[deliverable/linux.git] / arch / powerpc / platforms / pseries / ras.c

/*
 * Copyright (C) 2001 Dave Engebretsen IBM Corporation
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
 */

/* Change Activity:
 * 2001/09/21 : engebret : Created with minimal EPOW and HW exception support.
 * End Change Activity
 */

#include <linux/errno.h>
#include <linux/threads.h>
#include <linux/kernel_stat.h>
#include <linux/signal.h>
#include <linux/sched.h>
#include <linux/ioport.h>
#include <linux/interrupt.h>
#include <linux/timex.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/pci.h>
#include <linux/delay.h>
#include <linux/irq.h>
#include <linux/random.h>
#include <linux/sysrq.h>
#include <linux/bitops.h>

#include <asm/uaccess.h>
#include <asm/system.h>
#include <asm/io.h>
#include <asm/pgtable.h>
#include <asm/irq.h>
#include <asm/cache.h>
#include <asm/prom.h>
#include <asm/ptrace.h>
#include <asm/machdep.h>
#include <asm/rtas.h>
#include <asm/udbg.h>
#include <asm/firmware.h>

#include "ras.h"

static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX];
static DEFINE_SPINLOCK(ras_log_buf_lock);

char mce_data_buf[RTAS_ERROR_LOG_MAX];

static int ras_get_sensor_state_token;
static int ras_check_exception_token;

#define EPOW_SENSOR_TOKEN	9
#define EPOW_SENSOR_INDEX	0
#define RAS_VECTOR_OFFSET	0x500

static irqreturn_t ras_epow_interrupt(int irq, void *dev_id,
					struct pt_regs * regs);
static irqreturn_t ras_error_interrupt(int irq, void *dev_id,
					struct pt_regs * regs);

/* #define DEBUG */


static void request_ras_irqs(struct device_node *np,
			irqreturn_t (*handler)(int, void *, struct pt_regs *),
			const char *name)
{
	int i, index, count = 0;
	struct of_irq oirq;
	u32 *opicprop;
	unsigned int opicplen;
	unsigned int virqs[16];

	/* Check for obsolete "open-pic-interrupt" property. If present, then
	 * map those interrupts using the default interrupt host and default
	 * trigger
	 */
	opicprop = (u32 *)get_property(np, "open-pic-interrupt", &opicplen);
	if (opicprop) {
		opicplen /= sizeof(u32);
		for (i = 0; i < opicplen; i++) {
			if (count > 15)
				break;
			virqs[count] = irq_create_mapping(NULL, *(opicprop++),
							 IRQ_TYPE_NONE);
			if (virqs[count] == NO_IRQ)
				printk(KERN_ERR "Unable to allocate interrupt "
				       "number for %s\n", np->full_name);
			else
				count++;

		}
	}
	/* Else use normal interrupt tree parsing */
	else {
		/* First try to do a proper OF tree parsing */
		for (index = 0; of_irq_map_one(np, index, &oirq) == 0;
		     index++) {
			if (count > 15)
				break;
			virqs[count] = irq_create_of_mapping(oirq.controller,
							    oirq.specifier,
							    oirq.size);
			if (virqs[count] == NO_IRQ)
				printk(KERN_ERR "Unable to allocate interrupt "
				       "number for %s\n", np->full_name);
			else
				count++;
		}
	}

	/* Now request them */
	for (i = 0; i < count; i++) {
		if (request_irq(virqs[i], handler, 0, name, NULL)) {
			printk(KERN_ERR "Unable to request interrupt %d for "
			       "%s\n", virqs[i], np->full_name);
			return;
		}
	}
}

/*
 * Initialize handlers for the set of interrupts caused by hardware errors
 * and power system events.
 */
static int __init init_ras_IRQ(void)
{
	struct device_node *np;

	ras_get_sensor_state_token = rtas_token("get-sensor-state");
	ras_check_exception_token = rtas_token("check-exception");

	/* Internal Errors */
	np = of_find_node_by_path("/event-sources/internal-errors");
	if (np != NULL) {
		request_ras_irqs(np, ras_error_interrupt, "RAS_ERROR");
		of_node_put(np);
	}

	/* EPOW Events */
	np = of_find_node_by_path("/event-sources/epow-events");
	if (np != NULL) {
		request_ras_irqs(np, ras_epow_interrupt, "RAS_EPOW");
		of_node_put(np);
	}

	return 0;
}
__initcall(init_ras_IRQ);

