• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * coretemp.c - Linux kernel module for hardware monitoring
4  *
5  * Copyright (C) 2007 Rudolf Marek <r.marek@assembler.cz>
6  *
7  * Inspired from many hwmon drivers
8  */
9 
10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11 
12 #include <linux/module.h>
13 #include <linux/init.h>
14 #include <linux/slab.h>
15 #include <linux/jiffies.h>
16 #include <linux/hwmon.h>
17 #include <linux/sysfs.h>
18 #include <linux/hwmon-sysfs.h>
19 #include <linux/err.h>
20 #include <linux/mutex.h>
21 #include <linux/list.h>
22 #include <linux/platform_device.h>
23 #include <linux/cpu.h>
24 #include <linux/smp.h>
25 #include <linux/moduleparam.h>
26 #include <linux/pci.h>
27 #include <asm/msr.h>
28 #include <asm/processor.h>
29 #include <asm/cpu_device_id.h>
30 
31 #define DRVNAME	"coretemp"
32 
33 /*
34  * force_tjmax only matters when TjMax can't be read from the CPU itself.
35  * When set, it replaces the driver's suboptimal heuristic.
36  */
37 static int force_tjmax;
38 module_param_named(tjmax, force_tjmax, int, 0444);
39 MODULE_PARM_DESC(tjmax, "TjMax value in degrees Celsius");
40 
41 #define PKG_SYSFS_ATTR_NO	1	/* Sysfs attribute for package temp */
42 #define BASE_SYSFS_ATTR_NO	2	/* Sysfs Base attr no for coretemp */
43 #define NUM_REAL_CORES		512	/* Number of Real cores per cpu */
44 #define CORETEMP_NAME_LENGTH	28	/* String Length of attrs */
45 #define MAX_CORE_ATTRS		4	/* Maximum no of basic attrs */
46 #define TOTAL_ATTRS		(MAX_CORE_ATTRS + 1)
47 #define MAX_CORE_DATA		(NUM_REAL_CORES + BASE_SYSFS_ATTR_NO)
48 
49 #ifdef CONFIG_SMP
50 #define for_each_sibling(i, cpu) \
51 	for_each_cpu(i, topology_sibling_cpumask(cpu))
52 #else
53 #define for_each_sibling(i, cpu)	for (i = 0; false; )
54 #endif
55 
56 /*
57  * Per-Core Temperature Data
58  * @last_updated: The time when the current temperature value was updated
59  *		earlier (in jiffies).
60  * @cpu_core_id: The CPU Core from which temperature values should be read
61  *		This value is passed as "id" field to rdmsr/wrmsr functions.
62  * @status_reg: One of IA32_THERM_STATUS or IA32_PACKAGE_THERM_STATUS,
63  *		from where the temperature values should be read.
64  * @attr_size:  Total number of pre-core attrs displayed in the sysfs.
65  * @is_pkg_data: If this is 1, the temp_data holds pkgtemp data.
66  *		Otherwise, temp_data holds coretemp data.
67  * @valid: If this is 1, the current temperature is valid.
