2014-07-10 21:43:20

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 00/13] Tegra124 CL-DVFS / DFLL clocksource, plus cpufreq

This series implements the DFLL/CL-DVFS clock source for the fast CPU
cluster on Tegra124, and a cpufreq driver that uses the DFLL for
clocking the CPU. Most of this is based on Paul Walmsley's public patch
set from December 2013, which is available at
http://comments.gmane.org/gmane.linux.ports.tegra/15273

The DFLL clock hardware is a voltage-controlled oscillator plus
control logic that compares the generated output clock with a
51 MHz reference clock, and can make decisions to either lower
or raise the DFLL voltage to keep the output rate close to the
software-requested rate. The voltage changes are done by
communicating with an off-chip PMIC via either I2C or PWM.
As the DFLL oscillator is powered via the CPU rail, using
the DFLL as the CPU clocksource also gives us dynamic CPU
voltage scaling.

This series has been tested on the Jetson TK1 (Rev C). Before attempting
to port this to the Venice2, do note that there are two versions of the
AS3722 with different voltage tables for the CPU rail (and that Venice2
does not have active cooling).

Thanks,
Tuomas

Paul Walmsley (1):
clk: tegra: Add DFLL DVCO reset control for Tegra124

Tuomas Tynkkynen (12):
clk: tegra: Add binding for the Tegra124 DFLL clocksource
clk: tegra: Add library for the DFLL clock source (open-loop mode)
clk: tegra: Add closed loop support for the DFLL
clk: tegra: Add functions for parsing CVB tables
clk: tegra: Add Tegra124 DFLL clocksource platform driver
clk: tegra: Save/restore CCLKG_BURST_POLICY on suspend
clk: tegra: Add the DFLL as a possible parent of the cclk_g clock
ARM: tegra: Add the DFLL to Tegra124 device tree
ARM: tegra: Enable the DFLL on the Jetson-TK1
cpufreq: tegra124: Add device tree bindings
cpufreq: Add cpufreq driver for Tegra124
ARM: tegra: Add entries for cpufreq on Tegra124

.../bindings/clock/nvidia,tegra124-dfll.txt | 86 +
.../bindings/cpufreq/tegra124-cpufreq.txt | 37 +
arch/arm/boot/dts/tegra124-jetson-tk1.dts | 83 +-
arch/arm/boot/dts/tegra124.dtsi | 29 +
arch/arm/mach-tegra/Kconfig | 1 +
drivers/clk/tegra/Makefile | 3 +
drivers/clk/tegra/clk-dfll.c | 1759 ++++++++++++++++++++
drivers/clk/tegra/clk-dfll.h | 55 +
drivers/clk/tegra/clk-tegra-super-gen4.c | 4 +-
drivers/clk/tegra/clk-tegra124-dfll-fcpu.c | 156 ++
drivers/clk/tegra/clk-tegra124.c | 61 +
drivers/clk/tegra/clk.h | 3 +
drivers/clk/tegra/cvb.c | 133 ++
drivers/clk/tegra/cvb.h | 67 +
drivers/cpufreq/Makefile | 1 +
drivers/cpufreq/tegra124-cpufreq.c | 221 +++
16 files changed, 2697 insertions(+), 2 deletions(-)
create mode 100644 Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
create mode 100644 Documentation/devicetree/bindings/cpufreq/tegra124-cpufreq.txt
create mode 100644 drivers/clk/tegra/clk-dfll.c
create mode 100644 drivers/clk/tegra/clk-dfll.h
create mode 100644 drivers/clk/tegra/clk-tegra124-dfll-fcpu.c
create mode 100644 drivers/clk/tegra/cvb.c
create mode 100644 drivers/clk/tegra/cvb.h
create mode 100644 drivers/cpufreq/tegra124-cpufreq.c

--
1.8.1.5


2014-07-10 21:43:24

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

The DFLL is the main clocksource for the fast CPU cluster on Tegra124
and also provides automatic CPU rail voltage scaling as well. The DFLL
is a separate IP block from the usual Tegra124 clock-and-reset
controller, so it gets its own node in the device tree.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
.../bindings/clock/nvidia,tegra124-dfll.txt | 86 ++++++++++++++++++++++
1 file changed, 86 insertions(+)
create mode 100644 Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt

diff --git a/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
new file mode 100644
index 0000000..cf89802
--- /dev/null
+++ b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
@@ -0,0 +1,86 @@
+NVIDIA Tegra124 DFLL FCPU clocksource
+
+This binding uses the common clock binding:
+Documentation/devicetree/bindings/clock/clock-bindings.txt
+
+The DFLL IP block on Tegra is a root clocksource designed for clocking
+the fast CPU cluster. It consists of a free-running voltage controlled
+oscillator connected to the CPU voltage rail (VDD_CPU), and a closed loop
+control module that will automatically adjust the VDD_CPU voltage by
+communicating with an off-chip PMIC either via an I2C bus or via PWM signals.
+
+Required properties:
+- compatible : should be "nvidia,tegra124-dfll-fcpu"
+- reg : Defines the following set of registers, in the order listed:
+ - registers for the DFLL control logic.
+ - registers for the I2C output logic.
+ - registers for the integrated I2C master controller.
+ - look-up table RAM for voltage register values.
+- interrupts: Should contain the DFLL block interrupt.
+- clocks: Must contain an entry for each entry in clock-names.
+ See ../clocks/clock-bindings.txt for details.
+- clock-names: Must include the following entries:
+ - soc: Clock source for the DFLL control logic.
+ - ref: The closed loop reference clock
+ - i2c: Clock source for the integrated I2C master.
+- #clock-cells: Must be 0.
+- clock-output-names: Name of the clock output.
+- vdd_cpu-supply: Regulator for the CPU voltage rail that the DFLL
+ hardware will start controlling.
+
+Required properties for the control loop parameters:
+- nvidia,sample-rate: Sample rate of the DFLL control loop.
+- nvidia,droop-ctrl: See the register CL_DVFS_DROOP_CTRL in the TRM.
+- nvidia,force-mode: See the field DFLL_PARAMS_FORCE_MODE in the TRM.
+- nvidia,cf: Numeric value, see the field DFLL_PARAMS_CF_PARAM in the TRM.
+- nvidia,ci: Numeric value, see the field DFLL_PARAMS_CI_PARAM in the TRM.
+- nvidia,cg: Numeric value, see the field DFLL_PARAMS_CG_PARAM in the TRM.
+- nvidia,cg-scale: Boolean value, see the field DFLL_PARAMS_CG_SCALE in the TRM.
+
+Required properties for I2C mode:
+- nvidia,pmic-i2c-address: I2C address of the PMIC that controls the VDD_CPU
+ voltage.
+- nvidia,i2c-10-bit-addresses: Boolean, whether to use 10-bit I2C addressing.
+- nvidia,pmic-i2c-voltage-register: Register of the PMIC that controls the
+ VDD_CPU voltage.
+- nvidia,i2c-fs-rate: I2C transfer rate, if using FS mode.
+- nvidia,pmic-voltage-table: Array of 2-tuples. Each entry should have the
+ form <register-value voltage-in-uV>, indicating the register value that
+ needs to be programmed to the PMIC for changing the VDD_CPU voltage to
+ the specified voltage. The table must be in ascending order by the voltage.
+
+Example:
+
+dfll@0,70110000 {
+ compatible = "nvidia,tegra124-dfll";
+ reg = <0 0x70110000 0 0x100>, /* DFLL control */
+ <0 0x70110000 0 0x100>, /* I2C output control */
+ <0 0x70110100 0 0x100>, /* Integrated I2C controller */
+ <0 0x70110200 0 0x100>; /* Look-up table RAM */
+ interrupts = <GIC_SPI 62 IRQ_TYPE_LEVEL_HIGH>;
+ clocks = <&tegra_car TEGRA124_CLK_DFLL_SOC>,
+ <&tegra_car TEGRA124_CLK_DFLL_REF>,
+ <&tegra_car TEGRA124_CLK_I2C5>;
+ clock-names = "soc", "ref", "i2c";
+ #clock-cells = <0>;
+ clock-output-names = "dfllCPU_out";
+ vdd_cpu-supply = <&vdd_cpu>;
+ status = "okay";
+
+ nvidia,sample-rate = <12500>;
+ nvidia,droop-ctrl = <0x00000f00>;
+ nvidia,force-mode = <1>;
+ nvidia,cf = <10>;
+ nvidia,ci = <0>;
+ nvidia,cg = <2>;
+
+ nvidia,i2c-fs-rate = <400000>;
+ nvidia,pmic-i2c-address = <0x40>;
+ nvidia,pmic-i2c-voltage-register = <0x00>;
+ nvidia,pmic-voltage-table =
+ <0x1e 700000>,
+ <0x1f 710000>,
+ /* etc... */
+ <0x63 1390000>,
+ <0x64 1400000>;
+};
--
1.8.1.5

2014-07-10 21:43:29

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 02/13] clk: tegra: Add library for the DFLL clock source (open-loop mode)

Add shared code to support the Tegra DFLL clocksource in open-loop
mode. This root clocksource is present on the Tegra124 SoCs. The
DFLL is the intended primary clock source for the fast CPU cluster.

This code is very closely based on a patch by Paul Walmsley from
December (http://comments.gmane.org/gmane.linux.ports.tegra/15273),
which in turn comes from the internal driver by originally created
by Aleksandr Frid <[email protected]>.

Subsequent patches will add support for closed loop mode and drivers
for the Tegra124 fast CPU cluster DFLL devices, which rely on this
code.

Signed-off-by: Paul Walmsley <[email protected]>
Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
drivers/clk/tegra/Makefile | 1 +
drivers/clk/tegra/clk-dfll.c | 1074 ++++++++++++++++++++++++++++++++++++++++++
drivers/clk/tegra/clk-dfll.h | 55 +++
3 files changed, 1130 insertions(+)
create mode 100644 drivers/clk/tegra/clk-dfll.c
create mode 100644 drivers/clk/tegra/clk-dfll.h

