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 #include <linux/irq_work.h>
26
27 #include <asm/machdep.h>
28 #include <asm/rtas.h>
29 #include <asm/firmware.h>
30
31 #include "pseries.h"
32
33 static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX];
34 static DEFINE_SPINLOCK(ras_log_buf_lock);
35
36 static int ras_check_exception_token;
37
38 static void mce_process_errlog_event(struct irq_work *work);
39 static struct irq_work mce_errlog_process_work = {
40 .func = mce_process_errlog_event,
41 };
42
43 #define EPOW_SENSOR_TOKEN 9
44 #define EPOW_SENSOR_INDEX 0
45
46 /* EPOW events counter variable */
47 static int num_epow_events;
48
49 static irqreturn_t ras_hotplug_interrupt(int irq, void *dev_id);
50 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id);
51 static irqreturn_t ras_error_interrupt(int irq, void *dev_id);
52
53
54 /*
55 * Enable the hotplug interrupt late because processing them may touch other
56 * devices or systems (e.g. hugepages) that have not been initialized at the
57 * subsys stage.
58 */
init_ras_hotplug_IRQ(void)59 int __init init_ras_hotplug_IRQ(void)
60 {
61 struct device_node *np;
62
63 /* Hotplug Events */
64 np = of_find_node_by_path("/event-sources/hot-plug-events");
65 if (np != NULL) {
66 if (dlpar_workqueue_init() == 0)
67 request_event_sources_irqs(np, ras_hotplug_interrupt,
68 "RAS_HOTPLUG");
69 of_node_put(np);
70 }
71
72 return 0;
73 }
74 machine_late_initcall(pseries, init_ras_hotplug_IRQ);
75
76 /*
77 * Initialize handlers for the set of interrupts caused by hardware errors
78 * and power system events.
79 */
init_ras_IRQ(void)80 static int __init init_ras_IRQ(void)
81 {
82 struct device_node *np;
83
84 ras_check_exception_token = rtas_token("check-exception");
85
86 /* Internal Errors */
87 np = of_find_node_by_path("/event-sources/internal-errors");
88 if (np != NULL) {
89 request_event_sources_irqs(np, ras_error_interrupt,
90 "RAS_ERROR");
91 of_node_put(np);
92 }
93
94 /* EPOW Events */
95 np = of_find_node_by_path("/event-sources/epow-events");
96 if (np != NULL) {
97 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW");
98 of_node_put(np);
99 }
100
101 return 0;
102 }
103 machine_subsys_initcall(pseries, init_ras_IRQ);
104
105 #define EPOW_SHUTDOWN_NORMAL 1
106 #define EPOW_SHUTDOWN_ON_UPS 2
107 #define EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS 3
108 #define EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH 4
109
handle_system_shutdown(char event_modifier)110 static void handle_system_shutdown(char event_modifier)
111 {
112 switch (event_modifier) {
113 case EPOW_SHUTDOWN_NORMAL:
114 pr_emerg("Power off requested\n");
115 orderly_poweroff(true);
116 break;
117
118 case EPOW_SHUTDOWN_ON_UPS:
119 pr_emerg("Loss of system power detected. System is running on"
120 " UPS/battery. Check RTAS error log for details\n");
121 orderly_poweroff(true);
122 break;
123
124 case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS:
125 pr_emerg("Loss of system critical functions detected. Check"
126 " RTAS error log for details\n");
127 orderly_poweroff(true);
128 break;
129
130 case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH:
131 pr_emerg("High ambient temperature detected. Check RTAS"
132 " error log for details\n");
133 orderly_poweroff(true);
134 break;
135
136 default:
137 pr_err("Unknown power/cooling shutdown event (modifier = %d)\n",
138 event_modifier);
139 }
140 }
141
142 struct epow_errorlog {
143 unsigned char sensor_value;
144 unsigned char event_modifier;
145 unsigned char extended_modifier;
146 unsigned char reserved;
147 unsigned char platform_reason;
148 };
149
150 #define EPOW_RESET 0
151 #define EPOW_WARN_COOLING 1
152 #define EPOW_WARN_POWER 2
153 #define EPOW_SYSTEM_SHUTDOWN 3
154 #define EPOW_SYSTEM_HALT 4
