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 2009 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #include <cma.h> 28 29 #include <unistd.h> 30 #include <fcntl.h> 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 #include <sys/systeminfo.h> 37 #include <sys/utsname.h> 38 39 #ifdef sun4v 40 #include <sys/fm/ldom.h> 41 42 static fmd_hdl_t *init_hdl; 43 ldom_hdl_t *cma_lhp; 44 #endif 45 46 #ifdef i386 47 boolean_t cma_is_native; 48 #endif 49 50 extern const char *fmd_fmri_get_platform(); 51 52 cma_t cma; 53 54 cma_stats_t cma_stats = { 55 { "cpu_flts", FMD_TYPE_UINT64, "cpu faults resolved" }, 56 { "cpu_repairs", FMD_TYPE_UINT64, "cpu faults repaired" }, 57 { "cpu_fails", FMD_TYPE_UINT64, "cpu faults unresolveable" }, 58 { "cpu_blfails", FMD_TYPE_UINT64, "failed cpu blacklists" }, 59 { "cpu_supp", FMD_TYPE_UINT64, "cpu offlines suppressed" }, 60 { "cpu_blsupp", FMD_TYPE_UINT64, "cpu blacklists suppressed" }, 61 { "page_flts", FMD_TYPE_UINT64, "page faults resolved" }, 62 { "page_repairs", FMD_TYPE_UINT64, "page faults repaired" }, 63 { "page_fails", FMD_TYPE_UINT64, "page faults unresolveable" }, 64 { "page_supp", FMD_TYPE_UINT64, "page retires suppressed" }, 65 { "page_nonent", FMD_TYPE_UINT64, "retires for non-existent fmris" }, 66 { "bad_flts", FMD_TYPE_UINT64, "invalid fault events received" }, 67 { "nop_flts", FMD_TYPE_UINT64, "inapplicable fault events received" }, 68 { "auto_flts", FMD_TYPE_UINT64, "auto-close faults received" } 69 }; 70 71 typedef struct cma_subscriber { 72 const char *subr_class; 73 const char *subr_sname; 74 uint_t subr_svers; 75 int (*subr_func)(fmd_hdl_t *, nvlist_t *, nvlist_t *, const char *, 76 boolean_t); 77 } cma_subscriber_t; 78 79 static const cma_subscriber_t cma_subrs[] = { 80 #if defined(i386) 81 /* 82 * On x86, the ASRUs are expected to be in hc scheme. When 83 * cpumem-retire wants to retire a cpu or mem page, it calls the 84 * methods registered in the topo node to do that. The topo 85 * enumerator, which necessarily knows all the config info that 86 * we'd ever need in deciding what/how to retire etc. This takes 87 * away much of that complexity from the agent into the entity 88 * that knows all config/topo information. 89 */ 90 { "fault.memory.page", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 91 cma_page_retire }, 92 { "fault.memory.page_sb", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 93 cma_page_retire }, 94 { "fault.memory.page_ck", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 95 cma_page_retire }, 96 { "fault.memory.page_ue", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 97 cma_page_retire }, 98 { "fault.memory.generic-x86.page_ce", FM_FMRI_SCHEME_HC, 99 FM_HC_SCHEME_VERSION, cma_page_retire }, 100 { "fault.memory.generic-x86.page_ue", FM_FMRI_SCHEME_HC, 101 FM_HC_SCHEME_VERSION, cma_page_retire }, 102 { "fault.memory.intel.page_ce", FM_FMRI_SCHEME_HC, 103 FM_HC_SCHEME_VERSION, cma_page_retire }, 104 { "fault.memory.intel.page_ue", FM_FMRI_SCHEME_HC, 105 FM_HC_SCHEME_VERSION, cma_page_retire }, 106 { "fault.memory.dimm", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 107 NULL }, 108 { "fault.memory.dimm_sb", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 109 NULL }, 110 { "fault.memory.dimm_ck", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 111 NULL }, 112 { "fault.memory.dimm_ue", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 113 NULL }, 114 { "fault.memory.generic-x86.dimm_ce", FM_FMRI_SCHEME_HC, 115 FM_HC_SCHEME_VERSION, NULL }, 116 { "fault.memory.generic-x86.dimm_ue", FM_FMRI_SCHEME_HC, 117 FM_HC_SCHEME_VERSION, NULL }, 118 { "fault.memory.intel.dimm_ce", FM_FMRI_SCHEME_HC, 119 FM_HC_SCHEME_VERSION, NULL }, 120 { "fault.memory.intel.dimm_ue", FM_FMRI_SCHEME_HC, 121 FM_HC_SCHEME_VERSION, NULL }, 122 { "fault.memory.intel.fbd.