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 subsys_initcall(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");
94 case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS:
95 pr_emerg("Loss of system critical functions reported by "
97 pr_emerg("Check RTAS error log for details");
98 orderly_poweroff(true);
101 case EPOW_SHUTDOWN_AMBIENT_TEMPERATURE_TOO_HIGH:
102 pr_emerg("Ambient temperature too high reported by firmware");
103 pr_emerg("Check RTAS error log for details");
104 orderly_poweroff(true);
108 pr_err("Unknown power/cooling shutdown event (modifier %d)",
113 struct epow_errorlog {
114 unsigned char sensor_value;
115 unsigned char event_modifier;
116 unsigned char extended_modifier;
117 unsigned char reserved;
118 unsigned char platform_reason;
122 #define EPOW_WARN_COOLING 1
123 #define EPOW_WARN_POWER 2
124 #define EPOW_SYSTEM_SHUTDOWN 3
125 #define EPOW_SYSTEM_HALT 4
126 #define EPOW_MAIN_ENCLOSURE 5
127 #define EPOW_POWER_OFF 7
129 void rtas_parse_epow_errlog(struct rtas_error_log *log)
131 struct pseries_errorlog *pseries_log;
132 struct epow_errorlog *epow_log;
136 pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
137 if (pseries_log == NULL)
140 epow_log = (struct epow_errorlog *)pseries_log->data;
141 action_code = epow_log->sensor_value & 0xF; /* bottom 4 bits */
142 modifier = epow_log->event_modifier & 0xF; /* bottom 4 bits */
144 switch (action_code) {
146 pr_err("Non critical power or cooling issue cleared");
149 case EPOW_WARN_COOLING:
150 pr_err("Non critical cooling issue reported by firmware");
151 pr_err("Check RTAS error log for details");
154 case EPOW_WARN_POWER:
155 pr_err("Non critical power issue reported by firmware");
156 pr_err("Check RTAS error log for details");
159 case EPOW_SYSTEM_SHUTDOWN:
160 handle_system_shutdown(epow_log->event_modifier);
163 case EPOW_SYSTEM_HALT:
164 pr_emerg("Firmware initiated power off");
165 orderly_poweroff(true);
168 case EPOW_MAIN_ENCLOSURE:
170 pr_emerg("Critical power/cooling issue reported by firmware");
171 pr_emerg("Check RTAS error log for details");
172 pr_emerg("Immediate power off");
178 pr_err("Unknown power/cooling event (action code %d)",
183 /* Handle environmental and power warning (EPOW) interrupts. */
184 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
190 status = rtas_get_sensor(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX, &state);
193 critical = 1; /* Time Critical */
197 spin_lock(&ras_log_buf_lock);
199 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
200 RTAS_VECTOR_EXTERNAL_INTERRUPT,
203 critical, __pa(&ras_log_buf),
204 rtas_get_error_log_max());
206 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
208 rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
210 spin_unlock(&ras_log_buf_lock);
215 * Handle hardware error interrupts.
217 * RTAS check-exception is called to collect data on the exception. If
218 * the error is deemed recoverable, we log a warning and return.
219 * For nonrecoverable errors, an error is logged and we stop all processing
220 * as quickly as possible in order to prevent propagation of the failure.
222 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
224 struct rtas_error_log *rtas_elog;
228 spin_lock(&ras_log_buf_lock);
230 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
231 RTAS_VECTOR_EXTERNAL_INTERRUPT,
233 RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
235 rtas_get_error_log_max());
237 rtas_elog = (struct rtas_error_log *)ras_log_buf;
240 rtas_error_severity(rtas_elog) >= RTAS_SEVERITY_ERROR_SYNC)
245 /* format and print the extended information */
246 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
249 pr_emerg("Fatal hardware error reported by firmware");
250 pr_emerg("Check RTAS error log for details");
251 pr_emerg("Immediate power off");
255 pr_err("Recoverable hardware error reported by firmware");
258 spin_unlock(&ras_log_buf_lock);
263 * Some versions of FWNMI place the buffer inside the 4kB page starting at
264 * 0x7000. Other versions place it inside the rtas buffer. We check both.
