Contributors: 8
Author Tokens Token Proportion Commits Commit Proportion
Gregory CLEMENT 3082 89.51% 7 38.89%
Miquel Raynal 258 7.49% 4 22.22%
Pali Rohár 87 2.53% 2 11.11%
Wang Qing 6 0.17% 1 5.56%
Kees Cook 4 0.12% 1 5.56%
Michael Turquette 3 0.09% 1 5.56%
Uwe Kleine-König 2 0.06% 1 5.56%
Vasyl Gomonovych 1 0.03% 1 5.56%
Total 3443 18 


// SPDX-License-Identifier: GPL-2.0+
/*
 * Marvell Armada 37xx SoC Peripheral clocks
 *
 * Copyright (C) 2016 Marvell
 *
 * Gregory CLEMENT <gregory.clement@free-electrons.com>
 *
 * Most of the peripheral clocks can be modelled like this:
 *             _____    _______    _______
 * TBG-A-P  --|     |  |       |  |       |   ______
 * TBG-B-P  --| Mux |--| /div1 |--| /div2 |--| Gate |--> perip_clk
 * TBG-A-S  --|     |  |       |  |       |  |______|
 * TBG-B-S  --|_____|  |_______|  |_______|
 *
 * However some clocks may use only one or two block or and use the
 * xtal clock as parent.
 */

#include <linux/clk-provider.h>
#include <linux/io.h>
#include <linux/mfd/syscon.h>
#include <linux/of.h>
#include <linux/of_device.h>
#include <linux/platform_device.h>
#include <linux/regmap.h>
#include <linux/slab.h>
#include <linux/jiffies.h>

#define TBG_SEL		0x0
#define DIV_SEL0	0x4
#define DIV_SEL1	0x8
#define DIV_SEL2	0xC
#define CLK_SEL		0x10
#define CLK_DIS		0x14

#define  ARMADA_37XX_DVFS_LOAD_1 1
#define LOAD_LEVEL_NR	4

#define ARMADA_37XX_NB_L0L1	0x18
#define ARMADA_37XX_NB_L2L3	0x1C
#define		ARMADA_37XX_NB_TBG_DIV_OFF	13
#define		ARMADA_37XX_NB_TBG_DIV_MASK	0x7
#define		ARMADA_37XX_NB_CLK_SEL_OFF	11
#define		ARMADA_37XX_NB_CLK_SEL_MASK	0x1
#define		ARMADA_37XX_NB_TBG_SEL_OFF	9
#define		ARMADA_37XX_NB_TBG_SEL_MASK	0x3
#define		ARMADA_37XX_NB_CONFIG_SHIFT	16
#define ARMADA_37XX_NB_DYN_MOD	0x24
#define		ARMADA_37XX_NB_DFS_EN	31
#define ARMADA_37XX_NB_CPU_LOAD	0x30
#define		ARMADA_37XX_NB_CPU_LOAD_MASK	0x3
#define		ARMADA_37XX_DVFS_LOAD_0		0
#define		ARMADA_37XX_DVFS_LOAD_1		1
#define		ARMADA_37XX_DVFS_LOAD_2		2
#define		ARMADA_37XX_DVFS_LOAD_3		3

struct clk_periph_driver_data {
	struct clk_hw_onecell_data *hw_data;
	spinlock_t lock;
	void __iomem *reg;

	/* Storage registers for suspend/resume operations */
	u32 tbg_sel;
	u32 div_sel0;
	u32 div_sel1;
	u32 div_sel2;
	u32 clk_sel;
	u32 clk_dis;
};

struct clk_double_div {
	struct clk_hw hw;
	void __iomem *reg1;
	u8 shift1;
	void __iomem *reg2;
	u8 shift2;
};

struct clk_pm_cpu {
	struct clk_hw hw;
	void __iomem *reg_mux;
	u8 shift_mux;
	u32 mask_mux;
	void __iomem *reg_div;
	u8 shift_div;
	struct regmap *nb_pm_base;
	unsigned long l1_expiration;
};

