#define LANMAN_WKS_PARAM_KEY "SYSTEM\\CurrentControlSet\\Services\\lanmanworkstation\\parameters"
#define LANMAN_WKS_SESSION_TIMEOUT "SessTimeout"
+#define LANMAN_WKS_EXT_SESSION_TIMEOUT "ExtendedSessTimeout"
afs_uint32 cryptall = 0;
afs_uint32 cm_anonvldb = 0;
0, KEY_QUERY_VALUE, &parmKey);
if (code == ERROR_SUCCESS)
{
- dummyLen = sizeof(DWORD);
- code = RegQueryValueEx(parmKey, LANMAN_WKS_SESSION_TIMEOUT, NULL, NULL,
- (BYTE *) &dwValue, &dummyLen);
- if (code == ERROR_SUCCESS)
- RDRtimeout = dwValue;
+ BOOL extTimeouts = msftSMBRedirectorSupportsExtendedTimeouts();
+
+ if (extTimeouts) {
+ /*
+ * The default value is 1000 seconds. However, this default
+ * will not apply to "\\AFS" unless "AFS" is listed in
+ * ServersWithExtendedSessTimeout which we will add when we
+ * create the ExtendedSessTimeout value in smb_Init()
+ */
+ dummyLen = sizeof(DWORD);
+ code = RegQueryValueEx(parmKey,
+ LANMAN_WKS_EXT_SESSION_TIMEOUT,
+ NULL, NULL,
+ (BYTE *) &dwValue, &dummyLen);
+ if (code == ERROR_SUCCESS) {
+ RDRtimeout = dwValue;
+ afsi_log("lanmanworkstation : ExtSessTimeout %u", RDRtimeout);
+ }
+ }
+ if (!extTimeouts || code != ERROR_SUCCESS) {
+ dummyLen = sizeof(DWORD);
+ code = RegQueryValueEx(parmKey,
+ LANMAN_WKS_SESSION_TIMEOUT,
+ NULL, NULL,
+ (BYTE *) &dwValue, &dummyLen);
+ if (code == ERROR_SUCCESS) {
+ RDRtimeout = dwValue;
+ afsi_log("lanmanworkstation : SessTimeout %u", RDRtimeout);
+ }
+ }
RegCloseKey(parmKey);
}
RegCloseKey(parmKey);
}
- afsi_log("lanmanworkstation : SessTimeout %u", RDRtimeout);
+ /*
+ * If these values were not set via cpp macro or obtained
+ * from the registry, we use a value that is derived from
+ * the smb redirector session timeout (RDRtimeout).
+ *
+ * The UNIX cache manager uses 120 seconds for the hard dead
+ * timeout and 50 seconds for the connection and idle timeouts.
+ *
+ * We base our values on those while making sure we leave
+ * enough time for overhead.
+ */
if (ConnDeadtimeout == 0) {
- ConnDeadtimeout = (unsigned short) (RDRtimeout / 2);
+ ConnDeadtimeout = (unsigned short) ((RDRtimeout / 2) < 50 ? (RDRtimeout / 2) : 50);
afsi_log("ConnDeadTimeout is %d", ConnDeadtimeout);
}
if (HardDeadtimeout == 0) {
- HardDeadtimeout = (unsigned short) RDRtimeout;
+ HardDeadtimeout = (unsigned short) (RDRtimeout > 125 ? 120 : (RDRtimeout - 5));
afsi_log("HardDeadTimeout is %d", HardDeadtimeout);
}
- if (ConnDeadtimeout == 0) {
- IdleDeadtimeout = (unsigned short) RDRtimeout;
+ if (IdleDeadtimeout == 0) {
+ IdleDeadtimeout = (unsigned short) ConnDeadtimeout;
afsi_log("IdleDeadTimeout is %d", IdleDeadtimeout);
}
osi_EndOnce(&once);
if (code)
return code;