155 #define EPOW_MAIN_ENCLOSURE 5
156 #define EPOW_POWER_OFF 7
157
rtas_parse_epow_errlog(struct rtas_error_log * log)158 static void rtas_parse_epow_errlog(struct rtas_error_log *log)
159 {
160 struct pseries_errorlog *pseries_log;
161 struct epow_errorlog *epow_log;
162 char action_code;
163 char modifier;
164
165 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
166 if (pseries_log == NULL)
167 return;
168
169 epow_log = (struct epow_errorlog *)pseries_log->data;
170 action_code = epow_log->sensor_value & 0xF; /* bottom 4 bits */
171 modifier = epow_log->event_modifier & 0xF; /* bottom 4 bits */
172
173 switch (action_code) {
174 case EPOW_RESET:
175 if (num_epow_events) {
176 pr_info("Non critical power/cooling issue cleared\n");
177 num_epow_events--;
178 }
179 break;
180
181 case EPOW_WARN_COOLING:
182 pr_info("Non-critical cooling issue detected. Check RTAS error"
183 " log for details\n");
184 break;
185
186 case EPOW_WARN_POWER:
187 pr_info("Non-critical power issue detected. Check RTAS error"
188 " log for details\n");
189 break;
190
191 case EPOW_SYSTEM_SHUTDOWN:
192 handle_system_shutdown(epow_log->event_modifier);
193 break;
194
195 case EPOW_SYSTEM_HALT:
196 pr_emerg("Critical power/cooling issue detected. Check RTAS"
197 " error log for details. Powering off.\n");
198 orderly_poweroff(true);
199 break;
200
201 case EPOW_MAIN_ENCLOSURE:
202 case EPOW_POWER_OFF:
203 pr_emerg("System about to lose power. Check RTAS error log "
204 " for details. Powering off immediately.\n");
205 emergency_sync();
206 kernel_power_off();
207 break;
208
209 default:
210 pr_err("Unknown power/cooling event (action code = %d)\n",
211 action_code);
212 }
213
214 /* Increment epow events counter variable */
215 if (action_code != EPOW_RESET)
216 num_epow_events++;
217 }
218
ras_hotplug_interrupt(int irq,void * dev_id)219 static irqreturn_t ras_hotplug_interrupt(int irq, void *dev_id)
220 {
221 struct pseries_errorlog *pseries_log;
222 struct pseries_hp_errorlog *hp_elog;
223
224 spin_lock(&ras_log_buf_lock);
225
226 rtas_call(ras_check_exception_token, 6, 1, NULL,
227 RTAS_VECTOR_EXTERNAL_INTERRUPT, virq_to_hw(irq),
228 RTAS_HOTPLUG_EVENTS, 0, __pa(&ras_log_buf),
229 rtas_get_error_log_max());
230
231 pseries_log = get_pseries_errorlog((struct rtas_error_log *)ras_log_buf,
232 PSERIES_ELOG_SECT_ID_HOTPLUG);
233 hp_elog = (struct pseries_hp_errorlog *)pseries_log->data;
234
235 /*
236 * Since PCI hotplug is not currently supported on pseries, put PCI
237 * hotplug events on the ras_log_buf to be handled by rtas_errd.
238 */
239 if (hp_elog->resource == PSERIES_HP_ELOG_RESOURCE_MEM ||
240 hp_elog->resource == PSERIES_HP_ELOG_RESOURCE_CPU)
241 queue_hotplug_event(hp_elog, NULL, NULL);
242 else
243 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
244
245 spin_unlock(&ras_log_buf_lock);
246 return IRQ_HANDLED;
247 }
248
249 /* Handle environmental and power warning (EPOW) interrupts. */
ras_epow_interrupt(int irq,void * dev_id)250 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
251 {
252 int status;
253 int state;
254 int critical;
255
256 status = rtas_get_sensor_fast(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX,
257 &state);
258
259 if (state > 3)
260 critical = 1; /* Time Critical */
261 else
262 critical = 0;
263
264 spin_lock(&ras_log_buf_lock);
265
266 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
267 RTAS_VECTOR_EXTERNAL_INTERRUPT,
268 virq_to_hw(irq),
269 RTAS_EPOW_WARNING,
270 critical, __pa(&ras_log_buf),
271 rtas_get_error_log_max());
272
273 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
274
275 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
276
277 spin_unlock(&ras_log_buf_lock);
278 return IRQ_HANDLED;
279 }
280
281 /*
282 * Handle hardware error interrupts.
