2 * coretemp.c - Linux kernel module for hardware monitoring
4 * Copyright (C) 2007 Rudolf Marek <r.marek@assembler.cz>
6 * Inspired from many hwmon drivers
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; version 2 of the License.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
23 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
25 #include <linux/module.h>
26 #include <linux/init.h>
27 #include <linux/slab.h>
28 #include <linux/jiffies.h>
29 #include <linux/hwmon.h>
30 #include <linux/sysfs.h>
31 #include <linux/hwmon-sysfs.h>
32 #include <linux/err.h>
33 #include <linux/mutex.h>
34 #include <linux/list.h>
35 #include <linux/platform_device.h>
36 #include <linux/cpu.h>
37 #include <linux/pci.h>
38 #include <linux/smp.h>
39 #include <linux/moduleparam.h>
41 #include <asm/processor.h>
43 #define DRVNAME "coretemp"
46 * force_tjmax only matters when TjMax can't be read from the CPU itself.
47 * When set, it replaces the driver's suboptimal heuristic.
49 static int force_tjmax;
50 module_param_named(tjmax, force_tjmax, int, 0444);
51 MODULE_PARM_DESC(tjmax, "TjMax value in degrees Celsius");
53 #define BASE_SYSFS_ATTR_NO 2 /* Sysfs Base attr no for coretemp */
54 #define NUM_REAL_CORES 32 /* Number of Real cores per cpu */
55 #define CORETEMP_NAME_LENGTH 17 /* String Length of attrs */
56 #define MAX_CORE_ATTRS 4 /* Maximum no of basic attrs */
57 #define TOTAL_ATTRS (MAX_CORE_ATTRS + 1)
58 #define MAX_CORE_DATA (NUM_REAL_CORES + BASE_SYSFS_ATTR_NO)
60 #define TO_PHYS_ID(cpu) cpu_data(cpu).phys_proc_id
61 #define TO_CORE_ID(cpu) cpu_data(cpu).cpu_core_id
62 #define TO_ATTR_NO(cpu) (TO_CORE_ID(cpu) + BASE_SYSFS_ATTR_NO)
65 #define for_each_sibling(i, cpu) for_each_cpu(i, cpu_sibling_mask(cpu))
67 #define for_each_sibling(i, cpu) for (i = 0; false; )
71 * Per-Core Temperature Data
72 * @last_updated: The time when the current temperature value was updated
73 * earlier (in jiffies).
74 * @cpu_core_id: The CPU Core from which temperature values should be read
75 * This value is passed as "id" field to rdmsr/wrmsr functions.
76 * @status_reg: One of IA32_THERM_STATUS or IA32_PACKAGE_THERM_STATUS,
77 * from where the temperature values should be read.
78 * @attr_size: Total number of pre-core attrs displayed in the sysfs.
79 * @is_pkg_data: If this is 1, the temp_data holds pkgtemp data.
80 * Otherwise, temp_data holds coretemp data.
81 * @valid: If this is 1, the current temperature is valid.
87 unsigned long last_updated;
94 struct sensor_device_attribute sd_attrs[TOTAL_ATTRS];
95 char attr_name[TOTAL_ATTRS][CORETEMP_NAME_LENGTH];
96 struct mutex update_lock;
99 /* Platform Data per Physical CPU */
100 struct platform_data {
101 struct device *hwmon_dev;
103 struct temp_data *core_data[MAX_CORE_DATA];
104 struct device_attribute name_attr;
108 struct list_head list;
109 struct platform_device *pdev;
113 static LIST_HEAD(pdev_list);
114 static DEFINE_MUTEX(pdev_list_mutex);
116 static ssize_t show_name(struct device *dev,
117 struct device_attribute *devattr, char *buf)
119 return sprintf(buf, "%s\n", DRVNAME);
122 static ssize_t show_label(struct device *dev,
123 struct device_attribute *devattr, char *buf)
125 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
126 struct platform_data *pdata = dev_get_drvdata(dev);
127 struct temp_data *tdata = pdata->core_data[attr->index];
129 if (tdata->is_pkg_data)
130 return sprintf(buf, "Physical id %u\n", pdata->phys_proc_id);
132 return sprintf(buf, "Core %u\n", tdata->cpu_core_id);
135 static ssize_t show_crit_alarm(struct device *dev,
136 struct device_attribute *devattr, char *buf)
139 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
140 struct platform_data *pdata = dev_get_drvdata(dev);
141 struct temp_data *tdata = pdata->core_data[attr->index];
143 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
145 return sprintf(buf, "%d\n", (eax >> 5) & 1);