- *serversppp = cm_GetVolServers(volp, fidp->volume);
+ *serversppp = cm_GetVolServers(volp, fidp->volume, userp, reqp);
lock_ObtainRead(&cm_volumeLock);
cm_PutVolume(volp);
lock_ReleaseRead(&cm_volumeLock);
- return 0;
+ return (*serversppp ? 0 : CM_ERROR_NOSUCHVOLUME);
}
/*
/* if timeout - check that it did not exceed the HardDead timeout
* and retry */
- /* timeleft - get if from reqp the same way as cmXonnByMServers does */
+ /* timeleft - get it from reqp the same way as cm_ConnByMServers does */
timeUsed = (GetTickCount() - reqp->startTime) / 1000;
-
- /* leave 5 seconds margin for sleep */
- if (reqp->flags & CM_REQ_NORETRY)
- timeLeft = 0;
- else
- timeLeft = HardDeadtimeout - timeUsed;
+ timeLeft = HardDeadtimeout - timeUsed;
/* get a pointer to the cell */
if (errorCode) {
if (cellp == NULL && serversp) {
struct cm_serverRef * refp;
for ( refp=serversp ; cellp == NULL && refp != NULL; refp=refp->next) {
+ if (refp->status == srv_deleted)
+ continue;
if ( refp->server )
cellp = refp->server->cellp;
}
}
lock_ObtainWrite(&cm_serverLock);
for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == srv_deleted)
+ continue;
if (tsrp->status == srv_busy) {
tsrp->status = srv_not_busy;
}
}
lock_ReleaseWrite(&cm_serverLock);
if (free_svr_list) {
- cm_FreeServerList(&serversp, 0);
- *serverspp = serversp;
+ cm_FreeServerList(serverspp, 0);
+ serverspp = NULL;
+ serversp = NULL;
+ free_svr_list = 0;
}
cm_UpdateVolumeStatus(volp, fidp->volume);
if (serversp) {
lock_ObtainWrite(&cm_serverLock);
for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == srv_deleted)
+ continue;
if (tsrp->status == srv_busy) {
tsrp->status = srv_not_busy;
}
}
lock_ObtainWrite(&cm_serverLock);
for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == srv_deleted)
+ continue;
if (tsrp->server == serverp && tsrp->status == srv_not_busy) {
tsrp->status = srv_busy;
if (fidp) { /* File Server query */
}
if (free_svr_list) {
- cm_FreeServerList(&serversp, 0);
- *serverspp = serversp;
+ cm_FreeServerList(serverspp, 0);
+ serverspp = NULL;
+ serversp = NULL;
+ free_svr_list = 0;
}
retry = 1;
}
else if (errorCode == VNOVOL || errorCode == VMOVED || errorCode == VOFFLINE ||
errorCode == VSALVAGE || errorCode == VNOSERVICE || errorCode == VIO)
{
- char addr[16];
char *format;
DWORD msgID;
+
+ /* In case of timeout */
+ reqp->volumeError = errorCode;
+
switch ( errorCode ) {
case VNOVOL:
msgID = MSG_SERVER_REPORTS_VNOVOL;
LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume);
}
- /* Mark server offline for this volume */
+ /*
+ * Mark server offline for this volume or delete the volume
+ * from the server list if it was moved or is not present.
