2 * Copyright 2000, International Business Machines Corporation and others.
5 * This software has been released under the terms of the IBM Public
6 * License. For details, see the LICENSE file in the top-level source
7 * directory or online at http://www.openafs.org/dl/license10.html
10 #include <afsconfig.h>
11 #include <afs/param.h>
23 #include <afs/unified_afs.h>
24 #include <afs/vlserver.h>
25 #include <WINNT/afsreg.h>
27 osi_rwlock_t cm_connLock;
29 DWORD RDRtimeout = CM_CONN_DEFAULTRDRTIMEOUT;
30 unsigned short ConnDeadtimeout = CM_CONN_CONNDEADTIME;
31 unsigned short HardDeadtimeout = CM_CONN_HARDDEADTIME;
32 unsigned short IdleDeadtimeout = CM_CONN_IDLEDEADTIME;
33 unsigned short NatPingInterval = CM_CONN_NATPINGINTERVAL;
35 #define LANMAN_WKS_PARAM_KEY "SYSTEM\\CurrentControlSet\\Services\\lanmanworkstation\\parameters"
36 #define LANMAN_WKS_SESSION_TIMEOUT "SessTimeout"
37 #define LANMAN_WKS_EXT_SESSION_TIMEOUT "ExtendedSessTimeout"
39 afs_uint32 cryptall = 0;
40 afs_uint32 cm_anonvldb = 0;
41 afs_uint32 rx_pmtu_discovery = 0;
43 void cm_PutConn(cm_conn_t *connp)
45 afs_int32 refCount = InterlockedDecrement(&connp->refCount);
46 osi_assertx(refCount >= 0, "cm_conn_t refcount underflow");
49 void cm_InitConn(void)
51 static osi_once_t once;
57 if (osi_Once(&once)) {
58 lock_InitializeRWLock(&cm_connLock, "connection global lock",
59 LOCK_HIERARCHY_CONN_GLOBAL);
61 /* keisa - read timeout value for lanmanworkstation service.
63 * http://support.microsoft.com:80/support/kb/articles/Q102/0/67.asp&NoWebContent=1
64 * the SessTimeout is a minimum timeout not a maximum timeout. Therefore,
65 * I believe that the default should not be short. Instead, we should wait until
66 * RX times out before reporting a timeout to the SMB client.
68 code = RegOpenKeyEx(HKEY_LOCAL_MACHINE, LANMAN_WKS_PARAM_KEY,
69 0, KEY_QUERY_VALUE, &parmKey);
70 if (code == ERROR_SUCCESS)
72 BOOL extTimeouts = msftSMBRedirectorSupportsExtendedTimeouts();
76 * The default value is 1000 seconds. However, this default
77 * will not apply to "\\AFS" unless "AFS" is listed in
78 * ServersWithExtendedSessTimeout which we will add when we
79 * create the ExtendedSessTimeout value in smb_Init()
81 dummyLen = sizeof(DWORD);
82 code = RegQueryValueEx(parmKey,
83 LANMAN_WKS_EXT_SESSION_TIMEOUT,
85 (BYTE *) &dwValue, &dummyLen);
86 if (code == ERROR_SUCCESS) {
88 afsi_log("lanmanworkstation : ExtSessTimeout %u", RDRtimeout);
91 if (!extTimeouts || code != ERROR_SUCCESS) {
92 dummyLen = sizeof(DWORD);
93 code = RegQueryValueEx(parmKey,
94 LANMAN_WKS_SESSION_TIMEOUT,
96 (BYTE *) &dwValue, &dummyLen);
97 if (code == ERROR_SUCCESS) {
99 afsi_log("lanmanworkstation : SessTimeout %u", RDRtimeout);
102 RegCloseKey(parmKey);
105 code = RegOpenKeyEx(HKEY_LOCAL_MACHINE, AFSREG_CLT_SVC_PARAM_SUBKEY,
106 0, KEY_QUERY_VALUE, &parmKey);
107 if (code == ERROR_SUCCESS) {
108 dummyLen = sizeof(DWORD);
109 code = RegQueryValueEx(parmKey, "ConnDeadTimeout", NULL, NULL,
110 (BYTE *) &dwValue, &dummyLen);
111 if (code == ERROR_SUCCESS) {
112 ConnDeadtimeout = (unsigned short)dwValue;
113 afsi_log("ConnDeadTimeout is %d", ConnDeadtimeout);
115 dummyLen = sizeof(DWORD);
116 code = RegQueryValueEx(parmKey, "HardDeadTimeout", NULL, NULL,
117 (BYTE *) &dwValue, &dummyLen);
118 if (code == ERROR_SUCCESS) {
119 HardDeadtimeout = (unsigned short)dwValue;
120 afsi_log("HardDeadTimeout is %d", HardDeadtimeout);
122 dummyLen = sizeof(DWORD);
123 code = RegQueryValueEx(parmKey, "IdleDeadTimeout", NULL, NULL,
124 (BYTE *) &dwValue, &dummyLen);
125 if (code == ERROR_SUCCESS) {
126 IdleDeadtimeout = (unsigned short)dwValue;
127 afsi_log("IdleDeadTimeout is %d", IdleDeadtimeout);
129 dummyLen = sizeof(DWORD);
130 code = RegQueryValueEx(parmKey, "NatPingInterval", NULL, NULL,
131 (BYTE *) &dwValue, &dummyLen);
132 if (code == ERROR_SUCCESS) {
133 NatPingInterval = (unsigned short)dwValue;
135 afsi_log("NatPingInterval is %d", NatPingInterval);
136 RegCloseKey(parmKey);
140 * If these values were not set via cpp macro or obtained
141 * from the registry, we use a value that is derived from
142 * the smb redirector session timeout (RDRtimeout).
144 * The UNIX cache manager uses 120 seconds for the hard dead
145 * timeout and 50 seconds for the connection and idle timeouts.
147 * We base our values on those while making sure we leave
148 * enough time for overhead.
