Commit | Line | Data |
---|---|---|
f1a18a10 SP |
1 | /* |
2 | * x86_pkg_temp_thermal driver | |
3 | * Copyright (c) 2013, Intel Corporation. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify it | |
6 | * under the terms and conditions of the GNU General Public License, | |
7 | * version 2, as published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope it will be useful, but WITHOUT | |
10 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
11 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
12 | * more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License along with | |
15 | * this program; if not, write to the Free Software Foundation, Inc. | |
16 | * | |
17 | */ | |
18 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
19 | ||
20 | #include <linux/module.h> | |
21 | #include <linux/init.h> | |
22 | #include <linux/err.h> | |
23 | #include <linux/param.h> | |
24 | #include <linux/device.h> | |
25 | #include <linux/platform_device.h> | |
26 | #include <linux/cpu.h> | |
27 | #include <linux/smp.h> | |
28 | #include <linux/slab.h> | |
29 | #include <linux/pm.h> | |
30 | #include <linux/thermal.h> | |
31 | #include <linux/debugfs.h> | |
32 | #include <asm/cpu_device_id.h> | |
33 | #include <asm/mce.h> | |
34 | ||
35 | /* | |
36 | * Rate control delay: Idea is to introduce denounce effect | |
37 | * This should be long enough to avoid reduce events, when | |
38 | * threshold is set to a temperature, which is constantly | |
39 | * violated, but at the short enough to take any action. | |
40 | * The action can be remove threshold or change it to next | |
41 | * interesting setting. Based on experiments, in around | |
42 | * every 5 seconds under load will give us a significant | |
43 | * temperature change. | |
44 | */ | |
45 | #define PKG_TEMP_THERMAL_NOTIFY_DELAY 5000 | |
46 | static int notify_delay_ms = PKG_TEMP_THERMAL_NOTIFY_DELAY; | |
47 | module_param(notify_delay_ms, int, 0644); | |
48 | MODULE_PARM_DESC(notify_delay_ms, | |
49 | "User space notification delay in milli seconds."); | |
50 | ||
51 | /* Number of trip points in thermal zone. Currently it can't | |
52 | * be more than 2. MSR can allow setting and getting notifications | |
53 | * for only 2 thresholds. This define enforces this, if there | |
54 | * is some wrong values returned by cpuid for number of thresholds. | |
55 | */ | |
56 | #define MAX_NUMBER_OF_TRIPS 2 | |
94e791f5 SP |
57 | /* Limit number of package temp zones */ |
58 | #define MAX_PKG_TEMP_ZONE_IDS 256 | |
f1a18a10 SP |
59 | |
60 | struct phy_dev_entry { | |
61 | struct list_head list; | |
62 | u16 phys_proc_id; | |
63 | u16 first_cpu; | |
64 | u32 tj_max; | |
65 | int ref_cnt; | |
66 | u32 start_pkg_therm_low; | |
67 | u32 start_pkg_therm_high; | |
68 | struct thermal_zone_device *tzone; | |
69 | }; | |
70 | ||
71 | /* List maintaining number of package instances */ | |
72 | static LIST_HEAD(phy_dev_list); | |
73 | static DEFINE_MUTEX(phy_dev_list_mutex); | |
74 | ||
75 | /* Interrupt to work function schedule queue */ | |
76 | static DEFINE_PER_CPU(struct delayed_work, pkg_temp_thermal_threshold_work); | |
77 | ||
78 | /* To track if the work is already scheduled on a package */ | |
79 | static u8 *pkg_work_scheduled; | |
