[linux.git] / drivers / hwmon / fam15h_power.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * fam15h_power.c - AMD Family 15h processor power monitoring
 *
 * Copyright (c) 2011-2016 Advanced Micro Devices, Inc.
 * Author: Andreas Herrmann <[email protected]>
 */

#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>
#include <linux/bitops.h>
#include <linux/cpu.h>
#include <linux/cpumask.h>
#include <linux/time.h>
#include <linux/sched.h>
#include <asm/processor.h>
#include <asm/msr.h>

MODULE_DESCRIPTION("AMD Family 15h CPU processor power monitor");
MODULE_AUTHOR("Andreas Herrmann <[email protected]>");
MODULE_LICENSE("GPL");

/* D18F3 */
#define REG_NORTHBRIDGE_CAP		0xe8

/* D18F4 */
#define REG_PROCESSOR_TDP		0x1b8

/* D18F5 */
#define REG_TDP_RUNNING_AVERAGE		0xe0
#define REG_TDP_LIMIT3			0xe8

#define FAM15H_MIN_NUM_ATTRS		2
#define FAM15H_NUM_GROUPS		2
#define MAX_CUS				8

/* set maximum interval as 1 second */
#define MAX_INTERVAL			1000

#define PCI_DEVICE_ID_AMD_15H_M70H_NB_F4 0x15b4

struct fam15h_power_data {
	struct pci_dev *pdev;
	unsigned int tdp_to_watts;
	unsigned int base_tdp;
	unsigned int processor_pwr_watts;
	unsigned int cpu_pwr_sample_ratio;
	const struct attribute_group *groups[FAM15H_NUM_GROUPS];
	struct attribute_group group;
	/* maximum accumulated power of a compute unit */
	u64 max_cu_acc_power;
	/* accumulated power of the compute units */
	u64 cu_acc_power[MAX_CUS];
	/* performance timestamp counter */
	u64 cpu_sw_pwr_ptsc[MAX_CUS];
	/* online/offline status of current compute unit */
	int cu_on[MAX_CUS];
	unsigned long power_period;
};

static bool is_carrizo_or_later(void)
{
	return boot_cpu_data.x86 == 0x15 && boot_cpu_data.x86_model >= 0x60;
}

static ssize_t power1_input_show(struct device *dev,
				 struct device_attribute *attr, char *buf)
{
	u32 val, tdp_limit, running_avg_range;
	s32 running_avg_capture;
	u64 curr_pwr_watts;
	struct fam15h_power_data *data = dev_get_drvdata(dev);
	struct pci_dev *f4 = data->pdev;

	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 5),
				  REG_TDP_RUNNING_AVERAGE, &val);

	/*
	 * On Carrizo and later platforms, TdpRunAvgAccCap bit field
	 * is extended to 4:31 from 4:25.
	 */
	if (is_carrizo_or_later()) {
		running_avg_capture = val >> 4;
		running_avg_capture = sign_extend32(running_avg_capture, 27);
	} else {
		running_avg_capture = (val >> 4) & 0x3fffff;
		running_avg_capture = sign_extend32(running_avg_capture, 21);
	}

	running_avg_range = (val & 0xf) + 1;

	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 5),
				  REG_TDP_LIMIT3, &val);

	/*
	 * On Carrizo and later platforms, ApmTdpLimit bit field
	 * is extended to 16:31 from 16:28.
	 */
	if (is_carrizo_or_later())
		tdp_limit = val >> 16;
	else
		tdp_limit = (val >> 16) & 0x1fff;

	curr_pwr_watts = ((u64)(tdp_limit +
				data->base_tdp)) << running_avg_range;
	curr_pwr_watts -= running_avg_capture;
	curr_pwr_watts *= data->tdp_to_watts;

	/*
	 * Convert to microWatt
	 *
	 * power is in Watt provided as fixed point integer with
	 * scaling factor 1/(2^16).  For conversion we use
	 * (10^6)/(2^16) = 15625/(2^10)
	 */
	curr_pwr_watts = (curr_pwr_watts * 15625) >> (10 + running_avg_range);
	return sprintf(buf, "%u\n", (unsigned int) curr_pwr_watts);
}
static DEVICE_ATTR_RO(power1_input);

static ssize_t power1_crit_show(struct device *dev,
				struct device_attribute *attr, char *buf)
{
	struct fam15h_power_data *data = dev_get_drvdata(dev);

	return sprintf(buf, "%u\n", data->processor_pwr_watts);
}
static DEVICE_ATTR_RO(power1_crit);

static void do_read_registers_on_cu(void *_data)
{
	struct fam15h_power_data *data = _data;
	int cpu, cu;

	cpu = smp_processor_id();

