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 2007 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 #include <sys/systeminfo.h> 37 #include <sys/utsname.h> 38 39 cma_t cma; 40 41 cma_stats_t cma_stats = { 42 { "cpu_flts", FMD_TYPE_UINT64, "cpu faults resolved" }, 43 { "cpu_fails", FMD_TYPE_UINT64, "cpu faults unresolveable" }, 44 { "cpu_blfails", FMD_TYPE_UINT64, "failed cpu blacklists" }, 45 { "cpu_supp", FMD_TYPE_UINT64, "cpu offlines suppressed" }, 46 { "cpu_blsupp", FMD_TYPE_UINT64, "cpu blacklists suppressed" }, 47 { "page_flts", FMD_TYPE_UINT64, "page faults resolved" }, 48 { "page_fails", FMD_TYPE_UINT64, "page faults unresolveable" }, 49 { "page_supp", FMD_TYPE_UINT64, "page retires suppressed" }, 50 { "page_nonent", FMD_TYPE_UINT64, "retires for non-existent fmris" }, 51 { "page_retmax", FMD_TYPE_UINT64, "hit max retries for page retire" }, 52 { "bad_flts", FMD_TYPE_UINT64, "invalid fault events received" }, 53 { "nop_flts", FMD_TYPE_UINT64, "inapplicable fault events received" }, 54 { "auto_flts", FMD_TYPE_UINT64, "auto-close faults received" } 55 }; 56 57 typedef struct cma_subscriber { 58 const char *subr_class; 59 const char *subr_sname; 60 uint_t subr_svers; 61 int (*subr_func)(fmd_hdl_t *, nvlist_t *, nvlist_t *, const char *); 62 } cma_subscriber_t; 63 64 static const cma_subscriber_t cma_subrs[] = { 65 { "fault.memory.page", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 66 cma_page_retire }, 67 { "fault.memory.page_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 68 cma_page_retire }, 69 { "fault.memory.page_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 70 cma_page_retire }, 71 { "fault.memory.page_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 72 cma_page_retire }, 73 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 74 NULL }, 75 { "fault.memory.dimm_sb", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 76 NULL }, 77 { "fault.memory.dimm_ck", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 78 NULL }, 79 { "fault.memory.dimm_ue", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 80 NULL }, 81 { "fault.memory.dimm_testfail", FM_FMRI_SCHEME_MEM, 82 FM_MEM_SCHEME_VERSION, NULL }, 83 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 84 NULL }, 85 { "fault.memory.datapath", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 86 NULL }, 87 88 /* 89 * The following faults do NOT retire a cpu thread, 90 * and therefore must be intercepted before 91 * the default "fault.cpu.*" dispatch to cma_cpu_retire. 92 */ 93 { "fault.cpu.ultraSPARC-T1.freg", FM_FMRI_SCHEME_CPU, 94 FM_CPU_SCHEME_VERSION, NULL }, 95 { "fault.cpu.ultraSPARC-T1.l2cachedata", FM_FMRI_SCHEME_CPU, 96 FM_CPU_SCHEME_VERSION, NULL }, 97 { "fault.cpu.ultraSPARC-T1.l2cachetag", FM_FMRI_SCHEME_CPU, 98 FM_CPU_SCHEME_VERSION, NULL }, 99 { "fault.cpu.ultraSPARC-T1.l2cachectl", FM_FMRI_SCHEME_CPU, 100 FM_CPU_SCHEME_VERSION, NULL }, 101 { "fault.cpu.ultraSPARC-T1.mau", FM_FMRI_SCHEME_CPU, 102 FM_CPU_SCHEME_VERSION, NULL }, 103 { "fault.cpu.amd.dramchannel", FM_FMRI_SCHEME_HC, FM_HC_SCHEME_VERSION, 104 NULL }, 105 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 106 cma_cpu_retire }, 107 { NULL, NULL, 0, NULL } 108 }; 109 110 static const cma_subscriber_t * 111 nvl2subr(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t **asrup) 112 { 113 const cma_subscriber_t *sp; 114 nvlist_t *asru; 115 char *scheme; 116 uint8_t version; 117 char *fltclass = "(unknown)"; 118 119 if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) != 0 || 120 nvlist_lookup_string(asru, FM_FMRI_SCHEME, &scheme) != 0 || 121 nvlist_lookup_uint8(asru, FM_VERSION, &version) != 0) { 122 cma_stats.bad_flts.fmds_value.ui64++; 123 return (NULL); 124 } 125 126 for (sp = cma_subrs; sp->subr_class != NULL; sp++) { 127 if (fmd_nvl_class_match(hdl, nvl, sp->subr_class) && 128 strcmp(scheme, sp->subr_sname) == 0 && 129 version <= sp->subr_svers) { 130 *asrup = asru; 131 return (sp); 132 } 133 } 134 135 (void) nvlist_lookup_string(nvl, FM_CLASS, &fltclass); 136 fmd_hdl_error(hdl, "No handling disposition for %s with asru in " 137 "scheme \"%s\"\n", fltclass, scheme); 138 cma_stats.nop_flts.fmds_value.ui64++; 139 return (NULL); 140 } 141 142 static void 143 cma_recv_list(fmd_hdl_t *hdl, nvlist_t *nvl) 144 { 145 char *uuid = NULL; 146 nvlist_t **nva; 147 uint_t nvc = 0; 148 uint_t keepopen; 149 int err = 0; 150 151 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid); 152 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, 153 &nva, &nvc); 154 if (err != 0) { 155 cma_stats.bad_flts.fmds_value.ui64++; 156 return; 157 } 158 159 keepopen = nvc; 160 while (nvc-- != 0 && !fmd_case_uuclosed(hdl, uuid)) { 161 nvlist_t *nvl = *nva++; 162 const cma_subscriber_t *subr; 163 nvlist_t *asru; 164 165 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 166 continue; 167 168 /* 169 * A handler returns CMA_RA_SUCCESS to indicate that 170 * from this suspects point-of-view the case may be 171 * closed, CMA_RA_FAILURE otherwise. 