lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <153ef3e0-9978-d201-44ad-3a5e55eeef4f@linaro.org>
Date:   Mon, 27 Feb 2023 04:55:21 +0200
From:   Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
To:     Konrad Dybcio <konrad.dybcio@...aro.org>,
        AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@...labora.com>,
        Andy Gross <agross@...nel.org>,
        Bjorn Andersson <andersson@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Viresh Kumar <vireshk@...nel.org>, Nishanth Menon <nm@...com>,
        Stephen Boyd <sboyd@...nel.org>,
        Niklas Cassel <nks@...wful.org>,
        Liam Girdwood <lgirdwood@...il.com>,
        Mark Brown <broonie@...nel.org>
Cc:     Robert Marko <robimarko@...il.com>, linux-kernel@...r.kernel.org,
        linux-arm-msm@...r.kernel.org, devicetree@...r.kernel.org,
        linux-pm@...r.kernel.org,
        AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@...ainline.org>
Subject: Re: [PATCH v10 5/6] soc: qcom: Add support for Core Power Reduction
 v3, v4 and Hardened

On 17/02/2023 13:08, Konrad Dybcio wrote:
> From: AngeloGioacchino Del Regno <angelogioacchino.delregno@...ainline.org>
> 
> This commit introduces a new driver, based on the one for cpr v1,
> to enable support for the newer Qualcomm Core Power Reduction
> hardware, known downstream as CPR3, CPR4 and CPRh, and support
> for MSM8998 and SDM630 CPU power reduction.
> 
> In these new versions of the hardware, support for various new
> features was introduced, including voltage reduction for the GPU,
> security hardening and a new way of controlling CPU DVFS,
> consisting in internal communication between microcontrollers,
> specifically the CPR-Hardened and the Operating State Manager.
> 
> The CPR v3, v4 and CPRh are present in a broad range of SoCs,
> from the mid-range to the high end ones including, but not limited
> to, MSM8953/8996/8998, SDM630/636/660/845.
> 
> Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@...ainline.org>
> [Konrad: rebase, apply review comments]
> Signed-off-by: Konrad Dybcio <konrad.dybcio@...aro.org>
> ---
>   drivers/soc/qcom/Kconfig      |   22 +
>   drivers/soc/qcom/Makefile     |    4 +-
>   drivers/soc/qcom/cpr-common.h |    2 +
>   drivers/soc/qcom/cpr3.c       | 2923 +++++++++++++++++++++++++++++++++++++++++
>   include/soc/qcom/cpr.h        |   17 +
>   5 files changed, 2967 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/soc/qcom/Kconfig b/drivers/soc/qcom/Kconfig
> index a8f283086a21..1a662c323ddd 100644
> --- a/drivers/soc/qcom/Kconfig
> +++ b/drivers/soc/qcom/Kconfig
> @@ -29,6 +29,7 @@ config QCOM_COMMAND_DB
>   config QCOM_CPR
>   	tristate "QCOM Core Power Reduction (CPR) support"
>   	depends on ARCH_QCOM && HAS_IOMEM
> +	select QCOM_CPR_COMMON
>   	select PM_OPP
>   	select REGMAP
>   	help
> @@ -42,6 +43,27 @@ config QCOM_CPR
>   	  To compile this driver as a module, choose M here: the module will
>   	  be called qcom-cpr
>   
> +config QCOM_CPR_COMMON
> +	tristate
> +
> +config QCOM_CPR3
> +	tristate "QCOM Core Power Reduction (CPR v3/v4/Hardened) support"
> +	depends on ARCH_QCOM && HAS_IOMEM
> +	select QCOM_CPR_COMMON
> +	select PM_OPP
> +	select REGMAP
> +	help
> +	  Say Y here to enable support for the CPR hardware found on a broad
> +	  variety of Qualcomm SoCs like MSM8996, MSM8998, SDM630, SDM660,
> +	  SDM845 and others.
> +
> +	  This driver populates OPP tables and makes adjustments to them
> +	  based on feedback from the CPR hardware. If you want to do CPU
> +	  and/or GPU frequency scaling say Y here.
> +
> +	  To compile this driver as a module, choose M here: the module will
> +	  be called qcom-cpr3
> +
>   config QCOM_GENI_SE
>   	tristate "QCOM GENI Serial Engine Driver"
>   	depends on ARCH_QCOM || COMPILE_TEST
> diff --git a/drivers/soc/qcom/Makefile b/drivers/soc/qcom/Makefile
> index c3a5bf014265..26dc8a3a6cd6 100644
> --- a/drivers/soc/qcom/Makefile
> +++ b/drivers/soc/qcom/Makefile
> @@ -3,7 +3,9 @@ CFLAGS_rpmh-rsc.o := -I$(src)
>   obj-$(CONFIG_QCOM_AOSS_QMP) +=	qcom_aoss.o
>   obj-$(CONFIG_QCOM_GENI_SE) +=	qcom-geni-se.o
>   obj-$(CONFIG_QCOM_COMMAND_DB) += cmd-db.o
> -obj-$(CONFIG_QCOM_CPR)		+= cpr-common.o cpr.o
> +obj-$(CONFIG_QCOM_CPR)		+= cpr.o
> +obj-$(CONFIG_QCOM_CPR_COMMON)	+= cpr-common.o
> +obj-$(CONFIG_QCOM_CPR3)		+= cpr3.o
>   obj-$(CONFIG_QCOM_GSBI)	+=	qcom_gsbi.o
>   obj-$(CONFIG_QCOM_MDT_LOADER)	+= mdt_loader.o
>   obj-$(CONFIG_QCOM_OCMEM)	+= ocmem.o
> diff --git a/drivers/soc/qcom/cpr-common.h b/drivers/soc/qcom/cpr-common.h
> index 2cd15f7eac90..a90f6351d022 100644
> --- a/drivers/soc/qcom/cpr-common.h
> +++ b/drivers/soc/qcom/cpr-common.h
> @@ -65,6 +65,8 @@ struct corner {
>   struct corner_data {
>   	unsigned int fuse_corner;
>   	unsigned long freq;
> +	int oloop_vadj;
> +	int cloop_vadj;
>   };
>   
>   struct acc_desc {
> diff --git a/drivers/soc/qcom/cpr3.c b/drivers/soc/qcom/cpr3.c
> new file mode 100644
> index 000000000000..22fcf311ca01
> --- /dev/null
> +++ b/drivers/soc/qcom/cpr3.c
> @@ -0,0 +1,2923 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (c) 2013-2020, The Linux Foundation. All rights reserved.
> + * Copyright (c) 2019 Linaro Limited
> + * Copyright (c) 2021, AngeloGioacchino Del Regno
> + *                     <angelogioacchino.delregno@...ainline.org>
> + */
> +
> +#include <linux/bitops.h>
> +#include <linux/clk.h>
> +#include <linux/debugfs.h>
> +#include <linux/err.h>
> +#include <linux/init.h>
> +#include <linux/interrupt.h>
> +#include <linux/io.h>
> +#include <linux/kernel.h>
> +#include <linux/mfd/syscon.h>
> +#include <linux/module.h>
> +#include <linux/nvmem-consumer.h>
> +#include <linux/of_device.h>
> +#include <linux/of.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_domain.h>
> +#include <linux/pm_opp.h>
> +#include <linux/regmap.h>
> +#include <linux/regulator/consumer.h>
> +#include <linux/slab.h>
> +#include <linux/string.h>
> +#include <linux/workqueue.h>
> +#include <soc/qcom/cpr.h>
> +
> +#include "cpr-common.h"
> +
> +#define CPR3_RO_COUNT				16
> +#define CPR3_RO_MASK				GENMASK(CPR3_RO_COUNT - 1, 0)
> +
> +/* CPR3 registers */
> +#define CPR3_REG_CPR_VERSION			0x0
> +#define CPRH_CPR_VERSION_4P5			0x40050000
> +
> +#define CPR3_REG_CPR_CTL			0x4
> +#define CPR3_CPR_CTL_LOOP_EN_MASK		BIT(0)
> +#define CPR3_CPR_CTL_IDLE_CLOCKS_MASK		GENMASK(5, 1)
> +#define CPR3_CPR_CTL_IDLE_CLOCKS_SHIFT		1
> +#define CPR3_CPR_CTL_COUNT_MODE_MASK		GENMASK(7, 6)
> +#define CPR3_CPR_CTL_COUNT_MODE_SHIFT		6
> +#define CPR3_CPR_CTL_COUNT_MODE_ALL_AT_ONCE_MIN	0
> +#define CPR3_CPR_CTL_COUNT_MODE_ALL_AT_ONCE_MAX	1
> +#define CPR3_CPR_CTL_COUNT_MODE_STAGGERED	2
> +#define CPR3_CPR_CTL_COUNT_MODE_ALL_AT_ONCE_AGE	3
> +#define CPR3_CPR_CTL_COUNT_REPEAT_MASK		GENMASK(31, 9)
> +#define CPR3_CPR_CTL_COUNT_REPEAT_SHIFT		9
> +
> +#define CPR3_REG_CPR_STATUS			0x8
> +#define CPR3_CPR_STATUS_BUSY_MASK		BIT(0)
> +
> +/*
> + * This register is not present on controllers that support HW closed-loop
> + * except CPR4 APSS controller.
> + */
> +#define CPR3_REG_CPR_TIMER_AUTO_CONT		0xC
> +
> +#define CPR3_REG_CPR_STEP_QUOT			0x14
> +#define CPR3_CPR_STEP_QUOT_MIN_MASK		GENMASK(5, 0)
> +#define CPR3_CPR_STEP_QUOT_MIN_SHIFT		0
> +#define CPR3_CPR_STEP_QUOT_MAX_MASK		GENMASK(11, 6)
> +#define CPR3_CPR_STEP_QUOT_MAX_SHIFT		6
> +#define CPRH_DELTA_QUOT_STEP_FACTOR		4
> +
> +#define CPR3_REG_GCNT(ro)			(0xA0 + 0x4 * (ro))
> +#define CPR3_REG_SENSOR_OWNER(sensor)		(0x200 + 0x4 * (sensor))
> +
> +#define CPR3_REG_CONT_CMD			0x800
> +#define CPR3_CONT_CMD_ACK			0x1
> +#define CPR3_CONT_CMD_NACK			0x0
> +
> +#define CPR3_REG_THRESH(thread)			(0x808 + 0x440 * (thread))
> +#define CPR3_THRESH_CONS_DOWN_MASK		GENMASK(3, 0)
> +#define CPR3_THRESH_CONS_DOWN_SHIFT		0
> +#define CPR3_THRESH_CONS_UP_MASK		GENMASK(7, 4)
> +#define CPR3_THRESH_CONS_UP_SHIFT		4
> +#define CPR3_THRESH_DOWN_THRESH_MASK		GENMASK(12, 8)
> +#define CPR3_THRESH_DOWN_THRESH_SHIFT		8
> +#define CPR3_THRESH_UP_THRESH_MASK		GENMASK(17, 13)
> +#define CPR3_THRESH_UP_THRESH_SHIFT		13
> +
> +#define CPR3_REG_RO_MASK(thread)		(0x80C + 0x440 * (thread))
> +
> +#define CPR3_REG_RESULT0(thread)		(0x810 + 0x440 * (thread))
> +#define CPR3_RESULT0_BUSY_MASK			BIT(0)
> +#define CPR3_RESULT0_STEP_DN_MASK		BIT(1)
> +#define CPR3_RESULT0_STEP_UP_MASK		BIT(2)
> +#define CPR3_RESULT0_ERROR_STEPS_MASK		GENMASK(7, 3)
> +#define CPR3_RESULT0_ERROR_STEPS_SHIFT		3
> +#define CPR3_RESULT0_ERROR_MASK			GENMASK(19, 8)
> +#define CPR3_RESULT0_ERROR_SHIFT		8
> +
> +#define CPR3_REG_RESULT1(thread)		(0x814 + 0x440 * (thread))
> +#define CPR3_RESULT1_QUOT_MIN_MASK		GENMASK(11, 0)
> +#define CPR3_RESULT1_QUOT_MIN_SHIFT		0
> +#define CPR3_RESULT1_QUOT_MAX_MASK		GENMASK(23, 12)
> +#define CPR3_RESULT1_QUOT_MAX_SHIFT		12
> +#define CPR3_RESULT1_RO_MIN_MASK		GENMASK(27, 24)
> +#define CPR3_RESULT1_RO_MIN_SHIFT		24
> +#define CPR3_RESULT1_RO_MAX_MASK		GENMASK(31, 28)
> +#define CPR3_RESULT1_RO_MAX_SHIFT		28
> +
> +#define CPR3_REG_RESULT2(thread)		(0x818 + 0x440 * (thread))
> +#define CPR3_RESULT2_STEP_QUOT_MIN_MASK		GENMASK(5, 0)
> +#define CPR3_RESULT2_STEP_QUOT_MIN_SHIFT	0
> +#define CPR3_RESULT2_STEP_QUOT_MAX_MASK		GENMASK(11, 6)
> +#define CPR3_RESULT2_STEP_QUOT_MAX_SHIFT	6
> +#define CPR3_RESULT2_SENSOR_MIN_MASK		GENMASK(23, 16)
> +#define CPR3_RESULT2_SENSOR_MIN_SHIFT		16
> +#define CPR3_RESULT2_SENSOR_MAX_MASK		GENMASK(31, 24)
> +#define CPR3_RESULT2_SENSOR_MAX_SHIFT		24
> +
> +#define CPR3_REG_IRQ_EN				0x81C
> +#define CPR3_REG_IRQ_CLEAR			0x820
> +#define CPR3_REG_IRQ_STATUS			0x824
> +#define CPR3_IRQ_UP				BIT(3)
> +#define CPR3_IRQ_MID				BIT(2)
> +#define CPR3_IRQ_DOWN				BIT(1)
> +#define CPR3_IRQ_ALL				(CPR3_IRQ_UP | CPR3_IRQ_MID | CPR3_IRQ_DOWN)
> +
> +#define CPR3_REG_TARGET_QUOT(thread, ro)	(0x840 + 0x440 * (thread) + 0x4 * (ro))
> +
> +/* Registers found only on controllers that support HW closed-loop. */
> +#define CPR3_REG_PD_THROTTLE			0xE8
> +
> +#define CPR3_REG_HW_CLOSED_LOOP_DISABLED	0x3000
> +#define CPR3_REG_CPR_TIMER_MID_CONT		0x3004
> +#define CPR3_REG_CPR_TIMER_UP_DN_CONT		0x3008
> +
> +/* CPR4 controller specific registers and bit definitions */
> +#define CPR4_REG_CPR_TIMER_CLAMP			0x10
> +#define CPR4_CPR_TIMER_CLAMP_THREAD_AGGREGATION_EN	BIT(27)
> +
> +#define CPR4_REG_MISC				0x700
> +#define CPR4_MISC_RESET_STEP_QUOT_LOOP_EN	BIT(2)
> +#define CPR4_MISC_THREAD_HAS_ALWAYS_VOTE_EN	BIT(3)
> +
> +#define CPR4_REG_SAW_ERROR_STEP_LIMIT		0x7A4
> +#define CPR4_SAW_ERROR_STEP_LIMIT_UP_MASK	GENMASK(4, 0)
> +#define CPR4_SAW_ERROR_STEP_LIMIT_UP_SHIFT	0
> +#define CPR4_SAW_ERROR_STEP_LIMIT_DN_MASK	GENMASK(9, 5)
> +#define CPR4_SAW_ERROR_STEP_LIMIT_DN_SHIFT	5
> +
> +#define CPR4_REG_MARGIN_TEMP_CORE_TIMERS			0x7A8
> +#define CPR4_MARGIN_TEMP_CORE_TIMERS_SETTLE_VOLTAGE_COUNT_MASK	GENMASK(28, 18)
> +#define CPR4_MARGIN_TEMP_CORE_TIMERS_SETTLE_VOLTAGE_COUNT_SHFT	18
> +
> +#define CPR4_REG_MARGIN_ADJ_CTL				0x7F8
> +#define CPR4_MARGIN_ADJ_HW_CLOSED_LOOP_EN		BIT(4)
> +#define CPR4_MARGIN_ADJ_PER_RO_KV_MARGIN_EN		BIT(7)
> +#define CPR4_MARGIN_ADJ_PMIC_STEP_SIZE_MASK		GENMASK(16, 12)
> +#define CPR4_MARGIN_ADJ_PMIC_STEP_SIZE_SHIFT		12
> +#define CPR4_MARGIN_ADJ_KV_MARGIN_ADJ_STEP_QUOT_MASK	GENMASK(31, 26)
> +#define CPR4_MARGIN_ADJ_KV_MARGIN_ADJ_STEP_QUOT_SHIFT	26
> +
> +#define CPR4_REG_CPR_MASK_THREAD(thread)		(0x80C + 0x440 * (thread))
> +#define CPR4_CPR_MASK_THREAD_DISABLE_THREAD		BIT(31)
> +#define CPR4_CPR_MASK_THREAD_RO_MASK4THREAD_MASK	GENMASK(15, 0)
> +
> +/* CPRh controller specific registers and bit definitions */
> +#define __CPRH_REG_CORNER(rbase, tbase, tid, cnum) (rbase + (tbase * tid) + (0x4 * cnum))
> +#define CPRH_REG_CORNER(d, t, c) __CPRH_REG_CORNER(d->reg_corner, d->reg_corner_tid, t, c)
> +
> +#define CPRH_CTL_OSM_ENABLED			BIT(0)
> +#define CPRH_CTL_BASE_VOLTAGE_MASK		GENMASK(10, 1)
> +#define CPRH_CTL_BASE_VOLTAGE_SHIFT		1
> +#define CPRH_CTL_MODE_SWITCH_DELAY_MASK		GENMASK(24, 17)
> +#define CPRH_CTL_MODE_SWITCH_DELAY_SHIFT	17
> +#define CPRH_CTL_VOLTAGE_MULTIPLIER_MASK	GENMASK(28, 25)
> +#define CPRH_CTL_VOLTAGE_MULTIPLIER_SHIFT	25
> +
> +#define CPRH_CORNER_INIT_VOLTAGE_MASK		GENMASK(7, 0)
> +#define CPRH_CORNER_INIT_VOLTAGE_SHIFT		0
> +#define CPRH_CORNER_FLOOR_VOLTAGE_MASK		GENMASK(15, 8)
> +#define CPRH_CORNER_FLOOR_VOLTAGE_SHIFT		8
> +#define CPRH_CORNER_QUOT_DELTA_MASK		GENMASK(24, 16)
> +#define CPRH_CORNER_QUOT_DELTA_SHIFT		16
> +#define CPRH_CORNER_RO_SEL_MASK			GENMASK(28, 25)
> +#define CPRH_CORNER_RO_SEL_SHIFT		25
> +#define CPRH_CORNER_CPR_CL_DISABLE		BIT(29)
> +
> +#define CPRH_CORNER_INIT_VOLTAGE_MAX_VALUE	255
> +#define CPRH_CORNER_FLOOR_VOLTAGE_MAX_VALUE	255
> +#define CPRH_CORNER_QUOT_DELTA_MAX_VALUE	511
> +
> +enum cpr_type {
> +	CTRL_TYPE_CPR3,
> +	CTRL_TYPE_CPR4,
> +	CTRL_TYPE_CPRH,
> +	CTRL_TYPE_MAX,
> +};
> +
> +/*
> + * struct cpr_thread_desc - CPR Thread-specific parameters
> + *
> + * @controller_id:      Identifier of the CPR controller expected by the HW
> + * @ro_scaling_factor:  Scaling factor for each ring oscillator entry
> + * @hw_tid:             Identifier of the CPR thread expected by the HW
> + * @init_voltage_step:  Voltage in uV for number of steps read from fuse array
> + * @init_voltage_width: Bit-width of the voltage read from the fuse array
> + * @sensor_range_start: First sensor ID used by a thread
> + * @sensor_range_end:   Last sensor ID used by a thread
> + * @num_fuse_corners:   Number of valid entries in fuse_corner_data
> + * @step_quot_init_min: Minimum achievable step quotient for this corner
> + * @step_quot_init_max: Maximum achievable step quotient for this corner
> + * @fuse_corner_data:   Parameters for calculation of each fuse corner
> + */
> +struct cpr_thread_desc {
> +	u8		controller_id;
> +	u8		hw_tid;
> +	const int	(*ro_scaling_factor)[CPR3_RO_COUNT];
> +	int		ro_avail_corners;
> +	int		init_voltage_step;
> +	int		init_voltage_width;
> +	u8		sensor_range_start;
> +	u8		sensor_range_end;
> +	u8		step_quot_init_min;
> +	u8		step_quot_init_max;
> +	unsigned int	num_fuse_corners;
> +	struct fuse_corner_data *fuse_corner_data;
> +};
> +
> +/*
> + * struct cpr_desc - Driver instance-wide CPR parameters
> + *
> + * @cpr_type:              Type (base version) of the CPR controller
> + * @num_threads:           Max. number of threads supported by this controller
> + * @timer_delay_us:        Loop delay time in uS
> + * @timer_updn_delay_us:   Voltage after-up/before-down delay time in uS
> + * @timer_cons_up:         Wait between consecutive up requests in uS
> + * @timer_cons_down:       Wait between consecutive down requests in uS
> + * @up_threshold:          Generic corner up threshold
> + * @down_threshold:        Generic corner down threshold
> + * @idle_clocks:           CPR Sensor: idle timer in cpr clocks unit
> + * @count_mode:            CPR Sensor: counting mode
> + * @count_repeat:          CPR Sensor: number of times to repeat reading
> + * @gcnt_us:               CPR measurement interval in uS
> + * @vreg_step_fixed:       Regulator voltage per step (if vreg unusable)
> + * @vreg_step_up_limit:    Num. of steps up at once before re-measuring sensors
> + * @vreg_step_down_limit:  Num. of steps dn at once before re-measuring sensors
> + * @vdd_settle_time_us:    Settling timer to account for one VDD supply step
> + * @corner_settle_time_us: Settle time for corner switch request
> + * @mem_acc_threshold:     Memory Accelerator (MEM-ACC) voltage threshold
> + * @apm_threshold:         Array Power Mux (APM) voltage threshold
> + * @apm_crossover:         Array Power Mux (APM) corner crossover voltage
> + * @apm_hysteresis:        Hysteresis for APM V-threshold related calculations
> + * @cpr_base_voltage:      Safety: Absolute minimum voltage (uV) on this CPR
> + * @cpr_max_voltage:       Safety: Absolute maximum voltage (uV) on this CPR
> + * @pd_throttle_val:       CPR Power Domain throttle during voltage switch
> + * @threads:               Array containing "CPR Thread" specific parameters
> + * @reduce_to_fuse_uV:     Reduce corner max volts (if higher) to fuse ceiling
> + * @reduce_to_corner_uV:   Reduce corner max volts (if higher) to corner ceil.
> + * @hw_closed_loop_en:     Enable CPR HW Closed-Loop voltage auto-adjustment
> + */
> +struct cpr_desc {
> +	enum cpr_type		cpr_type;
> +	unsigned int		num_threads;
> +	unsigned int		timer_delay_us;
> +	u8			timer_updn_delay_us;
> +	u8			timer_cons_up;
> +	u8			timer_cons_down;
> +	u8			up_threshold;
> +	u8			down_threshold;
> +	u8			idle_clocks;
> +	u8			count_mode;
> +	u8			count_repeat;
> +	u8			gcnt_us;
> +	u16			vreg_step_fixed;
> +	u8			vreg_step_up_limit;
> +	u8			vreg_step_down_limit;
> +	u8			vdd_settle_time_us;
> +	u8			corner_settle_time_us;
> +	int			mem_acc_threshold;
> +	int			apm_threshold;
> +	int			apm_crossover;
> +	int			apm_hysteresis;
> +	u32			cpr_base_voltage;
> +	u32			cpr_max_voltage;
> +	u32			pd_throttle_val;
> +
> +	const struct cpr_thread_desc **threads;
> +	bool reduce_to_fuse_uV;
> +	bool reduce_to_corner_uV;
> +	bool hw_closed_loop_en;
> +};
> +
> +struct cpr_drv;
> +struct cpr_thread {
> +	int			num_corners;
> +	int			id;
> +	bool			enabled;
> +	void __iomem		*base;
> +	struct clk		*cpu_clk;
> +	struct corner		*corner;
> +	struct corner		*corners;
> +	struct fuse_corner	*fuse_corners;
> +	struct cpr_drv		*drv;
> +	struct cpr_ext_data	ext_data;
> +	struct generic_pm_domain pd;
> +	struct device		*attached_cpu_dev;
> +	struct work_struct	restart_work;
> +	bool			restarting;
> +
> +	const struct cpr_fuse	*cpr_fuses;
> +	const struct cpr_thread_desc *desc;
> +};
> +
> +struct cpr_drv {
> +	int			irq;
> +	unsigned int		ref_clk_khz;
> +	struct device		*dev;
> +	struct mutex		lock;
> +	struct regulator	*vreg;
> +	struct regmap		*tcsr;
> +	u32			gcnt;
> +	u32			speed_bin;
> +	u32			fusing_rev;
> +	u32			last_uV;
> +	u32			cpr_hw_rev;
> +	u32			reg_corner;
> +	u32			reg_corner_tid;
> +	u32			reg_ctl;
> +	u32			reg_status;
> +	int			fuse_level_set;
> +	int			extra_corners;
> +	unsigned int		vreg_step;
> +	bool			enabled;
> +
> +	struct cpr_thread	*threads;
> +	struct genpd_onecell_data cell_data;
> +
> +	const struct cpr_desc	*desc;
> +	const struct acc_desc	*acc_desc;
> +	struct dentry		*debugfs;
> +};
> +
> +/**
> + * cpr_get_corner_post_vadj() - Get corner post-voltage adjustment values
> + * @opp:         Pointer to the corresponding OPP struct
> + * @tid:         CPR thread ID
> + * @open_loop:   Pointer to the closed-loop adjustment value
> + * @closed_loop: Pointer to the open-loop adjustment value
> + */
> +void cpr_get_corner_post_vadj(struct dev_pm_opp *opp, u32 tid,
> +			      s32 *open_loop, s32 *closed_loop)
> +{
> +	struct device_node *np;
> +
> +	/*
> +	 * There is no of_property_read_s32_index, so we just store the
> +	 * result into a s32 variable. After all, the OF API is doing
> +	 * the exact same for of_property_read_s32...
> +	 */
> +	np = dev_pm_opp_get_of_node(opp);
> +	if (of_property_read_u32_index(np, "qcom,opp-oloop-vadj", tid, open_loop))
> +		*open_loop = 0;
> +
> +	if (of_property_read_u32_index(np, "qcom,opp-cloop-vadj", tid, closed_loop))
> +		*closed_loop = 0;

