1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #pragma ident "%Z%%M% %I% %E% SMI" 28 29 #include <cma.h> 30 31 #include <strings.h> 32 #include <errno.h> 33 #include <time.h> 34 #include <fm/fmd_api.h> 35 #include <sys/fm/protocol.h> 36 37 cma_t cma; 38 39 cma_stats_t cma_stats = { 40 { "cpu_flts", FMD_TYPE_UINT64, "cpu faults resolved" }, 41 { "cpu_fails", FMD_TYPE_UINT64, "cpu faults unresolveable" }, 42 { "cpu_blfails", FMD_TYPE_UINT64, "failed cpu blacklists" }, 43 { "cpu_supp", FMD_TYPE_UINT64, "cpu offlines suppressed" }, 44 { "cpu_blsupp", FMD_TYPE_UINT64, "cpu blacklists suppressed" }, 45 { "page_flts", FMD_TYPE_UINT64, "page faults resolved" }, 46 { "page_fails", FMD_TYPE_UINT64, "page faults unresolveable" }, 47 { "page_supp", FMD_TYPE_UINT64, "page retires suppressed" }, 48 { "page_nonent", FMD_TYPE_UINT64, "retires for non-existent fmris" }, 49 { "page_retmax", FMD_TYPE_UINT64, "hit max retries for page retire" }, 50 { "bad_flts", FMD_TYPE_UINT64, "invalid fault events received" }, 51 { "nop_flts", FMD_TYPE_UINT64, "inapplicable fault events received" }, 52 { "auto_flts", FMD_TYPE_UINT64, "auto-close faults received" } 53 }; 54 55 typedef struct cma_subscriber { 56 const char *subr_class; 57 const char *subr_sname; 58 uint_t subr_svers; 59 void (*subr_func)(fmd_hdl_t *, nvlist_t *, nvlist_t *, const char *); 60 } cma_subscriber_t; 61 62 static const cma_subscriber_t cma_subrs[] = { 63 { "fault.memory.page", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 64 cma_page_retire }, 65 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 66 NULL }, 67 { "fault.memory.dimm_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 68 NULL }, 69 { "fault.memory.dimm_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 70 NULL }, 71 { "fault.memory.dimm_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 72 NULL }, 73 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 74 NULL }, 75 { "fault.memory.datapath", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 76 NULL }, 77 78 /* 79 * The following ultraSPARC-T1 faults do NOT retire a cpu thread, 80 * and therefore must be intercepted before 81 * the default "fault.cpu.*" dispatch to cma_cpu_retire. 82 */ 83 { "fault.cpu.ultraSPARC-T1.freg", FM_FMRI_SCHEME_CPU, 84 FM_CPU_SCHEME_VERSION, NULL }, 85 { "fault.cpu.ultraSPARC-T1.l2cachedata", FM_FMRI_SCHEME_CPU, 86 FM_CPU_SCHEME_VERSION, NULL }, 87 { "fault.cpu.ultraSPARC-T1.l2cachetag", FM_FMRI_SCHEME_CPU, 88 FM_CPU_SCHEME_VERSION, NULL }, 89 { "fault.cpu.ultraSPARC-T1.l2cachectl", FM_FMRI_SCHEME_CPU, 90 FM_CPU_SCHEME_VERSION, NULL }, 91 { "fault.cpu.ultraSPARC-T1.mau", FM_FMRI_SCHEME_CPU, 92 FM_CPU_SCHEME_VERSION, NULL }, 93 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 94 cma_cpu_retire }, 95 { NULL, NULL, 0, NULL } 96 }; 97 98 static const cma_subscriber_t * 99 nvl2subr(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t **asrup) 100 { 101 const cma_subscriber_t *sp; 102 nvlist_t *asru; 103 char *scheme; 104 uint8_t version; 105 106 if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) != 0 || 107 nvlist_lookup_string(asru, FM_FMRI_SCHEME, &scheme) != 0 || 108 nvlist_lookup_uint8(asru, FM_VERSION, &version) != 0) { 109 cma_stats.bad_flts.fmds_value.ui64++; 110 return (NULL); 111 } 112 113 for (sp = cma_subrs; sp->subr_class != NULL; sp++) { 114 if (fmd_nvl_class_match(hdl, nvl, sp->subr_class) && 115 strcmp(scheme, sp->subr_sname) == 0 && 116 version <= sp->subr_svers) { 117 *asrup = asru; 118 return (sp); 119 } 120 } 121 122 cma_stats.nop_flts.fmds_value.ui64++; 123 return (NULL); 124 } 125 126 static void 127 cma_recv_list(fmd_hdl_t *hdl, nvlist_t *nvl) 128 { 129 char *uuid = NULL; 130 nvlist_t **nva; 131 uint_t nvc = 0; 132 int err = 0; 133 134 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid); 135 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, 