1 /* $Id: bbc_envctrl.c,v 1.4 2001/04/06 16:48:08 davem Exp $
2 * bbc_envctrl.c: UltraSPARC-III environment control driver.
4 * Copyright (C) 2001 David S. Miller (davem@redhat.com)
7 #include <linux/kernel.h>
8 #include <linux/sched.h>
9 #include <linux/slab.h>
10 #include <asm/oplib.h>
12 #define __KERNEL_SYSCALLS__
14 #include <asm/unistd.h>
21 /* WARNING: Making changes to this driver is very dangerous.
22 * If you misprogram the sensor chips they can
23 * cut the power on you instantly.
26 /* Two temperature sensors exist in the SunBLADE-1000 enclosure.
27 * Both are implemented using max1617 i2c devices. Each max1617
28 * monitors 2 temperatures, one for one of the cpu dies and the other
29 * for the ambient temperature.
31 * The max1617 is capable of being programmed with power-off
32 * temperature values, one low limit and one high limit. These
33 * can be controlled independently for the cpu or ambient temperature.
34 * If a limit is violated, the power is simply shut off. The frequency
35 * with which the max1617 does temperature sampling can be controlled
38 * Three fans exist inside the machine, all three are controlled with
39 * an i2c digital to analog converter. There is a fan directed at the
40 * two processor slots, another for the rest of the enclosure, and the
41 * third is for the power supply. The first two fans may be speed
42 * controlled by changing the voltage fed to them. The third fan may
43 * only be completely off or on. The third fan is meant to only be
44 * disabled/enabled when entering/exiting the lowest power-saving
45 * mode of the machine.
47 * An environmental control kernel thread periodically monitors all
48 * temperature sensors. Based upon the samples it will adjust the
49 * fan speeds to try and keep the system within a certain temperature
50 * range (the goal being to make the fans as quiet as possible without
51 * allowing the system to get too hot).
53 * If the temperature begins to rise/fall outside of the acceptable
54 * operating range, a periodic warning will be sent to the kernel log.
55 * The fans will be put on full blast to attempt to deal with this
56 * situation. After exceeding the acceptable operating range by a
57 * certain threshold, the kernel thread will shut down the system.
58 * Here, the thread is attempting to shut the machine down cleanly
59 * before the hardware based power-off event is triggered.
62 /* These settings are in Celsius. We use these defaults only
63 * if we cannot interrogate the cpu-fru SEEPROM.
66 s8 high_pwroff, high_shutdown, high_warn;
67 s8 low_warn, low_shutdown, low_pwroff;
70 static struct temp_limits cpu_temp_limits[2] = {
71 { 100, 85, 80, 5, -5, -10 },
72 { 100, 85, 80, 5, -5, -10 },
75 static struct temp_limits amb_temp_limits[2] = {
76 { 65, 55, 40, 5, -5, -10 },
77 { 65, 55, 40, 5, -5, -10 },
80 enum fan_action { FAN_SLOWER, FAN_SAME, FAN_FASTER, FAN_FULLBLAST, FAN_STATE_MAX };
82 struct bbc_cpu_temperature {
83 struct bbc_cpu_temperature *next;
85 struct bbc_i2c_client *client;
88 /* Current readings, and history. */
98 enum fan_action fan_todo[2];
103 struct bbc_cpu_temperature *all_bbc_temps;
105 struct bbc_fan_control {
106 struct bbc_fan_control *next;
108 struct bbc_i2c_client *client;
113 int system_fan_speed;
116 struct bbc_fan_control *all_bbc_fans;
118 #define CPU_FAN_REG 0xf0
119 #define SYS_FAN_REG 0xf2
120 #define PSUPPLY_FAN_REG 0xf4
122 #define FAN_SPEED_MIN 0x0c
123 #define FAN_SPEED_MAX 0x3f
125 #define PSUPPLY_FAN_ON 0x1f
126 #define PSUPPLY_FAN_OFF 0x00
128 static void set_fan_speeds(struct bbc_fan_control *fp)
130 /* Put temperatures into range so we don't mis-program
133 if (fp->cpu_fan_speed < FAN_SPEED_MIN)
134 fp->cpu_fan_speed = FAN_SPEED_MIN;
135 if (fp->cpu_fan_speed > FAN_SPEED_MAX)
136 fp->cpu_fan_speed = FAN_SPEED_MAX;
137 if (fp->system_fan_speed < FAN_SPEED_MIN)
138 fp->system_fan_speed = FAN_SPEED_MIN;
139 if (fp->system_fan_speed > FAN_SPEED_MAX)
140 fp->system_fan_speed = FAN_SPEED_MAX;
142 printk("fan%d: Changed fan speed to cpu(%02x) sys(%02x)\n",
144 fp->cpu_fan_speed, fp->system_fan_speed);
147 bbc_i2c_writeb(fp->client, fp->cpu_fan_speed, CPU_FAN_REG);
148 bbc_i2c_writeb(fp->client, fp->system_fan_speed, SYS_FAN_REG);
149 bbc_i2c_writeb(fp->client,
150 (fp->psupply_fan_on ?
