/*
* Copyright 2000, International Business Machines Corporation and others.
* All Rights Reserved.
- *
+ *
* This software has been released under the terms of the IBM Public
* License. For details, see the LICENSE file in the top-level source
* directory or online at http://www.openafs.org/dl/license10.html
*/
+#include <afsconfig.h>
#include <afs/param.h>
+#include <roken.h>
+
#include <afs/stds.h>
#include <windows.h>
afs_uint32 cryptall = 0;
afs_uint32 cm_anonvldb = 0;
+afs_uint32 rx_pmtu_discovery = 0;
void cm_PutConn(cm_conn_t *connp)
{
DWORD dwValue;
DWORD dummyLen;
HKEY parmKey;
-
+
if (osi_Once(&once)) {
lock_InitializeRWLock(&cm_connLock, "connection global lock",
LOCK_HIERARCHY_CONN_GLOBAL);
/* keisa - read timeout value for lanmanworkstation service.
- * jaltman - as per
+ * 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,
+ * 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.
*/
BOOL extTimeouts = msftSMBRedirectorSupportsExtendedTimeouts();
if (extTimeouts) {
- /*
+ /*
* The default value is 1000 seconds. However, this default
- * will not apply to "\\AFS" unless "AFS" is listed in
+ * 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,
+ code = RegQueryValueEx(parmKey,
LANMAN_WKS_EXT_SESSION_TIMEOUT,
NULL, NULL,
(BYTE *) &dwValue, &dummyLen);
RDRtimeout = dwValue;
afsi_log("lanmanworkstation : ExtSessTimeout %u", RDRtimeout);
}
- }
+ }
if (!extTimeouts || code != ERROR_SUCCESS) {
dummyLen = sizeof(DWORD);
- code = RegQueryValueEx(parmKey,
+ code = RegQueryValueEx(parmKey,
LANMAN_WKS_SESSION_TIMEOUT,
NULL, NULL,
(BYTE *) &dwValue, &dummyLen);
RegCloseKey(parmKey);
}
- /*
- * If these values were not set via cpp macro or obtained
+ /*
+ * 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.
+ * We base our values on those while making sure we leave
+ * enough time for overhead.
*/
- if (ConnDeadtimeout == 0) {
- ConnDeadtimeout = (unsigned short) ((RDRtimeout / 2) < 50 ? (RDRtimeout / 2) : 50);
- afsi_log("ConnDeadTimeout is %d", ConnDeadtimeout);
- }
- if (HardDeadtimeout == 0) {
- HardDeadtimeout = (unsigned short) (RDRtimeout > 125 ? 120 : (RDRtimeout - 5));
- afsi_log("HardDeadTimeout is %d", HardDeadtimeout);
- }
- if (IdleDeadtimeout == 0) {
- IdleDeadtimeout = (unsigned short) ConnDeadtimeout;
- afsi_log("IdleDeadTimeout is %d", IdleDeadtimeout);
+ if (smb_Enabled) {
+ afsi_log("lanmanworkstation : SessTimeout %u", RDRtimeout);
+ if (ConnDeadtimeout == 0) {
+ ConnDeadtimeout = (unsigned short) ((RDRtimeout / 2) < 50 ? (RDRtimeout / 2) : 50);
+ afsi_log("ConnDeadTimeout is %d", ConnDeadtimeout);
+ }
+ if (HardDeadtimeout == 0) {
+ HardDeadtimeout = (unsigned short) (RDRtimeout > 125 ? 120 : (RDRtimeout - 5));
+ afsi_log("HardDeadTimeout is %d", HardDeadtimeout);
+ }
+ if (IdleDeadtimeout == 0) {
+ IdleDeadtimeout = (unsigned short) ConnDeadtimeout;
+ afsi_log("IdleDeadTimeout is %d", IdleDeadtimeout);
+ }
+ } else {
+ if (ConnDeadtimeout == 0) {
+ ConnDeadtimeout = CM_CONN_IFS_CONNDEADTIME;
+ afsi_log("ConnDeadTimeout is %d", ConnDeadtimeout);
+ }
+ if (HardDeadtimeout == 0) {
+ HardDeadtimeout = CM_CONN_IFS_HARDDEADTIME;
+ afsi_log("HardDeadTimeout is %d", HardDeadtimeout);
+ }
+ if (IdleDeadtimeout == 0) {
