xref: /dragonfly/sys/dev/acpica/acpi_thermal.c (revision 117e566d)
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  * $FreeBSD: head/sys/dev/acpica/acpi_thermal.c 255077 2013-08-30 19:21:12Z dumbbell $
28  */
29 
30 #include "opt_acpi.h"
31 #include <sys/param.h>
32 #include <sys/kernel.h>
33 #include <sys/bus.h>
34 #include <sys/kthread.h>
35 #include <sys/malloc.h>
36 #include <sys/module.h>
37 #include <sys/proc.h>
38 #include <sys/reboot.h>
39 #include <sys/sysctl.h>
40 #include <sys/unistd.h>
41 #include <sys/power.h>
42 #include <sys/sensors.h>
43 
44 #include "acpi.h"
45 #include "accommon.h"
46 
47 #include <dev/acpica/acpivar.h>
48 
49 /* Hooks for the ACPICA debugging infrastructure */
50 #define _COMPONENT	ACPI_THERMAL
51 ACPI_MODULE_NAME("THERMAL")
52 
53 #define TZ_ZEROC	2732
54 #define TZ_KELVTOC(x)	(((x) - TZ_ZEROC) / 10), abs(((x) - TZ_ZEROC) % 10)
55 
56 #define TZ_NOTIFY_TEMPERATURE	0x80 /* Temperature changed. */
57 #define TZ_NOTIFY_LEVELS	0x81 /* Cooling levels changed. */
58 #define TZ_NOTIFY_DEVICES	0x82 /* Device lists changed. */
59 #define TZ_NOTIFY_CRITICAL	0xcc /* Fake notify that _CRT/_HOT reached. */
60 
61 /* Check for temperature changes every 10 seconds by default */
62 #define TZ_POLLRATE	10
63 
64 /* Make sure the reported temperature is valid for this number of polls. */
65 #define TZ_VALIDCHECKS	3
66 
67 /* Notify the user we will be shutting down in one more poll cycle. */
68 #define TZ_NOTIFYCOUNT	(TZ_VALIDCHECKS - 1)
69 
70 /* ACPI spec defines this */
71 #define TZ_NUMLEVELS	10
72 struct acpi_tz_zone {
73     int		ac[TZ_NUMLEVELS];
74     ACPI_BUFFER	al[TZ_NUMLEVELS];
75     int		crt;
76     int		hot;
77     ACPI_BUFFER	psl;
78     int		psv;
79     int		tc1;
80     int		tc2;
81     int		tsp;
82     int		tzp;
83 };
84 
85 struct acpi_tz_softc {
86     device_t			tz_dev;
87     ACPI_HANDLE			tz_handle;	/*Thermal zone handle*/
88     int				tz_temperature;	/*Current temperature*/
89     int				tz_active;	/*Current active cooling*/
90 #define TZ_ACTIVE_NONE		-1
91 #define TZ_ACTIVE_UNKNOWN	-2
92     int				tz_requested;	/*Minimum active cooling*/
93     int				tz_thflags;	/*Current temp-related flags*/
94 #define TZ_THFLAG_NONE		0
95 #define TZ_THFLAG_PSV		(1<<0)
96 #define TZ_THFLAG_HOT		(1<<2)
97 #define TZ_THFLAG_CRT		(1<<3)
98     int				tz_flags;
99 #define TZ_FLAG_NO_SCP		(1<<0)		/*No _SCP method*/
100 #define TZ_FLAG_GETPROFILE	(1<<1)		/*Get power_profile in timeout*/
101 #define TZ_FLAG_GETSETTINGS	(1<<2)		/*Get devs/setpoints*/
102     struct timespec		tz_cooling_started;
103 					/*Current cooling starting time*/
104 
105     struct sysctl_ctx_list	tz_sysctl_ctx;
106     struct sysctl_oid		*tz_sysctl_tree;
107     eventhandler_tag		tz_event;
108 
109     struct acpi_tz_zone 	tz_zone;	/*Thermal zone parameters*/
110     time_t			tz_error_time;	/*Lookup error timestamp*/
111     int				tz_validchecks;
112     int				tz_insane_tmp_notified;
113 
114     /* passive cooling */
115     struct thread		*tz_cooling_proc;
116     int				tz_cooling_proc_running;
117     int				tz_cooling_enabled;
118     int				tz_cooling_active;
119     int				tz_cooling_updated;
120     int				tz_cooling_saved_freq;
121     /* sensors(9) related */
122     struct ksensordev		sensordev;
123     struct ksensor		sensor;
124 };
125 
126 /* silence errors after X seconds, try again after Y seconds */
127 #define TZ_SILENCE_ERROR						\
128     ((acpi_tz_polling_rate <= 0 ? TZ_POLLRATE : acpi_tz_polling_rate) * 2 + 1)
129 #define TZ_RETRY_ERROR		7200
130 
131 #define	TZ_ACTIVE_LEVEL(act)	((act) >= 0 ? (act) : TZ_NUMLEVELS)
132 
133 #define CPUFREQ_MAX_LEVELS	64 /* XXX cpufreq should export this */
134 
135 static int	acpi_tz_probe(device_t dev);
136 static int	acpi_tz_attach(device_t dev);
137 static int	acpi_tz_establish(struct acpi_tz_softc *sc);
138 static void	acpi_tz_monitor(void *Context);
139 static void	acpi_tz_switch_cooler_off(ACPI_OBJECT *obj, void *arg);
140 static void	acpi_tz_switch_cooler_on(ACPI_OBJECT *obj, void *arg);
141 static void	acpi_tz_getparam(struct acpi_tz_softc *sc, char *node,
142 				 int *data);
143 static void	acpi_tz_sanity(struct acpi_tz_softc *sc, int *val, char *what);
144 static int	acpi_tz_polling_sysctl(SYSCTL_HANDLER_ARGS);
145 static int	acpi_tz_active_sysctl(SYSCTL_HANDLER_ARGS);
146 static int	acpi_tz_cooling_sysctl(SYSCTL_HANDLER_ARGS);
147 static int	acpi_tz_temp_sysctl(SYSCTL_HANDLER_ARGS);
148 static int	acpi_tz_passive_sysctl(SYSCTL_HANDLER_ARGS);
149 static void	acpi_tz_notify_handler(ACPI_HANDLE h, UINT32 notify,
150 				       void *context);
151 static void	acpi_tz_signal(struct acpi_tz_softc *sc, int flags);
152 static void	acpi_tz_timeout(struct acpi_tz_softc *sc, int flags);
153 static void	acpi_tz_power_profile(void *arg);
154 static void	acpi_tz_thread(void *arg);
155 static int	acpi_tz_cooling_is_available(struct acpi_tz_softc *sc);
156 static int	acpi_tz_cooling_thread_start(struct acpi_tz_softc *sc);
157 
158 static device_method_t acpi_tz_methods[] = {
159     /* Device interface */
160     DEVMETHOD(device_probe,	acpi_tz_probe),
161     DEVMETHOD(device_attach,	acpi_tz_attach),
162 
163     DEVMETHOD_END
164 };
165 
166 static driver_t acpi_tz_driver = {
167     "acpi_tz",
168     acpi_tz_methods,
169     sizeof(struct acpi_tz_softc),
170     .gpri = KOBJ_GPRI_ACPI
171 };
172 
173 static char *acpi_tz_tmp_name = "_TMP";
174 
175 static devclass_t acpi_tz_devclass;
176 DRIVER_MODULE(acpi_tz, acpi, acpi_tz_driver, acpi_tz_devclass, NULL, NULL);
177 MODULE_DEPEND(acpi_tz, acpi, 1, 1, 1);
178 
179 static struct sysctl_ctx_list	acpi_tz_sysctl_ctx;
180 static struct sysctl_oid	*acpi_tz_sysctl_tree;
181 
182 /* Minimum cooling run time */
183 static int			acpi_tz_min_runtime;
184 static int			acpi_tz_polling_rate = TZ_POLLRATE;
185 static int			acpi_tz_override;
186 
187 /* Timezone polling thread */
188 static struct thread		*acpi_tz_td;
189 ACPI_LOCK_DECL(thermal, "ACPI thermal zone");
190 
