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 2010 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #include <sys/fm/protocol.h> 28 29 #include <fmd_api.h> 30 #include <fmd_subr.h> 31 #include <fmd_string.h> 32 #include <fmd_protocol.h> 33 #include <fmd_module.h> 34 #include <fmd_error.h> 35 36 static struct { 37 fmd_stat_t nosub; 38 fmd_stat_t module; 39 } self_stats = { 40 { "nosub", FMD_TYPE_UINT64, "event classes with no subscribers seen" }, 41 { "module", FMD_TYPE_UINT64, "error events received from fmd modules" }, 42 }; 43 44 typedef struct self_case { 45 enum { SC_CLASS, SC_MODULE } sc_kind; 46 char *sc_name; 47 } self_case_t; 48 49 static self_case_t * 50 self_case_create(fmd_hdl_t *hdl, int kind, const char *name) 51 { 52 self_case_t *scp = fmd_hdl_alloc(hdl, sizeof (self_case_t), FMD_SLEEP); 53 54 scp->sc_kind = kind; 55 scp->sc_name = fmd_hdl_strdup(hdl, name, FMD_SLEEP); 56 57 return (scp); 58 } 59 60 static void 61 self_case_destroy(fmd_hdl_t *hdl, self_case_t *scp) 62 { 63 fmd_hdl_strfree(hdl, scp->sc_name); 64 fmd_hdl_free(hdl, scp, sizeof (self_case_t)); 65 } 66 67 static fmd_case_t * 68 self_case_lookup(fmd_hdl_t *hdl, int kind, const char *name) 69 { 70 fmd_case_t *cp = NULL; 71 72 while ((cp = fmd_case_next(hdl, cp)) != NULL) { 73 self_case_t *scp = fmd_case_getspecific(hdl, cp); 74 if (scp->sc_kind == kind && strcmp(scp->sc_name, name) == 0) 75 break; 76 } 77 78 return (cp); 79 } 80 81 /*ARGSUSED*/ 82 static void 83 self_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 84 { 85 fmd_case_t *cp; 86 nvlist_t *flt, *mod; 87 char *name; 88 int err = 0; 89 90 /* 91 * If we get an error report from another fmd module, then create a 92 * case for the module and add the ereport to it. The error is either 93 * from fmd_hdl_error() or from fmd_api_error(). If it is the latter, 94 * fmd_module_error() will send another event of class EFMD_MOD_FAIL 95 * when the module has failed, at which point we can solve the case. 96 * We can also close the case on EFMD_MOD_CONF (bad config file). 97 */ 98 if (strcmp(class, fmd_errclass(EFMD_MODULE)) == 0 && 99 nvlist_lookup_nvlist(nvl, FM_EREPORT_DETECTOR, &mod) == 0 && 100 nvlist_lookup_string(mod, FM_FMRI_FMD_NAME, &name) == 0) { 101 102 if ((cp = self_case_lookup(hdl, SC_MODULE, name)) == NULL) { 103 cp = fmd_case_open(hdl, 104 self_case_create(hdl, SC_MODULE, name)); 105 } 106 107 fmd_case_add_ereport(hdl, cp, ep); 108 self_stats.module.fmds_value.ui64++; 109 (void) nvlist_lookup_int32(nvl, FMD_ERR_MOD_ERRNO, &err); 110 111 if (err != EFMD_MOD_FAIL && err != EFMD_MOD_CONF) 112 return; /* module is still active, so keep case open */ 113 114 if (fmd_case_solved(hdl, cp)) 115 return; /* case is already closed but error in _fini */ 116 117 class = err == EFMD_MOD_FAIL ? FMD_FLT_MOD : FMD_FLT_CONF; 118 flt = fmd_protocol_fault(class, 100, mod, NULL, NULL, NULL); 119 120 fmd_case_add_suspect(hdl, cp, flt); 121 fmd_case_solve(hdl, cp); 122 123 return; 124 } 125 126 /* 127 * If we get an I/O DDI ereport, drop it for now until the I/O DE is 128 * implemented and integrated. Existing drivers in O/N have bugs that 129 * will trigger these and we don't want this producing FMD_FLT_NOSUB. 