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>
40 #include <asm/processor.h>
42 #define DRVNAME "coretemp"
44 #define BASE_SYSFS_ATTR_NO 2 /* Sysfs Base attr no for coretemp */
45 #define NUM_REAL_CORES 16 /* Number of Real cores per cpu */
46 #define CORETEMP_NAME_LENGTH 17 /* String Length of attrs */
47 #define MAX_ATTRS 5 /* Maximum no of per-core attrs */
48 #define MAX_CORE_DATA (NUM_REAL_CORES + BASE_SYSFS_ATTR_NO)
51 #define TO_PHYS_ID(cpu) cpu_data(cpu).phys_proc_id
52 #define TO_CORE_ID(cpu) cpu_data(cpu).cpu_core_id
53 #define TO_ATTR_NO(cpu) (TO_CORE_ID(cpu) + BASE_SYSFS_ATTR_NO)
54 #define for_each_sibling(i, cpu) for_each_cpu(i, cpu_sibling_mask(cpu))
56 #define TO_PHYS_ID(cpu) (cpu)
57 #define TO_CORE_ID(cpu) (cpu)
58 #define TO_ATTR_NO(cpu) (cpu)
59 #define for_each_sibling(i, cpu) for (i = 0; false; )
63 * Per-Core Temperature Data
64 * @last_updated: The time when the current temperature value was updated
65 * earlier (in jiffies).
66 * @cpu_core_id: The CPU Core from which temperature values should be read
67 * This value is passed as "id" field to rdmsr/wrmsr functions.
68 * @status_reg: One of IA32_THERM_STATUS or IA32_PACKAGE_THERM_STATUS,
69 * from where the temperature values should be read.
70 * @is_pkg_data: If this is 1, the temp_data holds pkgtemp data.
71 * Otherwise, temp_data holds coretemp data.
72 * @valid: If this is 1, the current temperature is valid.
78 unsigned long last_updated;
84 struct sensor_device_attribute sd_attrs[MAX_ATTRS];
85 char attr_name[MAX_ATTRS][CORETEMP_NAME_LENGTH];
86 struct mutex update_lock;
89 /* Platform Data per Physical CPU */
90 struct platform_data {
91 struct device *hwmon_dev;
93 struct temp_data *core_data[MAX_CORE_DATA];
94 struct device_attribute name_attr;
98 struct list_head list;
99 struct platform_device *pdev;
103 static LIST_HEAD(pdev_list);
104 static DEFINE_MUTEX(pdev_list_mutex);
106 static ssize_t show_name(struct device *dev,
107 struct device_attribute *devattr, char *buf)
109 return sprintf(buf, "%s\n", DRVNAME);
112 static ssize_t show_label(struct device *dev,
113 struct device_attribute *devattr, char *buf)
115 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
116 struct platform_data *pdata = dev_get_drvdata(dev);
117 struct temp_data *tdata = pdata->core_data[attr->index];
119 if (tdata->is_pkg_data)
120 return sprintf(buf, "Physical id %u\n", pdata->phys_proc_id);
122 return sprintf(buf, "Core %u\n", tdata->cpu_core_id);
125 static ssize_t show_crit_alarm(struct device *dev,
126 struct device_attribute *devattr, char *buf)
129 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
130 struct platform_data *pdata = dev_get_drvdata(dev);
131 struct temp_data *tdata = pdata->core_data[attr->index];
133 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
135 return sprintf(buf, "%d\n", (eax >> 5) & 1);
138 static ssize_t show_tjmax(struct device *dev,
139 struct device_attribute *devattr, char *buf)
141 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
142 struct platform_data *pdata = dev_get_drvdata(dev);
144 return sprintf(buf, "%d\n", pdata->core_data[attr->index]->tjmax);
147 static ssize_t show_ttarget(struct device *dev,
148 struct device_attribute *devattr, char *buf)
150 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
151 struct platform_data *pdata = dev_get_drvdata(dev);
153 return sprintf(buf, "%d\n", pdata->core_data[attr->index]->ttarget);
156 static ssize_t show_temp(struct device *dev,
157 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);
162 struct temp_data *tdata = pdata->core_data[attr->index];
164 mutex_lock(&tdata->update_lock);
166 /* Check whether the time interval has elapsed */
167 if (!tdata->valid || time_after(jiffies, tdata->last_updated + HZ)) {
168 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
170 /* Check whether the data is valid */
171 if (eax & 0x80000000) {
172 tdata->temp = tdata->tjmax -
173 ((eax >> 16) & 0x7f) * 1000;
176 tdata->last_updated = jiffies;
179 mutex_unlock(&tdata->update_lock);
180 return tdata->valid ? sprintf(buf, "%d\n", tdata->temp) : -EAGAIN;
183 static int adjust_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
185 /* The 100C is default for both mobile and non mobile CPUs */
