xref: /freebsd/sys/dev/acpica/acpi_thermal.c (revision 0b3105a37d7adcadcb720112fed4dc4e8040be99)
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/kernel.h>
34 #include <sys/bus.h>
35 #include <sys/cpu.h>
36 #include <sys/kthread.h>
37 #include <sys/malloc.h>
38 #include <sys/module.h>
39 #include <sys/proc.h>
40 #include <sys/reboot.h>
41 #include <sys/sysctl.h>
42 #include <sys/unistd.h>
43 #include <sys/power.h>
44 
45 #include "cpufreq_if.h"
46 
47 #include <contrib/dev/acpica/include/acpi.h>
48 #include <contrib/dev/acpica/include/accommon.h>
49 
50 #include <dev/acpica/acpivar.h>
51 
52 /* Hooks for the ACPI CA debugging infrastructure */
53 #define _COMPONENT	ACPI_THERMAL
54 ACPI_MODULE_NAME("THERMAL")
55 
56 #define TZ_ZEROC	2732
57 #define TZ_KELVTOC(x)	(((x) - TZ_ZEROC) / 10), abs(((x) - TZ_ZEROC) % 10)
58 
59 #define TZ_NOTIFY_TEMPERATURE	0x80 /* Temperature changed. */
60 #define TZ_NOTIFY_LEVELS	0x81 /* Cooling levels changed. */
61 #define TZ_NOTIFY_DEVICES	0x82 /* Device lists changed. */
62 #define TZ_NOTIFY_CRITICAL	0xcc /* Fake notify that _CRT/_HOT reached. */
63 
64 /* Check for temperature changes every 10 seconds by default */
65 #define TZ_POLLRATE	10
66 
67 /* Make sure the reported temperature is valid for this number of polls. */
68 #define TZ_VALIDCHECKS	3
69 
70 /* Notify the user we will be shutting down in one more poll cycle. */
71 #define TZ_NOTIFYCOUNT	(TZ_VALIDCHECKS - 1)
72 
73 /* ACPI spec defines this */
74 #define TZ_NUMLEVELS	10
75 struct acpi_tz_zone {
76     int		ac[TZ_NUMLEVELS];
77     ACPI_BUFFER	al[TZ_NUMLEVELS];
78     int		crt;
79     int		hot;
80     ACPI_BUFFER	psl;
81     int		psv;
82     int		tc1;
83     int		tc2;
84     int		tsp;
85     int		tzp;
86 };
87 
88 struct acpi_tz_softc {
89     device_t			tz_dev;
90     ACPI_HANDLE			tz_handle;	/*Thermal zone handle*/
91     int				tz_temperature;	/*Current temperature*/
92     int				tz_active;	/*Current active cooling*/
93 #define TZ_ACTIVE_NONE		-1
94 #define TZ_ACTIVE_UNKNOWN	-2
95     int				tz_requested;	/*Minimum active cooling*/
96     int				tz_thflags;	/*Current temp-related flags*/
97 #define TZ_THFLAG_NONE		0
98 #define TZ_THFLAG_PSV		(1<<0)
99 #define TZ_THFLAG_HOT		(1<<2)
100 #define TZ_THFLAG_CRT		(1<<3)
101     int				tz_flags;
102 #define TZ_FLAG_NO_SCP		(1<<0)		/*No _SCP method*/
103 #define TZ_FLAG_GETPROFILE	(1<<1)		/*Get power_profile in timeout*/
104 #define TZ_FLAG_GETSETTINGS	(1<<2)		/*Get devs/setpoints*/
105     struct timespec		tz_cooling_started;
106 					/*Current cooling starting time*/
107 
108     struct sysctl_ctx_list	tz_sysctl_ctx;
109     struct sysctl_oid		*tz_sysctl_tree;
110     eventhandler_tag		tz_event;
111 
112     struct acpi_tz_zone 	tz_zone;	/*Thermal zone parameters*/
113     int				tz_validchecks;
114     int				tz_insane_tmp_notified;
115 
116     /* passive cooling */
117     struct proc			*tz_cooling_proc;
118     int				tz_cooling_proc_running;
119     int				tz_cooling_enabled;
120     int				tz_cooling_active;
121     int				tz_cooling_updated;
122     int				tz_cooling_saved_freq;
123 };
124 
125 #define	TZ_ACTIVE_LEVEL(act)	((act) >= 0 ? (act) : TZ_NUMLEVELS)
126 
127 #define CPUFREQ_MAX_LEVELS	64 /* XXX cpufreq should export this */
128 
129 static int	acpi_tz_probe(device_t dev);
130 static int	acpi_tz_attach(device_t dev);
131 static int	acpi_tz_establish(struct acpi_tz_softc *sc);
132 static void	acpi_tz_monitor(void *Context);
133 static void	acpi_tz_switch_cooler_off(ACPI_OBJECT *obj, void *arg);
134 static void	acpi_tz_switch_cooler_on(ACPI_OBJECT *obj, void *arg);
135 static void	acpi_tz_getparam(struct acpi_tz_softc *sc, char *node,
136 				 int *data);
137 static void	acpi_tz_sanity(struct acpi_tz_softc *sc, int *val, char *what);
138 static int	acpi_tz_active_sysctl(SYSCTL_HANDLER_ARGS);
139 static int	acpi_tz_cooling_sysctl(SYSCTL_HANDLER_ARGS);
140 static int	acpi_tz_temp_sysctl(SYSCTL_HANDLER_ARGS);
141 static int	acpi_tz_passive_sysctl(SYSCTL_HANDLER_ARGS);
142 static void	acpi_tz_notify_handler(ACPI_HANDLE h, UINT32 notify,
143 				       void *context);
144 static void	acpi_tz_signal(struct acpi_tz_softc *sc, int flags);
145 static void	acpi_tz_timeout(struct acpi_tz_softc *sc, int flags);
146 static void	acpi_tz_power_profile(void *arg);
147 static void	acpi_tz_thread(void *arg);
148 static int	acpi_tz_cooling_is_available(struct acpi_tz_softc *sc);
149 static int	acpi_tz_cooling_thread_start(struct acpi_tz_softc *sc);
150 
151 static device_method_t acpi_tz_methods[] = {
152     /* Device interface */
153     DEVMETHOD(device_probe,	acpi_tz_probe),
154     DEVMETHOD(device_attach,	acpi_tz_attach),
155 
156     DEVMETHOD_END
157 };
158 
159 static driver_t acpi_tz_driver = {
160     "acpi_tz",
161     acpi_tz_methods,
162     sizeof(struct acpi_tz_softc),
163 };
164 
165 static char *acpi_tz_tmp_name = "_TMP";
166 
167 static devclass_t acpi_tz_devclass;
168 DRIVER_MODULE(acpi_tz, acpi, acpi_tz_driver, acpi_tz_devclass, 0, 0);
169 MODULE_DEPEND(acpi_tz, acpi, 1, 1, 1);
170 
171 static struct sysctl_ctx_list	acpi_tz_sysctl_ctx;
172 static struct sysctl_oid	*acpi_tz_sysctl_tree;
173 
174 /* Minimum cooling run time */
175 static int			acpi_tz_min_runtime;
176 static int			acpi_tz_polling_rate = TZ_POLLRATE;
177 static int			acpi_tz_override;
178 
179 /* Timezone polling thread */
180 static struct proc		*acpi_tz_proc;
181 ACPI_LOCK_DECL(thermal, "ACPI thermal zone");
182 
183 static int			acpi_tz_cooling_unit = -1;
184 
185 static int
186 acpi_tz_probe(device_t dev)
187 {
188     int		result;
189 
