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>
13 #include <afs/procmgmt.h> /* signal(), kill(), wait(), etc. */
17 #include <sys/statfs.h>
21 #include <afs/voldefs.h>
24 #include <rx/rx_queue.h>
25 #include <afs/vlserver.h>
27 #include <afs/cellconfig.h>
30 #include <afs/afsint.h>
34 #include <afs/com_err.h>
36 #include <afs/kautils.h>
38 #include <afs/ihandle.h>
40 #include <afs/ntops.h>
42 #include <afs/vnode.h>
43 #include <afs/volume.h>
44 #define ERRCODE_RANGE 8 /* from error_table.h */
45 #define CLOCKSKEW 2 /* not really skew, but resolution */
46 #define CLOCKADJ(x) (((x) < CLOCKSKEW) ? 0 : (x) - CLOCKSKEW)
48 /* for UV_MoveVolume() recovery */
52 #include "volser_internal.h"
53 #include "volser_prototypes.h"
54 #include "vsutils_prototypes.h"
55 #include "lockprocs_prototypes.h"
57 extern struct ubik_client *cstruct;
58 int verbose = 0, noresolve = 0;
63 afs_int32 vldbEntryIndex;
66 /* Utility macros used by rest of this source file */
67 #define EPRINT(ec, es) \
69 fprintf(STDERR, "\n"); \
70 fprintf(STDERR, (es)); \
74 #define EPRINT1(ec, es, ep1) \
76 fprintf(STDERR, "\n"); \
77 fprintf(STDERR, (es), (ep1)); \
81 #define EPRINT2(ec, es, ep1, ep2) \
83 fprintf(STDERR, "\n"); \
84 fprintf(STDERR, (es), (ep1), (ep2)); \
88 #define EPRINT3(ec, es, ep1, ep2, ep3) \
90 fprintf(STDERR, "\n"); \
91 fprintf(STDERR, (es), (ep1), (ep2), (ep3)); \
95 #define EGOTO(where, ec, es) \
104 #define EGOTO1(where, ec, es, ep1) \
107 EPRINT1((ec),(es),(ep1)); \
113 #define EGOTO2(where, ec, es, ep1, ep2) \
116 EPRINT2((ec),(es),(ep1),(ep2)); \
122 #define EGOTO3(where, ec, es, ep1, ep2, ep3) \
125 EPRINT3((ec),(es),(ep1),(ep2),(ep3)); \
132 { if (verbose) { fprintf(STDOUT, (es)); fflush(STDOUT); } }
133 #define VPRINT1(es, p) \
134 { if (verbose) { fprintf(STDOUT, (es), (p)); fflush(STDOUT); } }
135 #define VPRINT2(es, p1, p2) \
136 { if (verbose) { fprintf(STDOUT, (es), (p1), (p2)); fflush(STDOUT); } }
137 #define VPRINT3(es, p1, p2, p3) \
138 { if (verbose) { fprintf(STDOUT, (es), (p1), (p2), (p3)); fflush(STDOUT); } }
140 { if (verbose) { fprintf(STDOUT, " done\n"); fflush(STDOUT); } }
141 #define VEPRINT(es) \
142 { if (verbose) { fprintf(STDERR, (es)); fflush(STDERR); } }
143 #define VEPRINT1(es, p) \
144 { if (verbose) { fprintf(STDERR, (es), (p)); fflush(STDERR); } }
145 #define VEPRINT2(es, p1, p2) \
146 { if (verbose) { fprintf(STDERR, (es), (p1), (p2)); fflush(STDERR); } }
147 #define VEPRINT3(es, p1, p2, p3) \
148 { if (verbose) { fprintf(STDERR, (es), (p1), (p2), (p3)); fflush(STDERR); } }
150 { if (verbose) { fprintf(STDERR, " done\n"); fflush(STDERR); } }
154 /* getting rid of this */
155 #define ERROR_EXIT(code) do { \
161 /* Protos for static routines */
163 static afs_int32 CheckAndDeleteVolume(struct rx_connection *aconn,
164 afs_int32 apart, afs_uint32 okvol,
167 static int GetTrans(struct nvldbentry *vldbEntryPtr, afs_int32 index,
168 struct rx_connection **connPtr, afs_int32 * transPtr,
169 afs_uint32 * crtimePtr, afs_uint32 * uptimePtr,
170 afs_int32 *origflags, afs_uint32 tmpVolId);
171 static int SimulateForwardMultiple(struct rx_connection *fromconn,
172 afs_int32 fromtid, afs_int32 fromdate,
173 manyDests * tr, afs_int32 flags,
174 void *cookie, manyResults * results);
175 static int DoVolOnline(struct nvldbentry *vldbEntryPtr, afs_uint32 avolid,
176 int index, char *vname, struct rx_connection *connPtr);
177 static int DoVolClone(struct rx_connection *aconn, afs_uint32 avolid,
178 afs_int32 apart, int type, afs_uint32 cloneid,
179 char *typestring, char *pname, char *vname, char *suffix,
180 struct volser_status *volstatus, afs_int32 *transPtr);
181 static int DoVolDelete(struct rx_connection *aconn, afs_uint32 avolid,
182 afs_int32 apart, char *typestring, afs_uint32 atoserver,
183 struct volser_status *volstatus, char *pprefix);
184 static afs_int32 CheckVolume(volintInfo * volumeinfo, afs_uint32 aserver,
185 afs_int32 apart, afs_int32 * modentry,
186 afs_uint32 * maxvolid, struct nvldbentry *aentry);
187 static afs_int32 VolumeExists(afs_uint32 server, afs_int32 partition,
188 afs_uint32 volumeid);
189 static afs_int32 CheckVldbRWBK(struct nvldbentry * entry,
190 afs_int32 * modified);
191 static afs_int32 CheckVldbRO(struct nvldbentry *entry, afs_int32 * modified);
192 static afs_int32 CheckVldb(struct nvldbentry *entry, afs_int32 * modified,
194 static void dump_sig_handler(int x);
195 static int sortVolumes(const void *a, const void *b);
198 /*map the partition <partId> into partition name <partName>*/
200 MapPartIdIntoName(afs_int32 partId, char *partName)
202 if (partId < 26) { /* what if partId > = 26 ? */
203 strcpy(partName, "/vicep");
204 partName[6] = partId + 'a';
207 } else if (partId < VOLMAXPARTS) {
208 strcpy(partName, "/vicep");
210 partName[6] = 'a' + (partId / 26);
211 partName[7] = 'a' + (partId % 26);
218 PrintError(char *msg, afs_int32 errcode)
220 fprintf(STDERR, "%s", msg);
221 /*replace by a big switch statement */
226 fprintf(STDERR, "Possible communication failure\n");
229 fprintf(STDERR, "Volume needs to be salvaged\n");
232 fprintf(STDERR, "Bad vnode number quoted\n");
236 "Volume not attached, does not exist, or not on line\n");
239 fprintf(STDERR, "Volume already exists\n");
242 fprintf(STDERR, "Volume is not in service\n");
245 fprintf(STDERR, "Volume is off line\n");
248 fprintf(STDERR, "Volume is already on line\n");
251 fprintf(STDERR, "Partition is full\n");
254 fprintf(STDERR, "Volume max quota exceeded\n");
257 fprintf(STDERR, "Volume temporarily unavailable\n");
260 fprintf(STDERR, "Volume has moved to another server\n");
263 fprintf(STDERR, "VLDB: volume Id exists in the vldb\n");
266 fprintf(STDERR, "VLDB: a read terminated too early\n");
269 fprintf(STDERR, "VLDB: volume entry exists in the vldb\n");
272 fprintf(STDERR, "VLDB: internal creation failure\n");
275 fprintf(STDERR, "VLDB: no such entry\n");
278 fprintf(STDERR, "VLDB: vldb database is empty\n");
281 fprintf(STDERR, "VLDB: entry is deleted (soft delete)\n");
284 fprintf(STDERR, "VLDB: volume name is illegal\n");
287 fprintf(STDERR, "VLDB: index was out of range\n");
290 fprintf(STDERR, "VLDB: bad volume type\n");
293 fprintf(STDERR, "VLDB: illegal server number (not within limits)\n");
295 case VL_BADPARTITION:
296 fprintf(STDERR, "VLDB: bad partition number\n");
299 fprintf(STDERR, "VLDB: run out of space for replication sites\n");
302 fprintf(STDERR, "VLDB: no such repsite server exists\n");
304 case VL_DUPREPSERVER:
305 fprintf(STDERR, "VLDB: replication site server already exists\n");
308 fprintf(STDERR, "VLDB: parent r/w entry not found\n");
311 fprintf(STDERR, "VLDB: illegal reference count number\n");
313 case VL_SIZEEXCEEDED:
314 fprintf(STDERR, "VLDB: vldb size for attributes exceeded\n");
317 fprintf(STDERR, "VLDB: bad incoming vldb entry\n");
319 case VL_BADVOLIDBUMP:
320 fprintf(STDERR, "VLDB: illegal max volid increment\n");
322 case VL_IDALREADYHASHED:
323 fprintf(STDERR, "VLDB: (RO/BACK) Id already hashed\n");
326 fprintf(STDERR, "VLDB: vldb entry is already locked\n");
329 fprintf(STDERR, "VLDB: bad volume operation code\n");
331 case VL_BADRELLOCKTYPE:
332 fprintf(STDERR, "VLDB: bad release lock type\n");
335 fprintf(STDERR, "VLDB: status report: last release was aborted\n");
337 case VL_BADSERVERFLAG:
338 fprintf(STDERR, "VLDB: invalid replication site server flag\n");
341 fprintf(STDERR, "VLDB: no permission access for call\n");
343 case VOLSERREAD_DUMPERROR:
345 "VOLSER: Problems encountered in reading the dump file !\n");
347 case VOLSERDUMPERROR:
348 fprintf(STDERR, "VOLSER: Problems encountered in doing the dump !\n");
350 case VOLSERATTACH_ERROR:
351 fprintf(STDERR, "VOLSER: Could not attach the volume\n");
