1 // SPDX-License-Identifier: GPL-2.0 2 /* bbc_envctrl.c: UltraSPARC-III environment control driver. 3 * 4 * Copyright (C) 2001, 2008 David S. Miller (davem@davemloft.net) 5 */ 6 7 #include <linux/kthread.h> 8 #include <linux/delay.h> 9 #include <linux/kmod.h> 10 #include <linux/reboot.h> 11 #include <linux/of.h> 12 #include <linux/slab.h> 13 #include <linux/of_device.h> 14 #include <asm/oplib.h> 15 16 #include "bbc_i2c.h" 17 #include "max1617.h" 18 19 #undef ENVCTRL_TRACE 20 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. 24 */ 25 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. 30 * 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 36 * as well. 37 * 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. 46 * 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). 52 * 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. 60 */ 61 62 /* These settings are in Celsius. We use these defaults only 63 * if we cannot interrogate the cpu-fru SEEPROM. 64 */ 65 struct temp_limits { 66 s8 high_pwroff, high_shutdown, high_warn; 67 s8 low_warn, low_shutdown, low_pwroff; 68 }; 69 70 static struct temp_limits cpu_temp_limits[2] = { 71 { 100, 85, 80, 5, -5, -10 }, 72 { 100, 85, 80, 5, -5, -10 }, 73 }; 74 75 static struct temp_limits amb_temp_limits[2] = { 76 { 65, 55, 40, 5, -5, -10 }, 77 { 65, 55, 40, 5, -5, -10 }, 78 }; 79 80 static LIST_HEAD(all_temps); 81 static LIST_HEAD(all_fans); 82 83 #define CPU_FAN_REG 0xf0 84 #define SYS_FAN_REG 0xf2 85 #define PSUPPLY_FAN_REG 0xf4 86 87 #define FAN_SPEED_MIN 0x0c 88 #define FAN_SPEED_MAX 0x3f 89 90 #define PSUPPLY_FAN_ON 0x1f 91 #define PSUPPLY_FAN_OFF 0x00 92 93 static void set_fan_speeds(struct bbc_fan_control *fp) 94 { 95 /* Put temperatures into range so we don't mis-program 96 * the hardware. 97 */ 98 if (fp->cpu_fan_speed < FAN_SPEED_MIN) 99 fp->cpu_fan_speed = FAN_SPEED_MIN; 100 if (fp->cpu_fan_speed > FAN_SPEED_MAX) 101 fp->cpu_fan_speed = FAN_SPEED_MAX; 102 if (fp->system_fan_speed < FAN_SPEED_MIN) 103 fp->system_fan_speed = FAN_SPEED_MIN; 104 if (fp->system_fan_speed > FAN_SPEED_MAX) 105 fp->system_fan_speed = FAN_SPEED_MAX; 106 #ifdef ENVCTRL_TRACE 107 printk("fan%d: Changed fan speed to cpu(%02x) sys(%02x)\n", 108 fp->index, 109 fp->cpu_fan_speed, fp->system_fan_speed); 110 #endif 111 112 bbc_i2c_writeb(fp->client, fp->cpu_fan_speed, CPU_FAN_REG); 113 bbc_i2c_writeb(fp->client, fp->system_fan_speed, SYS_FAN_REG); 114 bbc_i2c_writeb(fp->client, 115 (fp->psupply_fan_on ? 116 PSUPPLY_FAN_ON : PSUPPLY_FAN_OFF), 117 PSUPPLY_FAN_REG); 118 } 119 120 static void get_current_temps(struct bbc_cpu_temperature *tp) 121 { 122 tp->prev_amb_temp = tp->curr_amb_temp; 123 bbc_i2c_readb(tp->client, 124 (unsigned char *) &tp->curr_amb_temp, 125 MAX1617_AMB_TEMP); 126 tp->prev_cpu_temp = tp->curr_cpu_temp; 127 bbc_i2c_readb(tp->client, 128 (unsigned char *) &tp->curr_cpu_temp, 129 MAX1617_CPU_TEMP); 130 #ifdef ENVCTRL_TRACE 131 printk("temp%d: cpu(%d C) amb(%d C)\n", 132 tp->index, 133 (int) tp->curr_cpu_temp, (int) tp->curr_amb_temp); 134 #endif 135 } 136 137 138 static void do_envctrl_shutdown(struct bbc_cpu_temperature *tp) 139 { 140 static int shutting_down = 0; 141 char *type = "???"; 142 s8 val = -1; 143 144 if (shutting_down != 0) 145 return; 146 147 if (tp->curr_amb_temp >= amb_temp_limits[tp->index].high_shutdown || 148 tp->curr_amb_temp < amb_temp_limits[tp->index].low_shutdown) { 149 type = "ambient"; 150 val = tp->curr_amb_temp; 151 } else if (tp->curr_cpu_temp >= cpu_temp_limits[tp->index].high_shutdown || 152 tp->curr_cpu_temp < cpu_temp_limits[tp->index].low_shutdown) { 153 type = "CPU"; 154 val = tp->curr_cpu_temp; 155 } 156 157 printk(KERN_CRIT "temp%d: Outside of safe %s " 158 "operating temperature, %d C.\n", 159 tp->index, type, val); 160 161 printk(KERN_CRIT "kenvctrld: Shutting down the system now.\n"); 162 163 shutting_down = 1; 164 orderly_poweroff(true); 165 } 166 167 #define WARN_INTERVAL (30 * HZ) 168 169 static void analyze_ambient_temp(struct bbc_cpu_temperature *tp, unsigned long *last_warn, int tick) 170 { 171 int ret = 0; 172 173 if (time_after(jiffies, (*last_warn + WARN_INTERVAL))) { 174 if (tp->curr_amb_temp >= 175 amb_temp_limits[tp->index].high_warn) { 176 printk(KERN_WARNING "temp%d: " 177 "Above safe ambient operating temperature, %d C.\n", 178 tp->index, (int) tp->curr_amb_temp); 179 ret = 1; 180 } else if (tp->curr_amb_temp < 181 amb_temp_limits[tp->index].low_warn) { 182 printk(KERN_WARNING "temp%d: " 183 "Below safe ambient operating temperature, %d C.\n", 184 tp->index, (int) tp->curr_amb_temp); 185 ret = 1; 186 } 187 if (ret) 188 *last_warn = jiffies; 189 } else if (tp->curr_amb_temp >= amb_temp_limits[tp->index].high_warn || 190 tp->curr_amb_temp < amb_temp_limits[tp->index].low_warn) 191 ret = 1; 192 193 /* Now check the shutdown limits. */ 194 if (tp->curr_amb_temp >= amb_temp_limits[tp->index].high_shutdown || 195 tp->curr_amb_temp < amb_temp_limits[tp->index].low_shutdown) { 196 do_envctrl_shutdown(tp); 197 ret = 1; 198 } 199 200 if (ret) { 201 tp->fan_todo[FAN_AMBIENT] = FAN_FULLBLAST; 202 } else if ((tick & (8 - 1)) == 0) { 203 s8 amb_goal_hi = amb_temp_limits[tp->index].high_warn - 10; 204 s8 amb_goal_lo; 205 206 amb_goal_lo = amb_goal_hi - 3; 207 208 /* We do not try to avoid 'too cold' events. Basically we 209 * only try to deal with over-heating and fan noise reduction. 210 */ 211 if (tp->avg_amb_temp < amb_goal_hi) { 212 if (tp->avg_amb_temp >= amb_goal_lo) 213 tp->fan_todo[FAN_AMBIENT] = FAN_SAME; 214 else 215 tp->fan_todo[FAN_AMBIENT] = FAN_SLOWER; 216 } else { 217 tp->fan_todo[FAN_AMBIENT] = FAN_FASTER; 218 } 219 } else { 220 tp->fan_todo[FAN_AMBIENT] = FAN_SAME; 221 } 222 } 223 224 static void analyze_cpu_temp(struct bbc_cpu_temperature *tp, unsigned long *last_warn, int tick) 225 { 226 int ret = 0; 227 228 if (time_after(jiffies, (*last_warn + WARN_INTERVAL))) { 229 if (tp->curr_cpu_temp >= 230 cpu_temp_limits[tp->index].high_warn) { 231 printk(KERN_WARNING "temp%d: " 232 "Above safe CPU operating temperature, %d C.\n", 233 tp->index, (int) tp->curr_cpu_temp); 234 ret = 1; 235 } else if (tp->curr_cpu_temp < 236 cpu_temp_limits[tp->index].low_warn) { 237 printk(KERN_WARNING "temp%d: " 238 "Below safe CPU operating temperature, %d C.