151 PSUPPLY_FAN_ON : PSUPPLY_FAN_OFF),
155 static void get_current_temps(struct bbc_cpu_temperature *tp)
157 tp->prev_amb_temp = tp->curr_amb_temp;
158 bbc_i2c_readb(tp->client,
159 (unsigned char *) &tp->curr_amb_temp,
161 tp->prev_cpu_temp = tp->curr_cpu_temp;
162 bbc_i2c_readb(tp->client,
163 (unsigned char *) &tp->curr_cpu_temp,
166 printk("temp%d: cpu(%d C) amb(%d C)\n",
168 (int) tp->curr_cpu_temp, (int) tp->curr_amb_temp);
173 static void do_envctrl_shutdown(struct bbc_cpu_temperature *tp)
175 static int shutting_down = 0;
176 static char *envp[] = { "HOME=/", "TERM=linux", "PATH=/sbin:/usr/sbin:/bin:/usr/bin", NULL };
177 char *argv[] = { "/sbin/shutdown", "-h", "now", NULL };
181 if (shutting_down != 0)
184 if (tp->curr_amb_temp >= amb_temp_limits[tp->index].high_shutdown ||
185 tp->curr_amb_temp < amb_temp_limits[tp->index].low_shutdown) {
187 val = tp->curr_amb_temp;
188 } else if (tp->curr_cpu_temp >= cpu_temp_limits[tp->index].high_shutdown ||
189 tp->curr_cpu_temp < cpu_temp_limits[tp->index].low_shutdown) {
191 val = tp->curr_cpu_temp;
194 printk(KERN_CRIT "temp%d: Outside of safe %s "
195 "operating temperature, %d C.\n",
196 tp->index, type, val);
198 printk(KERN_CRIT "kenvctrld: Shutting down the system now.\n");
201 if (execve("/sbin/shutdown", argv, envp) < 0)
202 printk(KERN_CRIT "envctrl: shutdown execution failed\n");
205 #define WARN_INTERVAL (30 * HZ)
207 static void analyze_ambient_temp(struct bbc_cpu_temperature *tp, unsigned long *last_warn, int tick)
211 if (time_after(jiffies, (*last_warn + WARN_INTERVAL))) {
212 if (tp->curr_amb_temp >=
213 amb_temp_limits[tp->index].high_warn) {
214 printk(KERN_WARNING "temp%d: "
215 "Above safe ambient operating temperature, %d C.\n",
216 tp->index, (int) tp->curr_amb_temp);
218 } else if (tp->curr_amb_temp <
219 amb_temp_limits[tp->index].low_warn) {
220 printk(KERN_WARNING "temp%d: "
221 "Below safe ambient operating temperature, %d C.\n",
222 tp->index, (int) tp->curr_amb_temp);
226 *last_warn = jiffies;
227 } else if (tp->curr_amb_temp >= amb_temp_limits[tp->index].high_warn ||
228 tp->curr_amb_temp < amb_temp_limits[tp->index].low_warn)
231 /* Now check the shutdown limits. */
232 if (tp->curr_amb_temp >= amb_temp_limits[tp->index].high_shutdown ||
233 tp->curr_amb_temp < amb_temp_limits[tp->index].low_shutdown) {
234 do_envctrl_shutdown(tp);
239 tp->fan_todo[FAN_AMBIENT] = FAN_FULLBLAST;
240 } else if ((tick & (8 - 1)) == 0) {
241 s8 amb_goal_hi = amb_temp_limits[tp->index].high_warn - 10;
244 amb_goal_lo = amb_goal_hi - 3;
246 /* We do not try to avoid 'too cold' events. Basically we
247 * only try to deal with over-heating and fan noise reduction.