diff --git a/drivers/clk/tegra/Makefile b/drivers/clk/tegra/Makefile
index f7dfb72..47320ca 100644
--- a/drivers/clk/tegra/Makefile
+++ b/drivers/clk/tegra/Makefile
@@ -1,5 +1,6 @@
obj-y += clk.o
obj-y += clk-audio-sync.o
+obj-y += clk-dfll.o
obj-y += clk-divider.o
obj-y += clk-periph.o
obj-y += clk-periph-gate.o
diff --git a/drivers/clk/tegra/clk-dfll.c b/drivers/clk/tegra/clk-dfll.c
new file mode 100644
index 0000000..5eca4ff
--- /dev/null
+++ b/drivers/clk/tegra/clk-dfll.c
@@ -0,0 +1,1074 @@
+/*
+ * clk-dfll.c - Tegra DFLL clock source common code
+ *
+ * Copyright (C) 2012-2014 NVIDIA Corporation. All rights reserved.
+ *
+ * Aleksandr Frid <[email protected]>
+ * Paul Walmsley <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * This library is for the DVCO and DFLL IP blocks on the Tegra124
+ * SoC. These IP blocks together are also known at NVIDIA as
+ * "CL-DVFS". To try to avoid confusion, this code refers to them
+ * collectively as the "DFLL."
+ *
+ * The DFLL is a root clocksource which tolerates some amount of
+ * supply voltage noise. Tegra124 uses it to clock the fast CPU
+ * complex when the target CPU speed is above a particular rate. The
+ * DFLL can be operated in either open-loop mode or closed-loop mode.
+ * In open-loop mode, the DFLL generates an output clock appropriate
+ * to the supply voltage. In closed-loop mode, when configured with a
+ * target frequency, the DFLL minimizes supply voltage while
+ * delivering an average frequency equal to the target.
+ *
+ * Devices clocked by the DFLL must be able to tolerate frequency
+ * variation. In the case of the CPU, it's important to note that the
+ * CPU cycle time will vary. This has implications for
+ * performance-measurement code and any code that relies on the CPU
+ * cycle time to delay for a certain length of time.
+ *
+ */
+
+#include <linux/clk.h>
+#include <linux/clk-provider.h>
+#include <linux/debugfs.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/pm_opp.h>
+#include <linux/pm_runtime.h>
+#include <linux/seq_file.h>
+
+#include "clk-dfll.h"
+
+/*
+ * DFLL control registers - access via dfll_{readl,writel}
+ */
+
+/* DFLL_CTRL: DFLL control register */
+#define DFLL_CTRL 0x00
+#define DFLL_CTRL_MODE_MASK 0x03
+
+/* DFLL_CONFIG: DFLL sample rate control */
+#define DFLL_CONFIG 0x04
+#define DFLL_CONFIG_DIV_MASK 0xff
+#define DFLL_CONFIG_DIV_PRESCALE 32
+
+/* DFLL_PARAMS: tuning coefficients for closed loop integrator */
+#define DFLL_PARAMS 0x08
+#define DFLL_PARAMS_CG_SCALE (0x1 << 24)
+#define DFLL_PARAMS_FORCE_MODE_SHIFT 22
+#define DFLL_PARAMS_FORCE_MODE_MASK (0x3 << DFLL_PARAMS_FORCE_MODE_SHIFT)
+#define DFLL_PARAMS_CF_PARAM_SHIFT 16
+#define DFLL_PARAMS_CF_PARAM_MASK (0x3f << DFLL_PARAMS_CF_PARAM_SHIFT)
+#define DFLL_PARAMS_CI_PARAM_SHIFT 8
+#define DFLL_PARAMS_CI_PARAM_MASK (0x7 << DFLL_PARAMS_CI_PARAM_SHIFT)
+#define DFLL_PARAMS_CG_PARAM_SHIFT 0
+#define DFLL_PARAMS_CG_PARAM_MASK (0xff << DFLL_PARAMS_CG_PARAM_SHIFT)
+
+/* DFLL_TUNE0: delay line configuration register 0 */
+#define DFLL_TUNE0 0x0c
+
+/* DFLL_TUNE1: delay line configuration register 1 */
+#define DFLL_TUNE1 0x10
+
+/* DFLL_FREQ_REQ: target DFLL frequency control */
+#define DFLL_FREQ_REQ 0x14
+#define DFLL_FREQ_REQ_FORCE_ENABLE (0x1 << 28)
+#define DFLL_FREQ_REQ_FORCE_SHIFT 16
+#define DFLL_FREQ_REQ_FORCE_MASK (0xfff << DFLL_FREQ_REQ_FORCE_SHIFT)
+#define FORCE_MAX 2047
+#define FORCE_MIN -2048
+#define DFLL_FREQ_REQ_SCALE_SHIFT 8
+#define DFLL_FREQ_REQ_SCALE_MASK (0xff << DFLL_FREQ_REQ_SCALE_SHIFT)
+#define DFLL_FREQ_REQ_SCALE_MAX 256
+#define DFLL_FREQ_REQ_FREQ_VALID (0x1 << 7)
+#define DFLL_FREQ_REQ_MULT_SHIFT 0
+#define DFLL_FREQ_REG_MULT_MASK (0x7f << DFLL_FREQ_REQ_MULT_SHIFT)
+#define FREQ_MAX 127
+
+/* DFLL_DROOP_CTRL: droop prevention control */
+#define DFLL_DROOP_CTRL 0x1c
+
+/* DFLL_OUTPUT_CFG: closed loop mode control registers */
+#define DFLL_OUTPUT_CFG 0x20
+#define DFLL_OUTPUT_CFG_I2C_ENABLE (0x1 << 30)
+#define OUT_MASK 0x3f
+#define DFLL_OUTPUT_CFG_SAFE_SHIFT 24
+#define DFLL_OUTPUT_CFG_SAFE_MASK \
+ (OUT_MASK << DFLL_OUTPUT_CFG_SAFE_SHIFT)
+#define DFLL_OUTPUT_CFG_MAX_SHIFT 16
+#define DFLL_OUTPUT_CFG_MAX_MASK \
+ (OUT_MASK << DFLL_OUTPUT_CFG_MAX_SHIFT)
+#define DFLL_OUTPUT_CFG_MIN_SHIFT 8
+#define DFLL_OUTPUT_CFG_MIN_MASK \
+ (OUT_MASK << DFLL_OUTPUT_CFG_MIN_SHIFT)
+#define DFLL_OUTPUT_CFG_PWM_DELTA (0x1 << 7)
+#define DFLL_OUTPUT_CFG_PWM_ENABLE (0x1 << 6)
+#define DFLL_OUTPUT_CFG_PWM_DIV_SHIFT 0
+#define DFLL_OUTPUT_CFG_PWM_DIV_MASK \
+ (OUT_MASK << DFLL_OUTPUT_CFG_PWM_DIV_SHIFT)
+
+/* DFLL_OUTPUT_FORCE: closed loop mode voltage forcing control */
+#define DFLL_OUTPUT_FORCE 0x24
+#define DFLL_OUTPUT_FORCE_ENABLE (0x1 << 6)
+#define DFLL_OUTPUT_FORCE_VALUE_SHIFT 0
+#define DFLL_OUTPUT_FORCE_VALUE_MASK \
+ (OUT_MASK << DFLL_OUTPUT_FORCE_VALUE_SHIFT)
+
+/* DFLL_MONITOR_CTRL: internal monitor data source control */
+#define DFLL_MONITOR_CTRL 0x28
+#define DFLL_MONITOR_CTRL_FREQ 6
+
+/* DFLL_MONITOR_DATA: internal monitor data output */
+#define DFLL_MONITOR_DATA 0x2c
+#define DFLL_MONITOR_DATA_NEW_MASK (0x1 << 16)
+#define DFLL_MONITOR_DATA_VAL_SHIFT 0
+#define DFLL_MONITOR_DATA_VAL_MASK (0xFFFF << DFLL_MONITOR_DATA_VAL_SHIFT)
+
+/*
+ * I2C output control registers - access via dfll_i2c_{readl,writel}
+ */
+
+/* DFLL_I2C_CFG: I2C controller configuration register */
+#define DFLL_I2C_CFG 0x40
+#define DFLL_I2C_CFG_ARB_ENABLE (0x1 << 20)
+#define DFLL_I2C_CFG_HS_CODE_SHIFT 16
+#define DFLL_I2C_CFG_HS_CODE_MASK (0x7 << DFLL_I2C_CFG_HS_CODE_SHIFT)
+#define DFLL_I2C_CFG_PACKET_ENABLE (0x1 << 15)
+#define DFLL_I2C_CFG_SIZE_SHIFT 12
+#define DFLL_I2C_CFG_SIZE_MASK (0x7 << DFLL_I2C_CFG_SIZE_SHIFT)
+#define DFLL_I2C_CFG_SLAVE_ADDR_10 (0x1 << 10)
+#define DFLL_I2C_CFG_SLAVE_ADDR_SHIFT_7BIT 1
+#define DFLL_I2C_CFG_SLAVE_ADDR_SHIFT_10BIT 0
+#define DFLL_I2C_CFG_SLAVE_ADDR_MASK (0x3ff << DFLL_I2C_CFG_SLAVE_ADDR_SHIFT)
+
+/* DFLL_I2C_VDD_REG_ADDR: PMIC I2C address for closed loop mode */
+#define DFLL_I2C_VDD_REG_ADDR 0x44
+
+/* DFLL_I2C_STS: I2C controller status */
+#define DFLL_I2C_STS 0x48
+#define DFLL_I2C_STS_I2C_LAST_SHIFT 1
+#define DFLL_I2C_STS_I2C_REQ_PENDING 0x1
+
+/* DFLL_INTR_STS: DFLL interrupt status register */
+#define DFLL_INTR_STS 0x5c
+
+/* DFLL_INTR_EN: DFLL interrupt enable register */
+#define DFLL_INTR_EN 0x60
+#define DFLL_INTR_MIN_MASK 0x1
+#define DFLL_INTR_MAX_MASK 0x2
+
+/*
+ * Integrated I2C controller registers - relative to td->i2c_controller_base
+ */
+
+/* DFLL_I2C_CLK_DIVISOR: I2C controller clock divisor */
+#define DFLL_I2C_CLK_DIVISOR 0x6c
+#define DFLL_I2C_CLK_DIVISOR_MASK 0xffff
+#define DFLL_I2C_CLK_DIVISOR_FS_SHIFT 16
+#define DFLL_I2C_CLK_DIVISOR_HS_SHIFT 0
+#define DFLL_I2C_CLK_DIVISOR_PREDIV 8
+#define DFLL_I2C_CLK_DIVISOR_HSMODE_PREDIV 12
+
+/*
+ * Other constants
+ */
+
+/* MAX_DFLL_VOLTAGES: number of LUT entries in the DFLL IP block */
+#define MAX_DFLL_VOLTAGES 33
+
+/*
+ * REF_CLK_CYC_PER_DVCO_SAMPLE: the number of ref_clk cycles that the hardware
+ * integrates the DVCO counter over - used for debug rate monitoring and
+ * droop control
+ */
+#define REF_CLK_CYC_PER_DVCO_SAMPLE 4
+
+/*
+ * REF_CLOCK_RATE: the DFLL reference clock rate currently supported by this
+ * driver, in Hz
+ */
+#define REF_CLOCK_RATE 51000000UL
+
+
+/**
+ * enum dfll_ctrl_mode - DFLL hardware operating mode
+ * @DFLL_UNINITIALIZED: (uninitialized state - not in hardware bitfield)
+ * @DFLL_DISABLED: DFLL not generating an output clock
+ * @DFLL_OPEN_LOOP: DVCO running, but DFLL not adjusting voltage
+ *
+ * The integer corresponding to the last two states, minus one, is
+ * written to the DFLL hardware to change operating modes.
+ */
+enum dfll_ctrl_mode {
+ DFLL_UNINITIALIZED = 0,
+ DFLL_DISABLED = 1,
+ DFLL_OPEN_LOOP = 2,
+};
+
+/**
+ * enum dfll_tune_range - voltage range that the driver believes it's in
+ * @DFLL_TUNE_UNINITIALIZED: DFLL tuning not yet programmed
+ * @DFLL_TUNE_LOW: DFLL in the low-voltage range (or open-loop mode)
+ *
+ * Some DFLL tuning parameters may need to change depending on the
+ * DVCO's voltage; these states represent the ranges that the driver
+ * supports. These are software states; these values are never
+ * written into registers.
+ */
+enum dfll_tune_range {
+ DFLL_TUNE_UNINITIALIZED = 0,
+ DFLL_TUNE_LOW = 1,
+};
+
+struct tegra_dfll {
+ struct device *dev;
+ struct tegra_dfll_soc_data *soc;
+
+ void __iomem *base;
+ void __iomem *i2c_base;
+ void __iomem *i2c_controller_base;
+ void __iomem *lut_base;
+
+ struct clk *soc_clk;
+ struct clk *ref_clk;
+ struct clk *i2c_clk;
+ struct clk *dfll_clk;
+ unsigned long ref_rate;
+ unsigned long i2c_clk_rate;
+ unsigned long dvco_rate_min;
+
+ enum dfll_ctrl_mode mode;
+ enum dfll_tune_range tune_range;
+ struct dentry *debugfs_dir;
+ struct clk_hw dfll_clk_hw;
+ const char *output_clock_name;
+
+ /* Parameters from DT */
+ u32 droop_ctrl;
+};
+
+#define clk_hw_to_dfll(_hw) container_of(_hw, struct tegra_dfll, dfll_clk_hw)
+
+/* mode_name: map numeric DFLL modes to names for friendly console messages */
+static const char * const mode_name[] = {
+ [DFLL_UNINITIALIZED] = "uninitialized",
+ [DFLL_DISABLED] = "disabled",
+ [DFLL_OPEN_LOOP] = "open_loop",
+};
+
+/*
+ * Register accessors
+ */
+
+static inline u32 dfll_readl(struct tegra_dfll *td, u32 offs)
+{
+ return __raw_readl(td->base + offs);
+}
+
+static inline void dfll_writel(struct tegra_dfll *td, u32 val, u32 offs)
+{
+ WARN_ON(offs >= DFLL_I2C_CFG);
+ __raw_writel(val, td->base + offs);
+}
+
+static inline void dfll_wmb(struct tegra_dfll *td)
+{
+ dfll_readl(td, DFLL_CTRL);
+}
+
+/* I2C output control registers - for addresses above DFLL_I2C_CFG */
+
+static inline u32 dfll_i2c_readl(struct tegra_dfll *td, u32 offs)
+{
+ return __raw_readl(td->i2c_base + offs);
+}
+
+static inline void dfll_i2c_writel(struct tegra_dfll *td, u32 val, u32 offs)
+{
+ __raw_writel(val, td->i2c_base + offs);
+}
+
+static inline void dfll_i2c_wmb(struct tegra_dfll *td)
+{
+ dfll_i2c_readl(td, DFLL_I2C_CFG);
+}
+
+/**
+ * dfll_is_running - is the DFLL currently generating a clock?
+ * @td: DFLL instance
+ *
+ * If the DFLL is currently generating an output clock signal, return
+ * true; otherwise return false.
+ */
+static bool dfll_is_running(struct tegra_dfll *td)
+{
+ return td->mode >= DFLL_OPEN_LOOP;
+}
+
+/*
+ * Runtime PM suspend/resume callbacks
+ */
+
+/**
+ * tegra_dfll_runtime_resume - enable all clocks needed by the DFLL
+ * @dev: DFLL device *
+ *
+ * Enable all clocks needed by the DFLL. Assumes that clk_prepare()
+ * has already been called on all the clocks.
+ *
+ * XXX Should also handle context restore when returning from off.
+ */
+int tegra_dfll_runtime_resume(struct device *dev)
+{
+ struct tegra_dfll *td = dev_get_drvdata(dev);
+ int ret;
+
+ ret = clk_enable(td->i2c_clk);
+ if (ret) {
+ dev_err(dev, "could not enable I2C clock: %d\n", ret);
+ return ret;
+ }
+
+ ret = clk_enable(td->ref_clk);
+ if (ret) {
+ dev_err(dev, "could not enable ref clock: %d\n", ret);
+ return ret;
+ }
+
+ ret = clk_enable(td->soc_clk);
+ if (ret) {
+ dev_err(dev, "could not enable register clock: %d\n", ret);
+ return ret;
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(tegra_dfll_runtime_resume);
+
+/**
+ * tegra_dfll_runtime_suspend - disable all clocks needed by the DFLL
+ * @dev: DFLL device *
+ *
+ * Disable all clocks needed by the DFLL. Assumes that other code
+ * will later call clk_unprepare().
+ */
+int tegra_dfll_runtime_suspend(struct device *dev)
+{
+ struct tegra_dfll *td = dev_get_drvdata(dev);
+
+ clk_disable(td->soc_clk);
+ clk_disable(td->ref_clk);
+ clk_disable(td->i2c_clk);
+
+ return 0;
+}
+EXPORT_SYMBOL(tegra_dfll_runtime_suspend);
+
+/*
+ * DFLL tuning operations (per-voltage-range tuning settings)
+ */
+
+/**
+ * dfll_tune_low - tune to DFLL and CPU settings valid for any voltage
+ * @td: DFLL instance
+ *
+ * Tune the DFLL oscillator parameters and the CPU clock shaper for
+ * the low-voltage range. These settings are valid for any voltage,
+ * but may not be optimal.
+ */
+static void dfll_tune_low(struct tegra_dfll *td)
+{
+ td->tune_range = DFLL_TUNE_LOW;
+
+ dfll_writel(td, td->soc->tune0_low, DFLL_TUNE0);
+ dfll_writel(td, td->soc->tune1, DFLL_TUNE1);
+ dfll_wmb(td);
+
+ if (td->soc->set_clock_trimmers_low)
+ td->soc->set_clock_trimmers_low();
+}
+
+/*
+ * Output clock scaler helpers
+ */
+
+/**
+ * dfll_scale_dvco_rate - calculate scaled rate from the DVCO rate
+ * @scale_bits: clock scaler value (bits in the DFLL_FREQ_REQ_SCALE field)
+ * @dvco_rate: the DVCO rate
+ *
+ * Apply the same scaling formula that the DFLL hardware uses to scale
+ * the DVCO rate.
+ */
+static unsigned long dfll_scale_dvco_rate(int scale_bits,
+ unsigned long dvco_rate)
+{
+ return (u64)dvco_rate * (scale_bits + 1) / DFLL_FREQ_REQ_SCALE_MAX;
+}
+
+/*
+ * Monitor control
+ */
+
+/**
+ * dfll_calc_monitored_rate - convert DFLL_MONITOR_DATA_VAL rate into real freq
+ * @monitor_data: value read from the DFLL_MONITOR_DATA_VAL bitfield
+ * @ref_rate: DFLL reference clock rate
+ *
+ * Convert @monitor_data from DFLL_MONITOR_DATA_VAL units into cycles
+ * per second. Returns the converted value.
+ */
+static u64 dfll_calc_monitored_rate(u32 monitor_data,
+ unsigned long ref_rate)
+{
+ return monitor_data * (ref_rate / REF_CLK_CYC_PER_DVCO_SAMPLE);
+}
+
+/**
+ * dfll_read_monitor_rate - return the DFLL's output rate from internal monitor
+ * @td: DFLL instance
+ *
+ * If the DFLL is enabled, return the last rate reported by the DFLL's
+ * internal monitoring hardware. This works in both open-loop and
+ * closed-loop mode, and takes the output scaler setting into account.
+ * Assumes that the monitor was programmed to monitor frequency before
+ * the sample period started. If the driver believes that the DFLL is
+ * currently uninitialized or disabled, it will return 0, since
+ * otherwise the DFLL monitor data register will return the last
+ * measured rate from when the DFLL was active.
+ */
+static u64 dfll_read_monitor_rate(struct tegra_dfll *td)
+{
+ u32 v, s;
+ u64 pre_scaler_rate, post_scaler_rate;
+
+ if (!dfll_is_running(td))
+ return 0;
+
+ v = dfll_readl(td, DFLL_MONITOR_DATA);
+ v = (v & DFLL_MONITOR_DATA_VAL_MASK) >> DFLL_MONITOR_DATA_VAL_SHIFT;
+ pre_scaler_rate = dfll_calc_monitored_rate(v, td->ref_rate);
+
+ s = dfll_readl(td, DFLL_FREQ_REQ);
+ s = (s & DFLL_FREQ_REQ_SCALE_MASK) >> DFLL_FREQ_REQ_SCALE_SHIFT;
+ post_scaler_rate = dfll_scale_dvco_rate(pre_scaler_rate, s);
+
+ return post_scaler_rate;
+}
+
+/*
+ * DFLL mode switching
+ */
+
+/**
+ * dfll_set_mode - change the DFLL control mode
+ * @td: DFLL instance
+ * @mode: DFLL control mode (see enum dfll_ctrl_mode)
+ *
+ * Change the DFLL's operating mode between disabled, open-loop mode,
+ * and closed-loop mode, or vice versa.
+ */
+static void dfll_set_mode(struct tegra_dfll *td,
+ enum dfll_ctrl_mode mode)
+{
+ td->mode = mode;
+ dfll_writel(td, mode - 1, DFLL_CTRL);
+ dfll_wmb(td);
+}
+
+/*
+ * DFLL enable/disable & open-loop <-> closed-loop transitions
+ */
+
+/**
+ * dfll_disable - switch from open-loop mode to disabled mode
+ * @td: DFLL instance
+ *
+ * Switch from OPEN_LOOP state to DISABLED state. Returns 0 upon success
+ * or -EPERM if the DFLL is not currently in open-loop mode.
+ */
+static int dfll_disable(struct tegra_dfll *td)
+{
+ if (td->mode != DFLL_OPEN_LOOP) {
+ dev_err(td->dev, "cannot disable DFLL in %s mode\n",
+ mode_name[td->mode]);
+ return -EINVAL;
+ }
+
+ dfll_set_mode(td, DFLL_DISABLED);
+ pm_runtime_put_sync(td->dev);
+
+ return 0;
+}
+
+/**
+ * dfll_enable - switch a disabled DFLL to open-loop mode
+ * @td: DFLL instance
+ *
+ * Switch from DISABLED state to OPEN_LOOP state. Returns 0 upon success
+ * or -EPERM if the DFLL is not currently disabled.
+ */
+static int dfll_enable(struct tegra_dfll *td)
+{
+ if (td->mode != DFLL_DISABLED) {
+ dev_err(td->dev, "cannot enable DFLL in %s mode\n",
+ mode_name[td->mode]);
+ return -EPERM;
+ }
+
+ pm_runtime_get_sync(td->dev);
+ dfll_set_mode(td, DFLL_OPEN_LOOP);
+
+ return 0;
+}
+
+/**
+ * dfll_set_open_loop_config - prepare to switch to open-loop mode
+ * @td: DFLL instance
+ *
+ * Prepare to switch the DFLL to open-loop mode. This switches the
+ * DFLL to the low-voltage tuning range, ensures that I2C output
+ * forcing is disabled, and disables the output clock rate scaler.
+ * The DFLL's low-voltage tuning range parameters must be
+ * characterized to keep the downstream device stable at any DVCO
+ * input voltage. No return value.
+ */
+static void dfll_set_open_loop_config(struct tegra_dfll *td)
+{
+ u32 val;
+
+ /* always tune low (safe) in open loop */
+ if (td->tune_range != DFLL_TUNE_LOW)
+ dfll_tune_low(td);
+
+ val = dfll_readl(td, DFLL_FREQ_REQ);
+ val |= DFLL_FREQ_REQ_SCALE_MASK;
+ val &= ~DFLL_FREQ_REQ_FORCE_ENABLE;
+ dfll_writel(td, val, DFLL_FREQ_REQ);
+}
+
+/*
+ * Clock framework integration
+ */
+
+static int dfll_clk_is_enabled(struct clk_hw *hw)
+{
+ struct tegra_dfll *td = clk_hw_to_dfll(hw);
+
+ return dfll_is_running(td);
+}
+
+static int dfll_clk_enable(struct clk_hw *hw)
+{
+ struct tegra_dfll *td = clk_hw_to_dfll(hw);
+
+ return dfll_enable(td);
+}
+
+static void dfll_clk_disable(struct clk_hw *hw)
+{
+ struct tegra_dfll *td = clk_hw_to_dfll(hw);
+
+ dfll_disable(td);
+}
+
+static const struct clk_ops dfll_clk_ops = {
+ .is_enabled = dfll_clk_is_enabled,
+ .enable = dfll_clk_enable,
+ .disable = dfll_clk_disable,
+};
+
+static struct clk_init_data dfll_clk_init_data = {
+ .flags = CLK_IS_ROOT,
+ .ops = &dfll_clk_ops,
+ .num_parents = 0,
+};
+
+/**
+ * dfll_register_clk - register the DFLL output clock with the clock framework
+ * @td: DFLL instance
+ *
+ * Register the DFLL's output clock with the Linux clock framework and register
+ * the DFLL driver as an OF clock provider. Returns 0 upon success or -EINVAL
+ * upon failure.
+ */
+static int dfll_register_clk(struct tegra_dfll *td)
+{
+ int ret;
+
+ dfll_clk_init_data.name = td->output_clock_name;
+ td->dfll_clk_hw.init = &dfll_clk_init_data;
+
+ td->dfll_clk = clk_register(td->dev, &td->dfll_clk_hw);
+ if (IS_ERR(td->dfll_clk)) {
+ dev_err(td->dev, "DFLL clock registration error\n");
+ return -EINVAL;
+ }
+
+ ret = of_clk_add_provider(td->dev->of_node, of_clk_src_simple_get,
+ td->dfll_clk);
+ if (ret) {
+ dev_err(td->dev, "of_clk_add_provider() failed\n");
+ goto out_unregister_clk;
+ }
+
+ return 0;
+
+out_unregister_clk:
+ clk_unregister(td->dfll_clk);
+
+ return ret;
+}
+
+/**
+ * dfll_unregister_clk - unregister the DFLL output clock
+ * @td: DFLL instance
+ *
+ * Unregister the DFLL's output clock from the Linux clock framework
+ * and from clkdev. No return value.
+ */
+static void dfll_unregister_clk(struct tegra_dfll *td)
+{
+ of_clk_del_provider(td->dev->of_node);
+ clk_unregister(td->dfll_clk);
+ td->dfll_clk = NULL;
+}
+
+/*
+ * Debugfs interface
+ */
+
+#ifdef CONFIG_DEBUG_FS
+
+static int attr_enable_get(void *data, u64 *val)
+{
+ struct tegra_dfll *td = data;
+
+ *val = dfll_is_running(td);
+
+ return 0;
+}
+static int attr_enable_set(void *data, u64 val)
+{
+ struct tegra_dfll *td = data;
+
+ return val ? dfll_enable(td) : dfll_disable(td);
+}
+DEFINE_SIMPLE_ATTRIBUTE(enable_fops, attr_enable_get, attr_enable_set,
+ "%llu\n");
+
+static int attr_rate_get(void *data, u64 *val)
+{
+ struct tegra_dfll *td = data;
+
+ *val = dfll_read_monitor_rate(td);
+
+ return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(rate_fops, attr_rate_get, NULL, "%llu\n");
+
+static int attr_registers_show(struct seq_file *s, void *data)
+{
+ u32 offs;
+ struct tegra_dfll *td = s->private;
+
+ seq_puts(s, "CONTROL REGISTERS:\n");
+ for (offs = 0; offs <= DFLL_MONITOR_DATA; offs += 4)
+ seq_printf(s, "[0x%02x] = 0x%08x\n", offs,
+ dfll_readl(td, offs));
+
+ seq_puts(s, "\nI2C and INTR REGISTERS:\n");
+ for (offs = DFLL_I2C_CFG; offs <= DFLL_I2C_STS; offs += 4)
+ seq_printf(s, "[0x%02x] = 0x%08x\n", offs,
+ dfll_i2c_readl(td, offs));
+ for (offs = DFLL_INTR_STS; offs <= DFLL_INTR_EN; offs += 4)
+ seq_printf(s, "[0x%02x] = 0x%08x\n", offs,
+ dfll_i2c_readl(td, offs));
+
+ seq_puts(s, "\nINTEGRATED I2C CONTROLLER REGISTERS:\n");
+ offs = DFLL_I2C_CLK_DIVISOR;
+ seq_printf(s, "[0x%02x] = 0x%08x\n", offs,
+ __raw_readl(td->i2c_controller_base + offs));
+
+ seq_puts(s, "\nLUT:\n");
+ for (offs = 0; offs < 4 * MAX_DFLL_VOLTAGES; offs += 4)
+ seq_printf(s, "[0x%02x] = 0x%08x\n", offs,
+ __raw_readl(td->lut_base + offs));
+
+ return 0;
+}
+
+static int attr_registers_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, attr_registers_show, inode->i_private);
+}
+
+static const struct file_operations attr_registers_fops = {
+ .open = attr_registers_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+static int dfll_debug_init(struct tegra_dfll *td)
+{
+ int ret;
+
+ if (!td || (td->mode == DFLL_UNINITIALIZED))
+ return 0;
+
+ td->debugfs_dir = debugfs_create_dir("tegra_dfll_fcpu", NULL);
+ if (!td->debugfs_dir)
+ return -ENOMEM;
+
+ ret = -ENOMEM;
+
+ if (!debugfs_create_file("enable", S_IRUGO | S_IWUSR,
+ td->debugfs_dir, td, &enable_fops))
+ goto err_out;
+
+ if (!debugfs_create_file("rate", S_IRUGO,
+ td->debugfs_dir, td, &rate_fops))
+ goto err_out;
+
+ if (!debugfs_create_file("registers", S_IRUGO,
+ td->debugfs_dir, td, &attr_registers_fops))
+ goto err_out;
+
+ return 0;
+
+err_out:
+ debugfs_remove_recursive(td->debugfs_dir);
+ return ret;
+}
+
+#endif /* CONFIG_DEBUG_FS */
+
+/*
+ * DFLL initialization
+ */
+
+/**
+ * dfll_set_default_params - program non-output related DFLL parameters
+ * @td: DFLL instance
+ *
+ * During DFLL driver initialization or resume from context loss,
+ * program parameters for the closed loop integrator, DVCO tuning,
+ * voltage droop control and monitor control.
+ */
+static void dfll_set_default_params(struct tegra_dfll *td)
+{
+ dfll_tune_low(td);
+ dfll_writel(td, td->droop_ctrl, DFLL_DROOP_CTRL);
+ dfll_writel(td, DFLL_MONITOR_CTRL_FREQ, DFLL_MONITOR_CTRL);
+}
+
+/**
+ * dfll_init_clks - clk_get() the DFLL source clocks
+ * @td: DFLL instance
+ *
+ * Call clk_get() on the DFLL source clocks and save the pointers for later
+ * use. Returns 0 upon success or -ENODEV if one or more of the clocks
+ * couldn't be looked up.
+ */
+static int dfll_init_clks(struct tegra_dfll *td)
+{
+ td->ref_clk = devm_clk_get(td->dev, "ref");
+ if (IS_ERR(td->ref_clk)) {
+ dev_err(td->dev, "missing ref clock\n");
+ return PTR_ERR(td->ref_clk);
+ }
+
+ td->soc_clk = devm_clk_get(td->dev, "soc");
+ if (IS_ERR(td->soc_clk)) {
+ dev_err(td->dev, "missing soc clock\n");
+ return PTR_ERR(td->soc_clk);
+ }
+
+ td->i2c_clk = devm_clk_get(td->dev, "i2c");
+ if (IS_ERR(td->i2c_clk)) {
+ dev_err(td->dev, "missing i2c clock\n");
+ return PTR_ERR(td->i2c_clk);
+ }
+ td->i2c_clk_rate = clk_get_rate(td->i2c_clk);
+
+ return 0;
+}
+
+/**
+ * dfll_init - Prepare the DFLL IP block for use
+ * @td: DFLL instance
+ *
+ * Do everything necessary to prepare the DFLL IP block for use. The
+ * DFLL will be left in DISABLED state. Called by dfll_probe().
+ * Returns 0 upon success, or passes along the error from whatever
+ * function returned it.
+ */
+static int dfll_init(struct tegra_dfll *td)
+{
+ int ret;
+
+ td->ref_rate = clk_get_rate(td->ref_clk);
+ if (td->ref_rate != REF_CLOCK_RATE) {
+ dev_err(td->dev, "unexpected ref clk rate %lu, expecting %lu",
+ td->ref_rate, REF_CLOCK_RATE);
+ return -EINVAL;
+ }
+
+ if (td->soc->deassert_dvco_reset)
+ td->soc->deassert_dvco_reset();
+
+ ret = clk_prepare(td->i2c_clk);
+ if (ret) {
+ dev_err(td->dev, "failed to prepare i2c_clk\n");
+ return ret;
+ }
+
+ ret = clk_prepare(td->ref_clk);
+ if (ret) {
+ dev_err(td->dev, "failed to prepare ref_clk\n");
+ goto di_err1;
+ }
+
+ ret = clk_prepare(td->soc_clk);
+ if (ret) {
+ dev_err(td->dev, "failed to prepare soc_clk\n");
+ goto di_err2;
+ }
+
+ pm_runtime_enable(td->dev);
+ pm_runtime_get_sync(td->dev);
+
+ dfll_set_mode(td, DFLL_DISABLED);
+ dfll_set_default_params(td);
+
+ if (td->soc->init_clock_trimmers)
+ td->soc->init_clock_trimmers();
+
+ dfll_set_open_loop_config(td);
+
+ pm_runtime_put_sync(td->dev);
+
+ return 0;
+
+di_err2:
+ clk_unprepare(td->ref_clk);
+di_err1:
+ clk_unprepare(td->i2c_clk);
+
+ if (td->soc->assert_dvco_reset)
+ td->soc->assert_dvco_reset();
+
+ return ret;
+}
+
+/*
+ * DT data fetch
+ */
+
+/**
+ * read_dt_param - helper function for reading required parameters from the DT
+ * @td: DFLL instance
+ * @param: DT property name
+ * @dest: output pointer for the value read
+ *
+ * Read a required numeric parameter from the DFLL device node, or complain
+ * if the property doesn't exist. Returns a boolean indicating success for
+ * easy chaining of multiple calls to this function.
+ */
+static bool read_dt_param(struct tegra_dfll *td, const char *param, u32 *dest)
+{
+ int err = of_property_read_u32(td->dev->of_node, param, dest);
+
+ if (err < 0) {
+ dev_err(td->dev, "failed to read DT parameter %s: %d\n",
+ param, err);
+ return false;
+ }
+
+ return true;
+}
+
+/**
+ * dfll_parse_dt_params - read DFLL parameters from the device tree
+ * @td: DFLL instance
+ *
+ * Read all the required parameters for the DFLL from the device tree.
+ * Returns 0 on success or -EINVAL on any failure.
+ */
+static int dfll_parse_dt_params(struct tegra_dfll *td)
+{
+ bool ok = true;
+
+ ok &= read_dt_param(td, "nvidia,droop-ctrl", &td->droop_ctrl);
+
+ if (of_property_read_string(td->dev->of_node, "clock-output-names",
+ &td->output_clock_name)) {
+ dev_err(td->dev, "missing clock-output-names property\n");
+ ok = false;
+ }
+
+ return ok ? 0 : -EINVAL;
+}
+
+/*
+ * API exported to per-SoC platform drivers
+ */
+
+/**
+ * tegra_dfll_register - probe a Tegra DFLL device
+ * @pdev: DFLL platform_device *
+ * @soc: Per-SoC integration and characterization data for this DFLL instance
+ *
+ * Probe and initialize a DFLL device instance. Intended to be called
+ * by a SoC-specific shim driver that passes in per-SoC integration
+ * and configuration data via @soc. Returns 0 on success or -err on failure.
+ */
+int tegra_dfll_register(struct platform_device *pdev,
+ struct tegra_dfll_soc_data *soc)
+{
+ struct resource *mem;
+ struct tegra_dfll *td;
+ int ret;
+
+ td = devm_kzalloc(&pdev->dev, sizeof(*td), GFP_KERNEL);
+ if (!td)
+ return -ENOMEM;
+ td->dev = &pdev->dev;
+ platform_set_drvdata(pdev, td);
+
+ if (!soc) {
+ dev_err(td->dev, "no tegra_dfll_soc_data provided\n");
+ return -EINVAL;
+ }
+ td->soc = soc;
+
+ ret = dfll_parse_dt_params(td);
+ if (ret) {
+ dev_err(td->dev, "couldn't parse device tree parameters\n");
+ return ret;
+ }
+
+ mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (!mem) {
+ dev_err(td->dev, "no control register resource\n");
+ return -ENODEV;
+ }
+
+ td->base = devm_ioremap(td->dev, mem->start, resource_size(mem));
+ if (!td->base) {
+ dev_err(td->dev, "couldn't ioremap DFLL control registers\n");
+ return -ENODEV;
+ }
+
+ mem = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+ if (!mem) {
+ dev_err(td->dev, "no i2c_base resource\n");
+ return -ENODEV;
+ }
+
+ td->i2c_base = devm_ioremap(td->dev, mem->start, resource_size(mem));
+ if (!td->i2c_base) {
+ dev_err(td->dev, "couldn't ioremap i2c_base resource\n");
+ return -ENODEV;
+ }
+
+ mem = platform_get_resource(pdev, IORESOURCE_MEM, 2);
+ if (!mem) {
+ dev_err(td->dev, "no i2c_controller_base resource\n");
+ return -ENODEV;
+ }
+
+ td->i2c_controller_base = devm_ioremap(td->dev, mem->start,
+ resource_size(mem));
+ if (!td->i2c_controller_base) {
+ dev_err(td->dev,
+ "couldn't ioremap i2c_controller_base resource\n");
+ return -ENODEV;
+ }
+
+ mem = platform_get_resource(pdev, IORESOURCE_MEM, 3);
+ if (!mem) {
+ dev_err(td->dev, "no lut_base resource\n");
+ return -ENODEV;
+ }
+
+ td->lut_base = devm_ioremap(td->dev, mem->start, resource_size(mem));
+ if (!td->lut_base) {
+ dev_err(td->dev,
+ "couldn't ioremap lut_base resource\n");
+ return -ENODEV;
+ }
+
+ ret = dfll_init_clks(td);
+ if (ret) {
+ dev_err(&pdev->dev, "DFLL clock init error\n");
+ return ret;
+ }
+
+ /* Enable the clocks and set the device up */
+ ret = dfll_init(td);
+ if (ret)
+ return ret;
+
+ ret = dfll_register_clk(td);
+ if (ret) {
+ dev_err(&pdev->dev, "DFLL clk registration failed\n");
+ return ret;
+ }
+
+#ifdef CONFIG_DEBUG_FS
+ dfll_debug_init(td);
+#endif
+
+ return 0;
+}
+EXPORT_SYMBOL(tegra_dfll_register);
+
+/**
+ * tegra_dfll_unregister - release all of the DFLL driver resources for a device
+ * @pdev: DFLL platform_device *
+ *
+ * Unbind this driver from the DFLL hardware device represented by
+ * @pdev. The DFLL must be disabled for this to succeed. Returns 0
+ * upon success or -EBUSY if the DFLL is still active.
+ */
+int tegra_dfll_unregister(struct platform_device *pdev)
+{
+ struct tegra_dfll *td = platform_get_drvdata(pdev);
+
+ /* Try to prevent removal while the DFLL is active */
+ if (td->mode != DFLL_DISABLED) {
+ dev_err(&pdev->dev,
+ "must disable DFLL before removing driver\n");
+ return -EBUSY;
+ }
+
+ debugfs_remove_recursive(td->debugfs_dir);
+
+ dfll_unregister_clk(td);
+ pm_runtime_disable(&pdev->dev);
+
+ clk_unprepare(td->soc_clk);
+ clk_unprepare(td->ref_clk);
+ clk_unprepare(td->i2c_clk);
+
+ if (td->soc->assert_dvco_reset)
+ td->soc->assert_dvco_reset();
+
+ return 0;
+}
+EXPORT_SYMBOL(tegra_dfll_unregister);
diff --git a/drivers/clk/tegra/clk-dfll.h b/drivers/clk/tegra/clk-dfll.h
new file mode 100644
index 0000000..fbf90c4
--- /dev/null
+++ b/drivers/clk/tegra/clk-dfll.h
@@ -0,0 +1,55 @@
+/*
+ * clk-dfll.h - prototypes and macros for the Tegra DFLL clocksource driver
+ * Copyright (C) 2013 NVIDIA Corporation. All rights reserved.
+ *
+ * Aleksandr Frid <[email protected]>
+ * Paul Walmsley <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ */
+
+#ifndef __DRIVERS_CLK_TEGRA_CLK_DFLL_H
+#define __DRIVERS_CLK_TEGRA_CLK_DFLL_H
+
+#include <linux/platform_device.h>
+#include <linux/types.h>
+
+/**
+ * struct tegra_dfll_soc - SoC-specific hooks/integration for the DFLL driver
+ * @opp_dev: struct device * that holds the OPP table for the DFLL
+ * @min_millivolts: minimum voltage (in mV) that the DFLL can operate
+ * @tune0_low: DFLL tuning register 0 (low voltage range)
+ * @tune0_high: DFLL tuning register 0 (high voltage range)
+ * @tune1: DFLL tuning register 1
+ * @assert_dvco_reset: fn ptr to place the DVCO in reset
+ * @deassert_dvco_reset: fn ptr to release the DVCO reset
+ * @set_clock_trimmers_high: fn ptr to tune clock trimmers for high voltage
+ * @set_clock_trimmers_low: fn ptr to tune clock trimmers for low voltage
+ */
+struct tegra_dfll_soc_data {
+ struct device *opp_dev;
+ unsigned int min_millivolts;
+ u32 tune0_low;
+ u32 tune0_high;
+ u32 tune1;
+ void (*assert_dvco_reset)(void);
+ void (*deassert_dvco_reset)(void);
+ void (*init_clock_trimmers)(void);
+ void (*set_clock_trimmers_high)(void);
+ void (*set_clock_trimmers_low)(void);
+};
+
+int tegra_dfll_register(struct platform_device *pdev,
+ struct tegra_dfll_soc_data *soc);
+int tegra_dfll_unregister(struct platform_device *pdev);
+int tegra_dfll_runtime_suspend(struct device *dev);
+int tegra_dfll_runtime_resume(struct device *dev);
+
+#endif /* __DRIVERS_CLK_TEGRA_CLK_DFLL_H */
--
1.8.1.5