\n", 239 tp->index, (int) tp->curr_cpu_temp); 240 ret = 1; 241 } 242 if (ret) 243 *last_warn = jiffies; 244 } else if (tp->curr_cpu_temp >= cpu_temp_limits[tp->index].high_warn || 245 tp->curr_cpu_temp < cpu_temp_limits[tp->index].low_warn) 246 ret = 1; 247 248 /* Now check the shutdown limits. */ 249 if (tp->curr_cpu_temp >= cpu_temp_limits[tp->index].high_shutdown || 250 tp->curr_cpu_temp < cpu_temp_limits[tp->index].low_shutdown) { 251 do_envctrl_shutdown(tp); 252 ret = 1; 253 } 254 255 if (ret) { 256 tp->fan_todo[FAN_CPU] = FAN_FULLBLAST; 257 } else if ((tick & (8 - 1)) == 0) { 258 s8 cpu_goal_hi = cpu_temp_limits[tp->index].high_warn - 10; 259 s8 cpu_goal_lo; 260 261 cpu_goal_lo = cpu_goal_hi - 3; 262 263 /* We do not try to avoid 'too cold' events. Basically we 264 * only try to deal with over-heating and fan noise reduction. 265 */ 266 if (tp->avg_cpu_temp < cpu_goal_hi) { 267 if (tp->avg_cpu_temp >= cpu_goal_lo) 268 tp->fan_todo[FAN_CPU] = FAN_SAME; 269 else 270 tp->fan_todo[FAN_CPU] = FAN_SLOWER; 271 } else { 272 tp->fan_todo[FAN_CPU] = FAN_FASTER; 273 } 274 } else { 275 tp->fan_todo[FAN_CPU] = FAN_SAME; 276 } 277 } 278 279 static void analyze_temps(struct bbc_cpu_temperature *tp, unsigned long *last_warn) 280 { 281 tp->avg_amb_temp = (s8)((int)((int)tp->avg_amb_temp + (int)tp->curr_amb_temp) / 2); 282 tp->avg_cpu_temp = (s8)((int)((int)tp->avg_cpu_temp + (int)tp->curr_cpu_temp) / 2); 283 284 analyze_ambient_temp(tp, last_warn, tp->sample_tick); 285 analyze_cpu_temp(tp, last_warn, tp->sample_tick); 286 287 tp->sample_tick++; 288 } 289 290 static enum fan_action prioritize_fan_action(int which_fan) 291 { 292 struct bbc_cpu_temperature *tp; 293 enum fan_action decision = FAN_STATE_MAX; 294 295 /* Basically, prioritize what the temperature sensors 296 * recommend we do, and perform that action on all the 297 * fans. 298 */ 299 list_for_each_entry(tp, &all_temps, glob_list) { 300 if (tp->fan_todo[which_fan] == FAN_FULLBLAST) { 301 decision = FAN_FULLBLAST; 302 break; 303 } 304 if (tp->fan_todo[which_fan] == FAN_SAME && 305 decision != FAN_FASTER) 306 decision = FAN_SAME; 307 else if (tp->fan_todo[which_fan] == FAN_FASTER) 308 decision = FAN_FASTER; 309 else if (decision != FAN_FASTER && 310 decision != FAN_SAME && 311 tp->fan_todo[which_fan] == FAN_SLOWER) 312 decision = FAN_SLOWER; 313 } 314 if (decision == FAN_STATE_MAX) 315 decision = FAN_SAME; 316 317 return decision; 318 } 319 320 static int maybe_new_ambient_fan_speed(struct bbc_fan_control *fp) 321 { 322 enum fan_action decision = prioritize_fan_action(FAN_AMBIENT); 323 int ret; 324 325 if (decision == FAN_SAME) 326 return 0; 327 328 ret = 1; 329 if (decision == FAN_FULLBLAST) { 330 if (fp->system_fan_speed >= FAN_SPEED_MAX) 331 ret = 0; 332 else 333 fp->system_fan_speed = FAN_SPEED_MAX; 334 } else { 335 if (decision == FAN_FASTER) { 336 if (fp->system_fan_speed >= FAN_SPEED_MAX) 337 ret = 0; 338 else 339 fp->system_fan_speed += 2; 340 } else { 341 int orig_speed = fp->system_fan_speed; 342 343 if (orig_speed <= FAN_SPEED_MIN || 344 orig_speed <= (fp->cpu_fan_speed - 3)) 345 ret = 0; 346 else 347 fp->system_fan_speed -= 1; 348 } 349 } 350 351 return ret; 352 } 353 354 static int maybe_new_cpu_fan_speed(struct