xref: /freebsd/sys/dev/acpica/acpi_thermal.c (revision c93b6e5fa24ba172ab271432c6692f9cc604e15a)
1 /*-
2  * Copyright (c) 2000, 2001 Michael Smith
3  * Copyright (c) 2000 BSDi
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 
31 #include "opt_acpi.h"
32 #include <sys/param.h>
33 #include <sys/eventhandler.h>
34 #include <sys/kernel.h>
35 #include <sys/bus.h>
36 #include <sys/cpu.h>
37 #include <sys/kthread.h>
38 #include <sys/malloc.h>
39 #include <sys/module.h>
40 #include <sys/proc.h>
41 #include <sys/reboot.h>
42 #include <sys/sysctl.h>
43 #include <sys/unistd.h>
44 #include <sys/power.h>
45 
46 #include "cpufreq_if.h"
47 
48 #include <contrib/dev/acpica/include/acpi.h>
49 #include <contrib/dev/acpica/include/accommon.h>
50 
51 #include <dev/acpica/acpivar.h>
52 
53 /* Hooks for the ACPI CA debugging infrastructure */
54 #define _COMPONENT	ACPI_THERMAL
55 ACPI_MODULE_NAME("THERMAL")
56 
57 #define TZ_ZEROC	2731
58 #define TZ_KELVTOC(x)	(((x) - TZ_ZEROC) / 10), abs(((x) - TZ_ZEROC) % 10)
59 
60 #define TZ_NOTIFY_TEMPERATURE	0x80 /* Temperature changed. */
61 #define TZ_NOTIFY_LEVELS	0x81 /* Cooling levels changed. */
62 #define TZ_NOTIFY_DEVICES	0x82 /* Device lists changed. */
63 #define TZ_NOTIFY_CRITICAL	0xcc /* Fake notify that _CRT/_HOT reached. */
64 
65 /* Check for temperature changes every 10 seconds by default */
66 #define TZ_POLLRATE	10
67 
68 /* Make sure the reported temperature is valid for this number of polls. */
69 #define TZ_VALIDCHECKS	3
70 
71 /* Notify the user we will be shutting down in one more poll cycle. */
72 #define TZ_NOTIFYCOUNT	(TZ_VALIDCHECKS - 1)
73 
74 /* ACPI spec defines this */
75 #define TZ_NUMLEVELS	10
76 struct acpi_tz_zone {
77     int		ac[TZ_NUMLEVELS];
78     ACPI_BUFFER	al[TZ_NUMLEVELS];
79     int		crt;
80     int		hot;
81     ACPI_BUFFER	psl;
82     int		psv;
83     int		tc1;
84     int		tc2;
85     int		tsp;
86     int		tzp;
87 };
88 
89 struct acpi_tz_softc {
90     device_t			tz_dev;
91     ACPI_HANDLE			tz_handle;	/*Thermal zone handle*/
92     int				tz_temperature;	/*Current temperature*/
93     int				tz_active;	/*Current active cooling*/
94 #define TZ_ACTIVE_NONE		-1
95 #define TZ_ACTIVE_UNKNOWN	-2
96     int				tz_requested;	/*Minimum active cooling*/
97     int				tz_thflags;	/*Current temp-related flags*/
98 #define TZ_THFLAG_NONE		0
99 #define TZ_THFLAG_PSV		(1<<0)
100 #define TZ_THFLAG_HOT		(1<<2)
101 #define TZ_THFLAG_CRT		(1<<3)
102     int				tz_flags;
103 #define TZ_FLAG_NO_SCP		(1<<0)		/*No _SCP method*/
104 #define TZ_FLAG_GETPROFILE	(1<<1)		/*Get power_profile in timeout*/
105 #define TZ_FLAG_GETSETTINGS	(1<<2)		/*Get devs/setpoints*/
106     struct timespec		tz_cooling_started;
107 					/*Current cooling starting time*/
108 
109     struct sysctl_ctx_list	tz_sysctl_ctx;
110     struct sysctl_oid		*tz_sysctl_tree;
111     eventhandler_tag		tz_event;
112 
113     struct acpi_tz_zone 	tz_zone;	/*Thermal zone parameters*/
114     int				tz_validchecks;
115     int				tz_insane_tmp_notified;
116 
117     /* passive cooling */
118     struct proc			*tz_cooling_proc;
119     int				tz_cooling_proc_running;
120     int				tz_cooling_enabled;
121     int				tz_cooling_active;
122     int				tz_cooling_updated;
123     int				tz_cooling_saved_freq;
124 };
125 
126 #define	TZ_ACTIVE_LEVEL(act)	((act) >= 0 ? (act) : TZ_NUMLEVELS)
127 
128 #define CPUFREQ_MAX_LEVELS	64 /* XXX cpufreq should export this */
129 
130 static int	acpi_tz_probe(device_t dev);
131 static int	acpi_tz_attach(device_t dev);
132 static int	acpi_tz_establish(struct acpi_tz_softc *sc);
133 static void	acpi_tz_monitor(void *Context);
134 static void	acpi_tz_switch_cooler_off(ACPI_OBJECT *obj, void *arg);
135 static void	acpi_tz_switch_cooler_on(ACPI_OBJECT *obj, void *arg);
136 static void	acpi_tz_getparam(struct acpi_tz_softc *sc, char *node,
137 				 int *data);
138 static void	acpi_tz_sanity(struct acpi_tz_softc *sc, int *val, char *what);
139 static int	acpi_tz_active_sysctl(SYSCTL_HANDLER_ARGS);
140 static int	acpi_tz_cooling_sysctl(SYSCTL_HANDLER_ARGS);
141 static int	acpi_tz_temp_sysctl(SYSCTL_HANDLER_ARGS);
142 static int	acpi_tz_passive_sysctl(SYSCTL_HANDLER_ARGS);
143 static void	acpi_tz_notify_handler(ACPI_HANDLE h, UINT32 notify,
144 				       void *context);
145 static void	acpi_tz_signal(struct acpi_tz_softc *sc, int flags);
146 static void	acpi_tz_timeout(struct acpi_tz_softc *sc, int flags);
147 static void	acpi_tz_power_profile(void *arg);
148 static void	acpi_tz_thread(void *arg);
149 static int	acpi_tz_cooling_is_available(struct acpi_tz_softc *sc);
150 static int	acpi_tz_cooling_thread_start(struct acpi_tz_softc *sc);
151 
152 static device_method_t acpi_tz_methods[] = {
153     /* Device interface */
154     DEVMETHOD(device_probe,	acpi_tz_probe),
155     DEVMETHOD(device_attach,	acpi_tz_attach),
156 
157     DEVMETHOD_END
158 };
159 
160 static driver_t acpi_tz_driver = {
161     "acpi_tz",
162     acpi_tz_methods,
163     sizeof(struct acpi_tz_softc),
164 };
165 
166 static char *acpi_tz_tmp_name = "_TMP";
167 
168 static devclass_t acpi_tz_devclass;
169 DRIVER_MODULE(acpi_tz, acpi, acpi_tz_driver, acpi_tz_devclass, 0, 0);
170 MODULE_DEPEND(acpi_tz, acpi, 1, 1, 1);
171 
172 static struct sysctl_ctx_list	acpi_tz_sysctl_ctx;
173 static struct sysctl_oid	*acpi_tz_sysctl_tree;
174 
175 /* Minimum cooling run time */
176 static int			acpi_tz_min_runtime;
177 static int			acpi_tz_polling_rate = TZ_POLLRATE;
178 static int			acpi_tz_override;
179 
180 /* Timezone polling thread */
181 static struct proc		*acpi_tz_proc;
182 ACPI_LOCK_DECL(thermal, "ACPI thermal zone");
183 
184 static int			acpi_tz_cooling_unit = -1;
185 
186 static int