2014-07-10 21:43:40

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 06/13] clk: tegra: Add Tegra124 DFLL clocksource platform driver

Add basic platform driver support for the fast CPU cluster DFLL
clocksource found on Tegra124 SoCs. This small driver selects the
appropriate Tegra124-specific characterization data and integration
code. It relies on the DFLL common code to do most of the work.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
drivers/clk/tegra/Makefile | 2 +
drivers/clk/tegra/clk-tegra124-dfll-fcpu.c | 156 +++++++++++++++++++++++++++++
2 files changed, 158 insertions(+)
create mode 100644 drivers/clk/tegra/clk-tegra124-dfll-fcpu.c

diff --git a/drivers/clk/tegra/Makefile b/drivers/clk/tegra/Makefile
index 47320ca..2f87188 100644
--- a/drivers/clk/tegra/Makefile
+++ b/drivers/clk/tegra/Makefile
@@ -16,3 +16,5 @@ obj-$(CONFIG_ARCH_TEGRA_2x_SOC) += clk-tegra20.o
obj-$(CONFIG_ARCH_TEGRA_3x_SOC) += clk-tegra30.o
obj-$(CONFIG_ARCH_TEGRA_114_SOC) += clk-tegra114.o
obj-$(CONFIG_ARCH_TEGRA_124_SOC) += clk-tegra124.o
+obj-$(CONFIG_ARCH_TEGRA_124_SOC) += clk-tegra124-dfll-fcpu.o
+obj-y += cvb.o
diff --git a/drivers/clk/tegra/clk-tegra124-dfll-fcpu.c b/drivers/clk/tegra/clk-tegra124-dfll-fcpu.c
new file mode 100644
index 0000000..58512d5
--- /dev/null
+++ b/drivers/clk/tegra/clk-tegra124-dfll-fcpu.c
@@ -0,0 +1,156 @@
+/*
+ * Tegra124 DFLL FCPU clock source driver
+ *
+ * Copyright (C) 2012-2014 NVIDIA Corporation. All rights reserved.
+ *
+ * Aleksandr Frid <[email protected]>
+ * Paul Walmsley <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <linux/cpu.h>
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/tegra-soc.h>
+
+#include "clk.h"
+#include "clk-dfll.h"
+#include "cvb.h"
+
+/* Maximum CPU frequency, indexed by CPU speedo id */
+static const unsigned long cpu_max_freq_table[] = {
+ [0] = 2014500000UL,
+ [1] = 2320500000UL,
+ [2] = 2116500000UL,
+ [3] = 2524500000UL,
+};
+
+static const struct cvb_table tegra124_cpu_cvb_tables[] = {
+ {
+ .speedo_id = -1,
+ .process_id = -1,
+ .min_millivolts = 900,
+ .max_millivolts = 1260,
+ .alignment = {
+ .step_uv = 10000, /* 10mV */
+ },
+ .speedo_scale = 100,
+ .voltage_scale = 1000,
+ .cvb_table = {
+ {204000000UL, {1112619, -29295, 402} },
+ {306000000UL, {1150460, -30585, 402} },
+ {408000000UL, {1190122, -31865, 402} },
+ {510000000UL, {1231606, -33155, 402} },
+ {612000000UL, {1274912, -34435, 402} },
+ {714000000UL, {1320040, -35725, 402} },
+ {816000000UL, {1366990, -37005, 402} },
+ {918000000UL, {1415762, -38295, 402} },
+ {1020000000UL, {1466355, -39575, 402} },
+ {1122000000UL, {1518771, -40865, 402} },
+ {1224000000UL, {1573009, -42145, 402} },
+ {1326000000UL, {1629068, -43435, 402} },
+ {1428000000UL, {1686950, -44715, 402} },
+ {1530000000UL, {1746653, -46005, 402} },
+ {1632000000UL, {1808179, -47285, 402} },
+ {1734000000UL, {1871526, -48575, 402} },
+ {1836000000UL, {1936696, -49855, 402} },
+ {1938000000UL, {2003687, -51145, 402} },
+ {2014500000UL, {2054787, -52095, 402} },
+ {2116500000UL, {2124957, -53385, 402} },
+ {2218500000UL, {2196950, -54665, 402} },
+ {2320500000UL, {2270765, -55955, 402} },
+ {2422500000UL, {2346401, -57235, 402} },
+ {2524500000UL, {2437299, -58535, 402} },
+ {0, { 0, 0, 0} },
+ },
+ .cpu_dfll_data = {
+ .tune0_low = 0x005020ff,
+ .tune0_high = 0x005040ff,
+ .tune1 = 0x00000060,
+ }
+ },
+};
+
+static struct tegra_dfll_soc_data soc = {
+ .assert_dvco_reset = tegra124_clock_assert_dfll_dvco_reset,
+ .deassert_dvco_reset = tegra124_clock_deassert_dfll_dvco_reset,
+};
+
+static int tegra124_dfll_fcpu_probe(struct platform_device *pdev)
+{
+ int process_id, speedo_id, speedo_value;
+ const struct cvb_table *cvb;
+
+ process_id = tegra_sku_info.cpu_process_id;
+ speedo_id = tegra_sku_info.cpu_speedo_id;
+ speedo_value = tegra_sku_info.cpu_speedo_value;
+
+ if (speedo_id >= ARRAY_SIZE(cpu_max_freq_table)) {
+ dev_err(&pdev->dev, "unknown max CPU freq for speedo_id=%d\n",
+ speedo_id);
+ return -ENODEV;
+ }
+
+ soc.opp_dev = get_cpu_device(0);
+ if (!soc.opp_dev) {
+ dev_err(&pdev->dev, "no CPU0 device");
+ return -ENODEV;
+ }
+
+ cvb = tegra_cvb_build_opp_table(tegra124_cpu_cvb_tables,
+ ARRAY_SIZE(tegra124_cpu_cvb_tables),
+ process_id, speedo_id, speedo_value,
+ cpu_max_freq_table[speedo_id],
+ soc.opp_dev);
+ if (IS_ERR(cvb)) {
+ dev_err(&pdev->dev, "couldn't build OPP table: %ld\n",
+ PTR_ERR(cvb));
+ return PTR_ERR(cvb);
+ }
+
+ soc.min_millivolts = cvb->min_millivolts;
+ soc.tune0_low = cvb->cpu_dfll_data.tune0_low;
+ soc.tune0_high = cvb->cpu_dfll_data.tune0_high;
+ soc.tune1 = cvb->cpu_dfll_data.tune1;
+
+ return tegra_dfll_register(pdev, &soc);
+}
+
+static struct of_device_id tegra124_dfll_fcpu_of_match[] = {
+ { .compatible = "nvidia,tegra124-dfll", },
+ { },
+};
+MODULE_DEVICE_TABLE(of, tegra124_dfll_fcpu_of_match);
+
+static const struct dev_pm_ops tegra124_dfll_pm_ops = {
+ SET_RUNTIME_PM_OPS(tegra_dfll_runtime_suspend,
+ tegra_dfll_runtime_resume, NULL)
+};
+
+static struct platform_driver tegra124_dfll_fcpu_driver = {
+ .probe = tegra124_dfll_fcpu_probe,
+ .remove = tegra_dfll_unregister,
+ .driver = {
+ .name = "tegra124-dfll",
+ .owner = THIS_MODULE,
+ .of_match_table = tegra124_dfll_fcpu_of_match,
+ .pm = &tegra124_dfll_pm_ops,
+ },
+};
+module_platform_driver(tegra124_dfll_fcpu_driver);
+
+MODULE_DESCRIPTION("Tegra124 DFLL clock source driver");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Aleksandr Frid <[email protected]>");
+MODULE_AUTHOR("Paul Walmsley <[email protected]>");
--
1.8.1.5

2014-07-10 21:43:52

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 09/13] ARM: tegra: Add the DFLL to Tegra124 device tree

The DFLL clocksource is a separate IP block from the usual
clock-and-reset controller, so it gets its own device tree node.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
arch/arm/boot/dts/tegra124.dtsi | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)

diff --git a/arch/arm/boot/dts/tegra124.dtsi b/arch/arm/boot/dts/tegra124.dtsi
index 7cc535b..582a0e2 100644
--- a/arch/arm/boot/dts/tegra124.dtsi
+++ b/arch/arm/boot/dts/tegra124.dtsi
@@ -534,6 +534,28 @@
status = "disabled";
};

+ dfll: dfll@0,70110000 {
+ compatible = "nvidia,tegra124-dfll";
+ reg = <0 0x70110000 0 0x100>, /* DFLL control */
+ <0 0x70110000 0 0x100>, /* I2C output control */
+ <0 0x70110100 0 0x100>, /* Integrated I2C controller */
+ <0 0x70110200 0 0x100>; /* Look-up table RAM */
+ interrupts = <GIC_SPI 62 IRQ_TYPE_LEVEL_HIGH>;
+ clocks = <&tegra_car TEGRA124_CLK_DFLL_SOC>,
+ <&tegra_car TEGRA124_CLK_DFLL_REF>,
+ <&tegra_car TEGRA124_CLK_I2C5>;
+ clock-names = "soc", "ref", "i2c";
+ #clock-cells = <0>;
+ clock-output-names = "dfllCPU_out";
+ nvidia,sample-rate = <12500>;
+ nvidia,droop-ctrl = <0x00000f00>;
+ nvidia,force-mode = <1>;
+ nvidia,cf = <10>;
+ nvidia,ci = <0>;
+ nvidia,cg = <2>;
+ status = "disabled";
+ };
+
ahub@0,70300000 {
compatible = "nvidia,tegra124-ahub";
reg = <0x0 0x70300000 0x0 0x200>,
--
1.8.1.5

2014-07-10 21:43:35

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 04/13] clk: tegra: Add functions for parsing CVB tables

Tegra CVB tables encode the relationship between operating voltage and
optimal frequency as a second-order polynomial of the so-called speedo
value. The speedo value is written to the on-chip fuses at the factory,
which allows the voltage-frequency operating points to be calculated on
an per-chip basis.