+ IdleDeadtimeout = CM_CONN_IFS_IDLEDEADTIME;
+ afsi_log("IdleDeadTimeout is %d", IdleDeadtimeout);
+ }
}
osi_EndOnce(&once);
}
}
cellp = cm_FindCellByID(fidp->cell, 0);
- if (!cellp)
+ if (!cellp)
return CM_ERROR_NOSUCHCELL;
code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp, CM_GETVOL_FLAG_CREATE, &volp);
- if (code)
+ if (code)
return code;
-
+
*serversppp = cm_GetVolServers(volp, fidp->volume, userp, reqp);
lock_ObtainRead(&cm_volumeLock);
*/
int
cm_Analyze(cm_conn_t *connp, cm_user_t *userp, cm_req_t *reqp,
- struct cm_fid *fidp,
- AFSVolSync *volSyncp,
+ struct cm_fid *fidp,
+ AFSVolSync *volSyncp,
cm_serverRef_t * serversp,
cm_callbackRequest_t *cbrp, long errorCode)
{
long code;
char addr[16]="unknown";
int forcing_new = 0;
+ int location_updated = 0;
char *format;
DWORD msgID;
/* if timeout - check that it did not exceed the HardDead timeout
* and retry */
-
+
/* timeleft - get it from reqp the same way as cm_ConnByMServers does */
timeUsed = (GetTickCount() - reqp->startTime) / 1000;
- timeLeft = HardDeadtimeout - timeUsed;
+ if ( reqp->flags & CM_REQ_SOURCE_SMB )
+ timeLeft = HardDeadtimeout - timeUsed;
+ else
+ timeLeft = 0x0FFFFFFF;
/* get a pointer to the cell */
if (errorCode) {
if ( refp->server )
cellp = refp->server->cellp;
}
- }
+ }
if (cellp == NULL && fidp) {
cellp = cm_FindCellByID(fidp->cell, 0);
}
}
}
- /* if there is nosuchvolume, then we have a situation in which a
- * previously known volume no longer has a set of servers
+ /* 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.");
- /*
+ /*
* The VNOVOL or VL_NOENT error has already been translated
* to CM_ERROR_NOSUCHVOLUME. There is nothing for us to do.
*/
else if (errorCode == CM_ERROR_ALLDOWN) {
/* Servers marked DOWN will be restored by the background daemon
- * thread as they become available. The volume status is
+ * thread as they become available. The volume status is
* updated as the server state changes.
*/
if (fidp) {
}
else if (errorCode == CM_ERROR_ALLOFFLINE) {
- /* Volume instances marked offline will be restored by the
- * background daemon thread as they become available
+ /* Volume instances marked offline will be restored by the
+ * background daemon thread as they become available
*/
if (fidp) {
osi_Log2(afsd_logp, "cm_Analyze passed CM_ERROR_ALLOFFLINE (FS cell %s vol 0x%x)",
format = "All servers are offline when accessing cell %s volume %d.";
LogEvent(EVENTLOG_WARNING_TYPE, msgID, cellp->name, fidp->volume);
- code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
- CM_GETVOL_FLAG_NO_LRU_UPDATE,
+ code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
+ CM_GETVOL_FLAG_NO_LRU_UPDATE,
&volp);
if (code == 0) {
- if (timeLeft > 7) {
+ /*
+ * Do not perform a cm_CheckOfflineVolume() if cm_Analyze()
+ * was called by cm_CheckOfflineVolumeState().
+ */
+ if (!(reqp->flags & CM_REQ_OFFLINE_VOL_CHK) && timeLeft > 7) {
thrd_Sleep(5000);
/* cm_CheckOfflineVolume() resets the serverRef state */
}
}
else if (errorCode == CM_ERROR_ALLBUSY) {
- /* Volumes that are busy cannot be determined to be non-busy
+ /* Volumes that are busy cannot be determined to be non-busy
* without actually attempting to access them.
