xref: /openbmc/linux/drivers/hwmon/coretemp.c (revision 5499b451)
1 /*
2  * coretemp.c - Linux kernel module for hardware monitoring
3  *
4  * Copyright (C) 2007 Rudolf Marek <r.marek@assembler.cz>
5  *
6  * Inspired from many hwmon drivers
7  *
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.
11  *
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.
16  *
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
20  * 02110-1301 USA.
21  */
22 
23 #include <linux/module.h>
24 #include <linux/delay.h>
25 #include <linux/init.h>
26 #include <linux/slab.h>
27 #include <linux/jiffies.h>
28 #include <linux/hwmon.h>
29 #include <linux/sysfs.h>
30 #include <linux/hwmon-sysfs.h>
31 #include <linux/err.h>
32 #include <linux/mutex.h>
33 #include <linux/list.h>
34 #include <linux/platform_device.h>
35 #include <linux/cpu.h>
36 #include <linux/pci.h>
37 #include <asm/msr.h>
38 #include <asm/processor.h>
39 
40 #define DRVNAME	"coretemp"
41 
42 typedef enum { SHOW_TEMP, SHOW_TJMAX, SHOW_TTARGET, SHOW_LABEL,
43 		SHOW_NAME } SHOW;
44 
45 /*
46  * Functions declaration
47  */
48 
49 static struct coretemp_data *coretemp_update_device(struct device *dev);
50 
51 struct coretemp_data {
52 	struct device *hwmon_dev;
53 	struct mutex update_lock;
54 	const char *name;
55 	u32 id;
56 	char valid;		/* zero until following fields are valid */
57 	unsigned long last_updated;	/* in jiffies */
58 	int temp;
59 	int tjmax;
60 	int ttarget;
61 	u8 alarm;
62 };
63 
64 /*
65  * Sysfs stuff
66  */
67 
68 static ssize_t show_name(struct device *dev, struct device_attribute
69 			  *devattr, char *buf)
70 {
71 	int ret;
72 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
73 	struct coretemp_data *data = dev_get_drvdata(dev);
74 
75 	if (attr->index == SHOW_NAME)
76 		ret = sprintf(buf, "%s\n", data->name);
77 	else	/* show label */
78 		ret = sprintf(buf, "Core %d\n", data->id);
79 	return ret;
80 }
81 
82 static ssize_t show_alarm(struct device *dev, struct device_attribute
83 			  *devattr, char *buf)
84 {
85 	struct coretemp_data *data = coretemp_update_device(dev);
86 	/* read the Out-of-spec log, never clear */
87 	return sprintf(buf, "%d\n", data->alarm);
88 }
89 
90 static ssize_t show_temp(struct device *dev,
91 			 struct device_attribute *devattr, char *buf)
92 {
93 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
94 	struct coretemp_data *data = coretemp_update_device(dev);
95 	int err;
96 
97 	if (attr->index == SHOW_TEMP)
98 		err = data->valid ? sprintf(buf, "%d\n", data->temp) : -EAGAIN;
99 	else if (attr->index == SHOW_TJMAX)
100 		err = sprintf(buf, "%d\n", data->tjmax);
101 	else
102 		err = sprintf(buf, "%d\n", data->ttarget);
103 	return err;
104 }
105 
106 static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, show_temp, NULL,
107 			  SHOW_TEMP);
108 static SENSOR_DEVICE_ATTR(temp1_crit, S_IRUGO, show_temp, NULL,
109 			  SHOW_TJMAX);
110 static SENSOR_DEVICE_ATTR(temp1_max, S_IRUGO, show_temp, NULL,