/*
 * Handle power subsystem events (EPOW).
 *
 * Presently we just log the event has occurred.  This should be fixed
 * to examine the type of power failure and take appropriate action where
 * the time horizon permits something useful to be done.
 */
static irqreturn_t
ras_epow_interrupt(int irq, void *dev_id, struct pt_regs * regs)
{
	int status = 0xdeadbeef;
	int state = 0;
	int critical;

	status = rtas_call(ras_get_sensor_state_token, 2, 2, &state,
			   EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX);

	if (state > 3)
		critical = 1;  /* Time Critical */
	else
		critical = 0;

	spin_lock(&ras_log_buf_lock);

	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
			   RAS_VECTOR_OFFSET,
			   irq_map[irq].hwirq,
			   RTAS_EPOW_WARNING | RTAS_POWERMGM_EVENTS,
			   critical, __pa(&ras_log_buf),
				rtas_get_error_log_max());

	udbg_printf("EPOW <0x%lx 0x%x 0x%x>\n",
		    *((unsigned long *)&ras_log_buf), status, state);
	printk(KERN_WARNING "EPOW <0x%lx 0x%x 0x%x>\n",
	       *((unsigned long *)&ras_log_buf), status, state);

	/* format and print the extended information */
	log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);

	spin_unlock(&ras_log_buf_lock);
	return IRQ_HANDLED;
}

/*
 * Handle hardware error interrupts.
 *
 * RTAS check-exception is called to collect data on the exception.  If
 * the error is deemed recoverable, we log a warning and return.
 * For nonrecoverable errors, an error is logged and we stop all processing
 * as quickly as possible in order to prevent propagation of the failure.
 */
static irqreturn_t
ras_error_interrupt(int irq, void *dev_id, struct pt_regs * regs)
{
	struct rtas_error_log *rtas_elog;
	int status = 0xdeadbeef;
	int fatal;

	spin_lock(&ras_log_buf_lock);

	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
			   RAS_VECTOR_OFFSET,
			   irq_map[irq].hwirq,
			   RTAS_INTERNAL_ERROR, 1 /*Time Critical */,
			   __pa(&ras_log_buf),
				rtas_get_error_log_max());

	rtas_elog = (struct rtas_error_log *)ras_log_buf;

	if ((status == 0) && (rtas_elog->severity >= RTAS_SEVERITY_ERROR_SYNC))
		fatal = 1;
	else
		fatal = 0;

	/* format and print the extended information */
	log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);

	if (fatal) {
		udbg_printf("Fatal HW Error <0x%lx 0x%x>\n",
			    *((unsigned long *)&ras_log_buf), status);
		printk(KERN_EMERG "Error: Fatal hardware error <0x%lx 0x%x>\n",
		       *((unsigned long *)&ras_log_buf), status);

#ifndef DEBUG
		/* Don't actually power off when debugging so we can test
		 * without actually failing while injecting errors.
		 * Error data will not be logged to syslog.
		 */
		ppc_md.power_off();
#endif
	} else {
		udbg_printf("Recoverable HW Error <0x%lx 0x%x>\n",
			    *((unsigned long *)&ras_log_buf), status);
		printk(KERN_WARNING
		       "Warning: Recoverable hardware error <0x%lx 0x%x>\n",
		       *((unsigned long *)&ras_log_buf), status);
	}

	spin_unlock(&ras_log_buf_lock);
	return IRQ_HANDLED;
}

/* Get the error information for errors coming through the
 * FWNMI vectors.  The pt_regs' r3 will be updated to reflect
 * the actual r3 if possible, and a ptr to the error log entry
 * will be returned if found.
 *
 * The mce_data_buf does not have any locks or protection around it,
 * if a second machine check comes in, or a system reset is done
 * before we have logged the error, then we will get corruption in the
 * error log.  This is preferable over holding off on calling
 * ibm,nmi-interlock which would result in us checkstopping if a
 * second machine check did come in.
 */
static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
{
	unsigned long errdata = regs->gpr[3];
	struct rtas_error_log *errhdr = NULL;
	unsigned long *savep;

	if ((errdata >= 0x7000 && errdata < 0x7fff0) ||
	    (errdata >= rtas.base && errdata < rtas.base + rtas.size - 16)) {
		savep = __va(errdata);
		regs->gpr[3] = savep[0];	/* restore original r3 */
		memset(mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
		memcpy(mce_data_buf, (char *)(savep + 1), RTAS_ERROR_LOG_MAX);
		errhdr = (struct rtas_error_log *)mce_data_buf;
	} else {
		printk("FWNMI: corrupt r3\n");
	}
	return errhdr;
}