*/
if (fidp) { /* File Server query */
format = "All servers are busy when accessing cell %s volume %d.";
LogEvent(EVENTLOG_WARNING_TYPE, msgID, cellp->name, fidp->volume);
- code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
- CM_GETVOL_FLAG_NO_LRU_UPDATE,
+ code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
+ CM_GETVOL_FLAG_NO_LRU_UPDATE,
&volp);
if (code == 0) {
if (timeLeft > 7) {
thrd_Sleep(5000);
-
+
statep = cm_VolumeStateByID(volp, fidp->volume);
- if (statep->state != vl_offline &&
+ if (statep->state != vl_offline &&
statep->state != vl_busy &&
statep->state != vl_unknown) {
retry = 1;
continue;
if (tsrp->status == srv_busy) {
tsrp->status = srv_not_busy;
- }
+ }
}
lock_ReleaseWrite(&cm_serverLock);
if (free_svr_list) {
cm_FreeServerList(serverspp, 0);
- serverspp = NULL;
serversp = NULL;
free_svr_list = 0;
}
switch ( errorCode ) {
case VBUSY:
msgID = MSG_SERVER_REPORTS_VBUSY;
- format = "Server %s reported busy when accessing volume %d.";
+ format = "Server %s reported busy when accessing volume %d in cell %s.";
break;
case VRESTARTING:
msgID = MSG_SERVER_REPORTS_VRESTARTING;
- format = "Server %s reported restarting when accessing volume %d.";
+ format = "Server %s reported restarting when accessing volume %d in cell %s.";
break;
}
((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);
- LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume);
+ osi_Log3(afsd_logp, format, osi_LogSaveString(afsd_logp,addr), fidp->volume, cellp->name);
+ LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume, cellp->name);
}
lock_ObtainWrite(&cm_serverLock);
tsrp->status = srv_busy;
if (fidp) { /* File Server query */
lock_ReleaseWrite(&cm_serverLock);
- code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
- CM_GETVOL_FLAG_NO_LRU_UPDATE,
+ code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
+ CM_GETVOL_FLAG_NO_LRU_UPDATE,
&volp);
if (code == 0)
statep = cm_VolumeStateByID(volp, fidp->volume);
}
}
lock_ReleaseWrite(&cm_serverLock);
-
+
if (statep) {
cm_UpdateVolumeStatus(volp, statep->ID);
lock_ObtainRead(&cm_volumeLock);
if (free_svr_list) {
cm_FreeServerList(serverspp, 0);
- serverspp = NULL;
serversp = NULL;
free_svr_list = 0;
}
/* special codes: missing volumes */
else if (errorCode == VNOVOL || errorCode == VMOVED || errorCode == VOFFLINE ||
- errorCode == VSALVAGE || errorCode == VNOSERVICE || errorCode == VIO)
- {
+ errorCode == VSALVAGE || errorCode == VNOSERVICE || errorCode == VIO)
+ {
/* In case of timeout */
reqp->volumeError = errorCode;
switch ( errorCode ) {
case VNOVOL:
msgID = MSG_SERVER_REPORTS_VNOVOL;
- format = "Server %s reported volume %d as not attached (does not exist).";
+ format = "Server %s reported volume %d in cell %s as not attached (may have been moved or deleted).";
break;
case VMOVED:
msgID = MSG_SERVER_REPORTS_VMOVED;
- format = "Server %s reported volume %d as moved.";
+ format = "Server %s reported volume %d in cell %s as moved.";
break;
case VOFFLINE:
msgID = MSG_SERVER_REPORTS_VOFFLINE;
- format = "Server %s reported volume %d as offline.";
+ format = "Server %s reported volume %d in cell %s as offline.";
break;
case VSALVAGE:
msgID = MSG_SERVER_REPORTS_VSALVAGE;
- format = "Server %s reported volume %d as needs salvage.";
+ format = "Server %s reported volume %d in cell %s as needs salvage.";
break;
case VNOSERVICE:
msgID = MSG_SERVER_REPORTS_VNOSERVICE;
- format = "Server %s reported volume %d as not in service.";
+ format = "Server %s reported volume %d in cell %s as not in service.";
break;
case VIO:
msgID = MSG_SERVER_REPORTS_VIO;
- format = "Server %s reported volume %d as temporarily unaccessible.";
+ format = "Server %s reported volume %d in cell %s as temporarily unaccessible.";
break;
}
- if (serverp && fidp) {
- /* 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));
+ if (fidp) { /* File Server query */
+ if (serverp) {
+ /* Log server being unavailable 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_Log3(afsd_logp, format, osi_LogSaveString(afsd_logp,addr), fidp->volume, cellp->name);
+ LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume, cellp->name);
+ }
- osi_Log2(afsd_logp, format, osi_LogSaveString(afsd_logp,addr), fidp->volume);
- LogEvent(EVENTLOG_WARNING_TYPE, msgID, addr, fidp->volume);
- }
+ code = cm_FindVolumeByID(cellp, fidp->volume, userp, reqp,
+ CM_GETVOL_FLAG_NO_LRU_UPDATE,
+ &volp);
+ if (code == 0)
+ statep = cm_VolumeStateByID(volp, fidp->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) {
- serversp = *serverspp;
- free_svr_list = 1;
+ if ((errorCode == VMOVED || errorCode == VNOVOL || errorCode == VOFFLINE) &&
+ !(reqp->flags & CM_REQ_VOLUME_UPDATED))
+ {
+ code = cm_ForceUpdateVolume(fidp, userp, reqp);
+ if (code == 0)
+ location_updated = 1;
+
+ /* Even if the update fails, there might still be another replica */
+
+ 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 (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;
+ }
+
+ /*
+ * Mark server offline for this volume or delete the volume
+ * from the server list if it was moved or is not present.