191 static int			acpi_tz_cooling_unit = -1;
192 
193 static int
194 acpi_tz_probe(device_t dev)
195 {
196     int		result;
197 
198     if (acpi_get_type(dev) == ACPI_TYPE_THERMAL && !acpi_disabled("thermal")) {
199 	device_set_desc(dev, "Thermal Zone");
200 	result = -10;
201     } else
202 	result = ENXIO;
203     return (result);
204 }
205 
206 static int
207 acpi_tz_attach(device_t dev)
208 {
209     struct acpi_tz_softc	*sc;
210     struct acpi_softc		*acpi_sc;
211     int				error;
212     char			oidname[8];
213 
214     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
215     if (device_get_unit(dev) == 0)
216 	ACPI_LOCK_INIT(thermal, "acpitz");
217 
218     ACPI_LOCK(thermal);
219 
220     sc = device_get_softc(dev);
221     sc->tz_dev = dev;
222     sc->tz_handle = acpi_get_handle(dev);
223     sc->tz_requested = TZ_ACTIVE_NONE;
224     sc->tz_active = TZ_ACTIVE_UNKNOWN;
225     sc->tz_thflags = TZ_THFLAG_NONE;
226     sc->tz_cooling_proc = NULL;
227     sc->tz_cooling_proc_running = FALSE;
228     sc->tz_cooling_active = FALSE;
229     sc->tz_cooling_updated = FALSE;
230     sc->tz_cooling_enabled = FALSE;
231 
232     /*
233      * Parse the current state of the thermal zone and build control
234      * structures.  We don't need to worry about interference with the
235      * control thread since we haven't fully attached this device yet.
236      */
237     if ((error = acpi_tz_establish(sc)) != 0) {
238 	ACPI_UNLOCK(thermal);
239 	return (error);
240     }
241 
242     /*
243      * Register for any Notify events sent to this zone.
244      */
245     AcpiInstallNotifyHandler(sc->tz_handle, ACPI_DEVICE_NOTIFY,
246 			     acpi_tz_notify_handler, sc);
247 
248     /*
249      * Create our sysctl nodes.
250      *
251      * XXX we need a mechanism for adding nodes under ACPI.
252      */
253     if (device_get_unit(dev) == 0) {
254 	acpi_sc = acpi_device_get_parent_softc(dev);
255 	sysctl_ctx_init(&acpi_tz_sysctl_ctx);
256 	acpi_tz_sysctl_tree = SYSCTL_ADD_NODE(&acpi_tz_sysctl_ctx,
257 			      SYSCTL_CHILDREN(acpi_sc->acpi_sysctl_tree),
258 			      OID_AUTO, "thermal", CTLFLAG_RD, 0, "");
259 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
260 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
261 		       OID_AUTO, "min_runtime", CTLFLAG_RW,
262 		       &acpi_tz_min_runtime, 0,
263 		       "minimum cooling run time in sec");
264 	SYSCTL_ADD_PROC(&acpi_tz_sysctl_ctx,
265 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
266 		       OID_AUTO, "polling_rate", CTLTYPE_INT | CTLFLAG_RW,
267 		       &acpi_tz_polling_rate, 0, acpi_tz_polling_sysctl,
268 		       "I", "monitor polling interval in seconds");
269 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
270 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree), OID_AUTO,
271 		       "user_override", CTLFLAG_RW, &acpi_tz_override, 0,
272 		       "allow override of thermal settings");
273     }
274     sysctl_ctx_init(&sc->tz_sysctl_ctx);
275     ksprintf(oidname, "tz%d", device_get_unit(dev));
276     sc->tz_sysctl_tree = SYSCTL_ADD_NODE(&sc->tz_sysctl_ctx,
277 					 SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
278 					 OID_AUTO, oidname, CTLFLAG_RD, 0, "");
279     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
280 		    OID_AUTO, "temperature", CTLTYPE_INT | CTLFLAG_RD,
281 		    &sc->tz_temperature, 0, sysctl_handle_int,
282 		    "IK", "current thermal zone temperature");
283     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
284 		    OID_AUTO, "active", CTLTYPE_INT | CTLFLAG_RW,
285 		    sc, 0, acpi_tz_active_sysctl, "I", "cooling is active");
286     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
287 		    OID_AUTO, "passive_cooling", CTLTYPE_INT | CTLFLAG_RW,
288 		    sc, 0, acpi_tz_cooling_sysctl, "I",
289 		    "enable passive (speed reduction) cooling");
290 
291     SYSCTL_ADD_INT(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
292 		   OID_AUTO, "thermal_flags", CTLFLAG_RD,
293 		   &sc->tz_thflags, 0, "thermal zone flags");
294     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
295 		    OID_AUTO, "_PSV", CTLTYPE_INT | CTLFLAG_RW,
296 		    sc, offsetof(struct acpi_tz_softc, tz_zone.psv),
297 		    acpi_tz_temp_sysctl, "IK", "passive cooling temp setpoint");
298     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
299 		    OID_AUTO, "_HOT", CTLTYPE_INT | CTLFLAG_RW,
300 		    sc, offsetof(struct acpi_tz_softc, tz_zone.hot),
301 		    acpi_tz_temp_sysctl, "IK",
302 		    "too hot temp setpoint (suspend now)");
303     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
304 		    OID_AUTO, "_CRT", CTLTYPE_INT | CTLFLAG_RW,
305 		    sc, offsetof(struct acpi_tz_softc, tz_zone.crt),
306 		    acpi_tz_temp_sysctl, "IK",
307 		    "critical temp setpoint (shutdown now)");
308     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
309 		    OID_AUTO, "_ACx", CTLTYPE_INT | CTLFLAG_RD,
310 		    &sc->tz_zone.ac, sizeof(sc->tz_zone.ac),
311 		    sysctl_handle_opaque, "IK", "");
312     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
313 		    OID_AUTO, "_TC1", CTLTYPE_INT | CTLFLAG_RW,
314 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tc1),
315 		    acpi_tz_passive_sysctl, "I",
316 		    "thermal constant 1 for passive cooling");
317     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
318 		    OID_AUTO, "_TC2", CTLTYPE_INT | CTLFLAG_RW,
319 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tc2),
320 		    acpi_tz_passive_sysctl, "I",
321 		    "thermal constant 2 for passive cooling");
322     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
323 		    OID_AUTO, "_TSP", CTLTYPE_INT | CTLFLAG_RW,
324 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tsp),
325 		    acpi_tz_passive_sysctl, "I",
326 		    "thermal sampling period for passive cooling");
327 
328     /*
329      * Create thread to service all of the thermal zones.  Register
330      * our power profile event handler.