148 static ssize_t show_tjmax(struct device *dev,
149 struct device_attribute *devattr, char *buf)
151 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
152 struct platform_data *pdata = dev_get_drvdata(dev);
154 return sprintf(buf, "%d\n", pdata->core_data[attr->index]->tjmax);
157 static ssize_t show_ttarget(struct device *dev,
158 struct device_attribute *devattr, char *buf)
160 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
161 struct platform_data *pdata = dev_get_drvdata(dev);
163 return sprintf(buf, "%d\n", pdata->core_data[attr->index]->ttarget);
166 static ssize_t show_temp(struct device *dev,
167 struct device_attribute *devattr, char *buf)
170 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
171 struct platform_data *pdata = dev_get_drvdata(dev);
172 struct temp_data *tdata = pdata->core_data[attr->index];
174 mutex_lock(&tdata->update_lock);
176 /* Check whether the time interval has elapsed */
177 if (!tdata->valid || time_after(jiffies, tdata->last_updated + HZ)) {
178 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
180 /* Check whether the data is valid */
181 if (eax & 0x80000000) {
182 tdata->temp = tdata->tjmax -
183 ((eax >> 16) & 0x7f) * 1000;
186 tdata->last_updated = jiffies;
189 mutex_unlock(&tdata->update_lock);
190 return tdata->valid ? sprintf(buf, "%d\n", tdata->temp) : -EAGAIN;
193 static int adjust_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
195 /* The 100C is default for both mobile and non mobile CPUs */
198 int tjmax_ee = 85000;
202 struct pci_dev *host_bridge;
204 /* Early chips have no MSR for TjMax */
206 if (c->x86_model == 0xf && c->x86_mask < 4)
211 if (c->x86_model == 0x1c || c->x86_model == 0x26
212 || c->x86_model == 0x27) {
215 host_bridge = pci_get_bus_and_slot(0, PCI_DEVFN(0, 0));
217 if (host_bridge && host_bridge->vendor == PCI_VENDOR_ID_INTEL
218 && (host_bridge->device == 0xa000 /* NM10 based nettop */
219 || host_bridge->device == 0xa010)) /* NM10 based netbook */
224 pci_dev_put(host_bridge);
225 } else if (c->x86_model == 0x36) {
230 if (c->x86_model > 0xe && usemsr_ee) {
234 * Now we can detect the mobile CPU using Intel provided table
235 * http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
236 * For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
238 err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
241 "Unable to access MSR 0x17, assuming desktop"
244 } else if (c->x86_model < 0x17 && !(eax & 0x10000000)) {
246 * Trust bit 28 up to Penryn, I could not find any
247 * documentation on that; if you happen to know
248 * someone at Intel please ask
252 /* Platform ID bits 52:50 (EDX starts at bit 32) */
253 platform_id = (edx >> 18) & 0x7;
256 * Mobile Penryn CPU seems to be platform ID 7 or 5
259 if (c->x86_model == 0x17 &&
260 (platform_id == 5 || platform_id == 7)) {
262 * If MSR EE bit is set, set it to 90 degrees C,
263 * otherwise 105 degrees C
272 err = rdmsr_safe_on_cpu(id, 0xee, &eax, &edx);
275 "Unable to access MSR 0xEE, for Tjmax, left"
277 } else if (eax & 0x40000000) {
280 } else if (tjmax == 100000) {
282 * If we don't use msr EE it means we are desktop CPU
283 * (with exeception of Atom)
285 dev_warn(dev, "Using relative temperature scale!\n");
291 static int get_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
298 * A new feature of current Intel(R) processors, the
299 * IA32_TEMPERATURE_TARGET contains the TjMax value
301 err = rdmsr_safe_on_cpu(id, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
303 if (c->x86_model > 0xe && c->x86_model != 0x1c)
304 dev_warn(dev, "Unable to read TjMax from CPU %u\n", id);
306 val = (eax >> 16) & 0xff;
308 * If the TjMax is not plausible, an assumption
312 dev_dbg(dev, "TjMax is %d degrees C\n", val);
318 dev_notice(dev, "TjMax forced to %d degrees C by user\n",
320 return force_tjmax * 1000;
324 * An assumption is made for early CPUs and unreadable MSR.