190     if (acpi_get_type(dev) == ACPI_TYPE_THERMAL && !acpi_disabled("thermal")) {
191 	device_set_desc(dev, "Thermal Zone");
192 	result = -10;
193     } else
194 	result = ENXIO;
195     return (result);
196 }
197 
198 static int
199 acpi_tz_attach(device_t dev)
200 {
201     struct acpi_tz_softc	*sc;
202     struct acpi_softc		*acpi_sc;
203     int				error;
204     char			oidname[8];
205 
206     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
207 
208     sc = device_get_softc(dev);
209     sc->tz_dev = dev;
210     sc->tz_handle = acpi_get_handle(dev);
211     sc->tz_requested = TZ_ACTIVE_NONE;
212     sc->tz_active = TZ_ACTIVE_UNKNOWN;
213     sc->tz_thflags = TZ_THFLAG_NONE;
214     sc->tz_cooling_proc = NULL;
215     sc->tz_cooling_proc_running = FALSE;
216     sc->tz_cooling_active = FALSE;
217     sc->tz_cooling_updated = FALSE;
218     sc->tz_cooling_enabled = FALSE;
219 
220     /*
221      * Parse the current state of the thermal zone and build control
222      * structures.  We don't need to worry about interference with the
223      * control thread since we haven't fully attached this device yet.
224      */
225     if ((error = acpi_tz_establish(sc)) != 0)
226 	return (error);
227 
228     /*
229      * Register for any Notify events sent to this zone.
230      */
231     AcpiInstallNotifyHandler(sc->tz_handle, ACPI_DEVICE_NOTIFY,
232 			     acpi_tz_notify_handler, sc);
233 
234     /*
235      * Create our sysctl nodes.
236      *
237      * XXX we need a mechanism for adding nodes under ACPI.
238      */
239     if (device_get_unit(dev) == 0) {
240 	acpi_sc = acpi_device_get_parent_softc(dev);
241 	sysctl_ctx_init(&acpi_tz_sysctl_ctx);
242 	acpi_tz_sysctl_tree = SYSCTL_ADD_NODE(&acpi_tz_sysctl_ctx,
243 			      SYSCTL_CHILDREN(acpi_sc->acpi_sysctl_tree),
244 			      OID_AUTO, "thermal", CTLFLAG_RD, 0, "");
245 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
246 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
247 		       OID_AUTO, "min_runtime", CTLFLAG_RW,
248 		       &acpi_tz_min_runtime, 0,
249 		       "minimum cooling run time in sec");
250 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
251 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
252 		       OID_AUTO, "polling_rate", CTLFLAG_RW,
253 		       &acpi_tz_polling_rate, 0, "monitor polling interval in seconds");
254 	SYSCTL_ADD_INT(&acpi_tz_sysctl_ctx,
255 		       SYSCTL_CHILDREN(acpi_tz_sysctl_tree), OID_AUTO,
256 		       "user_override", CTLFLAG_RW, &acpi_tz_override, 0,
257 		       "allow override of thermal settings");
258     }
259     sysctl_ctx_init(&sc->tz_sysctl_ctx);
260     sprintf(oidname, "tz%d", device_get_unit(dev));
261     sc->tz_sysctl_tree = SYSCTL_ADD_NODE(&sc->tz_sysctl_ctx,
262 					 SYSCTL_CHILDREN(acpi_tz_sysctl_tree),
263 					 OID_AUTO, oidname, CTLFLAG_RD, 0, "");
264     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
265 		    OID_AUTO, "temperature", CTLTYPE_INT | CTLFLAG_RD,
266 		    &sc->tz_temperature, 0, sysctl_handle_int,
267 		    "IK", "current thermal zone temperature");
268     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
269 		    OID_AUTO, "active", CTLTYPE_INT | CTLFLAG_RW,
270 		    sc, 0, acpi_tz_active_sysctl, "I", "cooling is active");
271     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
272 		    OID_AUTO, "passive_cooling", CTLTYPE_INT | CTLFLAG_RW,
273 		    sc, 0, acpi_tz_cooling_sysctl, "I",
274 		    "enable passive (speed reduction) cooling");
275 
276     SYSCTL_ADD_INT(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
277 		   OID_AUTO, "thermal_flags", CTLFLAG_RD,
278 		   &sc->tz_thflags, 0, "thermal zone flags");
279     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
280 		    OID_AUTO, "_PSV", CTLTYPE_INT | CTLFLAG_RW,
281 		    sc, offsetof(struct acpi_tz_softc, tz_zone.psv),
282 		    acpi_tz_temp_sysctl, "IK", "passive cooling temp setpoint");
283     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
284 		    OID_AUTO, "_HOT", CTLTYPE_INT | CTLFLAG_RW,
285 		    sc, offsetof(struct acpi_tz_softc, tz_zone.hot),
286 		    acpi_tz_temp_sysctl, "IK",
287 		    "too hot temp setpoint (suspend now)");
288     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
289 		    OID_AUTO, "_CRT", CTLTYPE_INT | CTLFLAG_RW,
290 		    sc, offsetof(struct acpi_tz_softc, tz_zone.crt),
291 		    acpi_tz_temp_sysctl, "IK",
292 		    "critical temp setpoint (shutdown now)");
293     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
294 		    OID_AUTO, "_ACx", CTLTYPE_INT | CTLFLAG_RD,
295 		    &sc->tz_zone.ac, sizeof(sc->tz_zone.ac),
296 		    sysctl_handle_opaque, "IK", "");
297     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
298 		    OID_AUTO, "_TC1", CTLTYPE_INT | CTLFLAG_RW,
299 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tc1),
300 		    acpi_tz_passive_sysctl, "I",
301 		    "thermal constant 1 for passive cooling");
302     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
303 		    OID_AUTO, "_TC2", CTLTYPE_INT | CTLFLAG_RW,
304 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tc2),
305 		    acpi_tz_passive_sysctl, "I",
306 		    "thermal constant 2 for passive cooling");
307     SYSCTL_ADD_PROC(&sc->tz_sysctl_ctx, SYSCTL_CHILDREN(sc->tz_sysctl_tree),
308 		    OID_AUTO, "_TSP", CTLTYPE_INT | CTLFLAG_RW,
309 		    sc, offsetof(struct acpi_tz_softc, tz_zone.tsp),
310 		    acpi_tz_passive_sysctl, "I",
311 		    "thermal sampling period for passive cooling");
312 
313     /*
314      * Create thread to service all of the thermal zones.  Register
315      * our power profile event handler.