/* Call this when done with the data returned by FWNMI_get_errinfo.
 * It will release the saved data area for other CPUs in the
 * partition to receive FWNMI errors.
 */
static void fwnmi_release_errinfo(void)
{
	int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
	if (ret != 0)
		printk("FWNMI: nmi-interlock failed: %d\n", ret);
}

int pSeries_system_reset_exception(struct pt_regs *regs)
{
	if (fwnmi_active) {
		struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
		if (errhdr) {
			/* XXX Should look at FWNMI information */
		}
		fwnmi_release_errinfo();
	}
	return 0; /* need to perform reset */
}

/*
 * See if we can recover from a machine check exception.
 * This is only called on power4 (or above) and only via
 * the Firmware Non-Maskable Interrupts (fwnmi) handler
 * which provides the error analysis for us.
 *
 * Return 1 if corrected (or delivered a signal).
 * Return 0 if there is nothing we can do.
 */
static int recover_mce(struct pt_regs *regs, struct rtas_error_log * err)
{
	int nonfatal = 0;

	if (err->disposition == RTAS_DISP_FULLY_RECOVERED) {
		/* Platform corrected itself */
		nonfatal = 1;
	} else if ((regs->msr & MSR_RI) &&
		   user_mode(regs) &&
		   err->severity == RTAS_SEVERITY_ERROR_SYNC &&
		   err->disposition == RTAS_DISP_NOT_RECOVERED &&
		   err->target == RTAS_TARGET_MEMORY &&
		   err->type == RTAS_TYPE_ECC_UNCORR &&
		   !(current->pid == 0 || current->pid == 1)) {
		/* Kill off a user process with an ECC error */
		printk(KERN_ERR "MCE: uncorrectable ecc error for pid %d\n",
		       current->pid);
		/* XXX something better for ECC error? */
		_exception(SIGBUS, regs, BUS_ADRERR, regs->nip);
		nonfatal = 1;
	}

	log_error((char *)err, ERR_TYPE_RTAS_LOG, !nonfatal);

	return nonfatal;
}

/*
 * Handle a machine check.
 *
 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
 * should be present.  If so the handler which called us tells us if the
 * error was recovered (never true if RI=0).
 *
 * On hardware prior to Power 4 these exceptions were asynchronous which
 * means we can't tell exactly where it occurred and so we can't recover.
 */
int pSeries_machine_check_exception(struct pt_regs *regs)
{
	struct rtas_error_log *errp;

	if (fwnmi_active) {
		errp = fwnmi_get_errinfo(regs);
		fwnmi_release_errinfo();
		if (errp && recover_mce(regs, errp))
			return 1;
	}

