[v12]
* Addressed the kbuild fail on arm architecture
[v11]
* Split the series into domains
[v9]
* Addressed comments from Viresh and Russel about the error handling
[v8]
* Reordered the patch series into 4 groups
* Addressed comments from Amit about the comments and commit messages
* Addressed comments from Amit and Viresh about the resourses deallocation
[v7]
* Addressed comments from Viresh about resourses deallocation
and DT compatible
[v6]
* Addressed comments from Viresh about:
** Comments style
** Kconfig bool instead of tristate
** DT and documentation style
** Resourses deallocation on an error
** Typos
[v5]
* Rebased
* Addressed comments from Bjorn about SPDX style,
functions and parameters naming
* Addressed comments from Viresh DT properties and style, comments style,
resourses deallocation, documentation placement
* Addressed comments from Sricharan about unnessesary include
* Addressed comments from Nicolas
* Addressed comments from Rob about the commit messages and acks
* Addressed comments from Mark
[v4]
* Adressed all comments from Stephen
* Added CPU regulator support
* Added qcom-cpufreq-kryo driver
[v3]
* Rebased on top of the latest PLL driver changes
* Addressed comment from Rob Herring for bindings
[v2]
* Addressed comments from Rob Herring for bindings
* Addressed comments from Mark Rutland for memory barrier
* Addressed comments from Julien Thierry for clock reenabling condition
* Tuned the HW configuration for clock frequencies below 600MHz
SOC (1/15):
Extracts the kryo l2 accessors driver from the QCOM PMU driver
Clocks (2/15-9/15):
This series adds support for the CPU clocks on msm8996 devices.
The driver uses the existing PLL drivers and is required to control
the CPU frequency scaling on the MSM8996.
Ilia Lin (6):
soc: qcom: Separate kryo l2 accessors from PMU driver
clk: Use devm_ in the register fixed factor clock
clk: qcom: Add CPU clock driver for msm8996
dt-bindings: clk: qcom: Add bindings for CPU clock for msm8996
clk: qcom: cpu-8996: Add support to switch below 600Mhz
clk: qcom: Add ACD path to CPU clock driver for msm8996
Rajendra Nayak (2):
clk: qcom: Make clk_alpha_pll_configure available to modules
clk: qcom: cpu-8996: Add support to switch to alternate PLL
.../devicetree/bindings/clock/qcom,kryocc.txt | 17 +
drivers/clk/clk-fixed-factor.c | 2 +-
drivers/clk/qcom/Kconfig | 10 +
drivers/clk/qcom/Makefile | 1 +
drivers/clk/qcom/clk-alpha-pll.c | 1 +
drivers/clk/qcom/clk-alpha-pll.h | 6 +
drivers/clk/qcom/clk-cpu-8996.c | 510 +++++++++++++++++++++
drivers/perf/Kconfig | 1 +
drivers/perf/qcom_l2_pmu.c | 90 +---
drivers/soc/qcom/Kconfig | 3 +
drivers/soc/qcom/Makefile | 1 +
drivers/soc/qcom/kryo-l2-accessors.c | 56 +++
include/soc/qcom/kryo-l2-accessors.h | 12 +
13 files changed, 643 insertions(+), 67 deletions(-)
create mode 100644 Documentation/devicetree/bindings/clock/qcom,kryocc.txt
create mode 100644 drivers/clk/qcom/clk-cpu-8996.c
create mode 100644 drivers/soc/qcom/kryo-l2-accessors.c
create mode 100644 include/soc/qcom/kryo-l2-accessors.h
--
1.9.1
From: Rajendra Nayak <[email protected]>
Allow clk_alpha_pll_configure to be called from loadable
kernel modules.
Signed-off-by: Rajendra Nayak <[email protected]>
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/clk/qcom/clk-alpha-pll.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/clk/qcom/clk-alpha-pll.c b/drivers/clk/qcom/clk-alpha-pll.c
index 9722b70..57f2084 100644
--- a/drivers/clk/qcom/clk-alpha-pll.c
+++ b/drivers/clk/qcom/clk-alpha-pll.c
@@ -228,6 +228,7 @@ void clk_alpha_pll_configure(struct clk_alpha_pll *pll, struct regmap *regmap,
if (pll->flags & SUPPORTS_FSM_MODE)
qcom_pll_set_fsm_mode(regmap, PLL_MODE(pll), 6, 0);
}
+EXPORT_SYMBOL_GPL(clk_alpha_pll_configure);
static int clk_alpha_pll_hwfsm_enable(struct clk_hw *hw)
{
--
1.9.1
From: Rajendra Nayak <[email protected]>
Each of the CPU clusters on msm8996 are powered via a primary
PLL and a secondary PLL. The primary PLL is what drives the
CPU clk, except for times when we are reprogramming the PLL
itself, when we temporarily switch to an alternate PLL.
Use clock rate change notifiers to support this.
Signed-off-by: Rajendra Nayak <[email protected]>
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/clk/qcom/clk-cpu-8996.c | 33 +++++++++++++++++++++++++++++++++
1 file changed, 33 insertions(+)
diff --git a/drivers/clk/qcom/clk-cpu-8996.c b/drivers/clk/qcom/clk-cpu-8996.c
index d92cad93..620fdc2 100644
--- a/drivers/clk/qcom/clk-cpu-8996.c
+++ b/drivers/clk/qcom/clk-cpu-8996.c
@@ -52,6 +52,7 @@
* detect voltage droops.
*/
+#include <linux/clk.h>
#include <linux/module.h>
#include <linux/platform_device.h>
#include <linux/regmap.h>
@@ -178,10 +179,14 @@ struct clk_cpu_8996_mux {
u32 reg;
u8 shift;
u8 width;
+ struct notifier_block nb;
struct clk_hw *pll;
struct clk_regmap clkr;
};
+#define to_clk_cpu_8996_mux_nb(_nb) \
+ container_of(_nb, struct clk_cpu_8996_mux, nb)
+
static inline
struct clk_cpu_8996_mux *to_clk_cpu_8996_mux_hw(struct clk_hw *hw)
{
@@ -227,6 +232,26 @@ static int clk_cpu_8996_mux_set_parent(struct clk_hw *hw, u8 index)
return 0;
}
+int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
+ void *data)
+{
+ int ret;
+ struct clk_cpu_8996_mux *cpuclk = to_clk_cpu_8996_mux_nb(nb);
+
+ switch (event) {
+ case PRE_RATE_CHANGE:
+ ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw, ALT_INDEX);
+ break;
+ case POST_RATE_CHANGE:
+ ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw, PLL_INDEX);
+ break;
+ default:
+ ret = 0;
+ break;
+ }
+
+ return notifier_from_errno(ret);
+};
const struct clk_ops clk_cpu_8996_mux_ops = {
.set_parent = clk_cpu_8996_mux_set_parent,
.get_parent = clk_cpu_8996_mux_get_parent,
@@ -270,6 +295,7 @@ static int clk_cpu_8996_mux_set_parent(struct clk_hw *hw, u8 index)
.shift = 0,
.width = 2,
.pll = &pwrcl_pll.clkr.hw,
+ .nb.notifier_call = cpu_clk_notifier_cb,
.clkr.hw.init = &(struct clk_init_data) {
.name = "pwrcl_pmux",
.parent_names = (const char *[]){
@@ -289,6 +315,7 @@ static int clk_cpu_8996_mux_set_parent(struct clk_hw *hw, u8 index)
.shift = 0,
.width = 2,
.pll = &perfcl_pll.clkr.hw,
+ .nb.notifier_call = cpu_clk_notifier_cb,
.clkr.hw.init = &(struct clk_init_data) {
.name = "perfcl_pmux",
.parent_names = (const char *[]){
@@ -347,6 +374,12 @@ struct clk_regmap *clks[] = {
clk_alpha_pll_configure(&perfcl_alt_pll, regmap, &altpll_config);
clk_alpha_pll_configure(&pwrcl_alt_pll, regmap, &altpll_config);
+ ret = clk_notifier_register(pwrcl_pmux.clkr.hw.clk, &pwrcl_pmux.nb);
+ if (ret)
+ return ret;
+
+ ret = clk_notifier_register(perfcl_pmux.clkr.hw.clk, &perfcl_pmux.nb);
+
return ret;
}
--
1.9.1
The PMUX for each duplex allows for selection of ACD clock source.
