[PATCH v3 0/5] hwmon: k10temp driver improvements

[PATCH v3 0/5] hwmon: k10temp driver improvements

[Guenter Roeck]

This patch series implements various improvements for the k10temp driver.

Patch 1/5 introduces the use of bit operations.

Patch 2/5 converts the driver to use the devm_hwmon_device_register_with_info
API. This not only simplifies the code and reduces its size, it also
makes the code easier to maintain and enhance. 

Patch 3/5 adds support for reporting Core Complex Die (CCD) temperatures
on Zen2 (Ryzen and Threadripper) CPUs (note that reporting is incomplete
for Threadripper CPUs - it is known that additional temperature sensors
exist, but the register locations are unknown).

Patch 4/5 adds support for reporting core and SoC current and voltage
information on Ryzen CPUs (note: voltage and current measurements for
Threadripper and EPYC CPUs are known to exist, but register locations
are unknown, and values are therefore not reported at this time).

Patch 5/5 removes the maximum temperature from Tdie for Ryzen CPUs.
It is inaccurate, misleading, and it just doesn't make sense to report
wrong information.

With all patches in place, output on Ryzen 3900X CPUs looks as follows
(with the system under load).

k10temp-pci-00c3
Adapter: PCI adapter
Vcore:        +1.36 V
Vsoc:         +1.18 V
Tdie:         +86.8°C
Tctl:         +86.8°C
Tccd1:        +80.0°C
Tccd2:        +81.8°C
Icore:       +44.14 A
Isoc:        +13.83 A

The voltage and current information is limited to Ryzen CPUs. Voltage
and current reporting on Threadripper and EPYC CPUs is different, and the
reported information is either incomplete or wrong. Exclude it for the time
being; it can always be added if/when more information becomes available.

Tested with the following Ryzen CPUs:
    1300X A user with this CPU in the system reported somewhat unexpected
          values for Vcore; it isn't entirely if at all clear why that is
          the case. Overall this does not warrant holding up the series.
    1600
    1800X
    2200G
    2400G
    2700
    2700X
    2950X
    3600X
    3800X
    3900X
    3950X
    3970X
    EPYC 7302
    EPYC 7742

Many thanks to everyone who helped to test this series.

---
v2: Added Tested-by: tags as received.
    Don't display voltage and current information for Threadripper and EPYC.
    Stop displaying the fixed (and wrong) maximum temperature of 70 degrees C
    for Tdie on model 17h/18h CPUs.

v3: Added more Tested-by: tags
    Added detection for 3970X, and report Tccd1 for this CPU.

[PATCH v3 1/5] hwmon: (k10temp) Use bitops

[Guenter Roeck]

Using bitops makes bit masks and shifts easier to read.

Tested-by: Brad Campbell <lists2009@fnarfbargle.com>
Tested-by: Bernhard Gebetsberger <bernhard.gebetsberger@gmx.at>
Tested-by: Holger Kiehl <holger.kiehl@dwd.de>
Tested-by: Michael Larabel <michael@phoronix.com>
Tested-by: Jonathan McDowell <noodles@earth.li>
Tested-by: Ken Moffat <zarniwhoop73@googlemail.com>
Tested-by: Darren Salt <devspam@moreofthesa.me.uk>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/k10temp.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/hwmon/k10temp.c b/drivers/hwmon/k10temp.c
index 5c1dddde193c..8807d7da68db 100644
--- a/drivers/hwmon/k10temp.c
+++ b/drivers/hwmon/k10temp.c
@@ -5,6 +5,7 @@
  * Copyright (c) 2009 Clemens Ladisch <clemens@ladisch.de>
  */
 
+#include <linux/bitops.h>
 #include <linux/err.h>
 #include <linux/hwmon.h>
 #include <linux/hwmon-sysfs.h>
@@ -31,22 +32,22 @@ static DEFINE_MUTEX(nb_smu_ind_mutex);
 #endif
 
 /* CPUID function 0x80000001, ebx */
-#define CPUID_PKGTYPE_MASK	0xf0000000
+#define CPUID_PKGTYPE_MASK	GENMASK(31, 28)
 #define CPUID_PKGTYPE_F		0x00000000
 #define CPUID_PKGTYPE_AM2R2_AM3	0x10000000
 
 /* DRAM controller (PCI function 2) */
 #define REG_DCT0_CONFIG_HIGH		0x094
-#define  DDR3_MODE			0x00000100
+#define  DDR3_MODE			BIT(8)
 
