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 
26 #include <asm/machdep.h>
27 #include <asm/rtas.h>
28 #include <asm/firmware.h>
29 
30 #include "pseries.h"
31 
32 static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX];
33 static DEFINE_SPINLOCK(ras_log_buf_lock);
34 
35 static char global_mce_data_buf[RTAS_ERROR_LOG_MAX];
36 static DEFINE_PER_CPU(__u64, mce_data_buf);
37 
38 static int ras_check_exception_token;
39 
40 #define EPOW_SENSOR_TOKEN	9
41 #define EPOW_SENSOR_INDEX	0
42 
43 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id);
44 static irqreturn_t ras_error_interrupt(int irq, void *dev_id);
45 
46 
47 /*
48  * Initialize handlers for the set of interrupts caused by hardware errors
49  * and power system events.
50  */
51 static int __init init_ras_IRQ(void)
52 {
53 	struct device_node *np;
54 
55 	ras_check_exception_token = rtas_token("check-exception");
56 
57 	/* Internal Errors */
58 	np = of_find_node_by_path("/event-sources/internal-errors");
59 	if (np != NULL) {
60 		request_event_sources_irqs(np, ras_error_interrupt,
61 					   "RAS_ERROR");
62 		of_node_put(np);
63 	}
64 
65 	/* EPOW Events */
66 	np = of_find_node_by_path("/event-sources/epow-events");
67 	if (np != NULL) {
68 		request_event_sources_irqs(np, ras_epow_interrupt, "RAS_EPOW");
69 		of_node_put(np);
70 	}
71 
72 	return 0;
73 }
74 subsys_initcall(init_ras_IRQ);
75 
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
80 
81 static void handle_system_shutdown(char event_modifier)
82 {
83 	switch (event_modifier) {
84 	case EPOW_SHUTDOWN_NORMAL:
85 		pr_emerg("Firmware initiated power off");
86 		orderly_poweroff(true);
87 		break;
88 
89 	case EPOW_SHUTDOWN_ON_UPS:
90 		pr_emerg("Loss of power reported by firmware, system is "
91 			"running on UPS/battery");
92 		break;
93 
94 	case EPOW_SHUTDOWN_LOSS_OF_CRITICAL_FUNCTIONS:
95 		pr_emerg("Loss of system critical functions reported by "
96 			"firmware");
97 		pr_emerg("Check RTAS error log for details");
98 		orderly_poweroff(true);
99 		break;
100 
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);
105 		break;
106 
107 	default:
108 		pr_err("Unknown power/cooling shutdown event (modifier %d)",
109 			event_modifier);
110 	}
111 }
112 
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;
119 };
120 
121 #define EPOW_RESET			0
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
128 
129 void rtas_parse_epow_errlog(struct rtas_error_log *log)
130 {
131 	struct pseries_errorlog *pseries_log;
132 	struct epow_errorlog *epow_log;
133 	char action_code;
134 	char modifier;
135 
136 	pseries_log = get_pseries_errorlog(log, PSERIES_ELOG_SECT_ID_EPOW);
137 	if (pseries_log == NULL)
138 		return;
139 
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 */
143 
144 	switch (action_code) {
145 	case EPOW_RESET:
146 		pr_err("Non critical power or cooling issue cleared");
147 		break;
148 
149 	case EPOW_WARN_COOLING:
150 		pr_err("Non critical cooling issue reported by firmware");
151 		pr_err("Check RTAS error log for details");
152 		break;
153 
154 	case EPOW_WARN_POWER:
155 		pr_err("Non critical power issue reported by firmware");
156 		pr_err("Check RTAS error log for details");
157 		break;
158 
159 	case EPOW_SYSTEM_SHUTDOWN:
160 		handle_system_shutdown(epow_log->event_modifier);
161 		break;
162 
163 	case EPOW_SYSTEM_HALT:
164 		pr_emerg("Firmware initiated power off");
165 		orderly_poweroff(true);
166 		break;
167 
168 	case EPOW_MAIN_ENCLOSURE:
169 	case EPOW_POWER_OFF:
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");
173 		emergency_sync();
174 		kernel_power_off();
175 		break;
176 
177 	default:
178 		pr_err("Unknown power/cooling event (action code %d)",
179 			action_code);
180 	}
181 }
182 
183 /* Handle environmental and power warning (EPOW) interrupts. */
184 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id)
185 {
186 	int status;
187 	int state;
188 	int critical;
189 
190 	status = rtas_get_sensor(EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX, &state);
191 
192 	if (state > 3)
193 		critical = 1;		/* Time Critical */
194 	else
195 		critical = 0;
196 
197 	spin_lock(&ras_log_buf_lock);
198 
199 	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
200 			   RTAS_VECTOR_EXTERNAL_INTERRUPT,
201 			   virq_to_hw(irq),
202 			   RTAS_EPOW_WARNING,
203 			   critical, __pa(&ras_log_buf),
204 				rtas_get_error_log_max());
205 
206 	log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
207 
208 	rtas_parse_epow_errlog((struct rtas_error_log *)ras_log_buf);
209 
210 	spin_unlock(&ras_log_buf_lock);
211 	return IRQ_HANDLED;
212 }
213 
214 /*
215  * Handle hardware error interrupts.