80 | ||
81 | /* Spin lock to prevent races with pkg_work_scheduled */ | |
82 | static spinlock_t pkg_work_lock; | |
83 | static u16 max_phy_id; | |
84 | ||
85 | /* Debug counters to show using debugfs */ | |
86 | static struct dentry *debugfs; | |
87 | static unsigned int pkg_interrupt_cnt; | |
88 | static unsigned int pkg_work_cnt; | |
89 | ||
90 | static int pkg_temp_debugfs_init(void) | |
91 | { | |
92 | struct dentry *d; | |
93 | ||
94 | debugfs = debugfs_create_dir("pkg_temp_thermal", NULL); | |
95 | if (!debugfs) | |
96 | return -ENOENT; | |
97 | ||
98 | d = debugfs_create_u32("pkg_thres_interrupt", S_IRUGO, debugfs, | |
99 | (u32 *)&pkg_interrupt_cnt); | |
100 | if (!d) | |
101 | goto err_out; | |
102 | ||
103 | d = debugfs_create_u32("pkg_thres_work", S_IRUGO, debugfs, | |
104 | (u32 *)&pkg_work_cnt); | |
105 | if (!d) | |
106 | goto err_out; | |
107 | ||
108 | return 0; | |
109 | ||
110 | err_out: | |
111 | debugfs_remove_recursive(debugfs); | |
112 | return -ENOENT; | |
113 | } | |
114 | ||
115 | static struct phy_dev_entry | |
116 | *pkg_temp_thermal_get_phy_entry(unsigned int cpu) | |
117 | { | |
118 | u16 phys_proc_id = topology_physical_package_id(cpu); | |
119 | struct phy_dev_entry *phy_ptr; | |
120 | ||
121 | mutex_lock(&phy_dev_list_mutex); | |
122 | ||
123 | list_for_each_entry(phy_ptr, &phy_dev_list, list) | |
124 | if (phy_ptr->phys_proc_id == phys_proc_id) { | |
125 | mutex_unlock(&phy_dev_list_mutex); | |
126 | return phy_ptr; | |
127 | } | |
128 | ||
129 | mutex_unlock(&phy_dev_list_mutex); | |
130 | ||
131 | return NULL; | |
132 | } | |
133 | ||
134 | /* | |
135 | * tj-max is is interesting because threshold is set relative to this | |
136 | * temperature. | |
137 | */ | |
138 | static int get_tj_max(int cpu, u32 *tj_max) | |
139 | { | |
140 | u32 eax, edx; | |
141 | u32 val; | |
142 | int err; | |
143 | ||
144 | err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx); | |
145 | if (err) | |
146 | goto err_ret; | |
147 | else { | |
148 | val = (eax >> 16) & 0xff; | |
149 | if (val) | |
150 | *tj_max = val * 1000; | |
151 | else { | |
152 | err = -EINVAL; | |
153 | goto err_ret; | |
154 | } | |
155 | } | |
156 | ||
157 | return 0; | |
158 | err_ret: | |
159 | *tj_max = 0; | |
160 | return err; | |
161 | } | |
162 | ||
163 | static int sys_get_curr_temp(struct thermal_zone_device *tzd, unsigned long *temp) | |
164 | { | |
165 | u32 eax, edx; | |
166 | struct phy_dev_entry *phy_dev_entry; | |
167 | ||
168 | phy_dev_entry = tzd->devdata; | |
169 | rdmsr_on_cpu(phy_dev_entry->first_cpu, MSR_IA32_PACKAGE_THERM_STATUS, | |
170 | &eax, &edx); | |
171 | if (eax & 0x80000000) { | |
172 | *temp = phy_dev_entry->tj_max - | |
173 | ((eax >> 16) & 0x7f) * 1000; | |
174 | pr_debug("sys_get_curr_temp %ld\n", *temp); | |
175 | return 0; | |
176 | } | |
177 | ||
178 | return -EINVAL; | |
179 | } | |
180 | ||
181 | static int sys_get_trip_temp(struct thermal_zone_device *tzd, | |
182 | int trip, unsigned long *temp) | |
183 | { | |
184 | u32 eax, edx; | |
185 | struct phy_dev_entry *phy_dev_entry; | |
186 | u32 mask, shift; | |
187 | unsigned long thres_reg_value; | |
188 | int ret; | |
189 | ||
190 | if (trip >= MAX_NUMBER_OF_TRIPS) | |
