mcdi_mon.c (15608B)
1// SPDX-License-Identifier: GPL-2.0-only 2/**************************************************************************** 3 * Driver for Solarflare network controllers and boards 4 * Copyright 2011-2013 Solarflare Communications Inc. 5 */ 6 7#include <linux/bitops.h> 8#include <linux/slab.h> 9#include <linux/hwmon.h> 10#include <linux/stat.h> 11 12#include "net_driver.h" 13#include "mcdi.h" 14#include "mcdi_pcol.h" 15#include "nic.h" 16 17enum efx_hwmon_type { 18 EFX_HWMON_UNKNOWN, 19 EFX_HWMON_TEMP, /* temperature */ 20 EFX_HWMON_COOL, /* cooling device, probably a heatsink */ 21 EFX_HWMON_IN, /* voltage */ 22 EFX_HWMON_CURR, /* current */ 23 EFX_HWMON_POWER, /* power */ 24 EFX_HWMON_TYPES_COUNT 25}; 26 27static const char *const efx_hwmon_unit[EFX_HWMON_TYPES_COUNT] = { 28 [EFX_HWMON_TEMP] = " degC", 29 [EFX_HWMON_COOL] = " rpm", /* though nonsense for a heatsink */ 30 [EFX_HWMON_IN] = " mV", 31 [EFX_HWMON_CURR] = " mA", 32 [EFX_HWMON_POWER] = " W", 33}; 34 35static const struct { 36 const char *label; 37 enum efx_hwmon_type hwmon_type; 38 int port; 39} efx_mcdi_sensor_type[] = { 40#define SENSOR(name, label, hwmon_type, port) \ 41 [MC_CMD_SENSOR_##name] = { label, EFX_HWMON_ ## hwmon_type, port } 42 SENSOR(CONTROLLER_TEMP, "Controller board temp.", TEMP, -1), 43 SENSOR(PHY_COMMON_TEMP, "PHY temp.", TEMP, -1), 44 SENSOR(CONTROLLER_COOLING, "Controller heat sink", COOL, -1), 45 SENSOR(PHY0_TEMP, "PHY temp.", TEMP, 0), 46 SENSOR(PHY0_COOLING, "PHY heat sink", COOL, 0), 47 SENSOR(PHY1_TEMP, "PHY temp.", TEMP, 1), 48 SENSOR(PHY1_COOLING, "PHY heat sink", COOL, 1), 49 SENSOR(IN_1V0, "1.0V supply", IN, -1), 50 SENSOR(IN_1V2, "1.2V supply", IN, -1), 51 SENSOR(IN_1V8, "1.8V supply", IN, -1), 52 SENSOR(IN_2V5, "2.5V supply", IN, -1), 53 SENSOR(IN_3V3, "3.3V supply", IN, -1), 54 SENSOR(IN_12V0, "12.0V supply", IN, -1), 55 SENSOR(IN_1V2A, "1.2V analogue supply", IN, -1), 56 SENSOR(IN_VREF, "Ref. voltage", IN, -1), 57 SENSOR(OUT_VAOE, "AOE FPGA supply", IN, -1), 58 SENSOR(AOE_TEMP, "AOE FPGA temp.", TEMP, -1), 59 SENSOR(PSU_AOE_TEMP, "AOE regulator temp.", TEMP, -1), 60 SENSOR(PSU_TEMP, "Controller regulator temp.", 61 TEMP, -1), 62 SENSOR(FAN_0, "Fan 0", COOL, -1), 63 SENSOR(FAN_1, "Fan 1", COOL, -1), 64 SENSOR(FAN_2, "Fan 2", COOL, -1), 65 SENSOR(FAN_3, "Fan 3", COOL, -1), 66 SENSOR(FAN_4, "Fan 4", COOL, -1), 67 SENSOR(IN_VAOE, "AOE input supply", IN, -1), 68 SENSOR(OUT_IAOE, "AOE output current", CURR, -1), 69 SENSOR(IN_IAOE, "AOE input current", CURR, -1), 70 SENSOR(NIC_POWER, "Board power use", POWER, -1), 71 SENSOR(IN_0V9, "0.9V supply", IN, -1), 72 SENSOR(IN_I0V9, "0.9V supply current", CURR, -1), 73 SENSOR(IN_I1V2, "1.2V supply