249 if (tp->avg_amb_temp < amb_goal_hi) {
250 if (tp->avg_amb_temp >= amb_goal_lo)
251 tp->fan_todo[FAN_AMBIENT] = FAN_SAME;
253 tp->fan_todo[FAN_AMBIENT] = FAN_SLOWER;
255 tp->fan_todo[FAN_AMBIENT] = FAN_FASTER;
258 tp->fan_todo[FAN_AMBIENT] = FAN_SAME;
262 static void analyze_cpu_temp(struct bbc_cpu_temperature *tp, unsigned long *last_warn, int tick)
266 if (time_after(jiffies, (*last_warn + WARN_INTERVAL))) {
267 if (tp->curr_cpu_temp >=
268 cpu_temp_limits[tp->index].high_warn) {
269 printk(KERN_WARNING "temp%d: "
270 "Above safe CPU operating temperature, %d C.\n",
271 tp->index, (int) tp->curr_cpu_temp);
273 } else if (tp->curr_cpu_temp <
274 cpu_temp_limits[tp->index].low_warn) {
275 printk(KERN_WARNING "temp%d: "
276 "Below safe CPU operating temperature, %d C.\n",
277 tp->index, (int) tp->curr_cpu_temp);
281 *last_warn = jiffies;
282 } else if (tp->curr_cpu_temp >= cpu_temp_limits[tp->index].high_warn ||
283 tp->curr_cpu_temp < cpu_temp_limits[tp->index].low_warn)
286 /* Now check the shutdown limits. */
287 if (tp->curr_cpu_temp >= cpu_temp_limits[tp->index].high_shutdown ||
288 tp->curr_cpu_temp < cpu_temp_limits[tp->index].low_shutdown) {
289 do_envctrl_shutdown(tp);
294 tp->fan_todo[FAN_CPU] = FAN_FULLBLAST;
295 } else if ((tick & (8 - 1)) == 0) {
296 s8 cpu_goal_hi = cpu_temp_limits[tp->index].high_warn - 10;
299 cpu_goal_lo = cpu_goal_hi - 3;
301 /* We do not try to avoid 'too cold' events. Basically we
302 * only try to deal with over-heating and fan noise reduction.
304 if (tp->avg_cpu_temp < cpu_goal_hi) {
305 if (tp->avg_cpu_temp >= cpu_goal_lo)
306 tp->fan_todo[FAN_CPU] = FAN_SAME;
308 tp->fan_todo[FAN_CPU] = FAN_SLOWER;
310 tp->fan_todo[FAN_CPU] = FAN_FASTER;
313 tp->fan_todo[FAN_CPU] = FAN_SAME;
317 static void analyze_temps(struct bbc_cpu_temperature *tp, unsigned long *last_warn)
319 tp->avg_amb_temp = (s8)((int)((int)tp->avg_amb_temp + (int)tp->curr_amb_temp) / 2);
320 tp->avg_cpu_temp = (s8)((int)((int)tp->avg_cpu_temp + (int)tp->curr_cpu_temp) / 2);
322 analyze_ambient_temp(tp, last_warn, tp->sample_tick);
323 analyze_cpu_temp(tp, last_warn, tp->sample_tick);
328 static enum fan_action prioritize_fan_action(int which_fan)
330 struct bbc_cpu_temperature *tp;
331 enum fan_action decision = FAN_STATE_MAX;
333 /* Basically, prioritize what the temperature sensors
334 * recommend we do, and perform that action on all the
337 for (tp = all_bbc_temps; tp; tp = tp->next) {
338 if (tp->fan_todo[which_fan] == FAN_FULLBLAST) {
339 decision = FAN_FULLBLAST;
342 if (tp->fan_todo[which_fan] == FAN_SAME &&
343 decision != FAN_FASTER)
345 else if (tp->fan_todo[which_fan] == FAN_FASTER)
346 decision = FAN_FASTER;
347 else if (decision != FAN_FASTER &&
348 decision != FAN_SAME &&
349 tp->fan_todo[which_fan] == FAN_SLOWER)
350 decision = FAN_SLOWER;