68  */
69 struct temp_data {
70 	int temp;
71 	int ttarget;
72 	int tjmax;
73 	unsigned long last_updated;
74 	unsigned int cpu;
75 	u32 cpu_core_id;
76 	u32 status_reg;
77 	int attr_size;
78 	bool is_pkg_data;
79 	bool valid;
80 	struct sensor_device_attribute sd_attrs[TOTAL_ATTRS];
81 	char attr_name[TOTAL_ATTRS][CORETEMP_NAME_LENGTH];
82 	struct attribute *attrs[TOTAL_ATTRS + 1];
83 	struct attribute_group attr_group;
84 	struct mutex update_lock;
85 };
86 
87 /* Platform Data per Physical CPU */
88 struct platform_data {
89 	struct device		*hwmon_dev;
90 	u16			pkg_id;
91 	u16			cpu_map[NUM_REAL_CORES];
92 	struct ida		ida;
93 	struct cpumask		cpumask;
94 	struct temp_data	*core_data[MAX_CORE_DATA];
95 	struct device_attribute name_attr;
96 };
97 
98 /* Keep track of how many zone pointers we allocated in init() */
99 static int max_zones __read_mostly;
100 /* Array of zone pointers. Serialized by cpu hotplug lock */
101 static struct platform_device **zone_devices;
102 
show_label(struct device * dev,struct device_attribute * devattr,char * buf)103 static ssize_t show_label(struct device *dev,
104 				struct device_attribute *devattr, char *buf)
105 {
106 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
107 	struct platform_data *pdata = dev_get_drvdata(dev);
108 	struct temp_data *tdata = pdata->core_data[attr->index];
109 
110 	if (tdata->is_pkg_data)
111 		return sprintf(buf, "Package id %u\n", pdata->pkg_id);
112 
113 	return sprintf(buf, "Core %u\n", tdata->cpu_core_id);
114 }
115 
show_crit_alarm(struct device * dev,struct device_attribute * devattr,char * buf)116 static ssize_t show_crit_alarm(struct device *dev,
117 				struct device_attribute *devattr, char *buf)
118 {
119 	u32 eax, edx;
120 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
121 	struct platform_data *pdata = dev_get_drvdata(dev);
122 	struct temp_data *tdata = pdata->core_data[attr->index];
123 
124 	mutex_lock(&tdata->update_lock);
125 	rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
126 	mutex_unlock(&tdata->update_lock);
127 
128 	return sprintf(buf, "%d\n", (eax >> 5) & 1);
129 }
130 
show_tjmax(struct device * dev,struct device_attribute * devattr,char * buf)131 static ssize_t show_tjmax(struct device *dev,
132 			struct device_attribute *devattr, char *buf)
133 {
134 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
135 	struct platform_data *pdata = dev_get_drvdata(dev);
136 
137 	return sprintf(buf, "%d\n", pdata->core_data[attr->index]->tjmax);
138 }
139 
show_ttarget(struct device * dev,struct device_attribute * devattr,char * buf)140 static ssize_t show_ttarget(struct device *dev,
141 				struct device_attribute *devattr, char *buf)
142 {
143 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
144 	struct platform_data *pdata = dev_get_drvdata(dev);
145 
146 	return sprintf(buf, "%d\n", pdata->core_data[attr->index]->ttarget);
147 }
148 
show_temp(struct device * dev,struct device_attribute * devattr,char * buf)149 static ssize_t show_temp(struct device *dev,
150 			struct device_attribute *devattr, char *buf)
151 {
152 	u32 eax, edx;
153 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
154 	struct platform_data *pdata = dev_get_drvdata(dev);
155 	struct temp_data *tdata = pdata->core_data[attr->index];
156 
157 	mutex_lock(&tdata->update_lock);
158 
159 	/* Check whether the time interval has elapsed */
160 	if (!tdata->valid || time_after(jiffies, tdata->last_updated + HZ)) {
161 		rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
162 		/*
163 		 * Ignore the valid bit. In all observed cases the register
164 		 * value is either low or zero if the valid bit is 0.
165 		 * Return it instead of reporting an error which doesn't
166 		 * really help at all.
