#include <malloc.h>
#include <osi.h>
#include <rx/rx.h>
-#ifndef DJGPP
-#include <rx/rxkad.h>
-#else
#include <rx/rxkad.h>
-#endif
-
#include "afsd.h"
osi_rwlock_t cm_connLock;
long RDRtimeout = CM_CONN_DEFAULTRDRTIMEOUT;
+unsigned short ConnDeadtimeout = CM_CONN_CONNDEADTIME;
+unsigned short HardDeadtimeout = CM_CONN_HARDDEADTIME;
#define LANMAN_WKS_PARAM_KEY "SYSTEM\\CurrentControlSet\\Services\\lanmanworkstation\\parameters"
#define LANMAN_WKS_SESSION_TIMEOUT "SessTimeout"
lock_InitializeRWLock(&cm_connLock, "connection global lock");
/* keisa - read timeout value for lanmanworkstation service.
- * It is used as hardtimeout for connections.
- * Default value is 45
+ * jaltman - as per
+ * http://support.microsoft.com:80/support/kb/articles/Q102/0/67.asp&NoWebContent=1
+ * the SessTimeout is a minimum timeout not a maximum timeout. Therefore,
+ * I believe that the default should not be short. Instead, we should wait until
+ * RX times out before reporting a timeout to the SMB client.
*/
code = RegOpenKeyEx(HKEY_LOCAL_MACHINE, LANMAN_WKS_PARAM_KEY,
0, KEY_QUERY_VALUE, &parmKey);
{
afsi_log("lanmanworkstation : SessTimeout %d", sessTimeout);
RDRtimeout = sessTimeout;
- }
- else
- {
- RDRtimeout = CM_CONN_DEFAULTRDRTIMEOUT;
+ if ( ConnDeadtimeout < RDRtimeout + 15 ) {
+ ConnDeadtimeout = RDRtimeout + 15;
+ afsi_log("ConnDeadTimeout increased to %d", ConnDeadtimeout);
+ }
+ if ( HardDeadtimeout < 2 * ConnDeadtimeout ) {
+ HardDeadtimeout = 2 * ConnDeadtimeout;
+ afsi_log("HardDeadTimeout increased to %d", HardDeadtimeout);
+ }
}
}
-
+
osi_EndOnce(&once);
}
}
{
memset((char *)reqp, 0, sizeof(cm_req_t));
#ifndef DJGPP
- reqp->startTime = GetCurrentTime();
+ reqp->startTime = GetTickCount();
#else
gettimeofday(&reqp->startTime, NULL);
#endif
-
}
-long cm_GetServerList(struct cm_fid *fidp, struct cm_user *userp,
- struct cm_req *reqp, cm_serverRef_t **serverspp)
+static long cm_GetServerList(struct cm_fid *fidp, struct cm_user *userp,
+ struct cm_req *reqp, cm_serverRef_t ***serversppp)
{
- long code;
- cm_volume_t *volp = NULL;
- cm_serverRef_t *serversp = NULL;
- cm_cell_t *cellp = NULL;
+ long code;
+ cm_volume_t *volp = NULL;
+ cm_cell_t *cellp = NULL;
- if (!fidp) {
- *serverspp = NULL;
- return 0;
- }
+ if (!fidp) {
+ *serversppp = NULL;
+ return 0;
+ }
- cellp = cm_FindCellByID(fidp->cell);
- if (!cellp) return CM_ERROR_NOSUCHCELL;
+ cellp = cm_FindCellByID(fidp->cell);
+ if (!cellp) return CM_ERROR_NOSUCHCELL;
- code = cm_GetVolumeByID(cellp, fidp->volume, userp, reqp, &volp);
- if (code) return code;
-
- if (fidp->volume == volp->rwID)
- serversp = volp->rwServersp;
- else if (fidp->volume == volp->roID)
- serversp = volp->roServersp;
- else if (fidp->volume == volp->bkID)
- serversp = volp->bkServersp;
- else
- serversp = NULL;
-
- cm_PutVolume(volp);
- *serverspp = serversp;
- return 0;
+ code = cm_GetVolumeByID(cellp, fidp->volume, userp, reqp, &volp);
+ if (code) return code;
+
+ *serversppp = cm_GetVolServers(volp, fidp->volume);
+
+ cm_PutVolume(volp);
+ return 0;
}
/*
*/
int
cm_Analyze(cm_conn_t *connp, cm_user_t *userp, cm_req_t *reqp,
- struct cm_fid *fidp,
- AFSVolSync *volSyncp, cm_callbackRequest_t *cbrp, long errorCode)
+ struct cm_fid *fidp,
+ AFSVolSync *volSyncp,
+ cm_serverRef_t * serversp,
+ cm_callbackRequest_t *cbrp, long errorCode)
{
- cm_server_t *serverp;
- cm_serverRef_t *serversp, *tsrp;
- cm_ucell_t *ucellp;
- int retry = 0;
- int dead_session;
-
- osi_Log2(afsd_logp, "cm_Analyze connp 0x%x, code %d",
- (long) connp, errorCode);
-
- /* no locking required, since connp->serverp never changes after
- * creation */
- dead_session = (userp->cellInfop == NULL);
- if (connp)