current", CURR, -1), 74 SENSOR(IN_0V9_ADC, "0.9V supply (ext. ADC)", IN, -1), 75 SENSOR(CONTROLLER_2_TEMP, "Controller board temp. 2", TEMP, -1), 76 SENSOR(VREG_INTERNAL_TEMP, "Regulator die temp.", TEMP, -1), 77 SENSOR(VREG_0V9_TEMP, "0.9V regulator temp.", TEMP, -1), 78 SENSOR(VREG_1V2_TEMP, "1.2V regulator temp.", TEMP, -1), 79 SENSOR(CONTROLLER_VPTAT, 80 "Controller PTAT voltage (int. ADC)", IN, -1), 81 SENSOR(CONTROLLER_INTERNAL_TEMP, 82 "Controller die temp. (int. ADC)", TEMP, -1), 83 SENSOR(CONTROLLER_VPTAT_EXTADC, 84 "Controller PTAT voltage (ext. ADC)", IN, -1), 85 SENSOR(CONTROLLER_INTERNAL_TEMP_EXTADC, 86 "Controller die temp. (ext. ADC)", TEMP, -1), 87 SENSOR(AMBIENT_TEMP, "Ambient temp.", TEMP, -1), 88 SENSOR(AIRFLOW, "Air flow raw", IN, -1), 89 SENSOR(VDD08D_VSS08D_CSR, "0.9V die (int. ADC)", IN, -1), 90 SENSOR(VDD08D_VSS08D_CSR_EXTADC, "0.9V die (ext. ADC)", IN, -1), 91 SENSOR(HOTPOINT_TEMP, "Controller board temp. (hotpoint)", TEMP, -1), 92#undef SENSOR 93}; 94 95static const char *const sensor_status_names[] = { 96 [MC_CMD_SENSOR_STATE_OK] = "OK", 97 [MC_CMD_SENSOR_STATE_WARNING] = "Warning", 98 [MC_CMD_SENSOR_STATE_FATAL] = "Fatal", 99 [MC_CMD_SENSOR_STATE_BROKEN] = "Device failure", 100 [MC_CMD_SENSOR_STATE_NO_READING] = "No reading", 101}; 102 103void efx_siena_mcdi_sensor_event(struct efx_nic *efx, efx_qword_t *ev) 104{ 105 unsigned int type, state, value; 106 enum efx_hwmon_type hwmon_type = EFX_HWMON_UNKNOWN; 107 const char *name = NULL, *state_txt, *unit; 108 109 type = EFX_QWORD_FIELD(*ev, MCDI_EVENT_SENSOREVT_MONITOR); 110 state = EFX_QWORD_FIELD(*ev, MCDI_EVENT_SENSOREVT_STATE); 111 value = EFX_QWORD_FIELD(*ev, MCDI_EVENT_SENSOREVT_VALUE); 112 113 /* Deal gracefully with the board having more drivers than we 114 * know about, but do not expect new sensor states. */ 115 if (type < ARRAY_SIZE(efx_mcdi_sensor_type)) { 116 name = efx_mcdi_sensor_type[type].label; 117 hwmon_type = efx_mcdi_sensor_type[type].hwmon_type; 118 } 119 if (!name) 120 name = "No sensor name available"; 121 EFX_WARN_ON_PARANOID(state >= ARRAY_SIZE(sensor_status_names)); 122 state_txt = sensor_status_names[state]; 123 EFX_WARN_ON_PARANOID(hwmon_type >= EFX_HWMON_TYPES_COUNT); 124 unit = efx_hwmon_unit[hwmon_type]; 125 if (!unit) 126 unit = ""; 127 128 netif_err(efx, hw, efx->net_dev, 129 "Sensor %d (%s) reports condition '%s' for value %d%s\n", 130 type, name, state_txt, value, unit); 131} 132 133#ifdef CONFIG_SFC_SIENA_MCDI_MON 134 135struct efx_mcdi_mon_attribute { 136 struct device_attribute dev_attr; 137 unsigned int index; 138 unsigned int type; 139 enum efx_hwmon_type hwmon_type; 140 unsigned int limit_value; 141 char name[12]; 142}; 143 144static int