Skip to contentSkip to conversation

clk: Add Allwinner CLK, RESET support

PATCH Superseded Jagan Teki 6 participants 49 messages Dec 31, 2018 – Jan 22, 2019

Subsystems (12)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore

Jagan Teki

author
#0 lore raw

v5 cover clk: Add Allwinner CLK, RESET support

Although the previous version[1] is properly handled the clock gates with enable and disable management, but this series is trying to add some more complex Allwinner CLK architecture by handling parent clock and other CLK attributes.

Allwinner Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management and these can be handled via ccu_clock_gate, which is almost same as previous version changes.

Tree clock has more Allwinner CLK attributes like clock type, fixed clocks, misc clocks, mp, nk, nkm, nkmp, pre/post div, flags etc and these can be managed via ccu_clock_tree.

On top of initial clock gates supported from previous version, this series is trying to add support for MP, NK, MISC, FIXED clock types with get_rate functionality and that can eventually used by uart driver.

On the summary, this would be an initial infrasture that can fit into remaining clock handle support like set_rate, so the rest of code will add on the requirement basics.

Once this is fine, I will try to add code for other parts especially for MMC since we have migration deadline for BLK, MMC, SCSI.

So, please do let me know if anyone have any inputs.

All these changes available at u-boot-sunxi/clk-next

thanks,
Jagan.

[1] https://patchwork.ozlabs.org/cover/962226/

Jagan Teki (26):
clk: Add Allwinner A64 CLK driver
reset: Add Allwinner RESET driver
clk: sunxi: Add Allwinner H3/H5 CLK driver
clk: sunxi: Add Allwinner A10/A20 CLK driver
clk: sunxi: Add Allwinner A10s/A13 CLK driver
clk: sunxi: Add Allwinner A31 CLK driver
clk: sunxi: Add Allwinner A23/A33 CLK driver
clk: sunxi: Add Allwinner A83T CLK driver
clk: sunxi: Add Allwinner R40 CLK driver
clk: sunxi: Add Allwinner V3S CLK driver
clk: sunxi: Implement UART clocks
clk: sunxi: Implement UART resets
clk: sunxi: Add Allwinner H6 CLK driver
sunxi: A64: Update sun50i-a64-ccu.h
clk: sunxi: Add ccu clock tree support
sunxi: Enable CLK
phy: sun4i-usb: Use CLK and RESET support
reset: Add reset valid
musb-new: sunxi: Use CLK and RESET support
sunxi: usb: Switch to Generic host controllers
usb: host: Drop [e-o]hci-sunxi drivers
clk: sunxi: Implement SPI clocks
spi: sun4i: Add CLK support
clk: sunxi: Implement A64 SPI clocks, resets
spi: Add Allwinner A31 SPI driver
board: sopine: Enable SPI/SPI-FLASH

 .../sun50i-a64-sopine-baseboard-u-boot.dtsi   |  16 +
 arch/arm/include/asm/arch-sunxi/ccu.h         | 284 +++++++++++++
 arch/arm/mach-sunxi/Kconfig                   |  12 +
 configs/A10-OLinuXino-Lime_defconfig          |   1 +
 configs/A10s-OLinuXino-M_defconfig            |   1 +
 configs/A13-OLinuXinoM_defconfig              |   1 +
 configs/A13-OLinuXino_defconfig               |   1 +
 configs/A20-OLinuXino-Lime2-eMMC_defconfig    |   1 +
 configs/A20-OLinuXino-Lime2_defconfig         |   1 +
 configs/A20-OLinuXino-Lime_defconfig          |   1 +
 configs/A20-Olimex-SOM204-EVB_defconfig       |   1 +
 configs/Auxtek-T003_defconfig                 |   1 +
 configs/Auxtek-T004_defconfig                 |   1 +
 configs/Bananapi_defconfig                    |   1 +
 configs/Bananapi_m2m_defconfig                |   1 +
 configs/Bananapro_defconfig                   |   1 +
 configs/CHIP_defconfig                        |   1 +
 configs/CHIP_pro_defconfig                    |   1 +
 configs/CSQ_CS908_defconfig                   |   1 +
 configs/Colombus_defconfig                    |   1 +
 configs/Cubieboard2_defconfig                 |   1 +
 configs/Cubieboard_defconfig                  |   1 +
 configs/Cubietruck_plus_defconfig             |   1 +
 configs/Hummingbird_A31_defconfig             |   1 +
 configs/Itead_Ibox_A20_defconfig              |   1 +
 configs/Linksprite_pcDuino3_Nano_defconfig    |   1 +
 configs/Linksprite_pcDuino3_defconfig         |   1 +
 configs/Linksprite_pcDuino_defconfig          |   1 +
 configs/MK808C_defconfig                      |   1 +
 configs/Marsboard_A10_defconfig               |   1 +
 configs/Mele_A1000G_quad_defconfig            |   1 +
 configs/Mele_A1000_defconfig                  |   1 +
 configs/Mele_I7_defconfig                     |   1 +
 configs/Mele_M3_defconfig                     |   1 +
 configs/Mele_M5_defconfig                     |   1 +
 configs/Mele_M9_defconfig                     |   1 +
 configs/Mini-X_defconfig                      |   1 +
 configs/Orangepi_defconfig                    |   1 +
 configs/Orangepi_mini_defconfig               |   1 +
 configs/Sinlinx_SinA31s_defconfig             |   1 +
 configs/Sinlinx_SinA33_defconfig              |   1 +
 configs/Sinovoip_BPI_M2_Plus_defconfig        |   1 +
 configs/Sinovoip_BPI_M2_defconfig             |   1 +
 configs/Sinovoip_BPI_M3_defconfig             |   1 +
 configs/Wexler_TAB7200_defconfig              |   1 +
 configs/Wobo_i5_defconfig                     |   1 +
 configs/a64-olinuxino_defconfig               |   1 +
 configs/ba10_tv_box_defconfig                 |   1 +
 configs/bananapi_m1_plus_defconfig            |   1 +
 configs/bananapi_m64_defconfig                |   1 +
 configs/ga10h_v1_1_defconfig                  |   1 +
 configs/h8_homlet_v2_defconfig                |   1 +
 configs/i12-tvbox_defconfig                   |   1 +
 configs/icnova-a20-swac_defconfig             |   1 +
 configs/inet1_defconfig                       |   1 +
 configs/inet_q972_defconfig                   |   1 +
 configs/jesurun_q5_defconfig                  |   1 +
 configs/libretech_all_h3_cc_h2_plus_defconfig |   1 +
 configs/libretech_all_h3_cc_h3_defconfig      |   1 +
 configs/libretech_all_h3_cc_h5_defconfig      |   1 +
 configs/mixtile_loftq_defconfig               |   1 +
 configs/mk802_a10s_defconfig                  |   1 +
 configs/mk802_defconfig                       |   1 +
 configs/mk802ii_defconfig                     |   1 +
 configs/nanopi_a64_defconfig                  |   1 +
 configs/nanopi_m1_defconfig                   |   1 +
 configs/nanopi_m1_plus_defconfig              |   1 +
 configs/nanopi_neo2_defconfig                 |   1 +
 configs/nanopi_neo_air_defconfig              |   1 +
 configs/nanopi_neo_defconfig                  |   1 +
 configs/nanopi_neo_plus2_defconfig            |   1 +
 configs/orangepi_2_defconfig                  |   1 +
 configs/orangepi_lite_defconfig               |   1 +
 configs/orangepi_one_defconfig                |   1 +
 configs/orangepi_pc2_defconfig                |   1 +
 configs/orangepi_pc_defconfig                 |   1 +
 configs/orangepi_pc_plus_defconfig            |   1 +
 configs/orangepi_plus2e_defconfig             |   1 +
 configs/orangepi_plus_defconfig               |   1 +
 configs/orangepi_prime_defconfig              |   1 +
 configs/orangepi_r1_defconfig                 |   1 +
 configs/orangepi_win_defconfig                |   1 +
 configs/orangepi_zero_defconfig               |   1 +
 configs/orangepi_zero_plus2_defconfig         |   1 +
 configs/orangepi_zero_plus_defconfig          |   1 +
 configs/parrot_r16_defconfig                  |   1 +
 configs/pine64_plus_defconfig                 |   1 +
 configs/r7-tv-dongle_defconfig                |   1 +
 configs/sopine_baseboard_defconfig            |   8 +
 configs/sun8i_a23_evb_defconfig               |   1 +
 configs/sunxi_Gemei_G9_defconfig              |   1 +
 configs/tbs_a711_defconfig                    |   1 +
 drivers/clk/Kconfig                           |   1 +
 drivers/clk/Makefile                          |   1 +
 drivers/clk/sunxi/Kconfig                     |  82 ++++
 drivers/clk/sunxi/Makefile                    |  18 +
 drivers/clk/sunxi/clk_a10.c                   |  77 ++++
 drivers/clk/sunxi/clk_a10s.c                  |  68 +++
 drivers/clk/sunxi/clk_a23.c                   |  75 ++++
 drivers/clk/sunxi/clk_a31.c                   |  82 ++++
 drivers/clk/sunxi/clk_a64.c                   | 125 ++++++
 drivers/clk/sunxi/clk_a83t.c                  |  75 ++++
 drivers/clk/sunxi/clk_h3.c                    |  89 ++++
 drivers/clk/sunxi/clk_h6.c                    |  53 +++
 drivers/clk/sunxi/clk_r40.c                   |  88 ++++
 drivers/clk/sunxi/clk_sunxi.c                 | 256 +++++++++++
 drivers/clk/sunxi/clk_v3s.c                   |  59 +++
 drivers/phy/allwinner/phy-sun4i-usb.c         |  77 +++-
 drivers/reset/Kconfig                         |   8 +
 drivers/reset/Makefile                        |   1 +
 drivers/reset/reset-sunxi.c                   | 125 ++++++
 drivers/spi/Kconfig                           |   6 +
 drivers/spi/Makefile                          |   1 +
 drivers/spi/sun4i_spi.c                       |  46 +-
 drivers/spi/sun6i_spi.c                       | 398 ++++++++++++++++++
 drivers/usb/host/Kconfig                      |   2 +
 drivers/usb/host/Makefile                     |   2 -
 drivers/usb/host/ehci-sunxi.c                 | 204 ---------
 drivers/usb/host/ohci-sunxi.c                 | 233 ----------
 drivers/usb/musb-new/sunxi.c                  |  81 ++--
 include/configs/sun4i.h                       |   4 -
 include/configs/sun50i.h                      |   5 -
 include/configs/sun5i.h                       |   4 -
 include/configs/sun6i.h                       |   4 -
 include/configs/sun7i.h                       |   4 -
 include/configs/sun8i.h                       |   4 -
 include/configs/sunxi-common.h                |   1 -
 include/dt-bindings/clock/sun50i-a64-ccu.h    |   2 +
 include/reset.h                               |  11 +
 scripts/config_whitelist.txt                  |   2 -
 130 files changed, 2248 insertions(+), 534 deletions(-)
 create mode 100644 arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi
 create mode 100644 arch/arm/include/asm/arch-sunxi/ccu.h
 create mode 100644 drivers/clk/sunxi/Kconfig
 create mode 100644 drivers/clk/sunxi/Makefile
 create mode 100644 drivers/clk/sunxi/clk_a10.c
 create mode 100644 drivers/clk/sunxi/clk_a10s.c
 create mode 100644 drivers/clk/sunxi/clk_a23.c
 create mode 100644 drivers/clk/sunxi/clk_a31.c
 create mode 100644 drivers/clk/sunxi/clk_a64.c
 create mode 100644 drivers/clk/sunxi/clk_a83t.c
 create mode 100644 drivers/clk/sunxi/clk_h3.c
 create mode 100644 drivers/clk/sunxi/clk_h6.c
 create mode 100644 drivers/clk/sunxi/clk_r40.c
 create mode 100644 drivers/clk/sunxi/clk_sunxi.c
 create mode 100644 drivers/clk/sunxi/clk_v3s.c
 create mode 100644 drivers/reset/reset-sunxi.c
 create mode 100644 drivers/spi/sun6i_spi.c
 delete mode 100644 drivers/usb/host/ehci-sunxi.c
 delete mode 100644 drivers/usb/host/ohci-sunxi.c
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#1 lore raw

PATCH v5 1/26 clk: Add Allwinner A64 CLK driver

Add initial clock driver for Allwinner A64.

Implement USB clock enable and disable functions for OHCI, EHCI, OTG and USBPHY gate and clock registers via ccu clk gate table.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 65 +++++++++++++++++++++++
 drivers/clk/Kconfig                   |  1 +
 drivers/clk/Makefile                  |  1 +
 drivers/clk/sunxi/Kconfig             | 18 +++++++
 drivers/clk/sunxi/Makefile            |  9 ++++
 drivers/clk/sunxi/clk_a64.c           | 46 +++++++++++++++++
 drivers/clk/sunxi/clk_sunxi.c         | 74 +++++++++++++++++++++++++++
 7 files changed, 214 insertions(+)
 create mode 100644 arch/arm/include/asm/arch-sunxi/ccu.h
 create mode 100644 drivers/clk/sunxi/Kconfig
 create mode 100644 drivers/clk/sunxi/Makefile
 create mode 100644 drivers/clk/sunxi/clk_a64.c
 create mode 100644 drivers/clk/sunxi/clk_sunxi.c
arch/arm/include/asm/arch-sunxi/ccu.h new file +65 -0
new file mode 100644
index 0000000000..db69c8f0d5
@@ -0,0 +1,65 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (C) 2018 Amarula Solutions.
+ * Author: Jagan Teki <[email protected]>
+ */
+
+#ifndef _ASM_ARCH_CCU_H
+#define _ASM_ARCH_CCU_H
+
+/**
+ * enum ccu_clk_flags - ccu clock flags
+ *
+ * @CCU_CLK_F_INIT_DONE:		clock gate init done check
+ */
+enum ccu_clk_flags {
+	CCU_CLK_F_INIT_DONE		= BIT(0),
+};
+
+/**
+ * struct ccu_clk_gate - ccu clock gate
+ * @off:	gate offset
+ * @bit:	gate bit
+ * @flags:	clock gate flags
+ */
+struct ccu_clk_gate {
+	u16 off;
+	u32 bit;
+	enum ccu_clk_flags flags;
+};
+
+#define GATE(_off, _bit) {			\
+	.off = _off,				\
+	.bit = _bit,				\
+	.flags = CCU_CLK_F_INIT_DONE,		\
+}
+
+/**
+ * struct ccu_desc - clock control unit descriptor
+ *
+ * @gates:	clock gates
+ */
+struct ccu_desc {
+	const struct ccu_clk_gate *gates;
+};
+
+/**
+ * struct ccu_priv - sunxi clock control unit
+ *
+ * @base:	base address
+ * @desc:	ccu descriptor
+ */
+struct ccu_priv {
+	void *base;
+	const struct ccu_desc *desc;
+};
+
+/**
+ * sunxi_clk_probe - common sunxi clock probe
+ * @dev:	clock device
+ */
+int sunxi_clk_probe(struct udevice *dev);
+
+extern struct clk_ops sunxi_clk_ops;
+
+#endif /* _ASM_ARCH_CCU_H */
drivers/clk/Kconfig +1 -0
index eadf7f8250..51c931b906 100644
@@ -104,6 +104,7 @@ source "drivers/clk/imx/Kconfig"
 source "drivers/clk/mvebu/Kconfig"
 source "drivers/clk/owl/Kconfig"
 source "drivers/clk/renesas/Kconfig"
+source "drivers/clk/sunxi/Kconfig"
 source "drivers/clk/tegra/Kconfig"
 source "drivers/clk/uniphier/Kconfig"
 
drivers/clk/Makefile +1 -0
index 9acbb1a650..6a4ff9143b 100644
@@ -22,6 +22,7 @@ obj-$(CONFIG_CLK_HSDK) += clk-hsdk-cgu.o
 obj-$(CONFIG_CLK_MPC83XX) += mpc83xx_clk.o
 obj-$(CONFIG_CLK_OWL) += owl/
 obj-$(CONFIG_CLK_RENESAS) += renesas/
+obj-$(CONFIG_ARCH_SUNXI) += sunxi/
 obj-$(CONFIG_CLK_STM32F) += clk_stm32f.o
 obj-$(CONFIG_CLK_STM32MP1) += clk_stm32mp1.o
 obj-$(CONFIG_CLK_UNIPHIER) += uniphier/
drivers/clk/sunxi/Kconfig new file +18 -0
new file mode 100644
index 0000000000..bf5ecb3801
@@ -0,0 +1,18 @@
+config CLK_SUNXI
+	bool "Clock support for Allwinner SoCs"
+	depends on CLK && ARCH_SUNXI
+	default y
+	help
+	  This enables support for common clock driver API on Allwinner
+	  SoCs.
+
+if CLK_SUNXI
+
+config CLK_SUN50I_A64
+	bool "Clock driver for Allwinner A64"
+	default MACH_SUN50I
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner A64 SoC.
+
+endif # CLK_SUNXI
drivers/clk/sunxi/Makefile new file +9 -0
new file mode 100644
index 0000000000..fb20d28333
@@ -0,0 +1,9 @@
+#
+# Copyright (C) 2018 Amarula Solutions.
+#
+# SPDX-License-Identifier:      GPL-2.0+
+#
+
+obj-$(CONFIG_CLK_SUNXI) += clk_sunxi.o
+
+obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_a64.c new file +46 -0

46 lines not on this page · raw message

drivers/clk/sunxi/clk_sunxi.c new file +74 -0

74 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#2 lore raw

PATCH v5 2/26 reset: Add Allwinner RESET driver

In reply to Jagan Teki

Add common reset driver for all Allwinner SoC's.

Since CLK and RESET share common DT compatible, it is CLK driver job is to bind the reset driver. So add CLK bind call on respective SoC driver by passing ccu map descriptor so-that reset deassert, deassert operations held based on ccu reset table defined from CLK driver.

Select DM_RESET via CLK_SUNXI, this make hidden section of RESET since CLK and RESET share common DT compatible and code.

---
 arch/arm/include/asm/arch-sunxi/ccu.h |  29 ++++++
 drivers/clk/sunxi/Kconfig             |   1 +
 drivers/clk/sunxi/clk_a64.c           |  20 +++++
 drivers/reset/Kconfig                 |   8 ++
 drivers/reset/Makefile                |   1 +
 drivers/reset/reset-sunxi.c           | 125 ++++++++++++++++++++++++++
 6 files changed, 184 insertions(+)
 create mode 100644 drivers/reset/reset-sunxi.c
arch/arm/include/asm/arch-sunxi/ccu.h +29 -0
index db69c8f0d5..3fdc26978d 100644
@@ -34,13 +34,33 @@ struct ccu_clk_gate {
 	.flags = CCU_CLK_F_INIT_DONE,		\
 }
 
+/**
+ * struct ccu_reset - ccu reset
+ * @off:	reset offset
+ * @bit:	reset bit
+ * @flags:	reset flags
+ */
+struct ccu_reset {
+	u16 off;
+	u32 bit;
+	enum ccu_clk_flags flags;
+};
+
+#define RESET(_off, _bit) {			\
+	.off = _off,				\
+	.bit = _bit,				\
+	.flags = CCU_CLK_F_INIT_DONE,		\
+}
+
 /**
  * struct ccu_desc - clock control unit descriptor
  *
  * @gates:	clock gates
+ * @resets:	reset unit
  */
 struct ccu_desc {
 	const struct ccu_clk_gate *gates;
+	const struct ccu_reset *resets;
 };
 
 /**
@@ -62,4 +82,13 @@ int sunxi_clk_probe(struct udevice *dev);
 
 extern struct clk_ops sunxi_clk_ops;
 
+/**
+ * sunxi_reset_bind() - reset binding
+ *
+ * @dev:       reset device
+ * @count:     reset count
+ * @return 0 success, or error value
+ */
+int sunxi_reset_bind(struct udevice *dev, ulong count);
+
 #endif /* _ASM_ARCH_CCU_H */
drivers/clk/sunxi/Kconfig +1 -0
index bf5ecb3801..041d711e58 100644
@@ -1,6 +1,7 @@
 config CLK_SUNXI
 	bool "Clock support for Allwinner SoCs"
 	depends on CLK && ARCH_SUNXI
+	select DM_RESET
 	default y
 	help
 	  This enables support for common clock driver API on Allwinner
drivers/clk/sunxi/clk_a64.c +20 -0
index 803a2f711d..28bda1f497 100644
@@ -10,6 +10,7 @@
 #include <errno.h>
 #include <asm/arch/ccu.h>
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
+#include <dt-bindings/reset/sun50i-a64-ccu.h>
 
 static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_BUS_OTG]		= GATE(0x060, BIT(23)),
@@ -26,10 +27,28 @@ static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_USB_OHCI1]		= GATE(0x0cc, BIT(17)),
 };
 
+static const struct ccu_reset a64_resets[] = {
+	[RST_USB_PHY0]          = RESET(0x0cc, BIT(0)),
+	[RST_USB_PHY1]          = RESET(0x0cc, BIT(1)),
+	[RST_USB_HSIC]          = RESET(0x0cc, BIT(2)),
+
+	[RST_BUS_OTG]           = RESET(0x2c0, BIT(23)),
+	[RST_BUS_EHCI0]         = RESET(0x2c0, BIT(24)),
+	[RST_BUS_EHCI1]         = RESET(0x2c0, BIT(25)),
+	[RST_BUS_OHCI0]         = RESET(0x2c0, BIT(28)),
+	[RST_BUS_OHCI1]         = RESET(0x2c0, BIT(29)),
+};
+
 static const struct ccu_desc a64_ccu_desc = {
 	.gates = a64_gates,
+	.resets = a64_resets,
 };
 
+static int a64_clk_bind(struct udevice *dev)
+{
+	return sunxi_reset_bind(dev, 50);
+}
+
 static const struct udevice_id a64_ccu_ids[] = {
 	{ .compatible = "allwinner,sun50i-a64-ccu",
 	  .data = (ulong)&a64_ccu_desc },
@@ -43,4 +62,5 @@ U_BOOT_DRIVER(clk_sun50i_a64) = {
 	.priv_auto_alloc_size	= sizeof(struct ccu_priv),
 	.ops		= &sunxi_clk_ops,
 	.probe		= sunxi_clk_probe,
+	.bind		= a64_clk_bind,
 };
drivers/reset/Kconfig +8 -0
index 9c5208b7da..b6b40b6ce9 100644
@@ -106,4 +106,12 @@ config RESET_SOCFPGA
 	help
 	  Support for reset controller on SoCFPGA platform.
 
+config RESET_SUNXI
+	bool "RESET support for Allwinner SoCs"
+	depends on DM_RESET && ARCH_SUNXI
+	default y
+	help
+	  This enables support for common reset driver for
+	  Allwinner SoCs.
+
 endmenu
drivers/reset/Makefile +1 -0
index f4520878b7..377c038163 100644
@@ -17,3 +17,4 @@ obj-$(CONFIG_AST2500_RESET) += ast2500-reset.o
 obj-$(CONFIG_RESET_ROCKCHIP) += reset-rockchip.o
 obj-$(CONFIG_RESET_MESON) += reset-meson.o
 obj-$(CONFIG_RESET_SOCFPGA) += reset-socfpga.o
+obj-$(CONFIG_RESET_SUNXI) += reset-sunxi.o
drivers/reset/reset-sunxi.c new file +125 -0

125 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
#3 lore raw

PATCH v5 2/26 reset: Add Allwinner RESET driver

Quoted 113 lines · Jagan Teki

Add common reset driver for all Allwinner SoC's.

Since CLK and RESET share common DT compatible, it is CLK driver job is to bind the reset driver. So add CLK bind call on respective SoC driver by passing ccu map descriptor so-that reset deassert, deassert operations held based on ccu reset table defined from CLK driver.

Select DM_RESET via CLK_SUNXI, this make hidden section of RESET since CLK and RESET share common DT compatible and code.

---
 arch/arm/include/asm/arch-sunxi/ccu.h |  29 ++++++
 drivers/clk/sunxi/Kconfig             |   1 +
 drivers/clk/sunxi/clk_a64.c           |  20 +++++
 drivers/reset/Kconfig                 |   8 ++
 drivers/reset/Makefile                |   1 +
 drivers/reset/reset-sunxi.c           | 125 ++++++++++++++++++++++++++
 6 files changed, 184 insertions(+)
 create mode 100644 drivers/reset/reset-sunxi.c
arch/arm/include/asm/arch-sunxi/ccu.h +29 -0
index db69c8f0d5..3fdc26978d 100644
@@ -34,13 +34,33 @@ struct ccu_clk_gate {
 	.flags = CCU_CLK_F_INIT_DONE,		\
 }
 
+/**
+ * struct ccu_reset - ccu reset
+ * @off:	reset offset
+ * @bit:	reset bit
+ * @flags:	reset flags
+ */
+struct ccu_reset {
+	u16 off;
+	u32 bit;
+	enum ccu_clk_flags flags;
+};
+
+#define RESET(_off, _bit) {			\
+	.off = _off,				\
+	.bit = _bit,				\
+	.flags = CCU_CLK_F_INIT_DONE,		\
+}
+
 /**
  * struct ccu_desc - clock control unit descriptor
  *
  * @gates:	clock gates
+ * @resets:	reset unit
  */
 struct ccu_desc {
 	const struct ccu_clk_gate *gates;
+	const struct ccu_reset *resets;
 };
 
 /**
@@ -62,4 +82,13 @@ int sunxi_clk_probe(struct udevice *dev);
 
 extern struct clk_ops sunxi_clk_ops;
 
+/**
+ * sunxi_reset_bind() - reset binding
+ *
+ * @dev:       reset device
+ * @count:     reset count
+ * @return 0 success, or error value
+ */
+int sunxi_reset_bind(struct udevice *dev, ulong count);
+
 #endif /* _ASM_ARCH_CCU_H */
drivers/clk/sunxi/Kconfig +1 -0
index bf5ecb3801..041d711e58 100644
@@ -1,6 +1,7 @@
 config CLK_SUNXI
 	bool "Clock support for Allwinner SoCs"
 	depends on CLK && ARCH_SUNXI
+	select DM_RESET
 	default y
 	help
 	  This enables support for common clock driver API on Allwinner
drivers/clk/sunxi/clk_a64.c +11 -0
index 803a2f711d..28bda1f497 100644
@@ -10,6 +10,7 @@
 #include <errno.h>
 #include <asm/arch/ccu.h>
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
+#include <dt-bindings/reset/sun50i-a64-ccu.h>
 
 static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_BUS_OTG]		= GATE(0x060, BIT(23)),
@@ -26,10 +27,28 @@ static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_USB_OHCI1]		= GATE(0x0cc, BIT(17)),
 };
 
+static const struct ccu_reset a64_resets[] = {
+	[RST_USB_PHY0]          = RESET(0x0cc, BIT(0)),
+	[RST_USB_PHY1]          = RESET(0x0cc, BIT(1)),
+	[RST_USB_HSIC]          = RESET(0x0cc, BIT(2)),
+
+	[RST_BUS_OTG]           = RESET(0x2c0, BIT(23)),
+	[RST_BUS_EHCI0]         = RESET(0x2c0, BIT(24)),
+	[RST_BUS_EHCI1]         = RESET(0x2c0, BIT(25)),
+	[RST_BUS_OHCI0]         = RESET(0x2c0, BIT(28)),
+	[RST_BUS_OHCI1]         = RESET(0x2c0, BIT(29)),

drivers/clk/sunxi/clk_a64.c static const struct ccu_clk_gate a64_gates[] = { · line 49

Quoted 10 lines
drivers/clk/sunxi/clk_a64.c +6 -0
+};
+
 static const struct ccu_desc a64_ccu_desc = {
 	.gates = a64_gates,
+	.resets = a64_resets,
 };
 
+static int a64_clk_bind(struct udevice *dev)
+{
+	return sunxi_reset_bind(dev, 50);

The second parameter is count, so it should be 51. But this is wrong either way, so just use ARRAY_SIZE(a64_reset), as this is what you are after with the check later on.
Same for the other SoCs, of course.

Quoted 67 lines
drivers/clk/sunxi/clk_a64.c +3 -0
+}
+
 static const struct udevice_id a64_ccu_ids[] = {
 	{ .compatible = "allwinner,sun50i-a64-ccu",
 	  .data = (ulong)&a64_ccu_desc },
@@ -43,4 +62,5 @@ U_BOOT_DRIVER(clk_sun50i_a64) = {
 	.priv_auto_alloc_size	= sizeof(struct ccu_priv),
 	.ops		= &sunxi_clk_ops,
 	.probe		= sunxi_clk_probe,
+	.bind		= a64_clk_bind,
 };
drivers/reset/Kconfig +8 -0
index 9c5208b7da..b6b40b6ce9 100644
@@ -106,4 +106,12 @@ config RESET_SOCFPGA
 	help
 	  Support for reset controller on SoCFPGA platform.
 
+config RESET_SUNXI
+	bool "RESET support for Allwinner SoCs"
+	depends on DM_RESET && ARCH_SUNXI
+	default y
+	help
+	  This enables support for common reset driver for
+	  Allwinner SoCs.
+
 endmenu
drivers/reset/Makefile +1 -0
index f4520878b7..377c038163 100644
@@ -17,3 +17,4 @@ obj-$(CONFIG_AST2500_RESET) += ast2500-reset.o
 obj-$(CONFIG_RESET_ROCKCHIP) += reset-rockchip.o
 obj-$(CONFIG_RESET_MESON) += reset-meson.o
 obj-$(CONFIG_RESET_SOCFPGA) += reset-socfpga.o
+obj-$(CONFIG_RESET_SUNXI) += reset-sunxi.o
drivers/reset/reset-sunxi.c new file +24 -0
new file mode 100644
index 0000000000..af63cac64e
@@ -0,0 +1,125 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR MIT)
+/*
+ * Copyright (C) 2018 Amarula Solutions.
+ * Author: Jagan Teki <[email protected]>
+ */
+
+#include <common.h>
+#include <dm.h>
+#include <errno.h>
+#include <reset-uclass.h>
+#include <asm/io.h>
+#include <dm/lists.h>
+#include <linux/log2.h>
+#include <asm/arch/ccu.h>
+
+struct sunxi_reset_priv {
+	void *base;
+	ulong count;
+	const struct ccu_desc *desc;
+};
+
+static const struct ccu_reset *priv_to_reset(struct sunxi_reset_priv *priv,
+					     unsigned long id)
+{

drivers/reset/reset-sunxi.c line 34

Quoted 10 lines
drivers/reset/reset-sunxi.c new file +10 -0
+	return	&priv->desc->resets[id];
+}
+
+static int sunxi_reset_request(struct reset_ctl *reset_ctl)
+{
+	struct sunxi_reset_priv *priv = dev_get_priv(reset_ctl->dev);
+
+	debug("%s: (RST#%ld)\n", __func__, reset_ctl->id);
+
+	/* check dt-bindings/reset/sun8i-h3-ccu.h for max id */

File name should be generic, but the comment is not correct anyway, since we are limited by the array size, not by the DT. So please just remove this comment.

Cheers,
Andre.

Quoted 92 lines
drivers/reset/reset-sunxi.c new file +91 -0
+	if (reset_ctl->id >= priv->count)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int sunxi_reset_free(struct reset_ctl *reset_ctl)
+{
+	debug("%s: (RST#%ld)\n", __func__, reset_ctl->id);
+
+	return 0;
+}
+
+static int sunxi_set_reset(struct reset_ctl *reset_ctl, bool on)
+{
+	struct sunxi_reset_priv *priv = dev_get_priv(reset_ctl->dev);
+	const struct ccu_reset *reset = priv_to_reset(priv, reset_ctl->id);
+	u32 reg;
+
+	if (!(reset->flags & CCU_CLK_F_INIT_DONE)) {
+		printf("%s: (RST#%ld) unhandled\n", __func__, reset_ctl->id);
+		return 0;
+	}
+
+	debug("%s: (RST#%ld) off#0x%x, BIT(%d)\n", __func__,
+	      reset_ctl->id, reset->off, ilog2(reset->bit));
+
+	reg = readl(priv->base + reset->off);
+	if (on)
+		reg |= reset->bit;
+	else
+		reg &= ~reset->bit;
+
+	writel(reg, priv->base + reset->off);
+
+	return 0;
+}
+
+static int sunxi_reset_assert(struct reset_ctl *reset_ctl)
+{
+	return sunxi_set_reset(reset_ctl, false);
+}
+
+static int sunxi_reset_deassert(struct reset_ctl *reset_ctl)
+{
+	return sunxi_set_reset(reset_ctl, true);
+}
+
+struct reset_ops sunxi_reset_ops = {
+	.request = sunxi_reset_request,
+	.free = sunxi_reset_free,
+	.rst_assert = sunxi_reset_assert,
+	.rst_deassert = sunxi_reset_deassert,
+};
+
+static int sunxi_reset_probe(struct udevice *dev)
+{
+	struct sunxi_reset_priv *priv = dev_get_priv(dev);
+
+	priv->base = dev_read_addr_ptr(dev);
+
+	return 0;
+}
+
+int sunxi_reset_bind(struct udevice *dev, ulong count)
+{
+	struct udevice *rst_dev;
+	struct sunxi_reset_priv *priv;
+	int ret;
+
+	ret = device_bind_driver_to_node(dev, "sunxi_reset", "reset",
+					 dev_ofnode(dev), &rst_dev);
+	if (ret) {
+		debug("failed to bind sunxi_reset driver (ret=%d)\n", ret);
+		return ret;
+	}
+	priv = malloc(sizeof(struct sunxi_reset_priv));
+	priv->count = count;
+	priv->desc = (const struct ccu_desc *)dev_get_driver_data(dev);
+	rst_dev->priv = priv;
+
+	return 0;
+}
+
+U_BOOT_DRIVER(sunxi_reset) = {
+	.name		= "sunxi_reset",
+	.id		= UCLASS_RESET,
+	.ops		= &sunxi_reset_ops,
+	.probe		= sunxi_reset_probe,
+	.priv_auto_alloc_size = sizeof(struct sunxi_reset_priv),
+};
author
#4 lore raw

PATCH v5 3/26 clk: sunxi: Add Allwinner H3/H5 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner H3/H5.