The DVM (Dynamic Variation Monitor) will flag an error
when a voltage droop event is detected. This flagged error
enables ACD to provide a div-by-2 clock, sourced from the primary PLL.
The duplex will be provided the divided clock
until a pre-programmed delay has expired.
This change configures ACD during the probe and switches
the PMUXes to the ACD clock source.
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/clk/qcom/clk-cpu-8996.c | 75 +++++++++++++++++++++++++++++++++++------
1 file changed, 65 insertions(+), 10 deletions(-)
diff --git a/drivers/clk/qcom/clk-cpu-8996.c b/drivers/clk/qcom/clk-cpu-8996.c
index ff5c0a5..0a908d8 100644
--- a/drivers/clk/qcom/clk-cpu-8996.c
+++ b/drivers/clk/qcom/clk-cpu-8996.c
@@ -53,9 +53,11 @@
*/
#include <linux/clk.h>
+#include <linux/clk-provider.h>
#include <linux/module.h>
#include <linux/platform_device.h>
#include <linux/regmap.h>
+#include <soc/qcom/kryo-l2-accessors.h>
#include "clk-alpha-pll.h"
#include "clk-regmap.h"
@@ -69,6 +71,11 @@ enum _pmux_input {
};
#define DIV_2_THRESHOLD 600000000
+#define PWRCL_REG_OFFSET 0x0
+#define PERFCL_REG_OFFSET 0x80000
+#define MUX_OFFSET 0x40
+#define ALT_PLL_OFFSET 0x100
+#define SSSCTL_OFFSET 0x160
static const u8 prim_pll_regs[PLL_OFF_MAX_REGS] = {
[PLL_OFF_L_VAL] = 0x04,
@@ -107,7 +114,7 @@ enum _pmux_input {
};
static struct clk_alpha_pll perfcl_pll = {
- .offset = 0x80000,
+ .offset = PERFCL_REG_OFFSET,
.regs = prim_pll_regs,
.flags = SUPPORTS_DYNAMIC_UPDATE | SUPPORTS_FSM_MODE,
.clkr.hw.init = &(struct clk_init_data){
@@ -119,7 +126,7 @@ enum _pmux_input {
};
static struct clk_alpha_pll pwrcl_pll = {
- .offset = 0x0,
+ .offset = PWRCL_REG_OFFSET,
.regs = prim_pll_regs,
.flags = SUPPORTS_DYNAMIC_UPDATE | SUPPORTS_FSM_MODE,
.clkr.hw.init = &(struct clk_init_data){
@@ -149,7 +156,7 @@ enum _pmux_input {
};
static struct clk_alpha_pll perfcl_alt_pll = {
- .offset = 0x80100,
+ .offset = PERFCL_REG_OFFSET + ALT_PLL_OFFSET,
.regs = alt_pll_regs,
.vco_table = alt_pll_vco_modes,
.num_vco = ARRAY_SIZE(alt_pll_vco_modes),
@@ -163,7 +170,7 @@ enum _pmux_input {
};
static struct clk_alpha_pll pwrcl_alt_pll = {
- .offset = 0x100,
+ .offset = PWRCL_REG_OFFSET + ALT_PLL_OFFSET,
.regs = alt_pll_regs,
.vco_table = alt_pll_vco_modes,
.num_vco = ARRAY_SIZE(alt_pll_vco_modes),
@@ -176,6 +183,9 @@ enum _pmux_input {
},
};
+void __iomem *base;
+static void qcom_cpu_clk_msm8996_acd_init(void __iomem *base);
+
/* Mux'es */
struct clk_cpu_8996_mux {
@@ -253,6 +263,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
switch (event) {
case PRE_RATE_CHANGE:
ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw, ALT_INDEX);
+ qcom_cpu_clk_msm8996_acd_init(base);
break;
case POST_RATE_CHANGE:
if (cnd->new_rate < DIV_2_THRESHOLD)
@@ -260,7 +271,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
DIV_2_INDEX);
else
ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw,
- PLL_INDEX);
+ ACD_INDEX);
break;
default:
ret = 0;
@@ -276,7 +287,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
};
static struct clk_cpu_8996_mux pwrcl_smux = {
- .reg = 0x40,
+ .reg = PWRCL_REG_OFFSET + MUX_OFFSET,
.shift = 2,
.width = 2,
.clkr.hw.init = &(struct clk_init_data) {
@@ -292,7 +303,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
};
static struct clk_cpu_8996_mux perfcl_smux = {
- .reg = 0x80040,
+ .reg = PERFCL_REG_OFFSET + MUX_OFFSET,
.shift = 2,
.width = 2,
.clkr.hw.init = &(struct clk_init_data) {
@@ -308,7 +319,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
};
static struct clk_cpu_8996_mux pwrcl_pmux = {
- .reg = 0x40,
+ .reg = PWRCL_REG_OFFSET + MUX_OFFSET,
.shift = 0,
.width = 2,
.pll = &pwrcl_pll.clkr.hw,
@@ -329,7 +340,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
};
static struct clk_cpu_8996_mux perfcl_pmux = {
- .reg = 0x80040,
+ .reg = PERFCL_REG_OFFSET + MUX_OFFSET,
.shift = 0,
.width = 2,
.pll = &perfcl_pll.clkr.hw,
@@ -393,6 +404,10 @@ struct clk_regmap *clks[] = {
clk_alpha_pll_configure(&perfcl_alt_pll, regmap, &altpll_config);
clk_alpha_pll_configure(&pwrcl_alt_pll, regmap, &altpll_config);
+ /* Enable alt PLLs */
+ clk_prepare_enable(pwrcl_alt_pll.clkr.hw.clk);
+ clk_prepare_enable(perfcl_alt_pll.clkr.hw.clk);
+
ret = clk_notifier_register(pwrcl_pmux.clkr.hw.clk, &pwrcl_pmux.nb);
if (ret)
return ret;
@@ -402,10 +417,48 @@ struct clk_regmap *clks[] = {
return ret;
}
+#define CPU_AFINITY_MASK 0xFFF
+#define PWRCL_CPU_REG_MASK 0x3
+#define PERFCL_CPU_REG_MASK 0x103
+
+#define L2ACDCR_REG 0x580ULL
+#define L2ACDTD_REG 0x581ULL
+#define L2ACDDVMRC_REG 0x584ULL
+#define L2ACDSSCR_REG 0x589ULL
+
+static DEFINE_SPINLOCK(acd_lock);
+
+static void qcom_cpu_clk_msm8996_acd_init(void __iomem *base)
+{
+ u64 hwid;
+ unsigned long flags;
+
+ spin_lock_irqsave(&acd_lock, flags);
+
+ hwid = read_cpuid_mpidr() & CPU_AFINITY_MASK;
+
+ kryo_l2_set_indirect_reg(L2ACDTD_REG, 0x00006A11);
+ kryo_l2_set_indirect_reg(L2ACDDVMRC_REG, 0x000E0F0F);
+ kryo_l2_set_indirect_reg(L2ACDSSCR_REG, 0x00000601);
+
+ if (PWRCL_CPU_REG_MASK == (hwid | PWRCL_CPU_REG_MASK)) {
+ writel(0xF, base + PWRCL_REG_OFFSET + SSSCTL_OFFSET);
+ wmb();
+ kryo_l2_set_indirect_reg(L2ACDCR_REG, 0x002C5FFD);
+ }
+
+ if (PERFCL_CPU_REG_MASK == (hwid | PERFCL_CPU_REG_MASK)) {
+ kryo_l2_set_indirect_reg(L2ACDCR_REG, 0x002C5FFD);
+ writel(0xF, base + PERFCL_REG_OFFSET + SSSCTL_OFFSET);
+ wmb();
+ }
+
+ spin_unlock_irqrestore(&acd_lock, flags);
+}
+
static int qcom_cpu_clk_msm8996_driver_probe(struct platform_device *pdev)
{
int ret;
- void __iomem *base;
struct resource *res;
struct regmap *regmap;
struct clk_hw_onecell_data *data;
@@ -429,6 +482,8 @@ static int qcom_cpu_clk_msm8996_driver_probe(struct platform_device *pdev)
if (ret)
return ret;
+ qcom_cpu_clk_msm8996_acd_init(base);
+
data->hws[0] = &pwrcl_pmux.clkr.hw;
data->hws[1] = &perfcl_pmux.clkr.hw;
data->num = 2;
--
1.9.1
The CPU clock controller's primary PLL operates on a single VCO range,
between 600MHz and 3GHz. However the CPUs do support OPPs with
frequencies between 300MHz and 600MHz. In order to support running the
CPUs at those frequencies we end up having to lock the PLL at twice the
rate and drive the CPU clk via the PLL/2 output and SMUX.