These need to be stored per tid and per fusing_rev. I'd suggest to split 
use something like qcom,opp-TID-oloop-vadj = <array or a single-value>;
So, a single property per thread ID, allowing either a single value or 
an array of values, one per fusing revision.

> +
> +	of_node_put(np);
> +}
> +
> +/**
> + * cpr_get_ro_factor() - Get fuse corner ring oscillator factor
> + * @tdesc:  CPR Thread-specific parameters
> + * @fnum:   Fuse corner
> + * @ro_idx: Ring Oscillator fuse number
> + *
> + * Not all threads have different scaling factors for each
> + * Fuse Corner: if the RO factors are the same for all corners,
> + * then only one is specified, instead of uselessly repeating
> + * the same array for FC-times.
> + * This function checks for the same and gives back the right
> + * factor for the requested ring oscillator.
> + *
> + * Return: Ring oscillator factor
> + */
> +static int cpr_get_ro_factor(const struct cpr_thread_desc *tdesc,
> +			     int fnum, int ro_idx)
> +{
> +	int ro_fnum;
> +
> +	if (tdesc->ro_avail_corners == tdesc->num_fuse_corners)
> +		ro_fnum = fnum;
> +	else
> +		ro_fnum = 0;
> +
> +	return tdesc->ro_scaling_factor[ro_fnum][ro_idx];
> +}
> +
> +static void cpr_write(struct cpr_thread *thread, u32 offset, u32 value)
> +{
> +	writel(value, thread->base + offset);
> +}
> +
> +static u32 cpr_read(struct cpr_thread *thread, u32 offset)
> +{
> +	return readl(thread->base + offset);
> +}
> +
> +static void
> +cpr_masked_write(struct cpr_thread *thread, u32 offset, u32 mask, u32 value)
> +{
> +	u32 val;
> +
> +	val = readl(thread->base + offset);
> +	val &= ~mask;
> +	val |= value & mask;
> +	writel(val, thread->base + offset);

This makes me wonder if it would be worth switching to regmap.