325 * NOTE: the calculated value may not be correct.
327 return adjust_tjmax(c, id, dev);
330 static int create_name_attr(struct platform_data *pdata, struct device *dev)
332 sysfs_attr_init(&pdata->name_attr.attr);
333 pdata->name_attr.attr.name = "name";
334 pdata->name_attr.attr.mode = S_IRUGO;
335 pdata->name_attr.show = show_name;
336 return device_create_file(dev, &pdata->name_attr);
339 static int create_core_attrs(struct temp_data *tdata, struct device *dev,
343 static ssize_t (*const rd_ptr[TOTAL_ATTRS]) (struct device *dev,
344 struct device_attribute *devattr, char *buf) = {
345 show_label, show_crit_alarm, show_temp, show_tjmax,
347 static const char *const names[TOTAL_ATTRS] = {
348 "temp%d_label", "temp%d_crit_alarm",
349 "temp%d_input", "temp%d_crit",
352 for (i = 0; i < tdata->attr_size; i++) {
353 snprintf(tdata->attr_name[i], CORETEMP_NAME_LENGTH, names[i],
355 sysfs_attr_init(&tdata->sd_attrs[i].dev_attr.attr);
356 tdata->sd_attrs[i].dev_attr.attr.name = tdata->attr_name[i];
357 tdata->sd_attrs[i].dev_attr.attr.mode = S_IRUGO;
358 tdata->sd_attrs[i].dev_attr.show = rd_ptr[i];
359 tdata->sd_attrs[i].index = attr_no;
360 err = device_create_file(dev, &tdata->sd_attrs[i].dev_attr);
368 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
373 static int __cpuinit chk_ucode_version(unsigned int cpu)
375 struct cpuinfo_x86 *c = &cpu_data(cpu);
378 * Check if we have problem with errata AE18 of Core processors:
379 * Readings might stop update when processor visited too deep sleep,
380 * fixed for stepping D0 (6EC).
382 if (c->x86_model == 0xe && c->x86_mask < 0xc && c->microcode < 0x39) {
383 pr_err("Errata AE18 not fixed, update BIOS or "
384 "microcode of the CPU!\n");
390 static struct platform_device *coretemp_get_pdev(unsigned int cpu)
392 u16 phys_proc_id = TO_PHYS_ID(cpu);
393 struct pdev_entry *p;
395 mutex_lock(&pdev_list_mutex);
397 list_for_each_entry(p, &pdev_list, list)
398 if (p->phys_proc_id == phys_proc_id) {
399 mutex_unlock(&pdev_list_mutex);
403 mutex_unlock(&pdev_list_mutex);
407 static struct temp_data *init_temp_data(unsigned int cpu, int pkg_flag)
409 struct temp_data *tdata;
411 tdata = kzalloc(sizeof(struct temp_data), GFP_KERNEL);
415 tdata->status_reg = pkg_flag ? MSR_IA32_PACKAGE_THERM_STATUS :
416 MSR_IA32_THERM_STATUS;
417 tdata->is_pkg_data = pkg_flag;
419 tdata->cpu_core_id = TO_CORE_ID(cpu);
420 tdata->attr_size = MAX_CORE_ATTRS;
421 mutex_init(&tdata->update_lock);
425 static int create_core_data(struct platform_device *pdev,
426 unsigned int cpu, int pkg_flag)
428 struct temp_data *tdata;
429 struct platform_data *pdata = platform_get_drvdata(pdev);
430 struct cpuinfo_x86 *c = &cpu_data(cpu);
435 * Find attr number for sysfs:
436 * We map the attr number to core id of the CPU
437 * The attr number is always core id + 2
438 * The Pkgtemp will always show up as temp1_*, if available
440 attr_no = pkg_flag ? 1 : TO_ATTR_NO(cpu);
442 if (attr_no > MAX_CORE_DATA - 1)
446 * Provide a single set of attributes for all HT siblings of a core
447 * to avoid duplicate sensors (the processor ID and core ID of all
448 * HT siblings of a core are the same).
449 * Skip if a HT sibling of this core is already registered.
450 * This is not an error.
452 if (pdata->core_data[attr_no] != NULL)
455 tdata = init_temp_data(cpu, pkg_flag);
459 /* Test if we can access the status register */
460 err = rdmsr_safe_on_cpu(cpu, tdata->status_reg, &eax, &edx);
464 /* We can access status register. Get Critical Temperature */
465 tdata->tjmax = get_tjmax(c, cpu, &pdev->dev);
468 * Read the still undocumented bits 8:15 of IA32_TEMPERATURE_TARGET.