- serverp = connp->serverp;
-
- /* Update callback pointer */
- if (cbrp && errorCode == 0) cbrp->serverp = connp->serverp;
-
- /* If not allowed to retry, don't */
- if (reqp->flags & CM_REQ_NORETRY)
- goto out;
-
- /* if timeout - check that is did not exceed the SMB timeout
- and retry */
- if (errorCode == CM_ERROR_TIMEDOUT)
- {
- long timeUsed, timeLeft;
- /* timeleft - get if from reqp the same way as cmXonnByMServers does */
+ cm_server_t *serverp = NULL;
+ cm_serverRef_t **serverspp = NULL;
+ cm_serverRef_t *tsrp;
+ cm_cell_t *cellp = NULL;
+ cm_ucell_t *ucellp;
+ int retry = 0;
+ int free_svr_list = 0;
+ int dead_session;
+ long timeUsed, timeLeft;
+ long code;
+ char addr[16];
+
+ osi_Log2(afsd_logp, "cm_Analyze connp 0x%p, code 0x%x",
+ connp, errorCode);
+
+ /* no locking required, since connp->serverp never changes after
+ * creation */
+ dead_session = (userp->cellInfop == NULL);
+ if (connp)
+ serverp = connp->serverp;
+
+ /* Update callback pointer */
+ if (cbrp && serverp && errorCode == 0) {
+ if (cbrp->serverp) {
+ if ( cbrp->serverp != serverp ) {
+ lock_ObtainWrite(&cm_serverLock);
+ cm_PutServerNoLock(cbrp->serverp);
+ cm_GetServerNoLock(serverp);
+ lock_ReleaseWrite(&cm_serverLock);
+ }
+ } else {
+ cm_GetServer(serverp);
+ }
+ lock_ObtainWrite(&cm_callbackLock);
+ cbrp->serverp = serverp;
+ lock_ReleaseWrite(&cm_callbackLock);
+ }
+
+ /* If not allowed to retry, don't */
+ if (reqp->flags & CM_REQ_NORETRY)
+ goto out;
+
+ /* if timeout - check that it did not exceed the SMB timeout
+ * and retry */
+
+ /* timeleft - get if from reqp the same way as cmXonnByMServers does */
#ifndef DJGPP
- timeUsed = (GetCurrentTime() - reqp->startTime) / 1000;
+ timeUsed = (GetTickCount() - reqp->startTime) / 1000;
#else
- gettimeofday(&now, NULL);
- timeUsed = sub_time(now, reqp->startTime) / 1000;
+ gettimeofday(&now, NULL);
+ timeUsed = sub_time(now, reqp->startTime) / 1000;
#endif
- /* leave 5 seconds margin for sleep */
- timeLeft = RDRtimeout - timeUsed;
- if (timeLeft > 5)
- {
+ /* leave 5 seconds margin for sleep */
+ timeLeft = RDRtimeout - timeUsed;
+
+ if (errorCode == CM_ERROR_TIMEDOUT) {
+ if (timeLeft > 5 ) {
thrd_Sleep(3000);
- cm_CheckServers(CM_FLAG_CHECKDOWNSERVERS, NULL);
+ if (cellp == NULL && serverp)
+ cellp = serverp->cellp;
+ if (cellp == NULL && serversp) {
+ struct cm_serverRef * refp;
+ for ( refp=serversp ; cellp == NULL && refp != NULL; refp=refp->next) {
+ if ( refp->server )
+ cellp = refp->server->cellp;
+ }
+ }
+ cm_CheckServers(CM_FLAG_CHECKDOWNSERVERS, cellp);
+ retry = 1;
+ }
+ }
+
+ /* if there is nosuchvolume, then we have a situation in which a
+ * previously known volume no longer has a set of servers
+ * associated with it. Either the volume has moved or
+ * the volume has been deleted. Try to find a new server list
+ * until the timeout period expires.
+ */
+ else if (errorCode == CM_ERROR_NOSUCHVOLUME) {
+ osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_NOSUCHVOLUME.");
+ if (timeLeft > 7) {
+ thrd_Sleep(5000);
+
retry = 1;
- }
+
+ if (fidp != NULL) /* Not a VLDB call */
+ cm_ForceUpdateVolume(fidp, userp, reqp);
+ }
}
- /* if all servers are offline, mark them non-busy and start over */
- if (errorCode == CM_ERROR_ALLOFFLINE) {
- osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLOFFLINE.");
- thrd_Sleep(5000);
- /* cm_ForceUpdateVolume marks all servers as non_busy */
- cm_ForceUpdateVolume(fidp, userp, reqp);
- retry = 1;
- }
+ else if (errorCode == CM_ERROR_ALLDOWN) {
+ osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLDOWN.");
+ /* Servers marked DOWN will be restored by the background daemon
+ * thread as they become available.