283 *
284 * RTAS check-exception is called to collect data on the exception. If
285 * the error is deemed recoverable, we log a warning and return.
286 * For nonrecoverable errors, an error is logged and we stop all processing
287 * as quickly as possible in order to prevent propagation of the failure.
288 */
ras_error_interrupt(int irq,void * dev_id)289 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
290 {
291 struct rtas_error_log *rtas_elog;
292 int status;
293 int fatal;
294
295 spin_lock(&ras_log_buf_lock);
296
297 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
298 RTAS_VECTOR_EXTERNAL_INTERRUPT,
299 virq_to_hw(irq),
300 RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
301 __pa(&ras_log_buf),
302 rtas_get_error_log_max());
303
304 rtas_elog = (struct rtas_error_log *)ras_log_buf;
305
306 if (status == 0 &&
307 rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC)
308 fatal = 1;
309 else
310 fatal = 0;
311
312 /* format and print the extended information */
313 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
314
315 if (fatal) {
316 pr_emerg("Fatal hardware error detected. Check RTAS error"
317 " log for details. Powering off immediately\n");
318 emergency_sync();
319 kernel_power_off();
320 } else {
321 pr_err("Recoverable hardware error detected\n");
322 }
323
324 spin_unlock(&ras_log_buf_lock);
325 return IRQ_HANDLED;
326 }
327
328 /*
329 * Some versions of FWNMI place the buffer inside the 4kB page starting at
330 * 0x7000. Other versions place it inside the rtas buffer. We check both.
331 */
332 #define VALID_FWNMI_BUFFER(A) \
333 ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \
334 (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16))))
335
fwnmi_get_errlog(void)336 static inline struct rtas_error_log *fwnmi_get_errlog(void)
337 {
338 return (struct rtas_error_log *)local_paca->mce_data_buf;
339 }
340
341 /*
342 * Get the error information for errors coming through the
343 * FWNMI vectors. The pt_regs' r3 will be updated to reflect
344 * the actual r3 if possible, and a ptr to the error log entry
345 * will be returned if found.
346 *
347 * Use one buffer mce_data_buf per cpu to store RTAS error.
348 *
349 * The mce_data_buf does not have any locks or protection around it,
350 * if a second machine check comes in, or a system reset is done
351 * before we have logged the error, then we will get corruption in the
352 * error log. This is preferable over holding off on calling
353 * ibm,nmi-interlock which would result in us checkstopping if a
354 * second machine check did come in.
355 */
fwnmi_get_errinfo(struct pt_regs * regs)356 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
357 {
358 unsigned long *savep;
359 struct rtas_error_log *h;
360
361 /* Mask top two bits */
362 regs->gpr[3] &= ~(0x3UL << 62);
363
364 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
365 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
366 return NULL;
367 }
368
369 savep = __va(regs->gpr[3]);
370 regs->gpr[3] = be64_to_cpu(savep[0]); /* restore original r3 */
371
372 h = (struct rtas_error_log *)&savep[1];
373 /* Use the per cpu buffer from paca to store rtas error log */
374 memset(local_paca->mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
375 if (!rtas_error_extended(h)) {
376 memcpy(local_paca->mce_data_buf, h, sizeof(__u64));
377 } else {
378 int len, error_log_length;
379
380 error_log_length = 8 + rtas_error_extended_log_length(h);
381 len = min_t(int, error_log_length, RTAS_ERROR_LOG_MAX);
382 memcpy(local_paca->mce_data_buf, h, len);
383 }
384
385 return (struct rtas_error_log *)local_paca->mce_data_buf;
386 }
387
388 /* Call this when done with the data returned by FWNMI_get_errinfo.
389 * It will release the saved data area for other CPUs in the
390 * partition to receive FWNMI errors.