	/*
	 * With the new x86 topology modelling, cpu core id actually
	 * is compute unit id.
	 */
	cu = cpu_data(cpu).cpu_core_id;

	rdmsrl_safe(MSR_F15H_CU_PWR_ACCUMULATOR, &data->cu_acc_power[cu]);
	rdmsrl_safe(MSR_F15H_PTSC, &data->cpu_sw_pwr_ptsc[cu]);

	data->cu_on[cu] = 1;
}

/*
 * This function is only able to be called when CPUID
 * Fn8000_0007:EDX[12] is set.
 */
static int read_registers(struct fam15h_power_data *data)
{
	int core, this_core;
	cpumask_var_t mask;
	int ret, cpu;

	ret = zalloc_cpumask_var(&mask, GFP_KERNEL);
	if (!ret)
		return -ENOMEM;

	memset(data->cu_on, 0, sizeof(int) * MAX_CUS);

	get_online_cpus();

	/*
	 * Choose the first online core of each compute unit, and then
	 * read their MSR value of power and ptsc in a single IPI,
	 * because the MSR value of CPU core represent the compute
	 * unit's.
	 */
	core = -1;

	for_each_online_cpu(cpu) {
		this_core = topology_core_id(cpu);

		if (this_core == core)
			continue;

		core = this_core;

		/* get any CPU on this compute unit */
		cpumask_set_cpu(cpumask_any(topology_sibling_cpumask(cpu)), mask);
	}

	on_each_cpu_mask(mask, do_read_registers_on_cu, data, true);

	put_online_cpus();
	free_cpumask_var(mask);

	return 0;
}

static ssize_t power1_average_show(struct device *dev,
				   struct device_attribute *attr, char *buf)
{
	struct fam15h_power_data *data = dev_get_drvdata(dev);
	u64 prev_cu_acc_power[MAX_CUS], prev_ptsc[MAX_CUS],
	    jdelta[MAX_CUS];
	u64 tdelta, avg_acc;
	int cu, cu_num, ret;
	signed long leftover;

	/*
	 * With the new x86 topology modelling, x86_max_cores is the
	 * compute unit number.
	 */
	cu_num = boot_cpu_data.x86_max_cores;

	ret = read_registers(data);
	if (ret)
		return 0;

	for (cu = 0; cu < cu_num; cu++) {
		prev_cu_acc_power[cu] = data->cu_acc_power[cu];
		prev_ptsc[cu] = data->cpu_sw_pwr_ptsc[cu];
	}

	leftover = schedule_timeout_interruptible(msecs_to_jiffies(data->power_period));
	if (leftover)
		return 0;

	ret = read_registers(data);
	if (ret)
		return 0;

	for (cu = 0, avg_acc = 0; cu < cu_num; cu++) {
		/* check if current compute unit is online */
		if (data->cu_on[cu] == 0)
			continue;

		if (data->cu_acc_power[cu] < prev_cu_acc_power[cu]) {
			jdelta[cu] = data->max_cu_acc_power + data->cu_acc_power[cu];
			jdelta[cu] -= prev_cu_acc_power[cu];
		} else {
			jdelta[cu] = data->cu_acc_power[cu] - prev_cu_acc_power[cu];
		}
		tdelta = data->cpu_sw_pwr_ptsc[cu] - prev_ptsc[cu];
		jdelta[cu] *= data->cpu_pwr_sample_ratio * 1000;
		do_div(jdelta[cu], tdelta);

		/* the unit is microWatt */
		avg_acc += jdelta[cu];
	}

	return sprintf(buf, "%llu\n", (unsigned long long)avg_acc);
}
static DEVICE_ATTR_RO(power1_average);

static ssize_t power1_average_interval_show(struct device *dev,
					    struct device_attribute *attr,
					    char *buf)
{
	struct fam15h_power_data *data = dev_get_drvdata(dev);

	return sprintf(buf, "%lu\n", data->power_period);
}

static ssize_t power1_average_interval_store(struct device *dev,
					     struct device_attribute *attr,
					     const char *buf, size_t count)
{
	struct fam15h_power_data *data = dev_get_drvdata(dev);
	unsigned long temp;
	int ret;

	ret = kstrtoul(buf, 10, &temp);
	if (ret)
		return ret;

	if (temp > MAX_INTERVAL)
		return -EINVAL;

	/* the interval value should be greater than 0 */
	if (temp <= 0)
		return -EINVAL;

	data->power_period = temp;

	return count;
}
static DEVICE_ATTR_RW(power1_average_interval);

static int fam15h_power_init_attrs(struct pci_dev *pdev,
				   struct fam15h_power_data *data)
{
	int n = FAM15H_MIN_NUM_ATTRS;
	struct attribute **fam15h_power_attrs;
	struct cpuinfo_x86 *c = &boot_cpu_data;

	if (c->x86 == 0x15 &&
	    (c->x86_model <= 0xf ||
	     (c->x86_model >= 0x60 && c->x86_model <= 0x7f)))
		n += 1;