bbc_fan_control *fp) 355 { 356 enum fan_action decision = prioritize_fan_action(FAN_CPU); 357 int ret; 358 359 if (decision == FAN_SAME) 360 return 0; 361 362 ret = 1; 363 if (decision == FAN_FULLBLAST) { 364 if (fp->cpu_fan_speed >= FAN_SPEED_MAX) 365 ret = 0; 366 else 367 fp->cpu_fan_speed = FAN_SPEED_MAX; 368 } else { 369 if (decision == FAN_FASTER) { 370 if (fp->cpu_fan_speed >= FAN_SPEED_MAX) 371 ret = 0; 372 else { 373 fp->cpu_fan_speed += 2; 374 if (fp->system_fan_speed < 375 (fp->cpu_fan_speed - 3)) 376 fp->system_fan_speed = 377 fp->cpu_fan_speed - 3; 378 } 379 } else { 380 if (fp->cpu_fan_speed <= FAN_SPEED_MIN) 381 ret = 0; 382 else 383 fp->cpu_fan_speed -= 1; 384 } 385 } 386 387 return ret; 388 } 389 390 static void maybe_new_fan_speeds(struct bbc_fan_control *fp) 391 { 392 int new; 393 394 new = maybe_new_ambient_fan_speed(fp); 395 new |= maybe_new_cpu_fan_speed(fp); 396 397 if (new) 398 set_fan_speeds(fp); 399 } 400 401 static void fans_full_blast(void) 402 { 403 struct bbc_fan_control *fp; 404 405 /* Since we will not be monitoring things anymore, put 406 * the fans on full blast. 407 */ 408 list_for_each_entry(fp, &all_fans, glob_list) { 409 fp->cpu_fan_speed = FAN_SPEED_MAX; 410 fp->system_fan_speed = FAN_SPEED_MAX; 411 fp->psupply_fan_on = 1; 412 set_fan_speeds(fp); 413 } 414 } 415 416 #define POLL_INTERVAL (5 * 1000) 417 static unsigned long last_warning_jiffies; 418 static struct task_struct *kenvctrld_task; 419 420 static int kenvctrld(void *__unused) 421 { 422 printk(KERN_INFO "bbc_envctrl: kenvctrld starting...\n"); 423 last_warning_jiffies = jiffies - WARN_INTERVAL; 424 for (;;) { 425 struct bbc_cpu_temperature *tp; 426 struct bbc_fan_control *fp; 427 428 msleep_interruptible(POLL_INTERVAL); 429 if (kthread_should_stop()) 430 break; 431 432 list_for_each_entry(tp, &all_temps, glob_list) { 433 get_current_temps(tp); 434 analyze_temps(tp, &last_warning_jiffies); 435 } 436 list_for_each_entry(fp, &all_fans, glob_list) 437 maybe_new_fan_speeds(fp); 438 } 439 printk(KERN_INFO "bbc_envctrl: kenvctrld exiting...\n"); 440 441 fans_full_blast(); 442 443 return 0; 444 } 445 446 static void attach_one_temp(struct bbc_i2c_bus *bp, struct platform_device *op, 447 int temp_idx) 448 { 449 struct bbc_cpu_temperature *tp; 450 451 tp = kzalloc(sizeof(*tp), GFP_KERNEL); 452 if (!tp) 453 return; 454 455 INIT_LIST_HEAD(&tp->bp_list); 456 INIT_LIST_HEAD(&tp->glob_list); 457 458 tp->client = bbc_i2c_attach(bp, op); 459 if (!tp->client) { 460 kfree(tp); 461 return; 462 } 463 464 465 tp->index = temp_idx; 466 467 list_add(&tp->glob_list, &all_temps); 468 list_add(&tp->bp_list, &bp->temps); 469 470 /* Tell it to convert once every 5 seconds, clear all cfg 471 * bits. 472 */ 473 bbc_i2c_writeb(tp->client, 0x00, MAX1617_WR_CFG_BYTE); 474 bbc_i2c_writeb(tp->client, 0x02, MAX1617_WR_CVRATE_BYTE); 475 476 /* Program the hard temperature limits into the chip. */ 477 bbc_i2c_writeb(tp->client, amb_temp_limits[tp->index].high_pwroff, 478 MAX1617_WR_AMB_HIGHLIM); 479 bbc_i2c_writeb(tp->client, amb_temp_limits[tp->index].low_pwroff, 480 MAX1617_WR_AMB_LOWLIM); 481 bbc_i2c_writeb(tp->client, cpu_temp_limits[tp->index].high_pwroff, 