151 afsi_log("lanmanworkstation : SessTimeout %u", RDRtimeout);
152 if (ConnDeadtimeout == 0) {
153 ConnDeadtimeout = (unsigned short) ((RDRtimeout / 2) < 50 ? (RDRtimeout / 2) : 50);
154 afsi_log("ConnDeadTimeout is %d", ConnDeadtimeout);
156 if (HardDeadtimeout == 0) {
157 HardDeadtimeout = (unsigned short) (RDRtimeout > 125 ? 120 : (RDRtimeout - 5));
158 afsi_log("HardDeadTimeout is %d", HardDeadtimeout);
160 if (IdleDeadtimeout == 0) {
161 IdleDeadtimeout = (unsigned short) ConnDeadtimeout;
162 afsi_log("IdleDeadTimeout is %d", IdleDeadtimeout);
169 void cm_InitReq(cm_req_t *reqp)
171 memset(reqp, 0, sizeof(cm_req_t));
172 reqp->startTime = GetTickCount();
175 static long cm_GetServerList(struct cm_fid *fidp, struct cm_user *userp,
176 struct cm_req *reqp, cm_serverRef_t ***serversppp)
179 cm_volume_t *volp = NULL;
180 cm_cell_t *cellp = NULL;
184 return CM_ERROR_INVAL;
187 cellp = cm_FindCellByID(fidp->cell, 0);
189 return CM_ERROR_NOSUCHCELL;
191 code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp, CM_GETVOL_FLAG_CREATE, &volp);
195 *serversppp = cm_GetVolServers(volp, fidp->volume, userp, reqp);
197 lock_ObtainRead(&cm_volumeLock);
199 lock_ReleaseRead(&cm_volumeLock);
200 return (*serversppp ? 0 : CM_ERROR_NOSUCHVOLUME);
204 * Analyze the error return from an RPC. Determine whether or not to retry,
205 * and if we're going to retry, determine whether failover is appropriate,
206 * and whether timed backoff is appropriate.
208 * If the error code is from cm_ConnFromFID() or friends, it will be a CM_ERROR code.
209 * Otherwise it will be an RPC code. This may be a UNIX code (e.g. EDQUOT), or
210 * it may be an RX code, or it may be a special code (e.g. VNOVOL), or it may
211 * be a security code (e.g. RXKADEXPIRED).
213 * If the error code is from cm_ConnFromFID() or friends, connp will be NULL.
215 * For VLDB calls, fidp will be NULL.
217 * volSyncp and/or cbrp may also be NULL.
220 cm_Analyze(cm_conn_t *connp, cm_user_t *userp, cm_req_t *reqp,
222 AFSVolSync *volSyncp,
223 cm_serverRef_t * serversp,
224 cm_callbackRequest_t *cbrp, long errorCode)
226 cm_server_t *serverp = NULL;
227 cm_serverRef_t **serverspp = NULL;
228 cm_serverRef_t *tsrp;
229 cm_cell_t *cellp = NULL;
231 cm_volume_t * volp = NULL;
232 cm_vol_state_t *statep = NULL;
234 int free_svr_list = 0;
236 long timeUsed, timeLeft;
238 char addr[16]="unknown";
240 int location_updated = 0;
244 osi_Log2(afsd_logp, "cm_Analyze connp 0x%p, code 0x%x",
247 /* no locking required, since connp->serverp never changes after
249 dead_session = (userp->cellInfop == NULL);
251 serverp = connp->serverp;
253 /* Update callback pointer */
254 if (cbrp && serverp && errorCode == 0) {
256 if ( cbrp->serverp != serverp ) {
257 lock_ObtainWrite(&cm_serverLock);
258 cm_PutServerNoLock(cbrp->serverp);
259 cm_GetServerNoLock(serverp);
260 lock_ReleaseWrite(&cm_serverLock);
263 cm_GetServer(serverp);
265 lock_ObtainWrite(&cm_callbackLock);
266 cbrp->serverp = serverp;
267 lock_ReleaseWrite(&cm_callbackLock);
270 /* if timeout - check that it did not exceed the HardDead timeout
273 /* timeleft - get it from reqp the same way as cm_ConnByMServers does */
274 timeUsed = (GetTickCount() - reqp->startTime) / 1000;
275 timeLeft = HardDeadtimeout - timeUsed;
277 /* get a pointer to the cell */
279 if (cellp == NULL && serverp)
280 cellp = serverp->cellp;
281 if (cellp == NULL && serversp) {
282 struct cm_serverRef * refp;
283 for ( refp=serversp ; cellp == NULL && refp != NULL; refp=refp->next) {
284 if (refp->status == srv_deleted)
287 cellp = refp->server->cellp;
290 if (cellp == NULL && fidp) {
291 cellp = cm_FindCellByID(fidp->cell, 0);
295 if (errorCode == CM_ERROR_TIMEDOUT) {
296 if ( timeLeft > 5 ) {
298 cm_CheckServers(CM_FLAG_CHECKDOWNSERVERS, cellp);
303 else if (errorCode == UAEWOULDBLOCK || errorCode == EWOULDBLOCK ||
304 errorCode == UAEAGAIN || errorCode == EAGAIN) {
305 osi_Log0(afsd_logp, "cm_Analyze passed EWOULDBLOCK or EAGAIN.");
306 if ( timeLeft > 5 ) {
312 /* if there is nosuchvolume, then we have a situation in which a
313 * previously known volume no longer has a set of servers
314 * associated with it. Either the volume has moved or
315 * the volume has been deleted. Try to find a new server list
316 * until the timeout period expires.
318 else if (errorCode == CM_ERROR_NOSUCHVOLUME) {
319 osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_NOSUCHVOLUME.");
321 * The VNOVOL or VL_NOENT error has already been translated
322 * to CM_ERROR_NOSUCHVOLUME. There is nothing for us to do.
326 else if (errorCode == CM_ERROR_ALLDOWN) {
327 /* Servers marked DOWN will be restored by the background daemon
328 * thread as they become available. The volume status is
329 * updated as the server state changes.
332 osi_Log2(afsd_logp, "cm_Analyze passed CM_ERROR_DOWN (FS cell %s vol 0x%x)",
333 cellp->name, fidp->volume);
334 msgID = MSG_ALL_SERVERS_DOWN;
335 format = "All servers are unreachable when accessing cell %s volume %d.";
336 LogEvent(EVENTLOG_WARNING_TYPE, msgID, cellp->name, fidp->volume);
338 osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLDOWN (VL Server)");
342 else if (errorCode == CM_ERROR_ALLOFFLINE) {
343 /* Volume instances marked offline will be restored by the
344 * background daemon thread as they become available
347 osi_Log2(afsd_logp, "cm_Analyze passed CM_ERROR_ALLOFFLINE (FS cell %s vol 0x%x)",
348 cellp->name, fidp->volume);
349 msgID = MSG_ALL_SERVERS_OFFLINE;
350 format = "All servers are offline when accessing cell %s volume %d.";
351 LogEvent(EVENTLOG_WARNING_TYPE, msgID, cellp->name, fidp->volume);
353 code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
354 CM_GETVOL_FLAG_NO_LRU_UPDATE,
358 * Do not perform a cm_CheckOfflineVolume() if cm_Analyze()
359 * was called by cm_CheckOfflineVolumeState().
