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 /* Change Activity: 20 * 2001/09/21 : engebret : Created with minimal EPOW and HW exception support. 21 * End Change Activity 22 */ 23 24 #include <linux/errno.h> 25 #include <linux/threads.h> 26 #include <linux/kernel_stat.h> 27 #include <linux/signal.h> 28 #include <linux/sched.h> 29 #include <linux/ioport.h> 30 #include <linux/interrupt.h> 31 #include <linux/timex.h> 32 #include <linux/init.h> 33 #include <linux/delay.h> 34 #include <linux/irq.h> 35 #include <linux/random.h> 36 #include <linux/sysrq.h> 37 #include <linux/bitops.h> 38 #include <linux/fs.h> 39 #include <linux/reboot.h> 40 41 #include <asm/uaccess.h> 42 #include <asm/system.h> 43 #include <asm/io.h> 44 #include <asm/pgtable.h> 45 #include <asm/irq.h> 46 #include <asm/cache.h> 47 #include <asm/prom.h> 48 #include <asm/ptrace.h> 49 #include <asm/machdep.h> 50 #include <asm/rtas.h> 51 #include <asm/udbg.h> 52 #include <asm/firmware.h> 53 54 #include "pseries.h" 55 56 static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX]; 57 static DEFINE_SPINLOCK(ras_log_buf_lock); 58 59 static char global_mce_data_buf[RTAS_ERROR_LOG_MAX]; 60 static DEFINE_PER_CPU(__u64, mce_data_buf); 61 62 static int ras_check_exception_token; 63 64 #define EPOW_SENSOR_TOKEN 9 65 #define EPOW_SENSOR_INDEX 0 66 67 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id); 68 static irqreturn_t ras_error_interrupt(int irq, void *dev_id); 69 70 71 /* 72 * Initialize handlers for the set of interrupts caused by hardware errors 73 * and power system events. 74 */ 75 static int __init init_ras_IRQ(void) 76 { 77 struct device_node *np; 78 79 ras_check_exception_token = rtas_token("check-exception"); 80 81 /* Internal Errors */ 82 np = of_find_node_by_path("/event-sources/internal-errors"); 83 if (np != NULL) { 84 request_event_sources_irqs(np, ras_error_interrupt, 85 "RAS_ERROR"); 86 of_node_put(np); 87 } 88 89 /* EPOW Events */ 90 np = of_find_node_by_path("/event-sources/epow-events"); 91 if (np != NULL) { 92 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW"); 93 of_node_put(np); 94 } 95 96 return 0; 97 } 98 subsys_initcall(init_ras_IRQ); 99 100 #define EPOW_SHUTDOWN_NORMAL 1 101 #define EPOW_SHUTDOWN_ON_UPS 2 102 #define EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS 3 103 #define EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH 4 104 105 static void handle_system_shutdown(char event_modifier) 106 { 107 switch (event_modifier) { 108 case EPOW_SHUTDOWN_NORMAL: 109 pr_emerg("Firmware initiated power off"); 110 orderly_poweroff(1); 111 break; 112 113 case EPOW_SHUTDOWN_ON_UPS: 114 pr_emerg("Loss of power reported by firmware, system is " 115 "running on UPS/battery"); 116 break; 117 118 case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS: 119 pr_emerg("Loss of system critical functions reported by " 120 "firmware"); 121 pr_emerg("Check RTAS error log for details"); 122 orderly_poweroff(1); 123 break; 124 125 case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH: 126 pr_emerg("Ambient temperature too high reported by firmware"); 127 pr_emerg("Check RTAS error log for details"); 128 orderly_poweroff(1); 129 break; 130 131 default: 132 pr_err("Unknown power/cooling shutdown event (modifier %d)", 133 event_modifier); 134 } 135 } 136 137 struct epow_errorlog { 138 unsigned char sensor_value; 139 unsigned char event_modifier; 140 unsigned char extended_modifier; 141 unsigned char reserved; 142 unsigned char platform_reason; 143 }; 144 145 #define EPOW_RESET 0 146 #define EPOW_WARN_COOLING 1 147 #define EPOW_WARN_POWER 2 148 #define EPOW_SYSTEM_SHUTDOWN 3 149 #define EPOW_SYSTEM_HALT 4 150 #define EPOW_MAIN_ENCLOSURE 5 151 #define EPOW_POWER_OFF 7 152 153 void rtas_parse_epow_errlog(struct rtas_error_log *log) 154 { 155 struct pseries_errorlog *pseries_log; 156 struct epow_errorlog *epow_log; 157 char action_code; 158 char modifier; 159 160 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW); 161 if (pseries_log == NULL) 162 return; 163 164 epow_log = (struct epow_errorlog *)pseries_log->data; 165 action_code = epow_log->sensor_value & 0xF; /* bottom 4 bits */ 166 modifier = epow_log->event_modifier & 0xF; /* bottom 4 bits */ 167 168 switch (action_code) { 169 case EPOW_RESET: 170 pr_err("Non critical power or