Home | History | Annotate | Line # | Download | only in dev
memerr.c revision 1.8
      1 /*	$NetBSD: memerr.c,v 1.8 1997/04/28 21:59:22 gwr Exp $ */
      2 
      3 /*
      4  * Copyright (c) 1992, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * This software was developed by the Computer Systems Engineering group
      8  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
      9  * contributed to Berkeley.
     10  *
     11  * All advertising materials mentioning features or use of this software
     12  * must display the following acknowledgement:
     13  *	This product includes software developed by the University of
     14  *	California, Lawrence Berkeley Laboratory.
     15  *
     16  * Redistribution and use in source and binary forms, with or without
     17  * modification, are permitted provided that the following conditions
     18  * are met:
     19  * 1. Redistributions of source code must retain the above copyright
     20  *    notice, this list of conditions and the following disclaimer.
     21  * 2. Redistributions in binary form must reproduce the above copyright
     22  *    notice, this list of conditions and the following disclaimer in the
     23  *    documentation and/or other materials provided with the distribution.
     24  * 3. All advertising materials mentioning features or use of this software
     25  *    must display the following acknowledgement:
     26  *	This product includes software developed by the University of
     27  *	California, Berkeley and its contributors.
     28  * 4. Neither the name of the University nor the names of its contributors
     29  *    may be used to endorse or promote products derived from this software
     30  *    without specific prior written permission.
     31  *
     32  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     33  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     34  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     35  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     36  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     37  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     38  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     39  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     40  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     41  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     42  * SUCH DAMAGE.
     43  *
     44  *	@(#)memreg.c	8.1 (Berkeley) 6/11/93
     45  */
     46 
     47 #include <sys/param.h>
     48 #include <sys/systm.h>
     49 #include <sys/device.h>
     50 
     51 #include <machine/autoconf.h>
     52 #include <machine/control.h>
     53 #include <machine/cpu.h>
     54 #include <machine/idprom.h>
     55 #include <machine/obio.h>
     56 #include <machine/pte.h>
     57 
     58 #include <sun3/dev/memerr.h>
     59 /* #include <sun3/dev/eccreg.h> - not yet */
     60 
     61 #define	ME_PRI	7	/* Interrupt level (NMI) */
     62 
     63 extern unsigned char cpu_machine_id;
     64 
     65 enum memerr_type { ME_PAR = 0, ME_ECC = 1 };
     66 
     67 struct memerr_softc {
     68 	struct device sc_dev;
     69 	struct memerr *sc_reg;
     70 	enum memerr_type sc_type;
     71 	char *sc_typename;	/* "Parity" or "ECC" */
     72 	char *sc_csrbits;	/* how to print csr bits */
     73 	/* XXX: counters? */
     74 };
     75 
     76 static int  memerr_match __P((struct device *, struct cfdata *, void *));
     77 static void memerr_attach __P((struct device *, struct device *, void *));
     78 static int  memerr_interrupt __P((void *));
     79 static void memerr_correctable __P((struct memerr_softc *));
     80 
     81 struct cfattach memerr_ca = {
     82 	sizeof(struct memerr_softc), memerr_match, memerr_attach
     83 };
     84 
     85 struct cfdriver memerr_cd = {
     86 	NULL, "memerr", DV_DULL
     87 };
     88 
     89 
     90 static int
     91 memerr_match(parent, cf, args)
     92     struct device *parent;
     93     struct cfdata *cf;
     94     void *args;
     95 {
     96 	struct confargs *ca = args;
     97 
     98 	/* This driver only supports one unit. */
     99 	if (cf->cf_unit != 0)
    100 		return (0);
    101 
    102 	/* We use obio_mapin(), so require OBIO. */
    103 	if (ca->ca_bustype != BUS_OBIO)
    104 		return (0);
    105 
    106 	/* Make sure there is something there... */
    107 	if (bus_peek(ca->ca_bustype, ca->ca_paddr, 1) == -1)
    108 		return (0);
    109 
    110 	/* Default interrupt priority. */
    111 	if (ca->ca_intpri == -1)
    112 		ca->ca_intpri = ME_PRI;
    113 
    114 	return (1);
    115 }
    116 
    117 static void
    118 memerr_attach(parent, self, args)
    119 	struct device *parent;
    120 	struct device *self;
    121 	void *args;
    122 {
    123 	struct memerr_softc *sc = (void *)self;
    124 	struct confargs *ca = args;
    125 	struct memerr *mer;
    126 
    127 	/*
    128 	 * Which type of memory subsystem do we have?