+ */
if (!serversp && fidp) {
code = cm_GetServerList(fidp, userp, reqp, &serverspp);
if (code == 0) {
lock_ObtainWrite(&cm_serverLock);
for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == srv_deleted)
+ continue;
+
+ sprintf(addr, "%d.%d.%d.%d",
+ ((tsrp->server->addr.sin_addr.s_addr & 0xff)),
+ ((tsrp->server->addr.sin_addr.s_addr & 0xff00)>> 8),
+ ((tsrp->server->addr.sin_addr.s_addr & 0xff0000)>> 16),
+ ((tsrp->server->addr.sin_addr.s_addr & 0xff000000)>> 24));
+
if (tsrp->server == serverp) {
/* REDIRECT */
if (errorCode == VMOVED || errorCode == VNOVOL) {
+ osi_Log2(afsd_logp, "volume %d not present on server %s",
+ fidp->volume, osi_LogSaveString(afsd_logp,addr));
tsrp->status = srv_deleted;
if (fidp)
cm_RemoveVolumeFromServer(serverp, fidp->volume);
} else {
+ osi_Log2(afsd_logp, "volume %d instance on server %s marked offline",
+ fidp->volume, osi_LogSaveString(afsd_logp,addr));
tsrp->status = srv_offline;
}
-
- if (fidp) { /* File Server query */
- lock_ReleaseWrite(&cm_serverLock);
- code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
- CM_GETVOL_FLAG_NO_LRU_UPDATE,
- &volp);
- if (code == 0)
- cm_VolumeStateByID(volp, fidp->volume);
- lock_ObtainWrite(&cm_serverLock);
- }
+ /* break; */
+ } else {
+ osi_Log3(afsd_logp, "volume %d exists on server %s with status %u",
+ fidp->volume, osi_LogSaveString(afsd_logp,addr), tsrp->status);
}
}
lock_ReleaseWrite(&cm_serverLock);
- if (fidp && (errorCode == VMOVED || errorCode == VNOVOL)) {
- code = cm_ForceUpdateVolume(fidp, userp, reqp);
- if (code)
- timeLeft = 0; /* prevent a retry on failure */
+ /* Free the server list before cm_ForceUpdateVolume is called */
+ if (free_svr_list) {
+ cm_FreeServerList(serverspp, 0);
+ serverspp = NULL;
+ serversp = NULL;
+ free_svr_list = 0;
}
- if (statep) {
- cm_UpdateVolumeStatus(volp, statep->ID);
- lock_ObtainRead(&cm_volumeLock);
- cm_PutVolume(volp);
- lock_ReleaseRead(&cm_volumeLock);
- volp = NULL;
- }
+ if (fidp) { /* File Server query */
+ code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
+ CM_GETVOL_FLAG_NO_LRU_UPDATE,
+ &volp);
+ if (code == 0)
+ statep = cm_VolumeStateByID(volp, fidp->volume);
+
+ if ((errorCode == VMOVED || errorCode == VNOVOL) &&
+ !(reqp->flags & CM_REQ_VOLUME_UPDATED))
+ {
+ code = cm_ForceUpdateVolume(fidp, userp, reqp);
+ if (code)
+ timeLeft = 0; /* prevent a retry on failure */
+ else
+ reqp->flags |= CM_REQ_VOLUME_UPDATED;
+ osi_Log3(afsd_logp, "cm_Analyze called cm_ForceUpdateVolume cell %u vol %u code 0x%x",
+ fidp->cell, fidp->volume, code);
+ }
- if (free_svr_list) {
- cm_FreeServerList(&serversp, 0);
- *serverspp = serversp;
+ if (statep) {
+ cm_UpdateVolumeStatus(volp, statep->ID);
+ osi_Log3(afsd_logp, "cm_Analyze NewVolState cell %u vol %u state %u",
+ fidp->cell, fidp->volume, statep->state);
+ }
+
+ if (volp) {
+ lock_ObtainRead(&cm_volumeLock);
+ cm_PutVolume(volp);
+ lock_ReleaseRead(&cm_volumeLock);
+ volp = NULL;
+ }
}
+
if ( timeLeft > 2 )
retry = 1;
} else if ( errorCode == VNOVNODE ) {
/* RX codes */
else if (errorCode == RX_CALL_TIMEOUT) {
- /* server took longer than hardDeadTime
+ /* RPC took longer than hardDeadTime or the server
+ * reported idle for longer than idleDeadTime
* 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
*/
if (serverp) {
- /* 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));
-
- LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_HARD_DEAD_TIME_EXCEEDED, addr);
-
- osi_Log1(afsd_logp, "cm_Analyze: hardDeadTime exceeded addr[%s]",
- osi_LogSaveString(afsd_logp,addr));
- reqp->tokenIdleErrorServp = serverp;
- reqp->idleError++;
- retry = 1;
- }
+ 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));
+
+ LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_HARD_DEAD_TIME_EXCEEDED, addr);
+ osi_Log1(afsd_logp, "cm_Analyze: hardDeadTime or idleDeadtime exceeded addr[%s]",
+ osi_LogSaveString(afsd_logp,addr));
+ reqp->tokenIdleErrorServp = serverp;
+ reqp->idleError++;
+
+ if (timeLeft > 2) {
+ if (!fidp) { /* vldb */
+ retry = 1;
+ } else { /* file */
+ cm_volume_t *volp = cm_GetVolumeByFID(fidp);
+ if (volp) {
+ if (fidp->volume == cm_GetROVolumeID(volp))
+ retry = 1;
+ cm_PutVolume(volp);
+ }
+ }
+ }
+ }
}
else if (errorCode >= -64 && errorCode < 0) {
/* mark server as down */
- lock_ObtainMutex(&serverp->mx);
- if (reqp->flags & CM_REQ_NEW_CONN_FORCED) {
- if (!(serverp->flags & CM_SERVERFLAG_DOWN)) {
- serverp->flags |= CM_SERVERFLAG_DOWN;
- serverp->downTime = time(NULL);
+ if (serverp)
+ 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));
+
+ if (errorCode == RX_CALL_DEAD)
+ osi_Log2(afsd_logp, "cm_Analyze: Rx Call Dead addr[%s] forcedNew[%s]",
+ osi_LogSaveString(afsd_logp,addr),
+ (reqp->flags & CM_REQ_NEW_CONN_FORCED ? "yes" : "no"));
+ else
+ osi_Log3(afsd_logp, "cm_Analyze: Rx Misc Error[%d] addr[%s] forcedNew[%s]",
+ errorCode,
+ osi_LogSaveString(afsd_logp,addr),
+ (reqp->flags & CM_REQ_NEW_CONN_FORCED ? "yes" : "no"));
+
+ if (serverp) {
+ lock_ObtainMutex(&serverp->mx);
+ if (errorCode == RX_CALL_DEAD &&
+ (reqp->flags & CM_REQ_NEW_CONN_FORCED)) {
+ if (!(serverp->flags & CM_SERVERFLAG_DOWN)) {
+ serverp->flags |= CM_SERVERFLAG_DOWN;
+ serverp->downTime = time(NULL);
+ }
+ } else {
+ if (reqp->flags & CM_REQ_NEW_CONN_FORCED) {
+ reqp->tokenIdleErrorServp = serverp;
+ reqp->tokenError = errorCode;
+ } else {
+ reqp->flags |= CM_REQ_NEW_CONN_FORCED;
+ forcing_new = 1;
+ }
}
- } else {
- reqp->flags |= CM_REQ_NEW_CONN_FORCED;
- forcing_new = 1;
- }
- lock_ReleaseMutex(&serverp->mx);
- cm_ForceNewConnections(serverp);
+ lock_ReleaseMutex(&serverp->mx);
+ cm_ForceNewConnections(serverp);
+ }
if ( timeLeft > 2 )
retry = 1;
}
else if (errorCode == RXKADEXPIRED) {
+ osi_Log1(afsd_logp, "cm_Analyze: rxkad error code 0x%x (RXKADEXPIRED)",
+ errorCode);
if (!dead_session) {
lock_ObtainMutex(&userp->mx);
ucellp = cm_GetUCell(userp, serverp->cellp);
retry = 1;
}
} else if (errorCode >= ERROR_TABLE_BASE_RXK && errorCode < ERROR_TABLE_BASE_RXK + 256) {
+ 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;
+ }
+ osi_Log2(afsd_logp, "cm_Analyze: rxkad error code 0x%x (%s)",
+ errorCode, s);
+
if (serverp) {
- reqp->tokenIdleErrorServp = serverp;
- reqp->tokenError = errorCode;
- retry = 1;
+ reqp->tokenIdleErrorServp = serverp;
+ reqp->tokenError = errorCode;
+ retry = 1;
+ }
+ } else if (errorCode >= ERROR_TABLE_BASE_U && errorCode < ERROR_TABLE_BASE_U + 256) {
+ /*
+ * We received a ubik error. its possible that the server we are
+ * communicating with has a corrupted database or is partitioned
+ * from the rest of the servers and another server might be able
+ * to answer our query. Therefore, we will retry the request
+ * and force the use of another server.