191 | return -EINVAL; | |
192 | ||
193 | phy_dev_entry = tzd->devdata; | |
194 | ||
195 | if (trip) { | |
196 | mask = THERM_MASK_THRESHOLD1; | |
197 | shift = THERM_SHIFT_THRESHOLD1; | |
198 | } else { | |
199 | mask = THERM_MASK_THRESHOLD0; | |
200 | shift = THERM_SHIFT_THRESHOLD0; | |
201 | } | |
202 | ||
203 | ret = rdmsr_on_cpu(phy_dev_entry->first_cpu, | |
204 | MSR_IA32_PACKAGE_THERM_INTERRUPT, &eax, &edx); | |
205 | if (ret < 0) | |
206 | return -EINVAL; | |
207 | ||
208 | thres_reg_value = (eax & mask) >> shift; | |
209 | if (thres_reg_value) | |
210 | *temp = phy_dev_entry->tj_max - thres_reg_value * 1000; | |
211 | else | |
212 | *temp = 0; | |
213 | pr_debug("sys_get_trip_temp %ld\n", *temp); | |
214 | ||
215 | return 0; | |
216 | } | |
217 | ||
218 | int sys_set_trip_temp(struct thermal_zone_device *tzd, int trip, | |
219 | unsigned long temp) | |
220 | { | |
221 | u32 l, h; | |
222 | struct phy_dev_entry *phy_dev_entry; | |
223 | u32 mask, shift, intr; | |
224 | int ret; | |
225 | ||
226 | phy_dev_entry = tzd->devdata; | |
227 | ||
228 | if (trip >= MAX_NUMBER_OF_TRIPS || temp >= phy_dev_entry->tj_max) | |
229 | return -EINVAL; | |
230 | ||
231 | ret = rdmsr_on_cpu(phy_dev_entry->first_cpu, | |
232 | MSR_IA32_PACKAGE_THERM_INTERRUPT, | |
233 | &l, &h); | |
234 | if (ret < 0) | |
235 | return -EINVAL; | |
236 | ||
237 | if (trip) { | |
238 | mask = THERM_MASK_THRESHOLD1; | |
239 | shift = THERM_SHIFT_THRESHOLD1; | |
240 | intr = THERM_INT_THRESHOLD1_ENABLE; | |
241 | } else { | |
242 | mask = THERM_MASK_THRESHOLD0; | |
243 | shift = THERM_SHIFT_THRESHOLD0; | |
244 | intr = THERM_INT_THRESHOLD0_ENABLE; | |
245 | } | |
246 | l &= ~mask; | |
247 | /* | |
248 | * When users space sets a trip temperature == 0, which is indication | |
249 | * that, it is no longer interested in receiving notifications. | |
250 | */ | |
251 | if (!temp) | |
252 | l &= ~intr; | |
253 | else { | |
254 | l |= (phy_dev_entry->tj_max - temp)/1000 << shift; | |
255 | l |= intr; | |
256 | } | |
257 | ||
258 | return wrmsr_on_cpu(phy_dev_entry->first_cpu, | |
259 | MSR_IA32_PACKAGE_THERM_INTERRUPT, | |
260 | l, h); | |
261 | } | |
262 | ||
263 | static int sys_get_trip_type(struct thermal_zone_device *thermal, | |
264 | int trip, enum thermal_trip_type *type) | |
265 | { | |
266 | ||
267 | *type = THERMAL_TRIP_PASSIVE; | |
268 | ||
269 | return 0; | |
270 | } | |
271 | ||
272 | /* Thermal zone callback registry */ | |
273 | static struct thermal_zone_device_ops tzone_ops = { | |
274 | .get_temp = sys_get_curr_temp, | |
275 | .get_trip_temp = sys_get_trip_temp, | |
276 | .get_trip_type = sys_get_trip_type, | |
277 | .set_trip_temp = sys_set_trip_temp, | |
278 | }; | |
279 | ||
280 | static bool pkg_temp_thermal_platform_thermal_rate_control(void) | |
281 | { | |
282 | return true; | |
283 | } | |
284 | ||
285 | /* Enable threshold interrupt on local package/cpu */ | |
286 | static inline void enable_pkg_thres_interrupt(void) | |
287 | { | |
288 | u32 l, h; | |
289 | u8 thres_0, thres_1; | |
290 | ||
291 | rdmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, l, h); | |
292 | /* only enable/disable if it had valid threshold value */ | |
293 | thres_0 = (l & THERM_MASK_THRESHOLD0) >> THERM_SHIFT_THRESHOLD0; | |