Add utility functions to parse the Tegra-specific tables and export the
voltage-frequency pairs to the generic OPP framework for other drivers
to use.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
arch/arm/mach-tegra/Kconfig | 1 +
drivers/clk/tegra/cvb.c | 133 ++++++++++++++++++++++++++++++++++++++++++++
drivers/clk/tegra/cvb.h | 67 ++++++++++++++++++++++
3 files changed, 201 insertions(+)
create mode 100644 drivers/clk/tegra/cvb.c
create mode 100644 drivers/clk/tegra/cvb.h

diff --git a/arch/arm/mach-tegra/Kconfig b/arch/arm/mach-tegra/Kconfig
index 0953996..0d5832f 100644
--- a/arch/arm/mach-tegra/Kconfig
+++ b/arch/arm/mach-tegra/Kconfig
@@ -7,6 +7,7 @@ menuconfig ARCH_TEGRA
select HAVE_ARM_SCU if SMP
select HAVE_ARM_TWD if SMP
select PINCTRL
+ select PM_OPP
select ARCH_HAS_RESET_CONTROLLER
select RESET_CONTROLLER
select SOC_BUS
diff --git a/drivers/clk/tegra/cvb.c b/drivers/clk/tegra/cvb.c
new file mode 100644
index 0000000..69c74ee
--- /dev/null
+++ b/drivers/clk/tegra/cvb.c
@@ -0,0 +1,133 @@
+/*
+ * Utility functions for parsing Tegra CVB voltage tables
+ *
+ * Copyright (C) 2012-2014 NVIDIA Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ */
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <linux/pm_opp.h>
+
+#include "cvb.h"
+
+/* cvb_mv = ((c2 * speedo / s_scale + c1) * speedo / s_scale + c0) */
+static inline int get_cvb_voltage(int speedo, int s_scale,
+ const struct cvb_coefficients *cvb)
+{
+ int mv;
+
+ /* apply only speedo scale: output mv = cvb_mv * v_scale */
+ mv = DIV_ROUND_CLOSEST(cvb->c2 * speedo, s_scale);
+ mv = DIV_ROUND_CLOSEST((mv + cvb->c1) * speedo, s_scale) + cvb->c0;
+ return mv;
+}
+
+static int round_cvb_voltage(int mv, int v_scale,
+ const struct rail_alignment *align)
+{
+ /* combined: apply voltage scale and round to cvb alignment step */
+ int uv;
+ int step = (align->step_uv ? : 1000) * v_scale;
+ int offset = align->offset_uv * v_scale;
+
+ uv = max(mv * 1000, offset) - offset;
+ uv = DIV_ROUND_UP(uv, step) * align->step_uv + align->offset_uv;
+ return uv / 1000;
+}
+
+enum {
+ DOWN,
+ UP
+};
+
+static int round_voltage(int mv, const struct rail_alignment *align, int up)
+{
+ if (align->step_uv) {
+ int uv;
+
+ uv = max(mv * 1000, align->offset_uv) - align->offset_uv;
+ uv = (uv + (up ? align->step_uv - 1 : 0)) / align->step_uv;
+ return (uv * align->step_uv + align->offset_uv) / 1000;
+ }
+ return mv;
+}
+
+static int build_opp_table(const struct cvb_table *d,
+ int speedo_value,
+ unsigned long max_freq,
+ struct device *opp_dev)
+{
+ int i, ret, dfll_mv, min_mv, max_mv;
+ const struct cvb_table_freq_entry *table = NULL;
+ const struct rail_alignment *align = &d->alignment;
+
+ min_mv = round_voltage(d->min_millivolts, align, UP);
+ max_mv = round_voltage(d->max_millivolts, align, DOWN);
+
+ for (i = 0; i < MAX_DVFS_FREQS; i++) {
+ table = &d->cvb_table[i];
+ if (!table->freq || (table->freq > max_freq))
+ break;
+
+ dfll_mv = get_cvb_voltage(
+ speedo_value, d->speedo_scale, &table->coefficients);
+ dfll_mv = round_cvb_voltage(dfll_mv, d->voltage_scale, align);
+ dfll_mv = clamp(dfll_mv, min_mv, max_mv);
+
+ ret = dev_pm_opp_add(opp_dev, table->freq, dfll_mv * 1000);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * tegra_cvb_build_opp_table - build OPP table from Tegra CVB tables
+ * @cvb_tables: array of CVB tables
+ * @sz: size of the previously mentioned array
+ * @process_id: process id of the HW module
+ * @speedo_id: speedo id of the HW module
+ * @speedo_value: speedo value of the HW module
+ * @max_rate: highest safe clock rate
+ * @opp_dev: the struct device * for which the OPP table is built
+ *
+ * On Tegra, a CVB table encodes the relationship between operating voltage
+ * and safe maximal frequency for a given module (e.g. GPU or CPU). This
+ * function calculates the optimal voltage-frequency operating points
+ * for the given arguments and exports them via the OPP library for the
+ * given @opp_dev. Returns a pointer to the struct cvb_table that matched
+ * or an ERR_PTR on failure.
+ */
+const struct cvb_table *tegra_cvb_build_opp_table(
+ const struct cvb_table *cvb_tables,
+ size_t sz, int process_id,
+ int speedo_id, int speedo_value,
+ unsigned long max_rate,
+ struct device *opp_dev)
+{
+ int i, ret;
+
+ for (i = 0; i < sz; i++) {
+ const struct cvb_table *d = &cvb_tables[i];
+
+ if (d->speedo_id != -1 && d->speedo_id != speedo_id)
+ continue;
+ if (d->process_id != -1 && d->process_id != process_id)
+ continue;
+
+ ret = build_opp_table(d, speedo_value, max_rate, opp_dev);
+ return ret ? ERR_PTR(ret) : d;
+ }
+
+ return ERR_PTR(-EINVAL);
+}
diff --git a/drivers/clk/tegra/cvb.h b/drivers/clk/tegra/cvb.h
new file mode 100644
index 0000000..f62cdc4
--- /dev/null
+++ b/drivers/clk/tegra/cvb.h
@@ -0,0 +1,67 @@
+/*
+ * Utility functions for parsing Tegra CVB voltage tables
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#ifndef __DRIVERS_CLK_TEGRA_CVB_H
+#define __DRIVERS_CLK_TEGRA_CVB_H
+
+#include <linux/types.h>
+
+struct device;
+
+#define MAX_DVFS_FREQS 40
+
+struct rail_alignment {
+ int offset_uv;
+ int step_uv;
+};
+
+struct cvb_coefficients {
+ int c0;
+ int c1;
+ int c2;
+};
+
+struct cvb_table_freq_entry {
+ unsigned long freq;
+ struct cvb_coefficients coefficients;
+};
+
+struct cvb_cpu_dfll_data {
+ u32 tune0_low;
+ u32 tune0_high;
+ u32 tune1;
+};
+
+struct cvb_table {
+ int speedo_id;
+ int process_id;
+
+ int min_millivolts;
+ int max_millivolts;
+ struct rail_alignment alignment;
+
+ int speedo_scale;
+ int voltage_scale;
+ struct cvb_table_freq_entry cvb_table[MAX_DVFS_FREQS];
+ struct cvb_cpu_dfll_data cpu_dfll_data;
+};
+
+const struct cvb_table *tegra_cvb_build_opp_table(
+ const struct cvb_table *cvb_tables,
+ size_t sz, int process_id,
+ int speedo_id, int speedo_value,
+ unsigned long max_rate,
+ struct device *opp_dev);
+
+#endif
--
1.8.1.5

2014-07-10 21:43:47

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 08/13] clk: tegra: Add the DFLL as a possible parent of the cclk_g clock

The DFLL clocksource was missing from the list of possible parents for
the fast CPU cluster. Add it to the list.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
drivers/clk/tegra/clk-tegra-super-gen4.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/clk/tegra/clk-tegra-super-gen4.c b/drivers/clk/tegra/clk-tegra-super-gen4.c
index feb3201..f1f4410 100644
--- a/drivers/clk/tegra/clk-tegra-super-gen4.c
+++ b/drivers/clk/tegra/clk-tegra-super-gen4.c
@@ -44,7 +44,9 @@ static const char *sclk_parents[] = { "clk_m", "pll_c_out1", "pll_p_out4",

static const char *cclk_g_parents[] = { "clk_m", "pll_c", "clk_32k", "pll_m",
"pll_p", "pll_p_out4", "unused",
- "unused", "pll_x" };
+ "unused", "pll_x", "unused", "unused",
+ "unused", "unused", "unused", "unused",
+ "dfllCPU_out" };

static const char *cclk_lp_parents[] = { "clk_m", "pll_c", "clk_32k", "pll_m",
"pll_p", "pll_p_out4", "unused",
--
1.8.1.5

2014-07-10 21:43:57

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 11/13] cpufreq: tegra124: Add device tree bindings

The cpufreq driver for Tegra124 will be a different one than the old
Tegra20 cpufreq driver (tegra-cpufreq), which does not use the device
tree.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
.../bindings/cpufreq/tegra124-cpufreq.txt | 37 ++++++++++++++++++++++
1 file changed, 37 insertions(+)
create mode 100644 Documentation/devicetree/bindings/cpufreq/tegra124-cpufreq.txt

diff --git a/Documentation/devicetree/bindings/cpufreq/tegra124-cpufreq.txt b/Documentation/devicetree/bindings/cpufreq/tegra124-cpufreq.txt
new file mode 100644
index 0000000..6f3f608
--- /dev/null
+++ b/Documentation/devicetree/bindings/cpufreq/tegra124-cpufreq.txt
@@ -0,0 +1,37 @@
+Tegra124 CPU frequency scaling driver bindings
+----------------------------------------------
+
+Both required and optional properties listed below must be defined
+under node /cpus/cpu@0.
+
+Required properties:
+- clocks: Must contain an entry for each entry in clock-names.
+ See ../clocks/clock-bindings.txt for details.
+- clock-names: Must include the following entries:
+ - cpu_g: Clock mux for the fast CPU cluster.
+ - cpu_lp: Clock mux for the low-power CPU cluster.
+ - pll_x: Fast PLL clocksource.
+ - pll_p: Auxiliary PLL used during fast PLL rate changes.
+ - dfll: Fast DFLL clocksource that also automatically scales CPU voltage.
+
+Example:
+--------
+cpus {
+ #address-cells = <1>;
+ #size-cells = <0>;
+
+ cpu@0 {
+ device_type = "cpu";
+ compatible = "arm,cortex-a15";
+ reg = <0>;
+
+ clocks = <&tegra_car TEGRA124_CLK_CCLK_G>,
+ <&tegra_car TEGRA124_CLK_CCLK_LP>,
+ <&tegra_car TEGRA124_CLK_PLL_X>,
+ <&tegra_car TEGRA124_CLK_PLL_P>,
+ <&dfll>;
+ clock-names = "cpu_g", "cpu_lp", "pll_x", "pll_p", "dfll";
+ };
+
+ <...>
+};
--
1.8.1.5

2014-07-10 21:44:03

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

Add a new cpufreq driver for Tegra124. Instead of using the PLLX as
the CPU clocksource, switch immediately to the DFLL. It allows the use
of higher clock rates, and will automatically scale the CPU voltage as
well. We also rely on the DFLL driver to determine the CPU clock
frequencies that the chip supports, so that we can directly build a
cpufreq table with the OPP library helper dev_pm_opp_init_cpufreq_table.

This driver is a completely independent of the old cpufreq driver
(tegra-cpufreq), which is only used on Tegra20.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
A platform_driver_register() followed by platform_device_register_simple()
looks pretty weird, but being a platform device is required for probe
deferral. Any better ways to handle this?

drivers/cpufreq/Makefile | 1 +
drivers/cpufreq/tegra124-cpufreq.c | 221 +++++++++++++++++++++++++++++++++++++
2 files changed, 222 insertions(+)
create mode 100644 drivers/cpufreq/tegra124-cpufreq.c

diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
index db6d9a2..3437d24 100644
--- a/drivers/cpufreq/Makefile
+++ b/drivers/cpufreq/Makefile
@@ -76,6 +76,7 @@ obj-$(CONFIG_ARM_SA1100_CPUFREQ) += sa1100-cpufreq.o
obj-$(CONFIG_ARM_SA1110_CPUFREQ) += sa1110-cpufreq.o
obj-$(CONFIG_ARM_SPEAR_CPUFREQ) += spear-cpufreq.o
obj-$(CONFIG_ARM_TEGRA_CPUFREQ) += tegra-cpufreq.o
+obj-$(CONFIG_ARM_TEGRA_CPUFREQ) += tegra124-cpufreq.o
obj-$(CONFIG_ARM_VEXPRESS_SPC_CPUFREQ) += vexpress-spc-cpufreq.o

##################################################################################
diff --git a/drivers/cpufreq/tegra124-cpufreq.c b/drivers/cpufreq/tegra124-cpufreq.c
new file mode 100644
index 0000000..3e29af4
--- /dev/null
+++ b/drivers/cpufreq/tegra124-cpufreq.c
@@ -0,0 +1,221 @@
+/*
+ * Tegra 124 cpufreq driver
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/clk.h>
+#include <linux/cpufreq.h>
+#include <linux/cpu.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/types.h>
+
+static struct cpufreq_frequency_table *freq_table;
+
+static struct device *cpu_dev;
+static struct clk *cpu_clk;
+static struct clk *pllp_clk;
+static struct clk *pllx_clk;
+static struct clk *dfll_clk;
+
+static int tegra124_target(struct cpufreq_policy *policy, unsigned int index)
+{
+ unsigned long cpu_rate = freq_table[index].frequency;
+ int ret;
+
+ ret = clk_set_rate(dfll_clk, cpu_rate * 1000);
+ if (ret)
+ pr_err("Failed to set cpu frequency to %lu kHz\n", cpu_rate);
+
+ return ret;
+}
+
+static int tegra124_cpu_init(struct cpufreq_policy *policy)
+{
+ int ret;
+
+ clk_prepare_enable(cpu_clk);
+
+ /* FIXME: what's the actual transition time? */
+ ret = cpufreq_generic_init(policy, freq_table, 300 * 1000);
+ if (ret) {
+ clk_disable_unprepare(cpu_clk);
+ return ret;
+ }
+
+ policy->clk = cpu_clk;
+ policy->suspend_freq = freq_table[0].frequency;
+ return 0;
+}
+
+static int tegra124_cpu_exit(struct cpufreq_policy *policy)
+{
+ clk_disable_unprepare(cpu_clk);
+ return 0;
+}
+
+static int tegra124_cpu_switch_to_dfll(void)
+{
+ struct clk *original_cpu_clk_parent;
+ unsigned long rate;
+ struct dev_pm_opp *opp;
+ int ret;
+
+ rate = clk_get_rate(cpu_clk);
+ opp = dev_pm_opp_find_freq_ceil(cpu_dev, &rate);
+ if (IS_ERR(opp))
+ return PTR_ERR(opp);
+
+ ret = clk_set_rate(dfll_clk, rate);
+ if (ret)
+ return ret;
+
+ original_cpu_clk_parent = clk_get_parent(cpu_clk);
+ clk_set_parent(cpu_clk, pllp_clk);
+ if (ret)
+ return ret;
+
+ ret = clk_prepare_enable(dfll_clk);
+ if (ret)
+ goto out_switch_to_original_parent;
+
+ clk_set_parent(cpu_clk, dfll_clk);
+
+ return 0;
+
+out_switch_to_original_parent:
+ clk_set_parent(cpu_clk, original_cpu_clk_parent);
+
+ return ret;
+}
+
+static struct cpufreq_driver tegra124_cpufreq_driver = {
+ .verify = cpufreq_generic_frequency_table_verify,
+ .target_index = tegra124_target,
+ .get = cpufreq_generic_get,
+ .init = tegra124_cpu_init,
+ .exit = tegra124_cpu_exit,
+ .name = "tegra124",
+ .attr = cpufreq_generic_attr,
+#ifdef CONFIG_PM
+ .suspend = cpufreq_generic_suspend,
+#endif
+};
+
+static int tegra124_cpufreq_probe(struct platform_device *pdev)
+{
+ int ret;
+
+ cpu_dev = get_cpu_device(0);
+ if (!cpu_dev)
+ return -ENODEV;
+
+ cpu_clk = of_clk_get_by_name(cpu_dev->of_node, "cpu_g");
+ if (IS_ERR(cpu_clk))
+ return PTR_ERR(cpu_clk);
+
+ dfll_clk = of_clk_get_by_name(cpu_dev->of_node, "dfll");
+ if (IS_ERR(dfll_clk)) {
+ ret = PTR_ERR(dfll_clk);
+ goto out_put_cpu_clk;
+ }
+
+ pllx_clk = of_clk_get_by_name(cpu_dev->of_node, "pll_x");
+ if (IS_ERR(pllx_clk)) {
+ ret = PTR_ERR(pllx_clk);
+ goto out_put_dfll_clk;
+ }
+
+ pllp_clk = of_clk_get_by_name(cpu_dev->of_node, "pll_p");
+ if (IS_ERR(pllp_clk)) {
+ ret = PTR_ERR(pllp_clk);
+ goto out_put_pllx_clk;
+ }
+
+ ret = dev_pm_opp_init_cpufreq_table(cpu_dev, &freq_table);
+ if (ret)
+ goto out_put_pllp_clk;
+
+ ret = tegra124_cpu_switch_to_dfll();
+ if (ret)
+ goto out_free_table;
+
+ ret = cpufreq_register_driver(&tegra124_cpufreq_driver);
+ if (ret) {
+ /*
+ * The VDD_CPU voltage may have been changed at this point and
+ * and switching back to PLLX might not be safe. Don't even try.
+ */
+ pr_err("failed to register cpufreq driver: %d\n", ret);
+ }
+
+ return ret;
+
+out_free_table:
+ dev_pm_opp_free_cpufreq_table(cpu_dev, &freq_table);
+out_put_pllp_clk:
+ clk_put(pllp_clk);
+out_put_pllx_clk:
+ clk_put(pllx_clk);
+out_put_dfll_clk:
+ clk_put(dfll_clk);
+out_put_cpu_clk:
+ clk_put(cpu_clk);
+
+ return ret;
+}
+
+static struct platform_driver tegra124_cpufreq_platdrv = {
+ .driver = {
+ .name = "cpufreq-tegra124",
+ .owner = THIS_MODULE,
+ },
+ .probe = tegra124_cpufreq_probe,
+};
+
+static const struct of_device_id soc_of_matches[] = {
+ { .compatible = "nvidia,tegra124", },
+ {}
+};
+
+static int __init tegra_cpufreq_init(void)
+{
+ int ret;
+ struct platform_device *pdev;
+
+ if (!of_find_matching_node(NULL, soc_of_matches))
+ return -ENODEV;
+
+ ret = platform_driver_register(&tegra124_cpufreq_platdrv);
+ if (ret)
+ return ret;
+
+ pdev = platform_device_register_simple("cpufreq-tegra124", -1, NULL, 0);
+ if (IS_ERR(pdev)) {
+ platform_driver_unregister(&tegra124_cpufreq_platdrv);
+ return PTR_ERR(pdev);
+ }
+
+ return 0;
+}
+
+MODULE_AUTHOR("Tuomas Tynkkynen <[email protected]>");
+MODULE_DESCRIPTION("cpufreq driver for nVIDIA Tegra124");
+MODULE_LICENSE("GPLv2");
+module_init(tegra_cpufreq_init);
--
1.8.1.5