+ */
+ if (!serversp || location_updated) {
+ code = cm_GetServerList(fidp, userp, reqp, &serverspp);
+ if (code == 0) {
+ serversp = *serverspp;
+ free_svr_list = 1;
+ }
}
}
((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));
+ ((tsrp->server->addr.sin_addr.s_addr & 0xff000000)>> 24));
- if (tsrp->server == serverp) {
+ if (cm_ServerEqual(tsrp->server, serverp)) {
/* REDIRECT */
if (errorCode == VMOVED || errorCode == VNOVOL) {
- osi_Log2(afsd_logp, "volume %d not present on server %s",
+ 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",
+ osi_Log2(afsd_logp, "volume %d instance on server %s marked offline",
fidp->volume, osi_LogSaveString(afsd_logp,addr));
tsrp->status = srv_offline;
}
/* break; */
} else {
- osi_Log3(afsd_logp, "volume %d exists on server %s with status %u",
+ 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);
/* 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 (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 (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 ) {
pscp = cm_FindSCacheParent(scp);
lock_ObtainWrite(&scp->rw);
+ scp->flags |= CM_SCACHEFLAG_DELETED;
lock_ObtainWrite(&cm_scacheLock);
- cm_RemoveSCacheFromHashTable(scp);
+ cm_AdjustScacheLRU(scp);
+ cm_RemoveSCacheFromHashTable(scp);
lock_ReleaseWrite(&cm_scacheLock);
cm_LockMarkSCacheLost(scp);
- scp->flags |= CM_SCACHEFLAG_DELETED;
lock_ReleaseWrite(&scp->rw);
cm_ReleaseSCache(scp);
* 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
+ * In addition, we log an event to the event log
*/
if (serverp) {
- sprintf(addr, "%d.%d.%d.%d",
+ 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));
+ ((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);
- }
+ 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 == RX_MSGSIZE) {
+ /*
+ * RPC failed because a transmitted rx packet
+ * may have grown larger than the path mtu.
+ * Force a retry and the Rx library will try
+ * with a smaller mtu size.
+ */
+
+ 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));
+
+ LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_MSGSIZE_EXCEEDED, addr);
+ osi_Log1(afsd_logp, "cm_Analyze: Path MTU may have been exceeded addr[%s]",
+ osi_LogSaveString(afsd_logp,addr));
+
+ retry = 1;
+ }
+ else if (errorCode == CM_RX_RETRY_BUSY_CALL) {
+ /*
+ * RPC failed because the selected call channel
+ * is currently busy on the server. Unconditionally
+ * retry the request so an alternate call channel can be used.