111 			  SHOW_TTARGET);
112 static DEVICE_ATTR(temp1_crit_alarm, S_IRUGO, show_alarm, NULL);
113 static SENSOR_DEVICE_ATTR(temp1_label, S_IRUGO, show_name, NULL, SHOW_LABEL);
114 static SENSOR_DEVICE_ATTR(name, S_IRUGO, show_name, NULL, SHOW_NAME);
115 
116 static struct attribute *coretemp_attributes[] = {
117 	&sensor_dev_attr_name.dev_attr.attr,
118 	&sensor_dev_attr_temp1_label.dev_attr.attr,
119 	&dev_attr_temp1_crit_alarm.attr,
120 	&sensor_dev_attr_temp1_input.dev_attr.attr,
121 	&sensor_dev_attr_temp1_crit.dev_attr.attr,
122 	NULL
123 };
124 
125 static const struct attribute_group coretemp_group = {
126 	.attrs = coretemp_attributes,
127 };
128 
129 static struct coretemp_data *coretemp_update_device(struct device *dev)
130 {
131 	struct coretemp_data *data = dev_get_drvdata(dev);
132 
133 	mutex_lock(&data->update_lock);
134 
135 	if (!data->valid || time_after(jiffies, data->last_updated + HZ)) {
136 		u32 eax, edx;
137 
138 		data->valid = 0;
139 		rdmsr_on_cpu(data->id, MSR_IA32_THERM_STATUS, &eax, &edx);
140 		data->alarm = (eax >> 5) & 1;
141 		/* update only if data has been valid */
142 		if (eax & 0x80000000) {
143 			data->temp = data->tjmax - (((eax >> 16)
144 							& 0x7f) * 1000);
145 			data->valid = 1;
146 		} else {
147 			dev_dbg(dev, "Temperature data invalid (0x%x)\n", eax);
148 		}
149 		data->last_updated = jiffies;
150 	}
151 
152 	mutex_unlock(&data->update_lock);
153 	return data;
154 }
155 
156 static int __devinit adjust_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
157 {
158 	/* The 100C is default for both mobile and non mobile CPUs */
159 
160 	int tjmax = 100000;
161 	int tjmax_ee = 85000;
162 	int usemsr_ee = 1;
163 	int err;
164 	u32 eax, edx;
165 	struct pci_dev *host_bridge;
166 
167 	/* Early chips have no MSR for TjMax */
168 
169 	if ((c->x86_model == 0xf) && (c->x86_mask < 4)) {
170 		usemsr_ee = 0;
171 	}
172 
173 	/* Atom CPUs */
174 
175 	if (c->x86_model == 0x1c) {
176 		usemsr_ee = 0;
177 
178 		host_bridge = pci_get_bus_and_slot(0, PCI_DEVFN(0, 0));
179 
180 		if (host_bridge && host_bridge->vendor == PCI_VENDOR_ID_INTEL
181 		    && (host_bridge->device == 0xa000	/* NM10 based nettop */
182 		    || host_bridge->device == 0xa010))	/* NM10 based netbook */
183 			tjmax = 100000;
184 		else
185 			tjmax = 90000;
186 
187 		pci_dev_put(host_bridge);
188 	}
189 
190 	if ((c->x86_model > 0xe) && (usemsr_ee)) {
191 		u8 platform_id;
192 
193 		/* Now we can detect the mobile CPU using Intel provided table
194 		   http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
195 		   For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
196 		*/
197 
198 		err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
199 		if (err) {
200 			dev_warn(dev,
201 				 "Unable to access MSR 0x17, assuming desktop"
202 				 " CPU\n");
203 			usemsr_ee = 0;
204 		} else if (c->x86_model < 0x17 && !(eax & 0x10000000)) {
205 			/* Trust bit 28 up to Penryn, I could not find any