187 acpi_tz_probe(device_t dev)
188 {
189     int		result;
190 
191     if (acpi_get_type(dev) == ACPI_TYPE_THERMAL && !acpi_disabled("thermal")) {
192 	device_set_desc(dev, "Thermal Zone");
193 	result = -10;
194     } else
195 	result = ENXIO;
196     return (result);
197 }
198 
199 static int
200 acpi_tz_attach(device_t dev)
201 {
202     struct acpi_tz_softc	*sc;
203     struct acpi_softc		*acpi_sc;
204     int				error;
205     char			oidname[8];
206 
207     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
208 
209     sc = device_get_softc(dev);
210     sc->tz_dev = dev;
211     sc->tz_handle = acpi_get_handle(dev);
212     sc->tz_requested = TZ_ACTIVE_NONE;
213     sc->tz_active = TZ_ACTIVE_UNKNOWN;
214     sc->tz_thflags = TZ_THFLAG_NONE;
215     sc->tz_cooling_proc = NULL;
216     sc->tz_cooling_proc_running = FALSE;
217     sc->tz_cooling_active = FALSE;
218     sc->tz_cooling_updated = FALSE;
219     sc->tz_cooling_enabled = FALSE;
220 
221     /*
222      * Parse the current state of the thermal zone and build control
223      * structures.  We don't need to worry about interference with the
224      * control thread since we haven't fully attached this device yet.
225      */
226     if ((error = acpi_tz_establish(sc)) != 0)
227 	return (error);
228 
229     /*
230      * Register for any Notify events sent to this zone.
231      */
232     AcpiInstallNotifyHandler(sc->tz_handle, ACPI_DEVICE_NOTIFY,
233 			     acpi_tz_notify_handler, sc);
234 
235     /*
236      * Create our sysctl nodes.
237      *
238      * XXX we need a mechanism for adding nodes under ACPI.
239      */
240     if (device_get_unit(dev) == 0) {
241 	acpi_sc = acpi_device_get_parent_softc(dev);
242 	sysctl_ctx_init(&acpi_tz_sysctl_ctx);
243 	acpi_tz_sysctl_tree = SYSCTL_ADD_NODE(&acpi_tz_sysctl_ctx,
244 			      SYSCTL_CHILDREN(acpi_sc->acpi_sysctl_tree),
245 			      OID_AUTO, "thermal", CTLFLAG_RD, 0, "");
246 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
247 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
248 		       OID_AUTO, "min_runtime", CTLFLAG_RW,
249 		       &acpi_tz_min_runtime, 0,
250 		       "minimum cooling run time in sec");
251 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
252 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
253 		       OID_AUTO, "polling_rate", CTLFLAG_RW,
254 		       &acpi_tz_polling_rate, 0, "monitor polling interval in seconds");
255 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
256 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree), OID_AUTO,
257 		       "user_override", CTLFLAG_RW, &acpi_tz_override, 0,
258 		       "allow override of thermal settings");
259     }
260     sysctl_ctx_init(&sc->tz_sysctl_ctx);
261     sprintf(oidname, "tz%d", device_get_unit(dev));
262     sc->tz_sysctl_tree = SYSCTL_ADD_NODE_WITH_LABEL(&sc->tz_sysctl_ctx,
263 			 SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
264 			 OID_AUTO, oidname, CTLFLAG_RD, 0, "", "thermal_zone");
265     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
266 		    OID_AUTO, "temperature", CTLTYPE_INT | CTLFLAG_RD,
267 		    &sc->tz_temperature, 0, sysctl_handle_int,
268 		    "IK", "current thermal zone temperature");
269     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
270 		    OID_AUTO, "active", CTLTYPE_INT | CTLFLAG_RW,
271 		    sc, 0, acpi_tz_active_sysctl, "I", "cooling is active");
272     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
273 		    OID_AUTO, "passive_cooling", CTLTYPE_INT | CTLFLAG_RW,
274 		    sc, 0, acpi_tz_cooling_sysctl, "I",
275 		    "enable passive (speed reduction) cooling");
276 
277     SYSCTL_ADD_INT(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
278 		   OID_AUTO, "thermal_flags", CTLFLAG_RD,
279 		   &sc->tz_thflags, 0, "thermal zone flags");
280     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
281 		    OID_AUTO, "_PSV", CTLTYPE_INT | CTLFLAG_RW,
282 		    sc, offsetof(struct acpi_tz_softc, tz_zone.psv),
283 		    acpi_tz_temp_sysctl, "IK", "passive cooling temp setpoint");
284     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
285 		    OID_AUTO, "_HOT", CTLTYPE_INT | CTLFLAG_RW,
286 		    sc, offsetof(struct acpi_tz_softc, tz_zone.hot),
287 		    acpi_tz_temp_sysctl, "IK",
288 		    "too hot temp setpoint (suspend now)");
289     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
290 		    OID_AUTO, "_CRT", CTLTYPE_INT | CTLFLAG_RW,
291 		    sc, offsetof(struct acpi_tz_softc, tz_zone.crt),
292 		    acpi_tz_temp_sysctl, "IK",
293 		    "critical temp setpoint (shutdown now)");
294     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
295 		    OID_AUTO, "_ACx", CTLTYPE_INT | CTLFLAG_RD,
296 		    &sc->tz_zone.ac, sizeof(sc->tz_zone.ac),
297 		    sysctl_handle_opaque, "IK", "");
298     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
299 		    OID_AUTO, "_TC1", CTLTYPE_INT | CTLFLAG_RW,
300 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tc1),
301 		    acpi_tz_passive_sysctl, "I",
302 		    "thermal constant 1 for passive cooling");
303     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
304 		    OID_AUTO, "_TC2", CTLTYPE_INT | CTLFLAG_RW,
305 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tc2),
306 		    acpi_tz_passive_sysctl, "I",
307 		    "thermal constant 2 for passive cooling");
308     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
309 		    OID_AUTO, "_TSP", CTLTYPE_INT | CTLFLAG_RW,
310 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tsp),
311 		    acpi_tz_passive_sysctl, "I",
312 		    "thermal sampling period for passive cooling");
313 
314     /*
315      * Register our power profile event handler.