216  *
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.
221  */
222 static irqreturn_t ras_error_interrupt(int irq, void *dev_id)
223 {
224 	struct rtas_error_log *rtas_elog;
225 	int status;
226 	int fatal;
227 
228 	spin_lock(&ras_log_buf_lock);
229 
230 	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
231 			   RTAS_VECTOR_EXTERNAL_INTERRUPT,
232 			   virq_to_hw(irq),
233 			   RTAS_INTERNAL_ERROR, 1 /* Time Critical */,
234 			   __pa(&ras_log_buf),
235 				rtas_get_error_log_max());
236 
237 	rtas_elog = (struct rtas_error_log *)ras_log_buf;
238 
239 	if ((status == 0) && (rtas_elog->severity >= RTAS_SEVERITY_ERROR_SYNC))
240 		fatal = 1;
241 	else
242 		fatal = 0;
243 
244 	/* format and print the extended information */
245 	log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
246 
247 	if (fatal) {
248 		pr_emerg("Fatal hardware error reported by firmware");
249 		pr_emerg("Check RTAS error log for details");
250 		pr_emerg("Immediate power off");
251 		emergency_sync();
252 		kernel_power_off();
253 	} else {
254 		pr_err("Recoverable hardware error reported by firmware");
255 	}
256 
257 	spin_unlock(&ras_log_buf_lock);
258 	return IRQ_HANDLED;
259 }
260 
261 /*
262  * Some versions of FWNMI place the buffer inside the 4kB page starting at
263  * 0x7000. Other versions place it inside the rtas buffer. We check both.
264  */
265 #define VALID_FWNMI_BUFFER(A) \
266 	((((A) >= 0x7000) && ((A) < 0x7ff0)) || \
267 	(((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16))))
268 
269 /*
270  * Get the error information for errors coming through the
271  * FWNMI vectors.  The pt_regs' r3 will be updated to reflect
272  * the actual r3 if possible, and a ptr to the error log entry
273  * will be returned if found.
274  *
275  * If the RTAS error is not of the extended type, then we put it in a per
276  * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf.
277  *
278  * The global_mce_data_buf does not have any locks or protection around it,
279  * if a second machine check comes in, or a system reset is done
280  * before we have logged the error, then we will get corruption in the
281  * error log.  This is preferable over holding off on calling
282  * ibm,nmi-interlock which would result in us checkstopping if a
283  * second machine check did come in.