 /* miscellaneous (PCI function 3) */
 #define REG_HARDWARE_THERMAL_CONTROL	0x64
-#define  HTC_ENABLE			0x00000001
+#define  HTC_ENABLE			BIT(0)
 
 #define REG_REPORTED_TEMPERATURE	0xa4
 
 #define REG_NORTHBRIDGE_CAPABILITIES	0xe8
-#define  NB_CAP_HTC			0x00000400
+#define  NB_CAP_HTC			BIT(10)
 
 /*
  * For F15h M60h and M70h, REG_HARDWARE_THERMAL_CONTROL
@@ -60,6 +61,9 @@ static DEFINE_MUTEX(nb_smu_ind_mutex);
 /* F17h M01h Access througn SMN */
 #define F17H_M01H_REPORTED_TEMP_CTRL_OFFSET	0x00059800
 
+#define CUR_TEMP_SHIFT				21
+#define CUR_TEMP_RANGE_SEL_MASK			BIT(19)
+
 struct k10temp_data {
 	struct pci_dev *pdev;
 	void (*read_htcreg)(struct pci_dev *pdev, u32 *regval);
@@ -129,7 +133,7 @@ static unsigned int get_raw_temp(struct k10temp_data *data)
 	u32 regval;
 
 	data->read_tempreg(data->pdev, &regval);
-	temp = (regval >> 21) * 125;
+	temp = (regval >> CUR_TEMP_SHIFT) * 125;
 	if (regval & data->temp_adjust_mask)
 		temp -= 49000;
 	return temp;
@@ -312,7 +316,7 @@ static int k10temp_probe(struct pci_dev *pdev,
 		data->read_htcreg = read_htcreg_nb_f15;
 		data->read_tempreg = read_tempreg_nb_f15;
 	} else if (boot_cpu_data.x86 == 0x17 || boot_cpu_data.x86 == 0x18) {
-		data->temp_adjust_mask = 0x80000;
+		data->temp_adjust_mask = CUR_TEMP_RANGE_SEL_MASK;
 		data->read_tempreg = read_tempreg_nb_f17;
 		data->show_tdie = true;
 	} else {
-- 
2.17.1

[PATCH v3 2/5] hmon: (k10temp) Convert to use devm_hwmon_device_register_with_info

[Guenter Roeck]

Convert driver to use devm_hwmon_device_register_with_info to simplify
the code and to reduce its size.

Old size (x86_64):
   text	   data	    bss	    dec	    hex	filename
   8247	   4488	     64	  12799	   31ff	drivers/hwmon/k10temp.o
New size:
   text	   data	    bss	    dec	    hex	filename
   6778	   2792	     64	   9634	   25a2	drivers/hwmon/k10temp.o

Tested-by: Brad Campbell <lists2009@fnarfbargle.com>
Tested-by: Bernhard Gebetsberger <bernhard.gebetsberger@gmx.at>
Tested-by: Holger Kiehl <holger.kiehl@dwd.de>
Tested-by: Michael Larabel <michael@phoronix.com>
Tested-by: Jonathan McDowell <noodles@earth.li>
Tested-by: Ken Moffat <zarniwhoop73@googlemail.com>
Tested-by: Darren Salt <devspam@moreofthesa.me.uk>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/k10temp.c | 213 +++++++++++++++++++++-------------------
 1 file changed, 112 insertions(+), 101 deletions(-)

diff --git a/drivers/hwmon/k10temp.c b/drivers/hwmon/k10temp.c
index 8807d7da68db..c45f6498a59b 100644
--- a/drivers/hwmon/k10temp.c
+++ b/drivers/hwmon/k10temp.c
@@ -1,14 +1,15 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 /*
- * k10temp.c - AMD Family 10h/11h/12h/14h/15h/16h processor hardware monitoring
+ * k10temp.c - AMD Family 10h/11h/12h/14h/15h/16h/17h
+ *		processor hardware monitoring
  *
  * Copyright (c) 2009 Clemens Ladisch <clemens@ladisch.de>
+ * Copyright (c) 2020 Guenter Roeck <linux@roeck-us.net>
  */
 