> +}
> +
> +static void cpr_irq_clr(struct cpr_thread *thread)
> +{
> +	cpr_write(thread, CPR3_REG_IRQ_CLEAR, CPR3_IRQ_ALL);
> +}
> +
> +static void cpr_irq_clr_nack(struct cpr_thread *thread)
> +{
> +	cpr_irq_clr(thread);
> +	cpr_write(thread, CPR3_REG_CONT_CMD, CPR3_CONT_CMD_NACK);
> +}
> +
> +static void cpr_irq_clr_ack(struct cpr_thread *thread)
> +{
> +	cpr_irq_clr(thread);
> +	cpr_write(thread, CPR3_REG_CONT_CMD, CPR3_CONT_CMD_ACK);
> +}
> +
> +static void cpr_irq_set(struct cpr_thread *thread, u32 int_bits)
> +{
> +	/* On CPR-hardened, interrupts are managed by and on firmware */
> +	if (thread->drv->desc->cpr_type == CTRL_TYPE_CPRH)
> +		return;
> +
> +	cpr_write(thread, CPR3_REG_IRQ_EN, int_bits);
> +}
> +
> +/**
> + * cpr_ctl_enable() - Enable CPR thread
> + * @thread:     Structure holding CPR thread-specific parameters
> + */
> +static void cpr_ctl_enable(struct cpr_thread *thread)
> +{
> +	if (thread->drv->enabled && !thread->restarting) {
> +		cpr_masked_write(thread, CPR3_REG_CPR_CTL,
> +				 CPR3_CPR_CTL_LOOP_EN_MASK,
> +				 CPR3_CPR_CTL_LOOP_EN_MASK);
> +	}
> +}
> +
> +/**
> + * cpr_ctl_disable() - Disable CPR thread
> + * @thread:     Structure holding CPR thread-specific parameters
> + */
> +static void cpr_ctl_disable(struct cpr_thread *thread)
> +{
> +	const struct cpr_desc *desc = thread->drv->desc;
> +
> +	if (desc->cpr_type != CTRL_TYPE_CPRH) {
> +		cpr_irq_set(thread, 0);
> +		cpr_irq_clr(thread);
> +	}
> +
> +	cpr_masked_write(thread, CPR3_REG_CPR_CTL,
> +			 CPR3_CPR_CTL_LOOP_EN_MASK, 0);
> +}
> +
> +/**
> + * cpr_ctl_is_enabled() - Check if thread is enabled
> + * @thread:     Structure holding CPR thread-specific parameters
> + *
> + * Return: true if the CPR is enabled, false if it is disabled.
> + */
> +static bool cpr_ctl_is_enabled(struct cpr_thread *thread)
> +{
> +	u32 reg_val;
> +
> +	reg_val = cpr_read(thread, CPR3_REG_CPR_CTL);
> +	return reg_val & CPR3_CPR_CTL_LOOP_EN_MASK;
> +}
> +
> +/**
> + * cpr_check_any_thread_busy() - Check if HW is done processing
> + * @thread:     Structure holding CPR thread-specific parameters
> + *
> + * Return: true if the CPR is busy, false if it is ready.
> + */
> +static bool cpr_check_any_thread_busy(struct cpr_thread *thread)
> +{
> +	int i;
> +
> +	for (i = 0; i < thread->drv->desc->num_threads; i++)
> +		if (cpr_read(thread, CPR3_REG_RESULT0(i)) &
> +		    CPR3_RESULT0_BUSY_MASK)
> +			return true;
> +
> +	return false;
> +}
> +
> +static void cpr_restart_worker(struct work_struct *work)
> +{
> +	struct cpr_thread *thread = container_of(work, struct cpr_thread,
> +						 restart_work);
> +	struct cpr_drv *drv = thread->drv;
> +	int i;
> +
> +	mutex_lock(&drv->lock);
> +
> +	thread->restarting = true;
> +	cpr_ctl_disable(thread);
> +	disable_irq(drv->irq);
> +
> +	mutex_unlock(&drv->lock);
> +
> +	for (i = 0; i < 20; i++) {
> +		u32 cpr_status = cpr_read(thread, CPR3_REG_CPR_STATUS);
> +		u32 ctl = cpr_read(thread, CPR3_REG_CPR_CTL);
> +
> +		if ((cpr_status & CPR3_CPR_STATUS_BUSY_MASK) &&
> +		   !(ctl & CPR3_CPR_CTL_LOOP_EN_MASK))
> +			break;
> +
> +		udelay(10);
> +	}
> +
> +	cpr_irq_clr(thread);
> +
> +	for (i = 0; i < 20; i++) {
> +		u32 status = cpr_read(thread, CPR3_REG_IRQ_STATUS);
> +
> +		if (!(status & CPR3_IRQ_ALL))
> +			break;
> +		udelay(10);
> +	}
> +
> +	mutex_lock(&drv->lock);
> +
> +	thread->restarting = false;
> +	enable_irq(drv->irq);
> +	cpr_ctl_enable(thread);
> +
> +	mutex_unlock(&drv->lock);
> +}
> +
> +/**
> + * cpr_corner_restore() - Restore saved corner level
> + * @thread: Structure holding CPR thread-specific parameters
> + * @corner: Structure holding the saved corner level
> + */
> +static void cpr_corner_restore(struct cpr_thread *thread,
> +			       struct corner *corner)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	struct fuse_corner *fuse = corner->fuse_corner;
> +	const struct cpr_thread_desc *tdesc = thread->desc;
> +	u32 ro_sel = fuse->ring_osc_idx;
> +
> +	cpr_write(thread, CPR3_REG_GCNT(ro_sel), drv->gcnt);
> +
> +	cpr_write(thread, CPR3_REG_RO_MASK(tdesc->hw_tid),
> +		  CPR3_RO_MASK & ~BIT(ro_sel));
> +
> +	cpr_write(thread, CPR3_REG_TARGET_QUOT(tdesc->hw_tid, ro_sel),
> +		  fuse->quot - corner->quot_adjust);
> +
> +	if (drv->desc->cpr_type == CTRL_TYPE_CPR4)
> +		cpr_masked_write(thread,
> +				 CPR4_REG_CPR_MASK_THREAD(tdesc->hw_tid),
> +				 CPR4_CPR_MASK_THREAD_DISABLE_THREAD |
> +				 CPR4_CPR_MASK_THREAD_RO_MASK4THREAD_MASK, 0);
> +
> +	thread->corner = corner;
> +	corner->last_uV = corner->uV;
> +}
> +
> +/**
> + * cpr_set_acc() - Set fuse level to the mem-acc
> + * @drv: Main driver structure
> + * @f:   Fuse level
> + */
> +static void cpr_set_acc(struct cpr_drv *drv, int f)
> +{
> +	const struct acc_desc *desc = drv->acc_desc;
> +	struct reg_sequence *s = desc->settings;
> +	int n = desc->num_regs_per_fuse;
> +
> +	if (!s || f == drv->fuse_level_set)
> +		return;
> +
> +	regmap_multi_reg_write(drv->tcsr, s + (n * f), n);
> +	drv->fuse_level_set = f;
> +}
> +
> +/**
> + * cpr_pre_voltage() - Actions to execute before setting voltage
> + * @thread:     Structure holding CPR thread-specific parameters
> + * @dir:        Enumeration for voltage change direction
> + * @fuse_level: Fuse corner for mem-acc, if supported.
> + *
> + * Return: Zero for success or negative number on errors.
> + */
> +static int cpr_pre_voltage(struct cpr_thread *thread,
> +			   enum voltage_change_dir dir,
> +			   int fuse_level)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +
> +	if (drv->desc->cpr_type == CTRL_TYPE_CPR3 &&
> +	    drv->desc->pd_throttle_val)
> +		cpr_write(thread, CPR3_REG_PD_THROTTLE,
> +			  drv->desc->pd_throttle_val);
> +
> +	if (drv->tcsr && dir == DOWN)
> +		cpr_set_acc(drv, fuse_level);
> +
> +	return 0;
> +}
> +
> +/**
> + * cpr_post_voltage() - Actions to execute after setting voltage
> + * @thread:     Structure holding CPR thread-specific parameters
> + * @dir:        Enumeration for voltage change direction
> + * @fuse_level: Fuse corner for mem-acc, if supported.
> + *
> + * Return: Zero for success or negative number on errors.
> + */
> +static int cpr_post_voltage(struct cpr_thread *thread,
> +			    enum voltage_change_dir dir,
> +			    int fuse_level)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +
> +	if (drv->tcsr && dir == UP)
> +		cpr_set_acc(drv, fuse_level);
> +
> +	if (drv->desc->cpr_type == CTRL_TYPE_CPR3)
> +		cpr_write(thread, CPR3_REG_PD_THROTTLE, 0);
> +
> +	return 0;
> +}
> +
> +/**
> + * cpr_commit_state() - Set the newly requested voltage
> + * @thread:     Structure holding CPR thread-specific parameters
> + *
> + * Return: IRQ_SUCCESS for success, IRQ_NONE if the CPR is disabled.
> + */
> +static int cpr_commit_state(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	int min_uV = 0, max_uV = 0, new_uV = 0, fuse_level = 0;
> +	enum voltage_change_dir dir;
> +	u32 next_irqmask = 0;
> +	int ret, i;
> +
> +	/* On CPRhardened, control states are managed in firmware */
> +	if (drv->desc->cpr_type == CTRL_TYPE_CPRH)
> +		return 0;
> +
> +	for (i = 0; i < drv->desc->num_threads; i++) {
> +		struct cpr_thread *thread = &drv->threads[i];
> +
> +		if (!thread->corner)
> +			continue;
> +
> +		fuse_level = max(fuse_level,
> +				 (int) (thread->corner->fuse_corner -
> +				 &thread->fuse_corners[0]));
> +
> +		max_uV = max(max_uV, thread->corner->max_uV);
> +		min_uV = max(min_uV, thread->corner->min_uV);
> +		new_uV = max(new_uV, thread->corner->last_uV);
> +	}
> +	dev_vdbg(drv->dev, "%s: new uV: %d, last uV: %d\n",
> +		 __func__, new_uV, drv->last_uV);
> +
> +	/*
> +	 * Safety measure: if the voltage is out of the globally allowed
> +	 * range, then go out and warn the user.
> +	 * This should *never* happen.
> +	 */
> +	if (new_uV > drv->desc->cpr_max_voltage ||
> +	    new_uV < drv->desc->cpr_base_voltage) {
> +		dev_warn(drv->dev, "Voltage (%u uV) out of range.", new_uV);
> +		return -EINVAL;
> +	}
> +
> +	if (new_uV == drv->last_uV || fuse_level == drv->fuse_level_set)
> +		goto out;
> +
> +	if (fuse_level > drv->fuse_level_set)
> +		dir = UP;
> +	else
> +		dir = DOWN;
> +
> +	ret = cpr_pre_voltage(thread, fuse_level, dir);
> +	if (ret)
> +		return ret;
> +
> +	dev_vdbg(drv->dev, "setting voltage: %d\n", new_uV);
> +
> +	ret = regulator_set_voltage(drv->vreg, new_uV, new_uV);

Shouldn't this be min/max instead of a single new_uV?

