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