 #include <linux/bitops.h>
 #include <linux/err.h>
 #include <linux/hwmon.h>
-#include <linux/hwmon-sysfs.h>
 #include <linux/init.h>
 #include <linux/module.h>
 #include <linux/pci.h>
@@ -127,10 +128,10 @@ static void read_tempreg_nb_f17(struct pci_dev *pdev, u32 *regval)
 		     F17H_M01H_REPORTED_TEMP_CTRL_OFFSET, regval);
 }
 
-static unsigned int get_raw_temp(struct k10temp_data *data)
+static long get_raw_temp(struct k10temp_data *data)
 {
-	unsigned int temp;
 	u32 regval;
+	long temp;
 
 	data->read_tempreg(data->pdev, &regval);
 	temp = (regval >> CUR_TEMP_SHIFT) * 125;
@@ -139,118 +140,108 @@ static unsigned int get_raw_temp(struct k10temp_data *data)
 	return temp;
 }
 
-static ssize_t temp1_input_show(struct device *dev,
-				struct device_attribute *attr, char *buf)
-{
-	struct k10temp_data *data = dev_get_drvdata(dev);
-	unsigned int temp = get_raw_temp(data);
-
-	if (temp > data->temp_offset)
-		temp -= data->temp_offset;
-	else
-		temp = 0;
-
-	return sprintf(buf, "%u\n", temp);
-}
-
-static ssize_t temp2_input_show(struct device *dev,
-				struct device_attribute *devattr, char *buf)
-{
-	struct k10temp_data *data = dev_get_drvdata(dev);
-	unsigned int temp = get_raw_temp(data);
-
-	return sprintf(buf, "%u\n", temp);
-}
-
-static ssize_t temp_label_show(struct device *dev,
-			       struct device_attribute *devattr, char *buf)
-{
-	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
-
-	return sprintf(buf, "%s\n", attr->index ? "Tctl" : "Tdie");
-}
+const char *k10temp_temp_label[] = {
+	"Tdie",
+	"Tctl",
+};
 
-static ssize_t temp1_max_show(struct device *dev,
-			      struct device_attribute *attr, char *buf)
+static int k10temp_read_labels(struct device *dev,
+			       enum hwmon_sensor_types type,
+			       u32 attr, int channel, const char **str)
 {
-	return sprintf(buf, "%d\n", 70 * 1000);
+	*str = k10temp_temp_label[channel];
+	return 0;
 }
 
-static ssize_t temp_crit_show(struct device *dev,
-			      struct device_attribute *devattr, char *buf)
+static int k10temp_read(struct device *dev, enum hwmon_sensor_types type,
+			u32 attr, int channel, long *val)
 {
-	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 	struct k10temp_data *data = dev_get_drvdata(dev);
-	int show_hyst = attr->index;
 	u32 regval;
-	int value;
 
-	data->read_htcreg(data->pdev, &regval);
-	value = ((regval >> 16) & 0x7f) * 500 + 52000;
-	if (show_hyst)
-		value -= ((regval >> 24) & 0xf) * 500;
-	return sprintf(buf, "%d\n", value);
+	switch (attr) {
+	case hwmon_temp_input:
+		switch (channel) {
+		case 0:		/* Tdie */
+			*val = get_raw_temp(data) - data->temp_offset;
+			if (*val < 0)
+				*val = 0;
+			break;
+		case 1:		/* Tctl */
+			*val = get_raw_temp(data);
+			if (*val < 0)
+				*val = 0;
+			break;
+		default:
+			return -EOPNOTSUPP;
+		}
+		break;
+	case hwmon_temp_max:
+		*val = 70 * 1000;
+		break;
+	case hwmon_temp_crit:
+		data->read_htcreg(data->pdev, &regval);
+		*val = ((regval >> 16) & 0x7f) * 500 + 52000;
+		break;
+	case hwmon_temp_crit_hyst:
+		data->read_htcreg(data->pdev, &regval);
+		*val = (((regval >> 16) & 0x7f)
+			- ((regval >> 24) & 0xf)) * 500 + 52000;
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+	return 0;
 }
 