469 * The target temperature is available on older CPUs but not in this
470 * register. Atoms don't have the register at all.
472 if (c->x86_model > 0xe && c->x86_model != 0x1c) {
473 err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET,
477 = tdata->tjmax - ((eax >> 8) & 0xff) * 1000;
482 pdata->core_data[attr_no] = tdata;
484 /* Create sysfs interfaces */
485 err = create_core_attrs(tdata, &pdev->dev, attr_no);
491 pdata->core_data[attr_no] = NULL;
496 static void coretemp_add_core(unsigned int cpu, int pkg_flag)
498 struct platform_device *pdev = coretemp_get_pdev(cpu);
504 err = create_core_data(pdev, cpu, pkg_flag);
506 dev_err(&pdev->dev, "Adding Core %u failed\n", cpu);
509 static void coretemp_remove_core(struct platform_data *pdata,
510 struct device *dev, int indx)
513 struct temp_data *tdata = pdata->core_data[indx];
515 /* Remove the sysfs attributes */
516 for (i = 0; i < tdata->attr_size; i++)
517 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
519 kfree(pdata->core_data[indx]);
520 pdata->core_data[indx] = NULL;
523 static int __devinit coretemp_probe(struct platform_device *pdev)
525 struct platform_data *pdata;
528 /* Initialize the per-package data structures */
529 pdata = kzalloc(sizeof(struct platform_data), GFP_KERNEL);
533 err = create_name_attr(pdata, &pdev->dev);
537 pdata->phys_proc_id = pdev->id;
538 platform_set_drvdata(pdev, pdata);
540 pdata->hwmon_dev = hwmon_device_register(&pdev->dev);
541 if (IS_ERR(pdata->hwmon_dev)) {
542 err = PTR_ERR(pdata->hwmon_dev);
543 dev_err(&pdev->dev, "Class registration failed (%d)\n", err);
549 device_remove_file(&pdev->dev, &pdata->name_attr);
550 platform_set_drvdata(pdev, NULL);
556 static int __devexit coretemp_remove(struct platform_device *pdev)
558 struct platform_data *pdata = platform_get_drvdata(pdev);
561 for (i = MAX_CORE_DATA - 1; i >= 0; --i)
562 if (pdata->core_data[i])
563 coretemp_remove_core(pdata, &pdev->dev, i);
565 device_remove_file(&pdev->dev, &pdata->name_attr);
566 hwmon_device_unregister(pdata->hwmon_dev);
567 platform_set_drvdata(pdev, NULL);
572 static struct platform_driver coretemp_driver = {
574 .owner = THIS_MODULE,
577 .probe = coretemp_probe,
578 .remove = __devexit_p(coretemp_remove),
581 static int __cpuinit coretemp_device_add(unsigned int cpu)
584 struct platform_device *pdev;
585 struct pdev_entry *pdev_entry;
587 mutex_lock(&pdev_list_mutex);
589 pdev = platform_device_alloc(DRVNAME, TO_PHYS_ID(cpu));
592 pr_err("Device allocation failed\n");
596 pdev_entry = kzalloc(sizeof(struct pdev_entry), GFP_KERNEL);
599 goto exit_device_put;
602 err = platform_device_add(pdev);
604 pr_err("Device addition failed (%d)\n", err);
605 goto exit_device_free;
608 pdev_entry->pdev = pdev;
609 pdev_entry->phys_proc_id = pdev->id;
611 list_add_tail(&pdev_entry->list, &pdev_list);
612 mutex_unlock(&pdev_list_mutex);
619 platform_device_put(pdev);
621 mutex_unlock(&pdev_list_mutex);
625 static void coretemp_device_remove(unsigned int cpu)
627 struct pdev_entry *p, *n;
628 u16 phys_proc_id = TO_PHYS_ID(cpu);
630 mutex_lock(&pdev_list_mutex);
631 list_for_each_entry_safe(p, n, &pdev_list, list) {
632 if (p->phys_proc_id != phys_proc_id)
634 platform_device_unregister(p->pdev);
638 mutex_unlock(&pdev_list_mutex);
641 static bool is_any_core_online(struct platform_data *pdata)
645 /* Find online cores, except pkgtemp data */
646 for (i = MAX_CORE_DATA - 1; i >= 0; --i) {
647 if (pdata->core_data[i] &&
648 !pdata->core_data[i]->is_pkg_data) {
655 static void __cpuinit get_core_online(unsigned int cpu)
657 struct cpuinfo_x86 *c = &cpu_data(cpu);
658 struct platform_device *pdev = coretemp_get_pdev(cpu);
662 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
663 * sensors. We check this bit only, all the early CPUs
664 * without thermal sensors will be filtered out.