352 if (decision == FAN_STATE_MAX)
358 static int maybe_new_ambient_fan_speed(struct bbc_fan_control *fp)
360 enum fan_action decision = prioritize_fan_action(FAN_AMBIENT);
363 if (decision == FAN_SAME)
367 if (decision == FAN_FULLBLAST) {
368 if (fp->system_fan_speed >= FAN_SPEED_MAX)
371 fp->system_fan_speed = FAN_SPEED_MAX;
373 if (decision == FAN_FASTER) {
374 if (fp->system_fan_speed >= FAN_SPEED_MAX)
377 fp->system_fan_speed += 2;
379 int orig_speed = fp->system_fan_speed;
381 if (orig_speed <= FAN_SPEED_MIN ||
382 orig_speed <= (fp->cpu_fan_speed - 3))
385 fp->system_fan_speed -= 1;
392 static int maybe_new_cpu_fan_speed(struct bbc_fan_control *fp)
394 enum fan_action decision = prioritize_fan_action(FAN_CPU);
397 if (decision == FAN_SAME)
401 if (decision == FAN_FULLBLAST) {
402 if (fp->cpu_fan_speed >= FAN_SPEED_MAX)
405 fp->cpu_fan_speed = FAN_SPEED_MAX;
407 if (decision == FAN_FASTER) {
408 if (fp->cpu_fan_speed >= FAN_SPEED_MAX)
411 fp->cpu_fan_speed += 2;
412 if (fp->system_fan_speed <
413 (fp->cpu_fan_speed - 3))
414 fp->system_fan_speed =
415 fp->cpu_fan_speed - 3;
418 if (fp->cpu_fan_speed <= FAN_SPEED_MIN)
421 fp->cpu_fan_speed -= 1;
428 static void maybe_new_fan_speeds(struct bbc_fan_control *fp)
432 new = maybe_new_ambient_fan_speed(fp);
433 new |= maybe_new_cpu_fan_speed(fp);
439 static void fans_full_blast(void)
441 struct bbc_fan_control *fp;
443 /* Since we will not be monitoring things anymore, put
444 * the fans on full blast.
446 for (fp = all_bbc_fans; fp; fp = fp->next) {
447 fp->cpu_fan_speed = FAN_SPEED_MAX;
448 fp->system_fan_speed = FAN_SPEED_MAX;
449 fp->psupply_fan_on = 1;
454 #define POLL_INTERVAL (5 * HZ)
455 static unsigned long last_warning_jiffies;
456 static struct task_struct *kenvctrld_task;
458 static int kenvctrld(void *__unused)
460 daemonize("kenvctrld");
461 allow_signal(SIGKILL);
462 kenvctrld_task = current;
464 printk(KERN_INFO "bbc_envctrl: kenvctrld starting...\n");
465 last_warning_jiffies = jiffies - WARN_INTERVAL;
467 struct bbc_cpu_temperature *tp;
468 struct bbc_fan_control *fp;
470 current->state = TASK_INTERRUPTIBLE;
471 schedule_timeout(POLL_INTERVAL);
472 if (signal_pending(current))
475 for (tp = all_bbc_temps; tp; tp = tp->next) {
476 get_current_temps(tp);
477 analyze_temps(tp, &last_warning_jiffies);
479 for (fp = all_bbc_fans; fp; fp = fp->next)
480 maybe_new_fan_speeds(fp);
482 printk(KERN_INFO "bbc_envctrl: kenvctrld exiting...\n");
489 static void attach_one_temp(struct linux_ebus_child *echild, int temp_idx)
491 struct bbc_cpu_temperature *tp = kmalloc(sizeof(*tp), GFP_KERNEL);
495 memset(tp, 0, sizeof(*tp));
496 tp->client = bbc_i2c_attach(echild);
502 tp->index = temp_idx;
504 struct bbc_cpu_temperature **tpp = &all_bbc_temps;
506 tpp = &((*tpp)->next);
511 /* Tell it to convert once every 5 seconds, clear all cfg