*", FM_FMRI_SCHEME_HC, 123 FM_HC_SCHEME_VERSION, NULL }, 124 { "fault.memory.dimm_testfail", FM_FMRI_SCHEME_HC, 125 FM_HC_SCHEME_VERSION, NULL }, 126 { "fault.memory.bank", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 127 NULL }, 128 { "fault.memory.datapath", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 129 NULL }, 130 { "fault.cpu.intel.quickpath.*", FM_FMRI_SCHEME_HC, 131 FM_HC_SCHEME_VERSION, NULL }, 132 { "fault.cpu.generic-x86.mc", FM_FMRI_SCHEME_HC, 133 FM_HC_SCHEME_VERSION, NULL }, 134 { "fault.cpu.intel.dma", FM_FMRI_SCHEME_HC, 135 FM_HC_SCHEME_VERSION, NULL }, 136 { "fault.cpu.intel.dma", FM_FMRI_SCHEME_CPU, 137 FM_CPU_SCHEME_VERSION, NULL }, 138 139 /* 140 * The ASRU for cpu faults are in cpu scheme on native and in hc 141 * scheme on xpv. So each cpu fault class needs to be listed twice. 142 */ 143 144 /* 145 * The following faults do NOT retire a cpu thread, 146 * and therefore must be intercepted before 147 * the default "fault.cpu.*" dispatch to cma_cpu_hc_retire. 148 */ 149 { "fault.cpu.amd.dramchannel", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 150 NULL }, 151 { "fault.cpu.amd.dramchannel", FM_FMRI_SCHEME_CPU, 152 FM_CPU_SCHEME_VERSION, NULL }, 153 { "fault.cpu.generic-x86.bus_interconnect_memory", FM_FMRI_SCHEME_HC, 154 FM_HC_SCHEME_VERSION, NULL }, 155 { "fault.cpu.generic-x86.bus_interconnect_memory", FM_FMRI_SCHEME_CPU, 156 FM_CPU_SCHEME_VERSION, NULL }, 157 { "fault.cpu.generic-x86.bus_interconnect_io", FM_FMRI_SCHEME_HC, 158 FM_HC_SCHEME_VERSION, NULL }, 159 { "fault.cpu.generic-x86.bus_interconnect_io", FM_FMRI_SCHEME_CPU, 160 FM_CPU_SCHEME_VERSION, NULL }, 161 { "fault.cpu.generic-x86.bus_interconnect", FM_FMRI_SCHEME_HC, 162 FM_HC_SCHEME_VERSION, NULL }, 163 { "fault.cpu.generic-x86.bus_interconnect", FM_FMRI_SCHEME_CPU, 164 FM_CPU_SCHEME_VERSION, NULL }, 165 { "fault.cpu.intel.bus_interconnect_memory", FM_FMRI_SCHEME_HC, 166 FM_HC_SCHEME_VERSION, NULL }, 167 { "fault.cpu.intel.bus_interconnect_memory", FM_FMRI_SCHEME_CPU, 168 FM_CPU_SCHEME_VERSION, NULL }, 169 { "fault.cpu.intel.bus_interconnect_io", FM_FMRI_SCHEME_HC, 170 FM_HC_SCHEME_VERSION, NULL }, 171 { "fault.cpu.intel.bus_interconnect_io", FM_FMRI_SCHEME_CPU, 172 FM_CPU_SCHEME_VERSION, NULL }, 173 { "fault.cpu.intel.bus_interconnect", FM_FMRI_SCHEME_HC, 174 FM_HC_SCHEME_VERSION, NULL }, 175 { "fault.cpu.intel.bus_interconnect", FM_FMRI_SCHEME_CPU, 176 FM_CPU_SCHEME_VERSION, NULL }, 177 { "fault.cpu.intel.nb.*", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 178 NULL }, 179 { "fault.cpu.intel.nb.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 180 NULL }, 181 { "fault.cpu.intel.dma", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 182 NULL }, 183 { "fault.cpu.intel.dma", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 184 NULL }, 185 { "fault.cpu.*", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 186 cma_cpu_hc_retire }, 187 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 188 cma_cpu_hc_retire }, 189 #elif defined(sun4v) 190 /* 191 * The following are PI sun4v faults 192 */ 193 { "fault.memory.memlink", FM_FMRI_SCHEME_HC, 194 FM_HC_SCHEME_VERSION, NULL }, 195 { "fault.memory.memlink-uc", FM_FMRI_SCHEME_HC, 196 FM_HC_SCHEME_VERSION, NULL }, 197 { "fault.memory.memlink-failover", FM_FMRI_SCHEME_HC, 198 FM_HC_SCHEME_VERSION, NULL }, 199 { "fault.memory.dimm-ue-imminent", FM_FMRI_SCHEME_HC, 200 FM_HC_SCHEME_VERSION, NULL }, 201 { "fault.memory.dram-ue-imminent", FM_FMRI_SCHEME_HC, 202 FM_HC_SCHEME_VERSION, NULL }, 203 { "fault.memory.dimm-page-retires-excessive", FM_FMRI_SCHEME_HC, 204 FM_HC_SCHEME_VERSION, NULL }, 205 { "fault.memory.page", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 206 cma_page_retire }, 207 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 208 NULL }, 209 { "fault.memory.dimm_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 210 NULL }, 211 { "fault.memory.dimm_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 212 NULL }, 213 { "fault.memory.dimm_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 214 NULL }, 215 { "fault.memory.dimm-page-retires-excessive", FM_FMRI_SCHEME_MEM, 216 FM_MEM_SCHEME_VERSION, NULL }, 217 { "fault.memory.dimm-ue-imminent", FM_FMRI_SCHEME_MEM, 218 FM_MEM_SCHEME_VERSION, NULL }, 219 { "fault.memory.dram-ue-imminent", FM_FMRI_SCHEME_MEM, 220 FM_MEM_SCHEME_VERSION, NULL }, 221 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 222 NULL }, 223 { "fault.memory.datapath", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 224 NULL }, 225 { "fault.memory.link-c", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 226 NULL }, 227 { "fault.memory.link-u", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 228 NULL }, 229 { "fault.memory.link-f", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 230 NULL }, 231 { "fault.memory.link-c", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 232 NULL }, 233 { "fault.memory.link-u", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 234 NULL }, 235 { "fault.memory.link-f", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 236 NULL }, 237 238 /* 239 * The following ultraSPARC-T1/T2 faults do NOT retire a cpu thread, 240 * and therefore must be intercepted before 241 * the default "fault.cpu.*" dispatch to cma_cpu_hc_retire. 242 */ 243 { "fault.cpu.*.l2cachedata", FM_FMRI_SCHEME_CPU, 244 FM_CPU_SCHEME_VERSION, NULL }, 245 { "fault.cpu.*.l2cachetag", FM_FMRI_SCHEME_CPU, 246 FM_CPU_SCHEME_VERSION, NULL }, 247 { "fault.cpu.*.l2cachectl", FM_FMRI_SCHEME_CPU, 248 FM_CPU_SCHEME_VERSION, NULL }, 249 { "fault.cpu.*.l2data-c", FM_FMRI_SCHEME_CPU, 250 FM_CPU_SCHEME_VERSION, NULL }, 251 { "fault.cpu.*.l2data-u", FM_FMRI_SCHEME_CPU, 252 FM_CPU_SCHEME_VERSION, NULL }, 253 { "fault.cpu.*.mau", FM_FMRI_SCHEME_CPU, 254 FM_CPU_SCHEME_VERSION, NULL }, 255 { "fault.cpu.*.lfu-u", FM_FMRI_SCHEME_CPU, 256 FM_CPU_SCHEME_VERSION, NULL }, 257 { "fault.cpu.*.lfu-f", FM_FMRI_SCHEME_CPU, 258 FM_CPU_SCHEME_VERSION, NULL }, 259 { "fault.cpu.*.lfu-p", FM_FMRI_SCHEME_CPU, 260 FM_CPU_SCHEME_VERSION, NULL }, 261 { "fault.cpu.ultraSPARC-T1.freg", FM_FMRI_SCHEME_CPU, 262 FM_CPU_SCHEME_VERSION, NULL }, 263 { "fault.cpu.ultraSPARC-T1.l2cachedata", FM_FMRI_SCHEME_CPU, 264 FM_CPU_SCHEME_VERSION, NULL }, 265 { "fault.cpu.ultraSPARC-T1.l2cachetag", FM_FMRI_SCHEME_CPU, 266 FM_CPU_SCHEME_VERSION, NULL }, 267 { "fault.cpu.ultraSPARC-T1.l2cachectl", FM_FMRI_SCHEME_CPU, 268 FM_CPU_SCHEME_VERSION, NULL }, 269 { "fault.cpu.ultraSPARC-T1.mau", FM_FMRI_SCHEME_CPU, 270 FM_CPU_SCHEME_VERSION, NULL }, 271 { "fault.cpu.ultraSPARC-T2plus.chip", FM_FMRI_SCHEME_HC, 272 FM_HC_SCHEME_VERSION, NULL }, 273 { "fault.cpu.*", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 274 cma_cpu_hc_retire }, 275 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 276 cma_cpu_hc_retire }, 277 #elif defined(opl) 278 { "fault.memory.page", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 279 cma_page_retire }, 280 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 281 NULL }, 282 { "fault.memory.dimm-page-retires-excessive", FM_FMRI_SCHEME_MEM, 283 FM_MEM_SCHEME_VERSION, NULL }, 284 { "fault.memory.dimm-ue-imminent", FM_FMRI_SCHEME_MEM, 285 FM_MEM_SCHEME_VERSION, NULL }, 286 { "fault.memory.dram-ue-imminent", FM_FMRI_SCHEME_MEM, 287 FM_MEM_SCHEME_VERSION, NULL }, 288 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 289 NULL }, 290 { "fault.cpu.SPARC64-VI.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 291 cma_cpu_cpu_retire }, 292 { "fault.cpu.SPARC64-VII.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 293 cma_cpu_cpu_retire }, 294 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VI.core.se", 295 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 296 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VI.core.se-offlinereq", 297 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 298 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VI.core.ce", 299 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 300 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VI.core.ce-offlinereq", 301 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 302 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VII.core.se", 303 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 304 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VII.core.se-offlinereq", 305 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 306 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VII.core.ce", 307 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 308 { "fault.chassis.SPARC-Enterprise.cpu.SPARC64-VII.core.ce-offlinereq", 309 FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, cma_cpu_hc_retire }, 310 #else 311 /* 312 * For platforms excluding i386, sun4v and opl. 313 */ 314 { "fault.memory.page", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 315 cma_page_retire }, 316 { "fault.memory.page_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 317 cma_page_retire }, 318 { "fault.memory.page_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 319 cma_page_retire }, 320 { "fault.memory.page_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 321 cma_page_retire }, 322 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 323 NULL }, 324 { "fault.memory.dimm_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 325 NULL }, 326 { "fault.memory.dimm_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 327 NULL }, 328 { "fault.memory.dimm_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 329 NULL }, 330 { "fault.memory.dimm-page-retires-excessive", FM_FMRI_SCHEME_MEM, 331 FM_MEM_SCHEME_VERSION, NULL }, 332 { "fault.memory.dimm-ue-imminent", FM_FMRI_SCHEME_MEM, 333 FM_MEM_SCHEME_VERSION, NULL }, 334 { "fault.memory.dram-ue-imminent", FM_FMRI_SCHEME_MEM, 335 FM_MEM_SCHEME_VERSION, NULL }, 336 { "fault.memory.dimm_testfail", FM_FMRI_SCHEME_MEM, 337 FM_MEM_SCHEME_VERSION, NULL }, 338 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 339 NULL }, 340 { "fault.memory.datapath", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 341 NULL }, 342 343 /* 344 * The following faults do NOT retire a cpu thread, 345 * and therefore must be intercepted before 346 * the default "fault.cpu.