Skip to contentSkip to conversation

imx6qp QuadPlus: support improved enet clocking

PATCH In review Sven Van Asbroeck 3 participants 8 messages Jul 13, 2020 – Aug 17, 2020

Subsystems (7)
lore
Share & tools
lore
#0 lore raw

cover imx6qp QuadPlus: support improved enet clocking

On the imx6qp QuadPlus, the h/w designers have improved enet clocking.

This patchset extends the clock tree to reflect the situation on QuadPlus.

This allows board designers to choose the enet clocking method by making simple clocktree changes in the devicetree.

Default setting: external routing of enet_ref from pad to pad.

Example, change the default to enet_ref @ 125MHz clock routed internally:

&fec {
	assigned-clocks = <&clks IMX6QDL_CLK_ENET_PTP>,
			  <&clks IMX6QDL_CLK_ENET_REF>;
	assigned-clock-parents = <&clks IMX6QDL_CLK_ENET_REF>;
	assigned-clock-rates = <0>, <125000000>;
};

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

Sven Van Asbroeck (5):
ARM: mach-imx6q: do not select enet PTP clock source on QuadPlus
clk: imx: add simple regmap-backed clock mux
dt-bindings: imx6qdl-clock: add QuadPlus enet clocks
clk: imx6q: support improved enet clocking on QuadPlus
ARM: dts: imx6qp: support improved enet clocking on QuadPlus

 .../bindings/clock/imx6q-clock.yaml           |   2 +
 arch/arm/boot/dts/imx6qp.dtsi                 |   3 +
 arch/arm/mach-imx/mach-imx6q.c                |   4 +
 drivers/clk/imx/Makefile                      |   1 +
 drivers/clk/imx/clk-imx6q.c                   |  46 ++++++++
 drivers/clk/imx/clk-mux-regmap.c              | 110 ++++++++++++++++++
 drivers/clk/imx/clk.h                         |   7 ++
 include/dt-bindings/clock/imx6qdl-clock.h     |   5 +-
 include/linux/mfd/syscon/imx6q-iomuxc-gpr.h   |   1 +
 9 files changed, 178 insertions(+), 1 deletion(-)
 create mode 100644 drivers/clk/imx/clk-mux-regmap.c
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/5 ARM: mach-imx6q: do not select enet PTP clock source on QuadPlus

On QuadPlus, the enet PTP clock source is selected via the clocktree. Therefore it should not be set in the machine layer (mach-imx6q).

---

Tree: v5.8-rc4

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

 arch/arm/mach-imx/mach-imx6q.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm/mach-imx/mach-imx6q.c +4 -0
index 85c084a716ab..fcf211492f6f 100644
@@ -169,6 +169,10 @@ static void __init imx6q_1588_init(void)
 	struct regmap *gpr;
 	u32 clksel;
 
+	/* QuadPlus selects PTP clock source via clocktree */
+	if (of_machine_is_compatible("fsl,imx6qp"))
+		return;
+
 	np = of_find_compatible_node(NULL, NULL, "fsl,imx6q-fec");
 	if (!np) {
 		pr_warn("%s: failed to find fec node\n", __func__);
Signature · 1 line
2.17.1
#2 lore raw

PATCH 2/5 clk: imx: add simple regmap-backed clock mux

In reply to Sven Van Asbroeck

On the imx6 QuadPlus, some clocktree muxes have control bits located in the GPR area. This area is already modelled as a syscon regmap.

To control these muxes, add a simple regmap-backed clock mux.

---

Tree: v5.8-rc4

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

 drivers/clk/imx/Makefile         |   1 +
 drivers/clk/imx/clk-mux-regmap.c | 110 +++++++++++++++++++++++++++++++
 drivers/clk/imx/clk.h            |   7 ++
 3 files changed, 118 insertions(+)
 create mode 100644 drivers/clk/imx/clk-mux-regmap.c
drivers/clk/imx/Makefile +1 -0
index 928f874c73d2..3d36740c1efd 100644
@@ -19,6 +19,7 @@ obj-$(CONFIG_MXC_CLK) += \
 	clk-pllv3.o \
 	clk-pllv4.o \
 	clk-sscg-pll.o \
+	clk-mux-regmap.o \
 	clk-pll14xx.o
 
 obj-$(CONFIG_MXC_CLK_SCU) += \
drivers/clk/imx/clk-mux-regmap.c new file +110 -0

110 lines not on this page · raw message

drivers/clk/imx/clk.h +7 -0
index 16adbc34e05f..4df766abbe16 100644
@@ -5,6 +5,8 @@
 #include <linux/spinlock.h>
 #include <linux/clk-provider.h>
 