- Implement USB bus and USB clocks via ccu_clk_gate table for H3/H5, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB bus and USB resets via ccu_reset table for H3/H5, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig  |  7 ++++
 drivers/clk/sunxi/Makefile |  1 +
 drivers/clk/sunxi/clk_h3.c | 79 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 87 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_h3.c
drivers/clk/sunxi/Kconfig +7 -0
index 041d711e58..c3713bbac2 100644
@@ -9,6 +9,13 @@ config CLK_SUNXI
 
 if CLK_SUNXI
 
+config CLK_SUN8I_H3
+	bool "Clock driver for Allwinner H3/H5"
+	default MACH_SUNXI_H3_H5
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner H3/H5 SoC.
+
 config CLK_SUN50I_A64
 	bool "Clock driver for Allwinner A64"
 	default MACH_SUN50I
drivers/clk/sunxi/Makefile +1 -0
index fb20d28333..dec49f27a1 100644
@@ -6,4 +6,5 @@
 
 obj-$(CONFIG_CLK_SUNXI) += clk_sunxi.o
 
+obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_h3.c new file +79 -0

79 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#5 lore raw

PATCH v5 4/26 clk: sunxi: Add Allwinner A10/A20 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner A10/A20.

- Implement USB ahb and USB clocks via ccu_clk_gate table for A10/A20, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB resets via ccu_reset table for A10/A20, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig   |  7 +++++
 drivers/clk/sunxi/Makefile  |  1 +
 drivers/clk/sunxi/clk_a10.c | 59 +++++++++++++++++++++++++++++++++++++
 3 files changed, 67 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_a10.c
drivers/clk/sunxi/Kconfig +7 -0
index c3713bbac2..fbbf94ef55 100644
@@ -9,6 +9,13 @@ config CLK_SUNXI
 
 if CLK_SUNXI
 
+config CLK_SUN4I_A10
+	bool "Clock driver for Allwinner A10/A20"
+	default MACH_SUN4I || MACH_SUN7I
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner A10/A20 SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index dec49f27a1..bba830922f 100644
@@ -6,5 +6,6 @@
 
 obj-$(CONFIG_CLK_SUNXI) += clk_sunxi.o
 
+obj-$(CONFIG_CLK_SUN4I_A10) += clk_a10.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_a10.c new file +59 -0

59 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#6 lore raw

PATCH v5 5/26 clk: sunxi: Add Allwinner A10s/A13 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner A10s/A13.

- Implement USB ahb and USB clocks via ccu_clk_gate table for A10s/A13, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB resets via ccu_reset table for A10s/A13, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig    |  7 +++++
 drivers/clk/sunxi/Makefile   |  1 +
 drivers/clk/sunxi/clk_a10s.c | 56 ++++++++++++++++++++++++++++++++++++
 3 files changed, 64 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_a10s.c
drivers/clk/sunxi/Kconfig +7 -0
index fbbf94ef55..b228c2fa3a 100644
@@ -16,6 +16,13 @@ config CLK_SUN4I_A10
 	  This enables common clock driver support for platforms based
 	  on Allwinner A10/A20 SoC.
 
+config CLK_SUN5I_A10S
+	bool "Clock driver for Allwinner A10s/A13"
+	default MACH_SUN5I
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner A10s/A13 SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index bba830922f..466d4b79d6 100644
@@ -7,5 +7,6 @@
 obj-$(CONFIG_CLK_SUNXI) += clk_sunxi.o
 
 obj-$(CONFIG_CLK_SUN4I_A10) += clk_a10.o
+obj-$(CONFIG_CLK_SUN5I_A10S) += clk_a10s.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_a10s.c new file +56 -0

56 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#7 lore raw

PATCH v5 6/26 clk: sunxi: Add Allwinner A31 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner A31.

- Implement USB ahb1 and USB clocks via ccu_clk_gate table for A31, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB ahb1 and USB resets via ccu_reset table for A31, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig   |  7 ++++
 drivers/clk/sunxi/Makefile  |  1 +
 drivers/clk/sunxi/clk_a31.c | 68 +++++++++++++++++++++++++++++++++++++
 3 files changed, 76 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_a31.c
drivers/clk/sunxi/Kconfig +7 -0
index b228c2fa3a..535b0dc02c 100644
@@ -23,6 +23,13 @@ config CLK_SUN5I_A10S
 	  This enables common clock driver support for platforms based
 	  on Allwinner A10s/A13 SoC.
 
+config CLK_SUN6I_A31
+	bool "Clock driver for Allwinner A31/A31s"
+	default MACH_SUN6I
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner A31/A31s SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index 466d4b79d6..3cf0071b0c 100644
@@ -8,5 +8,6 @@ obj-$(CONFIG_CLK_SUNXI) += clk_sunxi.o
 
 obj-$(CONFIG_CLK_SUN4I_A10) += clk_a10.o
 obj-$(CONFIG_CLK_SUN5I_A10S) += clk_a10s.o
+obj-$(CONFIG_CLK_SUN6I_A31) += clk_a31.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_a31.c new file +68 -0

68 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#8 lore raw

PATCH v5 7/26 clk: sunxi: Add Allwinner A23/A33 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner A23/A33.

- Implement USB bus and USB clocks via ccu_clk_gate table for A23/A33, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB bus and USB resets via ccu_reset table for A23/A33, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig   |  7 +++++
 drivers/clk/sunxi/Makefile  |  1 +
 drivers/clk/sunxi/clk_a23.c | 63 +++++++++++++++++++++++++++++++++++++
 3 files changed, 71 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_a23.c
drivers/clk/sunxi/Kconfig +7 -0
index 535b0dc02c..38ff99d345 100644
@@ -30,6 +30,13 @@ config CLK_SUN6I_A31
 	  This enables common clock driver support for platforms based
 	  on Allwinner A31/A31s SoC.
 
+config CLK_SUN8I_A23
+	bool "Clock driver for Allwinner A23/A33"
+	default MACH_SUN8I_A23 || MACH_SUN8I_A33
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner A23/A33 SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index 3cf0071b0c..6924897036 100644
@@ -9,5 +9,6 @@ obj-$(CONFIG_CLK_SUNXI) += clk_sunxi.o
 obj-$(CONFIG_CLK_SUN4I_A10) += clk_a10.o
 obj-$(CONFIG_CLK_SUN5I_A10S) += clk_a10s.o
 obj-$(CONFIG_CLK_SUN6I_A31) += clk_a31.o
+obj-$(CONFIG_CLK_SUN8I_A23) += clk_a23.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_a23.c new file +63 -0

63 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#9 lore raw

PATCH v5 8/26 clk: sunxi: Add Allwinner A83T CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner A83T.

- Implement USB bus and USB clocks via ccu_clk_gate table for A83T, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB bus and USB resets via ccu_reset table for A83T, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig    |  7 ++++
 drivers/clk/sunxi/Makefile   |  1 +
 drivers/clk/sunxi/clk_a83t.c | 63 ++++++++++++++++++++++++++++++++++++
 3 files changed, 71 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_a83t.c
drivers/clk/sunxi/Kconfig +7 -0
index 38ff99d345..90af70d171 100644
@@ -37,6 +37,13 @@ config CLK_SUN8I_A23
 	  This enables common clock driver support for platforms based
 	  on Allwinner A23/A33 SoC.
 
+config CLK_SUN8I_A83T
+	bool "Clock driver for Allwinner A83T"
+	default MACH_SUN8I_A83T
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner A83T SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index 6924897036..4a254c8671 100644
@@ -10,5 +10,6 @@ obj-$(CONFIG_CLK_SUN4I_A10) += clk_a10.o
 obj-$(CONFIG_CLK_SUN5I_A10S) += clk_a10s.o
 obj-$(CONFIG_CLK_SUN6I_A31) += clk_a31.o
 obj-$(CONFIG_CLK_SUN8I_A23) += clk_a23.o
+obj-$(CONFIG_CLK_SUN8I_A83T) += clk_a83t.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_a83t.c new file +63 -0

63 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#10 lore raw

PATCH v5 9/26 clk: sunxi: Add Allwinner R40 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner R40.

- Implement USB bus and USB clocks via ccu_clk_gate for R40, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB bus and USB resets via ccu_reset table for R40, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig   |  7 ++++
 drivers/clk/sunxi/Makefile  |  1 +
 drivers/clk/sunxi/clk_r40.c | 70 +++++++++++++++++++++++++++++++++++++
 3 files changed, 78 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_r40.c
drivers/clk/sunxi/Kconfig +7 -0
index 90af70d171..c45a4ba378 100644
@@ -44,6 +44,13 @@ config CLK_SUN8I_A83T
 	  This enables common clock driver support for platforms based
 	  on Allwinner A83T SoC.
 
+config CLK_SUN8I_R40
+	bool "Clock driver for Allwinner R40"
+	default MACH_SUN8I_R40
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner R40 SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index 4a254c8671..61f8b87396 100644
@@ -11,5 +11,6 @@ obj-$(CONFIG_CLK_SUN5I_A10S) += clk_a10s.o
 obj-$(CONFIG_CLK_SUN6I_A31) += clk_a31.o
 obj-$(CONFIG_CLK_SUN8I_A23) += clk_a23.o
 obj-$(CONFIG_CLK_SUN8I_A83T) += clk_a83t.o
+obj-$(CONFIG_CLK_SUN8I_R40) += clk_r40.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_r40.c new file +70 -0

70 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#11 lore raw

PATCH v5 10/26 clk: sunxi: Add Allwinner V3S CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner V3S.

- Implement USB bus and USB clocks via ccu_clk_gate table for V3S, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement USB bus and USB resets via ccu_reset table for V3S, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig   |  7 +++++
 drivers/clk/sunxi/Makefile  |  1 +
 drivers/clk/sunxi/clk_v3s.c | 51 +++++++++++++++++++++++++++++++++++++
 3 files changed, 59 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_v3s.c
drivers/clk/sunxi/Kconfig +7 -0
index c45a4ba378..a6f84e9e56 100644
@@ -51,6 +51,13 @@ config CLK_SUN8I_R40
 	  This enables common clock driver support for platforms based
 	  on Allwinner R40 SoC.
 
+config CLK_SUN8I_V3S
+	bool "Clock driver for Allwinner V3S"
+	default MACH_SUN8I_V3S
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner V3S SoC.
+
 config CLK_SUN8I_H3
 	bool "Clock driver for Allwinner H3/H5"
 	default MACH_SUNXI_H3_H5
drivers/clk/sunxi/Makefile +1 -0
index 61f8b87396..fbd43527a6 100644
@@ -12,5 +12,6 @@ obj-$(CONFIG_CLK_SUN6I_A31) += clk_a31.o
 obj-$(CONFIG_CLK_SUN8I_A23) += clk_a23.o
 obj-$(CONFIG_CLK_SUN8I_A83T) += clk_a83t.o
 obj-$(CONFIG_CLK_SUN8I_R40) += clk_r40.o
+obj-$(CONFIG_CLK_SUN8I_V3S) += clk_v3s.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_v3s.c new file +51 -0

51 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#12 lore raw

PATCH v5 11/26 clk: sunxi: Implement UART clocks

In reply to Jagan Teki

Implement UART clocks for all Allwinner SoC
clock drivers via ccu clock gate table.

---
 drivers/clk/sunxi/clk_a10.c  | 9 +++++++++
 drivers/clk/sunxi/clk_a10s.c | 5 +++++
 drivers/clk/sunxi/clk_a23.c  | 6 ++++++
 drivers/clk/sunxi/clk_a31.c  | 7 +++++++
 drivers/clk/sunxi/clk_a64.c  | 6 ++++++
 drivers/clk/sunxi/clk_a83t.c | 6 ++++++
 drivers/clk/sunxi/clk_h3.c   | 5 +++++
 drivers/clk/sunxi/clk_r40.c  | 9 +++++++++
 drivers/clk/sunxi/clk_v3s.c  | 4 ++++
 9 files changed, 57 insertions(+)
drivers/clk/sunxi/clk_a10.c +9 -0
index 59d9d90a19..1b074e3872 100644
@@ -19,6 +19,15 @@ static struct ccu_clk_gate a10_gates[] = {
 	[CLK_AHB_EHCI1]		= GATE(0x060, BIT(3)),
 	[CLK_AHB_OHCI1]		= GATE(0x060, BIT(4)),
 
+	[CLK_APB1_UART0]	= GATE(0x06c, BIT(16)),
+	[CLK_APB1_UART1]	= GATE(0x06c, BIT(17)),
+	[CLK_APB1_UART2]	= GATE(0x06c, BIT(18)),
+	[CLK_APB1_UART3]	= GATE(0x06c, BIT(19)),
+	[CLK_APB1_UART4]	= GATE(0x06c, BIT(20)),
+	[CLK_APB1_UART5]	= GATE(0x06c, BIT(21)),
+	[CLK_APB1_UART6]	= GATE(0x06c, BIT(22)),
+	[CLK_APB1_UART7]	= GATE(0x06c, BIT(23)),
+
 	[CLK_USB_OHCI0]		= GATE(0x0cc, BIT(6)),
 	[CLK_USB_OHCI1]		= GATE(0x0cc, BIT(7)),
 	[CLK_USB_PHY]		= GATE(0x0cc, BIT(8)),
drivers/clk/sunxi/clk_a10s.c +5 -0
index 474f37d607..a50c6e3f7f 100644
@@ -17,6 +17,11 @@ static struct ccu_clk_gate a10s_gates[] = {
 	[CLK_AHB_EHCI]		= GATE(0x060, BIT(1)),
 	[CLK_AHB_OHCI]		= GATE(0x060, BIT(2)),
 
+	[CLK_APB1_UART0]	= GATE(0x06c, BIT(16)),
+	[CLK_APB1_UART1]	= GATE(0x06c, BIT(17)),
+	[CLK_APB1_UART2]	= GATE(0x06c, BIT(18)),
+	[CLK_APB1_UART3]	= GATE(0x06c, BIT(19)),
+
 	[CLK_USB_OHCI]		= GATE(0x0cc, BIT(6)),
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
drivers/clk/sunxi/clk_a23.c +6 -0
index 0fc8a5821c..63004b881d 100644
@@ -17,6 +17,12 @@ static struct ccu_clk_gate a23_gates[] = {
 	[CLK_BUS_EHCI]		= GATE(0x060, BIT(26)),
 	[CLK_BUS_OHCI]		= GATE(0x060, BIT(29)),
 
+	[CLK_BUS_UART0]		= GATE(0x06c, BIT(16)),
+	[CLK_BUS_UART1]		= GATE(0x06c, BIT(17)),
+	[CLK_BUS_UART2]		= GATE(0x06c, BIT(18)),
+	[CLK_BUS_UART3]		= GATE(0x06c, BIT(19)),
+	[CLK_BUS_UART4]		= GATE(0x06c, BIT(20)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_HSIC]		= GATE(0x0cc, BIT(10)),
drivers/clk/sunxi/clk_a31.c +7 -0
index 03dd9ae913..c01d2a3549 100644
@@ -20,6 +20,13 @@ static struct ccu_clk_gate a31_gates[] = {
 	[CLK_AHB1_OHCI1]	= GATE(0x060, BIT(30)),
 	[CLK_AHB1_OHCI2]	= GATE(0x060, BIT(31)),
 
+	[CLK_APB2_UART0]	= GATE(0x06c, BIT(16)),
+	[CLK_APB2_UART1]	= GATE(0x06c, BIT(17)),
+	[CLK_APB2_UART2]	= GATE(0x06c, BIT(18)),
+	[CLK_APB2_UART3]	= GATE(0x06c, BIT(19)),
+	[CLK_APB2_UART4]	= GATE(0x06c, BIT(20)),
+	[CLK_APB2_UART5]	= GATE(0x06c, BIT(21)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_PHY2]		= GATE(0x0cc, BIT(10)),
drivers/clk/sunxi/clk_a64.c +6 -0
index 28bda1f497..b427e3adda 100644
@@ -19,6 +19,12 @@ static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_BUS_OHCI0]		= GATE(0x060, BIT(28)),
 	[CLK_BUS_OHCI1]		= GATE(0x060, BIT(29)),
 
+	[CLK_BUS_UART0]		= GATE(0x06c, BIT(16)),
+	[CLK_BUS_UART1]		= GATE(0x06c, BIT(17)),
+	[CLK_BUS_UART2]		= GATE(0x06c, BIT(18)),
+	[CLK_BUS_UART3]		= GATE(0x06c, BIT(19)),
+	[CLK_BUS_UART4]		= GATE(0x06c, BIT(20)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_HSIC]		= GATE(0x0cc, BIT(10)),
drivers/clk/sunxi/clk_a83t.c +6 -0
index 45faef2f11..fbffa3bf5f 100644
@@ -18,6 +18,12 @@ static struct ccu_clk_gate a83t_gates[] = {
 	[CLK_BUS_EHCI1]		= GATE(0x060, BIT(27)),
 	[CLK_BUS_OHCI0]		= GATE(0x060, BIT(29)),
 
+	[CLK_BUS_UART0]		= GATE(0x06c, BIT(16)),
+	[CLK_BUS_UART1]		= GATE(0x06c, BIT(17)),
+	[CLK_BUS_UART2]		= GATE(0x06c, BIT(18)),
+	[CLK_BUS_UART3]		= GATE(0x06c, BIT(19)),
+	[CLK_BUS_UART4]		= GATE(0x06c, BIT(20)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_HSIC]		= GATE(0x0cc, BIT(10)),
drivers/clk/sunxi/clk_h3.c +5 -0
index 283fc31b01..f706e77dee 100644
@@ -23,6 +23,11 @@ static struct ccu_clk_gate h3_gates[] = {
 	[CLK_BUS_OHCI2]		= GATE(0x060, BIT(30)),
 	[CLK_BUS_OHCI3]		= GATE(0x060, BIT(31)),
 
+	[CLK_BUS_UART0]		= GATE(0x06c, BIT(16)),
+	[CLK_BUS_UART1]		= GATE(0x06c, BIT(17)),
+	[CLK_BUS_UART2]		= GATE(0x06c, BIT(18)),
+	[CLK_BUS_UART3]		= GATE(0x06c, BIT(19)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_PHY2]		= GATE(0x0cc, BIT(10)),
drivers/clk/sunxi/clk_r40.c +9 -0
index 2a6000f4dc..3405b39c46 100644
@@ -21,6 +21,15 @@ static struct ccu_clk_gate r40_gates[] = {
 	[CLK_BUS_OHCI1]		= GATE(0x060, BIT(30)),
 	[CLK_BUS_OHCI2]		= GATE(0x060, BIT(31)),
 
+	[CLK_BUS_UART0]		= GATE(0x06c, BIT(16)),
+	[CLK_BUS_UART1]		= GATE(0x06c, BIT(17)),
+	[CLK_BUS_UART2]		= GATE(0x06c, BIT(18)),
+	[CLK_BUS_UART3]		= GATE(0x06c, BIT(19)),
+	[CLK_BUS_UART4]		= GATE(0x06c, BIT(20)),
+	[CLK_BUS_UART5]		= GATE(0x06c, BIT(21)),
+	[CLK_BUS_UART6]		= GATE(0x06c, BIT(22)),
+	[CLK_BUS_UART7]		= GATE(0x06c, BIT(23)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_PHY2]		= GATE(0x0cc, BIT(10)),
drivers/clk/sunxi/clk_v3s.c +4 -0
index 0eeea84c1a..a662bc3e93 100644
@@ -15,6 +15,10 @@
 static struct ccu_clk_gate v3s_gates[] = {
 	[CLK_BUS_OTG]		= GATE(0x060, BIT(24)),
 
+	[CLK_BUS_UART0]		= GATE(0x06c, BIT(16)),
+	[CLK_BUS_UART1]		= GATE(0x06c, BIT(17)),
+	[CLK_BUS_UART2]		= GATE(0x06c, BIT(18)),
+
 	[CLK_USB_PHY0]          = GATE(0x0cc, BIT(8)),
 };
 
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#13 lore raw

PATCH v5 12/26 clk: sunxi: Implement UART resets

In reply to Jagan Teki

Implement UART resets for all relevant Allwinner SoC clock drivers via ccu reset table.

---
 drivers/clk/sunxi/clk_a23.c  | 6 ++++++
 drivers/clk/sunxi/clk_a31.c  | 7 +++++++
 drivers/clk/sunxi/clk_a64.c  | 6 ++++++
 drivers/clk/sunxi/clk_a83t.c | 6 ++++++
 drivers/clk/sunxi/clk_h3.c   | 5 +++++
 drivers/clk/sunxi/clk_r40.c  | 9 +++++++++
 drivers/clk/sunxi/clk_v3s.c  | 4 ++++
 7 files changed, 43 insertions(+)
drivers/clk/sunxi/clk_a23.c +6 -0
index 63004b881d..4661a33191 100644
@@ -38,6 +38,12 @@ static struct ccu_reset a23_resets[] = {
 	[RST_BUS_OTG]		= RESET(0x2c0, BIT(24)),
 	[RST_BUS_EHCI]		= RESET(0x2c0, BIT(26)),
 	[RST_BUS_OHCI]		= RESET(0x2c0, BIT(29)),
+
+	[RST_BUS_UART0]		= RESET(0x2d8, BIT(16)),
+	[RST_BUS_UART1]		= RESET(0x2d8, BIT(17)),
+	[RST_BUS_UART2]		= RESET(0x2d8, BIT(18)),
+	[RST_BUS_UART3]		= RESET(0x2d8, BIT(19)),
+	[RST_BUS_UART4]		= RESET(0x2d8, BIT(20)),
 };
 
 static const struct ccu_desc a23_ccu_desc = {
drivers/clk/sunxi/clk_a31.c +7 -0
index c01d2a3549..b05bea5b82 100644
@@ -46,6 +46,13 @@ static struct ccu_reset a31_resets[] = {
 	[RST_AHB1_OHCI0]	= RESET(0x2c0, BIT(29)),
 	[RST_AHB1_OHCI1]	= RESET(0x2c0, BIT(30)),
 	[RST_AHB1_OHCI2]	= RESET(0x2c0, BIT(31)),
+
+	[RST_APB2_UART0]	= RESET(0x2d8, BIT(16)),
+	[RST_APB2_UART1]	= RESET(0x2d8, BIT(17)),
+	[RST_APB2_UART2]	= RESET(0x2d8, BIT(18)),
+	[RST_APB2_UART3]	= RESET(0x2d8, BIT(19)),
+	[RST_APB2_UART4]	= RESET(0x2d8, BIT(20)),
+	[RST_APB2_UART5]	= RESET(0x2d8, BIT(21)),
 };
 
 static const struct ccu_desc a31_ccu_desc = {
drivers/clk/sunxi/clk_a64.c +6 -0
index b427e3adda..162ec769d6 100644
@@ -43,6 +43,12 @@ static const struct ccu_reset a64_resets[] = {
 	[RST_BUS_EHCI1]         = RESET(0x2c0, BIT(25)),
 	[RST_BUS_OHCI0]         = RESET(0x2c0, BIT(28)),
 	[RST_BUS_OHCI1]         = RESET(0x2c0, BIT(29)),
+
+	[RST_BUS_UART0]		= RESET(0x2d8, BIT(16)),
+	[RST_BUS_UART1]		= RESET(0x2d8, BIT(17)),
+	[RST_BUS_UART2]		= RESET(0x2d8, BIT(18)),
+	[RST_BUS_UART3]		= RESET(0x2d8, BIT(19)),
+	[RST_BUS_UART4]		= RESET(0x2d8, BIT(20)),
 };
 
 static const struct ccu_desc a64_ccu_desc = {
drivers/clk/sunxi/clk_a83t.c +6 -0
index fbffa3bf5f..3fe96ccd84 100644
@@ -40,6 +40,12 @@ static struct ccu_reset a83t_resets[] = {
 	[RST_BUS_EHCI0]		= RESET(0x2c0, BIT(26)),
 	[RST_BUS_EHCI1]		= RESET(0x2c0, BIT(27)),
 	[RST_BUS_OHCI0]		= RESET(0x2c0, BIT(29)),
+
+	[RST_BUS_UART0]		= RESET(0x2d8, BIT(16)),
+	[RST_BUS_UART1]		= RESET(0x2d8, BIT(17)),
+	[RST_BUS_UART2]		= RESET(0x2d8, BIT(18)),
+	[RST_BUS_UART3]		= RESET(0x2d8, BIT(19)),
+	[RST_BUS_UART4]		= RESET(0x2d8, BIT(20)),
 };
 
 static const struct ccu_desc a83t_ccu_desc = {
drivers/clk/sunxi/clk_h3.c +5 -0
index f706e77dee..89f59dbea3 100644
@@ -53,6 +53,11 @@ static struct ccu_reset h3_resets[] = {
 	[RST_BUS_OHCI1]		= RESET(0x2c0, BIT(29)),
 	[RST_BUS_OHCI2]		= RESET(0x2c0, BIT(30)),
 	[RST_BUS_OHCI3]		= RESET(0x2c0, BIT(31)),
+
+	[RST_BUS_UART0]		= RESET(0x2d8, BIT(16)),
+	[RST_BUS_UART1]		= RESET(0x2d8, BIT(17)),
+	[RST_BUS_UART2]		= RESET(0x2d8, BIT(18)),
+	[RST_BUS_UART3]		= RESET(0x2d8, BIT(19)),
 };
 
 static const struct ccu_desc h3_ccu_desc = {
drivers/clk/sunxi/clk_r40.c +9 -0
index 3405b39c46..a95017f12e 100644
@@ -50,6 +50,15 @@ static struct ccu_reset r40_resets[] = {
 	[RST_BUS_OHCI0]		= RESET(0x2c0, BIT(29)),
 	[RST_BUS_OHCI1]		= RESET(0x2c0, BIT(30)),
 	[RST_BUS_OHCI2]		= RESET(0x2c0, BIT(31)),
+
+	[RST_BUS_UART0]		= RESET(0x2d8, BIT(16)),
+	[RST_BUS_UART1]		= RESET(0x2d8, BIT(17)),
+	[RST_BUS_UART2]		= RESET(0x2d8, BIT(18)),
+	[RST_BUS_UART3]		= RESET(0x2d8, BIT(19)),
+	[RST_BUS_UART4]		= RESET(0x2d8, BIT(20)),
+	[RST_BUS_UART5]		= RESET(0x2d8, BIT(21)),
+	[RST_BUS_UART6]		= RESET(0x2d8, BIT(22)),
+	[RST_BUS_UART7]		= RESET(0x2d8, BIT(23)),
 };
 
 static const struct ccu_desc r40_ccu_desc = {
drivers/clk/sunxi/clk_v3s.c +4 -0
index a662bc3e93..867b029172 100644
@@ -26,6 +26,10 @@ static struct ccu_reset v3s_resets[] = {
 	[RST_USB_PHY0]		= RESET(0x0cc, BIT(0)),
 
 	[RST_BUS_OTG]		= RESET(0x2c0, BIT(24)),
+
+	[RST_BUS_UART0]		= RESET(0x2d8, BIT(16)),
+	[RST_BUS_UART1]		= RESET(0x2d8, BIT(17)),
+	[RST_BUS_UART2]		= RESET(0x2d8, BIT(18)),
 };
 
 static const struct ccu_desc v3s_ccu_desc = {
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#14 lore raw

PATCH v5 13/26 clk: sunxi: Add Allwinner H6 CLK driver

In reply to Jagan Teki

Add initial clock driver for Allwinner H6.

- Implement UART bus clocks via ccu_clk_gate table for H6, so it can accessed in common clk enable and disable functions from clk_sunxi.c
- Implement UART bus resets via ccu_reset table for H6, so it can accessed in common reset deassert and assert functions from reset-sunxi.c

---
 drivers/clk/sunxi/Kconfig  |  7 +++++
 drivers/clk/sunxi/Makefile |  1 +
 drivers/clk/sunxi/clk_h6.c | 53 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 61 insertions(+)
 create mode 100644 drivers/clk/sunxi/clk_h6.c
drivers/clk/sunxi/Kconfig +7 -0
index a6f84e9e56..cb11c7c21e 100644
@@ -65,6 +65,13 @@ config CLK_SUN8I_H3
 	  This enables common clock driver support for platforms based
 	  on Allwinner H3/H5 SoC.
 
+config CLK_SUN50I_H6
+	bool "Clock driver for Allwinner H6"
+	default MACH_SUN50I_H6
+	help
+	  This enables common clock driver support for platforms based
+	  on Allwinner H6 SoC.
+
 config CLK_SUN50I_A64
 	bool "Clock driver for Allwinner A64"
 	default MACH_SUN50I
drivers/clk/sunxi/Makefile +1 -0
index fbd43527a6..794aa2461c 100644
@@ -14,4 +14,5 @@ obj-$(CONFIG_CLK_SUN8I_A83T) += clk_a83t.o
 obj-$(CONFIG_CLK_SUN8I_R40) += clk_r40.o
 obj-$(CONFIG_CLK_SUN8I_V3S) += clk_v3s.o
 obj-$(CONFIG_CLK_SUN8I_H3) += clk_h3.o
+obj-$(CONFIG_CLK_SUN50I_H6) += clk_h6.o
 obj-$(CONFIG_CLK_SUN50I_A64) += clk_a64.o
drivers/clk/sunxi/clk_h6.c new file +53 -0

53 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#15 lore raw

PATCH v5 14/26 sunxi: A64: Update sun50i-a64-ccu.h

In reply to Jagan Teki

Update sun50i-a64-ccu.h from the Linux sunxi/dt64-for-4.20 tree:
commit 679294497be31596e1c9c61507746d72b6b05f26

Author: Rodrigo Exterckötter Tjäder <[email protected]>
Date:   Wed Sep 26 19:48:24 2018 +0000

arm64: dts: allwinner: a64: a64-olinuxino: set the PHY TX delay

This should be a part of previous sync patch from
commit 1b39a1834ed182bbd8036a5cd74a9ea111fa4691

Author: Andre Przywara <[email protected]>
Date:   Mon Oct 29 00:56:47 2018 +0000

sunxi: A64: Update .dts/.dtsi files

---
 include/dt-bindings/clock/sun50i-a64-ccu.h | 2 ++
 1 file changed, 2 insertions(+)
include/dt-bindings/clock/sun50i-a64-ccu.h +2 -0
index 370c0a0473..d66432c6e6 100644
@@ -43,6 +43,8 @@
 #ifndef _DT_BINDINGS_CLK_SUN50I_A64_H_
 #define _DT_BINDINGS_CLK_SUN50I_A64_H_
 
+#define CLK_PLL_PERIPH0		11
+
 #define CLK_BUS_MIPI_DSI	28
 #define CLK_BUS_CE		29
 #define CLK_BUS_DMA		30
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#16 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

In reply to Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192 +++++++++++++++++++++++++-
 drivers/clk/sunxi/clk_a64.c           |  40 ++++++
 drivers/clk/sunxi/clk_sunxi.c         | 182 ++++++++++++++++++++++++
 3 files changed, 413 insertions(+), 1 deletion(-)
arch/arm/include/asm/arch-sunxi/ccu.h +191 -1
index 3fdc26978d..61b8c36b3b 100644
@@ -7,15 +7,204 @@
 #ifndef _ASM_ARCH_CCU_H
 #define _ASM_ARCH_CCU_H
 