So for frequencies above 600MHz we follow the following path
Primary PLL --> PLL_EARLY --> PMUX(1) --> CPU clk
and for frequencies between 300MHz and 600MHz we follow
Primary PLL --> PLL/2 --> SMUX(1) --> PMUX(0) --> CPU clk
Signed-off-by: Rajendra Nayak <[email protected]>
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/clk/qcom/clk-cpu-8996.c | 25 ++++++++++++++++++++++---
1 file changed, 22 insertions(+), 3 deletions(-)
diff --git a/drivers/clk/qcom/clk-cpu-8996.c b/drivers/clk/qcom/clk-cpu-8996.c
index 620fdc2..ff5c0a5 100644
--- a/drivers/clk/qcom/clk-cpu-8996.c
+++ b/drivers/clk/qcom/clk-cpu-8996.c
@@ -68,6 +68,8 @@ enum _pmux_input {
NUM_OF_PMUX_INPUTS
};
+#define DIV_2_THRESHOLD 600000000
+
static const u8 prim_pll_regs[PLL_OFF_MAX_REGS] = {
[PLL_OFF_L_VAL] = 0x04,
[PLL_OFF_ALPHA_VAL] = 0x08,
@@ -95,10 +97,11 @@ enum _pmux_input {
static const struct alpha_pll_config hfpll_config = {
.l = 60,
- .config_ctl_val = 0x200d4828,
+ .config_ctl_val = 0x200d4aa8,
.config_ctl_hi_val = 0x006,
.pre_div_mask = BIT(12),
.post_div_mask = 0x3 << 8,
+ .post_div_val = 0x1 << 8,
.main_output_mask = BIT(0),
.early_output_mask = BIT(3),
};
@@ -140,7 +143,7 @@ enum _pmux_input {
.vco_mask = 0x3 << 20,
.config_ctl_val = 0x4001051b,
.post_div_mask = 0x3 << 8,
- .post_div_val = 0x1,
+ .post_div_val = 0x1 << 8,
.main_output_mask = BIT(0),
.early_output_mask = BIT(3),
};
@@ -181,6 +184,7 @@ struct clk_cpu_8996_mux {
u8 width;
struct notifier_block nb;
struct clk_hw *pll;
+ struct clk_hw *pll_div_2;
struct clk_regmap clkr;
};
@@ -226,6 +230,13 @@ static int clk_cpu_8996_mux_set_parent(struct clk_hw *hw, u8 index)
struct clk_cpu_8996_mux *cpuclk = to_clk_cpu_8996_mux_hw(hw);
struct clk_hw *parent = cpuclk->pll;
+ if (cpuclk->pll_div_2 && req->rate < DIV_2_THRESHOLD) {
+ if (req->rate < (DIV_2_THRESHOLD / 2))
+ return -EINVAL;
+
+ parent = cpuclk->pll_div_2;
+ }
+
req->best_parent_rate = clk_hw_round_rate(parent, req->rate);
req->best_parent_hw = parent;
@@ -237,13 +248,19 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
{
int ret;
struct clk_cpu_8996_mux *cpuclk = to_clk_cpu_8996_mux_nb(nb);
+ struct clk_notifier_data *cnd = data;
switch (event) {
case PRE_RATE_CHANGE:
ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw, ALT_INDEX);
break;
case POST_RATE_CHANGE:
- ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw, PLL_INDEX);
+ if (cnd->new_rate < DIV_2_THRESHOLD)
+ ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw,
+ DIV_2_INDEX);
+ else
+ ret = clk_cpu_8996_mux_set_parent(&cpuclk->clkr.hw,
+ PLL_INDEX);
break;
default:
ret = 0;
@@ -295,6 +312,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
.shift = 0,
.width = 2,
.pll = &pwrcl_pll.clkr.hw,
+ .pll_div_2 = &pwrcl_smux.clkr.hw,
.nb.notifier_call = cpu_clk_notifier_cb,
.clkr.hw.init = &(struct clk_init_data) {
.name = "pwrcl_pmux",
@@ -315,6 +333,7 @@ int cpu_clk_notifier_cb(struct notifier_block *nb, unsigned long event,
.shift = 0,
.width = 2,
.pll = &perfcl_pll.clkr.hw,
+ .pll_div_2 = &perfcl_smux.clkr.hw,
.nb.notifier_call = cpu_clk_notifier_cb,
.clkr.hw.init = &(struct clk_init_data) {
.name = "perfcl_pmux",
--
1.9.1
Use devm_clk_hw_register instead of clk_hw_register
to simplify the usage of this API. This way drivers that call
the clk_hw_register_fixed_factor won't need to maintain
a data structure for further cleanup.
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/clk/clk-fixed-factor.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/clk/clk-fixed-factor.c b/drivers/clk/clk-fixed-factor.c
index a5d402d..8e39bda 100644
--- a/drivers/clk/clk-fixed-factor.c
+++ b/drivers/clk/clk-fixed-factor.c
@@ -94,7 +94,7 @@ struct clk_hw *clk_hw_register_fixed_factor(struct device *dev,
init.num_parents = 1;
hw = &fix->hw;
- ret = clk_hw_register(dev, hw);
+ ret = devm_clk_hw_register(dev, hw);
if (ret) {
kfree(fix);
hw = ERR_PTR(ret);
--
1.9.1
Each of the CPU clusters (Power and Perf) on msm8996 are
clocked via 2 PLLs, a primary and alternate. There are also
2 Mux'es, a primary and secondary all connected together
as shown below
+-------+
XO | |
+------------------>0 |
| |
PLL/2 | SMUX +----+
+------->1 | |
| | | |
| +-------+ | +-------+
| +---->0 |
| | |
+---------------+ | +----------->1 | CPU clk
|Primary PLL +----+ PLL_EARLY | | +------>
| +------+-----------+ +------>2 PMUX |
+---------------+ | | | |
| +------+ | +-->3 |
+--^+ ACD +-----+ | +-------+
+---------------+ +------+ |
|Alt PLL | |
| +---------------------------+
+---------------+ PLL_EARLY
The primary PLL is what drives the CPU clk, except for times
when we are reprogramming the PLL itself (for rate changes) when
we temporarily switch to an alternate PLL. A subsequent patch adds
support to switch between primary and alternate PLL during rate
changes.
The primary PLL operates on a single VCO range, between 600MHz
and 3GHz. However the CPUs do support OPPs with frequencies
between 300MHz and 600MHz. In order to support running the CPUs
at those frequencies we end up having to lock the PLL at twice
the rate and drive the CPU clk via the PLL/2 output and SMUX.
So for frequencies above 600MHz we follow the following path
Primary PLL --> PLL_EARLY --> PMUX(1) --> CPU clk
and for frequencies between 300MHz and 600MHz we follow
Primary PLL --> PLL/2 --> SMUX(1) --> PMUX(0) --> CPU clk
Support for this is added in a subsequent patch as well.