	/* check if processor supports accumulated power */
	if (boot_cpu_has(X86_FEATURE_ACC_POWER))
		n += 2;

	fam15h_power_attrs = devm_kcalloc(&pdev->dev, n,
					  sizeof(*fam15h_power_attrs),
					  GFP_KERNEL);

	if (!fam15h_power_attrs)
		return -ENOMEM;

	n = 0;
	fam15h_power_attrs[n++] = &dev_attr_power1_crit.attr;
	if (c->x86 == 0x15 &&
	    (c->x86_model <= 0xf ||
	     (c->x86_model >= 0x60 && c->x86_model <= 0x7f)))
		fam15h_power_attrs[n++] = &dev_attr_power1_input.attr;

	if (boot_cpu_has(X86_FEATURE_ACC_POWER)) {
		fam15h_power_attrs[n++] = &dev_attr_power1_average.attr;
		fam15h_power_attrs[n++] = &dev_attr_power1_average_interval.attr;
	}

	data->group.attrs = fam15h_power_attrs;

	return 0;
}

static bool should_load_on_this_node(struct pci_dev *f4)
{
	u32 val;

	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 3),
				  REG_NORTHBRIDGE_CAP, &val);
	if ((val & BIT(29)) && ((val >> 30) & 3))
		return false;

	return true;
}

/*
 * Newer BKDG versions have an updated recommendation on how to properly
 * initialize the running average range (was: 0xE, now: 0x9). This avoids
 * counter saturations resulting in bogus power readings.
 * We correct this value ourselves to cope with older BIOSes.
 */
static const struct pci_device_id affected_device[] = {
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_NB_F4) },
	{ 0 }
};

static void tweak_runavg_range(struct pci_dev *pdev)
{
	u32 val;

	/*
	 * let this quirk apply only to the current version of the
	 * northbridge, since future versions may change the behavior
	 */
	if (!pci_match_id(affected_device, pdev))
		return;

	pci_bus_read_config_dword(pdev->bus,
		PCI_DEVFN(PCI_SLOT(pdev->devfn), 5),
		REG_TDP_RUNNING_AVERAGE, &val);
	if ((val & 0xf) != 0xe)
		return;

	val &= ~0xf;
	val |=  0x9;
	pci_bus_write_config_dword(pdev->bus,
		PCI_DEVFN(PCI_SLOT(pdev->devfn), 5),
		REG_TDP_RUNNING_AVERAGE, val);
}

#ifdef CONFIG_PM
static int fam15h_power_resume(struct pci_dev *pdev)
{
	tweak_runavg_range(pdev);
	return 0;
}
#else
#define fam15h_power_resume NULL
#endif

static int fam15h_power_init_data(struct pci_dev *f4,
				  struct fam15h_power_data *data)
{
	u32 val;
	u64 tmp;
	int ret;

	pci_read_config_dword(f4, REG_PROCESSOR_TDP, &val);
	data->base_tdp = val >> 16;
	tmp = val & 0xffff;

	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 5),
				  REG_TDP_LIMIT3, &val);

	data->tdp_to_watts = ((val & 0x3ff) << 6) | ((val >> 10) & 0x3f);
	tmp *= data->tdp_to_watts;

	/* result not allowed to be >= 256W */
	if ((tmp >> 16) >= 256)
		dev_warn(&f4->dev,
			 "Bogus value for ProcessorPwrWatts (processor_pwr_watts>=%u)\n",
			 (unsigned int) (tmp >> 16));

	/* convert to microWatt */
	data->processor_pwr_watts = (tmp * 15625) >> 10;

	ret = fam15h_power_init_attrs(f4, data);
	if (ret)
		return ret;


	/* CPUID Fn8000_0007:EDX[12] indicates to support accumulated power */
	if (!boot_cpu_has(X86_FEATURE_ACC_POWER))
		return 0;

	/*
	 * determine the ratio of the compute unit power accumulator
	 * sample period to the PTSC counter period by executing CPUID
	 * Fn8000_0007:ECX
	 */
	data->cpu_pwr_sample_ratio = cpuid_ecx(0x80000007);

	if (rdmsrl_safe(MSR_F15H_CU_MAX_PWR_ACCUMULATOR, &tmp)) {
		pr_err("Failed to read max compute unit power accumulator MSR\n");
		return -ENODEV;
	}

	data->max_cu_acc_power = tmp;

	/*
	 * Milliseconds are a reasonable interval for the measurement.
	 * But it shouldn't set too long here, because several seconds
	 * would cause the read function to hang. So set default
	 * interval as 10 ms.
	 */
	data->power_period = 10;

	return read_registers(data);
}

static int fam15h_power_probe(struct pci_dev *pdev,
			      const struct pci_device_id *id)
{
	struct fam15h_power_data *data;
	struct device *dev = &pdev->dev;
	struct device *hwmon_dev;
	int ret;