#define to_clk_double_div(_hw) container_of(_hw, struct clk_double_div, hw)
#define to_clk_pm_cpu(_hw) container_of(_hw, struct clk_pm_cpu, hw)

struct clk_periph_data {
	const char *name;
	const char * const *parent_names;
	int num_parents;
	struct clk_hw *mux_hw;
	struct clk_hw *rate_hw;
	struct clk_hw *gate_hw;
	struct clk_hw *muxrate_hw;
	bool is_double_div;
};

static const struct clk_div_table clk_table6[] = {
	{ .val = 1, .div = 1, },
	{ .val = 2, .div = 2, },
	{ .val = 3, .div = 3, },
	{ .val = 4, .div = 4, },
	{ .val = 5, .div = 5, },
	{ .val = 6, .div = 6, },
	{ .val = 0, .div = 0, }, /* last entry */
};

static const struct clk_div_table clk_table1[] = {
	{ .val = 0, .div = 1, },
	{ .val = 1, .div = 2, },
	{ .val = 0, .div = 0, }, /* last entry */
};

static const struct clk_div_table clk_table2[] = {
	{ .val = 0, .div = 2, },
	{ .val = 1, .div = 4, },
	{ .val = 0, .div = 0, }, /* last entry */
};

static const struct clk_ops clk_double_div_ops;
static const struct clk_ops clk_pm_cpu_ops;

#define PERIPH_GATE(_name, _bit)		\
struct clk_gate gate_##_name = {		\
	.reg = (void *)CLK_DIS,			\
	.bit_idx = _bit,			\
	.hw.init = &(struct clk_init_data){	\
		.ops =  &clk_gate_ops,		\
	}					\
};

#define PERIPH_MUX(_name, _shift)		\
struct clk_mux mux_##_name = {			\
	.reg = (void *)TBG_SEL,			\
	.shift = _shift,			\
	.mask = 3,				\
	.hw.init = &(struct clk_init_data){	\
		.ops =  &clk_mux_ro_ops,	\
	}					\
};

#define PERIPH_DOUBLEDIV(_name, _reg1, _reg2, _shift1, _shift2)	\
struct clk_double_div rate_##_name = {		\
	.reg1 = (void *)_reg1,			\
	.reg2 = (void *)_reg2,			\
	.shift1 = _shift1,			\
	.shift2 = _shift2,			\
	.hw.init = &(struct clk_init_data){	\
		.ops =  &clk_double_div_ops,	\
	}					\
};

#define PERIPH_DIV(_name, _reg, _shift, _table)	\
struct clk_divider rate_##_name = {		\
	.reg = (void *)_reg,			\
	.table = _table,			\
	.shift = _shift,			\
	.hw.init = &(struct clk_init_data){	\
		.ops =  &clk_divider_ro_ops,	\
	}					\
};

#define PERIPH_PM_CPU(_name, _shift1, _reg, _shift2)	\
struct clk_pm_cpu muxrate_##_name = {		\
	.reg_mux = (void *)TBG_SEL,		\
	.mask_mux = 3,				\
	.shift_mux = _shift1,			\
	.reg_div = (void *)_reg,		\
	.shift_div = _shift2,			\
	.hw.init = &(struct clk_init_data){	\
		.ops =  &clk_pm_cpu_ops,	\
	}					\
};

#define PERIPH_CLK_FULL_DD(_name, _bit, _shift, _reg1, _reg2, _shift1, _shift2)\
static PERIPH_GATE(_name, _bit);			    \
static PERIPH_MUX(_name, _shift);			    \
static PERIPH_DOUBLEDIV(_name, _reg1, _reg2, _shift1, _shift2);

#define PERIPH_CLK_FULL(_name, _bit, _shift, _reg, _shift1, _table)	\
static PERIPH_GATE(_name, _bit);			    \
static PERIPH_MUX(_name, _shift);			    \
static PERIPH_DIV(_name, _reg, _shift1, _table);

#define PERIPH_CLK_GATE_DIV(_name, _bit,  _reg, _shift, _table)	\
static PERIPH_GATE(_name, _bit);			\
static PERIPH_DIV(_name, _reg, _shift, _table);

#define PERIPH_CLK_MUX_DD(_name, _shift, _reg1, _reg2, _shift1, _shift2)\
static PERIPH_MUX(_name, _shift);			    \
static PERIPH_DOUBLEDIV(_name, _reg1, _reg2, _shift1, _shift2);