+ */
+ }
- /* if all servers are busy, mark them non-busy and start over */
- if (errorCode == CM_ERROR_ALLBUSY) {
- cm_GetServerList(fidp, userp, reqp, &serversp);
- for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
- if (tsrp->status == busy)
- tsrp->status = not_busy;
+ else if (errorCode == CM_ERROR_ALLOFFLINE) {
+ if (timeLeft > 7) {
+ osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLOFFLINE.");
+ thrd_Sleep(5000);
+
+ if (fidp) { /* Not a VLDB call */
+ if (!serversp) {
+ code = cm_GetServerList(fidp, userp, reqp, &serverspp);
+ if (code == 0) {
+ serversp = *serverspp;
+ free_svr_list = 1;
+ }
}
- thrd_Sleep(5000);
- retry = 1;
- }
-
- /* special codes: VBUSY and VRESTARTING */
- if (errorCode == VBUSY || errorCode == VRESTARTING) {
- cm_GetServerList(fidp, userp, reqp, &serversp);
- for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
- if (tsrp->server == serverp
- && tsrp->status == not_busy) {
- tsrp->status = busy;
- break;
- }
+ if (serversp) {
+ lock_ObtainWrite(&cm_serverLock);
+ for (tsrp = serversp; tsrp; tsrp=tsrp->next)
+ tsrp->status = not_busy;
+ lock_ReleaseWrite(&cm_serverLock);
+ if (free_svr_list) {
+ cm_FreeServerList(&serversp);
+ *serverspp = serversp;
+ }
+ retry = 1;
}
- retry = 1;
- }
- /* special codes: missing volumes */
- if (errorCode == VNOVOL || errorCode == VMOVED || errorCode == VOFFLINE
- || errorCode == VSALVAGE || errorCode == VNOSERVICE) {
- /* Log server being offline for this volume */
- osi_Log4(afsd_logp, "cm_Analyze found server %d.%d.%d.%d marked offline for a volume",
- ((serverp->addr.sin_addr.s_addr & 0xff)),
- ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
- ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
- ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
- /* Create Event Log message */
- {
- HANDLE h;
- char *ptbuf[1];
- char s[100];
- h = RegisterEventSource(NULL, AFS_DAEMON_EVENT_NAME);
- sprintf(s, "cm_Analyze: Server %d.%d.%d.%d reported volume %d as missing.",
- ((serverp->addr.sin_addr.s_addr & 0xff)),
- ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
- ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
- ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24),
- fidp->volume);
- ptbuf[0] = s;
- ReportEvent(h, EVENTLOG_WARNING_TYPE, 0, 1009, NULL,
- 1, 0, ptbuf, NULL);
- DeregisterEventSource(h);
+ cm_ForceUpdateVolume(fidp, userp, reqp);
+ } else { /* VLDB call */
+ if (serversp) {
+ lock_ObtainWrite(&cm_serverLock);
+ for (tsrp = serversp; tsrp; tsrp=tsrp->next)
+ tsrp->status = not_busy;
+ lock_ReleaseWrite(&cm_serverLock);
+ if (free_svr_list) {
+ cm_FreeServerList(&serversp);
+ *serverspp = serversp;
+ }
}
+ }
+ }
+ }
- /* Mark server offline for this volume */
- cm_GetServerList(fidp, userp, reqp, &serversp);
+ /* if all servers are busy, mark them non-busy and start over */
+ else if (errorCode == CM_ERROR_ALLBUSY) {
+ osi_Log0(afsd_logp, "cm_Analyze passed CM_ERROR_ALLBUSY.");
+ if (timeLeft > 7) {
+ thrd_Sleep(5000);
+ if (!serversp) {
+ code = cm_GetServerList(fidp, userp, reqp, &serverspp);
+ if (code == 0) {
+ serversp = *serverspp;
+ free_svr_list = 1;
+ }
+ }
+ lock_ObtainWrite(&cm_serverLock);
+ for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == busy)
+ tsrp->status = not_busy;
+ }
+ lock_ReleaseWrite(&cm_serverLock);
+ if (free_svr_list) {
+ cm_FreeServerList(&serversp);
+ *serverspp = serversp;
+ }
+ retry = 1;
+ }
+ }
- for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
- if (tsrp->server == serverp)
- tsrp->status = offline;
- }
- retry = 1;
+ /* special codes: VBUSY and VRESTARTING */
+ else if (errorCode == VBUSY || errorCode == VRESTARTING) {
+ if (!serversp) {
+ code = cm_GetServerList(fidp, userp, reqp, &serverspp);