266 #define VALID_FWNMI_BUFFER(A) \
267 ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \
268 (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16))))
271 * Get the error information for errors coming through the
272 * FWNMI vectors. The pt_regs' r3 will be updated to reflect
273 * the actual r3 if possible, and a ptr to the error log entry
274 * will be returned if found.
276 * If the RTAS error is not of the extended type, then we put it in a per
277 * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf.
279 * The global_mce_data_buf does not have any locks or protection around it,
280 * if a second machine check comes in, or a system reset is done
281 * before we have logged the error, then we will get corruption in the
282 * error log. This is preferable over holding off on calling
283 * ibm,nmi-interlock which would result in us checkstopping if a
284 * second machine check did come in.
286 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
288 unsigned long *savep;
289 struct rtas_error_log *h, *errhdr = NULL;
291 /* Mask top two bits */
292 regs->gpr[3] &= ~(0x3UL << 62);
294 if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
295 printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
299 savep = __va(regs->gpr[3]);
300 regs->gpr[3] = savep[0]; /* restore original r3 */
302 /* If it isn't an extended log we can use the per cpu 64bit buffer */
303 h = (struct rtas_error_log *)&savep[1];
304 if (!rtas_error_extended(h)) {
305 memcpy(&__get_cpu_var(mce_data_buf), h, sizeof(__u64));
306 errhdr = (struct rtas_error_log *)&__get_cpu_var(mce_data_buf);
308 int len, error_log_length;
310 error_log_length = 8 + rtas_error_extended_log_length(h);
311 len = max_t(int, error_log_length, RTAS_ERROR_LOG_MAX);
312 memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
313 memcpy(global_mce_data_buf, h, len);
314 errhdr = (struct rtas_error_log *)global_mce_data_buf;
320 /* Call this when done with the data returned by FWNMI_get_errinfo.
321 * It will release the saved data area for other CPUs in the
322 * partition to receive FWNMI errors.
324 static void fwnmi_release_errinfo(void)
326 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
328 printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
331 int pSeries_system_reset_exception(struct pt_regs *regs)
334 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
336 /* XXX Should look at FWNMI information */
338 fwnmi_release_errinfo();
340 return 0; /* need to perform reset */
344 * See if we can recover from a machine check exception.
345 * This is only called on power4 (or above) and only via
346 * the Firmware Non-Maskable Interrupts (fwnmi) handler
347 * which provides the error analysis for us.
349 * Return 1 if corrected (or delivered a signal).
350 * Return 0 if there is nothing we can do.
352 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
355 int disposition = rtas_error_disposition(err);
357 if (!(regs->msr & MSR_RI)) {
358 /* If MSR_RI isn't set, we cannot recover */
361 } else if (disposition == RTAS_DISP_FULLY_RECOVERED) {
362 /* Platform corrected itself */
365 } else if (disposition == RTAS_DISP_LIMITED_RECOVERY) {
366 /* Platform corrected itself but could be degraded */
367 printk(KERN_ERR "MCE: limited recovery, system may "
371 } else if (user_mode(regs) && !is_global_init(current) &&
372 rtas_error_severity(err) == RTAS_SEVERITY_ERROR_SYNC) {
375 * If we received a synchronous error when in userspace
376 * kill the task. Firmware may report details of the fail
377 * asynchronously, so we can't rely on the target and type
378 * fields being valid here.
380 printk(KERN_ERR "MCE: uncorrectable error, killing task "
381 "%s:%d\n", current->comm, current->pid);
383 _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
387 log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
393 * Handle a machine check.
395 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
396 * should be present. If so the handler which called us tells us if the
397 * error was recovered (never true if RI=0).
399 * On hardware prior to Power 4 these exceptions were asynchronous which
400 * means we can't tell exactly where it occurred and so we can't recover.
402 int pSeries_machine_check_exception(struct pt_regs *regs)
404 struct rtas_error_log *errp;
407 errp = fwnmi_get_errinfo(regs);
408 fwnmi_release_errinfo();
409 if (errp && recover_mce(regs, errp))