284  */
285 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
286 {
287 	unsigned long *savep;
288 	struct rtas_error_log *h, *errhdr = NULL;
289 
290 	/* Mask top two bits */
291 	regs->gpr[3] &= ~(0x3UL << 62);
292 
293 	if (!VALID_FWNMI_BUFFER(regs->gpr[3])) {
294 		printk(KERN_ERR "FWNMI: corrupt r3 0x%016lx\n", regs->gpr[3]);
295 		return NULL;
296 	}
297 
298 	savep = __va(regs->gpr[3]);
299 	regs->gpr[3] = savep[0];	/* restore original r3 */
300 
301 	/* If it isn't an extended log we can use the per cpu 64bit buffer */
302 	h = (struct rtas_error_log *)&savep[1];
303 	if (!h->extended) {
304 		memcpy(&__get_cpu_var(mce_data_buf), h, sizeof(__u64));
305 		errhdr = (struct rtas_error_log *)&__get_cpu_var(mce_data_buf);
306 	} else {
307 		int len;
308 
309 		len = max_t(int, 8+h->extended_log_length, RTAS_ERROR_LOG_MAX);
310 		memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
311 		memcpy(global_mce_data_buf, h, len);
312 		errhdr = (struct rtas_error_log *)global_mce_data_buf;
313 	}
314 
315 	return errhdr;
316 }
317 
318 /* Call this when done with the data returned by FWNMI_get_errinfo.
319  * It will release the saved data area for other CPUs in the
320  * partition to receive FWNMI errors.
321  */
322 static void fwnmi_release_errinfo(void)
323 {
324 	int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
325 	if (ret != 0)
326 		printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret);
327 }
328 
329 int pSeries_system_reset_exception(struct pt_regs *regs)
330 {
331 	if (fwnmi_active) {
332 		struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
333 		if (errhdr) {
334 			/* XXX Should look at FWNMI information */
335 		}
336 		fwnmi_release_errinfo();
337 	}
338 	return 0; /* need to perform reset */
339 }
340 
341 /*
342  * See if we can recover from a machine check exception.
343  * This is only called on power4 (or above) and only via
344  * the Firmware Non-Maskable Interrupts (fwnmi) handler
345  * which provides the error analysis for us.
346  *
347  * Return 1 if corrected (or delivered a signal).
348  * Return 0 if there is nothing we can do.
349  */
350 static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err)
351 {
352 	int recovered = 0;
353 
354 	if (!(regs->msr & MSR_RI)) {
355 		/* If MSR_RI isn't set, we cannot recover */
356 		recovered = 0;
357 
358 	} else if (err->disposition == RTAS_DISP_FULLY_RECOVERED) {
359 		/* Platform corrected itself */
360 		recovered = 1;
361 
362 	} else if (err->disposition == RTAS_DISP_LIMITED_RECOVERY) {
363 		/* Platform corrected itself but could be degraded */
364 		printk(KERN_ERR "MCE: limited recovery, system may "
365 		       "be degraded\n");
366 		recovered = 1;
367 
368 	} else if (user_mode(regs) && !is_global_init(current) &&
369 		   err->severity == RTAS_SEVERITY_ERROR_SYNC) {
370 
371 		/*
372 		 * If we received a synchronous error when in userspace
373 		 * kill the task. Firmware may report details of the fail
374 		 * asynchronously, so we can't rely on the target and type
375 		 * fields being valid here.
376 		 */
377 		printk(KERN_ERR "MCE: uncorrectable error, killing task "
378 		       "%s:%d\n", current->comm, current->pid);
379 
380 		_exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip);
381 		recovered = 1;
382 	}
383 
384 	log_error((char *)err, ERR_TYPE_RTAS_LOG, 0);
385 
386 	return recovered;
387 }
388 
389 /*
390  * Handle a machine check.
391  *
392  * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
393  * should be present.  If so the handler which called us tells us if the
394  * error was recovered (never true if RI=0).
395  *
396  * On hardware prior to Power 4 these exceptions were asynchronous which
397  * means we can't tell exactly where it occurred and so we can't recover.
398  */
399 int pSeries_machine_check_exception(struct pt_regs *regs)
400 {
401 	struct rtas_error_log *errp;
402 
403 	if (fwnmi_active) {
404 		errp = fwnmi_get_errinfo(regs);
405 		fwnmi_release_errinfo();
406 		if (errp && recover_mce(regs, errp))
407 			return 1;
408 	}
409 
410 	return 0;
411 }
412