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.page_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 66 cma_page_retire }, 67 { "fault.memory.page_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 68 cma_page_retire }, 69 { "fault.memory.page_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 70 cma_page_retire }, 71 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 72 NULL }, 73 { "fault.memory.dimm_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 74 NULL }, 75 { "fault.memory.dimm_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 76 NULL }, 77 { "fault.memory.dimm_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 78 NULL }, 79 { "fault.memory.dimm_testfail", FM_FMRI_SCHEME_MEM, 80 FM_MEM_SCHEME_VERSION, NULL }, 81 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 82 NULL }, 83 { "fault.memory.datapath", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 84 NULL }, 85 86 /* 87 * The following faults do NOT retire a cpu thread, 88 * and therefore must be intercepted before 89 * the default "fault.cpu.*" dispatch to cma_cpu_retire. 90 */ 91 { "fault.cpu.ultraSPARC-T1.freg", FM_FMRI_SCHEME_CPU, 92 FM_CPU_SCHEME_VERSION, NULL }, 93 { "fault.cpu.ultraSPARC-T1.l2cachedata", FM_FMRI_SCHEME_CPU, 94 FM_CPU_SCHEME_VERSION, NULL }, 95 { "fault.cpu.ultraSPARC-T1.l2cachetag", FM_FMRI_SCHEME_CPU, 96 FM_CPU_SCHEME_VERSION, NULL }, 97 { "fault.cpu.ultraSPARC-T1.l2cachectl", FM_FMRI_SCHEME_CPU, 98 FM_CPU_SCHEME_VERSION, NULL }, 99 { "fault.cpu.ultraSPARC-T1.mau", FM_FMRI_SCHEME_CPU, 100 FM_CPU_SCHEME_VERSION, NULL }, 101 { "fault.cpu.amd.dramchannel", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 102 NULL }, 103 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 104 cma_cpu_retire }, 105 { NULL, NULL, 0, NULL } 106 }; 107 108 static const cma_subscriber_t * 109 nvl2subr(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t **asrup) 110 { 111 const cma_subscriber_t *sp; 112 nvlist_t *asru; 113 char *scheme; 114 uint8_t version; 115 char *fltclass = "(unknown)"; 116 117 if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) != 0 || 118 nvlist_lookup_string(asru, FM_FMRI_SCHEME, &scheme) != 0 || 119 nvlist_lookup_uint8(asru, FM_VERSION, &version) != 0) { 120 cma_stats.bad_flts.fmds_value.ui64++; 121 return (NULL); 122 } 123 124 for (sp = cma_subrs; sp->subr_class != NULL; sp++) { 125 if (fmd_nvl_class_match(hdl, nvl, sp->subr_class) && 126 strcmp(scheme, sp->subr_sname) == 0 && 127 version <= sp->subr_svers) { 128 *asrup = asru; 129 return (sp); 130 } 131 } 132 133 (void) nvlist_lookup_string(nvl, FM_CLASS, &fltclass); 134 fmd_hdl_error(hdl, "No handling disposition for %s with asru in " 135 "scheme \"%s\"\n", fltclass, scheme); 136 cma_stats.nop_flts.fmds_value.ui64++; 137 return (NULL); 138 } 139 140 static void 141 cma_recv_list(fmd_hdl_t *hdl, nvlist_t *nvl) 142 { 143 char *uuid = NULL; 144 nvlist_t **nva; 145 uint_t nvc = 0; 146 uint_t keepopen; 147 int err = 0; 148 149 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid); 150 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, 151 &nva, &nvc); 152 if (err != 0) { 153 cma_stats.bad_flts.fmds_value.ui64++; 154 return; 155 } 156 157 keepopen = nvc; 158 while (nvc-- != 0 && !fmd_case_uuclosed(hdl, uuid)) { 159 nvlist_t *nvl = *nva++; 160 const cma_subscriber_t *subr; 161 nvlist_t *asru; 162 163 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 164 continue; 165 166 /* 167 * A handler returns CMA_RA_SUCCESS to indicate that 168 * from this suspects point-of-view the case may be 169 * closed, CMA_RA_FAILURE otherwise. 170 * A handler must not close the case itself. 