-static DEVICE_ATTR_RO(temp1_input);
-static DEVICE_ATTR_RO(temp1_max);
-static SENSOR_DEVICE_ATTR_RO(temp1_crit, temp_crit, 0);
-static SENSOR_DEVICE_ATTR_RO(temp1_crit_hyst, temp_crit, 1);
-
-static SENSOR_DEVICE_ATTR_RO(temp1_label, temp_label, 0);
-static DEVICE_ATTR_RO(temp2_input);
-static SENSOR_DEVICE_ATTR_RO(temp2_label, temp_label, 1);
-
-static umode_t k10temp_is_visible(struct kobject *kobj,
-				  struct attribute *attr, int index)
+static umode_t k10temp_is_visible(const void *_data,
+				  enum hwmon_sensor_types type,
+				  u32 attr, int channel)
 {
-	struct device *dev = container_of(kobj, struct device, kobj);
-	struct k10temp_data *data = dev_get_drvdata(dev);
+	const struct k10temp_data *data = _data;
 	struct pci_dev *pdev = data->pdev;
 	u32 reg;
 
-	switch (index) {
-	case 0 ... 1:	/* temp1_input, temp1_max */
-	default:
-		break;
-	case 2 ... 3:	/* temp1_crit, temp1_crit_hyst */
-		if (!data->read_htcreg)
-			return 0;
-
-		pci_read_config_dword(pdev, REG_NORTHBRIDGE_CAPABILITIES,
-				      &reg);
-		if (!(reg & NB_CAP_HTC))
-			return 0;
-
-		data->read_htcreg(data->pdev, &reg);
-		if (!(reg & HTC_ENABLE))
-			return 0;
-		break;
-	case 4 ... 6:	/* temp1_label, temp2_input, temp2_label */
-		if (!data->show_tdie)
+	switch (type) {
+	case hwmon_temp:
+		switch (attr) {
+		case hwmon_temp_input:
+			if (channel && !data->show_tdie)
+				return 0;
+			break;
+		case hwmon_temp_max:
+			if (channel)
+				return 0;
+			break;
+		case hwmon_temp_crit:
+		case hwmon_temp_crit_hyst:
+			if (channel || !data->read_htcreg)
+				return 0;
+
+			pci_read_config_dword(pdev,
+					      REG_NORTHBRIDGE_CAPABILITIES,
+					      &reg);
+			if (!(reg & NB_CAP_HTC))
+				return 0;
+
+			data->read_htcreg(data->pdev, &reg);
+			if (!(reg & HTC_ENABLE))
+				return 0;
+			break;
+		case hwmon_temp_label:
+			if (!data->show_tdie)
+				return 0;
+			break;
+		default:
 			return 0;
+		}
 		break;
+	default:
+		return 0;
 	}
-	return attr->mode;
+	return 0444;
 }
 
-static struct attribute *k10temp_attrs[] = {
-	&dev_attr_temp1_input.attr,
-	&dev_attr_temp1_max.attr,
-	&sensor_dev_attr_temp1_crit.dev_attr.attr,
-	&sensor_dev_attr_temp1_crit_hyst.dev_attr.attr,
-	&sensor_dev_attr_temp1_label.dev_attr.attr,
-	&dev_attr_temp2_input.attr,
-	&sensor_dev_attr_temp2_label.dev_attr.attr,
-	NULL
-};
-
-static const struct attribute_group k10temp_group = {
-	.attrs = k10temp_attrs,
-	.is_visible = k10temp_is_visible,
-};
-__ATTRIBUTE_GROUPS(k10temp);
-
 static bool has_erratum_319(struct pci_dev *pdev)
 {
 	u32 pkg_type, reg_dram_cfg;
@@ -285,8 +276,27 @@ static bool has_erratum_319(struct pci_dev *pdev)
 	       (boot_cpu_data.x86_model == 4 && boot_cpu_data.x86_stepping <= 2);
 }
 
-static int k10temp_probe(struct pci_dev *pdev,
-				   const struct pci_device_id *id)
+static const struct hwmon_channel_info *k10temp_info[] = {
+	HWMON_CHANNEL_INFO(temp,
+			   HWMON_T_INPUT | HWMON_T_MAX |
+			   HWMON_T_CRIT | HWMON_T_CRIT_HYST |
+			   HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL),
+	NULL
+};
+
+static const struct hwmon_ops k10temp_hwmon_ops = {
+	.is_visible = k10temp_is_visible,
+	.read = k10temp_read,
+	.read_string = k10temp_read_labels,
+};
+
+static const struct hwmon_chip_info k10temp_chip_info = {
+	.ops = &k10temp_hwmon_ops,
+	.info = k10temp_info,
+};
+
+static int k10temp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	int unreliable = has_erratum_319(pdev);
 	struct device *dev = &pdev->dev;
@@ -334,8 +344,9 @@ static int k10temp_probe(struct pci_dev *pdev,
 		}
 	}
 