+#define OSC_32K_ULL		32000ULL
+#define OSC_24M_ULL		24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:			misc clock type
+ * @CCU_CLK_TYPE_FIXED:			fixed clock type
+ * @CCU_CLK_TYPE_MP:			mp clock type
+ * @CCU_CLK_TYPE_NK:			nk clock type
+ */
+enum ccu_clk_type {
+	CCU_CLK_TYPE_MISC		= 0,
+	CCU_CLK_TYPE_FIXED		= 1,
+	CCU_CLK_TYPE_MP			= 2,
+	CCU_CLK_TYPE_NK			= 3,
+};
+
 /**
  * enum ccu_clk_flags - ccu clock flags
  *
- * @CCU_CLK_F_INIT_DONE:		clock gate init done check
+ * @CCU_CLK_F_INIT_DONE:		clock tree/gate init done check
+ * @CCU_CLK_F_POSTDIV:			clock post divider
  */
 enum ccu_clk_flags {
 	CCU_CLK_F_INIT_DONE		= BIT(0),
+	CCU_CLK_F_POSTDIV		= BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:		multiplier shift value
+ * @width:		multiplier width value
+ * @offset:		multiplier offset
+ * @min:		minimum multiplier
+ * @max:		maximum multiplier
+ */
+struct ccu_mult {
+	u8 shift;
+	u8 width;
+	u8 offset;
+	u8 min;
+	u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width, _offset,		\
+			      _min, _max) {			\
+	.shift = _shift,					\
+	.width = _width,					\
+	.offset = _offset,					\
+	.min = _min,						\
+	.max = _max,						\
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)			\
+	_CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)				\
+	_CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:		multiplexer shift value
+ * @width:		multiplexer width value
+ */
+struct ccu_mux {
+	u8 shift;
+	u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {		\
+	.shift = _shift,			\
+	.width = _width,			\
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:		divider shift value
+ * @width:		divider width value
+ * @offset:		divider offset
+ * @max:		maximum divider value
+ */
+struct ccu_div {
+	u8 shift;
+	u8 width;
+	u32 offset;
+	u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {		\
+	.shift = _shift,			\
+	.width = _width,			\
+	.offset = 1,				\
+	.max = 0,				\
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:		parent clock tree
+ * @type:		clock type
+ * @off:		clock tree offset
+ * @m:			divider m
+ * @p:			divider p
+ * @mux:		multiplexer mux
+ * @post:		post divider value
+ * @n:			multiplier n
+ * @k:			multiplier k
+ * @fixed_rate:		fixed rate
+ * @flags:		clock tree flags
+ */
+struct ccu_clk_tree {
+	const unsigned long *parent;
+	enum ccu_clk_type type;
+	u16 off;
+
+	struct ccu_div m;
+	struct ccu_div p;
+	struct ccu_mux mux;
+	unsigned int postdiv;
+
+	struct ccu_mult n;
+	struct ccu_mult k;
+
+	ulong fixed_rate;
+	enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,				\
+	     _m, _p,						\
+	     _mux,						\
+	     _postdiv,						\
+	     _n, _k,						\
+	     _fixed_rate,					\
+	     _flags) {						\
+	.parent = _parent,					\
+	.type = _type,						\
+	.off = _off,						\
+	.m = _m,						\
+	.p = _p,						\
+	.mux = _mux,						\
+	.postdiv = _postdiv,					\
+	.n = _n,						\
+	.k = _k,						\
+	.fixed_rate = _fixed_rate,				\
+	.flags = _flags,					\
+}
+
+#define MISC(_parent)						\
+	TREE(_parent, CCU_CLK_TYPE_MISC, 0,			\
+	     {0}, {0},						\
+	     {0},						\
+	     0,							\
+	     {0}, {0},						\
+	     0,							\
+	     CCU_CLK_F_INIT_DONE)
+
+#define FIXED(_fixed_rate)					\
+	TREE(NULL, CCU_CLK_TYPE_FIXED, 0,			\
+	     {0}, {0},						\
+	     {0},						\
+	     0,							\
+	     {0}, {0},						\
+	     _fixed_rate,					\
+	     CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,					\
+	   _nshift, _nwidth,					\
+	   _kshift, _kwidth, _kmin,				\
+	   _postdiv,						\
+	   _flags)						\
+	TREE(_parent, CCU_CLK_TYPE_NK, _off,			\
+	     {0}, {0},						\
+	     {0},						\
+	     _postdiv,						\
+	     _CCU_MULT(_nshift, _nwidth),			\
+	     _CCU_MULT_MIN(_kshift, _kwidth, _kmin),		\
+	     0,							\
+	     CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,					\
+	   _mshift, _mwidth,					\
+	   _pshift, _pwidth,					\
+	   _muxshift, _muxwidth,				\
+	   _postdiv,						\
+	   _flags)						\
+	TREE(_parent, CCU_CLK_TYPE_MP, _off,			\
+	     _CCU_DIV(_mshift, _mwidth),			\
+	     _CCU_DIV(_pshift, _pwidth),			\
+	     _CCU_MUX(_muxshift, _muxwidth),			\
+	     _postdiv,						\
+	     {0}, {0},						\
+	     0,							\
+	     CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:	gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:	reset unit
  */
 struct ccu_desc {
+	const struct ccu_clk_tree *tree;
 	const struct ccu_clk_gate *gates;
 	const struct ccu_reset *resets;
 };
drivers/clk/sunxi/clk_a64.c +40 -0
index 162ec769d6..1d0cd98183 100644
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2			26
+#define CLK_OSC_32K			(CLK_GPU + 1)
+#define CLK_OSC_24M			(CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {
+	CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+	CLK_OSC_32K,
+	CLK_OSC_24M,
+	CLK_PLL_PERIPH0,
+	CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+	CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+	[CLK_OSC_32K]		= FIXED(OSC_32K_ULL),
+	[CLK_OSC_24M]		= FIXED(OSC_24M_ULL),
+
+	[CLK_PLL_PERIPH0]	= NK(periph0_parents, 0x028,
+				     8, 5,	/* N */
+				     4, 2, 2,	/* K */
+				     2,		/* post-div */
+				     CCU_CLK_F_POSTDIV),
+
+	[CLK_APB2]		= MP(apb2_parents, 0x058,
+				     0, 5,	/* M */
+				     16, 2,	/* P */
+				     24, 2,	/* mux */
+				     0,
+				     0),
+
+	[CLK_BUS_UART0]		= MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_BUS_OTG]		= GATE(0x060, BIT(23)),
 	[CLK_BUS_EHCI0]		= GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+	.tree = a64_tree,
 	.gates = a64_gates,
 	.resets = a64_resets,
 };
drivers/clk/sunxi/clk_sunxi.c +182 -0

194 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
#17 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

Quoted 6 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Quoted 2 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 22 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192 +++++++++++++++++++++++++-
 drivers/clk/sunxi/clk_a64.c           |  40 ++++++
 drivers/clk/sunxi/clk_sunxi.c         | 182 ++++++++++++++++++++++++
 3 files changed, 413 insertions(+), 1 deletion(-)
arch/arm/include/asm/arch-sunxi/ccu.h +0 -0
index 3fdc26978d..61b8c36b3b 100644

arch/arm/include/asm/arch-sunxi/ccu.h line 10

Quoted 4 lines
arch/arm/include/asm/arch-sunxi/ccu.h +1 -0
@@ -7,15 +7,204 @@
 #ifndef _ASM_ARCH_CCU_H
 #define _ASM_ARCH_CCU_H
 
+#define OSC_32K_ULL		32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL		24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:			misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

arch/arm/include/asm/arch-sunxi/ccu.h line 19

Quoted 3 lines
arch/arm/include/asm/arch-sunxi/ccu.h +3 -0
+ * @CCU_CLK_TYPE_FIXED:			fixed clock type
+ * @CCU_CLK_TYPE_MP:			mp clock type
+ * @CCU_CLK_TYPE_NK:			nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

arch/arm/include/asm/arch-sunxi/ccu.h line 31

Quoted 13 lines
arch/arm/include/asm/arch-sunxi/ccu.h +9 -1
+ */
+enum ccu_clk_type {
+	CCU_CLK_TYPE_MISC		= 0,
+	CCU_CLK_TYPE_FIXED		= 1,
+	CCU_CLK_TYPE_MP			= 2,
+	CCU_CLK_TYPE_NK			= 3,
+};
+
 /**
  * enum ccu_clk_flags - ccu clock flags
  *
- * @CCU_CLK_F_INIT_DONE:		clock gate init done check
+ * @CCU_CLK_F_INIT_DONE:		clock tree/gate init done check

Is this flag to tell implemented clocks apart from unimplemented ones, which would be reset to all zeroes by the compiler? Then it should be called something with VALID in it.

Quoted 84 lines
arch/arm/include/asm/arch-sunxi/ccu.h +79 -0
+ * @CCU_CLK_F_POSTDIV:			clock post divider
  */
 enum ccu_clk_flags {
 	CCU_CLK_F_INIT_DONE		= BIT(0),
+	CCU_CLK_F_POSTDIV		= BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:		multiplier shift value
+ * @width:		multiplier width value
+ * @offset:		multiplier offset
+ * @min:		minimum multiplier
+ * @max:		maximum multiplier
+ */
+struct ccu_mult {
+	u8 shift;
+	u8 width;
+	u8 offset;
+	u8 min;
+	u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width, _offset,		\
+			      _min, _max) {			\
+	.shift = _shift,					\
+	.width = _width,					\
+	.offset = _offset,					\
+	.min = _min,						\
+	.max = _max,						\
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)			\
+	_CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)				\
+	_CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:		multiplexer shift value
+ * @width:		multiplexer width value
+ */
+struct ccu_mux {
+	u8 shift;
+	u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {		\
+	.shift = _shift,			\
+	.width = _width,			\
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:		divider shift value
+ * @width:		divider width value
+ * @offset:		divider offset
+ * @max:		maximum divider value
+ */
+struct ccu_div {
+	u8 shift;
+	u8 width;
+	u32 offset;
+	u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {		\
+	.shift = _shift,			\
+	.width = _width,			\
+	.offset = 1,				\
+	.max = 0,				\
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:		parent clock tree
+ * @type:		clock type
+ * @off:		clock tree offset
+ * @m:			divider m

arch/arm/include/asm/arch-sunxi/ccu.h line 125

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+ * @p:			divider p
+ * @mux:		multiplexer mux
+ * @post:		post divider value
+ * @n:			multiplier n
+ * @k:			multiplier k
+ * @fixed_rate:		fixed rate
+ * @flags:		clock tree flags
+ */
+struct ccu_clk_tree {
+	const unsigned long *parent;

Shouldn't that be called "parents" instead?

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+	enum ccu_clk_type type;
+	u16 off;
+
+	struct ccu_div m;
+	struct ccu_div p;
+	struct ccu_mux mux;

arch/arm/include/asm/arch-sunxi/ccu.h line 141

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+	unsigned int postdiv;
+
+	struct ccu_mult n;
+	struct ccu_mult k;
+
+	ulong fixed_rate;
+	enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,				\

Just a nit, but TREE is somewhat confusing here, as this just constructs a single entry in an array. The tree property is realised through the parent array, if I get this correctly.
So should we name this ENTRY or CLK_ENTRY instead?

Quoted 17 lines
arch/arm/include/asm/arch-sunxi/ccu.h +17 -0
+	     _m, _p,						\
+	     _mux,						\
+	     _postdiv,						\
+	     _n, _k,						\
+	     _fixed_rate,					\
+	     _flags) {						\
+	.parent = _parent,					\
+	.type = _type,						\
+	.off = _off,						\
+	.m = _m,						\
+	.p = _p,						\
+	.mux = _mux,						\
+	.postdiv = _postdiv,					\
+	.n = _n,						\
+	.k = _k,						\
+	.fixed_rate = _fixed_rate,				\
+	.flags = _flags,					\

arch/arm/include/asm/arch-sunxi/ccu.h line 168

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+}
+
+#define MISC(_parent)						\
+	TREE(_parent, CCU_CLK_TYPE_MISC, 0,			\
+	     {0}, {0},						\
+	     {0},						\
+	     0,							\
+	     {0}, {0},						\
+	     0,							\
+	     CCU_CLK_F_INIT_DONE)

If MISC is really something like GATE or SIMPLE, would it be possible to construct the single element array here, so that the macro just takes the one parent clock ID here, instead of referring to an extra array?

Quoted 55 lines
arch/arm/include/asm/arch-sunxi/ccu.h +40 -0
+
+#define FIXED(_fixed_rate)					\
+	TREE(NULL, CCU_CLK_TYPE_FIXED, 0,			\
+	     {0}, {0},						\
+	     {0},						\
+	     0,							\
+	     {0}, {0},						\
+	     _fixed_rate,					\
+	     CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,					\
+	   _nshift, _nwidth,					\
+	   _kshift, _kwidth, _kmin,				\
+	   _postdiv,						\
+	   _flags)						\
+	TREE(_parent, CCU_CLK_TYPE_NK, _off,			\
+	     {0}, {0},						\
+	     {0},						\
+	     _postdiv,						\
+	     _CCU_MULT(_nshift, _nwidth),			\
+	     _CCU_MULT_MIN(_kshift, _kwidth, _kmin),		\
+	     0,							\
+	     CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,					\
+	   _mshift, _mwidth,					\
+	   _pshift, _pwidth,					\
+	   _muxshift, _muxwidth,				\
+	   _postdiv,						\
+	   _flags)						\
+	TREE(_parent, CCU_CLK_TYPE_MP, _off,			\
+	     _CCU_DIV(_mshift, _mwidth),			\
+	     _CCU_DIV(_pshift, _pwidth),			\
+	     _CCU_MUX(_muxshift, _muxwidth),			\
+	     _postdiv,						\
+	     {0}, {0},						\
+	     0,							\
+	     CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:	gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:	reset unit
  */
 struct ccu_desc {
+	const struct ccu_clk_tree *tree;
 	const struct ccu_clk_gate *gates;
 	const struct ccu_reset *resets;
 };
drivers/clk/sunxi/clk_a64.c +0 -0
index 162ec769d6..1d0cd98183 100644

drivers/clk/sunxi/clk_a64.c line 19

Quoted 8 lines
drivers/clk/sunxi/clk_a64.c +5 -0
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2			26
+#define CLK_OSC_32K			(CLK_GPU + 1)
+#define CLK_OSC_24M			(CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {

Why is this long? That's the DT clock index, which is 32 bit wide, right? Just unsigned int or uint32_t should be sufficient. long is quite a treacherous type to use, especially as we share code between 32 and 64-bit architectures.

Quoted 50 lines
drivers/clk/sunxi/clk_a64.c +35 -0
+	CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+	CLK_OSC_32K,
+	CLK_OSC_24M,
+	CLK_PLL_PERIPH0,
+	CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+	CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+	[CLK_OSC_32K]		= FIXED(OSC_32K_ULL),
+	[CLK_OSC_24M]		= FIXED(OSC_24M_ULL),
+
+	[CLK_PLL_PERIPH0]	= NK(periph0_parents, 0x028,
+				     8, 5,	/* N */
+				     4, 2, 2,	/* K */
+				     2,		/* post-div */
+				     CCU_CLK_F_POSTDIV),
+
+	[CLK_APB2]		= MP(apb2_parents, 0x058,
+				     0, 5,	/* M */
+				     16, 2,	/* P */
+				     24, 2,	/* mux */
+				     0,
+				     0),
+
+	[CLK_BUS_UART0]		= MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_BUS_OTG]		= GATE(0x060, BIT(23)),
 	[CLK_BUS_EHCI0]		= GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+	.tree = a64_tree,
 	.gates = a64_gates,
 	.resets = a64_resets,
 };
drivers/clk/sunxi/clk_sunxi.c +0 -0
index 345d706c2a..2aebd257d1 100644

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 22

Quoted 5 lines
drivers/clk/sunxi/clk_sunxi.c +2 -0
@@ -18,6 +18,187 @@ static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv,
 	return &priv->desc->gates[id];
 }
 
+static const struct ccu_clk_tree *priv_to_tree(struct ccu_priv *priv,
+					       unsigned long id)

Again, why long here? Especially as it's u32 in the function below.

Quoted 6 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+	return &priv->desc->tree[id];
+}
+
+static int sunxi_get_parent_idx(const struct ccu_clk_tree *tree, void *base)
+{

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 38

Quoted 10 lines
drivers/clk/sunxi/clk_sunxi.c +10 -0
+	u32 reg, idx;
+
+	reg = readl(base + tree->off);
+	idx = reg >> tree->mux.shift;
+	idx &= (1 << tree->mux.width) - 1;
+
+	return idx;
+}
+
+static ulong sunxi_fixed_get_rate(struct clk *clk, unsigned long id)

Same "long" question for both the return type and the id. And for everything below.

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 50

Quoted 12 lines
drivers/clk/sunxi/clk_sunxi.c +12 -0
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+
+	if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+		printf("%s: (CLK#%ld) unhandled\n", __func__, clk->id);
+		return 0;
+	}
+
+	return tree->fixed_rate;
+}
+

So why are there all those separate functions? Isn't that all the same
algorithm: adjust the parent rate based on the clock type? I reworked this with recursive calls and it's MUCH less code:

(fill the gaps, using your types for your convenience ;-)

static ulong sunxi_apply_pll(struct ccu_priv *priv, ulong id, ulong parent_rate) { NK algorithm of sunxi_nk_get_rate }
static ulong sunxi_apply_div(struct ccu_priv *priv, ulong id, ulong parent_rate) { MP algorithm of sunxi_mp_get_rate }
static ulong sunxi_calc_clk_rate(struct ccu_priv *priv, ulong clkid) {
....

	switch (tree->type) {
	case CCU_CLK_TYPE_MISC:
		return sunxi_calc_clk_rate(priv, tree->parent[0]);
	case CCU_CLK_TYPE_FIXED:
		return tree->fixed_rate;
	case CCU_CLK_TYPE_NK:
		rate = sunxi_calc_clk_rate(priv,
			sunxi_get_parent_id(tree, priv->base));
		return sunxi_apply_pll(priv, clkid, rate);
	(similar for _MP)

...
}

static ulong sunxi_clk_get_rate(struct clk *clk)

{
	struct ccu_priv *priv = dev_get_priv(clk->dev);

	return sunxi_calc_clk_rate(priv, clk->id);
}

This removes about 80 lines from that file.
Let me know if I should post my rework.

Cheers,
Andre.

Quoted 163 lines
drivers/clk/sunxi/clk_sunxi.c +152 -0
+static ulong sunxi_nk_get_parent_rate(struct clk *clk, unsigned long id)
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+	ulong rate = 0;
+
+	switch (tree->type) {
+	case CCU_CLK_TYPE_FIXED:
+		rate = sunxi_fixed_get_rate(clk, id);
+		break;
+	default:
+		printf("%s: Unknown (TYPE#%d)\n", __func__, tree->type);
+		break;
+	}
+
+	return rate;
+}
+
+static ulong sunxi_nk_get_rate(struct clk *clk, unsigned long id)
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+	ulong rate, parent_rate;
+	unsigned long n, k;
+	u32 reg;
+
+	parent_rate = sunxi_nk_get_parent_rate(clk, tree->parent[0]);
+
+	reg = readl(priv->base + tree->off);
+
+	n = reg >> tree->n.shift;
+	n &= (1 << tree->n.width) - 1;
+	n += tree->n.offset;
+	if (!n)
+		n++;
+
+	k = reg >> tree->k.shift;
+	k &= (1 << tree->k.width) - 1;
+	k += tree->k.offset;
+	if (!k)
+		k++;
+
+	rate = parent_rate * n * k;
+	if (tree->flags & CCU_CLK_F_POSTDIV)
+		rate /= tree->postdiv;
+
+	return rate;
+}
+
+static ulong sunxi_mp_get_parent_rate(struct clk *clk, unsigned long id)
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+	ulong rate = 0;
+
+	if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+		printf("%s: (CLK#%ld) unhandled\n", __func__, clk->id);
+		return 0;
+	}
+
+	switch (tree->type) {
+	case CCU_CLK_TYPE_FIXED:
+		rate = sunxi_fixed_get_rate(clk, id);
+		break;
+	case CCU_CLK_TYPE_NK:
+		rate = sunxi_nk_get_rate(clk, id);
+		break;
+	default:
+		printf("%s: (TYPE#%d) unhandled\n", __func__, tree->type);
+		break;
+	}
+
+	return rate;
+}
+
+static ulong sunxi_mp_get_rate(struct clk *clk, unsigned long id)
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+	unsigned int m, p;
+	ulong parent_rate;
+	u32 reg, idx;
+
+	idx = sunxi_get_parent_idx(tree, priv->base);
+	if (idx < 0) {
+		printf("%s: Wrong parent index %d\n", __func__, idx);
+		return 0;
+	}
+
+	parent_rate = sunxi_mp_get_parent_rate(clk, tree->parent[idx]);
+
+	reg = readl(priv->base + tree->off);
+
+	m = reg >> tree->m.shift;
+	m &= (1 << tree->m.width) - 1;
+	m += tree->m.offset;
+	if (!m)
+		m++;
+
+	p = reg >> tree->p.shift;
+	p &= (1 << tree->p.width) - 1;
+
+	return (parent_rate >> p) / m;
+}
+
+static ulong sunxi_misc_get_rate(struct clk *clk, unsigned long id)
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+	ulong rate = 0;
+
+	if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+		printf("%s: (CLK#%ld) unhandled\n", __func__, clk->id);
+		return 0;
+	}
+
+	switch (tree->type) {
+	case CCU_CLK_TYPE_MP:
+		rate = sunxi_mp_get_rate(clk, id);
+		break;
+	default:
+		printf("%s: (TYPE#%d) unhandled\n", __func__, tree->type);
+		break;
+	}
+
+	return rate;
+}
+
+static ulong sunxi_clk_get_rate(struct clk *clk)
+{
+	struct ccu_priv *priv = dev_get_priv(clk->dev);
+	const struct ccu_clk_tree *tree = priv_to_tree(priv, clk->id);
+	ulong rate = 0;
+
+	if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+		printf("%s: (CLK#%ld) unhandled\n", __func__, clk->id);
+		return 0;
+	}
+
+	switch (tree->type) {
+	case CCU_CLK_TYPE_MISC:
+		rate = sunxi_misc_get_rate(clk, tree->parent[0]);
+		break;
+	default:
+		printf("%s: (TYPE#%d) unhandled\n", __func__, tree->type);
+		break;
+	}
+
+	return rate;
+}
+
 static int sunxi_set_gate(struct clk *clk, bool on)
 {
 	struct ccu_priv *priv = dev_get_priv(clk->dev);
@@ -56,6 +237,7 @@ static int sunxi_clk_disable(struct clk *clk)
 struct clk_ops sunxi_clk_ops = {
 	.enable = sunxi_clk_enable,
 	.disable = sunxi_clk_disable,
+	.get_rate = sunxi_clk_get_rate,
 };
 
 int sunxi_clk_probe(struct udevice *dev)
#18 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

Quoted 68 lines · André Przywara
Quoted 6 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Quoted 2 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 22 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192 +++++++++++++++++++++++++-
 drivers/clk/sunxi/clk_a64.c           |  40 ++++++
 drivers/clk/sunxi/clk_sunxi.c         | 182 ++++++++++++++++++++++++
 3 files changed, 413 insertions(+), 1 deletion(-)
arch/arm/include/asm/arch-sunxi/ccu.h +0 -0
index 3fdc26978d..61b8c36b3b 100644

arch/arm/include/asm/arch-sunxi/ccu.h line 10

Quoted 4 lines
arch/arm/include/asm/arch-sunxi/ccu.h +1 -0
@@ -7,15 +7,204 @@
 #ifndef _ASM_ARCH_CCU_H
 #define _ASM_ARCH_CCU_H
 
+#define OSC_32K_ULL		32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL		24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:			misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 7 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 19

Quoted 3 lines
arch/arm/include/asm/arch-sunxi/ccu.h +3 -0
+ * @CCU_CLK_TYPE_FIXED:			fixed clock type
+ * @CCU_CLK_TYPE_MP:			mp clock type
+ * @CCU_CLK_TYPE_NK:			nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields

We have PLL with 2 multipliers, but also others with other factors sets, so that will end up being confusing. If the MP, NK and so on stuff is confusing, maybe we should just add a comment on top of that structure to explain what those factors are and what it actually means?

Maxime

Signature · 10 lines
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190107/26a42f3f/attachment.sig>
#19 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

On Mon, 7 Jan 2019 14:01:01 +0100
Maxime Ripard <[email protected]> wrote:

Hi,

Quoted 81 lines · André Przywara
Quoted 71 lines · Jagan Teki
Quoted 7 lines

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Quoted 3 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 19 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192

+++++++++++++++++++++++++- drivers/clk/sunxi/clk_a64.c
| 40 ++++++ drivers/clk/sunxi/clk_sunxi.c | 182
++++++++++++++++++++++++ 3 files changed, 413 insertions(+), 1 deletion(-)

(quoted diff) +0 -0
Quoted 8 lines

b/arch/arm/include/asm/arch-sunxi/ccu.h index 3fdc26978d..61b8c36b3b 100644 --- a/arch/arm/include/asm/arch-sunxi/ccu.h +++ b/arch/arm/include/asm/arch-sunxi/ccu.h @@ -7,15 +7,204 @@
#ifndef _ASM_ARCH_CCU_H
#define _ASM_ARCH_CCU_H

+#define OSC_32K_ULL 32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL		24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:			misc clock type  

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 8 lines
Quoted 3 lines

+ * @CCU_CLK_TYPE_FIXED: fixed clock type
+ * @CCU_CLK_TYPE_MP: mp clock type
+ * @CCU_CLK_TYPE_NK: nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier
fields

Quoted 5 lines

We have PLL with 2 multipliers, but also others with other factors sets, so that will end up being confusing. If the MP, NK and so on stuff is confusing, maybe we should just add a comment on top of that structure to explain what those factors are and what it actually means?

Fair enough, or we name it CCU_CLK_TYPE_PLL_NK, because this is what this type deals with. Point is that by chance I happened to know about those naming of factors in the manual, but other might be lost by just seeing "mp" or "nk", without any explanation - and the comment doesn't help here at all.

The other part is that the "TYPE_MP" is twice as confusing, as it can perfectly describe the MMC clocks, which use "N" and "M" in the A64 manual, for instance. That's why my suggestion for calling a spade a spade and saying it's a divider clock with a multiplexer. Happy to have the Linux naming in the comments.

Thanks,
Andre.

#20 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

Quoted 19 lines · Andre Przywara
Quoted 13 lines
Quoted 4 lines

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 8 lines
Quoted 3 lines

+ * @CCU_CLK_TYPE_FIXED: fixed clock type
+ * @CCU_CLK_TYPE_MP: mp clock type
+ * @CCU_CLK_TYPE_NK: nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier
fields

Quoted 5 lines

We have PLL with 2 multipliers, but also others with other factors sets, so that will end up being confusing. If the MP, NK and so on stuff is confusing, maybe we should just add a comment on top of that structure to explain what those factors are and what it actually means?

Quoted 5 lines

Fair enough, or we name it CCU_CLK_TYPE_PLL_NK, because this is what this type deals with. Point is that by chance I happened to know about those naming of factors in the manual, but other might be lost by just seeing "mp" or "nk", without any explanation - and the comment doesn't help here at all.

Either way, we should really document this properly.

Quoted 5 lines

The other part is that the "TYPE_MP" is twice as confusing, as it can perfectly describe the MMC clocks, which use "N" and "M" in the A64 manual, for instance. That's why my suggestion for calling a spade a spade and saying it's a divider clock with a multiplexer. Happy to have the Linux naming in the comments.

NM and MP aren't really the same though. NM is one multiplier and one divider, while MP is one divider and one right shift.

Maxime

Signature · 10 lines
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190107/6e86550c/attachment.sig>
author
#21 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

In reply to André Przywara
Quoted 9 lines · André Przywara
Quoted 6 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

Quoted 7 lines

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Exactly. Idea is to keep the macro's as simple as possible.

Adding gates clocks separately make easy and reasonable way to enable/disable clock operations. We even operate with single macro with all clock attributes along with gate(either another member or common structure like Linux does), but that seems not simple as per as my experince since there are many IP's like USB's just need enable/disable.

Quoted 42 lines
Quoted 3 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 22 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192 +++++++++++++++++++++++++-
 drivers/clk/sunxi/clk_a64.c           |  40 ++++++
 drivers/clk/sunxi/clk_sunxi.c         | 182 ++++++++++++++++++++++++
 3 files changed, 413 insertions(+), 1 deletion(-)
arch/arm/include/asm/arch-sunxi/ccu.h +0 -0
index 3fdc26978d..61b8c36b3b 100644

arch/arm/include/asm/arch-sunxi/ccu.h line 10

Quoted 4 lines
arch/arm/include/asm/arch-sunxi/ccu.h +1 -0
@@ -7,15 +7,204 @@
 #ifndef _ASM_ARCH_CCU_H
 #define _ASM_ARCH_CCU_H
 
+#define OSC_32K_ULL          32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

Quoted 10 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL          24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:                       misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Unlike MP type in MMC, UART doesn't need any clock attributes like dividers, mux, etc, just have attached parent. I don't think UART clock is simple one It has parent, that indeed have another parents and so...on, ie reason I named as MISC.

Quoted 21 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 19

Quoted 4 lines
arch/arm/include/asm/arch-sunxi/ccu.h +3 -0
+ * @CCU_CLK_TYPE_FIXED:                      fixed clock type
+ * @CCU_CLK_TYPE_MP:                 mp clock type
+ * @CCU_CLK_TYPE_NK:                 nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

Quoted 9 lines
arch/arm/include/asm/arch-sunxi/ccu.h +8 -0
+ */
+enum ccu_clk_type {
+     CCU_CLK_TYPE_MISC               = 0,
+     CCU_CLK_TYPE_FIXED              = 1,
+     CCU_CLK_TYPE_MP                 = 2,
+     CCU_CLK_TYPE_NK                 = 3,
+};
+
 /**
Quoted 8 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 31

Quoted 4 lines
arch/arm/include/asm/arch-sunxi/ccu.h +1 -1
  * enum ccu_clk_flags - ccu clock flags
  *
- * @CCU_CLK_F_INIT_DONE:             clock gate init done check
+ * @CCU_CLK_F_INIT_DONE:             clock tree/gate init done check

Is this flag to tell implemented clocks apart from unimplemented ones, which would be reset to all zeroes by the compiler? Then it should be called something with VALID in it.

Since the flags attached with real numeric initialized by macro itself rather than some other source like complier or any, so marked INIT_DONE seems to meaningful.
and eventually the same verified in code whether the init done or not.

Quoted 335 lines
Quoted 85 lines
arch/arm/include/asm/arch-sunxi/ccu.h +79 -0
+ * @CCU_CLK_F_POSTDIV:                       clock post divider
  */
 enum ccu_clk_flags {
      CCU_CLK_F_INIT_DONE             = BIT(0),
+     CCU_CLK_F_POSTDIV               = BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:           multiplier shift value
+ * @width:           multiplier width value
+ * @offset:          multiplier offset
+ * @min:             minimum multiplier
+ * @max:             maximum multiplier
+ */
+struct ccu_mult {
+     u8 shift;
+     u8 width;
+     u8 offset;
+     u8 min;
+     u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width, _offset,               \
+                           _min, _max) {                     \
+     .shift = _shift,                                        \
+     .width = _width,                                        \
+     .offset = _offset,                                      \
+     .min = _min,                                            \
+     .max = _max,                                            \
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)                  \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)                            \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:           multiplexer shift value
+ * @width:           multiplexer width value
+ */
+struct ccu_mux {
+     u8 shift;
+     u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:           divider shift value
+ * @width:           divider width value
+ * @offset:          divider offset
+ * @max:             maximum divider value
+ */
+struct ccu_div {
+     u8 shift;
+     u8 width;
+     u32 offset;
+     u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+     .offset = 1,                            \
+     .max = 0,                               \
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:          parent clock tree
+ * @type:            clock type
+ * @off:             clock tree offset
+ * @m:                       divider m

arch/arm/include/asm/arch-sunxi/ccu.h line 125

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+ * @p:                       divider p
+ * @mux:             multiplexer mux
+ * @post:            post divider value
+ * @n:                       multiplier n
+ * @k:                       multiplier k
+ * @fixed_rate:              fixed rate
+ * @flags:           clock tree flags
+ */
+struct ccu_clk_tree {
+     const unsigned long *parent;

Shouldn't that be called "parents" instead?

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+     enum ccu_clk_type type;
+     u16 off;
+
+     struct ccu_div m;
+     struct ccu_div p;
+     struct ccu_mux mux;

arch/arm/include/asm/arch-sunxi/ccu.h line 141

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+     unsigned int postdiv;
+
+     struct ccu_mult n;
+     struct ccu_mult k;
+
+     ulong fixed_rate;
+     enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,                           \

Just a nit, but TREE is somewhat confusing here, as this just constructs a single entry in an array. The tree property is realised through the parent array, if I get this correctly.
So should we name this ENTRY or CLK_ENTRY instead?

Quoted 17 lines
arch/arm/include/asm/arch-sunxi/ccu.h +17 -0
+          _m, _p,                                            \
+          _mux,                                              \
+          _postdiv,                                          \
+          _n, _k,                                            \
+          _fixed_rate,                                       \
+          _flags) {                                          \
+     .parent = _parent,                                      \
+     .type = _type,                                          \
+     .off = _off,                                            \
+     .m = _m,                                                \
+     .p = _p,                                                \
+     .mux = _mux,                                            \
+     .postdiv = _postdiv,                                    \
+     .n = _n,                                                \
+     .k = _k,                                                \
+     .fixed_rate = _fixed_rate,                              \
+     .flags = _flags,                                        \

arch/arm/include/asm/arch-sunxi/ccu.h line 168

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+}
+
+#define MISC(_parent)                                                \
+     TREE(_parent, CCU_CLK_TYPE_MISC, 0,                     \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE)

If MISC is really something like GATE or SIMPLE, would it be possible to construct the single element array here, so that the macro just takes the one parent clock ID here, instead of referring to an extra array?

Quoted 55 lines
arch/arm/include/asm/arch-sunxi/ccu.h +40 -0
+
+#define FIXED(_fixed_rate)                                   \
+     TREE(NULL, CCU_CLK_TYPE_FIXED, 0,                       \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          _fixed_rate,                                       \
+          CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,                                    \
+        _nshift, _nwidth,                                    \
+        _kshift, _kwidth, _kmin,                             \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_NK, _off,                    \
+          {0}, {0},                                          \
+          {0},                                               \
+          _postdiv,                                          \
+          _CCU_MULT(_nshift, _nwidth),                       \
+          _CCU_MULT_MIN(_kshift, _kwidth, _kmin),            \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,                                    \
+        _mshift, _mwidth,                                    \
+        _pshift, _pwidth,                                    \
+        _muxshift, _muxwidth,                                \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_MP, _off,                    \
+          _CCU_DIV(_mshift, _mwidth),                        \
+          _CCU_DIV(_pshift, _pwidth),                        \
+          _CCU_MUX(_muxshift, _muxwidth),                    \
+          _postdiv,                                          \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:     gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:  reset unit
  */
 struct ccu_desc {
+     const struct ccu_clk_tree *tree;
      const struct ccu_clk_gate *gates;
      const struct ccu_reset *resets;
 };
drivers/clk/sunxi/clk_a64.c +0 -0
index 162ec769d6..1d0cd98183 100644

drivers/clk/sunxi/clk_a64.c line 19

Quoted 8 lines
drivers/clk/sunxi/clk_a64.c +5 -0
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2                     26
+#define CLK_OSC_32K                  (CLK_GPU + 1)
+#define CLK_OSC_24M                  (CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {

Why is this long? That's the DT clock index, which is 32 bit wide, right? Just unsigned int or uint32_t should be sufficient. long is quite a treacherous type to use, especially as we share code between 32 and 64-bit architectures.

Quoted 50 lines
drivers/clk/sunxi/clk_a64.c +35 -0
+     CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+     CLK_OSC_32K,
+     CLK_OSC_24M,
+     CLK_PLL_PERIPH0,
+     CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+     CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+     [CLK_OSC_32K]           = FIXED(OSC_32K_ULL),
+     [CLK_OSC_24M]           = FIXED(OSC_24M_ULL),
+
+     [CLK_PLL_PERIPH0]       = NK(periph0_parents, 0x028,
+                                  8, 5,      /* N */
+                                  4, 2, 2,   /* K */
+                                  2,         /* post-div */
+                                  CCU_CLK_F_POSTDIV),
+
+     [CLK_APB2]              = MP(apb2_parents, 0x058,
+                                  0, 5,      /* M */
+                                  16, 2,     /* P */
+                                  24, 2,     /* mux */
+                                  0,
+                                  0),
+
+     [CLK_BUS_UART0]         = MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
      [CLK_BUS_OTG]           = GATE(0x060, BIT(23)),
      [CLK_BUS_EHCI0]         = GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+     .tree = a64_tree,
      .gates = a64_gates,
      .resets = a64_resets,
 };
drivers/clk/sunxi/clk_sunxi.c +0 -0
index 345d706c2a..2aebd257d1 100644

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 22

Quoted 5 lines
drivers/clk/sunxi/clk_sunxi.c +2 -0
@@ -18,6 +18,187 @@ static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv,
      return &priv->desc->gates[id];
 }
 
+static const struct ccu_clk_tree *priv_to_tree(struct ccu_priv *priv,
+                                            unsigned long id)

Again, why long here? Especially as it's u32 in the function below.