+ */
+ 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));
+
+ LogEvent(EVENTLOG_WARNING_TYPE, MSG_RX_BUSY_CALL_CHANNEL, addr);
+ osi_Log1(afsd_logp, "cm_Analyze: Retry RPC due to busy call channel addr[%s]",
+ osi_LogSaveString(afsd_logp,addr));
+ retry = 1;
+ }
else if (errorCode >= -64 && errorCode < 0) {
/* mark server as down */
if (serverp)
if (errorCode == RX_CALL_DEAD)
osi_Log2(afsd_logp, "cm_Analyze: Rx Call Dead addr[%s] forcedNew[%s]",
- osi_LogSaveString(afsd_logp,addr),
+ 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),
+ osi_LogSaveString(afsd_logp,addr),
(reqp->flags & CM_REQ_NEW_CONN_FORCED ? "yes" : "no"));
if (serverp) {
if (errorCode == RX_CALL_DEAD &&
(reqp->flags & CM_REQ_NEW_CONN_FORCED)) {
if (!(serverp->flags & CM_SERVERFLAG_DOWN)) {
- serverp->flags |= CM_SERVERFLAG_DOWN;
+ _InterlockedOr(&serverp->flags, CM_SERVERFLAG_DOWN);
serverp->downTime = time(NULL);
}
} else {
free(ucellp->ticketp);
ucellp->ticketp = NULL;
}
- ucellp->flags &= ~CM_UCELLFLAG_RXKAD;
+ _InterlockedAnd(&ucellp->flags, ~CM_UCELLFLAG_RXKAD);
ucellp->gen++;
lock_ReleaseMutex(&userp->mx);
if ( timeLeft > 2 )
case VL_INDEXERANGE : s = "VL_INDEXERANGE"; break;
case VL_MULTIPADDR : s = "VL_MULTIPADDR"; break;
case VL_BADMASK : s = "VL_BADMASK"; break;
- case CM_ERROR_NOSUCHCELL : s = "CM_ERROR_NOSUCHCELL"; break;
- case CM_ERROR_NOSUCHVOLUME : s = "CM_ERROR_NOSUCHVOLUME"; break;
- case CM_ERROR_TIMEDOUT : s = "CM_ERROR_TIMEDOUT"; break;
- case CM_ERROR_RETRY : s = "CM_ERROR_RETRY"; break;
- case CM_ERROR_NOACCESS : s = "CM_ERROR_NOACCESS"; break;
- case CM_ERROR_NOSUCHFILE : s = "CM_ERROR_NOSUCHFILE"; break;
- case CM_ERROR_STOPNOW : s = "CM_ERROR_STOPNOW"; break;
- case CM_ERROR_TOOBIG : s = "CM_ERROR_TOOBIG"; break;
- case CM_ERROR_INVAL : s = "CM_ERROR_INVAL"; break;
- case CM_ERROR_BADFD : s = "CM_ERROR_BADFD"; break;
- case CM_ERROR_BADFDOP : s = "CM_ERROR_BADFDOP"; break;
- case CM_ERROR_EXISTS : s = "CM_ERROR_EXISTS"; break;
- case CM_ERROR_CROSSDEVLINK : s = "CM_ERROR_CROSSDEVLINK"; break;
- case CM_ERROR_BADOP : s = "CM_ERROR_BADOP"; break;
- case CM_ERROR_BADPASSWORD : s = "CM_ERROR_BADPASSWORD"; break;
- case CM_ERROR_NOTDIR : s = "CM_ERROR_NOTDIR"; break;
- case CM_ERROR_ISDIR : s = "CM_ERROR_ISDIR"; break;
- case CM_ERROR_READONLY : s = "CM_ERROR_READONLY"; break;
- case CM_ERROR_WOULDBLOCK : s = "CM_ERROR_WOULDBLOCK"; break;
- case CM_ERROR_QUOTA : s = "CM_ERROR_QUOTA"; break;
- case CM_ERROR_SPACE : s = "CM_ERROR_SPACE"; break;
- case CM_ERROR_BADSHARENAME : s = "CM_ERROR_BADSHARENAME"; break;
- case CM_ERROR_BADTID : s = "CM_ERROR_BADTID"; break;
- case CM_ERROR_UNKNOWN : s = "CM_ERROR_UNKNOWN"; break;
- case CM_ERROR_NOMORETOKENS : s = "CM_ERROR_NOMORETOKENS"; break;
