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 /* 28 * FMD Transport Subsystem 29 * 30 * A transport module uses some underlying mechanism to transport events. 31 * This mechanism may use any underlying link-layer protocol and may support 32 * additional link-layer packets unrelated to FMA. Some appropriate link- 33 * layer mechanism to create the underlying connection is expected to be 34 * called prior to calling fmd_xprt_open() itself. Alternatively, a transport 35 * may be created in the suspended state by specifying the FMD_XPRT_SUSPENDED 36 * flag as part of the call to fmd_xprt_open(), and then may be resumed later. 37 * The underlying transport mechanism is *required* to provide ordering: that 38 * is, the sequences of bytes written across the transport must be read by 39 * the remote peer in the order that they are written, even across separate 40 * calls to fmdo_send(). As an example, the Internet TCP protocol would be 41 * a valid transport as it guarantees ordering, whereas the Internet UDP 42 * protocol would not because UDP datagrams may be delivered in any order 43 * as a result of delays introduced when datagrams pass through routers. 44 * 45 * Similar to sending events, a transport module receives events that are from 46 * its peer remote endpoint using some transport-specific mechanism that is 47 * unknown to FMD. As each event is received, the transport module is 48 * responsible for constructing a valid nvlist_t object from the data and then 49 * calling fmd_xprt_post() to post the event to the containing FMD's dispatch 50 * queue, making it available to all local modules that are not transport 51 * modules that have subscribed to the event. 52 * 53 * The following state machine is used for each transport. The initial state 54 * is either SYN, ACK, or RUN, depending on the flags specified to xprt_create. 55 * 56 * FMD_XPRT_ACCEPT !FMD_XPRT_ACCEPT 57 * | | 58 * waiting +--v--+ +--v--+ waiting 59 * for syn | SYN |--+ --+| ACK | for ack 60 * event +-----+ \ / +-----+ event 61 * | \ / | 62 * drop all +--v--+ X +--v--+ send subscriptions, 63 * events | ERR |<---+ +--->| SUB | recv subscriptions, 64 * +-----+ +-----+ wait for run event 65 * ^ | 66 * | +-----+ | 67 * +-----| RUN |<----+ 68 * +--^--+ 69 * | 70 * FMD_XPRT_RDONLY 71 * 72 * When fmd_xprt_open() is called without FMD_XPRT_ACCEPT, the Common Transport 73 * Layer enqueues a "syn" event for the module in its event queue and sets the 74 * state to ACK. In state ACK, we are waiting for the transport to get an 75 * "ack" event and call fmd_xprt_post() on this event. Other events will be 76 * discarded. If an "ack" is received, we transition to state SUB. If a 77 * configurable timeout occurs or if the "ack" is invalid (e.g. invalid version 78 * exchange), we transition to state ERR. Once in state ERR, no further 79 * operations are valid except fmd_xprt_close() and fmd_xprt_error() will 80 * return a non-zero value to the caller indicating the transport has failed. 81 * 82 * When fmd_xprt_open() is called with FMD_XPRT_ACCEPT, the Common Transport 83 * Layer assumes this transport is being used to accept a virtual connection 84 * from a remote peer that is sending a "syn", and sets the initial state to 85 * SYN. In this state, the transport waits for a "syn" event, validates it, 86 * and then transitions to state SUB if it is valid or state ERR if it is not. 87 * 88 * Once in state SUB, the transport module is expected to receive a sequence of 89 * zero or more "subscribe" events from the remote peer, followed by a "run" 90 * event. Once in state RUN, the transport is active and any events can be 91 * sent or received. The transport module is free to call fmd_xprt_close() 92 * from any state. The fmd_xprt_error() function will return zero if the 93 * transport is not in the ERR state, or non-zero if it is in the ERR state. 94 * 95 * Once the state machine reaches RUN, other FMA protocol events can be sent 96 * and received across the transport in addition to the various control events. 97 * 98 * Table of Common Transport Layer Control Events 99 * ============================================== 100 * 101 * FMA Class Payload 102 * --------- ------- 103 * resource.fm.xprt.uuclose string (uuid of case) 104 * resource.fm.xprt.uuresolved string (uuid of case) 105 * resource.fm.xprt.updated string (uuid of case) 106 * resource.fm.xprt.subscribe string (class pattern) 107 * resource.fm.xprt.unsubscribe string (class pattern) 108 * resource.fm.xprt.unsuback string (class pattern) 109 * resource.fm.xprt.syn version information 110 * resource.fm.xprt.ack version information 111 * resource.fm.xprt.run version information 112 * 113 * Control events are used to add and delete proxy subscriptions on the remote 114 * transport peer module, and to set up connections. When a "syn" event is 115 * sent, FMD will include in the payload the highest version of the FMA event 116 * protocol that is supported by the sender. When a "syn" event is received, 117 * the receiving FMD will use the minimum of this version and its version of 118 * the protocol, and reply with this new minimum version in the "ack" event. 119 * The receiver will then use this new minimum for subsequent event semantics. 120 */ 121 122 #include <sys/fm/protocol.h> 123 #include <strings.h> 124 #include <limits.h> 125 126 #include <fmd_alloc.h> 127 #include <fmd_error.h> 128 #include <fmd_conf.h> 129 #include <fmd_subr.h> 130 #include <fmd_string.h> 131 #include <fmd_protocol.h> 132 #include <fmd_thread.h> 133 #include <fmd_eventq.h> 134 #include <fmd_dispq.h> 135 #include <fmd_ctl.h> 136 #include <fmd_log.h> 137 #include <fmd_ustat.h> 138 #include <fmd_case.h> 139 #include <fmd_api.h> 140 #include <fmd_fmri.h> 141 #include <fmd_asru.h> 142 #include <fmd_xprt.h> 143 144 #include <fmd.h> 145 146 /* 147 * The states shown above in the transport state machine diagram are encoded 148 * using arrays of class patterns and a corresponding action function. These 149 * arrays are then passed to fmd_xprt_transition() to change transport states. 150 */ 151 152 const fmd_xprt_rule_t _fmd_xprt_state_syn[] = { 153 { "resource.fm.xprt.syn", fmd_xprt_event_syn }, 154 { "*", fmd_xprt_event_error }, 155 { NULL, NULL } 156 }; 157 158 const fmd_xprt_rule_t _fmd_xprt_state_ack[] = { 159 { "resource.fm.xprt.ack", fmd_xprt_event_ack }, 160 { "*", fmd_xprt_event_error }, 161 }; 162 163 const fmd_xprt_rule_t _fmd_xprt_state_err[] = { 164 { "*", fmd_xprt_event_drop }, 165 { NULL, NULL } 166 }; 167 168 const fmd_xprt_rule_t _fmd_xprt_state_sub[] = { 169 { "resource.fm.xprt.subscribe", fmd_xprt_event_sub }, 170 { "resource.fm.xprt.run", fmd_xprt_event_run }, 171 { "resource.fm.xprt.*", fmd_xprt_event_error }, 172 { "*", fmd_xprt_event_drop }, 173 { NULL, NULL } 174 }; 175 176 const fmd_xprt_rule_t _fmd_xprt_state_run[] = { 177 { "resource.fm.xprt.subscribe", fmd_xprt_event_sub }, 178 { "resource.fm.xprt.unsubscribe", fmd_xprt_event_unsub }, 179 { "resource.fm.xprt.unsuback", fmd_xprt_event_unsuback }, 180 { "resource.fm.xprt.uuclose", fmd_xprt_event_uuclose }, 181 { "resource.fm.xprt.uuresolved", fmd_xprt_event_uuresolved }, 182 { "resource.fm.xprt.updated", fmd_xprt_event_updated }, 183 { "resource.fm.xprt.*", fmd_xprt_event_error }, 184 { NULL, NULL } 185 }; 186 187 /* 188 * Template for per-transport statistics installed by fmd on behalf of each 189 * transport. These are used to initialize the per-transport xi_stats. For 190 * each statistic, the name is prepended with "fmd.xprt.%u", where %u is the 191 * transport ID (xi_id) and then are inserted into the per-module stats hash. 192 * The values in this array must match fmd_xprt_stat_t from <fmd_xprt.h>. 