353 case VOLSERDETACH_ERROR:
354 fprintf(STDERR, "VOLSER: Could not detach the volume\n");
356 case VOLSERILLEGAL_PARTITION:
357 fprintf(STDERR, "VOLSER: encountered illegal partition number\n");
359 case VOLSERBAD_ACCESS:
360 fprintf(STDERR, "VOLSER: permission denied, not a super user\n");
362 case VOLSERVLDB_ERROR:
363 fprintf(STDERR, "VOLSER: error detected in the VLDB\n");
366 fprintf(STDERR, "VOLSER: error in volume name\n");
369 fprintf(STDERR, "VOLSER: volume has moved\n");
372 fprintf(STDERR, "VOLSER: illegal operation\n");
374 case VOLSERBADRELEASE:
375 fprintf(STDERR, "VOLSER: release could not be completed\n");
378 fprintf(STDERR, "VOLSER: volume is busy\n");
380 case VOLSERNO_MEMORY:
381 fprintf(STDERR, "VOLSER: volume server is out of memory\n");
385 "VOLSER: no such volume - location specified incorrectly or volume does not exist\n");
387 case VOLSERMULTIRWVOL:
389 "VOLSER: multiple RW volumes with same ID, one of which should be deleted\n");
393 "VOLSER: not all entries were successfully processed\n");
397 initialize_RXK_error_table();
398 initialize_KTC_error_table();
399 initialize_ACFG_error_table();
400 initialize_VL_error_table();
402 fprintf(STDERR, "%s: %s\n", afs_error_table_name(errcode),
403 afs_error_message(errcode));
410 void init_volintInfo(struct volintInfo *vinfo) {
411 memset(vinfo, 0, sizeof(struct volintInfo));
413 vinfo->maxquota = -1;
415 vinfo->creationDate = -1;
416 vinfo->updateDate = -1;
424 static struct rx_securityClass *uvclass = 0;
425 static int uvindex = -1;
426 /* called by VLDBClient_Init to set the security module to be used in the RPC */
428 UV_SetSecurity(struct rx_securityClass *as, afs_int32 aindex)
435 /* bind to volser on <port> <aserver> */
436 /* takes server address in network order, port in host order. dumb */
437 struct rx_connection *
438 UV_Bind(afs_uint32 aserver, afs_int32 port)
440 struct rx_connection *tc;
442 tc = rx_NewConnection(aserver, htons(port), VOLSERVICE_ID, uvclass,
448 AFSVolCreateVolume_retry(struct rx_connection *z_conn,
449 afs_int32 partition, char *name, afs_int32 type,
450 afs_int32 parent, afs_uint32 *volid, afs_int32 *trans)
455 code = AFSVolCreateVolume(z_conn, partition, name, type, parent,
457 if (code != VOLSERVOLBUSY)
460 #ifdef AFS_PTHREAD_ENV
463 IOMGR_Sleep(3-retries);
470 AFSVolTransCreate_retry(struct rx_connection *z_conn,
471 afs_int32 volume, afs_int32 partition,
472 afs_int32 flags, afs_int32 * trans)
477 code = AFSVolTransCreate(z_conn, volume, partition, flags, trans);
478 if (code != VOLSERVOLBUSY)
481 #ifdef AFS_PTHREAD_ENV
484 IOMGR_Sleep(3-retries);
491 /* if <okvol> is allright(indicated by beibg able to
492 * start a transaction, delete the <delvol> */
494 CheckAndDeleteVolume(struct rx_connection *aconn, afs_int32 apart,
495 afs_uint32 okvol, afs_uint32 delvol)
497 afs_int32 error, code, tid, rcode;
502 code = AFSVolTransCreate_retry(aconn, delvol, apart, ITOffline, &tid);
505 code = AFSVolDeleteVolume(aconn, tid);
508 code = AFSVolEndTrans(aconn, tid, &rcode);
515 code = AFSVolTransCreate_retry(aconn, okvol, apart, ITOffline, &tid);
517 code = AFSVolEndTrans(aconn, tid, &rcode);
522 code = AFSVolTransCreate_retry(aconn, delvol, apart, ITOffline, &tid);
525 code = AFSVolDeleteVolume(aconn, tid);
528 code = AFSVolEndTrans(aconn, tid, &rcode);
541 /* called by EmuerateEntry, show vldb entry in a reasonable format */
543 SubEnumerateEntry(struct nvldbentry *entry)
551 fprintf(STDOUT, " readWriteID %-10u ", entry->volumeId[RWVOL]);
552 if (entry->flags & RW_EXISTS)
553 fprintf(STDOUT, " valid \n");
555 fprintf(STDOUT, " invalid \n");
556 fprintf(STDOUT, " readOnlyID %-10u ", entry->volumeId[ROVOL]);
557 if (entry->flags & RO_EXISTS)
558 fprintf(STDOUT, " valid \n");
560 fprintf(STDOUT, " invalid \n");
561 fprintf(STDOUT, " backUpID %-10u ", entry->volumeId[BACKVOL]);
562 if (entry->flags & BACK_EXISTS)
563 fprintf(STDOUT, " valid \n");
565 fprintf(STDOUT, " invalid \n");
566 if ((entry->cloneId != 0) && (entry->flags & RO_EXISTS))
567 fprintf(STDOUT, " releaseClone %-10u \n", entry->cloneId);
569 if (entry->flags & RW_EXISTS)
570 fprintf(STDOUT, " RWrite: %-10u", entry->volumeId[RWVOL]);
571 if (entry->flags & RO_EXISTS)
572 fprintf(STDOUT, " ROnly: %-10u", entry->volumeId[ROVOL]);
573 if (entry->flags & BACK_EXISTS)
574 fprintf(STDOUT, " Backup: %-10u", entry->volumeId[BACKVOL]);
575 if ((entry->cloneId != 0) && (entry->flags & RO_EXISTS))
576 fprintf(STDOUT, " RClone: %-10lu", (unsigned long)entry->cloneId);
577 fprintf(STDOUT, "\n");
579 fprintf(STDOUT, " number of sites -> %lu\n",
580 (unsigned long)entry->nServers);
581 for (i = 0; i < entry->nServers; i++) {
582 if (entry->serverFlags[i] & NEW_REPSITE)
585 for (i = 0; i < entry->nServers; i++) {
586 MapPartIdIntoName(entry->serverPartition[i], pname);
587 fprintf(STDOUT, " server %s partition %s ",
588 noresolve ? afs_inet_ntoa_r(entry->serverNumber[i], hoststr) :
589 hostutil_GetNameByINet(entry->serverNumber[i]), pname);
590 if (entry->serverFlags[i] & ITSRWVOL)
591 fprintf(STDOUT, "RW Site ");
593 fprintf(STDOUT, "RO Site ");
595 if (entry->serverFlags[i] & NEW_REPSITE)
596 fprintf(STDOUT," -- New release");
598 if (!(entry->serverFlags[i] & ITSRWVOL))
599 fprintf(STDOUT," -- Old release");
601 if (entry->serverFlags[i] & RO_DONTUSE)
602 fprintf(STDOUT, " -- Not released");
604 fprintf(STDOUT, "\n");
611 /*enumerate the vldb entry corresponding to <entry> */
613 EnumerateEntry(struct nvldbentry *entry)
616 fprintf(STDOUT, "\n");
617 fprintf(STDOUT, "%s \n", entry->name);
618 SubEnumerateEntry(entry);
622 /* forcibly remove a volume. Very dangerous call */
624 UV_NukeVolume(afs_uint32 server, afs_int32 partid, afs_uint32 volid)
626 struct rx_connection *tconn;
629 tconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
631 code = AFSVolNukeVolume(tconn, partid, volid);
632 rx_DestroyConnection(tconn);
638 /* like df. Return usage of <pname> on <server> in <partition> */
640 UV_PartitionInfo64(afs_uint32 server, char *pname,
641 struct diskPartition64 *partition)
643 struct rx_connection *aconn;
646 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
647 code = AFSVolPartitionInfo64(aconn, pname, partition);
648 if (code == RXGEN_OPCODE) {
649 struct diskPartition *dpp = malloc(sizeof(struct diskPartition));
650 code = AFSVolPartitionInfo(aconn, pname, dpp);
652 strncpy(partition->name, dpp->name, 32);
653 strncpy(partition->devName, dpp->devName, 32);
654 partition->lock_fd = dpp->lock_fd;
655 partition->free = dpp->free;
656 partition->minFree = dpp->minFree;
661 fprintf(STDERR, "Could not get information on partition %s\n", pname);
662 PrintError("", code);
665 rx_DestroyConnection(aconn);
669 /* old interface to create volumes */
671 UV_CreateVolume(afs_uint32 aserver, afs_int32 apart, char *aname,
676 code = UV_CreateVolume2(aserver, apart, aname, 5000, 0, 0, 0, 0, anewid);
680 /* less old interface to create volumes */
682 UV_CreateVolume2(afs_uint32 aserver, afs_int32 apart, char *aname,
683 afs_int32 aquota, afs_int32 aspare1, afs_int32 aspare2,
684 afs_int32 aspare3, afs_int32 aspare4, afs_uint32 * anewid)
686 afs_uint32 roid = 0, bkid = 0;
687 return UV_CreateVolume3(aserver, apart, aname, aquota, aspare1, aspare2,
688 aspare3, aspare4, anewid, &roid, &bkid);
692 * Create a volume on the given server and partition
694 * @param aserver server to create volume on
695 * @param spart partition to create volume on
696 * @param aname name of new volume
697 * @param aquota quota for new volume
698 * @param anewid contains the desired volume id for the new volume. If
699 * *anewid == 0, a new id will be chosen, and will be placed
700 * in *anewid when UV_CreateVolume3 returns.
