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
20 * 2001/09/21 : engebret : Created with minimal EPOW and HW exception support.
24 #include <linux/errno.h>
25 #include <linux/threads.h>
26 #include <linux/kernel_stat.h>
27 #include <linux/signal.h>
28 #include <linux/sched.h>
29 #include <linux/ioport.h>
30 #include <linux/interrupt.h>
31 #include <linux/timex.h>
32 #include <linux/init.h>
33 #include <linux/delay.h>
34 #include <linux/irq.h>
35 #include <linux/random.h>
36 #include <linux/sysrq.h>
37 #include <linux/bitops.h>
39 #include <asm/uaccess.h>
40 #include <asm/system.h>
42 #include <asm/pgtable.h>
44 #include <asm/cache.h>
46 #include <asm/ptrace.h>
47 #include <asm/machdep.h>
50 #include <asm/firmware.h>
54 static unsigned char ras_log_buf
[RTAS_ERROR_LOG_MAX
];
55 static DEFINE_SPINLOCK(ras_log_buf_lock
);
57 static char mce_data_buf
[RTAS_ERROR_LOG_MAX
];
59 static int ras_get_sensor_state_token
;
60 static int ras_check_exception_token
;
62 #define EPOW_SENSOR_TOKEN 9
63 #define EPOW_SENSOR_INDEX 0
65 static irqreturn_t
ras_epow_interrupt(int irq
, void *dev_id
);
66 static irqreturn_t
ras_error_interrupt(int irq
, void *dev_id
);
70 * Initialize handlers for the set of interrupts caused by hardware errors
71 * and power system events.
73 static int __init
init_ras_IRQ(void)
75 struct device_node
*np
;
77 ras_get_sensor_state_token
= rtas_token("get-sensor-state");
78 ras_check_exception_token
= rtas_token("check-exception");
81 np
= of_find_node_by_path("/event-sources/internal-errors");
83 request_event_sources_irqs(np
, ras_error_interrupt
,
89 np
= of_find_node_by_path("/event-sources/epow-events");
91 request_event_sources_irqs(np
, ras_epow_interrupt
, "RAS_EPOW");
97 __initcall(init_ras_IRQ
);
100 * Handle power subsystem events (EPOW).
102 * Presently we just log the event has occurred. This should be fixed
103 * to examine the type of power failure and take appropriate action where
104 * the time horizon permits something useful to be done.
106 static irqreturn_t
ras_epow_interrupt(int irq
, void *dev_id
)
108 int status
= 0xdeadbeef;
112 status
= rtas_call(ras_get_sensor_state_token
, 2, 2, &state
,
113 EPOW_SENSOR_TOKEN
, EPOW_SENSOR_INDEX
);
116 critical
= 1; /* Time Critical */
120 spin_lock(&ras_log_buf_lock
);
122 status
= rtas_call(ras_check_exception_token
, 6, 1, NULL
,
123 RTAS_VECTOR_EXTERNAL_INTERRUPT
,
125 RTAS_EPOW_WARNING
| RTAS_POWERMGM_EVENTS
,
126 critical
, __pa(&ras_log_buf
),
127 rtas_get_error_log_max());
129 udbg_printf("EPOW <0x%lx 0x%x 0x%x>\n",
130 *((unsigned long *)&ras_log_buf
), status
, state
);
131 printk(KERN_WARNING
"EPOW <0x%lx 0x%x 0x%x>\n",
132 *((unsigned long *)&ras_log_buf
), status
, state
);
134 /* format and print the extended information */
135 log_error(ras_log_buf
, ERR_TYPE_RTAS_LOG
, 0);
137 spin_unlock(&ras_log_buf_lock
);
142 * Handle hardware error interrupts.
144 * RTAS check-exception is called to collect data on the exception. If
145 * the error is deemed recoverable, we log a warning and return.
146 * For nonrecoverable errors, an error is logged and we stop all processing
147 * as quickly as possible in order to prevent propagation of the failure.
149 static irqreturn_t
ras_error_interrupt(int irq
, void *dev_id
)
151 struct rtas_error_log
*rtas_elog
;
152 int status
= 0xdeadbeef;
155 spin_lock(&ras_log_buf_lock
);
157 status
= rtas_call(ras_check_exception_token
, 6, 1, NULL
,
158 RTAS_VECTOR_EXTERNAL_INTERRUPT
,
160 RTAS_INTERNAL_ERROR
, 1 /*Time Critical */,
162 rtas_get_error_log_max());
164 rtas_elog
= (struct rtas_error_log
*)ras_log_buf
;
166 if ((status
== 0) && (rtas_elog
->severity
>= RTAS_SEVERITY_ERROR_SYNC
))
171 /* format and print the extended information */
172 log_error(ras_log_buf
, ERR_TYPE_RTAS_LOG
, fatal
);
175 udbg_printf("Fatal HW Error <0x%lx 0x%x>\n",
176 *((unsigned long *)&ras_log_buf
), status
);
177 printk(KERN_EMERG
"Error: Fatal hardware error <0x%lx 0x%x>\n",
178 *((unsigned long *)&ras_log_buf
), status
);
180 #ifndef DEBUG_RTAS_POWER_OFF
181 /* Don't actually power off when debugging so we can test
182 * without actually failing while injecting errors.
