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 32 /* 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 for_each_sibling(i, cpu) for_each_cpu(i, cpu_sibling_mask(cpu))
55 #define TO_PHYS_ID(cpu) (cpu)
56 #define TO_CORE_ID(cpu) (cpu)
57 #define for_each_sibling(i, cpu) for (i = 0; false; )
59 #define TO_ATTR_NO(cpu) (TO_CORE_ID(cpu) + BASE_SYSFS_ATTR_NO)
62 * Per-Core Temperature Data
63 * @last_updated: The time when the current temperature value was updated
64 * earlier (in jiffies).
65 * @cpu_core_id: The CPU Core from which temperature values should be read
66 * This value is passed as "id" field to rdmsr/wrmsr functions.
67 * @status_reg: One of IA32_THERM_STATUS or IA32_PACKAGE_THERM_STATUS,
68 * from where the temperature values should be read.
69 * @is_pkg_data: If this is 1, the temp_data holds pkgtemp data.
70 * Otherwise, temp_data holds coretemp data.
71 * @valid: If this is 1, the current temperature is valid.
77 unsigned long last_updated;
83 struct sensor_device_attribute sd_attrs[MAX_ATTRS];
84 char attr_name[MAX_ATTRS][CORETEMP_NAME_LENGTH];
85 struct mutex update_lock;
88 /* Platform Data per Physical CPU */
89 struct platform_data {
90 struct device *hwmon_dev;
92 struct temp_data *core_data[MAX_CORE_DATA];
93 struct device_attribute name_attr;
97 struct list_head list;
98 struct platform_device *pdev;
102 static LIST_HEAD(pdev_list);
103 static DEFINE_MUTEX(pdev_list_mutex);
105 static ssize_t show_name(struct device *dev,
106 struct device_attribute *devattr, char *buf)
108 return sprintf(buf, "%s\n", DRVNAME);
111 static ssize_t show_label(struct device *dev,
112 struct device_attribute *devattr, char *buf)
114 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
115 struct platform_data *pdata = dev_get_drvdata(dev);
116 struct temp_data *tdata = pdata->core_data[attr->index];
118 if (tdata->is_pkg_data)
119 return sprintf(buf, "Physical id %u\n", pdata->phys_proc_id);
121 return sprintf(buf, "Core %u\n", tdata->cpu_core_id);
124 static ssize_t show_crit_alarm(struct device *dev,
125 struct device_attribute *devattr, char *buf)
128 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
129 struct platform_data *pdata = dev_get_drvdata(dev);
130 struct temp_data *tdata = pdata->core_data[attr->index];
132 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
134 return sprintf(buf, "%d\n", (eax >> 5) & 1);
137 static ssize_t show_tjmax(struct device *dev,
138 struct device_attribute *devattr, char *buf)
140 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
141 struct platform_data *pdata = dev_get_drvdata(dev);
143 return sprintf(buf, "%d\n", pdata->core_data[attr->index]->tjmax);
146 static ssize_t show_ttarget(struct device *dev,
147 struct device_attribute *devattr, char *buf)
149 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
150 struct platform_data *pdata = dev_get_drvdata(dev);
152 return sprintf(buf, "%d\n", pdata->core_data[attr->index]->ttarget);
155 static ssize_t show_temp(struct device *dev,
156 struct device_attribute *devattr, char *buf)
159 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
160 struct platform_data *pdata = dev_get_drvdata(dev);
161 struct temp_data *tdata = pdata->core_data[attr->index];
163 mutex_lock(&tdata->update_lock);
165 /* Check whether the time interval has elapsed */
166 if (!tdata->valid || time_after(jiffies, tdata->last_updated + HZ)) {
167 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
169 /* Check whether the data is valid */
170 if (eax & 0x80000000) {
171 tdata->temp = tdata->tjmax -
172 ((eax >> 16) & 0x7f) * 1000;
175 tdata->last_updated = jiffies;
178 mutex_unlock(&tdata->update_lock);
179 return tdata->valid ? sprintf(buf, "%d\n", tdata->temp) : -EAGAIN;
182 static int adjust_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
184 /* The 100C is default for both mobile and non mobile CPUs */
187 int tjmax_ee = 85000;
191 struct pci_dev *host_bridge;
193 /* Early chips have no MSR for TjMax */
195 if (c->x86_model == 0xf && c->x86_mask < 4)
200 if (c->x86_model == 0x1c) {