331      */
332     sc->tz_event = EVENTHANDLER_REGISTER(power_profile_change,
333 					 acpi_tz_power_profile, sc, 0);
334     if (acpi_tz_td == NULL) {
335 	error = kthread_create(acpi_tz_thread, NULL, &acpi_tz_td,
336 	    "acpi_thermal");
337 	if (error != 0) {
338 	    device_printf(sc->tz_dev, "could not create thread - %d", error);
339 	    goto out;
340 	}
341     }
342 
343     /*
344      * Create a thread to handle passive cooling for 1st zone which
345      * has _PSV, _TSP, _TC1 and _TC2.  Users can enable it for other
346      * zones manually for now.
347      *
348      * XXX We enable only one zone to avoid multiple zones conflict
349      * with each other since cpufreq currently sets all CPUs to the
350      * given frequency whereas it's possible for different thermal
351      * zones to specify independent settings for multiple CPUs.
352      */
353     if (acpi_tz_cooling_unit < 0 && acpi_tz_cooling_is_available(sc))
354 	sc->tz_cooling_enabled = TRUE;
355     if (sc->tz_cooling_enabled) {
356 	error = acpi_tz_cooling_thread_start(sc);
357 	if (error != 0) {
358 	    sc->tz_cooling_enabled = FALSE;
359 	    goto out;
360 	}
361 	acpi_tz_cooling_unit = device_get_unit(dev);
362     }
363 
364     /*
365      * Flag the event handler for a manual invocation by our timeout.
366      * We defer it like this so that the rest of the subsystem has time
367      * to come up.  Don't bother evaluating/printing the temperature at
368      * this point; on many systems it'll be bogus until the EC is running.
369      */
370     sc->tz_flags |= TZ_FLAG_GETPROFILE;
371 
372     /* Attach sensors(9). */
373     strlcpy(sc->sensordev.xname, device_get_nameunit(sc->tz_dev),
374         sizeof(sc->sensordev.xname));
375 
376     sc->sensor.type = SENSOR_TEMP;
377     sensor_attach(&sc->sensordev, &sc->sensor);
378 
379     sensordev_install(&sc->sensordev);
380 
381 out:
382     if (error != 0) {
383 	EVENTHANDLER_DEREGISTER(power_profile_change, sc->tz_event);
384 	AcpiRemoveNotifyHandler(sc->tz_handle, ACPI_DEVICE_NOTIFY,
385 	    acpi_tz_notify_handler);
386 	sysctl_ctx_free(&sc->tz_sysctl_ctx);
387     }
388     ACPI_UNLOCK(thermal);
389 
390     return_VALUE (error);
391 }
392 
393 /*
394  * Parse the current state of this thermal zone and set up to use it.
395  *
396  * Note that we may have previous state, which will have to be discarded.
397  */
398 static int
399 acpi_tz_establish(struct acpi_tz_softc *sc)
400 {
401     ACPI_OBJECT	*obj;
402     int		i;
403     char	nbuf[8];
404 
405     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
406 
407     /* Erase any existing state. */
408     for (i = 0; i < TZ_NUMLEVELS; i++)
409 	if (sc->tz_zone.al[i].Pointer != NULL)
410 	    AcpiOsFree(sc->tz_zone.al[i].Pointer);
411     if (sc->tz_zone.psl.Pointer != NULL)
412 	AcpiOsFree(sc->tz_zone.psl.Pointer);
413 
414     /*
415      * XXX: We initialize only ACPI_BUFFER to avoid race condition
416      * with passive cooling thread which refers psv, tc1, tc2 and tsp.
417      */
418     bzero(sc->tz_zone.ac, sizeof(sc->tz_zone.ac));
419     bzero(sc->tz_zone.al, sizeof(sc->tz_zone.al));
420     bzero(&sc->tz_zone.psl, sizeof(sc->tz_zone.psl));
421 
422     /* Evaluate thermal zone parameters. */
423     for (i = 0; i < TZ_NUMLEVELS; i++) {
424 	ksprintf(nbuf, "_AC%d", i);
425 	acpi_tz_getparam(sc, nbuf, &sc->tz_zone.ac[i]);
426 	ksprintf(nbuf, "_AL%d", i);
427 	sc->tz_zone.al[i].Length = ACPI_ALLOCATE_BUFFER;
428 	sc->tz_zone.al[i].Pointer = NULL;
429 	AcpiEvaluateObject(sc->tz_handle, nbuf, NULL, &sc->tz_zone.al[i]);
430 	obj = (ACPI_OBJECT *)sc->tz_zone.al[i].Pointer;
431 	if (obj != NULL) {
432 	    /* Should be a package containing a list of power objects */
433 	    if (obj->Type != ACPI_TYPE_PACKAGE) {
434 		device_printf(sc->tz_dev, "%s has unknown type %d, rejecting\n",
435 			      nbuf, obj->Type);
436 		return_VALUE (ENXIO);
437 	    }
438 	}
439     }
440     acpi_tz_getparam(sc, "_CRT", &sc->tz_zone.crt);
441     acpi_tz_getparam(sc, "_HOT", &sc->tz_zone.hot);
442     sc->tz_zone.psl.Length = ACPI_ALLOCATE_BUFFER;
443     sc->tz_zone.psl.Pointer = NULL;
444     AcpiEvaluateObject(sc->tz_handle, "_PSL", NULL, &sc->tz_zone.psl);
445     acpi_tz_getparam(sc, "_PSV", &sc->tz_zone.psv);
446     acpi_tz_getparam(sc, "_TC1", &sc->tz_zone.tc1);
447     acpi_tz_getparam(sc, "_TC2", &sc->tz_zone.tc2);
448     acpi_tz_getparam(sc, "_TSP", &sc->tz_zone.tsp);
449     acpi_tz_getparam(sc, "_TZP", &sc->tz_zone.tzp);
450 
451     /*
452      * Sanity-check the values we've been given.