206 			   documentation on that; if you happen to know
207 			   someone at Intel please ask */
208 			usemsr_ee = 0;
209 		} else {
210 			/* Platform ID bits 52:50 (EDX starts at bit 32) */
211 			platform_id = (edx >> 18) & 0x7;
212 
213 			/* Mobile Penryn CPU seems to be platform ID 7 or 5
214 			  (guesswork) */
215 			if ((c->x86_model == 0x17) &&
216 			    ((platform_id == 5) || (platform_id == 7))) {
217 				/* If MSR EE bit is set, set it to 90 degrees C,
218 				   otherwise 105 degrees C */
219 				tjmax_ee = 90000;
220 				tjmax = 105000;
221 			}
222 		}
223 	}
224 
225 	if (usemsr_ee) {
226 
227 		err = rdmsr_safe_on_cpu(id, 0xee, &eax, &edx);
228 		if (err) {
229 			dev_warn(dev,
230 				 "Unable to access MSR 0xEE, for Tjmax, left"
231 				 " at default");
232 		} else if (eax & 0x40000000) {
233 			tjmax = tjmax_ee;
234 		}
235 	/* if we dont use msr EE it means we are desktop CPU (with exeception
236 	   of Atom) */
237 	} else if (tjmax == 100000) {
238 		dev_warn(dev, "Using relative temperature scale!\n");
239 	}
240 
241 	return tjmax;
242 }
243 
244 static int __devinit coretemp_probe(struct platform_device *pdev)
245 {
246 	struct coretemp_data *data;
247 	struct cpuinfo_x86 *c = &cpu_data(pdev->id);
248 	int err;
249 	u32 eax, edx;
250 
251 	if (!(data = kzalloc(sizeof(struct coretemp_data), GFP_KERNEL))) {
252 		err = -ENOMEM;
253 		dev_err(&pdev->dev, "Out of memory\n");
254 		goto exit;
255 	}
256 
257 	data->id = pdev->id;
258 	data->name = "coretemp";
259 	mutex_init(&data->update_lock);
260 
261 	/* test if we can access the THERM_STATUS MSR */
262 	err = rdmsr_safe_on_cpu(data->id, MSR_IA32_THERM_STATUS, &eax, &edx);
263 	if (err) {
264 		dev_err(&pdev->dev,
265 			"Unable to access THERM_STATUS MSR, giving up\n");
266 		goto exit_free;
267 	}
268 
269 	/* Check if we have problem with errata AE18 of Core processors:
270 	   Readings might stop update when processor visited too deep sleep,
271 	   fixed for stepping D0 (6EC).
272 	*/
273 
274 	if ((c->x86_model == 0xe) && (c->x86_mask < 0xc)) {
275 		/* check for microcode update */
276 		rdmsr_on_cpu(data->id, MSR_IA32_UCODE_REV, &eax, &edx);
277 		if (edx < 0x39) {
278 			err = -ENODEV;
279 			dev_err(&pdev->dev,
280 				"Errata AE18 not fixed, update BIOS or "
281 				"microcode of the CPU!\n");
282 			goto exit_free;
283 		}
284 	}
285 
286 	data->tjmax = adjust_tjmax(c, data->id, &pdev->dev);
287 	platform_set_drvdata(pdev, data);
288 
289 	/* read the still undocumented IA32_TEMPERATURE_TARGET it exists
290 	   on older CPUs but not in this register, Atoms don't have it either */
291 
292 	if ((c->x86_model > 0xe) && (c->x86_model != 0x1c)) {
293 		err = rdmsr_safe_on_cpu(data->id, 0x1a2, &eax, &edx);
294 		if (err) {
295 			dev_warn(&pdev->dev, "Unable to read"
296 					" IA32_TEMPERATURE_TARGET MSR\n");
297 		} else {
298 			data->ttarget = data->tjmax -
299 					(((eax >> 8) & 0xff) * 1000);
300 			err = device_create_file(&pdev->dev,
