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 * Enable the hotplug interrupt late because processing them may touch other
53 * devices or systems (e.g. hugepages) that have not been initialized at the
54 * subsys stage.
55 */
init_ras_hotplug_IRQ(void)56 int __init init_ras_hotplug_IRQ(void)
57 {
58 struct device_node *np;
59
60 /* Hotplug Events */
61 np = of_find_node_by_path("/event-sources/hot-plug-events");
62 if (np != NULL) {
63 if (dlpar_workqueue_init() == 0)
64 request_event_sources_irqs(np, ras_hotplug_interrupt,
65 "RAS_HOTPLUG");
66 of_node_put(np);
67 }
68
69 return 0;
70 }
71 machine_late_initcall(pseries, init_ras_hotplug_IRQ);
72
73 /*
74 * Initialize handlers for the set of interrupts caused by hardware errors
75 * and power system events.
76 */
init_ras_IRQ(void)77 static int __init init_ras_IRQ(void)
78 {
79 struct device_node *np;
80
81 ras_check_exception_token = rtas_token("check-exception");
82
83 /* Internal Errors */
84 np = of_find_node_by_path("/event-sources/internal-errors");
85 if (np != NULL) {
86 request_event_sources_irqs(np, ras_error_interrupt,
87 "RAS_ERROR");
88 of_node_put(np);
89 }
90
91 /* EPOW Events */
92 np = of_find_node_by_path("/event-sources/epow-events");
93 if (np != NULL) {
94 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW");
95 of_node_put(np);
96 }
97
98 return 0;
99 }
100 machine_subsys_initcall(pseries, init_ras_IRQ);
101
102 #define EPOW_SHUTDOWN_NORMAL 1
103 #define EPOW_SHUTDOWN_ON_UPS 2
104 #define EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS 3
105 #define EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH 4
106
handle_system_shutdown(char event_modifier)107 static void handle_system_shutdown(char event_modifier)
108 {
109 switch (event_modifier) {
110 case EPOW_SHUTDOWN_NORMAL:
111 pr_emerg("Power off requested\n");
112 orderly_poweroff(true);
113 break;
114
115 case EPOW_SHUTDOWN_ON_UPS:
116 pr_emerg("Loss of system power detected. System is running on"
117 " UPS/battery. Check RTAS error log for details\n");
118 orderly_poweroff(true);
119 break;
120
121 case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS:
122 pr_emerg("Loss of system critical functions detected. Check"
123 " RTAS error log for details\n");
124 orderly_poweroff(true);
125 break;
126
127 case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH:
128 pr_emerg("High ambient temperature detected. Check RTAS"
129 " error log for details\n");
130 orderly_poweroff(true);
131 break;
132
133 default:
134 pr_err("Unknown power/cooling shutdown event (modifier = %d)\n",
135 event_modifier);
136 }
137 }
138
139 struct epow_errorlog {
140 unsigned char sensor_value;
141 unsigned char event_modifier;
142 unsigned char extended_modifier;
143 unsigned char reserved;
144 unsigned char platform_reason;
145 };
146
147 #define EPOW_RESET 0
148 #define EPOW_WARN_COOLING 1
149 #define EPOW_WARN_POWER 2
150 #define EPOW_SYSTEM_SHUTDOWN 3
151 #define EPOW_SYSTEM_HALT 4
152 #define EPOW_MAIN_ENCLOSURE 5
153 #define EPOW_POWER_OFF 7
154
rtas_parse_epow_errlog(struct rtas_error_log * log)155 static void rtas_parse_epow_errlog(struct rtas_error_log *log)
156 {
157 struct pseries_errorlog *pseries_log;
158 struct epow_errorlog *epow_log;
159 char action_code;
160 char modifier;
161
162 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
163 if (pseries_log == NULL)
164 return;
165
166 epow_log = (struct epow_errorlog *)pseries_log->data;
167 action_code = epow_log->sensor_value & 0xF; /* bottom 4 bits */
168 modifier = epow_log->event_modifier & 0xF; /* bottom 4 bits */
169
170 switch (action_code) {
171 case EPOW_RESET:
172 if (num_epow_events) {
173 pr_info("Non critical power/cooling issue cleared\n");
174 num_epow_events--;
175 }
176 break;
177
178 case EPOW_WARN_COOLING:
179 pr_info("Non-critical cooling issue detected. Check RTAS error"
180 " log for details\n");
181 break;
182
183 case EPOW_WARN_POWER:
184 pr_info("Non-critical power issue detected. Check RTAS error"
185 " log for details\n");
186 break;
187
188 case EPOW_SYSTEM_SHUTDOWN:
189 handle_system_shutdown(epow_log->event_modifier);
190 break;
191
192 case EPOW_SYSTEM_HALT:
193 pr_emerg("Critical power/cooling issue detected. Check RTAS"
194 " error log for details. Powering off.\n");
195 orderly_poweroff(true);
196 break;
197
198 case EPOW_MAIN_ENCLOSURE:
199 case EPOW_POWER_OFF:
200 pr_emerg("System about to lose power. Check RTAS error log "
201 " for details. Powering off immediately.\n");
202 emergency_sync();
203 kernel_power_off();
204 break;
205
206 default:
207 pr_err("Unknown power/cooling event (action code = %d)\n",
208 action_code);
209 }
210
211 /* Increment epow events counter variable */
212 if (action_code != EPOW_RESET)
213 num_epow_events++;
214 }
215
ras_hotplug_interrupt(int irq,void * dev_id)216 static irqreturn_t ras_hotplug_interrupt(int irq, void *dev_id)
217 {
218 struct pseries_errorlog *pseries_log;
219 struct pseries_hp_errorlog *hp_elog;
220
221 spin_lock(&ras_log_buf_lock);
222
223 rtas_call(ras_check_exception_token, 6, 1, NULL,
224 RTAS_VECTOR_EXTERNAL_INTERRUPT, virq_to_hw(irq),
225 RTAS_HOTPLUG_EVENTS, 0, __pa(&ras_log_buf),
226 rtas_get_error_log_max());
227
228 pseries_log = get_pseries_errorlog((struct rtas_error_log *)ras_log_buf,
229 PSERIES_ELOG_SECT_ID_HOTPLUG);
230 hp_elog = (struct pseries_hp_errorlog *)pseries_log->data;
231
232 /*
233 * Since PCI hotplug is not currently supported on pseries, put PCI
234 * hotplug events on the ras_log_buf to be handled by rtas_errd.