482 MAX1617_WR_CPU_HIGHLIM); 483 bbc_i2c_writeb(tp->client, cpu_temp_limits[tp->index].low_pwroff, 484 MAX1617_WR_CPU_LOWLIM); 485 486 get_current_temps(tp); 487 tp->prev_cpu_temp = tp->avg_cpu_temp = tp->curr_cpu_temp; 488 tp->prev_amb_temp = tp->avg_amb_temp = tp->curr_amb_temp; 489 490 tp->fan_todo[FAN_AMBIENT] = FAN_SAME; 491 tp->fan_todo[FAN_CPU] = FAN_SAME; 492 } 493 494 static void attach_one_fan(struct bbc_i2c_bus *bp, struct platform_device *op, 495 int fan_idx) 496 { 497 struct bbc_fan_control *fp; 498 499 fp = kzalloc(sizeof(*fp), GFP_KERNEL); 500 if (!fp) 501 return; 502 503 INIT_LIST_HEAD(&fp->bp_list); 504 INIT_LIST_HEAD(&fp->glob_list); 505 506 fp->client = bbc_i2c_attach(bp, op); 507 if (!fp->client) { 508 kfree(fp); 509 return; 510 } 511 512 fp->index = fan_idx; 513 514 list_add(&fp->glob_list, &all_fans); 515 list_add(&fp->bp_list, &bp->fans); 516 517 /* The i2c device controlling the fans is write-only. 518 * So the only way to keep track of the current power 519 * level fed to the fans is via software. Choose half 520 * power for cpu/system and 'on' fo the powersupply fan 521 * and set it now. 522 */ 523 fp->psupply_fan_on = 1; 524 fp->cpu_fan_speed = (FAN_SPEED_MAX - FAN_SPEED_MIN) / 2; 525 fp->cpu_fan_speed += FAN_SPEED_MIN; 526 fp->system_fan_speed = (FAN_SPEED_MAX - FAN_SPEED_MIN) / 2; 527 fp->system_fan_speed += FAN_SPEED_MIN; 528 529 set_fan_speeds(fp); 530 } 531 532 static void destroy_one_temp(struct bbc_cpu_temperature *tp) 533 { 534 bbc_i2c_detach(tp->client); 535 kfree(tp); 536 } 537 538 static void destroy_all_temps(struct bbc_i2c_bus *bp) 539 { 540 struct bbc_cpu_temperature *tp, *tpos; 541 542 list_for_each_entry_safe(tp, tpos, &bp->temps, bp_list) { 543 list_del(&tp->bp_list); 544 list_del(&tp->glob_list); 545 destroy_one_temp(tp); 546 } 547 } 548 549 static void destroy_one_fan(struct bbc_fan_control *fp) 550 { 551 bbc_i2c_detach(fp->client); 552 kfree(fp); 553 } 554 555 static void destroy_all_fans(struct bbc_i2c_bus *bp) 556 { 557 struct bbc_fan_control *fp, *fpos; 558 559 list_for_each_entry_safe(fp, fpos, &bp->fans, bp_list) { 560 list_del(&fp->bp_list); 561 list_del(&fp->glob_list); 562 destroy_one_fan(fp); 563 } 564 } 565 566 int bbc_envctrl_init(struct bbc_i2c_bus *bp) 567 { 568 struct platform_device *op; 569 int temp_index = 0; 570 int fan_index = 0; 571 int devidx = 0; 572 573 while ((op = bbc_i2c_getdev(bp, devidx++)) != NULL) { 574 if (!strcmp(op->dev.of_node->name, "temperature")) 575 attach_one_temp(bp, op, temp_index++); 576 if (!strcmp(op->dev.of_node->name, "fan-control")) 577 attach_one_fan(bp, op, fan_index++); 578 } 579 if (temp_index != 0 && fan_index != 0) { 580 kenvctrld_task = kthread_run(kenvctrld, NULL, "kenvctrld"); 581 if (IS_ERR(kenvctrld_task)) { 582 int err = PTR_ERR(kenvctrld_task); 583 584 kenvctrld_task = NULL; 585 destroy_all_temps(bp); 586 destroy_all_fans(bp); 587 return err; 588 } 589 } 590 591 return 0; 592 } 593 594 void bbc_envctrl_cleanup(struct bbc_i2c_bus *bp) 595 { 596 if (kenvctrld_task) 597 kthread_stop(kenvctrld_task); 598 599 destroy_all_temps(bp); 600 destroy_all_fans(bp); 601 } 602