x86/hwmon: avoid deadlock on CPU removal in pkgtemp
[deliverable/linux.git] / drivers / hwmon / pkgtemp.c
CommitLineData
cb84b194
FY
1/*
2 * pkgtemp.c - Linux kernel module for processor package hardware monitoring
3 *
4 * Copyright (C) 2010 Fenghua Yu <fenghua.yu@intel.com>
5 *
6 * Inspired from many hwmon drivers especially coretemp.
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 "pkgtemp"
41
42enum { SHOW_TEMP, SHOW_TJMAX, SHOW_TTARGET, SHOW_LABEL, SHOW_NAME };
43
44/*
45 * Functions declaration
46 */
47
48static struct pkgtemp_data *pkgtemp_update_device(struct device *dev);
49
50struct pkgtemp_data {
51 struct device *hwmon_dev;
52 struct mutex update_lock;
53 const char *name;
54 u32 id;
55 u16 phys_proc_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
68static 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 pkgtemp_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, "physical id %d\n",
79 data->phys_proc_id);
80 return ret;
81}
82
83static ssize_t show_alarm(struct device *dev, struct device_attribute
84 *devattr, char *buf)
85{
86 struct pkgtemp_data *data = pkgtemp_update_device(dev);
87 /* read the Out-of-spec log, never clear */
88 return sprintf(buf, "%d\n", data->alarm);
89}
90
91static ssize_t show_temp(struct device *dev,
92 struct device_attribute *devattr, char *buf)
93{
94 struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
95 struct pkgtemp_data *data = pkgtemp_update_device(dev);
96 int err = 0;
97
98 if (attr->index == SHOW_TEMP)
99 err = data->valid ? sprintf(buf, "%d\n", data->temp) : -EAGAIN;
100 else if (attr->index == SHOW_TJMAX)
101 err = sprintf(buf, "%d\n", data->tjmax);
102 else
103 err = sprintf(buf, "%d\n", data->ttarget);
104 return err;
105}
106
107static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, show_temp, NULL, SHOW_TEMP);
108static SENSOR_DEVICE_ATTR(temp1_crit, S_IRUGO, show_temp, NULL, SHOW_TJMAX);
109static SENSOR_DEVICE_ATTR(temp1_max, S_IRUGO, show_temp, NULL, SHOW_TTARGET);
110static DEVICE_ATTR(temp1_crit_alarm, S_IRUGO, show_alarm, NULL);
111static SENSOR_DEVICE_ATTR(temp1_label, S_IRUGO, show_name, NULL, SHOW_LABEL);
112static SENSOR_DEVICE_ATTR(name, S_IRUGO, show_name, NULL, SHOW_NAME);
113
114static struct attribute *pkgtemp_attributes[] = {
115 &sensor_dev_attr_name.dev_attr.attr,
116 &sensor_dev_attr_temp1_label.dev_attr.attr,
117 &dev_attr_temp1_crit_alarm.attr,
118 &sensor_dev_attr_temp1_input.dev_attr.attr,
119 &sensor_dev_attr_temp1_crit.dev_attr.attr,
120 NULL
121};
122
123static const struct attribute_group pkgtemp_group = {
124 .attrs = pkgtemp_attributes,
125};
126
127static struct pkgtemp_data *pkgtemp_update_device(struct device *dev)
128{
129 struct pkgtemp_data *data = dev_get_drvdata(dev);
130 unsigned int cpu;
131 int err;
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 cpu = data->id;
140 err = rdmsr_on_cpu(cpu, MSR_IA32_PACKAGE_THERM_STATUS,
141 &eax, &edx);
142 if (!err) {
143 data->alarm = (eax >> 5) & 1;
144 data->temp = data->tjmax - (((eax >> 16)
145 & 0x7f) * 1000);
146 data->valid = 1;
147 } else
148 dev_dbg(dev, "Temperature data invalid (0x%x)\n", eax);
149
150 data->last_updated = jiffies;
151 }
152
153 mutex_unlock(&data->update_lock);
154 return data;
155}
156
157static int get_tjmax(int cpu, struct device *dev)
158{
159 int default_tjmax = 100000;
160 int err;
161 u32 eax, edx;
162 u32 val;
163
164 /* IA32_TEMPERATURE_TARGET contains the TjMax value */
165 err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
166 if (!err) {
167 val = (eax >> 16) & 0xff;