235 */
236 if (hp_elog->resource == PSERIES_HP_ELOG_RESOURCE_MEM ||
237 hp_elog->resource == PSERIES_HP_ELOG_RESOURCE_CPU)
238 queue_hotplug_event(hp_elog, NULL, NULL);
239 else
240 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
241
242 spin_unlock(&ras_log_buf_lock);
243 return IRQ_HANDLED;
244 }
245
246 /* Handle environmental and power warning (EPOW) interrupts. */
ras_epow_interrupt(int irq,void * dev_id)247 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
248 {
249 int status;
250 int state;
251 int critical;
252
253 status = rtas_get_sensor_fast(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX,
254 &state);
255
256 if (state > 3)
257 critical = 1; /* Time Critical */
258 else
259 critical = 0;
260
261 spin_lock(&ras_log_buf_lock);
262
263 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
264 RTAS_VECTOR_EXTERNAL_INTERRUPT,
265 virq_to_hw(irq),
266 RTAS_EPOW_WARNING,
267 critical, __pa(&ras_log_buf),
268 rtas_get_error_log_max());
269
270 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
271
272 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
273
274 spin_unlock(&ras_log_buf_lock);
275 return IRQ_HANDLED;
276 }
277
278 /*
279 * Handle hardware error interrupts.
280 *
281 * RTAS check-exception is called to collect data on the exception. If
282 * the error is deemed recoverable, we log a warning and return.
283 * For nonrecoverable errors, an error is logged and we stop all processing
284 * as quickly as possible in order to prevent propagation of the failure.
285 */
ras_error_interrupt(int irq,void * dev_id)286 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
287 {
288 struct rtas_error_log *rtas_elog;
289 int status;
290 int fatal;
291
292 spin_lock(&ras_log_buf_lock);
293
294 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
295 RTAS_VECTOR_EXTERNAL_INTERRUPT,
296 virq_to_hw(irq),
297 RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
298 __pa(&ras_log_buf),
299 rtas_get_error_log_max());
300
301 rtas_elog = (struct rtas_error_log *)ras_log_buf;
302
303 if (status == 0 &&
304 rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC)
305 fatal = 1;
306 else
307 fatal = 0;
308
309 /* format and print the extended information */
310 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
311
312 if (fatal) {
313 pr_emerg("Fatal hardware error detected. Check RTAS error"
314 " log for details. Powering off immediately\n");
315 emergency_sync();
316 kernel_power_off();
317 } else {
318 pr_err("Recoverable hardware error detected\n");
319 }
320
321 spin_unlock(&ras_log_buf_lock);
322 return IRQ_HANDLED;
323 }
324
325 /*
326 * Some versions of FWNMI place the buffer inside the 4kB page starting at
327 * 0x7000. Other versions place it inside the rtas buffer. We check both.
328 */
329 #define VALID_FWNMI_BUFFER(A) \
330 ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \
331 (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16))))
332
333 /*
334 * Get the error information for errors coming through the
335 * FWNMI vectors. The pt_regs' r3 will be updated to reflect
336 * the actual r3 if possible, and a ptr to the error log entry
337 * will be returned if found.
338 *
339 * If the RTAS error is not of the extended type, then we put it in a per
340 * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf.
341 *
342 * The global_mce_data_buf does not have any locks or protection around it,
343 * if a second machine check comes in, or a system reset is done
344 * before we have logged the error, then we will get corruption in the
345 * error log. This is preferable over holding off on calling
346 * ibm,nmi-interlock which would result in us checkstopping if a
347 * second machine check did come in.