	return 0;
}
Commit	Line	Data
1da177e4	1	/*
1da177e4	2	* Copyright (C) 2001 Dave Engebretsen IBM Corporation
d9953105	3	*
1da177e4 LT	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.
d9953105	8	*
1da177e4 LT	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.
d9953105	13	*
1da177e4 LT	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	/* Change Activity:
	20	* 2001/09/21 : engebret : Created with minimal EPOW and HW exception support.
d9953105	21	* End Change Activity
1da177e4 LT	22	*/
	23
	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/slab.h>
	34	#include <linux/pci.h>
	35	#include <linux/delay.h>
	36	#include <linux/irq.h>
	37	#include <linux/random.h>
	38	#include <linux/sysrq.h>
	39	#include <linux/bitops.h>
	40
	41	#include <asm/uaccess.h>
	42	#include <asm/system.h>
	43	#include <asm/io.h>
	44	#include <asm/pgtable.h>
	45	#include <asm/irq.h>
	46	#include <asm/cache.h>
	47	#include <asm/prom.h>
	48	#include <asm/ptrace.h>
1da177e4 LT	49	#include <asm/machdep.h>
1da177e4 LT	50	#include <asm/rtas.h>
dcad47fc	51	#include <asm/udbg.h>
8c4f1f29	52	#include <asm/firmware.h>
1da177e4	53
c902be71 AB	54	#include "ras.h"
c902be71 AB	55
1da177e4 LT	56	static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX];
	57	static DEFINE_SPINLOCK(ras_log_buf_lock);
	58
8c4f1f29	59	char mce_data_buf[RTAS_ERROR_LOG_MAX];
1da177e4	60
1da177e4 LT	61	static int ras_get_sensor_state_token;
	62	static int ras_check_exception_token;
	63
	64	#define EPOW_SENSOR_TOKEN 9
	65	#define EPOW_SENSOR_INDEX 0
	66	#define RAS_VECTOR_OFFSET 0x500
	67
	68	static irqreturn_t ras_epow_interrupt(int irq, void *dev_id,
	69	struct pt_regs * regs);
	70	static irqreturn_t ras_error_interrupt(int irq, void *dev_id,
	71	struct pt_regs * regs);
	72
	73	/* #define DEBUG */
	74
0ebfff14 BH	75
0ebfff14 BH	76	static void request_ras_irqs(struct device_node *np,
1da177e4 LT	77	irqreturn_t (handler)(int, void , struct pt_regs *),
	78	const char *name)
	79	{
0ebfff14 BH	80	int i, index, count = 0;
	81	struct of_irq oirq;
	82	u32 *opicprop;
	83	unsigned int opicplen;
	84	unsigned int virqs[16];
	85
	86	/* Check for obsolete "open-pic-interrupt" property. If present, then
	87	* map those interrupts using the default interrupt host and default
	88	* trigger
	89	*/
	90	opicprop = (u32 *)get_property(np, "open-pic-interrupt", &opicplen);
	91	if (opicprop) {
	92	opicplen /= sizeof(u32);
	93	for (i = 0; i < opicplen; i++) {
	94	if (count > 15)
	95	break;
	96	virqs[count] = irq_create_mapping(NULL, *(opicprop++),
	97	IRQ_TYPE_NONE);
	98	if (virqs[count] == NO_IRQ)
	99	printk(KERN_ERR "Unable to allocate interrupt "
	100	"number for %s\n", np->full_name);
	101	else
	102	count++;
	103
1da177e4	104	}
0ebfff14 BH	105	}
	106	/* Else use normal interrupt tree parsing */
	107	else {
	108	/* First try to do a proper OF tree parsing */
	109	for (index = 0; of_irq_map_one(np, index, &oirq) == 0;
	110	index++) {
	111	if (count > 15)
	112	break;
	113	virqs[count] = irq_create_of_mapping(oirq.controller,
	114	oirq.specifier,
	115	oirq.size);
	116	if (virqs[count] == NO_IRQ)
	117	printk(KERN_ERR "Unable to allocate interrupt "
	118	"number for %s\n", np->full_name);
	119	else
	120	count++;
	121	}
	122	}
	123
	124	/* Now request them */
	125	for (i = 0; i < count; i++) {
	126	if (request_irq(virqs[i], handler, 0, name, NULL)) {
1da177e4	127	printk(KERN_ERR "Unable to request interrupt %d for "
0ebfff14	128	"%s\n", virqs[i], np->full_name);
1da177e4 LT	129	return;
1da177e4 LT	130	}
1da177e4 LT	131	}
	132	}
	133
	134	/*
	135	* Initialize handlers for the set of interrupts caused by hardware errors
	136	* and power system events.
	137	*/
	138	static int __init init_ras_IRQ(void)
	139	{
	140	struct device_node *np;
	141
	142	ras_get_sensor_state_token = rtas_token("get-sensor-state");
	143	ras_check_exception_token = rtas_token("check-exception");
	144
	145	/* Internal Errors */