167 		 */
168 		tdata->temp = tdata->tjmax - ((eax >> 16) & 0x7f) * 1000;
169 		tdata->valid = 1;
170 		tdata->last_updated = jiffies;
171 	}
172 
173 	mutex_unlock(&tdata->update_lock);
174 	return sprintf(buf, "%d\n", tdata->temp);
175 }
176 
177 struct tjmax_pci {
178 	unsigned int device;
179 	int tjmax;
180 };
181 
182 static const struct tjmax_pci tjmax_pci_table[] = {
183 	{ 0x0708, 110000 },	/* CE41x0 (Sodaville ) */
184 	{ 0x0c72, 102000 },	/* Atom S1240 (Centerton) */
185 	{ 0x0c73, 95000 },	/* Atom S1220 (Centerton) */
186 	{ 0x0c75, 95000 },	/* Atom S1260 (Centerton) */
187 };
188 
189 struct tjmax {
190 	char const *id;
191 	int tjmax;
192 };
193 
194 static const struct tjmax tjmax_table[] = {
195 	{ "CPU  230", 100000 },		/* Model 0x1c, stepping 2	*/
196 	{ "CPU  330", 125000 },		/* Model 0x1c, stepping 2	*/
197 };
198 
199 struct tjmax_model {
200 	u8 model;
201 	u8 mask;
202 	int tjmax;
203 };
204 
205 #define ANY 0xff
206 
207 static const struct tjmax_model tjmax_model_table[] = {
208 	{ 0x1c, 10, 100000 },	/* D4xx, K4xx, N4xx, D5xx, K5xx, N5xx */
209 	{ 0x1c, ANY, 90000 },	/* Z5xx, N2xx, possibly others
210 				 * Note: Also matches 230 and 330,
211 				 * which are covered by tjmax_table
212 				 */
213 	{ 0x26, ANY, 90000 },	/* Atom Tunnel Creek (Exx), Lincroft (Z6xx)
214 				 * Note: TjMax for E6xxT is 110C, but CPU type
215 				 * is undetectable by software
216 				 */
217 	{ 0x27, ANY, 90000 },	/* Atom Medfield (Z2460) */
218 	{ 0x35, ANY, 90000 },	/* Atom Clover Trail/Cloverview (Z27x0) */
219 	{ 0x36, ANY, 100000 },	/* Atom Cedar Trail/Cedarview (N2xxx, D2xxx)
220 				 * Also matches S12x0 (stepping 9), covered by
221 				 * PCI table
222 				 */
223 };
224 
adjust_tjmax(struct cpuinfo_x86 * c,u32 id,struct device * dev)225 static int adjust_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
226 {
227 	/* The 100C is default for both mobile and non mobile CPUs */
228 
229 	int tjmax = 100000;
230 	int tjmax_ee = 85000;
231 	int usemsr_ee = 1;
232 	int err;
233 	u32 eax, edx;
234 	int i;
235 	u16 devfn = PCI_DEVFN(0, 0);
236 	struct pci_dev *host_bridge = pci_get_domain_bus_and_slot(0, 0, devfn);
237 
238 	/*
239 	 * Explicit tjmax table entries override heuristics.
240 	 * First try PCI host bridge IDs, followed by model ID strings
241 	 * and model/stepping information.
242 	 */
243 	if (host_bridge && host_bridge->vendor == PCI_VENDOR_ID_INTEL) {
244 		for (i = 0; i < ARRAY_SIZE(tjmax_pci_table); i++) {
245 			if (host_bridge->device == tjmax_pci_table[i].device) {
246 				pci_dev_put(host_bridge);
247 				return tjmax_pci_table[i].tjmax;
248 			}
249 		}
250 	}
251 	pci_dev_put(host_bridge);
252 
253 	for (i = 0; i < ARRAY_SIZE(tjmax_table); i++) {
254 		if (strstr(c->x86_model_id, tjmax_table[i].id))
255 			return tjmax_table[i].tjmax;
256 	}
257 
258 	for (i = 0; i < ARRAY_SIZE(tjmax_model_table); i++) {
259 		const struct tjmax_model *tm = &tjmax_model_table[i];
260 		if (c->x86_model == tm->model &&
261 		    (tm->mask == ANY || c->x86_stepping == tm->mask))
262 			return tm->tjmax;
263 	}
264 
265 	/* Early chips have no MSR for TjMax */
266 
267 	if (c->x86_model == 0xf && c->x86_stepping < 4)
