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, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 23 /* 24 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 25 * Use is subject to license terms. 26 */ 27 28 #pragma ident "%Z%%M% %I% %E% SMI" 29 30 #include <sys/fm/protocol.h> 31 32 #include <fmd_api.h> 33 #include <fmd_subr.h> 34 #include <fmd_string.h> 35 #include <fmd_protocol.h> 36 #include <fmd_module.h> 37 #include <fmd_error.h> 38 39 static struct { 40 fmd_stat_t nosub; 41 fmd_stat_t module; 42 } self_stats = { 43 { "nosub", FMD_TYPE_UINT64, "event classes with no subscribers seen" }, 44 { "module", FMD_TYPE_UINT64, "error events received from fmd modules" }, 45 }; 46 47 typedef struct self_case { 48 enum { SC_CLASS, SC_MODULE } sc_kind; 49 char *sc_name; 50 } self_case_t; 51 52 static self_case_t * 53 self_case_create(fmd_hdl_t *hdl, int kind, const char *name) 54 { 55 self_case_t *scp = fmd_hdl_alloc(hdl, sizeof (self_case_t), FMD_SLEEP); 56 57 scp->sc_kind = kind; 58 scp->sc_name = fmd_hdl_strdup(hdl, name, FMD_SLEEP); 59 60 return (scp); 61 } 62 63 static void 64 self_case_destroy(fmd_hdl_t *hdl, self_case_t *scp) 65 { 66 fmd_hdl_strfree(hdl, scp->sc_name); 67 fmd_hdl_free(hdl, scp, sizeof (self_case_t)); 68 } 69 70 static fmd_case_t * 71 self_case_lookup(fmd_hdl_t *hdl, int kind, const char *name) 72 { 73 fmd_case_t *cp = NULL; 74 75 while ((cp = fmd_case_next(hdl, cp)) != NULL) { 76 self_case_t *scp = fmd_case_getspecific(hdl, cp); 77 if (scp->sc_kind == kind && strcmp(scp->sc_name, name) == 0) 78 break; 79 } 80 81 return (cp); 82 } 83 84 /*ARGSUSED*/ 85 static void 86 self_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 87 { 88 fmd_case_t *cp; 89 nvlist_t *flt, *mod; 90 char *name; 91 int err = 0; 92 93 /* 94 * If we get an error report from another fmd module, then create a 95 * case for the module and add the ereport to it. The error is either 96 * from fmd_hdl_error() or from fmd_api_error(). If it is the latter, 97 * fmd_module_error() will send another event of class EFMD_MOD_FAIL 98 * when the module has failed, at which point we can solve the case. 99 * We can also close the case on EFMD_MOD_CONF (bad config file). 100 */ 101 if (strcmp(class, fmd_errclass(EFMD_MODULE)) == 0 && 102 nvlist_lookup_nvlist(nvl, FM_EREPORT_DETECTOR, &mod) == 0 && 103 nvlist_lookup_string(mod, FM_FMRI_FMD_NAME, &name) == 0) { 104 105 if ((cp = self_case_lookup(hdl, SC_MODULE, name)) == NULL) { 106 cp = fmd_case_open(hdl, 107 self_case_create(hdl, SC_MODULE, name)); 108 } 109 110 fmd_case_add_ereport(hdl, cp, ep); 111 self_stats.module.fmds_value.ui64++; 112 (void) nvlist_lookup_int32(nvl, FMD_ERR_MOD_ERRNO, &err); 113 114 if (err != EFMD_MOD_FAIL && err != EFMD_MOD_CONF) 115 return; /* module is still active, so keep case open */ 116 117 if (fmd_case_solved(hdl, cp)) 118 return; /* case is already closed but error in _fini */ 119 120 class = err == EFMD_MOD_FAIL ? FMD_FLT_MOD : FMD_FLT_CONF; 121 flt = fmd_protocol_fault(class, 100, mod, NULL, NULL); 122 123 fmd_case_add_suspect(hdl, cp, flt); 124 fmd_case_solve(hdl, cp); 125 126 return; 127 } 128 129 /* 130 * If we get an I/O DDI ereport, drop it for now until the I/O DE is 131 * implemented and integrated. Existing drivers in O/N have bugs that 132 * will trigger these and we don't want this producing FMD_FLT_NOSUB. 