203 host_bridge = pci_get_bus_and_slot(0, PCI_DEVFN(0, 0));
205 if (host_bridge && host_bridge->vendor == PCI_VENDOR_ID_INTEL
206 && (host_bridge->device == 0xa000 /* NM10 based nettop */
207 || host_bridge->device == 0xa010)) /* NM10 based netbook */
212 pci_dev_put(host_bridge);
215 if (c->x86_model > 0xe && usemsr_ee) {
219 * Now we can detect the mobile CPU using Intel provided table
220 * http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
221 * For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
223 err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
226 "Unable to access MSR 0x17, assuming desktop"
229 } else if (c->x86_model < 0x17 && !(eax & 0x10000000)) {
231 * Trust bit 28 up to Penryn, I could not find any
232 * documentation on that; if you happen to know
233 * someone at Intel please ask
237 /* Platform ID bits 52:50 (EDX starts at bit 32) */
238 platform_id = (edx >> 18) & 0x7;
241 * Mobile Penryn CPU seems to be platform ID 7 or 5
244 if (c->x86_model == 0x17 &&
245 (platform_id == 5 || platform_id == 7)) {
247 * If MSR EE bit is set, set it to 90 degrees C,
248 * otherwise 105 degrees C
257 err = rdmsr_safe_on_cpu(id, 0xee, &eax, &edx);
260 "Unable to access MSR 0xEE, for Tjmax, left"
262 } else if (eax & 0x40000000) {
265 } else if (tjmax == 100000) {
267 * If we don't use msr EE it means we are desktop CPU
268 * (with exeception of Atom)
270 dev_warn(dev, "Using relative temperature scale!\n");
276 static int get_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
278 /* The 100C is default for both mobile and non mobile CPUs */
284 * A new feature of current Intel(R) processors, the
285 * IA32_TEMPERATURE_TARGET contains the TjMax value
287 err = rdmsr_safe_on_cpu(id, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
289 dev_warn(dev, "Unable to read TjMax from CPU.\n");
291 val = (eax >> 16) & 0xff;
293 * If the TjMax is not plausible, an assumption
297 dev_info(dev, "TjMax is %d C.\n", val);
303 * An assumption is made for early CPUs and unreadable MSR.
304 * NOTE: the calculated value may not be correct.
306 return adjust_tjmax(c, id, dev);
309 static void __devinit get_ucode_rev_on_cpu(void *edx)
313 wrmsr(MSR_IA32_UCODE_REV, 0, 0);
315 rdmsr(MSR_IA32_UCODE_REV, eax, *(u32 *)edx);
318 static int get_pkg_tjmax(unsigned int cpu, struct device *dev)
323 err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
325 val = (eax >> 16) & 0xff;
329 dev_warn(dev, "Unable to read Pkg-TjMax from CPU:%u\n", cpu);
330 return 100000; /* Default TjMax: 100 degree celsius */
333 static int create_name_attr(struct platform_data *pdata, struct device *dev)
335 sysfs_attr_init(&pdata->name_attr.attr);
336 pdata->name_attr.attr.name = "name";
337 pdata->name_attr.attr.mode = S_IRUGO;
338 pdata->name_attr.show = show_name;
339 return device_create_file(dev, &pdata->name_attr);
342 static int create_core_attrs(struct temp_data *tdata, struct device *dev,
346 static ssize_t (*rd_ptr[MAX_ATTRS]) (struct device *dev,
347 struct device_attribute *devattr, char *buf) = {
348 show_label, show_crit_alarm, show_ttarget,
349 show_temp, show_tjmax };
350 static const char *names[MAX_ATTRS] = {
351 "temp%d_label", "temp%d_crit_alarm",
352 "temp%d_max", "temp%d_input",
355 for (i = 0; i < MAX_ATTRS; i++) {
356 snprintf(tdata->attr_name[i], CORETEMP_NAME_LENGTH, names[i],
358 sysfs_attr_init(&tdata->sd_attrs[i].dev_attr.attr);
359 tdata->sd_attrs[i].dev_attr.attr.name = tdata->attr_name[i];
360 tdata->sd_attrs[i].dev_attr.attr.mode = S_IRUGO;
361 tdata->sd_attrs[i].dev_attr.show = rd_ptr[i];
362 tdata->sd_attrs[i].dev_attr.store = NULL;
363 tdata->sd_attrs[i].index = attr_no;
364 err = device_create_file(dev, &tdata->sd_attrs[i].dev_attr);
372 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
376 static void update_ttarget(__u8 cpu_model, struct temp_data *tdata,
383 * Initialize ttarget value. Eventually this will be
384 * initialized with the value from MSR_IA32_THERM_INTERRUPT
385 * register. If IA32_TEMPERATURE_TARGET is supported, this
386 * value will be over written below.