2014-07-10 21:44:26

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 13/13] ARM: tegra: Add entries for cpufreq on Tegra124

The Tegra124 cpufreq driver relies on certain clocks being present
in the /cpus/cpu@0 node.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
arch/arm/boot/dts/tegra124.dtsi | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/arch/arm/boot/dts/tegra124.dtsi b/arch/arm/boot/dts/tegra124.dtsi
index 582a0e2..5e290bf 100644
--- a/arch/arm/boot/dts/tegra124.dtsi
+++ b/arch/arm/boot/dts/tegra124.dtsi
@@ -776,6 +776,13 @@
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0>;
+
+ clocks = <&tegra_car TEGRA124_CLK_CCLK_G>,
+ <&tegra_car TEGRA124_CLK_CCLK_LP>,
+ <&tegra_car TEGRA124_CLK_PLL_X>,
+ <&tegra_car TEGRA124_CLK_PLL_P>,
+ <&dfll>;
+ clock-names = "cpu_g", "cpu_lp", "pll_x", "pll_p", "dfll";
};

cpu@1 {
--
1.8.1.5

2014-07-10 21:45:20

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 10/13] ARM: tegra: Enable the DFLL on the Jetson TK1

Add the board-specific properties of the DFLL for the Jetson TK1 board.
On this board, the DFLL will take control of the sd0 regulator on the
on-board AS3722 PMIC.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
arch/arm/boot/dts/tegra124-jetson-tk1.dts | 83 ++++++++++++++++++++++++++++++-
1 file changed, 82 insertions(+), 1 deletion(-)

diff --git a/arch/arm/boot/dts/tegra124-jetson-tk1.dts b/arch/arm/boot/dts/tegra124-jetson-tk1.dts
index 89a772d..2dfd516 100644
--- a/arch/arm/boot/dts/tegra124-jetson-tk1.dts
+++ b/arch/arm/boot/dts/tegra124-jetson-tk1.dts
@@ -1453,7 +1453,7 @@
vin-ldo9-10-supply = <&vdd_5v0_sys>;
vin-ldo11-supply = <&vdd_3v3_run>;

- sd0 {
+ vdd_cpu: sd0 {
regulator-name = "+VDD_CPU_AP";
regulator-min-microvolt = <700000>;
regulator-max-microvolt = <1400000>;
@@ -1636,6 +1636,87 @@
non-removable;
};

+ dfll@0,70110000 {
+ status = "okay";
+ vdd_cpu-supply = <&vdd_cpu>;
+ nvidia,i2c-fs-rate = <400000>;
+ nvidia,pmic-i2c-address = <0x40>;
+ nvidia,pmic-i2c-voltage-register = <0x00>;
+
+ nvidia,pmic-voltage-table =
+ <0x1e 700000>,
+ <0x1f 710000>,
+ <0x20 720000>,
+ <0x21 730000>,
+ <0x22 740000>,
+ <0x23 750000>,
+ <0x24 760000>,
+ <0x25 770000>,
+ <0x26 780000>,
+ <0x27 790000>,
+ <0x28 800000>,
+ <0x29 810000>,
+ <0x2a 820000>,
+ <0x2b 830000>,
+ <0x2c 840000>,
+ <0x2d 850000>,
+ <0x2e 860000>,
+ <0x2f 870000>,
+ <0x30 880000>,
+ <0x31 890000>,
+ <0x32 900000>,
+ <0x33 910000>,
+ <0x34 920000>,
+ <0x35 930000>,
+ <0x36 940000>,
+ <0x37 950000>,
+ <0x38 960000>,
+ <0x39 970000>,
+ <0x3a 980000>,
+ <0x3b 990000>,
+ <0x3c 1000000>,
+ <0x3d 1010000>,
+ <0x3e 1020000>,
+ <0x3f 1030000>,
+ <0x40 1040000>,
+ <0x41 1050000>,
+ <0x42 1060000>,
+ <0x43 1070000>,
+ <0x44 1080000>,
+ <0x45 1090000>,
+ <0x46 1100000>,
+ <0x47 1110000>,
+ <0x48 1120000>,
+ <0x49 1130000>,
+ <0x4a 1140000>,
+ <0x4b 1150000>,
+ <0x4c 1160000>,
+ <0x4d 1170000>,
+ <0x4e 1180000>,
+ <0x4f 1190000>,
+ <0x50 1200000>,
+ <0x51 1210000>,
+ <0x52 1220000>,
+ <0x53 1230000>,
+ <0x54 1240000>,
+ <0x55 1250000>,
+ <0x56 1260000>,
+ <0x57 1270000>,
+ <0x58 1280000>,
+ <0x59 1290000>,
+ <0x5a 1300000>,
+ <0x5b 1310000>,
+ <0x5c 1320000>,
+ <0x5d 1330000>,
+ <0x5e 1340000>,
+ <0x5f 1350000>,
+ <0x60 1360000>,
+ <0x61 1370000>,
+ <0x62 1380000>,
+ <0x63 1390000>,
+ <0x64 1400000>;
+ };
+
ahub@0,70300000 {
i2s@0,70301100 {
status = "okay";
--
1.8.1.5

2014-07-10 21:43:44

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 07/13] clk: tegra: Save/restore CCLKG_BURST_POLICY on suspend

Save and restore this register since the LP1 restore assembly routines
fiddle with it. Otherwise the CPU would keep running on PLLX after
resume from suspend even when DFLL was the original clocksource.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
drivers/clk/tegra/clk-tegra124.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)

diff --git a/drivers/clk/tegra/clk-tegra124.c b/drivers/clk/tegra/clk-tegra124.c
index 77fbf38..758a4cf 100644
--- a/drivers/clk/tegra/clk-tegra124.c
+++ b/drivers/clk/tegra/clk-tegra124.c
@@ -89,6 +89,8 @@
#define PMC_PLLM_WB0_OVERRIDE 0x1dc
#define PMC_PLLM_WB0_OVERRIDE_2 0x2b0

+#define CCLKG_BURST_POLICY 0x368
+
#define UTMIP_PLL_CFG2 0x488
#define UTMIP_PLL_CFG2_STABLE_COUNT(x) (((x) & 0xffff) << 6)
#define UTMIP_PLL_CFG2_ACTIVE_DLY_COUNT(x) (((x) & 0x3f) << 18)
@@ -121,6 +123,8 @@
#ifdef CONFIG_PM_SLEEP
static struct cpu_clk_suspend_context {
u32 clk_csite_src;
+ u32 cclkg_burst;
+ u32 cclkg_divider;
} tegra124_cpu_clk_sctx;
#endif

@@ -1318,12 +1322,22 @@ static void tegra124_cpu_clock_suspend(void)
tegra124_cpu_clk_sctx.clk_csite_src =
readl(clk_base + CLK_SOURCE_CSITE);
writel(3 << 30, clk_base + CLK_SOURCE_CSITE);
+
+ tegra124_cpu_clk_sctx.cclkg_burst =
+ readl(clk_base + CCLKG_BURST_POLICY);
+ tegra124_cpu_clk_sctx.cclkg_divider =
+ readl(clk_base + CCLKG_BURST_POLICY + 4);
}

static void tegra124_cpu_clock_resume(void)
{
writel(tegra124_cpu_clk_sctx.clk_csite_src,
clk_base + CLK_SOURCE_CSITE);
+
+ writel(tegra124_cpu_clk_sctx.cclkg_burst,
+ clk_base + CCLKG_BURST_POLICY);
+ writel(tegra124_cpu_clk_sctx.cclkg_divider,
+ clk_base + CCLKG_BURST_POLICY + 4);
}
#endif

--
1.8.1.5

2014-07-10 21:46:46

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 05/13] clk: tegra: Add DFLL DVCO reset control for Tegra124

From: Paul Walmsley <[email protected]>

The DVCO present in the DFLL IP block has a separate reset line,
exposed via the CAR IP block. This reset line is asserted upon SoC
reset. Unless something (such as the DFLL driver) deasserts this
line, the DVCO will not oscillate, although reads and writes to the
DFLL IP block will complete.

Thanks to Aleksandr Frid <[email protected]> for identifying this and
saving hours of debugging time.

Signed-off-by: Paul Walmsley <[email protected]>
[ttynkkynen: ported to tegra124 from tegra114]
Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
drivers/clk/tegra/clk-tegra124.c | 47 ++++++++++++++++++++++++++++++++++++++++
drivers/clk/tegra/clk.h | 3 +++
2 files changed, 50 insertions(+)

diff --git a/drivers/clk/tegra/clk-tegra124.c b/drivers/clk/tegra/clk-tegra124.c
index 80efe51..77fbf38 100644
--- a/drivers/clk/tegra/clk-tegra124.c
+++ b/drivers/clk/tegra/clk-tegra124.c
@@ -31,6 +31,9 @@
#define CLK_SOURCE_CSITE 0x1d4
#define CLK_SOURCE_EMC 0x19c

+#define RST_DFLL_DVCO 0x2f4
+#define DVFS_DFLL_RESET_SHIFT 0
+
#define PLLC_BASE 0x80
#define PLLC_OUT 0x84
#define PLLC_MISC2 0x88
@@ -1378,6 +1381,50 @@ static void __init tegra124_clock_apply_init_table(void)
tegra_init_from_table(init_table, clks, TEGRA124_CLK_CLK_MAX);
}

+/**
+ * tegra124_car_barrier - wait for pending writes to the CAR to complete
+ *
+ * Wait for any outstanding writes to the CAR MMIO space from this CPU
+ * to complete before continuing execution. No return value.
+ */
+static void tegra124_car_barrier(void)
+{
+ readl_relaxed(clk_base + RST_DFLL_DVCO);
+}
+
+/**
+ * tegra124_clock_assert_dfll_dvco_reset - assert the DFLL's DVCO reset
+ *
+ * Assert the reset line of the DFLL's DVCO. No return value.
+ */
+void tegra124_clock_assert_dfll_dvco_reset(void)
+{
+ u32 v;
+
+ v = readl_relaxed(clk_base + RST_DFLL_DVCO);
+ v |= (1 << DVFS_DFLL_RESET_SHIFT);
+ writel_relaxed(v, clk_base + RST_DFLL_DVCO);
+ tegra124_car_barrier();
+}
+EXPORT_SYMBOL(tegra124_clock_assert_dfll_dvco_reset);
+
+/**
+ * tegra124_clock_deassert_dfll_dvco_reset - deassert the DFLL's DVCO reset
+ *
+ * Deassert the reset line of the DFLL's DVCO, allowing the DVCO to
+ * operate. No return value.
+ */
+void tegra124_clock_deassert_dfll_dvco_reset(void)
+{
+ u32 v;
+
+ v = readl_relaxed(clk_base + RST_DFLL_DVCO);
+ v &= ~(1 << DVFS_DFLL_RESET_SHIFT);
+ writel_relaxed(v, clk_base + RST_DFLL_DVCO);
+ tegra124_car_barrier();
+}
+EXPORT_SYMBOL(tegra124_clock_deassert_dfll_dvco_reset);
+
static void __init tegra124_clock_init(struct device_node *np)
{
struct device_node *node;
diff --git a/drivers/clk/tegra/clk.h b/drivers/clk/tegra/clk.h
index 16ec8d6..4b3f3d0 100644
--- a/drivers/clk/tegra/clk.h
+++ b/drivers/clk/tegra/clk.h
@@ -627,6 +627,9 @@ void tegra114_clock_tune_cpu_trimmers_init(void);
void tegra114_clock_assert_dfll_dvco_reset(void);
void tegra114_clock_deassert_dfll_dvco_reset(void);

+void tegra124_clock_assert_dfll_dvco_reset(void);
+void tegra124_clock_deassert_dfll_dvco_reset(void);
+
typedef void (*tegra_clk_apply_init_table_func)(void);
extern tegra_clk_apply_init_table_func tegra_clk_apply_init_table;

--
1.8.1.5

2014-07-10 21:47:09

by Tuomas Tynkkynen

[permalink] [raw]
Subject: [PATCH 03/13] clk: tegra: Add closed loop support for the DFLL

With closed loop support, the clock rate of the DFLL can be adjusted.

The oscillator itself in the DFLL is a free-running oscillator whose
rate is directly determined the supply voltage. However, the DFLL
module contains logic to compare the DFLL output rate to a fixed
reference clock (51 MHz) and make a decision to either lower or raise
the DFLL supply voltage. The DFLL module can then autonomously change
the supply voltage by communicating with an off-chip PMIC via either I2C
or PWM signals. This driver currently supports only I2C.

Signed-off-by: Tuomas Tynkkynen <[email protected]>
---
drivers/clk/tegra/clk-dfll.c | 691 ++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 688 insertions(+), 3 deletions(-)

diff --git a/drivers/clk/tegra/clk-dfll.c b/drivers/clk/tegra/clk-dfll.c
index 5eca4ff..024e71c 100644
--- a/drivers/clk/tegra/clk-dfll.c
+++ b/drivers/clk/tegra/clk-dfll.c
@@ -202,12 +202,16 @@
*/
#define REF_CLOCK_RATE 51000000UL

+#define DVCO_RATE_TO_MULT(rate, ref_rate) ((rate) / ((ref_rate) / 2))
+#define MULT_TO_DVCO_RATE(mult, ref_rate) ((mult) * ((ref_rate) / 2))

/**
* enum dfll_ctrl_mode - DFLL hardware operating mode
* @DFLL_UNINITIALIZED: (uninitialized state - not in hardware bitfield)
* @DFLL_DISABLED: DFLL not generating an output clock
* @DFLL_OPEN_LOOP: DVCO running, but DFLL not adjusting voltage
+ * @DFLL_CLOSED_LOOP: DVCO running, and DFLL adjusting voltage to match
+ * the requested rate
*
* The integer corresponding to the last two states, minus one, is
* written to the DFLL hardware to change operating modes.
@@ -216,6 +220,7 @@ enum dfll_ctrl_mode {
DFLL_UNINITIALIZED = 0,
DFLL_DISABLED = 1,
DFLL_OPEN_LOOP = 2,
+ DFLL_CLOSED_LOOP = 3,
};

/**
@@ -233,6 +238,35 @@ enum dfll_tune_range {
DFLL_TUNE_LOW = 1,
};

+/**
+ * struct voltage_reg_map - mapping from voltage to PMIC voltage register value
+ *
+ * The DFLL hardware is able to make I2C transactions to the PMIC to change
+ * the supply voltage of the DFLL, so the driver needs to know what voltage
+ * register value (@reg_value) needs to be used for a particular DFLL rail
+ * voltage (@microvolts).
+ */
+struct voltage_reg_map {
+ u8 reg_value;
+ int microvolts;
+};
+
+/**
+ * struct dfll_rate_req - target DFLL rate request data
+ * @rate: target frequency, after the postscaling
+ * @dvco_target_rate: target frequency, after the postscaling
+ * @lut_index: LUT index at which voltage the dvco_target_rate will be reached
+ * @mult_bits: value to program to the MULT bits of the DFLL_FREQ_REQ register
+ * @scale_bits: value to program to the SCALE bits of the DFLL_FREQ_REQ register
+ */
+struct dfll_rate_req {
+ unsigned long rate;
+ unsigned long dvco_target_rate;
+ int lut_index;
+ u8 mult_bits;
+ u8 scale_bits;
+};
+
struct tegra_dfll {
struct device *dev;
struct tegra_dfll_soc_data *soc;
@@ -255,9 +289,28 @@ struct tegra_dfll {
struct dentry *debugfs_dir;
struct clk_hw dfll_clk_hw;
const char *output_clock_name;
+ struct dfll_rate_req last_req;

/* Parameters from DT */
u32 droop_ctrl;
+ u32 sample_rate;
+ u32 force_mode;
+ u32 cf;
+ u32 ci;
+ u32 cg;
+ bool cg_scale;
+
+ /* I2C interface parameters from DT */
+ u32 i2c_fs_rate;
+ u32 i2c_reg;
+ u32 i2c_slave_addr;
+ bool i2c_10_bit_addresses;
+ struct voltage_reg_map *pmic_vdd_map;
+ size_t pmic_vdd_map_size;
+
+ struct voltage_reg_map *i2c_lut[MAX_DFLL_VOLTAGES];
+ int i2c_lut_size;
+ u8 lut_min, lut_max, lut_safe;
};

#define clk_hw_to_dfll(_hw) container_of(_hw, struct tegra_dfll, dfll_clk_hw)
@@ -267,6 +320,7 @@ static const char * const mode_name[] = {
[DFLL_UNINITIALIZED] = "uninitialized",
[DFLL_DISABLED] = "disabled",
[DFLL_OPEN_LOOP] = "open_loop",
+ [DFLL_CLOSED_LOOP] = "closed_loop",
};