701 * @param aroid contains the desired RO volume id. If NULL, the RO id entry
702 * will be unset. If *aroid == 0, an id will be chosen, and
703 * will be placed in *anewid when UV_CreateVolume3 returns.
704 * @param abkid same as aroid, except for the BK volume id instead of the
706 * @return 0 on success, error code otherwise.
709 UV_CreateVolume3(afs_uint32 aserver, afs_int32 apart, char *aname,
710 afs_int32 aquota, afs_int32 aspare1, afs_int32 aspare2,
711 afs_int32 aspare3, afs_int32 aspare4, afs_uint32 * anewid,
712 afs_uint32 * aroid, afs_uint32 * abkid)
714 struct rx_connection *aconn;
718 afs_int32 rcode, vcode;
720 struct nvldbentry entry, storeEntry; /*the new vldb entry */
721 struct volintInfo tstatus;
726 init_volintInfo(&tstatus);
727 tstatus.maxquota = aquota;
729 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
731 if (aroid && *aroid) {
732 VPRINT1("Using RO volume ID %d.\n", *aroid);
734 if (abkid && *abkid) {
735 VPRINT1("Using BK volume ID %d.\n", *abkid);
739 vcode = VLDB_GetEntryByID(*anewid, -1, &entry);
741 fprintf(STDERR, "Volume ID %d already exists\n", *anewid);
744 VPRINT1("Using volume ID %d.\n", *anewid);
746 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, anewid);
747 EGOTO1(cfail, vcode, "Could not get an Id for volume %s\n", aname);
749 if (aroid && *aroid == 0) {
750 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, aroid);
751 EGOTO1(cfail, vcode, "Could not get an RO Id for volume %s\n", aname);
754 if (abkid && *abkid == 0) {
755 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, abkid);
756 EGOTO1(cfail, vcode, "Could not get a BK Id for volume %s\n", aname);
760 /* rw,ro, bk id are related in the default case */
761 /* If caller specified RW id, but not RO/BK ids, have them be RW+1 and RW+2 */
763 if (aroid && *aroid != 0) {
764 lastid = max(lastid, *aroid);
766 if (abkid && *abkid != 0) {
767 lastid = max(lastid, *abkid);
769 if (aroid && *aroid == 0) {
772 if (abkid && *abkid == 0) {
777 AFSVolCreateVolume_retry(aconn, apart, aname, volser_RW, 0, anewid, &tid);
778 EGOTO2(cfail, code, "Failed to create the volume %s %u \n", aname,
781 code = AFSVolSetInfo(aconn, tid, &tstatus);
783 EPRINT(code, "Could not change quota, continuing...\n");
785 code = AFSVolSetFlags(aconn, tid, 0); /* bring it online (mark it InService */
786 EGOTO2(cfail, code, "Could not bring the volume %s %u online \n", aname,
789 VPRINT2("Volume %s %u created and brought online\n", aname, *anewid);
791 /* set up the vldb entry for this volume */
792 strncpy(entry.name, aname, VOLSER_OLDMAXVOLNAME);
794 entry.serverNumber[0] = aserver; /* this should have another
795 * level of indirection later */
796 entry.serverPartition[0] = apart; /* this should also have
797 * another indirection level */
798 entry.flags = RW_EXISTS; /* this records that rw volume exists */
799 entry.serverFlags[0] = ITSRWVOL; /*this rep site has rw vol */
800 entry.volumeId[RWVOL] = *anewid;
801 entry.volumeId[ROVOL] = aroid ? *aroid : 0;
802 entry.volumeId[BACKVOL] = abkid ? *abkid : 0;
804 /*map into right byte order, before passing to xdr, the stuff has to be in host
805 * byte order. Xdr converts it into network order */
806 MapNetworkToHost(&entry, &storeEntry);
807 /* create the vldb entry */
808 vcode = VLDB_CreateEntry(&storeEntry);
811 "Could not create a VLDB entry for the volume %s %lu\n",
812 aname, (unsigned long)*anewid);
813 /*destroy the created volume */
814 VPRINT1("Deleting the newly created volume %u\n", *anewid);
815 AFSVolDeleteVolume(aconn, tid);
819 VPRINT2("Created the VLDB entry for the volume %s %u\n", aname, *anewid);
820 /* volume created, now terminate the transaction and release the connection */
821 code = AFSVolEndTrans(aconn, tid, &rcode); /*if it crashes before this
822 * the volume will come online anyway when transaction timesout , so if
823 * vldb entry exists then the volume is guaranteed to exist too wrt create */
827 "Failed to end the transaction on the volume %s %lu\n", aname,
828 (unsigned long)*anewid);
835 code = AFSVolEndTrans(aconn, tid, &rcode);
837 fprintf(STDERR, "WARNING: could not end transaction\n");
840 rx_DestroyConnection(aconn);
841 PrintError("", error);
845 /* create a volume, given a server, partition number, volume name --> sends
846 * back new vol id in <anewid>*/
848 UV_AddVLDBEntry(afs_uint32 aserver, afs_int32 apart, char *aname,
851 struct rx_connection *aconn;
854 struct nvldbentry entry, storeEntry; /*the new vldb entry */
856 aconn = (struct rx_connection *)0;
859 /* set up the vldb entry for this volume */
860 strncpy(entry.name, aname, VOLSER_OLDMAXVOLNAME);
862 entry.serverNumber[0] = aserver; /* this should have another
863 * level of indirection later */
864 entry.serverPartition[0] = apart; /* this should also have
865 * another indirection level */
866 entry.flags = RW_EXISTS; /* this records that rw volume exists */
867 entry.serverFlags[0] = ITSRWVOL; /*this rep site has rw vol */
868 entry.volumeId[RWVOL] = aid;
870 entry.volumeId[ROVOL] = anewid + 1; /* rw,ro, bk id are related in the default case */
871 entry.volumeId[BACKVOL] = *anewid + 2;
873 entry.volumeId[ROVOL] = 0;
874 entry.volumeId[BACKVOL] = 0;
877 /*map into right byte order, before passing to xdr, the stuff has to be in host
878 * byte order. Xdr converts it into network order */
879 MapNetworkToHost(&entry, &storeEntry);
880 /* create the vldb entry */
881 vcode = VLDB_CreateEntry(&storeEntry);
884 "Could not create a VLDB entry for the volume %s %lu\n",
885 aname, (unsigned long)aid);
889 VPRINT2("Created the VLDB entry for the volume %s %u\n", aname, aid);
893 rx_DestroyConnection(aconn);
894 PrintError("", error);
898 /* Delete the volume <volid>on <aserver> <apart>
899 * the physical entry gets removed from the vldb only if the ref count
903 UV_DeleteVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid)
905 struct rx_connection *aconn = (struct rx_connection *)0;
907 afs_int32 code, rcode;
909 struct nvldbentry entry, storeEntry;
911 afs_int32 avoltype = -1, vtype;
912 int notondisk = 0, notinvldb = 0;
914 /* Find and read bhe VLDB entry for this volume */
915 code = ubik_VL_SetLock(cstruct, 0, avolid, avoltype, VLOP_DELETE);
917 if (code != VL_NOENT) {
918 EGOTO1(error_exit, code,
919 "Could not lock VLDB entry for the volume %u\n", avolid);
925 code = VLDB_GetEntryByID(avolid, avoltype, &entry);
926 EGOTO1(error_exit, code, "Could not fetch VLDB entry for volume %u\n",
928 MapHostToNetwork(&entry);
931 EnumerateEntry(&entry);
934 /* Whether volume is in the VLDB or not. Delete the volume on disk */
935 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
937 code = DoVolDelete(aconn, avolid, apart, "the", 0, NULL, NULL);
947 /* Now update the VLDB entry.