268 		usemsr_ee = 0;
269 
270 	if (c->x86_model > 0xe && usemsr_ee) {
271 		u8 platform_id;
272 
273 		/*
274 		 * Now we can detect the mobile CPU using Intel provided table
275 		 * http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
276 		 * For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
277 		 */
278 		err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
279 		if (err) {
280 			dev_warn(dev,
281 				 "Unable to access MSR 0x17, assuming desktop"
282 				 " CPU\n");
283 			usemsr_ee = 0;
284 		} else if (c->x86_model < 0x17 && !(eax & 0x10000000)) {
285 			/*
286 			 * Trust bit 28 up to Penryn, I could not find any
287 			 * documentation on that; if you happen to know
288 			 * someone at Intel please ask
289 			 */
290 			usemsr_ee = 0;
291 		} else {
292 			/* Platform ID bits 52:50 (EDX starts at bit 32) */
293 			platform_id = (edx >> 18) & 0x7;
294 
295 			/*
296 			 * Mobile Penryn CPU seems to be platform ID 7 or 5
297 			 * (guesswork)
298 			 */
299 			if (c->x86_model == 0x17 &&
300 			    (platform_id == 5 || platform_id == 7)) {
301 				/*
302 				 * If MSR EE bit is set, set it to 90 degrees C,
303 				 * otherwise 105 degrees C
304 				 */
305 				tjmax_ee = 90000;
306 				tjmax = 105000;
307 			}
308 		}
309 	}
310 
311 	if (usemsr_ee) {
312 		err = rdmsr_safe_on_cpu(id, 0xee, &eax, &edx);
313 		if (err) {
314 			dev_warn(dev,
315 				 "Unable to access MSR 0xEE, for Tjmax, left"
316 				 " at default\n");
317 		} else if (eax & 0x40000000) {
318 			tjmax = tjmax_ee;
319 		}
320 	} else if (tjmax == 100000) {
321 		/*
322 		 * If we don't use msr EE it means we are desktop CPU
323 		 * (with exeception of Atom)
324 		 */
325 		dev_warn(dev, "Using relative temperature scale!\n");
326 	}
327 
328 	return tjmax;
329 }
330 
cpu_has_tjmax(struct cpuinfo_x86 * c)331 static bool cpu_has_tjmax(struct cpuinfo_x86 *c)
332 {
333 	u8 model = c->x86_model;
334 
335 	return model > 0xe &&
336 	       model != 0x1c &&
337 	       model != 0x26 &&
338 	       model != 0x27 &&
339 	       model != 0x35 &&
340 	       model != 0x36;
341 }
342 
get_tjmax(struct cpuinfo_x86 * c,u32 id,struct device * dev)343 static int get_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *dev)
344 {
345 	int err;
346 	u32 eax, edx;
347 	u32 val;
348 
349 	/*
350 	 * A new feature of current Intel(R) processors, the
351 	 * IA32_TEMPERATURE_TARGET contains the TjMax value
352 	 */
353 	err = rdmsr_safe_on_cpu(id, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
354 	if (err) {
355 		if (cpu_has_tjmax(c))
356 			dev_warn(dev, "Unable to read TjMax from CPU %u\n", id);
357 	} else {
358 		val = (eax >> 16) & 0xff;
359 		/*
360 		 * If the TjMax is not plausible, an assumption
361 		 * will be used
362 		 */
363 		if (val) {
364 			dev_dbg(dev, "TjMax is %d degrees C\n", val);
365 			return val * 1000;
366 		}
367 	}
368 
369 	if (force_tjmax) {
370 		dev_notice(dev, "TjMax forced to %d degrees C by user\n",
371 			   force_tjmax);
372 		return force_tjmax * 1000;
373 	}
374 
375 	/*
376 	 * An assumption is made for early CPUs and unreadable MSR.
377 	 * NOTE: the calculated value may not be correct.