-	hwmon_dev = devm_hwmon_device_register_with_groups(dev, "k10temp", data,
-							   k10temp_groups);
+	hwmon_dev = devm_hwmon_device_register_with_info(dev, "k10temp", data,
+							 &k10temp_chip_info,
+							 NULL);
 	return PTR_ERR_OR_ZERO(hwmon_dev);
 }
 
-- 
2.17.1

[PATCH v3 3/5] hwmon: (k10temp) Report temperatures per CPU die

[Guenter Roeck]

Zen2 reports reporting temperatures per CPU die (called Core Complex Dies,
or CCD, by AMD). Add support for it to the k10temp driver.

Tested-by: Brad Campbell <lists2009@fnarfbargle.com>
Tested-by: Bernhard Gebetsberger <bernhard.gebetsberger@gmx.at>
Tested-by: Holger Kiehl <holger.kiehl@dwd.de>
Tested-by: Michael Larabel <michael@phoronix.com>
Tested-by: Jonathan McDowell <noodles@earth.li>
Tested-by: Ken Moffat <zarniwhoop73@googlemail.com>
Tested-by: Darren Salt <devspam@moreofthesa.me.uk>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/k10temp.c | 80 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 79 insertions(+), 1 deletion(-)

diff --git a/drivers/hwmon/k10temp.c b/drivers/hwmon/k10temp.c
index c45f6498a59b..0af096b061fa 100644
--- a/drivers/hwmon/k10temp.c
+++ b/drivers/hwmon/k10temp.c
@@ -5,6 +5,12 @@
  *
  * Copyright (c) 2009 Clemens Ladisch <clemens@ladisch.de>
  * Copyright (c) 2020 Guenter Roeck <linux@roeck-us.net>
+ *
+ * Implementation notes:
+ * - CCD1 and CCD2 register address information as well as the calculation to
+ *   convert raw register values is from https://github.com/ocerman/zenpower.
+ *   The information is not confirmed from chip datasheets, but experiments
+ *   suggest that it provides reasonable temperature values.
  */
 
 #include <linux/bitops.h>
@@ -61,6 +67,8 @@ static DEFINE_MUTEX(nb_smu_ind_mutex);
 
 /* F17h M01h Access througn SMN */
 #define F17H_M01H_REPORTED_TEMP_CTRL_OFFSET	0x00059800
+#define F17H_M70H_CCD1_TEMP			0x00059954
+#define F17H_M70H_CCD2_TEMP			0x00059958
 
 #define CUR_TEMP_SHIFT				21
 #define CUR_TEMP_RANGE_SEL_MASK			BIT(19)
@@ -72,6 +80,8 @@ struct k10temp_data {
 	int temp_offset;
 	u32 temp_adjust_mask;
 	bool show_tdie;
+	bool show_tccd1;
+	bool show_tccd2;
 };
 
 struct tctl_offset {
@@ -143,6 +153,8 @@ static long get_raw_temp(struct k10temp_data *data)
 const char *k10temp_temp_label[] = {
 	"Tdie",
 	"Tctl",
+	"Tccd1",
+	"Tccd2",
 };
 