*" dispatch to cma_cpu_cpu_retire. 347 */ 348 { "fault.cpu.ultraSPARC-IVplus.l2cachedata-line", 349 FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 350 cma_cache_way_retire }, 351 { "fault.cpu.ultraSPARC-IVplus.l3cachedata-line", 352 FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 353 cma_cache_way_retire }, 354 { "fault.cpu.ultraSPARC-IVplus.l2cachetag-line", 355 FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 356 cma_cache_way_retire }, 357 { "fault.cpu.ultraSPARC-IVplus.l3cachetag-line", 358 FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 359 cma_cache_way_retire }, 360 361 /* 362 * Default "fault.cpu.*" for "cpu" scheme ASRU dispatch. 363 */ 364 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 365 cma_cpu_cpu_retire }, 366 #endif 367 { NULL, NULL, 0, NULL } 368 }; 369 370 static const cma_subscriber_t * 371 nvl2subr(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t **asrup) 372 { 373 const cma_subscriber_t *sp; 374 nvlist_t *asru; 375 char *scheme; 376 uint8_t version; 377 boolean_t retire; 378 379 if (nvlist_lookup_boolean_value(nvl, FM_SUSPECT_RETIRE, &retire) == 0 && 380 retire == 0) { 381 fmd_hdl_debug(hdl, "cma_recv: retire suppressed"); 382 return (NULL); 383 } 384 385 if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) != 0 || 386 nvlist_lookup_string(asru, FM_FMRI_SCHEME, &scheme) != 0 || 387 nvlist_lookup_uint8(asru, FM_VERSION, &version) != 0) { 388 cma_stats.bad_flts.fmds_value.ui64++; 389 return (NULL); 390 } 391 392 for (sp = cma_subrs; sp->subr_class != NULL; sp++) { 393 if (fmd_nvl_class_match(hdl, nvl, sp->subr_class) && 394 strcmp(scheme, sp->subr_sname) == 0 && 395 version <= sp->subr_svers) { 396 *asrup = asru; 397 return (sp); 398 } 399 } 400 401 cma_stats.nop_flts.fmds_value.ui64++; 402 return (NULL); 403 } 404 405 static void 406 cma_recv_list(fmd_hdl_t *hdl, nvlist_t *nvl, const char *class) 407 { 408 char *uuid = NULL; 409 nvlist_t **nva, **save_nva; 410 uint_t nvc = 0, save_nvc; 411 uint_t keepopen; 412 int err = 0; 413 nvlist_t *asru = NULL; 414 uint32_t index; 415 416 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid); 417 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, 418 &nva, &nvc); 419 if (err != 0) { 420 cma_stats.bad_flts.fmds_value.ui64++; 421 return; 422 } 423 424 save_nvc = keepopen = nvc; 425 save_nva = nva; 426 while (nvc-- != 0 && (strcmp(class, FM_LIST_SUSPECT_CLASS) != 0 || 427 !fmd_case_uuclosed(hdl, uuid))) { 428 nvlist_t *nvl = *nva++; 429 const cma_subscriber_t *subr; 430 int has_fault; 431 432 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 433 continue; 434 435 /* 436 * A handler returns CMA_RA_SUCCESS to indicate that 437 * from this suspects point-of-view the case may be 438 * closed, CMA_RA_FAILURE otherwise. 439 * A handler must not close the case itself. 