316      */
317     sc->tz_event = EVENTHANDLER_REGISTER(power_profile_change,
318 	acpi_tz_power_profile, sc, 0);
319     if (acpi_tz_proc == NULL) {
320 	error = kproc_create(acpi_tz_thread, NULL, &acpi_tz_proc,
321 	    RFHIGHPID, 0, "acpi_thermal");
322 	if (error != 0) {
323 	    device_printf(sc->tz_dev, "could not create thread - %d", error);
324 	    goto out;
325 	}
326     }
327 
328     /*
329      * Create a thread to handle passive cooling for 1st zone which
330      * has _PSV, _TSP, _TC1 and _TC2.  Users can enable it for other
331      * zones manually for now.
332      *
333      * XXX We enable only one zone to avoid multiple zones conflict
334      * with each other since cpufreq currently sets all CPUs to the
335      * given frequency whereas it's possible for different thermal
336      * zones to specify independent settings for multiple CPUs.
337      */
338     if (acpi_tz_cooling_unit < 0 && acpi_tz_cooling_is_available(sc))
339 	sc->tz_cooling_enabled = TRUE;
340     if (sc->tz_cooling_enabled) {
341 	error = acpi_tz_cooling_thread_start(sc);
342 	if (error != 0) {
343 	    sc->tz_cooling_enabled = FALSE;
344 	    goto out;
345 	}
346 	acpi_tz_cooling_unit = device_get_unit(dev);
347     }
348 
349     /*
350      * Flag the event handler for a manual invocation by our timeout.
351      * We defer it like this so that the rest of the subsystem has time
352      * to come up.  Don't bother evaluating/printing the temperature at
353      * this point; on many systems it'll be bogus until the EC is running.
354      */
355     sc->tz_flags |= TZ_FLAG_GETPROFILE;
356 
357 out:
358     if (error != 0) {
359 	EVENTHANDLER_DEREGISTER(power_profile_change, sc->tz_event);
360 	AcpiRemoveNotifyHandler(sc->tz_handle, ACPI_DEVICE_NOTIFY,
361 	    acpi_tz_notify_handler);
362 	sysctl_ctx_free(&sc->tz_sysctl_ctx);
363     }
364     return_VALUE (error);
365 }
366 
367 /*
368  * Parse the current state of this thermal zone and set up to use it.
369  *
370  * Note that we may have previous state, which will have to be discarded.
371  */
372 static int
373 acpi_tz_establish(struct acpi_tz_softc *sc)
374 {
375     ACPI_OBJECT	*obj;
376     int		i;
377     char	nbuf[8];
378 
379     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
380 
381     /* Erase any existing state. */
382     for (i = 0; i < TZ_NUMLEVELS; i++)
383 	if (sc->tz_zone.al[i].Pointer != NULL)
384 	    AcpiOsFree(sc->tz_zone.al[i].Pointer);
385     if (sc->tz_zone.psl.Pointer != NULL)
386 	AcpiOsFree(sc->tz_zone.psl.Pointer);
387 
388     /*
389      * XXX: We initialize only ACPI_BUFFER to avoid race condition
390      * with passive cooling thread which refers psv, tc1, tc2 and tsp.
391      */
392     bzero(sc->tz_zone.ac, sizeof(sc->tz_zone.ac));
393     bzero(sc->tz_zone.al, sizeof(sc->tz_zone.al));
394     bzero(&sc->tz_zone.psl, sizeof(sc->tz_zone.psl));
395 
396     /* Evaluate thermal zone parameters. */
397     for (i = 0; i < TZ_NUMLEVELS; i++) {
398 	sprintf(nbuf, "_AC%d", i);
399 	acpi_tz_getparam(sc, nbuf, &sc->tz_zone.ac[i]);
400 	sprintf(nbuf, "_AL%d", i);
401 	sc->tz_zone.al[i].Length = ACPI_ALLOCATE_BUFFER;
402 	sc->tz_zone.al[i].Pointer = NULL;
403 	AcpiEvaluateObject(sc->tz_handle, nbuf, NULL, &sc->tz_zone.al[i]);
404 	obj = (ACPI_OBJECT *)sc->tz_zone.al[i].Pointer;
405 	if (obj != NULL) {
406 	    /* Should be a package containing a list of power objects */
407 	    if (obj->Type != ACPI_TYPE_PACKAGE) {
408 		device_printf(sc->tz_dev, "%s has unknown type %d, rejecting\n",
409 			      nbuf, obj->Type);
410 		return_VALUE (ENXIO);
411 	    }
412 	}
413     }
414     acpi_tz_getparam(sc, "_CRT", &sc->tz_zone.crt);
415     acpi_tz_getparam(sc, "_HOT", &sc->tz_zone.hot);
416     sc->tz_zone.psl.Length = ACPI_ALLOCATE_BUFFER;
417     sc->tz_zone.psl.Pointer = NULL;
418     AcpiEvaluateObject(sc->tz_handle, "_PSL", NULL, &sc->tz_zone.psl);
419     acpi_tz_getparam(sc, "_PSV", &sc->tz_zone.psv);
420     acpi_tz_getparam(sc, "_TC1", &sc->tz_zone.tc1);
421     acpi_tz_getparam(sc, "_TC2", &sc->tz_zone.tc2);
422     acpi_tz_getparam(sc, "_TSP", &sc->tz_zone.tsp);
423     acpi_tz_getparam(sc, "_TZP", &sc->tz_zone.tzp);
424 
425     /*
426      * Sanity-check the values we've been given.
