arch/powerpc/platforms/pseries/ras.c

   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_hotplug_interrupt(int irq, void *dev_id);
  47 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id);
  48 static irqreturn_t ras_error_interrupt(int irq, void *dev_id);
  49
  50
  51 /*
  52  * Initialize handlers for the set of interrupts caused by hardware errors
  53  * and power system events.
  54  */
  55 static int __init init_ras_IRQ(void)
  56 {
  57         struct device_node *np;
  58
  59         ras_check_exception_token = rtas_token("check-exception");
  60
  61         /* Internal Errors */
  62         np = of_find_node_by_path("/event-sources/internal-errors");
  63         if (np != NULL) {
  64                 request_event_sources_irqs(np, ras_error_interrupt,
  65                                            "RAS_ERROR");
  66                 of_node_put(np);
  67         }
  68
  69         /* Hotplug Events */
  70         np = of_find_node_by_path("/event-sources/hot-plug-events");
  71         if (np != NULL) {
  72                 request_event_sources_irqs(np, ras_hotplug_interrupt,
  73                                            "RAS_HOTPLUG");
  74                 of_node_put(np);
  75         }
  76
  77         /* EPOW Events */
  78         np = of_find_node_by_path("/event-sources/epow-events");
  79         if (np != NULL) {
  80                 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW");
  81                 of_node_put(np);
  82         }
  83
  84         return 0;
  85 }
  86 machine_subsys_initcall(pseries, init_ras_IRQ);
  87
  88 #define EPOW_SHUTDOWN_NORMAL                            1
  89 #define EPOW_SHUTDOWN_ON_UPS                            2
  90 #define EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS        3
  91 #define EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH      4
  92
  93 static void handle_system_shutdown(char event_modifier)
  94 {
  95         switch (event_modifier) {
  96         case EPOW_SHUTDOWN_NORMAL:
  97                 pr_emerg("Power off requested\n");
  98                 orderly_poweroff(true);
  99                 break;
 100
 101         case EPOW_SHUTDOWN_ON_UPS:
 102                 pr_emerg("Loss of system power detected. System is running on"
 103                          " UPS/battery. Check RTAS error log for details\n");
 104                 orderly_poweroff(true);
 105                 break;
 106
 107         case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS:
 108                 pr_emerg("Loss of system critical functions detected. Check"
 109                          " RTAS error log for details\n");
 110                 orderly_poweroff(true);
 111                 break;
 112
 113         case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH:
 114                 pr_emerg("High ambient temperature detected. Check RTAS"
 115                          " error log for details\n");
 116                 orderly_poweroff(true);
 117                 break;
 118
 119         default:
 120                 pr_err("Unknown power/cooling shutdown event (modifier = %d)\n",
 121                         event_modifier);
 122         }
 123 }
 124
 125 struct epow_errorlog {
 126         unsigned char sensor_value;
 127         unsigned char event_modifier;
 128         unsigned char extended_modifier;
 129         unsigned char reserved;
 130         unsigned char platform_reason;
 131 };
 132
 133 #define EPOW_RESET                      0
 134 #define EPOW_WARN_COOLING               1
 135 #define EPOW_WARN_POWER                 2
 136 #define EPOW_SYSTEM_SHUTDOWN            3
 137 #define EPOW_SYSTEM_HALT                4
 138 #define EPOW_MAIN_ENCLOSURE             5
 139 #define EPOW_POWER_OFF                  7
 140
 141 static void rtas_parse_epow_errlog(struct rtas_error_log *log)
 142 {
 143         struct pseries_errorlog *pseries_log;
 144         struct epow_errorlog *epow_log;
 145         char action_code;
 146         char modifier;
 147
 148         pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
 149         if (pseries_log == NULL)
 150                 return;
 151
 152         epow_log = (struct epow_errorlog *)pseries_log->data;
 153         action_code = epow_log->sensor_value & 0xF;     /* bottom 4 bits */