316      */
317     sc->tz_event = EVENTHANDLER_REGISTER(power_profile_change,
318 	acpi_tz_power_profile, sc, 0);
319 
320     /*
321      * Flag the event handler for a manual invocation by our timeout.
322      * We defer it like this so that the rest of the subsystem has time
323      * to come up.  Don't bother evaluating/printing the temperature at
324      * this point; on many systems it'll be bogus until the EC is running.
325      */
326     sc->tz_flags |= TZ_FLAG_GETPROFILE;
327 
328     return_VALUE (0);
329 }
330 
331 static void
332 acpi_tz_startup(void *arg __unused)
333 {
334     struct acpi_tz_softc *sc;
335     device_t *devs;
336     int devcount, error, i;
337 
338     devclass_get_devices(acpi_tz_devclass, &devs, &devcount);
339     if (devcount == 0) {
340 	free(devs, M_TEMP);
341 	return;
342     }
343 
344     /*
345      * Create thread to service all of the thermal zones.
346      */
347     error = kproc_create(acpi_tz_thread, NULL, &acpi_tz_proc, RFHIGHPID, 0,
348 	"acpi_thermal");
349     if (error != 0)
350 	printf("acpi_tz: could not create thread - %d", error);
351 
352     /*
353      * Create a thread to handle passive cooling for 1st zone which
354      * has _PSV, _TSP, _TC1 and _TC2.  Users can enable it for other
355      * zones manually for now.
356      *
357      * XXX We enable only one zone to avoid multiple zones conflict
358      * with each other since cpufreq currently sets all CPUs to the
359      * given frequency whereas it's possible for different thermal
360      * zones to specify independent settings for multiple CPUs.
361      */
362     for (i = 0; i < devcount; i++) {
363 	sc = device_get_softc(devs[i]);
364 	if (acpi_tz_cooling_is_available(sc)) {
365 	    sc->tz_cooling_enabled = TRUE;
366 	    error = acpi_tz_cooling_thread_start(sc);
367 	    if (error != 0) {
368 		sc->tz_cooling_enabled = FALSE;
369 		break;
370 	    }
371 	    acpi_tz_cooling_unit = device_get_unit(devs[i]);
372 	    break;
373 	}
374     }
375     free(devs, M_TEMP);
376 }
377 SYSINIT(acpi_tz, SI_SUB_KICK_SCHEDULER, SI_ORDER_ANY, acpi_tz_startup, NULL);
378 
379 /*
380  * Parse the current state of this thermal zone and set up to use it.
381  *
382  * Note that we may have previous state, which will have to be discarded.
383  */
384 static int
385 acpi_tz_establish(struct acpi_tz_softc *sc)
386 {
387     ACPI_OBJECT	*obj;
388     int		i;
389     char	nbuf[8];
390 
391     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
392 
393     /* Erase any existing state. */
394     for (i = 0; i < TZ_NUMLEVELS; i++)
395 	if (sc->tz_zone.al[i].Pointer != NULL)
396 	    AcpiOsFree(sc->tz_zone.al[i].Pointer);
397     if (sc->tz_zone.psl.Pointer != NULL)
398 	AcpiOsFree(sc->tz_zone.psl.Pointer);
399 
400     /*
401      * XXX: We initialize only ACPI_BUFFER to avoid race condition
402      * with passive cooling thread which refers psv, tc1, tc2 and tsp.
403      */
404     bzero(sc->tz_zone.ac, sizeof(sc->tz_zone.ac));
405     bzero(sc->tz_zone.al, sizeof(sc->tz_zone.al));
406     bzero(&sc->tz_zone.psl, sizeof(sc->tz_zone.psl));
407 
408     /* Evaluate thermal zone parameters. */
409     for (i = 0; i < TZ_NUMLEVELS; i++) {
410 	sprintf(nbuf, "_AC%d", i);
411 	acpi_tz_getparam(sc, nbuf, &sc->tz_zone.ac[i]);
412 	sprintf(nbuf, "_AL%d", i);
413 	sc->tz_zone.al[i].Length = ACPI_ALLOCATE_BUFFER;
414 	sc->tz_zone.al[i].Pointer = NULL;
415 	AcpiEvaluateObject(sc->tz_handle, nbuf, NULL, &sc->tz_zone.al[i]);
416 	obj = (ACPI_OBJECT *)sc->tz_zone.al[i].Pointer;
417 	if (obj != NULL) {
418 	    /* Should be a package containing a list of power objects */
419 	    if (obj->Type != ACPI_TYPE_PACKAGE) {
420 		device_printf(sc->tz_dev, "%s has unknown type %d, rejecting\n",
421 			      nbuf, obj->Type);
422 		return_VALUE (ENXIO);
423 	    }
424 	}
425     }
426     acpi_tz_getparam(sc, "_CRT", &sc->tz_zone.crt);
427     acpi_tz_getparam(sc, "_HOT", &sc->tz_zone.hot);
428     sc->tz_zone.psl.Length = ACPI_ALLOCATE_BUFFER;
429     sc->tz_zone.psl.Pointer = NULL;
430     AcpiEvaluateObject(sc->tz_handle, "_PSL", NULL, &sc->tz_zone.psl);
431     acpi_tz_getparam(sc, "_PSV", &sc->tz_zone.psv);
432     acpi_tz_getparam(sc, "_TC1", &sc->tz_zone.tc1);
433     acpi_tz_getparam(sc, "_TC2", &sc->tz_zone.tc2);
434     acpi_tz_getparam(sc, "_TSP", &sc->tz_zone.tsp);
435     acpi_tz_getparam(sc, "_TZP", &sc->tz_zone.tzp);
436 
437     /*
438      * Sanity-check the values we've been given.