#define REF_CLK_FULL(_name)				\
	{ .name = #_name,				\
	  .parent_names = (const char *[]){ "TBG-A-P",	\
	      "TBG-B-P", "TBG-A-S", "TBG-B-S"},		\
	  .num_parents = 4,				\
	  .mux_hw = &mux_##_name.hw,			\
	  .gate_hw = &gate_##_name.hw,			\
	  .rate_hw = &rate_##_name.hw,			\
	}

#define REF_CLK_FULL_DD(_name)				\
	{ .name = #_name,				\
	  .parent_names = (const char *[]){ "TBG-A-P",	\
	      "TBG-B-P", "TBG-A-S", "TBG-B-S"},		\
	  .num_parents = 4,				\
	  .mux_hw = &mux_##_name.hw,			\
	  .gate_hw = &gate_##_name.hw,			\
	  .rate_hw = &rate_##_name.hw,			\
	  .is_double_div = true,			\
	}

#define REF_CLK_GATE(_name, _parent_name)			\
	{ .name = #_name,					\
	  .parent_names = (const char *[]){ _parent_name},	\
	  .num_parents = 1,					\
	  .gate_hw = &gate_##_name.hw,				\
	}

#define REF_CLK_GATE_DIV(_name, _parent_name)			\
	{ .name = #_name,					\
	  .parent_names = (const char *[]){ _parent_name},	\
	  .num_parents = 1,					\
	  .gate_hw = &gate_##_name.hw,				\
	  .rate_hw = &rate_##_name.hw,				\
	}

#define REF_CLK_PM_CPU(_name)				\
	{ .name = #_name,				\
	  .parent_names = (const char *[]){ "TBG-A-P",	\
	      "TBG-B-P", "TBG-A-S", "TBG-B-S"},		\
	  .num_parents = 4,				\
	  .muxrate_hw = &muxrate_##_name.hw,		\
	}

#define REF_CLK_MUX_DD(_name)				\
	{ .name = #_name,				\
	  .parent_names = (const char *[]){ "TBG-A-P",	\
	      "TBG-B-P", "TBG-A-S", "TBG-B-S"},		\
	  .num_parents = 4,				\
	  .mux_hw = &mux_##_name.hw,			\
	  .rate_hw = &rate_##_name.hw,			\
	  .is_double_div = true,			\
	}

/* NB periph clocks */
PERIPH_CLK_FULL_DD(mmc, 2, 0, DIV_SEL2, DIV_SEL2, 16, 13);
PERIPH_CLK_FULL_DD(sata_host, 3, 2, DIV_SEL2, DIV_SEL2, 10, 7);
PERIPH_CLK_FULL_DD(sec_at, 6, 4, DIV_SEL1, DIV_SEL1, 3, 0);
PERIPH_CLK_FULL_DD(sec_dap, 7, 6, DIV_SEL1, DIV_SEL1, 9, 6);
PERIPH_CLK_FULL_DD(tscem, 8, 8, DIV_SEL1, DIV_SEL1, 15, 12);
PERIPH_CLK_FULL(tscem_tmx, 10, 10, DIV_SEL1, 18, clk_table6);
static PERIPH_GATE(avs, 11);
PERIPH_CLK_FULL_DD(pwm, 13, 14, DIV_SEL0, DIV_SEL0, 3, 0);
PERIPH_CLK_FULL_DD(sqf, 12, 12, DIV_SEL1, DIV_SEL1, 27, 24);
static PERIPH_GATE(i2c_2, 16);
static PERIPH_GATE(i2c_1, 17);
PERIPH_CLK_GATE_DIV(ddr_phy, 19, DIV_SEL0, 18, clk_table2);
PERIPH_CLK_FULL_DD(ddr_fclk, 21, 16, DIV_SEL0, DIV_SEL0, 15, 12);
PERIPH_CLK_FULL(trace, 22, 18, DIV_SEL0, 20, clk_table6);
PERIPH_CLK_FULL(counter, 23, 20, DIV_SEL0, 23, clk_table6);
PERIPH_CLK_FULL_DD(eip97, 24, 24, DIV_SEL2, DIV_SEL2, 22, 19);
static PERIPH_PM_CPU(cpu, 22, DIV_SEL0, 28);