193 */ 194 static const fmd_xprt_stat_t _fmd_xprt_stat_tmpl = { 195 { 196 { "dispatched", FMD_TYPE_UINT64, "total events dispatched to transport" }, 197 { "dequeued", FMD_TYPE_UINT64, "total events dequeued by transport" }, 198 { "prdequeued", FMD_TYPE_UINT64, "protocol events dequeued by transport" }, 199 { "dropped", FMD_TYPE_UINT64, "total events dropped on queue overflow" }, 200 { "wcnt", FMD_TYPE_UINT32, "count of events waiting on queue" }, 201 { "wtime", FMD_TYPE_TIME, "total wait time on queue" }, 202 { "wlentime", FMD_TYPE_TIME, "total wait length * time product" }, 203 { "wlastupdate", FMD_TYPE_TIME, "hrtime of last wait queue update" }, 204 { "dtime", FMD_TYPE_TIME, "total processing time after dequeue" }, 205 { "dlastupdate", FMD_TYPE_TIME, "hrtime of last event dequeue completion" }, 206 }, 207 { "module", FMD_TYPE_STRING, "module that owns this transport" }, 208 { "authority", FMD_TYPE_STRING, "authority associated with this transport" }, 209 { "state", FMD_TYPE_STRING, "current transport state" }, 210 { "received", FMD_TYPE_UINT64, "events received by transport" }, 211 { "discarded", FMD_TYPE_UINT64, "bad events discarded by transport" }, 212 { "retried", FMD_TYPE_UINT64, "retries requested of transport" }, 213 { "replayed", FMD_TYPE_UINT64, "events replayed by transport" }, 214 { "lost", FMD_TYPE_UINT64, "events lost by transport" }, 215 { "timeouts", FMD_TYPE_UINT64, "events received by transport with ttl=0" }, 216 { "subscriptions", FMD_TYPE_UINT64, "subscriptions registered to transport" }, 217 }; 218 219 static void 220 fmd_xprt_class_hash_create(fmd_xprt_class_hash_t *xch, fmd_eventq_t *eq) 221 { 222 uint_t hashlen = fmd.d_str_buckets; 223 224 xch->xch_queue = eq; 225 xch->xch_hashlen = hashlen; 226 xch->xch_hash = fmd_zalloc(sizeof (void *) * hashlen, FMD_SLEEP); 227 } 228 229 static void 230 fmd_xprt_class_hash_destroy(fmd_xprt_class_hash_t *xch) 231 { 232 fmd_eventq_t *eq = xch->xch_queue; 233 fmd_xprt_class_t *xcp, *ncp; 234 uint_t i; 235 236 for (i = 0; i < xch->xch_hashlen; i++) { 237 for (xcp = xch->xch_hash[i]; xcp != NULL; xcp = ncp) { 238 ncp = xcp->xc_next; 239 240 if (eq != NULL) 241 fmd_dispq_delete(fmd.d_disp, eq, xcp->xc_class); 242 243 fmd_strfree(xcp->xc_class); 244 fmd_free(xcp, sizeof (fmd_xprt_class_t)); 245 } 246 } 247 248 fmd_free(xch->xch_hash, sizeof (void *) * xch->xch_hashlen); 249 } 250 251 /* 252 * Insert the specified class into the specified class hash, and return the 253 * reference count. A return value of one indicates this is the first insert. 254 * If an eventq is associated with the hash, insert a dispq subscription for it. 255 */ 256 static uint_t 257 fmd_xprt_class_hash_insert(fmd_xprt_impl_t *xip, 258 fmd_xprt_class_hash_t *xch, const char *class) 259 { 260 uint_t h = fmd_strhash(class) % xch->xch_hashlen; 261 fmd_xprt_class_t *xcp; 262 263 ASSERT(MUTEX_HELD(&xip->xi_lock)); 264 265 for (xcp = xch->xch_hash[h]; xcp != NULL; xcp = xcp->xc_next) { 266 if (strcmp(class, xcp->xc_class) == 0) 267 return (++xcp->xc_refs); 268 } 269 270 xcp = fmd_alloc(sizeof (fmd_xprt_class_t), FMD_SLEEP); 271 xcp->xc_class = fmd_strdup(class, FMD_SLEEP); 272 xcp->xc_next = xch->xch_hash[h]; 273 xcp->xc_refs = 1; 274 xch->xch_hash[h] = xcp; 275 276 if (xch->xch_queue != NULL) 277 fmd_dispq_insert(fmd.d_disp, xch->xch_queue, class); 278 279 return (xcp->xc_refs); 280 } 281 282 /* 283 * Delete the specified class from the specified class hash, and return the 284 * reference count. A return value of zero indicates the class was deleted. 285 * If an eventq is associated with the hash, delete the dispq subscription. 286 */ 287 static uint_t 288 fmd_xprt_class_hash_delete(fmd_xprt_impl_t *xip, 289 fmd_xprt_class_hash_t *xch, const char *class) 290 { 291 uint_t h = fmd_strhash(class) % xch->xch_hashlen; 292 fmd_xprt_class_t *xcp, **pp; 293 294 ASSERT(MUTEX_HELD(&xip->xi_lock)); 295 pp = &xch->xch_hash[h]; 296 297 for (xcp = *pp; xcp != NULL; xcp = xcp->xc_next) { 298 if (strcmp(class, xcp->xc_class) == 0) 299 break; 300 else 301 pp = &xcp->xc_next; 302 } 303 304 if (xcp == NULL) 305 return (-1U); /* explicitly permit an invalid delete */ 306 307 if (--xcp->xc_refs != 0) 308 return (xcp->xc_refs); 309 310 ASSERT(xcp->xc_refs == 0); 311 *pp = xcp->xc_next; 312 313 fmd_strfree(xcp->xc_class); 314 fmd_free(xcp, sizeof (fmd_xprt_class_t)); 315 316 if (xch->xch_queue != NULL) 317 fmd_dispq_delete(fmd.d_disp, xch->xch_queue, class); 318 319 return (0); 320 } 321 322 /* 323 * Queue subscribe events for the specified transport corresponding to all of 324 * the active module subscriptions. This is an extremely heavyweight operation 325 * that we expect to take place rarely (i.e. when loading a transport module 326 * or when it establishes a connection). We lock all of the known modules to 327 * prevent them from adding or deleting subscriptions, then snapshot their 328 * subscriptions, and then unlock all of the modules. We hold the modhash 329 * lock for the duration of this operation to prevent new modules from loading. 330 */ 331 static void 332 fmd_xprt_subscribe_modhash(fmd_xprt_impl_t *xip, fmd_modhash_t *mhp) 333 { 334 fmd_xprt_t *xp = (fmd_xprt_t *)xip; 335 const fmd_conf_path_t *pap; 336 fmd_module_t *mp; 337 uint_t i, j; 338 339 (void) pthread_rwlock_rdlock(&mhp->mh_lock); 340 341 for (i = 0; i < mhp->mh_hashlen; i++) { 342 for (mp = mhp->mh_hash[i]; mp != NULL; mp = mp->mod_next) 343 fmd_module_lock(mp); 344 } 345 346 (void) pthread_mutex_lock(&xip->xi_lock); 347 ASSERT(!(xip->xi_flags & FMD_XPRT_SUBSCRIBER)); 348 xip->xi_flags |= FMD_XPRT_SUBSCRIBER; 349 (void) pthread_mutex_unlock(&xip->xi_lock); 350 351 for (i = 0; i < mhp->mh_hashlen; i++) { 352 for (mp = mhp->mh_hash[i]; mp != NULL; mp = mp->mod_next) { 353 (void) fmd_conf_getprop(mp->mod_conf, 354 FMD_PROP_SUBSCRIPTIONS, &pap); 355 for (j = 0; j < pap->cpa_argc; j++) 356 fmd_xprt_subscribe(xp, pap->cpa_argv[j]); 357 } 358 } 359 360 for (i = 0; i < mhp->mh_hashlen; i++) { 361 for (mp = mhp->mh_hash[i]; mp != NULL; mp = mp->mod_next) 362 fmd_module_unlock(mp); 363 } 364 365 (void) pthread_rwlock_unlock(&mhp->mh_lock); 366 } 367 368 static void 369 fmd_xprt_transition(fmd_xprt_impl_t *xip, 370 const fmd_xprt_rule_t *state, const char *tag) 371 { 372 fmd_xprt_t *xp = (fmd_xprt_t *)xip; 373 fmd_event_t *e; 374 nvlist_t *nvl; 375 char *s; 376 377 TRACE((FMD_DBG_XPRT, "xprt %u -> %s\n", xip->xi_id, tag)); 378 379 xip->xi_state = state; 380 s = fmd_strdup(tag, FMD_SLEEP); 381 382 (void) pthread_mutex_lock(&xip->xi_stats_lock); 383 fmd_strfree(xip->xi_stats->xs_state.fmds_value.str); 384 xip->xi_stats->xs_state.fmds_value.str = s; 385 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 386 387 /* 388 * If we've reached the SUB state, take out the big hammer and snapshot 389 * all of the subscriptions of all of the loaded modules. Then queue a 390 * run event for our remote peer indicating that it can enter RUN. 391 */ 392 if (state == _fmd_xprt_state_sub) { 393 fmd_xprt_subscribe_modhash(xip, fmd.d_mod_hash); 394 395 /* 396 * For read-write transports, we always want to set up remote 397 * subscriptions to the bultin list.* events, regardless of 398 * whether any agents have subscribed to them. 399 */ 400 if (xip->xi_flags & FMD_XPRT_RDWR) { 401 fmd_xprt_subscribe(xp, FM_LIST_SUSPECT_CLASS); 402 fmd_xprt_subscribe(xp, FM_LIST_ISOLATED_CLASS); 403 fmd_xprt_subscribe(xp, FM_LIST_UPDATED_CLASS); 404 fmd_xprt_subscribe(xp, FM_LIST_RESOLVED_CLASS); 405 fmd_xprt_subscribe(xp, FM_LIST_REPAIRED_CLASS); 406 } 407 408 nvl = fmd_protocol_xprt_ctl(xip->xi_queue->eq_mod, 409 "resource.fm.xprt.run", xip->xi_version); 410 411 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 412 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 413 fmd_eventq_insert_at_time(xip->xi_queue, e); 414 } 415 } 416 417 static void 418 fmd_xprt_authupdate(fmd_xprt_impl_t *xip) 419 { 420 char *s = fmd_fmri_auth2str(xip->xi_auth); 421 422 (void) pthread_mutex_lock(&xip->xi_stats_lock); 423 fmd_strfree(xip->xi_stats->xs_authority.fmds_value.str); 424 xip->xi_stats->xs_authority.fmds_value.str = s; 425 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 426 } 427 428 static int 429 fmd_xprt_vmismatch(fmd_xprt_impl_t *xip, nvlist_t *nvl, uint_t *rversionp) 430 { 431 uint8_t rversion; 432 433 if (nvlist_lookup_uint8(nvl, FM_VERSION, &rversion) != 0) { 434 (void) pthread_mutex_lock(&xip->xi_stats_lock); 435 xip->xi_stats->xs_discarded.fmds_value.ui64++; 