439      *
440      * XXX what do we do about systems that give us the same value for
441      *     more than one of these setpoints?
442      */
443     acpi_tz_sanity(sc, &sc->tz_zone.crt, "_CRT");
444     acpi_tz_sanity(sc, &sc->tz_zone.hot, "_HOT");
445     acpi_tz_sanity(sc, &sc->tz_zone.psv, "_PSV");
446     for (i = 0; i < TZ_NUMLEVELS; i++)
447 	acpi_tz_sanity(sc, &sc->tz_zone.ac[i], "_ACx");
448 
449     return_VALUE (0);
450 }
451 
452 static char *aclevel_string[] = {
453     "NONE", "_AC0", "_AC1", "_AC2", "_AC3", "_AC4",
454     "_AC5", "_AC6", "_AC7", "_AC8", "_AC9"
455 };
456 
457 static __inline const char *
458 acpi_tz_aclevel_string(int active)
459 {
460     if (active < -1 || active >= TZ_NUMLEVELS)
461 	return (aclevel_string[0]);
462 
463     return (aclevel_string[active + 1]);
464 }
465 
466 /*
467  * Get the current temperature.
468  */
469 static int
470 acpi_tz_get_temperature(struct acpi_tz_softc *sc)
471 {
472     int		temp;
473     ACPI_STATUS	status;
474 
475     ACPI_FUNCTION_NAME ("acpi_tz_get_temperature");
476 
477     /* Evaluate the thermal zone's _TMP method. */
478     status = acpi_GetInteger(sc->tz_handle, acpi_tz_tmp_name, &temp);
479     if (ACPI_FAILURE(status)) {
480 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
481 	    "error fetching current temperature -- %s\n",
482 	     AcpiFormatException(status));
483 	return (FALSE);
484     }
485 
486     /* Check it for validity. */
487     acpi_tz_sanity(sc, &temp, acpi_tz_tmp_name);
488     if (temp == -1)
489 	return (FALSE);
490 
491     ACPI_DEBUG_PRINT((ACPI_DB_VALUES, "got %d.%dC\n", TZ_KELVTOC(temp)));
492     sc->tz_temperature = temp;
493     return (TRUE);
494 }
495 
496 /*
497  * Evaluate the condition of a thermal zone, take appropriate actions.
498  */
499 static void
500 acpi_tz_monitor(void *Context)
501 {
502     struct acpi_tz_softc *sc;
503     struct	timespec curtime;
504     int		temp;
505     int		i;
506     int		newactive, newflags;
507 
508     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
509 
510     sc = (struct acpi_tz_softc *)Context;
511 
512     /* Get the current temperature. */
513     if (!acpi_tz_get_temperature(sc)) {
514 	/* XXX disable zone? go to max cooling? */
515 	return_VOID;
516     }
517     temp = sc->tz_temperature;
518 
519     /*
520      * Work out what we ought to be doing right now.
521      *
522      * Note that the _ACx levels sort from hot to cold.
523      */
524     newactive = TZ_ACTIVE_NONE;
525     for (i = TZ_NUMLEVELS - 1; i >= 0; i--) {
526 	if (sc->tz_zone.ac[i] != -1 && temp >= sc->tz_zone.ac[i])
527 	    newactive = i;
528     }
529 
530     /*
531      * We are going to get _ACx level down (colder side), but give a guaranteed
532      * minimum cooling run time if requested.
533      */
534     if (acpi_tz_min_runtime > 0 && sc->tz_active != TZ_ACTIVE_NONE &&
535 	sc->tz_active != TZ_ACTIVE_UNKNOWN &&
536 	(newactive == TZ_ACTIVE_NONE || newactive > sc->tz_active)) {
537 
538 	getnanotime(&curtime);
539 	timespecsub(&curtime, &sc->tz_cooling_started, &curtime);
540 	if (curtime.tv_sec < acpi_tz_min_runtime)
541 	    newactive = sc->tz_active;
542     }
543 
544     /* Handle user override of active mode */
545     if (sc->tz_requested != TZ_ACTIVE_NONE && (newactive == TZ_ACTIVE_NONE
546         || sc->tz_requested < newactive))
547 	newactive = sc->tz_requested;
548 
549     /* update temperature-related flags */
550     newflags = TZ_THFLAG_NONE;
551     if (sc->tz_zone.psv != -1 && temp >= sc->tz_zone.psv)
552 	newflags |= TZ_THFLAG_PSV;
553     if (sc->tz_zone.hot != -1 && temp >= sc->tz_zone.hot)
554 	newflags |= TZ_THFLAG_HOT;
555     if (sc->tz_zone.crt != -1 && temp >= sc->tz_zone.crt)
556 	newflags |= TZ_THFLAG_CRT;
557 
558     /* If the active cooling state has changed, we have to switch things. */
559     if (sc->tz_active == TZ_ACTIVE_UNKNOWN) {
560 	/*
561 	 * We don't know which cooling device is on or off,
562 	 * so stop them all, because we now know which
563 	 * should be on (if any).