453      *
454      * XXX what do we do about systems that give us the same value for
455      *     more than one of these setpoints?
456      */
457     acpi_tz_sanity(sc, &sc->tz_zone.crt, "_CRT");
458     acpi_tz_sanity(sc, &sc->tz_zone.hot, "_HOT");
459     acpi_tz_sanity(sc, &sc->tz_zone.psv, "_PSV");
460     for (i = 0; i < TZ_NUMLEVELS; i++)
461 	acpi_tz_sanity(sc, &sc->tz_zone.ac[i], "_ACx");
462 
463     return_VALUE (0);
464 }
465 
466 static char *aclevel_string[] = {
467     "NONE", "_AC0", "_AC1", "_AC2", "_AC3", "_AC4",
468     "_AC5", "_AC6", "_AC7", "_AC8", "_AC9"
469 };
470 
471 static __inline const char *
472 acpi_tz_aclevel_string(int active)
473 {
474     if (active < -1 || active >= TZ_NUMLEVELS)
475 	return (aclevel_string[0]);
476 
477     return (aclevel_string[active + 1]);
478 }
479 
480 /*
481  * Get the current temperature.
482  */
483 static int
484 acpi_tz_get_temperature(struct acpi_tz_softc *sc)
485 {
486     int		temp;
487     ACPI_STATUS	status;
488 
489     ACPI_FUNCTION_NAME ("acpi_tz_get_temperature");
490 
491     /*
492      * Silence lookup errors after 10 seconds, then retry every two hours.
493      */
494     if (sc->tz_error_time &&
495 	time_uptime - sc->tz_error_time > TZ_SILENCE_ERROR) {
496 	    if (time_uptime - sc->tz_error_time < TZ_RETRY_ERROR)
497 		return (FALSE);
498 	sc->tz_error_time = time_uptime - TZ_SILENCE_ERROR;
499     }
500 
501     /* Evaluate the thermal zone's _TMP method. */
502     status = acpi_GetInteger(sc->tz_handle, acpi_tz_tmp_name, &temp);
503     if (ACPI_FAILURE(status)) {
504 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
505 	    "error fetching current temperature -- %s\n",
506 	     AcpiFormatException(status));
507 	if (sc->tz_error_time == 0)
508 	    sc->tz_error_time = time_uptime;
509 	return (FALSE);
510     }
511 
512     /* Check it for validity. */
513     acpi_tz_sanity(sc, &temp, acpi_tz_tmp_name);
514     if (temp == -1) {
515 	if (sc->tz_error_time == 0)
516 	    sc->tz_error_time = time_uptime;
517 	return (FALSE);
518     }
519 
520     ACPI_DEBUG_PRINT((ACPI_DB_VALUES, "got %d.%dC\n", TZ_KELVTOC(temp)));
521     sc->tz_temperature = temp;
522     sc->tz_error_time = 0;
523     /* Update sensor */
524     if(sc->tz_temperature == -1)
525         sc->sensor.flags &= ~SENSOR_FINVALID;
526     sc->sensor.value = sc->tz_temperature * 100000 - 50000;
527     return (TRUE);
528 }
529 
530 /*
531  * Evaluate the condition of a thermal zone, take appropriate actions.
532  */
533 static void
534 acpi_tz_monitor(void *Context)
535 {
536     struct acpi_tz_softc *sc;
537     struct	timespec curtime;
538     int		temp;
539     int		i;
540     int		newactive, newflags;
541 
542     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
543 
544     sc = (struct acpi_tz_softc *)Context;
545 
546     /* Get the current temperature. */
547     if (!acpi_tz_get_temperature(sc)) {
548 	/* XXX disable zone? go to max cooling? */
549 	return_VOID;
550     }
551     temp = sc->tz_temperature;
552 
553     /*
554      * Work out what we ought to be doing right now.
555      *
556      * Note that the _ACx levels sort from hot to cold.
557      */
558     newactive = TZ_ACTIVE_NONE;
559     for (i = TZ_NUMLEVELS - 1; i >= 0; i--) {
560 	if (sc->tz_zone.ac[i] != -1 && temp >= sc->tz_zone.ac[i])
561 	    newactive = i;
562     }
563 
564     /*
565      * We are going to get _ACx level down (colder side), but give a guaranteed
566      * minimum cooling run time if requested.
567      */
568     if (acpi_tz_min_runtime > 0 && sc->tz_active != TZ_ACTIVE_NONE &&
569 	sc->tz_active != TZ_ACTIVE_UNKNOWN &&
570 	(newactive == TZ_ACTIVE_NONE || newactive > sc->tz_active)) {
571 
572 	getnanotime(&curtime);
573 	timespecsub(&curtime, &sc->tz_cooling_started);
574 	if (curtime.tv_sec < acpi_tz_min_runtime)
575 	    newactive = sc->tz_active;
576     }
577 
578     /* Handle user override of active mode */
579     if (sc->tz_requested != TZ_ACTIVE_NONE && (newactive == TZ_ACTIVE_NONE
580         || sc->tz_requested < newactive))
581 	newactive = sc->tz_requested;
582 
583     /* update temperature-related flags */
584     newflags = TZ_THFLAG_NONE;
585     if (sc->tz_zone.psv != -1 && temp >= sc->tz_zone.psv)
586 	newflags |= TZ_THFLAG_PSV;
587     if (sc->tz_zone.hot != -1 && temp >= sc->tz_zone.hot)
588 	newflags |= TZ_THFLAG_HOT;
589     if (sc->tz_zone.crt != -1 && temp >= sc->tz_zone.crt)
590 	newflags |= TZ_THFLAG_CRT;
591 
592     /* If the active cooling state has changed, we have to switch things. */
593     if (sc->tz_active == TZ_ACTIVE_UNKNOWN) {
594 	/*
595 	 * We don't know which cooling device is on or off,
596 	 * so stop them all, because we now know which
597 	 * should be on (if any).