static struct clk_periph_data data_nb[] = {
	REF_CLK_FULL_DD(mmc),
	REF_CLK_FULL_DD(sata_host),
	REF_CLK_FULL_DD(sec_at),
	REF_CLK_FULL_DD(sec_dap),
	REF_CLK_FULL_DD(tscem),
	REF_CLK_FULL(tscem_tmx),
	REF_CLK_GATE(avs, "xtal"),
	REF_CLK_FULL_DD(sqf),
	REF_CLK_FULL_DD(pwm),
	REF_CLK_GATE(i2c_2, "xtal"),
	REF_CLK_GATE(i2c_1, "xtal"),
	REF_CLK_GATE_DIV(ddr_phy, "TBG-A-S"),
	REF_CLK_FULL_DD(ddr_fclk),
	REF_CLK_FULL(trace),
	REF_CLK_FULL(counter),
	REF_CLK_FULL_DD(eip97),
	REF_CLK_PM_CPU(cpu),
	{ },
};

/* SB periph clocks */
PERIPH_CLK_MUX_DD(gbe_50, 6, DIV_SEL2, DIV_SEL2, 6, 9);
PERIPH_CLK_MUX_DD(gbe_core, 8, DIV_SEL1, DIV_SEL1, 18, 21);
PERIPH_CLK_MUX_DD(gbe_125, 10, DIV_SEL1, DIV_SEL1, 6, 9);
static PERIPH_GATE(gbe1_50, 0);
static PERIPH_GATE(gbe0_50, 1);
static PERIPH_GATE(gbe1_125, 2);
static PERIPH_GATE(gbe0_125, 3);
PERIPH_CLK_GATE_DIV(gbe1_core, 4, DIV_SEL1, 13, clk_table1);
PERIPH_CLK_GATE_DIV(gbe0_core, 5, DIV_SEL1, 14, clk_table1);
PERIPH_CLK_GATE_DIV(gbe_bm, 12, DIV_SEL1, 0, clk_table1);
PERIPH_CLK_FULL_DD(sdio, 11, 14, DIV_SEL0, DIV_SEL0, 3, 6);
PERIPH_CLK_FULL_DD(usb32_usb2_sys, 16, 16, DIV_SEL0, DIV_SEL0, 9, 12);
PERIPH_CLK_FULL_DD(usb32_ss_sys, 17, 18, DIV_SEL0, DIV_SEL0, 15, 18);
static PERIPH_GATE(pcie, 14);

static struct clk_periph_data data_sb[] = {
	REF_CLK_MUX_DD(gbe_50),
	REF_CLK_MUX_DD(gbe_core),
	REF_CLK_MUX_DD(gbe_125),
	REF_CLK_GATE(gbe1_50, "gbe_50"),
	REF_CLK_GATE(gbe0_50, "gbe_50"),
	REF_CLK_GATE(gbe1_125, "gbe_125"),
	REF_CLK_GATE(gbe0_125, "gbe_125"),
	REF_CLK_GATE_DIV(gbe1_core, "gbe_core"),
	REF_CLK_GATE_DIV(gbe0_core, "gbe_core"),
	REF_CLK_GATE_DIV(gbe_bm, "gbe_core"),
	REF_CLK_FULL_DD(sdio),
	REF_CLK_FULL_DD(usb32_usb2_sys),
	REF_CLK_FULL_DD(usb32_ss_sys),
	REF_CLK_GATE(pcie, "gbe_core"),
	{ },
};

static unsigned int get_div(void __iomem *reg, int shift)
{
	u32 val;

	val = (readl(reg) >> shift) & 0x7;
	if (val > 6)
		return 0;
	return val;
}

static unsigned long clk_double_div_recalc_rate(struct clk_hw *hw,
						unsigned long parent_rate)
{
	struct clk_double_div *double_div = to_clk_double_div(hw);
	unsigned int div;

	div = get_div(double_div->reg1, double_div->shift1);
	div *= get_div(double_div->reg2, double_div->shift2);

	return DIV_ROUND_UP_ULL((u64)parent_rate, div);
}

static const struct clk_ops clk_double_div_ops = {
	.recalc_rate = clk_double_div_recalc_rate,
};