> +	if (ret) {
> +		dev_err_ratelimited(drv->dev, "failed to set voltage %d: %d\n", new_uV, ret);
> +		return ret;
> +	}
> +
> +	ret = cpr_post_voltage(thread, fuse_level, dir);
> +	if (ret)
> +		return ret;
> +
> +	drv->last_uV = new_uV;
> +out:
> +	if (new_uV > min_uV)
> +		next_irqmask |= CPR3_IRQ_DOWN;
> +	if (new_uV < max_uV)
> +		next_irqmask |= CPR3_IRQ_UP;
> +
> +	cpr_irq_set(thread, next_irqmask);
> +
> +	return 0;
> +}
> +
> +static unsigned int cpr_get_cur_perf_state(struct cpr_thread *thread)
> +{
> +	return thread->corner ? thread->corner - thread->corners + 1 : 0;
> +}
> +
> +/**
> + * cpr_scale() - Calculate new voltage for the received direction
> + * @thread: Structure holding CPR thread-specific parameters
> + * @dir:    Enumeration for voltage change direction
> + *
> + * The CPR scales one by one: this function calculates the new
> + * voltage to set when a voltage-UP or voltage-DOWN request comes
> + * and stores it into the per-thread structure that gets passed.
> + */
> +static void cpr_scale(struct cpr_thread *thread, enum voltage_change_dir dir)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	const struct cpr_thread_desc *tdesc = thread->desc;
> +	u32 val, error_steps;
> +	int last_uV, new_uV;
> +	struct corner *corner;
> +
> +	if (dir != UP && dir != DOWN)
> +		return;
> +
> +	corner = thread->corner;
> +	val = cpr_read(thread, CPR3_REG_RESULT0(tdesc->hw_tid));
> +	error_steps = val >> CPR3_RESULT0_ERROR_STEPS_SHIFT;
> +	error_steps &= CPR3_RESULT0_ERROR_STEPS_MASK;
> +
> +	last_uV = corner->last_uV;
> +
> +	if (dir == UP) {
> +		if (!(val & CPR3_RESULT0_STEP_UP_MASK))
> +			return;
> +
> +		/* Calculate new voltage */
> +		new_uV = last_uV + drv->vreg_step;
> +		new_uV = min(new_uV, corner->max_uV);
> +
> +		dev_vdbg(drv->dev, "[T%u] UP - new_uV=%d last_uV=%d p-state=%u st=%u\n",
> +			 thread->id, new_uV, last_uV,
> +			 cpr_get_cur_perf_state(thread), error_steps);
> +	} else {
> +		if (!(val & CPR3_RESULT0_STEP_DN_MASK))
> +			return;
> +
> +		/* Calculate new voltage */
> +		new_uV = last_uV - drv->vreg_step;
> +		new_uV = max(new_uV, corner->min_uV);
> +		dev_vdbg(drv->dev, "[T%u] DOWN - new_uV=%d last_uV=%d p-state=%u st=%u\n",
> +			 thread->id, new_uV, last_uV,
> +			 cpr_get_cur_perf_state(thread), error_steps);
> +	}
> +	corner->last_uV = new_uV;
> +}
> +
> +/**
> + * cpr_irq_handler() - Handle CPR3/CPR4 status interrupts
> + * @irq: Number of the interrupt
> + * @dev: Pointer to the cpr_thread structure
> + *
> + * Handle the interrupts coming from non-hardened CPR HW as to get
> + * an ok to scale voltages immediately, or to pass error status to
> + * the hardware (either success/ACK or failure/NACK).
> + *
> + * Return: IRQ_SUCCESS for success, IRQ_NONE if the CPR is disabled.
> + */
> +static irqreturn_t cpr_irq_handler(int irq, void *dev)
> +{
> +	struct cpr_thread *thread = dev;
> +	struct cpr_drv *drv = thread->drv;
> +	irqreturn_t ret = IRQ_HANDLED;
> +	int i, rc;
> +	enum voltage_change_dir dir = NO_CHANGE;
> +	u32 val;
> +
> +	mutex_lock(&drv->lock);
> +
> +	val = cpr_read(thread, CPR3_REG_IRQ_STATUS);
> +
> +	dev_vdbg(drv->dev, "IRQ_STATUS = %#02x\n", val);
> +
> +	if (!cpr_ctl_is_enabled(thread)) {
> +		dev_vdbg(drv->dev, "CPR is disabled\n");
> +		ret = IRQ_NONE;
> +	} else if (cpr_check_any_thread_busy(thread)) {
> +		cpr_irq_clr_nack(thread);
> +		dev_dbg(drv->dev, "CPR measurement is not ready\n");
> +	} else {
> +		/*
> +		 * Following sequence of handling is as per each IRQ's
> +		 * priority
> +		 */
> +		if (val & CPR3_IRQ_UP)
> +			dir = UP;
> +		else if (val & CPR3_IRQ_DOWN)
> +			dir = DOWN;
> +
> +		if (dir != NO_CHANGE) {
> +			for (i = 0; i < drv->desc->num_threads; i++) {
> +				thread = &drv->threads[i];
> +				cpr_scale(thread, dir);
> +			}
> +
> +			rc = cpr_commit_state(thread);
> +			if (rc)
> +				cpr_irq_clr_nack(thread);
> +			else
> +				cpr_irq_clr_ack(thread);
> +		} else if (val & CPR3_IRQ_MID) {
> +			dev_dbg(drv->dev, "IRQ occurred for Mid Flag\n");
> +		} else {
> +			dev_warn(drv->dev, "IRQ occurred for unknown flag (%#08x)\n", val);
> +			schedule_work(&thread->restart_work);
> +		}
> +	}
> +
> +	mutex_unlock(&drv->lock);
> +
> +	return ret;
> +}
> +
> +static int cpr_switch(struct cpr_drv *drv)
> +{
> +	int i, ret;
> +	bool enabled = false;
> +
> +	if (drv->desc->cpr_type == CTRL_TYPE_CPRH)
> +		return 0;
> +
> +	for (i = 0; i < drv->desc->num_threads && !enabled; i++)
> +		enabled = drv->threads[i].enabled;
> +
> +	dev_vdbg(drv->dev, "%s: enabled = %d\n", __func__, enabled);
> +
> +	if (enabled == drv->enabled)
> +		return 0;
> +
> +	if (enabled) {
> +		ret = regulator_enable(drv->vreg);
> +		if (ret)
> +			return ret;
> +
> +		drv->enabled = enabled;
> +
> +		for (i = 0; i < drv->desc->num_threads; i++)
> +			if (drv->threads[i].corner)
> +				break;
> +
> +		if (i < drv->desc->num_threads) {
> +			cpr_irq_clr(&drv->threads[i]);
> +
> +			cpr_commit_state(&drv->threads[i]);
> +			cpr_ctl_enable(&drv->threads[i]);
> +		}
> +	} else {
> +		for (i = 0; i < drv->desc->num_threads && !enabled; i++)
> +			cpr_ctl_disable(&drv->threads[i]);
> +
> +		drv->enabled = enabled;
> +
> +		ret = regulator_disable(drv->vreg);
> +		if (ret < 0)
> +			return ret;
> +	}
> +
> +	return 0;
> +}
> +
> +/**
> + * cpr_enable() - Enables a CPR thread
> + * @thread: Structure holding CPR thread-specific parameters
> + *
> + * Return: Zero for success or negative number on errors.
> + */
> +static int cpr_enable(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	int ret;
> +
> +	dev_dbg(drv->dev, "Enabling thread %d\n", thread->id);
> +
> +	mutex_lock(&drv->lock);
> +
> +	thread->enabled = true;
> +	ret = cpr_switch(thread->drv);
> +
> +	mutex_unlock(&drv->lock);
> +
> +	return ret;
> +}
> +
> +/**
> + * cpr_disable() - Disables a CPR thread
> + * @thread: Structure holding CPR thread-specific parameters
> + *
> + * Return: Zero for success or negative number on errors.
> + */
> +static int cpr_disable(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	int ret;
> +
> +	dev_dbg(drv->dev, "Disabling thread %d\n", thread->id);
> +
> +	mutex_lock(&drv->lock);
> +
> +	thread->enabled = false;
> +	ret = cpr_switch(thread->drv);
> +
> +	mutex_unlock(&drv->lock);
> +
> +	return ret;
> +}
> +
> +/**
> + * cpr_configure() - Configure main HW parameters
> + * @thread: Structure holding CPR thread-specific parameters
> + *
> + * This function configures the main CPR hardware parameters, such as
> + * internal timers (and delays), sensor ownerships, activates and/or
> + * deactivates cpr-threads and others, as one sequence for all of the
> + * versions supported in this driver. By design, the function may
> + * return a success earlier if the sequence for "a previous version"
> + * has ended.
> + *
> + * Context: The CPR must be clocked before calling this function!
> + *
> + * Return: Zero for success or negative number on errors.
> + */
> +static int cpr_configure(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	const struct cpr_desc *desc = drv->desc;
> +	const struct cpr_thread_desc *tdesc = thread->desc;
> +	u32 val;
> +	int i;
> +
> +	/* Disable interrupt and CPR */
> +	cpr_irq_set(thread, 0);
> +	cpr_write(thread, CPR3_REG_CPR_CTL, 0);
> +
> +	/* Init and save gcnt */
> +	drv->gcnt = drv->ref_clk_khz * desc->gcnt_us;
> +	do_div(drv->gcnt, 1000);
> +
> +	/* Program the delay count for the timer */
> +	val = drv->ref_clk_khz * desc->timer_delay_us;
> +	do_div(val, 1000);
> +	if (desc->cpr_type == CTRL_TYPE_CPR3) {
> +		cpr_write(thread, CPR3_REG_CPR_TIMER_MID_CONT, val);
> +
> +		val = drv->ref_clk_khz * desc->timer_updn_delay_us;
> +		do_div(val, 1000);
> +		cpr_write(thread, CPR3_REG_CPR_TIMER_UP_DN_CONT, val);
> +	} else {
> +		cpr_write(thread, CPR3_REG_CPR_TIMER_AUTO_CONT, val);
> +	}
> +	dev_dbg(drv->dev, "Timer count: %#0x (for %d us)\n", val,
> +		desc->timer_delay_us);
> +
> +	/* Program the control register */
> +	val = desc->idle_clocks << CPR3_CPR_CTL_IDLE_CLOCKS_SHIFT;
> +	val |= desc->count_mode << CPR3_CPR_CTL_COUNT_MODE_SHIFT;
> +	val |= desc->count_repeat << CPR3_CPR_CTL_COUNT_REPEAT_SHIFT;
> +	cpr_write(thread, CPR3_REG_CPR_CTL, val);
> +
> +	/* Configure CPR default step quotients */
> +	val = tdesc->step_quot_init_min << CPR3_CPR_STEP_QUOT_MIN_SHIFT;
> +	val |= tdesc->step_quot_init_max << CPR3_CPR_STEP_QUOT_MAX_SHIFT;
> +
> +	cpr_write(thread, CPR3_REG_CPR_STEP_QUOT, val);
> +
> +	/*
> +	 * Configure the CPR sensor ownership always on thread 0
> +	 * TODO: SDM845 has different ownership for sensors!!
> +	 */
> +	for (i = tdesc->sensor_range_start; i < tdesc->sensor_range_end; i++)
> +		cpr_write(thread, CPR3_REG_SENSOR_OWNER(i), 0);
> +
> +	/* Program Consecutive Up & Down */
> +	val = desc->timer_cons_up << CPR3_THRESH_CONS_UP_SHIFT;
> +	val |= desc->timer_cons_down << CPR3_THRESH_CONS_DOWN_SHIFT;
> +	val |= desc->up_threshold << CPR3_THRESH_UP_THRESH_SHIFT;
> +	val |= desc->down_threshold << CPR3_THRESH_DOWN_THRESH_SHIFT;
> +	cpr_write(thread, CPR3_REG_THRESH(tdesc->hw_tid), val);
> +
> +	/* Mask all ring oscillators for all threads initially */
> +	cpr_write(thread, CPR3_REG_RO_MASK(tdesc->hw_tid), CPR3_RO_MASK);
> +
> +	/* HW Closed-loop control */
> +	if (desc->cpr_type == CTRL_TYPE_CPR3) {
> +		cpr_write(thread, CPR3_REG_HW_CLOSED_LOOP_DISABLED,
> +			  !desc->hw_closed_loop_en);
> +	} else {
> +		cpr_masked_write(thread, CPR4_REG_MARGIN_ADJ_CTL,
> +				 CPR4_MARGIN_ADJ_HW_CLOSED_LOOP_EN,
> +				 desc->hw_closed_loop_en ?
> +				 CPR4_MARGIN_ADJ_HW_CLOSED_LOOP_EN : 0);
> +	}
> +
> +	/* Additional configuration for CPR4 and beyond */
> +	if (desc->cpr_type < CTRL_TYPE_CPR4)
> +		return 0;
> +
> +	/* Disable threads initially only on non-hardened CPR4 */
> +	if (desc->cpr_type == CTRL_TYPE_CPR4)
> +		cpr_masked_write(thread, CPR4_REG_CPR_MASK_THREAD(1),
> +				 CPR4_CPR_MASK_THREAD_DISABLE_THREAD |
> +				 CPR4_CPR_MASK_THREAD_RO_MASK4THREAD_MASK,
> +				 CPR4_CPR_MASK_THREAD_DISABLE_THREAD |
> +				 CPR4_CPR_MASK_THREAD_RO_MASK4THREAD_MASK);
> +
> +	if (tdesc->hw_tid > 0)
> +		cpr_masked_write(thread, CPR4_REG_MISC,
> +				 CPR4_MISC_RESET_STEP_QUOT_LOOP_EN |
> +				 CPR4_MISC_THREAD_HAS_ALWAYS_VOTE_EN,
> +				 CPR4_MISC_RESET_STEP_QUOT_LOOP_EN |
> +				 CPR4_MISC_THREAD_HAS_ALWAYS_VOTE_EN);
> +
> +	val = drv->vreg_step;
> +	do_div(val, 1000);
> +	cpr_masked_write(thread, CPR4_REG_MARGIN_ADJ_CTL,
> +			 CPR4_MARGIN_ADJ_PMIC_STEP_SIZE_MASK,
> +			 val << CPR4_MARGIN_ADJ_PMIC_STEP_SIZE_SHIFT);
> +
> +	cpr_masked_write(thread, CPR4_REG_SAW_ERROR_STEP_LIMIT,
> +			 CPR4_SAW_ERROR_STEP_LIMIT_DN_MASK,
> +			 desc->vreg_step_down_limit <<
> +			 CPR4_SAW_ERROR_STEP_LIMIT_DN_SHIFT);
> +
> +	cpr_masked_write(thread, CPR4_REG_SAW_ERROR_STEP_LIMIT,
> +			 CPR4_SAW_ERROR_STEP_LIMIT_UP_MASK,
> +			 desc->vreg_step_up_limit <<
> +			 CPR4_SAW_ERROR_STEP_LIMIT_UP_SHIFT);
> +
> +	cpr_masked_write(thread, CPR4_REG_MARGIN_ADJ_CTL,
> +			 CPR4_MARGIN_ADJ_PER_RO_KV_MARGIN_EN,
> +			 CPR4_MARGIN_ADJ_PER_RO_KV_MARGIN_EN);
> +
> +	if (tdesc->hw_tid > 0)
> +		cpr_masked_write(thread, CPR4_REG_CPR_TIMER_CLAMP,
> +				 CPR4_CPR_TIMER_CLAMP_THREAD_AGGREGATION_EN,
> +				 CPR4_CPR_TIMER_CLAMP_THREAD_AGGREGATION_EN);
> +
> +	/* Settling timer to account for one VDD supply step */
> +	if (desc->vdd_settle_time_us > 0) {
> +		u32 m = CPR4_MARGIN_TEMP_CORE_TIMERS_SETTLE_VOLTAGE_COUNT_MASK;
> +		u32 s = CPR4_MARGIN_TEMP_CORE_TIMERS_SETTLE_VOLTAGE_COUNT_SHFT;
> +
> +		cpr_masked_write(thread, CPR4_REG_MARGIN_TEMP_CORE_TIMERS,
> +				 m, desc->vdd_settle_time_us << s);
> +	}
> +
> +	/* Additional configuration for CPR-hardened */
> +	if (desc->cpr_type < CTRL_TYPE_CPRH)
> +		return 0;
> +
> +	/* Settling timer to account for one corner-switch request */
> +	if (desc->corner_settle_time_us > 0)
> +		cpr_masked_write(thread, drv->reg_ctl,
> +				 CPRH_CTL_MODE_SWITCH_DELAY_MASK,
> +				 desc->corner_settle_time_us <<
> +				 CPRH_CTL_MODE_SWITCH_DELAY_SHIFT);
> +
> +	/* Base voltage and multiplier values for CPRh internal calculations */
> +	cpr_masked_write(thread, drv->reg_ctl,
> +			 CPRH_CTL_BASE_VOLTAGE_MASK,
> +			 (DIV_ROUND_UP(desc->cpr_base_voltage,
> +				       drv->vreg_step) <<
> +			  CPRH_CTL_BASE_VOLTAGE_SHIFT));
> +
> +	cpr_masked_write(thread, drv->reg_ctl,
> +			 CPRH_CTL_VOLTAGE_MULTIPLIER_MASK,
> +			 DIV_ROUND_UP(drv->vreg_step, 1000) <<
> +			 CPRH_CTL_VOLTAGE_MULTIPLIER_SHIFT);
> +
> +	return 0;
> +}
> +
> +static int cprh_dummy_set_performance_state(struct generic_pm_domain *domain,
> +					    unsigned int state)
> +{
> +	return 0;
> +}
> +
> +static int cpr_set_performance_state(struct generic_pm_domain *domain,
> +				     unsigned int state)
> +{
> +	struct cpr_thread *thread = container_of(domain, struct cpr_thread, pd);
> +	struct cpr_drv *drv = thread->drv;
> +	struct corner *corner, *end;
> +	int ret = 0;
> +
> +	mutex_lock(&drv->lock);
> +
> +	dev_dbg(drv->dev, "setting perf state: %u (prev state: %u thread: %u)\n",
> +		state, cpr_get_cur_perf_state(thread), thread->id);
> +
> +	/*
> +	 * Determine new corner we're going to.
> +	 * Remove one since lowest performance state is 1.
> +	 */
> +	corner = thread->corners + state - 1;
> +	end = &thread->corners[thread->num_corners - 1];
> +	if (corner > end || corner < thread->corners) {
> +		ret = -EINVAL;
> +		goto unlock;
> +	}
> +
> +	cpr_ctl_disable(thread);
> +
> +	cpr_irq_clr(thread);
> +	if (thread->corner != corner)
> +		cpr_corner_restore(thread, corner);
> +
> +	ret = cpr_commit_state(thread);
> +	if (ret)
> +		goto unlock;
> +
> +	cpr_ctl_enable(thread);
> +unlock:
> +	mutex_unlock(&drv->lock);
> +
> +	dev_dbg(drv->dev, "set perf state %u on thread %u\n", state, thread->id);
> +
> +	return ret;
> +}
> +
> +/**
> + * cpr3_adjust_quot - Adjust the closed-loop quotients
> + * @ring_osc_factor:  Ring oscillator adjustment factor
> + * @volt_closed_loop: Closed-loop voltage adjustment factor
> + *
> + * Calculates the quotient adjustment factor based on closed-loop
> + * quotients and ring oscillator factor.
> + *
> + * Return: Adjusted quotient
> + */
> +static int cpr3_adjust_quot(int ring_osc_factor, int volt_closed_loop)
> +{
> +	s64 temp;
> +
> +	if (ring_osc_factor == 0 || volt_closed_loop == 0)
> +		return 0;
> +
> +	temp = (s64)(ring_osc_factor * volt_closed_loop);
> +	return (int)div_s64(temp, 1000000);
> +}
> +
> +/**
> + * cpr_fuse_corner_init() - Calculate fuse corner table
> + * @thread: Structure holding CPR thread-specific parameters
> + *
> + * This function populates the fuse corners table by reading the
> + * values from the fuses, eventually adjusting them with a fixed
> + * per-corner offset and doing basic checks about them being
> + * supported by the regulator that is assigned to this CPR - if
> + * it is available (on CPR-Hardened, there is no usable vreg, as
> + * that is protected by the hypervisor).
> + *
> + * Return: Zero for success, negative number on error
> + */
> +static int cpr_fuse_corner_init(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	const struct cpr_thread_desc *desc = thread->desc;
> +	const struct cpr_fuse *cpr_fuse = thread->cpr_fuses;
> +	struct fuse_corner_data *fdata;
> +	struct fuse_corner *fuse, *prev_fuse, *end;
> +	int i, ret;
> +
> +	/* Populate fuse_corner members */
> +	fuse = thread->fuse_corners;
> +	prev_fuse = &fuse[0];
> +	end = &fuse[desc->num_fuse_corners - 1];
> +	fdata = desc->fuse_corner_data;
> +
> +	for (i = 0; fuse <= end; fuse++, cpr_fuse++, i++, fdata++) {
> +		int factor = cpr_get_ro_factor(desc, i, fuse->ring_osc_idx);
> +
> +		ret = cpr_populate_fuse_common(drv->dev, fdata, cpr_fuse,
> +					       fuse, drv->vreg_step,
> +					       desc->init_voltage_width,
> +					       desc->init_voltage_step);
> +		if (ret)
> +			return ret;
> +
> +		/*
> +		 * Adjust the fuse quot with per-fuse-corner closed-loop
> +		 * voltage adjustment parameters.
> +		 */
> +		fuse->quot += cpr3_adjust_quot(factor, fdata->volt_cloop_adjust);
> +
> +		/* CPRh: no regulator access... */
> +		if (drv->desc->cpr_type == CTRL_TYPE_CPRH)
> +			goto skip_pvs_restrict;
> +
> +		/* Re-check if corner voltage range is supported by regulator */
> +		ret = cpr_check_vreg_constraints(drv->dev, drv->vreg, fuse);
> +		if (ret)
> +			return ret;
> +
> +skip_pvs_restrict:
> +		if (fuse->uV < prev_fuse->uV)
> +			fuse->uV = prev_fuse->uV;
> +		prev_fuse = fuse;
> +		dev_dbg(drv->dev, "fuse corner %d: [%d %d %d] RO%u quot %d\n",
> +			i, fuse->min_uV, fuse->uV, fuse->max_uV,
> +			fuse->ring_osc_idx, fuse->quot);
> +
> +		/* Check if constraints are valid */
> +		if (fuse->uV < fuse->min_uV || fuse->uV > fuse->max_uV) {
> +			dev_err(drv->dev, "fuse corner %d: Bad voltage range.\n", i);
> +			return -EINVAL;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +static void cpr3_restrict_corner(struct corner *corner, int threshold,
> +				 int hysteresis, int step)
> +{
> +	if (threshold > corner->min_uV && threshold <= corner->max_uV) {
> +		if (corner->uV >= threshold) {
> +			corner->min_uV = max(corner->min_uV,
> +					     threshold - hysteresis);
> +			if (corner->min_uV > corner->uV)
> +				corner->uV = corner->min_uV;
> +		} else {
> +			corner->max_uV = threshold;
> +			corner->max_uV -= step;
> +		}
> +	}
> +}
> +
> +/*
> + * cprh_corner_adjust_opps() - Set voltage on each CPU OPP table entry

This should be called for non-CPRh targets too. For example, we can use 
this data to populate scaling tables for GPMU.

> + *
> + * On CPR-Hardened, the voltage level is controlled internally through
> + * the OSM hardware: in order to initialize the latter, we have to
> + * communicate the voltage to its driver, so that it will be able to
> + * write the right parameters (as they have to be set both on the CPRh
> + * and on the OSM) on it.
> + * This function is called only for CPRh.
> + *
> + * Return: Zero for success, negative number for error.
> + */
> +static int cprh_corner_adjust_opps(struct cpr_thread *thread)
> +{
> +	struct corner *corner = thread->corners;
> +	struct cpr_drv *drv = thread->drv;
> +	int i, ret;
> +
> +	for (i = 0; i < thread->num_corners; i++) {
> +		ret = dev_pm_opp_adjust_voltage(thread->attached_cpu_dev,

This will work if there are no other power domains set for the CPU/GPU 
devices. If e.g. the device has cpr and mx power domains, then 
pd_attach_dev callback will be called with the virtual genpd device.

I fixed this for me by adding base_dev pointer to pd.attach_dev() 
callback (and then using base_dev for the OPP calls).