/*
@@ -490,6 +544,282 @@ static void dfll_set_mode(struct tegra_dfll *td,
}

/*
+ * DFLL-to-I2C controller interface
+ */
+
+/**
+ * dfll_i2c_set_output_enabled - enable/disable I2C PMIC voltage requests
+ * @td: DFLL instance
+ * @enable: whether to enable or disable the I2C voltage requests
+ *
+ * Set the master enable control for I2C control value updates. If disabled,
+ * then I2C control messages are inhibited, regardless of the DFLL mode.
+ */
+static int dfll_i2c_set_output_enabled(struct tegra_dfll *td, bool enable)
+{
+ u32 val;
+
+ val = dfll_i2c_readl(td, DFLL_OUTPUT_CFG);
+
+ if (enable)
+ val |= DFLL_OUTPUT_CFG_I2C_ENABLE;
+ else
+ val &= ~DFLL_OUTPUT_CFG_I2C_ENABLE;
+
+ dfll_i2c_writel(td, val, DFLL_OUTPUT_CFG);
+ dfll_i2c_wmb(td);
+
+ return 0;
+}
+
+/**
+ * dfll_load_lut - load the voltage lookup table
+ * @td: struct tegra_dfll *
+ *
+ * Load the voltage-to-PMIC register value lookup table into the DFLL
+ * IP block memory. Look-up tables can be loaded at any time.
+ */
+static void dfll_load_i2c_lut(struct tegra_dfll *td)
+{
+ int i;
+ u32 val;
+
+ val = td->i2c_lut[td->lut_min]->reg_value;
+ for (i = 0; i <= td->lut_min; i++)
+ __raw_writel(val, td->lut_base + i * 4);
+
+ for (; i < td->lut_max; i++) {
+ val = td->i2c_lut[i]->reg_value;
+ __raw_writel(val, td->lut_base + i * 4);
+ }
+
+ val = td->i2c_lut[td->lut_max]->reg_value;
+ for (; i < MAX_DFLL_VOLTAGES; i++)
+ __raw_writel(val, td->lut_base + i * 4);
+
+ dfll_i2c_wmb(td);
+}
+
+/**
+ * dfll_init_i2c_if - set up the DFLL's DFLL-I2C interface
+ * @td: DFLL instance
+ *
+ * During DFLL driver initialization, program the DFLL-I2C interface
+ * with the PMU slave address, vdd register offset, and transfer mode.
+ * This data is used by the DFLL to automatically construct I2C
+ * voltage-set commands, which are then passed to the DFLL's internal
+ * I2C controller.
+ */
+static void dfll_init_i2c_if(struct tegra_dfll *td)
+{
+ u32 val;
+
+ if (td->i2c_10_bit_addresses) {
+ val = td->i2c_slave_addr << DFLL_I2C_CFG_SLAVE_ADDR_SHIFT_10BIT;
+ val |= DFLL_I2C_CFG_SLAVE_ADDR_10;
+ } else {
+ val = td->i2c_slave_addr << DFLL_I2C_CFG_SLAVE_ADDR_SHIFT_7BIT;
+ }
+ val |= DFLL_I2C_CFG_SIZE_MASK;
+ val |= DFLL_I2C_CFG_ARB_ENABLE;
+ dfll_i2c_writel(td, val, DFLL_I2C_CFG);
+
+ dfll_i2c_writel(td, td->i2c_reg, DFLL_I2C_VDD_REG_ADDR);
+
+ val = DIV_ROUND_UP(td->i2c_clk_rate, td->i2c_fs_rate * 8);
+ BUG_ON(!val || (val > DFLL_I2C_CLK_DIVISOR_MASK));
+ val = (val - 1) << DFLL_I2C_CLK_DIVISOR_FS_SHIFT;
+
+ /* default hs divisor just in case */
+ val |= 1 << DFLL_I2C_CLK_DIVISOR_HS_SHIFT;
+ __raw_writel(val, td->i2c_controller_base + DFLL_I2C_CLK_DIVISOR);
+ dfll_i2c_wmb(td);
+}
+
+/**
+ * dfll_init_out_if - prepare DFLL-to-PMIC interface
+ * @td: DFLL instance
+ *
+ * During DFLL driver initialization or resume from context loss,
+ * disable the I2C command output to the PMIC, set safe voltage and
+ * output limits, and disable and clear limit interrupts.
+ */
+static void dfll_init_out_if(struct tegra_dfll *td)
+{
+ u32 val;
+
+ td->lut_min = 0;
+ td->lut_max = td->i2c_lut_size - 1;
+ td->lut_safe = td->lut_min + 1;
+
+ dfll_i2c_writel(td, 0, DFLL_OUTPUT_CFG);
+ val = (td->lut_safe << DFLL_OUTPUT_CFG_SAFE_SHIFT) |
+ (td->lut_max << DFLL_OUTPUT_CFG_MAX_SHIFT) |
+ (td->lut_min << DFLL_OUTPUT_CFG_MIN_SHIFT);
+ dfll_writel(td, val, DFLL_OUTPUT_CFG);
+ dfll_wmb(td);
+
+ dfll_writel(td, 0, DFLL_OUTPUT_FORCE);
+ dfll_i2c_writel(td, 0, DFLL_INTR_EN);
+ dfll_i2c_writel(td, DFLL_INTR_MAX_MASK | DFLL_INTR_MIN_MASK,
+ DFLL_INTR_STS);
+
+ dfll_load_i2c_lut(td);
+ dfll_init_i2c_if(td);
+}
+
+/*
+ * Set/get the DFLL's targeted output clock rate
+ */
+
+/**
+ * find_lut_index_for_rate - determine I2C LUT index for given DFLL rate
+ * @td: DFLL instance
+ * @rate: clock rate
+ *
+ * Determines the index of a I2C LUT entry for a voltage that approximately
+ * produces the given DFLL clock rate. This is used when forcing a value
+ * to the integrator during rate changes. Returns -ENOENT if a suitable
+ * LUT index is not found.
+ */
+static int find_lut_index_for_rate(struct tegra_dfll *td, unsigned long rate)
+{
+ struct dev_pm_opp *opp;
+ int i, uv;
+
+ opp = dev_pm_opp_find_freq_ceil(td->soc->opp_dev, &rate);
+ if (IS_ERR(opp))
+ return PTR_ERR(opp);
+ uv = dev_pm_opp_get_voltage(opp);
+
+ for (i = 0; i < td->i2c_lut_size; i++) {
+ if (td->i2c_lut[i]->microvolts == uv)
+ return i;
+ }
+
+ return -ENOENT;
+}
+
+/**
+ * dfll_calculate_rate_request - calculate DFLL parameters for a given rate
+ * @td: DFLL instance
+ * @req: DFLL-rate-request structure
+ * @rate: the desired DFLL rate
+ *
+ * Populate the DFLL-rate-request record @req fields with the scale_bits
+ * and mult_bits fields, based on the target input rate. Returns 0 upon
+ * success, or -EINVAL if the requested rate in req->rate is too high
+ * or low for the DFLL to generate.
+ */
+static int dfll_calculate_rate_request(struct tegra_dfll *td,
+ struct dfll_rate_req *req,
+ unsigned long rate)
+{
+ u32 val;
+
+ /*
+ * If requested rate is below the minimum DVCO rate, active the scaler.
+ * In the future the DVCO minimum voltage should be selected based on
+ * chip temperature and the actual minimum rate should be calibrated
+ * at runtime.
+ */
+ req->scale_bits = DFLL_FREQ_REQ_SCALE_MAX - 1;
+ if (rate < td->dvco_rate_min) {
+ int scale;
+
+ scale = DIV_ROUND_CLOSEST(rate / 1000 * DFLL_FREQ_REQ_SCALE_MAX,
+ td->dvco_rate_min / 1000);
+ if (!scale) {
+ dev_err(td->dev, "%s: Rate %lu is too low\n",
+ __func__, rate);
+ return -EINVAL;
+ }
+ req->scale_bits = scale - 1;
+ rate = td->dvco_rate_min;
+ }
+
+ /* Convert requested rate into frequency request and scale settings */
+ val = DVCO_RATE_TO_MULT(rate, td->ref_rate);
+ if (val > FREQ_MAX) {
+ dev_err(td->dev, "%s: Rate %lu is above dfll range\n",
+ __func__, rate);
+ return -EINVAL;
+ }
+ req->mult_bits = val;
+ req->dvco_target_rate = MULT_TO_DVCO_RATE(req->mult_bits, td->ref_rate);
+ req->rate = dfll_scale_dvco_rate(req->dvco_target_rate,
+ req->scale_bits);
+ req->lut_index = find_lut_index_for_rate(td, req->dvco_target_rate);
+ if (req->lut_index < 0)
+ return req->lut_index;
+
+ return 0;
+}
+
+/**
+ * dfll_set_frequency_request - start the frequency change operation
+ * @td: DFLL instance
+ * @req: rate request structure
+ *
+ * Tell the DFLL to try to change its output frequency to the
+ * frequency represented by @req. DFLL must be in closed-loop mode.
+ */
+static void dfll_set_frequency_request(struct tegra_dfll *td,
+ struct dfll_rate_req *req)
+{
+ u32 val = 0;
+ int force_val;
+ int coef = 128; /* FIXME: td->cg_scale? */;
+
+ force_val = (req->lut_index - td->lut_safe) * coef / td->cg;
+ force_val = clamp(force_val, FORCE_MIN, FORCE_MAX);
+
+ val |= req->mult_bits << DFLL_FREQ_REQ_MULT_SHIFT;
+ val |= req->scale_bits << DFLL_FREQ_REQ_SCALE_SHIFT;
+ val |= ((u32)force_val << DFLL_FREQ_REQ_FORCE_SHIFT) &
+ DFLL_FREQ_REQ_FORCE_MASK;
+ val |= DFLL_FREQ_REQ_FREQ_VALID | DFLL_FREQ_REQ_FORCE_ENABLE;
+
+ dfll_writel(td, val, DFLL_FREQ_REQ);
+ dfll_wmb(td);
+}
+
+/**
+ * tegra_dfll_request_rate - set the next rate for the DFLL to tune to
+ * @td: DFLL instance
+ * @rate: clock rate to target
+ *
+ * Convert the requested clock rate @rate into the DFLL control logic
+ * settings. In closed-loop mode, update new settings immediately to
+ * adjust DFLL output rate accordingly. Otherwise, just save them
+ * until the next switch to closed loop. Returns 0 upon success,
+ * -EPERM if the DFLL driver has not yet been initialized, or -EINVAL
+ * if @rate is outside the DFLL's tunable range.
+ */
+static int dfll_request_rate(struct tegra_dfll *td, unsigned long rate)
+{
+ int ret;
+ struct dfll_rate_req req;
+
+ if (td->mode == DFLL_UNINITIALIZED) {
+ dev_err(td->dev, "%s: Cannot set DFLL rate in %s mode\n",
+ __func__, mode_name[td->mode]);
+ return -EPERM;
+ }
+
+ ret = dfll_calculate_rate_request(td, &req, rate);
+ if (ret)
+ return ret;
+
+ td->last_req = req;
+
+ if (td->mode == DFLL_CLOSED_LOOP)
+ dfll_set_frequency_request(td, &td->last_req);
+
+ return 0;
+}
+
+/*
* DFLL enable/disable & open-loop <-> closed-loop transitions
*/

@@ -558,10 +888,79 @@ static void dfll_set_open_loop_config(struct tegra_dfll *td)
val |= DFLL_FREQ_REQ_SCALE_MASK;
val &= ~DFLL_FREQ_REQ_FORCE_ENABLE;
dfll_writel(td, val, DFLL_FREQ_REQ);
+ dfll_wmb(td);
+}
+
+/**
+ * tegra_dfll_lock - switch from open-loop to closed-loop mode
+ * @td: DFLL instance
+ *
+ * Switch from OPEN_LOOP state to CLOSED_LOOP state. Returns 0 upon success,
+ * -EINVAL if the DFLL's target rate hasn't been set yet, or -EPERM if the
+ * DFLL is not currently in open-loop mode.
+ */
+static int dfll_lock(struct tegra_dfll *td)
+{
+ struct dfll_rate_req *req = &td->last_req;
+
+ switch (td->mode) {
+ case DFLL_CLOSED_LOOP:
+ return 0;
+
+ case DFLL_OPEN_LOOP:
+ if (req->rate == 0) {
+ dev_err(td->dev, "%s: Cannot lock DFLL at rate 0\n",
+ __func__);
+ return -EINVAL;
+ }
+
+ dfll_i2c_set_output_enabled(td, true);
+ dfll_set_mode(td, DFLL_CLOSED_LOOP);
+ dfll_set_frequency_request(td, req);
+ return 0;
+
+ default:
+ BUG_ON(td->mode > DFLL_CLOSED_LOOP);
+ dev_err(td->dev, "%s: Cannot lock DFLL in %s mode\n",
+ __func__, mode_name[td->mode]);
+ return -EPERM;
+ }
+}
+
+/**
+ * tegra_dfll_unlock - switch from closed-loop to open-loop mode
+ * @td: DFLL instance
+ *
+ * Switch from CLOSED_LOOP state to OPEN_LOOP state. Returns 0 upon success,
+ * or -EPERM if the DFLL is not currently in open-loop mode.
+ */
+static int dfll_unlock(struct tegra_dfll *td)
+{
+ switch (td->mode) {
+ case DFLL_CLOSED_LOOP:
+ dfll_set_open_loop_config(td);
+ dfll_set_mode(td, DFLL_OPEN_LOOP);
+ dfll_i2c_set_output_enabled(td, false);
+ return 0;
+
+ case DFLL_OPEN_LOOP:
+ return 0;
+
+ default:
+ BUG_ON(td->mode > DFLL_CLOSED_LOOP);
+ dev_err(td->dev, "%s: Cannot unlock DFLL in %s mode\n",
+ __func__, mode_name[td->mode]);
+ return -EPERM;
+ }
}

/*
* Clock framework integration
+ *
+ * When the DFLL is being controlled by the CCF, always enter closed loop
+ * mode when the clk is enabled. This requires that a DFLL rate request
+ * has been set beforehand, which implies that a clk_set_rate() call is
+ * always required before a clk_enable().
*/

static int dfll_clk_is_enabled(struct clk_hw *hw)
@@ -574,21 +973,67 @@ static int dfll_clk_is_enabled(struct clk_hw *hw)
static int dfll_clk_enable(struct clk_hw *hw)
{
struct tegra_dfll *td = clk_hw_to_dfll(hw);
+ int ret;
+
+ ret = dfll_enable(td);
+ if (ret)
+ return ret;
+
+ ret = dfll_lock(td);
+ if (ret)
+ dfll_disable(td);

- return dfll_enable(td);
+ return ret;
}

static void dfll_clk_disable(struct clk_hw *hw)
{
struct tegra_dfll *td = clk_hw_to_dfll(hw);
+ int ret;
+
+ ret = dfll_unlock(td);
+ if (!ret)
+ dfll_disable(td);
+}
+
+static unsigned long dfll_clk_recalc_rate(struct clk_hw *hw,
+ unsigned long parent_rate)
+{
+ struct tegra_dfll *td = clk_hw_to_dfll(hw);
+
+ return td->last_req.rate;
+}
+
+static long dfll_clk_round_rate(struct clk_hw *hw,
+ unsigned long rate,
+ unsigned long *parent_rate)
+{
+ struct tegra_dfll *td = clk_hw_to_dfll(hw);
+ struct dfll_rate_req req;
+ int ret;
+
+ ret = dfll_calculate_rate_request(td, &req, rate);
+ if (ret)
+ return ret;
+
+ return req.rate;
+}
+
+static int dfll_clk_set_rate(struct clk_hw *hw, unsigned long rate,
+ unsigned long parent_rate)
+{
+ struct tegra_dfll *td = clk_hw_to_dfll(hw);

- dfll_disable(td);
+ return dfll_request_rate(td, rate);
}

static const struct clk_ops dfll_clk_ops = {
.is_enabled = dfll_clk_is_enabled,
.enable = dfll_clk_enable,
.disable = dfll_clk_disable,
+ .recalc_rate = dfll_clk_recalc_rate,
+ .round_rate = dfll_clk_round_rate,
+ .set_rate = dfll_clk_set_rate,
};

static struct clk_init_data dfll_clk_init_data = {
@@ -670,6 +1115,23 @@ static int attr_enable_set(void *data, u64 val)
DEFINE_SIMPLE_ATTRIBUTE(enable_fops, attr_enable_get, attr_enable_set,
"%llu\n");

+static int attr_lock_get(void *data, u64 *val)
+{
+ struct tegra_dfll *td = data;
+
+ *val = (td->mode == DFLL_CLOSED_LOOP);
+
+ return 0;
+}
+static int attr_lock_set(void *data, u64 val)
+{
+ struct tegra_dfll *td = data;
+
+ return val ? dfll_lock(td) : dfll_unlock(td);
+}
+DEFINE_SIMPLE_ATTRIBUTE(lock_fops, attr_lock_get, attr_lock_set,
+ "%llu\n");
+
static int attr_rate_get(void *data, u64 *val)
{
struct tegra_dfll *td = data;
@@ -678,7 +1140,14 @@ static int attr_rate_get(void *data, u64 *val)

return 0;
}
-DEFINE_SIMPLE_ATTRIBUTE(rate_fops, attr_rate_get, NULL, "%llu\n");
+
+static int attr_rate_set(void *data, u64 val)
+{
+ struct tegra_dfll *td = data;
+
+ return dfll_request_rate(td, val);
+}
+DEFINE_SIMPLE_ATTRIBUTE(rate_fops, attr_rate_get, attr_rate_set, "%llu\n");

static int attr_registers_show(struct seq_file *s, void *data)
{
@@ -740,6 +1209,10 @@ static int dfll_debug_init(struct tegra_dfll *td)
td->debugfs_dir, td, &enable_fops))
goto err_out;

+ if (!debugfs_create_file("lock", S_IRUGO,
+ td->debugfs_dir, td, &lock_fops))
+ goto err_out;
+
if (!debugfs_create_file("rate", S_IRUGO,
td->debugfs_dir, td, &rate_fops))
goto err_out;
@@ -771,6 +1244,19 @@ err_out:
*/
static void dfll_set_default_params(struct tegra_dfll *td)
{
+ u32 val;
+
+ val = DIV_ROUND_UP(td->ref_rate, td->sample_rate * 32);
+ BUG_ON(val > DFLL_CONFIG_DIV_MASK);
+ dfll_writel(td, val, DFLL_CONFIG);
+
+ val = (td->force_mode << DFLL_PARAMS_FORCE_MODE_SHIFT) |
+ (td->cf << DFLL_PARAMS_CF_PARAM_SHIFT) |
+ (td->ci << DFLL_PARAMS_CI_PARAM_SHIFT) |
+ (td->cg << DFLL_PARAMS_CG_PARAM_SHIFT) |
+ (td->cg_scale ? DFLL_PARAMS_CG_SCALE : 0);
+ dfll_writel(td, val, DFLL_PARAMS);
+
dfll_tune_low(td);
dfll_writel(td, td->droop_ctrl, DFLL_DROOP_CTRL);
dfll_writel(td, DFLL_MONITOR_CTRL_FREQ, DFLL_MONITOR_CTRL);
@@ -860,6 +1346,8 @@ static int dfll_init(struct tegra_dfll *td)

dfll_set_open_loop_config(td);

+ dfll_init_out_if(td);
+
pm_runtime_put_sync(td->dev);

return 0;
@@ -879,6 +1367,179 @@ di_err1:
* DT data fetch
*/

+/*
+ * Find a PMIC voltage register-to-voltage mapping for the given voltage.
+ * An exact voltage match is required.
+ */
+static struct voltage_reg_map *find_vdd_map_entry_exact(
+ struct tegra_dfll *td, int uV)
+{
+ int i;
+
+ for (i = 0; i < td->pmic_vdd_map_size; i++) {
+ if (uV == td->pmic_vdd_map[i].microvolts)
+ return &td->pmic_vdd_map[i];
+ }
+
+ dev_err(td->dev, "no voltage map entry for %d uV\n", uV);
+ return NULL;
+}
+
+/*
+ * Find a PMIC voltage register-to-voltage mapping for the given voltage,
+ * rounding up to the closest supported voltage.
+ * */
+static struct voltage_reg_map *find_vdd_map_entry_min(
+ struct tegra_dfll *td, int uV)
+{
+ int i;
+
+ for (i = 0; i < td->pmic_vdd_map_size; i++) {
+ if (uV <= td->pmic_vdd_map[i].microvolts)
+ return &td->pmic_vdd_map[i];
+ }
+
+ dev_err(td->dev, "no voltage map entry rounding to %d uV\n", uV);
+ return NULL;
+}
+
+/**
+ * dfll_build_i2c_lut - build the I2C voltage register lookup table
+ * @td: DFLL instance
+ *
+ * The DFLL hardware has 33 bytes of look-up table RAM that must be filled with
+ * PMIC voltage register values that span the entire DFLL operating range.
+ * This function builds the look-up table based on the OPP table provided by
+ * the soc-specific platform driver (td->soc->opp_dev) and the PMIC
+ * register-to-voltage mapping (td->pmic_vdd_map) parsed from the DT.
+ *
+ * On success, fills in td->i2c_lut and returns 0, or -err on failure.
+ */
+static int dfll_build_i2c_lut(struct tegra_dfll *td)
+{
+ int ret = -EINVAL;
+ int j, v, v_max, v_opp;
+ unsigned long rate;
+ struct voltage_reg_map *m;
+ struct dev_pm_opp *opp;
+
+ rcu_read_lock();
+
+ rate = ULONG_MAX;
+ opp = dev_pm_opp_find_freq_floor(td->soc->opp_dev, &rate);
+ if (IS_ERR(opp)) {
+ dev_err(td->dev, "couldn't get vmax opp, empty opp table?\n");
+ goto out;
+ }
+ v_max = dev_pm_opp_get_voltage(opp);
+
+ v = td->soc->min_millivolts * 1000;
+ td->i2c_lut[0] = find_vdd_map_entry_exact(td, v);
+ if (!td->i2c_lut[0])
+ goto out;
+
+ for (j = 1, rate = 0; ; rate++) {
+ opp = dev_pm_opp_find_freq_ceil(td->soc->opp_dev, &rate);
+ if (IS_ERR(opp))
+ break;
+ v_opp = dev_pm_opp_get_voltage(opp);
+
+ if (v_opp <= td->soc->min_millivolts * 1000)
+ td->dvco_rate_min = dev_pm_opp_get_freq(opp);
+
+ for (;;) {
+ v += max(1, (v_max - v) / (MAX_DFLL_VOLTAGES - j));
+ if (v >= v_opp)
+ break;
+
+ m = find_vdd_map_entry_min(td, v);
+ if (!m)
+ goto out;
+ if (m != td->i2c_lut[j - 1])
+ td->i2c_lut[j++] = m;
+ }
+
+ v = (j == MAX_DFLL_VOLTAGES - 1) ? v_max : v_opp;
+ m = find_vdd_map_entry_exact(td, v);
+ if (!m)
+ goto out;
+ if (m != td->i2c_lut[j - 1])
+ td->i2c_lut[j++] = m;
+
+ if (v >= v_max)
+ break;
+ }
+ td->i2c_lut_size = j;
+
+ if (!td->dvco_rate_min)
+ dev_err(td->dev, "no opp above DFLL minimum voltage %d mV\n",
+ td->soc->min_millivolts);
+ else
+ ret = 0;
+
+out:
+ rcu_read_unlock();
+
+ return ret;
+}
+
+/**
+ * dfll_parse_vdd_map - parse the I2C PMIC voltage table from the device tree
+ * @td: DFLL instance
+ *
+ * Reads an array of 2-tuple entries from the nvidia,pmic-voltage-table
+ * property. Each entry should have the form <register-value voltage-in-uV>,
+ * indicating the register value that needs to be programmed to the PMIC for
+ * changing the VDD_CPU voltage to the given voltage. The array should be
+ * sorted in ascending order of voltage. For example:
+ *
+ * nvidia,pmic-voltage-table =
+ * <0x1e 700000>,
+ * <0x1f 710000>,
+ * ...
+ * <0x63 1390000>,
+ * <0x64 1400000>;
+ *
+ * On success, fills in td->pmic_vdd_map and td->pmic_vdd_map_size and
+ * returns true; on failure, return false.
+ */
+static bool dfll_parse_vdd_map(struct tegra_dfll *td)
+{
+ const struct property *prop;
+ const __be32 *val;
+ size_t i, num_longs, num_entries;
+ struct voltage_reg_map *map;
+
+ prop = of_find_property(td->dev->of_node,
+ "nvidia,pmic-voltage-table", NULL);
+ if (!prop || !prop->value) {
+ dev_err(td->dev, "missing nvidia,pmic-voltage-table property");
+ return false;
+ }
+
+ num_longs = prop->length / sizeof(u32);
+ if (num_longs % 2) {
+ dev_err(td->dev, "odd number of entries in voltage table");
+ return false;
+ }
+ num_entries = num_longs / 2;
+
+ map = devm_kcalloc(td->dev, num_entries, sizeof(*map), GFP_KERNEL);
+ if (!map)
+ return false;
+
+ val = prop->value;
+ for (i = 0; i < num_entries; i++) {
+ map[i].reg_value = be32_to_cpup(val++);
+ map[i].microvolts = be32_to_cpup(val++);
+ }
+
+ td->pmic_vdd_map = map;
+ td->pmic_vdd_map_size = num_entries;
+
+ return true;
+}
+
/**
* read_dt_param - helper function for reading required parameters from the DT
* @td: DFLL instance
@@ -914,6 +1575,24 @@ static int dfll_parse_dt_params(struct tegra_dfll *td)
bool ok = true;