948 * But first, verify we have a VLDB entry.
949 * Whether volume is on disk or not. Delete the volume in VLDB.
954 if (avolid == entry.volumeId[BACKVOL]) {
955 /* Its a backup volume, modify the VLDB entry. Check that the
956 * backup volume is on the server/partition we asked to delete.
958 if (!(entry.flags & BACK_EXISTS) || !Lp_Match(aserver, apart, &entry)) {
959 notinvldb = 2; /* Not on this server and partition */
963 VPRINT1("Marking the backup volume %u deleted in the VLDB\n", avolid);
965 entry.flags &= ~BACK_EXISTS;
969 else if (avolid == entry.volumeId[ROVOL]) {
970 /* Its a read-only volume, modify the VLDB entry. Check that the
971 * readonly volume is on the server/partition we asked to delete.
972 * If flags does not have RO_EIXSTS set, then this may mean the RO
973 * hasn't been released (and could exist in VLDB).
975 if (!Lp_ROMatch(aserver, apart, &entry)) {
976 notinvldb = 2; /* Not found on this server and partition */
982 "Marking the readonly volume %lu deleted in the VLDB\n",
983 (unsigned long)avolid);
985 Lp_SetROValue(&entry, aserver, apart, 0, 0); /* delete the site */
987 if (!Lp_ROMatch(0, 0, &entry))
988 entry.flags &= ~RO_EXISTS; /* This was the last ro volume */
992 else if (avolid == entry.volumeId[RWVOL]) {
993 /* It's a rw volume, delete the backup volume, modify the VLDB entry.
994 * Check that the readwrite volumes is on the server/partition we
997 if (!(entry.flags & RW_EXISTS) || !Lp_Match(aserver, apart, &entry)) {
998 notinvldb = 2; /* Not found on this server and partition */
1002 if (entry.volumeId[BACKVOL]) {
1003 /* Delete backup if it exists */
1004 code = DoVolDelete(aconn, entry.volumeId[BACKVOL], apart,
1005 "the backup", 0, NULL, NULL);
1006 if (code && code != VNOVOL) {
1014 "Marking the readwrite volume %lu%s deleted in the VLDB\n",
1015 (unsigned long)avolid,
1017 flags & BACK_EXISTS) ? ", and its backup volume," :
1020 Lp_SetRWValue(&entry, aserver, apart, 0L, 0L);
1022 entry.flags &= ~(BACK_EXISTS | RW_EXISTS);
1025 if (entry.flags & RO_EXISTS)
1026 fprintf(STDERR, "WARNING: ReadOnly copy(s) may still exist\n");
1030 notinvldb = 2; /* Not found on this server and partition */
1034 /* Either delete or replace the VLDB entry */
1035 if ((entry.nServers <= 0) || !(entry.flags & (RO_EXISTS | RW_EXISTS))) {
1038 "Last reference to the VLDB entry for %lu - deleting entry\n",
1039 (unsigned long)avolid);
1040 code = ubik_VL_DeleteEntry(cstruct, 0, avolid, vtype);
1041 EGOTO1(error_exit, code,
1042 "Could not delete the VLDB entry for the volume %u \n",
1045 MapNetworkToHost(&entry, &storeEntry);
1047 VLDB_ReplaceEntry(avolid, vtype, &storeEntry,
1048 (LOCKREL_OPCODE | LOCKREL_AFSID |
1049 LOCKREL_TIMESTAMP));
1050 EGOTO1(error_exit, code,
1051 "Could not update the VLDB entry for the volume %u \n",
1058 EPRINT(error, "\n");
1060 if (notondisk && notinvldb) {
1061 EPRINT2(VOLSERNOVOL, "Volume %u does not exist %s\n", avolid,
1062 ((notinvldb == 2) ? "on server and partition" : ""));
1064 error = VOLSERNOVOL;
1065 } else if (notondisk) {
1067 "WARNING: Volume %lu did not exist on the partition\n",
1068 (unsigned long)avolid);
1069 } else if (notinvldb) {
1070 fprintf(STDERR, "WARNING: Volume %lu does not exist in VLDB %s\n",
1071 (unsigned long)avolid,
1072 ((notinvldb == 2) ? "on server and partition" : ""));
1076 code = AFSVolEndTrans(aconn, ttid, &rcode);
1077 code = (code ? code : rcode);
1079 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
1080 (unsigned long)avolid);
1081 PrintError("", code);
1089 ubik_VL_ReleaseLock(cstruct, 0, avolid, -1,
1090 (LOCKREL_OPCODE | LOCKREL_AFSID |
1091 LOCKREL_TIMESTAMP));
1094 "Could not release the lock on the VLDB entry for the volume %u \n",
1102 rx_DestroyConnection(aconn);
1106 /* add recovery to UV_MoveVolume */
1108 #define TESTC 0 /* set to test recovery code, clear for production */
1114 do_interrupt(void * unused)
1117 #if !defined(AFS_PTHREAD_ENV) && !defined(AFS_NT40_ENV)
1118 /* Avoid UNIX LWP from getting confused that our stack has suddenly
1119 * changed. This will avoid some sanity checks, but until a better way
1120 * is found, the only alternative is always crashing and burning on at
1121 * least the stack-overflow check. */
1122 lwp_cpptr->stack = NULL;
1127 fprintf(STDOUT, "\nSIGINT handler: vos move operation in progress\n");
1129 "WARNING: may leave AFS storage and metadata in indeterminate state\n");
1130 fprintf(STDOUT, "enter second control-c to exit\n");
1138 sigint_handler(int x)
1140 #ifdef AFS_PTHREAD_ENV
1143 IOMGR_SoftSig(do_interrupt, 0);
1145 (void)signal(SIGINT, sigint_handler);
1149 DoVolDelete(struct rx_connection *aconn, afs_uint32 avolid,
1150 afs_int32 apart, char *ptypestring, afs_uint32 atoserver,
1151 struct volser_status *volstatus, char *pprefix)
1153 afs_int32 ttid = 0, code, rcode, error = 0;
1154 char *prefix, *typestring;
1163 typestring = ptypestring;
1169 VPRINT3("%sDeleting %s volume %u ...", prefix, typestring, avolid);
1172 AFSVolTransCreate_retry(aconn, avolid, apart, ITOffline, &ttid);
1174 EGOTO2(dfail, code, "%sFailed to start transaction on %u\n",
1178 code = AFSVolGetStatus(aconn, ttid, volstatus);
1179 EGOTO2(dfail, code, "%sCould not get timestamp from volume %u\n",
1184 AFSVolSetFlags(aconn, ttid,
1185 VTDeleteOnSalvage | VTOutOfService);
1187 EGOTO2(dfail, code, "%sCould not set flags on volume %u \n",
1191 VPRINT1("%sSetting volume forwarding pointer ...", prefix);
1192 AFSVolSetForwarding(aconn, ttid, atoserver);
1196 code = AFSVolDeleteVolume(aconn, ttid);
1197 EGOTO2(dfail, code, "%sCould not delete volume %u\n", prefix, avolid);
1201 code = AFSVolEndTrans(aconn, ttid, &rcode);
1206 fprintf(STDERR, "%sCould not end transaction on %s volume %lu \n",
1207 prefix, typestring, (unsigned long)avolid);
1213 if (beverbose && !error)
1219 DoVolClone(struct rx_connection *aconn, afs_uint32 avolid,
1220 afs_int32 apart, int type, afs_uint32 cloneid,
1221 char *typestring, char *pname, char *vname, char *suffix,
1222 struct volser_status *volstatus, afs_int32 *transPtr)
1225 afs_int32 ttid = 0, btid = 0;
1226 afs_int32 code = 0, rcode = 0;
1227 afs_int32 error = 0;
1228 int cloneexists = 1;
1230 /* Test to see if the clone volume exists by trying to create
1231 * a transaction on the clone volume. We've assumed the clone exists.