436 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 437 438 fmd_xprt_transition(xip, _fmd_xprt_state_err, "ERR"); 439 return (1); 440 } 441 442 if (rversion > xip->xi_version) { 443 fmd_dprintf(FMD_DBG_XPRT, "xprt %u protocol mismatch: %u>%u\n", 444 xip->xi_id, rversion, xip->xi_version); 445 446 (void) pthread_mutex_lock(&xip->xi_stats_lock); 447 xip->xi_stats->xs_discarded.fmds_value.ui64++; 448 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 449 450 fmd_xprt_transition(xip, _fmd_xprt_state_err, "ERR"); 451 return (1); 452 } 453 454 if (rversionp != NULL) 455 *rversionp = rversion; 456 457 return (0); 458 } 459 460 void 461 fmd_xprt_event_syn(fmd_xprt_impl_t *xip, nvlist_t *nvl) 462 { 463 fmd_event_t *e; 464 uint_t vers; 465 char *class; 466 467 if (fmd_xprt_vmismatch(xip, nvl, &vers)) 468 return; /* transitioned to error state */ 469 470 /* 471 * If the transport module didn't specify an authority, extract the 472 * one that is passed along with the xprt.syn event and use that. 473 */ 474 if (xip->xi_auth == NULL && 475 nvlist_lookup_nvlist(nvl, FM_RSRC_RESOURCE, &nvl) == 0 && 476 nvlist_lookup_nvlist(nvl, FM_FMRI_AUTHORITY, &nvl) == 0) { 477 (void) nvlist_xdup(nvl, &xip->xi_auth, &fmd.d_nva); 478 fmd_xprt_authupdate(xip); 479 } 480 481 nvl = fmd_protocol_xprt_ctl(xip->xi_queue->eq_mod, 482 "resource.fm.xprt.ack", xip->xi_version); 483 484 (void) nvlist_lookup_string(nvl, FM_CLASS, &class); 485 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, class); 486 fmd_eventq_insert_at_time(xip->xi_queue, e); 487 488 xip->xi_version = MIN(FM_RSRC_XPRT_VERSION, vers); 489 fmd_xprt_transition(xip, _fmd_xprt_state_sub, "SUB"); 490 } 491 492 void 493 fmd_xprt_event_ack(fmd_xprt_impl_t *xip, nvlist_t *nvl) 494 { 495 uint_t vers; 496 497 if (fmd_xprt_vmismatch(xip, nvl, &vers)) 498 return; /* transitioned to error state */ 499 500 /* 501 * If the transport module didn't specify an authority, extract the 502 * one that is passed along with the xprt.syn event and use that. 503 */ 504 if (xip->xi_auth == NULL && 505 nvlist_lookup_nvlist(nvl, FM_RSRC_RESOURCE, &nvl) == 0 && 506 nvlist_lookup_nvlist(nvl, FM_FMRI_AUTHORITY, &nvl) == 0) { 507 (void) nvlist_xdup(nvl, &xip->xi_auth, &fmd.d_nva); 508 fmd_xprt_authupdate(xip); 509 } 510 511 xip->xi_version = MIN(FM_RSRC_XPRT_VERSION, vers); 512 fmd_xprt_transition(xip, _fmd_xprt_state_sub, "SUB"); 513 } 514 515 /* 516 * Upon transition to RUN, we take every solved case and resend a list.suspect 517 * event for it to our remote peer. If a case transitions from solved to a 518 * future state (CLOSE_WAIT, CLOSED, or REPAIRED) while we are iterating over 519 * the case hash, we will get it as part of examining the resource cache, next. 520 */ 521 static void 522 fmd_xprt_send_case(fmd_case_t *cp, void *arg) 523 { 524 fmd_case_impl_t *cip = (fmd_case_impl_t *)cp; 525 fmd_xprt_impl_t *xip = arg; 526 527 fmd_event_t *e; 528 nvlist_t *nvl; 529 char *class; 530 531 if (cip->ci_state == FMD_CASE_UNSOLVED) 532 return; 533 534 nvl = fmd_case_mkevent(cp, FM_LIST_SUSPECT_CLASS); 535 (void) nvlist_lookup_string(nvl, FM_CLASS, &class); 536 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, class); 537 538 fmd_dprintf(FMD_DBG_XPRT, "re-send %s for %s to transport %u\n", 539 FM_LIST_SUSPECT_CLASS, cip->ci_uuid, xip->xi_id); 540 541 fmd_dispq_dispatch_gid(fmd.d_disp, e, class, xip->xi_queue->eq_sgid); 542 } 543 544 void 545 fmd_xprt_event_run(fmd_xprt_impl_t *xip, nvlist_t *nvl) 546 { 547 if (!fmd_xprt_vmismatch(xip, nvl, NULL)) { 548 fmd_xprt_transition(xip, _fmd_xprt_state_run, "RUN"); 549 fmd_case_hash_apply(fmd.d_cases, fmd_xprt_send_case, xip); 550 } 551 } 552 553 void 554 fmd_xprt_event_sub(fmd_xprt_impl_t *xip, nvlist_t *nvl) 555 { 556 char *class; 557 558 if (fmd_xprt_vmismatch(xip, nvl, NULL)) 559 return; /* transitioned to error state */ 560 561 if (nvlist_lookup_string(nvl, FM_RSRC_XPRT_SUBCLASS, &class) != 0) 562 return; /* malformed protocol event */ 563 564 (void) pthread_mutex_lock(&xip->xi_lock); 565 (void) fmd_xprt_class_hash_insert(xip, &xip->xi_lsub, class); 566 (void) pthread_mutex_unlock(&xip->xi_lock); 567 568 (void) pthread_mutex_lock(&xip->xi_stats_lock); 569 xip->xi_stats->xs_subscriptions.fmds_value.ui64++; 570 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 571 } 572 573 void 574 fmd_xprt_event_unsub(fmd_xprt_impl_t *xip, nvlist_t *nvl) 575 { 576 fmd_event_t *e; 577 char *class; 578 579 if (fmd_xprt_vmismatch(xip, nvl, NULL)) 580 return; /* transitioned to error state */ 581 582 if (nvlist_lookup_string(nvl, FM_RSRC_XPRT_SUBCLASS, &class) != 0) 583 return; /* malformed protocol event */ 584 585 (void) pthread_mutex_lock(&xip->xi_lock); 586 (void) fmd_xprt_class_hash_delete(xip, &xip->xi_lsub, class); 587 (void) pthread_mutex_unlock(&xip->xi_lock); 588 589 (void) pthread_mutex_lock(&xip->xi_stats_lock); 590 xip->xi_stats->xs_subscriptions.fmds_value.ui64--; 591 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 592 593 nvl = fmd_protocol_xprt_sub(xip->xi_queue->eq_mod, 594 "resource.fm.xprt.unsuback", xip->xi_version, class); 595 596 (void) nvlist_lookup_string(nvl, FM_CLASS, &class); 597 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, class); 598 fmd_eventq_insert_at_time(xip->xi_queue, e); 599 } 600 601 void 602 fmd_xprt_event_unsuback(fmd_xprt_impl_t *xip, nvlist_t *nvl) 603 { 604 char *class; 605 606 if (fmd_xprt_vmismatch(xip, nvl, NULL)) 607 return; /* transitioned to error state */ 608 609 if (nvlist_lookup_string(nvl, FM_RSRC_XPRT_SUBCLASS, &class) != 0) 610 return; /* malformed protocol event */ 611 612 (void) pthread_mutex_lock(&xip->xi_lock); 613 (void) fmd_xprt_class_hash_delete(xip, &xip->xi_usub, class); 614 (void) pthread_mutex_unlock(&xip->xi_lock); 615 } 616 617 /* 618 * on diagnosing side, receive a uuclose from the proxy. 619 */ 620 void 621 fmd_xprt_event_uuclose(fmd_xprt_impl_t *xip, nvlist_t *nvl) 622 { 623 fmd_case_t *cp; 624 char *uuid; 625 626 if (fmd_xprt_vmismatch(xip, nvl, NULL)) 627 return; /* transitioned to error state */ 628 629 if (nvlist_lookup_string(nvl, FM_RSRC_XPRT_UUID, &uuid) == 0 && 630 (cp = fmd_case_hash_lookup(fmd.d_cases, uuid)) != NULL) { 631 /* 632 * update resource cache status and transition case 633 */ 634 fmd_case_close_status(cp); 635 fmd_case_transition(cp, FMD_CASE_CLOSE_WAIT, FMD_CF_ISOLATED); 636 fmd_case_rele(cp); 637 } 638 } 639 640 /* 641 * on diagnosing side, receive a uuresolved from the proxy. 642 */ 643 void 644 fmd_xprt_event_uuresolved(fmd_xprt_impl_t *xip, nvlist_t *nvl) 645 { 646 fmd_case_t *cp; 647 char *uuid; 648 649 if (fmd_xprt_vmismatch(xip, nvl, NULL)) 650 return; /* transitioned to error state */ 651 652 if (nvlist_lookup_string(nvl, FM_RSRC_XPRT_UUID, &uuid) == 0 && 653 (cp = fmd_case_hash_lookup(fmd.d_cases, uuid)) != NULL) { 654 fmd_case_impl_t *cip = (fmd_case_impl_t *)cp; 655 656 fmd_case_transition(cp, (cip->ci_state == FMD_CASE_REPAIRED) ? 657 FMD_CASE_RESOLVED : (cip->ci_state == FMD_CASE_CLOSED) ? 658 FMD_CASE_REPAIRED : FMD_CASE_CLOSE_WAIT, FMD_CF_RESOLVED); 659 fmd_case_rele(cp); 660 } 661 } 662 663 /* 664 * on diagnosing side, receive a repair/acquit from the proxy. 665 */ 666 void 667 fmd_xprt_event_updated(fmd_xprt_impl_t *xip, nvlist_t *nvl) 668 { 669 fmd_case_t *cp; 670 char *uuid; 671 672 if (fmd_xprt_vmismatch(xip, nvl, NULL)) 673 return; /* transitioned to error state */ 674 675 if (nvlist_lookup_string(nvl, FM_RSRC_XPRT_UUID, &uuid) == 0 && 676 (cp = fmd_case_hash_lookup(fmd.d_cases, uuid)) != NULL) { 677 uint8_t *statusp, *proxy_asrup = NULL; 678 uint_t nelem = 0; 679 680 /* 681 * Only update status with new repairs if "no remote repair" 682 * is not set. Do the case_update anyway though (as this will 683 * refresh the status on the proxy side). 684 */ 685 if (!