1 /* 2 * Copyright (C) 2001 Dave Engebretsen IBM Corporation 3 * 4 * This program is free software; you can redistribute it and/or modify 5 * it under the terms of the GNU General Public License as published by 6 * the Free Software Foundation; either version 2 of the License, or 7 * (at your option) any later version. 8 * 9 * This program is distributed in the hope that it will be useful, 10 * but WITHOUT ANY WARRANTY; without even the implied warranty of 11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 * GNU General Public License for more details. 13 * 14 * You should have received a copy of the GNU General Public License 15 * along with this program; if not, write to the Free Software 16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA 17 */ 18 19 #include <linux/sched.h> 20 #include <linux/interrupt.h> 21 #include <linux/irq.h> 22 #include <linux/of.h> 23 #include <linux/fs.h> 24 #include <linux/reboot.h> 25 26 #include <asm/machdep.h> 27 #include <asm/rtas.h> 28 #include <asm/firmware.h> 29 30 #include "pseries.h" 31 32 static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX]; 33 static DEFINE_SPINLOCK(ras_log_buf_lock); 34 35 static char global_mce_data_buf[RTAS_ERROR_LOG_MAX]; 36 static DEFINE_PER_CPU(__u64, mce_data_buf); 37 38 static int ras_check_exception_token; 39 40 #define EPOW_SENSOR_TOKEN 9 41 #define EPOW_SENSOR_INDEX 0 42 43 /* EPOW events counter variable */ 44 static int num_epow_events; 45 46 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id); 47 static irqreturn_t ras_error_interrupt(int irq, void *dev_id); 48 49 50 /* 51 * Initialize handlers for the set of interrupts caused by hardware errors 52 * and power system events. 53 */ 54 static int __init init_ras_IRQ(void) 55 { 56 struct device_node *np; 57 58 ras_check_exception_token = rtas_token("check-exception"); 59 60 /* Internal Errors */ 61 np = of_find_node_by_path("/event-sources/internal-errors"); 62 if (np != NULL) { 63 request_event_sources_irqs(np, ras_error_interrupt, 64 "RAS_ERROR"); 65 of_node_put(np); 66 } 67 68 /* EPOW Events */ 69 np = of_find_node_by_path("/event-sources/epow-events"); 70 if (np != NULL) { 71 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW"); 72 of_node_put(np); 73 } 74 75 return 0; 76 } 77 machine_subsys_initcall(pseries, init_ras_IRQ); 78 79 #define EPOW_SHUTDOWN_NORMAL 1 80 #define EPOW_SHUTDOWN_ON_UPS 2 81 #define EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS 3 82 #define EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH 4 83 84 static void handle_system_shutdown(char event_modifier) 85 { 86 switch (event_modifier) { 87 case EPOW_SHUTDOWN_NORMAL: 88 pr_emerg("Power off requested\n"); 89 orderly_poweroff(true); 90 break; 91 92 case EPOW_SHUTDOWN_ON_UPS: 93 pr_emerg("Loss of system power detected. System is running on" 94 " UPS/battery. Check RTAS error log for details\n"); 95 orderly_poweroff(true); 96 break; 97 98 case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS: 99 pr_emerg("Loss of system critical functions detected. Check" 100 " RTAS error log for details\n"); 101 orderly_poweroff(true); 102 break; 103 104 case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH: 105 pr_emerg("High ambient temperature detected. Check RTAS" 106 " error log for details\n"); 107 orderly_poweroff(true); 108 break; 109 110 default: 111 pr_err("Unknown power/cooling shutdown event (modifier = %d)\n", 112 event_modifier); 113 } 114 } 115 116 struct epow_errorlog { 117 unsigned char sensor_value; 118 unsigned char event_modifier; 119 unsigned char extended_modifier; 120 unsigned char reserved; 121 unsigned char platform_reason; 122 }; 123 124 #define EPOW_RESET 0 125 #define EPOW_WARN_COOLING 1 126 #define EPOW_WARN_POWER 2 127 #define EPOW_SYSTEM_SHUTDOWN 3 128 #define EPOW_SYSTEM_HALT 4 129 #define EPOW_MAIN_ENCLOSURE 5 130 #define EPOW_POWER_OFF 7 131 132 static void rtas_parse_epow_errlog(struct rtas_error_log *log) 133 { 134 struct pseries_errorlog *pseries_log; 135 struct epow_errorlog *epow_log; 136 char action_code; 137 char modifier; 138 139 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW); 140 if (pseries_log == NULL) 141 return; 142 143 epow_log = (struct epow_errorlog *)pseries_log->data; 144 action_code = epow_log->sensor_value & 0xF; /* bottom 4 bits */ 145 modifier = epow_log->event_modifier & 0xF; /* bottom 4 bits */ 146 147 switch (action_code) { 148 case EPOW_RESET: 149 if (num_epow_events) { 150 pr_info("Non critical power/cooling issue cleared\n"); 151 num_epow_events--; 152 } 153 break; 154 155 case EPOW_WARN_COOLING: 156 pr_info("Non-critical cooling issue detected. Check RTAS error" 157 " log for details\n"); 158 break; 159 160 case EPOW_WARN_POWER: 161 pr_info("Non-critical power issue detected. Check RTAS error" 162 " log for details\n"); 163 break; 164 165 case EPOW_SYSTEM_SHUTDOWN: 166 handle_system_shutdown(epow_log->event_modifier); 167 break; 168 169 case EPOW_SYSTEM_HALT: 170 pr_emerg("Critical power/cooling issue detected. Check RTAS" 171 " error log for details. Powering off.\n"); 172 orderly_poweroff(true); 173 break; 174 175 case EPOW_MAIN_ENCLOSURE: 176 case EPOW_POWER_OFF: 177 pr_emerg("System about to lose power. Check RTAS error log " 178 " for details. Powering off immediately.\n"); 179 emergency_sync(); 180 kernel_power_off(); 181 break; 182 183 default: 184 pr_err("Unknown power/cooling event (action code = %d)\n", 185 action_code); 186 } 187 188 /* Increment epow events counter variable */ 189 if (action_code != EPOW_RESET) 190 num_epow_events++; 191 } 192 193 /* Handle environmental and power warning (EPOW) interrupts. */ 194 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id) 195 { 196 int status; 197 int state; 198 int critical; 199 200 status = rtas_get_sensor_fast(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX, 201 &state); 202 203 if (state > 3) 204 critical = 1; /* Time Critical */ 205 else 206 critical = 0; 207 208 spin_lock(&ras_log_buf_lock); 209 210 status = rtas_call(ras_check_exception_token, 6, 1, NULL, 211 RTAS_VECTOR_EXTERNAL_INTERRUPT, 212 virq_to_hw(irq), 213 RTAS_EPOW_WARNING, 214 critical, __pa(&ras_log_buf), 215 rtas_get_error_log_max()); 216 217 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0); 218 219 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf); 220 221 spin_unlock(&ras_log_buf_lock); 222 return IRQ_HANDLED; 223 } 224 225 /* 226 * Handle hardware error interrupts. 227 * 228 * RTAS check-exception is called to collect data on the exception. If 229 * the error is deemed recoverable, we log a warning and return. 230 * For nonrecoverable errors, an error is logged and we stop all processing 231 * as quickly as possible in order to prevent propagation of the failure. 232 */ 233 static irqreturn_t ras_error_interrupt(int irq, void *dev_id) 234 { 235 struct rtas_error_log *rtas_elog; 236 int status; 237 int fatal; 238 239 spin_lock(&ras_log_buf_lock); 240 241 status = rtas_call(ras_check_exception_token, 6, 1, NULL, 242 RTAS_VECTOR_EXTERNAL_INTERRUPT, 243 virq_to_hw(irq), 244 RTAS_INTERNAL_ERROR, 1 /* Time Critical */, 245 __pa(&ras_log_buf), 246 rtas_get_error_log_max()); 247 248 rtas_elog = (struct rtas_error_log *)ras_log_buf; 249 250 if (status == 0 && 251 rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC) 252 fatal = 1; 253 else 254 fatal = 0; 255 256 /* format and print the extended information */ 257 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal); 258 259 if (fatal) { 260 pr_emerg("Fatal hardware error detected. Check RTAS error" 261 " log for details. Powering off immediately\n"); 262 emergency_sync(); 263 kernel_power_off(); 264 } else { 265 pr_err("Recoverable hardware error detected\n"); 266 } 267 268 spin_unlock(&ras_log_buf_lock); 269 return IRQ_HANDLED; 270 } 271 272 /* 273 * Some versions of FWNMI place the buffer inside the 4kB page starting at 274 * 0x7000. Other versions place it inside the rtas buffer. We check both. 275 */ 276 #define VALID_FWNMI_BUFFER(A) \ 277 ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \ 278 (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16)))) 279 280 /* 281 * Get the error information for errors coming through the 282 * FWNMI vectors. The pt_regs' r3 will be updated to reflect 283 * the actual r3 if possible, and a ptr to the error log entry 284 * will be returned if found. 