136 &nva, &nvc); 137 if (err != 0) { 138 cma_stats.bad_flts.fmds_value.ui64++; 139 return; 140 } 141 142 while (nvc-- != 0 && !fmd_case_uuclosed(hdl, uuid)) { 143 nvlist_t *nvl = *nva++; 144 const cma_subscriber_t *subr; 145 nvlist_t *asru; 146 147 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 148 continue; 149 150 if (subr->subr_func != NULL) 151 subr->subr_func(hdl, nvl, asru, uuid); 152 } 153 } 154 155 static void 156 cma_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl) 157 { 158 const cma_subscriber_t *subr; 159 nvlist_t *asru; 160 161 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 162 return; 163 164 if (subr->subr_func != NULL) 165 subr->subr_func(hdl, nvl, asru, NULL); 166 } 167 168 /*ARGSUSED*/ 169 static void 170 cma_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 171 { 172 fmd_hdl_debug(hdl, "received %s\n", class); 173 174 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0) 175 cma_recv_list(hdl, nvl); 176 else 177 cma_recv_one(hdl, nvl); 178 } 179 180 /*ARGSUSED*/ 181 static void 182 cma_timeout(fmd_hdl_t *hdl, id_t id, void *arg) 183 { 184 if (id == cma.cma_page_timerid) 185 cma_page_retry(hdl); 186 } 187 188 static const fmd_hdl_ops_t fmd_ops = { 189 cma_recv, /* fmdo_recv */ 190 cma_timeout, /* fmdo_timeout */ 191 NULL, /* fmdo_close */ 192 NULL, /* fmdo_stats */ 193 NULL, /* fmdo_gc */ 194 }; 195 196 static const fmd_prop_t fmd_props[] = { 197 { "cpu_tries", FMD_TYPE_UINT32, "10" }, 198 { "cpu_delay", FMD_TYPE_TIME, "1sec" }, 199 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" }, 200 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" }, 201 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "true" }, 202 { "page_ret_mindelay", FMD_TYPE_TIME, "1sec" }, 203 { "page_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 204 { "page_retire_enable", FMD_TYPE_BOOL, "true" }, 205 #ifdef i386 206 /* 207 * On i386, leaving cases open while we retry the 208 * retire can cause the eft module to use large amounts 209 * of memory. Until eft is fixed, we set a maximum number 210 * of retries on page retires, after which the case will 211 * be closed. 212 */ 213 { "page_retire_maxretries", FMD_TYPE_UINT32, "8" }, 214 #else 215 { "page_retire_maxretries", FMD_TYPE_UINT32, "0" }, 216 #endif /* i386 */ 217 { NULL, 0, NULL } 218 }; 219 220 static const fmd_hdl_info_t fmd_info = { 221 "CPU/Memory Retire Agent", CMA_VERSION, &fmd_ops, fmd_props 222 }; 223 224 void 225 _fmd_init(fmd_hdl_t *hdl) 226 { 227 hrtime_t nsec; 228 229 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) 230 return; /* invalid data in configuration file */ 231 232 fmd_hdl_subscribe(hdl, "fault.cpu.*"); 233 fmd_hdl_subscribe(hdl, "fault.memory.*"); 234 235 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cma_stats) / 236 sizeof (fmd_stat_t), (fmd_stat_t *)&cma_stats); 237 238 cma.cma_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries"); 239 240 nsec = fmd_prop_get_int64(hdl, "cpu_delay"); 241 cma.cma_cpu_delay.tv_sec = nsec / NANOSEC; 242 cma.cma_cpu_delay.tv_nsec = nsec % NANOSEC; 243 244 cma.cma_page_mindelay = fmd_prop_get_int64(hdl, "page_ret_mindelay"); 245 cma.cma_page_maxdelay = fmd_prop_get_int64(hdl, "page_ret_maxdelay"); 246 247 cma.cma_cpu_dooffline = fmd_prop_get_int32(hdl, "cpu_offline_enable"); 248 cma.cma_cpu_forcedoffline = fmd_prop_get_int32(hdl, 249 "cpu_forced_offline"); 250 cma.cma_cpu_doblacklist = fmd_prop_get_int32(hdl, 251 "cpu_blacklist_enable"); 252 cma.cma_page_doretire = fmd_prop_get_int32(hdl, "page_retire_enable"); 253 cma.cma_page_maxretries = 254 fmd_prop_get_int32(hdl, "page_retire_maxretries"); 255 256 if (cma.cma_page_maxdelay < cma.cma_page_mindelay) 257 fmd_hdl_abort(hdl, "page retirement delays conflict\n"); 258 } 259 260 void 261 _fmd_fini(fmd_hdl_t *hdl) 262 { 263 cma_page_fini(hdl); 264 } 265