378 	 */
379 	return adjust_tjmax(c, id, dev);
380 }
381 
create_core_attrs(struct temp_data * tdata,struct device * dev,int index)382 static int create_core_attrs(struct temp_data *tdata, struct device *dev,
383 			     int index)
384 {
385 	int i;
386 	static ssize_t (*const rd_ptr[TOTAL_ATTRS]) (struct device *dev,
387 			struct device_attribute *devattr, char *buf) = {
388 			show_label, show_crit_alarm, show_temp, show_tjmax,
389 			show_ttarget };
390 	static const char *const suffixes[TOTAL_ATTRS] = {
391 		"label", "crit_alarm", "input", "crit", "max"
392 	};
393 
394 	for (i = 0; i < tdata->attr_size; i++) {
395 		/*
396 		 * We map the attr number to core id of the CPU
397 		 * The attr number is always core id + 2
398 		 * The Pkgtemp will always show up as temp1_*, if available
399 		 */
400 		int attr_no = tdata->is_pkg_data ? 1 : tdata->cpu_core_id + 2;
401 
402 		snprintf(tdata->attr_name[i], CORETEMP_NAME_LENGTH,
403 			 "temp%d_%s", attr_no, suffixes[i]);
404 		sysfs_attr_init(&tdata->sd_attrs[i].dev_attr.attr);
405 		tdata->sd_attrs[i].dev_attr.attr.name = tdata->attr_name[i];
406 		tdata->sd_attrs[i].dev_attr.attr.mode = 0444;
407 		tdata->sd_attrs[i].dev_attr.show = rd_ptr[i];
408 		tdata->sd_attrs[i].index = index;
409 		tdata->attrs[i] = &tdata->sd_attrs[i].dev_attr.attr;
410 	}
411 	tdata->attr_group.attrs = tdata->attrs;
412 	return sysfs_create_group(&dev->kobj, &tdata->attr_group);
413 }
414 
415 
chk_ucode_version(unsigned int cpu)416 static int chk_ucode_version(unsigned int cpu)
417 {
418 	struct cpuinfo_x86 *c = &cpu_data(cpu);
419 
420 	/*
421 	 * Check if we have problem with errata AE18 of Core processors:
422 	 * Readings might stop update when processor visited too deep sleep,
423 	 * fixed for stepping D0 (6EC).
424 	 */
425 	if (c->x86_model == 0xe && c->x86_stepping < 0xc && c->microcode < 0x39) {
426 		pr_err("Errata AE18 not fixed, update BIOS or microcode of the CPU!\n");
427 		return -ENODEV;
428 	}
429 	return 0;
430 }
431 
coretemp_get_pdev(unsigned int cpu)432 static struct platform_device *coretemp_get_pdev(unsigned int cpu)
433 {
434 	int id = topology_logical_die_id(cpu);
435 
436 	if (id >= 0 && id < max_zones)
437 		return zone_devices[id];
438 	return NULL;
439 }
440 
init_temp_data(unsigned int cpu,int pkg_flag)441 static struct temp_data *init_temp_data(unsigned int cpu, int pkg_flag)
442 {
443 	struct temp_data *tdata;
444 
445 	tdata = kzalloc(sizeof(struct temp_data), GFP_KERNEL);
446 	if (!tdata)
447 		return NULL;
448 
449 	tdata->status_reg = pkg_flag ? MSR_IA32_PACKAGE_THERM_STATUS :
450 							MSR_IA32_THERM_STATUS;
451 	tdata->is_pkg_data = pkg_flag;
452 	tdata->cpu = cpu;
453 	tdata->cpu_core_id = topology_core_id(cpu);
454 	tdata->attr_size = MAX_CORE_ATTRS;
455 	mutex_init(&tdata->update_lock);
456 	return tdata;
457 }
458 
create_core_data(struct platform_device * pdev,unsigned int cpu,int pkg_flag)459 static int create_core_data(struct platform_device *pdev, unsigned int cpu,
460 			    int pkg_flag)
461 {
462 	struct temp_data *tdata;
463 	struct platform_data *pdata = platform_get_drvdata(pdev);
464 	struct cpuinfo_x86 *c = &cpu_data(cpu);
465 	u32 eax, edx;
466 	int err, index;
467 
468 	/*
469 	 * Get the index of tdata in pdata->core_data[]
470 	 * tdata for package: pdata->core_data[1]
471 	 * tdata for core: pdata->core_data[2] .. pdata->core_data[NUM_REAL_CORES + 1]
472 	 */
473 	if (pkg_flag) {
474 		index = PKG_SYSFS_ATTR_NO;
475 	} else {
476 		index = ida_alloc_max(&pdata->ida, NUM_REAL_CORES - 1, GFP_KERNEL);
477 		if (index < 0)
478 			return index;
479 
480 		pdata->cpu_map[index] = topology_core_id(cpu);
481 		index += BASE_SYSFS_ATTR_NO;
482 	}
483 
484 	tdata = init_temp_data(cpu, pkg_flag);
485 	if (!tdata) {
486 		err = -ENOMEM;
487 		goto ida_free;
488 	}
489 
490 	/* Test if we can access the status register */
491 	err = rdmsr_safe_on_cpu(cpu, tdata->status_reg, &eax, &edx);
492 	if (err)
493 		goto exit_free;
494 
495 	/* We can access status register. Get Critical Temperature */
496 	tdata->tjmax = get_tjmax(c, cpu, &pdev->dev);
497 
498 	/*
499 	 * Read the still undocumented bits 8:15 of IA32_TEMPERATURE_TARGET.