Quoted 6 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     return &priv->desc->tree[id];
+}
+
+static int sunxi_get_parent_idx(const struct ccu_clk_tree *tree, void *base)
+{

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 38

Quoted 10 lines
drivers/clk/sunxi/clk_sunxi.c +10 -0
+     u32 reg, idx;
+
+     reg = readl(base + tree->off);
+     idx = reg >> tree->mux.shift;
+     idx &= (1 << tree->mux.width) - 1;
+
+     return idx;
+}
+
+static ulong sunxi_fixed_get_rate(struct clk *clk, unsigned long id)

Same "long" question for both the return type and the id. And for everything below.

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 50

Quoted 12 lines
drivers/clk/sunxi/clk_sunxi.c +12 -0
+{
+     struct ccu_priv *priv = dev_get_priv(clk->dev);
+     const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+
+     if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+             printf("%s: (CLK#%ld) unhandled\n", __func__, clk->id);
+             return 0;
+     }
+
+     return tree->fixed_rate;
+}
+

So why are there all those separate functions? Isn't that all the same
algorithm: adjust the parent rate based on the clock type? I reworked this with recursive calls and it's MUCH less code:

(fill the gaps, using your types for your convenience ;-)

static ulong sunxi_apply_pll(struct ccu_priv *priv, ulong id, ulong parent_rate) { NK algorithm of sunxi_nk_get_rate }
static ulong sunxi_apply_div(struct ccu_priv *priv, ulong id, ulong parent_rate) { MP algorithm of sunxi_mp_get_rate }
static ulong sunxi_calc_clk_rate(struct ccu_priv *priv, ulong clkid) {
....

        switch (tree->type) {
        case CCU_CLK_TYPE_MISC:
                return sunxi_calc_clk_rate(priv, tree->parent[0]);
        case CCU_CLK_TYPE_FIXED:
                return tree->fixed_rate;
        case CCU_CLK_TYPE_NK:
                rate = sunxi_calc_clk_rate(priv,
                        sunxi_get_parent_id(tree, priv->base));
                return sunxi_apply_pll(priv, clkid, rate);
        (similar for _MP)
Quoted 2 lines

...
}

Initially I would tried the recursive and yes code can reduce but using recursive can leed more disadvantage in-terms of code tracing during long run. Due to all these factors I used simple function calls.

#22 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

On Tue, 8 Jan 2019 16:27:14 +0530
Jagan Teki <[email protected]> wrote:

Hi,

Quoted 86 lines
Quoted 10 lines · André Przywara
Quoted 7 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

Quoted 8 lines

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Exactly. Idea is to keep the macro's as simple as possible.

Adding gates clocks separately make easy and reasonable way to enable/disable clock operations. We even operate with single macro with all clock attributes along with gate(either another member or common structure like Linux does), but that seems not simple as per as my experince since there are many IP's like USB's just need enable/disable.

Quoted 44 lines
Quoted 4 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 19 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192

+++++++++++++++++++++++++- drivers/clk/sunxi/clk_a64.c
| 40 ++++++ drivers/clk/sunxi/clk_sunxi.c | 182
++++++++++++++++++++++++ 3 files changed, 413 insertions(+), 1 deletion(-)

(quoted diff) +0 -0
Quoted 8 lines

b/arch/arm/include/asm/arch-sunxi/ccu.h index 3fdc26978d..61b8c36b3b 100644 --- a/arch/arm/include/asm/arch-sunxi/ccu.h +++ b/arch/arm/include/asm/arch-sunxi/ccu.h @@ -7,15 +7,204 @@
#ifndef _ASM_ARCH_CCU_H
#define _ASM_ARCH_CCU_H

+#define OSC_32K_ULL 32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

Quoted 11 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL          24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:                       misc clock type  

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 4 lines

Unlike MP type in MMC, UART doesn't need any clock attributes like dividers, mux, etc, just have attached parent. I don't think UART clock is simple one It has parent, that indeed have another parents and so...on, ie reason I named as MISC.

Not really, as far as I can see the UART clock is a just a gate clock as many others, with one parent (APB2).
The fact that APB2 in turn can have multiple parents doesn't affect the UART clock itself, as you model this via the clock tree.

In fact we could have similar clocks in the tree structure for the other gate clocks (USB, for instance), it's just that the UART is the only user so far which actually queries the clock rate.

So MISC is way too generic, I would still prefer CCU_CLK_TYPE_GATE.

Quoted 30 lines
Quoted 21 lines
Quoted 3 lines

+ * @CCU_CLK_TYPE_FIXED: fixed clock type
+ * @CCU_CLK_TYPE_MP: mp clock type
+ * @CCU_CLK_TYPE_NK: nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

Quoted 10 lines

+ */

arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+enum ccu_clk_type {
+     CCU_CLK_TYPE_MISC               = 0,
+     CCU_CLK_TYPE_FIXED              = 1,
+     CCU_CLK_TYPE_MP                 = 2,
+     CCU_CLK_TYPE_NK                 = 3,
+};
+
 /**
  * enum ccu_clk_flags - ccu clock flags
Quoted 8 lines
Quoted 4 lines

*
- * @CCU_CLK_F_INIT_DONE: clock gate init done check
+ * @CCU_CLK_F_INIT_DONE: clock tree/gate init done
check

Is this flag to tell implemented clocks apart from unimplemented ones, which would be reset to all zeroes by the compiler? Then it should be called something with VALID in it.

Quoted 3 lines

Since the flags attached with real numeric initialized by macro itself rather than some other source like complier or any, so marked INIT_DONE seems to meaningful.

When I read INIT_DONE I understand some code has initialised this clock at some point, which isn't true. I don't fight the flag itself, just the name.

Quoted 345 lines

and eventually the same verified in code whether the init done or not.

Quoted 340 lines
Quoted 86 lines

+ * @CCU_CLK_F_POSTDIV: clock post divider
*/
enum ccu_clk_flags {
CCU_CLK_F_INIT_DONE = BIT(0),

arch/arm/include/asm/arch-sunxi/ccu.h +19 -0
+     CCU_CLK_F_POSTDIV               = BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:           multiplier shift value
+ * @width:           multiplier width value
+ * @offset:          multiplier offset
+ * @min:             minimum multiplier
+ * @max:             maximum multiplier
+ */
+struct ccu_mult {
+     u8 shift;
+     u8 width;
+     u8 offset;
+     u8 min;
+     u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width,

_offset, \

arch/arm/include/asm/arch-sunxi/ccu.h +59 -0
+                           _min, _max) {                     \
+     .shift = _shift,                                        \
+     .width = _width,                                        \
+     .offset = _offset,                                      \
+     .min = _min,                                            \
+     .max = _max,                                            \
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)                  \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)                            \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:           multiplexer shift value
+ * @width:           multiplexer width value
+ */
+struct ccu_mux {
+     u8 shift;
+     u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:           divider shift value
+ * @width:           divider width value
+ * @offset:          divider offset
+ * @max:             maximum divider value
+ */
+struct ccu_div {
+     u8 shift;
+     u8 width;
+     u32 offset;
+     u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+     .offset = 1,                            \
+     .max = 0,                               \
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:          parent clock tree
+ * @type:            clock type
+ * @off:             clock tree offset
+ * @m:                       divider m

arch/arm/include/asm/arch-sunxi/ccu.h line 125

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+ * @p:                       divider p
+ * @mux:             multiplexer mux
+ * @post:            post divider value
+ * @n:                       multiplier n
+ * @k:                       multiplier k
+ * @fixed_rate:              fixed rate
+ * @flags:           clock tree flags
+ */
+struct ccu_clk_tree {
+     const unsigned long *parent;  

Shouldn't that be called "parents" instead?

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+     enum ccu_clk_type type;
+     u16 off;
+
+     struct ccu_div m;
+     struct ccu_div p;
+     struct ccu_mux mux;

arch/arm/include/asm/arch-sunxi/ccu.h line 141

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+     unsigned int postdiv;
+
+     struct ccu_mult n;
+     struct ccu_mult k;
+
+     ulong fixed_rate;
+     enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,                           \  

Just a nit, but TREE is somewhat confusing here, as this just constructs a single entry in an array. The tree property is realised through the parent array, if I get this correctly. So should we name this ENTRY or CLK_ENTRY instead?

Quoted 21 lines
arch/arm/include/asm/arch-sunxi/ccu.h +20 -0
+          _m, _p,                                            \
+          _mux,                                              \
+          _postdiv,                                          \
+          _n, _k,                                            \
+          _fixed_rate,                                       \
+          _flags) {                                          \
+     .parent = _parent,                                      \
+     .type = _type,                                          \
+     .off = _off,                                            \
+     .m = _m,                                                \
+     .p = _p,                                                \
+     .mux = _mux,                                            \
+     .postdiv = _postdiv,                                    \
+     .n = _n,                                                \
+     .k = _k,                                                \
+     .fixed_rate = _fixed_rate,                              \
+     .flags = _flags,                                        \
+}
+
+#define

MISC(_parent) \

arch/arm/include/asm/arch-sunxi/ccu.h line 168

Quoted 7 lines
arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+     TREE(_parent, CCU_CLK_TYPE_MISC, 0,                     \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE)  

If MISC is really something like GATE or SIMPLE, would it be possible to construct the single element array here, so that the macro just takes the one parent clock ID here, instead of referring to an extra array?

Quoted 55 lines
arch/arm/include/asm/arch-sunxi/ccu.h +40 -0
+
+#define FIXED(_fixed_rate)                                   \
+     TREE(NULL, CCU_CLK_TYPE_FIXED, 0,                       \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          _fixed_rate,                                       \
+          CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,                                    \
+        _nshift, _nwidth,                                    \
+        _kshift, _kwidth, _kmin,                             \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_NK, _off,                    \
+          {0}, {0},                                          \
+          {0},                                               \
+          _postdiv,                                          \
+          _CCU_MULT(_nshift, _nwidth),                       \
+          _CCU_MULT_MIN(_kshift, _kwidth, _kmin),            \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,                                    \
+        _mshift, _mwidth,                                    \
+        _pshift, _pwidth,                                    \
+        _muxshift, _muxwidth,                                \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_MP, _off,                    \
+          _CCU_DIV(_mshift, _mwidth),                        \
+          _CCU_DIV(_pshift, _pwidth),                        \
+          _CCU_MUX(_muxshift, _muxwidth),                    \
+          _postdiv,                                          \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:     gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:  reset unit
  */
 struct ccu_desc {
+     const struct ccu_clk_tree *tree;
      const struct ccu_clk_gate *gates;
      const struct ccu_reset *resets;
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_a64.c index 162ec769d6..1d0cd98183 100644

drivers/clk/sunxi/clk_a64.c line 19

Quoted 8 lines
drivers/clk/sunxi/clk_a64.c +5 -0
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2                     26
+#define CLK_OSC_32K                  (CLK_GPU + 1)
+#define CLK_OSC_24M                  (CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {  

Why is this long? That's the DT clock index, which is 32 bit wide, right? Just unsigned int or uint32_t should be sufficient. long is quite a treacherous type to use, especially as we share code between 32 and 64-bit architectures.

Quoted 50 lines
drivers/clk/sunxi/clk_a64.c +35 -0
+     CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+     CLK_OSC_32K,
+     CLK_OSC_24M,
+     CLK_PLL_PERIPH0,
+     CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+     CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+     [CLK_OSC_32K]           = FIXED(OSC_32K_ULL),
+     [CLK_OSC_24M]           = FIXED(OSC_24M_ULL),
+
+     [CLK_PLL_PERIPH0]       = NK(periph0_parents, 0x028,
+                                  8, 5,      /* N */
+                                  4, 2, 2,   /* K */
+                                  2,         /* post-div */
+                                  CCU_CLK_F_POSTDIV),
+
+     [CLK_APB2]              = MP(apb2_parents, 0x058,
+                                  0, 5,      /* M */
+                                  16, 2,     /* P */
+                                  24, 2,     /* mux */
+                                  0,
+                                  0),
+
+     [CLK_BUS_UART0]         = MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
      [CLK_BUS_OTG]           = GATE(0x060, BIT(23)),
      [CLK_BUS_EHCI0]         = GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+     .tree = a64_tree,
      .gates = a64_gates,
      .resets = a64_resets,
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_sunxi.c index 345d706c2a..2aebd257d1 100644 --- a/drivers/clk/sunxi/clk_sunxi.c
+++ b/drivers/clk/sunxi/clk_sunxi.c

Review static const struct ccu_clk_gate · line 23

Quoted 6 lines
(quoted diff) +2 -0
@@ -18,6 +18,187 @@ static const struct ccu_clk_gate
 *priv_to_gate(struct ccu_priv *priv, return
 &priv->desc->gates[id]; }
 
+static const struct ccu_clk_tree *priv_to_tree(struct ccu_priv
 *priv,
+                                            unsigned long id)  

Again, why long here? Especially as it's u32 in the function below.

Quoted 7 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     return &priv->desc->tree[id];
+}
+
+static int sunxi_get_parent_idx(const struct ccu_clk_tree *tree,
 void *base) +{
+     u32 reg, idx;

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv,

Quoted 10 lines
drivers/clk/sunxi/clk_sunxi.c +9 -0
+
+     reg = readl(base + tree->off);
+     idx = reg >> tree->mux.shift;
+     idx &= (1 << tree->mux.width) - 1;
+
+     return idx;
+}
+
+static ulong sunxi_fixed_get_rate(struct clk *clk, unsigned long
 id)  

Same "long" question for both the return type and the id. And for everything below.

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 50

Quoted 13 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     struct ccu_priv *priv = dev_get_priv(clk->dev);
+     const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+
+     if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+             printf("%s: (CLK#%ld) unhandled\n", __func__,

clk->id);

drivers/clk/sunxi/clk_sunxi.c +6 -0
+             return 0;
+     }
+
+     return tree->fixed_rate;
+}
+  

So why are there all those separate functions? Isn't that all the same algorithm: adjust the parent rate based on the clock type? I reworked this with recursive calls and it's MUCH less code:

(fill the gaps, using your types for your convenience ;-)

static ulong sunxi_apply_pll(struct ccu_priv *priv, ulong id, ulong parent_rate) { NK algorithm of sunxi_nk_get_rate }
static ulong sunxi_apply_div(struct ccu_priv *priv, ulong id, ulong parent_rate) { MP algorithm of sunxi_mp_get_rate }
static ulong sunxi_calc_clk_rate(struct ccu_priv *priv, ulong clkid) {
....

        switch (tree->type) {
        case CCU_CLK_TYPE_MISC:
                return sunxi_calc_clk_rate(priv, tree->parent[0]);
        case CCU_CLK_TYPE_FIXED:
                return tree->fixed_rate;
        case CCU_CLK_TYPE_NK:
                rate = sunxi_calc_clk_rate(priv,
                        sunxi_get_parent_id(tree, priv->base));
                return sunxi_apply_pll(priv, clkid, rate);
        (similar for _MP)
Quoted 2 lines

...
}

Quoted 4 lines

Initially I would tried the recursive and yes code can reduce but using recursive can leed more disadvantage in-terms of code tracing during long run. Due to all these factors I used simple function calls.

But I find those extra functions much more confusing, due to the similar names and their very similar functionality. Also it seems that you just implemented what we need so far, so you will probably need to extend those functions, making them even more similar and duplicating more code. Basically you try to roll out the tree structure.

Since the clocks are organised in a tree-like structure, I believe this recursive definition is a much better fit: A clock takes one of possibly multiple input clocks and adjusts this rate. Full stop. The rest is then just connecting them to other clocks.
The code looks much simpler and is much smaller this way:
https://gist.github.com/apritzel/db93dd06b4defb46504bccbfe4fc2c20#file-sunxi_clk-c-L86-L112

Typically the recursion depth is just two or three levels, so I don't buy the argument of code tracing.

Cheers,
Andre.

author
#23 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

Quoted 102 lines · Andre Przywara

On Tue, 8 Jan 2019 16:27:14 +0530
Jagan Teki <[email protected]> wrote:

Hi,

Quoted 86 lines
Quoted 10 lines · André Przywara
Quoted 7 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

Quoted 8 lines

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Exactly. Idea is to keep the macro's as simple as possible.

Adding gates clocks separately make easy and reasonable way to enable/disable clock operations. We even operate with single macro with all clock attributes along with gate(either another member or common structure like Linux does), but that seems not simple as per as my experince since there are many IP's like USB's just need enable/disable.

Quoted 44 lines
Quoted 4 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 19 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192

+++++++++++++++++++++++++- drivers/clk/sunxi/clk_a64.c
| 40 ++++++ drivers/clk/sunxi/clk_sunxi.c | 182
++++++++++++++++++++++++ 3 files changed, 413 insertions(+), 1 deletion(-)

(quoted diff) +0 -0
Quoted 8 lines

b/arch/arm/include/asm/arch-sunxi/ccu.h index 3fdc26978d..61b8c36b3b 100644 --- a/arch/arm/include/asm/arch-sunxi/ccu.h +++ b/arch/arm/include/asm/arch-sunxi/ccu.h @@ -7,15 +7,204 @@
#ifndef _ASM_ARCH_CCU_H
#define _ASM_ARCH_CCU_H

+#define OSC_32K_ULL 32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

Quoted 11 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL          24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:                       misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 4 lines

Unlike MP type in MMC, UART doesn't need any clock attributes like dividers, mux, etc, just have attached parent. I don't think UART clock is simple one It has parent, that indeed have another parents and so...on, ie reason I named as MISC.

Not really, as far as I can see the UART clock is a just a gate clock as many others, with one parent (APB2).
The fact that APB2 in turn can have multiple parents doesn't affect the UART clock itself, as you model this via the clock tree.

Quoted 5 lines

In fact we could have similar clocks in the tree structure for the other gate clocks (USB, for instance), it's just that the UART is the only user so far which actually queries the clock rate.

So MISC is way too generic, I would still prefer CCU_CLK_TYPE_GATE.

TYPE_GATE is more sense. fine for me.

Quoted 31 lines
Quoted 31 lines
Quoted 30 lines
Quoted 4 lines

+ * @CCU_CLK_TYPE_FIXED: fixed clock type
+ * @CCU_CLK_TYPE_MP: mp clock type
+ * @CCU_CLK_TYPE_NK: nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

Quoted 13 lines

+ */

arch/arm/include/asm/arch-sunxi/ccu.h +8 -1
+enum ccu_clk_type {
+     CCU_CLK_TYPE_MISC               = 0,
+     CCU_CLK_TYPE_FIXED              = 1,
+     CCU_CLK_TYPE_MP                 = 2,
+     CCU_CLK_TYPE_NK                 = 3,
+};
+
 /**
  * enum ccu_clk_flags - ccu clock flags
  *
- * @CCU_CLK_F_INIT_DONE:             clock gate init done check
+ * @CCU_CLK_F_INIT_DONE:             clock tree/gate init done
Quoted 1 line

check

Is this flag to tell implemented clocks apart from unimplemented ones, which would be reset to all zeroes by the compiler? Then it should be called something with VALID in it.

Quoted 7 lines
Quoted 3 lines

Since the flags attached with real numeric initialized by macro itself rather than some other source like complier or any, so marked INIT_DONE seems to meaningful.

When I read INIT_DONE I understand some code has initialised this clock at some point, which isn't true. I don't fight the flag itself, just the name.

OK, I have updated INIT_DONE with IS_VALID and sent new version changes based on this.

Quoted 364 lines
Quoted 346 lines

and eventually the same verified in code whether the init done or not.

Quoted 340 lines
Quoted 86 lines

+ * @CCU_CLK_F_POSTDIV: clock post divider
*/
enum ccu_clk_flags {
CCU_CLK_F_INIT_DONE = BIT(0),

arch/arm/include/asm/arch-sunxi/ccu.h +19 -0
+     CCU_CLK_F_POSTDIV               = BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:           multiplier shift value
+ * @width:           multiplier width value
+ * @offset:          multiplier offset
+ * @min:             minimum multiplier
+ * @max:             maximum multiplier
+ */
+struct ccu_mult {
+     u8 shift;
+     u8 width;
+     u8 offset;
+     u8 min;
+     u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width,

_offset, \

arch/arm/include/asm/arch-sunxi/ccu.h +59 -0
+                           _min, _max) {                     \
+     .shift = _shift,                                        \
+     .width = _width,                                        \
+     .offset = _offset,                                      \
+     .min = _min,                                            \
+     .max = _max,                                            \
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)                  \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)                            \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:           multiplexer shift value
+ * @width:           multiplexer width value
+ */
+struct ccu_mux {
+     u8 shift;
+     u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:           divider shift value
+ * @width:           divider width value
+ * @offset:          divider offset
+ * @max:             maximum divider value
+ */
+struct ccu_div {
+     u8 shift;
+     u8 width;
+     u32 offset;
+     u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+     .offset = 1,                            \
+     .max = 0,                               \
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:          parent clock tree
+ * @type:            clock type
+ * @off:             clock tree offset
+ * @m:                       divider m

arch/arm/include/asm/arch-sunxi/ccu.h line 125

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+ * @p:                       divider p
+ * @mux:             multiplexer mux
+ * @post:            post divider value
+ * @n:                       multiplier n
+ * @k:                       multiplier k
+ * @fixed_rate:              fixed rate
+ * @flags:           clock tree flags
+ */
+struct ccu_clk_tree {
+     const unsigned long *parent;

Shouldn't that be called "parents" instead?

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+     enum ccu_clk_type type;
+     u16 off;
+
+     struct ccu_div m;
+     struct ccu_div p;
+     struct ccu_mux mux;

arch/arm/include/asm/arch-sunxi/ccu.h line 141

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+     unsigned int postdiv;
+
+     struct ccu_mult n;
+     struct ccu_mult k;
+
+     ulong fixed_rate;
+     enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,                           \

Just a nit, but TREE is somewhat confusing here, as this just constructs a single entry in an array. The tree property is realised through the parent array, if I get this correctly. So should we name this ENTRY or CLK_ENTRY instead?

Quoted 21 lines
arch/arm/include/asm/arch-sunxi/ccu.h +20 -0
+          _m, _p,                                            \
+          _mux,                                              \
+          _postdiv,                                          \
+          _n, _k,                                            \
+          _fixed_rate,                                       \
+          _flags) {                                          \
+     .parent = _parent,                                      \
+     .type = _type,                                          \
+     .off = _off,                                            \
+     .m = _m,                                                \
+     .p = _p,                                                \
+     .mux = _mux,                                            \
+     .postdiv = _postdiv,                                    \
+     .n = _n,                                                \
+     .k = _k,                                                \
+     .fixed_rate = _fixed_rate,                              \
+     .flags = _flags,                                        \
+}
+
+#define

MISC(_parent) \

arch/arm/include/asm/arch-sunxi/ccu.h line 168

Quoted 7 lines
arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+     TREE(_parent, CCU_CLK_TYPE_MISC, 0,                     \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE)

If MISC is really something like GATE or SIMPLE, would it be possible to construct the single element array here, so that the macro just takes the one parent clock ID here, instead of referring to an extra array?

Quoted 55 lines
arch/arm/include/asm/arch-sunxi/ccu.h +40 -0
+
+#define FIXED(_fixed_rate)                                   \
+     TREE(NULL, CCU_CLK_TYPE_FIXED, 0,                       \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          _fixed_rate,                                       \
+          CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,                                    \
+        _nshift, _nwidth,                                    \
+        _kshift, _kwidth, _kmin,                             \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_NK, _off,                    \
+          {0}, {0},                                          \
+          {0},                                               \
+          _postdiv,                                          \
+          _CCU_MULT(_nshift, _nwidth),                       \
+          _CCU_MULT_MIN(_kshift, _kwidth, _kmin),            \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,                                    \
+        _mshift, _mwidth,                                    \
+        _pshift, _pwidth,                                    \
+        _muxshift, _muxwidth,                                \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_MP, _off,                    \
+          _CCU_DIV(_mshift, _mwidth),                        \
+          _CCU_DIV(_pshift, _pwidth),                        \
+          _CCU_MUX(_muxshift, _muxwidth),                    \
+          _postdiv,                                          \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:     gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:  reset unit
  */
 struct ccu_desc {
+     const struct ccu_clk_tree *tree;
      const struct ccu_clk_gate *gates;
      const struct ccu_reset *resets;
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_a64.c index 162ec769d6..1d0cd98183 100644

drivers/clk/sunxi/clk_a64.c line 19

Quoted 8 lines
drivers/clk/sunxi/clk_a64.c +5 -0
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2                     26
+#define CLK_OSC_32K                  (CLK_GPU + 1)
+#define CLK_OSC_24M                  (CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {

Why is this long? That's the DT clock index, which is 32 bit wide, right? Just unsigned int or uint32_t should be sufficient. long is quite a treacherous type to use, especially as we share code between 32 and 64-bit architectures.

Quoted 50 lines
drivers/clk/sunxi/clk_a64.c +35 -0
+     CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+     CLK_OSC_32K,
+     CLK_OSC_24M,
+     CLK_PLL_PERIPH0,
+     CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+     CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+     [CLK_OSC_32K]           = FIXED(OSC_32K_ULL),
+     [CLK_OSC_24M]           = FIXED(OSC_24M_ULL),
+
+     [CLK_PLL_PERIPH0]       = NK(periph0_parents, 0x028,
+                                  8, 5,      /* N */
+                                  4, 2, 2,   /* K */
+                                  2,         /* post-div */
+                                  CCU_CLK_F_POSTDIV),
+
+     [CLK_APB2]              = MP(apb2_parents, 0x058,
+                                  0, 5,      /* M */
+                                  16, 2,     /* P */
+                                  24, 2,     /* mux */
+                                  0,
+                                  0),
+
+     [CLK_BUS_UART0]         = MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
      [CLK_BUS_OTG]           = GATE(0x060, BIT(23)),
      [CLK_BUS_EHCI0]         = GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+     .tree = a64_tree,
      .gates = a64_gates,
      .resets = a64_resets,
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_sunxi.c index 345d706c2a..2aebd257d1 100644 --- a/drivers/clk/sunxi/clk_sunxi.c
+++ b/drivers/clk/sunxi/clk_sunxi.c

Review static const struct ccu_clk_gate · line 23

Quoted 6 lines
(quoted diff) +2 -0
@@ -18,6 +18,187 @@ static const struct ccu_clk_gate
 *priv_to_gate(struct ccu_priv *priv, return
 &priv->desc->gates[id]; }
 
+static const struct ccu_clk_tree *priv_to_tree(struct ccu_priv
 *priv,
+                                            unsigned long id)

Again, why long here? Especially as it's u32 in the function below.

Quoted 7 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     return &priv->desc->tree[id];
+}
+
+static int sunxi_get_parent_idx(const struct ccu_clk_tree *tree,
 void *base) +{
+     u32 reg, idx;

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv,

Quoted 10 lines
drivers/clk/sunxi/clk_sunxi.c +9 -0
+
+     reg = readl(base + tree->off);
+     idx = reg >> tree->mux.shift;
+     idx &= (1 << tree->mux.width) - 1;
+
+     return idx;
+}
+
+static ulong sunxi_fixed_get_rate(struct clk *clk, unsigned long
 id)

Same "long" question for both the return type and the id. And for everything below.

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 50

Quoted 13 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     struct ccu_priv *priv = dev_get_priv(clk->dev);
+     const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+
+     if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+             printf("%s: (CLK#%ld) unhandled\n", __func__,

clk->id);

drivers/clk/sunxi/clk_sunxi.c +6 -0
+             return 0;
+     }
+
+     return tree->fixed_rate;
+}
+

So why are there all those separate functions? Isn't that all the same algorithm: adjust the parent rate based on the clock type? I reworked this with recursive calls and it's MUCH less code:

(fill the gaps, using your types for your convenience ;-)

static ulong sunxi_apply_pll(struct ccu_priv *priv, ulong id, ulong parent_rate) { NK algorithm of sunxi_nk_get_rate }
static ulong sunxi_apply_div(struct ccu_priv *priv, ulong id, ulong parent_rate) { MP algorithm of sunxi_mp_get_rate }
static ulong sunxi_calc_clk_rate(struct ccu_priv *priv, ulong clkid) {
....

        switch (tree->type) {
        case CCU_CLK_TYPE_MISC:
                return sunxi_calc_clk_rate(priv, tree->parent[0]);
        case CCU_CLK_TYPE_FIXED:
                return tree->fixed_rate;
        case CCU_CLK_TYPE_NK:
                rate = sunxi_calc_clk_rate(priv,
                        sunxi_get_parent_id(tree, priv->base));
                return sunxi_apply_pll(priv, clkid, rate);
        (similar for _MP)
Quoted 2 lines

...
}

Quoted 4 lines

Initially I would tried the recursive and yes code can reduce but using recursive can leed more disadvantage in-terms of code tracing during long run. Due to all these factors I used simple function calls.

But I find those extra functions much more confusing, due to the similar names and their very similar functionality. Also it seems that you just implemented what we need so far, so you will probably need to extend those functions, making them even more similar and duplicating more code. Basically you try to roll out the tree structure.

Since the clocks are organised in a tree-like structure, I believe this recursive definition is a much better fit: A clock takes one of possibly multiple input clocks and adjusts this rate. Full stop. The rest is then just connecting them to other clocks.
The code looks much simpler and is much smaller this way:
https://gist.github.com/apritzel/db93dd06b4defb46504bccbfe4fc2c20#file-sunxi_clk-c-L86-L112

Quoted 2 lines

Typically the recursion depth is just two or three levels, so I don't buy the argument of code tracing.

OK. agreed thanks.

I shall take this recursive and try to mark separate patch on behalf of you if possible, will that be fine?

He is my next version TODO.
- Fixing clock types with meaning full values
- Adopt recursive function calls
- Try to add full MMC blot.

Let me know if you have any comments or question so-that we can collaborate smooth to get CLK in as soon as possible.

Jagan.

#24 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

Quoted 146 lines · Jagan Teki
Quoted 102 lines · Andre Przywara

On Tue, 8 Jan 2019 16:27:14 +0530
Jagan Teki <[email protected]> wrote:

Hi,

Quoted 86 lines
Quoted 10 lines · André Przywara
Quoted 7 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

Quoted 8 lines

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Exactly. Idea is to keep the macro's as simple as possible.

Adding gates clocks separately make easy and reasonable way to enable/disable clock operations. We even operate with single macro with all clock attributes along with gate(either another member or common structure like Linux does), but that seems not simple as per as my experince since there are many IP's like USB's just need enable/disable.

Quoted 44 lines
Quoted 4 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 19 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192

+++++++++++++++++++++++++- drivers/clk/sunxi/clk_a64.c
| 40 ++++++ drivers/clk/sunxi/clk_sunxi.c | 182
++++++++++++++++++++++++ 3 files changed, 413 insertions(+), 1 deletion(-)

(quoted diff) +0 -0
Quoted 8 lines

b/arch/arm/include/asm/arch-sunxi/ccu.h index 3fdc26978d..61b8c36b3b 100644 --- a/arch/arm/include/asm/arch-sunxi/ccu.h +++ b/arch/arm/include/asm/arch-sunxi/ccu.h @@ -7,15 +7,204 @@
#ifndef _ASM_ARCH_CCU_H
#define _ASM_ARCH_CCU_H

+#define OSC_32K_ULL 32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

Quoted 11 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL          24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:                       misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 4 lines

Unlike MP type in MMC, UART doesn't need any clock attributes like dividers, mux, etc, just have attached parent. I don't think UART clock is simple one It has parent, that indeed have another parents and so...on, ie reason I named as MISC.

Not really, as far as I can see the UART clock is a just a gate clock as many others, with one parent (APB2).
The fact that APB2 in turn can have multiple parents doesn't affect the UART clock itself, as you model this via the clock tree.

Quoted 5 lines

In fact we could have similar clocks in the tree structure for the other gate clocks (USB, for instance), it's just that the UART is the only user so far which actually queries the clock rate.

So MISC is way too generic, I would still prefer CCU_CLK_TYPE_GATE.

TYPE_GATE is more sense. fine for me.

Quoted 35 lines
Quoted 34 lines
Quoted 22 lines
Quoted 4 lines

+ * @CCU_CLK_TYPE_FIXED: fixed clock type
+ * @CCU_CLK_TYPE_MP: mp clock type
+ * @CCU_CLK_TYPE_NK: nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

Quoted 10 lines

+ */

arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+enum ccu_clk_type {
+     CCU_CLK_TYPE_MISC               = 0,
+     CCU_CLK_TYPE_FIXED              = 1,
+     CCU_CLK_TYPE_MP                 = 2,
+     CCU_CLK_TYPE_NK                 = 3,
+};
+
 /**
  * enum ccu_clk_flags - ccu clock flags
Quoted 8 lines
Quoted 4 lines

*
- * @CCU_CLK_F_INIT_DONE: clock gate init done check
+ * @CCU_CLK_F_INIT_DONE: clock tree/gate init done
check

Is this flag to tell implemented clocks apart from unimplemented ones, which would be reset to all zeroes by the compiler? Then it should be called something with VALID in it.

Since the flags attached with real numeric initialized by macro itself rather than some other source like complier or any, so marked INIT_DONE seems to meaningful.

Quoted 6 lines
Quoted 3 lines

When I read INIT_DONE I understand some code has initialised this clock at some point, which isn't true. I don't fight the flag itself, just the name.

OK, I have updated INIT_DONE with IS_VALID and sent new version changes based on this.

Great, thanks!

Quoted 363 lines
Quoted 363 lines
Quoted 345 lines

and eventually the same verified in code whether the init done or not.