(xip->xi_flags & FMD_XPRT_NO_REMOTE_REPAIR)) { 686 if (nvlist_lookup_uint8_array(nvl, 687 FM_RSRC_XPRT_FAULT_STATUS, &statusp, &nelem) == 0 && 688 nelem != 0) { 689 (void) nvlist_lookup_uint8_array(nvl, 690 FM_RSRC_XPRT_FAULT_HAS_ASRU, &proxy_asrup, 691 &nelem); 692 fmd_case_update_status(cp, statusp, 693 proxy_asrup, NULL); 694 } 695 fmd_case_update_containees(cp); 696 } 697 fmd_case_update(cp); 698 fmd_case_rele(cp); 699 } 700 } 701 702 void 703 fmd_xprt_event_error(fmd_xprt_impl_t *xip, nvlist_t *nvl) 704 { 705 char *class = "<unknown>"; 706 707 (void) pthread_mutex_lock(&xip->xi_stats_lock); 708 xip->xi_stats->xs_discarded.fmds_value.ui64++; 709 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 710 711 (void) nvlist_lookup_string(nvl, FM_CLASS, &class); 712 TRACE((FMD_DBG_XPRT, "xprt %u bad event %s\n", xip->xi_id, class)); 713 714 fmd_xprt_transition(xip, _fmd_xprt_state_err, "ERR"); 715 } 716 717 void 718 fmd_xprt_event_drop(fmd_xprt_impl_t *xip, nvlist_t *nvl) 719 { 720 char *class = "<unknown>"; 721 722 (void) pthread_mutex_lock(&xip->xi_stats_lock); 723 xip->xi_stats->xs_discarded.fmds_value.ui64++; 724 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 725 726 (void) nvlist_lookup_string(nvl, FM_CLASS, &class); 727 TRACE((FMD_DBG_XPRT, "xprt %u drop event %s\n", xip->xi_id, class)); 728 729 } 730 731 fmd_xprt_t * 732 fmd_xprt_create(fmd_module_t *mp, uint_t flags, nvlist_t *auth, void *data) 733 { 734 fmd_xprt_impl_t *xip = fmd_zalloc(sizeof (fmd_xprt_impl_t), FMD_SLEEP); 735 fmd_stat_t *statv; 736 uint_t i, statc; 737 738 char buf[PATH_MAX]; 739 fmd_event_t *e; 740 nvlist_t *nvl; 741 char *s; 742 743 (void) pthread_mutex_init(&xip->xi_lock, NULL); 744 (void) pthread_cond_init(&xip->xi_cv, NULL); 745 (void) pthread_mutex_init(&xip->xi_stats_lock, NULL); 746 747 xip->xi_auth = auth; 748 xip->xi_data = data; 749 xip->xi_version = FM_RSRC_XPRT_VERSION; 750 xip->xi_flags = flags; 751 752 /* 753 * Grab fmd.d_xprt_lock to block fmd_xprt_suspend_all() and then create 754 * a transport ID and make it visible in fmd.d_xprt_ids. If transports 755 * were previously suspended, set the FMD_XPRT_DSUSPENDED flag on us to 756 * ensure that this transport will not run until fmd_xprt_resume_all(). 757 */ 758 (void) pthread_mutex_lock(&fmd.d_xprt_lock); 759 xip->xi_id = fmd_idspace_alloc(fmd.d_xprt_ids, xip); 760 761 if (fmd.d_xprt_suspend != 0) 762 xip->xi_flags |= FMD_XPRT_DSUSPENDED; 763 764 (void) pthread_mutex_unlock(&fmd.d_xprt_lock); 765 766 /* 767 * If the module has not yet finished _fmd_init(), set the ISUSPENDED 768 * bit so that fmdo_send() is not called until _fmd_init() completes. 769 */ 770 if (!(mp->mod_flags & FMD_MOD_INIT)) 771 xip->xi_flags |= FMD_XPRT_ISUSPENDED; 772 773 /* 774 * Initialize the transport statistics that we keep on behalf of fmd. 775 * These are set up using a template defined at the top of this file. 776 * We rename each statistic with a prefix ensuring its uniqueness. 777 */ 778 statc = sizeof (_fmd_xprt_stat_tmpl) / sizeof (fmd_stat_t); 779 statv = fmd_alloc(sizeof (_fmd_xprt_stat_tmpl), FMD_SLEEP); 780 bcopy(&_fmd_xprt_stat_tmpl, statv, sizeof (_fmd_xprt_stat_tmpl)); 781 782 for (i = 0; i < statc; i++) { 783 (void) snprintf(statv[i].fmds_name, 784 sizeof (statv[i].fmds_name), "fmd.xprt.%u.%s", xip->xi_id, 785 ((fmd_stat_t *)&_fmd_xprt_stat_tmpl + i)->fmds_name); 786 } 787 788 xip->xi_stats = (fmd_xprt_stat_t *)fmd_ustat_insert( 789 mp->mod_ustat, FMD_USTAT_NOALLOC, statc, statv, NULL); 790 791 if (xip->xi_stats == NULL) 792 fmd_panic("failed to create xi_stats (%p)\n", (void *)statv); 793 794 xip->xi_stats->xs_module.fmds_value.str = 795 fmd_strdup(mp->mod_name, FMD_SLEEP); 796 797 if (xip->xi_auth != NULL) 798 fmd_xprt_authupdate(xip); 799 800 /* 801 * Create the outbound eventq for this transport and link to its stats. 802 * If any suspend bits were set above, suspend the eventq immediately. 803 */ 804 xip->xi_queue = fmd_eventq_create(mp, &xip->xi_stats->xs_evqstat, 805 &xip->xi_stats_lock, mp->mod_stats->ms_xprtqlimit.fmds_value.ui32); 806 807 if (xip->xi_flags & FMD_XPRT_SMASK) 808 fmd_eventq_suspend(xip->xi_queue); 809 810 /* 811 * Create our subscription hashes: local subscriptions go to xi_queue, 812 * remote subscriptions are tracked only for protocol requests, and 813 * pending unsubscriptions are associated with the /dev/null eventq. 814 */ 815 fmd_xprt_class_hash_create(&xip->xi_lsub, xip->xi_queue); 816 fmd_xprt_class_hash_create(&xip->xi_rsub, NULL); 817 fmd_xprt_class_hash_create(&xip->xi_usub, fmd.d_rmod->mod_queue); 818 819 /* 820 * Determine our initial state based upon the creation flags. If we're 821 * read-only, go directly to RUN. If we're accepting a new connection, 822 * wait for a SYN. Otherwise send a SYN and wait for an ACK. 823 */ 824 if ((flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 825 fmd_xprt_transition(xip, _fmd_xprt_state_run, "RUN"); 826 else if (flags & FMD_XPRT_ACCEPT) 827 fmd_xprt_transition(xip, _fmd_xprt_state_syn, "SYN"); 828 else 829 fmd_xprt_transition(xip, _fmd_xprt_state_ack, "ACK"); 830 831 /* 832 * If client.xprtlog is set to TRUE, create a debugging log for the 833 * events received by the transport in var/fm/fmd/xprt/. 834 */ 835 (void) fmd_conf_getprop(fmd.d_conf, "client.xprtlog", &i); 836 (void) fmd_conf_getprop(fmd.d_conf, "log.xprt", &s); 837 838 if (i) { 839 (void) snprintf(buf, sizeof (buf), "%s/%u.log", s, xip->xi_id); 840 xip->xi_log = fmd_log_open(fmd.d_rootdir, buf, FMD_LOG_XPRT); 841 } 842 843 ASSERT(fmd_module_locked(mp)); 844 fmd_list_append(&mp->mod_transports, xip); 845 846 (void) pthread_mutex_lock(&mp->mod_stats_lock); 847 mp->mod_stats->ms_xprtopen.fmds_value.ui32++; 848 (void) pthread_mutex_unlock(&mp->mod_stats_lock); 849 850 /* 851 * If this is a read-only transport, return without creating a send 852 * queue thread and setting up any connection events in our queue. 853 */ 854 if ((flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 855 goto out; 856 857 /* 858 * Once the transport is fully initialized, create a send queue thread 859 * and start any connect events flowing to complete our initialization. 860 */ 861 if ((xip->xi_thread = fmd_thread_create(mp, 862 (fmd_thread_f *)fmd_xprt_send, xip)) == NULL) { 863 864 fmd_error(EFMD_XPRT_THR, 865 "failed to create thread for transport %u", xip->xi_id); 866 867 fmd_xprt_destroy((fmd_xprt_t *)xip); 868 (void) fmd_set_errno(EFMD_XPRT_THR); 869 return (NULL); 870 } 871 872 /* 873 * If the transport is not being opened to accept an inbound connect, 874 * start an outbound connection by enqueuing a SYN event for our peer. 875 */ 876 if (!(flags & FMD_XPRT_ACCEPT)) { 877 nvl = fmd_protocol_xprt_ctl(mp, 878 "resource.fm.xprt.syn", FM_RSRC_XPRT_VERSION); 879 880 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 881 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 882 fmd_eventq_insert_at_time(xip->xi_queue, e); 883 } 884 out: 885 fmd_dprintf(FMD_DBG_XPRT, "opened transport %u\n", xip->xi_id); 886 return ((fmd_xprt_t *)xip); 887 } 888 889 void 890 fmd_xprt_destroy(fmd_xprt_t *xp) 891 { 892 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 893 fmd_module_t *mp = xip->xi_queue->eq_mod; 894 uint_t id = xip->xi_id; 895 896 fmd_case_impl_t *cip, *nip; 897 fmd_stat_t *sp; 898 uint_t i, n; 899 900 ASSERT(fmd_module_locked(mp)); 901 fmd_list_delete(&mp->mod_transports, xip); 902 903 (void) pthread_mutex_lock(&mp->mod_stats_lock); 904 mp->mod_stats->ms_xprtopen.fmds_value.ui32--; 905 (void) pthread_mutex_unlock(&mp->mod_stats_lock); 906 907 (void) pthread_mutex_lock(&xip->xi_lock); 908 909 while (xip->xi_busy != 0) 910 (void) pthread_cond_wait(&xip->xi_cv, &xip->xi_lock); 911 912 /* 913 * Remove the transport from global visibility, cancel its send-side 914 * thread, join with it, and then remove the transport from module 915 * visibility. Once all this is done, destroy and free the transport. 