ok &= read_dt_param(td, "nvidia,droop-ctrl", &td->droop_ctrl);
+ ok &= read_dt_param(td, "nvidia,sample-rate", &td->sample_rate);
+ ok &= read_dt_param(td, "nvidia,force-mode", &td->force_mode);
+ ok &= read_dt_param(td, "nvidia,cf", &td->cf);
+ ok &= read_dt_param(td, "nvidia,ci", &td->ci);
+ ok &= read_dt_param(td, "nvidia,cg", &td->cg);
+ td->cg_scale = of_property_read_bool(td->dev->of_node,
+ "nvidia,cg-scale");
+
+ ok &= read_dt_param(td, "nvidia,i2c-fs-rate", &td->i2c_fs_rate);
+ td->i2c_10_bit_addresses = of_property_read_bool(td->dev->of_node,
+ "nvidia,i2c-10-bit-addresses");
+
+ ok &= read_dt_param(td, "nvidia,pmic-i2c-address",
+ &td->i2c_slave_addr);
+ ok &= read_dt_param(td, "nvidia,pmic-i2c-voltage-register",
+ &td->i2c_reg);
+
+ ok &= dfll_parse_vdd_map(td);

if (of_property_read_string(td->dev->of_node, "clock-output-names",
&td->output_clock_name)) {
@@ -962,6 +1641,12 @@ int tegra_dfll_register(struct platform_device *pdev,
return ret;
}

+ ret = dfll_build_i2c_lut(td);
+ if (ret) {
+ dev_err(td->dev, "couldn't build I2C LUT\n");
+ return ret;
+ }
+
mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
if (!mem) {
dev_err(td->dev, "no control register resource\n");
--
1.8.1.5

2014-07-11 04:35:59

by Viresh Kumar

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

Hi Tuomas,

On 11 July 2014 03:12, Tuomas Tynkkynen <[email protected]> wrote:
> Add a new cpufreq driver for Tegra124. Instead of using the PLLX as
> the CPU clocksource, switch immediately to the DFLL. It allows the use
> of higher clock rates, and will automatically scale the CPU voltage as
> well. We also rely on the DFLL driver to determine the CPU clock
> frequencies that the chip supports, so that we can directly build a
> cpufreq table with the OPP library helper dev_pm_opp_init_cpufreq_table.
>
> This driver is a completely independent of the old cpufreq driver
> (tegra-cpufreq), which is only used on Tegra20.
>
> Signed-off-by: Tuomas Tynkkynen <[email protected]>

Please reuse cpufreq-cpu0 instead of adding a new driver. Similar
is being adopted by all platforms now: krait, mvebu, etc..

2014-07-11 07:14:49

by Mikko Perttunen

[permalink] [raw]
Subject: Re: [PATCH 10/13] ARM: tegra: Enable the DFLL on the Jetson TK1

It might look nicer if the voltage table was in a separate file. A bit
of a border case, maybe.

On 11/07/14 00:42, Tuomas Tynkkynen wrote:
> ...
> + nvidia,pmic-voltage-table =
> + <0x1e 700000>,
> + <0x1f 710000>,
> + <0x20 720000>,
> + <0x21 730000>,
> + <0x22 740000>,
> + <0x23 750000>,
> + <0x24 760000>,
> + <0x25 770000>,
> + <0x26 780000>,
> + <0x27 790000>,
> + <0x28 800000>,
> + <0x29 810000>,
> + <0x2a 820000>,
> + <0x2b 830000>,
> + <0x2c 840000>,
> + <0x2d 850000>,
> + <0x2e 860000>,
> + <0x2f 870000>,
> + <0x30 880000>,
> + <0x31 890000>,
> + <0x32 900000>,
> + <0x33 910000>,
> + <0x34 920000>,
> + <0x35 930000>,
> + <0x36 940000>,
> + <0x37 950000>,
> + <0x38 960000>,
> + <0x39 970000>,
> + <0x3a 980000>,
> + <0x3b 990000>,
> + <0x3c 1000000>,
> + <0x3d 1010000>,
> + <0x3e 1020000>,
> + <0x3f 1030000>,
> + <0x40 1040000>,
> + <0x41 1050000>,
> + <0x42 1060000>,
> + <0x43 1070000>,
> + <0x44 1080000>,
> + <0x45 1090000>,
> + <0x46 1100000>,
> + <0x47 1110000>,
> + <0x48 1120000>,
> + <0x49 1130000>,
> + <0x4a 1140000>,
> + <0x4b 1150000>,
> + <0x4c 1160000>,
> + <0x4d 1170000>,
> + <0x4e 1180000>,
> + <0x4f 1190000>,
> + <0x50 1200000>,
> + <0x51 1210000>,
> + <0x52 1220000>,
> + <0x53 1230000>,
> + <0x54 1240000>,
> + <0x55 1250000>,
> + <0x56 1260000>,
> + <0x57 1270000>,
> + <0x58 1280000>,
> + <0x59 1290000>,
> + <0x5a 1300000>,
> + <0x5b 1310000>,
> + <0x5c 1320000>,
> + <0x5d 1330000>,
> + <0x5e 1340000>,
> + <0x5f 1350000>,
> + <0x60 1360000>,
> + <0x61 1370000>,
> + <0x62 1380000>,
> + <0x63 1390000>,
> + <0x64 1400000>;
> ...

2014-07-11 09:12:29

by Peter De Schrijver

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On Fri, Jul 11, 2014 at 06:35:56AM +0200, Viresh Kumar wrote:
> Hi Tuomas,
>
> On 11 July 2014 03:12, Tuomas Tynkkynen <[email protected]> wrote:
> > Add a new cpufreq driver for Tegra124. Instead of using the PLLX as
> > the CPU clocksource, switch immediately to the DFLL. It allows the use
> > of higher clock rates, and will automatically scale the CPU voltage as
> > well. We also rely on the DFLL driver to determine the CPU clock
> > frequencies that the chip supports, so that we can directly build a
> > cpufreq table with the OPP library helper dev_pm_opp_init_cpufreq_table.
> >
> > This driver is a completely independent of the old cpufreq driver
> > (tegra-cpufreq), which is only used on Tegra20.
> >
> > Signed-off-by: Tuomas Tynkkynen <[email protected]>
>
> Please reuse cpufreq-cpu0 instead of adding a new driver. Similar
> is being adopted by all platforms now: krait, mvebu, etc..

I don't think that's going to work? The voltage scaling is handled in hw.
cpufreq-cpu0 seems to assume explicit sw control of the vdd_cpu regulator?

Cheers,

Peter.

2014-07-11 09:14:10

by Viresh Kumar

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On 11 July 2014 14:42, Peter De Schrijver <[email protected]> wrote:
> I don't think that's going to work? The voltage scaling is handled in hw.

Ahh..

> cpufreq-cpu0 seems to assume explicit sw control of the vdd_cpu regulator?

Its optional. If it fails to find a regulator, it doesn't depend or
play with it.

2014-07-11 14:14:17

by Tuomas Tynkkynen

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124



On 11/07/14 07:35, Viresh Kumar wrote:
> Hi Tuomas,
>
> On 11 July 2014 03:12, Tuomas Tynkkynen <[email protected]> wrote:
>> Add a new cpufreq driver for Tegra124. Instead of using the PLLX as
>> the CPU clocksource, switch immediately to the DFLL. It allows the use
>> of higher clock rates, and will automatically scale the CPU voltage as
>> well. We also rely on the DFLL driver to determine the CPU clock
>> frequencies that the chip supports, so that we can directly build a
>> cpufreq table with the OPP library helper dev_pm_opp_init_cpufreq_table.
>>
>> This driver is a completely independent of the old cpufreq driver
>> (tegra-cpufreq), which is only used on Tegra20.
>>
>> Signed-off-by: Tuomas Tynkkynen <[email protected]>
>
> Please reuse cpufreq-cpu0 instead of adding a new driver. Similar
> is being adopted by all platforms now: krait, mvebu, etc..
>

Sure, I can do the CPU clock parent change first and then instantiate
the cpufreq-cpu0 driver, like highbank-cpufreq. That'll depend on the
patch 'cpufreq: cpu0: OPPs can be populated at runtime' from your
'Extend support beyond CPU0' series though, any idea when that patch
will land in?


--
nvpublic

2014-07-11 14:37:45

by Viresh Kumar

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On 11 July 2014 19:44, Tuomas Tynkkynen <[email protected]> wrote:
> Sure, I can do the CPU clock parent change first and then instantiate the
> cpufreq-cpu0 driver, like highbank-cpufreq. That'll depend on the patch
> 'cpufreq: cpu0: OPPs can be populated at runtime' from your 'Extend support
> beyond CPU0' series though, any idea when that patch will land in?

So, after your mail I had a chat with Rafael. As this patch happens to be a
fix, we will try to take it into 3.16 only (I am going to send few patches out
of that series separately in few mins). Otherwise 3.17 is where most of it
should go.

Thanks for your agreement to make our life easier :)

2014-07-11 14:57:48

by Thierry Reding

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On Fri, Jul 11, 2014 at 12:12:07PM +0300, Peter De Schrijver wrote:
> On Fri, Jul 11, 2014 at 06:35:56AM +0200, Viresh Kumar wrote:
> > Hi Tuomas,
> >
> > On 11 July 2014 03:12, Tuomas Tynkkynen <[email protected]> wrote:
> > > Add a new cpufreq driver for Tegra124. Instead of using the PLLX as
> > > the CPU clocksource, switch immediately to the DFLL. It allows the use
> > > of higher clock rates, and will automatically scale the CPU voltage as
> > > well. We also rely on the DFLL driver to determine the CPU clock
> > > frequencies that the chip supports, so that we can directly build a
> > > cpufreq table with the OPP library helper dev_pm_opp_init_cpufreq_table.
> > >
> > > This driver is a completely independent of the old cpufreq driver
> > > (tegra-cpufreq), which is only used on Tegra20.
> > >
> > > Signed-off-by: Tuomas Tynkkynen <[email protected]>
> >
> > Please reuse cpufreq-cpu0 instead of adding a new driver. Similar
> > is being adopted by all platforms now: krait, mvebu, etc..
>
> I don't think that's going to work? The voltage scaling is handled in hw.

Do we have to handle it in hardware or can we opt to do it in software,
too?

Thierry


Attachments:
(No filename) (1.16 kB)
(No filename) (819.00 B)
Download all attachments

2014-07-11 15:11:48

by Tuomas Tynkkynen

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On 11/07/14 17:57, Thierry Reding wrote:

>> I don't think that's going to work? The voltage scaling is handled in hw.
>
> Do we have to handle it in hardware or can we opt to do it in software,
> too?
>

With the PLLX, voltage scaling is done entirely in SW. With the DFLL,
it's possible to stay in open-loop mode and do it in SW, but there's
not much point in that.

--
nvpublic

2014-07-11 15:15:14

by Thierry Reding

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On Fri, Jul 11, 2014 at 06:11:41PM +0300, Tuomas Tynkkynen wrote:
> On 11/07/14 17:57, Thierry Reding wrote:
>
> >>I don't think that's going to work? The voltage scaling is handled in hw.
> >
> >Do we have to handle it in hardware or can we opt to do it in software,
> >too?
> >
>
> With the PLLX, voltage scaling is done entirely in SW. With the DFLL,
> it's possible to stay in open-loop mode and do it in SW, but there's
> not much point in that.

It's kind of ugly how we need to pass the address of the PMU and the
offset of the voltage control register to the DFLL which will then
initiate I2C transactions itself. I'm wondering if that plays well with
the I2C traffic originating from within the kernel.

Thierry


Attachments:
(No filename) (723.00 B)
(No filename) (819.00 B)
Download all attachments

2014-07-11 15:29:11

by Tuomas Tynkkynen

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124



On 11/07/14 18:15, Thierry Reding wrote:
> * PGP Signed by an unknown key
>
> On Fri, Jul 11, 2014 at 06:11:41PM +0300, Tuomas Tynkkynen wrote:
>> On 11/07/14 17:57, Thierry Reding wrote:
>>
>>>> I don't think that's going to work? The voltage scaling is handled in hw.
>>>
>>> Do we have to handle it in hardware or can we opt to do it in software,
>>> too?
>>>
>>
>> With the PLLX, voltage scaling is done entirely in SW. With the DFLL,
>> it's possible to stay in open-loop mode and do it in SW, but there's
>> not much point in that.
>
> It's kind of ugly how we need to pass the address of the PMU and the
> offset of the voltage control register to the DFLL which will then
> initiate I2C transactions itself. I'm wondering if that plays well with
> the I2C traffic originating from within the kernel.
>

On the hardware level, the two I2C controllers sharing the same pins
have knowledge of each other and won't start transmitting if the bus
is busy (something different from the usual I2C arbitration, that is).
I guess on the kernel side there could be a problem if the voltage
register is marked cached in the PMIC driver's regmap.

--
nvpublic

2014-07-11 15:33:00

by Mike Turquette

[permalink] [raw]
Subject: Re: [PATCH 00/13] Tegra124 CL-DVFS / DFLL clocksource, plus cpufreq

Quoting Tuomas Tynkkynen (2014-07-10 14:42:36)
> This series implements the DFLL/CL-DVFS clock source for the fast CPU
> cluster on Tegra124, and a cpufreq driver that uses the DFLL for
> clocking the CPU. Most of this is based on Paul Walmsley's public patch
> set from December 2013, which is available at
> http://comments.gmane.org/gmane.linux.ports.tegra/15273
>
> The DFLL clock hardware is a voltage-controlled oscillator plus
> control logic that compares the generated output clock with a
> 51 MHz reference clock, and can make decisions to either lower
> or raise the DFLL voltage to keep the output rate close to the
> software-requested rate. The voltage changes are done by
> communicating with an off-chip PMIC via either I2C or PWM.
> As the DFLL oscillator is powered via the CPU rail, using
> the DFLL as the CPU clocksource also gives us dynamic CPU
> voltage scaling.

Clock driver bits look good to me. CVB table stuff is pretty neat.

Regards,
Mike

>
> This series has been tested on the Jetson TK1 (Rev C). Before attempting
> to port this to the Venice2, do note that there are two versions of the
> AS3722 with different voltage tables for the CPU rail (and that Venice2
> does not have active cooling).
>
> Thanks,
> Tuomas
>
> Paul Walmsley (1):
> clk: tegra: Add DFLL DVCO reset control for Tegra124
>
> Tuomas Tynkkynen (12):
> clk: tegra: Add binding for the Tegra124 DFLL clocksource
> clk: tegra: Add library for the DFLL clock source (open-loop mode)
> clk: tegra: Add closed loop support for the DFLL
> clk: tegra: Add functions for parsing CVB tables
> clk: tegra: Add Tegra124 DFLL clocksource platform driver
> clk: tegra: Save/restore CCLKG_BURST_POLICY on suspend
> clk: tegra: Add the DFLL as a possible parent of the cclk_g clock
> ARM: tegra: Add the DFLL to Tegra124 device tree
> ARM: tegra: Enable the DFLL on the Jetson-TK1
> cpufreq: tegra124: Add device tree bindings
> cpufreq: Add cpufreq driver for Tegra124
> ARM: tegra: Add entries for cpufreq on Tegra124
>
> .../bindings/clock/nvidia,tegra124-dfll.txt | 86 +
> .../bindings/cpufreq/tegra124-cpufreq.txt | 37 +
> arch/arm/boot/dts/tegra124-jetson-tk1.dts | 83 +-
> arch/arm/boot/dts/tegra124.dtsi | 29 +
> arch/arm/mach-tegra/Kconfig | 1 +
> drivers/clk/tegra/Makefile | 3 +
> drivers/clk/tegra/clk-dfll.c | 1759 ++++++++++++++++++++
> drivers/clk/tegra/clk-dfll.h | 55 +
> drivers/clk/tegra/clk-tegra-super-gen4.c | 4 +-
> drivers/clk/tegra/clk-tegra124-dfll-fcpu.c | 156 ++
> drivers/clk/tegra/clk-tegra124.c | 61 +
> drivers/clk/tegra/clk.h | 3 +
> drivers/clk/tegra/cvb.c | 133 ++
> drivers/clk/tegra/cvb.h | 67 +
> drivers/cpufreq/Makefile | 1 +
> drivers/cpufreq/tegra124-cpufreq.c | 221 +++
> 16 files changed, 2697 insertions(+), 2 deletions(-)
> create mode 100644 Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
> create mode 100644 Documentation/devicetree/bindings/cpufreq/tegra124-cpufreq.txt
> create mode 100644 drivers/clk/tegra/clk-dfll.c
> create mode 100644 drivers/clk/tegra/clk-dfll.h
> create mode 100644 drivers/clk/tegra/clk-tegra124-dfll-fcpu.c
> create mode 100644 drivers/clk/tegra/cvb.c
> create mode 100644 drivers/clk/tegra/cvb.h
> create mode 100644 drivers/cpufreq/tegra124-cpufreq.c
>
> --
> 1.8.1.5
>

2014-07-11 16:28:14

by Andrew Bresticker

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Thu, Jul 10, 2014 at 2:42 PM, Tuomas Tynkkynen <[email protected]> wrote:
> The DFLL is the main clocksource for the fast CPU cluster on Tegra124
> and also provides automatic CPU rail voltage scaling as well. The DFLL
> is a separate IP block from the usual Tegra124 clock-and-reset
> controller, so it gets its own node in the device tree.

> diff --git a/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt

> +- nvidia,pmic-voltage-table: Array of 2-tuples. Each entry should have the
> + form <register-value voltage-in-uV>, indicating the register value that
> + needs to be programmed to the PMIC for changing the VDD_CPU voltage to
> + the specified voltage. The table must be in ascending order by the voltage.