598 	 */
599 	for (i = 0; i < TZ_NUMLEVELS; i++) {
600 	    if (sc->tz_zone.al[i].Pointer != NULL) {
601 		acpi_ForeachPackageObject(
602 		    (ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
603 		    acpi_tz_switch_cooler_off, sc);
604 	    }
605 	}
606 	/* now we know that all devices are off */
607 	sc->tz_active = TZ_ACTIVE_NONE;
608     }
609 
610     if (newactive != sc->tz_active) {
611 	/* Turn off unneeded cooling devices that are on, if any are */
612 	for (i = TZ_ACTIVE_LEVEL(sc->tz_active);
613 	     i < TZ_ACTIVE_LEVEL(newactive); i++) {
614 	    acpi_ForeachPackageObject(
615 		(ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
616 		acpi_tz_switch_cooler_off, sc);
617 	}
618 	/* Turn on cooling devices that are required, if any are */
619 	for (i = TZ_ACTIVE_LEVEL(sc->tz_active) - 1;
620 	     i >= TZ_ACTIVE_LEVEL(newactive); i--) {
621 	    acpi_ForeachPackageObject(
622 		(ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
623 		acpi_tz_switch_cooler_on, sc);
624 	}
625 
626 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
627 		    "switched from %s to %s: %d.%dC\n",
628 		    acpi_tz_aclevel_string(sc->tz_active),
629 		    acpi_tz_aclevel_string(newactive), TZ_KELVTOC(temp));
630 	sc->tz_active = newactive;
631 	getnanotime(&sc->tz_cooling_started);
632     }
633 
634     /* XXX (de)activate any passive cooling that may be required. */
635 
636     /*
637      * If the temperature is at _HOT or _CRT, increment our event count.
638      * If it has occurred enough times, shutdown the system.  This is
639      * needed because some systems will report an invalid high temperature
640      * for one poll cycle.  It is suspected this is due to the embedded
641      * controller timing out.  A typical value is 138C for one cycle on
642      * a system that is otherwise 65C.
643      *
644      * If we're almost at that threshold, notify the user through devd(8).
645      */
646     if ((newflags & (TZ_THFLAG_HOT | TZ_THFLAG_CRT)) != 0) {
647 	sc->tz_validchecks++;
648 	if (sc->tz_validchecks == TZ_VALIDCHECKS) {
649 	    device_printf(sc->tz_dev,
650 		"WARNING - current temperature (%d.%dC) exceeds safe limits\n",
651 		TZ_KELVTOC(sc->tz_temperature));
652 	    shutdown_nice(RB_POWEROFF);
653 	} else if (sc->tz_validchecks == TZ_NOTIFYCOUNT)
654 	    acpi_UserNotify("Thermal", sc->tz_handle, TZ_NOTIFY_CRITICAL);
655     } else {
656 	sc->tz_validchecks = 0;
657     }
658     sc->tz_thflags = newflags;
659 
660     return_VOID;
661 }
662 
663 /*
664  * Given an object, verify that it's a reference to a device of some sort,
665  * and try to switch it off.
666  */
667 static void
668 acpi_tz_switch_cooler_off(ACPI_OBJECT *obj, void *arg)
669 {
670     ACPI_HANDLE			cooler;
671 
672     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
673 
674     cooler = acpi_GetReference(NULL, obj);
675     if (cooler == NULL) {
676 	ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "can't get handle\n"));
677 	return_VOID;
678     }
679 
680     ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "called to turn %s off\n",
681 		     acpi_name(cooler)));
682     acpi_pwr_switch_consumer(cooler, ACPI_STATE_D3);
683 
684     return_VOID;
685 }
686 
687 /*
688  * Given an object, verify that it's a reference to a device of some sort,
689  * and try to switch it on.
690  *
691  * XXX replication of off/on function code is bad.
692  */
693 static void
694 acpi_tz_switch_cooler_on(ACPI_OBJECT *obj, void *arg)
695 {
696     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)arg;
697     ACPI_HANDLE			cooler;
698     ACPI_STATUS			status;
699 
700     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
701 
702     cooler = acpi_GetReference(NULL, obj);
703     if (cooler == NULL) {
704 	ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "can't get handle\n"));
705 	return_VOID;
706     }
707 
708     ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "called to turn %s on\n",
709 		     acpi_name(cooler)));
710     status = acpi_pwr_switch_consumer(cooler, ACPI_STATE_D0);
711     if (ACPI_FAILURE(status)) {
712 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
713 		    "failed to activate %s - %s\n", acpi_name(cooler),
714 		    AcpiFormatException(status));
715     }
716 
717     return_VOID;
718 }
719 
720 /*
721  * Read/debug-print a parameter, default it to -1.
722  */
723 static void
724 acpi_tz_getparam(struct acpi_tz_softc *sc, char *node, int *data)
725 {
726 
727     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
728 
729     if (ACPI_FAILURE(acpi_GetInteger(sc->tz_handle, node, data))) {
730 	*data = -1;
731     } else {
732 	ACPI_DEBUG_PRINT((ACPI_DB_VALUES, "%s.%s = %d\n",
733 			 acpi_name(sc->tz_handle), node, *data));
734     }
735 
736     return_VOID;
737 }
738 
739 /*
740  * Handle sysctl for reading and changing the thermal-zone polling rate.
741  */
742 static int
743 acpi_tz_polling_sysctl(SYSCTL_HANDLER_ARGS)
744 {
745     int val, error;
746 
747     val = acpi_tz_polling_rate;
748     error = sysctl_handle_int(oidp, &val, 0, req);
749 
750     /* Error or no new value */
751     if (error != 0 || req->newptr == NULL)
752 	return (error);
753     if (val < 0 || val > 3600)
754 	return (EINVAL);
755 
756     acpi_tz_polling_rate = val;
757     wakeup(&acpi_tz_td);
758     return (error);
759 }
760 
761 /*
762  * Sanity-check a temperature value.  Assume that setpoints
763  * should be between 0C and 200C.
764  */
765 static void
766 acpi_tz_sanity(struct acpi_tz_softc *sc, int *val, char *what)
767 {
768     if (*val != -1 && (*val < TZ_ZEROC || *val > TZ_ZEROC + 2000)) {
769 	/*
770 	 * If the value we are checking is _TMP, warn the user only
771 	 * once. This avoids spamming messages if, for instance, the
772 	 * sensor is broken and always returns an invalid temperature.
773 	 *
774 	 * This is only done for _TMP; other values always emit a
775 	 * warning.