361 if (!(reqp->flags & CM_REQ_OFFLINE_VOL_CHK) && timeLeft > 7) {
364 /* cm_CheckOfflineVolume() resets the serverRef state */
365 if (cm_CheckOfflineVolume(volp, fidp->volume))
368 cm_UpdateVolumeStatus(volp, fidp->volume);
370 lock_ObtainRead(&cm_volumeLock);
372 lock_ReleaseRead(&cm_volumeLock);
376 osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLOFFLINE (VL Server)");
379 else if (errorCode == CM_ERROR_ALLBUSY) {
380 /* Volumes that are busy cannot be determined to be non-busy
381 * without actually attempting to access them.
383 if (fidp) { /* File Server query */
384 osi_Log2(afsd_logp, "cm_Analyze passed CM_ERROR_ALLBUSY (FS cell %s vol 0x%x)",
385 cellp->name, fidp->volume);
386 msgID = MSG_ALL_SERVERS_BUSY;
387 format = "All servers are busy when accessing cell %s volume %d.";
388 LogEvent(EVENTLOG_WARNING_TYPE, msgID, cellp->name, fidp->volume);
390 code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
391 CM_GETVOL_FLAG_NO_LRU_UPDATE,
397 statep = cm_VolumeStateByID(volp, fidp->volume);
398 if (statep->state != vl_offline &&
399 statep->state != vl_busy &&
400 statep->state != vl_unknown) {
404 code = cm_GetServerList(fidp, userp, reqp, &serverspp);
406 serversp = *serverspp;
410 lock_ObtainWrite(&cm_serverLock);
411 for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
412 if (tsrp->status == srv_deleted)
414 if (tsrp->status == srv_busy) {
415 tsrp->status = srv_not_busy;
418 lock_ReleaseWrite(&cm_serverLock);
420 cm_FreeServerList(serverspp, 0);
425 cm_UpdateVolumeStatus(volp, fidp->volume);
429 cm_UpdateVolumeStatus(volp, fidp->volume);
432 lock_ObtainRead(&cm_volumeLock);
434 lock_ReleaseRead(&cm_volumeLock);
437 } else { /* VL Server query */
438 osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLBUSY (VL Server).");
444 lock_ObtainWrite(&cm_serverLock);
445 for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
446 if (tsrp->status == srv_deleted)
448 if (tsrp->status == srv_busy) {
449 tsrp->status = srv_not_busy;
452 lock_ReleaseWrite(&cm_serverLock);
459 /* special codes: VBUSY and VRESTARTING */
460 else if (errorCode == VBUSY || errorCode == VRESTARTING) {
461 if (!serversp && fidp) {
462 code = cm_GetServerList(fidp, userp, reqp, &serverspp);
464 serversp = *serverspp;
469 switch ( errorCode ) {
471 msgID = MSG_SERVER_REPORTS_VBUSY;
472 format = "Server %s reported busy when accessing volume %d in cell %s.";
475 msgID = MSG_SERVER_REPORTS_VRESTARTING;
476 format = "Server %s reported restarting when accessing volume %d in cell %s.";
480 if (serverp && fidp) {
481 /* Log server being offline for this volume */
482 sprintf(addr, "%d.%d.%d.%d",
483 ((serverp->addr.sin_addr.s_addr & 0xff)),
484 ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
485 ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
486 ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
488 osi_Log3(afsd_logp, format, osi_LogSaveString(afsd_logp,addr), fidp->volume, cellp->name);
489 LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume, cellp->name);
492 lock_ObtainWrite(&cm_serverLock);
493 for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
494 if (tsrp->status == srv_deleted)
496 if (tsrp->server == serverp && tsrp->status == srv_not_busy) {
497 tsrp->status = srv_busy;
498 if (fidp) { /* File Server query */
499 lock_ReleaseWrite(&cm_serverLock);
500 code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
501 CM_GETVOL_FLAG_NO_LRU_UPDATE,
504 statep = cm_VolumeStateByID(volp, fidp->volume);
505 lock_ObtainWrite(&cm_serverLock);
510 lock_ReleaseWrite(&cm_serverLock);
513 cm_UpdateVolumeStatus(volp, statep->ID);
514 lock_ObtainRead(&cm_volumeLock);
516 lock_ReleaseRead(&cm_volumeLock);
521 cm_FreeServerList(serverspp, 0);
528 /* special codes: missing volumes */
529 else if (errorCode == VNOVOL || errorCode == VMOVED || errorCode == VOFFLINE ||
530 errorCode == VSALVAGE || errorCode == VNOSERVICE || errorCode == VIO)
532 /* In case of timeout */
533 reqp->volumeError = errorCode;
535 switch ( errorCode ) {
537 msgID = MSG_SERVER_REPORTS_VNOVOL;
538 format = "Server %s reported volume %d in cell %s as not attached (may have been moved or deleted).";
541 msgID = MSG_SERVER_REPORTS_VMOVED;
542 format = "Server %s reported volume %d in cell %s as moved.";
545 msgID = MSG_SERVER_REPORTS_VOFFLINE;
546 format = "Server %s reported volume %d in cell %s as offline.";
549 msgID = MSG_SERVER_REPORTS_VSALVAGE;
550 format = "Server %s reported volume %d in cell %s as needs salvage.";
553 msgID = MSG_SERVER_REPORTS_VNOSERVICE;
554 format = "Server %s reported volume %d in cell %s as not in service.";
557 msgID = MSG_SERVER_REPORTS_VIO;
558 format = "Server %s reported volume %d in cell %s as temporarily unaccessible.";
562 if (fidp) { /* File Server query */
564 /* Log server being unavailable for this volume */
565 sprintf(addr, "%d.%d.%d.%d",
566 ((serverp->addr.sin_addr.s_addr & 0xff)),
567 ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
568 ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
569 ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
571 osi_Log3(afsd_logp, format, osi_LogSaveString(afsd_logp,addr), fidp->volume, cellp->name);
572 LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume, cellp->name);
575 code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
576 CM_GETVOL_FLAG_NO_LRU_UPDATE,
579 statep = cm_VolumeStateByID(volp, fidp->volume);
581 if ((errorCode == VMOVED || errorCode == VNOVOL || errorCode == VOFFLINE) &&
582 !(reqp->flags & CM_REQ_VOLUME_UPDATED))
584 code = cm_ForceUpdateVolume(fidp, userp, reqp);
586 location_updated = 1;
588 /* Even if the update fails, there might still be another replica */
590 reqp->flags |= CM_REQ_VOLUME_UPDATED;
591 osi_Log3(afsd_logp, "cm_Analyze called cm_ForceUpdateVolume cell %u vol %u code 0x%x",
592 fidp->cell, fidp->volume, code);
596 cm_UpdateVolumeStatus(volp, statep->ID);
597 osi_Log3(afsd_logp, "cm_Analyze NewVolState cell %u vol %u state %u",
598 fidp->cell, fidp->volume, statep->state);
602 lock_ObtainRead(&cm_volumeLock);
604 lock_ReleaseRead(&cm_volumeLock);
609 * Mark server offline for this volume or delete the volume
610 * from the server list if it was moved or is not present.