	146	np = of_find_node_by_path("/event-sources/internal-errors");
	147	if (np != NULL) {
0ebfff14	148	request_ras_irqs(np, ras_error_interrupt, "RAS_ERROR");
1da177e4 LT	149	of_node_put(np);
	150	}
	151
	152	/* EPOW Events */
	153	np = of_find_node_by_path("/event-sources/epow-events");
	154	if (np != NULL) {
0ebfff14	155	request_ras_irqs(np, ras_epow_interrupt, "RAS_EPOW");
1da177e4 LT	156	of_node_put(np);
	157	}
	158
69ed3324	159	return 0;
1da177e4 LT	160	}
	161	__initcall(init_ras_IRQ);
	162
	163	/*
	164	* Handle power subsystem events (EPOW).
	165	*
	166	* Presently we just log the event has occurred. This should be fixed
	167	* to examine the type of power failure and take appropriate action where
	168	* the time horizon permits something useful to be done.
	169	*/
	170	static irqreturn_t
	171	ras_epow_interrupt(int irq, void dev_id, struct pt_regs regs)
	172	{
	173	int status = 0xdeadbeef;
	174	int state = 0;
	175	int critical;
	176
	177	status = rtas_call(ras_get_sensor_state_token, 2, 2, &state,
	178	EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX);
	179
	180	if (state > 3)
	181	critical = 1; /* Time Critical */
	182	else
	183	critical = 0;
	184
	185	spin_lock(&ras_log_buf_lock);
	186
	187	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
	188	RAS_VECTOR_OFFSET,
0ebfff14	189	irq_map[irq].hwirq,
1da177e4 LT	190	RTAS_EPOW_WARNING \| RTAS_POWERMGM_EVENTS,
	191	critical, __pa(&ras_log_buf),
	192	rtas_get_error_log_max());
	193
	194	udbg_printf("EPOW <0x%lx 0x%x 0x%x>\n",
	195	((unsigned long )&ras_log_buf), status, state);
	196	printk(KERN_WARNING "EPOW <0x%lx 0x%x 0x%x>\n",
	197	((unsigned long )&ras_log_buf), status, state);
	198
	199	/* format and print the extended information */
	200	log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
	201
	202	spin_unlock(&ras_log_buf_lock);
	203	return IRQ_HANDLED;
	204	}
	205
	206	/*
	207	* Handle hardware error interrupts.
	208	*
	209	* RTAS check-exception is called to collect data on the exception. If
	210	* the error is deemed recoverable, we log a warning and return.
	211	* For nonrecoverable errors, an error is logged and we stop all processing
	212	* as quickly as possible in order to prevent propagation of the failure.
	213	*/
	214	static irqreturn_t
	215	ras_error_interrupt(int irq, void dev_id, struct pt_regs regs)
	216	{
	217	struct rtas_error_log *rtas_elog;
	218	int status = 0xdeadbeef;
	219	int fatal;
	220
	221	spin_lock(&ras_log_buf_lock);
	222
	223	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
	224	RAS_VECTOR_OFFSET,
0ebfff14	225	irq_map[irq].hwirq,
1da177e4 LT	226	RTAS_INTERNAL_ERROR, 1 /Time Critical /,
	227	__pa(&ras_log_buf),
	228	rtas_get_error_log_max());
	229
	230	rtas_elog = (struct rtas_error_log *)ras_log_buf;
	231
	232	if ((status == 0) && (rtas_elog->severity >= RTAS_SEVERITY_ERROR_SYNC))
	233	fatal = 1;
	234	else
	235	fatal = 0;
	236
	237	/* format and print the extended information */
	238	log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
	239
	240	if (fatal) {
	241	udbg_printf("Fatal HW Error <0x%lx 0x%x>\n",
	242	((unsigned long )&ras_log_buf), status);
	243	printk(KERN_EMERG "Error: Fatal hardware error <0x%lx 0x%x>\n",
	244	((unsigned long )&ras_log_buf), status);
	245
	246	#ifndef DEBUG
	247	/* Don't actually power off when debugging so we can test
	248	* without actually failing while injecting errors.
	249	* Error data will not be logged to syslog.
	250	*/
	251	ppc_md.power_off();
	252	#endif
	253	} else {
	254	udbg_printf("Recoverable HW Error <0x%lx 0x%x>\n",
	255	((unsigned long )&ras_log_buf), status);
	256	printk(KERN_WARNING
	257	"Warning: Recoverable hardware error <0x%lx 0x%x>\n",
	258	((unsigned long )&ras_log_buf), status);
	259	}
	260
	261	spin_unlock(&ras_log_buf_lock);
	262	return IRQ_HANDLED;
	263	}
	264
	265	/* Get the error information for errors coming through the
	266	* FWNMI vectors. The pt_regs' r3 will be updated to reflect