1233 code = AFSVolTransCreate_retry(aconn, cloneid, apart, ITOffline, &btid);
1235 if (code != VNOVOL) {
1236 EPRINT2(code, "Could not reach the %s volume %lu\n",
1237 typestring, (unsigned long)cloneid);
1241 cloneexists = 0; /* clone volume does not exist */
1244 code = AFSVolEndTrans(aconn, btid, &rcode);
1246 if (code || rcode) {
1248 "Could not end transaction on the previous %s volume %lu\n",
1249 typestring, (unsigned long)cloneid);
1250 error = (code ? code : rcode);
1255 /* Now go ahead and try to clone the RW volume.
1256 * First start a transaction on the RW volume
1258 code = AFSVolTransCreate_retry(aconn, avolid, apart, ITBusy, &ttid);
1260 fprintf(STDERR, "Could not start a transaction on the volume %lu\n",
1261 (unsigned long)avolid);
1266 /* Clone or reclone the volume, depending on whether the clone
1267 * volume exists or not
1270 VPRINT2("Re-cloning %s volume %u ...", typestring, cloneid);
1272 code = AFSVolReClone(aconn, ttid, cloneid);
1274 EPRINT2(code, "Could not re-clone %s volume %lu\n",
1275 typestring, (unsigned long)cloneid);
1280 VPRINT2("Creating a new %s clone %u ...", typestring, cloneid);
1283 strcpy(cname, pname);
1284 strcat(cname, suffix);
1287 code = AFSVolClone(aconn, ttid, 0, type, vname?vname:cname,
1290 fprintf(STDERR, "Failed to clone the volume %lu\n",
1291 (unsigned long)avolid);
1300 VPRINT1("Getting status of parent volume %u...", avolid);
1301 code = AFSVolGetStatus(aconn, ttid, volstatus);
1303 fprintf(STDERR, "Failed to get the status of the parent volume %lu\n",
1304 (unsigned long)avolid);
1313 code = AFSVolEndTrans(aconn, ttid, &rcode);
1314 if (code || rcode) {
1315 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
1316 (unsigned long)avolid);
1318 error = (code ? code : rcode);
1323 code = AFSVolEndTrans(aconn, btid, &rcode);
1324 if (code || rcode) {
1326 "Could not end transaction on the %s volume %lu\n",
1327 typestring, (unsigned long)cloneid);
1329 error = (code ? code : rcode);
1335 /* Move volume <afromvol> on <afromserver> <afrompart> to <atoserver>
1336 * <atopart>. The operation is almost idempotent. The following
1337 * flags are recognized:
1339 * RV_NOCLONE - don't use a copy clone
1343 UV_MoveVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
1344 afs_uint32 atoserver, afs_int32 atopart, int flags)
1346 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
1347 * be changing during the move */
1348 struct rx_connection * volatile toconn;
1349 struct rx_connection * volatile fromconn;
1350 afs_int32 volatile fromtid;
1351 afs_int32 volatile totid;
1352 afs_int32 volatile clonetid;
1353 afs_uint32 volatile newVol;
1354 afs_uint32 volatile volid;
1355 afs_uint32 volatile backupId;
1356 int volatile islocked;
1361 char tmpName[VOLSER_MAXVOLNAME + 1];
1366 struct restoreCookie cookie;
1367 afs_int32 vcode, code;
1368 struct volser_status tstatus;
1369 struct destServer destination;
1371 struct nvldbentry entry, storeEntry;
1374 char in, lf; /* for test code */
1378 #ifdef ENABLE_BUGFIX_1165
1379 volEntries volumeInfo;
1380 struct volintInfo *infop = 0;
1384 fromconn = (struct rx_connection *)0;
1385 toconn = (struct rx_connection *)0;
1395 /* support control-c processing */
1398 (void)signal(SIGINT, sigint_handler);
1402 "\nThere are three tests points - verifies all code paths through recovery.\n");
1403 fprintf(STDOUT, "First test point - operation not started.\n");
1404 fprintf(STDOUT, "...test here (y, n)? ");
1406 fscanf(stdin, "%c", &in);
1407 fscanf(stdin, "%c", &lf); /* toss away */
1409 fprintf(STDOUT, "type control-c\n");
1411 fprintf(stdout, ".");
1416 /* or drop through */
1419 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
1420 EGOTO1(mfail, vcode,
1421 "Could not fetch the entry for the volume %u from the VLDB \n",
1424 if (entry.volumeId[RWVOL] != afromvol) {
1425 fprintf(STDERR, "Only RW volume can be moved\n");
1429 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_MOVE);
1430 EGOTO1(mfail, vcode, "Could not lock entry for volume %u \n", afromvol);
1433 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
1434 EGOTO1(mfail, vcode,
1435 "Could not fetch the entry for the volume %u from the VLDB \n",
1438 backupId = entry.volumeId[BACKVOL];
1439 MapHostToNetwork(&entry);
1441 if (!Lp_Match(afromserver, afrompart, &entry)) {
1442 /* the from server and partition do not exist in the vldb entry corresponding to volid */
1443 if (!Lp_Match(atoserver, atopart, &entry)) {
1444 /* the to server and partition do not exist in the vldb entry corresponding to volid */
1445 fprintf(STDERR, "The volume %lu is not on the specified site. \n",
1446 (unsigned long)afromvol);
1447 fprintf(STDERR, "The current site is :");
1448 for (i = 0; i < entry.nServers; i++) {
1449 if (entry.serverFlags[i] == ITSRWVOL) {
1451 MapPartIdIntoName(entry.serverPartition[i], pname);
1452 fprintf(STDERR, " server %s partition %s \n",
1453 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
1454 hostutil_GetNameByINet(entry.serverNumber[i]),
1459 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1460 (LOCKREL_OPCODE | LOCKREL_AFSID |
1461 LOCKREL_TIMESTAMP));
1462 EGOTO1(mfail, vcode,
1463 " Could not release lock on the VLDB entry for the volume %u \n",
1466 return VOLSERVOLMOVED;
1469 /* delete the volume afromvol on src_server */
1470 /* from-info does not exist but to-info does =>
1471 * we have already done the move, but the volume
1472 * may still be existing physically on from fileserver
1474 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1477 code = DoVolDelete(fromconn, afromvol, afrompart,
1478 "leftover", 0, NULL, NULL);
1479 if (code && code != VNOVOL) {
1484 code = DoVolDelete(fromconn, backupId, afrompart,
1485 "leftover backup", 0, NULL, NULL);
1486 if (code && code != VNOVOL) {
1496 /* From-info matches the vldb info about volid,
1497 * its ok start the move operation, the backup volume
1498 * on the old site is deleted in the process
1500 if (afrompart == atopart) {
1501 same = VLDB_IsSameAddrs(afromserver, atoserver, &error);
1502 EGOTO2(mfail, error,
1503 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
1504 afromserver, error);
1507 EGOTO1(mfail, VOLSERVOLMOVED,
1508 "Warning: Moving volume %u to its home partition ignored!\n",
1514 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
1515 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1516 totid = 0; /* initialize to uncreated */
1519 * clone the read/write volume locally.
1522 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1523 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1525 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
1529 if (!(flags & RV_NOCLONE)) {
1530 /* Get a clone id */
1531 VPRINT1("Allocating new volume id for clone of volume %u ...",
1534 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
1536 EGOTO1(mfail, vcode,
1537 "Could not get an ID for the clone of volume %u from the VLDB\n",
1541 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
1542 VPRINT1("Cloning source volume %u ...", afromvol);
1543 strcpy(vname, "move-clone-temp");
1545 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname, &tmpVol);
1547 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
1552 /* lookup the name of the volume we just cloned */
1554 code = AFSVolGetName(fromconn, fromtid, &volName);
1555 EGOTO1(mfail, code, "Failed to get the name of the volume %u\n",
1558 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
1560 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1565 "Failed to end the transaction on the source volume %u\n",
1570 * Create the destination volume
1573 if (!(flags & RV_NOCLONE)) {
1574 /* All of this is to get the fromDate */
1575 VPRINT1("Starting transaction on the cloned volume %u ...", newVol);
1578 AFSVolTransCreate_retry(fromconn, newVol, afrompart, ITOffline,
1582 "Failed to start a transaction on the cloned volume%u\n",
1586 VPRINT1("Setting flags on cloned volume %u ...", newVol);
1588 AFSVolSetFlags(fromconn, clonetid,
1589 VTDeleteOnSalvage | VTOutOfService); /*redundant */
1590 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
1594 /* remember time from which we've dumped the volume */
1595 VPRINT1("Getting status of cloned volume %u ...", newVol);
1596 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
1598 "Failed to get the status of the cloned volume %u\n",
1602 fromDate = CLOCKADJ(tstatus.creationDate);
1604 /* With RV_NOCLONE, just do a full copy from the source */
1609 #ifdef ENABLE_BUGFIX_1165
1611 * Get the internal volume state from the source volume. We'll use such info (i.e. dayUse)
1612 * to copy it to the new volume (via AFSSetInfo later on) so that when we move volumes we
1613 * don't use this information...