172 * A handler must not close the case itself. 173 */ 174 if (subr->subr_func != NULL) { 175 err = subr->subr_func(hdl, nvl, asru, uuid); 176 177 if (err == CMA_RA_SUCCESS) 178 keepopen--; 179 } 180 } 181 182 if (!keepopen) 183 fmd_case_uuclose(hdl, uuid); 184 } 185 186 static void 187 cma_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl) 188 { 189 const cma_subscriber_t *subr; 190 nvlist_t *asru; 191 192 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 193 return; 194 195 if (subr->subr_func != NULL) 196 (void) subr->subr_func(hdl, nvl, asru, NULL); 197 198 } 199 200 /*ARGSUSED*/ 201 static void 202 cma_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 203 { 204 fmd_hdl_debug(hdl, "received %s\n", class); 205 206 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0) 207 cma_recv_list(hdl, nvl); 208 else 209 cma_recv_one(hdl, nvl); 210 } 211 212 /*ARGSUSED*/ 213 static void 214 cma_timeout(fmd_hdl_t *hdl, id_t id, void *arg) 215 { 216 if (id == cma.cma_page_timerid) 217 cma_page_retry(hdl); 218 } 219 220 static const fmd_hdl_ops_t fmd_ops = { 221 cma_recv, /* fmdo_recv */ 222 cma_timeout, /* fmdo_timeout */ 223 NULL, /* fmdo_close */ 224 NULL, /* fmdo_stats */ 225 NULL, /* fmdo_gc */ 226 }; 227 228 static const fmd_prop_t fmd_props[] = { 229 { "cpu_tries", FMD_TYPE_UINT32, "10" }, 230 { "cpu_delay", FMD_TYPE_TIME, "1sec" }, 231 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" }, 232 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" }, 233 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "true" }, 234 { "page_ret_mindelay", FMD_TYPE_TIME, "1sec" }, 235 { "page_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 236 { "page_retire_enable", FMD_TYPE_BOOL, "true" }, 237 #ifdef i386 238 /* 239 * On i386, leaving cases open while we retry the 240 * retire can cause the eft module to use large amounts 241 * of memory. Until eft is fixed, we set a maximum number 242 * of retries on page retires, after which the case will 243 * be closed. 244 */ 245 { "page_retire_maxretries", FMD_TYPE_UINT32, "5" }, 246 #else 247 { "page_retire_maxretries", FMD_TYPE_UINT32, "0" }, 248 #endif /* i386 */ 249 { NULL, 0, NULL } 250 }; 251 252 static const fmd_hdl_info_t fmd_info = { 253 "CPU/Memory Retire Agent", CMA_VERSION, &fmd_ops, fmd_props 254 }; 255 256 void 257 _fmd_init(fmd_hdl_t *hdl) 258 { 259 hrtime_t nsec; 260 char buf[SYS_NMLN]; 261 int ret; 262 263 /* 264 * Abort the cpumem-retire module if Solaris is running under the Xen 265 * hypervisor. 266 */ 267 ret = sysinfo(SI_PLATFORM, buf, sizeof (buf)); 268 if (ret == -1 || (strncmp(buf, "i86xpv", sizeof (buf)) == 0)) 269 return; 270 271 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) 272 return; /* invalid data in configuration file */ 273 274 fmd_hdl_subscribe(hdl, "fault.cpu.*"); 275 fmd_hdl_subscribe(hdl, "fault.memory.*"); 276 277 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cma_stats) / 278 sizeof (fmd_stat_t), (fmd_stat_t *)&cma_stats); 279 280 cma.cma_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries"); 281 282 nsec = fmd_prop_get_int64(hdl, "cpu_delay"); 283 cma.cma_cpu_delay.tv_sec = nsec / NANOSEC; 284 cma.cma_cpu_delay.tv_nsec = nsec % NANOSEC; 285 286 cma.cma_page_mindelay = fmd_prop_get_int64(hdl, "page_ret_mindelay"); 287 cma.cma_page_maxdelay = fmd_prop_get_int64(hdl, "page_ret_maxdelay"); 288 289 cma.cma_cpu_dooffline = fmd_prop_get_int32(hdl, "cpu_offline_enable"); 290 cma.cma_cpu_forcedoffline = fmd_prop_get_int32(hdl, 291 "cpu_forced_offline"); 292 cma.cma_cpu_doblacklist = fmd_prop_get_int32(hdl, 293 "cpu_blacklist_enable"); 294 cma.cma_page_doretire = fmd_prop_get_int32(hdl, "page_retire_enable"); 295 cma.cma_page_maxretries = 296 fmd_prop_get_int32(hdl, "page_retire_maxretries"); 297 298 if (cma.cma_page_maxdelay < cma.cma_page_mindelay) 299 fmd_hdl_abort(hdl, "page retirement delays conflict\n"); 300 } 301 302 void 303 _fmd_fini(fmd_hdl_t *hdl) 304 { 305 cma_page_fini(hdl); 306 } 307