500 	 * The target temperature is available on older CPUs but not in this
501 	 * register. Atoms don't have the register at all.
502 	 */
503 	if (c->x86_model > 0xe && c->x86_model != 0x1c) {
504 		err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET,
505 					&eax, &edx);
506 		if (!err) {
507 			tdata->ttarget
508 			  = tdata->tjmax - ((eax >> 8) & 0xff) * 1000;
509 			tdata->attr_size++;
510 		}
511 	}
512 
513 	pdata->core_data[index] = tdata;
514 
515 	/* Create sysfs interfaces */
516 	err = create_core_attrs(tdata, pdata->hwmon_dev, index);
517 	if (err)
518 		goto exit_free;
519 
520 	return 0;
521 exit_free:
522 	pdata->core_data[index] = NULL;
523 	kfree(tdata);
524 ida_free:
525 	if (!pkg_flag)
526 		ida_free(&pdata->ida, index - BASE_SYSFS_ATTR_NO);
527 	return err;
528 }
529 
530 static void
coretemp_add_core(struct platform_device * pdev,unsigned int cpu,int pkg_flag)531 coretemp_add_core(struct platform_device *pdev, unsigned int cpu, int pkg_flag)
532 {
533 	if (create_core_data(pdev, cpu, pkg_flag))
534 		dev_err(&pdev->dev, "Adding Core %u failed\n", cpu);
535 }
536 
coretemp_remove_core(struct platform_data * pdata,int indx)537 static void coretemp_remove_core(struct platform_data *pdata, int indx)
538 {
539 	struct temp_data *tdata = pdata->core_data[indx];
540 
541 	/* if we errored on add then this is already gone */
542 	if (!tdata)
543 		return;
544 
545 	/* Remove the sysfs attributes */
546 	sysfs_remove_group(&pdata->hwmon_dev->kobj, &tdata->attr_group);
547 
548 	kfree(pdata->core_data[indx]);
549 	pdata->core_data[indx] = NULL;
550 
551 	if (indx >= BASE_SYSFS_ATTR_NO)
552 		ida_free(&pdata->ida, indx - BASE_SYSFS_ATTR_NO);
553 }
554 
coretemp_device_add(int zoneid)555 static int coretemp_device_add(int zoneid)
556 {
557 	struct platform_device *pdev;
558 	struct platform_data *pdata;
559 	int err;
560 
561 	/* Initialize the per-zone data structures */
562 	pdata = kzalloc(sizeof(*pdata), GFP_KERNEL);
563 	if (!pdata)
564 		return -ENOMEM;
565 
566 	pdata->pkg_id = zoneid;
567 	ida_init(&pdata->ida);
568 
569 	pdev = platform_device_alloc(DRVNAME, zoneid);
570 	if (!pdev) {
571 		err = -ENOMEM;
572 		goto err_free_pdata;
573 	}
574 
575 	err = platform_device_add(pdev);
576 	if (err)
577 		goto err_put_dev;
578 
579 	platform_set_drvdata(pdev, pdata);
580 	zone_devices[zoneid] = pdev;
581 	return 0;
582 
583 err_put_dev:
584 	platform_device_put(pdev);
585 err_free_pdata:
586 	kfree(pdata);
587 	return err;
588 }
589 
coretemp_device_remove(int zoneid)590 static void coretemp_device_remove(int zoneid)
591 {
592 	struct platform_device *pdev = zone_devices[zoneid];
593 	struct platform_data *pdata = platform_get_drvdata(pdev);
594 
595 	ida_destroy(&pdata->ida);
596 	kfree(pdata);
597 	platform_device_unregister(pdev);
598 }
599 