ACD stands for Adaptive Clock Distribution and is used to
detect voltage droops.
Signed-off-by: Rajendra Nayak <[email protected]>
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/clk/qcom/Kconfig | 10 +
drivers/clk/qcom/Makefile | 1 +
drivers/clk/qcom/clk-alpha-pll.h | 6 +
drivers/clk/qcom/clk-cpu-8996.c | 403 +++++++++++++++++++++++++++++++++++++++
4 files changed, 420 insertions(+)
create mode 100644 drivers/clk/qcom/clk-cpu-8996.c
diff --git a/drivers/clk/qcom/Kconfig b/drivers/clk/qcom/Kconfig
index e42e1af..42e84b5 100644
--- a/drivers/clk/qcom/Kconfig
+++ b/drivers/clk/qcom/Kconfig
@@ -33,6 +33,16 @@ config QCOM_CLK_APCS_MSM8916
Say Y if you want to support CPU frequency scaling on devices
such as msm8916.
+config QCOM_CLK_APCC_MSM8996
+ tristate "MSM8996 CPU Clock Controller"
+ depends on ARM64
+ depends on COMMON_CLK_QCOM
+ select QCOM_KRYO_L2_ACCESSORS
+ help
+ Support for the CPU clock controller on msm8996 devices.
+ Say Y if you want to support CPU clock scaling using CPUfreq
+ drivers for dyanmic power management.
+
config QCOM_CLK_RPM
tristate "RPM based Clock Controller"
depends on COMMON_CLK_QCOM && MFD_QCOM_RPM
diff --git a/drivers/clk/qcom/Makefile b/drivers/clk/qcom/Makefile
index 7c09ab1..a822fc8 100644
--- a/drivers/clk/qcom/Makefile
+++ b/drivers/clk/qcom/Makefile
@@ -36,6 +36,7 @@ obj-$(CONFIG_MSM_MMCC_8974) += mmcc-msm8974.o
obj-$(CONFIG_MSM_MMCC_8996) += mmcc-msm8996.o
obj-$(CONFIG_QCOM_A53PLL) += a53-pll.o
obj-$(CONFIG_QCOM_CLK_APCS_MSM8916) += apcs-msm8916.o
+obj-$(CONFIG_QCOM_CLK_APCC_MSM8996) += clk-cpu-8996.o
obj-$(CONFIG_QCOM_CLK_RPM) += clk-rpm.o
obj-$(CONFIG_QCOM_CLK_SMD_RPM) += clk-smd-rpm.o
obj-$(CONFIG_SPMI_PMIC_CLKDIV) += clk-spmi-pmic-div.o
diff --git a/drivers/clk/qcom/clk-alpha-pll.h b/drivers/clk/qcom/clk-alpha-pll.h
index f981b48..9ce2a32 100644
--- a/drivers/clk/qcom/clk-alpha-pll.h
+++ b/drivers/clk/qcom/clk-alpha-pll.h
@@ -50,6 +50,12 @@ struct pll_vco {
u32 val;
};
+#define VCO(a, b, c) { \
+ .val = a,\
+ .min_freq = b,\
+ .max_freq = c,\
+}
+
/**
* struct clk_alpha_pll - phase locked loop (PLL)
* @offset: base address of registers
diff --git a/drivers/clk/qcom/clk-cpu-8996.c b/drivers/clk/qcom/clk-cpu-8996.c
new file mode 100644
index 0000000..d92cad93
--- /dev/null
+++ b/drivers/clk/qcom/clk-cpu-8996.c
@@ -0,0 +1,403 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018, The Linux Foundation. All rights reserved.
+ */
+
+/*
+ * Each of the CPU clusters (Power and Perf) on msm8996 are
+ * clocked via 2 PLLs, a primary and alternate. There are also
+ * 2 Mux'es, a primary and secondary all connected together
+ * as shown below
+ *
+ * +-------+
+ * XO | |
+ * +------------------>0 |
+ * | |
+ * PLL/2 | SMUX +----+
+ * +------->1 | |
+ * | | | |
+ * | +-------+ | +-------+
+ * | +---->0 |
+ * | | |
+ * +---------------+ | +----------->1 | CPU clk
+ * |Primary PLL +----+ PLL_EARLY | | +------>
+ * | +------+-----------+ +------>2 PMUX |
+ * +---------------+ | | | |
+ * | +------+ | +-->3 |
+ * +--^+ ACD +-----+ | +-------+
+ * +---------------+ +------+ |
+ * |Alt PLL | |
+ * | +---------------------------+
+ * +---------------+ PLL_EARLY
+ *
+ * The primary PLL is what drives the CPU clk, except for times
+ * when we are reprogramming the PLL itself (for rate changes) when
+ * we temporarily switch to an alternate PLL. A subsequent patch adds
+ * support to switch between primary and alternate PLL during rate
+ * changes.
+ *
+ * The primary PLL operates on a single VCO range, between 600MHz
+ * and 3GHz. However the CPUs do support OPPs with frequencies
+ * between 300MHz and 600MHz. In order to support running the CPUs
+ * at those frequencies we end up having to lock the PLL at twice
+ * the rate and drive the CPU clk via the PLL/2 output and SMUX.
+ *
+ * So for frequencies above 600MHz we follow the following path
+ * Primary PLL --> PLL_EARLY --> PMUX(1) --> CPU clk
+ * and for frequencies between 300MHz and 600MHz we follow
+ * Primary PLL --> PLL/2 --> SMUX(1) --> PMUX(0) --> CPU clk
+ * Support for this is added in a subsequent patch as well.
+ *
+ * ACD stands for Adaptive Clock Distribution and is used to
+ * detect voltage droops.
+ */
+
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+
+#include "clk-alpha-pll.h"
+#include "clk-regmap.h"
+
+enum _pmux_input {
+ DIV_2_INDEX = 0,
+ PLL_INDEX,
+ ACD_INDEX,
+ ALT_INDEX,
+ NUM_OF_PMUX_INPUTS
+};
+
+static const u8 prim_pll_regs[PLL_OFF_MAX_REGS] = {
+ [PLL_OFF_L_VAL] = 0x04,
+ [PLL_OFF_ALPHA_VAL] = 0x08,
+ [PLL_OFF_USER_CTL] = 0x10,
+ [PLL_OFF_CONFIG_CTL] = 0x18,
+ [PLL_OFF_CONFIG_CTL_U] = 0x1c,
+ [PLL_OFF_TEST_CTL] = 0x20,
+ [PLL_OFF_TEST_CTL_U] = 0x24,
+ [PLL_OFF_STATUS] = 0x28,
+};
+
+static const u8 alt_pll_regs[PLL_OFF_MAX_REGS] = {
+ [PLL_OFF_L_VAL] = 0x04,
+ [PLL_OFF_ALPHA_VAL] = 0x08,
+ [PLL_OFF_ALPHA_VAL_U] = 0x0c,
+ [PLL_OFF_USER_CTL] = 0x10,
+ [PLL_OFF_USER_CTL_U] = 0x14,
+ [PLL_OFF_CONFIG_CTL] = 0x18,
+ [PLL_OFF_TEST_CTL] = 0x20,
+ [PLL_OFF_TEST_CTL_U] = 0x24,
+ [PLL_OFF_STATUS] = 0x28,
+};
+
+/* PLLs */
+
+static const struct alpha_pll_config hfpll_config = {
+ .l = 60,