+struct regmap;
+
 #define IMX_CLK_GATE2_SINGLE_BIT	1
 
 extern spinlock_t imx_ccm_lock;
@@ -447,6 +449,11 @@ static inline struct clk_hw *imx_dev_clk_hw_mux(struct device *dev,
 			reg, shift, width, 0, &imx_ccm_lock);
 }
 
+struct clk_hw *
+imx_clk_hw_mux_regmap(const char *name, struct regmap *regmap, unsigned int reg,
+		      u8 shift, u8 width, const char * const *parent_names,
+		      int num_parents);
+
 static inline struct clk *imx_clk_mux2(const char *name, void __iomem *reg,
 			u8 shift, u8 width, const char * const *parents,
 			int num_parents)
Signature · 1 line
2.17.1
#3 lore raw

PATCH 3/5 dt-bindings: imx6qdl-clock: add QuadPlus enet clocks

In reply to Sven Van Asbroeck

On the QuadPlus, hardware designers have improved the enet clocking. Add these QuadPlus-specific clocks to the dt-bindings.

---

Tree: v5.8-rc4

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

 Documentation/devicetree/bindings/clock/imx6q-clock.yaml | 2 ++
 include/dt-bindings/clock/imx6qdl-clock.h                | 5 ++++-
 2 files changed, 6 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/clock/imx6q-clock.yaml +2 -0
index 92a8e545e212..ecd28403eeff 100644
@@ -34,6 +34,7 @@ properties:
       - description: ckih1 clock input
       - description: anaclk1 clock input
       - description: anaclk2 clock input
+      - description: anaclk3 clock input
 
   clock-names:
     items:
@@ -42,6 +43,7 @@ properties:
       - const: ckih1
       - const: anaclk1
       - const: anaclk2
+      - const: anaclk3
 
   fsl,pmic-stby-poweroff:
     $ref: /schemas/types.yaml#/definitions/flag
include/dt-bindings/clock/imx6qdl-clock.h +4 -1
index e20c43cc36f6..afc6b7b1e3e6 100644
@@ -273,6 +273,9 @@
 #define IMX6QDL_CLK_MMDC_P0_IPG			263
 #define IMX6QDL_CLK_DCIC1			264
 #define IMX6QDL_CLK_DCIC2			265
-#define IMX6QDL_CLK_END				266
+#define IMX6QDL_CLK_ENET_PAD			266
+#define IMX6QDL_CLK_ENET_PTP			267
+#define IMX6QDL_CLK_ANACLK3			268
+#define IMX6QDL_CLK_END				269
 
 #endif /* __DT_BINDINGS_CLOCK_IMX6QDL_H */
Signature · 1 line
2.17.1
#4 lore raw

PATCH 4/5 clk: imx6q: support improved enet clocking on QuadPlus

In reply to Sven Van Asbroeck

On the imx6q QuadPlus, hardware designers have improved the enet clocking:

                          _
	 anaclk3---------| \
	                 |  |              _
	                 |M |--enet_pad-->| \
	                 |  |             |  |
	 enet_ref---o--->|_/              |M |----enet_ptp
	            |                     |  |
	            o---------------------|_/

Where anaclk3 is an external clock fed into the imx6q via a pad, typically RGMII_TX_CTL or GPIO_16, when these are configured in mux mode ENET_REF_CLK.

Board designers can now change the clock tree (via the devicetree) to choose between all three supported enet clocking methods.

See follow-up devicetree patch for details.

---

Tree: v5.8-rc4

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

 drivers/clk/imx/clk-imx6q.c                 | 46 +++++++++++++++++++++
 include/linux/mfd/syscon/imx6q-iomuxc-gpr.h |  1 +
 2 files changed, 47 insertions(+)
drivers/clk/imx/clk-imx6q.c +46 -0
index ba33c79158de..0edffa306a28 100644
@@ -15,6 +15,8 @@
 #include <linux/of_address.h>
 #include <linux/of_irq.h>
 #include <soc/imx/revision.h>
+#include <linux/mfd/syscon.h>
+#include <linux/mfd/syscon/imx6q-iomuxc-gpr.h>
 #include <dt-bindings/clock/imx6qdl-clock.h>
 
 #include "clk.h"
@@ -87,6 +89,8 @@ static const char *pll4_bypass_sels[] = { "pll4", "pll4_bypass_src", };
 static const char *pll5_bypass_sels[] = { "pll5", "pll5_bypass_src", };
 static const char *pll6_bypass_sels[] = { "pll6", "pll6_bypass_src", };
 static const char *pll7_bypass_sels[] = { "pll7", "pll7_bypass_src", };
+static const char * const enet_pad_sels[] = { "anaclk3", "enet_ref", };
+static const char * const enet_ptp_sels[] = { "enet_pad", "enet_ref", };
 
 static struct clk_hw **hws;
 static struct clk_hw_onecell_data *clk_hw_data;
@@ -436,6 +440,7 @@ static struct clk_hw * __init imx6q_obtain_fixed_clk_hw(struct device_node *np,
 
 static void __init imx6q_clocks_init(struct device_node *ccm_node)
 {
+	struct regmap *gpr;
 	struct device_node *np;
 	void __iomem *anatop_base, *base;
 	int ret;
@@ -464,6 +469,9 @@ static void __init imx6q_clocks_init(struct device_node *ccm_node)
 	WARN_ON(!base);
 	of_node_put(np);
 