916 */ 917 (void) fmd_idspace_free(fmd.d_xprt_ids, xip->xi_id); 918 919 if (xip->xi_thread != NULL) { 920 fmd_eventq_abort(xip->xi_queue); 921 fmd_module_unlock(mp); 922 fmd_thread_destroy(xip->xi_thread, FMD_THREAD_JOIN); 923 fmd_module_lock(mp); 924 } 925 926 if (xip->xi_log != NULL) 927 fmd_log_rele(xip->xi_log); 928 929 /* 930 * Release every case handle in the module that was cached by this 931 * transport. This will result in these cases disappearing from the 932 * local case hash so that fmd_case_uuclose() and fmd_case_repaired() 933 * etc can no longer be used. 934 */ 935 for (cip = fmd_list_next(&mp->mod_cases); cip != NULL; cip = nip) { 936 nip = fmd_list_next(cip); 937 if (cip->ci_xprt == xp) 938 fmd_case_discard((fmd_case_t *)cip, B_TRUE); 939 } 940 941 /* 942 * Destroy every class in the various subscription hashes and remove 943 * any corresponding subscriptions from the event dispatch queue. 944 */ 945 fmd_xprt_class_hash_destroy(&xip->xi_lsub); 946 fmd_xprt_class_hash_destroy(&xip->xi_rsub); 947 fmd_xprt_class_hash_destroy(&xip->xi_usub); 948 949 /* 950 * Uniquify the stat names exactly as was done in fmd_xprt_create() 951 * before calling fmd_ustat_insert(), otherwise fmd_ustat_delete() 952 * won't find the entries in the hash table. 953 */ 954 n = sizeof (_fmd_xprt_stat_tmpl) / sizeof (fmd_stat_t); 955 sp = fmd_alloc(sizeof (_fmd_xprt_stat_tmpl), FMD_SLEEP); 956 bcopy(&_fmd_xprt_stat_tmpl, sp, sizeof (_fmd_xprt_stat_tmpl)); 957 for (i = 0; i < n; i++) { 958 (void) snprintf(sp[i].fmds_name, 959 sizeof (sp[i].fmds_name), "fmd.xprt.%u.%s", xip->xi_id, 960 ((fmd_stat_t *)&_fmd_xprt_stat_tmpl + i)->fmds_name); 961 } 962 fmd_ustat_delete(mp->mod_ustat, n, sp); 963 fmd_free(sp, sizeof (_fmd_xprt_stat_tmpl)); 964 965 fmd_free(xip->xi_stats, sizeof (fmd_xprt_stat_t)); 966 fmd_eventq_destroy(xip->xi_queue); 967 nvlist_free(xip->xi_auth); 968 fmd_free(xip, sizeof (fmd_xprt_impl_t)); 969 970 fmd_dprintf(FMD_DBG_XPRT, "closed transport %u\n", id); 971 } 972 973 void 974 fmd_xprt_xsuspend(fmd_xprt_t *xp, uint_t flags) 975 { 976 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 977 uint_t oflags; 978 979 ASSERT((flags & ~FMD_XPRT_SMASK) == 0); 980 (void) pthread_mutex_lock(&xip->xi_lock); 981 982 oflags = xip->xi_flags; 983 xip->xi_flags |= flags; 984 985 if (!(oflags & FMD_XPRT_SMASK) && (xip->xi_flags & FMD_XPRT_SMASK) != 0) 986 fmd_eventq_suspend(xip->xi_queue); 987 988 (void) pthread_cond_broadcast(&xip->xi_cv); 989 990 while (xip->xi_busy != 0) 991 (void) pthread_cond_wait(&xip->xi_cv, &xip->xi_lock); 992 993 (void) pthread_mutex_unlock(&xip->xi_lock); 994 } 995 996 void 997 fmd_xprt_xresume(fmd_xprt_t *xp, uint_t flags) 998 { 999 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1000 uint_t oflags; 1001 1002 ASSERT((flags & ~FMD_XPRT_SMASK) == 0); 1003 (void) pthread_mutex_lock(&xip->xi_lock); 1004 1005 oflags = xip->xi_flags; 1006 xip->xi_flags &= ~flags; 1007 1008 if ((oflags & FMD_XPRT_SMASK) != 0 && !(xip->xi_flags & FMD_XPRT_SMASK)) 1009 fmd_eventq_resume(xip->xi_queue); 1010 1011 (void) pthread_cond_broadcast(&xip->xi_cv); 1012 (void) pthread_mutex_unlock(&xip->xi_lock); 1013 } 1014 1015 void 1016 fmd_xprt_send(fmd_xprt_t *xp) 1017 { 1018 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1019 fmd_module_t *mp = xip->xi_queue->eq_mod; 1020 fmd_event_t *ep; 1021 int err; 1022 1023 while ((ep = fmd_eventq_delete(xip->xi_queue)) != NULL) { 1024 if (FMD_EVENT_TTL(ep) == 0) { 1025 fmd_event_rele(ep); 1026 continue; 1027 } 1028 1029 fmd_dprintf(FMD_DBG_XPRT, "xprt %u sending %s\n", 1030 xip->xi_id, (char *)FMD_EVENT_DATA(ep)); 1031 1032 err = mp->mod_ops->mop_transport(mp, xp, ep); 1033 fmd_eventq_done(xip->xi_queue); 1034 1035 if (err == FMD_SEND_RETRY) { 1036 fmd_eventq_insert_at_time(xip->xi_queue, ep); 1037 (void) pthread_mutex_lock(&xip->xi_stats_lock); 1038 xip->xi_stats->xs_retried.fmds_value.ui64++; 1039 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 1040 } 1041 1042 if (err != FMD_SEND_SUCCESS && err != FMD_SEND_RETRY) { 1043 (void) pthread_mutex_lock(&xip->xi_stats_lock); 1044 xip->xi_stats->xs_lost.fmds_value.ui64++; 1045 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 1046 } 1047 1048 fmd_event_rele(ep); 1049 } 1050 } 1051 1052 /* 1053 * This function creates a local suspect list. This is used when a suspect list 1054 * is created directly by an external source like fminject. 1055 */ 1056 static void 1057 fmd_xprt_list_suspect_local(fmd_xprt_t *xp, nvlist_t *nvl) 1058 { 1059 nvlist_t **nvlp; 1060 nvlist_t *de_fmri, *de_fmri_dup = NULL; 1061 int64_t *diag_time; 1062 char *code = NULL; 1063 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1064 fmd_case_t *cp; 1065 uint_t nelem = 0, nelem2 = 0, i; 1066 1067 fmd_module_lock(xip->xi_queue->eq_mod); 1068 cp = fmd_case_create(xip->xi_queue->eq_mod, NULL); 1069 if (cp == NULL) { 1070 fmd_module_unlock(xip->xi_queue->eq_mod); 1071 return; 1072 } 1073 1074 /* 1075 * copy diag_code if present 1076 */ 1077 (void) nvlist_lookup_string(nvl, FM_SUSPECT_DIAG_CODE, &code); 1078 if (code != NULL) { 1079 fmd_case_impl_t *cip = (fmd_case_impl_t *)cp; 1080 1081 cip->ci_precanned = 1; 1082 fmd_case_setcode(cp, code); 1083 } 1084 1085 /* 1086 * copy suspects 1087 */ 1088 (void) nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, &nvlp, 1089 &nelem); 1090 for (i = 0; i < nelem; i++) { 1091 nvlist_t *flt_copy, *asru = NULL, *fru = NULL, *rsrc = NULL; 1092 topo_hdl_t *thp; 1093 char *loc = NULL; 1094 int err; 1095 1096 thp = fmd_fmri_topo_hold(TOPO_VERSION); 1097 (void) nvlist_xdup(nvlp[i], &flt_copy, &fmd.d_nva); 1098 (void) nvlist_lookup_nvlist(nvlp[i], FM_FAULT_RESOURCE, &rsrc); 1099 1100 /* 1101 * If no fru specified, get it from topo 1102 */ 1103 if (nvlist_lookup_nvlist(nvlp[i], FM_FAULT_FRU, &fru) != 0 && 1104 rsrc && topo_fmri_fru(thp, rsrc, &fru, &err) == 0) 1105 (void) nvlist_add_nvlist(flt_copy, FM_FAULT_FRU, fru); 1106 /* 1107 * If no asru specified, get it from topo 1108 */ 1109 if (nvlist_lookup_nvlist(nvlp[i], FM_FAULT_ASRU, &asru) != 0 && 1110 rsrc && topo_fmri_asru(thp, rsrc, &asru, &err) == 0) 1111 (void) nvlist_add_nvlist(flt_copy, FM_FAULT_ASRU, asru); 1112 /* 1113 * If no location specified, get it from topo 1114 */ 1115 if (nvlist_lookup_string(nvlp[i], FM_FAULT_LOCATION, 1116 &loc) != 0) { 1117 if (fru && topo_fmri_label(thp, fru, &loc, &err) == 0) 1118 (void) nvlist_add_string(flt_copy, 1119 FM_FAULT_LOCATION, loc); 1120 else if (rsrc && topo_fmri_label(thp, rsrc, &loc, 1121 &err) == 0) 1122 (void) nvlist_add_string(flt_copy, 1123 FM_FAULT_LOCATION, loc); 1124 if (loc) 1125 topo_hdl_strfree(thp, loc); 1126 } 1127 if (fru) 1128 nvlist_free(fru); 1129 if (asru) 1130 nvlist_free(asru); 1131 if (rsrc) 1132 nvlist_free(rsrc); 1133 fmd_fmri_topo_rele(thp); 1134 fmd_case_insert_suspect(cp, flt_copy); 1135 } 1136 1137 /* 1138 * copy diag_time if present 1139 */ 1140 if (nvlist_lookup_int64_array(nvl, FM_SUSPECT_DIAG_TIME, &diag_time, 1141 &nelem2) == 0 && nelem2 >= 2) 1142 fmd_case_settime(cp, diag_time[0], diag_time[1]); 1143 1144 /* 1145 * copy DE fmri if present 1146 */ 1147 if (nvlist_lookup_nvlist(nvl, FM_SUSPECT_DE, &de_fmri) == 0) { 1148 (void) nvlist_xdup(de_fmri, &de_fmri_dup, &fmd.d_nva); 1149 fmd_case_set_de_fmri(cp, de_fmri_dup); 1150 } 1151 1152 fmd_case_transition(cp, FMD_CASE_SOLVED, FMD_CF_SOLVED); 1153 fmd_module_unlock(xip->xi_queue->eq_mod); 1154 } 1155 1156 /* 1157 * This function is called to create a proxy case on receipt of a list.suspect 1158 * from the diagnosing side of the transport. 