+ .config_ctl_val = 0x200d4828,
+ .config_ctl_hi_val = 0x006,
+ .pre_div_mask = BIT(12),
+ .post_div_mask = 0x3 << 8,
+ .main_output_mask = BIT(0),
+ .early_output_mask = BIT(3),
+};
+
+static struct clk_alpha_pll perfcl_pll = {
+ .offset = 0x80000,
+ .regs = prim_pll_regs,
+ .flags = SUPPORTS_DYNAMIC_UPDATE | SUPPORTS_FSM_MODE,
+ .clkr.hw.init = &(struct clk_init_data){
+ .name = "perfcl_pll",
+ .parent_names = (const char *[]){ "xo" },
+ .num_parents = 1,
+ .ops = &clk_alpha_pll_huayra_ops,
+ },
+};
+
+static struct clk_alpha_pll pwrcl_pll = {
+ .offset = 0x0,
+ .regs = prim_pll_regs,
+ .flags = SUPPORTS_DYNAMIC_UPDATE | SUPPORTS_FSM_MODE,
+ .clkr.hw.init = &(struct clk_init_data){
+ .name = "pwrcl_pll",
+ .parent_names = (const char *[]){ "xo" },
+ .num_parents = 1,
+ .ops = &clk_alpha_pll_huayra_ops,
+ },
+};
+
+static const struct pll_vco alt_pll_vco_modes[] = {
+ VCO(3, 250000000, 500000000),
+ VCO(2, 500000000, 750000000),
+ VCO(1, 750000000, 1000000000),
+ VCO(0, 1000000000, 2150400000),
+};
+
+static const struct alpha_pll_config altpll_config = {
+ .l = 16,
+ .vco_val = 0x3 << 20,
+ .vco_mask = 0x3 << 20,
+ .config_ctl_val = 0x4001051b,
+ .post_div_mask = 0x3 << 8,
+ .post_div_val = 0x1,
+ .main_output_mask = BIT(0),
+ .early_output_mask = BIT(3),
+};
+
+static struct clk_alpha_pll perfcl_alt_pll = {
+ .offset = 0x80100,
+ .regs = alt_pll_regs,
+ .vco_table = alt_pll_vco_modes,
+ .num_vco = ARRAY_SIZE(alt_pll_vco_modes),
+ .flags = SUPPORTS_OFFLINE_REQ | SUPPORTS_FSM_MODE,
+ .clkr.hw.init = &(struct clk_init_data) {
+ .name = "perfcl_alt_pll",
+ .parent_names = (const char *[]){ "xo" },
+ .num_parents = 1,
+ .ops = &clk_alpha_pll_hwfsm_ops,
+ },
+};
+
+static struct clk_alpha_pll pwrcl_alt_pll = {
+ .offset = 0x100,
+ .regs = alt_pll_regs,
+ .vco_table = alt_pll_vco_modes,
+ .num_vco = ARRAY_SIZE(alt_pll_vco_modes),
+ .flags = SUPPORTS_OFFLINE_REQ | SUPPORTS_FSM_MODE,
+ .clkr.hw.init = &(struct clk_init_data) {
+ .name = "pwrcl_alt_pll",
+ .parent_names = (const char *[]){ "xo" },
+ .num_parents = 1,
+ .ops = &clk_alpha_pll_hwfsm_ops,
+ },
+};
+
+/* Mux'es */
+
+struct clk_cpu_8996_mux {
+ u32 reg;
+ u8 shift;
+ u8 width;
+ struct clk_hw *pll;
+ struct clk_regmap clkr;
+};
+
+static inline
+struct clk_cpu_8996_mux *to_clk_cpu_8996_mux_hw(struct clk_hw *hw)
+{
+ return container_of(to_clk_regmap(hw), struct clk_cpu_8996_mux, clkr);
+}
+
+static u8 clk_cpu_8996_mux_get_parent(struct clk_hw *hw)
+{
+ u32 val;
+ struct clk_regmap *clkr = to_clk_regmap(hw);
+ struct clk_cpu_8996_mux *cpuclk = to_clk_cpu_8996_mux_hw(hw);
+ u32 mask = (u32)GENMASK(cpuclk->width - 1, 0);
+
+ regmap_read(clkr->regmap, cpuclk->reg, &val);
+ val >>= (u32)(cpuclk->shift);
+
+ return (u8)(val & mask);
+}
+
+static int clk_cpu_8996_mux_set_parent(struct clk_hw *hw, u8 index)
+{
+ u32 val;
+ struct clk_regmap *clkr = to_clk_regmap(hw);
+ struct clk_cpu_8996_mux *cpuclk = to_clk_cpu_8996_mux_hw(hw);
+ unsigned int mask = GENMASK(cpuclk->width + cpuclk->shift - 1,
+ cpuclk->shift);
+
+ val = (u32)index;
+ val <<= (u32)(cpuclk->shift);
+
+ return regmap_update_bits(clkr->regmap, cpuclk->reg, mask, val);
+}
+
+static int
+clk_cpu_8996_mux_determine_rate(struct clk_hw *hw, struct clk_rate_request *req)
+{
+ struct clk_cpu_8996_mux *cpuclk = to_clk_cpu_8996_mux_hw(hw);
+ struct clk_hw *parent = cpuclk->pll;
+
+ req->best_parent_rate = clk_hw_round_rate(parent, req->rate);
+ req->best_parent_hw = parent;
+
+ return 0;
+}
+
+const struct clk_ops clk_cpu_8996_mux_ops = {
+ .set_parent = clk_cpu_8996_mux_set_parent,
+ .get_parent = clk_cpu_8996_mux_get_parent,
+ .determine_rate = clk_cpu_8996_mux_determine_rate,
+};
+
+static struct clk_cpu_8996_mux pwrcl_smux = {
+ .reg = 0x40,
+ .shift = 2,
+ .width = 2,
+ .clkr.hw.init = &(struct clk_init_data) {
+ .name = "pwrcl_smux",
+ .parent_names = (const char *[]){
+ "xo",
+ "pwrcl_pll_main",
+ },
+ .num_parents = 2,
+ .ops = &clk_cpu_8996_mux_ops,
+ .flags = CLK_SET_RATE_PARENT,
+ },
+};
+
+static struct clk_cpu_8996_mux perfcl_smux = {
+ .reg = 0x80040,
+ .shift = 2,
+ .width = 2,
+ .clkr.hw.init = &(struct clk_init_data) {
+ .name = "perfcl_smux",
+ .parent_names = (const char *[]){
+ "xo",
+ "perfcl_pll_main",
+ },
+ .num_parents = 2,
+ .ops = &clk_cpu_8996_mux_ops,
+ .flags = CLK_SET_RATE_PARENT,
+ },
+};
+
+static struct clk_cpu_8996_mux pwrcl_pmux = {
+ .reg = 0x40,
+ .shift = 0,
+ .width = 2,
+ .pll = &pwrcl_pll.clkr.hw,
+ .clkr.hw.init = &(struct clk_init_data) {
+ .name = "pwrcl_pmux",
+ .parent_names = (const char *[]){
+ "pwrcl_smux",
+ "pwrcl_pll",
+ "pwrcl_pll_acd",
+ "pwrcl_alt_pll",
+ },
+ .num_parents = 4,
+ .ops = &clk_cpu_8996_mux_ops,
+ .flags = CLK_SET_RATE_PARENT | CLK_IGNORE_UNUSED,
+ },
+};
+
+static struct clk_cpu_8996_mux perfcl_pmux = {
+ .reg = 0x80040,
+ .shift = 0,
+ .width = 2,
+ .pll = &perfcl_pll.clkr.hw,
+ .clkr.hw.init = &(struct clk_init_data) {
+ .name = "perfcl_pmux",
+ .parent_names = (const char *[]){
+ "perfcl_smux",
+ "perfcl_pll",
+ "perfcl_pll_acd",
+ "perfcl_alt_pll",
+ },
+ .num_parents = 4,
+ .ops = &clk_cpu_8996_mux_ops,
+ .flags = CLK_SET_RATE_PARENT | CLK_IGNORE_UNUSED,