440 */ 441 if (subr->subr_func != NULL) { 442 has_fault = fmd_nvl_fmri_has_fault(hdl, asru, 443 FMD_HAS_FAULT_ASRU, NULL); 444 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0) { 445 if (has_fault == 1) 446 err = subr->subr_func(hdl, nvl, asru, 447 uuid, 0); 448 } else { 449 if (has_fault == 0) 450 err = subr->subr_func(hdl, nvl, asru, 451 uuid, 1); 452 } 453 if (err == CMA_RA_SUCCESS) 454 keepopen--; 455 } 456 } 457 458 /* 459 * Run though again to catch any new faults in list.updated. 460 */ 461 while (save_nvc-- != 0 && (strcmp(class, FM_LIST_UPDATED_CLASS) == 0)) { 462 nvlist_t *nvl = *save_nva++; 463 const cma_subscriber_t *subr; 464 int has_fault; 465 466 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 467 continue; 468 if (subr->subr_func != NULL) { 469 has_fault = fmd_nvl_fmri_has_fault(hdl, asru, 470 FMD_HAS_FAULT_ASRU, NULL); 471 if (has_fault == 1) 472 err = subr->subr_func(hdl, nvl, asru, uuid, 0); 473 } 474 } 475 476 /* 477 * Do not close the case if we are handling cache faults. 478 */ 479 if (asru != NULL) { 480 if (nvlist_lookup_uint32(asru, FM_FMRI_CPU_CACHE_INDEX, 481 &index) != 0) { 482 if (!keepopen && strcmp(class, 483 FM_LIST_SUSPECT_CLASS) == 0) { 484 fmd_case_uuclose(hdl, uuid); 485 } 486 } 487 } 488 489 if (!keepopen && strcmp(class, FM_LIST_REPAIRED_CLASS) == 0) 490 fmd_case_uuresolved(hdl, uuid); 491 } 492 493 static void 494 cma_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl) 495 { 496 const cma_subscriber_t *subr; 497 nvlist_t *asru; 498 499 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 500 return; 501 502 if (subr->subr_func != NULL) { 503 if (fmd_nvl_fmri_has_fault(hdl, asru, 504 FMD_HAS_FAULT_ASRU, NULL) == 1) 505 (void) subr->subr_func(hdl, nvl, asru, NULL, 0); 506 } 507 } 508 509 /*ARGSUSED*/ 510 static void 511 cma_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 512 { 513 fmd_hdl_debug(hdl, "received %s\n", class); 514 515 if (strcmp(class, FM_LIST_RESOLVED_CLASS) == 0) 516 return; 517 518 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0 || 519 strcmp(class, FM_LIST_REPAIRED_CLASS) == 0 || 520 strcmp(class, FM_LIST_UPDATED_CLASS) == 0) 521 cma_recv_list(hdl, nvl, class); 522 else 523 cma_recv_one(hdl, nvl); 524 } 525 526 /*ARGSUSED*/ 527 static void 528 cma_timeout(fmd_hdl_t *hdl, id_t id, void *arg) 529 { 530 if (id == cma.cma_page_timerid) 531 cma_page_retry(hdl); 532 #ifdef sun4v 533 /* 534 * cpu offline/online needs to be retried on sun4v because 535 * ldom request can be asynchronous. 536 */ 537 else if (id == cma.cma_cpu_timerid) 538 cma_cpu_retry(hdl); 539 #endif 540 } 541 542 #ifdef sun4v 543 static void * 544 cma_init_alloc(size_t size) 545 { 546 return (fmd_hdl_alloc(init_hdl, size, FMD_SLEEP)); 547 } 548 549 static void 550 cma_init_free(void *addr, size_t size) 551 { 552 fmd_hdl_free(init_hdl, addr, size); 553 } 554 #endif 555 556 static const fmd_hdl_ops_t fmd_ops = { 557 cma_recv, /* fmdo_recv */ 558 cma_timeout, /* fmdo_timeout */ 559 NULL, /* fmdo_close */ 560 NULL, /* fmdo_stats */ 561 NULL, /* fmdo_gc */ 562 }; 563 564 static const fmd_prop_t fmd_props[] = { 565 { "cpu_tries", FMD_TYPE_UINT32, "10" }, 566 { "cpu_delay", FMD_TYPE_TIME, "1sec" }, 567 #ifdef sun4v 568 { "cpu_ret_mindelay", FMD_TYPE_TIME, "5sec" }, 569 { "cpu_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 570 #endif /* sun4v */ 571 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" }, 572 { "cpu_online_enable", FMD_TYPE_BOOL, "true" }, 573 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" }, 