	/*
	 * though we ignore every other northbridge, we still have to
	 * do the tweaking on _each_ node in MCM processors as the counters
	 * are working hand-in-hand
	 */
	tweak_runavg_range(pdev);

	if (!should_load_on_this_node(pdev))
		return -ENODEV;

	data = devm_kzalloc(dev, sizeof(struct fam15h_power_data), GFP_KERNEL);
	if (!data)
		return -ENOMEM;

	ret = fam15h_power_init_data(pdev, data);
	if (ret)
		return ret;

	data->pdev = pdev;

	data->groups[0] = &data->group;

	hwmon_dev = devm_hwmon_device_register_with_groups(dev, "fam15h_power",
							   data,
							   &data->groups[0]);
	return PTR_ERR_OR_ZERO(hwmon_dev);
}

static const struct pci_device_id fam15h_power_id_table[] = {
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_NB_F4) },
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M30H_NB_F4) },
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M60H_NB_F4) },
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M70H_NB_F4) },
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_16H_NB_F4) },
	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_16H_M30H_NB_F4) },
	{}
};
MODULE_DEVICE_TABLE(pci, fam15h_power_id_table);

static struct pci_driver fam15h_power_driver = {
	.name = "fam15h_power",
	.id_table = fam15h_power_id_table,
	.probe = fam15h_power_probe,
	.resume = fam15h_power_resume,
};