static void armada_3700_pm_dvfs_update_regs(unsigned int load_level,
					    unsigned int *reg,
					    unsigned int *offset)
{
	if (load_level <= ARMADA_37XX_DVFS_LOAD_1)
		*reg = ARMADA_37XX_NB_L0L1;
	else
		*reg = ARMADA_37XX_NB_L2L3;

	if (load_level == ARMADA_37XX_DVFS_LOAD_0 ||
	    load_level ==  ARMADA_37XX_DVFS_LOAD_2)
		*offset += ARMADA_37XX_NB_CONFIG_SHIFT;
}

static bool armada_3700_pm_dvfs_is_enabled(struct regmap *base)
{
	unsigned int val, reg = ARMADA_37XX_NB_DYN_MOD;

	if (IS_ERR(base))
		return false;

	regmap_read(base, reg, &val);

	return !!(val & BIT(ARMADA_37XX_NB_DFS_EN));
}

static unsigned int armada_3700_pm_dvfs_get_cpu_div(struct regmap *base)
{
	unsigned int reg = ARMADA_37XX_NB_CPU_LOAD;
	unsigned int offset = ARMADA_37XX_NB_TBG_DIV_OFF;
	unsigned int load_level, div;

	/*
	 * This function is always called after the function
	 * armada_3700_pm_dvfs_is_enabled, so no need to check again
	 * if the base is valid.
	 */
	regmap_read(base, reg, &load_level);

	/*
	 * The register and the offset inside this register accessed to
	 * read the current divider depend on the load level
	 */
	load_level &= ARMADA_37XX_NB_CPU_LOAD_MASK;
	armada_3700_pm_dvfs_update_regs(load_level, &reg, &offset);

	regmap_read(base, reg, &div);

	return (div >> offset) & ARMADA_37XX_NB_TBG_DIV_MASK;
}

static unsigned int armada_3700_pm_dvfs_get_cpu_parent(struct regmap *base)
{
	unsigned int reg = ARMADA_37XX_NB_CPU_LOAD;
	unsigned int offset = ARMADA_37XX_NB_TBG_SEL_OFF;
	unsigned int load_level, sel;

	/*
	 * This function is always called after the function
	 * armada_3700_pm_dvfs_is_enabled, so no need to check again
	 * if the base is valid
	 */
	regmap_read(base, reg, &load_level);

	/*
	 * The register and the offset inside this register accessed to
	 * read the current divider depend on the load level
	 */
	load_level &= ARMADA_37XX_NB_CPU_LOAD_MASK;
	armada_3700_pm_dvfs_update_regs(load_level, &reg, &offset);

	regmap_read(base, reg, &sel);

	return (sel >> offset) & ARMADA_37XX_NB_TBG_SEL_MASK;
}

static u8 clk_pm_cpu_get_parent(struct clk_hw *hw)
{
	struct clk_pm_cpu *pm_cpu = to_clk_pm_cpu(hw);
	u32 val;

	if (armada_3700_pm_dvfs_is_enabled(pm_cpu->nb_pm_base)) {
		val = armada_3700_pm_dvfs_get_cpu_parent(pm_cpu->nb_pm_base);
	} else {
		val = readl(pm_cpu->reg_mux) >> pm_cpu->shift_mux;
		val &= pm_cpu->mask_mux;
	}

	return val;
}

static unsigned long clk_pm_cpu_recalc_rate(struct clk_hw *hw,
					    unsigned long parent_rate)
{
	struct clk_pm_cpu *pm_cpu = to_clk_pm_cpu(hw);
	unsigned int div;

	if (armada_3700_pm_dvfs_is_enabled(pm_cpu->nb_pm_base))
		div = armada_3700_pm_dvfs_get_cpu_div(pm_cpu->nb_pm_base);
	else
		div = get_div(pm_cpu->reg_div, pm_cpu->shift_div);
	return DIV_ROUND_UP_ULL((u64)parent_rate, div);
}

static long clk_pm_cpu_round_rate(struct clk_hw *hw, unsigned long rate,
				  unsigned long *parent_rate)
{
	struct clk_pm_cpu *pm_cpu = to_clk_pm_cpu(hw);
	struct regmap *base = pm_cpu->nb_pm_base;
	unsigned int div = *parent_rate / rate;
	unsigned int load_level;
	/* only available when DVFS is enabled */
	if (!armada_3700_pm_dvfs_is_enabled(base))
		return -EINVAL;