+ if (code == 0) {
+ serversp = *serverspp;
+ free_svr_list = 1;
+ }
+ }
+ lock_ObtainWrite(&cm_serverLock);
+ for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->server == serverp
+ && tsrp->status == not_busy) {
+ tsrp->status = busy;
+ break;
+ }
+ }
+ lock_ReleaseWrite(&cm_serverLock);
+ if (free_svr_list) {
+ cm_FreeServerList(&serversp);
+ *serverspp = serversp;
+ }
+ retry = 1;
+ }
+
+ /* special codes: missing volumes */
+ else if (errorCode == VNOVOL || errorCode == VMOVED || errorCode == VOFFLINE ||
+ errorCode == VSALVAGE || errorCode == VNOSERVICE || errorCode == VIO)
+ {
+ char addr[16];
+ char *format;
+#ifndef DJGPP
+ DWORD msgID;
+#endif
+ switch ( errorCode ) {
+ case VNOVOL:
+#ifndef DJGPP
+ msgID = MSG_SERVER_REPORTS_VNOVOL;
+#endif
+ format = "Server %s reported volume %d as not attached.";
+ break;
+ case VMOVED:
+#ifndef DJGPP
+ msgID = MSG_SERVER_REPORTS_VMOVED;
+#endif
+ format = "Server %s reported volume %d as moved.";
+ break;
+ case VOFFLINE:
+#ifndef DJGPP
+ msgID = MSG_SERVER_REPORTS_VOFFLINE;
+#endif
+ format = "Server %s reported volume %d as offline.";
+ break;
+ case VSALVAGE:
+#ifndef DJGPP
+ msgID = MSG_SERVER_REPORTS_VSALVAGE;
+#endif
+ format = "Server %s reported volume %d as needs salvage.";
+ break;
+ case VNOSERVICE:
+#ifndef DJGPP
+ msgID = MSG_SERVER_REPORTS_VNOSERVICE;
+#endif
+ format = "Server %s reported volume %d as not in service.";
+ break;
+ case VIO:
+#ifndef DJGPP
+ msgID = MSG_SERVER_REPORTS_VIO;
+#endif
+ format = "Server %s reported volume %d as temporarily unaccessible.";
+ break;
+ }
+
+ /* Log server being offline for this volume */
+ sprintf(addr, "%d.%d.%d.%d",
+ ((serverp->addr.sin_addr.s_addr & 0xff)),
+ ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
+ ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
+ ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
+
+ osi_Log2(afsd_logp, format, osi_LogSaveString(afsd_logp,addr), fidp->volume);
+#ifndef DJGPP
+ LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume);
+#endif
+
+ /* Mark server offline for this volume */
+ if (!serversp) {
+ code = cm_GetServerList(fidp, userp, reqp, &serverspp);
+ if (code == 0) {
+ serversp = *serverspp;
+ free_svr_list = 1;
+ }
+ }
+ for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->server == serverp)
+ tsrp->status = offline;
+ }
+ if (free_svr_list) {
+ cm_FreeServerList(&serversp);
+ *serverspp = serversp;
+ }
+ if ( timeLeft > 2 )
+ retry = 1;
+ } else if ( errorCode == VNOVNODE ) {
+ if ( fidp ) {
+ cm_scache_t * scp;
+ osi_Log4(afsd_logp, "cm_Analyze passed VNOVNODE cell %u vol %u vn %u uniq %u.",
+ fidp->cell, fidp->volume, fidp->vnode, fidp->unique);
+ if (!cm_GetSCache(fidp, &scp, userp, reqp)) {
+ cm_FlushParent(scp, userp, reqp);
+ cm_FlushFile(scp, userp, reqp);
+ cm_ReleaseSCache(scp);
+ }
+ } else {
+ osi_Log0(afsd_logp, "cm_Analyze passed VNOVNODE unknown fid.");
}
+ }
+
+ /* RX codes */
+ else if (errorCode == RX_CALL_TIMEOUT) {
+ /* server took longer than hardDeadTime
+ * don't mark server as down but don't retry
+ * this is to prevent the SMB session from timing out
+ * In addition, we log an event to the event log
+ */
+
+ /* Log server being offline for this volume */
+ sprintf(addr, "%d.%d.%d.%d",
+ ((serverp->addr.sin_addr.s_addr & 0xff)),
+ ((serverp->addr.sin_addr.s_addr & 0xff00)>> 8),
+ ((serverp->addr.sin_addr.s_addr & 0xff0000)>> 16),
+ ((serverp->addr.sin_addr.s_addr & 0xff000000)>> 24));
- /* RX codes */
- if (errorCode == RX_CALL_TIMEOUT) {
- /* server took longer than hardDeadTime
- * don't mark server as down but don't retry
- * this is to prevent the SMB session from timing out
- * In addition, we log an event to the event log