+ },
+};
+
+static const struct regmap_config cpu_msm8996_regmap_config = {
+ .reg_bits = 32,
+ .reg_stride = 4,
+ .val_bits = 32,
+ .max_register = 0x80210,
+ .fast_io = true,
+ .val_format_endian = REGMAP_ENDIAN_LITTLE,
+};
+
+struct clk_regmap *clks[] = {
+ &perfcl_pll.clkr,
+ &pwrcl_pll.clkr,
+ &perfcl_alt_pll.clkr,
+ &pwrcl_alt_pll.clkr,
+ &perfcl_smux.clkr,
+ &pwrcl_smux.clkr,
+ &perfcl_pmux.clkr,
+ &pwrcl_pmux.clkr,
+};
+
+static int
+qcom_cpu_clk_msm8996_register_clks(struct device *dev, struct regmap *regmap)
+{
+ int i, ret;
+
+ perfcl_smux.pll = clk_hw_register_fixed_factor(dev, "perfcl_pll_main",
+ "perfcl_pll",
+ CLK_SET_RATE_PARENT, 1, 2);
+
+ pwrcl_smux.pll = clk_hw_register_fixed_factor(dev, "pwrcl_pll_main",
+ "pwrcl_pll",
+ CLK_SET_RATE_PARENT, 1, 2);
+
+ for (i = 0; i < ARRAY_SIZE(clks); i++) {
+ ret = devm_clk_register_regmap(dev, clks[i]);
+ if (ret)
+ return ret;
+ }
+
+ clk_alpha_pll_configure(&perfcl_pll, regmap, &hfpll_config);
+ clk_alpha_pll_configure(&pwrcl_pll, regmap, &hfpll_config);
+ clk_alpha_pll_configure(&perfcl_alt_pll, regmap, &altpll_config);
+ clk_alpha_pll_configure(&pwrcl_alt_pll, regmap, &altpll_config);
+
+ return ret;
+}
+
+static int qcom_cpu_clk_msm8996_driver_probe(struct platform_device *pdev)
+{
+ int ret;
+ void __iomem *base;
+ struct resource *res;
+ struct regmap *regmap;
+ struct clk_hw_onecell_data *data;
+ struct device *dev = &pdev->dev;
+
+ data = devm_kzalloc(dev, sizeof(*data) + 2 * sizeof(struct clk_hw *),
+ GFP_KERNEL);
+ if (!data)
+ return -ENOMEM;
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ base = devm_ioremap_resource(dev, res);
+ if (IS_ERR(base))
+ return PTR_ERR(base);
+
+ regmap = devm_regmap_init_mmio(dev, base, &cpu_msm8996_regmap_config);
+ if (IS_ERR(regmap))
+ return PTR_ERR(regmap);
+
+ ret = qcom_cpu_clk_msm8996_register_clks(dev, regmap);
+ if (ret)
+ return ret;
+
+ data->hws[0] = &pwrcl_pmux.clkr.hw;
+ data->hws[1] = &perfcl_pmux.clkr.hw;
+ data->num = 2;
+
+ return devm_of_clk_add_hw_provider(dev, of_clk_hw_onecell_get, data);
+}
+
+static const struct of_device_id qcom_cpu_clk_msm8996_match_table[] = {
+ { .compatible = "qcom,msm8996-apcc" },
+ {}
+};
+
+static struct platform_driver qcom_cpu_clk_msm8996_driver = {
+ .probe = qcom_cpu_clk_msm8996_driver_probe,
+ .driver = {
+ .name = "qcom-msm8996-apcc",
+ .of_match_table = qcom_cpu_clk_msm8996_match_table,
+ },
+};
+module_platform_driver(qcom_cpu_clk_msm8996_driver);
+
+MODULE_ALIAS("platform:msm8996-apcc");
+MODULE_DESCRIPTION("QCOM MSM8996 CPU Clock Driver");
+MODULE_LICENSE("GPL v2");
--
1.9.1
The driver provides kernel level API for other drivers
to access the MSM8996 L2 cache registers.
Separating the L2 access code from the PMU driver and
making it public to allow other drivers use it.
The accesses must be separated with a single spinlock,
maintained in this driver.
Signed-off-by: Ilia Lin <[email protected]>
---
drivers/perf/Kconfig | 1 +
drivers/perf/qcom_l2_pmu.c | 90 ++++++++++--------------------------
drivers/soc/qcom/Kconfig | 3 ++
drivers/soc/qcom/Makefile | 1 +
drivers/soc/qcom/kryo-l2-accessors.c | 56 ++++++++++++++++++++++
include/soc/qcom/kryo-l2-accessors.h | 12 +++++
6 files changed, 97 insertions(+), 66 deletions(-)
create mode 100644 drivers/soc/qcom/kryo-l2-accessors.c
create mode 100644 include/soc/qcom/kryo-l2-accessors.h
diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig
index 28bb5a0..561252a 100644
--- a/drivers/perf/Kconfig
+++ b/drivers/perf/Kconfig
@@ -69,6 +69,7 @@ config HISI_PMU
config QCOM_L2_PMU
bool "Qualcomm Technologies L2-cache PMU"
depends on ARCH_QCOM && ARM64 && ACPI
+ select QCOM_KRYO_L2_ACCESSORS
help
Provides support for the L2 cache performance monitor unit (PMU)
in Qualcomm Technologies processors.
diff --git a/drivers/perf/qcom_l2_pmu.c b/drivers/perf/qcom_l2_pmu.c
index 842135c..cc31f51 100644
--- a/drivers/perf/qcom_l2_pmu.c
+++ b/drivers/perf/qcom_l2_pmu.c
@@ -31,6 +31,7 @@
#include <asm/barrier.h>
#include <asm/local64.h>
#include <asm/sysreg.h>
+#include <soc/qcom/kryo-l2-accessors.h>
#define MAX_L2_CTRS 9
@@ -87,8 +88,6 @@
#define L2_COUNTER_RELOAD BIT_ULL(31)
#define L2_CYCLE_COUNTER_RELOAD BIT_ULL(63)
-#define L2CPUSRSELR_EL1 sys_reg(3, 3, 15, 0, 6)
-#define L2CPUSRDR_EL1 sys_reg(3, 3, 15, 0, 7)
#define reg_idx(reg, i) (((i) * IA_L2_REG_OFFSET) + reg##_BASE)
@@ -107,48 +106,7 @@
#define L2_EVENT_STREX 0x421
#define L2_EVENT_CLREX 0x422
-static DEFINE_RAW_SPINLOCK(l2_access_lock);
-/**
- * set_l2_indirect_reg: write value to an L2 register
- * @reg: Address of L2 register.
- * @value: Value to be written to register.
- *
- * Use architecturally required barriers for ordering between system register
- * accesses
- */
-static void set_l2_indirect_reg(u64 reg, u64 val)
-{
- unsigned long flags;
-
- raw_spin_lock_irqsave(&l2_access_lock, flags);
- write_sysreg_s(reg, L2CPUSRSELR_EL1);
- isb();
- write_sysreg_s(val, L2CPUSRDR_EL1);
- isb();
- raw_spin_unlock_irqrestore(&l2_access_lock, flags);
-}
-
-/**
- * get_l2_indirect_reg: read an L2 register value
- * @reg: Address of L2 register.