Quoted 340 lines
Quoted 86 lines

+ * @CCU_CLK_F_POSTDIV: clock post divider
*/
enum ccu_clk_flags {
CCU_CLK_F_INIT_DONE = BIT(0),

arch/arm/include/asm/arch-sunxi/ccu.h +19 -0
+     CCU_CLK_F_POSTDIV               = BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:           multiplier shift value
+ * @width:           multiplier width value
+ * @offset:          multiplier offset
+ * @min:             minimum multiplier
+ * @max:             maximum multiplier
+ */
+struct ccu_mult {
+     u8 shift;
+     u8 width;
+     u8 offset;
+     u8 min;
+     u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width,

_offset, \

arch/arm/include/asm/arch-sunxi/ccu.h +59 -0
+                           _min, _max) {                     \
+     .shift = _shift,                                        \
+     .width = _width,                                        \
+     .offset = _offset,                                      \
+     .min = _min,                                            \
+     .max = _max,                                            \
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)                  \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)                            \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:           multiplexer shift value
+ * @width:           multiplexer width value
+ */
+struct ccu_mux {
+     u8 shift;
+     u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:           divider shift value
+ * @width:           divider width value
+ * @offset:          divider offset
+ * @max:             maximum divider value
+ */
+struct ccu_div {
+     u8 shift;
+     u8 width;
+     u32 offset;
+     u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+     .offset = 1,                            \
+     .max = 0,                               \
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:          parent clock tree
+ * @type:            clock type
+ * @off:             clock tree offset
+ * @m:                       divider m

arch/arm/include/asm/arch-sunxi/ccu.h line 125

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+ * @p:                       divider p
+ * @mux:             multiplexer mux
+ * @post:            post divider value
+ * @n:                       multiplier n
+ * @k:                       multiplier k
+ * @fixed_rate:              fixed rate
+ * @flags:           clock tree flags
+ */
+struct ccu_clk_tree {
+     const unsigned long *parent;

Shouldn't that be called "parents" instead?

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+     enum ccu_clk_type type;
+     u16 off;
+
+     struct ccu_div m;
+     struct ccu_div p;
+     struct ccu_mux mux;

arch/arm/include/asm/arch-sunxi/ccu.h line 141

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+     unsigned int postdiv;
+
+     struct ccu_mult n;
+     struct ccu_mult k;
+
+     ulong fixed_rate;
+     enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,                           \

Just a nit, but TREE is somewhat confusing here, as this just constructs a single entry in an array. The tree property is realised through the parent array, if I get this correctly. So should we name this ENTRY or CLK_ENTRY instead?

Quoted 21 lines
arch/arm/include/asm/arch-sunxi/ccu.h +20 -0
+          _m, _p,                                            \
+          _mux,                                              \
+          _postdiv,                                          \
+          _n, _k,                                            \
+          _fixed_rate,                                       \
+          _flags) {                                          \
+     .parent = _parent,                                      \
+     .type = _type,                                          \
+     .off = _off,                                            \
+     .m = _m,                                                \
+     .p = _p,                                                \
+     .mux = _mux,                                            \
+     .postdiv = _postdiv,                                    \
+     .n = _n,                                                \
+     .k = _k,                                                \
+     .fixed_rate = _fixed_rate,                              \
+     .flags = _flags,                                        \
+}
+
+#define

MISC(_parent) \

arch/arm/include/asm/arch-sunxi/ccu.h line 168

Quoted 7 lines
arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+     TREE(_parent, CCU_CLK_TYPE_MISC, 0,                     \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE)

If MISC is really something like GATE or SIMPLE, would it be possible to construct the single element array here, so that the macro just takes the one parent clock ID here, instead of referring to an extra array?

Quoted 55 lines
arch/arm/include/asm/arch-sunxi/ccu.h +40 -0
+
+#define FIXED(_fixed_rate)                                   \
+     TREE(NULL, CCU_CLK_TYPE_FIXED, 0,                       \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          _fixed_rate,                                       \
+          CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,                                    \
+        _nshift, _nwidth,                                    \
+        _kshift, _kwidth, _kmin,                             \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_NK, _off,                    \
+          {0}, {0},                                          \
+          {0},                                               \
+          _postdiv,                                          \
+          _CCU_MULT(_nshift, _nwidth),                       \
+          _CCU_MULT_MIN(_kshift, _kwidth, _kmin),            \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,                                    \
+        _mshift, _mwidth,                                    \
+        _pshift, _pwidth,                                    \
+        _muxshift, _muxwidth,                                \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_MP, _off,                    \
+          _CCU_DIV(_mshift, _mwidth),                        \
+          _CCU_DIV(_pshift, _pwidth),                        \
+          _CCU_MUX(_muxshift, _muxwidth),                    \
+          _postdiv,                                          \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:     gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:  reset unit
  */
 struct ccu_desc {
+     const struct ccu_clk_tree *tree;
      const struct ccu_clk_gate *gates;
      const struct ccu_reset *resets;
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_a64.c index 162ec769d6..1d0cd98183 100644

drivers/clk/sunxi/clk_a64.c line 19

Quoted 8 lines
drivers/clk/sunxi/clk_a64.c +5 -0
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2                     26
+#define CLK_OSC_32K                  (CLK_GPU + 1)
+#define CLK_OSC_24M                  (CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {

Why is this long? That's the DT clock index, which is 32 bit wide, right? Just unsigned int or uint32_t should be sufficient. long is quite a treacherous type to use, especially as we share code between 32 and 64-bit architectures.

Quoted 50 lines
drivers/clk/sunxi/clk_a64.c +35 -0
+     CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+     CLK_OSC_32K,
+     CLK_OSC_24M,
+     CLK_PLL_PERIPH0,
+     CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+     CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+     [CLK_OSC_32K]           = FIXED(OSC_32K_ULL),
+     [CLK_OSC_24M]           = FIXED(OSC_24M_ULL),
+
+     [CLK_PLL_PERIPH0]       = NK(periph0_parents, 0x028,
+                                  8, 5,      /* N */
+                                  4, 2, 2,   /* K */
+                                  2,         /* post-div */
+                                  CCU_CLK_F_POSTDIV),
+
+     [CLK_APB2]              = MP(apb2_parents, 0x058,
+                                  0, 5,      /* M */
+                                  16, 2,     /* P */
+                                  24, 2,     /* mux */
+                                  0,
+                                  0),
+
+     [CLK_BUS_UART0]         = MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
      [CLK_BUS_OTG]           = GATE(0x060, BIT(23)),
      [CLK_BUS_EHCI0]         = GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+     .tree = a64_tree,
      .gates = a64_gates,
      .resets = a64_resets,
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_sunxi.c index 345d706c2a..2aebd257d1 100644 --- a/drivers/clk/sunxi/clk_sunxi.c
+++ b/drivers/clk/sunxi/clk_sunxi.c

Review static const struct ccu_clk_gate · line 23

Quoted 6 lines
(quoted diff) +2 -0
@@ -18,6 +18,187 @@ static const struct ccu_clk_gate
 *priv_to_gate(struct ccu_priv *priv, return
 &priv->desc->gates[id]; }
 
+static const struct ccu_clk_tree *priv_to_tree(struct ccu_priv
 *priv,
+                                            unsigned long id)

Again, why long here? Especially as it's u32 in the function below.

Quoted 7 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     return &priv->desc->tree[id];
+}
+
+static int sunxi_get_parent_idx(const struct ccu_clk_tree *tree,
 void *base) +{
+     u32 reg, idx;

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv,

Quoted 10 lines
drivers/clk/sunxi/clk_sunxi.c +9 -0
+
+     reg = readl(base + tree->off);
+     idx = reg >> tree->mux.shift;
+     idx &= (1 << tree->mux.width) - 1;
+
+     return idx;
+}
+
+static ulong sunxi_fixed_get_rate(struct clk *clk, unsigned long
 id)

Same "long" question for both the return type and the id. And for everything below.

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 50

Quoted 13 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     struct ccu_priv *priv = dev_get_priv(clk->dev);
+     const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+
+     if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+             printf("%s: (CLK#%ld) unhandled\n", __func__,

clk->id);

drivers/clk/sunxi/clk_sunxi.c +6 -0
+             return 0;
+     }
+
+     return tree->fixed_rate;
+}
+

So why are there all those separate functions? Isn't that all the same algorithm: adjust the parent rate based on the clock type? I reworked this with recursive calls and it's MUCH less code:

(fill the gaps, using your types for your convenience ;-)

static ulong sunxi_apply_pll(struct ccu_priv *priv, ulong id, ulong parent_rate) { NK algorithm of sunxi_nk_get_rate }
static ulong sunxi_apply_div(struct ccu_priv *priv, ulong id, ulong parent_rate) { MP algorithm of sunxi_mp_get_rate }
static ulong sunxi_calc_clk_rate(struct ccu_priv *priv, ulong clkid) {
....

        switch (tree->type) {
        case CCU_CLK_TYPE_MISC:
                return sunxi_calc_clk_rate(priv, tree->parent[0]);
        case CCU_CLK_TYPE_FIXED:
                return tree->fixed_rate;
        case CCU_CLK_TYPE_NK:
                rate = sunxi_calc_clk_rate(priv,
                        sunxi_get_parent_id(tree, priv->base));
                return sunxi_apply_pll(priv, clkid, rate);
        (similar for _MP)
Quoted 2 lines

...
}

Quoted 4 lines

Initially I would tried the recursive and yes code can reduce but using recursive can leed more disadvantage in-terms of code tracing during long run. Due to all these factors I used simple function calls.

But I find those extra functions much more confusing, due to the similar names and their very similar functionality. Also it seems that you just implemented what we need so far, so you will probably need to extend those functions, making them even more similar and duplicating more code. Basically you try to roll out the tree structure.

Since the clocks are organised in a tree-like structure, I believe this recursive definition is a much better fit: A clock takes one of possibly multiple input clocks and adjusts this rate. Full stop. The rest is then just connecting them to other clocks.
The code looks much simpler and is much smaller this way:
https://gist.github.com/apritzel/db93dd06b4defb46504bccbfe4fc2c20#file-sunxi_clk-c-L86-L112

Quoted 7 lines
Quoted 2 lines

Typically the recursion depth is just two or three levels, so I don't buy the argument of code tracing.

OK. agreed thanks.

I shall take this recursive and try to mark separate patch on behalf of you if possible, will that be fine?

No need, just rework this into this patch and keep your authorship. You did the bulk of the work anyway. If someone complains about the code, send them to me ;-)

Quoted 7 lines

He is my next version TODO.
- Fixing clock types with meaning full values
- Adopt recursive function calls
- Try to add full MMC blot.

Let me know if you have any comments or question so-that we can collaborate smooth to get CLK in as soon as possible.

Sure, lets nail this in the upcoming merge window.

Btw. I still get the warning for DM_USB, shouldn't this be fixed already by this series?

Cheers,
Andre.

author
#25 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

Quoted 537 lines · André Przywara
Quoted 146 lines · Jagan Teki
Quoted 102 lines · Andre Przywara

On Tue, 8 Jan 2019 16:27:14 +0530
Jagan Teki <[email protected]> wrote:

Hi,

Quoted 86 lines
Quoted 10 lines · André Przywara
Quoted 7 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

Quoted 8 lines

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Exactly. Idea is to keep the macro's as simple as possible.

Adding gates clocks separately make easy and reasonable way to enable/disable clock operations. We even operate with single macro with all clock attributes along with gate(either another member or common structure like Linux does), but that seems not simple as per as my experince since there are many IP's like USB's just need enable/disable.

Quoted 44 lines
Quoted 4 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 19 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192

+++++++++++++++++++++++++- drivers/clk/sunxi/clk_a64.c
| 40 ++++++ drivers/clk/sunxi/clk_sunxi.c | 182
++++++++++++++++++++++++ 3 files changed, 413 insertions(+), 1 deletion(-)

(quoted diff) +0 -0
Quoted 8 lines

b/arch/arm/include/asm/arch-sunxi/ccu.h index 3fdc26978d..61b8c36b3b 100644 --- a/arch/arm/include/asm/arch-sunxi/ccu.h +++ b/arch/arm/include/asm/arch-sunxi/ccu.h @@ -7,15 +7,204 @@
#ifndef _ASM_ARCH_CCU_H
#define _ASM_ARCH_CCU_H

+#define OSC_32K_ULL 32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

Quoted 11 lines

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL          24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:                       misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

Quoted 4 lines

Unlike MP type in MMC, UART doesn't need any clock attributes like dividers, mux, etc, just have attached parent. I don't think UART clock is simple one It has parent, that indeed have another parents and so...on, ie reason I named as MISC.

Not really, as far as I can see the UART clock is a just a gate clock as many others, with one parent (APB2).
The fact that APB2 in turn can have multiple parents doesn't affect the UART clock itself, as you model this via the clock tree.

Quoted 5 lines

In fact we could have similar clocks in the tree structure for the other gate clocks (USB, for instance), it's just that the UART is the only user so far which actually queries the clock rate.

So MISC is way too generic, I would still prefer CCU_CLK_TYPE_GATE.

TYPE_GATE is more sense. fine for me.

Quoted 35 lines
Quoted 34 lines
Quoted 22 lines
Quoted 4 lines

+ * @CCU_CLK_TYPE_FIXED: fixed clock type
+ * @CCU_CLK_TYPE_MP: mp clock type
+ * @CCU_CLK_TYPE_NK: nk clock type

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

Quoted 10 lines

+ */

arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+enum ccu_clk_type {
+     CCU_CLK_TYPE_MISC               = 0,
+     CCU_CLK_TYPE_FIXED              = 1,
+     CCU_CLK_TYPE_MP                 = 2,
+     CCU_CLK_TYPE_NK                 = 3,
+};
+
 /**
  * enum ccu_clk_flags - ccu clock flags
Quoted 8 lines
Quoted 4 lines

*
- * @CCU_CLK_F_INIT_DONE: clock gate init done check
+ * @CCU_CLK_F_INIT_DONE: clock tree/gate init done
check

Is this flag to tell implemented clocks apart from unimplemented ones, which would be reset to all zeroes by the compiler? Then it should be called something with VALID in it.

Since the flags attached with real numeric initialized by macro itself rather than some other source like complier or any, so marked INIT_DONE seems to meaningful.

Quoted 6 lines
Quoted 3 lines

When I read INIT_DONE I understand some code has initialised this clock at some point, which isn't true. I don't fight the flag itself, just the name.

OK, I have updated INIT_DONE with IS_VALID and sent new version changes based on this.

Great, thanks!

Quoted 363 lines
Quoted 363 lines
Quoted 345 lines

and eventually the same verified in code whether the init done or not.

Quoted 340 lines
Quoted 86 lines

+ * @CCU_CLK_F_POSTDIV: clock post divider
*/
enum ccu_clk_flags {
CCU_CLK_F_INIT_DONE = BIT(0),

arch/arm/include/asm/arch-sunxi/ccu.h +19 -0
+     CCU_CLK_F_POSTDIV               = BIT(1),
 };
 
+/**
+ * struct ccu_mult - ccu clock multiplier
+ *
+ * @shift:           multiplier shift value
+ * @width:           multiplier width value
+ * @offset:          multiplier offset
+ * @min:             minimum multiplier
+ * @max:             maximum multiplier
+ */
+struct ccu_mult {
+     u8 shift;
+     u8 width;
+     u8 offset;
+     u8 min;
+     u8 max;
+};
+
+#define _CCU_MULT_OFF_MIN_MAX(_shift, _width,

_offset, \

arch/arm/include/asm/arch-sunxi/ccu.h +59 -0
+                           _min, _max) {                     \
+     .shift = _shift,                                        \
+     .width = _width,                                        \
+     .offset = _offset,                                      \
+     .min = _min,                                            \
+     .max = _max,                                            \
+}
+
+#define _CCU_MULT_MIN(_shift, _width, _min)                  \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, _min, 0)
+
+#define _CCU_MULT(_shift, _width)                            \
+     _CCU_MULT_OFF_MIN_MAX(_shift, _width, 1, 1, 0)
+
+/**
+ * struct ccu_mux - ccu clock multiplexer
+ *
+ * @shift:           multiplexer shift value
+ * @width:           multiplexer width value
+ */
+struct ccu_mux {
+     u8 shift;
+     u8 width;
+};
+
+#define _CCU_MUX(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+}
+
+/**
+ * struct ccu_div - ccu clock divider
+ *
+ * @shift:           divider shift value
+ * @width:           divider width value
+ * @offset:          divider offset
+ * @max:             maximum divider value
+ */
+struct ccu_div {
+     u8 shift;
+     u8 width;
+     u32 offset;
+     u32 max;
+};
+
+#define _CCU_DIV(_shift, _width) {           \
+     .shift = _shift,                        \
+     .width = _width,                        \
+     .offset = 1,                            \
+     .max = 0,                               \
+}
+
+/**
+ * struct ccu_clk_tree - ccu clock tree
+ *
+ * @parent:          parent clock tree
+ * @type:            clock type
+ * @off:             clock tree offset
+ * @m:                       divider m

arch/arm/include/asm/arch-sunxi/ccu.h line 125

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+ * @p:                       divider p
+ * @mux:             multiplexer mux
+ * @post:            post divider value
+ * @n:                       multiplier n
+ * @k:                       multiplier k
+ * @fixed_rate:              fixed rate
+ * @flags:           clock tree flags
+ */
+struct ccu_clk_tree {
+     const unsigned long *parent;

Shouldn't that be called "parents" instead?

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+     enum ccu_clk_type type;
+     u16 off;
+
+     struct ccu_div m;
+     struct ccu_div p;
+     struct ccu_mux mux;

arch/arm/include/asm/arch-sunxi/ccu.h line 141

Quoted 10 lines
arch/arm/include/asm/arch-sunxi/ccu.h +10 -0
+     unsigned int postdiv;
+
+     struct ccu_mult n;
+     struct ccu_mult k;
+
+     ulong fixed_rate;
+     enum ccu_clk_flags flags;
+};
+
+#define TREE(_parent, _type, _off,                           \

Just a nit, but TREE is somewhat confusing here, as this just constructs a single entry in an array. The tree property is realised through the parent array, if I get this correctly. So should we name this ENTRY or CLK_ENTRY instead?

Quoted 21 lines
arch/arm/include/asm/arch-sunxi/ccu.h +20 -0
+          _m, _p,                                            \
+          _mux,                                              \
+          _postdiv,                                          \
+          _n, _k,                                            \
+          _fixed_rate,                                       \
+          _flags) {                                          \
+     .parent = _parent,                                      \
+     .type = _type,                                          \
+     .off = _off,                                            \
+     .m = _m,                                                \
+     .p = _p,                                                \
+     .mux = _mux,                                            \
+     .postdiv = _postdiv,                                    \
+     .n = _n,                                                \
+     .k = _k,                                                \
+     .fixed_rate = _fixed_rate,                              \
+     .flags = _flags,                                        \
+}
+
+#define

MISC(_parent) \

arch/arm/include/asm/arch-sunxi/ccu.h line 168

Quoted 7 lines
arch/arm/include/asm/arch-sunxi/ccu.h +7 -0
+     TREE(_parent, CCU_CLK_TYPE_MISC, 0,                     \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE)

If MISC is really something like GATE or SIMPLE, would it be possible to construct the single element array here, so that the macro just takes the one parent clock ID here, instead of referring to an extra array?

Quoted 55 lines
arch/arm/include/asm/arch-sunxi/ccu.h +40 -0
+
+#define FIXED(_fixed_rate)                                   \
+     TREE(NULL, CCU_CLK_TYPE_FIXED, 0,                       \
+          {0}, {0},                                          \
+          {0},                                               \
+          0,                                                 \
+          {0}, {0},                                          \
+          _fixed_rate,                                       \
+          CCU_CLK_F_INIT_DONE)
+
+#define NK(_parent, _off,                                    \
+        _nshift, _nwidth,                                    \
+        _kshift, _kwidth, _kmin,                             \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_NK, _off,                    \
+          {0}, {0},                                          \
+          {0},                                               \
+          _postdiv,                                          \
+          _CCU_MULT(_nshift, _nwidth),                       \
+          _CCU_MULT_MIN(_kshift, _kwidth, _kmin),            \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
+#define MP(_parent, _off,                                    \
+        _mshift, _mwidth,                                    \
+        _pshift, _pwidth,                                    \
+        _muxshift, _muxwidth,                                \
+        _postdiv,                                            \
+        _flags)                                              \
+     TREE(_parent, CCU_CLK_TYPE_MP, _off,                    \
+          _CCU_DIV(_mshift, _mwidth),                        \
+          _CCU_DIV(_pshift, _pwidth),                        \
+          _CCU_MUX(_muxshift, _muxwidth),                    \
+          _postdiv,                                          \
+          {0}, {0},                                          \
+          0,                                                 \
+          CCU_CLK_F_INIT_DONE | _flags)
+
 /**
  * struct ccu_clk_gate - ccu clock gate
  * @off:     gate offset
@@ -59,6 +248,7 @@ struct ccu_reset {
  * @resets:  reset unit
  */
 struct ccu_desc {
+     const struct ccu_clk_tree *tree;
      const struct ccu_clk_gate *gates;
      const struct ccu_reset *resets;
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_a64.c index 162ec769d6..1d0cd98183 100644

drivers/clk/sunxi/clk_a64.c line 19

Quoted 8 lines
drivers/clk/sunxi/clk_a64.c +5 -0
@@ -12,6 +12,45 @@
 #include <dt-bindings/clock/sun50i-a64-ccu.h>
 #include <dt-bindings/reset/sun50i-a64-ccu.h>
 
+#define CLK_APB2                     26
+#define CLK_OSC_32K                  (CLK_GPU + 1)
+#define CLK_OSC_24M                  (CLK_OSC_32K + 1)
+
+static const unsigned long periph0_parents[] = {

Why is this long? That's the DT clock index, which is 32 bit wide, right? Just unsigned int or uint32_t should be sufficient. long is quite a treacherous type to use, especially as we share code between 32 and 64-bit architectures.

Quoted 50 lines
drivers/clk/sunxi/clk_a64.c +35 -0
+     CLK_OSC_24M,
+};
+
+static const unsigned long apb2_parents[] = {
+     CLK_OSC_32K,
+     CLK_OSC_24M,
+     CLK_PLL_PERIPH0,
+     CLK_PLL_PERIPH0,
+};
+
+static const unsigned long uart_parents[] = {
+     CLK_APB2,
+};
+
+static const struct ccu_clk_tree a64_tree[] = {
+     [CLK_OSC_32K]           = FIXED(OSC_32K_ULL),
+     [CLK_OSC_24M]           = FIXED(OSC_24M_ULL),
+
+     [CLK_PLL_PERIPH0]       = NK(periph0_parents, 0x028,
+                                  8, 5,      /* N */
+                                  4, 2, 2,   /* K */
+                                  2,         /* post-div */
+                                  CCU_CLK_F_POSTDIV),
+
+     [CLK_APB2]              = MP(apb2_parents, 0x058,
+                                  0, 5,      /* M */
+                                  16, 2,     /* P */
+                                  24, 2,     /* mux */
+                                  0,
+                                  0),
+
+     [CLK_BUS_UART0]         = MISC(uart_parents),
+};
+
 static const struct ccu_clk_gate a64_gates[] = {
      [CLK_BUS_OTG]           = GATE(0x060, BIT(23)),
      [CLK_BUS_EHCI0]         = GATE(0x060, BIT(24)),
@@ -52,6 +91,7 @@ static const struct ccu_reset a64_resets[] = {
 };
 
 static const struct ccu_desc a64_ccu_desc = {
+     .tree = a64_tree,
      .gates = a64_gates,
      .resets = a64_resets,
 };
(quoted diff) +0 -0

b/drivers/clk/sunxi/clk_sunxi.c index 345d706c2a..2aebd257d1 100644 --- a/drivers/clk/sunxi/clk_sunxi.c
+++ b/drivers/clk/sunxi/clk_sunxi.c

Review static const struct ccu_clk_gate · line 23

Quoted 6 lines
(quoted diff) +2 -0
@@ -18,6 +18,187 @@ static const struct ccu_clk_gate
 *priv_to_gate(struct ccu_priv *priv, return
 &priv->desc->gates[id]; }
 
+static const struct ccu_clk_tree *priv_to_tree(struct ccu_priv
 *priv,
+                                            unsigned long id)

Again, why long here? Especially as it's u32 in the function below.

Quoted 7 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     return &priv->desc->tree[id];
+}
+
+static int sunxi_get_parent_idx(const struct ccu_clk_tree *tree,
 void *base) +{
+     u32 reg, idx;

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv,

Quoted 10 lines
drivers/clk/sunxi/clk_sunxi.c +9 -0
+
+     reg = readl(base + tree->off);
+     idx = reg >> tree->mux.shift;
+     idx &= (1 << tree->mux.width) - 1;
+
+     return idx;
+}
+
+static ulong sunxi_fixed_get_rate(struct clk *clk, unsigned long
 id)

Same "long" question for both the return type and the id. And for everything below.

drivers/clk/sunxi/clk_sunxi.c static const struct ccu_clk_gate *priv_to_gate(struct ccu_priv *priv, · line 50

Quoted 13 lines
drivers/clk/sunxi/clk_sunxi.c +6 -0
+{
+     struct ccu_priv *priv = dev_get_priv(clk->dev);
+     const struct ccu_clk_tree *tree = priv_to_tree(priv, id);
+
+     if (!(tree->flags & CCU_CLK_F_INIT_DONE)) {
+             printf("%s: (CLK#%ld) unhandled\n", __func__,

clk->id);

drivers/clk/sunxi/clk_sunxi.c +6 -0
+             return 0;
+     }
+
+     return tree->fixed_rate;
+}
+

So why are there all those separate functions? Isn't that all the same algorithm: adjust the parent rate based on the clock type? I reworked this with recursive calls and it's MUCH less code:

(fill the gaps, using your types for your convenience ;-)

static ulong sunxi_apply_pll(struct ccu_priv *priv, ulong id, ulong parent_rate) { NK algorithm of sunxi_nk_get_rate }
static ulong sunxi_apply_div(struct ccu_priv *priv, ulong id, ulong parent_rate) { MP algorithm of sunxi_mp_get_rate }
static ulong sunxi_calc_clk_rate(struct ccu_priv *priv, ulong clkid) {
....

        switch (tree->type) {
        case CCU_CLK_TYPE_MISC:
                return sunxi_calc_clk_rate(priv, tree->parent[0]);
        case CCU_CLK_TYPE_FIXED:
                return tree->fixed_rate;
        case CCU_CLK_TYPE_NK:
                rate = sunxi_calc_clk_rate(priv,
                        sunxi_get_parent_id(tree, priv->base));
                return sunxi_apply_pll(priv, clkid, rate);
        (similar for _MP)
Quoted 2 lines

...
}

Quoted 4 lines

Initially I would tried the recursive and yes code can reduce but using recursive can leed more disadvantage in-terms of code tracing during long run. Due to all these factors I used simple function calls.

But I find those extra functions much more confusing, due to the similar names and their very similar functionality. Also it seems that you just implemented what we need so far, so you will probably need to extend those functions, making them even more similar and duplicating more code. Basically you try to roll out the tree structure.

Since the clocks are organised in a tree-like structure, I believe this recursive definition is a much better fit: A clock takes one of possibly multiple input clocks and adjusts this rate. Full stop. The rest is then just connecting them to other clocks.
The code looks much simpler and is much smaller this way:
https://gist.github.com/apritzel/db93dd06b4defb46504bccbfe4fc2c20#file-sunxi_clk-c-L86-L112

Quoted 7 lines
Quoted 2 lines

Typically the recursion depth is just two or three levels, so I don't buy the argument of code tracing.

OK. agreed thanks.

I shall take this recursive and try to mark separate patch on behalf of you if possible, will that be fine?

No need, just rework this into this patch and keep your authorship. You did the bulk of the work anyway. If someone complains about the code, send them to me ;-)

Quoted 5 lines

He is my next version TODO.
- Fixing clock types with meaning full values
- Adopt recursive function calls
- Try to add full MMC blot.

Quoted 7 lines
Quoted 2 lines

Let me know if you have any comments or question so-that we can collaborate smooth to get CLK in as soon as possible.

Sure, lets nail this in the upcoming merge window.

Btw. I still get the warning for DM_USB, shouldn't this be fixed already by this series?

We already have DM_USB enabled, since the check will also look for BLK it's still showing the warning. Once we move DM_MMC it will gone.

author
#26 lore raw

PATCH v5 15/26 clk: sunxi: Add ccu clock tree support

In reply to André Przywara
Quoted 73 lines · André Przywara
Quoted 6 lines · Jagan Teki

Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management which can be handle via ccu_clock_gate.

So if I understand this correctly, you implement the gate functionality separately from the complex clock code, even if they are the same clock from the DT point of view? So if one wants to enable the MMC0 clock, which is a mux/divider clock, one needs to specify an extra entry in the gate array to describe the enable bit in this special clock register? Sounds a bit surprising, but is probably a neat trick to keep things simple. There should be a comment in the code to this regard then.

Quoted 2 lines

Tree clocks are parent clock type, fixed clocks, mp, nk, nkm, nkmp, pre/post div, flags etc. which were managed via ccu_clock_tree.

For a start, can we use more descriptive names than those very specific MP/NK names? DIV_MUX and PLL sound more descriptive to me. I understand that Linux uses those terms, but it would be great if uninitiated people have a chance to understand this as well.

Quoted 22 lines

This patch add support for MP, NK, MISC, FIXED clock types as part of ccu clock tree with get_rate functionality this eventually used by uart driver. and rest of the infrastructure will try to add while CLK is being used on respective peripherals.

Note that few of the tree type clock would require to enable gates on their specific clock, in that case we need to add the gate details via ccu_clock_gate, example: MP with gate so the gate offset, bit value should add as part of ccu_clock_gate.

---
 arch/arm/include/asm/arch-sunxi/ccu.h | 192 +++++++++++++++++++++++++-
 drivers/clk/sunxi/clk_a64.c           |  40 ++++++
 drivers/clk/sunxi/clk_sunxi.c         | 182 ++++++++++++++++++++++++
 3 files changed, 413 insertions(+), 1 deletion(-)
arch/arm/include/asm/arch-sunxi/ccu.h +0 -0
index 3fdc26978d..61b8c36b3b 100644

arch/arm/include/asm/arch-sunxi/ccu.h line 10

Quoted 4 lines
arch/arm/include/asm/arch-sunxi/ccu.h +1 -0
@@ -7,15 +7,204 @@
 #ifndef _ASM_ARCH_CCU_H
 #define _ASM_ARCH_CCU_H
 
+#define OSC_32K_ULL          32000ULL

32768

And why ULL? The whole Allwinner clock system works with 32-bit values, so just U would be totally sufficient. This avoid blowing this up to 64 bit unnecessarily, which sounds painful for those poor ARMv7 parts.

arch/arm/include/asm/arch-sunxi/ccu.h line 16

Quoted 6 lines
arch/arm/include/asm/arch-sunxi/ccu.h +6 -0
+#define OSC_24M_ULL          24000000ULL
+
+/**
+ * enum ccu_clk_type - ccu clock types
+ *
+ * @CCU_CLK_TYPE_MISC:                       misc clock type

What is MISC, exactly? Seems like an artefact clock to me, some placeholder you need because gate clocks are handled separately in the gates struct. Should this be called something with SIMPLE instead, or GATE?

arch/arm/include/asm/arch-sunxi/ccu.h line 19

Quoted 3 lines
arch/arm/include/asm/arch-sunxi/ccu.h +3 -0
+ * @CCU_CLK_TYPE_FIXED:                      fixed clock type
+ * @CCU_CLK_TYPE_MP:                 mp clock type
+ * @CCU_CLK_TYPE_NK:                 nk clock type
Quoted 6 lines

What is the point of those comments, as you are basically repeating the enum name? What about:
* @CCU_CLK_TYPE_PLL: PLL clock with two multiplier fields
* @CCU_CLK_TYPE_MUX_DIV: multiple parents, two divider fields

This gives more speaking names, plus some explanation.

My idea is to give generic name for a given clock type for example MP clock has different varients like MP clock, MP with DIV, MP with Postdiv, MP can be MMC etc. same like NK clock type as NK with postdiv, NK can be PLL.

With this we can expose the base name to outside and keep fill the require variants during macro initialization. This can avoid to many names on the same clock based on the different variants. Yes we can add proper full detailed comments on the given type.

author
#27 lore raw

PATCH v5 16/26 sunxi: Enable CLK

In reply to Jagan Teki

CLK and DM_RESET drivers are now available for most of the Allwinner platforms, so enable in mach-sunxi/Kconfig

Enabling CLK will select DM_RESET by default.