- */
#ifndef DJGPP
- HANDLE h;
- char *ptbuf[1];
- char s[100];
- h = RegisterEventSource(NULL, AFS_DAEMON_EVENT_NAME);
- sprintf(s, "cm_Analyze: HardDeadTime exceeded.");
- ptbuf[0] = s;
- ReportEvent(h, EVENTLOG_WARNING_TYPE, 0, 1009, NULL,
- 1, 0, ptbuf, NULL);
- DeregisterEventSource(h);
+ LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_HARD_DEAD_TIME_EXCEEDED, addr);
#endif /* !DJGPP */
- retry = 0;
- osi_Log0(afsd_logp, "cm_Analyze: hardDeadTime exceeded");
- }
- else if (errorCode >= -64 && errorCode < 0) {
- /* mark server as down */
- lock_ObtainMutex(&serverp->mx);
- serverp->flags |= CM_SERVERFLAG_DOWN;
- lock_ReleaseMutex(&serverp->mx);
+ retry = 0;
+ osi_Log1(afsd_logp, "cm_Analyze: hardDeadTime exceeded addr[%s]",
+ osi_LogSaveString(afsd_logp,addr));
+ }
+ else if (errorCode >= -64 && errorCode < 0) {
+ /* mark server as down */
+ lock_ObtainMutex(&serverp->mx);
+ serverp->flags |= CM_SERVERFLAG_DOWN;
+ lock_ReleaseMutex(&serverp->mx);
+ cm_ForceNewConnections(serverp);
+ if ( timeLeft > 2 )
+ retry = 1;
+ }
+ else if (errorCode == RXKADEXPIRED ||
+ errorCode == RXKADBADTICKET) {
+ if (!dead_session) {
+ lock_ObtainMutex(&userp->mx);
+ ucellp = cm_GetUCell(userp, serverp->cellp);
+ if (ucellp->ticketp) {
+ free(ucellp->ticketp);
+ ucellp->ticketp = NULL;
+ }
+ ucellp->flags &= ~CM_UCELLFLAG_RXKAD;
+ ucellp->gen++;
+ lock_ReleaseMutex(&userp->mx);
+ if ( timeLeft > 2 )
retry = 1;
}
+ } else {
+ if (errorCode) {
+ char * s = "unknown error";
+ switch ( errorCode ) {
+ case RXKADINCONSISTENCY: s = "RXKADINCONSISTENCY"; break;
+ case RXKADPACKETSHORT : s = "RXKADPACKETSHORT"; break;
+ case RXKADLEVELFAIL : s = "RXKADLEVELFAIL"; break;
+ case RXKADTICKETLEN : s = "RXKADTICKETLEN"; break;
+ case RXKADOUTOFSEQUENCE: s = "RXKADOUTOFSEQUENCE"; break;
+ case RXKADNOAUTH : s = "RXKADNOAUTH"; break;
+ case RXKADBADKEY : s = "RXKADBADKEY"; break;
+ case RXKADBADTICKET : s = "RXKADBADTICKET"; break;
+ case RXKADUNKNOWNKEY : s = "RXKADUNKNOWNKEY"; break;
+ case RXKADEXPIRED : s = "RXKADEXPIRED"; break;
+ case RXKADSEALEDINCON : s = "RXKADSEALEDINCON"; break;
+ case RXKADDATALEN : s = "RXKADDATALEN"; break;
+ case RXKADILLEGALLEVEL : s = "RXKADILLEGALLEVEL"; break;
+ case VSALVAGE : s = "VSALVAGE"; break;
+ case VNOVNODE : s = "VNOVNODE"; break;
+ case VNOVOL : s = "VNOVOL"; break;
+ case VVOLEXISTS : s = "VVOLEXISTS"; break;
+ case VNOSERVICE : s = "VNOSERVICE"; break;
+ case VOFFLINE : s = "VOFFLINE"; break;
+ case VONLINE : s = "VONLINE"; break;
+ case VDISKFULL : s = "VDISKFULL"; break;
+ case VOVERQUOTA : s = "VOVERQUOTA"; break;
+ case VBUSY : s = "VBUSY"; break;
+ case VMOVED : s = "VMOVED"; break;
+ case VIO : s = "VIO"; break;
+ case VRESTRICTED : s = "VRESTRICTED"; break;
+ case VRESTARTING : s = "VRESTARTING"; break;
+ case VREADONLY : s = "VREADONLY"; break;
+ case EAGAIN : s = "EAGAIN"; break;
+ case EACCES : s = "EACCES"; break;
+ }
+ osi_Log2(afsd_logp, "cm_Analyze: ignoring error code 0x%x (%s)",
+ errorCode, s);
+ retry = 0;
+ }
+ }
- if (errorCode == RXKADEXPIRED && !dead_session) {
- lock_ObtainMutex(&userp->mx);
- ucellp = cm_GetUCell(userp, serverp->cellp);
- if (ucellp->ticketp) {
- free(ucellp->ticketp);
- ucellp->ticketp = NULL;
- }
- ucellp->flags &= ~CM_UCELLFLAG_RXKAD;
- ucellp->gen++;
- lock_ReleaseMutex(&userp->mx);
- retry = 1;
- }
+ if (retry && dead_session)
+ retry = 0;
- if (retry && dead_session)
- retry = 0;
-
-out:
- /* drop this on the way out */
- if (connp)
- cm_PutConn(connp);
+ out:
+ /* drop this on the way out */
+ if (connp)
+ cm_PutConn(connp);
- /* retry until we fail to find a connection */
- return retry;
+ /* retry until we fail to find a connection */
+ return retry;