130 */ 131 if (strncmp(class, "ereport.io.ddi.", strlen("ereport.io.ddi.")) == 0) 132 return; /* if we got a DDI ereport, drop it for now */ 133 134 /* 135 * If we get any other type of event then it is of a class for which 136 * there are no subscribers. Some of these correspond to internal fmd 137 * errors, which we ignore. Otherwise we keep one case per class and 138 * use it to produce a message indicating that something is awry. 139 */ 140 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0 || 141 strcmp(class, FM_LIST_ISOLATED_CLASS) == 0 || 142 strcmp(class, FM_LIST_UPDATED_CLASS) == 0 || 143 strcmp(class, FM_LIST_RESOLVED_CLASS) == 0 || 144 strcmp(class, FM_LIST_REPAIRED_CLASS) == 0 || 145 strncmp(class, FM_FAULT_CLASS, strlen(FM_FAULT_CLASS)) == 0 || 146 strncmp(class, FM_DEFECT_CLASS, strlen(FM_DEFECT_CLASS)) == 0) 147 return; /* if no agents are present just drop list.* */ 148 149 if (strncmp(class, FMD_ERR_CLASS, FMD_ERR_CLASS_LEN) == 0) 150 return; /* if fmd itself produced the error just drop it */ 151 152 if (strncmp(class, FMD_RSRC_CLASS, FMD_RSRC_CLASS_LEN) == 0) 153 return; /* if fmd itself produced the event just drop it */ 154 155 if (strncmp(class, SYSEVENT_RSRC_CLASS, SYSEVENT_RSRC_CLASS_LEN) == 0) 156 return; /* sysvent resources are auto generated by fmd */ 157 158 if (self_case_lookup(hdl, SC_CLASS, class) != NULL) 159 return; /* case is already open against this class */ 160 161 cp = fmd_case_open(hdl, self_case_create(hdl, SC_CLASS, class)); 162 fmd_case_add_ereport(hdl, cp, ep); 163 self_stats.nosub.fmds_value.ui64++; 164 165 flt = fmd_protocol_fault(FMD_FLT_NOSUB, 100, NULL, NULL, NULL, NULL); 166 (void) nvlist_add_string(flt, "nosub_class", class); 167 fmd_case_add_suspect(hdl, cp, flt); 168 fmd_case_solve(hdl, cp); 169 } 170 171 static void 172 self_close(fmd_hdl_t *hdl, fmd_case_t *cp) 173 { 174 self_case_destroy(hdl, fmd_case_getspecific(hdl, cp)); 175 } 176 177 static const fmd_hdl_ops_t self_ops = { 178 self_recv, /* fmdo_recv */ 179 NULL, /* fmdo_timeout */ 180 self_close, /* fmdo_close */ 181 NULL, /* fmdo_stats */ 182 NULL, /* fmdo_gc */ 183 }; 184 185 void 186 self_init(fmd_hdl_t *hdl) 187 { 188 fmd_module_t *mp = (fmd_module_t *)hdl; /* see below */ 189 190 fmd_hdl_info_t info = { 191 "Fault Manager Self-Diagnosis", "1.0", &self_ops, NULL 192 }; 193 194 /* 195 * Unlike other modules, fmd-self-diagnosis has some special needs that 196 * fall outside of what we want in the module API. Manually disable 197 * checkpointing for this module by tweaking the mod_stats values. 198 * The self-diagnosis world relates to fmd's running state and modules 199 * which all change when it restarts, so don't bother w/ checkpointing. 200 */ 201 (void) pthread_mutex_lock(&mp->mod_stats_lock); 202 mp->mod_stats->ms_ckpt_save.fmds_value.bool = FMD_B_FALSE; 203 mp->mod_stats->ms_ckpt_restore.fmds_value.bool = FMD_B_FALSE; 204 (void) pthread_mutex_unlock(&mp->mod_stats_lock); 205 206 if (fmd_hdl_register(hdl, FMD_API_VERSION, &info) != 0) 207 return; /* failed to register with fmd */ 208 209 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (self_stats) / 210 sizeof (fmd_stat_t), (fmd_stat_t *)&self_stats); 211 } 212 213 void 214 self_fini(fmd_hdl_t *hdl) 215 { 216 fmd_case_t *cp = NULL; 217 218 while ((cp = fmd_case_next(hdl, cp)) != NULL) 219 self_case_destroy(hdl, fmd_case_getspecific(hdl, cp)); 220 } 221