612 if (!serversp || location_updated) {
613 code = cm_GetServerList(fidp, userp, reqp, &serverspp);
615 serversp = *serverspp;
621 lock_ObtainWrite(&cm_serverLock);
622 for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
623 if (tsrp->status == srv_deleted)
626 sprintf(addr, "%d.%d.%d.%d",
627 ((tsrp->server->addr.sin_addr.s_addr & 0xff)),
628 ((tsrp->server->addr.sin_addr.s_addr & 0xff00)>> 8),
629 ((tsrp->server->addr.sin_addr.s_addr & 0xff0000)>> 16),
630 ((tsrp->server->addr.sin_addr.s_addr & 0xff000000)>> 24));
632 if (cm_ServerEqual(tsrp->server, serverp)) {
634 if (errorCode == VMOVED || errorCode == VNOVOL) {
635 osi_Log2(afsd_logp, "volume %d not present on server %s",
636 fidp->volume, osi_LogSaveString(afsd_logp,addr));
637 tsrp->status = srv_deleted;
639 cm_RemoveVolumeFromServer(serverp, fidp->volume);
641 osi_Log2(afsd_logp, "volume %d instance on server %s marked offline",
642 fidp->volume, osi_LogSaveString(afsd_logp,addr));
643 tsrp->status = srv_offline;
647 osi_Log3(afsd_logp, "volume %d exists on server %s with status %u",
648 fidp->volume, osi_LogSaveString(afsd_logp,addr), tsrp->status);
651 lock_ReleaseWrite(&cm_serverLock);
653 /* Free the server list before cm_ForceUpdateVolume is called */
655 cm_FreeServerList(serverspp, 0);
662 } else if ( errorCode == VNOVNODE ) {
665 osi_Log4(afsd_logp, "cm_Analyze passed VNOVNODE cell %u vol %u vn %u uniq %u.",
666 fidp->cell, fidp->volume, fidp->vnode, fidp->unique);
668 scp = cm_FindSCache(fidp);
670 cm_scache_t *pscp = NULL;
672 if (scp->fileType != CM_SCACHETYPE_DIRECTORY)
673 pscp = cm_FindSCacheParent(scp);
675 lock_ObtainWrite(&scp->rw);
676 scp->flags |= CM_SCACHEFLAG_DELETED;
677 lock_ObtainWrite(&cm_scacheLock);
678 cm_AdjustScacheLRU(scp);
679 cm_RemoveSCacheFromHashTable(scp);
680 lock_ReleaseWrite(&cm_scacheLock);
681 cm_LockMarkSCacheLost(scp);
682 lock_ReleaseWrite(&scp->rw);
683 cm_ReleaseSCache(scp);
686 if (cm_HaveCallback(pscp)) {
687 lock_ObtainWrite(&pscp->rw);
688 cm_DiscardSCache(pscp);
689 lock_ReleaseWrite(&pscp->rw);
691 cm_ReleaseSCache(pscp);
695 osi_Log0(afsd_logp, "cm_Analyze passed VNOVNODE unknown fid.");
700 else if (errorCode == RX_CALL_TIMEOUT) {
701 /* RPC took longer than hardDeadTime or the server
702 * reported idle for longer than idleDeadTime
703 * don't mark server as down but don't retry
704 * this is to prevent the SMB session from timing out
705 * In addition, we log an event to the event log
709 sprintf(addr, "%d.%d.%d.%d",
710 ((serverp->addr.sin_addr.s_addr & 0xff)),
711 ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
712 ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
713 ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
715 LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_HARD_DEAD_TIME_EXCEEDED, addr);
716 osi_Log1(afsd_logp, "cm_Analyze: hardDeadTime or idleDeadtime exceeded addr[%s]",
717 osi_LogSaveString(afsd_logp,addr));
718 reqp->tokenIdleErrorServp = serverp;
723 if (!fidp) { /* vldb */
726 cm_volume_t *volp = cm_GetVolumeByFID(fidp);
728 if (fidp->volume == cm_GetROVolumeID(volp))
735 else if (errorCode == RX_MSGSIZE) {
737 * RPC failed because a transmitted rx packet
738 * may have grown larger than the path mtu.
739 * Force a retry and the Rx library will try
740 * with a smaller mtu size.
744 sprintf(addr, "%d.%d.%d.%d",
745 ((serverp->addr.sin_addr.s_addr & 0xff)),
746 ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
747 ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
748 ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
750 LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_MSGSIZE_EXCEEDED, addr);
751 osi_Log1(afsd_logp, "cm_Analyze: Path MTU may have been exceeded addr[%s]",
752 osi_LogSaveString(afsd_logp,addr));
756 else if (errorCode == CM_RX_RETRY_BUSY_CALL) {
758 * RPC failed because the selected call channel
759 * is currently busy on the server. Unconditionally
760 * retry the request so an alternate call channel can be used.