> +						corner[i].freq,
> +						corner[i].uV,
> +						corner[i].min_uV,
> +						corner[i].max_uV);
> +		if (ret)
> +			break;
> +
> +		dev_dbg(drv->dev, "OPP voltage adjusted for %lu kHz, %d uV\n",
> +			corner[i].freq, corner[i].uV);
> +	}
> +
> +	/* If we couldn't adjust voltage for all corners, something went wrong */
> +	if (i < thread->num_corners)
> +		return -EINVAL;
> +
> +	return ret;
> +}
> +
> +/**
> + * cpr3_corner_init() - Calculate and set-up corners for the CPR HW
> + * @thread: Structure holding CPR thread-specific parameters
> + *
> + * This function calculates all the corner parameters by comparing
> + * and interpolating the values read from the various set-points
> + * read from the fuses (also called "fuse corners") to generate and
> + * program to the CPR a lookup table that describes each voltage
> + * step, mapped to a performance level (or corner number).
> + *
> + * It also programs other essential parameters on the CPR and - if
> + * we are dealing with CPR-Hardened, it will also enable the internal
> + * interface between the Operating State Manager (OSM) and the CPRh
> + * in order to achieve CPU DVFS.
> + *
> + * Return: Zero for success, negative number on error
> + */
> +static int cpr3_corner_init(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	const struct cpr_desc *desc = drv->desc;
> +	const struct cpr_thread_desc *tdesc = thread->desc;
> +	const struct cpr_fuse *fuses = thread->cpr_fuses;
> +	int i, ret, total_corners, extra_corners, level, scaling = 0;
> +	unsigned int fnum, fc;
> +	const char *quot_offset;
> +	const struct fuse_corner_data *fdata;
> +	struct fuse_corner *fuse, *prev_fuse;
> +	struct corner *corner, *prev_corner, *end;
> +	struct corner_data *cdata;
> +	struct dev_pm_opp *opp;
> +	unsigned long freq;
> +	u32 ring_osc_mask = CPR3_RO_MASK, min_quotient = U32_MAX;
> +
> +	corner = thread->corners;
> +	prev_corner = &thread->corners[0];
> +	end = &corner[thread->num_corners - 1];
> +
> +	cdata = devm_kcalloc(drv->dev, thread->num_corners + drv->extra_corners,
> +			     sizeof(struct corner_data), GFP_KERNEL);
> +	if (!cdata)
> +		return -ENOMEM;
> +
> +	for (level = 1; level <= thread->num_corners; level++) {
> +		opp = dev_pm_opp_find_level_exact(&thread->pd.dev, level);
> +		if (IS_ERR(opp))
> +			return -EINVAL;
> +
> +		/*
> +		 * If there is only one specified qcom,opp-fuse-level, then
> +		 * it is assumed that this only one is global and valid for
> +		 * all IDs, so try to get the specific one but, on failure,
> +		 * go for the global one.
> +		 */
> +		fc = cpr_get_fuse_corner(opp, thread->id);
> +		if (fc == 0) {
> +			fc = cpr_get_fuse_corner(opp, 0);
> +			if (fc == 0) {
> +				dev_err(drv->dev, "qcom,opp-fuse-level is missing!\n");
> +				dev_pm_opp_put(opp);
> +				return -EINVAL;
> +			}
> +		}
> +		fnum = fc - 1;
> +
> +		freq = cpr_get_opp_hz_for_req(opp, thread->attached_cpu_dev);
> +		if (!freq) {
> +			thread->num_corners = max(level - 1, 0);
> +			end = &thread->corners[thread->num_corners - 1];
> +			break;
> +		}
> +
> +		/*
> +		 * If any post-vadj (open/closed loop) is not specified, then
> +		 * it's zero, meaning that it is not required for this corner.
> +		 */
> +		cpr_get_corner_post_vadj(opp, thread->id,
> +					 &cdata[level - 1].oloop_vadj,
> +					 &cdata[level - 1].cloop_vadj);
> +		cdata[level - 1].fuse_corner = fnum;
> +		cdata[level - 1].freq = freq;
> +
> +		fuse = &thread->fuse_corners[fnum];
> +		dev_dbg(drv->dev, "freq: %lu level: %u fuse level: %u\n",
> +			freq, dev_pm_opp_get_level(opp) - 1, fnum);
> +		if (freq > fuse->max_freq)
> +			fuse->max_freq = freq;
> +		dev_pm_opp_put(opp);
> +
> +		/*
> +		 * Make sure that the frequencies in the table are in ascending
> +		 * order, as this is critical for the algorithm to work.
> +		 */
> +		if (cdata[level - 2].freq > freq) {
> +			dev_err(drv->dev, "Frequency table not in ascending order.\n");
> +			return -EINVAL;
> +		}
> +	}
> +
> +	if (thread->num_corners < 2) {
> +		dev_err(drv->dev, "need at least 2 OPPs to use CPR\n");
> +		return -EINVAL;
> +	}
> +
> +	/*
> +	 * Get the quotient adjustment scaling factor, according to:
> +	 *
> +	 * scaling = min(1000 * (QUOT(corner_N) - QUOT(corner_N-1))
> +	 *		/ (freq(corner_N) - freq(corner_N-1)), max_factor)
> +	 *
> +	 * QUOT(corner_N):	quotient read from fuse for fuse corner N
> +	 * QUOT(corner_N-1):	quotient read from fuse for fuse corner (N - 1)
> +	 * freq(corner_N):	max frequency in MHz supported by fuse corner N
> +	 * freq(corner_N-1):	max frequency in MHz supported by fuse corner
> +	 *			 (N - 1)
> +	 *
> +	 * Then walk through the corners mapped to each fuse corner
> +	 * and calculate the quotient adjustment for each one using the
> +	 * following formula:
> +	 *
> +	 * quot_adjust = (freq_max - freq_corner) * scaling / 1000
> +	 *
> +	 * freq_max: max frequency in MHz supported by the fuse corner
> +	 * freq_corner: frequency in MHz corresponding to the corner
> +	 * scaling: calculated from above equation
> +	 *
> +	 *
> +	 *     +                           +
> +	 *     |                         v |
> +	 *   q |           f c           o |           f c
> +	 *   u |         c               l |         c
> +	 *   o |       f                 t |       f
> +	 *   t |     c                   a |     c
> +	 *     | c f                     g | c f
> +	 *     |                         e |
> +	 *     +---------------            +----------------
> +	 *       0 1 2 3 4 5 6               0 1 2 3 4 5 6
> +	 *          corner                      corner
> +	 *
> +	 *    c = corner
> +	 *    f = fuse corner
> +	 *
> +	 */
> +	for (i = 0; corner <= end; corner++, i++) {
> +		unsigned long freq_diff_mhz;
> +		int ro_fac, vadj, prev_quot;
> +
> +		fnum = cdata[i].fuse_corner;
> +		fdata = &tdesc->fuse_corner_data[fnum];
> +		quot_offset = fuses[fnum].quotient_offset;
> +		fuse = &thread->fuse_corners[fnum];
> +		ring_osc_mask &= (u16)(~BIT(fuse->ring_osc_idx));
> +		if (fnum)
> +			prev_fuse = &thread->fuse_corners[fnum - 1];
> +		else
> +			prev_fuse = NULL;
> +
> +		corner->fuse_corner = fuse;
> +		corner->freq = cdata[i].freq;
> +		corner->uV = fuse->uV;
> +
> +		if (prev_fuse) {
> +			if (prev_fuse->ring_osc_idx == fuse->ring_osc_idx)
> +				quot_offset = NULL;
> +
> +			scaling = cpr_calculate_scaling(drv->dev, quot_offset,
> +							fdata, corner);

CPR for GFX uses several RO for each corner. Can you possibly split this 
into separate pieces: one for scaling coefficinent calculation and 
another one for calculating quot adjustment?

> +			if (scaling < 0)
> +				return scaling;
> +
> +			freq_diff_mhz = fuse->max_freq - corner->freq;
> +			do_div(freq_diff_mhz, 1000000); /* now in MHz */
> +
> +			corner->quot_adjust = scaling * freq_diff_mhz;
> +			do_div(corner->quot_adjust, 1000);

Rather than having corner->quot_adjust which is used to adjust 
fuse->quot, it might be better to add an array quot[CPR3_RO_COUNT] to 
corner and then set relevant quot directly (this will make it easier to 
use it for GFX CPR).

> +
> +			/* Fine-tune QUOT (closed-loop) based on fixed values */
> +			ro_fac = cpr_get_ro_factor(tdesc, fnum, fuse->ring_osc_idx);
> +			vadj = cdata[i].cloop_vadj;
> +			corner->quot_adjust -= cpr3_adjust_quot(ro_fac, vadj);
> +			dev_vdbg(drv->dev, "Quot fine-tuning to %d for post-vadj=%d\n",
> +				 corner->quot_adjust, vadj);
> +
> +			/*
> +			 * Make sure that we scale (up) monotonically.
> +			 * P.S.: Fuse quots can never be descending.
> +			 */
> +			prev_quot = prev_corner->fuse_corner->quot;
> +			prev_quot -= prev_corner->quot_adjust;
> +			if (fuse->quot - corner->quot_adjust < prev_quot) {
> +				int new_adj = prev_corner->fuse_corner->quot;
> +
> +				new_adj -= fuse->quot;
> +				dev_vdbg(drv->dev, "Monotonic increase forced: %d->%d\n",
> +					 corner->quot_adjust, new_adj);
> +				corner->quot_adjust = new_adj;
> +			}
> +
> +			corner->uV = cpr_interpolate(corner,
> +						     drv->vreg_step, fdata);
> +		}
> +		/* Negative fuse quotients are nonsense. */
> +		if (fuse->quot < corner->quot_adjust)
> +			return -EINVAL;
> +
> +		min_quotient = min(min_quotient,
> +				   (u32)(fuse->quot - corner->quot_adjust));
> +
> +		/* Fine-tune voltages (open-loop) based on fixed values */
> +		corner->uV += cdata[i].oloop_vadj;
> +		dev_dbg(drv->dev, "Voltage fine-tuning to %d for post-vadj=%d\n",
> +			corner->uV, cdata[i].oloop_vadj);
> +
> +		corner->max_uV = fuse->max_uV;
> +		corner->min_uV = fuse->min_uV;
> +		corner->uV = clamp(corner->uV, corner->min_uV, corner->max_uV);
> +		dev_vdbg(drv->dev, "Clamped after interpolation: [%d %d %d]\n",
> +			 corner->min_uV, corner->uV, corner->max_uV);
> +
> +		/* Make sure that we scale monotonically here, too. */
> +		if (corner->uV < prev_corner->uV)
> +			corner->uV = prev_corner->uV;
> +
> +		corner->last_uV = corner->uV;
> +
> +		/* Reduce the ceiling voltage if needed */
> +		if (desc->reduce_to_corner_uV && corner->uV < corner->max_uV)
> +			corner->max_uV = corner->uV;
> +		else if (desc->reduce_to_fuse_uV && fuse->uV < corner->max_uV)
> +			corner->max_uV = max(corner->min_uV, fuse->uV);
> +
> +		corner->min_uV = max(corner->max_uV - fdata->range_uV,
> +				     corner->min_uV);
> +
> +		/*
> +		 * Adjust per-corner floor and ceiling voltages so that
> +		 * they do not overlap the memory Array Power Mux (APM)
> +		 * nor the Memory Accelerator (MEM-ACC) threshold voltages.
> +		 */
> +		if (desc->apm_threshold)
> +			cpr3_restrict_corner(corner, desc->apm_threshold,
> +					     desc->apm_hysteresis,
> +					     drv->vreg_step);
> +		if (desc->mem_acc_threshold)
> +			cpr3_restrict_corner(corner, desc->mem_acc_threshold,
> +					     0, drv->vreg_step);
> +
> +		prev_corner = corner;
> +		dev_dbg(drv->dev, "corner %d: [%d %d %d] scaling %d quot %d\n", i,
> +			corner->min_uV, corner->uV, corner->max_uV, scaling,
> +			fuse->quot - corner->quot_adjust);
> +	}
> +
> +	/* Additional setup for CPRh only */
> +	if (desc->cpr_type < CTRL_TYPE_CPRH)
> +		return 0;
> +
> +	/* If the OPPs can't be adjusted, programming the CPRh is useless */
> +	ret = cprh_corner_adjust_opps(thread);
> +	if (ret) {
> +		dev_err(drv->dev, "Cannot adjust CPU OPP voltages: %d\n", ret);
> +		return ret;
> +	}
> +
> +	total_corners = thread->num_corners;
> +	extra_corners = drv->extra_corners;
> +
> +	/* If the APM extra corner exists, add it now. */
> +	if (desc->apm_crossover && desc->apm_threshold && extra_corners) {
> +		/* Program the APM crossover corner on the CPR-Hardened */

I'd again ask to split generic code and CPRh code. CPRh code differs 
quite significantly. It would be easier to comprehend CPR driver if CPRh 
and CPR[34] code is split apart (yes, I had to do this while 
fighting/finding my way through the driver).

