On 2015-05-11 07:09, Nikolay Dimitrov wrote:
Hi gang,

While porting kernel 3.10.53 to an imx6d-based board, I stumbled upon
the following issue - my board has a 100 MBit/s RMII PHY which needs 50
MHz reference clock from the SOC, but during the kernel initialization
the clock is configured as 125 MHz (observed with a oscilloscope on a
test pad). The reference clock is sent from pin GPIO_16 to the PHY.

I assume there's no hardware issue with the board, as the old kernel
3.10.17 works fine on the same board. Also U-Boot successfully
downloads the 3.10.53 kernel via network, just before observing the
fore-mentioned issue.

Here are the relevant parts of my DT:


&fec {
     pinctrl-names = "default";
     pinctrl-0 = <&pinctrl_enet_4>;
     phy-mode = "rmii";
     phy-reset-gpios = <&gpio4 5 0>;
     phy-reset-duration = <1>;
     status = "okay";
};

&iomuxc {
     pinctrl-names = "default";
     pinctrl-0 = <&pinctrl_hog_1>;

     enet {
         pinctrl_enet_4: enetgrp-4 {
             fsl,pins = <
                 MX6QDL_PAD_GPIO_16__ENET_REF_CLK    0x4001b0a8
                 MX6QDL_PAD_ENET_MDIO__ENET_MDIO        0x1b0b0
                 MX6QDL_PAD_ENET_MDC__ENET_MDC        0x1b0b0
                 MX6QDL_PAD_ENET_TXD0__ENET_TX_DATA0    0x1b0b0
                 MX6QDL_PAD_ENET_TXD1__ENET_TX_DATA1    0x1b0b0
                 MX6QDL_PAD_ENET_TX_EN__ENET_TX_EN    0x1b0b0
                 MX6QDL_PAD_ENET_RXD0__ENET_RX_DATA0    0x1b0b0
                 MX6QDL_PAD_ENET_RXD1__ENET_RX_DATA1    0x1b0b0
                 MX6QDL_PAD_ENET_RX_ER__ENET_RX_ER    0x1b0b0
                 MX6QDL_PAD_ENET_CRS_DV__ENET_RX_EN    0x1b0b0
             >;
         };
     };
};


Here's a bootlog extract that shows the FEC-related messages:


...
libphy: fec_enet_mii_bus: probed
fec 2188000.ethernet eth0: registered PHC device 0
...
fec 2188000.ethernet eth0: Freescale FEC PHY driver [SMSC LAN8710/LAN8720] 
(mii_bus:phy_addr=2188000.ethernet:01, irq
=-1)
IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
Waiting up to 110 more seconds for network.
...


In general, I'm looking at the appropriate way to set the
CCM_ANALOG_PLL_ENET.DIV_SELECT to 50 MHz.

Thanks in advance for your comments. Regards,
Nikolay

Here's how I solved this - I added a "ref-clock" property to
my fec node and set it to 50MHz.

--
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------
>From f7b6e2268d7107e9e260c67644691e64d9432dfd Mon Sep 17 00:00:00 2001
From: Gary Thomas <[email protected]>
Date: Wed, 29 Apr 2015 09:19:21 -0600
Subject: [PATCH] Set up and start enet_ref clock - needed by Micrel PHY

---
 arch/arm/mach-imx/clk-imx6q.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/arm/mach-imx/clk-imx6q.c b/arch/arm/mach-imx/clk-imx6q.c
index 608c9bcf..21dc524 100644
--- a/arch/arm/mach-imx/clk-imx6q.c
+++ b/arch/arm/mach-imx/clk-imx6q.c
@@ -753,16 +753,18 @@ static void __init imx6q_clocks_init(struct device_node *ccm_node)
 	imx_clk_set_rate(clk[pll4_audio_div], 541900800);
 
         /* Set [default] enet_ref clock to 125M to supply for RGMII tx_clk */
-        clk_set_rate(clk[enet_ref], 125000000);
+        imx_clk_set_rate(clk[enet_ref], 125000000);
         np = of_find_compatible_node(NULL, NULL, "fsl,imx6q-fec");
         if (np) {
                 u32 clock_frequency;
                 int ret;
                 ret = of_property_read_u32(np, "ref-clock", &clock_frequency);
                 if (ret == 0) {
-                        clk_set_rate(clk[enet_ref], clock_frequency);
+                        imx_clk_set_rate(clk[enet_ref], clock_frequency);
                 }
         }
+        // This clock may be needed very early on
+        imx_clk_prepare_enable(clk[enet_ref]);
 
 #ifdef CONFIG_MX6_VPU_352M
 	/*
-- 
1.9.1

-- 
_______________________________________________
meta-freescale mailing list
[email protected]
https://lists.yoctoproject.org/listinfo/meta-freescale

Reply via email to