763 sprintf(addr, "%d.%d.%d.%d",
764 ((serverp->addr.sin_addr.s_addr & 0xff)),
765 ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
766 ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
767 ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
769 LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_BUSY_CALL_CHANNEL, addr);
770 osi_Log1(afsd_logp, "cm_Analyze: Retry RPC due to busy call channel addr[%s]",
771 osi_LogSaveString(afsd_logp,addr));
774 else if (errorCode >= -64 && errorCode < 0) {
775 /* mark server as down */
777 sprintf(addr, "%d.%d.%d.%d",
778 ((serverp->addr.sin_addr.s_addr & 0xff)),
779 ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
780 ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
781 ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
783 if (errorCode == RX_CALL_DEAD)
784 osi_Log2(afsd_logp, "cm_Analyze: Rx Call Dead addr[%s] forcedNew[%s]",
785 osi_LogSaveString(afsd_logp,addr),
786 (reqp->flags & CM_REQ_NEW_CONN_FORCED ? "yes" : "no"));
788 osi_Log3(afsd_logp, "cm_Analyze: Rx Misc Error[%d] addr[%s] forcedNew[%s]",
790 osi_LogSaveString(afsd_logp,addr),
791 (reqp->flags & CM_REQ_NEW_CONN_FORCED ? "yes" : "no"));
794 lock_ObtainMutex(&serverp->mx);
795 if (errorCode == RX_CALL_DEAD &&
796 (reqp->flags & CM_REQ_NEW_CONN_FORCED)) {
797 if (!(serverp->flags & CM_SERVERFLAG_DOWN)) {
798 _InterlockedOr(&serverp->flags, CM_SERVERFLAG_DOWN);
799 serverp->downTime = time(NULL);
802 if (reqp->flags & CM_REQ_NEW_CONN_FORCED) {
803 reqp->tokenIdleErrorServp = serverp;
804 reqp->tokenError = errorCode;
806 reqp->flags |= CM_REQ_NEW_CONN_FORCED;
810 lock_ReleaseMutex(&serverp->mx);
811 cm_ForceNewConnections(serverp);
816 else if (errorCode == RXKADEXPIRED) {
817 osi_Log1(afsd_logp, "cm_Analyze: rxkad error code 0x%x (RXKADEXPIRED)",
820 lock_ObtainMutex(&userp->mx);
821 ucellp = cm_GetUCell(userp, serverp->cellp);
822 if (ucellp->ticketp) {
823 free(ucellp->ticketp);
824 ucellp->ticketp = NULL;
826 _InterlockedAnd(&ucellp->flags, ~CM_UCELLFLAG_RXKAD);
828 lock_ReleaseMutex(&userp->mx);
832 } else if (errorCode >= ERROR_TABLE_BASE_RXK && errorCode < ERROR_TABLE_BASE_RXK + 256) {
833 char * s = "unknown error";
834 switch ( errorCode ) {
835 case RXKADINCONSISTENCY: s = "RXKADINCONSISTENCY"; break;
836 case RXKADPACKETSHORT : s = "RXKADPACKETSHORT"; break;
837 case RXKADLEVELFAIL : s = "RXKADLEVELFAIL"; break;
838 case RXKADTICKETLEN : s = "RXKADTICKETLEN"; break;
839 case RXKADOUTOFSEQUENCE: s = "RXKADOUTOFSEQUENCE"; break;
840 case RXKADNOAUTH : s = "RXKADNOAUTH"; break;
841 case RXKADBADKEY : s = "RXKADBADKEY"; break;
842 case RXKADBADTICKET : s = "RXKADBADTICKET"; break;
843 case RXKADUNKNOWNKEY : s = "RXKADUNKNOWNKEY"; break;
844 case RXKADEXPIRED : s = "RXKADEXPIRED"; break;
845 case RXKADSEALEDINCON : s = "RXKADSEALEDINCON"; break;
846 case RXKADDATALEN : s = "RXKADDATALEN"; break;
847 case RXKADILLEGALLEVEL : s = "RXKADILLEGALLEVEL"; break;
849 osi_Log2(afsd_logp, "cm_Analyze: rxkad error code 0x%x (%s)",
853 reqp->tokenIdleErrorServp = serverp;
854 reqp->tokenError = errorCode;
857 } else if (errorCode >= ERROR_TABLE_BASE_U && errorCode < ERROR_TABLE_BASE_U + 256) {
859 * We received a ubik error. its possible that the server we are
860 * communicating with has a corrupted database or is partitioned
861 * from the rest of the servers and another server might be able
862 * to answer our query. Therefore, we will retry the request
863 * and force the use of another server.
866 reqp->tokenIdleErrorServp = serverp;
867 reqp->tokenError = errorCode;
870 } else if (errorCode == VICECONNBAD || errorCode == VICETOKENDEAD) {
871 cm_ForceNewConnections(serverp);
876 char * s = "unknown error";
877 switch ( errorCode ) {
878 case VSALVAGE : s = "VSALVAGE"; break;
879 case VNOVNODE : s = "VNOVNODE"; break;
880 case VNOVOL : s = "VNOVOL"; break;
881 case VVOLEXISTS : s = "VVOLEXISTS"; break;
882 case VNOSERVICE : s = "VNOSERVICE"; break;
883 case VOFFLINE : s = "VOFFLINE"; break;
884 case VONLINE : s = "VONLINE"; break;
885 case VDISKFULL : s = "VDISKFULL"; break;
886 case VOVERQUOTA : s = "VOVERQUOTA"; break;
887 case VBUSY : s = "VBUSY"; break;
888 case VMOVED : s = "VMOVED"; break;
889 case VIO : s = "VIO"; break;
890 case VRESTRICTED : s = "VRESTRICTED"; break;
891 case VRESTARTING : s = "VRESTARTING"; break;
892 case VREADONLY : s = "VREADONLY"; break;
893 case EAGAIN : s = "EAGAIN"; break;
894 case UAEAGAIN : s = "UAEAGAIN"; break;
895 case EINVAL : s = "EINVAL"; break;
896 case UAEINVAL : s = "UAEINVAL"; break;
897 case EACCES : s = "EACCES"; break;
898 case UAEACCES : s = "UAEACCES"; break;
899 case ENOENT : s = "ENOENT"; break;
900 case UAENOENT : s = "UAENOENT"; break;
901 case EEXIST : s = "EEXIST"; break;
902 case UAEEXIST : s = "UAEEXIST"; break;
903 case VICECONNBAD : s = "VICECONNBAD"; break;
904 case VICETOKENDEAD : s = "VICETOKENDEAD"; break;
905 case WSAEWOULDBLOCK : s = "WSAEWOULDBLOCK"; break;
906 case UAEWOULDBLOCK : s = "UAEWOULDBLOCK"; break;
907 case VL_IDEXIST : s = "VL_IDEXIST"; break;
908 case VL_IO : s = "VL_IO"; break;
909 case VL_NAMEEXIST : s = "VL_NAMEEXIST"; break;