+ */
+ if (serverp) {
+ reqp->tokenIdleErrorServp = serverp;
+ reqp->tokenError = errorCode;
+ retry = 1;
}
} else if (errorCode == VICECONNBAD || errorCode == VICETOKENDEAD) {
cm_ForceNewConnections(serverp);
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;
if (connp)
cm_PutConn(connp);
+ /*
+ * clear the volume updated flag if we succeed.
+ * this way the flag will not prevent a subsequent volume
+ * from being updated if necessary.
+ */
+ if (errorCode == 0)
+ {
+ reqp->flags &= ~CM_REQ_VOLUME_UPDATED;
+ }
+
/* retry until we fail to find a connection */
return retry;
}
lock_ObtainRead(&cm_serverLock);
for (tsrp = serversp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == srv_deleted)
+ continue;
+
tsp = tsrp->server;
if (reqp->tokenIdleErrorServp) {
/*
if (tsp) {
cm_GetServerNoLock(tsp);
lock_ReleaseRead(&cm_serverLock);
- if ((tsrp->status != srv_deleted) && !(tsp->flags & CM_SERVERFLAG_DOWN)) {
+ if (!(tsp->flags & CM_SERVERFLAG_DOWN)) {
allDown = 0;
if (tsrp->status == srv_busy) {
allOffline = 0;
int secIndex;
struct rx_securityClass *secObjp;
- if (serverp->type == CM_SERVER_VLDB) {
- port = htons(7003);
+ port = serverp->addr.sin_port;
+ switch (serverp->type) {
+ case CM_SERVER_VLDB:
+ if (port == 0)
+ port = htons(7003);
serviceID = 52;
- }
- else {
- osi_assertx(serverp->type == CM_SERVER_FILE, "incorrect server type");
- port = htons(7000);
+ break;
+ case CM_SERVER_FILE:
+ if (port == 0)
+ port = htons(7000);
serviceID = 1;
+ break;
+ default:
+ osi_panic("unknown server type", __FILE__, __LINE__);
}
+
if (ucellp->flags & CM_UCELLFLAG_RXKAD) {
secIndex = 2;
switch (cryptall) {
}
osi_assertx(secObjp != NULL, "null rx_securityClass");
tcp->rxconnp = rx_NewConnection(serverp->addr.sin_addr.s_addr,
- port,
- serviceID,
- secObjp,
- secIndex);
+ port,
+ serviceID,
+ secObjp,
+ secIndex);
rx_SetConnDeadTime(tcp->rxconnp, ConnDeadtimeout);
rx_SetConnHardDeadTime(tcp->rxconnp, HardDeadtimeout);
rx_SetConnIdleDeadTime(tcp->rxconnp, IdleDeadtimeout);
lock_ObtainRead(&cm_serverLock);
for (tsrp = *serverspp; tsrp; tsrp=tsrp->next) {
+ if (tsrp->status == srv_deleted)
+ continue;
tsp = tsrp->server;
- if ((tsrp->status != srv_deleted) && !(tsp->flags & CM_SERVERFLAG_DOWN)) {
+ if (!(tsp->flags & CM_SERVERFLAG_DOWN)) {
allDown = 0;
if (tsrp->status == srv_busy) {
allOffline = 0;
*connpp = NULL;
- serverspp = cm_GetVolServers(volp, volid);
+ serverspp = cm_GetVolServers(volp, volid, userp, reqp);
code = cm_ConnByMServers(*serverspp, userp, reqp, connpp);
cm_FreeServerList(serverspp, 0);
git://git.openafs.org / openafs.git / blobdiff