}
long cm_ConnByMServers(cm_serverRef_t *serversp, cm_user_t *usersp,
cm_req_t *reqp, cm_conn_t **connpp)
{
- long code;
- cm_serverRef_t *tsrp;
- cm_server_t *tsp;
- long firstError = 0;
- int someBusy = 0, someOffline = 0, allDown = 1;
- long timeUsed, timeLeft, hardTimeLeft;
+ long code;
+ cm_serverRef_t *tsrp;
+ cm_server_t *tsp;
+ long firstError = 0;
+ int someBusy = 0, someOffline = 0, allOffline = 1, allBusy = 1, allDown = 1;
+ long timeUsed, timeLeft, hardTimeLeft;
#ifdef DJGPP
- struct timeval now;
+ struct timeval now;
#endif /* DJGPP */
- *connpp = NULL;
+ if (serversp == NULL) {
+ osi_Log1(afsd_logp, "cm_ConnByMServers returning 0x%x", CM_ERROR_NOSUCHVOLUME);
+ return CM_ERROR_NOSUCHVOLUME;
+ }
+
+ *connpp = NULL;
#ifndef DJGPP
- timeUsed = (GetCurrentTime() - reqp->startTime) / 1000;
+ timeUsed = (GetTickCount() - reqp->startTime) / 1000;
#else
- gettimeofday(&now, NULL);
- timeUsed = sub_time(now, reqp->startTime) / 1000;
+ gettimeofday(&now, NULL);
+ timeUsed = sub_time(now, reqp->startTime) / 1000;
#endif
- /* leave 5 seconds margin of safety */
- timeLeft = RDRtimeout - timeUsed - 5;
- hardTimeLeft = timeLeft;
+ /* leave 5 seconds margin of safety */
+ timeLeft = ConnDeadtimeout - timeUsed - 5;
+ hardTimeLeft = HardDeadtimeout - timeUsed - 5;
- /* Time enough to do an RPC? */
- if (timeLeft < 1) {
- return CM_ERROR_TIMEDOUT;
- }
-
- lock_ObtainWrite(&cm_serverLock);
-
- for(tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ lock_ObtainWrite(&cm_serverLock);
+ for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
tsp = tsrp->server;
- tsp->refCount++;
+ cm_GetServerNoLock(tsp);
lock_ReleaseWrite(&cm_serverLock);
if (!(tsp->flags & CM_SERVERFLAG_DOWN)) {
- allDown = 0;
- if (tsrp->status == busy)
+ allDown = 0;
+ if (tsrp->status == busy) {
+ allOffline = 0;
someBusy = 1;
- else if (tsrp->status == offline)
- someOffline = 1;
- else {
+ } else if (tsrp->status == offline) {
+ allBusy = 0;
+ someOffline = 1;
+ } else {
+ allOffline = 0;
+ allBusy = 0;
code = cm_ConnByServer(tsp, usersp, connpp);
- if (code == 0) {
+ if (code == 0) { /* cm_CBS only returns 0 */
cm_PutServer(tsp);
/* Set RPC timeout */
- if (timeLeft > CM_CONN_CONNDEADTIME)
- timeLeft = CM_CONN_CONNDEADTIME;
+ if (timeLeft > ConnDeadtimeout)
+ timeLeft = ConnDeadtimeout;
- if (hardTimeLeft > CM_CONN_HARDDEADTIME)
- hardTimeLeft = CM_CONN_HARDDEADTIME;
+ if (hardTimeLeft > HardDeadtimeout)
+ hardTimeLeft = HardDeadtimeout;
lock_ObtainMutex(&(*connpp)->mx);
- rx_SetConnDeadTime((*connpp)->callp,
- timeLeft);
- rx_SetConnHardDeadTime((*connpp)->callp,
- (u_short) hardTimeLeft);
+ rx_SetConnDeadTime((*connpp)->callp, timeLeft);
+ rx_SetConnHardDeadTime((*connpp)->callp, (u_short) hardTimeLeft);
lock_ReleaseMutex(&(*connpp)->mx);
-
return 0;
}
- if (firstError == 0)
+
+ /* therefore, this code is never executed */
+ if (firstError == 0)
firstError = code;
}
- }
+ }
lock_ObtainWrite(&cm_serverLock);
- osi_assert(tsp->refCount-- > 0);
+ cm_PutServerNoLock(tsp);
}
+ lock_ReleaseWrite(&cm_serverLock);
+
+ if (firstError == 0) {
+ if (allDown)
+ firstError = CM_ERROR_ALLDOWN;
+ else if (allBusy)
+ firstError = CM_ERROR_ALLBUSY;
+ else if (allOffline || (someBusy && someOffline))
+ firstError = CM_ERROR_ALLOFFLINE;
+ else {
+ osi_Log0(afsd_logp, "cm_ConnByMServers returning impossible error TIMEDOUT");
+ firstError = CM_ERROR_TIMEDOUT;
+ }
+ }
- lock_ReleaseWrite(&cm_serverLock);
- if (firstError == 0) {
- if (someBusy)
- firstError = CM_ERROR_ALLBUSY;
- else if (someOffline)
- firstError = CM_ERROR_ALLOFFLINE;
- else if (!allDown && serversp)
- firstError = CM_ERROR_TIMEDOUT;
- /* Only return CM_ERROR_NOSUCHVOLUME if there are no
- servers for this volume */