---
 arch/arm/mach-sunxi/Kconfig | 12 ++++++++++++
 1 file changed, 12 insertions(+)
arch/arm/mach-sunxi/Kconfig +12 -0
index 3c54f5106d..5f51bbbed0 100644
@@ -132,6 +132,7 @@ endif
 
 config MACH_SUNXI_H3_H5
 	bool
+	select CLK
 	select DM_I2C
 	select PHY_SUN4I_USB
 	select SUNXI_DE2
@@ -154,6 +155,7 @@ config MACH_SUN4I
 	bool "sun4i (Allwinner A10)"
 	select CPU_V7A
 	select ARM_CORTEX_CPU_IS_UP
+	select CLK
 	select DM_MMC if MMC
 	select DM_SCSI if SCSI
 	select PHY_SUN4I_USB
@@ -165,6 +167,7 @@ config MACH_SUN5I
 	bool "sun5i (Allwinner A13)"
 	select CPU_V7A
 	select ARM_CORTEX_CPU_IS_UP
+	select CLK
 	select DRAM_SUN4I
 	select PHY_SUN4I_USB
 	select SUNXI_GEN_SUN4I
@@ -177,6 +180,7 @@ config MACH_SUN6I
 	select CPU_V7_HAS_NONSEC
 	select CPU_V7_HAS_VIRT
 	select ARCH_SUPPORT_PSCI
+	select CLK
 	select DRAM_SUN6I
 	select PHY_SUN4I_USB
 	select SUN6I_P2WI
@@ -191,6 +195,7 @@ config MACH_SUN7I
 	select CPU_V7_HAS_NONSEC
 	select CPU_V7_HAS_VIRT
 	select ARCH_SUPPORT_PSCI
+	select CLK
 	select DRAM_SUN4I
 	select PHY_SUN4I_USB
 	select SUNXI_GEN_SUN4I
@@ -203,6 +208,7 @@ config MACH_SUN8I_A23
 	select CPU_V7_HAS_NONSEC
 	select CPU_V7_HAS_VIRT
 	select ARCH_SUPPORT_PSCI
+	select CLK
 	select DRAM_SUN8I_A23
 	select PHY_SUN4I_USB
 	select SUNXI_GEN_SUN6I
@@ -216,6 +222,7 @@ config MACH_SUN8I_A33
 	select CPU_V7_HAS_NONSEC
 	select CPU_V7_HAS_VIRT
 	select ARCH_SUPPORT_PSCI
+	select CLK
 	select DRAM_SUN8I_A33
 	select PHY_SUN4I_USB
 	select SUNXI_GEN_SUN6I
@@ -226,6 +233,7 @@ config MACH_SUN8I_A33
 config MACH_SUN8I_A83T
 	bool "sun8i (Allwinner A83T)"
 	select CPU_V7A
+	select CLK
 	select DRAM_SUN8I_A83T
 	select PHY_SUN4I_USB
 	select SUNXI_GEN_SUN6I
@@ -248,6 +256,7 @@ config MACH_SUN8I_R40
 	select CPU_V7_HAS_NONSEC
 	select CPU_V7_HAS_VIRT
 	select ARCH_SUPPORT_PSCI
+	select CLK
 	select SUNXI_GEN_SUN6I
 	select SUPPORT_SPL
 	select SUNXI_DRAM_DW
@@ -259,6 +268,7 @@ config MACH_SUN8I_V3S
 	select CPU_V7_HAS_NONSEC
 	select CPU_V7_HAS_VIRT
 	select ARCH_SUPPORT_PSCI
+	select CLK
 	select SUNXI_GEN_SUN6I
 	select SUNXI_DRAM_DW
 	select SUNXI_DRAM_DW_16BIT
@@ -277,6 +287,7 @@ config MACH_SUN9I
 config MACH_SUN50I
 	bool "sun50i (Allwinner A64)"
 	select ARM64
+	select CLK
 	select DM_I2C
 	select PHY_SUN4I_USB
 	select SUN6I_PRCM
@@ -300,6 +311,7 @@ config MACH_SUN50I_H5
 config MACH_SUN50I_H6
 	bool "sun50i (Allwinner H6)"
 	select ARM64
+	select CLK
 	select SUPPORT_SPL
 	select FIT
 	select SPL_LOAD_FIT
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#28 lore raw

PATCH v5 17/26 phy: sun4i-usb: Use CLK and RESET support

In reply to Jagan Teki

Now clock and reset drivers are available for respective SoC's so use clk and reset ops on phy driver.

---
 drivers/phy/allwinner/phy-sun4i-usb.c | 77 ++++++++++++++++++++-------
 1 file changed, 57 insertions(+), 20 deletions(-)
drivers/phy/allwinner/phy-sun4i-usb.c +57 -20
index a7d7e3f044..f206fa3f5d 100644
@@ -11,10 +11,12 @@
  */
 
 #include <common.h>
+#include <clk.h>
 #include <dm.h>
 #include <dm/device.h>
 #include <generic-phy.h>
 #include <phy-sun4i-usb.h>
+#include <reset.h>
 #include <asm/gpio.h>
 #include <asm/io.h>
 #include <asm/arch/clock.h>
@@ -80,6 +82,7 @@ struct sun4i_usb_phy_cfg {
 	enum sun4i_usb_phy_type type;
 	u32 disc_thresh;
 	u8 phyctl_offset;
+	bool dedicated_clocks;
 	bool enable_pmu_unk1;
 	bool phy0_dual_route;
 };
@@ -88,30 +91,21 @@ struct sun4i_usb_phy_info {
 	const char *gpio_vbus;
 	const char *gpio_vbus_det;
 	const char *gpio_id_det;
-	int rst_mask;
 } phy_info[] = {
 	{
 		.gpio_vbus = CONFIG_USB0_VBUS_PIN,
 		.gpio_vbus_det = CONFIG_USB0_VBUS_DET,
 		.gpio_id_det = CONFIG_USB0_ID_DET,
-		.rst_mask = (CCM_USB_CTRL_PHY0_RST | CCM_USB_CTRL_PHY0_CLK),
 	},
 	{
 		.gpio_vbus = CONFIG_USB1_VBUS_PIN,
 		.gpio_vbus_det = NULL,
 		.gpio_id_det = NULL,
-		.rst_mask = (CCM_USB_CTRL_PHY1_RST | CCM_USB_CTRL_PHY1_CLK),
 	},
 	{
 		.gpio_vbus = CONFIG_USB2_VBUS_PIN,
 		.gpio_vbus_det = NULL,
 		.gpio_id_det = NULL,
-#ifdef CONFIG_MACH_SUN8I_A83T
-		.rst_mask = (CCM_USB_CTRL_HSIC_RST | CCM_USB_CTRL_HSIC_CLK |
-			     CCM_USB_CTRL_12M_CLK),
-#else
-		.rst_mask = (CCM_USB_CTRL_PHY2_RST | CCM_USB_CTRL_PHY2_CLK),
-#endif
 	},
 	{
 		.gpio_vbus = CONFIG_USB3_VBUS_PIN,
@@ -126,13 +120,13 @@ struct sun4i_usb_phy_plat {
 	int gpio_vbus;
 	int gpio_vbus_det;
 	int gpio_id_det;
-	int rst_mask;
+	struct clk clocks;
+	struct reset_ctl resets;
 	int id;
 };
 
 struct sun4i_usb_phy_data {
 	void __iomem *base;
-	struct sunxi_ccm_reg *ccm;
 	const struct sun4i_usb_phy_cfg *cfg;
 	struct sun4i_usb_phy_plat *usb_phy;
 };
@@ -266,8 +260,19 @@ static int sun4i_usb_phy_init(struct phy *phy)
 	struct sun4i_usb_phy_data *data = dev_get_priv(phy->dev);
 	struct sun4i_usb_phy_plat *usb_phy = &data->usb_phy[phy->id];
 	u32 val;
+	int ret;
 
-	setbits_le32(&data->ccm->usb_clk_cfg, usb_phy->rst_mask);
+	ret = clk_enable(&usb_phy->clocks);
+	if (ret) {
+		dev_err(dev, "failed to enable usb_%ldphy clock\n", phy->id);
+		return ret;
+	}
+
+	ret = reset_deassert(&usb_phy->resets);
+	if (ret) {
+		dev_err(dev, "failed to deassert usb_%ldreset reset\n", phy->id);
+		return ret;
+	}
 
 	if (data->cfg->type == sun8i_a83t_phy) {
 		if (phy->id == 0) {
@@ -308,6 +313,7 @@ static int sun4i_usb_phy_exit(struct phy *phy)
 {
 	struct sun4i_usb_phy_data *data = dev_get_priv(phy->dev);
 	struct sun4i_usb_phy_plat *usb_phy = &data->usb_phy[phy->id];
+	int ret;
 
 	if (phy->id == 0) {
 		if (data->cfg->type == sun8i_a83t_phy) {
@@ -320,7 +326,17 @@ static int sun4i_usb_phy_exit(struct phy *phy)
 
 	sun4i_usb_phy_passby(phy, false);
 
-	clrbits_le32(&data->ccm->usb_clk_cfg, usb_phy->rst_mask);
+	ret = clk_disable(&usb_phy->clocks);
+	if (ret) {
+		dev_err(dev, "failed to disable usb_%ldphy clock\n", phy->id);
+		return ret;
+	}
+
+	ret = reset_assert(&usb_phy->resets);
+	if (ret) {
+		dev_err(dev, "failed to assert usb_%ldreset reset\n", phy->id);
+		return ret;
+	}
 
 	return 0;
 }
@@ -407,10 +423,6 @@ static int sun4i_usb_phy_probe(struct udevice *dev)
 	if (IS_ERR(data->base))
 		return PTR_ERR(data->base);
 
-	data->ccm = (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
-	if (IS_ERR(data->ccm))
-		return PTR_ERR(data->ccm);
-
 	data->usb_phy = plat;
 	for (i = 0; i < data->cfg->num_phys; i++) {
 		struct sun4i_usb_phy_plat *phy = &plat[i];
@@ -448,6 +460,24 @@ static int sun4i_usb_phy_probe(struct udevice *dev)
 			sunxi_gpio_set_pull(phy->gpio_id_det, SUNXI_GPIO_PULL_UP);
 		}
 
+		if (data->cfg->dedicated_clocks)
+			snprintf(name, sizeof(name), "usb%d_phy", i);
+		else
+			strlcpy(name, "usb_phy", sizeof(name));
+
+		ret = clk_get_by_name(dev, name, &phy->clocks);
+		if (ret) {
+			dev_err(dev, "failed to get usb%d_phy clock phandle\n", i);
+			return ret;
+		}
+
+		snprintf(name, sizeof(name), "usb%d_reset", i);
+		ret = reset_get_by_name(dev, name, &phy->resets);
+		if (ret) {
+			dev_err(dev, "failed to get usb%d_reset reset phandle\n", i);
+			return ret;
+		}
+
 		if (i || data->cfg->phy0_dual_route) {
 			snprintf(name, sizeof(name), "pmu%d", i);
 			phy->pmu = (void __iomem *)devfdt_get_addr_name(dev, name);
@@ -456,9 +486,6 @@ static int sun4i_usb_phy_probe(struct udevice *dev)
 		}
 
 		phy->id = i;
-		phy->rst_mask = info->rst_mask;
-		if ((data->cfg->type == sun8i_h3_phy) && (phy->id == 3))
-			phy->rst_mask = (BIT(3) | BIT(11));
 	};
 
 	debug("Allwinner Sun4I USB PHY driver loaded\n");
@@ -470,6 +497,7 @@ static const struct sun4i_usb_phy_cfg sun4i_a10_cfg = {
 	.type = sun4i_a10_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A10,
+	.dedicated_clocks = false,
 	.enable_pmu_unk1 = false,
 };
 
@@ -478,6 +506,7 @@ static const struct sun4i_usb_phy_cfg sun5i_a13_cfg = {
 	.type = sun4i_a10_phy,
 	.disc_thresh = 2,
 	.phyctl_offset = REG_PHYCTL_A10,
+	.dedicated_clocks = false,
 	.enable_pmu_unk1 = false,
 };
 
@@ -486,6 +515,7 @@ static const struct sun4i_usb_phy_cfg sun6i_a31_cfg = {
 	.type = sun6i_a31_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A10,
+	.dedicated_clocks = true,
 	.enable_pmu_unk1 = false,
 };
 
@@ -494,6 +524,7 @@ static const struct sun4i_usb_phy_cfg sun7i_a20_cfg = {
 	.type = sun4i_a10_phy,
 	.disc_thresh = 2,
 	.phyctl_offset = REG_PHYCTL_A10,
+	.dedicated_clocks = false,
 	.enable_pmu_unk1 = false,
 };
 
@@ -502,6 +533,7 @@ static const struct sun4i_usb_phy_cfg sun8i_a23_cfg = {
 	.type = sun4i_a10_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A10,
+	.dedicated_clocks = true,
 	.enable_pmu_unk1 = false,
 };
 
@@ -510,6 +542,7 @@ static const struct sun4i_usb_phy_cfg sun8i_a33_cfg = {
 	.type = sun8i_a33_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A33,
+	.dedicated_clocks = true,
 	.enable_pmu_unk1 = false,
 };
 
@@ -517,6 +550,7 @@ static const struct sun4i_usb_phy_cfg sun8i_a83t_cfg = {
 	.num_phys = 3,
 	.type = sun8i_a83t_phy,
 	.phyctl_offset = REG_PHYCTL_A33,
+	.dedicated_clocks = true,
 };
 
 static const struct sun4i_usb_phy_cfg sun8i_h3_cfg = {
@@ -524,6 +558,7 @@ static const struct sun4i_usb_phy_cfg sun8i_h3_cfg = {
 	.type = sun8i_h3_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A33,
+	.dedicated_clocks = true,
 	.enable_pmu_unk1 = true,
 	.phy0_dual_route = true,
 };
@@ -533,6 +568,7 @@ static const struct sun4i_usb_phy_cfg sun8i_v3s_cfg = {
 	.type = sun8i_v3s_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A33,
+	.dedicated_clocks = true,
 	.enable_pmu_unk1 = true,
 	.phy0_dual_route = true,
 };
@@ -542,6 +578,7 @@ static const struct sun4i_usb_phy_cfg sun50i_a64_cfg = {
 	.type = sun50i_a64_phy,
 	.disc_thresh = 3,
 	.phyctl_offset = REG_PHYCTL_A33,
+	.dedicated_clocks = true,
 	.enable_pmu_unk1 = true,
 	.phy0_dual_route = true,
 };
Signature · 1 line
2.18.0.321.gffc6fa0e3
#29 lore raw

PATCH v5 17/26 phy: sun4i-usb: Use CLK and RESET support

Quoted 6 lines · Jagan Teki

Now clock and reset drivers are available for respective SoC's so use clk and reset ops on phy driver.

btw I hope this is planned for _after_ 2019.01 release ?

[...]

Signature · 2 lines
Best regards,
Marek Vasut
author
#30 lore raw

PATCH v5 17/26 phy: sun4i-usb: Use CLK and RESET support

Quoted 5 lines · Marek Vasut
Quoted 3 lines · Jagan Teki

Now clock and reset drivers are available for respective SoC's so use clk and reset ops on phy driver.

Quoted 7 lines
Quoted 3 lines

btw I hope this is planned for _after_ 2019.01 release ?

Yes.

author
#31 lore raw

PATCH v5 18/26 reset: Add reset valid

In reply to Jagan Teki

Add reset_valid to check whether given reset is valid or not.

---
 include/reset.h | 11 +++++++++++
 1 file changed, 11 insertions(+)
include/reset.h +11 -0
index bc495a90c2..65aa7a4ce5 100644
@@ -306,4 +306,15 @@ static inline int reset_release_bulk(struct reset_ctl_bulk *bulk)
 }
 #endif
 
+/**
+ * reset_valid() - check if reset is valid
+ *
+ * @reset_ctl:		the reset to check
+ * @return TRUE if valid, or FALSE
+ */
+static inline bool reset_valid(struct reset_ctl *reset_ctl)
+{
+	return !!reset_ctl->dev;
+}
+
 #endif
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#32 lore raw

PATCH v5 19/26 musb-new: sunxi: Use CLK and RESET support

In reply to Jagan Teki

Now clock and reset drivers are available for respective SoC's so use clk and reset ops on musb driver.

---
 drivers/usb/musb-new/sunxi.c | 81 ++++++++++++++++++------------------
 1 file changed, 40 insertions(+), 41 deletions(-)
drivers/usb/musb-new/sunxi.c +40 -41
index f542a181fa..f79b3eedcf 100644
@@ -16,9 +16,11 @@
  * This file is part of the Inventra Controller Driver for Linux.
  */
 #include <common.h>
+#include <clk.h>
 #include <dm.h>
 #include <generic-phy.h>
 #include <phy-sun4i-usb.h>
+#include <reset.h>
 #include <asm/arch/cpu.h>
 #include <asm/arch/clock.h>
 #include <asm/arch/gpio.h>
@@ -80,16 +82,12 @@
 
 struct sunxi_musb_config {
 	struct musb_hdrc_config *config;
-	bool has_reset;
-	u8 rst_bit;
-	u8 clkgate_bit;
-	u32 off_reset0;
 };
 
 struct sunxi_glue {
 	struct musb_host_data mdata;
-	struct sunxi_ccm_reg *ccm;
-	u32 *reg_reset0;
+	struct clk clk;
+	struct reset_ctl rst;
 	struct sunxi_musb_config *cfg;
 	struct device dev;
 	struct phy phy;
@@ -296,24 +294,27 @@ static int sunxi_musb_init(struct musb *musb)
 
 	pr_debug("%s():\n", __func__);
 
-	ret = generic_phy_init(&glue->phy);
+	ret = clk_enable(&glue->clk);
 	if (ret) {
-		pr_err("failed to init USB PHY\n");
+		pr_err("failed to enable clock\n");
 		return ret;
 	}
 
-	musb->isr = sunxi_musb_interrupt;
-
-	setbits_le32(&glue->ccm->ahb_gate0, BIT(AHB_GATE_OFFSET_USB0));
-	if (glue->cfg->clkgate_bit)
-		setbits_le32(&glue->ccm->ahb_gate0,
-			     BIT(glue->cfg->clkgate_bit));
+	if (reset_valid(&glue->rst)) {
+		ret = reset_deassert(&glue->rst);
+		if (ret) {
+			pr_err("failed to deassert reset\n");
+			goto err_clk;
+		}
+	}
 
-	if (glue->cfg->has_reset)
-		setbits_le32(glue->reg_reset0, BIT(AHB_GATE_OFFSET_USB0));
+	ret = generic_phy_init(&glue->phy);
+	if (ret) {
+		pr_err("failed to init USB PHY\n");
+		goto err_rst;
+	}
 
-	if (glue->cfg->rst_bit)
-		setbits_le32(glue->reg_reset0, BIT(glue->cfg->rst_bit));
+	musb->isr = sunxi_musb_interrupt;
 
 	USBC_ConfigFIFO_Base();
 	USBC_EnableDpDmPullUp(musb->mregs);
@@ -329,6 +330,13 @@ static int sunxi_musb_init(struct musb *musb)
 	USBC_ForceVbusValidToHigh(musb->mregs);
 
 	return 0;
+
+err_rst:
+	if (reset_valid(&glue->rst))
+		reset_assert(&glue->rst);
+err_clk:
+	clk_disable(&glue->clk);
+	return ret;
 }
 
 static int sunxi_musb_exit(struct musb *musb)
@@ -339,21 +347,14 @@ static int sunxi_musb_exit(struct musb *musb)
 	if (generic_phy_valid(&glue->phy)) {
 		ret = generic_phy_exit(&glue->phy);
 		if (ret) {
-			dev_err(dev, "failed to power off usb phy\n");
+			pr_err("failed to power off usb phy\n");
 			return ret;
 		}
 	}
 
-	if (glue->cfg->has_reset)
-		clrbits_le32(glue->reg_reset0, BIT(AHB_GATE_OFFSET_USB0));
-
-	if (glue->cfg->rst_bit)
-		clrbits_le32(glue->reg_reset0, BIT(glue->cfg->rst_bit));
-
-	clrbits_le32(&glue->ccm->ahb_gate0, BIT(AHB_GATE_OFFSET_USB0));
-	if (glue->cfg->clkgate_bit)
-		clrbits_le32(&glue->ccm->ahb_gate0,
-			     BIT(glue->cfg->clkgate_bit));
+	if (reset_valid(&glue->rst))
+		reset_assert(&glue->rst);
+	clk_disable(&glue->clk);
 
 	return 0;
 }
@@ -450,11 +451,17 @@ static int musb_usb_probe(struct udevice *dev)
 	if (!glue->cfg)
 		return -EINVAL;
 
-	glue->ccm = (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
-	if (IS_ERR(glue->ccm))
-		return PTR_ERR(glue->ccm);
+	ret = clk_get_by_index(dev, 0, &glue->clk);
+	if (ret) {
+		pr_err("failed to get clock\n");
+		return ret;
+	}
 
-	glue->reg_reset0 = (void *)glue->ccm + glue->cfg->off_reset0;
+	ret = reset_get_by_index(dev, 0, &glue->rst);
+	if (ret && ret != -ENOENT) {
+		pr_err("failed to get reset\n");
+		return ret;
+	}
 
 	ret = generic_phy_get_by_name(dev, "usb", &glue->phy);
 	if (ret) {
@@ -462,7 +469,6 @@ static int musb_usb_probe(struct udevice *dev)
 		return ret;
 	}
 
-
 	memset(&pdata, 0, sizeof(pdata));
 	pdata.power = 250;
 	pdata.platform_ops = &sunxi_musb_ops;
@@ -505,21 +511,14 @@ static int musb_usb_remove(struct udevice *dev)
 
 static const struct sunxi_musb_config sun4i_a10_cfg = {
 	.config = &musb_config,
-	.has_reset = false,
 };
 
 static const struct sunxi_musb_config sun6i_a31_cfg = {
 	.config = &musb_config,
-	.has_reset = true,
-	.off_reset0 = OFF_SUN6I_AHB_RESET0,
 };
 
 static const struct sunxi_musb_config sun8i_h3_cfg = {
 	.config = &musb_config_h3,
-	.has_reset = true,
-	.rst_bit = 23,
-	.clkgate_bit = 23,
-	.off_reset0 = OFF_SUN6I_AHB_RESET0,
 };
 
 static const struct udevice_id sunxi_musb_ids[] = {
Signature · 1 line
2.18.0.321.gffc6fa0e3
#33 lore raw

PATCH v5 19/26 musb-new: sunxi: Use CLK and RESET support

Quoted 57 lines · Jagan Teki

Now clock and reset drivers are available for respective SoC's so use clk and reset ops on musb driver.

---
 drivers/usb/musb-new/sunxi.c | 81 ++++++++++++++++++------------------
 1 file changed, 40 insertions(+), 41 deletions(-)
drivers/usb/musb-new/sunxi.c +6 -8
index f542a181fa..f79b3eedcf 100644
@@ -16,9 +16,11 @@
  * This file is part of the Inventra Controller Driver for Linux.
  */
 #include <common.h>
+#include <clk.h>
 #include <dm.h>
 #include <generic-phy.h>
 #include <phy-sun4i-usb.h>
+#include <reset.h>
 #include <asm/arch/cpu.h>
 #include <asm/arch/clock.h>
 #include <asm/arch/gpio.h>
@@ -80,16 +82,12 @@
 
 struct sunxi_musb_config {
 	struct musb_hdrc_config *config;
-	bool has_reset;
-	u8 rst_bit;
-	u8 clkgate_bit;
-	u32 off_reset0;
 };
 
 struct sunxi_glue {
 	struct musb_host_data mdata;
-	struct sunxi_ccm_reg *ccm;
-	u32 *reg_reset0;
+	struct clk clk;
+	struct reset_ctl rst;
 	struct sunxi_musb_config *cfg;
 	struct device dev;
 	struct phy phy;
@@ -296,24 +294,27 @@ static int sunxi_musb_init(struct musb *musb)
 
 	pr_debug("%s():\n", __func__);
 
-	ret = generic_phy_init(&glue->phy);
+	ret = clk_enable(&glue->clk);
 	if (ret) {
-		pr_err("failed to init USB PHY\n");
+		pr_err("failed to enable clock\n");
 		return ret;
 	}
 

drivers/usb/musb-new/sunxi.c static int sunxi_musb_init(struct musb *musb) · line 306

Quoted 10 lines
drivers/usb/musb-new/sunxi.c +4 -6
-	musb->isr = sunxi_musb_interrupt;
-
-	setbits_le32(&glue->ccm->ahb_gate0, BIT(AHB_GATE_OFFSET_USB0));
-	if (glue->cfg->clkgate_bit)
-		setbits_le32(&glue->ccm->ahb_gate0,
-			     BIT(glue->cfg->clkgate_bit));
+	if (reset_valid(&glue->rst)) {
+		ret = reset_deassert(&glue->rst);
+		if (ret) {
+			pr_err("failed to deassert reset\n");

dev_err(dev, ...), fix globally

[...]

Signature · 2 lines
Best regards,
Marek Vasut
author
#34 lore raw

PATCH v5 20/26 sunxi: usb: Switch to Generic host controllers

In reply to Jagan Teki

Once of key blocker for using USB Generic host controller drivers in Allwinner are CLK and RESET drivers, now these available for USB usage.

So switch sunxi USB use EHCI and OHCI Generic controllers.

Enabling USB is wisely a board choise, So Enable USB_OHCI_HCD where it already have USB_EHCI_HCD