> +		thread->corners[total_corners].uV = desc->apm_crossover;
> +		thread->corners[total_corners].min_uV = desc->apm_crossover;
> +		thread->corners[total_corners].max_uV = desc->apm_crossover;
> +		thread->corners[total_corners].is_open_loop = true;
> +
> +		/*
> +		 * We have calculated the APM parameters for this clock plan:
> +		 * make the APM *threshold* available to external callers.
> +		 * The crossover is used only internally in the CPR.
> +		 */
> +		thread->ext_data.apm_threshold_uV = desc->apm_threshold;
> +
> +		dev_dbg(drv->dev, "corner %d (APM): [%d %d %d] Open-Loop\n",
> +			total_corners, desc->apm_crossover,
> +			desc->apm_crossover, desc->apm_crossover);
> +
> +		total_corners++;
> +		extra_corners--;
> +	}
> +
> +	if (desc->mem_acc_threshold && extra_corners) {
> +		/* Program the Memory Accelerator threshold corner to CPRh */
> +		thread->corners[total_corners].uV = desc->mem_acc_threshold;
> +		thread->corners[total_corners].min_uV = desc->mem_acc_threshold;
> +		thread->corners[total_corners].max_uV = desc->mem_acc_threshold;
> +		thread->corners[total_corners].is_open_loop = true;
> +
> +		/*
> +		 * We have calculated a mem-acc threshold for this clock plan:
> +		 * make it available to external callers.
> +		 */
> +		thread->ext_data.mem_acc_threshold_uV = desc->mem_acc_threshold;
> +
> +		dev_dbg(drv->dev, "corner %d (MEMACC): [%d %d %d] Open-Loop\n",
> +			total_corners, desc->mem_acc_threshold,
> +			desc->mem_acc_threshold, desc->mem_acc_threshold);
> +
> +		total_corners++;
> +		extra_corners--;
> +	}
> +
> +	/*
> +	 * If there are any extra corners left, it means that even though we
> +	 * expect to fill in both APM and MEM-ACC crossovers, one couldn't
> +	 * satisfy requirements, which means that the specified parameters
> +	 * are wrong: in this case, inform the user and bail out, otherwise
> +	 * if we go on writing the (invalid) table to the CPR-Hardened, the
> +	 * hardware (in this case, the CPU) will surely freeze and crash.
> +	 */
> +	if (unlikely(extra_corners)) {
> +		dev_err(drv->dev, "APM/MEM-ACC corners: bad parameters.\n");
> +		return -EINVAL;
> +	}
> +	/* Reassign extra_corners, as we have to exclude delta_quot for them */
> +	extra_corners = drv->extra_corners;
> +
> +	/* Disable the interface between OSM and CPRh */
> +	cpr_masked_write(thread, drv->reg_ctl,
> +			 CPRH_CTL_OSM_ENABLED, 0);
> +
> +	/* Program the GCNT before unmasking ring oscillator(s) */
> +	for (i = 0; i < CPR3_RO_COUNT; i++) {
> +		if (!(ring_osc_mask & BIT(i))) {
> +			cpr_write(thread, CPR3_REG_GCNT(i), drv->gcnt);
> +			dev_vdbg(drv->dev, "RO%d gcnt=%d\n", i, drv->gcnt);
> +		}
> +	}
> +
> +	/*
> +	 * Unmask the ring oscillator(s) that we're going to use: it seems
> +	 * to be mandatory to do this *before* sending the rest of the
> +	 * CPRhardened specific configuration.
> +	 */
> +	dev_dbg(drv->dev, "Unmasking ring oscillators with mask 0x%x\n", ring_osc_mask);
> +	cpr_write(thread, CPR3_REG_RO_MASK(tdesc->hw_tid), ring_osc_mask);
> +
> +	/* Setup minimum quotients for ring oscillators */
> +	for (i = 0; i < CPR3_RO_COUNT; i++) {
> +		u32 tgt_quot_reg = CPR3_REG_TARGET_QUOT(tdesc->hw_tid, i);
> +		u32 tgt_quot_val = 0;
> +
> +		if (!(ring_osc_mask & BIT(i)))
> +			tgt_quot_val = min_quotient;
> +
> +		cpr_write(thread, tgt_quot_reg, tgt_quot_val);
> +		dev_vdbg(drv->dev, "Programmed min quotient %u for Ring Oscillator %d\n",
> +			 tgt_quot_val, tgt_quot_reg);
> +	}
> +
> +	for (i = 0; i < total_corners; i++) {
> +		int volt_oloop_steps, volt_floor_steps, delta_quot_steps;
> +		int ring_osc;
> +		u32 val;
> +
> +		fnum = cdata[i].fuse_corner;
> +		fuse = &thread->fuse_corners[fnum];
> +
> +		val = thread->corners[i].uV - desc->cpr_base_voltage;
> +		volt_oloop_steps = DIV_ROUND_UP(val, drv->vreg_step);
> +
> +		val = thread->corners[i].min_uV - desc->cpr_base_voltage;
> +		volt_floor_steps = DIV_ROUND_UP(val, drv->vreg_step);
> +
> +		/*
> +		 * If we are accessing corners that are not used as
> +		 * an active DCVS set-point, then always select RO 0
> +		 * and zero out the delta quotient.
> +		 */
> +		if (i >= thread->num_corners) {
> +			ring_osc = 0;
> +			delta_quot_steps = 0;
> +		} else {
> +			ring_osc = fuse->ring_osc_idx;
> +			val = fuse->quot - thread->corners[i].quot_adjust;
> +			val -= min_quotient;
> +			delta_quot_steps = DIV_ROUND_UP(val,
> +						CPRH_DELTA_QUOT_STEP_FACTOR);
> +		}
> +
> +		if (volt_oloop_steps > CPRH_CORNER_INIT_VOLTAGE_MAX_VALUE  ||
> +		    volt_floor_steps > CPRH_CORNER_FLOOR_VOLTAGE_MAX_VALUE ||
> +		    delta_quot_steps > CPRH_CORNER_QUOT_DELTA_MAX_VALUE) {
> +			dev_err(drv->dev, "Invalid cfg: oloop=%d, floor=%d, delta=%d\n",
> +				volt_oloop_steps, volt_floor_steps,
> +				delta_quot_steps);
> +			return -EINVAL;
> +		}
> +		/* Green light: Go, Go, Go! */
> +
> +		/* Set number of open-loop steps */
> +		val = volt_oloop_steps << CPRH_CORNER_INIT_VOLTAGE_SHIFT;
> +		val &= CPRH_CORNER_INIT_VOLTAGE_MASK;
> +
> +		/* Set number of floor voltage steps */
> +		val |= (volt_floor_steps << CPRH_CORNER_FLOOR_VOLTAGE_SHIFT) &
> +		       CPRH_CORNER_FLOOR_VOLTAGE_MASK;
> +
> +		/* Set number of target quotient delta steps */
> +		val |= (delta_quot_steps << CPRH_CORNER_QUOT_DELTA_SHIFT) &
> +		       CPRH_CORNER_QUOT_DELTA_MASK;
> +
> +		/* Select ring oscillator for this corner */
> +		val |= (ring_osc << CPRH_CORNER_RO_SEL_SHIFT) &
> +		       CPRH_CORNER_RO_SEL_MASK;
> +
> +		/* Open loop corner is usually APM/ACC crossover */
> +		if (thread->corners[i].is_open_loop) {
> +			dev_dbg(drv->dev, "Disabling Closed-Loop on corner %d\n", i);
> +			val |= CPRH_CORNER_CPR_CL_DISABLE;
> +		}
> +		cpr_write(thread, CPRH_REG_CORNER(drv, tdesc->hw_tid, i), val);
> +
> +		dev_dbg(drv->dev, "steps [%d]: open-loop %d, floor %d, delta_quot %d\n",
> +			i, volt_oloop_steps, volt_floor_steps,
> +			delta_quot_steps);
> +	}
> +
> +	/* YAY! Setup is done! Enable the internal loop to start CPR. */
> +	cpr_masked_write(thread, CPR3_REG_CPR_CTL,
> +			CPR3_CPR_CTL_LOOP_EN_MASK,
> +			CPR3_CPR_CTL_LOOP_EN_MASK);
> +
> +	/*
> +	 * All the writes are going through before enabling internal
> +	 * communication between the OSM and the CPRh controllers
> +	 * because we are never using relaxed accessors, but should
> +	 * we use them, it would be critical to issue a barrier here,
> +	 * otherwise there is a high risk of hardware lockups due to
> +	 * under-voltage for the selected CPU clock.
> +	 *
> +	 * Please note that the CPR-hardened gets set-up in Linux but
> +	 * then gets actually used in firmware (and only by the OSM);
> +	 * after handing it off we will have no more control on it.
> +	 */
> +
> +	/* Enable the interface between OSM and CPRh */
> +	cpr_masked_write(thread, drv->reg_ctl,
> +			 CPRH_CTL_OSM_ENABLED,
> +			 CPRH_CTL_OSM_ENABLED);
> +
> +	/* On success, free cdata manually */
> +	devm_kfree(drv->dev, cdata);
> +	return 0;
> +}
> +
> +/**
> + * cpr3_init_parameters() - Initialize CPR global parameters
> + * @drv: Main driver structure
> + *
> + * Initial "integrity" checks and setup for the thread-independent parameters.
> + *
> + * Return: Zero for success, negative number on error
> + */
> +static int cpr3_init_parameters(struct cpr_drv *drv)
> +{
> +	const struct cpr_desc *desc = drv->desc;
> +	struct clk *clk;
> +
> +	clk = devm_clk_get(drv->dev, "ref");
> +	if (IS_ERR(clk))
> +		return PTR_ERR(clk);
> +
> +	drv->ref_clk_khz = clk_get_rate(clk);
> +	do_div(drv->ref_clk_khz, 1000);
> +
> +	/* On CPRh this clock is not always-on... */
> +	if (desc->cpr_type == CTRL_TYPE_CPRH)
> +		clk_prepare_enable(clk);
> +	else
> +		devm_clk_put(drv->dev, clk);
> +
> +	if (desc->timer_cons_up > CPR3_THRESH_CONS_UP_MASK ||
> +	    desc->timer_cons_down > CPR3_THRESH_CONS_DOWN_MASK ||
> +	    desc->up_threshold > CPR3_THRESH_UP_THRESH_MASK ||
> +	    desc->down_threshold > CPR3_THRESH_DOWN_THRESH_MASK ||
> +	    desc->idle_clocks > CPR3_CPR_CTL_IDLE_CLOCKS_MASK ||
> +	    desc->count_mode > CPR3_CPR_CTL_COUNT_MODE_MASK ||
> +	    desc->count_repeat > CPR3_CPR_CTL_COUNT_REPEAT_MASK)
> +		return -EINVAL;
> +
> +	/*
> +	 * Read the CPR version register only from CPR3 onwards:
> +	 * this is needed to get the additional register offsets.
> +	 *
> +	 * Note: When threaded, even if multi-controller, there
> +	 *       is no chance to have different versions at the
> +	 *       same time in the same domain, so it is safe to
> +	 *       check this only on the first controller/thread.
> +	 */
> +	drv->cpr_hw_rev = cpr_read(&drv->threads[0],
> +				   CPR3_REG_CPR_VERSION);
> +	dev_dbg(drv->dev, "CPR hardware revision: 0x%x\n", drv->cpr_hw_rev);
> +
> +	if (drv->cpr_hw_rev >= CPRH_CPR_VERSION_4P5) {
> +		drv->reg_corner = 0x3500;
> +		drv->reg_corner_tid = 0xa0;
> +		drv->reg_ctl = 0x3a80;
> +		drv->reg_status = 0x3a84;
> +	} else {
> +		drv->reg_corner = 0x3a00;
> +		drv->reg_corner_tid = 0;
> +		drv->reg_ctl = 0x3aa0;
> +		drv->reg_status = 0x3aa4;
> +	}
> +
> +	dev_dbg(drv->dev, "up threshold = %u, down threshold = %u\n",
> +		desc->up_threshold, desc->down_threshold);
> +
> +	return 0;
> +}
> +
> +/**
> + * cpr3_find_initial_corner() - Finds boot-up p-state and enables CPR
> + * @thread: Structure holding CPR thread-specific parameters
> + *
> + * Differently from CPRv1, from CPRv3 onwards when we successfully find
> + * the target boot-up performance state, we must refresh the HW
> + * immediately to guarantee system stability and to avoid overheating
> + * during the boot process, thing that would more likely happen without
> + * this driver doing its job.
> + *
> + * Return: Zero for success, negative number on error
> + */
> +static int cpr3_find_initial_corner(struct cpr_thread *thread)
> +{
> +	struct cpr_drv *drv = thread->drv;
> +	struct corner *corner;
> +	int uV, idx;
> +
> +	idx = cpr_find_initial_corner(drv->dev, thread->cpu_clk,
> +				      thread->corners,
> +				      thread->num_corners);
> +	if (idx < 0)
> +		return idx;
> +
> +	cpr_ctl_disable(thread);
> +
> +	corner = &thread->corners[idx];
> +	cpr_corner_restore(thread, corner);
> +
> +	uV = regulator_get_voltage(drv->vreg);
> +	uV = clamp(uV, corner->min_uV, corner->max_uV);
> +
> +	corner->last_uV = uV;
> +	if (!drv->last_uV)
> +		drv->last_uV = uV;
> +
> +	cpr_commit_state(thread);
> +	thread->enabled = true;
> +	cpr_switch(drv);
> +
> +	return 0;
> +}
> +
> +static const int msm8998_gold_scaling_factor[][CPR3_RO_COUNT] = {
> +	/* Fuse Corner 0 */
> +	{
> +		2857, 3057, 2828, 2952, 2699, 2798, 2446, 2631,
> +		2629, 2578, 2244, 3344, 3289, 3137, 3164, 2655
> +	},
> +	/* Fuse Corner 1 */
> +	{
> +		2857, 3057, 2828, 2952, 2699, 2798, 2446, 2631,
> +		2629, 2578, 2244, 3344, 3289, 3137, 3164, 2655
> +	},
> +	/* Fuse Corner 2 */
> +	{
> +		2603, 2755, 2676, 2777, 2573, 2685, 2465, 2610,
> +		2312, 2423, 2243, 3104, 3022, 3036, 2740, 2303
> +	},
> +	/* Fuse Corner 3 */
> +	{
> +		1901, 2016, 2096, 2228, 2034, 2161, 2077, 2188,
> +		1565, 1870, 1925, 2235, 2205, 2413, 1762, 1478
> +	}
> +};
> +
> +static const int msm8998_silver_scaling_factor[][CPR3_RO_COUNT] = {
> +	/* Fuse Corner 0 */
> +	{
> +		2595, 2794, 2577, 2762, 2471, 2674, 2199, 2553,
> +		3189, 3255, 3192, 2962, 3054, 2982, 2042, 2945
> +	},
> +	/* Fuse Corner 1 */
> +	{
> +		2595, 2794, 2577, 2762, 2471, 2674, 2199, 2553,
> +		3189, 3255, 3192, 2962, 3054, 2982, 2042, 2945
> +	},
> +	/* Fuse Corner 2 */
> +	{
> +		2391, 2550, 2483, 2638, 2382, 2564, 2259, 2555,
> +		2766, 3041, 2988, 2935, 2873, 2688, 2013, 2784
> +	},
> +	/* Fuse Corner 3 */
> +	{
> +		2066, 2153, 2300, 2434, 2220, 2386, 2288, 2465,
> +		2028, 2511, 2487, 2734, 2554, 2117, 1892, 2377
> +	}
> +};
> +
> +static const struct cpr_thread_desc msm8998_thread_gold = {
> +	.controller_id = 1,
> +	.hw_tid = 0,
> +	.ro_scaling_factor = msm8998_gold_scaling_factor,
> +	.ro_avail_corners = ARRAY_SIZE(msm8998_gold_scaling_factor),
> +	.sensor_range_start = 0,
> +	.sensor_range_end = 9,
> +	.init_voltage_step = 10000,
> +	.init_voltage_width = 6,
> +	.step_quot_init_min = 9,
> +	.step_quot_init_max = 14,
> +	.num_fuse_corners = 4,
> +	.fuse_corner_data = (struct fuse_corner_data[]){
> +		/* fuse corner 0 */
> +		{
> +			.ref_uV = 756000,
> +			.max_uV = 828000,
> +			.min_uV = 568000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = 0,
> +			.volt_oloop_adjust = 8000,
> +			.max_volt_scale = 4,
> +			.max_quot_scale = 10,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 1 */
> +		{
> +			.ref_uV = 756000,
> +			.max_uV = 900000,
> +			.min_uV = 624000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = 0,
> +			.volt_oloop_adjust = 0,
> +			.max_volt_scale = 320,
> +			.max_quot_scale = 350,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 2 */
> +		{
> +			.ref_uV = 828000,
> +			.max_uV = 952000,
> +			.min_uV = 632000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = 12000,
> +			.volt_oloop_adjust = 12000,
> +			.max_volt_scale = 620,
> +			.max_quot_scale = 750,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 3 */
> +		{
> +			.ref_uV = 1056000,
> +			.max_uV = 1136000,
> +			.min_uV = 772000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = 50000,
> +			.volt_oloop_adjust = 52000,
> +			.max_volt_scale = 580,
> +			.max_quot_scale = 1040,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +	},
> +};
> +
> +static const struct cpr_thread_desc msm8998_thread_silver = {
> +	.controller_id = 0,
> +	.hw_tid = 0,
> +	.ro_scaling_factor = msm8998_silver_scaling_factor,
> +	.ro_avail_corners = ARRAY_SIZE(msm8998_silver_scaling_factor),
> +	.sensor_range_start = 0,
> +	.sensor_range_end = 6,