910 case VL_CREATEFAIL : s = "VL_CREATEFAIL"; break;
911 case VL_NOENT : s = "VL_NOENT"; break;
912 case VL_EMPTY : s = "VL_EMPTY"; break;
913 case VL_ENTDELETED : s = "VL_ENTDELETED"; break;
914 case VL_BADNAME : s = "VL_BADNAME"; break;
915 case VL_BADINDEX : s = "VL_BADINDEX"; break;
916 case VL_BADVOLTYPE : s = "VL_BADVOLTYPE"; break;
917 case VL_BADSERVER : s = "VL_BADSERVER"; break;
918 case VL_BADPARTITION : s = "VL_BADPARTITION"; break;
919 case VL_REPSFULL : s = "VL_REPSFULL"; break;
920 case VL_NOREPSERVER : s = "VL_NOREPSERVER"; break;
921 case VL_DUPREPSERVER : s = "VL_DUPREPSERVER"; break;
922 case VL_RWNOTFOUND : s = "VL_RWNOTFOUND"; break;
923 case VL_BADREFCOUNT : s = "VL_BADREFCOUNT"; break;
924 case VL_SIZEEXCEEDED : s = "VL_SIZEEXCEEDED"; break;
925 case VL_BADENTRY : s = "VL_BADENTRY"; break;
926 case VL_BADVOLIDBUMP : s = "VL_BADVOLIDBUMP"; break;
927 case VL_IDALREADYHASHED: s = "VL_IDALREADYHASHED"; break;
928 case VL_ENTRYLOCKED : s = "VL_ENTRYLOCKED"; break;
929 case VL_BADVOLOPER : s = "VL_BADVOLOPER"; break;
930 case VL_BADRELLOCKTYPE : s = "VL_BADRELLOCKTYPE"; break;
931 case VL_RERELEASE : s = "VL_RERELEASE"; break;
932 case VL_BADSERVERFLAG : s = "VL_BADSERVERFLAG"; break;
933 case VL_PERM : s = "VL_PERM"; break;
934 case VL_NOMEM : s = "VL_NOMEM"; break;
935 case VL_BADVERSION : s = "VL_BADVERSION"; break;
936 case VL_INDEXERANGE : s = "VL_INDEXERANGE"; break;
937 case VL_MULTIPADDR : s = "VL_MULTIPADDR"; break;
938 case VL_BADMASK : s = "VL_BADMASK"; break;
939 case CM_ERROR_NOSUCHCELL : s = "CM_ERROR_NOSUCHCELL"; break;
940 case CM_ERROR_NOSUCHVOLUME : s = "CM_ERROR_NOSUCHVOLUME"; break;
941 case CM_ERROR_TIMEDOUT : s = "CM_ERROR_TIMEDOUT"; break;
942 case CM_ERROR_RETRY : s = "CM_ERROR_RETRY"; break;
943 case CM_ERROR_NOACCESS : s = "CM_ERROR_NOACCESS"; break;
944 case CM_ERROR_NOSUCHFILE : s = "CM_ERROR_NOSUCHFILE"; break;
945 case CM_ERROR_STOPNOW : s = "CM_ERROR_STOPNOW"; break;
946 case CM_ERROR_TOOBIG : s = "CM_ERROR_TOOBIG"; break;
947 case CM_ERROR_INVAL : s = "CM_ERROR_INVAL"; break;
948 case CM_ERROR_BADFD : s = "CM_ERROR_BADFD"; break;
949 case CM_ERROR_BADFDOP : s = "CM_ERROR_BADFDOP"; break;
950 case CM_ERROR_EXISTS : s = "CM_ERROR_EXISTS"; break;
951 case CM_ERROR_CROSSDEVLINK : s = "CM_ERROR_CROSSDEVLINK"; break;
952 case CM_ERROR_BADOP : s = "CM_ERROR_BADOP"; break;
953 case CM_ERROR_BADPASSWORD : s = "CM_ERROR_BADPASSWORD"; break;
954 case CM_ERROR_NOTDIR : s = "CM_ERROR_NOTDIR"; break;
955 case CM_ERROR_ISDIR : s = "CM_ERROR_ISDIR"; break;
956 case CM_ERROR_READONLY : s = "CM_ERROR_READONLY"; break;
957 case CM_ERROR_WOULDBLOCK : s = "CM_ERROR_WOULDBLOCK"; break;
958 case CM_ERROR_QUOTA : s = "CM_ERROR_QUOTA"; break;
959 case CM_ERROR_SPACE : s = "CM_ERROR_SPACE"; break;
960 case CM_ERROR_BADSHARENAME : s = "CM_ERROR_BADSHARENAME"; break;
961 case CM_ERROR_BADTID : s = "CM_ERROR_BADTID"; break;
962 case CM_ERROR_UNKNOWN : s = "CM_ERROR_UNKNOWN"; break;
963 case CM_ERROR_NOMORETOKENS : s = "CM_ERROR_NOMORETOKENS"; break;
964 case CM_ERROR_NOTEMPTY : s = "CM_ERROR_NOTEMPTY"; break;
965 case CM_ERROR_USESTD : s = "CM_ERROR_USESTD"; break;
966 case CM_ERROR_REMOTECONN : s = "CM_ERROR_REMOTECONN"; break;
967 case CM_ERROR_ATSYS : s = "CM_ERROR_ATSYS"; break;
968 case CM_ERROR_NOSUCHPATH : s = "CM_ERROR_NOSUCHPATH"; break;
969 case CM_ERROR_CLOCKSKEW : s = "CM_ERROR_CLOCKSKEW"; break;
970 case CM_ERROR_BADSMB : s = "CM_ERROR_BADSMB"; break;
971 case CM_ERROR_ALLBUSY : s = "CM_ERROR_ALLBUSY"; break;
972 case CM_ERROR_NOFILES : s = "CM_ERROR_NOFILES"; break;
973 case CM_ERROR_PARTIALWRITE : s = "CM_ERROR_PARTIALWRITE"; break;
974 case CM_ERROR_NOIPC : s = "CM_ERROR_NOIPC"; break;
975 case CM_ERROR_BADNTFILENAME : s = "CM_ERROR_BADNTFILENAME"; break;
976 case CM_ERROR_BUFFERTOOSMALL : s = "CM_ERROR_BUFFERTOOSMALL"; break;
977 case CM_ERROR_RENAME_IDENTICAL : s = "CM_ERROR_RENAME_IDENTICAL"; break;
978 case CM_ERROR_ALLOFFLINE : s = "CM_ERROR_ALLOFFLINE"; break;
979 case CM_ERROR_AMBIGUOUS_FILENAME: s = "CM_ERROR_AMBIGUOUS_FILENAME"; break;
980 case CM_ERROR_BADLOGONTYPE : s = "CM_ERROR_BADLOGONTYPE"; break;
981 case CM_ERROR_GSSCONTINUE : s = "CM_ERROR_GSSCONTINUE"; break;
982 case CM_ERROR_TIDIPC : s = "CM_ERROR_TIDIPC"; break;
983 case CM_ERROR_TOO_MANY_SYMLINKS : s = "CM_ERROR_TOO_MANY_SYMLINKS"; break;
984 case CM_ERROR_PATH_NOT_COVERED : s = "CM_ERROR_PATH_NOT_COVERED"; break;
985 case CM_ERROR_LOCK_CONFLICT : s = "CM_ERROR_LOCK_CONFLICT"; break;
986 case CM_ERROR_SHARING_VIOLATION : s = "CM_ERROR_SHARING_VIOLATION"; break;
987 case CM_ERROR_ALLDOWN : s = "CM_ERROR_ALLDOWN"; break;
988 case CM_ERROR_TOOFEWBUFS : s = "CM_ERROR_TOOFEWBUFS"; break;
989 case CM_ERROR_TOOMANYBUFS : s = "CM_ERROR_TOOMANYBUFS"; break;
991 osi_Log2(afsd_logp, "cm_Analyze: ignoring error code 0x%x (%s)",
997 /* If not allowed to retry, don't */
998 if (!forcing_new && (reqp->flags & CM_REQ_NORETRY) &&
999 (errorCode != RX_MSGSIZE && errorCode != CM_RX_RETRY_BUSY_CALL))
1001 else if (retry && dead_session)
1004 /* drop this on the way out */
1009 * clear the volume updated flag if we succeed.