	for (load_level = 0; load_level < LOAD_LEVEL_NR; load_level++) {
		unsigned int reg, val, offset = ARMADA_37XX_NB_TBG_DIV_OFF;

		armada_3700_pm_dvfs_update_regs(load_level, &reg, &offset);

		regmap_read(base, reg, &val);

		val >>= offset;
		val &= ARMADA_37XX_NB_TBG_DIV_MASK;
		if (val == div)
			/*
			 * We found a load level matching the target
			 * divider, switch to this load level and
			 * return.
			 */
			return *parent_rate / div;
	}

	/* We didn't find any valid divider */
	return -EINVAL;
}

/*
 * Workaround when base CPU frequnecy is 1000 or 1200 MHz
 *
 * Switching the CPU from the L2 or L3 frequencies (250/300 or 200 MHz
 * respectively) to L0 frequency (1/1.2 GHz) requires a significant
 * amount of time to let VDD stabilize to the appropriate
 * voltage. This amount of time is large enough that it cannot be
 * covered by the hardware countdown register. Due to this, the CPU
 * might start operating at L0 before the voltage is stabilized,
 * leading to CPU stalls.
 *
 * To work around this problem, we prevent switching directly from the
 * L2/L3 frequencies to the L0 frequency, and instead switch to the L1
 * frequency in-between. The sequence therefore becomes:
 * 1. First switch from L2/L3 (200/250/300 MHz) to L1 (500/600 MHz)
 * 2. Sleep 20ms for stabling VDD voltage
 * 3. Then switch from L1 (500/600 MHz) to L0 (1000/1200 MHz).
 */
static void clk_pm_cpu_set_rate_wa(struct clk_pm_cpu *pm_cpu,
				   unsigned int new_level, unsigned long rate,
				   struct regmap *base)
{
	unsigned int cur_level;

	regmap_read(base, ARMADA_37XX_NB_CPU_LOAD, &cur_level);
	cur_level &= ARMADA_37XX_NB_CPU_LOAD_MASK;

	if (cur_level == new_level)
		return;

	/*
	 * System wants to go to L1 on its own. If we are going from L2/L3,
	 * remember when 20ms will expire. If from L0, set the value so that
	 * next switch to L0 won't have to wait.
	 */
	if (new_level == ARMADA_37XX_DVFS_LOAD_1) {
		if (cur_level == ARMADA_37XX_DVFS_LOAD_0)
			pm_cpu->l1_expiration = jiffies;
		else
			pm_cpu->l1_expiration = jiffies + msecs_to_jiffies(20);
		return;
	}

	/*
	 * If we are setting to L2/L3, just invalidate L1 expiration time,
	 * sleeping is not needed.
	 */
	if (rate < 1000*1000*1000)
		goto invalidate_l1_exp;

	/*
	 * We are going to L0 with rate >= 1GHz. Check whether we have been at
	 * L1 for long enough time. If not, go to L1 for 20ms.
	 */
	if (pm_cpu->l1_expiration && time_is_before_eq_jiffies(pm_cpu->l1_expiration))
		goto invalidate_l1_exp;

	regmap_update_bits(base, ARMADA_37XX_NB_CPU_LOAD,
			   ARMADA_37XX_NB_CPU_LOAD_MASK,
			   ARMADA_37XX_DVFS_LOAD_1);
	msleep(20);

invalidate_l1_exp:
	pm_cpu->l1_expiration = 0;
}

static int clk_pm_cpu_set_rate(struct clk_hw *hw, unsigned long rate,
			       unsigned long parent_rate)
{
	struct clk_pm_cpu *pm_cpu = to_clk_pm_cpu(hw);
	struct regmap *base = pm_cpu->nb_pm_base;
	unsigned int div = parent_rate / rate;
	unsigned int load_level;

