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 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id);
44 static irqreturn_t ras_error_interrupt(int irq, void *dev_id);
48 * Initialize handlers for the set of interrupts caused by hardware errors
49 * and power system events.
51 static int __init init_ras_IRQ(void)
53 struct device_node *np;
55 ras_check_exception_token = rtas_token("check-exception");
58 np = of_find_node_by_path("/event-sources/internal-errors");
60 request_event_sources_irqs(np, ras_error_interrupt,
66 np = of_find_node_by_path("/event-sources/epow-events");
68 request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW");
74 machine_subsys_initcall(pseries, init_ras_IRQ);
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
81 static void handle_system_shutdown(char event_modifier)
83 switch (event_modifier) {
84 case EPOW_SHUTDOWN_NORMAL:
85 pr_emerg("Firmware initiated power off");
86 orderly_poweroff(true);
89 case EPOW_SHUTDOWN_ON_UPS:
90 pr_emerg("Loss of power reported by firmware, system is "
91 "running on UPS/battery");
92 pr_emerg("Check RTAS error log for details");
95 case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS:
96 pr_emerg("Loss of system critical functions reported by "
98 pr_emerg("Check RTAS error log for details");
99 orderly_poweroff(true);
102 case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH:
103 pr_emerg("Ambient temperature too high reported by firmware");
104 pr_emerg("Check RTAS error log for details");
105 orderly_poweroff(true);
109 pr_err("Unknown power/cooling shutdown event (modifier %d)",
114 struct epow_errorlog {
115 unsigned char sensor_value;
116 unsigned char event_modifier;
117 unsigned char extended_modifier;
118 unsigned char reserved;
119 unsigned char platform_reason;
123 #define EPOW_WARN_COOLING 1
124 #define EPOW_WARN_POWER 2
125 #define EPOW_SYSTEM_SHUTDOWN 3
126 #define EPOW_SYSTEM_HALT 4
127 #define EPOW_MAIN_ENCLOSURE 5
128 #define EPOW_POWER_OFF 7
130 static void rtas_parse_epow_errlog(struct rtas_error_log *log)
132 struct pseries_errorlog *pseries_log;
133 struct epow_errorlog *epow_log;
137 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
138 if (pseries_log == NULL)
141 epow_log = (struct epow_errorlog *)pseries_log->data;
142 action_code = epow_log->sensor_value & 0xF; /* bottom 4 bits */
143 modifier = epow_log->event_modifier & 0xF; /* bottom 4 bits */
145 switch (action_code) {
147 pr_err("Non critical power or cooling issue cleared");
150 case EPOW_WARN_COOLING:
151 pr_err("Non critical cooling issue reported by firmware");
152 pr_err("Check RTAS error log for details");
155 case EPOW_WARN_POWER:
156 pr_err("Non critical power issue reported by firmware");
157 pr_err("Check RTAS error log for details");
160 case EPOW_SYSTEM_SHUTDOWN:
161 handle_system_shutdown(epow_log->event_modifier);
164 case EPOW_SYSTEM_HALT:
165 pr_emerg("Firmware initiated power off");
166 orderly_poweroff(true);
169 case EPOW_MAIN_ENCLOSURE:
171 pr_emerg("Critical power/cooling issue reported by firmware");
172 pr_emerg("Check RTAS error log for details");
173 pr_emerg("Immediate power off");
179 pr_err("Unknown power/cooling event (action code %d)",
184 /* Handle environmental and power warning (EPOW) interrupts. */
185 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
191 status = rtas_get_sensor_fast(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX,
195 critical = 1; /* Time Critical */
199 spin_lock(&ras_log_buf_lock);
201 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
202 RTAS_VECTOR_EXTERNAL_INTERRUPT,
205 critical, __pa(&ras_log_buf),
206 rtas_get_error_log_max());
208 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
210 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
212 spin_unlock(&ras_log_buf_lock);
217 * Handle hardware error interrupts.
219 * RTAS check-exception is called to collect data on the exception. If
220 * the error is deemed recoverable, we log a warning and return.
221 * For nonrecoverable errors, an error is logged and we stop all processing
222 * as quickly as possible in order to prevent propagation of the failure.