188 int tjmax_ee = 85000;
192 struct pci_dev *host_bridge;
194 /* Early chips have no MSR for TjMax */
196 if (c->x86_model == 0xf && c->x86_mask < 4)
201 if (c->x86_model == 0x1c) {
204 host_bridge = pci_get_bus_and_slot(0, PCI_DEVFN(0, 0));
206 if (host_bridge && host_bridge->vendor == PCI_VENDOR_ID_INTEL
207 && (host_bridge->device == 0xa000 /* NM10 based nettop */
208 || host_bridge->device == 0xa010)) /* NM10 based netbook */
213 pci_dev_put(host_bridge);
216 if (c->x86_model > 0xe && usemsr_ee) {
220 * Now we can detect the mobile CPU using Intel provided table
221 * http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
222 * For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
224 err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
227 "Unable to access MSR 0x17, assuming desktop"
230 } else if (c->x86_model < 0x17 && !(eax & 0x10000000)) {
232 * Trust bit 28 up to Penryn, I could not find any
233 * documentation on that; if you happen to know
234 * someone at Intel please ask
238 /* Platform ID bits 52:50 (EDX starts at bit 32) */
239 platform_id = (edx >> 18) & 0x7;
242 * Mobile Penryn CPU seems to be platform ID 7 or 5
245 if (c->x86_model == 0x17 &&
246 (platform_id == 5 || platform_id == 7)) {
248 * If MSR EE bit is set, set it to 90 degrees C,
249 * otherwise 105 degrees C
258 err = rdmsr_safe_on_cpu(id, 0xee, &eax, &edx);
261 "Unable to access MSR 0xEE, for Tjmax, left"
263 } else if (eax & 0x40000000) {
266 } else if (tjmax == 100000) {
268 * If we don't use msr EE it means we are desktop CPU
269 * (with exeception of Atom)
271 dev_warn(dev, "Using relative temperature scale!\n");
277 static int get_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
279 /* The 100C is default for both mobile and non mobile CPUs */
285 * A new feature of current Intel(R) processors, the
286 * IA32_TEMPERATURE_TARGET contains the TjMax value
288 err = rdmsr_safe_on_cpu(id, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
290 dev_warn(dev, "Unable to read TjMax from CPU.\n");
292 val = (eax >> 16) & 0xff;
294 * If the TjMax is not plausible, an assumption
298 dev_info(dev, "TjMax is %d C.\n", val);
304 * An assumption is made for early CPUs and unreadable MSR.
305 * NOTE: the calculated value may not be correct.
307 return adjust_tjmax(c, id, dev);
310 static void __devinit get_ucode_rev_on_cpu(void *edx)
314 wrmsr(MSR_IA32_UCODE_REV, 0, 0);
316 rdmsr(MSR_IA32_UCODE_REV, eax, *(u32 *)edx);
319 static int get_pkg_tjmax(unsigned int cpu, struct device *dev)
324 err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
326 val = (eax >> 16) & 0xff;
330 dev_warn(dev, "Unable to read Pkg-TjMax from CPU:%u\n", cpu);
331 return 100000; /* Default TjMax: 100 degree celsius */
334 static int create_name_attr(struct platform_data *pdata, struct device *dev)
336 sysfs_attr_init(&pdata->name_attr.attr);
337 pdata->name_attr.attr.name = "name";
338 pdata->name_attr.attr.mode = S_IRUGO;
339 pdata->name_attr.show = show_name;
340 return device_create_file(dev, &pdata->name_attr);
343 static int create_core_attrs(struct temp_data *tdata, struct device *dev,
347 static ssize_t (*rd_ptr[MAX_ATTRS]) (struct device *dev,
348 struct device_attribute *devattr, char *buf) = {
349 show_label, show_crit_alarm, show_ttarget,
350 show_temp, show_tjmax };
351 static const char *names[MAX_ATTRS] = {
352 "temp%d_label", "temp%d_crit_alarm",
353 "temp%d_max", "temp%d_input",
356 for (i = 0; i < MAX_ATTRS; i++) {
357 snprintf(tdata->attr_name[i], CORETEMP_NAME_LENGTH, names[i],
359 sysfs_attr_init(&tdata->sd_attrs[i].dev_attr.attr);
360 tdata->sd_attrs[i].dev_attr.attr.name = tdata->attr_name[i];
361 tdata->sd_attrs[i].dev_attr.attr.mode = S_IRUGO;
362 tdata->sd_attrs[i].dev_attr.show = rd_ptr[i];
363 tdata->sd_attrs[i].dev_attr.store = NULL;
364 tdata->sd_attrs[i].index = attr_no;
365 err = device_create_file(dev, &tdata->sd_attrs[i].dev_attr);
373 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
377 static void update_ttarget(__u8 cpu_model, struct temp_data *tdata,
384 * Initialize ttarget value. Eventually this will be
385 * initialized with the value from MSR_IA32_THERM_INTERRUPT
386 * register. If IA32_TEMPERATURE_TARGET is supported, this
387 * value will be over written below.
