2 * Copyright (C) 2001 Dave Engebretsen IBM Corporation
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.
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.
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
19 #include <linux/sched.h>
20 #include <linux/interrupt.h>
21 #include <linux/irq.h>
24 #include <linux/reboot.h>
26 #include <asm/machdep.h>
28 #include <asm/firmware.h>
32 static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX];
33 static DEFINE_SPINLOCK(ras_log_buf_lock);
35 static char global_mce_data_buf[RTAS_ERROR_LOG_MAX];
36 static DEFINE_PER_CPU(__u64, mce_data_buf);
38 static int ras_check_exception_token;
40 #define EPOW_SENSOR_TOKEN 9
41 #define EPOW_SENSOR_INDEX 0
43 /* EPOW events counter variable */
44 static int num_epow_events;
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);
52 * Initialize handlers for the set of interrupts caused by hardware errors
53 * and power system events.
55 static int __init init_ras_IRQ(void)
57 struct device_node *np;
59 ras_check_exception_token = rtas_token("check-exception");
62 np = of_find_node_by_path("/event-sources/internal-errors");
64 request_event_sources_irqs(np, ras_error_interrupt,
70 np = of_find_node_by_path("/event-sources/hot-plug-events");
72 request_event_sources_irqs(np, ras_hotplug_interrupt,
78 np = of_find_node_by_path("/event-sources/epow-events");
80 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW");
86 machine_subsys_initcall(pseries, init_ras_IRQ);
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
93 static void handle_system_shutdown(char event_modifier)
95 switch (event_modifier) {
96 case EPOW_SHUTDOWN_NORMAL:
97 pr_emerg("Power off requested\n");
98 orderly_poweroff(true);
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);
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);
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);
120 pr_err("Unknown power/cooling shutdown event (modifier = %d)\n",
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;
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
141 static void rtas_parse_epow_errlog(struct rtas_error_log *log)
143 struct pseries_errorlog *pseries_log;
144 struct epow_errorlog *epow_log;
148 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
149 if (pseries_log == NULL)
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 */
156 switch (action_code) {
158 if (num_epow_events) {
159 pr_info("Non critical power/cooling issue cleared\n");
164 case EPOW_WARN_COOLING:
165 pr_info("Non-critical cooling issue detected. Check RTAS error"
166 " log for details\n");
169 case EPOW_WARN_POWER:
170 pr_info("Non-critical power issue detected. Check RTAS error"
171 " log for details\n");
174 case EPOW_SYSTEM_SHUTDOWN:
175 handle_system_shutdown(epow_log->event_modifier);
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);
184 case EPOW_MAIN_ENCLOSURE:
186 pr_emerg("System about to lose power. Check RTAS error log "
187 " for details. Powering off immediately.\n");
193 pr_err("Unknown power/cooling event (action code = %d)\n",
197 /* Increment epow events counter variable */
198 if (action_code != EPOW_RESET)
202 static irqreturn_t ras_hotplug_interrupt(int irq, void *dev_id)
204 struct pseries_errorlog *pseries_log;
205 struct pseries_hp_errorlog *hp_elog;
207 spin_lock(&ras_log_buf_lock);
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());
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;
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.
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);
226 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
228 spin_unlock(&ras_log_buf_lock);
232 /* Handle environmental and power warning (EPOW) interrupts. */
233 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
239 status = rtas_get_sensor_fast(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX,
243 critical = 1; /* Time Critical */
247 spin_lock(&ras_log_buf_lock);
249 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
250 RTAS_VECTOR_EXTERNAL_INTERRUPT,
253 critical, __pa(&ras_log_buf),
254 rtas_get_error_log_max());
256 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
258 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
260 spin_unlock(&ras_log_buf_lock);
265 * Handle hardware error interrupts.
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.
272 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
274 struct rtas_error_log *rtas_elog;
278 spin_lock(&ras_log_buf_lock);
280 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
281 RTAS_VECTOR_EXTERNAL_INTERRUPT,
283 RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
285 rtas_get_error_log_max());
287 rtas_elog = (struct rtas_error_log *)ras_log_buf;
290 rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC)
295 /* format and print the extended information */
296 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
299 pr_emerg("Fatal hardware error detected. Check RTAS error"
300 " log for details. Powering off immediately\n");
304 pr_err("Recoverable hardware error detected\n");
307 spin_unlock(&ras_log_buf_lock);
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.
315 #define VALID_FWNMI_BUFFER(A) \
316 ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \
317 (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16))))
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.
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.
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.
335 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
337 unsigned long *savep;
338 struct rtas_error_log *h, *errhdr = NULL;
340 /* Mask top two bits */
341 regs->gpr[3] &= ~(0x3UL << 62);
343 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
344 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
348 savep = __va(regs->gpr[3]);
349 regs->gpr[3] = savep[0]; /* restore original r3 */
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);
357 int len, error_log_length;
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;
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.
373 static void fwnmi_release_errinfo(void)
375 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
377 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
380 int pSeries_system_reset_exception(struct pt_regs *regs)
383 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
385 /* XXX Should look at FWNMI information */
387 fwnmi_release_errinfo();
390 if (smp_handle_nmi_ipi(regs))
393 return 0; /* need to perform reset */
397 * See if we can recover from a machine check exception.
398 * This is only called on power4 (or above) and only via
399 * the Firmware Non-Maskable Interrupts (fwnmi) handler
400 * which provides the error analysis for us.
402 * Return 1 if corrected (or delivered a signal).
403 * Return 0 if there is nothing we can do.
405 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
408 int disposition = rtas_error_disposition(err);
410 if (!(regs->msr & MSR_RI)) {
411 /* If MSR_RI isn't set, we cannot recover */
414 } else if (disposition == RTAS_DISP_FULLY_RECOVERED) {
415 /* Platform corrected itself */
418 } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) {
419 /* Platform corrected itself but could be degraded */
420 printk(KERN_ERR "MCE: limited recovery, system may "
424 } else if (user_mode(regs) && !is_global_init(current) &&
425 rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) {
428 * If we received a synchronous error when in userspace
429 * kill the task. Firmware may report details of the fail
430 * asynchronously, so we can't rely on the target and type
431 * fields being valid here.
433 printk(KERN_ERR "MCE: uncorrectable error, killing task "
434 "%s:%d\n", current->comm, current->pid);
436 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
440 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
446 * Handle a machine check.
448 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
449 * should be present. If so the handler which called us tells us if the
450 * error was recovered (never true if RI=0).
452 * On hardware prior to Power 4 these exceptions were asynchronous which
453 * means we can't tell exactly where it occurred and so we can't recover.
455 int pSeries_machine_check_exception(struct pt_regs *regs)
457 struct rtas_error_log *errp;
460 errp = fwnmi_get_errinfo(regs);
461 fwnmi_release_errinfo();
462 if (errp && recover_mce(regs, errp))