 static int k10temp_read_labels(struct device *dev,
@@ -172,6 +184,16 @@ static int k10temp_read(struct device *dev, enum hwmon_sensor_types type,
 			if (*val < 0)
 				*val = 0;
 			break;
+		case 2:		/* Tccd1 */
+			amd_smn_read(amd_pci_dev_to_node_id(data->pdev),
+				     F17H_M70H_CCD1_TEMP, &regval);
+			*val = (regval & 0xfff) * 125 - 305000;
+			break;
+		case 3:		/* Tccd2 */
+			amd_smn_read(amd_pci_dev_to_node_id(data->pdev),
+				     F17H_M70H_CCD2_TEMP, &regval);
+			*val = (regval & 0xfff) * 125 - 305000;
+			break;
 		default:
 			return -EOPNOTSUPP;
 		}
@@ -206,8 +228,24 @@ static umode_t k10temp_is_visible(const void *_data,
 	case hwmon_temp:
 		switch (attr) {
 		case hwmon_temp_input:
-			if (channel && !data->show_tdie)
+			switch (channel) {
+			case 0:		/* Tdie, or Tctl if we don't show it */
+				break;
+			case 1:		/* Tctl */
+				if (!data->show_tdie)
+					return 0;
+				break;
+			case 2:		/* Tccd1 */
+				if (!data->show_tccd1)
+					return 0;
+				break;
+			case 3:		/* Tccd2 */
+				if (!data->show_tccd2)
+					return 0;
+				break;
+			default:
 				return 0;
+			}
 			break;
 		case hwmon_temp_max:
 			if (channel)
@@ -229,8 +267,24 @@ static umode_t k10temp_is_visible(const void *_data,
 				return 0;
 			break;
 		case hwmon_temp_label:
+			/* No labels if we don't show the die temperature */
 			if (!data->show_tdie)
 				return 0;
+			switch (channel) {
+			case 0:		/* Tdie */
+			case 1:		/* Tctl */
+				break;
+			case 2:		/* Tccd1 */
+				if (!data->show_tccd1)
+					return 0;
+				break;
+			case 3:		/* Tccd2 */
+				if (!data->show_tccd2)
+					return 0;
+				break;
+			default:
+				return 0;
+			}
 			break;
 		default:
 			return 0;
@@ -281,6 +335,8 @@ static const struct hwmon_channel_info *k10temp_info[] = {
 			   HWMON_T_INPUT | HWMON_T_MAX |
 			   HWMON_T_CRIT | HWMON_T_CRIT_HYST |
 			   HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
+			   HWMON_T_INPUT | HWMON_T_LABEL,
 			   HWMON_T_INPUT | HWMON_T_LABEL),
 	NULL
 };
@@ -326,9 +382,31 @@ static int k10temp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		data->read_htcreg = read_htcreg_nb_f15;
 		data->read_tempreg = read_tempreg_nb_f15;
 	} else if (boot_cpu_data.x86 == 0x17 || boot_cpu_data.x86 == 0x18) {
+		u32 regval;
+
 		data->temp_adjust_mask = CUR_TEMP_RANGE_SEL_MASK;
 		data->read_tempreg = read_tempreg_nb_f17;
 		data->show_tdie = true;
+
+		switch (boot_cpu_data.x86_model) {
+		case 0x1:	/* Zen */
+		case 0x8:	/* Zen+ */
+		case 0x11:	/* Zen APU */
+		case 0x18:	/* Zen+ APU */
+			break;
+		case 0x31:	/* Zen2 Threadripper */
+		case 0x71:	/* Zen2 */
+			amd_smn_read(amd_pci_dev_to_node_id(pdev),
+				     F17H_M70H_CCD1_TEMP, &regval);
+			if (regval & 0xfff)
+				data->show_tccd1 = true;
+
+			amd_smn_read(amd_pci_dev_to_node_id(pdev),
+				     F17H_M70H_CCD2_TEMP, &regval);
+			if (regval & 0xfff)
+				data->show_tccd2 = true;
+			break;
+		}
 	} else {
 		data->read_htcreg = read_htcreg_pci;
 		data->read_tempreg = read_tempreg_pci;
-- 
2.17.1

[PATCH v3 4/5] hwmon: (k10temp) Show core and SoC current and voltages on Ryzen CPUs

[Guenter Roeck]

Ryzen CPUs report core and SoC voltages and currents. Add support
for it to the k10temp driver.

For the time being, only report voltages and currents for Ryzen
CPUs. Threadripper and EPYC appear to use a different mechanism.

Tested-by: Brad Campbell <lists2009@fnarfbargle.com>
Tested-by: Bernhard Gebetsberger <bernhard.gebetsberger@gmx.at>
Tested-by: Holger Kiehl <holger.kiehl@dwd.de>
Tested-by: Michael Larabel <michael@phoronix.com>
Tested-by: Jonathan McDowell <noodles@earth.li>
Tested-by: Ken Moffat <zarniwhoop73@googlemail.com>
Tested-by: Darren Salt <devspam@moreofthesa.me.uk>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/k10temp.c | 128 +++++++++++++++++++++++++++++++++++++++-
 1 file changed, 125 insertions(+), 3 deletions(-)

diff --git a/drivers/hwmon/k10temp.c b/drivers/hwmon/k10temp.c
index 0af096b061fa..9e8b1c1bcbbd 100644
--- a/drivers/hwmon/k10temp.c
+++ b/drivers/hwmon/k10temp.c
@@ -11,6 +11,15 @@
  *   convert raw register values is from https://github.com/ocerman/zenpower.
  *   The information is not confirmed from chip datasheets, but experiments
  *   suggest that it provides reasonable temperature values.
+ * - Register addresses to read chip voltage and current is also from
+ *   https://github.com/ocerman/zenpower, and not confirmed from chip
+ *   datasheets. Experiments suggest that reported current and voltage
+ *   information is reasonable.
+ * - It is unknown if the mechanism to read CCD1/CCD2 temperature as well as
+ *   current and voltage information works on higher-end Ryzen CPUs.
+ *   Information reported by Windows tools suggests that additional sensors
+ *   (both temperature and voltage/current) are supported, but their register
+ *   location is currently unknown.
  */
 