391 */
fwnmi_release_errinfo(void)392 static void fwnmi_release_errinfo(void)
393 {
394 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
395 if (ret != 0)
396 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
397 }
398
pSeries_system_reset_exception(struct pt_regs * regs)399 int pSeries_system_reset_exception(struct pt_regs *regs)
400 {
401 #ifdef __LITTLE_ENDIAN__
402 /*
403 * Some firmware byteswaps SRR registers and gives incorrect SRR1. Try
404 * to detect the bad SRR1 pattern here. Flip the NIP back to correct
405 * endian for reporting purposes. Unfortunately the MSR can't be fixed,
406 * so clear it. It will be missing MSR_RI so we won't try to recover.
407 */
408 if ((be64_to_cpu(regs->msr) &
409 (MSR_LE|MSR_RI|MSR_DR|MSR_IR|MSR_ME|MSR_PR|
410 MSR_ILE|MSR_HV|MSR_SF)) == (MSR_DR|MSR_SF)) {
411 regs->nip = be64_to_cpu((__be64)regs->nip);
412 regs->msr = 0;
413 }
414 #endif
415
416 if (fwnmi_active) {
417 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
418 if (errhdr) {
419 /* XXX Should look at FWNMI information */
420 }
421 fwnmi_release_errinfo();
422 }
423
424 if (smp_handle_nmi_ipi(regs))
425 return 1;
426
427 return 0; /* need to perform reset */
428 }
429
430 /*
431 * Process MCE rtas errlog event.
432 */
mce_process_errlog_event(struct irq_work * work)433 static void mce_process_errlog_event(struct irq_work *work)
434 {
435 struct rtas_error_log *err;
436
437 err = fwnmi_get_errlog();
438 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
439 }
440
441 /*
442 * See if we can recover from a machine check exception.
443 * This is only called on power4 (or above) and only via
444 * the Firmware Non-Maskable Interrupts (fwnmi) handler
445 * which provides the error analysis for us.
446 *
447 * Return 1 if corrected (or delivered a signal).
448 * Return 0 if there is nothing we can do.
449 */
recover_mce(struct pt_regs * regs,struct rtas_error_log * err)450 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
451 {
452 int recovered = 0;
453 int disposition = rtas_error_disposition(err);
454
455 if (!(regs->msr & MSR_RI)) {
456 /* If MSR_RI isn't set, we cannot recover */
457 recovered = 0;
458
459 } else if (disposition == RTAS_DISP_FULLY_RECOVERED) {
460 /* Platform corrected itself */
461 recovered = 1;
462
463 } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) {
464 /* Platform corrected itself but could be degraded */
465 printk(KERN_ERR "MCE: limited recovery, system may "
466 "be degraded\n");
467 recovered = 1;
468
469 } else if (user_mode(regs) && !is_global_init(current) &&
470 rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) {
471
472 /*
473 * If we received a synchronous error when in userspace
474 * kill the task. Firmware may report details of the fail
475 * asynchronously, so we can't rely on the target and type
476 * fields being valid here.
477 */
478 printk(KERN_ERR "MCE: uncorrectable error, killing task "
479 "%s:%d\n", current->comm, current->pid);
480
481 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
482 recovered = 1;
483 }
484
485 /* Queue irq work to log this rtas event later. */
486 irq_work_queue(&mce_errlog_process_work);
487
488 return recovered;
489 }
490
491 /*
492 * Handle a machine check.
493 *
494 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
495 * should be present. If so the handler which called us tells us if the
496 * error was recovered (never true if RI=0).
497 *
498 * On hardware prior to Power 4 these exceptions were asynchronous which
499 * means we can't tell exactly where it occurred and so we can't recover.
500 */
pSeries_machine_check_exception(struct pt_regs * regs)501 int pSeries_machine_check_exception(struct pt_regs *regs)
502 {
503 struct rtas_error_log *errp;
504
505 if (fwnmi_active) {
506 errp = fwnmi_get_errinfo(regs);
507 fwnmi_release_errinfo();
508 if (errp && recover_mce(regs, errp))
509 return 1;
510 }
511
512 return 0;
513 }
514