module_pci_driver(fam15h_power_driver);
Commit	Line	Data
6e7c1094	1	// SPDX-License-Identifier: GPL-2.0-or-later
512d1027 AH	2	/*
	3	* fam15h_power.c - AMD Family 15h processor power monitoring
	4	*
a6e232f7	5	* Copyright (c) 2011-2016 Advanced Micro Devices, Inc.
d034fbf0	6	* Author: Andreas Herrmann <[email protected]>
512d1027 AH	7	*/
	8
	9	#include <linux/err.h>
	10	#include <linux/hwmon.h>
	11	#include <linux/hwmon-sysfs.h>
	12	#include <linux/init.h>
	13	#include <linux/module.h>
	14	#include <linux/pci.h>
	15	#include <linux/bitops.h>
fa794344 HR	16	#include <linux/cpu.h>
fa794344 HR	17	#include <linux/cpumask.h>
11bf0d78 HR	18	#include <linux/time.h>
11bf0d78 HR	19	#include <linux/sched.h>
512d1027	20	#include <asm/processor.h>
3b5ea47d	21	#include <asm/msr.h>
512d1027 AH	22
512d1027 AH	23	MODULE_DESCRIPTION("AMD Family 15h CPU processor power monitor");
d034fbf0	24	MODULE_AUTHOR("Andreas Herrmann <[email protected]>");
512d1027 AH	25	MODULE_LICENSE("GPL");
	26
	27	/* D18F3 */
	28	#define REG_NORTHBRIDGE_CAP 0xe8
	29
	30	/* D18F4 */
	31	#define REG_PROCESSOR_TDP 0x1b8
	32
	33	/* D18F5 */
	34	#define REG_TDP_RUNNING_AVERAGE 0xe0
	35	#define REG_TDP_LIMIT3 0xe8
	36
7deb14b1 HR	37	#define FAM15H_MIN_NUM_ATTRS 2
7deb14b1 HR	38	#define FAM15H_NUM_GROUPS 2
fa794344	39	#define MAX_CUS 8
7deb14b1	40
11bf0d78 HR	41	/* set maximum interval as 1 second */
	42	#define MAX_INTERVAL 1000
	43
eff2a945 HR	44	#define PCI_DEVICE_ID_AMD_15H_M70H_NB_F4 0x15b4
eff2a945 HR	45
512d1027	46	struct fam15h_power_data {
562dc973	47	struct pci_dev *pdev;
512d1027 AH	48	unsigned int tdp_to_watts;
	49	unsigned int base_tdp;
	50	unsigned int processor_pwr_watts;
1ed32160	51	unsigned int cpu_pwr_sample_ratio;
7deb14b1 HR	52	const struct attribute_group *groups[FAM15H_NUM_GROUPS];
7deb14b1 HR	53	struct attribute_group group;
3b5ea47d HR	54	/* maximum accumulated power of a compute unit */
3b5ea47d HR	55	u64 max_cu_acc_power;
fa794344 HR	56	/* accumulated power of the compute units */
fa794344 HR	57	u64 cu_acc_power[MAX_CUS];
cdb9e110 HR	58	/* performance timestamp counter */
cdb9e110 HR	59	u64 cpu_sw_pwr_ptsc[MAX_CUS];
11bf0d78 HR	60	/* online/offline status of current compute unit */
	61	int cu_on[MAX_CUS];
	62	unsigned long power_period;
512d1027 AH	63	};
512d1027 AH	64
1d28e016 HR	65	static bool is_carrizo_or_later(void)
	66	{
	67	return boot_cpu_data.x86 == 0x15 && boot_cpu_data.x86_model >= 0x60;
	68	}
	69
d013f7f5 JL	70	static ssize_t power1_input_show(struct device *dev,
d013f7f5 JL	71	struct device_attribute attr, char buf)
512d1027 AH	72	{
	73	u32 val, tdp_limit, running_avg_range;
	74	s32 running_avg_capture;
	75	u64 curr_pwr_watts;
512d1027	76	struct fam15h_power_data *data = dev_get_drvdata(dev);
562dc973	77	struct pci_dev *f4 = data->pdev;
512d1027 AH	78
	79	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 5),
	80	REG_TDP_RUNNING_AVERAGE, &val);
e9cd4d55 HR	81
	82	/*
	83	* On Carrizo and later platforms, TdpRunAvgAccCap bit field
	84	* is extended to 4:31 from 4:25.
	85	*/
1d28e016	86	if (is_carrizo_or_later()) {
e9cd4d55 HR	87	running_avg_capture = val >> 4;
	88	running_avg_capture = sign_extend32(running_avg_capture, 27);
	89	} else {
	90	running_avg_capture = (val >> 4) & 0x3fffff;
	91	running_avg_capture = sign_extend32(running_avg_capture, 21);
	92	}
	93
941a956b	94	running_avg_range = (val & 0xf) + 1;
512d1027 AH	95
	96	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 5),
	97	REG_TDP_LIMIT3, &val);
	98
60dee3ca GK	99	/*
	100	* On Carrizo and later platforms, ApmTdpLimit bit field
	101	* is extended to 16:31 from 16:28.
	102	*/
1d28e016	103	if (is_carrizo_or_later())
60dee3ca GK	104	tdp_limit = val >> 16;
	105	else
	106	tdp_limit = (val >> 16) & 0x1fff;
	107
62867d49 GR	108	curr_pwr_watts = ((u64)(tdp_limit +
62867d49 GR	109	data->base_tdp)) << running_avg_range;
941a956b	110	curr_pwr_watts -= running_avg_capture;
512d1027 AH	111	curr_pwr_watts *= data->tdp_to_watts;
	112
	113	/*
	114	* Convert to microWatt
	115	*
	116	* power is in Watt provided as fixed point integer with
	117	* scaling factor 1/(2^16). For conversion we use