168 if ((val > 80) && (val < 120)) {
169 dev_info(dev, "TjMax is %d C.\n", val);
170 return val * 1000;
171 }
172 }
173 dev_warn(dev, "Unable to read TjMax from CPU.\n");
174 return default_tjmax;
175}
176
177static int __devinit pkgtemp_probe(struct platform_device *pdev)
178{
179 struct pkgtemp_data *data;
180 int err;
181 u32 eax, edx;
182#ifdef CONFIG_SMP
183 struct cpuinfo_x86 *c = &cpu_data(pdev->id);
184#endif
185
186 data = kzalloc(sizeof(struct pkgtemp_data), GFP_KERNEL);
187 if (!data) {
188 err = -ENOMEM;
189 dev_err(&pdev->dev, "Out of memory\n");
190 goto exit;
191 }
192
193 data->id = pdev->id;
194#ifdef CONFIG_SMP
195 data->phys_proc_id = c->phys_proc_id;
196#endif
197 data->name = "pkgtemp";
198 mutex_init(&data->update_lock);
199
200 /* test if we can access the THERM_STATUS MSR */
201 err = rdmsr_safe_on_cpu(data->id, MSR_IA32_PACKAGE_THERM_STATUS,
202 &eax, &edx);
203 if (err) {
204 dev_err(&pdev->dev,
205 "Unable to access THERM_STATUS MSR, giving up\n");
206 goto exit_free;
207 }
208
209 data->tjmax = get_tjmax(data->id, &pdev->dev);
210 platform_set_drvdata(pdev, data);
211
212 err = rdmsr_safe_on_cpu(data->id, MSR_IA32_TEMPERATURE_TARGET,
213 &eax, &edx);
214 if (err) {
215 dev_warn(&pdev->dev, "Unable to read"
216 " IA32_TEMPERATURE_TARGET MSR\n");
217 } else {
218 data->ttarget = data->tjmax - (((eax >> 8) & 0xff) * 1000);
219 err = device_create_file(&pdev->dev,
220 &sensor_dev_attr_temp1_max.dev_attr);
221 if (err)
222 goto exit_free;
223 }
224
225 err = sysfs_create_group(&pdev->dev.kobj, &pkgtemp_group);
226 if (err)
227 goto exit_free;
228
229 data->hwmon_dev = hwmon_device_register(&pdev->dev);
230 if (IS_ERR(data->hwmon_dev)) {
231 err = PTR_ERR(data->hwmon_dev);
232 dev_err(&pdev->dev, "Class registration failed (%d)\n",
233 err);
234 goto exit_class;
235 }
236
237 return 0;
238
239exit_class:
240 sysfs_remove_group(&pdev->dev.kobj, &pkgtemp_group);
241exit_free:
242 kfree(data);
243exit:
244 return err;
245}
246
247static int __devexit pkgtemp_remove(struct platform_device *pdev)
248{
249 struct pkgtemp_data *data = platform_get_drvdata(pdev);
250
251 hwmon_device_unregister(data->hwmon_dev);
252 sysfs_remove_group(&pdev->dev.kobj, &pkgtemp_group);
253 platform_set_drvdata(pdev, NULL);
254 kfree(data);
255 return 0;
256}
257
258static struct platform_driver pkgtemp_driver = {
259 .driver = {
260 .owner = THIS_MODULE,
261 .name = DRVNAME,
262 },
263 .probe = pkgtemp_probe,
264 .remove = __devexit_p(pkgtemp_remove),
265};
266
267struct pdev_entry {
268 struct list_head list;
269 struct platform_device *pdev;
270 unsigned int cpu;
271#ifdef CONFIG_SMP
272 u16 phys_proc_id;
273#endif
274};
275
276static LIST_HEAD(pdev_list);
277static DEFINE_MUTEX(pdev_list_mutex);
278
279static int __cpuinit pkgtemp_device_add(unsigned int cpu)
280{
281 int err;
282 struct platform_device *pdev;
283 struct pdev_entry *pdev_entry;
284#ifdef CONFIG_SMP
285 struct cpuinfo_x86 *c = &cpu_data(cpu);
286#endif
287
288 mutex_lock(&pdev_list_mutex);
289
290#ifdef CONFIG_SMP
291 /* Only keep the first entry in each package */
292 list_for_each_entry(pdev_entry, &pdev_list, list) {
293 if (c->phys_proc_id == pdev_entry->phys_proc_id) {
294 err = 0; /* Not an error */
295 goto exit;
296 }
297 }
298#endif
299
300 pdev = platform_device_alloc(DRVNAME, cpu);
301 if (!pdev) {
302 err = -ENOMEM;
303 printk(KERN_ERR DRVNAME ": Device allocation failed\n");
304 goto exit;
305 }
306
307 pdev_entry = kzalloc(sizeof(struct pdev_entry), GFP_KERNEL);
308 if (!pdev_entry) {
309 err = -ENOMEM;
310 goto exit_device_put;