427      *
428      * XXX what do we do about systems that give us the same value for
429      *     more than one of these setpoints?
430      */
431     acpi_tz_sanity(sc, &sc->tz_zone.crt, "_CRT");
432     acpi_tz_sanity(sc, &sc->tz_zone.hot, "_HOT");
433     acpi_tz_sanity(sc, &sc->tz_zone.psv, "_PSV");
434     for (i = 0; i < TZ_NUMLEVELS; i++)
435 	acpi_tz_sanity(sc, &sc->tz_zone.ac[i], "_ACx");
436 
437     return_VALUE (0);
438 }
439 
440 static char *aclevel_string[] = {
441     "NONE", "_AC0", "_AC1", "_AC2", "_AC3", "_AC4",
442     "_AC5", "_AC6", "_AC7", "_AC8", "_AC9"
443 };
444 
445 static __inline const char *
446 acpi_tz_aclevel_string(int active)
447 {
448     if (active < -1 || active >= TZ_NUMLEVELS)
449 	return (aclevel_string[0]);
450 
451     return (aclevel_string[active + 1]);
452 }
453 
454 /*
455  * Get the current temperature.
456  */
457 static int
458 acpi_tz_get_temperature(struct acpi_tz_softc *sc)
459 {
460     int		temp;
461     ACPI_STATUS	status;
462 
463     ACPI_FUNCTION_NAME ("acpi_tz_get_temperature");
464 
465     /* Evaluate the thermal zone's _TMP method. */
466     status = acpi_GetInteger(sc->tz_handle, acpi_tz_tmp_name, &temp);
467     if (ACPI_FAILURE(status)) {
468 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
469 	    "error fetching current temperature -- %s\n",
470 	     AcpiFormatException(status));
471 	return (FALSE);
472     }
473 
474     /* Check it for validity. */
475     acpi_tz_sanity(sc, &temp, acpi_tz_tmp_name);
476     if (temp == -1)
477 	return (FALSE);
478 
479     ACPI_DEBUG_PRINT((ACPI_DB_VALUES, "got %d.%dC\n", TZ_KELVTOC(temp)));
480     sc->tz_temperature = temp;
481     return (TRUE);
482 }
483 
484 /*
485  * Evaluate the condition of a thermal zone, take appropriate actions.
486  */
487 static void
488 acpi_tz_monitor(void *Context)
489 {
490     struct acpi_tz_softc *sc;
491     struct	timespec curtime;
492     int		temp;
493     int		i;
494     int		newactive, newflags;
495 
496     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
497 
498     sc = (struct acpi_tz_softc *)Context;
499 
500     /* Get the current temperature. */
501     if (!acpi_tz_get_temperature(sc)) {
502 	/* XXX disable zone? go to max cooling? */
503 	return_VOID;
504     }
505     temp = sc->tz_temperature;
506 
507     /*
508      * Work out what we ought to be doing right now.
509      *
510      * Note that the _ACx levels sort from hot to cold.
511      */
512     newactive = TZ_ACTIVE_NONE;
513     for (i = TZ_NUMLEVELS - 1; i >= 0; i--) {
514 	if (sc->tz_zone.ac[i] != -1 && temp >= sc->tz_zone.ac[i])
515 	    newactive = i;
516     }
517 
518     /*
519      * We are going to get _ACx level down (colder side), but give a guaranteed
520      * minimum cooling run time if requested.
521      */
522     if (acpi_tz_min_runtime > 0 && sc->tz_active != TZ_ACTIVE_NONE &&
523 	sc->tz_active != TZ_ACTIVE_UNKNOWN &&
524 	(newactive == TZ_ACTIVE_NONE || newactive > sc->tz_active)) {
525 
526 	getnanotime(&curtime);
527 	timespecsub(&curtime, &sc->tz_cooling_started);
528 	if (curtime.tv_sec < acpi_tz_min_runtime)
529 	    newactive = sc->tz_active;
530     }
531 
532     /* Handle user override of active mode */
533     if (sc->tz_requested != TZ_ACTIVE_NONE && (newactive == TZ_ACTIVE_NONE
534         || sc->tz_requested < newactive))
535 	newactive = sc->tz_requested;
536 
537     /* update temperature-related flags */
538     newflags = TZ_THFLAG_NONE;
539     if (sc->tz_zone.psv != -1 && temp >= sc->tz_zone.psv)
540 	newflags |= TZ_THFLAG_PSV;
541     if (sc->tz_zone.hot != -1 && temp >= sc->tz_zone.hot)
542 	newflags |= TZ_THFLAG_HOT;
543     if (sc->tz_zone.crt != -1 && temp >= sc->tz_zone.crt)
544 	newflags |= TZ_THFLAG_CRT;
545 
546     /* If the active cooling state has changed, we have to switch things. */
547     if (sc->tz_active == TZ_ACTIVE_UNKNOWN) {
548 	/*
549 	 * We don't know which cooling device is on or off,
550 	 * so stop them all, because we now know which
551 	 * should be on (if any).