- else
- firstError = CM_ERROR_NOSUCHVOLUME;
- }
- osi_Log1(afsd_logp, "cm_ConnByMServers returning %x", firstError);
+ osi_Log1(afsd_logp, "cm_ConnByMServers returning 0x%x", firstError);
return firstError;
}
/* called with a held server to GC all bad connections hanging off of the server */
void cm_GCConnections(cm_server_t *serverp)
{
- cm_conn_t *tcp;
- cm_conn_t **lcpp;
- cm_user_t *userp;
-
- lock_ObtainWrite(&cm_connLock);
- lcpp = &serverp->connsp;
- for(tcp = *lcpp; tcp; tcp = *lcpp) {
- userp = tcp->userp;
- if (userp && tcp->refCount == 0 && (userp->vcRefs == 0)) {
- /* do the deletion of this guy */
- cm_ReleaseUser(userp);
- *lcpp = tcp->nextp;
- rx_DestroyConnection(tcp->callp);
- lock_FinalizeMutex(&tcp->mx);
- free(tcp);
- }
- else {
- /* just advance to the next */
- lcpp = &tcp->nextp;
- }
+ cm_conn_t *tcp;
+ cm_conn_t **lcpp;
+ cm_user_t *userp;
+
+ lock_ObtainWrite(&cm_connLock);
+ lcpp = &serverp->connsp;
+ for (tcp = *lcpp; tcp; tcp = *lcpp) {
+ userp = tcp->userp;
+ if (userp && tcp->refCount == 0 && (userp->vcRefs == 0)) {
+ /* do the deletion of this guy */
+ cm_PutServer(tcp->serverp);
+ cm_ReleaseUser(userp);
+ *lcpp = tcp->nextp;
+ rx_DestroyConnection(tcp->callp);
+ lock_FinalizeMutex(&tcp->mx);
+ free(tcp);
}
- lock_ReleaseWrite(&cm_connLock);
+ else {
+ /* just advance to the next */
+ lcpp = &tcp->nextp;
+ }
+ }
+ lock_ReleaseWrite(&cm_connLock);
}
static void cm_NewRXConnection(cm_conn_t *tcp, cm_ucell_t *ucellp,
- cm_server_t *serverp)
+ cm_server_t *serverp)
{
- unsigned short port;
- int serviceID;
- int secIndex;
- struct rx_securityClass *secObjp;
- afs_int32 level;
-
- if (serverp->type == CM_SERVER_VLDB) {
- port = htons(7003);
- serviceID = 52;
- }
- else {
- osi_assert(serverp->type == CM_SERVER_FILE);
- port = htons(7000);
- serviceID = 1;
- }
- if (ucellp->flags & CM_UCELLFLAG_RXKAD) {
- secIndex = 2;
- if (cryptall) {
- level = rxkad_crypt;
- tcp->cryptlevel = rxkad_crypt;
- } else {
- level = rxkad_clear;
- }
- secObjp = rxkad_NewClientSecurityObject(level,
- &ucellp->sessionKey, ucellp->kvno,
- ucellp->ticketLen, ucellp->ticketp);
- }
- else {
- /* normal auth */
- secIndex = 0;
- secObjp = rxnull_NewClientSecurityObject();
+ unsigned short port;
+ int serviceID;
+ int secIndex;
+ struct rx_securityClass *secObjp;
+
+ if (serverp->type == CM_SERVER_VLDB) {
+ port = htons(7003);
+ serviceID = 52;
+ }
+ else {
+ osi_assert(serverp->type == CM_SERVER_FILE);
+ port = htons(7000);
+ serviceID = 1;
+ }
+ if (ucellp->flags & CM_UCELLFLAG_RXKAD) {
+ secIndex = 2;
+ if (cryptall) {
+ tcp->cryptlevel = rxkad_crypt;
+ } else {
+ tcp->cryptlevel = rxkad_clear;
}
- osi_assert(secObjp != NULL);
- tcp->callp = rx_NewConnection(serverp->addr.sin_addr.s_addr,
- port,
- serviceID,
- secObjp,
- secIndex);
- rx_SetConnDeadTime(tcp->callp, CM_CONN_CONNDEADTIME);
- rx_SetConnHardDeadTime(tcp->callp, CM_CONN_HARDDEADTIME);
- tcp->ucgen = ucellp->gen;
+ secObjp = rxkad_NewClientSecurityObject(tcp->cryptlevel,
+ &ucellp->sessionKey, ucellp->kvno,
+ ucellp->ticketLen, ucellp->ticketp);
+ } else {
+ /* normal auth */
+ secIndex = 0;
+ tcp->cryptlevel = rxkad_clear;
+ secObjp = rxnull_NewClientSecurityObject();
+ }
+ osi_assert(secObjp != NULL);
+ tcp->callp = rx_NewConnection(serverp->addr.sin_addr.s_addr,
+ port,
+ serviceID,
+ secObjp,
+ secIndex);
+ rx_SetConnDeadTime(tcp->callp, ConnDeadtimeout);
+ rx_SetConnHardDeadTime(tcp->callp, HardDeadtimeout);
+ tcp->ucgen = ucellp->gen;
if (secObjp)
rxs_Release(secObjp); /* Decrement the initial refCount */
}
long cm_ConnByServer(cm_server_t *serverp, cm_user_t *userp, cm_conn_t **connpp)
{
- cm_conn_t *tcp;
- cm_ucell_t *ucellp;
-
- lock_ObtainMutex(&userp->mx);