171 */ 172 if (subr->subr_func != NULL) { 173 err = subr->subr_func(hdl, nvl, asru, uuid); 174 175 if (err == CMA_RA_SUCCESS) 176 keepopen--; 177 } 178 } 179 180 if (!keepopen) 181 fmd_case_uuclose(hdl, uuid); 182 } 183 184 static void 185 cma_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl) 186 { 187 const cma_subscriber_t *subr; 188 nvlist_t *asru; 189 190 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 191 return; 192 193 if (subr->subr_func != NULL) 194 (void) subr->subr_func(hdl, nvl, asru, NULL); 195 196 } 197 198 /*ARGSUSED*/ 199 static void 200 cma_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 201 { 202 fmd_hdl_debug(hdl, "received %s\n", class); 203 204 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0) 205 cma_recv_list(hdl, nvl); 206 else 207 cma_recv_one(hdl, nvl); 208 } 209 210 /*ARGSUSED*/ 211 static void 212 cma_timeout(fmd_hdl_t *hdl, id_t id, void *arg) 213 { 214 if (id == cma.cma_page_timerid) 215 cma_page_retry(hdl); 216 } 217 218 static const fmd_hdl_ops_t fmd_ops = { 219 cma_recv, /* fmdo_recv */ 220 cma_timeout, /* fmdo_timeout */ 221 NULL, /* fmdo_close */ 222 NULL, /* fmdo_stats */ 223 NULL, /* fmdo_gc */ 224 }; 225 226 static const fmd_prop_t fmd_props[] = { 227 { "cpu_tries", FMD_TYPE_UINT32, "10" }, 228 { "cpu_delay", FMD_TYPE_TIME, "1sec" }, 229 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" }, 230 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" }, 231 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "true" }, 232 { "page_ret_mindelay", FMD_TYPE_TIME, "1sec" }, 233 { "page_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 234 { "page_retire_enable", FMD_TYPE_BOOL, "true" }, 235 #ifdef i386 236 /* 237 * On i386, leaving cases open while we retry the 238 * retire can cause the eft module to use large amounts 239 * of memory. Until eft is fixed, we set a maximum number 240 * of retries on page retires, after which the case will 241 * be closed. 242 */ 243 { "page_retire_maxretries", FMD_TYPE_UINT32, "5" }, 244 #else 245 { "page_retire_maxretries", FMD_TYPE_UINT32, "0" }, 246 #endif /* i386 */ 247 { NULL, 0, NULL } 248 }; 249 250 static const fmd_hdl_info_t fmd_info = { 251 "CPU/Memory Retire Agent", CMA_VERSION, &fmd_ops, fmd_props 252 }; 253 254 void 255 _fmd_init(fmd_hdl_t *hdl) 256 { 257 hrtime_t nsec; 258 259 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) 260 return; /* invalid data in configuration file */ 261 262 fmd_hdl_subscribe(hdl, "fault.cpu.*"); 263 fmd_hdl_subscribe(hdl, "fault.memory.*"); 264 265 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cma_stats) / 266 sizeof (fmd_stat_t), (fmd_stat_t *)&cma_stats); 267 268 cma.cma_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries"); 269 270 nsec = fmd_prop_get_int64(hdl, "cpu_delay"); 271 cma.cma_cpu_delay.tv_sec = nsec / NANOSEC; 272 cma.cma_cpu_delay.tv_nsec = nsec % NANOSEC; 273 274 cma.cma_page_mindelay = fmd_prop_get_int64(hdl, "page_ret_mindelay"); 275 cma.cma_page_maxdelay = fmd_prop_get_int64(hdl, "page_ret_maxdelay"); 276 277 cma.cma_cpu_dooffline = fmd_prop_get_int32(hdl, "cpu_offline_enable"); 278 cma.cma_cpu_forcedoffline = fmd_prop_get_int32(hdl, 279 "cpu_forced_offline"); 280 cma.cma_cpu_doblacklist = fmd_prop_get_int32(hdl, 281 "cpu_blacklist_enable"); 282 cma.cma_page_doretire = fmd_prop_get_int32(hdl, "page_retire_enable"); 283 cma.cma_page_maxretries = 284 fmd_prop_get_int32(hdl, "page_retire_maxretries"); 285 286 if (cma.cma_page_maxdelay < cma.cma_page_mindelay) 287 fmd_hdl_abort(hdl, "page retirement delays conflict\n"); 288 } 289 290 void 291 _fmd_fini(fmd_hdl_t *hdl) 292 { 293 cma_page_fini(hdl); 294 } 295