552 	 */
553 	for (i = 0; i < TZ_NUMLEVELS; i++) {
554 	    if (sc->tz_zone.al[i].Pointer != NULL) {
555 		acpi_ForeachPackageObject(
556 		    (ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
557 		    acpi_tz_switch_cooler_off, sc);
558 	    }
559 	}
560 	/* now we know that all devices are off */
561 	sc->tz_active = TZ_ACTIVE_NONE;
562     }
563 
564     if (newactive != sc->tz_active) {
565 	/* Turn off unneeded cooling devices that are on, if any are */
566 	for (i = TZ_ACTIVE_LEVEL(sc->tz_active);
567 	     i < TZ_ACTIVE_LEVEL(newactive); i++) {
568 	    acpi_ForeachPackageObject(
569 		(ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
570 		acpi_tz_switch_cooler_off, sc);
571 	}
572 	/* Turn on cooling devices that are required, if any are */
573 	for (i = TZ_ACTIVE_LEVEL(sc->tz_active) - 1;
574 	     i >= TZ_ACTIVE_LEVEL(newactive); i--) {
575 	    acpi_ForeachPackageObject(
576 		(ACPI_OBJECT *)sc->tz_zone.al[i].Pointer,
577 		acpi_tz_switch_cooler_on, sc);
578 	}
579 
580 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
581 		    "switched from %s to %s: %d.%dC\n",
582 		    acpi_tz_aclevel_string(sc->tz_active),
583 		    acpi_tz_aclevel_string(newactive), TZ_KELVTOC(temp));
584 	sc->tz_active = newactive;
585 	getnanotime(&sc->tz_cooling_started);
586     }
587 
588     /* XXX (de)activate any passive cooling that may be required. */
589 
590     /*
591      * If the temperature is at _HOT or _CRT, increment our event count.
592      * If it has occurred enough times, shutdown the system.  This is
593      * needed because some systems will report an invalid high temperature
594      * for one poll cycle.  It is suspected this is due to the embedded
595      * controller timing out.  A typical value is 138C for one cycle on
596      * a system that is otherwise 65C.
597      *
598      * If we're almost at that threshold, notify the user through devd(8).
599      */
600     if ((newflags & (TZ_THFLAG_HOT | TZ_THFLAG_CRT)) != 0) {
601 	sc->tz_validchecks++;
602 	if (sc->tz_validchecks == TZ_VALIDCHECKS) {
603 	    device_printf(sc->tz_dev,
604 		"WARNING - current temperature (%d.%dC) exceeds safe limits\n",
605 		TZ_KELVTOC(sc->tz_temperature));
606 	    shutdown_nice(RB_POWEROFF);
607 	} else if (sc->tz_validchecks == TZ_NOTIFYCOUNT)
608 	    acpi_UserNotify("Thermal", sc->tz_handle, TZ_NOTIFY_CRITICAL);
609     } else {
610 	sc->tz_validchecks = 0;
611     }
612     sc->tz_thflags = newflags;
613 
614     return_VOID;
615 }
616 
617 /*
618  * Given an object, verify that it's a reference to a device of some sort,
619  * and try to switch it off.
620  */
621 static void
622 acpi_tz_switch_cooler_off(ACPI_OBJECT *obj, void *arg)
623 {
624     ACPI_HANDLE			cooler;
625 
626     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
627 
628     cooler = acpi_GetReference(NULL, obj);
629     if (cooler == NULL) {
630 	ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "can't get handle\n"));
631 	return_VOID;
632     }
633 
634     ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "called to turn %s off\n",
635 		     acpi_name(cooler)));
636     acpi_pwr_switch_consumer(cooler, ACPI_STATE_D3);
637 
638     return_VOID;
639 }
640 
641 /*
642  * Given an object, verify that it's a reference to a device of some sort,
643  * and try to switch it on.
644  *
645  * XXX replication of off/on function code is bad.
646  */
647 static void
648 acpi_tz_switch_cooler_on(ACPI_OBJECT *obj, void *arg)
649 {
650     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)arg;
651     ACPI_HANDLE			cooler;
652     ACPI_STATUS			status;
653 
654     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
655 
656     cooler = acpi_GetReference(NULL, obj);
657     if (cooler == NULL) {
658 	ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "can't get handle\n"));
659 	return_VOID;
660     }
661 
662     ACPI_DEBUG_PRINT((ACPI_DB_OBJECTS, "called to turn %s on\n",
663 		     acpi_name(cooler)));
664     status = acpi_pwr_switch_consumer(cooler, ACPI_STATE_D0);
665     if (ACPI_FAILURE(status)) {
666 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
667 		    "failed to activate %s - %s\n", acpi_name(cooler),
668 		    AcpiFormatException(status));
669     }
670 
671     return_VOID;
672 }
673 
674 /*
675  * Read/debug-print a parameter, default it to -1.
676  */
677 static void
678 acpi_tz_getparam(struct acpi_tz_softc *sc, char *node, int *data)
679 {
680 
681     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
682 
683     if (ACPI_FAILURE(acpi_GetInteger(sc->tz_handle, node, data))) {
684 	*data = -1;
685     } else {
686 	ACPI_DEBUG_PRINT((ACPI_DB_VALUES, "%s.%s = %d\n",
687 			 acpi_name(sc->tz_handle), node, *data));
688     }
689 
690     return_VOID;
691 }
692 
693 /*
694  * Sanity-check a temperature value.  Assume that setpoints
695  * should be between 0C and 200C.
696  */
697 static void
698 acpi_tz_sanity(struct acpi_tz_softc *sc, int *val, char *what)
699 {
700     if (*val != -1 && (*val < TZ_ZEROC || *val > TZ_ZEROC + 2000)) {
701 	/*
702 	 * If the value we are checking is _TMP, warn the user only
703 	 * once. This avoids spamming messages if, for instance, the
704 	 * sensor is broken and always returns an invalid temperature.
705 	 *
706 	 * This is only done for _TMP; other values always emit a
707 	 * warning.
708 	 */
709 	if (what != acpi_tz_tmp_name || !sc->tz_insane_tmp_notified) {
710 	    device_printf(sc->tz_dev, "%s value is absurd, ignored (%d.%dC)\n",
711 			  what, TZ_KELVTOC(*val));
712 
713 	    /* Don't warn the user again if the read value doesn't improve. */
714 	    if (what == acpi_tz_tmp_name)
715 		sc->tz_insane_tmp_notified = 1;
716 	}
717 	*val = -1;
718 	return;
719     }
720 
721     /* This value is correct. Warn if it's incorrect again. */
722     if (what == acpi_tz_tmp_name)
723 	sc->tz_insane_tmp_notified = 0;
724 }
725 
726 /*
727  * Respond to a sysctl on the active state node.