514 bbc_i2c_writeb(tp->client, 0x00, MAX1617_WR_CFG_BYTE);
515 bbc_i2c_writeb(tp->client, 0x02, MAX1617_WR_CVRATE_BYTE);
517 /* Program the hard temperature limits into the chip. */
518 bbc_i2c_writeb(tp->client, amb_temp_limits[tp->index].high_pwroff,
519 MAX1617_WR_AMB_HIGHLIM);
520 bbc_i2c_writeb(tp->client, amb_temp_limits[tp->index].low_pwroff,
521 MAX1617_WR_AMB_LOWLIM);
522 bbc_i2c_writeb(tp->client, cpu_temp_limits[tp->index].high_pwroff,
523 MAX1617_WR_CPU_HIGHLIM);
524 bbc_i2c_writeb(tp->client, cpu_temp_limits[tp->index].low_pwroff,
525 MAX1617_WR_CPU_LOWLIM);
527 get_current_temps(tp);
528 tp->prev_cpu_temp = tp->avg_cpu_temp = tp->curr_cpu_temp;
529 tp->prev_amb_temp = tp->avg_amb_temp = tp->curr_amb_temp;
531 tp->fan_todo[FAN_AMBIENT] = FAN_SAME;
532 tp->fan_todo[FAN_CPU] = FAN_SAME;
535 static void attach_one_fan(struct linux_ebus_child *echild, int fan_idx)
537 struct bbc_fan_control *fp = kmalloc(sizeof(*fp), GFP_KERNEL);
541 memset(fp, 0, sizeof(*fp));
542 fp->client = bbc_i2c_attach(echild);
551 struct bbc_fan_control **fpp = &all_bbc_fans;
553 fpp = &((*fpp)->next);
558 /* The i2c device controlling the fans is write-only.
559 * So the only way to keep track of the current power
560 * level fed to the fans is via software. Choose half
561 * power for cpu/system and 'on' fo the powersupply fan
564 fp->psupply_fan_on = 1;
565 fp->cpu_fan_speed = (FAN_SPEED_MAX - FAN_SPEED_MIN) / 2;
566 fp->cpu_fan_speed += FAN_SPEED_MIN;
567 fp->system_fan_speed = (FAN_SPEED_MAX - FAN_SPEED_MIN) / 2;
568 fp->system_fan_speed += FAN_SPEED_MIN;
573 int bbc_envctrl_init(void)
575 struct linux_ebus_child *echild;
581 while ((echild = bbc_i2c_getdev(devidx++)) != NULL) {
582 if (!strcmp(echild->prom_name, "temperature"))
583 attach_one_temp(echild, temp_index++);
584 if (!strcmp(echild->prom_name, "fan-control"))
585 attach_one_fan(echild, fan_index++);
587 if (temp_index != 0 && fan_index != 0)
588 err = kernel_thread(kenvctrld, NULL, CLONE_FS | CLONE_FILES);
592 static void destroy_one_temp(struct bbc_cpu_temperature *tp)
594 bbc_i2c_detach(tp->client);
598 static void destroy_one_fan(struct bbc_fan_control *fp)
600 bbc_i2c_detach(fp->client);
604 void bbc_envctrl_cleanup(void)
606 struct bbc_cpu_temperature *tp;
607 struct bbc_fan_control *fp;
609 if (kenvctrld_task != NULL) {
610 force_sig(SIGKILL, kenvctrld_task);
612 struct task_struct *p;
615 read_lock(&tasklist_lock);
616 for_each_process(p) {
617 if (p == kenvctrld_task) {
622 read_unlock(&tasklist_lock);
625 current->state = TASK_INTERRUPTIBLE;
626 schedule_timeout(HZ);
627 current->state = TASK_RUNNING;
629 kenvctrld_task = NULL;
634 struct bbc_cpu_temperature *next = tp->next;
635 destroy_one_temp(tp);
638 all_bbc_temps = NULL;
642 struct bbc_fan_control *next = fp->next;