388 * To Do: Patch to initialize ttarget from MSR_IA32_THERM_INTERRUPT
390 tdata->ttarget = tdata->tjmax - 20000;
393 * Read the still undocumented IA32_TEMPERATURE_TARGET. It exists
394 * on older CPUs but not in this register,
395 * Atoms don't have it either.
397 if (cpu_model > 0xe && cpu_model != 0x1c) {
398 err = rdmsr_safe_on_cpu(tdata->cpu,
399 MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
402 "Unable to read IA32_TEMPERATURE_TARGET MSR\n");
404 tdata->ttarget = tdata->tjmax -
405 ((eax >> 8) & 0xff) * 1000;
410 static int __devinit chk_ucode_version(struct platform_device *pdev)
412 struct cpuinfo_x86 *c = &cpu_data(pdev->id);
417 * Check if we have problem with errata AE18 of Core processors:
418 * Readings might stop update when processor visited too deep sleep,
419 * fixed for stepping D0 (6EC).
421 if (c->x86_model == 0xe && c->x86_mask < 0xc) {
422 /* check for microcode update */
423 err = smp_call_function_single(pdev->id, get_ucode_rev_on_cpu,
427 "Cannot determine microcode revision of "
428 "CPU#%u (%d)!\n", pdev->id, err);
430 } else if (edx < 0x39) {
432 "Errata AE18 not fixed, update BIOS or "
433 "microcode of the CPU!\n");
440 static struct platform_device *coretemp_get_pdev(unsigned int cpu)
442 u16 phys_proc_id = TO_PHYS_ID(cpu);
443 struct pdev_entry *p;
445 mutex_lock(&pdev_list_mutex);
447 list_for_each_entry(p, &pdev_list, list)
448 if (p->phys_proc_id == phys_proc_id) {
449 mutex_unlock(&pdev_list_mutex);
453 mutex_unlock(&pdev_list_mutex);
457 static struct temp_data *init_temp_data(unsigned int cpu, int pkg_flag)
459 struct temp_data *tdata;
461 tdata = kzalloc(sizeof(struct temp_data), GFP_KERNEL);
465 tdata->status_reg = pkg_flag ? MSR_IA32_PACKAGE_THERM_STATUS :
466 MSR_IA32_THERM_STATUS;
467 tdata->is_pkg_data = pkg_flag;
469 tdata->cpu_core_id = TO_CORE_ID(cpu);
470 mutex_init(&tdata->update_lock);
474 static int create_core_data(struct platform_data *pdata,
475 struct platform_device *pdev,
476 unsigned int cpu, int pkg_flag)
478 struct temp_data *tdata;
479 struct cpuinfo_x86 *c = &cpu_data(cpu);
484 * Find attr number for sysfs:
485 * We map the attr number to core id of the CPU
486 * The attr number is always core id + 2
487 * The Pkgtemp will always show up as temp1_*, if available
489 attr_no = pkg_flag ? 1 : TO_ATTR_NO(cpu);
491 if (attr_no > MAX_CORE_DATA - 1)
495 * Provide a single set of attributes for all HT siblings of a core
496 * to avoid duplicate sensors (the processor ID and core ID of all
497 * HT siblings of a core are the same).
498 * Skip if a HT sibling of this core is already registered.
499 * This is not an error.