776 	 */
777 	if (what != acpi_tz_tmp_name || !sc->tz_insane_tmp_notified) {
778 	    device_printf(sc->tz_dev, "%s value is absurd, ignored (%d.%dC)\n",
779 			  what, TZ_KELVTOC(*val));
780 
781 	    /* Don't warn the user again if the read value doesn't improve. */
782 	    if (what == acpi_tz_tmp_name)
783 		sc->tz_insane_tmp_notified = 1;
784 	}
785 	*val = -1;
786 	return;
787     }
788 
789     /* This value is correct. Warn if it's incorrect again. */
790     if (what == acpi_tz_tmp_name)
791 	sc->tz_insane_tmp_notified = 0;
792 }
793 
794 /*
795  * Respond to a sysctl on the active state node.
796  */
797 static int
798 acpi_tz_active_sysctl(SYSCTL_HANDLER_ARGS)
799 {
800     struct acpi_tz_softc	*sc;
801     int				active;
802     int		 		error;
803 
804     sc = (struct acpi_tz_softc *)oidp->oid_arg1;
805     active = sc->tz_active;
806     error = sysctl_handle_int(oidp, &active, 0, req);
807 
808     /* Error or no new value */
809     if (error != 0 || req->newptr == NULL)
810 	return (error);
811     if (active < -1 || active >= TZ_NUMLEVELS)
812 	return (EINVAL);
813 
814     /* Set new preferred level and re-switch */
815     sc->tz_requested = active;
816     acpi_tz_signal(sc, 0);
817     return (0);
818 }
819 
820 static int
821 acpi_tz_cooling_sysctl(SYSCTL_HANDLER_ARGS)
822 {
823     struct acpi_tz_softc *sc;
824     int enabled, error;
825 
826     sc = (struct acpi_tz_softc *)oidp->oid_arg1;
827     enabled = sc->tz_cooling_enabled;
828     error = sysctl_handle_int(oidp, &enabled, 0, req);
829 
830     /* Error or no new value */
831     if (error != 0 || req->newptr == NULL)
832 	return (error);
833     if (enabled != TRUE && enabled != FALSE)
834 	return (EINVAL);
835 
836     if (enabled) {
837 	if (acpi_tz_cooling_is_available(sc))
838 	    error = acpi_tz_cooling_thread_start(sc);
839 	else
840 	    error = ENODEV;
841 	if (error)
842 	    enabled = FALSE;
843     }
844     sc->tz_cooling_enabled = enabled;
845     return (error);
846 }
847 
848 static int
849 acpi_tz_temp_sysctl(SYSCTL_HANDLER_ARGS)
850 {
851     struct acpi_tz_softc	*sc;
852     int				temp, *temp_ptr;
853     int		 		error;
854 
855     sc = oidp->oid_arg1;
856     temp_ptr = (int *)((uintptr_t)sc + oidp->oid_arg2);
857     temp = *temp_ptr;
858     error = sysctl_handle_int(oidp, &temp, 0, req);
859 
860     /* Error or no new value */
861     if (error != 0 || req->newptr == NULL)
862 	return (error);
863 
864     /* Only allow changing settings if override is set. */
865     if (!acpi_tz_override)
866 	return (EPERM);
867 
868     /* Check user-supplied value for sanity. */
869     acpi_tz_sanity(sc, &temp, "user-supplied temp");
870     if (temp == -1)
871 	return (EINVAL);
872 
873     *temp_ptr = temp;
874     return (0);
875 }
876 
877 static int
878 acpi_tz_passive_sysctl(SYSCTL_HANDLER_ARGS)
879 {
880     struct acpi_tz_softc	*sc;
881     int				val, *val_ptr;
882     int				error;
883 
884     sc = oidp->oid_arg1;
885     val_ptr = (int *)((uintptr_t)sc + oidp->oid_arg2);
886     val = *val_ptr;
887     error = sysctl_handle_int(oidp, &val, 0, req);
888 
889     /* Error or no new value */
890     if (error != 0 || req->newptr == NULL)
891 	return (error);
892 
893     /* Only allow changing settings if override is set. */
894     if (!acpi_tz_override)
895 	return (EPERM);
896 
897     *val_ptr = val;
898     return (0);
899 }
900 
901 static void
902 acpi_tz_notify_handler(ACPI_HANDLE h, UINT32 notify, void *context)
903 {
904     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)context;
905 
906     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
907 
908     switch (notify) {
909     case TZ_NOTIFY_TEMPERATURE:
910 	/* Temperature change occurred */
911 	acpi_tz_signal(sc, 0);
912 	break;
913     case TZ_NOTIFY_DEVICES:
914     case TZ_NOTIFY_LEVELS:
915 	/* Zone devices/setpoints changed */
916 	acpi_tz_signal(sc, TZ_FLAG_GETSETTINGS);
917 	break;
918     default:
919 	device_printf(sc->tz_dev, "unknown notify: %#x\n", notify);
920 	break;
921     }
922 
923     acpi_UserNotify("Thermal", h, notify);
924 
925     return_VOID;
926 }
927 
928 static void
929 acpi_tz_signal(struct acpi_tz_softc *sc, int flags)
930 {
931     ACPI_LOCK(thermal);
932     sc->tz_flags |= flags;
933     ACPI_UNLOCK(thermal);
934     wakeup(&acpi_tz_td);
935 }
936 
937 /*
938  * Notifies can be generated asynchronously but have also been seen to be
939  * triggered by other thermal methods.  One system generates a notify of
940  * 0x81 when the fan is turned on or off.  Another generates it when _SCP
941  * is called.  To handle these situations, we check the zone via
942  * acpi_tz_monitor() before evaluating changes to setpoints or the cooling
943  * policy.
944  */
945 static void
946 acpi_tz_timeout(struct acpi_tz_softc *sc, int flags)
947 {
948 
949     /* Check the current temperature and take action based on it */
950     acpi_tz_monitor(sc);
951 
952     /* If requested, get the power profile settings. */
953     if (flags & TZ_FLAG_GETPROFILE)
954 	acpi_tz_power_profile(sc);
955 
956     /*
957      * If requested, check for new devices/setpoints.  After finding them,
958      * check if we need to switch fans based on the new values.
959      */
960     if (flags & TZ_FLAG_GETSETTINGS) {
961 	acpi_tz_establish(sc);
962 	acpi_tz_monitor(sc);
963     }
964 
965     /* XXX passive cooling actions? */
966 }
967 
968 /*
969  * System power profile may have changed; fetch and notify the
970  * thermal zone accordingly.
971  *
972  * Since this can be called from an arbitrary eventhandler, it needs
973  * to get the ACPI lock itself.