coretemp_cpu_online(unsigned int cpu)600 static int coretemp_cpu_online(unsigned int cpu)
601 {
602 	struct platform_device *pdev = coretemp_get_pdev(cpu);
603 	struct cpuinfo_x86 *c = &cpu_data(cpu);
604 	struct platform_data *pdata;
605 
606 	/*
607 	 * Don't execute this on resume as the offline callback did
608 	 * not get executed on suspend.
609 	 */
610 	if (cpuhp_tasks_frozen)
611 		return 0;
612 
613 	/*
614 	 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
615 	 * sensors. We check this bit only, all the early CPUs
616 	 * without thermal sensors will be filtered out.
617 	 */
618 	if (!cpu_has(c, X86_FEATURE_DTHERM))
619 		return -ENODEV;
620 
621 	pdata = platform_get_drvdata(pdev);
622 	if (!pdata->hwmon_dev) {
623 		struct device *hwmon;
624 
625 		/* Check the microcode version of the CPU */
626 		if (chk_ucode_version(cpu))
627 			return -EINVAL;
628 
629 		/*
630 		 * Alright, we have DTS support.
631 		 * We are bringing the _first_ core in this pkg
632 		 * online. So, initialize per-pkg data structures and
633 		 * then bring this core online.
634 		 */
635 		hwmon = hwmon_device_register_with_groups(&pdev->dev, DRVNAME,
636 							  pdata, NULL);
637 		if (IS_ERR(hwmon))
638 			return PTR_ERR(hwmon);
639 		pdata->hwmon_dev = hwmon;
640 
641 		/*
642 		 * Check whether pkgtemp support is available.
643 		 * If so, add interfaces for pkgtemp.
644 		 */
645 		if (cpu_has(c, X86_FEATURE_PTS))
646 			coretemp_add_core(pdev, cpu, 1);
647 	}
648 
649 	/*
650 	 * Check whether a thread sibling is already online. If not add the
651 	 * interface for this CPU core.
652 	 */
653 	if (!cpumask_intersects(&pdata->cpumask, topology_sibling_cpumask(cpu)))
654 		coretemp_add_core(pdev, cpu, 0);
655 
656 	cpumask_set_cpu(cpu, &pdata->cpumask);
657 	return 0;
658 }
659 
coretemp_cpu_offline(unsigned int cpu)660 static int coretemp_cpu_offline(unsigned int cpu)
661 {
662 	struct platform_device *pdev = coretemp_get_pdev(cpu);
663 	struct platform_data *pd;
664 	struct temp_data *tdata;
665 	int i, indx = -1, target;
666 
667 	/* No need to tear down any interfaces for suspend */
668 	if (cpuhp_tasks_frozen)
669 		return 0;
670 
671 	/* If the physical CPU device does not exist, just return */
672 	pd = platform_get_drvdata(pdev);
673 	if (!pd->hwmon_dev)
674 		return 0;
675 
676 	for (i = 0; i < NUM_REAL_CORES; i++) {
677 		if (pd->cpu_map[i] == topology_core_id(cpu)) {
678 			indx = i + BASE_SYSFS_ATTR_NO;
679 			break;
680 		}
681 	}
682 
683 	/* Too many cores and this core is not populated, just return */
684 	if (indx < 0)
685 		return 0;
686 
687 	tdata = pd->core_data[indx];
688 
689 	cpumask_clear_cpu(cpu, &pd->cpumask);
690 
691 	/*
692 	 * If this is the last thread sibling, remove the CPU core
693 	 * interface, If there is still a sibling online, transfer the
694 	 * target cpu of that core interface to it.