- lock_ObtainWrite(&cm_connLock);
- for(tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
- if (tcp->userp == userp) break;
- }
- /* find ucell structure */
- ucellp = cm_GetUCell(userp, serverp->cellp);
- if (!tcp) {
- tcp = malloc(sizeof(*tcp));
- memset(tcp, 0, sizeof(*tcp));
- tcp->nextp = serverp->connsp;
- serverp->connsp = tcp;
- tcp->userp = userp;
- cm_HoldUser(userp);
- lock_InitializeMutex(&tcp->mx, "cm_conn_t mutex");
- tcp->serverp = serverp;
- tcp->cryptlevel = rxkad_clear;
- cm_NewRXConnection(tcp, ucellp, serverp);
- tcp->refCount = 1;
+ cm_conn_t *tcp;
+ cm_ucell_t *ucellp;
+
+ lock_ObtainMutex(&userp->mx);
+ lock_ObtainWrite(&cm_connLock);
+ for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
+ if (tcp->userp == userp)
+ break;
+ }
+
+ /* find ucell structure */
+ ucellp = cm_GetUCell(userp, serverp->cellp);
+ if (!tcp) {
+ cm_GetServer(serverp);
+ tcp = malloc(sizeof(*tcp));
+ memset(tcp, 0, sizeof(*tcp));
+ tcp->nextp = serverp->connsp;
+ serverp->connsp = tcp;
+ cm_HoldUser(userp);
+ tcp->userp = userp;
+ lock_InitializeMutex(&tcp->mx, "cm_conn_t mutex");
+ lock_ObtainMutex(&tcp->mx);
+ tcp->serverp = serverp;
+ tcp->cryptlevel = rxkad_clear;
+ cm_NewRXConnection(tcp, ucellp, serverp);
+ tcp->refCount = 1;
+ lock_ReleaseMutex(&tcp->mx);
+ } else {
+ if ((tcp->flags & CM_CONN_FLAG_FORCE_NEW) ||
+ (tcp->ucgen < ucellp->gen) ||
+ (tcp->cryptlevel != (cryptall ? (ucellp->flags & CM_UCELLFLAG_RXKAD ? rxkad_crypt : rxkad_clear) : rxkad_clear)))
+ {
+ if (tcp->ucgen < ucellp->gen)
+ osi_Log0(afsd_logp, "cm_ConnByServer replace connection due to token update");
+ else
+ osi_Log0(afsd_logp, "cm_ConnByServer replace connection due to crypt change");
+ lock_ObtainMutex(&tcp->mx);
+ tcp->flags &= ~CM_CONN_FLAG_FORCE_NEW;
+ rx_DestroyConnection(tcp->callp);
+ cm_NewRXConnection(tcp, ucellp, serverp);
+ lock_ReleaseMutex(&tcp->mx);
}
- else {
- if ((tcp->ucgen < ucellp->gen) || (tcp->cryptlevel != cryptall))
- {
- rx_DestroyConnection(tcp->callp);
- cm_NewRXConnection(tcp, ucellp, serverp);
- }
- tcp->refCount++;
- }
- lock_ReleaseWrite(&cm_connLock);
- lock_ReleaseMutex(&userp->mx);
+ tcp->refCount++;
+ }
+ lock_ReleaseWrite(&cm_connLock);
+ lock_ReleaseMutex(&userp->mx);
- /* return this pointer to our caller */
- osi_Log1(afsd_logp, "cm_ConnByServer returning conn 0x%x", (long) tcp);
- *connpp = tcp;
+ /* return this pointer to our caller */
+ osi_Log1(afsd_logp, "cm_ConnByServer returning conn 0x%p", tcp);
+ *connpp = tcp;
- return 0;
+ return 0;
}
long cm_Conn(struct cm_fid *fidp, struct cm_user *userp, cm_req_t *reqp,
- cm_conn_t **connpp)
+ cm_conn_t **connpp)
{
- long code;
+ long code;
- cm_serverRef_t *serversp;
+ cm_serverRef_t **serverspp;
- code = cm_GetServerList(fidp, userp, reqp, &serversp);
- if (code) {
- *connpp = NULL;
- return code;
- }
-
- code = cm_ConnByMServers(serversp, userp, reqp, connpp);
+ code = cm_GetServerList(fidp, userp, reqp, &serverspp);
+ if (code) {
+ *connpp = NULL;
return code;
+ }
+
+ code = cm_ConnByMServers(*serverspp, userp, reqp, connpp);
+ cm_FreeServerList(serverspp);
+ return code;
+}
+
+extern struct rx_connection *
+cm_GetRxConn(cm_conn_t *connp)
+{
+ struct rx_connection * rxconn;
+ lock_ObtainMutex(&connp->mx);
+ rxconn = connp->callp;
+ rx_GetConnection(rxconn);
+ lock_ReleaseMutex(&connp->mx);
+ return rxconn;
+}
+
+void cm_ForceNewConnections(cm_server_t *serverp)
+{
+ cm_conn_t *tcp;
+
+ lock_ObtainWrite(&cm_connLock);
+ for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
+ lock_ObtainMutex(&tcp->mx);
+ tcp->flags |= CM_CONN_FLAG_FORCE_NEW;
+ lock_ReleaseMutex(&tcp->mx);
+ }
+ lock_ReleaseWrite(&cm_connLock);
}
git://git.openafs.org / openafs.git / blobdiff