- case CM_ERROR_NOTEMPTY : s = "CM_ERROR_NOTEMPTY"; break;
- case CM_ERROR_USESTD : s = "CM_ERROR_USESTD"; break;
- case CM_ERROR_REMOTECONN : s = "CM_ERROR_REMOTECONN"; break;
- case CM_ERROR_ATSYS : s = "CM_ERROR_ATSYS"; break;
- case CM_ERROR_NOSUCHPATH : s = "CM_ERROR_NOSUCHPATH"; break;
- case CM_ERROR_CLOCKSKEW : s = "CM_ERROR_CLOCKSKEW"; break;
- case CM_ERROR_BADSMB : s = "CM_ERROR_BADSMB"; break;
- case CM_ERROR_ALLBUSY : s = "CM_ERROR_ALLBUSY"; break;
- case CM_ERROR_NOFILES : s = "CM_ERROR_NOFILES"; break;
- case CM_ERROR_PARTIALWRITE : s = "CM_ERROR_PARTIALWRITE"; break;
- case CM_ERROR_NOIPC : s = "CM_ERROR_NOIPC"; break;
- case CM_ERROR_BADNTFILENAME : s = "CM_ERROR_BADNTFILENAME"; break;
- case CM_ERROR_BUFFERTOOSMALL : s = "CM_ERROR_BUFFERTOOSMALL"; break;
- case CM_ERROR_RENAME_IDENTICAL : s = "CM_ERROR_RENAME_IDENTICAL"; break;
- case CM_ERROR_ALLOFFLINE : s = "CM_ERROR_ALLOFFLINE"; break;
- case CM_ERROR_AMBIGUOUS_FILENAME: s = "CM_ERROR_AMBIGUOUS_FILENAME"; break;
- case CM_ERROR_BADLOGONTYPE : s = "CM_ERROR_BADLOGONTYPE"; break;
- case CM_ERROR_GSSCONTINUE : s = "CM_ERROR_GSSCONTINUE"; break;
- case CM_ERROR_TIDIPC : s = "CM_ERROR_TIDIPC"; break;
- case CM_ERROR_TOO_MANY_SYMLINKS : s = "CM_ERROR_TOO_MANY_SYMLINKS"; break;
- case CM_ERROR_PATH_NOT_COVERED : s = "CM_ERROR_PATH_NOT_COVERED"; break;
- case CM_ERROR_LOCK_CONFLICT : s = "CM_ERROR_LOCK_CONFLICT"; break;
- case CM_ERROR_SHARING_VIOLATION : s = "CM_ERROR_SHARING_VIOLATION"; break;
- case CM_ERROR_ALLDOWN : s = "CM_ERROR_ALLDOWN"; break;
- case CM_ERROR_TOOFEWBUFS : s = "CM_ERROR_TOOFEWBUFS"; break;
- case CM_ERROR_TOOMANYBUFS : s = "CM_ERROR_TOOMANYBUFS"; break;
+ case CM_ERROR_NOSUCHCELL : s = "CM_ERROR_NOSUCHCELL"; break;
+ case CM_ERROR_NOSUCHVOLUME : s = "CM_ERROR_NOSUCHVOLUME"; break;
+ case CM_ERROR_TIMEDOUT : s = "CM_ERROR_TIMEDOUT"; break;
+ case CM_ERROR_RETRY : s = "CM_ERROR_RETRY"; break;
+ case CM_ERROR_NOACCESS : s = "CM_ERROR_NOACCESS"; break;
+ case CM_ERROR_NOSUCHFILE : s = "CM_ERROR_NOSUCHFILE"; break;
+ case CM_ERROR_STOPNOW : s = "CM_ERROR_STOPNOW"; break;
+ case CM_ERROR_TOOBIG : s = "CM_ERROR_TOOBIG"; break;
+ case CM_ERROR_INVAL : s = "CM_ERROR_INVAL"; break;
+ case CM_ERROR_BADFD : s = "CM_ERROR_BADFD"; break;
+ case CM_ERROR_BADFDOP : s = "CM_ERROR_BADFDOP"; break;
+ case CM_ERROR_EXISTS : s = "CM_ERROR_EXISTS"; break;
+ case CM_ERROR_CROSSDEVLINK : s = "CM_ERROR_CROSSDEVLINK"; break;
+ case CM_ERROR_BADOP : s = "CM_ERROR_BADOP"; break;
+ case CM_ERROR_BADPASSWORD : s = "CM_ERROR_BADPASSWORD"; break;
+ case CM_ERROR_NOTDIR : s = "CM_ERROR_NOTDIR"; break;
+ case CM_ERROR_ISDIR : s = "CM_ERROR_ISDIR"; break;
+ case CM_ERROR_READONLY : s = "CM_ERROR_READONLY"; break;
+ case CM_ERROR_WOULDBLOCK : s = "CM_ERROR_WOULDBLOCK"; break;