 154         modifier = epow_log->event_modifier & 0xF;      /* bottom 4 bits */
 155
 156         switch (action_code) {
 157         case EPOW_RESET:
 158                 if (num_epow_events) {
 159                         pr_info("Non critical power/cooling issue cleared\n");
 160                         num_epow_events--;
 161                 }
 162                 break;
 163
 164         case EPOW_WARN_COOLING:
 165                 pr_info("Non-critical cooling issue detected. Check RTAS error"
 166                         " log for details\n");
 167                 break;
 168
 169         case EPOW_WARN_POWER:
 170                 pr_info("Non-critical power issue detected. Check RTAS error"
 171                         " log for details\n");
 172                 break;
 173
 174         case EPOW_SYSTEM_SHUTDOWN:
 175                 handle_system_shutdown(epow_log->event_modifier);
 176                 break;
 177
 178         case EPOW_SYSTEM_HALT:
 179                 pr_emerg("Critical power/cooling issue detected. Check RTAS"
 180                          " error log for details. Powering off.\n");
 181                 orderly_poweroff(true);
 182                 break;
 183
 184         case EPOW_MAIN_ENCLOSURE:
 185         case EPOW_POWER_OFF:
 186                 pr_emerg("System about to lose power. Check RTAS error log "
 187                          " for details. Powering off immediately.\n");
 188                 emergency_sync();
 189                 kernel_power_off();
 190                 break;
 191
 192         default:
 193                 pr_err("Unknown power/cooling event (action code  = %d)\n",
 194                         action_code);
 195         }
 196
 197         /* Increment epow events counter variable */
 198         if (action_code != EPOW_RESET)
 199                 num_epow_events++;
 200 }
 201
 202 static irqreturn_t ras_hotplug_interrupt(int irq, void *dev_id)
 203 {
 204         struct pseries_errorlog *pseries_log;
 205         struct pseries_hp_errorlog *hp_elog;
 206
 207         spin_lock(&ras_log_buf_lock);
 208
 209         rtas_call(ras_check_exception_token, 6, 1, NULL,
 210                   RTAS_VECTOR_EXTERNAL_INTERRUPT, virq_to_hw(irq),
 211                   RTAS_HOTPLUG_EVENTS, 0, __pa(&ras_log_buf),
 212                   rtas_get_error_log_max());
 213
 214         pseries_log = get_pseries_errorlog((struct rtas_error_log *)ras_log_buf,
 215                                            PSERIES_ELOG_SECT_ID_HOTPLUG);
 216         hp_elog = (struct pseries_hp_errorlog *)pseries_log->data;
 217
 218         /*
 219          * Since PCI hotplug is not currently supported on pseries, put PCI
 220          * hotplug events on the ras_log_buf to be handled by rtas_errd.
 221          */
 222         if (hp_elog->resource == PSERIES_HP_ELOG_RESOURCE_MEM ||
 223             hp_elog->resource == PSERIES_HP_ELOG_RESOURCE_CPU)
 224                 queue_hotplug_event(hp_elog, NULL, NULL);
 225         else
 226                 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
 227
 228         spin_unlock(&ras_log_buf_lock);
 229         return IRQ_HANDLED;
 230 }
 231
 232 /* Handle environmental and power warning (EPOW) interrupts. */
 233 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
 234 {
 235         int status;
 236         int state;
 237         int critical;
 238
 239         status = rtas_get_sensor_fast(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX,
 240                                       &state);
 241
 242         if (state > 3)
 243                 critical = 1;           /* Time Critical */
 244         else
 245                 critical = 0;
 246
 247         spin_lock(&ras_log_buf_lock);
 248
 249         status = rtas_call(ras_check_exception_token, 6, 1, NULL,
 250                            RTAS_VECTOR_EXTERNAL_INTERRUPT,
 251                            virq_to_hw(irq),
 252                            RTAS_EPOW_WARNING,
 253                            critical, __pa(&ras_log_buf),
 254                                 rtas_get_error_log_max());
 255
 256         log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
 257
 258         rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
 259
 260         spin_unlock(&ras_log_buf_lock);
 261         return IRQ_HANDLED;
 262 }
 263
 264 /*
 265  * Handle hardware error interrupts.