    129 	 */
    130 	switch (cpu_machine_id) {
    131 	case SUN3_MACH_160:		/* XXX: correct? */
    132 	case SUN3_MACH_260:
    133 		sc->sc_type = ME_ECC;
    134 		sc->sc_typename = "ECC";
    135 		sc->sc_csrbits = ME_ECC_STR;
    136 		break;
    137 
    138 	default:
    139 		sc->sc_type = ME_PAR;
    140 		sc->sc_typename = "Parity";
    141 		sc->sc_csrbits = ME_PAR_STR;
    142 		break;
    143 	}
    144 	printf(": (%s memory)\n", sc->sc_typename);
    145 
    146 	mer = (struct memerr *)
    147 	    obio_mapin(ca->ca_paddr, sizeof(*mer));
    148 	if (mer == NULL)
    149 		panic("memerr: can not map register");
    150 	sc->sc_reg = mer;
    151 
    152 	/* Install interrupt handler. */
    153 	isr_add_autovect(memerr_interrupt,
    154 		(void *)sc, ca->ca_intpri);
    155 
    156 	/* Enable error interrupt (and checking). */
    157 	if (sc->sc_type == ME_PAR)
    158 		mer->me_csr = ME_CSR_IENA | ME_PAR_CHECK;
    159 	else {
    160 		/*
    161 		 * XXX:  Some day, figure out how to decode
    162 		 * correctable errors and set ME_ECC_CE_ENA
    163 		 * here so we can log them...
    164 		 */
    165 		mer->me_csr = ME_CSR_IENA; /* | ME_ECC_CE_ENA */
    166 	}
    167 }
    168 
    169 /*****************************************************************
    170  * Functions for ECC memory
    171  *****************************************************************/
    172 
    173 static int
    174 memerr_interrupt(arg)
    175 	void *arg;
    176 {
    177 	struct memerr_softc *sc = arg;
    178 	volatile struct memerr *me = sc->sc_reg;
    179 	u_char csr, ctx;
    180 	u_int pa, va;
    181 	int pte;
    182 	char bits[64];
    183 
    184 	csr = me->me_csr;
    185 	if ((csr & ME_CSR_IPEND) == 0)
    186 		return (0);
    187 
    188 	va = me->me_vaddr;
    189  	ctx = (va >> 28) & 0xF;
    190 	va &= 0x0FFFffff;
    191 	pte = get_pte(va);
    192 	pa = PG_PA(pte);
    193 
    194 	printf("\nMemory error on %s cycle!\n",
    195 		(ctx & 8) ? "DVMA" : "CPU");
    196 	printf(" ctx=%d, vaddr=0x%x, paddr=0x%x\n",
    197 		   (ctx & 7), va, pa);
    198 	printf(" csr=%s\n", bitmask_snprintf(csr, sc->sc_csrbits,
    199 	    bits, sizeof(bits)));
    200 
    201 	/*
    202 	 * If we have parity-checked memory, there is
    203 	 * not much to be done.  Any error is fatal.
    204 	 */
    205 	if (sc->sc_type == ME_PAR) {
    206 		if (csr & ME_PAR_EMASK) {
    207 			/* Parity errors are fatal. */
    208 			goto die;
    209 		}
    210 		/* The IPEND bit was set, but no error bits. */
    211 		goto noerror;
    212 	}
    213 
    214 	/*
    215 	 * We have ECC memory.  More complicated...
    216 	 */
    217 	if (csr & (ME_ECC_WBTMO | ME_ECC_WBERR)) {
    218 		printf(" write-back failed, pte=0x%x\n", pte);
    219 		goto die;
    220 	}
    221 	if (csr & ME_ECC_UE) {
    222 		printf(" uncorrectable ECC error\n");
    223 		goto die;
    224 	}
    225 	if (csr & ME_ECC_CE) {
    226 		/* Just log this and continue. */
    227 		memerr_correctable(sc);
    228 		goto recover;
    229 	}
    230 	/* The IPEND bit was set, but no error bits. */
    231 	goto noerror;
    232 
    233 die:
    234 	panic("all bets are off...");
    235 
    236 noerror:
    237 	printf("memerr: no error bits set?\n");
    238 
    239 recover:
    240 	/* Clear the error by writing the address register. */
    241 	me->me_vaddr = 0;
    242 	return (1);
    243 }
    244 
    245 /*
    246  * Announce (and log) a correctable ECC error.
    247  * Need to look at the ECC syndrome register on
    248  * the memory board that caused the error...
    249  */
    250 void
    251 memerr_correctable(sc)
    252 	struct memerr_softc *sc;
    253 {
    254 	/* XXX: Not yet... */
    255 }
    256