+ case CM_ERROR_QUOTA : s = "CM_ERROR_QUOTA"; break;
+ case CM_ERROR_SPACE : s = "CM_ERROR_SPACE"; break;
+ case CM_ERROR_BADSHARENAME : s = "CM_ERROR_BADSHARENAME"; break;
+ case CM_ERROR_BADTID : s = "CM_ERROR_BADTID"; break;
+ case CM_ERROR_UNKNOWN : s = "CM_ERROR_UNKNOWN"; break;
+ case CM_ERROR_NOMORETOKENS : s = "CM_ERROR_NOMORETOKENS"; break;
+ case CM_ERROR_NOTEMPTY : s = "CM_ERROR_NOTEMPTY"; break;
+ case CM_ERROR_USESTD : s = "CM_ERROR_USESTD"; break;
+ case CM_ERROR_REMOTECONN : s = "CM_ERROR_REMOTECONN"; break;
+ case CM_ERROR_ATSYS : s = "CM_ERROR_ATSYS"; break;
+ case CM_ERROR_NOSUCHPATH : s = "CM_ERROR_NOSUCHPATH"; break;
+ case CM_ERROR_CLOCKSKEW : s = "CM_ERROR_CLOCKSKEW"; break;
+ case CM_ERROR_BADSMB : s = "CM_ERROR_BADSMB"; break;
+ case CM_ERROR_ALLBUSY : s = "CM_ERROR_ALLBUSY"; break;
+ case CM_ERROR_NOFILES : s = "CM_ERROR_NOFILES"; break;
+ case CM_ERROR_PARTIALWRITE : s = "CM_ERROR_PARTIALWRITE"; break;
+ case CM_ERROR_NOIPC : s = "CM_ERROR_NOIPC"; break;
+ case CM_ERROR_BADNTFILENAME : s = "CM_ERROR_BADNTFILENAME"; break;
+ case CM_ERROR_BUFFERTOOSMALL : s = "CM_ERROR_BUFFERTOOSMALL"; break;
+ case CM_ERROR_RENAME_IDENTICAL : s = "CM_ERROR_RENAME_IDENTICAL"; break;
+ case CM_ERROR_ALLOFFLINE : s = "CM_ERROR_ALLOFFLINE"; break;
+ case CM_ERROR_AMBIGUOUS_FILENAME: s = "CM_ERROR_AMBIGUOUS_FILENAME"; break;
+ case CM_ERROR_BADLOGONTYPE : s = "CM_ERROR_BADLOGONTYPE"; break;
+ case CM_ERROR_GSSCONTINUE : s = "CM_ERROR_GSSCONTINUE"; break;
+ case CM_ERROR_TIDIPC : s = "CM_ERROR_TIDIPC"; break;
+ case CM_ERROR_TOO_MANY_SYMLINKS : s = "CM_ERROR_TOO_MANY_SYMLINKS"; break;
+ case CM_ERROR_PATH_NOT_COVERED : s = "CM_ERROR_PATH_NOT_COVERED"; break;
+ case CM_ERROR_LOCK_CONFLICT : s = "CM_ERROR_LOCK_CONFLICT"; break;
+ case CM_ERROR_SHARING_VIOLATION : s = "CM_ERROR_SHARING_VIOLATION"; break;
+ case CM_ERROR_ALLDOWN : s = "CM_ERROR_ALLDOWN"; break;
+ case CM_ERROR_TOOFEWBUFS : s = "CM_ERROR_TOOFEWBUFS"; break;
+ case CM_ERROR_TOOMANYBUFS : s = "CM_ERROR_TOOMANYBUFS"; break;
}
- osi_Log2(afsd_logp, "cm_Analyze: ignoring error code 0x%x (%s)",
+ osi_Log2(afsd_logp, "cm_Analyze: ignoring error code 0x%x (%s)",
errorCode, s);
retry = 0;
}
}
/* If not allowed to retry, don't */
- if (!forcing_new && (reqp->flags & CM_REQ_NORETRY))
+ if (!forcing_new && (reqp->flags & CM_REQ_NORETRY) &&
+ (errorCode != RX_MSGSIZE && errorCode != CM_RX_RETRY_BUSY_CALL))
retry = 0;
else if (retry && dead_session)
retry = 0;
- out:
/* drop this on the way out */
if (connp)
cm_PutConn(connp);
- /*
+ /*
* clear the volume updated flag if we succeed.
- * this way the flag will not prevent a subsequent volume
+ * this way the flag will not prevent a subsequent volume
* from being updated if necessary.