1159 */ 1160 static void 1161 fmd_xprt_list_suspect(fmd_xprt_t *xp, nvlist_t *nvl) 1162 { 1163 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1164 nvlist_t **nvlp; 1165 uint_t nelem = 0, nelem2 = 0, i; 1166 int64_t *diag_time; 1167 topo_hdl_t *thp; 1168 char *class; 1169 nvlist_t *rsrc, *asru, *de_fmri, *de_fmri_dup = NULL; 1170 nvlist_t *flt_copy; 1171 int err; 1172 nvlist_t **asrua; 1173 uint8_t *proxy_asru = NULL; 1174 int got_proxy_asru = 0; 1175 int got_hc_rsrc = 0; 1176 int got_present_rsrc = 0; 1177 uint8_t *diag_asru = NULL; 1178 char *scheme; 1179 uint8_t *statusp; 1180 char *uuid, *code; 1181 fmd_case_t *cp; 1182 fmd_case_impl_t *cip; 1183 int need_update = 0; 1184 1185 if (nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) != 0) 1186 return; 1187 if (nvlist_lookup_string(nvl, FM_SUSPECT_DIAG_CODE, &code) != 0) 1188 return; 1189 (void) nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, &nvlp, 1190 &nelem); 1191 1192 /* 1193 * In order to implement FMD_XPRT_HCONLY and FMD_XPRT_HC_PRESENT_ONLY 1194 * etc we first scan the suspects to see if 1195 * - there was an asru in the received fault 1196 * - there was an hc-scheme resource in the received fault 1197 * - any hc-scheme resource in the received fault is present in the 1198 * local topology 1199 * - any hc-scheme resource in the received fault has an asru in the 1200 * local topology 1201 */ 1202 if (nelem > 0) { 1203 asrua = fmd_zalloc(sizeof (nvlist_t *) * nelem, FMD_SLEEP); 1204 proxy_asru = fmd_zalloc(sizeof (uint8_t) * nelem, FMD_SLEEP); 1205 diag_asru = fmd_zalloc(sizeof (uint8_t) * nelem, FMD_SLEEP); 1206 thp = fmd_fmri_topo_hold(TOPO_VERSION); 1207 for (i = 0; i < nelem; i++) { 1208 if (nvlist_lookup_nvlist(nvlp[i], FM_FAULT_ASRU, 1209 &asru) == 0 && asru != NULL) 1210 diag_asru[i] = 1; 1211 if (nvlist_lookup_string(nvlp[i], FM_CLASS, 1212 &class) != 0 || strncmp(class, "fault", 5) != 0) 1213 continue; 1214 /* 1215 * If there is an hc-scheme asru, use that to find the 1216 * real asru. Otherwise if there is an hc-scheme 1217 * resource, work out the old asru from that. 1218 * This order is to allow a two stage evaluation 1219 * of the asru where a fault in the diagnosing side 1220 * is in a component not visible to the proxy side, 1221 * but prevents a component that is visible from 1222 * working. So the diagnosing side sets the asru to 1223 * the latter component (in hc-scheme as the diagnosing 1224 * side doesn't know about the proxy side's virtual 1225 * schemes), and then the proxy side can convert that 1226 * to a suitable virtual scheme asru. 1227 */ 1228 if (nvlist_lookup_nvlist(nvlp[i], FM_FAULT_ASRU, 1229 &asru) == 0 && asru != NULL && 1230 nvlist_lookup_string(asru, FM_FMRI_SCHEME, 1231 &scheme) == 0 && 1232 strcmp(scheme, FM_FMRI_SCHEME_HC) == 0) { 1233 got_hc_rsrc = 1; 1234 if (xip->xi_flags & FMD_XPRT_EXTERNAL) 1235 continue; 1236 if (topo_fmri_present(thp, asru, &err) == 0) 1237 got_present_rsrc = 1; 1238 if (topo_fmri_asru(thp, asru, &asrua[i], 1239 &err) == 0) { 1240 proxy_asru[i] = 1241 FMD_PROXY_ASRU_FROM_ASRU; 1242 got_proxy_asru = 1; 1243 } 1244 } else if (nvlist_lookup_nvlist(nvlp[i], 1245 FM_FAULT_RESOURCE, &rsrc) == 0 && rsrc != NULL && 1246 nvlist_lookup_string(rsrc, FM_FMRI_SCHEME, 1247 &scheme) == 0 && 1248 strcmp(scheme, FM_FMRI_SCHEME_HC) == 0) { 1249 got_hc_rsrc = 1; 1250 if (xip->xi_flags & FMD_XPRT_EXTERNAL) 1251 continue; 1252 if (topo_fmri_present(thp, rsrc, &err) == 0) 1253 got_present_rsrc = 1; 1254 if (topo_fmri_asru(thp, rsrc, &asrua[i], 1255 &err) == 0) { 1256 proxy_asru[i] = 1257 FMD_PROXY_ASRU_FROM_RSRC; 1258 got_proxy_asru = 1; 1259 } 1260 } 1261 } 1262 fmd_fmri_topo_rele(thp); 1263 } 1264 1265 /* 1266 * If we're set up only to report hc-scheme faults, and 1267 * there aren't any, then just drop the event. 1268 */ 1269 if (got_hc_rsrc == 0 && (xip->xi_flags & FMD_XPRT_HCONLY)) { 1270 if (nelem > 0) { 1271 fmd_free(proxy_asru, sizeof (uint8_t) * nelem); 1272 fmd_free(diag_asru, sizeof (uint8_t) * nelem); 1273 fmd_free(asrua, sizeof (nvlist_t *) * nelem); 1274 } 1275 return; 1276 } 1277 1278 /* 1279 * If we're set up only to report locally present hc-scheme 1280 * faults, and there aren't any, then just drop the event. 1281 */ 1282 if (got_present_rsrc == 0 && 1283 (xip->xi_flags & FMD_XPRT_HC_PRESENT_ONLY)) { 1284 if (nelem > 0) { 1285 for (i = 0; i < nelem; i++) 1286 if (asrua[i]) 1287 nvlist_free(asrua[i]); 1288 fmd_free(proxy_asru, sizeof (uint8_t) * nelem); 1289 fmd_free(diag_asru, sizeof (uint8_t) * nelem); 1290 fmd_free(asrua, sizeof (nvlist_t *) * nelem); 1291 } 1292 return; 1293 } 1294 1295 /* 1296 * If fmd_case_recreate() returns NULL, UUID is already known. 1297 */ 1298 fmd_module_lock(xip->xi_queue->eq_mod); 1299 if ((cp = fmd_case_recreate(xip->xi_queue->eq_mod, xp, 1300 FMD_CASE_UNSOLVED, uuid, code)) == NULL) { 1301 if (nelem > 0) { 1302 for (i = 0; i < nelem; i++) 1303 if (asrua[i]) 1304 nvlist_free(asrua[i]); 1305 fmd_free(proxy_asru, sizeof (uint8_t) * nelem); 1306 fmd_free(diag_asru, sizeof (uint8_t) * nelem); 1307 fmd_free(asrua, sizeof (nvlist_t *) * nelem); 1308 } 1309 fmd_module_unlock(xip->xi_queue->eq_mod); 1310 return; 1311 } 1312 1313 cip = (fmd_case_impl_t *)cp; 1314 cip->ci_diag_asru = diag_asru; 1315 cip->ci_proxy_asru = proxy_asru; 1316 for (i = 0; i < nelem; i++) { 1317 (void) nvlist_xdup(nvlp[i], &flt_copy, &fmd.d_nva); 1318 if (proxy_asru[i] != FMD_PROXY_ASRU_NOT_NEEDED) { 1319 /* 1320 * Copy suspects, but remove/replace asru first. Also if 1321 * the original asru was hc-scheme use that as resource. 1322 */ 1323 if (proxy_asru[i] == FMD_PROXY_ASRU_FROM_ASRU) { 1324 (void) nvlist_remove(flt_copy, 1325 FM_FAULT_RESOURCE, DATA_TYPE_NVLIST); 1326 (void) nvlist_lookup_nvlist(flt_copy, 1327 FM_FAULT_ASRU, &asru); 1328 (void) nvlist_add_nvlist(flt_copy, 1329 FM_FAULT_RESOURCE, asru); 1330 } 1331 (void) nvlist_remove(flt_copy, FM_FAULT_ASRU, 1332 DATA_TYPE_NVLIST); 1333 (void) nvlist_add_nvlist(flt_copy, FM_FAULT_ASRU, 1334 asrua[i]); 1335 nvlist_free(asrua[i]); 1336 } else if (nvlist_lookup_nvlist(flt_copy, FM_FAULT_ASRU, 1337 &asru) == 0 && asru != NULL) { 1338 /* 1339 * keep asru from diag side, but but mark as no retire 1340 */ 1341 (void) nvlist_add_boolean_value(flt_copy, 1342 FM_SUSPECT_RETIRE, B_FALSE); 1343 } 1344 fmd_case_insert_suspect(cp, flt_copy); 1345 } 1346 /* 1347 * copy diag_time 1348 */ 1349 if (nvlist_lookup_int64_array(nvl, FM_SUSPECT_DIAG_TIME, &diag_time, 1350 &nelem2) == 0 && nelem2 >= 2) 1351 fmd_case_settime(cp, diag_time[0], diag_time[1]); 1352 /* 1353 * copy DE fmri 1354 */ 1355 if (nvlist_lookup_nvlist(nvl, FM_SUSPECT_DE, &de_fmri) == 0) { 1356 (void) nvlist_xdup(de_fmri, &de_fmri_dup, &fmd.d_nva); 1357 fmd_case_set_de_fmri(cp, de_fmri_dup); 1358 } 1359 1360 /* 1361 * Transition to solved. This will log the suspect list and create 1362 * the resource cache entries. 1363 */ 1364 fmd_case_transition(cp, FMD_CASE_SOLVED, FMD_CF_SOLVED); 1365 1366 /* 1367 * Update status if it is not simply "all faulty" (can happen if 1368 * list.suspects are being re-sent when the transport has reconnected). 1369 */ 1370 (void) nvlist_lookup_uint8_array(nvl, FM_SUSPECT_FAULT_STATUS, &statusp, 1371 &nelem); 1372 for (i = 0; i < nelem; i++) { 1373 if ((statusp[i] & (FM_SUSPECT_FAULTY | FM_SUSPECT_UNUSABLE | 1374 FM_SUSPECT_NOT_PRESENT | FM_SUSPECT_DEGRADED)) != 1375 FM_SUSPECT_FAULTY) 1376 need_update = 1; 1377 } 1378 if (need_update) { 1379 fmd_case_update_status(cp, statusp, cip->ci_proxy_asru, 1380 cip->ci_diag_asru); 1381 fmd_case_update_containees(cp); 1382 fmd_case_update(cp); 1383 } 1384 1385 /* 1386 * if asru on proxy side, send an update back to the diagnosing side to 1387 * update UNUSABLE/DEGRADED. 1388 */ 1389 if (got_proxy_asru) 1390 fmd_case_xprt_updated(cp); 1391 1392 if (nelem > 0) 1393 fmd_free(asrua, sizeof (nvlist_t *) * nelem); 1394 fmd_module_unlock(xip->xi_queue->eq_mod); 1395 } 1396 1397 void 1398 fmd_xprt_recv(fmd_xprt_t *xp, nvlist_t *nvl, hrtime_t hrt, boolean_t logonly) 1399 { 1400 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1401 const fmd_xprt_rule_t *xrp; 1402 fmd_t *dp = &fmd; 1403 1404 fmd_event_t *e; 1405 char *class, *uuid; 1406 boolean_t isproto, isereport; 1407 1408 uint64_t *tod; 1409 uint8_t ttl; 1410 uint_t n; 1411 fmd_case_t *cp; 1412 1413 /* 1414 * Grab the transport lock and set the busy flag to indicate we are 1415 * busy receiving an event. If [DI]SUSPEND is pending, wait until fmd 1416 * resumes the transport before continuing on with the receive. 