	118	* (10^6)/(2^16) = 15625/(2^10)
	119	*/
941a956b	120	curr_pwr_watts = (curr_pwr_watts * 15625) >> (10 + running_avg_range);
512d1027 AH	121	return sprintf(buf, "%u\n", (unsigned int) curr_pwr_watts);
512d1027 AH	122	}
d013f7f5	123	static DEVICE_ATTR_RO(power1_input);
512d1027	124
d013f7f5 JL	125	static ssize_t power1_crit_show(struct device *dev,
d013f7f5 JL	126	struct device_attribute attr, char buf)
512d1027 AH	127	{
	128	struct fam15h_power_data *data = dev_get_drvdata(dev);
	129
	130	return sprintf(buf, "%u\n", data->processor_pwr_watts);
	131	}
d013f7f5	132	static DEVICE_ATTR_RO(power1_crit);
512d1027	133
fa794344 HR	134	static void do_read_registers_on_cu(void *_data)
	135	{
	136	struct fam15h_power_data *data = _data;
	137	int cpu, cu;
	138
	139	cpu = smp_processor_id();
	140
	141	/*
	142	* With the new x86 topology modelling, cpu core id actually
	143	* is compute unit id.
	144	*/
	145	cu = cpu_data(cpu).cpu_core_id;
	146
	147	rdmsrl_safe(MSR_F15H_CU_PWR_ACCUMULATOR, &data->cu_acc_power[cu]);
cdb9e110	148	rdmsrl_safe(MSR_F15H_PTSC, &data->cpu_sw_pwr_ptsc[cu]);
11bf0d78 HR	149
11bf0d78 HR	150	data->cu_on[cu] = 1;
fa794344 HR	151	}
	152
	153	/*
	154	* This function is only able to be called when CPUID
	155	* Fn8000_0007:EDX[12] is set.
	156	*/
	157	static int read_registers(struct fam15h_power_data *data)
	158	{
fa794344 HR	159	int core, this_core;
fa794344 HR	160	cpumask_var_t mask;
7be48818	161	int ret, cpu;
fa794344 HR	162
	163	ret = zalloc_cpumask_var(&mask, GFP_KERNEL);
	164	if (!ret)
	165	return -ENOMEM;
	166
11bf0d78 HR	167	memset(data->cu_on, 0, sizeof(int) * MAX_CUS);
11bf0d78 HR	168
fa794344	169	get_online_cpus();
fa794344 HR	170
	171	/*
	172	* Choose the first online core of each compute unit, and then
	173	* read their MSR value of power and ptsc in a single IPI,
	174	* because the MSR value of CPU core represent the compute
	175	* unit's.
	176	*/
	177	core = -1;
	178
	179	for_each_online_cpu(cpu) {
	180	this_core = topology_core_id(cpu);
	181
	182	if (this_core == core)
	183	continue;
	184
	185	core = this_core;
	186
	187	/* get any CPU on this compute unit */
	188	cpumask_set_cpu(cpumask_any(topology_sibling_cpumask(cpu)), mask);
	189	}
	190
7be48818	191	on_each_cpu_mask(mask, do_read_registers_on_cu, data, true);
fa794344	192
fa794344	193	put_online_cpus();
fa794344 HR	194	free_cpumask_var(mask);
	195
	196	return 0;
	197	}
	198
d013f7f5 JL	199	static ssize_t power1_average_show(struct device *dev,
d013f7f5 JL	200	struct device_attribute attr, char buf)
11bf0d78 HR	201	{
	202	struct fam15h_power_data *data = dev_get_drvdata(dev);
	203	u64 prev_cu_acc_power[MAX_CUS], prev_ptsc[MAX_CUS],
	204	jdelta[MAX_CUS];
	205	u64 tdelta, avg_acc;
	206	int cu, cu_num, ret;
	207	signed long leftover;
	208
	209	/*
	210	* With the new x86 topology modelling, x86_max_cores is the
	211	* compute unit number.
	212	*/
	213	cu_num = boot_cpu_data.x86_max_cores;
	214
	215	ret = read_registers(data);
	216	if (ret)
	217	return 0;
	218
	219	for (cu = 0; cu < cu_num; cu++) {
	220	prev_cu_acc_power[cu] = data->cu_acc_power[cu];
	221	prev_ptsc[cu] = data->cpu_sw_pwr_ptsc[cu];
	222	}
	223
	224	leftover = schedule_timeout_interruptible(msecs_to_jiffies(data->power_period));
	225	if (leftover)
	226	return 0;
	227
	228	ret = read_registers(data);
	229	if (ret)
	230	return 0;
	231
	232	for (cu = 0, avg_acc = 0; cu < cu_num; cu++) {
	233	/* check if current compute unit is online */
	234	if (data->cu_on[cu] == 0)
	235	continue;
	236
	237	if (data->cu_acc_power[cu] < prev_cu_acc_power[cu]) {
	238	jdelta[cu] = data->max_cu_acc_power + data->cu_acc_power[cu];
	239	jdelta[cu] -= prev_cu_acc_power[cu];
	240	} else {
	241	jdelta[cu] = data->cu_acc_power[cu] - prev_cu_acc_power[cu];
	242	}
	243	tdelta = data->cpu_sw_pwr_ptsc[cu] - prev_ptsc[cu];
	244	jdelta[cu] = data->cpu_pwr_sample_ratio 1000;
	245	do_div(jdelta[cu], tdelta);
	246
	247	/* the unit is microWatt */
	248	avg_acc += jdelta[cu];
	249	}
	250
	251	return sprintf(buf, "%llu\n", (unsigned long long)avg_acc);
	252	}
d013f7f5	253	static DEVICE_ATTR_RO(power1_average);