*/
if (errorCode == 0)
#ifdef SET_RX_TIMEOUTS_TO_TIMELEFT
timeUsed = (GetTickCount() - reqp->startTime) / 1000;
-
+
/* leave 5 seconds margin of safety */
timeLeft = ConnDeadtimeout - timeUsed - 5;
hardTimeLeft = HardDeadtimeout - timeUsed - 5;
tsp = tsrp->server;
if (reqp->tokenIdleErrorServp) {
- /*
+ /*
* search the list until we find the server
* that failed last time. When we find it
* clear the error, skip it and try the one
if (timeLeft > ConnDeadtimeout)
timeLeft = ConnDeadtimeout;
- if (hardTimeLeft > HardDeadtimeout)
+ if (hardTimeLeft > HardDeadtimeout)
hardTimeLeft = HardDeadtimeout;
lock_ObtainMutex(&(*connpp)->mx);
lock_ObtainRead(&cm_serverLock);
cm_PutServerNoLock(tsp);
}
- }
+ }
lock_ReleaseRead(&cm_serverLock);
if (firstError == 0) {
if (allDown) {
- firstError = (reqp->tokenError ? reqp->tokenError :
+ firstError = (reqp->tokenError ? reqp->tokenError :
(reqp->idleError ? RX_CALL_TIMEOUT : CM_ERROR_ALLDOWN));
/*
* if we experienced either a token error or and idle dead time error
cm_PutServer(tcp->serverp);
cm_ReleaseUser(userp);
*lcpp = tcp->nextp;
+ rx_SetConnSecondsUntilNatPing(tcp->rxconnp, 0);
rx_DestroyConnection(tcp->rxconnp);
lock_FinalizeMutex(&tcp->mx);
free(tcp);
}
secObjp = rxkad_NewClientSecurityObject(tcp->cryptlevel,
&ucellp->sessionKey, ucellp->kvno,
- ucellp->ticketLen, ucellp->ticketp);
+ ucellp->ticketLen, ucellp->ticketp);
} else {
/* normal auth */
secIndex = 0;
rx_SetConnDeadTime(tcp->rxconnp, ConnDeadtimeout);
rx_SetConnHardDeadTime(tcp->rxconnp, HardDeadtimeout);
rx_SetConnIdleDeadTime(tcp->rxconnp, IdleDeadtimeout);
- if (NatPingInterval)
+
+ /*
+ * Register the error to be returned on an idle dead timeout
+ */
+ rx_SetServerConnIdleDeadErr(tcp->rxconnp, RX_CALL_DEAD);
+
+ /*
+ * Let the Rx library know that we can auto-retry if an
+ * RX_MSGSIZE error is returned.
+ */
+ if (rx_pmtu_discovery)
+ rx_SetMsgsizeRetryErr(tcp->rxconnp, RX_MSGSIZE);
+
+ /*
+ * Attempt to limit NAT pings to the anonymous file server connections.
+ * Only file servers implement client callbacks and we only need one ping
+ * to be sent to each server.
+ */
+ if (NatPingInterval && serverp->type == CM_SERVER_FILE && secIndex == 0)
rx_SetConnSecondsUntilNatPing(tcp->rxconnp, NatPingInterval);
+
tcp->ucgen = ucellp->gen;
if (secObjp)
rxs_Release(secObjp); /* Decrement the initial refCount */
lock_ObtainMutex(&userp->mx);
lock_ObtainRead(&cm_connLock);
for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
- if (tcp->userp == userp)
+ if (tcp->userp == userp)
break;
}
-
+
/* find ucell structure */
ucellp = cm_GetUCell(userp, serverp->cellp);
if (!tcp) {
lock_ConvertRToW(&cm_connLock);
for (tcp = serverp->connsp; tcp; tcp=tcp->nextp) {
- if (tcp->userp == userp)
+ if (tcp->userp == userp)
break;
}
if (tcp) {
else
osi_Log0(afsd_logp, "cm_ConnByServer replace connection due to crypt change");
tcp->flags &= ~CM_CONN_FLAG_FORCE_NEW;
+ rx_SetConnSecondsUntilNatPing(tcp->rxconnp, 0);
rx_DestroyConnection(tcp->rxconnp);
cm_NewRXConnection(tcp, ucellp, serverp);
}
allBusy = 0;
}
}
- }
+ }
lock_ReleaseRead(&cm_serverLock);
cm_FreeServerList(serverspp, 0);
if (allDown)
return 0;
- else if (allBusy)
+ else if (allBusy)
return 0;
else if (allOffline || (someBusy && someOffline))
return 0;
return 1;
}
-/*
+/*
* The returned cm_conn_t ** object is released in the subsequent call
- * to cm_Analyze().
+ * to cm_Analyze().
*/
long cm_ConnFromFID(struct cm_fid *fidp, struct cm_user *userp, cm_req_t *reqp,
cm_conn_t **connpp)
git://git.openafs.org / openafs.git / blobdiff