Instead of listing the register values for each voltage in the DT,
can't you use regulator_list_voltage() to create this map?

2014-07-11 16:33:51

by Andrew Bresticker

[permalink] [raw]
Subject: Re: [PATCH 12/13] cpufreq: Add cpufreq driver for Tegra124

On Fri, Jul 11, 2014 at 8:29 AM, Tuomas Tynkkynen <[email protected]> wrote:
>
> On the hardware level, the two I2C controllers sharing the same pins
> have knowledge of each other and won't start transmitting if the bus
> is busy (something different from the usual I2C arbitration, that is).
> I guess on the kernel side there could be a problem if the voltage register
> is marked cached in the PMIC driver's regmap.

Yeah, in our tree we have a hack to disable regcache for SD0_VOLTAGE.
Other than the value reported to userspace being wrong, leaving it as
cacheable shouldn't be an issue if no other drivers try to read/write
that register.

2014-07-11 16:48:31

by Tuomas Tynkkynen

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource



On 11/07/14 19:28, Andrew Bresticker wrote:
> On Thu, Jul 10, 2014 at 2:42 PM, Tuomas Tynkkynen <[email protected]> wrote:
>> The DFLL is the main clocksource for the fast CPU cluster on Tegra124
>> and also provides automatic CPU rail voltage scaling as well. The DFLL
>> is a separate IP block from the usual Tegra124 clock-and-reset
>> controller, so it gets its own node in the device tree.
>
>> diff --git a/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
>
>> +- nvidia,pmic-voltage-table: Array of 2-tuples. Each entry should have the
>> + form <register-value voltage-in-uV>, indicating the register value that
>> + needs to be programmed to the PMIC for changing the VDD_CPU voltage to
>> + the specified voltage. The table must be in ascending order by the voltage.
>
> Instead of listing the register values for each voltage in the DT,
> can't you use regulator_list_voltage() to create this map?
>

I don't see a way to get the register values that way, unless we assume
that the mapping is linear and doesn't have holes.

--
nvpublic

2014-07-11 17:08:58

by Andrew Bresticker

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Fri, Jul 11, 2014 at 9:48 AM, Tuomas Tynkkynen <[email protected]> wrote:
>
>
> On 11/07/14 19:28, Andrew Bresticker wrote:
>>
>> On Thu, Jul 10, 2014 at 2:42 PM, Tuomas Tynkkynen <[email protected]>
>> wrote:
>>>
>>> The DFLL is the main clocksource for the fast CPU cluster on Tegra124
>>> and also provides automatic CPU rail voltage scaling as well. The DFLL
>>> is a separate IP block from the usual Tegra124 clock-and-reset
>>> controller, so it gets its own node in the device tree.
>>
>>
>>> diff --git
>>> a/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
>>> b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
>>
>>
>>> +- nvidia,pmic-voltage-table: Array of 2-tuples. Each entry should have
>>> the
>>> + form <register-value voltage-in-uV>, indicating the register value
>>> that
>>> + needs to be programmed to the PMIC for changing the VDD_CPU voltage to
>>> + the specified voltage. The table must be in ascending order by the
>>> voltage.
>>
>>
>> Instead of listing the register values for each voltage in the DT,
>> can't you use regulator_list_voltage() to create this map?
>>
>
> I don't see a way to get the register values that way, unless we assume that
> the mapping is linear and doesn't have holes.

Hmm... I guess if you don't assume it's linear and continuous you'd
have to iterate over all 256 selectors.

2014-07-11 17:21:34

by Tuomas Tynkkynen

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource



On 11/07/14 20:08, Andrew Bresticker wrote:
> On Fri, Jul 11, 2014 at 9:48 AM, Tuomas Tynkkynen <[email protected]> wrote:
>>
>>
>> On 11/07/14 19:28, Andrew Bresticker wrote:
>>>
>>> On Thu, Jul 10, 2014 at 2:42 PM, Tuomas Tynkkynen <[email protected]>
>>> wrote:
>>>>
>>>> The DFLL is the main clocksource for the fast CPU cluster on Tegra124
>>>> and also provides automatic CPU rail voltage scaling as well. The DFLL
>>>> is a separate IP block from the usual Tegra124 clock-and-reset
>>>> controller, so it gets its own node in the device tree.
>>>
>>>
>>>> diff --git
>>>> a/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
>>>> b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
>>>
>>>
>>>> +- nvidia,pmic-voltage-table: Array of 2-tuples. Each entry should have
>>>> the
>>>> + form <register-value voltage-in-uV>, indicating the register value
>>>> that
>>>> + needs to be programmed to the PMIC for changing the VDD_CPU voltage to
>>>> + the specified voltage. The table must be in ascending order by the
>>>> voltage.
>>>
>>>
>>> Instead of listing the register values for each voltage in the DT,
>>> can't you use regulator_list_voltage() to create this map?
>>>
>>
>> I don't see a way to get the register values that way, unless we assume that
>> the mapping is linear and doesn't have holes.
>
> Hmm... I guess if you don't assume it's linear and continuous you'd
> have to iterate over all 256 selectors.
>

I don't think we can assume that each selector maps to a concrete
register value, though I'm not sure. include/linux/regulator/driver.h
documents for @list_voltage "Selectors range from zero to one less
regulator_desc.n_voltages." but maybe the consumer API could take
different values.

--
nvpublic

2014-07-14 08:39:05

by Thierry Reding

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Fri, Jul 11, 2014 at 08:21:27PM +0300, Tuomas Tynkkynen wrote:
>
>
> On 11/07/14 20:08, Andrew Bresticker wrote:
> >On Fri, Jul 11, 2014 at 9:48 AM, Tuomas Tynkkynen <[email protected]> wrote:
> >>
> >>
> >>On 11/07/14 19:28, Andrew Bresticker wrote:
> >>>
> >>>On Thu, Jul 10, 2014 at 2:42 PM, Tuomas Tynkkynen <[email protected]>
> >>>wrote:
> >>>>
> >>>>The DFLL is the main clocksource for the fast CPU cluster on Tegra124
> >>>>and also provides automatic CPU rail voltage scaling as well. The DFLL
> >>>>is a separate IP block from the usual Tegra124 clock-and-reset
> >>>>controller, so it gets its own node in the device tree.
> >>>
> >>>
> >>>>diff --git
> >>>>a/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
> >>>>b/Documentation/devicetree/bindings/clock/nvidia,tegra124-dfll.txt
> >>>
> >>>
> >>>>+- nvidia,pmic-voltage-table: Array of 2-tuples. Each entry should have
> >>>>the
> >>>>+ form <register-value voltage-in-uV>, indicating the register value
> >>>>that
> >>>>+ needs to be programmed to the PMIC for changing the VDD_CPU voltage to
> >>>>+ the specified voltage. The table must be in ascending order by the
> >>>>voltage.
> >>>
> >>>
> >>>Instead of listing the register values for each voltage in the DT,
> >>>can't you use regulator_list_voltage() to create this map?
> >>>
> >>
> >>I don't see a way to get the register values that way, unless we assume that
> >>the mapping is linear and doesn't have holes.
> >
> >Hmm... I guess if you don't assume it's linear and continuous you'd
> >have to iterate over all 256 selectors.
> >
>
> I don't think we can assume that each selector maps to a concrete register
> value, though I'm not sure. include/linux/regulator/driver.h documents for
> @list_voltage "Selectors range from zero to one less
> regulator_desc.n_voltages." but maybe the consumer API could take different
> values.

I don't think the regulator API makes any guarantees that the selector
corresponds to a register value. Adding Mark Brown, maybe he can help
figure out the best way to do this.

Thierry


Attachments:
(No filename) (2.03 kB)
(No filename) (819.00 B)
Download all attachments

2014-07-14 09:13:31

by Mark Brown

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Mon, Jul 14, 2014 at 10:38:56AM +0200, Thierry Reding wrote:
> On Fri, Jul 11, 2014 at 08:21:27PM +0300, Tuomas Tynkkynen wrote:

> > I don't think we can assume that each selector maps to a concrete register
> > value, though I'm not sure. include/linux/regulator/driver.h documents for
> > @list_voltage "Selectors range from zero to one less
> > regulator_desc.n_voltages." but maybe the consumer API could take different
> > values.

> I don't think the regulator API makes any guarantees that the selector
> corresponds to a register value. Adding Mark Brown, maybe he can help
> figure out the best way to do this.

The selector value is opaque, it's entirely up to the driver to define
it. If you could tell me what "this" is I might be able to advise on
how to do it.


Attachments:
(No filename) (780.00 B)
signature.asc (819.00 B)
Digital signature
Download all attachments

2014-07-14 09:24:50

by Thierry Reding

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Mon, Jul 14, 2014 at 10:12:33AM +0100, Mark Brown wrote:
> On Mon, Jul 14, 2014 at 10:38:56AM +0200, Thierry Reding wrote:
> > On Fri, Jul 11, 2014 at 08:21:27PM +0300, Tuomas Tynkkynen wrote:
>
> > > I don't think we can assume that each selector maps to a concrete register
> > > value, though I'm not sure. include/linux/regulator/driver.h documents for
> > > @list_voltage "Selectors range from zero to one less
> > > regulator_desc.n_voltages." but maybe the consumer API could take different
> > > values.
>
> > I don't think the regulator API makes any guarantees that the selector
> > corresponds to a register value. Adding Mark Brown, maybe he can help
> > figure out the best way to do this.
>
> The selector value is opaque, it's entirely up to the driver to define
> it. If you could tell me what "this" is I might be able to advise on
> how to do it.

Tegra124 (and later, also some earlier variants) have this DFLL clock
that can program a PMIC automatically depending on the CPU frequency.
This DT binding did propose putting this into device tree as a table of
<frequency value> pairs where the frequency corresponds to the CPU
frequency and the value is the register value to be programmed into the
PMIC by the DFLL hardware (there are two additional properties to define
the slave address and the register offset).

Andrew proposed that this table could instead be built by using
regulator_list_voltage() instead. However, due to the fact that the DFLL
hardware needs to know the immediate value to write into a register, the
requirement would be for a 1:1 mapping between selector and register
value. Given that the API needs to cover the general case I don't see
how it could practically ensure this.

Thierry


Attachments:
(No filename) (1.70 kB)
(No filename) (819.00 B)
Download all attachments

2014-07-14 10:22:34

by Mark Brown

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Mon, Jul 14, 2014 at 11:24:35AM +0200, Thierry Reding wrote:
> On Mon, Jul 14, 2014 at 10:12:33AM +0100, Mark Brown wrote:

> > The selector value is opaque, it's entirely up to the driver to define
> > it. If you could tell me what "this" is I might be able to advise on
> > how to do it.

> Tegra124 (and later, also some earlier variants) have this DFLL clock
> that can program a PMIC automatically depending on the CPU frequency.
> This DT binding did propose putting this into device tree as a table of
> <frequency value> pairs where the frequency corresponds to the CPU
> frequency and the value is the register value to be programmed into the
> PMIC by the DFLL hardware (there are two additional properties to define
> the slave address and the register offset).

> Andrew proposed that this table could instead be built by using
> regulator_list_voltage() instead. However, due to the fact that the DFLL
> hardware needs to know the immediate value to write into a register, the
> requirement would be for a 1:1 mapping between selector and register
> value. Given that the API needs to cover the general case I don't see
> how it could practically ensure this.

Well, if you're going to do that you've already created a private API
between the regulator driver and the device since you're assuming that
the device is controlled only by register writes to a single register
bitfield which isn't always the case.

As with all these things it would also be better to extend the regulator
API so that users like this can discover the register address and so on
too rather than having to replicate that information in the device tree.
No sense in having to specify this information multiple times.


Attachments:
(No filename) (1.67 kB)
signature.asc (819.00 B)
Digital signature
Download all attachments

2014-07-15 20:23:47

by Tuomas Tynkkynen

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On 14/07/14 13:22, Mark Brown wrote:
> * PGP Signed by an unknown key
>
> On Mon, Jul 14, 2014 at 11:24:35AM +0200, Thierry Reding wrote:
>> On Mon, Jul 14, 2014 at 10:12:33AM +0100, Mark Brown wrote:
>
>>> The selector value is opaque, it's entirely up to the driver to define
>>> it. If you could tell me what "this" is I might be able to advise on
>>> how to do it.
>
>> Tegra124 (and later, also some earlier variants) have this DFLL clock
>> that can program a PMIC automatically depending on the CPU frequency.
>> This DT binding did propose putting this into device tree as a table of
>> <frequency value> pairs where the frequency corresponds to the CPU
>> frequency and the value is the register value to be programmed into the
>> PMIC by the DFLL hardware (there are two additional properties to define
>> the slave address and the register offset).
>
>> Andrew proposed that this table could instead be built by using
>> regulator_list_voltage() instead. However, due to the fact that the DFLL
>> hardware needs to know the immediate value to write into a register, the
>> requirement would be for a 1:1 mapping between selector and register
>> value. Given that the API needs to cover the general case I don't see
>> how it could practically ensure this.
>
> Well, if you're going to do that you've already created a private API
> between the regulator driver and the device since you're assuming that
> the device is controlled only by register writes to a single register
> bitfield which isn't always the case.
>
> As with all these things it would also be better to extend the regulator
> API so that users like this can discover the register address and so on
> too rather than having to replicate that information in the device tree.
> No sense in having to specify this information multiple times.
>

That sounds indeed useful for this case. How'd the following interface
sound for the register offset / selector-to-register-value conversion?
The I2C address would be a bit trickier to get as it would touch the
regmap stuff as well, but perhaps it would be a good idea to have a
phandle to the I2C device itself, and then parse the reg field for
the address.

diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
index c563d93..a5efb96 100644
--- a/drivers/regulator/core.c
+++ b/drivers/regulator/core.c
@@ -2228,6 +2228,63 @@ int regulator_list_voltage(struct regulator *regulator, unsigned selector)
EXPORT_SYMBOL_GPL(regulator_list_voltage);

/**
+ * regulator_get_hardware_vsel_register - get the HW voltage selector register
+ * @regulator: regulator source
+ * @vsel_reg: voltage selector register, output parameter
+ * @vsel_mask: mask for voltage selector bitfield, output parameter
+ *
+ * Returns the hardware register offset and bitmask used for setting the
+ * regulator voltage. This might be useful when configuring voltage-scaling
+ * hardware or firmware that can make I2C requests behind the kernel's back,
+ * for example.
+ *
+ * On success, the output parameters @vsel_reg and @vsel_mask are filled in
+ * and 0 is returned, otherwise a negative errno is returned.
+ */
+int regulator_get_hardware_vsel_register(struct regulator *regulator,
+ unsigned *vsel_reg,
+ unsigned *vsel_mask)
+{
+ struct regulator_dev *rdev = regulator->rdev;
+ struct regulator_ops *ops = rdev->desc->ops;
+
+ if (ops->set_voltage_sel != regulator_set_voltage_sel_regmap)
+ return -EOPNOTSUPP;
+
+ *vsel_reg = rdev->desc->vsel_reg;
+ *vsel_mask = rdev->desc->vsel_mask;
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(regulator_get_hardware_vsel_register);
+
+/**
+ * regulator_list_hardware_vsel - get the HW-specific register value for a selector
+ * @regulator: regulator source
+ * @selector: identify voltage to list
+ *
+ * Converts the selector to a hardware-specific voltage selector that can be
+ * directly written to the regulator registers. The address of the voltage
+ * register can be determined by calling @regulator_get_hardware_vsel_register.
+ *
+ * On error a negative errno is returned.
+ */
+int regulator_list_hardware_vsel(struct regulator *regulator,
+ unsigned selector)
+{
+ struct regulator_dev *rdev = regulator->rdev;
+ struct regulator_ops *ops = rdev->desc->ops;
+
+ if (selector >= rdev->desc->n_voltages)
+ return -EINVAL;
+ if (ops->set_voltage_sel != regulator_set_voltage_sel_regmap)
+ return -EOPNOTSUPP;
+
+ return selector;
+}
+EXPORT_SYMBOL_GPL(regulator_list_hardware_vsel);
+
+/**
* regulator_get_linear_step - return the voltage step size between VSEL values
* @regulator: regulator source
*
diff --git a/include/linux/regulator/consumer.h b/include/linux/regulator/consumer.h
index 14ec18d..fe4cdb2 100644
--- a/include/linux/regulator/consumer.h
+++ b/include/linux/regulator/consumer.h
@@ -215,6 +215,12 @@ int regulator_set_optimum_mode(struct regulator *regulator, int load_uA);

int regulator_allow_bypass(struct regulator *regulator, bool allow);

+int regulator_get_hardware_vsel_register(struct regulator *regulator,
+ unsigned *vsel_reg,
+ unsigned *vsel_mask);
+int regulator_list_hardware_vsel(struct regulator *regulator,
+ unsigned selector);
+
/* regulator notifier block */
int regulator_register_notifier(struct regulator *regulator,
struct notifier_block *nb);

--
nvpublic

2014-07-15 22:53:29

by Mark Brown

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Tue, Jul 15, 2014 at 11:23:39PM +0300, Tuomas Tynkkynen wrote:

> That sounds indeed useful for this case. How'd the following interface
> sound for the register offset / selector-to-register-value conversion?
> The I2C address would be a bit trickier to get as it would touch the
> regmap stuff as well, but perhaps it would be a good idea to have a
> phandle to the I2C device itself, and then parse the reg field for
> the address.

This looks fine, can you submit properly please? For the I2C address
why not just have an interface to get the regmap and then provide a way
to get the underlying device back from the regmap?


Attachments:
(No filename) (632.00 B)
signature.asc (819.00 B)
Digital signature
Download all attachments

2014-07-16 08:01:50

by Thierry Reding

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Tue, Jul 15, 2014 at 11:52:52PM +0100, Mark Brown wrote:
> On Tue, Jul 15, 2014 at 11:23:39PM +0300, Tuomas Tynkkynen wrote:
>
> > That sounds indeed useful for this case. How'd the following interface
> > sound for the register offset / selector-to-register-value conversion?
> > The I2C address would be a bit trickier to get as it would touch the
> > regmap stuff as well, but perhaps it would be a good idea to have a
> > phandle to the I2C device itself, and then parse the reg field for
> > the address.
>
> This looks fine, can you submit properly please? For the I2C address
> why not just have an interface to get the regmap and then provide a way
> to get the underlying device back from the regmap?

Is it mandatory for regulators to use regmap? Also I'm not sure how this
will work with MFDs, since the device may not be the actual bus device,
but rather a child of the MFD (and what we want access to is the MFD).
Perhaps for regmaps that would work in most cases since MFDs seem to
aften share the regmap with their children. However, the code in
regulator_register() at least indicates that even then it's possible to
have a regmap in children that's different from the top-level MFD.

Thierry


Attachments:
(No filename) (1.19 kB)
(No filename) (819.00 B)
Download all attachments

2014-07-16 11:00:49

by Mark Brown

[permalink] [raw]
Subject: Re: [PATCH 01/13] clk: tegra: Add binding for the Tegra124 DFLL clocksource

On Wed, Jul 16, 2014 at 10:01:34AM +0200, Thierry Reding wrote:
> On Tue, Jul 15, 2014 at 11:52:52PM +0100, Mark Brown wrote:

> > This looks fine, can you submit properly please? For the I2C address
> > why not just have an interface to get the regmap and then provide a way
> > to get the underlying device back from the regmap?

> Is it mandatory for regulators to use regmap? Also I'm not sure how this

No, but this is for a limited subset of devices that the hardware knows
how to write to directly which means that they're restricted to things
that can be supported with regmap (or already know how to interact with
the hardware and don't need this interface at all). The proposed patch
already relies on regmap - it's passing back the information the regmap
helpers use.

> will work with MFDs, since the device may not be the actual bus device,
> but rather a child of the MFD (and what we want access to is the MFD).

> Perhaps for regmaps that would work in most cases since MFDs seem to
> aften share the regmap with their children. However, the code in
> regulator_register() at least indicates that even then it's possible to
> have a regmap in children that's different from the top-level MFD.

Right, drivers can say exactly which regmap to use.


Attachments:
(No filename) (1.23 kB)
signature.asc (819.00 B)
Digital signature
Download all attachments