---
 configs/A10-OLinuXino-Lime_defconfig          | 1 +
 configs/A10s-OLinuXino-M_defconfig            | 1 +
 configs/A13-OLinuXinoM_defconfig              | 1 +
 configs/A13-OLinuXino_defconfig               | 1 +
 configs/A20-OLinuXino-Lime2-eMMC_defconfig    | 1 +
 configs/A20-OLinuXino-Lime2_defconfig         | 1 +
 configs/A20-OLinuXino-Lime_defconfig          | 1 +
 configs/A20-Olimex-SOM204-EVB_defconfig       | 1 +
 configs/Auxtek-T003_defconfig                 | 1 +
 configs/Auxtek-T004_defconfig                 | 1 +
 configs/Bananapi_defconfig                    | 1 +
 configs/Bananapi_m2m_defconfig                | 1 +
 configs/Bananapro_defconfig                   | 1 +
 configs/CHIP_defconfig                        | 1 +
 configs/CHIP_pro_defconfig                    | 1 +
 configs/CSQ_CS908_defconfig                   | 1 +
 configs/Colombus_defconfig                    | 1 +
 configs/Cubieboard2_defconfig                 | 1 +
 configs/Cubieboard_defconfig                  | 1 +
 configs/Cubietruck_plus_defconfig             | 1 +
 configs/Hummingbird_A31_defconfig             | 1 +
 configs/Itead_Ibox_A20_defconfig              | 1 +
 configs/Linksprite_pcDuino3_Nano_defconfig    | 1 +
 configs/Linksprite_pcDuino3_defconfig         | 1 +
 configs/Linksprite_pcDuino_defconfig          | 1 +
 configs/MK808C_defconfig                      | 1 +
 configs/Marsboard_A10_defconfig               | 1 +
 configs/Mele_A1000G_quad_defconfig            | 1 +
 configs/Mele_A1000_defconfig                  | 1 +
 configs/Mele_I7_defconfig                     | 1 +
 configs/Mele_M3_defconfig                     | 1 +
 configs/Mele_M5_defconfig                     | 1 +
 configs/Mele_M9_defconfig                     | 1 +
 configs/Mini-X_defconfig                      | 1 +
 configs/Orangepi_defconfig                    | 1 +
 configs/Orangepi_mini_defconfig               | 1 +
 configs/Sinlinx_SinA31s_defconfig             | 1 +
 configs/Sinlinx_SinA33_defconfig              | 1 +
 configs/Sinovoip_BPI_M2_Plus_defconfig        | 1 +
 configs/Sinovoip_BPI_M2_defconfig             | 1 +
 configs/Sinovoip_BPI_M3_defconfig             | 1 +
 configs/Wexler_TAB7200_defconfig              | 1 +
 configs/Wobo_i5_defconfig                     | 1 +
 configs/a64-olinuxino_defconfig               | 1 +
 configs/ba10_tv_box_defconfig                 | 1 +
 configs/bananapi_m1_plus_defconfig            | 1 +
 configs/bananapi_m64_defconfig                | 1 +
 configs/ga10h_v1_1_defconfig                  | 1 +
 configs/h8_homlet_v2_defconfig                | 1 +
 configs/i12-tvbox_defconfig                   | 1 +
 configs/icnova-a20-swac_defconfig             | 1 +
 configs/inet1_defconfig                       | 1 +
 configs/inet_q972_defconfig                   | 1 +
 configs/jesurun_q5_defconfig                  | 1 +
 configs/libretech_all_h3_cc_h2_plus_defconfig | 1 +
 configs/libretech_all_h3_cc_h3_defconfig      | 1 +
 configs/libretech_all_h3_cc_h5_defconfig      | 1 +
 configs/mixtile_loftq_defconfig               | 1 +
 configs/mk802_a10s_defconfig                  | 1 +
 configs/mk802_defconfig                       | 1 +
 configs/mk802ii_defconfig                     | 1 +
 configs/nanopi_a64_defconfig                  | 1 +
 configs/nanopi_m1_defconfig                   | 1 +
 configs/nanopi_m1_plus_defconfig              | 1 +
 configs/nanopi_neo2_defconfig                 | 1 +
 configs/nanopi_neo_air_defconfig              | 1 +
 configs/nanopi_neo_defconfig                  | 1 +
 configs/nanopi_neo_plus2_defconfig            | 1 +
 configs/orangepi_2_defconfig                  | 1 +
 configs/orangepi_lite_defconfig               | 1 +
 configs/orangepi_one_defconfig                | 1 +
 configs/orangepi_pc2_defconfig                | 1 +
 configs/orangepi_pc_defconfig                 | 1 +
 configs/orangepi_pc_plus_defconfig            | 1 +
 configs/orangepi_plus2e_defconfig             | 1 +
 configs/orangepi_plus_defconfig               | 1 +
 configs/orangepi_prime_defconfig              | 1 +
 configs/orangepi_r1_defconfig                 | 1 +
 configs/orangepi_win_defconfig                | 1 +
 configs/orangepi_zero_defconfig               | 1 +
 configs/orangepi_zero_plus2_defconfig         | 1 +
 configs/orangepi_zero_plus_defconfig          | 1 +
 configs/parrot_r16_defconfig                  | 1 +
 configs/pine64_plus_defconfig                 | 1 +
 configs/r7-tv-dongle_defconfig                | 1 +
 configs/sopine_baseboard_defconfig            | 1 +
 configs/sun8i_a23_evb_defconfig               | 1 +
 configs/sunxi_Gemei_G9_defconfig              | 1 +
 configs/tbs_a711_defconfig                    | 1 +
 drivers/usb/host/Kconfig                      | 2 ++
 include/configs/sun4i.h                       | 4 ----
 include/configs/sun50i.h                      | 5 -----
 include/configs/sun5i.h                       | 4 ----
 include/configs/sun6i.h                       | 4 ----
 include/configs/sun7i.h                       | 4 ----
 include/configs/sun8i.h                       | 4 ----
 include/configs/sunxi-common.h                | 1 -
 97 files changed, 91 insertions(+), 26 deletions(-)
configs/A10-OLinuXino-Lime_defconfig +1 -0
index 55134a9002..9b8891eed5 100644
@@ -22,5 +22,6 @@ CONFIG_SUN4I_EMAC=y
 CONFIG_AXP_ALDO3_VOLT=2800
 CONFIG_AXP_ALDO4_VOLT=2800
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A10s-OLinuXino-M_defconfig +1 -0
index d0fb4d7860..fcf4890c3a 100644
@@ -18,5 +18,6 @@ CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
 CONFIG_AXP152_POWER=y
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A13-OLinuXinoM_defconfig +1 -0
index 88374551e3..aedb0c3f0a 100644
@@ -19,5 +19,6 @@ CONFIG_NR_DRAM_BANKS=1
 CONFIG_DEFAULT_DEVICE_TREE="sun5i-a13-olinuxino-micro"
 CONFIG_SUNXI_NO_PMIC=y
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A13-OLinuXino_defconfig +1 -0
index 7b836ddf9c..eb76bf8fea 100644
@@ -26,6 +26,7 @@ CONFIG_DFU_RAM=y
 CONFIG_FASTBOOT_CMD_OEM_FORMAT=y
 CONFIG_AXP_ALDO3_VOLT=3300
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A20-OLinuXino-Lime2-eMMC_defconfig +1 -0
index dee8d028ee..41aed6edd9 100644
@@ -31,6 +31,7 @@ CONFIG_AXP_ALDO3_VOLT=2800
 CONFIG_AXP_ALDO3_VOLT_SLOPE_08=y
 CONFIG_AXP_ALDO4_VOLT=2800
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A20-OLinuXino-Lime2_defconfig +1 -0
index 0c04ae6491..e54049087a 100644
@@ -30,6 +30,7 @@ CONFIG_AXP_ALDO3_VOLT=2800
 CONFIG_AXP_ALDO3_VOLT_SLOPE_08=y
 CONFIG_AXP_ALDO4_VOLT=2800
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A20-OLinuXino-Lime_defconfig +1 -0
index 91f5210762..9e13792f1a 100644
@@ -21,5 +21,6 @@ CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO3_VOLT=2800
 CONFIG_AXP_ALDO4_VOLT=2800
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/A20-Olimex-SOM204-EVB_defconfig +1 -0
index 1181832b2e..e853c9301c 100644
@@ -29,6 +29,7 @@ CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO3_VOLT=2800
 CONFIG_AXP_ALDO4_VOLT=2800
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Auxtek-T003_defconfig +1 -0
index a693cfdd41..c2a681c569 100644
@@ -15,5 +15,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun5i-a10s-auxtek-t003"
 CONFIG_AXP152_POWER=y
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Auxtek-T004_defconfig +1 -0
index 5c88b2bec0..bc5fc8cee8 100644
@@ -13,5 +13,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun5i-a10s-auxtek-t004"
 CONFIG_AXP152_POWER=y
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Bananapi_defconfig +1 -0
index 41cb9fc225..67daec6d7f 100644
@@ -21,5 +21,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Bananapi_m2m_defconfig +1 -0
index 990c791b76..c8dfa7d51f 100644
@@ -15,6 +15,7 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_PARTITION_UUIDS is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-r16-bananapi-m2m"
 CONFIG_FASTBOOT_CMD_OEM_FORMAT=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_USB_FUNCTION_MASS_STORAGE=y
configs/Bananapro_defconfig +1 -0
index 41fc6a6eee..cd5d8b6cdd 100644
@@ -24,5 +24,6 @@ CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO4_VOLT=2500
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/CHIP_defconfig +1 -0
index ffdffcdb06..67fe58b21f 100644
@@ -17,6 +17,7 @@ CONFIG_DFU_RAM=y
 CONFIG_AXP_ALDO3_VOLT=3300
 CONFIG_AXP_ALDO4_VOLT=3300
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/CHIP_pro_defconfig +1 -0
index c607adb65b..618f113ff6 100644
@@ -22,6 +22,7 @@ CONFIG_SYS_NAND_OOBSIZE=0x100
 CONFIG_AXP_ALDO3_VOLT=3300
 CONFIG_AXP_ALDO4_VOLT=3300
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/CSQ_CS908_defconfig +1 -0
index ffc64d6aa4..9036ffc96a 100644
@@ -16,6 +16,7 @@ CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO1_VOLT=3300
 CONFIG_AXP_DLDO1_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Colombus_defconfig +1 -0
index 77c7e9ee29..cb65647c7a 100644
@@ -26,5 +26,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO1_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Cubieboard2_defconfig +1 -0
index 366f55b815..7fe5b83fa2 100644
@@ -18,5 +18,6 @@ CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Cubieboard_defconfig +1 -0
index efc3e2567a..ccb612a9c7 100644
@@ -17,5 +17,6 @@ CONFIG_SCSI_AHCI=y
 CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Cubietruck_plus_defconfig +1 -0
index 2f43bbd947..cecb428b02 100644
@@ -23,6 +23,7 @@ CONFIG_DEFAULT_DEVICE_TREE="sun8i-a83t-cubietruck-plus"
 CONFIG_AXP_DLDO3_VOLT=2500
 CONFIG_AXP_DLDO4_VOLT=3300
 CONFIG_AXP_FLDO1_VOLT=1200
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Hummingbird_A31_defconfig +1 -0
index b3e9d0a530..0f0b1c2251 100644
@@ -18,5 +18,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO1_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Itead_Ibox_A20_defconfig +1 -0
index 721bc87677..784c4de384 100644
@@ -18,5 +18,6 @@ CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Linksprite_pcDuino3_Nano_defconfig +1 -0
index 09aae78205..72d9b03929 100644
@@ -21,5 +21,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Linksprite_pcDuino3_defconfig +1 -0
index 00a49d1077..9156f132d1 100644
@@ -20,5 +20,6 @@ CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
 CONFIG_DM_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Linksprite_pcDuino_defconfig +1 -0
index 81dbb72c75..21cfea133d 100644
@@ -13,5 +13,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-pcduino"
 CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/MK808C_defconfig +1 -0
index f634ac894a..367a164554 100644
@@ -10,5 +10,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun7i-a20-mk808c"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Marsboard_A10_defconfig +1 -0
index ae7c26615f..c40e452aa4 100644
@@ -14,5 +14,6 @@ CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
 CONFIG_SUNXI_NO_PMIC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mele_A1000G_quad_defconfig +1 -0
index a2c613521e..feabbbf244 100644
@@ -19,6 +19,7 @@ CONFIG_AXP_DCDC1_VOLT=3300
 CONFIG_AXP_ALDO1_VOLT=3300
 CONFIG_AXP_DLDO1_VOLT=3300
 CONFIG_AXP_DLDO4_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mele_A1000_defconfig +1 -0
index 49672092e2..9375c5d424 100644
@@ -17,5 +17,6 @@ CONFIG_SCSI_AHCI=y
 CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mele_I7_defconfig +1 -0
index aac4fa12b2..eecf89b76e 100644
@@ -18,5 +18,6 @@ CONFIG_AXP_DCDC1_VOLT=3300
 CONFIG_AXP_ALDO1_VOLT=3300
 CONFIG_AXP_DLDO1_VOLT=3300
 CONFIG_AXP_DLDO4_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mele_M3_defconfig +1 -0
index acd98bc960..9738f8ba81 100644
@@ -17,5 +17,6 @@ CONFIG_DEFAULT_DEVICE_TREE="sun7i-a20-m3"
 CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mele_M5_defconfig +1 -0
index e473da772e..c76bc9f91f 100644
@@ -19,5 +19,6 @@ CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mele_M9_defconfig +1 -0
index 63bf3fd5b2..d3136c67cb 100644
@@ -18,5 +18,6 @@ CONFIG_AXP_DCDC1_VOLT=3300
 CONFIG_AXP_ALDO1_VOLT=3300
 CONFIG_AXP_DLDO1_VOLT=3300
 CONFIG_AXP_DLDO4_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Mini-X_defconfig +1 -0
index 161feea519..54b40d5696 100644
@@ -12,5 +12,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-mini-xplus"
 CONFIG_USB_EHCI_HCD=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Orangepi_defconfig +1 -0
index 3ea1d03d96..dae821a8d2 100644
@@ -23,5 +23,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Orangepi_mini_defconfig +1 -0
index 737c7a958e..e1cbbbd4e9 100644
@@ -25,5 +25,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Sinlinx_SinA31s_defconfig +1 -0
index dbc8c08e57..efa3b86b1f 100644
@@ -19,5 +19,6 @@ CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_DLDO1_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Sinlinx_SinA33_defconfig +1 -0
index df078c4b64..b92d62117c 100644
@@ -20,6 +20,7 @@ CONFIG_CMD_DFU=y
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-a33-sinlinx-sina33"
 CONFIG_DFU_RAM=y
 CONFIG_FASTBOOT_CMD_OEM_FORMAT=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Sinovoip_BPI_M2_Plus_defconfig +1 -0
index 35c4718e2d..c2025ee0a4 100644
@@ -14,6 +14,7 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-bananapi-m2-plus"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Sinovoip_BPI_M2_defconfig +1 -0
index 83ef4b4dce..fd69ac5170 100644
@@ -18,5 +18,6 @@ CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO1_VOLT=3300
 CONFIG_AXP_ALDO2_VOLT=1800
 CONFIG_AXP_DLDO1_VOLT=3000
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Sinovoip_BPI_M3_defconfig +1 -0
index 9f04731543..ec59feb4e4 100644
@@ -24,6 +24,7 @@ CONFIG_DEFAULT_DEVICE_TREE="sun8i-a83t-bananapi-m3"
 CONFIG_AXP_DCDC5_VOLT=1200
 CONFIG_AXP_DLDO3_VOLT=2500
 CONFIG_AXP_SW_ON=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Wexler_TAB7200_defconfig +1 -0
index 58c65b72bb..a2351dc925 100644
@@ -18,6 +18,7 @@ CONFIG_SPL_I2C_SUPPORT=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun7i-a20-wexler-tab7200"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/Wobo_i5_defconfig +1 -0
index a0dca6aa67..17f03c8579 100644
@@ -15,5 +15,6 @@ CONFIG_DEFAULT_DEVICE_TREE="sun5i-a10s-wobo-i5"
 CONFIG_AXP_ALDO3_VOLT=3300
 CONFIG_AXP_ALDO4_VOLT=3300
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/a64-olinuxino_defconfig +1 -0
index 01fcb86599..94ac94c6ab 100644
@@ -11,5 +11,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-olinuxino"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/ba10_tv_box_defconfig +1 -0
index 846fab2f5f..76a980e60f 100644
@@ -16,6 +16,7 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-ba10-tvbox"
 CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/bananapi_m1_plus_defconfig +1 -0
index 7286d2160b..46baf346c5 100644
@@ -21,4 +21,5 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_SCSI=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
configs/bananapi_m64_defconfig +1 -0
index aa0ad3c5f4..7dc768ff9f 100644
@@ -12,6 +12,7 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-bananapi-m64"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/ga10h_v1_1_defconfig +1 -0
index 4c88285e24..0049982407 100644
@@ -23,6 +23,7 @@ CONFIG_NR_DRAM_BANKS=1
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-a33-ga10h-v1.1"
 CONFIG_AXP_DLDO1_VOLT=3300
 CONFIG_CONS_INDEX=5
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/h8_homlet_v2_defconfig +1 -0
index 7aad6f0cdf..6e1cc00357 100644
@@ -16,6 +16,7 @@ CONFIG_CONSOLE_MUX=y
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-a83t-allwinner-h8homlet-v2"
 CONFIG_AXP_DLDO4_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/i12-tvbox_defconfig +1 -0
index 4b3a8a3d50..fea490d74d 100644
@@ -15,5 +15,6 @@ CONFIG_DEFAULT_DEVICE_TREE="sun7i-a20-i12-tvbox"
 CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/icnova-a20-swac_defconfig +1 -0
index 07c89a41e7..99cc7e889f 100644
@@ -22,5 +22,6 @@ CONFIG_DEFAULT_DEVICE_TREE="sun7i-a20-icnova-swac"
 CONFIG_ETH_DESIGNWARE=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/inet1_defconfig +1 -0
index 79e594bc15..b5b94d9c95 100644
@@ -19,6 +19,7 @@ CONFIG_SPL_I2C_SUPPORT=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-inet1"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/inet_q972_defconfig +1 -0
index d15f71ae7a..7f3d04db6d 100644
@@ -20,6 +20,7 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun6i-a31s-inet-q972"
 CONFIG_AXP_DLDO1_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/jesurun_q5_defconfig +1 -0
index ce485376f3..e6f90c09f2 100644
@@ -15,6 +15,7 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-jesurun-q5"
 CONFIG_MII=y
 CONFIG_SUN4I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_HOST=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/libretech_all_h3_cc_h2_plus_defconfig +1 -0
index 2c7a3ff9ef..9c2e28f1e8 100644
@@ -13,5 +13,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h2-plus-libretech-all-h3-cc"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/libretech_all_h3_cc_h3_defconfig +1 -0
index bd2c7083b0..9130a5980b 100644
@@ -13,5 +13,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-libretech-all-h3-cc"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/libretech_all_h3_cc_h5_defconfig +1 -0
index 001f53d0fe..eb0d5dc7af 100644
@@ -13,5 +13,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-libretech-all-h3-cc"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/mixtile_loftq_defconfig +1 -0
index 9ebb55c678..55d06497c9 100644
@@ -18,5 +18,6 @@ CONFIG_RGMII=y
 CONFIG_MII=y
 CONFIG_SUN7I_GMAC=y
 CONFIG_AXP_ALDO1_VOLT=3300
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/mk802_a10s_defconfig +1 -0
index 3fba55e70f..5a0b502145 100644
@@ -14,5 +14,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun5i-a10s-mk802"
 CONFIG_AXP152_POWER=y
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/mk802_defconfig +1 -0
index 1b59348f94..14f8feee20 100644
@@ -10,5 +10,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-mk802"
 CONFIG_SUNXI_NO_PMIC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/mk802ii_defconfig +1 -0
index 7a3f3d1f74..abd261db2f 100644
@@ -9,5 +9,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-mk802ii"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_a64_defconfig +1 -0
index cd3b59f2dc..4dfd36ee3c 100644
@@ -10,5 +10,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-nanopi-a64"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_m1_defconfig +1 -0
index bf4e874b12..c8efce81e4 100644
@@ -11,5 +11,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-nanopi-m1"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_m1_plus_defconfig +1 -0
index 66e90916f7..79b1c6913e 100644
@@ -13,5 +13,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-nanopi-m1-plus"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_neo2_defconfig +1 -0
index ee693be1a8..8892f6cd83 100644
@@ -11,5 +11,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-nanopi-neo2"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_neo_air_defconfig +1 -0
index 8c362fa885..149530416b 100644
@@ -13,5 +13,6 @@ CONFIG_CONSOLE_MUX=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-nanopi-neo-air"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_neo_defconfig +1 -0
index ffb854b4b8..b963a28ecc 100644
@@ -14,5 +14,6 @@ CONFIG_CONSOLE_MUX=y
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-nanopi-neo"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/nanopi_neo_plus2_defconfig +1 -0
index 331dc6a8f7..efb1737a75 100644
@@ -13,4 +13,5 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-nanopi-neo-plus2"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
configs/orangepi_2_defconfig +1 -0
index b86c0ab380..3ad13cb307 100644
@@ -16,5 +16,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-2"
 CONFIG_SUN8I_EMAC=y
 CONFIG_SY8106A_POWER=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_lite_defconfig +1 -0
index de08a8a2c6..e30df9be26 100644
@@ -11,5 +11,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-lite"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_one_defconfig +1 -0
index 003f2cc7cb..346a55f1d1 100644
@@ -12,5 +12,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-one"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_pc2_defconfig +1 -0
index 7f7e6d3264..a74fedda3f 100644
@@ -13,6 +13,7 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-orangepi-pc2"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_pc_defconfig +1 -0
index fd41e1a45b..14fcb66113 100644
@@ -14,5 +14,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-pc"
 CONFIG_SUN8I_EMAC=y
 CONFIG_SY8106A_POWER=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_pc_plus_defconfig +1 -0
index 1f0bf73149..0e7b4d3b1b 100644
@@ -15,5 +15,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-pc-plus"
 CONFIG_SUN8I_EMAC=y
 CONFIG_SY8106A_POWER=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_plus2e_defconfig +1 -0
index 9baccdef78..3e79c848c3 100644
@@ -16,5 +16,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-plus2e"
 CONFIG_SUN8I_EMAC=y
 CONFIG_SY8106A_POWER=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_plus_defconfig +1 -0
index 0588799b59..26bd9c439a 100644
@@ -18,5 +18,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h3-orangepi-plus"
 CONFIG_SUN8I_EMAC=y
 CONFIG_SY8106A_POWER=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_prime_defconfig +1 -0
index 7e10ebe796..d2429b767b 100644
@@ -11,6 +11,7 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-orangepi-prime"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_r1_defconfig +1 -0
index f587bec2ec..cd3355153c 100644
@@ -13,5 +13,6 @@ CONFIG_CONSOLE_MUX=y
 # CONFIG_CMD_FLASH is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h2-plus-orangepi-r1"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_win_defconfig +1 -0
index ab889ea8f5..78c55326c6 100644
@@ -11,5 +11,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-orangepi-win"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_zero_defconfig +1 -0
index 0da9c65093..e5a4c1d9fc 100644
@@ -13,5 +13,6 @@ CONFIG_CONSOLE_MUX=y
 # CONFIG_CMD_FLASH is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-h2-plus-orangepi-zero"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_zero_plus2_defconfig +1 -0
index 3c6ee03a53..7ae922774c 100644
@@ -13,5 +13,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-orangepi-zero-plus2"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/orangepi_zero_plus_defconfig +1 -0
index 5d783c5981..74133fa03b 100644
@@ -13,5 +13,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-h5-orangepi-zero-plus"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/parrot_r16_defconfig +1 -0
index a6071612c4..80c1bf1fbe 100644
@@ -18,6 +18,7 @@ CONFIG_NR_DRAM_BANKS=1
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-r16-parrot"
 CONFIG_FASTBOOT_CMD_OEM_FORMAT=y
 CONFIG_CONS_INDEX=5
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/pine64_plus_defconfig +1 -0
index a5b87b9063..ded4d134f1 100644
@@ -14,5 +14,6 @@ CONFIG_OF_LIST="sun50i-a64-pine64 sun50i-a64-pine64-plus"
 CONFIG_PHY_REALTEK=y
 CONFIG_RTL8211E_PINE64_GIGABIT_FIX=y
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/r7-tv-dongle_defconfig +1 -0
index 9c797c5512..d993b5d256 100644
@@ -13,5 +13,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 CONFIG_DEFAULT_DEVICE_TREE="sun5i-a10s-r7-tv-dongle"
 CONFIG_AXP152_POWER=y
 CONFIG_CONS_INDEX=2
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/sopine_baseboard_defconfig +1 -0
index 9ede081c08..0a189fc03d 100644
@@ -16,5 +16,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-sopine-baseboard"
 CONFIG_SUN8I_EMAC=y
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/sun8i_a23_evb_defconfig +1 -0
index 70c4072e22..563275b65f 100644
@@ -14,5 +14,6 @@ CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-a23-evb"
 CONFIG_CONS_INDEX=5
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/sunxi_Gemei_G9_defconfig +1 -0
index e07d7c7d9e..06a07df051 100644
@@ -16,5 +16,6 @@ CONFIG_SPL_I2C_SUPPORT=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun4i-a10-gemei-g9"
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_SYS_USB_EVENT_POLL_VIA_INT_QUEUE=y
configs/tbs_a711_defconfig +1 -0
index b2bcf3be57..103ce9be46 100644
@@ -19,6 +19,7 @@ CONFIG_NR_DRAM_BANKS=1
 CONFIG_DEFAULT_DEVICE_TREE="sun8i-a83t-tbs-a711"
 CONFIG_FASTBOOT_CMD_OEM_FORMAT=y
 CONFIG_AXP_DCDC5_VOLT=1200
+CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
 CONFIG_USB_MUSB_GADGET=y
 CONFIG_USB_FUNCTION_MASS_STORAGE=y
drivers/usb/host/Kconfig +2 -0
index a213c918bc..60f37f40fd 100644
@@ -190,6 +190,7 @@ config USB_EHCI_GENERIC
 	bool "Support for generic EHCI USB controller"
 	depends on OF_CONTROL
 	depends on DM_USB
+	default ARCH_SUNXI
 	default n
 	---help---
 	  Enables support for generic EHCI controller.
@@ -220,6 +221,7 @@ config USB_OHCI_GENERIC
 	bool "Support for generic OHCI USB controller"
 	depends on OF_CONTROL
 	depends on DM_USB
+	default ARCH_SUNXI
 	select USB_HOST
 	---help---
 	  Enables support for generic OHCI controller.
include/configs/sun4i.h +0 -4
index af079a71ee..6033760583 100644
@@ -11,10 +11,6 @@
  * A10 specific configuration
  */
 
-#ifdef CONFIG_USB_EHCI_HCD
-#define CONFIG_USB_EHCI_SUNXI
-#endif
-
 /*
  * Include common sunxi configuration where most the settings are
  */
include/configs/sun50i.h +0 -5
index 2d73c75b8c..e050a5299f 100644
@@ -10,11 +10,6 @@
  * A64 specific configuration
  */
 
-#ifdef CONFIG_USB_EHCI_HCD
-#define CONFIG_USB_EHCI_SUNXI
-#define CONFIG_USB_MAX_CONTROLLER_COUNT 1
-#endif
-
 #ifndef CONFIG_MACH_SUN50I_H6
 #define GICD_BASE		0x1c81000
 #define GICC_BASE		0x1c82000
include/configs/sun5i.h +0 -4
index c3692caa73..ee42af80d4 100644
@@ -11,10 +11,6 @@
  * High Level Configuration Options
  */
 
-#ifdef CONFIG_USB_EHCI_HCD
-#define CONFIG_USB_EHCI_SUNXI
-#endif
-
 /*
  * Include common sunxi configuration where most the settings are
  */
include/configs/sun6i.h +0 -4
index 1523684fad..1e490daac1 100644
@@ -14,10 +14,6 @@
  * A31 specific configuration
  */
 
-#ifdef CONFIG_USB_EHCI_HCD
-#define CONFIG_USB_EHCI_SUNXI
-#endif
-
 #define CONFIG_ARMV7_SECURE_BASE	SUNXI_SRAM_B_BASE
 #define CONFIG_ARMV7_SECURE_MAX_SIZE    (64 * 1024) /* 64 KB */
 
include/configs/sun7i.h +0 -4
index bb8f217b25..d2fd586672 100644
@@ -12,10 +12,6 @@
  * A20 specific configuration
  */
 
-#ifdef CONFIG_USB_EHCI_HCD
-#define CONFIG_USB_EHCI_SUNXI
-#endif
-
 #define CONFIG_ARMV7_SECURE_BASE	SUNXI_SRAM_B_BASE
 #define CONFIG_ARMV7_SECURE_MAX_SIZE	(64 * 1024) /* 64 KB */
 
include/configs/sun8i.h +0 -4
index 7dc8693b76..9b4675e4c3 100644
@@ -12,10 +12,6 @@
  * A23 specific configuration
  */
 
-#ifdef CONFIG_USB_EHCI_HCD
-#define CONFIG_USB_EHCI_SUNXI
-#endif
-
 /*
  * Include common sunxi configuration where most the settings are
  */
include/configs/sunxi-common.h +0 -1
index 9819d9980c..ed0cfc24f5 100644
@@ -295,7 +295,6 @@ extern int soft_i2c_gpio_scl;
 
 #ifdef CONFIG_USB_EHCI_HCD
 #define CONFIG_USB_OHCI_NEW
-#define CONFIG_USB_OHCI_SUNXI
 #define CONFIG_SYS_USB_OHCI_MAX_ROOT_PORTS 1
 #endif
 
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#35 lore raw

PATCH v5 21/26 usb: host: Drop [e-o]hci-sunxi drivers

In reply to Jagan Teki

Now Allwinner platform is all set to use Generic USB controller drivers, so remove the legacy sunxi drivers.

---
 drivers/usb/host/Makefile     |   2 -
 drivers/usb/host/ehci-sunxi.c | 204 -----------------------------
 drivers/usb/host/ohci-sunxi.c | 233 ----------------------------------
 scripts/config_whitelist.txt  |   2 -
 4 files changed, 441 deletions(-)
 delete mode 100644 drivers/usb/host/ehci-sunxi.c
 delete mode 100644 drivers/usb/host/ohci-sunxi.c
drivers/usb/host/Makefile +0 -2
index 948683a52b..6aa574f6f7 100644
@@ -15,7 +15,6 @@ obj-$(CONFIG_USB_OHCI_DA8XX) += ohci-da8xx.o
 obj-$(CONFIG_USB_R8A66597_HCD) += r8a66597-hcd.o
 obj-$(CONFIG_USB_SL811HS) += sl811-hcd.o
 obj-$(CONFIG_USB_OHCI_EP93XX) += ohci-ep93xx.o
-obj-$(CONFIG_USB_OHCI_SUNXI) += ohci-sunxi.o
 obj-$(CONFIG_USB_OHCI_LPC32XX) += ohci-lpc32xx.o
 obj-$(CONFIG_USB_OHCI_GENERIC) += ohci-generic.o
 
@@ -37,7 +36,6 @@ obj-$(CONFIG_USB_EHCI_MARVELL) += ehci-marvell.o
 obj-$(CONFIG_USB_EHCI_MSM) += ehci-msm.o
 obj-$(CONFIG_USB_EHCI_PCI) += ehci-pci.o
 obj-$(CONFIG_USB_EHCI_SPEAR) += ehci-spear.o
-obj-$(CONFIG_USB_EHCI_SUNXI) += ehci-sunxi.o
 obj-$(CONFIG_USB_EHCI_TEGRA) += ehci-tegra.o
 obj-$(CONFIG_USB_EHCI_VCT) += ehci-vct.o
 obj-$(CONFIG_USB_EHCI_VF) += ehci-vf.o
drivers/usb/host/ehci-sunxi.c deleted +0 -204

204 lines not on this page · raw message

drivers/usb/host/ohci-sunxi.c deleted +0 -233

233 lines not on this page · raw message

scripts/config_whitelist.txt +0 -2
index b8addeaf69..5ff10ee31e 100644
@@ -4548,7 +4548,6 @@ CONFIG_USB_EHCI_MX5
 CONFIG_USB_EHCI_MXC
 CONFIG_USB_EHCI_MXS
 CONFIG_USB_EHCI_SPEAR
-CONFIG_USB_EHCI_SUNXI
 CONFIG_USB_EHCI_TEGRA
 CONFIG_USB_EHCI_TXFIFO_THRESH
 CONFIG_USB_EHCI_VCT
@@ -4587,7 +4586,6 @@ CONFIG_USB_MUSB_TUSB6010
 CONFIG_USB_OHCI_EP93XX
 CONFIG_USB_OHCI_LPC32XX
 CONFIG_USB_OHCI_NEW
-CONFIG_USB_OHCI_SUNXI
 CONFIG_USB_OTG
 CONFIG_USB_OTG_BLACKLIST_HUB
 CONFIG_USB_PHY_CFG_BASE
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#36 lore raw

PATCH v5 22/26 clk: sunxi: Implement SPI clocks

In reply to Jagan Teki

Implement SPI AHB and MOD clocks for Allwinner A10/A20 and A10s/A13 SoC clock drivers via ccu clock gate table.

---
 drivers/clk/sunxi/clk_a10.c  | 9 +++++++++
 drivers/clk/sunxi/clk_a10s.c | 7 +++++++
 2 files changed, 16 insertions(+)
drivers/clk/sunxi/clk_a10.c +9 -0
index 1b074e3872..2b834856b3 100644
@@ -18,6 +18,15 @@ static struct ccu_clk_gate a10_gates[] = {
 	[CLK_AHB_OHCI0]		= GATE(0x060, BIT(2)),
 	[CLK_AHB_EHCI1]		= GATE(0x060, BIT(3)),
 	[CLK_AHB_OHCI1]		= GATE(0x060, BIT(4)),
+	[CLK_AHB_SPI0]		= GATE(0x060, BIT(20)),
+	[CLK_AHB_SPI1]		= GATE(0x060, BIT(21)),
+	[CLK_AHB_SPI2]		= GATE(0x060, BIT(22)),
+	[CLK_AHB_SPI3]		= GATE(0x060, BIT(23)),
+
+	[CLK_SPI0]		= GATE(0x0a0, BIT(31)),
+	[CLK_SPI1]		= GATE(0x0a4, BIT(31)),
+	[CLK_SPI2]		= GATE(0x0a8, BIT(31)),
+	[CLK_SPI3]		= GATE(0x0d4, BIT(31)),
 
 	[CLK_APB1_UART0]	= GATE(0x06c, BIT(16)),
 	[CLK_APB1_UART1]	= GATE(0x06c, BIT(17)),
drivers/clk/sunxi/clk_a10s.c +7 -0
index a50c6e3f7f..0cc48ed2e8 100644
@@ -16,6 +16,13 @@ static struct ccu_clk_gate a10s_gates[] = {
 	[CLK_AHB_OTG]		= GATE(0x060, BIT(0)),
 	[CLK_AHB_EHCI]		= GATE(0x060, BIT(1)),
 	[CLK_AHB_OHCI]		= GATE(0x060, BIT(2)),
+	[CLK_AHB_SPI0]		= GATE(0x060, BIT(20)),
+	[CLK_AHB_SPI1]		= GATE(0x060, BIT(21)),
+	[CLK_AHB_SPI2]		= GATE(0x060, BIT(22)),
+
+	[CLK_SPI0]		= GATE(0x0a0, BIT(31)),
+	[CLK_SPI1]		= GATE(0x0a4, BIT(31)),
+	[CLK_SPI2]		= GATE(0x0a8, BIT(31)),
 
 	[CLK_APB1_UART0]	= GATE(0x06c, BIT(16)),
 	[CLK_APB1_UART1]	= GATE(0x06c, BIT(17)),
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#37 lore raw

PATCH v5 23/26 spi: sun4i: Add CLK support

In reply to Jagan Teki

Add CLK support to enable AHB and MOD SPI clocks
on sun4i_spi driver.

---
 drivers/spi/sun4i_spi.c | 46 +++++++++++++++++++++++++++++++++++------
 1 file changed, 40 insertions(+), 6 deletions(-)
drivers/spi/sun4i_spi.c +40 -6
index b86b5a00ad..5b01aa4226 100644
@@ -19,6 +19,7 @@
  */
 
 #include <common.h>
+#include <clk.h>
 #include <dm.h>
 #include <spi.h>
 #include <errno.h>
@@ -114,6 +115,8 @@ struct sun4i_spi_platdata {
 
 struct sun4i_spi_priv {
 	struct sun4i_spi_regs *regs;
+	struct clk ahb_clk;
+	struct clk mod_clk;
 	u32 freq;
 	u32 mode;
 
@@ -238,13 +241,40 @@ static int sun4i_spi_parse_pins(struct udevice *dev)
 	return 0;
 }
 
-static inline void sun4i_spi_enable_clock(void)
+static inline int sun4i_spi_enable_clock(struct udevice *dev)
 {
-	struct sunxi_ccm_reg *const ccm =
-		(struct sunxi_ccm_reg *const)SUNXI_CCM_BASE;
+	struct sun4i_spi_priv *priv = dev_get_priv(dev);
+	int ret;
+
+	ret = clk_get_by_name(dev, "ahb", &priv->ahb_clk);
+	if (ret) {
+		dev_err(dev, "failed to get ahb clock\n");
+		return ret;
+	}
+
+	ret = clk_get_by_name(dev, "mod", &priv->mod_clk);
+	if (ret) {
+		dev_err(dev, "failed to get mod clock\n");
+		return ret;
+	}
+
+	ret = clk_enable(&priv->ahb_clk);
+	if (ret) {
+		dev_err(dev, "failed to enable ahb clock (ret=%d)\n", ret);
+		return ret;
+	}
+
+	ret = clk_enable(&priv->mod_clk);
+	if (ret) {
+		dev_err(dev, "failed to enable mod clock (ret=%d)\n", ret);
+		goto err_ahb;
+	}
 
-	setbits_le32(&ccm->ahb_gate0, (1 << AHB_GATE_OFFSET_SPI0));
-	writel((1 << 31), &ccm->spi0_clk_cfg);
+	return 0;
+
+err_ahb:
+	clk_disable(&priv->ahb_clk);
+	return ret;
 }
 
 static int sun4i_spi_ofdata_to_platdata(struct udevice *bus)
@@ -267,8 +297,12 @@ static int sun4i_spi_probe(struct udevice *bus)
 {
 	struct sun4i_spi_platdata *plat = dev_get_platdata(bus);
 	struct sun4i_spi_priv *priv = dev_get_priv(bus);
+	int ret;
+
+	ret = sun4i_spi_enable_clock(bus);
+	if (ret)
+		return ret;
 
-	sun4i_spi_enable_clock();
 	sun4i_spi_parse_pins(bus);
 
 	priv->regs = (struct sun4i_spi_regs *)(uintptr_t)plat->base_addr;
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#38 lore raw

PATCH v5 24/26 clk: sunxi: Implement A64 SPI clocks, resets

In reply to Jagan Teki

Implement SPI AHB, MOD clocks and resets for Allwinner A64.

---
 drivers/clk/sunxi/clk_a64.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/clk/sunxi/clk_a64.c +7 -0
index 1d0cd98183..09ff871aee 100644
@@ -52,6 +52,8 @@ static const struct ccu_clk_tree a64_tree[] = {
 };
 
 static const struct ccu_clk_gate a64_gates[] = {
+	[CLK_BUS_SPI0]		= GATE(0x060, BIT(20)),
+	[CLK_BUS_SPI1]		= GATE(0x060, BIT(21)),
 	[CLK_BUS_OTG]		= GATE(0x060, BIT(23)),
 	[CLK_BUS_EHCI0]		= GATE(0x060, BIT(24)),
 	[CLK_BUS_EHCI1]		= GATE(0x060, BIT(25)),
@@ -64,6 +66,9 @@ static const struct ccu_clk_gate a64_gates[] = {
 	[CLK_BUS_UART3]		= GATE(0x06c, BIT(19)),
 	[CLK_BUS_UART4]		= GATE(0x06c, BIT(20)),
 
+	[CLK_SPI0]		= GATE(0x0a0, BIT(31)),
+	[CLK_SPI1]		= GATE(0x0a4, BIT(31)),
+
 	[CLK_USB_PHY0]		= GATE(0x0cc, BIT(8)),
 	[CLK_USB_PHY1]		= GATE(0x0cc, BIT(9)),
 	[CLK_USB_HSIC]		= GATE(0x0cc, BIT(10)),
@@ -77,6 +82,8 @@ static const struct ccu_reset a64_resets[] = {
 	[RST_USB_PHY1]          = RESET(0x0cc, BIT(1)),
 	[RST_USB_HSIC]          = RESET(0x0cc, BIT(2)),
 
+	[RST_BUS_SPI0]		= RESET(0x2c0, BIT(20)),
+	[RST_BUS_SPI1]		= RESET(0x2c0, BIT(21)),
 	[RST_BUS_OTG]           = RESET(0x2c0, BIT(23)),
 	[RST_BUS_EHCI0]         = RESET(0x2c0, BIT(24)),
 	[RST_BUS_EHCI1]         = RESET(0x2c0, BIT(25)),
Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#39 lore raw

PATCH v5 25/26 spi: Add Allwinner A31 SPI driver

In reply to Jagan Teki

Add Allwinner sun6i SPI driver for A31, H3/H5 an A64.

---
 drivers/spi/Kconfig     |   6 +
 drivers/spi/Makefile    |   1 +
 drivers/spi/sun6i_spi.c | 398 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 405 insertions(+)
 create mode 100644 drivers/spi/sun6i_spi.c
drivers/spi/Kconfig +6 -0
index a7bb5b35c2..a7d5f70bc3 100644
@@ -223,6 +223,12 @@ config SUN4I_SPI
 	help
 	  SPI driver for Allwinner sun4i, sun5i and sun7i SoCs
 
+config SUN6I_SPI
+	bool "Allwinner A31 SPI controller"
+	depends on ARCH_SUNXI
+	help
+	  This enables using the SPI controller on the Allwinner A31 SoCs.
+
 config TEGRA114_SPI
 	bool "nVidia Tegra114 SPI driver"
 	help
drivers/spi/Makefile +1 -0
index 392a925795..54826171b5 100644
@@ -50,6 +50,7 @@ obj-$(CONFIG_SH_SPI) += sh_spi.o
 obj-$(CONFIG_SH_QSPI) += sh_qspi.o
 obj-$(CONFIG_STM32_QSPI) += stm32_qspi.o
 obj-$(CONFIG_SUN4I_SPI) += sun4i_spi.o
+obj-$(CONFIG_SUN6I_SPI) += sun6i_spi.o
 obj-$(CONFIG_TEGRA114_SPI) += tegra114_spi.o
 obj-$(CONFIG_TEGRA20_SFLASH) += tegra20_sflash.o
 obj-$(CONFIG_TEGRA20_SLINK) += tegra20_slink.o
drivers/spi/sun6i_spi.c new file +398 -0

398 lines not on this page · raw message

Signature · 1 line
2.18.0.321.gffc6fa0e3
author
#40 lore raw

PATCH v5 26/26 board: sopine: Enable SPI/SPI-FLASH

In reply to Jagan Teki

Sopine has Winbond SPI flash, so enable the same to use flash on Sopine board.

---
 .../dts/sun50i-a64-sopine-baseboard-u-boot.dtsi  | 16 ++++++++++++++++
 configs/sopine_baseboard_defconfig               |  7 +++++++
 2 files changed, 23 insertions(+)
 create mode 100644 arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi
arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file +16 -0
new file mode 100644
index 0000000000..9661b95d15
@@ -0,0 +1,16 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (C) 2018 Jagan Teki <[email protected]>
+ */
+
+/ {
+	aliases {
+		spi0 = &spi0;
+	};
+};
+
+&spi0  {
+	flash at 0 {
+		compatible = "spi-flash";
+	};
+};
configs/sopine_baseboard_defconfig +7 -0
index 0a189fc03d..5e6dc33d4d 100644
@@ -12,9 +12,16 @@ CONFIG_SPL_SPI_SUNXI=y
 CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SYS_MALLOC_CLEAR_ON_INIT is not set
 # CONFIG_CMD_FLASH is not set
+CONFIG_CMD_SF=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-sopine-baseboard"
+CONFIG_SPI=y
+CONFIG_DM_SPI=y
+CONFIG_SUN6I_SPI=y
+CONFIG_DM_SPI_FLASH=y
+CONFIG_SPI_FLASH=y
+CONFIG_SPI_FLASH_WINBOND=y
 CONFIG_SUN8I_EMAC=y
 CONFIG_USB_OHCI_HCD=y
 CONFIG_USB_EHCI_HCD=y
Signature · 1 line
2.18.0.321.gffc6fa0e3
#41 lore raw

PATCH v5 26/26 board: sopine: Enable SPI/SPI-FLASH

Quoted 41 lines · Jagan Teki

Sopine has Winbond SPI flash, so enable the same to use flash on Sopine board.