1417 */ 1418 (void) pthread_mutex_lock(&xip->xi_lock); 1419 1420 while (xip->xi_flags & (FMD_XPRT_DSUSPENDED | FMD_XPRT_ISUSPENDED)) { 1421 1422 if (fmd.d_signal != 0) { 1423 (void) pthread_mutex_unlock(&xip->xi_lock); 1424 return; /* fmd_destroy() is in progress */ 1425 } 1426 1427 (void) pthread_cond_wait(&xip->xi_cv, &xip->xi_lock); 1428 } 1429 1430 xip->xi_busy++; 1431 ASSERT(xip->xi_busy != 0); 1432 1433 (void) pthread_mutex_unlock(&xip->xi_lock); 1434 1435 (void) pthread_mutex_lock(&xip->xi_stats_lock); 1436 xip->xi_stats->xs_received.fmds_value.ui64++; 1437 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 1438 1439 if (nvlist_lookup_string(nvl, FM_CLASS, &class) != 0) { 1440 fmd_error(EFMD_XPRT_PAYLOAD, "discarding nvlist %p: missing " 1441 "required \"%s\" payload element", (void *)nvl, FM_CLASS); 1442 1443 (void) pthread_mutex_lock(&xip->xi_stats_lock); 1444 xip->xi_stats->xs_discarded.fmds_value.ui64++; 1445 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 1446 1447 nvlist_free(nvl); 1448 goto done; 1449 } 1450 1451 fmd_dprintf(FMD_DBG_XPRT, "xprt %u %s %s\n", xip->xi_id, 1452 ((logonly == FMD_B_TRUE) ? "logging" : "posting"), class); 1453 1454 isereport = (strncmp(class, FM_EREPORT_CLASS, 1455 sizeof (FM_EREPORT_CLASS - 1)) == 0) ? FMD_B_TRUE : FMD_B_FALSE; 1456 1457 /* 1458 * The logonly flag should only be set for ereports. 1459 */ 1460 if ((logonly == FMD_B_TRUE) && (isereport == FMD_B_FALSE)) { 1461 fmd_error(EFMD_XPRT_INVAL, "discarding nvlist %p: " 1462 "logonly flag is not valid for class %s", 1463 (void *)nvl, class); 1464 1465 (void) pthread_mutex_lock(&xip->xi_stats_lock); 1466 xip->xi_stats->xs_discarded.fmds_value.ui64++; 1467 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 1468 1469 nvlist_free(nvl); 1470 goto done; 1471 } 1472 1473 /* 1474 * If a time-to-live value is present in the event and is zero, drop 1475 * the event and bump xs_timeouts. Otherwise decrement the TTL value. 1476 */ 1477 if (nvlist_lookup_uint8(nvl, FMD_EVN_TTL, &ttl) == 0) { 1478 if (ttl == 0) { 1479 fmd_dprintf(FMD_DBG_XPRT, "xprt %u nvlist %p (%s) " 1480 "timeout: event received with ttl=0\n", 1481 xip->xi_id, (void *)nvl, class); 1482 1483 (void) pthread_mutex_lock(&xip->xi_stats_lock); 1484 xip->xi_stats->xs_timeouts.fmds_value.ui64++; 1485 (void) pthread_mutex_unlock(&xip->xi_stats_lock); 1486 1487 nvlist_free(nvl); 1488 goto done; 1489 } 1490 (void) nvlist_remove(nvl, FMD_EVN_TTL, DATA_TYPE_UINT8); 1491 (void) nvlist_add_uint8(nvl, FMD_EVN_TTL, ttl - 1); 1492 } 1493 1494 /* 1495 * If we are using the native system clock, the underlying transport 1496 * code can provide a tighter event time bound by telling us when the 1497 * event was enqueued. If we're using simulated clocks, this time 1498 * has no meaning to us, so just reset the value to use HRT_NOW. 1499 */ 1500 if (dp->d_clockops != &fmd_timeops_native) 1501 hrt = FMD_HRT_NOW; 1502 1503 /* 1504 * If an event's class is in the FMD_CTL_CLASS family, then create a 1505 * control event. If a FMD_EVN_TOD member is found, create a protocol 1506 * event using this time. Otherwise create a protocol event using hrt. 1507 */ 1508 isproto = (strncmp(class, FMD_CTL_CLASS, FMD_CTL_CLASS_LEN) == 0) ? 1509 FMD_B_FALSE : FMD_B_TRUE; 1510 if (isproto == FMD_B_FALSE) 1511 e = fmd_event_create(FMD_EVT_CTL, hrt, nvl, fmd_ctl_init(nvl)); 1512 else if (nvlist_lookup_uint64_array(nvl, FMD_EVN_TOD, &tod, &n) != 0) 1513 e = fmd_event_create(FMD_EVT_PROTOCOL, hrt, nvl, class); 1514 else { 1515 e = fmd_event_recreate(FMD_EVT_PROTOCOL, 1516 NULL, nvl, class, NULL, 0, 0); 1517 } 1518 1519 /* 1520 * If the debug log is enabled, create a temporary event, log it to the 1521 * debug log, and then reset the underlying state of the event. 1522 */ 1523 if (xip->xi_log != NULL) { 1524 fmd_event_impl_t *ep = (fmd_event_impl_t *)e; 1525 1526 fmd_log_append(xip->xi_log, e, NULL); 1527 1528 ep->ev_flags |= FMD_EVF_VOLATILE; 1529 ep->ev_off = 0; 1530 ep->ev_len = 0; 1531 1532 if (ep->ev_log != NULL) { 1533 fmd_log_rele(ep->ev_log); 1534 ep->ev_log = NULL; 1535 } 1536 } 1537 1538 /* 1539 * Iterate over the rules for the current state trying to match the 1540 * event class to one of our special rules. If a rule is matched, the 1541 * event is consumed and not dispatched to other modules. If the rule 1542 * set ends without matching an event, we fall through to dispatching. 1543 */ 1544 for (xrp = xip->xi_state; xrp->xr_class != NULL; xrp++) { 1545 if (fmd_event_match(e, FMD_EVT_PROTOCOL, xrp->xr_class)) { 1546 fmd_event_hold(e); 1547 xrp->xr_func(xip, nvl); 1548 fmd_event_rele(e); 1549 goto done; 1550 } 1551 } 1552 1553 /* 1554 * Record the event in the errlog if it is an ereport. This code will 1555 * be replaced later with a per-transport intent log instead. 1556 */ 1557 if (isereport == FMD_B_TRUE) { 1558 (void) pthread_rwlock_rdlock(&dp->d_log_lock); 1559 fmd_log_append(dp->d_errlog, e, NULL); 1560 (void) pthread_rwlock_unlock(&dp->d_log_lock); 1561 } 1562 1563 /* 1564 * If a list.suspect event is received, create a case for the specified 1565 * UUID in the case hash, with the transport module as its owner. 1566 */ 1567 if (fmd_event_match(e, FMD_EVT_PROTOCOL, FM_LIST_SUSPECT_CLASS)) { 1568 if (xip->xi_flags & FMD_XPRT_CACHE_AS_LOCAL) 1569 fmd_xprt_list_suspect_local(xp, nvl); 1570 else 1571 fmd_xprt_list_suspect(xp, nvl); 1572 fmd_event_hold(e); 1573 fmd_event_rele(e); 1574 goto done; 1575 } 1576 1577 /* 1578 * If a list.updated or list.repaired event is received, update the 1579 * resource cache status and the local case. 1580 */ 1581 if (fmd_event_match(e, FMD_EVT_PROTOCOL, FM_LIST_REPAIRED_CLASS) || 1582 fmd_event_match(e, FMD_EVT_PROTOCOL, FM_LIST_UPDATED_CLASS)) { 1583 uint8_t *statusp; 1584 uint_t nelem = 0; 1585 1586 (void) nvlist_lookup_uint8_array(nvl, FM_SUSPECT_FAULT_STATUS, 1587 &statusp, &nelem); 1588 fmd_module_lock(xip->xi_queue->eq_mod); 1589 if (nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) == 0 && 1590 (cp = fmd_case_hash_lookup(fmd.d_cases, uuid)) != NULL) { 1591 fmd_case_impl_t *cip = (fmd_case_impl_t *)cp; 1592 if (cip->ci_xprt != NULL) { 1593 fmd_case_update_status(cp, statusp, 1594 cip->ci_proxy_asru, cip->ci_diag_asru); 1595 fmd_case_update_containees(cp); 1596 fmd_case_update(cp); 1597 } 1598 fmd_case_rele(cp); 1599 } 1600 fmd_module_unlock(xip->xi_queue->eq_mod); 1601 fmd_event_hold(e); 1602 fmd_event_rele(e); 1603 goto done; 1604 } 1605 1606 /* 1607 * If a list.isolated event is received, update resource cache status 1608 */ 1609 if (fmd_event_match(e, FMD_EVT_PROTOCOL, FM_LIST_ISOLATED_CLASS)) { 1610 uint8_t *statusp; 1611 uint_t nelem = 0; 1612 1613 (void) nvlist_lookup_uint8_array(nvl, FM_SUSPECT_FAULT_STATUS, 1614 &statusp, &nelem); 1615 fmd_module_lock(xip->xi_queue->eq_mod); 1616 if (nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) == 0 && 1617 (cp = fmd_case_hash_lookup(fmd.d_cases, uuid)) != NULL) { 1618 fmd_case_impl_t *cip = (fmd_case_impl_t *)cp; 1619 if (cip->ci_xprt != NULL) 1620 fmd_case_update_status(cp, statusp, 1621 cip->ci_proxy_asru, cip->ci_diag_asru); 1622 fmd_case_rele(cp); 1623 } 1624 fmd_module_unlock(xip->xi_queue->eq_mod); 1625 fmd_event_hold(e); 1626 fmd_event_rele(e); 1627 goto done; 1628 } 1629 1630 /* 1631 * If a list.resolved event is received, resolve the local case. 1632 */ 1633 if (fmd_event_match(e, FMD_EVT_PROTOCOL, FM_LIST_RESOLVED_CLASS)) { 1634 fmd_module_lock(xip->xi_queue->eq_mod); 1635 if (nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) == 0 && 1636 (cp = fmd_case_hash_lookup(fmd.d_cases, uuid)) != NULL) { 1637 fmd_case_impl_t *cip = (fmd_case_impl_t *)cp; 1638 if (cip->ci_xprt != NULL) 1639 fmd_case_transition(cp, (cip->ci_state == 1640 FMD_CASE_REPAIRED) ? FMD_CASE_RESOLVED : 1641 (cip->ci_state == FMD_CASE_CLOSED) ? 1642 FMD_CASE_REPAIRED : FMD_CASE_CLOSE_WAIT, 1643 FMD_CF_RESOLVED); 1644 fmd_case_rele(cp); 1645 } 1646 fmd_module_unlock(xip->xi_queue->eq_mod); 1647 fmd_event_hold(e); 1648 fmd_event_rele(e); 1649 goto done; 1650 } 1651 1652 if (logonly == FMD_B_TRUE || (xip->xi_flags & FMD_XPRT_EXTERNAL)) { 1653 /* 1654 * Don't proxy ereports on an EXTERNAL transport - we won't 1655 * know how to diagnose them with the wrong topology. Note 1656 * that here (and above) we have to hold/release the event in 1657 * order for it to be freed. 