728  */
729 static int
730 acpi_tz_active_sysctl(SYSCTL_HANDLER_ARGS)
731 {
732     struct acpi_tz_softc	*sc;
733     int				active;
734     int		 		error;
735 
736     sc = (struct acpi_tz_softc *)oidp->oid_arg1;
737     active = sc->tz_active;
738     error = sysctl_handle_int(oidp, &active, 0, req);
739 
740     /* Error or no new value */
741     if (error != 0 || req->newptr == NULL)
742 	return (error);
743     if (active < -1 || active >= TZ_NUMLEVELS)
744 	return (EINVAL);
745 
746     /* Set new preferred level and re-switch */
747     sc->tz_requested = active;
748     acpi_tz_signal(sc, 0);
749     return (0);
750 }
751 
752 static int
753 acpi_tz_cooling_sysctl(SYSCTL_HANDLER_ARGS)
754 {
755     struct acpi_tz_softc *sc;
756     int enabled, error;
757 
758     sc = (struct acpi_tz_softc *)oidp->oid_arg1;
759     enabled = sc->tz_cooling_enabled;
760     error = sysctl_handle_int(oidp, &enabled, 0, req);
761 
762     /* Error or no new value */
763     if (error != 0 || req->newptr == NULL)
764 	return (error);
765     if (enabled != TRUE && enabled != FALSE)
766 	return (EINVAL);
767 
768     if (enabled) {
769 	if (acpi_tz_cooling_is_available(sc))
770 	    error = acpi_tz_cooling_thread_start(sc);
771 	else
772 	    error = ENODEV;
773 	if (error)
774 	    enabled = FALSE;
775     }
776     sc->tz_cooling_enabled = enabled;
777     return (error);
778 }
779 
780 static int
781 acpi_tz_temp_sysctl(SYSCTL_HANDLER_ARGS)
782 {
783     struct acpi_tz_softc	*sc;
784     int				temp, *temp_ptr;
785     int		 		error;
786 
787     sc = oidp->oid_arg1;
788     temp_ptr = (int *)((uintptr_t)sc + oidp->oid_arg2);
789     temp = *temp_ptr;
790     error = sysctl_handle_int(oidp, &temp, 0, req);
791 
792     /* Error or no new value */
793     if (error != 0 || req->newptr == NULL)
794 	return (error);
795 
796     /* Only allow changing settings if override is set. */
797     if (!acpi_tz_override)
798 	return (EPERM);
799 
800     /* Check user-supplied value for sanity. */
801     acpi_tz_sanity(sc, &temp, "user-supplied temp");
802     if (temp == -1)
803 	return (EINVAL);
804 
805     *temp_ptr = temp;
806     return (0);
807 }
808 
809 static int
810 acpi_tz_passive_sysctl(SYSCTL_HANDLER_ARGS)
811 {
812     struct acpi_tz_softc	*sc;
813     int				val, *val_ptr;
814     int				error;
815 
816     sc = oidp->oid_arg1;
817     val_ptr = (int *)((uintptr_t)sc + oidp->oid_arg2);
818     val = *val_ptr;
819     error = sysctl_handle_int(oidp, &val, 0, req);
820 
821     /* Error or no new value */
822     if (error != 0 || req->newptr == NULL)
823 	return (error);
824 
825     /* Only allow changing settings if override is set. */
826     if (!acpi_tz_override)
827 	return (EPERM);
828 
829     *val_ptr = val;
830     return (0);
831 }
832 
833 static void
834 acpi_tz_notify_handler(ACPI_HANDLE h, UINT32 notify, void *context)
835 {
836     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)context;
837 
838     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
839 
840     switch (notify) {
841     case TZ_NOTIFY_TEMPERATURE:
842 	/* Temperature change occurred */
843 	acpi_tz_signal(sc, 0);
844 	break;
845     case TZ_NOTIFY_DEVICES:
846     case TZ_NOTIFY_LEVELS:
847 	/* Zone devices/setpoints changed */
848 	acpi_tz_signal(sc, TZ_FLAG_GETSETTINGS);
849 	break;
850     default:
851 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
852 		    "unknown Notify event 0x%x\n", notify);
853 	break;
854     }
855 
856     acpi_UserNotify("Thermal", h, notify);
857 
858     return_VOID;
859 }
860 
861 static void
862 acpi_tz_signal(struct acpi_tz_softc *sc, int flags)
863 {
864     ACPI_LOCK(thermal);
865     sc->tz_flags |= flags;
866     ACPI_UNLOCK(thermal);
867     wakeup(&acpi_tz_proc);
868 }
869 
870 /*
871  * Notifies can be generated asynchronously but have also been seen to be
872  * triggered by other thermal methods.  One system generates a notify of
873  * 0x81 when the fan is turned on or off.  Another generates it when _SCP
874  * is called.  To handle these situations, we check the zone via
875  * acpi_tz_monitor() before evaluating changes to setpoints or the cooling
876  * policy.
877  */
878 static void
879 acpi_tz_timeout(struct acpi_tz_softc *sc, int flags)
880 {
881 
882     /* Check the current temperature and take action based on it */
883     acpi_tz_monitor(sc);
884 
885     /* If requested, get the power profile settings. */
886     if (flags & TZ_FLAG_GETPROFILE)
887 	acpi_tz_power_profile(sc);
888 
889     /*
890      * If requested, check for new devices/setpoints.  After finding them,
891      * check if we need to switch fans based on the new values.
892      */
893     if (flags & TZ_FLAG_GETSETTINGS) {
894 	acpi_tz_establish(sc);
895 	acpi_tz_monitor(sc);
896     }
897 
898     /* XXX passive cooling actions? */
899 }
900 
901 /*
902  * System power profile may have changed; fetch and notify the
903  * thermal zone accordingly.
904  *
905  * Since this can be called from an arbitrary eventhandler, it needs
906  * to get the ACPI lock itself.