133 */ 134 if (strncmp(class, "ereport.io.ddi.", strlen("ereport.io.ddi.")) == 0) 135 return; /* if we got a DDI ereport, drop it for now */ 136 137 /* 138 * If we get any other type of event then it is of a class for which 139 * there are no subscribers. Some of these correspond to internal fmd 140 * errors, which we ignore. Otherwise we keep one case per class and 141 * use it to produce a message indicating that something is awry. 142 */ 143 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0 || 144 strcmp(class, FM_LIST_ISOLATED_CLASS) == 0 || 145 strcmp(class, FM_LIST_REPAIRED_CLASS) == 0) 146 return; /* if no agents are present just drop list.suspect */ 147 148 if (strncmp(class, FMD_ERR_CLASS, FMD_ERR_CLASS_LEN) == 0) 149 return; /* if fmd itself produced the error just drop it */ 150 151 if (strncmp(class, FMD_RSRC_CLASS, FMD_RSRC_CLASS_LEN) == 0) 152 return; /* if fmd itself produced the event just drop it */ 153 154 if (self_case_lookup(hdl, SC_CLASS, class) != NULL) 155 return; /* case is already open against this class */ 156 157 cp = fmd_case_open(hdl, self_case_create(hdl, SC_CLASS, class)); 158 fmd_case_add_ereport(hdl, cp, ep); 159 self_stats.nosub.fmds_value.ui64++; 160 161 flt = fmd_protocol_fault(FMD_FLT_NOSUB, 100, NULL, NULL, NULL); 162 fmd_case_add_suspect(hdl, cp, flt); 163 fmd_case_solve(hdl, cp); 164 } 165 166 static void 167 self_close(fmd_hdl_t *hdl, fmd_case_t *cp) 168 { 169 self_case_destroy(hdl, fmd_case_getspecific(hdl, cp)); 170 } 171 172 static const fmd_hdl_ops_t self_ops = { 173 self_recv, /* fmdo_recv */ 174 NULL, /* fmdo_timeout */ 175 self_close, /* fmdo_close */ 176 NULL, /* fmdo_stats */ 177 NULL, /* fmdo_gc */ 178 }; 179 180 void 181 self_init(fmd_hdl_t *hdl) 182 { 183 fmd_module_t *mp = (fmd_module_t *)hdl; /* see below */ 184 185 fmd_hdl_info_t info = { 186 "Fault Manager Self-Diagnosis", "1.0", &self_ops, NULL 187 }; 188 189 /* 190 * Unlike other modules, fmd-self-diagnosis has some special needs that 191 * fall outside of what we want in the module API. Manually disable 192 * checkpointing for this module by tweaking the mod_stats values. 193 * The self-diagnosis world relates to fmd's running state and modules 194 * which all change when it restarts, so don't bother w/ checkpointing. 195 */ 196 (void) pthread_mutex_lock(&mp->mod_stats_lock); 197 mp->mod_stats->ms_ckpt_save.fmds_value.bool = FMD_B_FALSE; 198 mp->mod_stats->ms_ckpt_restore.fmds_value.bool = FMD_B_FALSE; 199 (void) pthread_mutex_unlock(&mp->mod_stats_lock); 200 201 if (fmd_hdl_register(hdl, FMD_API_VERSION, &info) != 0) 202 return; /* failed to register with fmd */ 203 204 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (self_stats) / 205 sizeof (fmd_stat_t), (fmd_stat_t *)&self_stats); 206 } 207 208 void 209 self_fini(fmd_hdl_t *hdl) 210 { 211 fmd_case_t *cp = NULL; 212 213 while ((cp = fmd_case_next(hdl, cp)) != NULL) 214 self_case_destroy(hdl, fmd_case_getspecific(hdl, cp)); 215 } 216