 266  *
 267  * RTAS check-exception is called to collect data on the exception.  If
 268  * the error is deemed recoverable, we log a warning and return.
 269  * For nonrecoverable errors, an error is logged and we stop all processing
 270  * as quickly as possible in order to prevent propagation of the failure.
 271  */
 272 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
 273 {
 274         struct rtas_error_log *rtas_elog;
 275         int status;
 276         int fatal;
 277
 278         spin_lock(&ras_log_buf_lock);
 279
 280         status = rtas_call(ras_check_exception_token, 6, 1, NULL,
 281                            RTAS_VECTOR_EXTERNAL_INTERRUPT,
 282                            virq_to_hw(irq),
 283                            RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
 284                            __pa(&ras_log_buf),
 285                                 rtas_get_error_log_max());
 286
 287         rtas_elog = (struct rtas_error_log *)ras_log_buf;
 288
 289         if (status == 0 &&
 290             rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC)
 291                 fatal = 1;
 292         else
 293                 fatal = 0;
 294
 295         /* format and print the extended information */
 296         log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
 297
 298         if (fatal) {
 299                 pr_emerg("Fatal hardware error detected. Check RTAS error"
 300                          " log for details. Powering off immediately\n");
 301                 emergency_sync();
 302                 kernel_power_off();
 303         } else {
 304                 pr_err("Recoverable hardware error detected\n");
 305         }
 306
 307         spin_unlock(&ras_log_buf_lock);
 308         return IRQ_HANDLED;
 309 }
 310
 311 /*
 312  * Some versions of FWNMI place the buffer inside the 4kB page starting at
 313  * 0x7000. Other versions place it inside the rtas buffer. We check both.
 314  */
 315 #define VALID_FWNMI_BUFFER(A) \
 316         ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \
 317         (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16))))
 318
 319 /*
 320  * Get the error information for errors coming through the
 321  * FWNMI vectors.  The pt_regs' r3 will be updated to reflect
 322  * the actual r3 if possible, and a ptr to the error log entry
 323  * will be returned if found.
 324  *
 325  * If the RTAS error is not of the extended type, then we put it in a per
 326  * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf.
 327  *
 328  * The global_mce_data_buf does not have any locks or protection around it,
 329  * if a second machine check comes in, or a system reset is done
 330  * before we have logged the error, then we will get corruption in the
 331  * error log.  This is preferable over holding off on calling
 332  * ibm,nmi-interlock which would result in us checkstopping if a
 333  * second machine check did come in.
 334  */
 335 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
 336 {
 337         unsigned long *savep;
 338         struct rtas_error_log *h, *errhdr = NULL;
 339
 340         /* Mask top two bits */
 341         regs->gpr[3] &= ~(0x3UL << 62);
 342
 343         if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
 344                 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
 345                 return NULL;
 346         }
 347
 348         savep = __va(regs->gpr[3]);
 349         regs->gpr[3] = savep[0];        /* restore original r3 */
 350
 351         /* If it isn't an extended log we can use the per cpu 64bit buffer */
 352         h = (struct rtas_error_log *)&savep[1];
 353         if (!rtas_error_extended(h)) {
 354                 memcpy(this_cpu_ptr(&mce_data_buf), h, sizeof(__u64));
 355                 errhdr = (struct rtas_error_log *)this_cpu_ptr(&mce_data_buf);
 356         } else {
 357                 int len, error_log_length;
 358
 359                 error_log_length = 8 + rtas_error_extended_log_length(h);
 360                 len = max_t(int, error_log_length, RTAS_ERROR_LOG_MAX);
 361                 memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
 362                 memcpy(global_mce_data_buf, h, len);
 363                 errhdr = (struct rtas_error_log *)global_mce_data_buf;
 364         }
 365
 366         return errhdr;
 367 }
 368
 369 /* Call this when done with the data returned by FWNMI_get_errinfo.