695 	 */
696 	target = cpumask_any_and(&pd->cpumask, topology_sibling_cpumask(cpu));
697 	if (target >= nr_cpu_ids) {
698 		coretemp_remove_core(pd, indx);
699 	} else if (tdata && tdata->cpu == cpu) {
700 		mutex_lock(&tdata->update_lock);
701 		tdata->cpu = target;
702 		mutex_unlock(&tdata->update_lock);
703 	}
704 
705 	/*
706 	 * If all cores in this pkg are offline, remove the interface.
707 	 */
708 	tdata = pd->core_data[PKG_SYSFS_ATTR_NO];
709 	if (cpumask_empty(&pd->cpumask)) {
710 		if (tdata)
711 			coretemp_remove_core(pd, PKG_SYSFS_ATTR_NO);
712 		hwmon_device_unregister(pd->hwmon_dev);
713 		pd->hwmon_dev = NULL;
714 		return 0;
715 	}
716 
717 	/*
718 	 * Check whether this core is the target for the package
719 	 * interface. We need to assign it to some other cpu.
720 	 */
721 	if (tdata && tdata->cpu == cpu) {
722 		target = cpumask_first(&pd->cpumask);
723 		mutex_lock(&tdata->update_lock);
724 		tdata->cpu = target;
725 		mutex_unlock(&tdata->update_lock);
726 	}
727 	return 0;
728 }
729 static const struct x86_cpu_id __initconst coretemp_ids[] = {
730 	X86_MATCH_VENDOR_FEATURE(INTEL, X86_FEATURE_DTHERM, NULL),
731 	{}
732 };
733 MODULE_DEVICE_TABLE(x86cpu, coretemp_ids);
734 
735 static enum cpuhp_state coretemp_hp_online;
736 
coretemp_init(void)737 static int __init coretemp_init(void)
738 {
739 	int i, err;
740 
741 	/*
742 	 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
743 	 * sensors. We check this bit only, all the early CPUs
744 	 * without thermal sensors will be filtered out.
745 	 */
746 	if (!x86_match_cpu(coretemp_ids))
747 		return -ENODEV;
748 
749 	max_zones = topology_max_packages() * topology_max_die_per_package();
750 	zone_devices = kcalloc(max_zones, sizeof(struct platform_device *),
751 			      GFP_KERNEL);
752 	if (!zone_devices)
753 		return -ENOMEM;
754 
755 	for (i = 0; i < max_zones; i++) {
756 		err = coretemp_device_add(i);
757 		if (err)
758 			goto outzone;
759 	}
760 
761 	err = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "hwmon/coretemp:online",
762 				coretemp_cpu_online, coretemp_cpu_offline);
763 	if (err < 0)
764 		goto outzone;
765 	coretemp_hp_online = err;
766 	return 0;
767 
768 outzone:
769 	while (i--)
770 		coretemp_device_remove(i);
771 	kfree(zone_devices);
772 	return err;
773 }
module_init(coretemp_init)774 module_init(coretemp_init)
775 
776 static void __exit coretemp_exit(void)
777 {
778 	int i;
779 
780 	cpuhp_remove_state(coretemp_hp_online);
781 	for (i = 0; i < max_zones; i++)
782 		coretemp_device_remove(i);
783 	kfree(zone_devices);
784 }
785 module_exit(coretemp_exit)
786 
787 MODULE_AUTHOR("Rudolf Marek <r.marek@assembler.cz>");
788 MODULE_DESCRIPTION("Intel Core temperature monitor");
789 MODULE_LICENSE("GPL");
790