efx_mcdi_mon_update(struct efx_nic *efx) 145{ 146 struct efx_mcdi_mon *hwmon = efx_mcdi_mon(efx); 147 MCDI_DECLARE_BUF(inbuf, MC_CMD_READ_SENSORS_EXT_IN_LEN); 148 int rc; 149 150 MCDI_SET_QWORD(inbuf, READ_SENSORS_EXT_IN_DMA_ADDR, 151 hwmon->dma_buf.dma_addr); 152 MCDI_SET_DWORD(inbuf, READ_SENSORS_EXT_IN_LENGTH, hwmon->dma_buf.len); 153 154 rc = efx_siena_mcdi_rpc(efx, MC_CMD_READ_SENSORS, 155 inbuf, sizeof(inbuf), NULL, 0, NULL); 156 if (rc == 0) 157 hwmon->last_update = jiffies; 158 return rc; 159} 160 161static int efx_mcdi_mon_get_entry(struct device *dev, unsigned int index, 162 efx_dword_t *entry) 163{ 164 struct efx_nic *efx = dev_get_drvdata(dev->parent); 165 struct efx_mcdi_mon *hwmon = efx_mcdi_mon(efx); 166 int rc; 167 168 BUILD_BUG_ON(MC_CMD_READ_SENSORS_OUT_LEN != 0); 169 170 mutex_lock(&hwmon->update_lock); 171 172 /* Use cached value if last update was < 1 s ago */ 173 if (time_before(jiffies, hwmon->last_update + HZ)) 174 rc = 0; 175 else 176 rc = efx_mcdi_mon_update(efx); 177 178 /* Copy out the requested entry */ 179 *entry = ((efx_dword_t *)hwmon->dma_buf.addr)[index]; 180 181 mutex_unlock(&hwmon->update_lock); 182 183 return rc; 184} 185 186static ssize_t efx_mcdi_mon_show_value(struct device *dev, 187 struct device_attribute *attr, 188 char *buf) 189{ 190 struct efx_mcdi_mon_attribute *mon_attr = 191 container_of(attr, struct efx_mcdi_mon_attribute, dev_attr); 192 efx_dword_t entry; 193 unsigned int value, state; 194 int rc; 195 196 rc = efx_mcdi_mon_get_entry(dev, mon_attr->index, &entry); 197 if (rc) 198 return rc; 199 200 state = EFX_DWORD_FIELD(entry, MC_CMD_SENSOR_VALUE_ENTRY_TYPEDEF_STATE); 201 if (state == MC_CMD_SENSOR_STATE_NO_READING) 202 return -EBUSY; 203 204 value = EFX_DWORD_FIELD(entry, MC_CMD_SENSOR_VALUE_ENTRY_TYPEDEF_VALUE); 205 206 switch (mon_attr->hwmon_type) { 207 case EFX_HWMON_TEMP: 208 /* Convert temperature from degrees to milli-degrees Celsius */ 209 value *= 1000; 210 break; 211 case EFX_HWMON_POWER: 212 /* Convert power from watts to microwatts */ 213 value *= 1000000; 214 break; 215 default: 216 /* No conversion needed */ 217 break; 218 } 219 220 return sprintf(buf, "%u\n", value); 221} 222 223static ssize_t efx_mcdi_mon_show_limit(struct device *dev, 224 struct device_attribute *attr, 225 char *buf) 226{ 227 struct efx_mcdi_mon_attribute *mon_attr = 228 container_of(attr, struct efx_mcdi_mon_attribute, dev_attr); 229 unsigned int value; 230 231 value = mon_attr->limit_value; 232 233 switch (mon_attr->hwmon_type) { 234 case EFX_HWMON_TEMP: 235 /* Convert temperature from degrees to milli-degrees Celsius */ 236 value *= 1000; 237 break; 238 case EFX_HWMON_POWER: 239 /* Convert power from watts to microwatts */ 240 value *= 1000000; 241 break; 242 default: 243 /* No conversion needed */ 244 break; 245 } 246 247 return sprintf(buf, "%u\n", value); 248} 249 250static ssize_t efx_mcdi_mon_show_alarm(struct device *dev, 251 struct device_attribute *attr, 252 char *buf) 253{ 254 struct efx_mcdi_mon_attribute *mon_attr = 255 container_of(attr, struct efx_mcdi_mon_attribute, dev_attr); 256 efx_dword_t entry; 257 int state; 258 int rc; 259 260 rc = efx_mcdi_mon_get_entry(dev, mon_attr->index, &entry); 261 if (rc) 262 return rc; 263 264 state = EFX_DWORD_FIELD(entry, MC_CMD_SENSOR_VALUE_ENTRY_TYPEDEF_STATE); 265 return sprintf(buf, "%d\n", state != MC_CMD_SENSOR_STATE_OK); 266} 267 268static ssize_t efx_mcdi_mon_show_label(struct device *dev, 269 struct device_attribute *attr, 270 char *buf) 271{ 272 struct efx_mcdi_mon_attribute *mon_attr = 273 container_of(attr, struct efx_mcdi_mon_attribute, dev_attr); 274 return sprintf(buf, "%s\n", 275 efx_mcdi_sensor_type[mon_attr->type].label); 276} 277 278static void 279efx_mcdi_mon_add_attr(struct efx_nic *efx, const char *name, 280 ssize_t (*reader)(struct device *, 281 struct device_attribute *, char *), 282 unsigned int index, unsigned int type, 283 unsigned int limit_value) 284{ 285 struct efx_mcdi_mon *hwmon = efx_mcdi_mon(efx); 286 struct efx_mcdi_mon_attribute *attr = &hwmon->attrs[hwmon->n_attrs]; 287 288 strlcpy(attr->name, name, sizeof(attr->name)); 289 attr->index = index; 290 attr->type = type; 291 if (type < ARRAY_SIZE(efx_mcdi_sensor_type)) 292 attr->hwmon_type = efx_mcdi_sensor_type[type].hwmon_type; 293 else 294 attr->hwmon_type = EFX_HWMON_UNKNOWN; 295 attr->limit_value = limit_value; 296 sysfs_attr_init(&attr->dev_attr.attr); 297 attr->dev_attr.attr.name = attr->name; 298 attr->dev_attr.attr.mode = 0444; 299 attr->dev_attr.show = reader; 300 hwmon->group.attrs[hwmon->n_attrs++] = &attr->dev_attr.attr; 301} 302 303int efx_siena_mcdi_mon_probe(struct efx_nic *efx) 304{ 305 unsigned int n_temp = 0, n_cool = 0, n_in = 0, n_curr = 0, n_power = 0; 306 struct efx_mcdi_mon *hwmon = efx_mcdi_mon(efx); 307 MCDI_DECLARE_BUF(inbuf, MC_CMD_SENSOR_INFO_EXT_IN_LEN); 308 MCDI_DECLARE_BUF(outbuf, MC_CMD_SENSOR_INFO_OUT_LENMAX); 309 unsigned int n_pages, n_sensors, n_attrs, page; 310 size_t outlen; 311 char name[12]; 312 u32 mask; 313 int rc, i, j, type; 314 315 /* Find out how many sensors are present */ 316 n_sensors = 0; 317 page = 0; 318 do { 319 MCDI_SET_DWORD(inbuf, SENSOR_INFO_EXT_IN_PAGE, page); 320 321 rc = efx_siena_mcdi_rpc(efx, MC_CMD_SENSOR_INFO, inbuf, 322 sizeof(inbuf), outbuf, sizeof(outbuf), 323 &outlen); 324 if (rc) 325 return rc; 326 if (outlen < MC_CMD_SENSOR_INFO_OUT_LENMIN) 327 return -EIO; 328 