1615 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
1616 volumeInfo.volEntries_len = 0;
1617 code = AFSVolListOneVolume(fromconn, afrompart, afromvol, &volumeInfo);
1619 "Failed to get the volint Info of the cloned volume %u\n",
1622 infop = (volintInfo *) volumeInfo.volEntries_val;
1623 infop->maxquota = -1; /* Else it will replace the default quota */
1624 infop->creationDate = -1; /* Else it will use the source creation date */
1625 infop->updateDate = -1; /* Else it will use the source update date */
1628 /* create a volume on the target machine */
1630 code = DoVolDelete(toconn, volid, atopart,
1631 "pre-existing destination", 0, NULL, NULL);
1632 if (code && code != VNOVOL) {
1637 VPRINT1("Creating the destination volume %u ...", volid);
1641 AFSVolCreateVolume(toconn, atopart, volName, volser_RW, volid, &tmpVol,
1645 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
1649 strncpy(tmpName, volName, VOLSER_OLDMAXVOLNAME);
1653 VPRINT1("Setting volume flags on destination volume %u ...", volid);
1655 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
1657 "Failed to set the flags on the destination volume %u\n", volid);
1661 * Now dump the clone to the new volume
1664 destination.destHost = ntohl(atoserver);
1665 destination.destPort = AFSCONF_VOLUMEPORT;
1666 destination.destSSID = 1;
1668 strncpy(cookie.name, tmpName, VOLSER_OLDMAXVOLNAME);
1669 cookie.type = RWVOL;
1670 cookie.parent = entry.volumeId[RWVOL];
1673 if (!(flags & RV_NOCLONE)) {
1674 /* Copy the clone to the new volume */
1675 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
1678 AFSVolForward(fromconn, clonetid, 0, &destination, totid,
1680 EGOTO1(mfail, code, "Failed to move data for the volume %u\n", volid);
1683 VPRINT1("Ending transaction on cloned volume %u ...", newVol);
1684 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
1689 "Failed to end the transaction on the cloned volume %u\n",
1695 * reattach to the main-line volume, and incrementally dump it.
1698 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1700 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1703 "Failed to create a transaction on the source volume %u\n",
1707 /* now do the incremental */
1709 ("Doing the%s dump from source to destination for volume %u ... ",
1710 (flags & RV_NOCLONE) ? "" : " incremental",
1713 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
1716 "Failed to do the%s dump from rw volume on old site to rw volume on newsite\n",
1717 (flags & RV_NOCLONE) ? "" : " incremental");
1720 /* now adjust the flags so that the new volume becomes official */
1721 VPRINT1("Setting volume flags on old source volume %u ...", afromvol);
1722 code = AFSVolSetFlags(fromconn, fromtid, VTOutOfService);
1724 "Failed to set the flags to make old source volume offline\n");
1727 VPRINT1("Setting volume flags on new source volume %u ...", afromvol);
1728 code = AFSVolSetFlags(toconn, totid, 0);
1730 "Failed to set the flags to make new source volume online\n");
1733 #ifdef ENABLE_BUGFIX_1165
1734 VPRINT1("Setting volume status on destination volume %u ...", volid);
1735 code = AFSVolSetInfo(toconn, totid, infop);
1737 "Failed to set volume status on the destination volume %u\n",
1742 /* put new volume online */
1743 VPRINT1("Ending transaction on destination volume %u ...", afromvol);
1744 code = AFSVolEndTrans(toconn, totid, &rcode);
1749 "Failed to end the transaction on the volume %u on the new site\n",
1753 Lp_SetRWValue(&entry, afromserver, afrompart, atoserver, atopart);
1754 MapNetworkToHost(&entry, &storeEntry);
1755 storeEntry.flags &= ~BACK_EXISTS;
1759 "Second test point - operation in progress but not complete.\n");
1760 fprintf(STDOUT, "...test here (y, n)? ");
1762 fscanf(stdin, "%c", &in);
1763 fscanf(stdin, "%c", &lf); /* toss away */
1765 fprintf(STDOUT, "type control-c\n");
1767 fprintf(stdout, ".");
1772 /* or drop through */
1775 VPRINT1("Releasing lock on VLDB entry for volume %u ...", afromvol);
1777 VLDB_ReplaceEntry(afromvol, -1, &storeEntry,
1778 (LOCKREL_OPCODE | LOCKREL_AFSID |
1779 LOCKREL_TIMESTAMP));
1782 " Could not release the lock on the VLDB entry for the volume %s %lu \n",
1783 storeEntry.name, (unsigned long)afromvol);
1792 "Third test point - operation complete but no cleanup.\n");
1793 fprintf(STDOUT, "...test here (y, n)? ");
1795 fscanf(stdin, "%c", &in);
1796 fscanf(stdin, "%c", &lf); /* toss away */
1798 fprintf(STDOUT, "type control-c\n");
1800 fprintf(stdout, ".");
1805 /* or drop through */
1808 /* This is tricky. File server is very stupid, and if you mark the volume
1809 * as VTOutOfService, it may mark the *good* instance (if you're moving
1810 * between partitions on the same machine) as out of service. Since
1811 * we're cleaning this code up in DEcorum, we're just going to kludge around
1812 * it for now by removing this call. */
1813 /* already out of service, just zap it now */
1815 AFSVolSetFlags(fromconn, fromtid, VTDeleteOnSalvage | VTOutOfService);
1818 "Failed to set the flags to make the old source volume offline\n");
1822 if (atoserver != afromserver) {
1823 /* set forwarding pointer for moved volumes */
1824 VPRINT1("Setting forwarding pointer for volume %u ...", afromvol);
1825 code = AFSVolSetForwarding(fromconn, fromtid, atoserver);
1827 "Failed to set the forwarding pointer for the volume %u\n",
1832 VPRINT1("Deleting old volume %u on source ...", afromvol);
1833 code = AFSVolDeleteVolume(fromconn, fromtid); /* zap original volume */
1834 EGOTO1(mfail, code, "Failed to delete the old volume %u on source\n",
1838 VPRINT1("Ending transaction on old volume %u on the source ...",
1840 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1845 "Failed to end the transaction on the old volume %u on the source\n",
1849 code = DoVolDelete(fromconn, backupId, afrompart,
1850 "source backup", 0, NULL, NULL);
1851 if (code && code != VNOVOL) {
1856 code = 0; /* no backup volume? that's okay */
1859 if (!(flags & RV_NOCLONE)) {
1860 code = DoVolDelete(fromconn, newVol, afrompart,
1861 "cloned", 0, NULL, NULL);
1872 fprintf(STDOUT, "Fourth test point - operation complete.\n");
1873 fprintf(STDOUT, "...test here (y, n)? ");
1875 fscanf(stdin, "%c", &in);
1876 fscanf(stdin, "%c", &lf); /* toss away */
1878 fprintf(STDOUT, "type control-c\n");
1880 fprintf(stdout, ".");
1885 /* or drop through */
1888 /* normal cleanup code */
1890 if (entry.flags & RO_EXISTS)
1891 fprintf(STDERR, "WARNING : readOnly copies still exist \n");
1894 VPRINT1("Cleanup: Releasing VLDB lock on volume %u ...", afromvol);
1896 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1897 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
1901 " Could not release the lock on the VLDB entry for the volume %lu \n",
1902 (unsigned long)afromvol);
1910 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
1912 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1913 if (code || rcode) {
1916 "Could not end transaction on the source volume %lu\n",
1917 (unsigned long)afromvol);
1919 error = (code ? code : rcode);
1925 VPRINT1("Cleanup: Ending transaction on clone volume %u ...", newVol);
1926 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
1927 if (code || rcode) {
1930 "Could not end transaction on the source's clone volume %lu\n",
1931 (unsigned long)newVol);
1933 error = (code ? code : rcode);
1939 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
1941 code = AFSVolEndTrans(toconn, totid, &rcode);
1945 "Could not end transaction on destination volume %lu\n",
1946 (unsigned long)afromvol);
1948 error = (code ? code : rcode);
1954 #ifdef ENABLE_BUGFIX_1165
1959 rx_DestroyConnection(fromconn);
1961 rx_DestroyConnection(toconn);
1962 PrintError("", error);
1965 /* come here only when the sky falls */
1970 "vos move: operation interrupted, cleanup in progress...\n");
1971 fprintf(STDOUT, "clear transaction contexts\n");
1975 /* unlock VLDB entry */
1977 VPRINT1("Recovery: Releasing VLDB lock on volume %u ...", afromvol);
1978 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1979 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
1985 VPRINT("Recovery: Ending transaction on clone volume ...");
1986 AFSVolEndTrans(fromconn, clonetid, &rcode);
1990 VPRINT("Recovery: Ending transaction on destination volume ...");
1991 AFSVolEndTrans(toconn, totid, &rcode);
1994 if (fromtid) { /* put it on-line */
1995 VPRINT("Recovery: Setting volume flags on source volume ...");
1996 AFSVolSetFlags(fromconn, fromtid, 0);
1999 VPRINT("Recovery: Ending transaction on source volume ...");
2000 AFSVolEndTrans(fromconn, fromtid, &rcode);
2004 VPRINT("Recovery: Accessing VLDB.\n");
2005 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2007 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2011 MapHostToNetwork(&entry);
2013 /* Delete either the volume on the source location or the target location.