311 }
312
313 err = platform_device_add(pdev);
314 if (err) {
315 printk(KERN_ERR DRVNAME ": Device addition failed (%d)\n",
316 err);
317 goto exit_device_free;
318 }
319
320#ifdef CONFIG_SMP
321 pdev_entry->phys_proc_id = c->phys_proc_id;
322#endif
323 pdev_entry->pdev = pdev;
324 pdev_entry->cpu = cpu;
325 list_add_tail(&pdev_entry->list, &pdev_list);
326 mutex_unlock(&pdev_list_mutex);
327
328 return 0;
329
330exit_device_free:
331 kfree(pdev_entry);
332exit_device_put:
333 platform_device_put(pdev);
334exit:
335 mutex_unlock(&pdev_list_mutex);
336 return err;
337}
338
339#ifdef CONFIG_HOTPLUG_CPU
340static void pkgtemp_device_remove(unsigned int cpu)
341{
d9bca435 342 struct pdev_entry *p;
cb84b194
FY
343 unsigned int i;
344 int err;
345
346 mutex_lock(&pdev_list_mutex);
d9bca435 347 list_for_each_entry(p, &pdev_list, list) {
cb84b194
FY
348 if (p->cpu != cpu)
349 continue;
350
351 platform_device_unregister(p->pdev);
352 list_del(&p->list);
d9bca435 353 mutex_unlock(&pdev_list_mutex);
cb84b194
FY
354 kfree(p);
355 for_each_cpu(i, cpu_core_mask(cpu)) {
356 if (i != cpu) {
357 err = pkgtemp_device_add(i);
358 if (!err)
359 break;
360 }
361 }
d9bca435 362 return;
cb84b194
FY
363 }
364 mutex_unlock(&pdev_list_mutex);
365}
366
367static int __cpuinit pkgtemp_cpu_callback(struct notifier_block *nfb,
368 unsigned long action, void *hcpu)
369{
370 unsigned int cpu = (unsigned long) hcpu;
371
372 switch (action) {
373 case CPU_ONLINE:
374 case CPU_DOWN_FAILED:
375 pkgtemp_device_add(cpu);
376 break;
377 case CPU_DOWN_PREPARE:
378 pkgtemp_device_remove(cpu);
379 break;
380 }
381 return NOTIFY_OK;
382}
383
384static struct notifier_block pkgtemp_cpu_notifier __refdata = {
385 .notifier_call = pkgtemp_cpu_callback,
386};
387#endif /* !CONFIG_HOTPLUG_CPU */
388
389static int __init pkgtemp_init(void)
390{
391 int i, err = -ENODEV;
392 struct pdev_entry *p, *n;
393
394 /* quick check if we run Intel */
395 if (cpu_data(0).x86_vendor != X86_VENDOR_INTEL)
396 goto exit;
397
398 err = platform_driver_register(&pkgtemp_driver);
399 if (err)
400 goto exit;
401
402 for_each_online_cpu(i) {
403 struct cpuinfo_x86 *c = &cpu_data(i);
404
405 if (!cpu_has(c, X86_FEATURE_PTS))
406 continue;
407
408 err = pkgtemp_device_add(i);
409 if (err)
410 goto exit_devices_unreg;
411 }
412 if (list_empty(&pdev_list)) {
413 err = -ENODEV;
414 goto exit_driver_unreg;
415 }
416
417#ifdef CONFIG_HOTPLUG_CPU
418 register_hotcpu_notifier(&pkgtemp_cpu_notifier);
419#endif
420 return 0;
421
422exit_devices_unreg:
423 mutex_lock(&pdev_list_mutex);
424 list_for_each_entry_safe(p, n, &pdev_list, list) {
425 platform_device_unregister(p->pdev);
426 list_del(&p->list);
427 kfree(p);
428 }
429 mutex_unlock(&pdev_list_mutex);
430exit_driver_unreg:
431 platform_driver_unregister(&pkgtemp_driver);
432exit:
433 return err;
434}
435
436static void __exit pkgtemp_exit(void)
437{
438 struct pdev_entry *p, *n;
439#ifdef CONFIG_HOTPLUG_CPU
440 unregister_hotcpu_notifier(&pkgtemp_cpu_notifier);
441#endif
442 mutex_lock(&pdev_list_mutex);
443 list_for_each_entry_safe(p, n, &pdev_list, list) {
444 platform_device_unregister(p->pdev);
445 list_del(&p->list);
446 kfree(p);
447 }
448 mutex_unlock(&pdev_list_mutex);
449 platform_driver_unregister(&pkgtemp_driver);
450}
451
452MODULE_AUTHOR("Fenghua Yu <fenghua.yu@intel.com>");
453MODULE_DESCRIPTION("Intel processor package temperature monitor");
454MODULE_LICENSE("GPL");
455
456module_init(pkgtemp_init)
457module_exit(pkgtemp_exit)
This page took 0.057179 seconds and 5 git commands to generate.