501 if (pdata->core_data[attr_no] != NULL)
504 tdata = init_temp_data(cpu, pkg_flag);
508 /* Test if we can access the status register */
509 err = rdmsr_safe_on_cpu(cpu, tdata->status_reg, &eax, &edx);
513 /* We can access status register. Get Critical Temperature */
515 tdata->tjmax = get_pkg_tjmax(pdev->id, &pdev->dev);
517 tdata->tjmax = get_tjmax(c, cpu, &pdev->dev);
519 update_ttarget(c->x86_model, tdata, &pdev->dev);
520 pdata->core_data[attr_no] = tdata;
522 /* Create sysfs interfaces */
523 err = create_core_attrs(tdata, &pdev->dev, attr_no);
533 static void coretemp_add_core(unsigned int cpu, int pkg_flag)
535 struct platform_data *pdata;
536 struct platform_device *pdev = coretemp_get_pdev(cpu);
542 pdata = platform_get_drvdata(pdev);
544 err = create_core_data(pdata, pdev, cpu, pkg_flag);
546 dev_err(&pdev->dev, "Adding Core %u failed\n", cpu);
549 static void coretemp_remove_core(struct platform_data *pdata,
550 struct device *dev, int indx)
553 struct temp_data *tdata = pdata->core_data[indx];
555 /* Remove the sysfs attributes */
556 for (i = 0; i < MAX_ATTRS; i++)
557 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
559 kfree(pdata->core_data[indx]);
560 pdata->core_data[indx] = NULL;
563 static int __devinit coretemp_probe(struct platform_device *pdev)
565 struct platform_data *pdata;
568 /* Check the microcode version of the CPU */
569 err = chk_ucode_version(pdev);
573 /* Initialize the per-package data structures */
574 pdata = kzalloc(sizeof(struct platform_data), GFP_KERNEL);
578 err = create_name_attr(pdata, &pdev->dev);
582 pdata->phys_proc_id = TO_PHYS_ID(pdev->id);
583 platform_set_drvdata(pdev, pdata);
585 pdata->hwmon_dev = hwmon_device_register(&pdev->dev);
586 if (IS_ERR(pdata->hwmon_dev)) {
587 err = PTR_ERR(pdata->hwmon_dev);
588 dev_err(&pdev->dev, "Class registration failed (%d)\n", err);
594 device_remove_file(&pdev->dev, &pdata->name_attr);
595 platform_set_drvdata(pdev, NULL);
601 static int __devexit coretemp_remove(struct platform_device *pdev)
603 struct platform_data *pdata = platform_get_drvdata(pdev);
606 for (i = MAX_CORE_DATA - 1; i >= 0; --i)
607 if (pdata->core_data[i])
608 coretemp_remove_core(pdata, &pdev->dev, i);
610 device_remove_file(&pdev->dev, &pdata->name_attr);
611 hwmon_device_unregister(pdata->hwmon_dev);
612 platform_set_drvdata(pdev, NULL);
617 static struct platform_driver coretemp_driver = {
619 .owner = THIS_MODULE,
622 .probe = coretemp_probe,
623 .remove = __devexit_p(coretemp_remove),
626 static int __cpuinit coretemp_device_add(unsigned int cpu)
629 struct platform_device *pdev;
630 struct pdev_entry *pdev_entry;
632 mutex_lock(&pdev_list_mutex);
634 pdev = platform_device_alloc(DRVNAME, cpu);
637 pr_err("Device allocation failed\n");
641 pdev_entry = kzalloc(sizeof(struct pdev_entry), GFP_KERNEL);
644 goto exit_device_put;
647 err = platform_device_add(pdev);
649 pr_err("Device addition failed (%d)\n", err);
650 goto exit_device_free;
653 pdev_entry->pdev = pdev;
654 pdev_entry->phys_proc_id = TO_PHYS_ID(cpu);
656 list_add_tail(&pdev_entry->list, &pdev_list);
657 mutex_unlock(&pdev_list_mutex);
664 platform_device_put(pdev);
666 mutex_unlock(&pdev_list_mutex);
670 static void coretemp_device_remove(unsigned int cpu)
672 struct pdev_entry *p, *n;
673 u16 phys_proc_id = TO_PHYS_ID(cpu);
675 mutex_lock(&pdev_list_mutex);
676 list_for_each_entry_safe(p, n, &pdev_list, list) {
677 if (p->phys_proc_id != phys_proc_id)
679 platform_device_unregister(p->pdev);
683 mutex_unlock(&pdev_list_mutex);
686 static bool is_any_core_online(struct platform_data *pdata)
690 /* Find online cores, except pkgtemp data */
691 for (i = MAX_CORE_DATA - 1; i >= 0; --i) {
692 if (pdata->core_data[i] &&
693 !pdata->core_data[i]->is_pkg_data) {
700 static void __cpuinit get_core_online(unsigned int cpu)
702 struct cpuinfo_x86 *c = &cpu_data(cpu);
703 struct platform_device *pdev = coretemp_get_pdev(cpu);
707 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
708 * sensors. We check this bit only, all the early CPUs
709 * without thermal sensors will be filtered out.
