AUTHOR("Johnson Leung");

MODULE_LICENSE("GPL");

 struct rtl821x_priv {
+       u16 phycr1;
        u16 phycr2;
 };

@@ -98,6 +99,14 @@ static int rtl821x_probe(struct phy_device

*phydev)

quoted

quoted

quoted

        if (!priv)
                return -ENOMEM;

+       priv->phycr1 = phy_read_paged(phydev, 0xa43,

RTL8211F_PHYCR1);

quoted

+       if (priv->phycr1 < 0)
+               return priv->phycr1;
+
+       priv->phycr1 &= (RTL8211F_ALDPS_PLL_OFF |
+ RTL8211F_ALDPS_ENABLE | RTL8211F_ALDPS_XTAL_OFF);

I believe your intention is

priv->phycr1 &= ~(RTL8211F_ALDPS_PLL_OFF | RTL8211F_ALDPS_ENABLE |
priv->RTL8211F_ALDPS_XTAL_OFF);
However, this is not necessary. See below.

No, mine intention is to read back this three bits what the register contained.

quoted

quoted

priv->phycr1 is 0 by default, so above 5 LoCs can be removed

The intention of this is to take bootloader into account. Such as uboot

configure the PHY before.

The last param "set" of phy_modify_paged_changed() means *bit mask of bits
to set* If we don't want to enable ALDPS, 0 is enough.

Even if uboot configured the PHY before linux, I believe
phy_modify_paged_changed() can clear ALDPS bits w/o above 5 LoCs.

The logic is:
1) read back these three bits from the register.
2) if linux set "realtek,aldps-enable", assert these three bit; if not, keep these three bits read before.
3) call phy_modify_paged_changed() to configure, "mask" parameter to clear these three bits first, "set" parameter to assert these three bits per the result of step 2.

So, if step 1 read back the value is that these three bits are asserted, then in step 3, it will first clear these three bits and assert these three bits again. The result is ALDPS is enabled even without " realtek,aldps-enable " in DT.

Best Regards,
Joakim Zhang

Thanks

quoted

Best Regards,
Joakim Zhang

quoted

quoted

+       if (of_property_read_bool(dev->of_node,

"realtek,aldps-enable"))

quoted

quoted

quoted

+               priv->phycr1 |= RTL8211F_ALDPS_PLL_OFF |
+ RTL8211F_ALDPS_ENABLE | RTL8211F_ALDPS_XTAL_OFF;
+
        priv->phycr2 = phy_read_paged(phydev, 0xa43,

RTL8211F_PHYCR2);

quoted

        if (priv->phycr2 < 0)
                return priv->phycr2; @@ -324,11 +333,16 @@ static
int rtl8211f_config_init(struct phy_device

*phydev)

quoted

        struct rtl821x_priv *priv = phydev->priv;
        struct device *dev = &phydev->mdio.dev;
        u16 val_txdly, val_rxdly;
-       u16 val;
        int ret;

-       val = RTL8211F_ALDPS_ENABLE | RTL8211F_ALDPS_PLL_OFF |

RTL8211F_ALDPS_XTAL_OFF;

quoted

-       phy_modify_paged_changed(phydev, 0xa43, RTL8211F_PHYCR1,

val,

quoted

quoted

val);

quoted

+       ret = phy_modify_paged_changed(phydev, 0xa43,

RTL8211F_PHYCR1,

quoted

+

RTL8211F_ALDPS_PLL_OFF |

quoted

quoted

RTL8211F_ALDPS_ENABLE | RTL8211F_ALDPS_XTAL_OFF,

quoted

+                                      priv->phycr1);
+       if (ret < 0) {
+               dev_err(dev, "aldps mode  configuration

failed: %pe\n",

quoted

quoted

quoted

+                       ERR_PTR(ret));
+               return ret;
+       }

        switch (phydev->interface) {
        case PHY_INTERFACE_MODE_RGMII:
--
2.17.1