387 * To Do: Patch to initialize ttarget from MSR_IA32_THERM_INTERRUPT
389 tdata->ttarget = tdata->tjmax - 20000;
392 * Read the still undocumented IA32_TEMPERATURE_TARGET. It exists
393 * on older CPUs but not in this register,
394 * Atoms don't have it either.
396 if (cpu_model > 0xe && cpu_model != 0x1c) {
397 err = rdmsr_safe_on_cpu(tdata->cpu,
398 MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
401 "Unable to read IA32_TEMPERATURE_TARGET MSR\n");
403 tdata->ttarget = tdata->tjmax -
404 ((eax >> 8) & 0xff) * 1000;
409 static int __devinit chk_ucode_version(struct platform_device *pdev)
411 struct cpuinfo_x86 *c = &cpu_data(pdev->id);
416 * Check if we have problem with errata AE18 of Core processors:
417 * Readings might stop update when processor visited too deep sleep,
418 * fixed for stepping D0 (6EC).
420 if (c->x86_model == 0xe && c->x86_mask < 0xc) {
421 /* check for microcode update */
422 err = smp_call_function_single(pdev->id, get_ucode_rev_on_cpu,
426 "Cannot determine microcode revision of "
427 "CPU#%u (%d)!\n", pdev->id, err);
429 } else if (edx < 0x39) {
431 "Errata AE18 not fixed, update BIOS or "
432 "microcode of the CPU!\n");
439 static struct platform_device *coretemp_get_pdev(unsigned int cpu)
441 u16 phys_proc_id = TO_PHYS_ID(cpu);
442 struct pdev_entry *p;
444 mutex_lock(&pdev_list_mutex);
446 list_for_each_entry(p, &pdev_list, list)
447 if (p->phys_proc_id == phys_proc_id) {
448 mutex_unlock(&pdev_list_mutex);
452 mutex_unlock(&pdev_list_mutex);
456 static struct temp_data *init_temp_data(unsigned int cpu, int pkg_flag)
458 struct temp_data *tdata;
460 tdata = kzalloc(sizeof(struct temp_data), GFP_KERNEL);
464 tdata->status_reg = pkg_flag ? MSR_IA32_PACKAGE_THERM_STATUS :
465 MSR_IA32_THERM_STATUS;
466 tdata->is_pkg_data = pkg_flag;
468 tdata->cpu_core_id = TO_CORE_ID(cpu);
469 mutex_init(&tdata->update_lock);
473 static int create_core_data(struct platform_data *pdata,
474 struct platform_device *pdev,
475 unsigned int cpu, int pkg_flag)
477 struct temp_data *tdata;
478 struct cpuinfo_x86 *c = &cpu_data(cpu);
483 * Find attr number for sysfs:
484 * We map the attr number to core id of the CPU
485 * The attr number is always core id + 2
486 * The Pkgtemp will always show up as temp1_*, if available
488 attr_no = pkg_flag ? 1 : TO_ATTR_NO(cpu);
490 if (attr_no > MAX_CORE_DATA - 1)
494 * Provide a single set of attributes for all HT siblings of a core
495 * to avoid duplicate sensors (the processor ID and core ID of all
496 * HT siblings of a core are the same).
497 * Skip if a HT sibling of this core is already registered.
498 * This is not an error.