 370  * It will release the saved data area for other CPUs in the
 371  * partition to receive FWNMI errors.
 372  */
 373 static void fwnmi_release_errinfo(void)
 374 {
 375         int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
 376         if (ret != 0)
 377                 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
 378 }
 379
 380 int pSeries_system_reset_exception(struct pt_regs *regs)
 381 {
 382 #ifdef __LITTLE_ENDIAN__
 383         /*
 384          * Some firmware byteswaps SRR registers and gives incorrect SRR1. Try
 385          * to detect the bad SRR1 pattern here. Flip the NIP back to correct
 386          * endian for reporting purposes. Unfortunately the MSR can't be fixed,
 387          * so clear it. It will be missing MSR_RI so we won't try to recover.
 388          */
 389         if ((be64_to_cpu(regs->msr) &
 390                         (MSR_LE|MSR_RI|MSR_DR|MSR_IR|MSR_ME|MSR_PR|
 391                          MSR_ILE|MSR_HV|MSR_SF)) == (MSR_DR|MSR_SF)) {
 392                 regs->nip = be64_to_cpu((__be64)regs->nip);
 393                 regs->msr = 0;
 394         }
 395 #endif
 396
 397         if (fwnmi_active) {
 398                 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
 399                 if (errhdr) {
 400                         /* XXX Should look at FWNMI information */
 401                 }
 402                 fwnmi_release_errinfo();
 403         }
 404
 405         if (smp_handle_nmi_ipi(regs))
 406                 return 1;
 407
 408         return 0; /* need to perform reset */
 409 }
 410
 411 /*
 412  * See if we can recover from a machine check exception.
 413  * This is only called on power4 (or above) and only via
 414  * the Firmware Non-Maskable Interrupts (fwnmi) handler
 415  * which provides the error analysis for us.
 416  *
 417  * Return 1 if corrected (or delivered a signal).
 418  * Return 0 if there is nothing we can do.
 419  */
 420 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
 421 {
 422         int recovered = 0;
 423         int disposition = rtas_error_disposition(err);
 424
 425         if (!(regs->msr & MSR_RI)) {
 426                 /* If MSR_RI isn't set, we cannot recover */
 427                 recovered = 0;
 428
 429         } else if (disposition == RTAS_DISP_FULLY_RECOVERED) {
 430                 /* Platform corrected itself */
 431                 recovered = 1;
 432
 433         } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) {
 434                 /* Platform corrected itself but could be degraded */
 435                 printk(KERN_ERR "MCE: limited recovery, system may "
 436                        "be degraded\n");
 437                 recovered = 1;
 438
 439         } else if (user_mode(regs) && !is_global_init(current) &&
 440                    rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) {
 441
 442                 /*
 443                  * If we received a synchronous error when in userspace
 444                  * kill the task. Firmware may report details of the fail
 445                  * asynchronously, so we can't rely on the target and type
 446                  * fields being valid here.
 447                  */
 448                 printk(KERN_ERR "MCE: uncorrectable error, killing task "
 449                        "%s:%d\n", current->comm, current->pid);
 450
 451                 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
 452                 recovered = 1;
 453         }
 454
 455         log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
 456
 457         return recovered;
 458 }
 459
 460 /*
 461  * Handle a machine check.
 462  *
 463  * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
 464  * should be present.  If so the handler which called us tells us if the
 465  * error was recovered (never true if RI=0).
 466  *
 467  * On hardware prior to Power 4 these exceptions were asynchronous which
 468  * means we can't tell exactly where it occurred and so we can't recover.
 469  */
 470 int pSeries_machine_check_exception(struct pt_regs *regs)
 471 {
 472         struct rtas_error_log *errp;
 473
 474         if (fwnmi_active) {
 475                 errp = fwnmi_get_errinfo(regs);
 476                 fwnmi_release_errinfo();
 477                 if (errp && recover_mce(regs, errp))
 478                         return 1;
 479         }
 480
 481         return 0;
 482 }