1010 * this way the flag will not prevent a subsequent volume
1011 * from being updated if necessary.
1015 reqp->flags &= ~CM_REQ_VOLUME_UPDATED;
1018 /* retry until we fail to find a connection */
1022 long cm_ConnByMServers(cm_serverRef_t *serversp, cm_user_t *usersp,
1023 cm_req_t *reqp, cm_conn_t **connpp)
1026 cm_serverRef_t *tsrp;
1028 long firstError = 0;
1029 int someBusy = 0, someOffline = 0, allOffline = 1, allBusy = 1, allDown = 1;
1030 #ifdef SET_RX_TIMEOUTS_TO_TIMELEFT
1031 long timeUsed, timeLeft, hardTimeLeft;
1035 if (serversp == NULL) {
1036 osi_Log1(afsd_logp, "cm_ConnByMServers returning 0x%x", CM_ERROR_ALLDOWN);
1037 return CM_ERROR_ALLDOWN;
1040 #ifdef SET_RX_TIMEOUTS_TO_TIMELEFT
1041 timeUsed = (GetTickCount() - reqp->startTime) / 1000;
1043 /* leave 5 seconds margin of safety */
1044 timeLeft = ConnDeadtimeout - timeUsed - 5;
1045 hardTimeLeft = HardDeadtimeout - timeUsed - 5;
1048 lock_ObtainRead(&cm_serverLock);
1049 for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
1050 if (tsrp->status == srv_deleted)
1054 if (reqp->tokenIdleErrorServp) {
1056 * search the list until we find the server
1057 * that failed last time. When we find it
1058 * clear the error, skip it and try the one
1061 if (tsp == reqp->tokenIdleErrorServp)
1062 reqp->tokenIdleErrorServp = NULL;
1066 cm_GetServerNoLock(tsp);
1067 lock_ReleaseRead(&cm_serverLock);
1068 if (!(tsp->flags & CM_SERVERFLAG_DOWN)) {
1070 if (tsrp->status == srv_busy) {
1073 } else if (tsrp->status == srv_offline) {
1079 code = cm_ConnByServer(tsp, usersp, connpp);
1080 if (code == 0) { /* cm_CBS only returns 0 */
1082 #ifdef SET_RX_TIMEOUTS_TO_TIMELEFT
1083 /* Set RPC timeout */
1084 if (timeLeft > ConnDeadtimeout)
1085 timeLeft = ConnDeadtimeout;
1087 if (hardTimeLeft > HardDeadtimeout)
1088 hardTimeLeft = HardDeadtimeout;
1090 lock_ObtainMutex(&(*connpp)->mx);
1091 rx_SetConnDeadTime((*connpp)->rxconnp, timeLeft);
1092 rx_SetConnHardDeadTime((*connpp)->rxconnp, (u_short) hardTimeLeft);
1093 lock_ReleaseMutex(&(*connpp)->mx);
1098 /* therefore, this code is never executed */
1099 if (firstError == 0)
1103 lock_ObtainRead(&cm_serverLock);
1104 cm_PutServerNoLock(tsp);
1107 lock_ReleaseRead(&cm_serverLock);
1109 if (firstError == 0) {
1111 firstError = (reqp->tokenError ? reqp->tokenError :
1112 (reqp->idleError ? RX_CALL_TIMEOUT : CM_ERROR_ALLDOWN));
1114 * if we experienced either a token error or and idle dead time error
1115 * and now all of the servers are down, we have either tried them
1116 * all or lost connectivity. Clear the error we are returning so
1117 * we will not return it indefinitely if the request is retried.
1119 reqp->idleError = reqp->tokenError = 0;
1120 } else if (allBusy) {
1121 firstError = CM_ERROR_ALLBUSY;
1122 } else if (allOffline || (someBusy && someOffline)) {
1123 firstError = CM_ERROR_ALLOFFLINE;
1125 osi_Log0(afsd_logp, "cm_ConnByMServers returning impossible error TIMEDOUT");
1126 firstError = CM_ERROR_TIMEDOUT;
1130 osi_Log1(afsd_logp, "cm_ConnByMServers returning 0x%x", firstError);
1134 /* called with a held server to GC all bad connections hanging off of the server */
1135 void cm_GCConnections(cm_server_t *serverp)
1141 lock_ObtainWrite(&cm_connLock);
1142 lcpp = &serverp->connsp;
1143 for (tcp = *lcpp; tcp; tcp = *lcpp) {
1145 if (userp && tcp->refCount == 0 && (userp->vcRefs == 0)) {
1146 /* do the deletion of this guy */
1147 cm_PutServer(tcp->serverp);
1148 cm_ReleaseUser(userp);
1150 rx_SetConnSecondsUntilNatPing(tcp->rxconnp, 0);
1151 rx_DestroyConnection(tcp->rxconnp);
1152 lock_FinalizeMutex(&tcp->mx);
1156 /* just advance to the next */
1160 lock_ReleaseWrite(&cm_connLock);
1163 static void cm_NewRXConnection(cm_conn_t *tcp, cm_ucell_t *ucellp,
1164 cm_server_t *serverp)
1166 unsigned short port;
1169 struct rx_securityClass *secObjp;
1171 port = serverp->addr.sin_port;
1172 switch (serverp->type) {
1173 case CM_SERVER_VLDB:
1178 case CM_SERVER_FILE:
1184 osi_panic("unknown server type", __FILE__, __LINE__);
1187 if (ucellp->flags & CM_UCELLFLAG_RXKAD) {
1191 tcp->cryptlevel = rxkad_clear;
1194 tcp->cryptlevel = rxkad_auth;
1197 tcp->cryptlevel = rxkad_crypt;
1199 secObjp = rxkad_NewClientSecurityObject(tcp->cryptlevel,
1200 &ucellp->sessionKey, ucellp->kvno,
1201 ucellp->ticketLen, ucellp->ticketp);
1205 tcp->cryptlevel = rxkad_clear;
1206 secObjp = rxnull_NewClientSecurityObject();
1208 osi_assertx(secObjp != NULL, "null rx_securityClass");
1209 tcp->rxconnp = rx_NewConnection(serverp->addr.sin_addr.s_addr,
1214 rx_SetConnDeadTime(tcp->rxconnp, ConnDeadtimeout);
1215 rx_SetConnHardDeadTime(tcp->rxconnp, HardDeadtimeout);
1216 rx_SetConnIdleDeadTime(tcp->rxconnp, IdleDeadtimeout);
1219 * Register the error to be returned on an idle dead timeout
1221 rx_SetServerConnIdleDeadErr(tcp->rxconnp, RX_CALL_DEAD);
1224 * Let the Rx library know that we can auto-retry if an
1225 * RX_MSGSIZE error is returned.