974  */
975 static void
976 acpi_tz_power_profile(void *arg)
977 {
978     ACPI_STATUS			status;
979     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)arg;
980     int				state;
981 
982     state = power_profile_get_state();
983     if (state != POWER_PROFILE_PERFORMANCE && state != POWER_PROFILE_ECONOMY)
984 	return;
985 
986     /* check that we haven't decided there's no _SCP method */
987     if ((sc->tz_flags & TZ_FLAG_NO_SCP) == 0) {
988 
989 	/* Call _SCP to set the new profile */
990 	status = acpi_SetInteger(sc->tz_handle, "_SCP",
991 	    (state == POWER_PROFILE_PERFORMANCE) ? 0 : 1);
992 	if (ACPI_FAILURE(status)) {
993 	    if (status != AE_NOT_FOUND)
994 		ACPI_VPRINT(sc->tz_dev,
995 			    acpi_device_get_parent_softc(sc->tz_dev),
996 			    "can't evaluate %s._SCP - %s\n",
997 			    acpi_name(sc->tz_handle),
998 			    AcpiFormatException(status));
999 	    sc->tz_flags |= TZ_FLAG_NO_SCP;
1000 	} else {
1001 	    /* We have to re-evaluate the entire zone now */
1002 	    acpi_tz_signal(sc, TZ_FLAG_GETSETTINGS);
1003 	}
1004     }
1005 }
1006 
1007 /*
1008  * Thermal zone monitor thread.
1009  */
1010 static void
1011 acpi_tz_thread(void *arg)
1012 {
1013     device_t	*devs;
1014     int		devcount, i;
1015     int		flags;
1016     struct acpi_tz_softc **sc;
1017 
1018     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
1019 
1020     devs = NULL;
1021     devcount = 0;
1022     sc = NULL;
1023 
1024     ACPI_LOCK(acpi);		/* wait for ACPI to finish nominal attach */
1025     ACPI_UNLOCK(acpi);
1026 
1027     lwkt_gettoken(&acpi_token);
1028     for (;;) {
1029 	/* If the number of devices has changed, re-evaluate. */
1030 	if (devclass_get_count(acpi_tz_devclass) != devcount) {
1031 	    if (devs != NULL) {
1032 		kfree(devs, M_TEMP);
1033 		kfree(sc, M_TEMP);
1034 	    }
1035 	    devclass_get_devices(acpi_tz_devclass, &devs, &devcount);
1036 	    sc = kmalloc(sizeof(struct acpi_tz_softc *) * devcount, M_TEMP,
1037 			M_WAITOK | M_ZERO);
1038 	    for (i = 0; i < devcount; i++)
1039 		sc[i] = device_get_softc(devs[i]);
1040 	}
1041 
1042 	/* Check for temperature events and act on them. */
1043 	for (i = 0; i < devcount; i++) {
1044 	    ACPI_LOCK(thermal);
1045 	    flags = sc[i]->tz_flags;
1046 	    sc[i]->tz_flags &= TZ_FLAG_NO_SCP;
1047 	    ACPI_UNLOCK(thermal);
1048 	    acpi_tz_timeout(sc[i], flags);
1049 	}
1050 
1051 	/* If more work to do, don't go to sleep yet. */
1052 	ACPI_LOCK(thermal);
1053 	for (i = 0; i < devcount; i++) {
1054 	    if (sc[i]->tz_flags & ~TZ_FLAG_NO_SCP)
1055 		break;
1056 	}
1057 
1058 	/*
1059 	 * Interlocked sleep until signaled or we timeout.
1060 	 */
1061 	if (i == devcount) {
1062 	    tsleep_interlock(&acpi_tz_td, 0);
1063 	    ACPI_UNLOCK(thermal);
1064 	    tsleep(&acpi_tz_td, PINTERLOCKED, "tzpoll",
1065 		(acpi_tz_polling_rate <= 0 ? 0 : hz * acpi_tz_polling_rate));
1066 	} else {
1067 	    ACPI_UNLOCK(thermal);
1068 	}
1069     }
1070     lwkt_reltoken(&acpi_token);
1071 }
1072 
1073 #ifdef __FreeBSD__
1074 static int
1075 acpi_tz_cpufreq_restore(struct acpi_tz_softc *sc)
1076 {
1077     device_t dev;
1078     int error;
1079 
1080     if (!sc->tz_cooling_updated)
1081 	return (0);
1082     if ((dev = devclass_get_device(devclass_find("cpufreq"), 0)) == NULL)
1083 	return (ENXIO);
1084     ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
1085 	"temperature %d.%dC: resuming previous clock speed (%d MHz)\n",
1086 	TZ_KELVTOC(sc->tz_temperature), sc->tz_cooling_saved_freq);
1087     error = CPUFREQ_SET(dev, NULL, CPUFREQ_PRIO_KERN);
1088     if (error == 0)
1089 	sc->tz_cooling_updated = FALSE;
1090     return (error);
1091 }
1092 
1093 static int
1094 acpi_tz_cpufreq_update(struct acpi_tz_softc *sc, int req)
1095 {
1096     device_t dev;
1097     struct cf_level *levels;
1098     int num_levels, error, freq, desired_freq, perf, i;
1099 
1100     levels = kmalloc(CPUFREQ_MAX_LEVELS * sizeof(*levels), M_TEMP, M_NOWAIT);
1101     if (levels == NULL)
1102 	return (ENOMEM);
1103 
1104     /*
1105      * Find the main device, cpufreq0.  We don't yet support independent
1106      * CPU frequency control on SMP.