 #include <linux/bitops.h>
@@ -70,6 +79,10 @@ static DEFINE_MUTEX(nb_smu_ind_mutex);
 #define F17H_M70H_CCD1_TEMP			0x00059954
 #define F17H_M70H_CCD2_TEMP			0x00059958
 
+#define F17H_M01H_SVI				0x0005A000
+#define F17H_M01H_SVI_TEL_PLANE0		(F17H_M01H_SVI + 0xc)
+#define F17H_M01H_SVI_TEL_PLANE1		(F17H_M01H_SVI + 0x10)
+
 #define CUR_TEMP_SHIFT				21
 #define CUR_TEMP_RANGE_SEL_MASK			BIT(19)
 
@@ -82,6 +95,9 @@ struct k10temp_data {
 	bool show_tdie;
 	bool show_tccd1;
 	bool show_tccd2;
+	u32 svi_addr[2];
+	bool show_current;
+	int cfactor[2];
 };
 
 struct tctl_offset {
@@ -99,6 +115,16 @@ static const struct tctl_offset tctl_offset_table[] = {
 	{ 0x17, "AMD Ryzen Threadripper 29", 27000 }, /* 29{20,50,70,90}[W]X */
 };
 
+static bool is_threadripper(void)
+{
+	return strstr(boot_cpu_data.x86_model_id, "Threadripper");
+}
+
+static bool is_epyc(void)
+{
+	return strstr(boot_cpu_data.x86_model_id, "EPYC");
+}
+
 static void read_htcreg_pci(struct pci_dev *pdev, u32 *regval)
 {
 	pci_read_config_dword(pdev, REG_HARDWARE_THERMAL_CONTROL, regval);
@@ -157,16 +183,76 @@ const char *k10temp_temp_label[] = {
 	"Tccd2",
 };
 
+const char *k10temp_in_label[] = {
+	"Vcore",
+	"Vsoc",
+};
+
+const char *k10temp_curr_label[] = {
+	"Icore",
+	"Isoc",
+};
+
 static int k10temp_read_labels(struct device *dev,
 			       enum hwmon_sensor_types type,
 			       u32 attr, int channel, const char **str)
 {
-	*str = k10temp_temp_label[channel];
+	switch (type) {
+	case hwmon_temp:
+		*str = k10temp_temp_label[channel];
+		break;
+	case hwmon_in:
+		*str = k10temp_in_label[channel];
+		break;
+	case hwmon_curr:
+		*str = k10temp_curr_label[channel];
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
 	return 0;
 }
 
-static int k10temp_read(struct device *dev, enum hwmon_sensor_types type,
-			u32 attr, int channel, long *val)
+static int k10temp_read_curr(struct device *dev, u32 attr, int channel,
+			     long *val)
+{
+	struct k10temp_data *data = dev_get_drvdata(dev);
+	u32 regval;
+
+	switch (attr) {
+	case hwmon_curr_input:
+		amd_smn_read(amd_pci_dev_to_node_id(data->pdev),
+			     data->svi_addr[channel], &regval);
+		*val = DIV_ROUND_CLOSEST(data->cfactor[channel] *
+					 (regval & 0xff),
+					 1000);
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+	return 0;
+}
+
+static int k10temp_read_in(struct device *dev, u32 attr, int channel, long *val)
+{
+	struct k10temp_data *data = dev_get_drvdata(dev);
+	u32 regval;
+
+	switch (attr) {
+	case hwmon_in_input:
+		amd_smn_read(amd_pci_dev_to_node_id(data->pdev),
+			     data->svi_addr[channel], &regval);
+		regval = (regval >> 16) & 0xff;
+		*val = DIV_ROUND_CLOSEST(155000 - regval * 625, 100);
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+	return 0;
+}
+
+static int k10temp_read_temp(struct device *dev, u32 attr, int channel,
+			     long *val)
 {
 	struct k10temp_data *data = dev_get_drvdata(dev);
 	u32 regval;
@@ -216,6 +302,21 @@ static int k10temp_read(struct device *dev, enum hwmon_sensor_types type,
 	return 0;
 }
 