cooling issue cleared"); 171 break; 172 173 case EPOW_WARN_COOLING: 174 pr_err("Non critical cooling issue reported by firmware"); 175 pr_err("Check RTAS error log for details"); 176 break; 177 178 case EPOW_WARN_POWER: 179 pr_err("Non critical power issue reported by firmware"); 180 pr_err("Check RTAS error log for details"); 181 break; 182 183 case EPOW_SYSTEM_SHUTDOWN: 184 handle_system_shutdown(epow_log->event_modifier); 185 break; 186 187 case EPOW_SYSTEM_HALT: 188 pr_emerg("Firmware initiated power off"); 189 orderly_poweroff(1); 190 break; 191 192 case EPOW_MAIN_ENCLOSURE: 193 case EPOW_POWER_OFF: 194 pr_emerg("Critical power/cooling issue reported by firmware"); 195 pr_emerg("Check RTAS error log for details"); 196 pr_emerg("Immediate power off"); 197 emergency_sync(); 198 kernel_power_off(); 199 break; 200 201 default: 202 pr_err("Unknown power/cooling event (action code %d)", 203 action_code); 204 } 205 } 206 207 /* Handle environmental and power warning (EPOW) interrupts. */ 208 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id) 209 { 210 int status; 211 int state; 212 int critical; 213 214 status = rtas_get_sensor(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX, &state); 215 216 if (state > 3) 217 critical = 1; /* Time Critical */ 218 else 219 critical = 0; 220 221 spin_lock(&ras_log_buf_lock); 222 223 status = rtas_call(ras_check_exception_token, 6, 1, NULL, 224 RTAS_VECTOR_EXTERNAL_INTERRUPT, 225 virq_to_hw(irq), 226 RTAS_EPOW_WARNING | RTAS_POWERMGM_EVENTS, 227 critical, __pa(&ras_log_buf), 228 rtas_get_error_log_max()); 229 230 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0); 231 232 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf); 233 234 spin_unlock(&ras_log_buf_lock); 235 return IRQ_HANDLED; 236 } 237 238 /* 239 * Handle hardware error interrupts. 240 * 241 * RTAS check-exception is called to collect data on the exception. If 242 * the error is deemed recoverable, we log a warning and return. 243 * For nonrecoverable errors, an error is logged and we stop all processing 244 * as quickly as possible in order to prevent propagation of the failure. 245 */ 246 static irqreturn_t ras_error_interrupt(int irq, void *dev_id) 247 { 248 struct rtas_error_log *rtas_elog; 249 int status = 0xdeadbeef; 250 int fatal; 251 252 spin_lock(&ras_log_buf_lock); 253 254 status = rtas_call(ras_check_exception_token, 6, 1, NULL, 255 RTAS_VECTOR_EXTERNAL_INTERRUPT, 256 virq_to_hw(irq), 257 RTAS_INTERNAL_ERROR, 1 /*Time Critical */, 258 __pa(&ras_log_buf), 259 rtas_get_error_log_max()); 260 261 rtas_elog = (struct rtas_error_log *)ras_log_buf; 262 263 if ((status == 0) && (rtas_elog->severity >= RTAS_SEVERITY_ERROR_SYNC)) 264 fatal = 1; 265 else 266 fatal = 0; 267 268 /* format and print the extended information */ 269 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal); 270 271 if (fatal) { 272 udbg_printf("Fatal HW Error <0x%lx 0x%x>\n", 273 *((unsigned long *)&ras_log_buf), status); 274 printk(KERN_EMERG "Error: Fatal hardware error <0x%lx 0x%x>\n", 275 *((unsigned long *)&ras_log_buf), status); 276 277 #ifndef DEBUG_RTAS_POWER_OFF 278 /* Don't actually power off when debugging so we can test 279 * without actually failing while injecting errors. 280 * Error data will not be logged to syslog. 281 */ 282 ppc_md.power_off(); 283 #endif 284 } else { 285 udbg_printf("Recoverable HW Error <0x%lx 0x%x>\n", 286 *((unsigned long *)&ras_log_buf), status); 287 printk(KERN_WARNING 288 "Warning: Recoverable hardware error <0x%lx 0x%x>\n", 289 *((unsigned long *)&ras_log_buf), status); 290 } 291 292 spin_unlock(&ras_log_buf_lock); 293 return IRQ_HANDLED; 294 } 295 296 /* 297 * Some versions of FWNMI place the buffer inside the 4kB page starting at 298 * 0x7000. Other versions place it inside the rtas buffer. We check both. 299 */ 300 #define VALID_FWNMI_BUFFER(A) \ 301 ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \ 302 (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16)))) 303 304 /* 305 * Get the error information for errors coming through the 306 * FWNMI vectors. The pt_regs' r3 will be updated to reflect 307 * the actual r3 if possible, and a ptr to the error log entry 308 * will be returned if found. 