- *
- * Use architecturally required barriers for ordering between system register
- * accesses
- */
-static u64 get_l2_indirect_reg(u64 reg)
-{
- u64 val;
- unsigned long flags;
-
- raw_spin_lock_irqsave(&l2_access_lock, flags);
- write_sysreg_s(reg, L2CPUSRSELR_EL1);
- isb();
- val = read_sysreg_s(L2CPUSRDR_EL1);
- raw_spin_unlock_irqrestore(&l2_access_lock, flags);
-
- return val;
-}
struct cluster_pmu;
@@ -219,28 +177,28 @@ static inline struct cluster_pmu *get_cluster_pmu(
static void cluster_pmu_reset(void)
{
/* Reset all counters */
- set_l2_indirect_reg(L2PMCR, L2PMCR_RESET_ALL);
- set_l2_indirect_reg(L2PMCNTENCLR, l2_counter_present_mask);
- set_l2_indirect_reg(L2PMINTENCLR, l2_counter_present_mask);
- set_l2_indirect_reg(L2PMOVSCLR, l2_counter_present_mask);
+ kryo_l2_set_indirect_reg(L2PMCR, L2PMCR_RESET_ALL);
+ kryo_l2_set_indirect_reg(L2PMCNTENCLR, l2_counter_present_mask);
+ kryo_l2_set_indirect_reg(L2PMINTENCLR, l2_counter_present_mask);
+ kryo_l2_set_indirect_reg(L2PMOVSCLR, l2_counter_present_mask);
}
static inline void cluster_pmu_enable(void)
{
- set_l2_indirect_reg(L2PMCR, L2PMCR_COUNTERS_ENABLE);
+ kryo_l2_set_indirect_reg(L2PMCR, L2PMCR_COUNTERS_ENABLE);
}
static inline void cluster_pmu_disable(void)
{
- set_l2_indirect_reg(L2PMCR, L2PMCR_COUNTERS_DISABLE);
+ kryo_l2_set_indirect_reg(L2PMCR, L2PMCR_COUNTERS_DISABLE);
}
static inline void cluster_pmu_counter_set_value(u32 idx, u64 value)
{
if (idx == l2_cycle_ctr_idx)
- set_l2_indirect_reg(L2PMCCNTR, value);
+ kryo_l2_set_indirect_reg(L2PMCCNTR, value);
else
- set_l2_indirect_reg(reg_idx(IA_L2PMXEVCNTR, idx), value);
+ kryo_l2_set_indirect_reg(reg_idx(IA_L2PMXEVCNTR, idx), value);
}
static inline u64 cluster_pmu_counter_get_value(u32 idx)
@@ -248,46 +206,46 @@ static inline u64 cluster_pmu_counter_get_value(u32 idx)
u64 value;
if (idx == l2_cycle_ctr_idx)
- value = get_l2_indirect_reg(L2PMCCNTR);
+ value = kryo_l2_get_indirect_reg(L2PMCCNTR);
else
- value = get_l2_indirect_reg(reg_idx(IA_L2PMXEVCNTR, idx));
+ value = kryo_l2_get_indirect_reg(reg_idx(IA_L2PMXEVCNTR, idx));
return value;
}
static inline void cluster_pmu_counter_enable(u32 idx)
{
- set_l2_indirect_reg(L2PMCNTENSET, idx_to_reg_bit(idx));
+ kryo_l2_set_indirect_reg(L2PMCNTENSET, idx_to_reg_bit(idx));
}
static inline void cluster_pmu_counter_disable(u32 idx)
{
- set_l2_indirect_reg(L2PMCNTENCLR, idx_to_reg_bit(idx));
+ kryo_l2_set_indirect_reg(L2PMCNTENCLR, idx_to_reg_bit(idx));
}
static inline void cluster_pmu_counter_enable_interrupt(u32 idx)
{
- set_l2_indirect_reg(L2PMINTENSET, idx_to_reg_bit(idx));
+ kryo_l2_set_indirect_reg(L2PMINTENSET, idx_to_reg_bit(idx));
}
static inline void cluster_pmu_counter_disable_interrupt(u32 idx)
{
- set_l2_indirect_reg(L2PMINTENCLR, idx_to_reg_bit(idx));
+ kryo_l2_set_indirect_reg(L2PMINTENCLR, idx_to_reg_bit(idx));
}
static inline void cluster_pmu_set_evccntcr(u32 val)
{
- set_l2_indirect_reg(L2PMCCNTCR, val);
+ kryo_l2_set_indirect_reg(L2PMCCNTCR, val);
}
static inline void cluster_pmu_set_evcntcr(u32 ctr, u32 val)
{
- set_l2_indirect_reg(reg_idx(IA_L2PMXEVCNTCR, ctr), val);
+ kryo_l2_set_indirect_reg(reg_idx(IA_L2PMXEVCNTCR, ctr), val);
}
static inline void cluster_pmu_set_evtyper(u32 ctr, u32 val)
{
- set_l2_indirect_reg(reg_idx(IA_L2PMXEVTYPER, ctr), val);
+ kryo_l2_set_indirect_reg(reg_idx(IA_L2PMXEVTYPER, ctr), val);
}
static void cluster_pmu_set_resr(struct cluster_pmu *cluster,
@@ -303,11 +261,11 @@ static void cluster_pmu_set_resr(struct cluster_pmu *cluster,
spin_lock_irqsave(&cluster->pmu_lock, flags);
- resr_val = get_l2_indirect_reg(L2PMRESR);
+ resr_val = kryo_l2_get_indirect_reg(L2PMRESR);
resr_val &= ~(L2PMRESR_GROUP_MASK << shift);
resr_val |= field;
resr_val |= L2PMRESR_EN;
- set_l2_indirect_reg(L2PMRESR, resr_val);
+ kryo_l2_set_indirect_reg(L2PMRESR, resr_val);
spin_unlock_irqrestore(&cluster->pmu_lock, flags);
}
@@ -323,14 +281,14 @@ static inline void cluster_pmu_set_evfilter_sys_mode(u32 ctr)
L2PMXEVFILTER_ORGFILTER_IDINDEP |
L2PMXEVFILTER_ORGFILTER_ALL;
- set_l2_indirect_reg(reg_idx(IA_L2PMXEVFILTER, ctr), val);
+ kryo_l2_set_indirect_reg(reg_idx(IA_L2PMXEVFILTER, ctr), val);
}
static inline u32 cluster_pmu_getreset_ovsr(void)
{
- u32 result = get_l2_indirect_reg(L2PMOVSSET);
+ u32 result = kryo_l2_get_indirect_reg(L2PMOVSSET);
- set_l2_indirect_reg(L2PMOVSCLR, result);
+ kryo_l2_set_indirect_reg(L2PMOVSCLR, result);
return result;
}
@@ -783,7 +741,7 @@ static int get_num_counters(void)
{
int val;
- val = get_l2_indirect_reg(L2PMCR);
+ val = kryo_l2_get_indirect_reg(L2PMCR);
/*
* Read number of counters from L2PMCR and add 1
diff --git a/drivers/soc/qcom/Kconfig b/drivers/soc/qcom/Kconfig
index 7093fe7..0567dff 100644
--- a/drivers/soc/qcom/Kconfig
+++ b/drivers/soc/qcom/Kconfig
@@ -39,6 +39,9 @@ config QCOM_GSBI
functions for connecting the underlying serial UART, SPI, and I2C
devices to the output pins.
+config QCOM_KRYO_L2_ACCESSORS
+ bool
+
config QCOM_MDT_LOADER
tristate
select QCOM_SCM
diff --git a/drivers/soc/qcom/Makefile b/drivers/soc/qcom/Makefile
index cbf414c..e4d3f5a 100644
--- a/drivers/soc/qcom/Makefile
+++ b/drivers/soc/qcom/Makefile
@@ -14,3 +14,4 @@ obj-$(CONFIG_QCOM_SMEM_STATE) += smem_state.o
obj-$(CONFIG_QCOM_SMP2P) += smp2p.o
obj-$(CONFIG_QCOM_SMSM) += smsm.o
obj-$(CONFIG_QCOM_WCNSS_CTRL) += wcnss_ctrl.o
+obj-$(CONFIG_QCOM_KRYO_L2_ACCESSORS) += kryo-l2-accessors.o
diff --git a/drivers/soc/qcom/kryo-l2-accessors.c b/drivers/soc/qcom/kryo-l2-accessors.c
new file mode 100644
index 0000000..75fd07a
--- /dev/null
+++ b/drivers/soc/qcom/kryo-l2-accessors.c
@@ -0,0 +1,56 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/spinlock.h>
+#include <asm/sysreg.h>
+#include <soc/qcom/kryo-l2-accessors.h>
+
+#define L2CPUSRSELR_EL1 sys_reg(3, 3, 15, 0, 6)
+#define L2CPUSRDR_EL1 sys_reg(3, 3, 15, 0, 7)
+
+static DEFINE_RAW_SPINLOCK(l2_access_lock);
+
+/**
+ * kryo_l2_set_indirect_reg() - write value to an L2 register
+ * @reg: Address of L2 register.
+ * @value: Value to be written to register.