564 	 */
565 	for (i = 0; i < TZ_NUMLEVELS; i++) {
566 	    if (sc->tz_zone.al[i].Pointer != NULL) {
567 		acpi_ForeachPackageObject(
568 		    (ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
569 		    acpi_tz_switch_cooler_off, sc);
570 	    }
571 	}
572 	/* now we know that all devices are off */
573 	sc->tz_active = TZ_ACTIVE_NONE;
574     }
575 
576     if (newactive != sc->tz_active) {
577 	/* Turn off unneeded cooling devices that are on, if any are */
578 	for (i = TZ_ACTIVE_LEVEL(sc->tz_active);
579 	     i < TZ_ACTIVE_LEVEL(newactive); i++) {
580 	    acpi_ForeachPackageObject(
581 		(ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
582 		acpi_tz_switch_cooler_off, sc);
583 	}
584 	/* Turn on cooling devices that are required, if any are */
585 	for (i = TZ_ACTIVE_LEVEL(sc->tz_active) - 1;
586 	     i >= TZ_ACTIVE_LEVEL(newactive); i--) {
587 	    acpi_ForeachPackageObject(
588 		(ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
589 		acpi_tz_switch_cooler_on, sc);
590 	}
591 
592 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
593 		    "switched from %s to %s: %d.%dC\n",
594 		    acpi_tz_aclevel_string(sc->tz_active),
595 		    acpi_tz_aclevel_string(newactive), TZ_KELVTOC(temp));
596 	sc->tz_active = newactive;
597 	getnanotime(&sc->tz_cooling_started);
598     }
599 
600     /* XXX (de)activate any passive cooling that may be required. */
601 
602     /*
603      * If the temperature is at _HOT or _CRT, increment our event count.
604      * If it has occurred enough times, shutdown the system.  This is
605      * needed because some systems will report an invalid high temperature
606      * for one poll cycle.  It is suspected this is due to the embedded
607      * controller timing out.  A typical value is 138C for one cycle on
608      * a system that is otherwise 65C.
609      *
610      * If we're almost at that threshold, notify the user through devd(8).
611      */
612     if ((newflags & (TZ_THFLAG_HOT | TZ_THFLAG_CRT)) != 0) {
613 	sc->tz_validchecks++;
614 	if (sc->tz_validchecks == TZ_VALIDCHECKS) {
615 	    device_printf(sc->tz_dev,
616 		"WARNING - current temperature (%d.%dC) exceeds safe limits\n",
617 		TZ_KELVTOC(sc->tz_temperature));
618 	    shutdown_nice(RB_POWEROFF);
619 	} else if (sc->tz_validchecks == TZ_NOTIFYCOUNT)
620 	    acpi_UserNotify("Thermal", sc->tz_handle, TZ_NOTIFY_CRITICAL);
621     } else {
622 	sc->tz_validchecks = 0;
623     }
624     sc->tz_thflags = newflags;
625 
626     return_VOID;
627 }
628 
629 /*
630  * Given an object, verify that it's a reference to a device of some sort,
631  * and try to switch it off.
632  */
633 static void
634 acpi_tz_switch_cooler_off(ACPI_OBJECT *obj, void *arg)
635 {
636     ACPI_HANDLE			cooler;
637 
638     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
639 
640     cooler = acpi_GetReference(NULL, obj);
641     if (cooler == NULL) {
642 	ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "can't get handle\n"));
643 	return_VOID;
644     }
645 
646     ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "called to turn %s off\n",
647 		     acpi_name(cooler)));
648     acpi_pwr_switch_consumer(cooler, ACPI_STATE_D3);
649 
650     return_VOID;
651 }
652 
653 /*
654  * Given an object, verify that it's a reference to a device of some sort,
655  * and try to switch it on.
656  *
657  * XXX replication of off/on function code is bad.
658  */
659 static void
660 acpi_tz_switch_cooler_on(ACPI_OBJECT *obj, void *arg)
661 {
662     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)arg;
663     ACPI_HANDLE			cooler;
664     ACPI_STATUS			status;
665 
666     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
667 
668     cooler = acpi_GetReference(NULL, obj);
669     if (cooler == NULL) {
670 	ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "can't get handle\n"));
671 	return_VOID;
672     }
673 
674     ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "called to turn %s on\n",
675 		     acpi_name(cooler)));
676     status = acpi_pwr_switch_consumer(cooler, ACPI_STATE_D0);
677     if (ACPI_FAILURE(status)) {
678 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
679 		    "failed to activate %s - %s\n", acpi_name(cooler),
680 		    AcpiFormatException(status));
681     }
682 
683     return_VOID;
684 }
685 
686 /*
687  * Read/debug-print a parameter, default it to -1.
688  */
689 static void
690 acpi_tz_getparam(struct acpi_tz_softc *sc, char *node, int *data)
691 {
692 
693     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
694 
695     if (ACPI_FAILURE(acpi_GetInteger(sc->tz_handle, node, data))) {
696 	*data = -1;
697     } else {
698 	ACPI_DEBUG_PRINT((ACPI_DB_VALUES, "%s.%s = %d\n",
699 			 acpi_name(sc->tz_handle), node, *data));
700     }
701 
702     return_VOID;
703 }
704 
705 /*
706  * Sanity-check a temperature value.  Assume that setpoints
707  * should be between 0C and 200C.
708  */
709 static void
710 acpi_tz_sanity(struct acpi_tz_softc *sc, int *val, char *what)
711 {
712     if (*val != -1 && (*val < TZ_ZEROC || *val > TZ_ZEROC + 2000)) {
713 	/*
714 	 * If the value we are checking is _TMP, warn the user only
715 	 * once. This avoids spamming messages if, for instance, the
716 	 * sensor is broken and always returns an invalid temperature.
717 	 *
718 	 * This is only done for _TMP; other values always emit a
719 	 * warning.
720 	 */
721 	if (what != acpi_tz_tmp_name || !sc->tz_insane_tmp_notified) {
722 	    device_printf(sc->tz_dev, "%s value is absurd, ignored (%d.%dC)\n",
723 			  what, TZ_KELVTOC(*val));
724 
725 	    /* Don't warn the user again if the read value doesn't improve. */
726 	    if (what == acpi_tz_tmp_name)
727 		sc->tz_insane_tmp_notified = 1;
728 	}
729 	*val = -1;
730 	return;
731     }
732 
733     /* This value is correct. Warn if it's incorrect again. */
734     if (what == acpi_tz_tmp_name)
735 	sc->tz_insane_tmp_notified = 0;
736 }
737 
738 /*
739  * Respond to a sysctl on the active state node.