574 #ifdef opl 575 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "false" }, 576 { "cpu_unblacklist_enable", FMD_TYPE_BOOL, "false" }, 577 #else 578 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "true" }, 579 { "cpu_unblacklist_enable", FMD_TYPE_BOOL, "true" }, 580 #endif /* opl */ 581 { "page_ret_mindelay", FMD_TYPE_TIME, "1sec" }, 582 { "page_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 583 { "page_retire_enable", FMD_TYPE_BOOL, "true" }, 584 { "page_unretire_enable", FMD_TYPE_BOOL, "true" }, 585 { NULL, 0, NULL } 586 }; 587 588 static const fmd_hdl_info_t fmd_info = { 589 "CPU/Memory Retire Agent", CMA_VERSION, &fmd_ops, fmd_props 590 }; 591 592 void 593 _fmd_init(fmd_hdl_t *hdl) 594 { 595 hrtime_t nsec; 596 #ifdef i386 597 char buf[BUFSIZ]; 598 const char *dom0 = "control_d"; 599 600 /* 601 * Abort the cpumem-retire module if Solaris is running under DomU. 602 */ 603 if (sysinfo(SI_PLATFORM, buf, sizeof (buf)) == -1) 604 return; 605 606 if (strncmp(buf, "i86pc", sizeof (buf)) == 0) { 607 cma_is_native = B_TRUE; 608 } else if (strncmp(buf, "i86xpv", sizeof (buf)) != 0) { 609 return; 610 } else { 611 int fd = open("/dev/xen/domcaps", O_RDONLY); 612 613 if (fd != -1) { 614 if (read(fd, buf, sizeof (buf)) <= 0 || 615 strncmp(buf, dom0, strlen(dom0)) != 0) { 616 (void) close(fd); 617 return; 618 } 619 (void) close(fd); 620 } 621 cma_is_native = B_FALSE; 622 } 623 #endif /* i386 */ 624 625 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) 626 return; /* invalid data in configuration file */ 627 628 fmd_hdl_subscribe(hdl, "fault.cpu.*"); 629 fmd_hdl_subscribe(hdl, "fault.memory.*"); 630 #ifdef opl 631 fmd_hdl_subscribe(hdl, "fault.chassis.SPARC-Enterprise.cpu.*"); 632 #endif 633 634 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cma_stats) / 635 sizeof (fmd_stat_t), (fmd_stat_t *)&cma_stats); 636 637 cma.cma_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries"); 638 639 nsec = fmd_prop_get_int64(hdl, "cpu_delay"); 640 cma.cma_cpu_delay.tv_sec = nsec / NANOSEC; 641 cma.cma_cpu_delay.tv_nsec = nsec % NANOSEC; 642 643 cma.cma_page_mindelay = fmd_prop_get_int64(hdl, "page_ret_mindelay"); 644 cma.cma_page_maxdelay = fmd_prop_get_int64(hdl, "page_ret_maxdelay"); 645 646 #ifdef sun4v 647 cma.cma_cpu_mindelay = fmd_prop_get_int64(hdl, "cpu_ret_mindelay"); 648 cma.cma_cpu_maxdelay = fmd_prop_get_int64(hdl, "cpu_ret_maxdelay"); 649 #endif 650 651 cma.cma_cpu_dooffline = fmd_prop_get_int32(hdl, "cpu_offline_enable"); 652 cma.cma_cpu_forcedoffline = fmd_prop_get_int32(hdl, 653 "cpu_forced_offline"); 654 cma.cma_cpu_doonline = fmd_prop_get_int32(hdl, "cpu_online_enable"); 655 cma.cma_cpu_doblacklist = fmd_prop_get_int32(hdl, 656 "cpu_blacklist_enable"); 657 cma.cma_cpu_dounblacklist = fmd_prop_get_int32(hdl, 658 "cpu_unblacklist_enable"); 659 cma.cma_page_doretire = fmd_prop_get_int32(hdl, "page_retire_enable"); 660 cma.cma_page_dounretire = fmd_prop_get_int32(hdl, 661 "page_unretire_enable"); 662 663 if (cma.cma_page_maxdelay < cma.cma_page_mindelay) 664 fmd_hdl_abort(hdl, "page retirement delays conflict\n"); 665 666 #ifdef sun4v 667 init_hdl = hdl; 668 cma_lhp = ldom_init(cma_init_alloc, cma_init_free); 669 #endif 670 } 671 672 void 673 _fmd_fini(fmd_hdl_t *hdl) 674 { 675 #ifdef sun4v 676 ldom_fini(cma_lhp); 677 cma_cpu_fini(hdl); 678 #endif 679 cma_page_fini(hdl); 680 } 681