907  */
908 static void
909 acpi_tz_power_profile(void *arg)
910 {
911     ACPI_STATUS			status;
912     struct acpi_tz_softc	*sc = (struct acpi_tz_softc *)arg;
913     int				state;
914 
915     state = power_profile_get_state();
916     if (state != POWER_PROFILE_PERFORMANCE && state != POWER_PROFILE_ECONOMY)
917 	return;
918 
919     /* check that we haven't decided there's no _SCP method */
920     if ((sc->tz_flags & TZ_FLAG_NO_SCP) == 0) {
921 
922 	/* Call _SCP to set the new profile */
923 	status = acpi_SetInteger(sc->tz_handle, "_SCP",
924 	    (state == POWER_PROFILE_PERFORMANCE) ? 0 : 1);
925 	if (ACPI_FAILURE(status)) {
926 	    if (status != AE_NOT_FOUND)
927 		ACPI_VPRINT(sc->tz_dev,
928 			    acpi_device_get_parent_softc(sc->tz_dev),
929 			    "can't evaluate %s._SCP - %s\n",
930 			    acpi_name(sc->tz_handle),
931 			    AcpiFormatException(status));
932 	    sc->tz_flags |= TZ_FLAG_NO_SCP;
933 	} else {
934 	    /* We have to re-evaluate the entire zone now */
935 	    acpi_tz_signal(sc, TZ_FLAG_GETSETTINGS);
936 	}
937     }
938 }
939 
940 /*
941  * Thermal zone monitor thread.
942  */
943 static void
944 acpi_tz_thread(void *arg)
945 {
946     device_t	*devs;
947     int		devcount, i;
948     int		flags;
949     struct acpi_tz_softc **sc;
950 
951     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
952 
953     devs = NULL;
954     devcount = 0;
955     sc = NULL;
956 
957     for (;;) {
958 	/* If the number of devices has changed, re-evaluate. */
959 	if (devclass_get_count(acpi_tz_devclass) != devcount) {
960 	    if (devs != NULL) {
961 		free(devs, M_TEMP);
962 		free(sc, M_TEMP);
963 	    }
964 	    devclass_get_devices(acpi_tz_devclass, &devs, &devcount);
965 	    sc = malloc(sizeof(struct acpi_tz_softc *) * devcount, M_TEMP,
966 			M_WAITOK | M_ZERO);
967 	    for (i = 0; i < devcount; i++)
968 		sc[i] = device_get_softc(devs[i]);
969 	}
970 
971 	/* Check for temperature events and act on them. */
972 	for (i = 0; i < devcount; i++) {
973 	    ACPI_LOCK(thermal);
974 	    flags = sc[i]->tz_flags;
975 	    sc[i]->tz_flags &= TZ_FLAG_NO_SCP;
976 	    ACPI_UNLOCK(thermal);
977 	    acpi_tz_timeout(sc[i], flags);
978 	}
979 
980 	/* If more work to do, don't go to sleep yet. */
981 	ACPI_LOCK(thermal);
982 	for (i = 0; i < devcount; i++) {
983 	    if (sc[i]->tz_flags & ~TZ_FLAG_NO_SCP)
984 		break;
985 	}
986 
987 	/*
988 	 * If we have no more work, sleep for a while, setting PDROP so that
989 	 * the mutex will not be reacquired.  Otherwise, drop the mutex and
990 	 * loop to handle more events.
991 	 */
992 	if (i == devcount)
993 	    msleep(&acpi_tz_proc, &thermal_mutex, PZERO | PDROP, "tzpoll",
994 		hz * acpi_tz_polling_rate);
995 	else
996 	    ACPI_UNLOCK(thermal);
997     }
998 }
999 
1000 static int
1001 acpi_tz_cpufreq_restore(struct acpi_tz_softc *sc)
1002 {
1003     device_t dev;
1004     int error;
1005 
1006     if (!sc->tz_cooling_updated)
1007 	return (0);
1008     if ((dev = devclass_get_device(devclass_find("cpufreq"), 0)) == NULL)
1009 	return (ENXIO);
1010     ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
1011 	"temperature %d.%dC: resuming previous clock speed (%d MHz)\n",
1012 	TZ_KELVTOC(sc->tz_temperature), sc->tz_cooling_saved_freq);
1013     error = CPUFREQ_SET(dev, NULL, CPUFREQ_PRIO_KERN);
1014     if (error == 0)
1015 	sc->tz_cooling_updated = FALSE;
1016     return (error);
1017 }
1018 
1019 static int
1020 acpi_tz_cpufreq_update(struct acpi_tz_softc *sc, int req)
1021 {
1022     device_t dev;
1023     struct cf_level *levels;
1024     int num_levels, error, freq, desired_freq, perf, i;
1025 
1026     levels = malloc(CPUFREQ_MAX_LEVELS * sizeof(*levels), M_TEMP, M_NOWAIT);
1027     if (levels == NULL)
1028 	return (ENOMEM);
1029 
1030     /*
1031      * Find the main device, cpufreq0.  We don't yet support independent
1032      * CPU frequency control on SMP.