740  */
741 static int
742 acpi_tz_active_sysctl(SYSCTL_HANDLER_ARGS)
743 {
744     struct acpi_tz_softc	*sc;
745     int				active;
746     int		 		error;
747 
748     sc = (struct acpi_tz_softc *)oidp->oid_arg1;
749     active = sc->tz_active;
750     error = sysctl_handle_int(oidp, &active, 0, req);
751 
752     /* Error or no new value */
753     if (error != 0 || req->newptr == NULL)
754 	return (error);
755     if (active < -1 || active >= TZ_NUMLEVELS)
756 	return (EINVAL);
757 
758     /* Set new preferred level and re-switch */
759     sc->tz_requested = active;
760     acpi_tz_signal(sc, 0);
761     return (0);
762 }
763 
764 static int
765 acpi_tz_cooling_sysctl(SYSCTL_HANDLER_ARGS)
766 {
767     struct acpi_tz_softc *sc;
768     int enabled, error;
769 
770     sc = (struct acpi_tz_softc *)oidp->oid_arg1;
771     enabled = sc->tz_cooling_enabled;
772     error = sysctl_handle_int(oidp, &enabled, 0, req);
773 
774     /* Error or no new value */
775     if (error != 0 || req->newptr == NULL)
776 	return (error);
777     if (enabled != TRUE && enabled != FALSE)
778 	return (EINVAL);
779 
780     if (enabled) {
781 	if (acpi_tz_cooling_is_available(sc))
782 	    error = acpi_tz_cooling_thread_start(sc);
783 	else
784 	    error = ENODEV;
785 	if (error)
786 	    enabled = FALSE;
787     }
788     sc->tz_cooling_enabled = enabled;
789     return (error);
790 }
791 
792 static int
793 acpi_tz_temp_sysctl(SYSCTL_HANDLER_ARGS)
794 {
795     struct acpi_tz_softc	*sc;
796     int				temp, *temp_ptr;
797     int		 		error;
798 
799     sc = oidp->oid_arg1;
800     temp_ptr = (int *)(void *)(uintptr_t)((uintptr_t)sc + oidp->oid_arg2);
801     temp = *temp_ptr;
802     error = sysctl_handle_int(oidp, &temp, 0, req);
803 
804     /* Error or no new value */
805     if (error != 0 || req->newptr == NULL)
806 	return (error);
807 
808     /* Only allow changing settings if override is set. */
809     if (!acpi_tz_override)
810 	return (EPERM);
811 
812     /* Check user-supplied value for sanity. */
813     acpi_tz_sanity(sc, &temp, "user-supplied temp");
814     if (temp == -1)
815 	return (EINVAL);
816 
817     *temp_ptr = temp;
818     return (0);
819 }
820 
821 static int
822 acpi_tz_passive_sysctl(SYSCTL_HANDLER_ARGS)
823 {
824     struct acpi_tz_softc	*sc;
825     int				val, *val_ptr;
826     int				error;
827 
828     sc = oidp->oid_arg1;
829     val_ptr = (int *)(void *)(uintptr_t)((uintptr_t)sc + oidp->oid_arg2);
830     val = *val_ptr;
831     error = sysctl_handle_int(oidp, &val, 0, req);
832 
833     /* Error or no new value */
834     if (error != 0 || req->newptr == NULL)
835 	return (error);
836 
837     /* Only allow changing settings if override is set. */
838     if (!acpi_tz_override)
839 	return (EPERM);
840 
841     *val_ptr = val;
842     return (0);
843 }
844 
845 static void
846 acpi_tz_notify_handler(ACPI_HANDLE h, UINT32 notify, void *context)
847 {
848     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)context;
849 
850     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
851 
852     switch (notify) {
853     case TZ_NOTIFY_TEMPERATURE:
854 	/* Temperature change occurred */
855 	acpi_tz_signal(sc, 0);
856 	break;
857     case TZ_NOTIFY_DEVICES:
858     case TZ_NOTIFY_LEVELS:
859 	/* Zone devices/setpoints changed */
860 	acpi_tz_signal(sc, TZ_FLAG_GETSETTINGS);
861 	break;
862     default:
863 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
864 		    "unknown Notify event 0x%x\n", notify);
865 	break;
866     }
867 
868     acpi_UserNotify("Thermal", h, notify);
869 
870     return_VOID;
871 }
872 
873 static void
874 acpi_tz_signal(struct acpi_tz_softc *sc, int flags)
875 {
876     ACPI_LOCK(thermal);
877     sc->tz_flags |= flags;
878     ACPI_UNLOCK(thermal);
879     wakeup(&acpi_tz_proc);
880 }
881 
882 /*
883  * Notifies can be generated asynchronously but have also been seen to be
884  * triggered by other thermal methods.  One system generates a notify of
885  * 0x81 when the fan is turned on or off.  Another generates it when _SCP
886  * is called.  To handle these situations, we check the zone via
887  * acpi_tz_monitor() before evaluating changes to setpoints or the cooling
888  * policy.
889  */
890 static void
891 acpi_tz_timeout(struct acpi_tz_softc *sc, int flags)
892 {
893 
894     /* Check the current temperature and take action based on it */
895     acpi_tz_monitor(sc);
896 
897     /* If requested, get the power profile settings. */
898     if (flags & TZ_FLAG_GETPROFILE)
899 	acpi_tz_power_profile(sc);
900 
901     /*
902      * If requested, check for new devices/setpoints.  After finding them,
903      * check if we need to switch fans based on the new values.
904      */
905     if (flags & TZ_FLAG_GETSETTINGS) {
906 	acpi_tz_establish(sc);
907 	acpi_tz_monitor(sc);
908     }
909 
910     /* XXX passive cooling actions? */
911 }
912 
913 /*
914  * System power profile may have changed; fetch and notify the
915  * thermal zone accordingly.
916  *
917  * Since this can be called from an arbitrary eventhandler, it needs
918  * to get the ACPI lock itself.