1658 */ 1659 fmd_event_hold(e); 1660 fmd_event_rele(e); 1661 } else if (isproto == FMD_B_TRUE) 1662 fmd_dispq_dispatch(dp->d_disp, e, class); 1663 else 1664 fmd_modhash_dispatch(dp->d_mod_hash, e); 1665 done: 1666 (void) pthread_mutex_lock(&xip->xi_lock); 1667 1668 ASSERT(xip->xi_busy != 0); 1669 xip->xi_busy--; 1670 1671 (void) pthread_cond_broadcast(&xip->xi_cv); 1672 (void) pthread_mutex_unlock(&xip->xi_lock); 1673 } 1674 1675 void 1676 fmd_xprt_uuclose(fmd_xprt_t *xp, const char *uuid) 1677 { 1678 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1679 1680 fmd_event_t *e; 1681 nvlist_t *nvl; 1682 char *s; 1683 1684 if ((xip->xi_flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 1685 return; /* read-only transports do not proxy uuclose */ 1686 1687 TRACE((FMD_DBG_XPRT, "xprt %u closing case %s\n", xip->xi_id, uuid)); 1688 1689 nvl = fmd_protocol_xprt_uuclose(xip->xi_queue->eq_mod, 1690 "resource.fm.xprt.uuclose", xip->xi_version, uuid); 1691 1692 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 1693 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 1694 fmd_eventq_insert_at_time(xip->xi_queue, e); 1695 } 1696 1697 /* 1698 * On proxy side, send back uuresolved request to diagnosing side 1699 */ 1700 void 1701 fmd_xprt_uuresolved(fmd_xprt_t *xp, const char *uuid) 1702 { 1703 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1704 1705 fmd_event_t *e; 1706 nvlist_t *nvl; 1707 char *s; 1708 1709 if ((xip->xi_flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 1710 return; /* read-only transports do not proxy uuresolved */ 1711 1712 TRACE((FMD_DBG_XPRT, "xprt %u resolving case %s\n", xip->xi_id, uuid)); 1713 1714 nvl = fmd_protocol_xprt_uuresolved(xip->xi_queue->eq_mod, 1715 "resource.fm.xprt.uuresolved", xip->xi_version, uuid); 1716 1717 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 1718 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 1719 fmd_eventq_insert_at_time(xip->xi_queue, e); 1720 } 1721 1722 /* 1723 * On proxy side, send back repair/acquit/etc request to diagnosing side 1724 */ 1725 void 1726 fmd_xprt_updated(fmd_xprt_t *xp, const char *uuid, uint8_t *statusp, 1727 uint8_t *has_asrup, uint_t nelem) 1728 { 1729 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1730 1731 fmd_event_t *e; 1732 nvlist_t *nvl; 1733 char *s; 1734 1735 if ((xip->xi_flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 1736 return; /* read-only transports do not support remote repairs */ 1737 1738 TRACE((FMD_DBG_XPRT, "xprt %u updating case %s\n", xip->xi_id, uuid)); 1739 1740 nvl = fmd_protocol_xprt_updated(xip->xi_queue->eq_mod, 1741 "resource.fm.xprt.updated", xip->xi_version, uuid, statusp, 1742 has_asrup, nelem); 1743 1744 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 1745 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 1746 fmd_eventq_insert_at_time(xip->xi_queue, e); 1747 } 1748 1749 /* 1750 * Insert the specified class into our remote subscription hash. If the class 1751 * is already present, bump the reference count; otherwise add it to the hash 1752 * and then enqueue an event for our remote peer to proxy our subscription. 1753 */ 1754 void 1755 fmd_xprt_subscribe(fmd_xprt_t *xp, const char *class) 1756 { 1757 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1758 1759 uint_t refs; 1760 nvlist_t *nvl; 1761 fmd_event_t *e; 1762 char *s; 1763 1764 if ((xip->xi_flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 1765 return; /* read-only transports do not proxy subscriptions */ 1766 1767 if (!(xip->xi_flags & FMD_XPRT_SUBSCRIBER)) 1768 return; /* transport is not yet an active subscriber */ 1769 1770 (void) pthread_mutex_lock(&xip->xi_lock); 1771 refs = fmd_xprt_class_hash_insert(xip, &xip->xi_rsub, class); 1772 (void) pthread_mutex_unlock(&xip->xi_lock); 1773 1774 if (refs > 1) 1775 return; /* we've already asked our peer for this subscription */ 1776 1777 fmd_dprintf(FMD_DBG_XPRT, 1778 "xprt %u subscribing to %s\n", xip->xi_id, class); 1779 1780 nvl = fmd_protocol_xprt_sub(xip->xi_queue->eq_mod, 1781 "resource.fm.xprt.subscribe", xip->xi_version, class); 1782 1783 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 1784 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 1785 fmd_eventq_insert_at_time(xip->xi_queue, e); 1786 } 1787 1788 /* 1789 * Delete the specified class from the remote subscription hash. If the 1790 * reference count drops to zero, ask our remote peer to unsubscribe by proxy. 1791 */ 1792 void 1793 fmd_xprt_unsubscribe(fmd_xprt_t *xp, const char *class) 1794 { 1795 fmd_xprt_impl_t *xip = (fmd_xprt_impl_t *)xp; 1796 1797 uint_t refs; 1798 nvlist_t *nvl; 1799 fmd_event_t *e; 1800 char *s; 1801 1802 if ((xip->xi_flags & FMD_XPRT_RDWR) == FMD_XPRT_RDONLY) 1803 return; /* read-only transports do not proxy subscriptions */ 1804 1805 if (!(xip->xi_flags & FMD_XPRT_SUBSCRIBER)) 1806 return; /* transport is not yet an active subscriber */ 1807 1808 /* 1809 * If the subscription reference count drops to zero in xi_rsub, insert 1810 * an entry into the xi_usub hash indicating we await an unsuback event. 1811 */ 1812 (void) pthread_mutex_lock(&xip->xi_lock); 1813 1814 if ((refs = fmd_xprt_class_hash_delete(xip, &xip->xi_rsub, class)) == 0) 1815 (void) fmd_xprt_class_hash_insert(xip, &xip->xi_usub, class); 1816 1817 (void) pthread_mutex_unlock(&xip->xi_lock); 1818 1819 if (refs != 0) 1820 return; /* other subscriptions for this class still active */ 1821 1822 fmd_dprintf(FMD_DBG_XPRT, 1823 "xprt %u unsubscribing from %s\n", xip->xi_id, class); 1824 1825 nvl = fmd_protocol_xprt_sub(xip->xi_queue->eq_mod, 1826 "resource.fm.xprt.unsubscribe", xip->xi_version, class); 1827 1828 (void) nvlist_lookup_string(nvl, FM_CLASS, &s); 1829 e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, s); 1830 fmd_eventq_insert_at_time(xip->xi_queue, e); 1831 } 1832 1833 static void 1834 fmd_xprt_subscribe_xid(fmd_idspace_t *ids, id_t id, void *class) 1835 { 1836 fmd_xprt_t *xp; 1837 1838 if ((xp = fmd_idspace_hold(ids, id)) != NULL) { 1839 fmd_xprt_subscribe(xp, class); 1840 fmd_idspace_rele(ids, id); 1841 } 1842 } 1843 1844 void 1845 fmd_xprt_subscribe_all(const char *class) 1846 { 1847 fmd_idspace_t *ids = fmd.d_xprt_ids; 1848 1849 if (ids->ids_count != 0) 1850 fmd_idspace_apply(ids, fmd_xprt_subscribe_xid, (void *)class); 1851 } 1852 1853 static void 1854 fmd_xprt_unsubscribe_xid(fmd_idspace_t *ids, id_t id, void *class) 1855 { 1856 fmd_xprt_t *xp; 1857 1858 if ((xp = fmd_idspace_hold(ids, id)) != NULL) { 1859 fmd_xprt_unsubscribe(xp, class); 1860 fmd_idspace_rele(ids, id); 1861 } 1862 } 1863 1864 void 1865 fmd_xprt_unsubscribe_all(const char *class) 1866 { 1867 fmd_idspace_t *ids = fmd.d_xprt_ids; 1868 1869 if (ids->ids_count != 0) 1870 fmd_idspace_apply(ids, fmd_xprt_unsubscribe_xid, (void *)class); 1871 } 1872 1873 /*ARGSUSED*/ 1874 static void 1875 fmd_xprt_suspend_xid(fmd_idspace_t *ids, id_t id, void *arg) 1876 { 1877 fmd_xprt_t *xp; 1878 1879 if ((xp = fmd_idspace_hold(ids, id)) != NULL) { 1880 fmd_xprt_xsuspend(xp, FMD_XPRT_DSUSPENDED); 1881 fmd_idspace_rele(ids, id); 1882 } 1883 } 1884 1885 void 1886 fmd_xprt_suspend_all(void) 1887 { 1888 fmd_idspace_t *ids = fmd.d_xprt_ids; 1889 1890 (void) pthread_mutex_lock(&fmd.d_xprt_lock); 1891 1892 if (fmd.d_xprt_suspend++ != 0) { 1893 (void) pthread_mutex_unlock(&fmd.d_xprt_lock); 1894 return; /* already suspended */ 1895 } 1896 1897 if (ids->ids_count != 0) 1898 fmd_idspace_apply(ids, fmd_xprt_suspend_xid, NULL); 1899 1900 (void) pthread_mutex_unlock(&fmd.d_xprt_lock); 1901 } 1902 1903 /*ARGSUSED*/ 1904 static void 1905 fmd_xprt_resume_xid(fmd_idspace_t *ids, id_t id, void *arg) 1906 { 1907 fmd_xprt_t *xp; 1908 1909 if ((xp = fmd_idspace_hold(ids, id)) != NULL) { 1910 fmd_xprt_xresume(xp, FMD_XPRT_DSUSPENDED); 1911 fmd_idspace_rele(ids, id); 1912 } 1913 } 1914 1915 void 1916 fmd_xprt_resume_all(void) 1917 { 1918 fmd_idspace_t *ids = fmd.d_xprt_ids; 1919 1920 (void) pthread_mutex_lock(&fmd.d_xprt_lock); 1921 1922 if (fmd.d_xprt_suspend == 0) 1923 fmd_panic("fmd_xprt_suspend/resume_all mismatch\n"); 1924 1925 if (--fmd.d_xprt_suspend != 0) { 1926 (void) pthread_mutex_unlock(&fmd.d_xprt_lock); 1927 return; /* not ready to be resumed */ 1928 } 1929 1930 if (ids->ids_count != 0) 1931 fmd_idspace_apply(ids, fmd_xprt_resume_xid, NULL); 1932 1933 (void) pthread_mutex_unlock(&fmd.d_xprt_lock); 1934 } 1935