2014 * If the vldb entry still points to the source location, then we know the
2015 * volume move didn't finish so we remove the volume from the target
2016 * location. Otherwise, we remove the volume from the source location.
2018 if (Lp_Match(afromserver, afrompart, &entry)) { /* didn't move - delete target volume */
2021 "move incomplete - attempt cleanup of target partition - no guarantee\n");
2025 if (volid && toconn)
2026 DoVolDelete(toconn, volid, atopart,
2027 "destination", 0, NULL, "Recovery:");
2029 /* put source volume on-line */
2031 VPRINT1("Recovery: Creating transaction on source volume %u ...",
2035 AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2041 VPRINT1("Recovery: Setting flags on source volume %u ...",
2043 AFSVolSetFlags(fromconn, fromtid, 0);
2047 ("Recovery: Ending transaction on source volume %u ...",
2049 AFSVolEndTrans(fromconn, fromtid, &rcode);
2053 ("\nRecovery: Unable to start transaction on source volume %u.\n",
2057 } else { /* yep, move complete */
2060 "move complete - attempt cleanup of source partition - no guarantee\n");
2064 /* delete backup volume */
2066 DoVolDelete(fromconn, backupId, afrompart,
2067 "backup", 0, NULL, "Recovery:");
2069 DoVolDelete(fromconn, afromvol, afrompart, "source",
2070 (atoserver != afromserver)?atoserver:0,
2075 /* common cleanup - delete local clone */
2077 DoVolDelete(fromconn, newVol, afrompart,
2078 "clone", 0, NULL, "Recovery:");
2080 /* unlock VLDB entry */
2082 VPRINT1("Recovery: Releasing lock on VLDB entry for volume %u ...",
2084 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2085 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2088 done: /* routine cleanup */
2091 #ifdef ENABLE_BUGFIX_1165
2096 rx_DestroyConnection(fromconn);
2098 rx_DestroyConnection(toconn);
2101 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2109 UV_MoveVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2110 afs_uint32 atoserver, afs_int32 atopart)
2112 return UV_MoveVolume2(afromvol, afromserver, afrompart,
2113 atoserver, atopart, 0);
2117 /* Copy volume <afromvol> from <afromserver> <afrompart> to <atoserver>
2118 * <atopart>. The new volume is named by <atovolname>. The new volume
2119 * has ID <atovolid> if that is nonzero; otherwise a new ID is allocated
2120 * from the VLDB. the following flags are supported:
2122 * RV_RDONLY - target volume is RO
2123 * RV_OFFLINE - leave target volume offline
2124 * RV_CPINCR - do incremental dump if target exists
2125 * RV_NOVLDB - don't create/update VLDB entry
2126 * RV_NOCLONE - don't use a copy clone
2129 UV_CopyVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2130 char *atovolname, afs_uint32 atoserver, afs_int32 atopart,
2131 afs_uint32 atovolid, int flags)
2133 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
2134 * be changing during the copy */
2136 afs_int32 volatile clonetid;
2137 afs_int32 volatile totid;
2138 afs_int32 volatile fromtid;
2139 struct rx_connection * volatile fromconn;
2140 struct rx_connection * volatile toconn;
2141 afs_uint32 volatile cloneVol;
2145 afs_int32 fromDate, cloneFromDate;
2146 struct restoreCookie cookie;
2147 afs_int32 vcode, code;
2150 struct volser_status tstatus;
2151 struct destServer destination;
2152 struct nvldbentry entry, newentry, storeEntry;
2157 fromconn = (struct rx_connection *)0;
2158 toconn = (struct rx_connection *)0;
2166 /* support control-c processing */
2169 (void)signal(SIGINT, sigint_handler);
2171 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2172 EGOTO1(mfail, vcode,
2173 "Could not fetch the entry for the volume %u from the VLDB \n",
2175 MapHostToNetwork(&entry);
2178 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
2179 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
2180 fromtid = totid = 0; /* initialize to uncreated */
2183 * clone the read/write volume locally.
2187 if (!(flags & RV_NOCLONE)) {
2188 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2190 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2193 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
2197 /* Get a clone id */
2198 VPRINT1("Allocating new volume id for clone of volume %u ...",
2202 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
2204 EGOTO1(mfail, vcode,
2205 "Could not get an ID for the clone of volume %u from the VLDB\n",
2213 /* Get a new volume id */
2214 VPRINT1("Allocating new volume id for copy of volume %u ...", afromvol);
2216 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &newVol);
2217 EGOTO1(mfail, vcode,
2218 "Could not get an ID for the copy of volume %u from the VLDB\n",
2223 if (!(flags & RV_NOCLONE)) {
2224 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
2225 VPRINT1("Cloning source volume %u ...", afromvol);
2226 strcpy(vname, "copy-clone-temp");
2229 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname,
2232 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
2236 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
2238 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2243 "Failed to end the transaction on the source volume %u\n",
2249 * Create the destination volume
2252 if (!(flags & RV_NOCLONE)) {
2253 VPRINT1("Starting transaction on the cloned volume %u ...", cloneVol);
2256 AFSVolTransCreate_retry(fromconn, cloneVol, afrompart, ITOffline,
2260 "Failed to start a transaction on the cloned volume%u\n",
2264 VPRINT1("Setting flags on cloned volume %u ...", cloneVol);
2266 AFSVolSetFlags(fromconn, clonetid,
2267 VTDeleteOnSalvage | VTOutOfService); /*redundant */
2268 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
2272 /* remember time from which we've dumped the volume */
2273 VPRINT1("Getting status of cloned volume %u ...", cloneVol);
2274 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
2276 "Failed to get the status of the cloned volume %u\n",
2280 fromDate = CLOCKADJ(tstatus.creationDate);
2285 /* create a volume on the target machine */
2288 code = AFSVolTransCreate_retry(toconn, newVol, atopart, ITOffline, &tmp);
2291 if ((flags & RV_CPINCR)) {
2292 VPRINT1("Getting status of pre-existing volume %u ...", newVol);
2293 code = AFSVolGetStatus(toconn, totid, &tstatus);
2295 "Failed to get the status of the pre-existing volume %u\n",
2299 /* Using the update date should be OK here, but add some fudge */
2300 cloneFromDate = CLOCKADJ(tstatus.updateDate);
2301 if ((flags & RV_NOCLONE))
2302 fromDate = cloneFromDate;
2304 /* XXX We should check that the source volume's creationDate is
2305 * XXX not newer than the existing target volume, and if not,
2306 * XXX throw away the existing target and do a full dump. */
2311 /* Delete the existing volume.
2312 * While we are deleting the volume in these steps, the transaction
2313 * we started against the cloned volume (clonetid above) will be
2314 * sitting idle. It will get cleaned up after 600 seconds
2316 VPRINT1("Deleting pre-existing volume %u on destination ...", newVol);
2317 code = AFSVolDeleteVolume(toconn, totid);
2319 "Could not delete the pre-existing volume %u on destination\n",
2324 ("Ending transaction on pre-existing volume %u on destination ...",
2326 code = AFSVolEndTrans(toconn, totid, &rcode);
2331 "Could not end the transaction on pre-existing volume %u on destination\n",
2336 VPRINT1("Creating the destination volume %u ...", newVol);
2339 AFSVolCreateVolume(toconn, atopart, atovolname,
2340 (flags & RV_RDONLY) ? volser_RO : volser_RW,
2341 newVol, &newVol, &tmp);
2343 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
2347 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2349 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
2351 "Failed to set the flags on the destination volume %u\n", newVol);
2356 destination.destHost = ntohl(atoserver);
2357 destination.destPort = AFSCONF_VOLUMEPORT;
2358 destination.destSSID = 1;
2360 strncpy(cookie.name, atovolname, VOLSER_OLDMAXVOLNAME);
2361 cookie.type = (flags & RV_RDONLY) ? ROVOL : RWVOL;
2366 * Now dump the clone to the new volume
2369 if (!(flags & RV_NOCLONE)) {
2370 /* XXX probably should have some code here that checks to see if
2371 * XXX we are copying to same server and partition - if so, just
2372 * XXX use a clone to save disk space */
2374 /* Copy the clone to the new volume */
2375 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
2378 AFSVolForward(fromconn, clonetid, cloneFromDate, &destination,
2380 EGOTO1(mfail, code, "Failed to move data for the volume %u\n",
2384 VPRINT1("Ending transaction on cloned volume %u ...", cloneVol);
2385 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2390 "Failed to end the transaction on the cloned volume %u\n",
2396 * reattach to the main-line volume, and incrementally dump it.