301 					&sensor_dev_attr_temp1_max.dev_attr);
302 			if (err)
303 				goto exit_free;
304 		}
305 	}
306 
307 	if ((err = sysfs_create_group(&pdev->dev.kobj, &coretemp_group)))
308 		goto exit_dev;
309 
310 	data->hwmon_dev = hwmon_device_register(&pdev->dev);
311 	if (IS_ERR(data->hwmon_dev)) {
312 		err = PTR_ERR(data->hwmon_dev);
313 		dev_err(&pdev->dev, "Class registration failed (%d)\n",
314 			err);
315 		goto exit_class;
316 	}
317 
318 	return 0;
319 
320 exit_class:
321 	sysfs_remove_group(&pdev->dev.kobj, &coretemp_group);
322 exit_dev:
323 	device_remove_file(&pdev->dev, &sensor_dev_attr_temp1_max.dev_attr);
324 exit_free:
325 	kfree(data);
326 exit:
327 	return err;
328 }
329 
330 static int __devexit coretemp_remove(struct platform_device *pdev)
331 {
332 	struct coretemp_data *data = platform_get_drvdata(pdev);
333 
334 	hwmon_device_unregister(data->hwmon_dev);
335 	sysfs_remove_group(&pdev->dev.kobj, &coretemp_group);
336 	device_remove_file(&pdev->dev, &sensor_dev_attr_temp1_max.dev_attr);
337 	platform_set_drvdata(pdev, NULL);
338 	kfree(data);
339 	return 0;
340 }
341 
342 static struct platform_driver coretemp_driver = {
343 	.driver = {
344 		.owner = THIS_MODULE,
345 		.name = DRVNAME,
346 	},
347 	.probe = coretemp_probe,
348 	.remove = __devexit_p(coretemp_remove),
349 };
350 
351 struct pdev_entry {
352 	struct list_head list;
353 	struct platform_device *pdev;
354 	unsigned int cpu;
355 };
356 
357 static LIST_HEAD(pdev_list);
358 static DEFINE_MUTEX(pdev_list_mutex);
359 
360 static int __cpuinit coretemp_device_add(unsigned int cpu)
361 {
362 	int err;
363 	struct platform_device *pdev;
364 	struct pdev_entry *pdev_entry;
365 
366 	pdev = platform_device_alloc(DRVNAME, cpu);
367 	if (!pdev) {
368 		err = -ENOMEM;
369 		printk(KERN_ERR DRVNAME ": Device allocation failed\n");
370 		goto exit;
371 	}
372 
373 	pdev_entry = kzalloc(sizeof(struct pdev_entry), GFP_KERNEL);
374 	if (!pdev_entry) {
375 		err = -ENOMEM;
376 		goto exit_device_put;
377 	}
378 
379 	err = platform_device_add(pdev);
380 	if (err) {
381 		printk(KERN_ERR DRVNAME ": Device addition failed (%d)\n",
382 		       err);
383 		goto exit_device_free;
384 	}
385 
386 	pdev_entry->pdev = pdev;
387 	pdev_entry->cpu = cpu;
388 	mutex_lock(&pdev_list_mutex);
389 	list_add_tail(&pdev_entry->list, &pdev_list);
390 	mutex_unlock(&pdev_list_mutex);
391 
392 	return 0;
393 
394 exit_device_free:
395 	kfree(pdev_entry);
396 exit_device_put:
397 	platform_device_put(pdev);
398 exit:
399 	return err;
400 }
401 
402 #ifdef CONFIG_HOTPLUG_CPU
403 static void coretemp_device_remove(unsigned int cpu)
404 {
405 	struct pdev_entry *p, *n;
406 	mutex_lock(&pdev_list_mutex);
407 	list_for_each_entry_safe(p, n, &pdev_list, list) {
408 		if (p->cpu == cpu) {
409 			platform_device_unregister(p->pdev);
410 			list_del(&p->list);
411 			kfree(p);
412 		}
413 	}
414 	mutex_unlock(&pdev_list_mutex);
415 }
416 
417 static int __cpuinit coretemp_cpu_callback(struct notifier_block *nfb,