666 if (!cpu_has(c, X86_FEATURE_DTHERM))
670 /* Check the microcode version of the CPU */
671 if (chk_ucode_version(cpu))
675 * Alright, we have DTS support.
676 * We are bringing the _first_ core in this pkg
677 * online. So, initialize per-pkg data structures and
678 * then bring this core online.
680 err = coretemp_device_add(cpu);
684 * Check whether pkgtemp support is available.
685 * If so, add interfaces for pkgtemp.
687 if (cpu_has(c, X86_FEATURE_PTS))
688 coretemp_add_core(cpu, 1);
691 * Physical CPU device already exists.
692 * So, just add interfaces for this core.
694 coretemp_add_core(cpu, 0);
697 static void __cpuinit put_core_offline(unsigned int cpu)
700 struct platform_data *pdata;
701 struct platform_device *pdev = coretemp_get_pdev(cpu);
703 /* If the physical CPU device does not exist, just return */
707 pdata = platform_get_drvdata(pdev);
709 indx = TO_ATTR_NO(cpu);
711 /* The core id is too big, just return */
712 if (indx > MAX_CORE_DATA - 1)
715 if (pdata->core_data[indx] && pdata->core_data[indx]->cpu == cpu)
716 coretemp_remove_core(pdata, &pdev->dev, indx);
719 * If a HT sibling of a core is taken offline, but another HT sibling
720 * of the same core is still online, register the alternate sibling.
721 * This ensures that exactly one set of attributes is provided as long
722 * as at least one HT sibling of a core is online.
724 for_each_sibling(i, cpu) {
728 * Display temperature sensor data for one HT sibling
729 * per core only, so abort the loop after one such
730 * sibling has been found.
736 * If all cores in this pkg are offline, remove the device.
737 * coretemp_device_remove calls unregister_platform_device,
738 * which in turn calls coretemp_remove. This removes the
739 * pkgtemp entry and does other clean ups.
741 if (!is_any_core_online(pdata))
742 coretemp_device_remove(cpu);
745 static int __cpuinit coretemp_cpu_callback(struct notifier_block *nfb,
746 unsigned long action, void *hcpu)
748 unsigned int cpu = (unsigned long) hcpu;
752 case CPU_DOWN_FAILED:
753 get_core_online(cpu);
755 case CPU_DOWN_PREPARE:
756 put_core_offline(cpu);
762 static struct notifier_block coretemp_cpu_notifier __refdata = {
763 .notifier_call = coretemp_cpu_callback,
766 static int __init coretemp_init(void)
768 int i, err = -ENODEV;
770 /* quick check if we run Intel */
771 if (cpu_data(0).x86_vendor != X86_VENDOR_INTEL)
774 err = platform_driver_register(&coretemp_driver);
778 for_each_online_cpu(i)
781 #ifndef CONFIG_HOTPLUG_CPU
782 if (list_empty(&pdev_list)) {
784 goto exit_driver_unreg;
788 register_hotcpu_notifier(&coretemp_cpu_notifier);
791 #ifndef CONFIG_HOTPLUG_CPU
793 platform_driver_unregister(&coretemp_driver);
799 static void __exit coretemp_exit(void)
801 struct pdev_entry *p, *n;
803 unregister_hotcpu_notifier(&coretemp_cpu_notifier);
804 mutex_lock(&pdev_list_mutex);
805 list_for_each_entry_safe(p, n, &pdev_list, list) {
806 platform_device_unregister(p->pdev);
810 mutex_unlock(&pdev_list_mutex);
811 platform_driver_unregister(&coretemp_driver);
814 MODULE_AUTHOR("Rudolf Marek <r.marek@assembler.cz>");
815 MODULE_DESCRIPTION("Intel Core temperature monitor");
816 MODULE_LICENSE("GPL");
818 module_init(coretemp_init)
819 module_exit(coretemp_exit)