309 * 310 * If the RTAS error is not of the extended type, then we put it in a per 311 * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf. 312 * 313 * The global_mce_data_buf does not have any locks or protection around it, 314 * if a second machine check comes in, or a system reset is done 315 * before we have logged the error, then we will get corruption in the 316 * error log. This is preferable over holding off on calling 317 * ibm,nmi-interlock which would result in us checkstopping if a 318 * second machine check did come in. 319 */ 320 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs) 321 { 322 unsigned long *savep; 323 struct rtas_error_log *h, *errhdr = NULL; 324 325 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) { 326 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]); 327 return NULL; 328 } 329 330 savep = __va(regs->gpr[3]); 331 regs->gpr[3] = savep[0]; /* restore original r3 */ 332 333 /* If it isn't an extended log we can use the per cpu 64bit buffer */ 334 h = (struct rtas_error_log *)&savep[1]; 335 if (!h->extended) { 336 memcpy(&__get_cpu_var(mce_data_buf), h, sizeof(__u64)); 337 errhdr = (struct rtas_error_log *)&__get_cpu_var(mce_data_buf); 338 } else { 339 int len; 340 341 len = max_t(int, 8+h->extended_log_length, RTAS_ERROR_LOG_MAX); 342 memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX); 343 memcpy(global_mce_data_buf, h, len); 344 errhdr = (struct rtas_error_log *)global_mce_data_buf; 345 } 346 347 return errhdr; 348 } 349 350 /* Call this when done with the data returned by FWNMI_get_errinfo. 351 * It will release the saved data area for other CPUs in the 352 * partition to receive FWNMI errors. 353 */ 354 static void fwnmi_release_errinfo(void) 355 { 356 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL); 357 if (ret != 0) 358 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret); 359 } 360 361 int pSeries_system_reset_exception(struct pt_regs *regs) 362 { 363 if (fwnmi_active) { 364 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs); 365 if (errhdr) { 366 /* XXX Should look at FWNMI information */ 367 } 368 fwnmi_release_errinfo(); 369 } 370 return 0; /* need to perform reset */ 371 } 372 373 /* 374 * See if we can recover from a machine check exception. 375 * This is only called on power4 (or above) and only via 376 * the Firmware Non-Maskable Interrupts (fwnmi) handler 377 * which provides the error analysis for us. 378 * 379 * Return 1 if corrected (or delivered a signal). 380 * Return 0 if there is nothing we can do. 381 */ 382 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err) 383 { 384 int recovered = 0; 385 386 if (!(regs->msr & MSR_RI)) { 387 /* If MSR_RI isn't set, we cannot recover */ 388 recovered = 0; 389 390 } else if (err->disposition == RTAS_DISP_FULLY_RECOVERED) { 391 /* Platform corrected itself */ 392 recovered = 1; 393 394 } else if (err->disposition == RTAS_DISP_LIMITED_RECOVERY) { 395 /* Platform corrected itself but could be degraded */ 396 printk(KERN_ERR "MCE: limited recovery, system may " 397 "be degraded\n"); 398 recovered = 1; 399 400 } else if (user_mode(regs) && !is_global_init(current) && 401 err->severity == RTAS_SEVERITY_ERROR_SYNC) { 402 403 /* 404 * If we received a synchronous error when in userspace 405 * kill the task. Firmware may report details of the fail 406 * asynchronously, so we can't rely on the target and type 407 * fields being valid here. 408 */ 409 printk(KERN_ERR "MCE: uncorrectable error, killing task " 410 "%s:%d\n", current->comm, current->pid); 411 412 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip); 413 recovered = 1; 414 } 415 416 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0); 417 418 return recovered; 419 } 420 421 /* 422 * Handle a machine check. 423 * 424 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi) 425 * should be present. If so the handler which called us tells us if the 426 * error was recovered (never true if RI=0). 427 * 428 * On hardware prior to Power 4 these exceptions were asynchronous which 429 * means we can't tell exactly where it occurred and so we can't recover. 430 */ 431 int pSeries_machine_check_exception(struct pt_regs *regs) 432 { 433 struct rtas_error_log *errp; 434 435 if (fwnmi_active) { 436 errp = fwnmi_get_errinfo(regs); 437 fwnmi_release_errinfo(); 438 if (errp && recover_mce(regs, errp)) 439 return 1; 440 } 441 442 return 0; 443 } 444