11bf0d78	254
d013f7f5 JL	255	static ssize_t power1_average_interval_show(struct device *dev,
	256	struct device_attribute *attr,
	257	char *buf)
11bf0d78 HR	258	{
	259	struct fam15h_power_data *data = dev_get_drvdata(dev);
	260
	261	return sprintf(buf, "%lu\n", data->power_period);
	262	}
	263
d013f7f5 JL	264	static ssize_t power1_average_interval_store(struct device *dev,
	265	struct device_attribute *attr,
	266	const char *buf, size_t count)
11bf0d78 HR	267	{
	268	struct fam15h_power_data *data = dev_get_drvdata(dev);
	269	unsigned long temp;
	270	int ret;
	271
	272	ret = kstrtoul(buf, 10, &temp);
	273	if (ret)
	274	return ret;
	275
	276	if (temp > MAX_INTERVAL)
	277	return -EINVAL;
	278
	279	/* the interval value should be greater than 0 */
	280	if (temp <= 0)
	281	return -EINVAL;
	282
	283	data->power_period = temp;
	284
	285	return count;
	286	}
d013f7f5	287	static DEVICE_ATTR_RW(power1_average_interval);
11bf0d78	288
7deb14b1 HR	289	static int fam15h_power_init_attrs(struct pci_dev *pdev,
7deb14b1 HR	290	struct fam15h_power_data *data)
961a2378	291	{
7deb14b1 HR	292	int n = FAM15H_MIN_NUM_ATTRS;
7deb14b1 HR	293	struct attribute **fam15h_power_attrs;
46f29c2b	294	struct cpuinfo_x86 *c = &boot_cpu_data;
961a2378	295
46f29c2b HR	296	if (c->x86 == 0x15 &&
46f29c2b HR	297	(c->x86_model <= 0xf \|\|
eff2a945	298	(c->x86_model >= 0x60 && c->x86_model <= 0x7f)))
7deb14b1	299	n += 1;
961a2378	300
11bf0d78 HR	301	/* check if processor supports accumulated power */
	302	if (boot_cpu_has(X86_FEATURE_ACC_POWER))
	303	n += 2;
	304
7deb14b1 HR	305	fam15h_power_attrs = devm_kcalloc(&pdev->dev, n,
	306	sizeof(*fam15h_power_attrs),
	307	GFP_KERNEL);
512d1027	308
7deb14b1 HR	309	if (!fam15h_power_attrs)
	310	return -ENOMEM;
	311
	312	n = 0;
	313	fam15h_power_attrs[n++] = &dev_attr_power1_crit.attr;
46f29c2b HR	314	if (c->x86 == 0x15 &&
46f29c2b HR	315	(c->x86_model <= 0xf \|\|
eff2a945	316	(c->x86_model >= 0x60 && c->x86_model <= 0x7f)))
7deb14b1 HR	317	fam15h_power_attrs[n++] = &dev_attr_power1_input.attr;
7deb14b1 HR	318
11bf0d78 HR	319	if (boot_cpu_has(X86_FEATURE_ACC_POWER)) {
	320	fam15h_power_attrs[n++] = &dev_attr_power1_average.attr;
	321	fam15h_power_attrs[n++] = &dev_attr_power1_average_interval.attr;
	322	}
	323
7deb14b1 HR	324	data->group.attrs = fam15h_power_attrs;
	325
	326	return 0;
	327	}
512d1027	328
d83e92b3	329	static bool should_load_on_this_node(struct pci_dev *f4)
512d1027 AH	330	{
	331	u32 val;
	332
	333	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 3),
	334	REG_NORTHBRIDGE_CAP, &val);
	335	if ((val & BIT(29)) && ((val >> 30) & 3))
	336	return false;
	337
	338	return true;
	339	}
	340
00250ec9 AP	341	/*
	342	* Newer BKDG versions have an updated recommendation on how to properly
	343	* initialize the running average range (was: 0xE, now: 0x9). This avoids
	344	* counter saturations resulting in bogus power readings.
	345	* We correct this value ourselves to cope with older BIOSes.
	346	*/
5f0ecb90	347	static const struct pci_device_id affected_device[] = {
c3e40a99 GR	348	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_NB_F4) },
	349	{ 0 }
	350	};
	351
5f0ecb90	352	static void tweak_runavg_range(struct pci_dev *pdev)
00250ec9 AP	353	{
00250ec9 AP	354	u32 val;
00250ec9 AP	355
	356	/*
	357	* let this quirk apply only to the current version of the
	358	* northbridge, since future versions may change the behavior
	359	*/
c3e40a99	360	if (!pci_match_id(affected_device, pdev))
00250ec9 AP	361	return;
	362
	363	pci_bus_read_config_dword(pdev->bus,
	364	PCI_DEVFN(PCI_SLOT(pdev->devfn), 5),
	365	REG_TDP_RUNNING_AVERAGE, &val);
	366	if ((val & 0xf) != 0xe)
	367	return;
	368
	369	val &= ~0xf;
	370	val \|= 0x9;
	371	pci_bus_write_config_dword(pdev->bus,
	372	PCI_DEVFN(PCI_SLOT(pdev->devfn), 5),
	373	REG_TDP_RUNNING_AVERAGE, val);
	374	}
	375
5f0ecb90 AH	376	#ifdef CONFIG_PM
	377	static int fam15h_power_resume(struct pci_dev *pdev)
	378	{
	379	tweak_runavg_range(pdev);
	380	return 0;
	381	}
	382	#else
	383	#define fam15h_power_resume NULL
	384	#endif
	385