+ *
+ * Use architecturally required barriers for ordering between system register
+ * accesses, and system registers with respect to device memory
+ */
+void kryo_l2_set_indirect_reg(u64 reg, u64 val)
+{
+ unsigned long flags;
+
+ raw_spin_lock_irqsave(&l2_access_lock, flags);
+ write_sysreg_s(reg, L2CPUSRSELR_EL1);
+ isb();
+ write_sysreg_s(val, L2CPUSRDR_EL1);
+ isb();
+ raw_spin_unlock_irqrestore(&l2_access_lock, flags);
+}
+EXPORT_SYMBOL(kryo_l2_set_indirect_reg);
+
+/**
+ * kryo_l2_get_indirect_reg() - read an L2 register value
+ * @reg: Address of L2 register.
+ *
+ * Use architecturally required barriers for ordering between system register
+ * accesses, and system registers with respect to device memory
+ */
+u64 kryo_l2_get_indirect_reg(u64 reg)
+{
+ u64 val;
+ unsigned long flags;
+
+ raw_spin_lock_irqsave(&l2_access_lock, flags);
+ write_sysreg_s(reg, L2CPUSRSELR_EL1);
+ isb();
+ val = read_sysreg_s(L2CPUSRDR_EL1);
+ raw_spin_unlock_irqrestore(&l2_access_lock, flags);
+
+ return val;
+}
+EXPORT_SYMBOL(kryo_l2_get_indirect_reg);
diff --git a/include/soc/qcom/kryo-l2-accessors.h b/include/soc/qcom/kryo-l2-accessors.h
new file mode 100644
index 0000000..673c534
--- /dev/null
+++ b/include/soc/qcom/kryo-l2-accessors.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2018, The Linux Foundation. All rights reserved.
+ */
+
+#ifndef __SOC_ARCH_QCOM_KRYO_L2_ACCESSORS_H
+#define __SOC_ARCH_QCOM_KRYO_L2_ACCESSORS_H
+
+void kryo_l2_set_indirect_reg(u64 reg, u64 val);
+u64 kryo_l2_get_indirect_reg(u64 reg);
+
+#endif
--
1.9.1
Each of the CPU clusters (Power and Perf) on msm8996 are
clocked via 2 PLLs, a primary and alternate. There are also
2 Mux'es, a primary and secondary all connected together
as shown below
+-------+
XO | |
+------------------>0 |
| |
PLL/2 | SMUX +----+
+------->1 | |
| | | |
| +-------+ | +-------+
| +---->0 |
| | |
+---------------+ | +----------->1 | CPU clk
|Primary PLL +----+ PLL_EARLY | | +------>
| +------+-----------+ +------>2 PMUX |
+---------------+ | | | |
| +------+ | +-->3 |
+--^+ ACD +-----+ | +-------+
+---------------+ +------+ |
|Alt PLL | |
| +---------------------------+
+---------------+ PLL_EARLY
The primary PLL is what drives the CPU clk, except for times
when we are reprogramming the PLL itself (for rate changes) when
we temporarily switch to an alternate PLL. A subsequent patch adds
support to switch between primary and alternate PLL during rate
changes.
The primary PLL operates on a single VCO range, between 600MHz
and 3GHz. However the CPUs do support OPPs with frequencies
between 300MHz and 600MHz. In order to support running the CPUs
at those frequencies we end up having to lock the PLL at twice
the rate and drive the CPU clk via the PLL/2 output and SMUX.
Signed-off-by: Ilia Lin <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
---
Documentation/devicetree/bindings/clock/qcom,kryocc.txt | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
create mode 100644 Documentation/devicetree/bindings/clock/qcom,kryocc.txt
diff --git a/Documentation/devicetree/bindings/clock/qcom,kryocc.txt b/Documentation/devicetree/bindings/clock/qcom,kryocc.txt
new file mode 100644
index 0000000..8458783
--- /dev/null
+++ b/Documentation/devicetree/bindings/clock/qcom,kryocc.txt
@@ -0,0 +1,17 @@
+Qualcomm CPUSS clock controller for Kryo CPUs
+----------------------------------------------------
+
+Required properties :
+- compatible : shall contain only one of the following:
+
+ "qcom,msm8996-apcc"
+
+- reg : shall contain base register location and length
+- #clock-cells : shall contain 1
+
+Example:
+ kryocc: clock-controller@6400000 {
+ compatible = "qcom,msm8996-apcc";
+ reg = <0x6400000 0x90000>;
+ #clock-cells = <1>;
+ };
--
1.9.1
On Thu, May 24, 2018 at 10:50 AM, Ilia Lin <[email protected]> wrote:
> [v12]
> * Addressed the kbuild fail on arm architecture
>
> [v11]
> * Split the series into domains
>
> [v9]
> * Addressed comments from Viresh and Russel about the error handling
>
> [v8]
> * Reordered the patch series into 4 groups
> * Addressed comments from Amit about the comments and commit messages
> * Addressed comments from Amit and Viresh about the resourses deallocation
>
> [v7]
> * Addressed comments from Viresh about resourses deallocation
> and DT compatible
>
> [v6]
> * Addressed comments from Viresh about:
> ** Comments style
> ** Kconfig bool instead of tristate
> ** DT and documentation style
> ** Resourses deallocation on an error
> ** Typos
>
> [v5]
> * Rebased
> * Addressed comments from Bjorn about SPDX style,
> functions and parameters naming
> * Addressed comments from Viresh DT properties and style, comments style,
> resourses deallocation, documentation placement
> * Addressed comments from Sricharan about unnessesary include
> * Addressed comments from Nicolas
> * Addressed comments from Rob about the commit messages and acks
> * Addressed comments from Mark
>
> [v4]
> * Adressed all comments from Stephen
> * Added CPU regulator support
> * Added qcom-cpufreq-kryo driver
>
> [v3]
> * Rebased on top of the latest PLL driver changes
> * Addressed comment from Rob Herring for bindings
>
> [v2]
> * Addressed comments from Rob Herring for bindings
> * Addressed comments from Mark Rutland for memory barrier
> * Addressed comments from Julien Thierry for clock reenabling condition
> * Tuned the HW configuration for clock frequencies below 600MHz
>
> SOC (1/15):
> Extracts the kryo l2 accessors driver from the QCOM PMU driver
>
> Clocks (2/15-9/15):
> This series adds support for the CPU clocks on msm8996 devices.
> The driver uses the existing PLL drivers and is required to control
> the CPU frequency scaling on the MSM8996.
>
> Ilia Lin (6):
> soc: qcom: Separate kryo l2 accessors from PMU driver
> clk: Use devm_ in the register fixed factor clock
> clk: qcom: Add CPU clock driver for msm8996
> dt-bindings: clk: qcom: Add bindings for CPU clock for msm8996
> clk: qcom: cpu-8996: Add support to switch below 600Mhz
> clk: qcom: Add ACD path to CPU clock driver for msm8996
>
> Rajendra Nayak (2):
> clk: qcom: Make clk_alpha_pll_configure available to modules
> clk: qcom: cpu-8996: Add support to switch to alternate PLL
>
> .../devicetree/bindings/clock/qcom,kryocc.txt | 17 +
> drivers/clk/clk-fixed-factor.c | 2 +-
> drivers/clk/qcom/Kconfig | 10 +
> drivers/clk/qcom/Makefile | 1 +
> drivers/clk/qcom/clk-alpha-pll.c | 1 +
> drivers/clk/qcom/clk-alpha-pll.h | 6 +
> drivers/clk/qcom/clk-cpu-8996.c | 510 +++++++++++++++++++++
> drivers/perf/Kconfig | 1 +
> drivers/perf/qcom_l2_pmu.c | 90 +---
> drivers/soc/qcom/Kconfig | 3 +
> drivers/soc/qcom/Makefile | 1 +
> drivers/soc/qcom/kryo-l2-accessors.c | 56 +++
> include/soc/qcom/kryo-l2-accessors.h | 12 +
For the perf/l2-accessors part, Reviewed-by: Amit Kucheria
<[email protected]>
For the entire series, Tested-by: Amit Kucheria <[email protected]>