224 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
226 struct rtas_error_log *rtas_elog;
230 spin_lock(&ras_log_buf_lock);
232 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
233 RTAS_VECTOR_EXTERNAL_INTERRUPT,
235 RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
237 rtas_get_error_log_max());
239 rtas_elog = (struct rtas_error_log *)ras_log_buf;
242 rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC)
247 /* format and print the extended information */
248 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
251 pr_emerg("Fatal hardware error reported by firmware");
252 pr_emerg("Check RTAS error log for details");
253 pr_emerg("Immediate power off");
257 pr_err("Recoverable hardware error reported by firmware");
260 spin_unlock(&ras_log_buf_lock);
265 * Some versions of FWNMI place the buffer inside the 4kB page starting at
266 * 0x7000. Other versions place it inside the rtas buffer. We check both.
267 * Minimum size of the buffer is 16 bytes.
269 #define VALID_FWNMI_BUFFER(A) \
270 ((((A) >= 0x7000) && ((A) <= 0x8000 - 16)) || \
271 (((A) >= rtas.base) && ((A) <= (rtas.base + rtas.size - 16))))
274 * Get the error information for errors coming through the
275 * FWNMI vectors. The pt_regs' r3 will be updated to reflect
276 * the actual r3 if possible, and a ptr to the error log entry
277 * will be returned if found.
279 * If the RTAS error is not of the extended type, then we put it in a per
280 * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf.
282 * The global_mce_data_buf does not have any locks or protection around it,
283 * if a second machine check comes in, or a system reset is done
284 * before we have logged the error, then we will get corruption in the
285 * error log. This is preferable over holding off on calling
286 * ibm,nmi-interlock which would result in us checkstopping if a
287 * second machine check did come in.
289 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
291 unsigned long *savep;
292 struct rtas_error_log *h, *errhdr = NULL;
294 /* Mask top two bits */
295 regs->gpr[3] &= ~(0x3UL << 62);
297 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
298 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
302 savep = __va(regs->gpr[3]);
303 regs->gpr[3] = be64_to_cpu(savep[0]); /* restore original r3 */
305 /* If it isn't an extended log we can use the per cpu 64bit buffer */
306 h = (struct rtas_error_log *)&savep[1];
307 if (!rtas_error_extended(h)) {
308 memcpy(this_cpu_ptr(&mce_data_buf), h, sizeof(__u64));
309 errhdr = (struct rtas_error_log *)this_cpu_ptr(&mce_data_buf);
311 int len, error_log_length;
313 error_log_length = 8 + rtas_error_extended_log_length(h);
314 len = min_t(int, error_log_length, RTAS_ERROR_LOG_MAX);
315 memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
316 memcpy(global_mce_data_buf, h, len);
317 errhdr = (struct rtas_error_log *)global_mce_data_buf;
323 /* Call this when done with the data returned by FWNMI_get_errinfo.
324 * It will release the saved data area for other CPUs in the
325 * partition to receive FWNMI errors.
327 static void fwnmi_release_errinfo(void)
329 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
331 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
334 int pSeries_system_reset_exception(struct pt_regs *regs)
337 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
339 /* XXX Should look at FWNMI information */
341 fwnmi_release_errinfo();
343 return 0; /* need to perform reset */
347 * See if we can recover from a machine check exception.
348 * This is only called on power4 (or above) and only via
349 * the Firmware Non-Maskable Interrupts (fwnmi) handler
350 * which provides the error analysis for us.
352 * Return 1 if corrected (or delivered a signal).
353 * Return 0 if there is nothing we can do.
355 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
358 int disposition = rtas_error_disposition(err);
360 if (!(regs->msr & MSR_RI)) {
361 /* If MSR_RI isn't set, we cannot recover */
364 } else if (disposition == RTAS_DISP_FULLY_RECOVERED) {
365 /* Platform corrected itself */
368 } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) {
369 /* Platform corrected itself but could be degraded */
370 printk(KERN_ERR "MCE: limited recovery, system may "
374 } else if (user_mode(regs) && !is_global_init(current) &&
375 rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) {
378 * If we received a synchronous error when in userspace
379 * kill the task. Firmware may report details of the fail
380 * asynchronously, so we can't rely on the target and type
381 * fields being valid here.
383 printk(KERN_ERR "MCE: uncorrectable error, killing task "
384 "%s:%d\n", current->comm, current->pid);
386 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
390 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
396 * Handle a machine check.
398 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
399 * should be present. If so the handler which called us tells us if the
400 * error was recovered (never true if RI=0).
402 * On hardware prior to Power 4 these exceptions were asynchronous which
403 * means we can't tell exactly where it occurred and so we can't recover.
405 int pSeries_machine_check_exception(struct pt_regs *regs)
407 struct rtas_error_log *errp;
410 errp = fwnmi_get_errinfo(regs);
411 fwnmi_release_errinfo();
412 if (errp && recover_mce(regs, errp))