500 if (pdata->core_data[attr_no] != NULL)
503 tdata = init_temp_data(cpu, pkg_flag);
507 /* Test if we can access the status register */
508 err = rdmsr_safe_on_cpu(cpu, tdata->status_reg, &eax, &edx);
512 /* We can access status register. Get Critical Temperature */
514 tdata->tjmax = get_pkg_tjmax(pdev->id, &pdev->dev);
516 tdata->tjmax = get_tjmax(c, cpu, &pdev->dev);
518 update_ttarget(c->x86_model, tdata, &pdev->dev);
519 pdata->core_data[attr_no] = tdata;
521 /* Create sysfs interfaces */
522 err = create_core_attrs(tdata, &pdev->dev, attr_no);
532 static void coretemp_add_core(unsigned int cpu, int pkg_flag)
534 struct platform_data *pdata;
535 struct platform_device *pdev = coretemp_get_pdev(cpu);
541 pdata = platform_get_drvdata(pdev);
545 err = create_core_data(pdata, pdev, cpu, pkg_flag);
547 dev_err(&pdev->dev, "Adding Core %u failed\n", cpu);
550 static void coretemp_remove_core(struct platform_data *pdata,
551 struct device *dev, int indx)
554 struct temp_data *tdata = pdata->core_data[indx];
556 /* Remove the sysfs attributes */
557 for (i = 0; i < MAX_ATTRS; i++)
558 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
560 kfree(pdata->core_data[indx]);
561 pdata->core_data[indx] = NULL;
564 static int __devinit coretemp_probe(struct platform_device *pdev)
566 struct platform_data *pdata;
569 /* Check the microcode version of the CPU */
570 err = chk_ucode_version(pdev);
574 /* Initialize the per-package data structures */
575 pdata = kzalloc(sizeof(struct platform_data), GFP_KERNEL);
579 err = create_name_attr(pdata, &pdev->dev);
583 pdata->phys_proc_id = TO_PHYS_ID(pdev->id);
584 platform_set_drvdata(pdev, pdata);
586 pdata->hwmon_dev = hwmon_device_register(&pdev->dev);
587 if (IS_ERR(pdata->hwmon_dev)) {
588 err = PTR_ERR(pdata->hwmon_dev);
589 dev_err(&pdev->dev, "Class registration failed (%d)\n", err);
595 device_remove_file(&pdev->dev, &pdata->name_attr);
596 platform_set_drvdata(pdev, NULL);
602 static int __devexit coretemp_remove(struct platform_device *pdev)
604 struct platform_data *pdata = platform_get_drvdata(pdev);
607 for (i = MAX_CORE_DATA - 1; i >= 0; --i)
608 if (pdata->core_data[i])
609 coretemp_remove_core(pdata, &pdev->dev, i);
611 device_remove_file(&pdev->dev, &pdata->name_attr);
612 hwmon_device_unregister(pdata->hwmon_dev);
613 platform_set_drvdata(pdev, NULL);
618 static struct platform_driver coretemp_driver = {
620 .owner = THIS_MODULE,
623 .probe = coretemp_probe,
624 .remove = __devexit_p(coretemp_remove),
627 static int __cpuinit coretemp_device_add(unsigned int cpu)
630 struct platform_device *pdev;
631 struct pdev_entry *pdev_entry;
633 mutex_lock(&pdev_list_mutex);
635 pdev = platform_device_alloc(DRVNAME, cpu);
638 pr_err("Device allocation failed\n");
642 pdev_entry = kzalloc(sizeof(struct pdev_entry), GFP_KERNEL);
645 goto exit_device_put;
648 err = platform_device_add(pdev);
650 pr_err("Device addition failed (%d)\n", err);
651 goto exit_device_free;
654 pdev_entry->pdev = pdev;
655 pdev_entry->phys_proc_id = TO_PHYS_ID(cpu);
657 list_add_tail(&pdev_entry->list, &pdev_list);
658 mutex_unlock(&pdev_list_mutex);
665 platform_device_put(pdev);
667 mutex_unlock(&pdev_list_mutex);
671 static void coretemp_device_remove(unsigned int cpu)
673 struct pdev_entry *p, *n;
674 u16 phys_proc_id = TO_PHYS_ID(cpu);
676 mutex_lock(&pdev_list_mutex);
677 list_for_each_entry_safe(p, n, &pdev_list, list) {
678 if (p->phys_proc_id != phys_proc_id)
680 platform_device_unregister(p->pdev);
684 mutex_unlock(&pdev_list_mutex);
687 static bool is_any_core_online(struct platform_data *pdata)
691 /* Find online cores, except pkgtemp data */
692 for (i = MAX_CORE_DATA - 1; i >= 0; --i) {
693 if (pdata->core_data[i] &&
694 !pdata->core_data[i]->is_pkg_data) {
701 static void __cpuinit get_core_online(unsigned int cpu)
703 struct cpuinfo_x86 *c = &cpu_data(cpu);
704 struct platform_device *pdev = coretemp_get_pdev(cpu);
708 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
709 * sensors. We check this bit only, all the early CPUs
710 * without thermal sensors will be filtered out.