1227 if (rx_pmtu_discovery)
1228 rx_SetMsgsizeRetryErr(tcp->rxconnp, RX_MSGSIZE);
1231 * Attempt to limit NAT pings to the anonymous file server connections.
1232 * Only file servers implement client callbacks and we only need one ping
1233 * to be sent to each server.
1235 if (NatPingInterval && serverp->type == CM_SERVER_FILE && secIndex == 0)
1236 rx_SetConnSecondsUntilNatPing(tcp->rxconnp, NatPingInterval);
1238 tcp->ucgen = ucellp->gen;
1240 rxs_Release(secObjp); /* Decrement the initial refCount */
1243 long cm_ConnByServer(cm_server_t *serverp, cm_user_t *userp, cm_conn_t **connpp)
1250 if (cm_anonvldb && serverp->type == CM_SERVER_VLDB)
1251 userp = cm_rootUserp;
1253 lock_ObtainMutex(&userp->mx);
1254 lock_ObtainRead(&cm_connLock);
1255 for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
1256 if (tcp->userp == userp)
1260 /* find ucell structure */
1261 ucellp = cm_GetUCell(userp, serverp->cellp);
1263 lock_ConvertRToW(&cm_connLock);
1264 for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
1265 if (tcp->userp == userp)
1269 lock_ReleaseWrite(&cm_connLock);
1272 cm_GetServer(serverp);
1273 tcp = malloc(sizeof(*tcp));
1274 memset(tcp, 0, sizeof(*tcp));
1275 tcp->nextp = serverp->connsp;
1276 serverp->connsp = tcp;
1279 lock_InitializeMutex(&tcp->mx, "cm_conn_t mutex", LOCK_HIERARCHY_CONN);
1280 lock_ObtainMutex(&tcp->mx);
1281 tcp->serverp = serverp;
1282 tcp->cryptlevel = rxkad_clear;
1283 cm_NewRXConnection(tcp, ucellp, serverp);
1285 lock_ReleaseMutex(&tcp->mx);
1286 lock_ReleaseWrite(&cm_connLock);
1288 lock_ReleaseRead(&cm_connLock);
1290 InterlockedIncrement(&tcp->refCount);
1292 lock_ObtainMutex(&tcp->mx);
1293 if ((tcp->flags & CM_CONN_FLAG_FORCE_NEW) ||
1294 (tcp->ucgen < ucellp->gen) ||
1295 (tcp->cryptlevel != (ucellp->flags & CM_UCELLFLAG_RXKAD ? (cryptall == 1 ? rxkad_crypt : (cryptall == 2 ? rxkad_auth : rxkad_clear)) : rxkad_clear)))
1297 if (tcp->ucgen < ucellp->gen)
1298 osi_Log0(afsd_logp, "cm_ConnByServer replace connection due to token update");
1300 osi_Log0(afsd_logp, "cm_ConnByServer replace connection due to crypt change");
1301 tcp->flags &= ~CM_CONN_FLAG_FORCE_NEW;
1302 rx_SetConnSecondsUntilNatPing(tcp->rxconnp, 0);
1303 rx_DestroyConnection(tcp->rxconnp);
1304 cm_NewRXConnection(tcp, ucellp, serverp);
1306 lock_ReleaseMutex(&tcp->mx);
1308 lock_ReleaseMutex(&userp->mx);
1310 /* return this pointer to our caller */
1311 osi_Log1(afsd_logp, "cm_ConnByServer returning conn 0x%p", tcp);
1317 long cm_ServerAvailable(struct cm_fid *fidp, struct cm_user *userp)
1321 cm_serverRef_t **serverspp;
1322 cm_serverRef_t *tsrp;
1324 int someBusy = 0, someOffline = 0, allOffline = 1, allBusy = 1, allDown = 1;
1328 code = cm_GetServerList(fidp, userp, &req, &serverspp);
1332 lock_ObtainRead(&cm_serverLock);
1333 for (tsrp = *serverspp; tsrp; tsrp=tsrp->next) {
1334 if (tsrp->status == srv_deleted)
1337 if (!(tsp->flags & CM_SERVERFLAG_DOWN)) {
1339 if (tsrp->status == srv_busy) {
1342 } else if (tsrp->status == srv_offline) {
1351 lock_ReleaseRead(&cm_serverLock);
1352 cm_FreeServerList(serverspp, 0);
1358 else if (allOffline || (someBusy && someOffline))
1365 * The returned cm_conn_t ** object is released in the subsequent call
1368 long cm_ConnFromFID(struct cm_fid *fidp, struct cm_user *userp, cm_req_t *reqp,
1372 cm_serverRef_t **serverspp;
1376 code = cm_GetServerList(fidp, userp, reqp, &serverspp);
1381 code = cm_ConnByMServers(*serverspp, userp, reqp, connpp);
1382 cm_FreeServerList(serverspp, 0);
1387 long cm_ConnFromVolume(struct cm_volume *volp, unsigned long volid, struct cm_user *userp, cm_req_t *reqp,
1391 cm_serverRef_t **serverspp;
1395 serverspp = cm_GetVolServers(volp, volid, userp, reqp);
1397 code = cm_ConnByMServers(*serverspp, userp, reqp, connpp);
1398 cm_FreeServerList(serverspp, 0);
1403 extern struct rx_connection *
1404 cm_GetRxConn(cm_conn_t *connp)
1406 struct rx_connection * rxconnp;
1407 lock_ObtainMutex(&connp->mx);
1408 rxconnp = connp->rxconnp;
1409 rx_GetConnection(rxconnp);
1410 lock_ReleaseMutex(&connp->mx);
1414 void cm_ForceNewConnections(cm_server_t *serverp)
1418 lock_ObtainWrite(&cm_connLock);
1419 for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
1420 lock_ObtainMutex(&tcp->mx);
1421 tcp->flags |= CM_CONN_FLAG_FORCE_NEW;
1422 lock_ReleaseMutex(&tcp->mx);
1424 lock_ReleaseWrite(&cm_connLock);