7deb14b1 HR	386	static int fam15h_power_init_data(struct pci_dev *f4,
7deb14b1 HR	387	struct fam15h_power_data *data)
512d1027	388	{
11bf0d78	389	u32 val;
512d1027	390	u64 tmp;
7deb14b1	391	int ret;
512d1027 AH	392
	393	pci_read_config_dword(f4, REG_PROCESSOR_TDP, &val);
	394	data->base_tdp = val >> 16;
	395	tmp = val & 0xffff;
	396
	397	pci_bus_read_config_dword(f4->bus, PCI_DEVFN(PCI_SLOT(f4->devfn), 5),
	398	REG_TDP_LIMIT3, &val);
	399
	400	data->tdp_to_watts = ((val & 0x3ff) << 6) \| ((val >> 10) & 0x3f);
	401	tmp *= data->tdp_to_watts;
	402
	403	/* result not allowed to be >= 256W */
	404	if ((tmp >> 16) >= 256)
b55f3757 GR	405	dev_warn(&f4->dev,
b55f3757 GR	406	"Bogus value for ProcessorPwrWatts (processor_pwr_watts>=%u)\n",
512d1027 AH	407	(unsigned int) (tmp >> 16));
	408
	409	/* convert to microWatt */
	410	data->processor_pwr_watts = (tmp * 15625) >> 10;
1ed32160	411
7deb14b1 HR	412	ret = fam15h_power_init_attrs(f4, data);
	413	if (ret)
	414	return ret;
	415
1ed32160 HR	416
1ed32160 HR	417	/* CPUID Fn8000_0007:EDX[12] indicates to support accumulated power */
11bf0d78	418	if (!boot_cpu_has(X86_FEATURE_ACC_POWER))
7deb14b1	419	return 0;
1ed32160 HR	420
	421	/*
	422	* determine the ratio of the compute unit power accumulator
	423	* sample period to the PTSC counter period by executing CPUID
	424	* Fn8000_0007:ECX
	425	*/
11bf0d78	426	data->cpu_pwr_sample_ratio = cpuid_ecx(0x80000007);
7deb14b1	427
3b5ea47d HR	428	if (rdmsrl_safe(MSR_F15H_CU_MAX_PWR_ACCUMULATOR, &tmp)) {
	429	pr_err("Failed to read max compute unit power accumulator MSR\n");
	430	return -ENODEV;
	431	}
	432
	433	data->max_cu_acc_power = tmp;
	434
11bf0d78 HR	435	/*
	436	* Milliseconds are a reasonable interval for the measurement.
	437	* But it shouldn't set too long here, because several seconds
	438	* would cause the read function to hang. So set default
	439	* interval as 10 ms.
	440	*/
	441	data->power_period = 10;
	442
fa794344	443	return read_registers(data);
512d1027 AH	444	}
512d1027 AH	445
6c931ae1	446	static int fam15h_power_probe(struct pci_dev *pdev,
7deb14b1	447	const struct pci_device_id *id)
512d1027 AH	448	{
512d1027 AH	449	struct fam15h_power_data *data;
87432a2e	450	struct device *dev = &pdev->dev;
562dc973	451	struct device *hwmon_dev;
7deb14b1	452	int ret;
512d1027	453
00250ec9 AP	454	/*
	455	* though we ignore every other northbridge, we still have to
	456	* do the tweaking on _each_ node in MCM processors as the counters
	457	* are working hand-in-hand
	458	*/
	459	tweak_runavg_range(pdev);
	460
d83e92b3	461	if (!should_load_on_this_node(pdev))
87432a2e GR	462	return -ENODEV;
	463
	464	data = devm_kzalloc(dev, sizeof(struct fam15h_power_data), GFP_KERNEL);
	465	if (!data)
	466	return -ENOMEM;
512d1027	467
7deb14b1 HR	468	ret = fam15h_power_init_data(pdev, data);
	469	if (ret)
	470	return ret;
	471
562dc973	472	data->pdev = pdev;
512d1027	473
7deb14b1 HR	474	data->groups[0] = &data->group;
7deb14b1 HR	475
562dc973 AL	476	hwmon_dev = devm_hwmon_device_register_with_groups(dev, "fam15h_power",
562dc973 AL	477	data,
7deb14b1	478	&data->groups[0]);
562dc973	479	return PTR_ERR_OR_ZERO(hwmon_dev);
512d1027 AH	480	}
512d1027 AH	481
cd9bb056	482	static const struct pci_device_id fam15h_power_id_table[] = {
512d1027	483	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_NB_F4) },
0a0039ad	484	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M30H_NB_F4) },
5dc08725	485	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M60H_NB_F4) },
eff2a945	486	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M70H_NB_F4) },
22e32f4f	487	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_16H_NB_F4) },
0bd52941	488	{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_16H_M30H_NB_F4) },
512d1027 AH	489	{}
	490	};
	491	MODULE_DEVICE_TABLE(pci, fam15h_power_id_table);
	492
	493	static struct pci_driver fam15h_power_driver = {
	494	.name = "fam15h_power",
	495	.id_table = fam15h_power_id_table,
	496	.probe = fam15h_power_probe,
5f0ecb90	497	.resume = fam15h_power_resume,
512d1027 AH	498	};
512d1027 AH	499
f71f5a55	500	module_pci_driver(fam15h_power_driver);