[PATCH v2] ARM: imx: imx8mn-ddr4-evk: Add ethernet support
Add support for ethernet on the imx8mn-ddr4-evk.
Signed-off-by: Marek Vasut marex@denx.de Cc: Fabio Estevam festevam@gmail.com Cc: Peng Fan peng.fan@nxp.com Cc: Stefano Babic sbabic@denx.de --- V2: Get rid of the IS_ENABLED macros, the ethernet on this board is always present --- arch/arm/dts/imx8mn-evk.dtsi | 1 + board/freescale/imx8mn_evk/imx8mn_evk.c | 36 ++++++++++++++++++++++--- configs/imx8mn_ddr4_evk_defconfig | 5 ++++ 3 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/arm/dts/imx8mn-evk.dtsi b/arch/arm/dts/imx8mn-evk.dtsi index 76d042a4cf0..416fadb22b1 100644 --- a/arch/arm/dts/imx8mn-evk.dtsi +++ b/arch/arm/dts/imx8mn-evk.dtsi @@ -53,6 +53,7 @@ pinctrl-0 = <&pinctrl_fec1>; phy-mode = "rgmii-id"; phy-handle = <ðphy0>; + phy-reset-gpios = <&gpio4 22 GPIO_ACTIVE_LOW>; fsl,magic-packet; status = "okay";
diff --git a/board/freescale/imx8mn_evk/imx8mn_evk.c b/board/freescale/imx8mn_evk/imx8mn_evk.c index 9a0a0488bf4..b24342fd5ca 100644 --- a/board/freescale/imx8mn_evk/imx8mn_evk.c +++ b/board/freescale/imx8mn_evk/imx8mn_evk.c @@ -7,17 +7,47 @@ #include <env.h> #include <init.h> #include <asm/global_data.h> +#include <miiphy.h> +#include <netdev.h> +#include <asm/io.h>
DECLARE_GLOBAL_DATA_PTR;
-int board_init(void) +int board_mmc_get_env_dev(int devno) { + return devno; +} + +static void setup_fec(void) +{ + struct iomuxc_gpr_base_regs *gpr = + (struct iomuxc_gpr_base_regs *)IOMUXC_GPR_BASE_ADDR; + + /* Use 125M anatop REF_CLK1 for ENET1, not from external */ + clrsetbits_le32(&gpr->gpr[1], 0x2000, 0); +} + +int board_phy_config(struct phy_device *phydev) +{ + /* enable rgmii rxc skew and phy mode select to RGMII copper */ + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8); + + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100); + + if (phydev->drv->config) + phydev->drv->config(phydev); return 0; }
-int board_mmc_get_env_dev(int devno) +int board_init(void) { - return devno; + setup_fec(); + + return 0; }
int board_late_init(void) diff --git a/configs/imx8mn_ddr4_evk_defconfig b/configs/imx8mn_ddr4_evk_defconfig index 28cc551dc41..7b032c0ed34 100644 --- a/configs/imx8mn_ddr4_evk_defconfig +++ b/configs/imx8mn_ddr4_evk_defconfig @@ -67,7 +67,12 @@ CONFIG_MMC_HS400_ES_SUPPORT=y CONFIG_MMC_HS400_SUPPORT=y CONFIG_FSL_USDHC=y CONFIG_PHYLIB=y +CONFIG_PHY_ATHEROS=y CONFIG_DM_ETH=y +CONFIG_DM_ETH_PHY=y +CONFIG_PHY_GIGE=y +CONFIG_FEC_MXC=y +CONFIG_MII=y CONFIG_PINCTRL=y CONFIG_SPL_PINCTRL=y CONFIG_PINCTRL_IMX8M=y
Add support for ethernet on the imx8mn-ddr4-evk. Signed-off-by: Marek Vasut marex@denx.de Cc: Fabio Estevam festevam@gmail.com Cc: Peng Fan peng.fan@nxp.com Cc: Stefano Babic sbabic@denx.de
Applied to u-boot-imx, master, thanks !
Best regards, Stefano Babic
Hi Marek,
+int board_phy_config(struct phy_device *phydev) +{
- /* enable rgmii rxc skew and phy mode select to RGMII copper */
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100);
This is a AR8031, correct? Is there a reason why you don't use the device tree binding to set RGMII voltage, and RX and tx delays?
-michael
On 2/21/22 11:10, Michael Walle wrote:
Hi,
+int board_phy_config(struct phy_device *phydev) +{
- /* enable rgmii rxc skew and phy mode select to RGMII copper */
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100);
This is a AR8031, correct? Is there a reason why you don't use the device tree binding to set RGMII voltage, and RX and tx delays?
I quickly cobbled this together from another board, since I needed ethernet to do other tests with this board and the ethernet support was missing. If you can send subsequent patch to flip this to DT-only setup, I can test the patch on the board.
PHY should be AR8031, yes.