> +	.init_voltage_step = 10000,
> +	.init_voltage_width = 6,
> +	.step_quot_init_min = 11,
> +	.step_quot_init_max = 12,
> +	.num_fuse_corners = 4,
> +	.fuse_corner_data = (struct fuse_corner_data[]){
> +		/* fuse corner 0 */
> +		{
> +			.ref_uV = 688000,
> +			.max_uV = 828000,
> +			.min_uV = 568000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = 20000,
> +			.volt_oloop_adjust = 40000,
> +			.max_volt_scale = 4,
> +			.max_quot_scale = 10,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 1 */
> +		{
> +			.ref_uV = 756000,
> +			.max_uV = 900000,
> +			.min_uV = 632000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = 26000,
> +			.volt_oloop_adjust = 24000,
> +			.max_volt_scale = 500,
> +			.max_quot_scale = 800,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 2 */
> +		{
> +			.ref_uV = 828000,
> +			.max_uV = 952000,
> +			.min_uV = 664000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = 12000,
> +			.volt_oloop_adjust = 12000,
> +			.max_volt_scale = 280,
> +			.max_quot_scale = 650,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +
> +		},
> +		/* fuse corner 3 */
> +		{
> +			.ref_uV = 1056000,
> +			.max_uV = 1056000,
> +			.min_uV = 772000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = 30000,
> +			.volt_oloop_adjust = 30000,
> +			.max_volt_scale = 430,
> +			.max_quot_scale = 800,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +	},
> +};
> +
> +static const struct cpr_desc msm8998_cpr_desc = {
> +	.cpr_type = CTRL_TYPE_CPRH,
> +	.num_threads = 2,
> +	.mem_acc_threshold = 852000,
> +	.apm_threshold = 800000,
> +	.apm_crossover = 880000,
> +	.apm_hysteresis = 0,
> +	.cpr_base_voltage = 352000,
> +	.cpr_max_voltage = 1200000,
> +	.timer_delay_us = 5000,
> +	.timer_cons_up = 0,
> +	.timer_cons_down = 2,
> +	.up_threshold = 2,
> +	.down_threshold = 2,
> +	.idle_clocks = 15,
> +	.count_mode = CPR3_CPR_CTL_COUNT_MODE_ALL_AT_ONCE_MIN,
> +	.count_repeat = 14,
> +	.gcnt_us = 1,
> +	.vreg_step_fixed = 4000,
> +	.vreg_step_up_limit = 1,
> +	.vreg_step_down_limit = 1,
> +	.vdd_settle_time_us = 34,
> +	.corner_settle_time_us = 6,
> +	.reduce_to_corner_uV = true,
> +	.hw_closed_loop_en = true,
> +	.threads = (const struct cpr_thread_desc *[]) {
> +		&msm8998_thread_silver,
> +		&msm8998_thread_gold,
> +	},
> +};
> +
> +static const struct cpr_acc_desc msm8998_cpr_acc_desc = {
> +	.cpr_desc = &msm8998_cpr_desc,
> +};
> +
> +static const int sdm630_gold_scaling_factor[][CPR3_RO_COUNT] = {
> +	/* Same RO factors for all fuse corners */
> +	{
> +		4040, 3230,    0, 2210, 2560, 2450, 2230, 2220,
> +		2410, 2300, 2560, 2470, 1600, 3120, 2620, 2280
> +	}
> +};
> +
> +static const int sdm630_silver_scaling_factor[][CPR3_RO_COUNT] = {
> +	/* Same RO factors for all fuse corners */
> +	{
> +		3600, 3600, 3830, 2430, 2520, 2700, 1790, 1760,
> +		1970, 1880, 2110, 2010, 2510, 4900, 4370, 4780,
> +	}
> +};
> +
> +static const struct cpr_thread_desc sdm630_thread_gold = {
> +	.controller_id = 0,
> +	.hw_tid = 0,
> +	.ro_scaling_factor = sdm630_gold_scaling_factor,
> +	.ro_avail_corners = ARRAY_SIZE(sdm630_gold_scaling_factor),
> +	.sensor_range_start = 0,
> +	.sensor_range_end = 6,
> +	.init_voltage_step = 10000,
> +	.init_voltage_width = 6,
> +	.step_quot_init_min = 12,
> +	.step_quot_init_max = 14,
> +	.num_fuse_corners = 5,
> +	.fuse_corner_data = (struct fuse_corner_data[]){
> +		/* fuse corner 0 */
> +		{
> +			.ref_uV = 644000,
> +			.max_uV = 724000,
> +			.min_uV = 588000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 15000,
> +			.max_volt_scale = 10,
> +			.max_quot_scale = 300,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 1 */
> +		{
> +			.ref_uV = 788000,
> +			.max_uV = 788000,
> +			.min_uV = 652000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 5000,
> +			.max_volt_scale = 320,
> +			.max_quot_scale = 275,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 2 */
> +		{
> +			.ref_uV = 868000,
> +			.max_uV = 868000,
> +			.min_uV = 712000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 5000,
> +			.max_volt_scale = 350,
> +			.max_quot_scale = 800,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 3 */
> +		{
> +			.ref_uV = 988000,
> +			.max_uV = 988000,
> +			.min_uV = 784000,
> +			.range_uV = 66000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 0,
> +			.max_volt_scale = 868,
> +			.max_quot_scale = 980,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 4 */
> +		{
> +			.ref_uV = 1068000,
> +			.max_uV = 1068000,
> +			.min_uV = 844000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 0,
> +			.max_volt_scale = 868,
> +			.max_quot_scale = 980,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +	},
> +};
> +
> +static const struct cpr_thread_desc sdm630_thread_silver = {
> +	.controller_id = 1,
> +	.hw_tid = 0,
> +	.ro_scaling_factor = sdm630_silver_scaling_factor,
> +	.ro_avail_corners = ARRAY_SIZE(sdm630_silver_scaling_factor),
> +	.sensor_range_start = 0,
> +	.sensor_range_end = 6,
> +	.init_voltage_step = 10000,
> +	.init_voltage_width = 6,
> +	.step_quot_init_min = 12,
> +	.step_quot_init_max = 14,
> +	.num_fuse_corners = 3,
> +	.fuse_corner_data = (struct fuse_corner_data[]){
> +		/* fuse corner 0 */
> +		{
> +			.ref_uV = 644000,
> +			.max_uV = 724000,
> +			.min_uV = 588000,
> +			.range_uV = 32000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 0,
> +			.max_volt_scale = 10,
> +			.max_quot_scale = 360,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 1 */
> +		{
> +			.ref_uV = 788000,
> +			.max_uV = 788000,
> +			.min_uV = 652000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 0,
> +			.max_volt_scale = 500,
> +			.max_quot_scale = 550,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +		/* fuse corner 2 */
> +		{
> +			.ref_uV = 1068000,
> +			.max_uV = 1068000,
> +			.min_uV = 800000,
> +			.range_uV = 40000,
> +			.volt_cloop_adjust = -30000,
> +			.volt_oloop_adjust = 0,
> +			.max_volt_scale = 2370,
> +			.max_quot_scale = 550,
> +			.quot_offset = 0,
> +			.quot_scale = 1,
> +			.quot_adjust = 0,
> +			.quot_offset_scale = 5,
> +			.quot_offset_adjust = 0,
> +		},
> +	},
> +};
> +
> +static const struct cpr_desc sdm630_cpr_desc = {
> +	.cpr_type = CTRL_TYPE_CPRH,
> +	.num_threads = 2,
> +	.apm_threshold = 872000,
> +	.apm_crossover = 872000,
> +	.apm_hysteresis = 20000,
> +	.cpr_base_voltage = 400000,
> +	.cpr_max_voltage = 1300000,
> +	.timer_delay_us = 5000,
> +	.timer_cons_up = 0,
> +	.timer_cons_down = 2,
> +	.up_threshold = 2,
> +	.down_threshold = 2,
> +	.idle_clocks = 15,
> +	.count_mode = CPR3_CPR_CTL_COUNT_MODE_ALL_AT_ONCE_MIN,
> +	.count_repeat = 14,
> +	.gcnt_us = 1,
> +	.vreg_step_fixed = 4000,
> +	.vreg_step_up_limit = 1,
> +	.vreg_step_down_limit = 1,
> +	.vdd_settle_time_us = 34,
> +	.corner_settle_time_us = 5,
> +	.reduce_to_corner_uV = true,
> +	.hw_closed_loop_en = true,
> +	.threads = (const struct cpr_thread_desc *[]) {
> +		&sdm630_thread_gold,
> +		&sdm630_thread_silver,
> +	},
> +};
> +
> +static const struct cpr_acc_desc sdm630_cpr_acc_desc = {
> +	.cpr_desc = &sdm630_cpr_desc,
> +};
> +
> +static unsigned int cpr_get_performance_state(struct generic_pm_domain *genpd,
> +					      struct dev_pm_opp *opp)
> +{
> +	return dev_pm_opp_get_level(opp);
> +}
> +
> +static int cpr_power_off(struct generic_pm_domain *domain)
> +{
> +	struct cpr_thread *thread = container_of(domain, struct cpr_thread, pd);
> +
> +	return cpr_disable(thread);
> +}
> +
> +static int cpr_power_on(struct generic_pm_domain *domain)
> +{
> +	struct cpr_thread *thread = container_of(domain, struct cpr_thread, pd);
> +
> +	return cpr_enable(thread);
> +}
> +
> +static void cpr_pd_detach_dev(struct generic_pm_domain *domain,
> +			      struct device *dev)
> +{
> +	struct cpr_thread *thread = container_of(domain, struct cpr_thread, pd);
> +	struct cpr_drv *drv = thread->drv;
> +
> +	mutex_lock(&drv->lock);
> +
> +	dev_dbg(drv->dev, "detach callback for: %s\n", dev_name(dev));
> +	thread->attached_cpu_dev = NULL;
> +
> +	mutex_unlock(&drv->lock);
> +}
> +
> +static int cpr_pd_attach_dev(struct generic_pm_domain *domain,
> +			     struct device *dev)
> +{
> +	struct cpr_thread *thread = container_of(domain, struct cpr_thread, pd);
> +	struct cpr_drv *drv = thread->drv;
> +	const struct acc_desc *acc_desc = drv->acc_desc;
> +	bool cprh_opp_remove_table = false;
> +	int ret = 0;
> +
> +	mutex_lock(&drv->lock);
> +
> +	dev_dbg(drv->dev, "attach callback for: %s\n", dev_name(dev));
> +
> +	/*
> +	 * This driver only supports scaling voltage for a CPU cluster
> +	 * where all CPUs in the cluster share a single regulator.
> +	 * Therefore, save the struct device pointer only for the first
> +	 * CPU device that gets attached. There is no need to do any
> +	 * additional initialization when further CPUs get attached.
> +	 * This is not an error condition.
> +	 */
> +	if (thread->attached_cpu_dev)
> +		goto unlock;
> +
> +	/*
> +	 * cpr_scale_voltage() requires the direction (if we are changing
> +	 * to a higher or lower OPP). The first time
> +	 * cpr_set_performance_state() is called, there is no previous
> +	 * performance state defined. Therefore, we call
> +	 * cpr_find_initial_corner() that gets the CPU clock frequency
> +	 * set by the bootloader, so that we can determine the direction
> +	 * the first time cpr_set_performance_state() is called.
> +	 */
> +	thread->cpu_clk = devm_clk_get(dev, NULL);
> +	if (drv->desc->cpr_type < CTRL_TYPE_CPRH && IS_ERR(thread->cpu_clk)) {
> +		ret = PTR_ERR(thread->cpu_clk);
> +		if (ret != -EPROBE_DEFER)
> +			dev_err(drv->dev, "could not get cpu clk: %d\n", ret);
> +		goto unlock;
> +	}
> +	thread->attached_cpu_dev = dev;
> +
> +	/*
> +	 * We are exporting the APM and MEM-ACC thresholds to the caller;
> +	 * while APM is necessary in the CPU CPR case, MEM-ACC may not be,
> +	 * depending on the SoC and on fuses.
> +	 * Initialize both to an invalid value, so that the caller can check
> +	 * if they got calculated or read from fuses in this driver.
> +	 */
> +	thread->ext_data.apm_threshold_uV = -1;
> +	thread->ext_data.mem_acc_threshold_uV = -1;
> +	dev_set_drvdata(thread->attached_cpu_dev, &thread->ext_data);
> +
> +	dev_dbg(drv->dev, "using cpu clk from: %s\n",
> +		dev_name(thread->attached_cpu_dev));
> +
> +	/*
> +	 * Everything related to (virtual) corners has to be initialized
> +	 * here, when attaching to the power domain, since we need to know
> +	 * the maximum frequency for each fuse corner, and this is only
> +	 * available after the cpufreq driver has attached to us.
> +	 * The reason for this is that we need to know the highest
> +	 * frequency associated with each fuse corner.
> +	 */
> +	ret = dev_pm_opp_get_opp_count(&thread->pd.dev);
> +	if (ret < 0) {
> +		dev_err(drv->dev, "could not get OPP count\n");
> +		thread->attached_cpu_dev = NULL;
> +		goto unlock;
> +	}
> +	thread->num_corners = ret;
> +
> +	thread->corners = devm_kcalloc(drv->dev,
> +				       thread->num_corners +
> +				       drv->extra_corners,
> +				       sizeof(*thread->corners),
> +				       GFP_KERNEL);
> +	if (!thread->corners) {
> +		ret = -ENOMEM;
> +		goto unlock;
> +	}
> +
> +	/*
> +	 * If we are on CPR-Hardened we have to make sure that the attached
> +	 * device has a OPP table installed, as we're going to modify it here
> +	 * with our calculations based on qfprom values.
> +	 */
> +	if (drv->desc->cpr_type == CTRL_TYPE_CPRH) {
> +		ret = dev_pm_opp_of_add_table(dev);
> +		if (ret && ret != -EEXIST) {
> +			dev_err(drv->dev, "Cannot add table: %d\n", ret);
> +			goto unlock;
> +		}
> +		cprh_opp_remove_table = true;
> +	}
> +
> +	ret = cpr3_corner_init(thread);
> +	if (ret)
> +		goto exit;
> +
> +	if (drv->desc->cpr_type < CTRL_TYPE_CPRH) {
> +		ret = cpr3_find_initial_corner(thread);
> +		if (ret)
> +			goto exit;
> +
> +		if (acc_desc->config)
> +			regmap_multi_reg_write(drv->tcsr, acc_desc->config,
> +					       acc_desc->num_regs_per_fuse);
> +
> +		/* Enable ACC if required */
> +		if (acc_desc->enable_mask)
> +			regmap_update_bits(drv->tcsr, acc_desc->enable_reg,
> +					   acc_desc->enable_mask,
> +					   acc_desc->enable_mask);
> +	}
> +	dev_info(drv->dev, "thread %d initialized with %u OPPs\n",
> +		 thread->id, thread->num_corners);
> +exit:
> +	/*
> +	 * If we are on CPRh and we reached an error condition, we installed
> +	 * the OPP table but we haven't done any setup on it, nor we ever will.
> +	 * In order to leave a clean state, remove the table.
> +	 */
> +	if (ret && cprh_opp_remove_table)
> +		dev_pm_opp_of_remove_table(thread->attached_cpu_dev);
> +unlock:
> +	mutex_unlock(&drv->lock);
> +
> +	return ret;
> +}
> +
> +static int cpr3_debug_info_show(struct seq_file *s, void *unused)
> +{
> +	u32 ro_sel, ctl, irq_status, reg, quot;
> +	struct cpr_thread *thread = s->private;
> +	struct corner *corner = thread->corners;
> +	struct fuse_corner *fuse = thread->fuse_corners;
> +	unsigned int i;
> +
> +	const struct {
> +		const char *name;
> +		uint32_t mask;
> +		uint8_t shift;
> +	} result0_fields[] = {
> +		{ "busy", 1, 0 },
> +		{ "step_dn", 1, 1 },
> +		{ "step_up", 1, 2 },
> +		{ "error_steps", CPR3_RESULT0_ERROR_STEPS_MASK,
> +				 CPR3_RESULT0_ERROR_STEPS_SHIFT },
> +		{ "error", CPR3_RESULT0_ERROR_MASK, CPR3_RESULT0_ERROR_SHIFT },
> +		{ "negative", 1, 20 },
> +	}, result1_fields[] = {
> +		{ "quot_min", CPR3_RESULT1_QUOT_MIN_MASK,
> +			      CPR3_RESULT1_QUOT_MIN_SHIFT },
> +		{ "quot_max", CPR3_RESULT1_QUOT_MAX_MASK,
> +			      CPR3_RESULT1_QUOT_MAX_SHIFT },
> +		{ "ro_min", CPR3_RESULT1_RO_MIN_MASK,
> +			    CPR3_RESULT1_RO_MIN_SHIFT },