294 | thres_1 = (l & THERM_MASK_THRESHOLD1) >> THERM_SHIFT_THRESHOLD1; | |
295 | if (thres_0) | |
296 | l |= THERM_INT_THRESHOLD0_ENABLE; | |
297 | if (thres_1) | |
298 | l |= THERM_INT_THRESHOLD1_ENABLE; | |
299 | wrmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, l, h); | |
300 | } | |
301 | ||
302 | /* Disable threshold interrupt on local package/cpu */ | |
303 | static inline void disable_pkg_thres_interrupt(void) | |
304 | { | |
305 | u32 l, h; | |
306 | rdmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, l, h); | |
307 | wrmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, | |
308 | l & (~THERM_INT_THRESHOLD0_ENABLE) & | |
309 | (~THERM_INT_THRESHOLD1_ENABLE), h); | |
310 | } | |
311 | ||
312 | static void pkg_temp_thermal_threshold_work_fn(struct work_struct *work) | |
313 | { | |
314 | __u64 msr_val; | |
315 | int cpu = smp_processor_id(); | |
316 | int phy_id = topology_physical_package_id(cpu); | |
317 | struct phy_dev_entry *phdev = pkg_temp_thermal_get_phy_entry(cpu); | |
318 | bool notify = false; | |
7bed1b3c | 319 | unsigned long flags; |
f1a18a10 SP |
320 | |
321 | if (!phdev) | |
322 | return; | |
323 | ||
7bed1b3c | 324 | spin_lock_irqsave(&pkg_work_lock, flags); |
f1a18a10 SP |
325 | ++pkg_work_cnt; |
326 | if (unlikely(phy_id > max_phy_id)) { | |
7bed1b3c | 327 | spin_unlock_irqrestore(&pkg_work_lock, flags); |
f1a18a10 SP |
328 | return; |
329 | } | |
330 | pkg_work_scheduled[phy_id] = 0; | |
7bed1b3c | 331 | spin_unlock_irqrestore(&pkg_work_lock, flags); |
f1a18a10 SP |
332 | |
333 | enable_pkg_thres_interrupt(); | |
334 | rdmsrl(MSR_IA32_PACKAGE_THERM_STATUS, msr_val); | |
335 | if (msr_val & THERM_LOG_THRESHOLD0) { | |
336 | wrmsrl(MSR_IA32_PACKAGE_THERM_STATUS, | |
337 | msr_val & ~THERM_LOG_THRESHOLD0); | |
338 | notify = true; | |
339 | } | |
340 | if (msr_val & THERM_LOG_THRESHOLD1) { | |
341 | wrmsrl(MSR_IA32_PACKAGE_THERM_STATUS, | |
342 | msr_val & ~THERM_LOG_THRESHOLD1); | |
343 | notify = true; | |
344 | } | |
345 | if (notify) { | |
346 | pr_debug("thermal_zone_device_update\n"); | |
347 | thermal_zone_device_update(phdev->tzone); | |
348 | } | |
349 | } | |
350 | ||
351 | static int pkg_temp_thermal_platform_thermal_notify(__u64 msr_val) | |
352 | { | |
353 | unsigned long flags; | |
354 | int cpu = smp_processor_id(); | |
355 | int phy_id = topology_physical_package_id(cpu); | |
356 | ||
357 | /* | |
358 | * When a package is in interrupted state, all CPU's in that package | |
359 | * are in the same interrupt state. So scheduling on any one CPU in | |
360 | * the package is enough and simply return for others. | |
361 | */ | |
362 | spin_lock_irqsave(&pkg_work_lock, flags); | |
363 | ++pkg_interrupt_cnt; | |
364 | if (unlikely(phy_id > max_phy_id) || unlikely(!pkg_work_scheduled) || | |
365 | pkg_work_scheduled[phy_id]) { | |
366 | disable_pkg_thres_interrupt(); | |
367 | spin_unlock_irqrestore(&pkg_work_lock, flags); | |
368 | return -EINVAL; | |
369 | } | |
370 | pkg_work_scheduled[phy_id] = 1; | |
371 | spin_unlock_irqrestore(&pkg_work_lock, flags); | |
372 | ||
373 | disable_pkg_thres_interrupt(); | |
374 | schedule_delayed_work_on(cpu, | |
375 | &per_cpu(pkg_temp_thermal_threshold_work, cpu), | |