329 mask = MCDI_DWORD(outbuf, SENSOR_INFO_OUT_MASK); 330 n_sensors += hweight32(mask & ~(1 << MC_CMD_SENSOR_PAGE0_NEXT)); 331 ++page; 332 } while (mask & (1 << MC_CMD_SENSOR_PAGE0_NEXT)); 333 n_pages = page; 334 335 /* Don't create a device if there are none */ 336 if (n_sensors == 0) 337 return 0; 338 339 rc = efx_siena_alloc_buffer(efx, &hwmon->dma_buf, 340 n_sensors * MC_CMD_SENSOR_VALUE_ENTRY_TYPEDEF_LEN, 341 GFP_KERNEL); 342 if (rc) 343 return rc; 344 345 mutex_init(&hwmon->update_lock); 346 efx_mcdi_mon_update(efx); 347 348 /* Allocate space for the maximum possible number of 349 * attributes for this set of sensors: 350 * value, min, max, crit, alarm and label for each sensor. 351 */ 352 n_attrs = 6 * n_sensors; 353 hwmon->attrs = kcalloc(n_attrs, sizeof(*hwmon->attrs), GFP_KERNEL); 354 if (!hwmon->attrs) { 355 rc = -ENOMEM; 356 goto fail; 357 } 358 hwmon->group.attrs = kcalloc(n_attrs + 1, sizeof(struct attribute *), 359 GFP_KERNEL); 360 if (!hwmon->group.attrs) { 361 rc = -ENOMEM; 362 goto fail; 363 } 364 365 for (i = 0, j = -1, type = -1; ; i++) { 366 enum efx_hwmon_type hwmon_type; 367 const char *hwmon_prefix; 368 unsigned hwmon_index; 369 u16 min1, max1, min2, max2; 370 371 /* Find next sensor type or exit if there is none */ 372 do { 373 type++; 374 375 if ((type % 32) == 0) { 376 page = type / 32; 377 j = -1; 378 if (page == n_pages) 379 goto hwmon_register; 380 381 MCDI_SET_DWORD(inbuf, SENSOR_INFO_EXT_IN_PAGE, 382 page); 383 rc = efx_siena_mcdi_rpc(efx, MC_CMD_SENSOR_INFO, 384 inbuf, sizeof(inbuf), 385 outbuf, sizeof(outbuf), 386 &outlen); 387 if (rc) 388 goto fail; 389 if (outlen < MC_CMD_SENSOR_INFO_OUT_LENMIN) { 390 rc = -EIO; 391 goto fail; 392 } 393 394 mask = (MCDI_DWORD(outbuf, 395 SENSOR_INFO_OUT_MASK) & 396 ~(1 << MC_CMD_SENSOR_PAGE0_NEXT)); 397 398 /* Check again for short response */ 399 if (outlen < 400 MC_CMD_SENSOR_INFO_OUT_LEN(hweight32(mask))) { 401 rc = -EIO; 402 goto fail; 403 } 404 } 405 } while (!(mask & (1 << type % 32))); 406 j++; 407 408 if (type < ARRAY_SIZE(efx_mcdi_sensor_type)) { 409 hwmon_type = efx_mcdi_sensor_type[type].hwmon_type; 410 411 /* Skip sensors specific to a different port */ 412 if (hwmon_type != EFX_HWMON_UNKNOWN && 413 efx_mcdi_sensor_type[type].port >= 0 && 414 efx_mcdi_sensor_type[type].port != 415 efx_port_num(efx)) 416 continue; 417 } else { 418 hwmon_type = EFX_HWMON_UNKNOWN; 419 } 420 421 switch (hwmon_type) { 422 case EFX_HWMON_TEMP: 423 hwmon_prefix = "temp"; 424 hwmon_index = ++n_temp; /* 1-based */ 425 break; 426 case EFX_HWMON_COOL: 427 /* This is likely to be a heatsink, but there 428 * is no convention for representing cooling 429 * devices other than fans. 430 */ 431 hwmon_prefix = "fan"; 432 hwmon_index = ++n_cool; /* 1-based */ 433 break; 434 default: 435 hwmon_prefix = "in"; 436 hwmon_index = n_in++; /* 0-based */ 437 break; 438 case EFX_HWMON_CURR: 439 hwmon_prefix = "curr"; 440 hwmon_index = ++n_curr; /* 1-based */ 441 break; 442 case EFX_HWMON_POWER: 443 hwmon_prefix = "power"; 444 hwmon_index = ++n_power; /* 1-based */ 445 break; 446 } 447 448 min1 = MCDI_ARRAY_FIELD(outbuf, SENSOR_ENTRY, 449 SENSOR_INFO_ENTRY, j, MIN1); 450 max1 = MCDI_ARRAY_FIELD(outbuf, SENSOR_ENTRY, 451 SENSOR_INFO_ENTRY, j, MAX1); 452 min2 = MCDI_ARRAY_FIELD(outbuf, SENSOR_ENTRY, 453 SENSOR_INFO_ENTRY, j, MIN2); 454 max2 = MCDI_ARRAY_FIELD(outbuf, SENSOR_ENTRY, 455 SENSOR_INFO_ENTRY, j, MAX2); 456 457 if (min1 != max1) { 458 snprintf(name, sizeof(name), "%s%u_input", 459 hwmon_prefix, hwmon_index); 460 efx_mcdi_mon_add_attr( 461 efx, name, efx_mcdi_mon_show_value, i, type, 0); 462 463 if (hwmon_type != EFX_HWMON_POWER) { 464 snprintf(name, sizeof(name), "%s%u_min", 465 hwmon_prefix, hwmon_index); 466 efx_mcdi_mon_add_attr( 467 efx, name, efx_mcdi_mon_show_limit, 468 i, type, min1); 469 } 470 471 snprintf(name, sizeof(name), "%s%u_max", 472 hwmon_prefix, hwmon_index); 473 efx_mcdi_mon_add_attr( 474 efx, name, efx_mcdi_mon_show_limit, 475 i, type, max1); 476 477 if (min2 != max2) { 478 /* Assume max2 is critical value. 479 * But we have no good way to expose min2. 480 */ 481 snprintf(name, sizeof(name), "%s%u_crit", 482 hwmon_prefix, hwmon_index); 483 efx_mcdi_mon_add_attr( 484 efx, name, efx_mcdi_mon_show_limit, 485 i, type, max2); 486 } 487 } 488 489 snprintf(name, sizeof(name), "%s%u_alarm", 490 hwmon_prefix, hwmon_index); 491 efx_mcdi_mon_add_attr( 492 efx, name, efx_mcdi_mon_show_alarm, i, type, 0); 493 494 if (type < ARRAY_SIZE(efx_mcdi_sensor_type) && 495 efx_mcdi_sensor_type[type].label) { 496 snprintf(name, sizeof(name), "%s%u_label", 497 hwmon_prefix, hwmon_index); 498 efx_mcdi_mon_add_attr( 499 efx, name, efx_mcdi_mon_show_label, i, type, 0); 500 } 501 } 502 503hwmon_register: 504 hwmon->groups[0] = &hwmon->group; 505 hwmon->device = hwmon_device_register_with_groups(&efx->pci_dev->dev, 506 KBUILD_MODNAME, NULL, 507 hwmon->groups); 508 if (IS_ERR(hwmon->device)) { 509 rc = PTR_ERR(hwmon->device); 510 goto fail; 511 } 512 513 return 0; 514 515fail: 516 efx_siena_mcdi_mon_remove(efx); 517 return rc; 518} 519 520void efx_siena_mcdi_mon_remove(struct efx_nic *efx) 521{ 522 struct efx_mcdi_mon *hwmon = efx_mcdi_mon(efx); 523 524 if (hwmon->device) 525 hwmon_device_unregister(hwmon->device); 526 kfree(hwmon->attrs); 527 kfree(hwmon->group.attrs); 528 efx_siena_free_buffer(efx, &hwmon->dma_buf); 529} 530 531#endif /* CONFIG_SFC_SIENA_MCDI_MON */