	267	* the actual r3 if possible, and a ptr to the error log entry
	268	* will be returned if found.
	269	*
	270	* The mce_data_buf does not have any locks or protection around it,
	271	* if a second machine check comes in, or a system reset is done
	272	* before we have logged the error, then we will get corruption in the
	273	* error log. This is preferable over holding off on calling
	274	* ibm,nmi-interlock which would result in us checkstopping if a
	275	* second machine check did come in.
	276	*/
	277	static struct rtas_error_log fwnmi_get_errinfo(struct pt_regs regs)
	278	{
	279	unsigned long errdata = regs->gpr[3];
	280	struct rtas_error_log *errhdr = NULL;
	281	unsigned long *savep;
	282
	283	if ((errdata >= 0x7000 && errdata < 0x7fff0) \|\|
	284	(errdata >= rtas.base && errdata < rtas.base + rtas.size - 16)) {
	285	savep = __va(errdata);
	286	regs->gpr[3] = savep[0]; /* restore original r3 */
	287	memset(mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
	288	memcpy(mce_data_buf, (char *)(savep + 1), RTAS_ERROR_LOG_MAX);
	289	errhdr = (struct rtas_error_log *)mce_data_buf;
290	} else {
291	printk("FWNMI: corrupt r3\n");
292	}
293	return errhdr;
294	}
295
296	/* Call this when done with the data returned by FWNMI_get_errinfo.
297	* It will release the saved data area for other CPUs in the
298	* partition to receive FWNMI errors.
299	*/
300	static void fwnmi_release_errinfo(void)
301	{
302	int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
303	if (ret != 0)
304	printk("FWNMI: nmi-interlock failed: %d\n", ret);
305	}
306
c902be71	307	int pSeries_system_reset_exception(struct pt_regs *regs)
1da177e4 LT	308	{
	309	if (fwnmi_active) {
	310	struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
	311	if (errhdr) {
	312	/* XXX Should look at FWNMI information */
	313	}
	314	fwnmi_release_errinfo();
	315	}
c902be71	316	return 0; /* need to perform reset */
1da177e4 LT	317	}
	318
	319	/*
	320	* See if we can recover from a machine check exception.
	321	* This is only called on power4 (or above) and only via
	322	* the Firmware Non-Maskable Interrupts (fwnmi) handler
	323	* which provides the error analysis for us.
	324	*
	325	* Return 1 if corrected (or delivered a signal).
	326	* Return 0 if there is nothing we can do.
	327	*/
	328	static int recover_mce(struct pt_regs regs, struct rtas_error_log err)
	329	{
	330	int nonfatal = 0;
	331
	332	if (err->disposition == RTAS_DISP_FULLY_RECOVERED) {
	333	/* Platform corrected itself */
	334	nonfatal = 1;
	335	} else if ((regs->msr & MSR_RI) &&
	336	user_mode(regs) &&
	337	err->severity == RTAS_SEVERITY_ERROR_SYNC &&
	338	err->disposition == RTAS_DISP_NOT_RECOVERED &&
	339	err->target == RTAS_TARGET_MEMORY &&
	340	err->type == RTAS_TYPE_ECC_UNCORR &&
	341	!(current->pid == 0 \|\| current->pid == 1)) {
	342	/* Kill off a user process with an ECC error */
	343	printk(KERN_ERR "MCE: uncorrectable ecc error for pid %d\n",
	344	current->pid);
	345	/* XXX something better for ECC error? */
	346	_exception(SIGBUS, regs, BUS_ADRERR, regs->nip);
	347	nonfatal = 1;
	348	}
	349
d9953105	350	log_error((char *)err, ERR_TYPE_RTAS_LOG, !nonfatal);
1da177e4 LT	351
	352	return nonfatal;
	353	}
	354
	355	/*
	356	* Handle a machine check.
	357	*
	358	* Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
	359	* should be present. If so the handler which called us tells us if the
	360	* error was recovered (never true if RI=0).
	361	*
	362	* On hardware prior to Power 4 these exceptions were asynchronous which
	363	* means we can't tell exactly where it occurred and so we can't recover.
	364	*/
	365	int pSeries_machine_check_exception(struct pt_regs *regs)
	366	{
	367	struct rtas_error_log *errp;
	368
	369	if (fwnmi_active) {
	370	errp = fwnmi_get_errinfo(regs);
	371	fwnmi_release_errinfo();
	372	if (errp && recover_mce(regs, errp))
	373	return 1;
	374	}
	375
	376	return 0;
	377	}