376 | msecs_to_jiffies(notify_delay_ms)); | |
377 | return 0; | |
378 | } | |
379 | ||
380 | static int find_siblings_cpu(int cpu) | |
381 | { | |
382 | int i; | |
383 | int id = topology_physical_package_id(cpu); | |
384 | ||
385 | for_each_online_cpu(i) | |
386 | if (i != cpu && topology_physical_package_id(i) == id) | |
387 | return i; | |
388 | ||
389 | return 0; | |
390 | } | |
391 | ||
392 | static int pkg_temp_thermal_device_add(unsigned int cpu) | |
393 | { | |
394 | int err; | |
395 | u32 tj_max; | |
396 | struct phy_dev_entry *phy_dev_entry; | |
397 | char buffer[30]; | |
398 | int thres_count; | |
399 | u32 eax, ebx, ecx, edx; | |
c7c1b311 | 400 | u8 *temp; |
7bed1b3c | 401 | unsigned long flags; |
f1a18a10 SP |
402 | |
403 | cpuid(6, &eax, &ebx, &ecx, &edx); | |
404 | thres_count = ebx & 0x07; | |
405 | if (!thres_count) | |
406 | return -ENODEV; | |
407 | ||
94e791f5 SP |
408 | if (topology_physical_package_id(cpu) > MAX_PKG_TEMP_ZONE_IDS) |
409 | return -ENODEV; | |
410 | ||
f1a18a10 SP |
411 | thres_count = clamp_val(thres_count, 0, MAX_NUMBER_OF_TRIPS); |
412 | ||
413 | err = get_tj_max(cpu, &tj_max); | |
414 | if (err) | |
415 | goto err_ret; | |
416 | ||
417 | mutex_lock(&phy_dev_list_mutex); | |
418 | ||
419 | phy_dev_entry = kzalloc(sizeof(*phy_dev_entry), GFP_KERNEL); | |
420 | if (!phy_dev_entry) { | |
421 | err = -ENOMEM; | |
422 | goto err_ret_unlock; | |
423 | } | |
424 | ||
7bed1b3c | 425 | spin_lock_irqsave(&pkg_work_lock, flags); |
f1a18a10 SP |
426 | if (topology_physical_package_id(cpu) > max_phy_id) |
427 | max_phy_id = topology_physical_package_id(cpu); | |
c7c1b311 WY |
428 | temp = krealloc(pkg_work_scheduled, |
429 | (max_phy_id+1) * sizeof(u8), GFP_ATOMIC); | |
430 | if (!temp) { | |
7bed1b3c | 431 | spin_unlock_irqrestore(&pkg_work_lock, flags); |
f1a18a10 SP |
432 | err = -ENOMEM; |
433 | goto err_ret_free; | |
434 | } | |
c7c1b311 | 435 | pkg_work_scheduled = temp; |
f1a18a10 | 436 | pkg_work_scheduled[topology_physical_package_id(cpu)] = 0; |
7bed1b3c | 437 | spin_unlock_irqrestore(&pkg_work_lock, flags); |
f1a18a10 SP |
438 | |
439 | phy_dev_entry->phys_proc_id = topology_physical_package_id(cpu); | |
440 | phy_dev_entry->first_cpu = cpu; | |
441 | phy_dev_entry->tj_max = tj_max; | |
442 | phy_dev_entry->ref_cnt = 1; | |
443 | snprintf(buffer, sizeof(buffer), "pkg-temp-%d\n", | |
444 | phy_dev_entry->phys_proc_id); | |
445 | phy_dev_entry->tzone = thermal_zone_device_register(buffer, | |
446 | thres_count, | |
447 | (thres_count == MAX_NUMBER_OF_TRIPS) ? | |
448 | 0x03 : 0x01, | |
449 | phy_dev_entry, &tzone_ops, NULL, 0, 0); | |
450 | if (IS_ERR(phy_dev_entry->tzone)) { | |
451 | err = PTR_ERR(phy_dev_entry->tzone); | |
452 | goto err_ret_free; | |
453 | } | |
454 | /* Store MSR value for package thermal interrupt, to restore at exit */ | |
455 | rdmsr_on_cpu(cpu, MSR_IA32_PACKAGE_THERM_INTERRUPT, | |
456 | &phy_dev_entry->start_pkg_therm_low, | |
457 | &phy_dev_entry->start_pkg_therm_high); | |
458 | ||
459 | list_add_tail(&phy_dev_entry->list, &phy_dev_list); | |
460 | pr_debug("pkg_temp_thermal_device_add :phy_id %d cpu %d\n", | |
461 | phy_dev_entry->phys_proc_id, cpu); | |
462 | ||