> +		{ "ro_max", CPR3_RESULT1_RO_MAX_MASK,
> +			    CPR3_RESULT1_RO_MAX_SHIFT },
> +	}, result2_fields[] = {
> +		{ "qout_step_min", CPR3_RESULT2_STEP_QUOT_MIN_MASK,
> +				   CPR3_RESULT2_STEP_QUOT_MIN_SHIFT },
> +		{ "qout_step_max", CPR3_RESULT2_STEP_QUOT_MAX_MASK,
> +				   CPR3_RESULT2_STEP_QUOT_MAX_SHIFT },
> +		{ "sensor_min", CPR3_RESULT2_SENSOR_MIN_MASK,
> +				CPR3_RESULT2_SENSOR_MIN_SHIFT },
> +		{ "sensor_max", CPR3_RESULT2_SENSOR_MAX_MASK,
> +				CPR3_RESULT2_SENSOR_MAX_SHIFT },
> +	};
> +
> +	if (thread->drv->desc->cpr_type < CTRL_TYPE_CPRH)
> +		seq_printf(s, "current_volt = %d uV\n", thread->drv->last_uV);
> +
> +	irq_status = cpr_read(thread, CPR3_REG_IRQ_STATUS);
> +	seq_printf(s, "irq_status = %#02X\n", irq_status);
> +
> +	ctl = cpr_read(thread, CPR3_REG_CPR_CTL);
> +	seq_printf(s, "cpr_ctl = %#02X\n", ctl);
> +
> +	seq_printf(s, "thread %d - hw tid: %u - enabled: %d:\n",
> +		   thread->id, thread->desc->hw_tid, thread->enabled);
> +	seq_printf(s, "%d corners, derived from %d fuse corners\n",
> +		   thread->num_corners, thread->desc->num_fuse_corners);
> +
> +	for (i = 0; i < thread->num_corners; i++, corner++)
> +		seq_printf(s, "corner %d - uV=[%d %d %d] quot=%d freq=%lu\n",
> +			   i, corner->min_uV, corner->uV, corner->max_uV,
> +			   corner->quot_adjust, corner->freq);
> +
> +	for (i = 0; i < thread->desc->num_fuse_corners; i++, fuse++)
> +		seq_printf(s, "fuse %d - uV=[%d %d %d] quot=%d freq=%lu\n",
> +			   i, fuse->min_uV, fuse->uV, fuse->max_uV,
> +			   fuse->quot, corner->freq);
> +
> +	seq_printf(s, "requested voltage: %d uV\n", thread->corner->last_uV);
> +
> +	ro_sel = corner->fuse_corner->ring_osc_idx;
> +	quot = cpr_read(thread, CPR3_REG_TARGET_QUOT(i, ro_sel));
> +	seq_printf(s, "quot_target (%u) = %#02X\n", ro_sel, quot);
> +
> +	reg = cpr_read(thread, CPR3_REG_RESULT0(i));
> +	seq_printf(s, "cpr_result_0 = %#02X\n  [", reg);
> +	for (i = 0; i < ARRAY_SIZE(result0_fields); i++)
> +		seq_printf(s, "%s%s = %u",
> +			   i ? ", " : "",
> +			   result0_fields[i].name,
> +			   (reg >> result0_fields[i].shift) &
> +				result0_fields[i].mask);
> +	seq_puts(s, "]\n");
> +	reg = cpr_read(thread, CPR3_REG_RESULT1(i));
> +	seq_printf(s, "cpr_result_1 = %#02X\n  [", reg);
> +	for (i = 0; i < ARRAY_SIZE(result1_fields); i++)
> +		seq_printf(s, "%s%s = %u",
> +			   i ? ", " : "",
> +			   result1_fields[i].name,
> +			   (reg >> result1_fields[i].shift) &
> +				result1_fields[i].mask);
> +	seq_puts(s, "]\n");
> +	reg = cpr_read(thread, CPR3_REG_RESULT2(i));
> +	seq_printf(s, "cpr_result_2 = %#02X\n  [", reg);
> +	for (i = 0; i < ARRAY_SIZE(result2_fields); i++)
> +		seq_printf(s, "%s%s = %u",
> +			   i ? ", " : "",
> +			   result2_fields[i].name,
> +			   (reg >> result2_fields[i].shift) &
> +				result2_fields[i].mask);
> +	seq_puts(s, "]\n");
> +
> +	return 0;
> +}
> +DEFINE_SHOW_ATTRIBUTE(cpr3_debug_info);
> +
> +static void cpr3_debugfs_init(struct cpr_drv *drv)
> +{
> +	int i;
> +
> +	drv->debugfs = debugfs_create_dir("qcom_cpr3", NULL);
> +
> +	for (i = 0; i < drv->desc->num_threads; i++) {
> +		char buf[50];
> +
> +		snprintf(buf, sizeof(buf), "thread%d", i);
> +
> +		debugfs_create_file(buf, 0444, drv->debugfs, &drv->threads[i],
> +				    &cpr3_debug_info_fops);
> +	}
> +}
> +
> +/**
> + * cpr_thread_init() - Initialize CPR thread related parameters
> + * @drv: Main driver structure
> + * @tid: Thread ID
> + *
> + * Return: Zero for success, negative number on error
> + */
> +static int cpr_thread_init(struct cpr_drv *drv, int tid)
> +{
> +	const struct cpr_desc *desc = drv->desc;
> +	const struct cpr_thread_desc *tdesc = desc->threads[tid];
> +	struct cpr_thread *thread = &drv->threads[tid];
> +	int ret;
> +
> +	if (tdesc->step_quot_init_min > CPR3_CPR_STEP_QUOT_MIN_MASK ||
> +	    tdesc->step_quot_init_max > CPR3_CPR_STEP_QUOT_MAX_MASK)
> +		return -EINVAL;
> +
> +	thread->id = tid;
> +	thread->drv = drv;
> +	thread->desc = tdesc;
> +	thread->fuse_corners = devm_kcalloc(drv->dev,
> +					    tdesc->num_fuse_corners +
> +					    drv->extra_corners,
> +					    sizeof(*thread->fuse_corners),
> +					    GFP_KERNEL);
> +	if (!thread->fuse_corners)
> +		return -ENOMEM;
> +
> +	thread->cpr_fuses = cpr_get_fuses(drv->dev, tid,
> +					  tdesc->num_fuse_corners);
> +	if (IS_ERR(thread->cpr_fuses))
> +		return PTR_ERR(thread->cpr_fuses);
> +
> +	ret = cpr_populate_ring_osc_idx(thread->drv->dev, thread->fuse_corners,
> +					thread->cpr_fuses,
> +					tdesc->num_fuse_corners);
> +	if (ret)
> +		return ret;
> +
> +	ret = cpr_fuse_corner_init(thread);
> +	if (ret)
> +		return ret;
> +
> +	thread->pd.name = devm_kasprintf(drv->dev, GFP_KERNEL,
> +					 "%s_thread%d",
> +					 drv->dev->of_node->full_name,
> +					 thread->id);
> +	if (!thread->pd.name)
> +		return -EINVAL;
> +
> +	thread->pd.power_off = cpr_power_off;
> +	thread->pd.power_on = cpr_power_on;
> +	thread->pd.opp_to_performance_state = cpr_get_performance_state;
> +	thread->pd.attach_dev = cpr_pd_attach_dev;
> +	thread->pd.detach_dev = cpr_pd_detach_dev;
> +
> +	/* CPR-Hardened performance states are managed in firmware */
> +	if (desc->cpr_type == CTRL_TYPE_CPRH)
> +		thread->pd.set_performance_state = cprh_dummy_set_performance_state;
> +	else
> +		thread->pd.set_performance_state = cpr_set_performance_state;
> +
> +	/* Anything later than CPR1 must be always-on for now */
> +	thread->pd.flags = GENPD_FLAG_ALWAYS_ON;
> +
> +	drv->cell_data.domains[tid] = &thread->pd;
> +
> +	ret = pm_genpd_init(&thread->pd, NULL, false);
> +	if (ret)
> +		return ret;
> +
> +	/* On CPRhardened, the interrupts are managed in firmware */
> +	if (desc->cpr_type < CTRL_TYPE_CPRH) {
> +		INIT_WORK(&thread->restart_work, cpr_restart_worker);
> +
> +		ret = devm_request_threaded_irq(drv->dev, drv->irq,
> +						NULL, cpr_irq_handler,
> +						IRQF_ONESHOT |
> +						IRQF_TRIGGER_RISING,
> +						"cpr", drv);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	return 0;
> +}
> +
> +/**
> + * cpr3_resources_init() - Initialize resources used by this driver
> + * @pdev: Platform device
> + * @drv:  Main driver structure
> + *
> + * Return: Zero for success, negative number on error
> + */
> +static int cpr3_resources_init(struct platform_device *pdev,
> +			       struct cpr_drv *drv)
> +{
> +	const struct cpr_desc *desc = drv->desc;
> +	struct cpr_thread *threads = drv->threads;
> +	unsigned int i;
> +	u8 cid_mask = 0;
> +
> +	/*
> +	 * Here, we are accounting for the following usecases:
> +	 * - One controller
> +	 *   - One or multiple threads on the same iospace
> +	 *
> +	 * - Multiple controllers
> +	 *   - Each controller has its own iospace and each
> +	 *     may have one or multiple threads in their
> +	 *     parent controller's iospace
> +	 *
> +	 * Then, to avoid complicating the code for no reason,
> +	 * this also needs a mandatory order in the list of
> +	 * threads which implies that all of them from the same
> +	 * controllers are specified sequentially. As an example:
> +	 *
> +	 *      C0-T0, C0-T1...C0-Tn, C1-T0, C1-T1...C1-Tn
> +	 */
> +	for (i = 0; i < desc->num_threads; i++) {
> +		u8 cid = desc->threads[i]->controller_id;
> +
> +		if (cid_mask & BIT(cid)) {
> +			if (desc->threads[i - 1]->controller_id != cid) {
> +				dev_err(drv->dev, "Bad threads order. Please fix!\n");
> +				return -EINVAL;
> +			}
> +			threads[i].base = threads[i - 1].base;
> +			continue;
> +		}
> +		threads[i].base = devm_platform_ioremap_resource(pdev, cid);
> +		if (IS_ERR(threads[i].base))
> +			return PTR_ERR(threads[i].base);
> +		cid_mask |= BIT(cid);
> +	}
> +	return 0;
> +}
> +
> +static int cpr_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct cpr_drv *drv;
> +	const struct cpr_desc *desc;
> +	const struct cpr_acc_desc *data;
> +	struct device_node *np;
> +	unsigned int i;
> +	int ret;
> +
> +	data = of_device_get_match_data(dev);
> +	if (!data || !data->cpr_desc)
> +		return -EINVAL;
> +
> +	desc = data->cpr_desc;
> +
> +	/* CPRh disallows MEM-ACC access from the HLOS */
> +	if (!data->acc_desc && desc->cpr_type < CTRL_TYPE_CPRH)
> +		return -EINVAL;
> +
> +	drv = devm_kzalloc(dev, sizeof(*drv), GFP_KERNEL);
> +	if (!drv)
> +		return -ENOMEM;
> +
> +	drv->dev = dev;
> +	drv->desc = desc;
> +	drv->threads = devm_kcalloc(dev, desc->num_threads,
> +				    sizeof(*drv->threads), GFP_KERNEL);
> +	if (!drv->threads)
> +		return -ENOMEM;
> +
> +	drv->cell_data.num_domains = desc->num_threads;
> +	drv->cell_data.domains = devm_kcalloc(drv->dev,
> +					      drv->cell_data.num_domains,
> +					      sizeof(*drv->cell_data.domains),
> +					      GFP_KERNEL);
> +	if (!drv->cell_data.domains)
> +		return -ENOMEM;
> +
> +	if (data->acc_desc)
> +		drv->acc_desc = data->acc_desc;
> +
> +	mutex_init(&drv->lock);
> +
> +	if (desc->cpr_type < CTRL_TYPE_CPRH) {
> +		np = of_parse_phandle(dev->of_node, "acc-syscon", 0);
> +		if (!np)
> +			return -ENODEV;
> +
> +		drv->tcsr = syscon_node_to_regmap(np);
> +		of_node_put(np);
> +		if (IS_ERR(drv->tcsr))
> +			return PTR_ERR(drv->tcsr);
> +	}
> +
> +	ret = cpr3_resources_init(pdev, drv);
> +	if (ret)
> +		return ret;
> +
> +	drv->irq = platform_get_irq_optional(pdev, 0);
> +	if (desc->cpr_type != CTRL_TYPE_CPRH && drv->irq < 0)
> +		return -EINVAL;
> +
> +	/* On CPRhardened, vreg access it not allowed */
> +	drv->vreg = devm_regulator_get_optional(dev, "vdd");
> +	if (desc->cpr_type != CTRL_TYPE_CPRH && IS_ERR(drv->vreg))
> +		return PTR_ERR(drv->vreg);
> +
> +	/*
> +	 * On at least CPRhardened, vreg is unaccessible and there is no
> +	 * way to read linear step from that regulator, hence it is hardcoded
> +	 * in the driver;
> +	 * When the vreg_step is not declared in the cpr data (or is zero),
> +	 * then having access to the vreg regulator is mandatory, as this
> +	 * will be retrieved through the regulator API.
> +	 */
> +	if (desc->vreg_step_fixed)
> +		drv->vreg_step = desc->vreg_step_fixed;
> +	else
> +		drv->vreg_step = regulator_get_linear_step(drv->vreg);
> +
> +	if (!drv->vreg_step)
> +		return -EINVAL;
> +
> +	/*
> +	 * Initialize fuse corners, since it simply depends
> +	 * on data in efuses.
> +	 * Everything related to (virtual) corners has to be
> +	 * initialized after attaching to the power domain,
> +	 * since it depends on the CPU's OPP table.
> +	 */
> +	ret = nvmem_cell_read_variable_le_u32(dev, "cpr_fuse_revision", &drv->fusing_rev);
> +	if (ret)
> +		return ret;
> +
> +	ret = nvmem_cell_read_variable_le_u32(dev, "cpr_speed_bin", &drv->speed_bin);
> +	if (ret)
> +		return ret;
> +
> +	/*
> +	 * Some SoCs require extra corners for MEM-ACC or APM: if
> +	 * the related parameters have been specified, then reserve
> +	 * a corner for the APM and/or MEM-ACC crossover, used by
> +	 * OSM and CPRh HW to set the supply voltage during the APM
> +	 * and/or MEM-ACC switch routine.
> +	 */
> +	if (desc->cpr_type == CTRL_TYPE_CPRH) {
> +		if (desc->apm_crossover && desc->apm_hysteresis >= 0)
> +			drv->extra_corners++;
> +
> +		if (desc->mem_acc_threshold)
> +			drv->extra_corners++;
> +	}
> +
> +	/* Initialize all threads */
> +	for (i = 0; i < desc->num_threads; i++) {
> +		ret = cpr_thread_init(drv, i);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	/* Initialize global parameters */
> +	ret = cpr3_init_parameters(drv);
> +	if (ret)
> +		return ret;
> +
> +	/* Write initial configuration on all threads */
> +	for (i = 0; i < desc->num_threads; i++) {
> +		ret = cpr_configure(&drv->threads[i]);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	ret = of_genpd_add_provider_onecell(dev->of_node, &drv->cell_data);
> +	if (ret)
> +		return ret;
> +
> +	platform_set_drvdata(pdev, drv);
> +	cpr3_debugfs_init(drv);
> +
> +	return 0;
> +}
> +
> +static int cpr_remove(struct platform_device *pdev)
> +{
> +	struct cpr_drv *drv = platform_get_drvdata(pdev);
> +	int i;
> +
> +	of_genpd_del_provider(pdev->dev.of_node);
> +
> +	for (i = 0; i < drv->desc->num_threads; i++) {
> +		cpr_ctl_disable(&drv->threads[i]);
> +		cpr_irq_set(&drv->threads[i], 0);
> +		pm_genpd_remove(&drv->threads[i].pd);
> +	}
> +
> +	debugfs_remove_recursive(drv->debugfs);
> +
> +	return 0;
> +}
> +
> +static const struct of_device_id cpr3_match_table[] = {
> +	{ .compatible = "qcom,msm8998-cprh", .data = &msm8998_cpr_acc_desc },
> +	{ .compatible = "qcom,sdm630-cprh", .data = &sdm630_cpr_acc_desc },
> +	{ }
> +};
> +MODULE_DEVICE_TABLE(of, cpr3_match_table);
> +
> +static struct platform_driver cpr3_driver = {
> +	.probe		= cpr_probe,
> +	.remove		= cpr_remove,
> +	.driver		= {
> +		.name	= "qcom-cpr3",
> +		.of_match_table = cpr3_match_table,
> +	},
> +};
> +module_platform_driver(cpr3_driver)
> +
> +MODULE_DESCRIPTION("Core Power Reduction (CPR) v3/v4 driver");
> +MODULE_LICENSE("GPL");
> diff --git a/include/soc/qcom/cpr.h b/include/soc/qcom/cpr.h
> new file mode 100644
> index 000000000000..2ba4324d18f6
> --- /dev/null
> +++ b/include/soc/qcom/cpr.h
> @@ -0,0 +1,17 @@
> +/* SPDX-License-Identifier: GPL-2.0-only */
> +/*
> + * Copyright (c) 2013-2020, The Linux Foundation. All rights reserved.
> + * Copyright (c) 2019 Linaro Limited
> + * Copyright (c) 2021, AngeloGioacchino Del Regno
> + *                     <angelogioacchino.delregno@...ainline.org>
> + */
> +
> +#ifndef __CPR_H__
> +#define __CPR_H__
> +
> +struct cpr_ext_data {
> +	int mem_acc_threshold_uV;
> +	int apm_threshold_uV;
> +};

Who is going to use this? Is it the cpufreq driver or some other driver?
We are adding an API without a clean user, can we drop it for now?

> +
> +#endif /* __CPR_H__ */
> 

-- 
With best wishes
Dmitry

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