	/* only available when DVFS is enabled */
	if (!armada_3700_pm_dvfs_is_enabled(base))
		return -EINVAL;

	for (load_level = 0; load_level < LOAD_LEVEL_NR; load_level++) {
		unsigned int reg, mask, val,
			offset = ARMADA_37XX_NB_TBG_DIV_OFF;

		armada_3700_pm_dvfs_update_regs(load_level, &reg, &offset);

		regmap_read(base, reg, &val);
		val >>= offset;
		val &= ARMADA_37XX_NB_TBG_DIV_MASK;

		if (val == div) {
			/*
			 * We found a load level matching the target
			 * divider, switch to this load level and
			 * return.
			 */
			reg = ARMADA_37XX_NB_CPU_LOAD;
			mask = ARMADA_37XX_NB_CPU_LOAD_MASK;

			/* Apply workaround when base CPU frequency is 1000 or 1200 MHz */
			if (parent_rate >= 1000*1000*1000)
				clk_pm_cpu_set_rate_wa(pm_cpu, load_level, rate, base);

			regmap_update_bits(base, reg, mask, load_level);

			return rate;
		}
	}

	/* We didn't find any valid divider */
	return -EINVAL;
}

static const struct clk_ops clk_pm_cpu_ops = {
	.get_parent = clk_pm_cpu_get_parent,
	.round_rate = clk_pm_cpu_round_rate,
	.set_rate = clk_pm_cpu_set_rate,
	.recalc_rate = clk_pm_cpu_recalc_rate,
};

static const struct of_device_id armada_3700_periph_clock_of_match[] = {
	{ .compatible = "marvell,armada-3700-periph-clock-nb",
	  .data = data_nb, },
	{ .compatible = "marvell,armada-3700-periph-clock-sb",
	.data = data_sb, },
	{ }
};

static int armada_3700_add_composite_clk(const struct clk_periph_data *data,
					 void __iomem *reg, spinlock_t *lock,
					 struct device *dev, struct clk_hw **hw)
{
	const struct clk_ops *mux_ops = NULL, *gate_ops = NULL,
		*rate_ops = NULL;
	struct clk_hw *mux_hw = NULL, *gate_hw = NULL, *rate_hw = NULL;

	if (data->mux_hw) {
		struct clk_mux *mux;

		mux_hw = data->mux_hw;
		mux = to_clk_mux(mux_hw);
		mux->lock = lock;
		mux_ops = mux_hw->init->ops;
		mux->reg = reg + (u64)mux->reg;
	}

	if (data->gate_hw) {
		struct clk_gate *gate;

		gate_hw = data->gate_hw;
		gate = to_clk_gate(gate_hw);
		gate->lock = lock;
		gate_ops = gate_hw->init->ops;
		gate->reg = reg + (u64)gate->reg;
		gate->flags = CLK_GATE_SET_TO_DISABLE;
	}

	if (data->rate_hw) {
		rate_hw = data->rate_hw;
		rate_ops = rate_hw->init->ops;
		if (data->is_double_div) {
			struct clk_double_div *rate;

			rate =  to_clk_double_div(rate_hw);
			rate->reg1 = reg + (u64)rate->reg1;
			rate->reg2 = reg + (u64)rate->reg2;
		} else {
			struct clk_divider *rate = to_clk_divider(rate_hw);
			const struct clk_div_table *clkt;
			int table_size = 0;

			rate->reg = reg + (u64)rate->reg;
			for (clkt = rate->table; clkt->div; clkt++)
				table_size++;
			rate->width = order_base_2(table_size);
			rate->lock = lock;
		}
	}

	if (data->muxrate_hw) {
		struct clk_pm_cpu *pmcpu_clk;
		struct clk_hw *muxrate_hw = data->muxrate_hw;
		struct regmap *map;

		pmcpu_clk =  to_clk_pm_cpu(muxrate_hw);
		pmcpu_clk->reg_mux = reg + (u64)pmcpu_clk->reg_mux;
		pmcpu_clk->reg_div = reg + (u64)pmcpu_clk->reg_div;

		mux_hw = muxrate_hw;
		rate_hw = muxrate_hw;
		mux_ops = muxrate_hw->init->ops;
		rate_ops = muxrate_hw->init->ops;

		map = syscon_regmap_lookup_by_compatible(
				"marvell,armada-3700-nb-pm");
		pmcpu_clk->nb_pm_base = map;
	}