+static int k10temp_read(struct device *dev, enum hwmon_sensor_types type,
+			u32 attr, int channel, long *val)
+{
+	switch (type) {
+	case hwmon_temp:
+		return k10temp_read_temp(dev, attr, channel, val);
+	case hwmon_in:
+		return k10temp_read_in(dev, attr, channel, val);
+	case hwmon_curr:
+		return k10temp_read_curr(dev, attr, channel, val);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static umode_t k10temp_is_visible(const void *_data,
 				  enum hwmon_sensor_types type,
 				  u32 attr, int channel)
@@ -290,6 +391,11 @@ static umode_t k10temp_is_visible(const void *_data,
 			return 0;
 		}
 		break;
+	case hwmon_in:
+	case hwmon_curr:
+		if (!data->show_current)
+			return 0;
+		break;
 	default:
 		return 0;
 	}
@@ -338,6 +444,12 @@ static const struct hwmon_channel_info *k10temp_info[] = {
 			   HWMON_T_INPUT | HWMON_T_LABEL,
 			   HWMON_T_INPUT | HWMON_T_LABEL,
 			   HWMON_T_INPUT | HWMON_T_LABEL),
+	HWMON_CHANNEL_INFO(in,
+			   HWMON_I_INPUT | HWMON_I_LABEL,
+			   HWMON_I_INPUT | HWMON_I_LABEL),
+	HWMON_CHANNEL_INFO(curr,
+			   HWMON_C_INPUT | HWMON_C_LABEL,
+			   HWMON_C_INPUT | HWMON_C_LABEL),
 	NULL
 };
 
@@ -393,9 +505,19 @@ static int k10temp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		case 0x8:	/* Zen+ */
 		case 0x11:	/* Zen APU */
 		case 0x18:	/* Zen+ APU */
+			data->show_current = !is_threadripper() && !is_epyc();
+			data->svi_addr[0] = F17H_M01H_SVI_TEL_PLANE0;
+			data->svi_addr[1] = F17H_M01H_SVI_TEL_PLANE1;
+			data->cfactor[0] = 1039211;	/* core */
+			data->cfactor[1] = 360772;	/* SoC */
 			break;
 		case 0x31:	/* Zen2 Threadripper */
 		case 0x71:	/* Zen2 */
+			data->show_current = !is_threadripper() && !is_epyc();
+			data->cfactor[0] = 658823;	/* core */
+			data->cfactor[1] = 294300;	/* SoC */
+			data->svi_addr[0] = F17H_M01H_SVI_TEL_PLANE1;
+			data->svi_addr[1] = F17H_M01H_SVI_TEL_PLANE0;
 			amd_smn_read(amd_pci_dev_to_node_id(pdev),
 				     F17H_M70H_CCD1_TEMP, &regval);
 			if (regval & 0xfff)
-- 
2.17.1

[PATCH v3 5/5] hwmon: (k10temp) Don't show temperature limits on Ryzen (Zen) CPUs

[Guenter Roeck]

The maximum Tdie or Tctl is not published for Ryzen CPUs. What is
known, however, is that the traditional value of 70 degrees C is no
longer correct. Displaying it is meaningless, confusing, and wrong.
Stop doing it.

Tested-by: Brad Campbell <lists2009@fnarfbargle.com>
Tested-by: Holger Kiehl <holger.kiehl@dwd.de>
Tested-by: Michael Larabel <michael@phoronix.com>
Tested-by: Jonathan McDowell <noodles@earth.li>
Tested-by: Ken Moffat <zarniwhoop73@googlemail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/k10temp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/hwmon/k10temp.c b/drivers/hwmon/k10temp.c
index 9e8b1c1bcbbd..6076a84cb3ae 100644
--- a/drivers/hwmon/k10temp.c
+++ b/drivers/hwmon/k10temp.c
@@ -349,7 +349,7 @@ static umode_t k10temp_is_visible(const void *_data,
 			}
 			break;
 		case hwmon_temp_max:
-			if (channel)
+			if (channel || data->show_tdie)
 				return 0;
 			break;
 		case hwmon_temp_crit:
-- 
2.17.1