+	gpr = syscon_regmap_lookup_by_compatible("fsl,imx6q-iomuxc-gpr");
+	WARN_ON(IS_ERR_OR_NULL(gpr));
+
 	/* Audio/video PLL post dividers do not work on i.MX6q revision 1.0 */
 	if (clk_on_imx6q() && imx_get_soc_revision() == IMX_CHIP_REVISION_1_0) {
 		post_div_table[1].div = 1;
@@ -959,6 +967,44 @@ static void __init imx6q_clocks_init(struct device_node *ccm_node)
 	/* Audio-related clocks configuration */
 	clk_set_parent(hws[IMX6QDL_CLK_SPDIF_SEL]->clk, hws[IMX6QDL_CLK_PLL3_PFD3_454M]->clk);
 
+	/*
+	 * The QuadPlus has improved enet clocking:
+	 *                  _
+	 * anaclk3---------| \
+	 *                 |  |              _
+	 *                 |M |--enet_pad-->| \
+	 *                 |  |             |  |
+	 * enet_ref---o--->|_/              |M |----enet_ptp
+	 *            |                     |  |
+	 *            o---------------------|_/
+	 *
+	 * On the plus, three options to generate the enet ptp clock:
+	 *
+	 * a) route enet_ref externally from pad to pad (the default):
+	 *    enet_ptp from enet_pad
+	 *    enet_pad from enet_ref
+	 * b) route internally on SoC from enet_ref:
+	 *    enet_ptp from enet_ref
+	 * c) route external clock (from PHY or oscillator) via pad:
+	 *    enet_ptp from enet_pad
+	 *    enet_pad from anaclk3
+	 *    anaclk3 from PHY or oscillator, add devicetree node
+	 */
+	if (clk_on_imx6qp()) {
+		/*
+		 * anaclk3: clock source from external clock via:
+		 * - RGMII_TX_CTL PAD (mux mode ENET_REF_CLK), or
+		 * - GPIO_16      PAD (mux mode ENET_REF_CLK)
+		 */
+		hws[IMX6QDL_CLK_ANACLK3] = imx6q_obtain_fixed_clk_hw(ccm_node, "anaclk3", 0);
+		hws[IMX6QDL_CLK_ENET_PAD] = imx_clk_hw_mux_regmap("enet_pad", gpr, IOMUXC_GPR1, ffs(IMX6Q_GPR1_ENET_CLK_SEL_MASK) - 1,
+								  1, enet_pad_sels, ARRAY_SIZE(enet_pad_sels));
+		hws[IMX6QDL_CLK_ENET_PTP] = imx_clk_hw_mux_regmap("enet_ptp", gpr, IOMUXC_GPR5, ffs(IMX6Q_GPR5_ENET_TXCLK_SEL_MASK) - 1,
+								  1, enet_ptp_sels, ARRAY_SIZE(enet_ptp_sels));
+		clk_set_parent(hws[IMX6QDL_CLK_ENET_PAD]->clk, hws[IMX6QDL_CLK_ENET_REF]->clk);
+		clk_set_parent(hws[IMX6QDL_CLK_ENET_PTP]->clk, hws[IMX6QDL_CLK_ENET_PAD]->clk);
+	}
+
 	/* All existing boards with PCIe use LVDS1 */
 	if (IS_ENABLED(CONFIG_PCI_IMX6))
 		clk_set_parent(hws[IMX6QDL_CLK_LVDS1_SEL]->clk, hws[IMX6QDL_CLK_SATA_REF_100M]->clk);
include/linux/mfd/syscon/imx6q-iomuxc-gpr.h +1 -0
index d4b5e527a7a3..ac1bb3695933 100644
@@ -240,6 +240,7 @@
 #define IMX6Q_GPR4_IPU_RD_CACHE_CTL		BIT(0)
 
 #define IMX6Q_GPR5_L2_CLK_STOP			BIT(8)
+#define IMX6Q_GPR5_ENET_TXCLK_SEL_MASK		BIT(9)
 #define IMX6Q_GPR5_SATA_SW_PD			BIT(10)
 #define IMX6Q_GPR5_SATA_SW_RST			BIT(11)
 
