[PATCH v2 1/2] phy: mv88e61xx: add support for RGMII TX/RX delay
Chris Packham
judge.packham at gmail.com
Mon Jan 4 07:11:15 CET 2021
(Note: viewing on mobile device so my comment below could be wrong)
On Mon, 4 Jan 2021, 9:35 AM Pawel Dembicki, <paweldembicki at gmail.com> wrote:
> Clock delay in RGMII is required for some boards.
>
> Clock delay is read from phy-mode dts property. Delay is configured via
> proper bits in PORT_REG_PHYS_CTRL register.
>
> Cc: Chris Packham <judge.packham at gmail.com>
> Cc: Joe Hershberger <joe.hershberger at ni.com>
> Cc: Anatolij Gustschin <agust at denx.de>
> Cc: Tim Harvey <tharvey at gateworks.com>
> Cc: Tom Rini <trini at konsulko.com>
> Signed-off-by: Pawel Dembicki <paweldembicki at gmail.com>
> ---
> Changes in v2:
> - change source info about delay: from hardcode to phy-mode propoperty in
> dts
>
> drivers/net/phy/mv88e61xx.c | 62 ++++++++++++++++++++++++++++++++++++-
> 1 file changed, 61 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/phy/mv88e61xx.c b/drivers/net/phy/mv88e61xx.c
> index 7eff37b244..70adec6578 100644
> --- a/drivers/net/phy/mv88e61xx.c
> +++ b/drivers/net/phy/mv88e61xx.c
> @@ -97,6 +97,8 @@
> #define PORT_REG_STATUS_CMODE_1000BASE_X 0x9
> #define PORT_REG_STATUS_CMODE_SGMII 0xa
>
> +#define PORT_REG_PHYS_CTRL_RGMII_RX_DELAY BIT(15)
> +#define PORT_REG_PHYS_CTRL_RGMII_TX_DELAY BIT(14)
> #define PORT_REG_PHYS_CTRL_PCS_AN_EN BIT(10)
> #define PORT_REG_PHYS_CTRL_PCS_AN_RST BIT(9)
> #define PORT_REG_PHYS_CTRL_FC_VALUE BIT(7)
> @@ -729,7 +731,45 @@ unforce:
> static int mv88e61xx_fixed_port_setup(struct phy_device *phydev, u8 port)
> {
> struct mv88e61xx_phy_priv *priv = phydev->priv;
> - int val;
> + ofnode node;
> + const char *str;
> + int val, phy_mode;
> + u32 phy_handle;
> +
> + if (!ofnode_valid(phydev->node)) {
> + node = dev_ofnode(phydev->dev);
> + phy_handle = ofnode_read_u32_default(node, "phy-handle",
> -ENXIO);
> +
> + if (phy_handle == -ENXIO) {
> + node = ofnode_first_subnode(node);
> +
> + while (ofnode_valid(node)) {
> + phy_handle = ofnode_read_u32_default(node,
> "phy-handle", -ENXIO);
> + if (phy_handle != -ENXIO)
> + break;
> + node = ofnode_next_subnode(node);
> + }
> + }
> +
> + if (phy_handle != -ENXIO)
> + node = ofnode_get_by_phandle(phy_handle);
> + else
> + node = ofnode_null();
> + } else {
> + node = phy_get_ofnode(phydev);
> + }
> +
> + node = ofnode_find_subnode(node, "ports");
> + node = ofnode_first_subnode(node);
> +
> + while (ofnode_valid(node)) {
> + u8 port_no = (u8)ofnode_read_u32_default(node, "reg", -1);
> +
> + if (port_no == port)
> + break;
> +
> + node = ofnode_next_subnode(node);
> + }
>
> val = mv88e61xx_port_read(phydev, port, PORT_REG_PHYS_CTRL);
> if (val < 0)
> @@ -754,6 +794,26 @@ static int mv88e61xx_fixed_port_setup(struct
> phy_device *phydev, u8 port)
> val |= PORT_REG_PHYS_CTRL_LINK_VALUE |
> PORT_REG_PHYS_CTRL_LINK_FORCE;
>
Should the 2 bits be cleared before we set them. In case there is config
from an earlier boot stage or eeprom?
> + if (ofnode_valid(node)) {
> + str = ofnode_read_string(node, "phy-mode");
> + if (str)
> + phy_mode = phy_get_interface_by_name(str);
>
This would probably be the sensible place to clear the bits as we know we
have a phy-mode at this point.
+ switch (phy_mode) {
> + case PHY_INTERFACE_MODE_RGMII_ID:
> + val |= PORT_REG_PHYS_CTRL_RGMII_RX_DELAY;
> + val |= PORT_REG_PHYS_CTRL_RGMII_TX_DELAY;
> + break;
> + case PHY_INTERFACE_MODE_RGMII_RXID:
> + val |= PORT_REG_PHYS_CTRL_RGMII_RX_DELAY;
> + break;
> + case PHY_INTERFACE_MODE_RGMII_TXID:
> + val |= PORT_REG_PHYS_CTRL_RGMII_TX_DELAY;
> + break;
> + default:
> + break;
> + }
> + }
> +
> return mv88e61xx_port_write(phydev, port, PORT_REG_PHYS_CTRL,
> val);
> }
> --
> 2.25.1
>
>
More information about the U-Boot
mailing list