---
 .../dts/sun50i-a64-sopine-baseboard-u-boot.dtsi  | 16 ++++++++++++++++
 configs/sopine_baseboard_defconfig               |  7 +++++++
 2 files changed, 23 insertions(+)
 create mode 100644 arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi
arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file +16 -0
new file mode 100644
index 0000000000..9661b95d15
@@ -0,0 +1,16 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (C) 2018 Jagan Teki <[email protected]>
+ */
+
+/ {
+	aliases {
+		spi0 = &spi0;
+	};
+};
+
+&spi0  {
+	flash at 0 {
+		compatible = "spi-flash";
+	};
+};
configs/sopine_baseboard_defconfig +0 -0
index 0a189fc03d..5e6dc33d4d 100644
@@ -12,9 +12,16 @@ CONFIG_SPL_SPI_SUNXI=y
 CONFIG_NR_DRAM_BANKS=1
 # CONFIG_SYS_MALLOC_CLEAR_ON_INIT is not set
 # CONFIG_CMD_FLASH is not set

configs/sopine_baseboard_defconfig CONFIG_SPL_SPI_SUNXI=y · line 24

Quoted 10 lines
configs/sopine_baseboard_defconfig +7 -0
+CONFIG_CMD_SF=y
 # CONFIG_SPL_DOS_PARTITION is not set
 # CONFIG_SPL_EFI_PARTITION is not set
 CONFIG_DEFAULT_DEVICE_TREE="sun50i-a64-sopine-baseboard"
+CONFIG_SPI=y
+CONFIG_DM_SPI=y
+CONFIG_SUN6I_SPI=y
+CONFIG_DM_SPI_FLASH=y
+CONFIG_SPI_FLASH=y
+CONFIG_SPI_FLASH_WINBOND=y

SPI, DM_SPI and SUN6I_SPI seem redundant to me, just like SPI_FLASH_WINBOND, DM_SPI_FLASH, and SPI_FLASH.

We should do better to provide the needed Kconfig select/depends on/defaults to have to enable only the SPI_FLASH_WINBOND. Everything else can be derived from that.

Maxime

Signature · 10 lines
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190107/04c82823/attachment.sig>
#42 lore raw

PATCH v5 26/26 board: sopine: Enable SPI/SPI-FLASH

In reply to Jagan Teki

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi

Quoted 16 lines · Jagan Teki

Sopine has Winbond SPI flash, so enable the same to use flash on Sopine board.

---
 .../dts/sun50i-a64-sopine-baseboard-u-boot.dtsi  | 16 ++++++++++++++++
 configs/sopine_baseboard_defconfig               |  7 +++++++
 2 files changed, 23 insertions(+)
 create mode 100644 arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi
arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file +0 -0
new file mode 100644
index 0000000000..9661b95d15

Where does this dtsi get included?

Alex

#43 lore raw

PATCH v5 26/26 board: sopine: Enable SPI/SPI-FLASH

Quoted 13 lines · On Tue, 22 Jan 2019 17:32:27 +0100 Alexander Graf <[email protected]> wrote:
Quoted 12 lines · Jagan Teki

Sopine has Winbond SPI flash, so enable the same to use flash on Sopine board.

---
 .../dts/sun50i-a64-sopine-baseboard-u-boot.dtsi  | 16
++++++++++++++++ configs/sopine_baseboard_defconfig
|  7 +++++++ 2 files changed, 23 insertions(+)
 create mode 100644

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi

Quoted 7 lines

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi

Quoted 5 lines
(quoted diff) +0 -0

b/arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file mode 100644 index 0000000000..9661b95d15

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file +0 -0

Where does this dtsi get included?

This is some U-Boot build system magic, more precisely by binman:
http://git.denx.de/?p=u-boot.git;a=blob;f=tools/binman/README;hb=HEAD#l564

Cheers,
Andre.

#44 lore raw

PATCH v5 26/26 board: sopine: Enable SPI/SPI-FLASH

Quoted 24 lines · Andre Przywara
Quoted 13 lines · On Tue, 22 Jan 2019 17:32:27 +0100 Alexander Graf <[email protected]> wrote:
Quoted 12 lines · Jagan Teki

Sopine has Winbond SPI flash, so enable the same to use flash on Sopine board.

---
 .../dts/sun50i-a64-sopine-baseboard-u-boot.dtsi  | 16
++++++++++++++++ configs/sopine_baseboard_defconfig
|  7 +++++++ 2 files changed, 23 insertions(+)
 create mode 100644

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi

Quoted 7 lines

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi

Quoted 5 lines
(quoted diff) +0 -0

b/arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file mode 100644 index 0000000000..9661b95d15

arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi new file +0 -0

Where does this dtsi get included?

Quoted 2 lines

This is some U-Boot build system magic, more precisely by binman:
http://git.denx.de/?p=u-boot.git;a=blob;f=tools/binman/README;hb=HEAD#l564

Well, scripts/Makefile.lib has the logic for searching for the most appropriate -u-boot.dtsi file.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190122/bf5be876/attachment.sig>
author
#45 lore raw

v5 cover clk: Add Allwinner CLK, RESET support

In reply to Jagan Teki
Quoted 57 lines · Jagan Teki

Although the previous version[1] is properly handled the clock gates with enable and disable management, but this series is trying to add some more complex Allwinner CLK architecture by handling parent clock and other CLK attributes.

Allwinner Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management and these can be handled via ccu_clock_gate, which is almost same as previous version changes.

Tree clock has more Allwinner CLK attributes like clock type, fixed clocks, misc clocks, mp, nk, nkm, nkmp, pre/post div, flags etc and these can be managed via ccu_clock_tree.

On top of initial clock gates supported from previous version, this series is trying to add support for MP, NK, MISC, FIXED clock types with get_rate functionality and that can eventually used by uart driver.

On the summary, this would be an initial infrasture that can fit into remaining clock handle support like set_rate, so the rest of code will add on the requirement basics.

Once this is fine, I will try to add code for other parts especially for MMC since we have migration deadline for BLK, MMC, SCSI.

So, please do let me know if anyone have any inputs.

All these changes available at u-boot-sunxi/clk-next

thanks,
Jagan.

[1] https://patchwork.ozlabs.org/cover/962226/

Jagan Teki (26):
clk: Add Allwinner A64 CLK driver
reset: Add Allwinner RESET driver
clk: sunxi: Add Allwinner H3/H5 CLK driver
clk: sunxi: Add Allwinner A10/A20 CLK driver
clk: sunxi: Add Allwinner A10s/A13 CLK driver
clk: sunxi: Add Allwinner A31 CLK driver
clk: sunxi: Add Allwinner A23/A33 CLK driver
clk: sunxi: Add Allwinner A83T CLK driver
clk: sunxi: Add Allwinner R40 CLK driver
clk: sunxi: Add Allwinner V3S CLK driver
clk: sunxi: Implement UART clocks
clk: sunxi: Implement UART resets
clk: sunxi: Add Allwinner H6 CLK driver
sunxi: A64: Update sun50i-a64-ccu.h
clk: sunxi: Add ccu clock tree support
sunxi: Enable CLK
phy: sun4i-usb: Use CLK and RESET support
reset: Add reset valid

Quoted 8 lines

musb-new: sunxi: Use CLK and RESET support
sunxi: usb: Switch to Generic host controllers
usb: host: Drop [e-o]hci-sunxi drivers
clk: sunxi: Implement SPI clocks
spi: sun4i: Add CLK support
clk: sunxi: Implement A64 SPI clocks, resets
spi: Add Allwinner A31 SPI driver
board: sopine: Enable SPI/SPI-FLASH

I'm planning to pick gate clock patches and keep tree based clock for next version, let me know if anyone have any questions.

#46 lore raw

v5 cover clk: Add Allwinner CLK, RESET support

In reply to Jagan Teki

On 31/12/2018 16:59, Jagan Teki wrote:

Hi Jagan,

many thanks for picking this up, I was about to come back to this myself. I am looking at the pinctrl part at the moment, so good you are working on the clocks!

TL;DR: I am good with the first patches, but would like to drop the last five 5 patches from this series, and discuss the whole tree approach (or at least the patch) further.

Quoted 5 lines

Although the previous version[1] is properly handled the clock gates with enable and disable management, but this series is trying to add some more complex Allwinner CLK architecture by handling parent clock and other CLK attributes.

Quoted 7 lines

Allwinner Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management and these can be handled via ccu_clock_gate, which is almost same as previous version changes.

So I like this part very much, all those gates and resets are nicely readable.
Also the removal of the special sunxi USB drivers is very welcome. So up until patch including 21/26, except 15/26, I am all in.

Quoted 3 lines

Tree clock has more Allwinner CLK attributes like clock type, fixed clocks, misc clocks, mp, nk, nkm, nkmp, pre/post div, flags etc and these can be managed via ccu_clock_tree.

Mmmh, this looks quite complex for U-Boot's very limited needs to me. As far as I can see we basically have a static setup for most devices, the ones where we actually change the parent clocks are only clock consumers, namely DE/TCON and MMC. The former chooses a parent once, based on the calculated clock rate. Only for MMC we switch it from OSC24 (for the initial SPI-like setup phase) to PLL_PERIPH0 at runtime. But however this does not really affect the tree, and also we never do re-parenting.
Under this premises, can't we do this somewhat simpler? At the end of the day this is U-Boot, not a sophisticated kernel, and we got away with quite simple clock code so far.

Meanwhile I am struggling to understand your approach, which looks impressive on the first glance, but leave me scratching my head when looking at details. I will comment on 15/26 for the parts I am wondering about. I was trying to add MMC support and hit some issues.

On the SPI part:
I am bit puzzled about the last 5 patches dealing with SPI. I see that they depend on the others, but would like to deal with them separately.

First, patch 25/26 (the sun6i SPI driver) is a bit surprising to find in here. I guess you want to have an easy user for another clock, but I believe this should be a separate series. There were other posts already for a sun6i driver, also I think we should have a unified driver instead of duplicating most of it again. See Oskari's latest post.

Secondly: I don't think the SPI clocks are correct. You model them as simple gates only, kind of ignoring the divider settings. This is what the current sun4i driver does as well, but: it writes 1U << 31 into the clock register, which explicitly writes 0 to the other bits, setting 24 MHz as the input rate. Your approach would read-modify-write the register, just setting bit 31 and leaving the other ones as they were before.
So why isn't that using your new tree approach? Looks like a standard mux/divider clock to me.

Cheers,
Andre.

Quoted 199 lines

On top of initial clock gates supported from previous version, this series is trying to add support for MP, NK, MISC, FIXED clock types with get_rate functionality and that can eventually used by uart driver.

On the summary, this would be an initial infrasture that can fit into remaining clock handle support like set_rate, so the rest of code will add on the requirement basics.

Once this is fine, I will try to add code for other parts especially for MMC since we have migration deadline for BLK, MMC, SCSI.

So, please do let me know if anyone have any inputs.

All these changes available at u-boot-sunxi/clk-next

thanks,
Jagan.

[1] https://patchwork.ozlabs.org/cover/962226/

Jagan Teki (26):
clk: Add Allwinner A64 CLK driver
reset: Add Allwinner RESET driver
clk: sunxi: Add Allwinner H3/H5 CLK driver
clk: sunxi: Add Allwinner A10/A20 CLK driver
clk: sunxi: Add Allwinner A10s/A13 CLK driver
clk: sunxi: Add Allwinner A31 CLK driver
clk: sunxi: Add Allwinner A23/A33 CLK driver
clk: sunxi: Add Allwinner A83T CLK driver
clk: sunxi: Add Allwinner R40 CLK driver
clk: sunxi: Add Allwinner V3S CLK driver
clk: sunxi: Implement UART clocks
clk: sunxi: Implement UART resets
clk: sunxi: Add Allwinner H6 CLK driver
sunxi: A64: Update sun50i-a64-ccu.h
clk: sunxi: Add ccu clock tree support
sunxi: Enable CLK
phy: sun4i-usb: Use CLK and RESET support
reset: Add reset valid
musb-new: sunxi: Use CLK and RESET support
sunxi: usb: Switch to Generic host controllers
usb: host: Drop [e-o]hci-sunxi drivers
clk: sunxi: Implement SPI clocks
spi: sun4i: Add CLK support
clk: sunxi: Implement A64 SPI clocks, resets
spi: Add Allwinner A31 SPI driver
board: sopine: Enable SPI/SPI-FLASH

 .../sun50i-a64-sopine-baseboard-u-boot.dtsi   |  16 +
 arch/arm/include/asm/arch-sunxi/ccu.h         | 284 +++++++++++++
 arch/arm/mach-sunxi/Kconfig                   |  12 +
 configs/A10-OLinuXino-Lime_defconfig          |   1 +
 configs/A10s-OLinuXino-M_defconfig            |   1 +
 configs/A13-OLinuXinoM_defconfig              |   1 +
 configs/A13-OLinuXino_defconfig               |   1 +
 configs/A20-OLinuXino-Lime2-eMMC_defconfig    |   1 +
 configs/A20-OLinuXino-Lime2_defconfig         |   1 +
 configs/A20-OLinuXino-Lime_defconfig          |   1 +
 configs/A20-Olimex-SOM204-EVB_defconfig       |   1 +
 configs/Auxtek-T003_defconfig                 |   1 +
 configs/Auxtek-T004_defconfig                 |   1 +
 configs/Bananapi_defconfig                    |   1 +
 configs/Bananapi_m2m_defconfig                |   1 +
 configs/Bananapro_defconfig                   |   1 +
 configs/CHIP_defconfig                        |   1 +
 configs/CHIP_pro_defconfig                    |   1 +
 configs/CSQ_CS908_defconfig                   |   1 +
 configs/Colombus_defconfig                    |   1 +
 configs/Cubieboard2_defconfig                 |   1 +
 configs/Cubieboard_defconfig                  |   1 +
 configs/Cubietruck_plus_defconfig             |   1 +
 configs/Hummingbird_A31_defconfig             |   1 +
 configs/Itead_Ibox_A20_defconfig              |   1 +
 configs/Linksprite_pcDuino3_Nano_defconfig    |   1 +
 configs/Linksprite_pcDuino3_defconfig         |   1 +
 configs/Linksprite_pcDuino_defconfig          |   1 +
 configs/MK808C_defconfig                      |   1 +
 configs/Marsboard_A10_defconfig               |   1 +
 configs/Mele_A1000G_quad_defconfig            |   1 +
 configs/Mele_A1000_defconfig                  |   1 +
 configs/Mele_I7_defconfig                     |   1 +
 configs/Mele_M3_defconfig                     |   1 +
 configs/Mele_M5_defconfig                     |   1 +
 configs/Mele_M9_defconfig                     |   1 +
 configs/Mini-X_defconfig                      |   1 +
 configs/Orangepi_defconfig                    |   1 +
 configs/Orangepi_mini_defconfig               |   1 +
 configs/Sinlinx_SinA31s_defconfig             |   1 +
 configs/Sinlinx_SinA33_defconfig              |   1 +
 configs/Sinovoip_BPI_M2_Plus_defconfig        |   1 +
 configs/Sinovoip_BPI_M2_defconfig             |   1 +
 configs/Sinovoip_BPI_M3_defconfig             |   1 +
 configs/Wexler_TAB7200_defconfig              |   1 +
 configs/Wobo_i5_defconfig                     |   1 +
 configs/a64-olinuxino_defconfig               |   1 +
 configs/ba10_tv_box_defconfig                 |   1 +
 configs/bananapi_m1_plus_defconfig            |   1 +
 configs/bananapi_m64_defconfig                |   1 +
 configs/ga10h_v1_1_defconfig                  |   1 +
 configs/h8_homlet_v2_defconfig                |   1 +
 configs/i12-tvbox_defconfig                   |   1 +
 configs/icnova-a20-swac_defconfig             |   1 +
 configs/inet1_defconfig                       |   1 +
 configs/inet_q972_defconfig                   |   1 +
 configs/jesurun_q5_defconfig                  |   1 +
 configs/libretech_all_h3_cc_h2_plus_defconfig |   1 +
 configs/libretech_all_h3_cc_h3_defconfig      |   1 +
 configs/libretech_all_h3_cc_h5_defconfig      |   1 +
 configs/mixtile_loftq_defconfig               |   1 +
 configs/mk802_a10s_defconfig                  |   1 +
 configs/mk802_defconfig                       |   1 +
 configs/mk802ii_defconfig                     |   1 +
 configs/nanopi_a64_defconfig                  |   1 +
 configs/nanopi_m1_defconfig                   |   1 +
 configs/nanopi_m1_plus_defconfig              |   1 +
 configs/nanopi_neo2_defconfig                 |   1 +
 configs/nanopi_neo_air_defconfig              |   1 +
 configs/nanopi_neo_defconfig                  |   1 +
 configs/nanopi_neo_plus2_defconfig            |   1 +
 configs/orangepi_2_defconfig                  |   1 +
 configs/orangepi_lite_defconfig               |   1 +
 configs/orangepi_one_defconfig                |   1 +
 configs/orangepi_pc2_defconfig                |   1 +
 configs/orangepi_pc_defconfig                 |   1 +
 configs/orangepi_pc_plus_defconfig            |   1 +
 configs/orangepi_plus2e_defconfig             |   1 +
 configs/orangepi_plus_defconfig               |   1 +
 configs/orangepi_prime_defconfig              |   1 +
 configs/orangepi_r1_defconfig                 |   1 +
 configs/orangepi_win_defconfig                |   1 +
 configs/orangepi_zero_defconfig               |   1 +
 configs/orangepi_zero_plus2_defconfig         |   1 +
 configs/orangepi_zero_plus_defconfig          |   1 +
 configs/parrot_r16_defconfig                  |   1 +
 configs/pine64_plus_defconfig                 |   1 +
 configs/r7-tv-dongle_defconfig                |   1 +
 configs/sopine_baseboard_defconfig            |   8 +
 configs/sun8i_a23_evb_defconfig               |   1 +
 configs/sunxi_Gemei_G9_defconfig              |   1 +
 configs/tbs_a711_defconfig                    |   1 +
 drivers/clk/Kconfig                           |   1 +
 drivers/clk/Makefile                          |   1 +
 drivers/clk/sunxi/Kconfig                     |  82 ++++
 drivers/clk/sunxi/Makefile                    |  18 +
 drivers/clk/sunxi/clk_a10.c                   |  77 ++++
 drivers/clk/sunxi/clk_a10s.c                  |  68 +++
 drivers/clk/sunxi/clk_a23.c                   |  75 ++++
 drivers/clk/sunxi/clk_a31.c                   |  82 ++++
 drivers/clk/sunxi/clk_a64.c                   | 125 ++++++
 drivers/clk/sunxi/clk_a83t.c                  |  75 ++++
 drivers/clk/sunxi/clk_h3.c                    |  89 ++++
 drivers/clk/sunxi/clk_h6.c                    |  53 +++
 drivers/clk/sunxi/clk_r40.c                   |  88 ++++
 drivers/clk/sunxi/clk_sunxi.c                 | 256 +++++++++++
 drivers/clk/sunxi/clk_v3s.c                   |  59 +++
 drivers/phy/allwinner/phy-sun4i-usb.c         |  77 +++-
 drivers/reset/Kconfig                         |   8 +
 drivers/reset/Makefile                        |   1 +
 drivers/reset/reset-sunxi.c                   | 125 ++++++
 drivers/spi/Kconfig                           |   6 +
 drivers/spi/Makefile                          |   1 +
 drivers/spi/sun4i_spi.c                       |  46 +-
 drivers/spi/sun6i_spi.c                       | 398 ++++++++++++++++++
 drivers/usb/host/Kconfig                      |   2 +
 drivers/usb/host/Makefile                     |   2 -
 drivers/usb/host/ehci-sunxi.c                 | 204 ---------
 drivers/usb/host/ohci-sunxi.c                 | 233 ----------
 drivers/usb/musb-new/sunxi.c                  |  81 ++--
 include/configs/sun4i.h                       |   4 -
 include/configs/sun50i.h                      |   5 -
 include/configs/sun5i.h                       |   4 -
 include/configs/sun6i.h                       |   4 -
 include/configs/sun7i.h                       |   4 -
 include/configs/sun8i.h                       |   4 -
 include/configs/sunxi-common.h                |   1 -
 include/dt-bindings/clock/sun50i-a64-ccu.h    |   2 +
 include/reset.h                               |  11 +
 scripts/config_whitelist.txt                  |   2 -
 130 files changed, 2248 insertions(+), 534 deletions(-)
 create mode 100644 arch/arm/dts/sun50i-a64-sopine-baseboard-u-boot.dtsi
 create mode 100644 arch/arm/include/asm/arch-sunxi/ccu.h
 create mode 100644 drivers/clk/sunxi/Kconfig
 create mode 100644 drivers/clk/sunxi/Makefile
 create mode 100644 drivers/clk/sunxi/clk_a10.c
 create mode 100644 drivers/clk/sunxi/clk_a10s.c
 create mode 100644 drivers/clk/sunxi/clk_a23.c
 create mode 100644 drivers/clk/sunxi/clk_a31.c
 create mode 100644 drivers/clk/sunxi/clk_a64.c
 create mode 100644 drivers/clk/sunxi/clk_a83t.c
 create mode 100644 drivers/clk/sunxi/clk_h3.c
 create mode 100644 drivers/clk/sunxi/clk_h6.c
 create mode 100644 drivers/clk/sunxi/clk_r40.c
 create mode 100644 drivers/clk/sunxi/clk_sunxi.c
 create mode 100644 drivers/clk/sunxi/clk_v3s.c
 create mode 100644 drivers/reset/reset-sunxi.c
 create mode 100644 drivers/spi/sun6i_spi.c
 delete mode 100644 drivers/usb/host/ehci-sunxi.c
 delete mode 100644 drivers/usb/host/ohci-sunxi.c
author
#47 lore raw

v5 cover clk: Add Allwinner CLK, RESET support

Quoted 26 lines · André Przywara

On 31/12/2018 16:59, Jagan Teki wrote:

Hi Jagan,

many thanks for picking this up, I was about to come back to this myself. I am looking at the pinctrl part at the moment, so good you are working on the clocks!

TL;DR: I am good with the first patches, but would like to drop the last five 5 patches from this series, and discuss the whole tree approach (or at least the patch) further.

Quoted 5 lines

Although the previous version[1] is properly handled the clock gates with enable and disable management, but this series is trying to add some more complex Allwinner CLK architecture by handling parent clock and other CLK attributes.

Quoted 7 lines

Allwinner Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management and these can be handled via ccu_clock_gate, which is almost same as previous version changes.

Quoted 4 lines

So I like this part very much, all those gates and resets are nicely readable.
Also the removal of the special sunxi USB drivers is very welcome. So up until patch including 21/26, except 15/26, I am all in.

Yes my plan is to do the same.

Quoted 5 lines
Quoted 4 lines

Tree clock has more Allwinner CLK attributes like clock type, fixed clocks, misc clocks, mp, nk, nkm, nkmp, pre/post div, flags etc and these can be managed via ccu_clock_tree.

Quoted 11 lines

Mmmh, this looks quite complex for U-Boot's very limited needs to me. As far as I can see we basically have a static setup for most devices, the ones where we actually change the parent clocks are only clock consumers, namely DE/TCON and MMC. The former chooses a parent once, based on the calculated clock rate. Only for MMC we switch it from OSC24 (for the initial SPI-like setup phase) to PLL_PERIPH0 at runtime. But however this does not really affect the tree, and also we never do re-parenting.
Under this premises, can't we do this somewhat simpler? At the end of the day this is U-Boot, not a sophisticated kernel, and we got away with quite simple clock code so far.

Yes, Idea is to go with limited clock setups but the thing is as you know Allwinner deals numerous number of clock attributes and I found it difficult to handle them w/o traversing the re-parenting stuff. I have tried simple version of set_rate in previous version[2] but it wouldn't ended as we desired. These operations need re-parenting, so we have to handle parents in clock attributes.

So finally I came up with tree structure.

My idea here is to support get and set rate with minimal re-parenting like OSC24, PLL_PERIPH0, PLL_PERIPH1 not adding much code like Linux does. Like GATE macro, TREE macro is also a readable structure with all possible/needed clk attributes at one glance. The code that handle these tree attributes may look confusing, but it is obvious because it would need to handle re-parenting.

If SoC has a big CLK architecture like this and to support all or needed clock operations, re-parenting may be one of painful code for SoC driver in U-Boot, Since CLK framework doesn't handle these generically(may be no need) as compared to Linux. and the existing SoC like mediatek clock code does the same.

At, this point, I think this tree approach can be possible way to move since it handle all clock attribute types and yes we can only support clock re-parenting which are required for U-Boot.

Quoted 4 lines

Meanwhile I am struggling to understand your approach, which looks impressive on the first glance, but leave me scratching my head when looking at details. I will comment on 15/26 for the parts I am wondering about. I was trying to add MMC support and hit some issues.

As I said above there is no generic way to handle re-parenting in U-Boot, and we need to traverse based on the SoC Clock architecture.

15/26 handling get_rate, say for example UART request a get_rate.

                              |==> OSC_32K
UART-->APB2--->|==> OSC_24M
                              |==> PLL_PH0   ===> OSC_24M
                              |==> PLL_PH0

Each clock tree, has clock type so UART is marked as MISC since it can't not usual clock tree unlike MMC. MMC can be MP clock type and so..on

So, if driver requesting get_rate with UART clock, then we need to traverse all the UART clock tree and get the rate based on the type and identified parent clock. UART is MISC so it can directly trigger parent clock abp2, apb2 need to find the parent using mux shift, width, once apb2 found the parent it will go to next as so on.

Quoted 4 lines

On the SPI part:
I am bit puzzled about the last 5 patches dealing with SPI. I see that they depend on the others, but would like to deal with them separately.

True, but these are previous version changes ie reason I grouped it here. anyway will handle accordingly in next versions.

Quoted 5 lines

First, patch 25/26 (the sun6i SPI driver) is a bit surprising to find in here. I guess you want to have an easy user for another clock, but I believe this should be a separate series. There were other posts already for a sun6i driver, also I think we should have a unified driver instead of duplicating most of it again. See Oskari's latest post.

Yes, I saw this sun6i driver is lasting for many releases and my plan it to unified the driver w/o any ifdef.

Quoted 10 lines

Secondly: I don't think the SPI clocks are correct. You model them as simple gates only, kind of ignoring the divider settings. This is what the current sun4i driver does as well, but: it writes 1U << 31 into the clock register, which explicitly writes 0 to the other bits, setting 24 MHz as the input rate. Your approach would read-modify-write the register, just setting bit 31 and leaving the other ones as they were before.
So why isn't that using your new tree approach? Looks like a standard mux/divider clock to me.

Yes these yet to implement, ie reason I have just added clk tree only on A64. Idea is to send next version for full support on all SoC types.

[2] https://patchwork.ozlabs.org/patch/955993/

#48 lore raw

v5 cover clk: Add Allwinner CLK, RESET support

Quoted 51 lines · Jagan Teki
Quoted 26 lines · André Przywara

On 31/12/2018 16:59, Jagan Teki wrote:

Hi Jagan,

many thanks for picking this up, I was about to come back to this myself. I am looking at the pinctrl part at the moment, so good you are working on the clocks!

TL;DR: I am good with the first patches, but would like to drop the last five 5 patches from this series, and discuss the whole tree approach (or at least the patch) further.

Quoted 5 lines

Although the previous version[1] is properly handled the clock gates with enable and disable management, but this series is trying to add some more complex Allwinner CLK architecture by handling parent clock and other CLK attributes.

Quoted 7 lines

Allwinner Clock control unit comprises of parent clocks, gates, multiplexers, dividers, multipliers, pre/post dividers and flags etc.

So, the U-Boot implementation of ccu has divided into gates and tree. gates are generic clock configuration of enable/disable bit management and these can be handled via ccu_clock_gate, which is almost same as previous version changes.

Quoted 4 lines

So I like this part very much, all those gates and resets are nicely readable.
Also the removal of the special sunxi USB drivers is very welcome. So up until patch including 21/26, except 15/26, I am all in.

Yes my plan is to do the same.

Quoted 5 lines
Quoted 4 lines

Tree clock has more Allwinner CLK attributes like clock type, fixed clocks, misc clocks, mp, nk, nkm, nkmp, pre/post div, flags etc and these can be managed via ccu_clock_tree.

Quoted 11 lines

Mmmh, this looks quite complex for U-Boot's very limited needs to me. As far as I can see we basically have a static setup for most devices, the ones where we actually change the parent clocks are only clock consumers, namely DE/TCON and MMC. The former chooses a parent once, based on the calculated clock rate. Only for MMC we switch it from OSC24 (for the initial SPI-like setup phase) to PLL_PERIPH0 at runtime. But however this does not really affect the tree, and also we never do re-parenting.
Under this premises, can't we do this somewhat simpler? At the end of the day this is U-Boot, not a sophisticated kernel, and we got away with quite simple clock code so far.

Quoted 6 lines

Yes, Idea is to go with limited clock setups but the thing is as you know Allwinner deals numerous number of clock attributes and I found it difficult to handle them w/o traversing the re-parenting stuff. I have tried simple version of set_rate in previous version[2] but it wouldn't ended as we desired. These operations need re-parenting, so we have to handle parents in clock attributes.

Just for clarification: with re-parenting I meant selecting a different parent clock *at runtime*, potentially adjusting other clocks which depend on that. We would not need that in U-Boot, fortunately.

Quoted 1 line

So finally I came up with tree structure.

Yeah, so looking at patch 15/26 in more detail this is probably the way to go. I was just hit by that wall of code. Implementing get_rate in a recursive fashion reduces the file size drastically, see the reply to that patch.
I was just concerned that making this all data structure based (as opposed to handle them all in a giant switch/case) would require to abstract this too much, up to a point where it gets more complicated. Especially when it comes to special clocks like the MMC beast with its new clock mode.

Quoted 13 lines

My idea here is to support get and set rate with minimal re-parenting like OSC24, PLL_PERIPH0, PLL_PERIPH1 not adding much code like Linux does. Like GATE macro, TREE macro is also a readable structure with all possible/needed clk attributes at one glance. The code that handle these tree attributes may look confusing, but it is obvious because it would need to handle re-parenting.

If SoC has a big CLK architecture like this and to support all or needed clock operations, re-parenting may be one of painful code for SoC driver in U-Boot, Since CLK framework doesn't handle these generically(may be no need) as compared to Linux. and the existing SoC like mediatek clock code does the same.

Quoted 3 lines

At, this point, I think this tree approach can be possible way to move since it handle all clock attribute types and yes we can only support clock re-parenting which are required for U-Boot.

Yeah, so we need to traverse a clock tree to get and set the parent rate. I think with the recursive rework this is now quite neat, especially as we can keep the per-SoC code small.
We just have to see how the special clocks blend in there.

Quoted 20 lines
Quoted 4 lines

Meanwhile I am struggling to understand your approach, which looks impressive on the first glance, but leave me scratching my head when looking at details. I will comment on 15/26 for the parts I am wondering about. I was trying to add MMC support and hit some issues.

As I said above there is no generic way to handle re-parenting in U-Boot, and we need to traverse based on the SoC Clock architecture.

15/26 handling get_rate, say for example UART request a get_rate.

                              |==> OSC_32K
UART-->APB2--->|==> OSC_24M
                              |==> PLL_PH0   ===> OSC_24M
                              |==> PLL_PH0

Each clock tree, has clock type so UART is marked as MISC since it can't not usual clock tree unlike MMC. MMC can be MP clock type and so..on

Quoted 5 lines

So, if driver requesting get_rate with UART clock, then we need to traverse all the UART clock tree and get the rate based on the type and identified parent clock. UART is MISC so it can directly trigger parent clock abp2, apb2 need to find the parent using mux shift, width, once apb2 found the parent it will go to next as so on.

Yeah, I understand this, but was hoping we could handle this in a switch/case, for clarity. Those macro designs look nice, but are hard to understand for outsiders.
But I now think that your data driven approach is better to support multiple SoCs. I can write up some documentation for this.

Quoted 7 lines
Quoted 3 lines

On the SPI part:
I am bit puzzled about the last 5 patches dealing with SPI. I see that they depend on the others, but would like to deal with them separately.

True, but these are previous version changes ie reason I grouped it here. anyway will handle accordingly in next versions.

Quoted 8 lines
Quoted 5 lines

First, patch 25/26 (the sun6i SPI driver) is a bit surprising to find in here. I guess you want to have an easy user for another clock, but I believe this should be a separate series. There were other posts already for a sun6i driver, also I think we should have a unified driver instead of duplicating most of it again. See Oskari's latest post.

Yes, I saw this sun6i driver is lasting for many releases and my plan it to unified the driver w/o any ifdef.

So I did this, basically by introducing a register map for both versions. That works, but is not very nice. Oskari's patch has the advantage of being quite small, and we will probably never need to support both SPI devices in one build.

Quoted 10 lines
Quoted 9 lines

Secondly: I don't think the SPI clocks are correct. You model them as simple gates only, kind of ignoring the divider settings. This is what the current sun4i driver does as well, but: it writes 1U << 31 into the clock register, which explicitly writes 0 to the other bits, setting 24 MHz as the input rate. Your approach would read-modify-write the register, just setting bit 31 and leaving the other ones as they were before.
So why isn't that using your new tree approach? Looks like a standard mux/divider clock to me.

Quoted 3 lines

Yes these yet to implement, ie reason I have just added clk tree only on A64. Idea is to send next version for full support on all SoC types.

Do we need to merge this for all SoCs in the beginning? I think it would be safer to test this on one SoC (A64) first, to avoid breaking everyone in case of problems. Especially if we can't test this on every board.

Cheers,
Andre.

Quoted 2 lines

[2] https://patchwork.ozlabs.org/patch/955993/