Signature · 1 line
2.17.1
#5 lore raw

PATCH 5/5 ARM: dts: imx6qp: support improved enet clocking on QuadPlus

In reply to Sven Van Asbroeck

Board designers are now able to choose between all three supported enet clocking methods, by changing the clocktree via the devicetree:

a) route enet_ref externally from pad to pad (the default):

no clock tree changes required

b) route internally on SoC from enet_ref

&fec {
    assigned-clocks = <&clks IMX6QDL_CLK_ENET_PTP>,
                      <&clks IMX6QDL_CLK_ENET_REF>;
    assigned-clock-parents = <&clks IMX6QDL_CLK_ENET_REF>;
    assigned-clock-rates = <0>, <125000000>;
};

c) route external clock (from PHY or oscillator) via pad

/ {
    clocks {
        phy_osc: anaclk3 {
            compatible = "fixed-clock";
            #clock-cells = <0>;
            clock-frequency = <125000000>;
       };

};
};

&fec {
assigned-clocks = <&clks IMX6QDL_CLK_ENET_PTP>,
<&clks IMX6QDL_CLK_ENET_PAD>;
assigned-clock-parents = <&clks IMX6QDL_CLK_ENET_PAD>,
<&clks IMX6QDL_CLK_ANACLK3>;
};

---

Tree: v5.8-rc4

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

 arch/arm/boot/dts/imx6qp.dtsi | 3 +++
 1 file changed, 3 insertions(+)
arch/arm/boot/dts/imx6qp.dtsi +3 -0
index b310f13a53f2..7f81d35f56c9 100644
@@ -79,6 +79,9 @@
 &fec {
 	interrupts = <0 118 IRQ_TYPE_LEVEL_HIGH>,
 		     <0 119 IRQ_TYPE_LEVEL_HIGH>;
+	clocks = <&clks IMX6QDL_CLK_ENET>,
+		 <&clks IMX6QDL_CLK_ENET>,
+		 <&clks IMX6QDL_CLK_ENET_PTP>;
 };
 
 &gpc {
Signature · 1 line
2.17.1
#6 lore raw
In reply to Sven Van Asbroeck
Quoted 27 lines · Sven Van Asbroeck

On the imx6qp QuadPlus, the h/w designers have improved enet clocking.

This patchset extends the clock tree to reflect the situation on QuadPlus.

This allows board designers to choose the enet clocking method by making simple clocktree changes in the devicetree.

Default setting: external routing of enet_ref from pad to pad.

Example, change the default to enet_ref @ 125MHz clock routed internally:

&fec {
	assigned-clocks = <&clks IMX6QDL_CLK_ENET_PTP>,
			  <&clks IMX6QDL_CLK_ENET_REF>;
	assigned-clock-parents = <&clks IMX6QDL_CLK_ENET_REF>;
	assigned-clock-rates = <0>, <125000000>;
};

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

Quoted 6 lines

Sven Van Asbroeck (5):
ARM: mach-imx6q: do not select enet PTP clock source on QuadPlus
clk: imx: add simple regmap-backed clock mux
dt-bindings: imx6qdl-clock: add QuadPlus enet clocks
clk: imx6q: support improved enet clocking on QuadPlus
ARM: dts: imx6qp: support improved enet clocking on QuadPlus

Hi Fugang,

Can you take a look at the series?

Shawn

#7 lore raw

From: Shawn Guo <[email protected]> Sent: Monday, August 17, 2020 3:17 PM

Quoted 34 lines · Sven Van Asbroeck
Quoted 26 lines

On the imx6qp QuadPlus, the h/w designers have improved enet clocking.

This patchset extends the clock tree to reflect the situation on QuadPlus.

This allows board designers to choose the enet clocking method by making simple clocktree changes in the devicetree.

Default setting: external routing of enet_ref from pad to pad.

Example, change the default to enet_ref @ 125MHz clock routed internally:

&fec {
      assigned-clocks = <&clks IMX6QDL_CLK_ENET_PTP>,
                        <&clks IMX6QDL_CLK_ENET_REF>;
      assigned-clock-parents = <&clks IMX6QDL_CLK_ENET_REF>;
      assigned-clock-rates = <0>, <125000000>; };

To: Shawn Guo <[email protected]>
To: Sascha Hauer <[email protected]>
Cc: Pengutronix Kernel Team <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: NXP Linux Team <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

Quoted 6 lines

Sven Van Asbroeck (5):
ARM: mach-imx6q: do not select enet PTP clock source on QuadPlus
clk: imx: add simple regmap-backed clock mux
dt-bindings: imx6qdl-clock: add QuadPlus enet clocks
clk: imx6q: support improved enet clocking on QuadPlus
ARM: dts: imx6qp: support improved enet clocking on QuadPlus

Quoted 5 lines

Hi Fugang,

Can you take a look at the series?

Shawn

In fact, Sven already sent the patch list to me in private for review due to miss cc to me. We discussed the series for long time and suppose Sven will send v2 patch set.

Regards,
Fugang