183 * Error data will not be logged to syslog.
188 udbg_printf("Recoverable HW Error <0x%lx 0x%x>\n",
189 *((unsigned long *)&ras_log_buf
), status
);
191 "Warning: Recoverable hardware error <0x%lx 0x%x>\n",
192 *((unsigned long *)&ras_log_buf
), status
);
195 spin_unlock(&ras_log_buf_lock
);
199 /* Get the error information for errors coming through the
200 * FWNMI vectors. The pt_regs' r3 will be updated to reflect
201 * the actual r3 if possible, and a ptr to the error log entry
202 * will be returned if found.
204 * The mce_data_buf does not have any locks or protection around it,
205 * if a second machine check comes in, or a system reset is done
206 * before we have logged the error, then we will get corruption in the
207 * error log. This is preferable over holding off on calling
208 * ibm,nmi-interlock which would result in us checkstopping if a
209 * second machine check did come in.
211 static struct rtas_error_log
*fwnmi_get_errinfo(struct pt_regs
*regs
)
213 unsigned long errdata
= regs
->gpr
[3];
214 struct rtas_error_log
*errhdr
= NULL
;
215 unsigned long *savep
;
217 if ((errdata
>= 0x7000 && errdata
< 0x7fff0) ||
218 (errdata
>= rtas
.base
&& errdata
< rtas
.base
+ rtas
.size
- 16)) {
219 savep
= __va(errdata
);
220 regs
->gpr
[3] = savep
[0]; /* restore original r3 */
221 memset(mce_data_buf
, 0, RTAS_ERROR_LOG_MAX
);
222 memcpy(mce_data_buf
, (char *)(savep
+ 1), RTAS_ERROR_LOG_MAX
);
223 errhdr
= (struct rtas_error_log
*)mce_data_buf
;
225 printk("FWNMI: corrupt r3\n");
230 /* Call this when done with the data returned by FWNMI_get_errinfo.
231 * It will release the saved data area for other CPUs in the
232 * partition to receive FWNMI errors.
234 static void fwnmi_release_errinfo(void)
236 int ret
= rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL
);
238 printk("FWNMI: nmi-interlock failed: %d\n", ret
);
241 int pSeries_system_reset_exception(struct pt_regs
*regs
)
244 struct rtas_error_log
*errhdr
= fwnmi_get_errinfo(regs
);
246 /* XXX Should look at FWNMI information */
248 fwnmi_release_errinfo();
250 return 0; /* need to perform reset */
254 * See if we can recover from a machine check exception.
255 * This is only called on power4 (or above) and only via
256 * the Firmware Non-Maskable Interrupts (fwnmi) handler
257 * which provides the error analysis for us.
259 * Return 1 if corrected (or delivered a signal).
260 * Return 0 if there is nothing we can do.
262 static int recover_mce(struct pt_regs
*regs
, struct rtas_error_log
*err
)
266 if (!(regs
->msr
& MSR_RI
)) {
267 /* If MSR_RI isn't set, we cannot recover */
270 } else if (err
->disposition
== RTAS_DISP_FULLY_RECOVERED
) {
271 /* Platform corrected itself */
274 } else if (err
->disposition
== RTAS_DISP_LIMITED_RECOVERY
) {
275 /* Platform corrected itself but could be degraded */
276 printk(KERN_ERR
"MCE: limited recovery, system may "
280 } else if (user_mode(regs
) && !is_global_init(current
) &&
281 err
->severity
== RTAS_SEVERITY_ERROR_SYNC
) {
284 * If we received a synchronous error when in userspace
285 * kill the task. Firmware may report details of the fail
286 * asynchronously, so we can't rely on the target and type
287 * fields being valid here.
289 printk(KERN_ERR
"MCE: uncorrectable error, killing task "
290 "%s:%d\n", current
->comm
, current
->pid
);
292 _exception(SIGBUS
, regs
, BUS_MCEERR_AR
, regs
->nip
);
296 log_error((char *)err
, ERR_TYPE_RTAS_LOG
, 0);
302 * Handle a machine check.
304 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
305 * should be present. If so the handler which called us tells us if the
306 * error was recovered (never true if RI=0).
308 * On hardware prior to Power 4 these exceptions were asynchronous which
309 * means we can't tell exactly where it occurred and so we can't recover.
311 int pSeries_machine_check_exception(struct pt_regs
*regs
)
313 struct rtas_error_log
*errp
;
316 errp
= fwnmi_get_errinfo(regs
);
317 fwnmi_release_errinfo();
318 if (errp
&& recover_mce(regs
, errp
))