1107      */
1108     if ((dev = devclass_get_device(devclass_find("cpufreq"), 0)) == NULL) {
1109 	error = ENXIO;
1110 	goto out;
1111     }
1112 
1113     /* Get the current frequency. */
1114     error = CPUFREQ_GET(dev, &levels[0]);
1115     if (error)
1116 	goto out;
1117     freq = levels[0].total_set.freq;
1118 
1119     /* Get the current available frequency levels. */
1120     num_levels = CPUFREQ_MAX_LEVELS;
1121     error = CPUFREQ_LEVELS(dev, levels, &num_levels);
1122     if (error) {
1123 	if (error == E2BIG)
1124 	    printf("cpufreq: need to increase CPUFREQ_MAX_LEVELS\n");
1125 	goto out;
1126     }
1127 
1128     /* Calculate the desired frequency as a percent of the max frequency. */
1129     perf = 100 * freq / levels[0].total_set.freq - req;
1130     if (perf < 0)
1131 	perf = 0;
1132     else if (perf > 100)
1133 	perf = 100;
1134     desired_freq = levels[0].total_set.freq * perf / 100;
1135 
1136     if (desired_freq < freq) {
1137 	/* Find the closest available frequency, rounding down. */
1138 	for (i = 0; i < num_levels; i++)
1139 	    if (levels[i].total_set.freq <= desired_freq)
1140 		break;
1141 
1142 	/* If we didn't find a relevant setting, use the lowest. */
1143 	if (i == num_levels)
1144 	    i--;
1145     } else {
1146 	/* If we didn't decrease frequency yet, don't increase it. */
1147 	if (!sc->tz_cooling_updated) {
1148 	    sc->tz_cooling_active = FALSE;
1149 	    goto out;
1150 	}
1151 
1152 	/* Use saved cpu frequency as maximum value. */
1153 	if (desired_freq > sc->tz_cooling_saved_freq)
1154 	    desired_freq = sc->tz_cooling_saved_freq;
1155 
1156 	/* Find the closest available frequency, rounding up. */
1157 	for (i = num_levels - 1; i >= 0; i--)
1158 	    if (levels[i].total_set.freq >= desired_freq)
1159 		break;
1160 
1161 	/* If we didn't find a relevant setting, use the highest. */
1162 	if (i == -1)
1163 	    i++;
1164 
1165 	/* If we're going to the highest frequency, restore the old setting. */
1166 	if (i == 0 || desired_freq == sc->tz_cooling_saved_freq) {
1167 	    error = acpi_tz_cpufreq_restore(sc);
1168 	    if (error == 0)
1169 		sc->tz_cooling_active = FALSE;
1170 	    goto out;
1171 	}
1172     }
1173 
1174     /* If we are going to a new frequency, activate it. */
1175     if (levels[i].total_set.freq != freq) {
1176 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
1177 	    "temperature %d.%dC: %screasing clock speed "
1178 	    "from %d MHz to %d MHz\n",
1179 	    TZ_KELVTOC(sc->tz_temperature),
1180 	    (freq > levels[i].total_set.freq) ? "de" : "in",
1181 	    freq, levels[i].total_set.freq);
1182 	error = CPUFREQ_SET(dev, &levels[i], CPUFREQ_PRIO_KERN);
1183 	if (error == 0 && !sc->tz_cooling_updated) {
1184 	    sc->tz_cooling_saved_freq = freq;
1185 	    sc->tz_cooling_updated = TRUE;
1186 	}
1187     }
1188 
1189 out:
1190     if (levels)
1191 	free(levels, M_TEMP);
1192     return (error);
1193 }
1194 #endif
1195 
1196 /*
1197  * Passive cooling thread; monitors current temperature according to the
1198  * cooling interval and calculates whether to scale back CPU frequency.
1199  */
1200 static void
1201 acpi_tz_cooling_thread(void *arg)
1202 {
1203     struct acpi_tz_softc *sc;
1204     int perf, curr_temp, prev_temp;
1205 #ifdef __FreeBSD__
1206     int error;
1207 #endif
1208 
1209     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
1210 
1211     ACPI_LOCK(acpi);		/* wait for ACPI to finish nominal attach */
1212     ACPI_UNLOCK(acpi);
1213 
1214     sc = (struct acpi_tz_softc *)arg;
1215     lwkt_gettoken(&acpi_token);
1216 
1217     prev_temp = sc->tz_temperature;
1218     while (sc->tz_cooling_enabled) {
1219 	if (sc->tz_cooling_active)
1220 	    (void)acpi_tz_get_temperature(sc);
1221 	curr_temp = sc->tz_temperature;
1222 	if (curr_temp >= sc->tz_zone.psv)
1223 	    sc->tz_cooling_active = TRUE;
1224 	if (sc->tz_cooling_active) {
1225 	    perf = sc->tz_zone.tc1 * (curr_temp - prev_temp) +
1226 		   sc->tz_zone.tc2 * (curr_temp - sc->tz_zone.psv);
1227 	    perf /= 10;
1228 
1229 	    if (perf != 0) {
1230 #ifdef __FreeBSD__
1231 		error = acpi_tz_cpufreq_update(sc, perf);
1232 
1233 		/*
1234 		 * If error and not simply a higher priority setting was
1235 		 * active, disable cooling.
1236 		 */
1237 		if (error != 0 && error != EPERM) {
1238 		    device_printf(sc->tz_dev,
1239 			"failed to set new freq, disabling passive cooling\n");
1240 		    sc->tz_cooling_enabled = FALSE;
1241 		}
1242 #endif
1243 	    }
1244 	}
1245 	prev_temp = curr_temp;
1246 	tsleep(&sc->tz_cooling_proc, 0, "cooling",
1247 	    hz * sc->tz_zone.tsp / 10);
1248     }
1249     if (sc->tz_cooling_active) {
1250 #ifdef __FreeBSD__
1251 	acpi_tz_cpufreq_restore(sc);
1252 #endif
1253 	sc->tz_cooling_active = FALSE;
1254     }
1255     sc->tz_cooling_proc = NULL;
1256     ACPI_LOCK(thermal);
1257     sc->tz_cooling_proc_running = FALSE;
1258     ACPI_UNLOCK(thermal);
1259 
1260     lwkt_reltoken(&acpi_token);
1261 }
1262 
1263 /*
1264  * TODO: We ignore _PSL (list of cooling devices) since cpufreq enumerates
1265  * all CPUs for us.  However, it's possible in the future _PSL will
1266  * reference non-CPU devices so we may want to support it then.
1267  */
1268 static int
1269 acpi_tz_cooling_is_available(struct acpi_tz_softc *sc)
1270 {
1271     return (sc->tz_zone.tc1 != -1 && sc->tz_zone.tc2 != -1 &&
1272 	sc->tz_zone.tsp != -1 && sc->tz_zone.tsp != 0 &&
1273 	sc->tz_zone.psv != -1);
1274 }
1275 
1276 static int
1277 acpi_tz_cooling_thread_start(struct acpi_tz_softc *sc)
1278 {
1279     int error;
1280 
1281     ACPI_LOCK(thermal);
1282     if (sc->tz_cooling_proc_running) {
1283 	ACPI_UNLOCK(thermal);
1284 	return (0);
1285     }
1286     sc->tz_cooling_proc_running = TRUE;
1287     ACPI_UNLOCK(thermal);
1288     error = 0;
1289     if (sc->tz_cooling_proc == NULL) {
1290 	error = kthread_create(acpi_tz_cooling_thread, sc,
1291 	    &sc->tz_cooling_proc,
1292 	    "acpi_cooling%d", device_get_unit(sc->tz_dev));
1293 	if (error != 0) {
1294 	    device_printf(sc->tz_dev, "could not create thread - %d", error);
1295 	    ACPI_LOCK(thermal);
1296 	    sc->tz_cooling_proc_running = FALSE;
1297 	    ACPI_UNLOCK(thermal);
1298 	}
1299     }
1300     return (error);
1301 }
1302