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 int (*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 uint_t keepopen; 133 int err = 0; 134 135 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid); 136 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, 137 &nva, &nvc); 138 if (err != 0) { 139 cma_stats.bad_flts.fmds_value.ui64++; 140 return; 141 } 142 143 keepopen = nvc; 144 while (nvc-- != 0 && !fmd_case_uuclosed(hdl, uuid)) { 145 nvlist_t *nvl = *nva++; 146 const cma_subscriber_t *subr; 147 nvlist_t *asru; 148 149 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 150 continue; 151 152 /* 153 * A handler returns CMA_RA_SUCCESS to indicate that 154 * from this suspects point-of-view the case may be 155 * closed, CMA_RA_FAILURE otherwise. 156 * A handler must not close the case itself. 157 */ 158 if (subr->subr_func != NULL) { 159 err = subr->subr_func(hdl, nvl, asru, uuid); 160 161 if (err == CMA_RA_SUCCESS) 162 keepopen--; 163 } 164 } 165 166 if (!keepopen) 167 fmd_case_uuclose(hdl, uuid); 168 } 169 170 static void 171 cma_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl) 172 { 173 const cma_subscriber_t *subr; 174 nvlist_t *asru; 175 176 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 177 return; 178 179 if (subr->subr_func != NULL) 180 (void) subr->subr_func(hdl, nvl, asru, NULL); 181 182 } 183 184 /*ARGSUSED*/ 185 static void 186 cma_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 187 { 188 fmd_hdl_debug(hdl, "received %s\n", class); 189 190 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0) 191 cma_recv_list(hdl, nvl); 192 else 193 cma_recv_one(hdl, nvl); 194 } 195 196 /*ARGSUSED*/ 197 static void 198 cma_timeout(fmd_hdl_t *hdl, id_t id, void *arg) 199 { 200 if (id == cma.cma_page_timerid) 201 cma_page_retry(hdl); 202 } 203 204 static const fmd_hdl_ops_t fmd_ops = { 205 cma_recv, /* fmdo_recv */ 206 cma_timeout, /* fmdo_timeout */ 207 NULL, /* fmdo_close */ 208 NULL, /* fmdo_stats */ 209 NULL, /* fmdo_gc */ 210 }; 211 212 static const fmd_prop_t fmd_props[] = { 213 { "cpu_tries", FMD_TYPE_UINT32, "10" }, 214 { "cpu_delay", FMD_TYPE_TIME, "1sec" }, 215 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" }, 216 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" }, 217 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "true" }, 218 { "page_ret_mindelay", FMD_TYPE_TIME, "1sec" }, 219 { "page_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 220 { "page_retire_enable", FMD_TYPE_BOOL, "true" }, 221 #ifdef i386 222 /* 223 * On i386, leaving cases open while we retry the 224 * retire can cause the eft module to use large amounts 225 * of memory. Until eft is fixed, we set a maximum number 226 * of retries on page retires, after which the case will 227 * be closed. 228 */ 229 { "page_retire_maxretries", FMD_TYPE_UINT32, "8" }, 230 #else 231 { "page_retire_maxretries", FMD_TYPE_UINT32, "0" }, 232 #endif /* i386 */ 233 { NULL, 0, NULL } 234 }; 235 236 static const fmd_hdl_info_t fmd_info = { 237 "CPU/Memory Retire Agent", CMA_VERSION, &fmd_ops, fmd_props 238 }; 239 240 void 241 _fmd_init(fmd_hdl_t *hdl) 242 { 243 hrtime_t nsec; 244 245 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) 246 return; /* invalid data in configuration file */ 247 248 fmd_hdl_subscribe(hdl, "fault.cpu.*"); 249 fmd_hdl_subscribe(hdl, "fault.memory.*"); 250 251 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cma_stats) / 252 sizeof (fmd_stat_t), (fmd_stat_t *)&cma_stats); 253 254 cma.cma_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries"); 255 256 nsec = fmd_prop_get_int64(hdl, "cpu_delay"); 257 cma.cma_cpu_delay.tv_sec = nsec / NANOSEC; 258 cma.cma_cpu_delay.tv_nsec = nsec % NANOSEC; 259 260 cma.cma_page_mindelay = fmd_prop_get_int64(hdl, "page_ret_mindelay"); 261 cma.cma_page_maxdelay = fmd_prop_get_int64(hdl, "page_ret_maxdelay"); 262 263 cma.cma_cpu_dooffline = fmd_prop_get_int32(hdl, "cpu_offline_enable"); 264 cma.cma_cpu_forcedoffline = fmd_prop_get_int32(hdl, 265 "cpu_forced_offline"); 266 cma.cma_cpu_doblacklist = fmd_prop_get_int32(hdl, 267 "cpu_blacklist_enable"); 268 cma.cma_page_doretire = fmd_prop_get_int32(hdl, "page_retire_enable"); 269 cma.cma_page_maxretries = 270 fmd_prop_get_int32(hdl, "page_retire_maxretries"); 271 272 if (cma.cma_page_maxdelay < cma.cma_page_mindelay) 273 fmd_hdl_abort(hdl, "page retirement delays conflict\n"); 274 } 275 276 void 277 _fmd_fini(fmd_hdl_t *hdl) 278 { 279 cma_page_fini(hdl); 280 } 281