1033      */
1034     if ((dev = devclass_get_device(devclass_find("cpufreq"), 0)) == NULL) {
1035 	error = ENXIO;
1036 	goto out;
1037     }
1038 
1039     /* Get the current frequency. */
1040     error = CPUFREQ_GET(dev, &levels[0]);
1041     if (error)
1042 	goto out;
1043     freq = levels[0].total_set.freq;
1044 
1045     /* Get the current available frequency levels. */
1046     num_levels = CPUFREQ_MAX_LEVELS;
1047     error = CPUFREQ_LEVELS(dev, levels, &num_levels);
1048     if (error) {
1049 	if (error == E2BIG)
1050 	    printf("cpufreq: need to increase CPUFREQ_MAX_LEVELS\n");
1051 	goto out;
1052     }
1053 
1054     /* Calculate the desired frequency as a percent of the max frequency. */
1055     perf = 100 * freq / levels[0].total_set.freq - req;
1056     if (perf < 0)
1057 	perf = 0;
1058     else if (perf > 100)
1059 	perf = 100;
1060     desired_freq = levels[0].total_set.freq * perf / 100;
1061 
1062     if (desired_freq < freq) {
1063 	/* Find the closest available frequency, rounding down. */
1064 	for (i = 0; i < num_levels; i++)
1065 	    if (levels[i].total_set.freq <= desired_freq)
1066 		break;
1067 
1068 	/* If we didn't find a relevant setting, use the lowest. */
1069 	if (i == num_levels)
1070 	    i--;
1071     } else {
1072 	/* If we didn't decrease frequency yet, don't increase it. */
1073 	if (!sc->tz_cooling_updated) {
1074 	    sc->tz_cooling_active = FALSE;
1075 	    goto out;
1076 	}
1077 
1078 	/* Use saved cpu frequency as maximum value. */
1079 	if (desired_freq > sc->tz_cooling_saved_freq)
1080 	    desired_freq = sc->tz_cooling_saved_freq;
1081 
1082 	/* Find the closest available frequency, rounding up. */
1083 	for (i = num_levels - 1; i >= 0; i--)
1084 	    if (levels[i].total_set.freq >= desired_freq)
1085 		break;
1086 
1087 	/* If we didn't find a relevant setting, use the highest. */
1088 	if (i == -1)
1089 	    i++;
1090 
1091 	/* If we're going to the highest frequency, restore the old setting. */
1092 	if (i == 0 || desired_freq == sc->tz_cooling_saved_freq) {
1093 	    error = acpi_tz_cpufreq_restore(sc);
1094 	    if (error == 0)
1095 		sc->tz_cooling_active = FALSE;
1096 	    goto out;
1097 	}
1098     }
1099 
1100     /* If we are going to a new frequency, activate it. */
1101     if (levels[i].total_set.freq != freq) {
1102 	ACPI_VPRINT(sc->tz_dev, acpi_device_get_parent_softc(sc->tz_dev),
1103 	    "temperature %d.%dC: %screasing clock speed "
1104 	    "from %d MHz to %d MHz\n",
1105 	    TZ_KELVTOC(sc->tz_temperature),
1106 	    (freq > levels[i].total_set.freq) ? "de" : "in",
1107 	    freq, levels[i].total_set.freq);
1108 	error = CPUFREQ_SET(dev, &levels[i], CPUFREQ_PRIO_KERN);
1109 	if (error == 0 && !sc->tz_cooling_updated) {
1110 	    sc->tz_cooling_saved_freq = freq;
1111 	    sc->tz_cooling_updated = TRUE;
1112 	}
1113     }
1114 
1115 out:
1116     if (levels)
1117 	free(levels, M_TEMP);
1118     return (error);
1119 }
1120 
1121 /*
1122  * Passive cooling thread; monitors current temperature according to the
1123  * cooling interval and calculates whether to scale back CPU frequency.
1124  */
1125 static void
1126 acpi_tz_cooling_thread(void *arg)
1127 {
1128     struct acpi_tz_softc *sc;
1129     int error, perf, curr_temp, prev_temp;
1130 
1131     ACPI_FUNCTION_TRACE((char *)(uintptr_t)__func__);
1132 
1133     sc = (struct acpi_tz_softc *)arg;
1134 
1135     prev_temp = sc->tz_temperature;
1136     while (sc->tz_cooling_enabled) {
1137 	if (sc->tz_cooling_active)
1138 	    (void)acpi_tz_get_temperature(sc);
1139 	curr_temp = sc->tz_temperature;
1140 	if (curr_temp >= sc->tz_zone.psv)
1141 	    sc->tz_cooling_active = TRUE;
1142 	if (sc->tz_cooling_active) {
1143 	    perf = sc->tz_zone.tc1 * (curr_temp - prev_temp) +
1144 		   sc->tz_zone.tc2 * (curr_temp - sc->tz_zone.psv);
1145 	    perf /= 10;
1146 
1147 	    if (perf != 0) {
1148 		error = acpi_tz_cpufreq_update(sc, perf);
1149 
1150 		/*
1151 		 * If error and not simply a higher priority setting was
1152 		 * active, disable cooling.
1153 		 */
1154 		if (error != 0 && error != EPERM) {
1155 		    device_printf(sc->tz_dev,
1156 			"failed to set new freq, disabling passive cooling\n");
1157 		    sc->tz_cooling_enabled = FALSE;
1158 		}
1159 	    }
1160 	}
1161 	prev_temp = curr_temp;
1162 	tsleep(&sc->tz_cooling_proc, PZERO, "cooling",
1163 	    hz * sc->tz_zone.tsp / 10);
1164     }
1165     if (sc->tz_cooling_active) {
1166 	acpi_tz_cpufreq_restore(sc);
1167 	sc->tz_cooling_active = FALSE;
1168     }
1169     sc->tz_cooling_proc = NULL;
1170     ACPI_LOCK(thermal);
1171     sc->tz_cooling_proc_running = FALSE;
1172     ACPI_UNLOCK(thermal);
1173     kproc_exit(0);
1174 }
1175 
1176 /*
1177  * TODO: We ignore _PSL (list of cooling devices) since cpufreq enumerates
1178  * all CPUs for us.  However, it's possible in the future _PSL will
1179  * reference non-CPU devices so we may want to support it then.
1180  */
1181 static int
1182 acpi_tz_cooling_is_available(struct acpi_tz_softc *sc)
1183 {
1184     return (sc->tz_zone.tc1 != -1 && sc->tz_zone.tc2 != -1 &&
1185 	sc->tz_zone.tsp != -1 && sc->tz_zone.tsp != 0 &&
1186 	sc->tz_zone.psv != -1);
1187 }
1188 
1189 static int
1190 acpi_tz_cooling_thread_start(struct acpi_tz_softc *sc)
1191 {
1192     int error;
1193 
1194     ACPI_LOCK(thermal);
1195     if (sc->tz_cooling_proc_running) {
1196 	ACPI_UNLOCK(thermal);
1197 	return (0);
1198     }
1199     sc->tz_cooling_proc_running = TRUE;
1200     ACPI_UNLOCK(thermal);
1201     error = 0;
1202     if (sc->tz_cooling_proc == NULL) {
1203 	error = kproc_create(acpi_tz_cooling_thread, sc,
1204 	    &sc->tz_cooling_proc, RFHIGHPID, 0, "acpi_cooling%d",
1205 	    device_get_unit(sc->tz_dev));
1206 	if (error != 0) {
1207 	    device_printf(sc->tz_dev, "could not create thread - %d", error);
1208 	    ACPI_LOCK(thermal);
1209 	    sc->tz_cooling_proc_running = FALSE;
1210 	    ACPI_UNLOCK(thermal);
1211 	}
1212     }
1213     return (error);
1214 }
1215