418 				 unsigned long action, void *hcpu)
419 {
420 	unsigned int cpu = (unsigned long) hcpu;
421 
422 	switch (action) {
423 	case CPU_ONLINE:
424 	case CPU_DOWN_FAILED:
425 		coretemp_device_add(cpu);
426 		break;
427 	case CPU_DOWN_PREPARE:
428 		coretemp_device_remove(cpu);
429 		break;
430 	}
431 	return NOTIFY_OK;
432 }
433 
434 static struct notifier_block coretemp_cpu_notifier __refdata = {
435 	.notifier_call = coretemp_cpu_callback,
436 };
437 #endif				/* !CONFIG_HOTPLUG_CPU */
438 
439 static int __init coretemp_init(void)
440 {
441 	int i, err = -ENODEV;
442 	struct pdev_entry *p, *n;
443 
444 	/* quick check if we run Intel */
445 	if (cpu_data(0).x86_vendor != X86_VENDOR_INTEL)
446 		goto exit;
447 
448 	err = platform_driver_register(&coretemp_driver);
449 	if (err)
450 		goto exit;
451 
452 	for_each_online_cpu(i) {
453 		struct cpuinfo_x86 *c = &cpu_data(i);
454 
455 		/* check if family 6, models 0xe (Pentium M DC),
456 		  0xf (Core 2 DC 65nm), 0x16 (Core 2 SC 65nm),
457 		  0x17 (Penryn 45nm), 0x1a (Nehalem), 0x1c (Atom),
458 		  0x1e (Lynnfield) */
459 		if ((c->cpuid_level < 0) || (c->x86 != 0x6) ||
460 		    !((c->x86_model == 0xe) || (c->x86_model == 0xf) ||
461 			(c->x86_model == 0x16) || (c->x86_model == 0x17) ||
462 			(c->x86_model == 0x1a) || (c->x86_model == 0x1c) ||
463 			(c->x86_model == 0x1e))) {
464 
465 			/* supported CPU not found, but report the unknown
466 			   family 6 CPU */
467 			if ((c->x86 == 0x6) && (c->x86_model > 0xf))
468 				printk(KERN_WARNING DRVNAME ": Unknown CPU "
469 					"model %x\n", c->x86_model);
470 			continue;
471 		}
472 
473 		err = coretemp_device_add(i);
474 		if (err)
475 			goto exit_devices_unreg;
476 	}
477 	if (list_empty(&pdev_list)) {
478 		err = -ENODEV;
479 		goto exit_driver_unreg;
480 	}
481 
482 #ifdef CONFIG_HOTPLUG_CPU
483 	register_hotcpu_notifier(&coretemp_cpu_notifier);
484 #endif
485 	return 0;
486 
487 exit_devices_unreg:
488 	mutex_lock(&pdev_list_mutex);
489 	list_for_each_entry_safe(p, n, &pdev_list, list) {
490 		platform_device_unregister(p->pdev);
491 		list_del(&p->list);
492 		kfree(p);
493 	}
494 	mutex_unlock(&pdev_list_mutex);
495 exit_driver_unreg:
496 	platform_driver_unregister(&coretemp_driver);
497 exit:
498 	return err;
499 }
500 
501 static void __exit coretemp_exit(void)
502 {
503 	struct pdev_entry *p, *n;
504 #ifdef CONFIG_HOTPLUG_CPU
505 	unregister_hotcpu_notifier(&coretemp_cpu_notifier);
506 #endif
507 	mutex_lock(&pdev_list_mutex);
508 	list_for_each_entry_safe(p, n, &pdev_list, list) {
509 		platform_device_unregister(p->pdev);
510 		list_del(&p->list);
511 		kfree(p);
512 	}
513 	mutex_unlock(&pdev_list_mutex);
514 	platform_driver_unregister(&coretemp_driver);
515 }
516 
517 MODULE_AUTHOR("Rudolf Marek <r.marek@assembler.cz>");
518 MODULE_DESCRIPTION("Intel Core temperature monitor");
519 MODULE_LICENSE("GPL");
520 
521 module_init(coretemp_init)
522 module_exit(coretemp_exit)
523