712 if (!cpu_has(c, X86_FEATURE_DTS))
717 * Alright, we have DTS support.
718 * We are bringing the _first_ core in this pkg
719 * online. So, initialize per-pkg data structures and
720 * then bring this core online.
722 err = coretemp_device_add(cpu);
726 * Check whether pkgtemp support is available.
727 * If so, add interfaces for pkgtemp.
729 if (cpu_has(c, X86_FEATURE_PTS))
730 coretemp_add_core(cpu, 1);
733 * Physical CPU device already exists.
734 * So, just add interfaces for this core.
736 coretemp_add_core(cpu, 0);
739 static void __cpuinit put_core_offline(unsigned int cpu)
742 struct platform_data *pdata;
743 struct platform_device *pdev = coretemp_get_pdev(cpu);
745 /* If the physical CPU device does not exist, just return */
749 pdata = platform_get_drvdata(pdev);
753 indx = TO_ATTR_NO(cpu);
755 /* The core id is too big, just return */
756 if (indx > MAX_CORE_DATA - 1)
759 if (pdata->core_data[indx] && pdata->core_data[indx]->cpu == cpu)
760 coretemp_remove_core(pdata, &pdev->dev, indx);
763 * If a HT sibling of a core is taken offline, but another HT sibling
764 * of the same core is still online, register the alternate sibling.
765 * This ensures that exactly one set of attributes is provided as long
766 * as at least one HT sibling of a core is online.
768 for_each_sibling(i, cpu) {
772 * Display temperature sensor data for one HT sibling
773 * per core only, so abort the loop after one such
774 * sibling has been found.
780 * If all cores in this pkg are offline, remove the device.
781 * coretemp_device_remove calls unregister_platform_device,
782 * which in turn calls coretemp_remove. This removes the
783 * pkgtemp entry and does other clean ups.
785 if (!is_any_core_online(pdata))
786 coretemp_device_remove(cpu);
789 static int __cpuinit coretemp_cpu_callback(struct notifier_block *nfb,
790 unsigned long action, void *hcpu)
792 unsigned int cpu = (unsigned long) hcpu;
796 case CPU_DOWN_FAILED:
797 get_core_online(cpu);
799 case CPU_DOWN_PREPARE:
800 put_core_offline(cpu);
806 static struct notifier_block coretemp_cpu_notifier __refdata = {
807 .notifier_call = coretemp_cpu_callback,
810 static int __init coretemp_init(void)
812 int i, err = -ENODEV;
814 /* quick check if we run Intel */
815 if (cpu_data(0).x86_vendor != X86_VENDOR_INTEL)
818 err = platform_driver_register(&coretemp_driver);
822 for_each_online_cpu(i)
825 #ifndef CONFIG_HOTPLUG_CPU
826 if (list_empty(&pdev_list)) {
828 goto exit_driver_unreg;
832 register_hotcpu_notifier(&coretemp_cpu_notifier);
835 #ifndef CONFIG_HOTPLUG_CPU
837 platform_driver_unregister(&coretemp_driver);
843 static void __exit coretemp_exit(void)
845 struct pdev_entry *p, *n;
847 unregister_hotcpu_notifier(&coretemp_cpu_notifier);
848 mutex_lock(&pdev_list_mutex);
849 list_for_each_entry_safe(p, n, &pdev_list, list) {
850 platform_device_unregister(p->pdev);
854 mutex_unlock(&pdev_list_mutex);
855 platform_driver_unregister(&coretemp_driver);
858 MODULE_AUTHOR("Rudolf Marek <r.marek@assembler.cz>");
859 MODULE_DESCRIPTION("Intel Core temperature monitor");
860 MODULE_LICENSE("GPL");
862 module_init(coretemp_init)
863 module_exit(coretemp_exit)