463 | mutex_unlock(&phy_dev_list_mutex); | |
464 | ||
465 | return 0; | |
466 | ||
467 | err_ret_free: | |
468 | kfree(phy_dev_entry); | |
469 | err_ret_unlock: | |
470 | mutex_unlock(&phy_dev_list_mutex); | |
471 | ||
472 | err_ret: | |
473 | return err; | |
474 | } | |
475 | ||
476 | static int pkg_temp_thermal_device_remove(unsigned int cpu) | |
477 | { | |
478 | struct phy_dev_entry *n; | |
479 | u16 phys_proc_id = topology_physical_package_id(cpu); | |
480 | struct phy_dev_entry *phdev = | |
481 | pkg_temp_thermal_get_phy_entry(cpu); | |
482 | ||
483 | if (!phdev) | |
484 | return -ENODEV; | |
485 | ||
486 | mutex_lock(&phy_dev_list_mutex); | |
487 | /* If we are loosing the first cpu for this package, we need change */ | |
488 | if (phdev->first_cpu == cpu) { | |
489 | phdev->first_cpu = find_siblings_cpu(cpu); | |
490 | pr_debug("thermal_device_remove: first cpu switched %d\n", | |
491 | phdev->first_cpu); | |
492 | } | |
493 | /* | |
494 | * It is possible that no siblings left as this was the last cpu | |
495 | * going offline. We don't need to worry about this assignment | |
496 | * as the phydev entry will be removed in this case and | |
497 | * thermal zone is removed. | |
498 | */ | |
499 | --phdev->ref_cnt; | |
500 | pr_debug("thermal_device_remove: pkg: %d cpu %d ref_cnt %d\n", | |
501 | phys_proc_id, cpu, phdev->ref_cnt); | |
502 | if (!phdev->ref_cnt) | |
503 | list_for_each_entry_safe(phdev, n, &phy_dev_list, list) { | |
504 | if (phdev->phys_proc_id == phys_proc_id) { | |
505 | thermal_zone_device_unregister(phdev->tzone); | |
506 | list_del(&phdev->list); | |
507 | kfree(phdev); | |
508 | break; | |
509 | } | |
510 | } | |
511 | mutex_unlock(&phy_dev_list_mutex); | |
512 | ||
513 | return 0; | |
514 | } | |
515 | ||
516 | static int get_core_online(unsigned int cpu) | |
517 | { | |
518 | struct cpuinfo_x86 *c = &cpu_data(cpu); | |
519 | struct phy_dev_entry *phdev = pkg_temp_thermal_get_phy_entry(cpu); | |
520 | ||
521 | /* Check if there is already an instance for this package */ | |
522 | if (!phdev) { | |
f3ed0a17 | 523 | if (!cpu_has(c, X86_FEATURE_DTHERM) || |
f1a18a10 SP |
524 | !cpu_has(c, X86_FEATURE_PTS)) |
525 | return -ENODEV; | |
526 | if (pkg_temp_thermal_device_add(cpu)) | |
527 | return -ENODEV; | |
528 | } else { | |
529 | mutex_lock(&phy_dev_list_mutex); | |
530 | ++phdev->ref_cnt; | |
531 | pr_debug("get_core_online: cpu %d ref_cnt %d\n", | |
532 | cpu, phdev->ref_cnt); | |
533 | mutex_unlock(&phy_dev_list_mutex); | |
534 | } | |
535 | INIT_DELAYED_WORK(&per_cpu(pkg_temp_thermal_threshold_work, cpu), | |
536 | pkg_temp_thermal_threshold_work_fn); | |
537 | ||
538 | pr_debug("get_core_online: cpu %d successful\n", cpu); | |
539 | ||
540 | return 0; | |
541 | } | |
542 | ||
543 | static void put_core_offline(unsigned int cpu) | |
544 | { | |
545 | if (!pkg_temp_thermal_device_remove(cpu)) | |
546 | cancel_delayed_work_sync( | |
547 | &per_cpu(pkg_temp_thermal_threshold_work, cpu)); | |
548 | ||
549 | pr_debug("put_core_offline: cpu %d\n", cpu); | |
550 | } | |
551 | ||
552 | static int pkg_temp_thermal_cpu_callback(struct notifier_block *nfb, | |
553 | unsigned long action, void *hcpu) | |
554 | { | |
555 | unsigned int cpu = (unsigned long) hcpu; | |