Hi Marek,
Am 21.02.22 um 17:49 schrieb Marek Vasut:
On 2/21/22 11:10, Michael Walle wrote:
Hi,
+int board_phy_config(struct phy_device *phydev) +{ + /* enable rgmii rxc skew and phy mode select to RGMII copper */ + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8);
+ phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100);
This is a AR8031, correct? Is there a reason why you don't use the device tree binding to set RGMII voltage, and RX and tx delays?
I quickly cobbled this together from another board, since I needed ethernet to do other tests with this board and the ethernet support was missing. If you can send subsequent patch to flip this to DT-only setup, I can test the patch on the board.
So why don't you post it as RFC? To be honest it seems a bit contradictory that you normally take a critical position on other people's patches when they use legacy code instead of DM/DT (which is alright) and then send a half-baked approach yourself and expect other people to fix it for you afterwards.
I don't want to complain, it just occurred to me.
Frieder
On 2/22/22 08:35, Frieder Schrempf wrote:
Hi Marek,
Hi,
+int board_phy_config(struct phy_device *phydev) +{ + /* enable rgmii rxc skew and phy mode select to RGMII copper */ + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8);
+ phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100);
This is a AR8031, correct? Is there a reason why you don't use the device tree binding to set RGMII voltage, and RX and tx delays?
I quickly cobbled this together from another board, since I needed ethernet to do other tests with this board and the ethernet support was missing. If you can send subsequent patch to flip this to DT-only setup, I can test the patch on the board.
So why don't you post it as RFC? To be honest it seems a bit contradictory that you normally take a critical position on other people's patches when they use legacy code instead of DM/DT (which is alright) and then send a half-baked approach yourself and expect other people to fix it for you afterwards.
I don't want to complain, it just occurred to me.
It's just a patch I had half-baked in a tree for over a year, so I sent it. It didn't occur to me there is now a way to do this in DT, since the patch was in my tree for so long and I only dealt with the ar803x PHY recently. I will try and convert this to DT next time I have the board on my desk, but if someone is faster ... even better.
Hi Marek,
Am 2022-02-21 17:49, schrieb Marek Vasut:
On 2/21/22 11:10, Michael Walle wrote:
+int board_phy_config(struct phy_device *phydev) +{
- /* enable rgmii rxc skew and phy mode select to RGMII copper */
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05);
- phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100);
This is a AR8031, correct? Is there a reason why you don't use the device tree binding to set RGMII voltage, and RX and tx delays?
I quickly cobbled this together from another board, since I needed ethernet to do other tests with this board and the ethernet support was missing. If you can send subsequent patch to flip this to DT-only setup, I can test the patch on the board.
I neither have this board, nor do I care much about it. I just saw the usual hardcoded atheros phy settings being done in the board setup. That being said, you can find an example in: arch/arm/dts/fsl-ls1028a-kontron-sl28-var4.dts
But you have to figure out what you need. I guess the last four writes are the rgmii delays which are handled by phy-mode = "rgmii-id";
The 0x8 to 0x1f is 1.8V RGMII voltage, you'd need to look at the schematics if that voltage is actually used. A Zero means 1.5V. If it's using 2.5V then you don't have to do anything because an external regulator is used.
-michael
On 2/22/22 08:49, Michael Walle wrote:
Hi Marek,
Hi,
Am 2022-02-21 17:49, schrieb Marek Vasut:
On 2/21/22 11:10, Michael Walle wrote:
+int board_phy_config(struct phy_device *phydev) +{ + /* enable rgmii rxc skew and phy mode select to RGMII copper */ + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x1f); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x8);
+ phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x00); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x82ee); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1d, 0x05); + phy_write(phydev, MDIO_DEVAD_NONE, 0x1e, 0x100);
This is a AR8031, correct? Is there a reason why you don't use the device tree binding to set RGMII voltage, and RX and tx delays?
I quickly cobbled this together from another board, since I needed ethernet to do other tests with this board and the ethernet support was missing. If you can send subsequent patch to flip this to DT-only setup, I can test the patch on the board.
I neither have this board, nor do I care much about it. I just saw the usual hardcoded atheros phy settings being done in the board setup. That being said, you can find an example in: arch/arm/dts/fsl-ls1028a-kontron-sl28-var4.dts
But you have to figure out what you need. I guess the last four writes are the rgmii delays which are handled by phy-mode = "rgmii-id";
The 0x8 to 0x1f is 1.8V RGMII voltage, you'd need to look at the schematics if that voltage is actually used. A Zero means 1.5V. If it's using 2.5V then you don't have to do anything because an external regulator is used.
One might almost think that's what NXP should've done for a while, considering this is a devkit and missing ethernet is a problem for any network boot.
So unless NXP is faster, I will try and convert this to DT next time I have the board on my desk.
participants (4)
-
Frieder Schrempf -
Marek Vasut -
Michael Walle -
sbabic＠denx.de