919  */
920 static void
921 acpi_tz_power_profile(void *arg)
922 {
923     ACPI_STATUS			status;
924     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)arg;
925     int				state;
926 
927     state = power_profile_get_state();
928     if (state != POWER_PROFILE_PERFORMANCE && state != POWER_PROFILE_ECONOMY)
929 	return;
930 
931     /* check that we haven't decided there's no _SCP method */
932     if ((sc->tz_flags & TZ_FLAG_NO_SCP) == 0) {
933 
934 	/* Call _SCP to set the new profile */
935 	status = acpi_SetInteger(sc->tz_handle, "_SCP",
936 	    (state == POWER_PROFILE_PERFORMANCE) ? 0 : 1);
937 	if (ACPI_FAILURE(status)) {
938 	    if (status != AE_NOT_FOUND)
939 		ACPI_VPRINT(sc->tz_dev,
940 			    acpi_device_get_parent_softc(sc->tz_dev),
941 			    "can't evaluate %s._SCP - %s\n",
942 			    acpi_name(sc->tz_handle),
943 			    AcpiFormatException(status));
944 	    sc->tz_flags |= TZ_FLAG_NO_SCP;
945 	} else {
946 	    /* We have to re-evaluate the entire zone now */
947 	    acpi_tz_signal(sc, TZ_FLAG_GETSETTINGS);
948 	}
949     }
950 }
951 
952 /*
953  * Thermal zone monitor thread.
954  */
955 static void
956 acpi_tz_thread(void *arg)
957 {
958     device_t	*devs;
959     int		devcount, i;
960     int		flags;
961     struct acpi_tz_softc **sc;
962 
963     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
964 
965     devs = NULL;
966     devcount = 0;
967     sc = NULL;
968 
969     for (;;) {
970 	/* If the number of devices has changed, re-evaluate. */
971 	if (devclass_get_count(acpi_tz_devclass) != devcount) {
972 	    if (devs != NULL) {
973 		free(devs, M_TEMP);
974 		free(sc, M_TEMP);
975 	    }
976 	    devclass_get_devices(acpi_tz_devclass, &devs, &devcount);
977 	    sc = malloc(sizeof(struct acpi_tz_softc *) * devcount, M_TEMP,
978 			M_WAITOK | M_ZERO);
979 	    for (i = 0; i < devcount; i++)
980 		sc[i] = device_get_softc(devs[i]);
981 	}
982 
983 	/* Check for temperature events and act on them. */
984 	for (i = 0; i < devcount; i++) {
985 	    ACPI_LOCK(thermal);
986 	    flags = sc[i]->tz_flags;
987 	    sc[i]->tz_flags &= TZ_FLAG_NO_SCP;
988 	    ACPI_UNLOCK(thermal);
989 	    acpi_tz_timeout(sc[i], flags);
990 	}
991 
992 	/* If more work to do, don't go to sleep yet. */
993 	ACPI_LOCK(thermal);
994 	for (i = 0; i < devcount; i++) {
995 	    if (sc[i]->tz_flags & ~TZ_FLAG_NO_SCP)
996 		break;
997 	}
998 
999 	/*
1000 	 * If we have no more work, sleep for a while, setting PDROP so that
1001 	 * the mutex will not be reacquired.  Otherwise, drop the mutex and
1002 	 * loop to handle more events.
1003 	 */
1004 	if (i == devcount)
1005 	    msleep(&acpi_tz_proc, &thermal_mutex, PZERO | PDROP, "tzpoll",
1006 		hz * acpi_tz_polling_rate);
1007 	else
1008 	    ACPI_UNLOCK(thermal);
1009     }
1010 }
1011 
1012 static int
1013 acpi_tz_cpufreq_restore(struct acpi_tz_softc *sc)
1014 {
1015     device_t dev;
1016     int error;
1017 
1018     if (!sc->tz_cooling_updated)
1019 	return (0);
1020     if ((dev = devclass_get_device(devclass_find("cpufreq"), 0)) == NULL)
1021 	return (ENXIO);
1022     ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
1023 	"temperature %d.%dC: resuming previous clock speed (%d MHz)\n",
1024 	TZ_KELVTOC(sc->tz_temperature), sc->tz_cooling_saved_freq);
1025     error = CPUFREQ_SET(dev, NULL, CPUFREQ_PRIO_KERN);
1026     if (error == 0)
1027 	sc->tz_cooling_updated = FALSE;
1028     return (error);
1029 }
1030 
1031 static int
1032 acpi_tz_cpufreq_update(struct acpi_tz_softc *sc, int req)
1033 {
1034     device_t dev;
1035     struct cf_level *levels;
1036     int num_levels, error, freq, desired_freq, perf, i;
1037 
1038     levels = malloc(CPUFREQ_MAX_LEVELS * sizeof(*levels), M_TEMP, M_NOWAIT);
1039     if (levels == NULL)
1040 	return (ENOMEM);
1041 
1042     /*
1043      * Find the main device, cpufreq0.  We don't yet support independent
1044      * CPU frequency control on SMP.