348 */
fwnmi_get_errinfo(struct pt_regs * regs)349 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
350 {
351 unsigned long *savep;
352 struct rtas_error_log *h, *errhdr = NULL;
353
354 /* Mask top two bits */
355 regs->gpr[3] &= ~(0x3UL << 62);
356
357 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
358 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
359 return NULL;
360 }
361
362 savep = __va(regs->gpr[3]);
363 regs->gpr[3] = be64_to_cpu(savep[0]); /* restore original r3 */
364
365 /* If it isn't an extended log we can use the per cpu 64bit buffer */
366 h = (struct rtas_error_log *)&savep[1];
367 if (!rtas_error_extended(h)) {
368 memcpy(this_cpu_ptr(&mce_data_buf), h, sizeof(__u64));
369 errhdr = (struct rtas_error_log *)this_cpu_ptr(&mce_data_buf);
370 } else {
371 int len, error_log_length;
372
373 error_log_length = 8 + rtas_error_extended_log_length(h);
374 len = min_t(int, error_log_length, RTAS_ERROR_LOG_MAX);
375 memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
376 memcpy(global_mce_data_buf, h, len);
377 errhdr = (struct rtas_error_log *)global_mce_data_buf;
378 }
379
380 return errhdr;
381 }
382
383 /* Call this when done with the data returned by FWNMI_get_errinfo.
384 * It will release the saved data area for other CPUs in the
385 * partition to receive FWNMI errors.
386 */
fwnmi_release_errinfo(void)387 static void fwnmi_release_errinfo(void)
388 {
389 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
390 if (ret != 0)
391 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
392 }
393
pSeries_system_reset_exception(struct pt_regs * regs)394 int pSeries_system_reset_exception(struct pt_regs *regs)
395 {
396 #ifdef __LITTLE_ENDIAN__
397 /*
398 * Some firmware byteswaps SRR registers and gives incorrect SRR1. Try
399 * to detect the bad SRR1 pattern here. Flip the NIP back to correct
400 * endian for reporting purposes. Unfortunately the MSR can't be fixed,
401 * so clear it. It will be missing MSR_RI so we won't try to recover.
402 */
403 if ((be64_to_cpu(regs->msr) &
404 (MSR_LE|MSR_RI|MSR_DR|MSR_IR|MSR_ME|MSR_PR|
405 MSR_ILE|MSR_HV|MSR_SF)) == (MSR_DR|MSR_SF)) {
406 regs->nip = be64_to_cpu((__be64)regs->nip);
407 regs->msr = 0;
408 }
409 #endif
410
411 if (fwnmi_active) {
412 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
413 if (errhdr) {
414 /* XXX Should look at FWNMI information */
415 }
416 fwnmi_release_errinfo();
417 }
418
419 if (smp_handle_nmi_ipi(regs))
420 return 1;
421
422 return 0; /* need to perform reset */
423 }
424
425 /*
426 * See if we can recover from a machine check exception.
427 * This is only called on power4 (or above) and only via
428 * the Firmware Non-Maskable Interrupts (fwnmi) handler
429 * which provides the error analysis for us.
430 *
431 * Return 1 if corrected (or delivered a signal).
432 * Return 0 if there is nothing we can do.
433 */
recover_mce(struct pt_regs * regs,struct rtas_error_log * err)434 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
435 {
436 int recovered = 0;
437 int disposition = rtas_error_disposition(err);
438
439 if (!(regs->msr & MSR_RI)) {
440 /* If MSR_RI isn't set, we cannot recover */
441 recovered = 0;
442
443 } else if (disposition == RTAS_DISP_FULLY_RECOVERED) {
444 /* Platform corrected itself */
445 recovered = 1;
446
447 } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) {
448 /* Platform corrected itself but could be degraded */
449 printk(KERN_ERR "MCE: limited recovery, system may "
450 "be degraded\n");
451 recovered = 1;
452
453 } else if (user_mode(regs) && !is_global_init(current) &&
454 rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) {
455
456 /*
457 * If we received a synchronous error when in userspace
458 * kill the task. Firmware may report details of the fail
459 * asynchronously, so we can't rely on the target and type
460 * fields being valid here.
461 */
462 printk(KERN_ERR "MCE: uncorrectable error, killing task "
463 "%s:%d\n", current->comm, current->pid);
464
465 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
466 recovered = 1;
467 }
468
469 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
470
471 return recovered;
472 }
473
474 /*
475 * Handle a machine check.
476 *
477 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
478 * should be present. If so the handler which called us tells us if the
479 * error was recovered (never true if RI=0).
480 *
481 * On hardware prior to Power 4 these exceptions were asynchronous which
482 * means we can't tell exactly where it occurred and so we can't recover.
483 */
pSeries_machine_check_exception(struct pt_regs * regs)484 int pSeries_machine_check_exception(struct pt_regs *regs)
485 {
486 struct rtas_error_log *errp;
487
488 if (fwnmi_active) {
489 errp = fwnmi_get_errinfo(regs);
490 fwnmi_release_errinfo();
491 if (errp && recover_mce(regs, errp))
492 return 1;
493 }
494
495 return 0;
496 }
497