556 | ||
557 | switch (action) { | |
558 | case CPU_ONLINE: | |
559 | case CPU_DOWN_FAILED: | |
560 | get_core_online(cpu); | |
561 | break; | |
562 | case CPU_DOWN_PREPARE: | |
563 | put_core_offline(cpu); | |
564 | break; | |
565 | } | |
566 | return NOTIFY_OK; | |
567 | } | |
568 | ||
569 | static struct notifier_block pkg_temp_thermal_notifier __refdata = { | |
570 | .notifier_call = pkg_temp_thermal_cpu_callback, | |
571 | }; | |
572 | ||
573 | static const struct x86_cpu_id __initconst pkg_temp_thermal_ids[] = { | |
f3ed0a17 | 574 | { X86_VENDOR_INTEL, X86_FAMILY_ANY, X86_MODEL_ANY, X86_FEATURE_PTS }, |
f1a18a10 SP |
575 | {} |
576 | }; | |
577 | MODULE_DEVICE_TABLE(x86cpu, pkg_temp_thermal_ids); | |
578 | ||
579 | static int __init pkg_temp_thermal_init(void) | |
580 | { | |
581 | int i; | |
582 | ||
583 | if (!x86_match_cpu(pkg_temp_thermal_ids)) | |
584 | return -ENODEV; | |
585 | ||
586 | spin_lock_init(&pkg_work_lock); | |
587 | platform_thermal_package_notify = | |
588 | pkg_temp_thermal_platform_thermal_notify; | |
589 | platform_thermal_package_rate_control = | |
590 | pkg_temp_thermal_platform_thermal_rate_control; | |
591 | ||
592 | get_online_cpus(); | |
593 | for_each_online_cpu(i) | |
594 | if (get_core_online(i)) | |
595 | goto err_ret; | |
596 | register_hotcpu_notifier(&pkg_temp_thermal_notifier); | |
597 | put_online_cpus(); | |
598 | ||
599 | pkg_temp_debugfs_init(); /* Don't care if fails */ | |
600 | ||
601 | return 0; | |
602 | ||
603 | err_ret: | |
f1a18a10 SP |
604 | for_each_online_cpu(i) |
605 | put_core_offline(i); | |
606 | put_online_cpus(); | |
607 | kfree(pkg_work_scheduled); | |
608 | platform_thermal_package_notify = NULL; | |
609 | platform_thermal_package_rate_control = NULL; | |
610 | ||
611 | return -ENODEV; | |
612 | } | |
613 | ||
614 | static void __exit pkg_temp_thermal_exit(void) | |
615 | { | |
616 | struct phy_dev_entry *phdev, *n; | |
617 | int i; | |
618 | ||
619 | get_online_cpus(); | |
620 | unregister_hotcpu_notifier(&pkg_temp_thermal_notifier); | |
621 | mutex_lock(&phy_dev_list_mutex); | |
622 | list_for_each_entry_safe(phdev, n, &phy_dev_list, list) { | |
623 | /* Retore old MSR value for package thermal interrupt */ | |
624 | wrmsr_on_cpu(phdev->first_cpu, | |
625 | MSR_IA32_PACKAGE_THERM_INTERRUPT, | |
626 | phdev->start_pkg_therm_low, | |
627 | phdev->start_pkg_therm_high); | |
628 | thermal_zone_device_unregister(phdev->tzone); | |
629 | list_del(&phdev->list); | |
630 | kfree(phdev); | |
631 | } | |
632 | mutex_unlock(&phy_dev_list_mutex); | |
633 | platform_thermal_package_notify = NULL; | |
634 | platform_thermal_package_rate_control = NULL; | |
635 | for_each_online_cpu(i) | |
636 | cancel_delayed_work_sync( | |
637 | &per_cpu(pkg_temp_thermal_threshold_work, i)); | |
638 | put_online_cpus(); | |
639 | ||
640 | kfree(pkg_work_scheduled); | |
641 | ||
642 | debugfs_remove_recursive(debugfs); | |
643 | } | |
644 | ||
645 | module_init(pkg_temp_thermal_init) | |
646 | module_exit(pkg_temp_thermal_exit) | |
647 | ||
648 | MODULE_DESCRIPTION("X86 PKG TEMP Thermal Driver"); | |
649 | MODULE_AUTHOR("Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>"); | |
650 | MODULE_LICENSE("GPL v2"); |