711 if (!cpu_has(c, X86_FEATURE_DTS))
716 * Alright, we have DTS support.
717 * We are bringing the _first_ core in this pkg
718 * online. So, initialize per-pkg data structures and
719 * then bring this core online.
721 err = coretemp_device_add(cpu);
725 * Check whether pkgtemp support is available.
726 * If so, add interfaces for pkgtemp.
728 if (cpu_has(c, X86_FEATURE_PTS))
729 coretemp_add_core(cpu, 1);
732 * Physical CPU device already exists.
733 * So, just add interfaces for this core.
735 coretemp_add_core(cpu, 0);
738 static void __cpuinit put_core_offline(unsigned int cpu)
741 struct platform_data *pdata;
742 struct platform_device *pdev = coretemp_get_pdev(cpu);
744 /* If the physical CPU device does not exist, just return */
748 pdata = platform_get_drvdata(pdev);
750 indx = TO_ATTR_NO(cpu);
752 if (pdata->core_data[indx] && pdata->core_data[indx]->cpu == cpu)
753 coretemp_remove_core(pdata, &pdev->dev, indx);
756 * If a HT sibling of a core is taken offline, but another HT sibling
757 * of the same core is still online, register the alternate sibling.
758 * This ensures that exactly one set of attributes is provided as long
759 * as at least one HT sibling of a core is online.
761 for_each_sibling(i, cpu) {
765 * Display temperature sensor data for one HT sibling
766 * per core only, so abort the loop after one such
767 * sibling has been found.
773 * If all cores in this pkg are offline, remove the device.
774 * coretemp_device_remove calls unregister_platform_device,
775 * which in turn calls coretemp_remove. This removes the
776 * pkgtemp entry and does other clean ups.
778 if (!is_any_core_online(pdata))
779 coretemp_device_remove(cpu);
782 static int __cpuinit coretemp_cpu_callback(struct notifier_block *nfb,
783 unsigned long action, void *hcpu)
785 unsigned int cpu = (unsigned long) hcpu;
789 case CPU_DOWN_FAILED:
790 get_core_online(cpu);
792 case CPU_DOWN_PREPARE:
793 put_core_offline(cpu);
799 static struct notifier_block coretemp_cpu_notifier __refdata = {
800 .notifier_call = coretemp_cpu_callback,
803 static int __init coretemp_init(void)
805 int i, err = -ENODEV;
807 /* quick check if we run Intel */
808 if (cpu_data(0).x86_vendor != X86_VENDOR_INTEL)
811 err = platform_driver_register(&coretemp_driver);
815 for_each_online_cpu(i)
818 #ifndef CONFIG_HOTPLUG_CPU
819 if (list_empty(&pdev_list)) {
821 goto exit_driver_unreg;
825 register_hotcpu_notifier(&coretemp_cpu_notifier);
828 #ifndef CONFIG_HOTPLUG_CPU
830 platform_driver_unregister(&coretemp_driver);
836 static void __exit coretemp_exit(void)
838 struct pdev_entry *p, *n;
840 unregister_hotcpu_notifier(&coretemp_cpu_notifier);
841 mutex_lock(&pdev_list_mutex);
842 list_for_each_entry_safe(p, n, &pdev_list, list) {
843 platform_device_unregister(p->pdev);
847 mutex_unlock(&pdev_list_mutex);
848 platform_driver_unregister(&coretemp_driver);
851 MODULE_AUTHOR("Rudolf Marek <r.marek@assembler.cz>");
852 MODULE_DESCRIPTION("Intel Core temperature monitor");
853 MODULE_LICENSE("GPL");
855 module_init(coretemp_init)
856 module_exit(coretemp_exit)