	*hw = clk_hw_register_composite(dev, data->name, data->parent_names,
					data->num_parents, mux_hw,
					mux_ops, rate_hw, rate_ops,
					gate_hw, gate_ops, CLK_IGNORE_UNUSED);

	return PTR_ERR_OR_ZERO(*hw);
}

static int __maybe_unused armada_3700_periph_clock_suspend(struct device *dev)
{
	struct clk_periph_driver_data *data = dev_get_drvdata(dev);

	data->tbg_sel = readl(data->reg + TBG_SEL);
	data->div_sel0 = readl(data->reg + DIV_SEL0);
	data->div_sel1 = readl(data->reg + DIV_SEL1);
	data->div_sel2 = readl(data->reg + DIV_SEL2);
	data->clk_sel = readl(data->reg + CLK_SEL);
	data->clk_dis = readl(data->reg + CLK_DIS);

	return 0;
}

static int __maybe_unused armada_3700_periph_clock_resume(struct device *dev)
{
	struct clk_periph_driver_data *data = dev_get_drvdata(dev);

	/* Follow the same order than what the Cortex-M3 does (ATF code) */
	writel(data->clk_dis, data->reg + CLK_DIS);
	writel(data->div_sel0, data->reg + DIV_SEL0);
	writel(data->div_sel1, data->reg + DIV_SEL1);
	writel(data->div_sel2, data->reg + DIV_SEL2);
	writel(data->tbg_sel, data->reg + TBG_SEL);
	writel(data->clk_sel, data->reg + CLK_SEL);

	return 0;
}

static const struct dev_pm_ops armada_3700_periph_clock_pm_ops = {
	SET_NOIRQ_SYSTEM_SLEEP_PM_OPS(armada_3700_periph_clock_suspend,
				      armada_3700_periph_clock_resume)
};

static int armada_3700_periph_clock_probe(struct platform_device *pdev)
{
	struct clk_periph_driver_data *driver_data;
	struct device_node *np = pdev->dev.of_node;
	const struct clk_periph_data *data;
	struct device *dev = &pdev->dev;
	int num_periph = 0, i, ret;
	struct resource *res;

	data = of_device_get_match_data(dev);
	if (!data)
		return -ENODEV;

	while (data[num_periph].name)
		num_periph++;

	driver_data = devm_kzalloc(dev, sizeof(*driver_data), GFP_KERNEL);
	if (!driver_data)
		return -ENOMEM;

	driver_data->hw_data = devm_kzalloc(dev,
					    struct_size(driver_data->hw_data,
							hws, num_periph),
					    GFP_KERNEL);
	if (!driver_data->hw_data)
		return -ENOMEM;
	driver_data->hw_data->num = num_periph;

	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
	driver_data->reg = devm_ioremap_resource(dev, res);
	if (IS_ERR(driver_data->reg))
		return PTR_ERR(driver_data->reg);

	spin_lock_init(&driver_data->lock);

	for (i = 0; i < num_periph; i++) {
		struct clk_hw **hw = &driver_data->hw_data->hws[i];
		if (armada_3700_add_composite_clk(&data[i], driver_data->reg,
						  &driver_data->lock, dev, hw))
			dev_err(dev, "Can't register periph clock %s\n",
				data[i].name);
	}

	ret = of_clk_add_hw_provider(np, of_clk_hw_onecell_get,
				     driver_data->hw_data);
	if (ret) {
		for (i = 0; i < num_periph; i++)
			clk_hw_unregister(driver_data->hw_data->hws[i]);
		return ret;
	}

	platform_set_drvdata(pdev, driver_data);
	return 0;
}

static void armada_3700_periph_clock_remove(struct platform_device *pdev)
{
	struct clk_periph_driver_data *data = platform_get_drvdata(pdev);
	struct clk_hw_onecell_data *hw_data = data->hw_data;
	int i;

	of_clk_del_provider(pdev->dev.of_node);

	for (i = 0; i < hw_data->num; i++)
		clk_hw_unregister(hw_data->hws[i]);
}

static struct platform_driver armada_3700_periph_clock_driver = {
	.probe = armada_3700_periph_clock_probe,
	.remove_new = armada_3700_periph_clock_remove,
	.driver		= {
		.name	= "marvell-armada-3700-periph-clock",
		.of_match_table = armada_3700_periph_clock_of_match,
		.pm	= &armada_3700_periph_clock_pm_ops,
	},
};

builtin_platform_driver(armada_3700_periph_clock_driver);