1045      */
1046     if ((dev = devclass_get_device(devclass_find("cpufreq"), 0)) == NULL) {
1047 	error = ENXIO;
1048 	goto out;
1049     }
1050 
1051     /* Get the current frequency. */
1052     error = CPUFREQ_GET(dev, &levels[0]);
1053     if (error)
1054 	goto out;
1055     freq = levels[0].total_set.freq;
1056 
1057     /* Get the current available frequency levels. */
1058     num_levels = CPUFREQ_MAX_LEVELS;
1059     error = CPUFREQ_LEVELS(dev, levels, &num_levels);
1060     if (error) {
1061 	if (error == E2BIG)
1062 	    printf("cpufreq: need to increase CPUFREQ_MAX_LEVELS\n");
1063 	goto out;
1064     }
1065 
1066     /* Calculate the desired frequency as a percent of the max frequency. */
1067     perf = 100 * freq / levels[0].total_set.freq - req;
1068     if (perf < 0)
1069 	perf = 0;
1070     else if (perf > 100)
1071 	perf = 100;
1072     desired_freq = levels[0].total_set.freq * perf / 100;
1073 
1074     if (desired_freq < freq) {
1075 	/* Find the closest available frequency, rounding down. */
1076 	for (i = 0; i < num_levels; i++)
1077 	    if (levels[i].total_set.freq <= desired_freq)
1078 		break;
1079 
1080 	/* If we didn't find a relevant setting, use the lowest. */
1081 	if (i == num_levels)
1082 	    i--;
1083     } else {
1084 	/* If we didn't decrease frequency yet, don't increase it. */
1085 	if (!sc->tz_cooling_updated) {
1086 	    sc->tz_cooling_active = FALSE;
1087 	    goto out;
1088 	}
1089 
1090 	/* Use saved cpu frequency as maximum value. */
1091 	if (desired_freq > sc->tz_cooling_saved_freq)
1092 	    desired_freq = sc->tz_cooling_saved_freq;
1093 
1094 	/* Find the closest available frequency, rounding up. */
1095 	for (i = num_levels - 1; i >= 0; i--)
1096 	    if (levels[i].total_set.freq >= desired_freq)
1097 		break;
1098 
1099 	/* If we didn't find a relevant setting, use the highest. */
1100 	if (i == -1)
1101 	    i++;
1102 
1103 	/* If we're going to the highest frequency, restore the old setting. */
1104 	if (i == 0 || desired_freq == sc->tz_cooling_saved_freq) {
1105 	    error = acpi_tz_cpufreq_restore(sc);
1106 	    if (error == 0)
1107 		sc->tz_cooling_active = FALSE;
1108 	    goto out;
1109 	}
1110     }
1111 
1112     /* If we are going to a new frequency, activate it. */
1113     if (levels[i].total_set.freq != freq) {
1114 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
1115 	    "temperature %d.%dC: %screasing clock speed "
1116 	    "from %d MHz to %d MHz\n",
1117 	    TZ_KELVTOC(sc->tz_temperature),
1118 	    (freq > levels[i].total_set.freq) ? "de" : "in",
1119 	    freq, levels[i].total_set.freq);
1120 	error = CPUFREQ_SET(dev, &levels[i], CPUFREQ_PRIO_KERN);
1121 	if (error == 0 && !sc->tz_cooling_updated) {
1122 	    sc->tz_cooling_saved_freq = freq;
1123 	    sc->tz_cooling_updated = TRUE;
1124 	}
1125     }
1126 
1127 out:
1128     if (levels)
1129 	free(levels, M_TEMP);
1130     return (error);
1131 }
1132 
1133 /*
1134  * Passive cooling thread; monitors current temperature according to the
1135  * cooling interval and calculates whether to scale back CPU frequency.
1136  */
1137 static void
1138 acpi_tz_cooling_thread(void *arg)
1139 {
1140     struct acpi_tz_softc *sc;
1141     int error, perf, curr_temp, prev_temp;
1142 
1143     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
1144 
1145     sc = (struct acpi_tz_softc *)arg;
1146 
1147     prev_temp = sc->tz_temperature;
1148     while (sc->tz_cooling_enabled) {
1149 	if (sc->tz_cooling_active)
1150 	    (void)acpi_tz_get_temperature(sc);
1151 	curr_temp = sc->tz_temperature;
1152 	if (curr_temp >= sc->tz_zone.psv)
1153 	    sc->tz_cooling_active = TRUE;
1154 	if (sc->tz_cooling_active) {
1155 	    perf = sc->tz_zone.tc1 * (curr_temp - prev_temp) +
1156 		   sc->tz_zone.tc2 * (curr_temp - sc->tz_zone.psv);
1157 	    perf /= 10;
1158 
1159 	    if (perf != 0) {
1160 		error = acpi_tz_cpufreq_update(sc, perf);
1161 
1162 		/*
1163 		 * If error and not simply a higher priority setting was
1164 		 * active, disable cooling.
1165 		 */
1166 		if (error != 0 && error != EPERM) {
1167 		    device_printf(sc->tz_dev,
1168 			"failed to set new freq, disabling passive cooling\n");
1169 		    sc->tz_cooling_enabled = FALSE;
1170 		}
1171 	    }
1172 	}
1173 	prev_temp = curr_temp;
1174 	tsleep(&sc->tz_cooling_proc, PZERO, "cooling",
1175 	    hz * sc->tz_zone.tsp / 10);
1176     }
1177     if (sc->tz_cooling_active) {
1178 	acpi_tz_cpufreq_restore(sc);
1179 	sc->tz_cooling_active = FALSE;
1180     }
1181     sc->tz_cooling_proc = NULL;
1182     ACPI_LOCK(thermal);
1183     sc->tz_cooling_proc_running = FALSE;
1184     ACPI_UNLOCK(thermal);
1185     kproc_exit(0);
1186 }
1187 
1188 /*
1189  * TODO: We ignore _PSL (list of cooling devices) since cpufreq enumerates
1190  * all CPUs for us.  However, it's possible in the future _PSL will
1191  * reference non-CPU devices so we may want to support it then.
1192  */
1193 static int
1194 acpi_tz_cooling_is_available(struct acpi_tz_softc *sc)
1195 {
1196     return (sc->tz_zone.tc1 != -1 && sc->tz_zone.tc2 != -1 &&
1197 	sc->tz_zone.tsp != -1 && sc->tz_zone.tsp != 0 &&
1198 	sc->tz_zone.psv != -1);
1199 }
1200 
1201 static int
1202 acpi_tz_cooling_thread_start(struct acpi_tz_softc *sc)
1203 {
1204     int error;
1205 
1206     ACPI_LOCK(thermal);
1207     if (sc->tz_cooling_proc_running) {
1208 	ACPI_UNLOCK(thermal);
1209 	return (0);
1210     }
1211     sc->tz_cooling_proc_running = TRUE;
1212     ACPI_UNLOCK(thermal);
1213     error = 0;
1214     if (sc->tz_cooling_proc == NULL) {
1215 	error = kproc_create(acpi_tz_cooling_thread, sc,
1216 	    &sc->tz_cooling_proc, RFHIGHPID, 0, "acpi_cooling%d",
1217 	    device_get_unit(sc->tz_dev));
1218 	if (error != 0) {
1219 	    device_printf(sc->tz_dev, "could not create thread - %d", error);
1220 	    ACPI_LOCK(thermal);
1221 	    sc->tz_cooling_proc_running = FALSE;
1222 	    ACPI_UNLOCK(thermal);
1223 	}
1224     }
1225     return (error);
1226 }
1227