285 * 286 * If the RTAS error is not of the extended type, then we put it in a per 287 * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf. 288 * 289 * The global_mce_data_buf does not have any locks or protection around it, 290 * if a second machine check comes in, or a system reset is done 291 * before we have logged the error, then we will get corruption in the 292 * error log. This is preferable over holding off on calling 293 * ibm,nmi-interlock which would result in us checkstopping if a 294 * second machine check did come in. 295 */ 296 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs) 297 { 298 unsigned long *savep; 299 struct rtas_error_log *h, *errhdr = NULL; 300 301 /* Mask top two bits */ 302 regs->gpr[3] &= ~(0x3UL << 62); 303 304 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) { 305 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]); 306 return NULL; 307 } 308 309 savep = __va(regs->gpr[3]); 310 regs->gpr[3] = savep[0]; /* restore original r3 */ 311 312 /* If it isn't an extended log we can use the per cpu 64bit buffer */ 313 h = (struct rtas_error_log *)&savep[1]; 314 if (!rtas_error_extended(h)) { 315 memcpy(this_cpu_ptr(&mce_data_buf), h, sizeof(__u64)); 316 errhdr = (struct rtas_error_log *)this_cpu_ptr(&mce_data_buf); 317 } else { 318 int len, error_log_length; 319 320 error_log_length = 8 + rtas_error_extended_log_length(h); 321 len = max_t(int, error_log_length, RTAS_ERROR_LOG_MAX); 322 memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX); 323 memcpy(global_mce_data_buf, h, len); 324 errhdr = (struct rtas_error_log *)global_mce_data_buf; 325 } 326 327 return errhdr; 328 } 329 330 /* Call this when done with the data returned by FWNMI_get_errinfo. 331 * It will release the saved data area for other CPUs in the 332 * partition to receive FWNMI errors. 333 */ 334 static void fwnmi_release_errinfo(void) 335 { 336 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL); 337 if (ret != 0) 338 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret); 339 } 340 341 int pSeries_system_reset_exception(struct pt_regs *regs) 342 { 343 if (fwnmi_active) { 344 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs); 345 if (errhdr) { 346 /* XXX Should look at FWNMI information */ 347 } 348 fwnmi_release_errinfo(); 349 } 350 return 0; /* need to perform reset */ 351 } 352 353 /* 354 * See if we can recover from a machine check exception. 355 * This is only called on power4 (or above) and only via 356 * the Firmware Non-Maskable Interrupts (fwnmi) handler 357 * which provides the error analysis for us. 358 * 359 * Return 1 if corrected (or delivered a signal). 360 * Return 0 if there is nothing we can do. 361 */ 362 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err) 363 { 364 int recovered = 0; 365 int disposition = rtas_error_disposition(err); 366 367 if (!(regs->msr & MSR_RI)) { 368 /* If MSR_RI isn't set, we cannot recover */ 369 recovered = 0; 370 371 } else if (disposition == RTAS_DISP_FULLY_RECOVERED) { 372 /* Platform corrected itself */ 373 recovered = 1; 374 375 } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) { 376 /* Platform corrected itself but could be degraded */ 377 printk(KERN_ERR "MCE: limited recovery, system may " 378 "be degraded\n"); 379 recovered = 1; 380 381 } else if (user_mode(regs) && !is_global_init(current) && 382 rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) { 383 384 /* 385 * If we received a synchronous error when in userspace 386 * kill the task. Firmware may report details of the fail 387 * asynchronously, so we can't rely on the target and type 388 * fields being valid here. 389 */ 390 printk(KERN_ERR "MCE: uncorrectable error, killing task " 391 "%s:%d\n", current->comm, current->pid); 392 393 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip); 394 recovered = 1; 395 } 396 397 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0); 398 399 return recovered; 400 } 401 402 /* 403 * Handle a machine check. 404 * 405 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi) 406 * should be present. If so the handler which called us tells us if the 407 * error was recovered (never true if RI=0). 408 * 409 * On hardware prior to Power 4 these exceptions were asynchronous which 410 * means we can't tell exactly where it occurred and so we can't recover. 411 */ 412 int pSeries_machine_check_exception(struct pt_regs *regs) 413 { 414 struct rtas_error_log *errp; 415 416 if (fwnmi_active) { 417 errp = fwnmi_get_errinfo(regs); 418 fwnmi_release_errinfo(); 419 if (errp && recover_mce(regs, errp)) 420 return 1; 421 } 422 423 return 0; 424 } 425