2399 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2401 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
2404 "Failed to create a transaction on the source volume %u\n",
2408 /* now do the incremental */
2410 ("Doing the%s dump from source to destination for volume %u ... ",
2411 (flags & RV_NOCLONE) ? "" : " incremental",
2414 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
2417 "Failed to do the%s dump from old site to new site\n",
2418 (flags & RV_NOCLONE) ? "" : " incremental");
2421 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2422 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
2423 code = AFSVolSetFlags(toconn, totid, volflag);
2425 "Failed to set the flags to make destination volume online\n");
2428 /* put new volume online */
2429 VPRINT1("Ending transaction on destination volume %u ...", newVol);
2430 code = AFSVolEndTrans(toconn, totid, &rcode);
2435 "Failed to end the transaction on the destination volume %u\n",
2439 VPRINT1("Ending transaction on source volume %u ...", afromvol);
2440 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2445 "Failed to end the transaction on the source volume %u\n",
2451 if (!(flags & RV_NOCLONE)) {
2452 code = DoVolDelete(fromconn, cloneVol, afrompart,
2453 "cloned", 0, NULL, NULL);
2460 if (!(flags & RV_NOVLDB)) {
2461 /* create the vldb entry for the copied volume */
2462 strncpy(newentry.name, atovolname, VOLSER_OLDMAXVOLNAME);
2463 newentry.nServers = 1;
2464 newentry.serverNumber[0] = atoserver;
2465 newentry.serverPartition[0] = atopart;
2466 newentry.flags = (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
2467 newentry.serverFlags[0] = (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
2468 newentry.volumeId[RWVOL] = newVol;
2469 newentry.volumeId[ROVOL] = (flags & RV_RDONLY) ? newVol : 0;
2470 newentry.volumeId[BACKVOL] = 0;
2471 newentry.cloneId = 0;
2472 /*map into right byte order, before passing to xdr, the stuff has to be in host
2473 * byte order. Xdr converts it into network order */
2474 MapNetworkToHost(&newentry, &storeEntry);
2475 /* create the vldb entry */
2476 vcode = VLDB_CreateEntry(&storeEntry);
2479 "Could not create a VLDB entry for the volume %s %lu\n",
2480 atovolname, (unsigned long)newVol);
2481 /*destroy the created volume */
2482 VPRINT1("Deleting the newly created volume %u\n", newVol);
2483 AFSVolDeleteVolume(toconn, totid);
2487 VPRINT2("Created the VLDB entry for the volume %s %u\n", atovolname,
2491 /* normal cleanup code */
2494 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
2496 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2497 if (code || rcode) {
2500 "Could not end transaction on the source volume %lu\n",
2501 (unsigned long)afromvol);
2503 error = (code ? code : rcode);
2509 VPRINT1("Cleanup: Ending transaction on clone volume %u ...",
2511 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2512 if (code || rcode) {
2515 "Could not end transaction on the source's clone volume %lu\n",
2516 (unsigned long)cloneVol);
2518 error = (code ? code : rcode);
2524 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
2526 code = AFSVolEndTrans(toconn, totid, &rcode);
2530 "Could not end transaction on destination volume %lu\n",
2531 (unsigned long)newVol);
2533 error = (code ? code : rcode);
2538 rx_DestroyConnection(fromconn);
2540 rx_DestroyConnection(toconn);
2541 PrintError("", error);
2544 /* come here only when the sky falls */
2549 "vos copy: operation interrupted, cleanup in progress...\n");
2550 fprintf(STDOUT, "clear transaction contexts\n");
2555 VPRINT("Recovery: Ending transaction on clone volume ...");
2556 AFSVolEndTrans(fromconn, clonetid, &rcode);
2560 VPRINT("Recovery: Ending transaction on destination volume ...");
2561 AFSVolEndTrans(toconn, totid, &rcode);
2564 if (fromtid) { /* put it on-line */
2565 VPRINT("Recovery: Ending transaction on source volume ...");
2566 AFSVolEndTrans(fromconn, fromtid, &rcode);
2570 VPRINT("Recovery: Accessing VLDB.\n");
2571 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2573 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2577 MapHostToNetwork(&entry);
2579 /* common cleanup - delete local clone */
2581 DoVolDelete(fromconn, cloneVol, afrompart,
2582 "clone", 0, NULL, "Recovery:");
2584 done: /* routine cleanup */
2586 rx_DestroyConnection(fromconn);
2588 rx_DestroyConnection(toconn);
2591 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2599 UV_CopyVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2600 char *atovolname, afs_uint32 atoserver, afs_int32 atopart)
2602 return UV_CopyVolume2(afromvol, afromserver, afrompart,
2603 atovolname, atoserver, atopart, 0, 0);
2608 /* Make a new backup of volume <avolid> on <aserver> and <apart>
2609 * if one already exists, update it
2613 UV_BackupVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid)
2615 struct rx_connection *aconn = (struct rx_connection *)0;
2616 afs_int32 ttid = 0, btid = 0;
2617 afs_uint32 backupID;
2618 afs_int32 code = 0, rcode = 0;
2619 struct nvldbentry entry, storeEntry;
2620 afs_int32 error = 0;
2621 int vldblocked = 0, vldbmod = 0;
2623 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
2625 /* the calls to VLDB will succeed only if avolid is a RW volume,
2626 * since we are following the RW hash chain for searching */
2627 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2630 "Could not fetch the entry for the volume %lu from the VLDB \n",
2631 (unsigned long)avolid);
2635 MapHostToNetwork(&entry);
2637 /* These operations require the VLDB be locked since it means the VLDB
2638 * will change or the vldb is already locked.
2640 if (!(entry.flags & BACK_EXISTS) || /* backup volume doesnt exist */
2641 (entry.flags & VLOP_ALLOPERS) || /* vldb lock already held */
2642 (entry.volumeId[BACKVOL] == INVALID_BID)) { /* no assigned backup volume id */
2644 code = ubik_VL_SetLock(cstruct, 0, avolid, RWVOL, VLOP_BACKUP);
2647 "Could not lock the VLDB entry for the volume %lu\n",
2648 (unsigned long)avolid);
2654 /* Reread the vldb entry */
2655 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2658 "Could not fetch the entry for the volume %lu from the VLDB \n",
2659 (unsigned long)avolid);
2663 MapHostToNetwork(&entry);
2666 if (!ISNAMEVALID(entry.name)) {
2667 fprintf(STDERR, "Name of the volume %s exceeds the size limit\n",
2669 error = VOLSERBADNAME;
2673 backupID = entry.volumeId[BACKVOL];
2674 if (backupID == INVALID_BID) {
2675 /* Get a backup volume id from the VLDB and update the vldb
2678 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &backupID);
2681 "Could not allocate ID for the backup volume of %lu from the VLDB\n",
2682 (unsigned long)avolid);
2686 entry.volumeId[BACKVOL] = backupID;
2690 code = DoVolClone(aconn, avolid, apart, backupVolume, backupID, "backup",
2691 entry.name, NULL, ".backup", NULL, NULL);
2697 /* Mark vldb as backup exists */
2698 if (!(entry.flags & BACK_EXISTS)) {
2699 entry.flags |= BACK_EXISTS;
2703 /* Now go back to the backup volume and bring it on line */
2704 code = AFSVolTransCreate_retry(aconn, backupID, apart, ITOffline, &btid);
2707 "Failed to start a transaction on the backup volume %lu\n",
2708 (unsigned long)backupID);
2713 code = AFSVolSetFlags(aconn, btid, 0);
2715 fprintf(STDERR, "Could not mark the backup volume %lu on line \n",
2716 (unsigned long)backupID);
2721 code = AFSVolEndTrans(aconn, btid, &rcode);
2723 if (code || rcode) {
2725 "Failed to end the transaction on the backup volume %lu\n",
2726 (unsigned long)backupID);
2727 error = (code ? code : rcode);
2733 /* Will update the vldb below */
2737 code = AFSVolEndTrans(aconn, ttid, &rcode);
2738 if (code || rcode) {
2739 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
2740 (unsigned long)avolid);
2742 error = (code ? code : rcode);
2747 code = AFSVolEndTrans(aconn, btid, &rcode);
2748 if (code || rcode) {
2750 "Could not end transaction the backup volume %lu\n",
2751 (unsigned long)backupID);
2753 error = (code ? code : rcode);
2757 /* Now update the vldb - if modified */
2760 MapNetworkToHost(&entry, &storeEntry);
2762 VLDB_ReplaceEntry(avolid, RWVOL, &storeEntry,
2763 (LOCKREL_OPCODE | LOCKREL_AFSID |
2764 LOCKREL_TIMESTAMP));
2767 "Could not update the VLDB entry for the volume %lu \n",
2768 (unsigned long)avolid);
2774 ubik_VL_ReleaseLock(cstruct, 0, avolid, RWVOL,
2775 (LOCKREL_OPCODE | LOCKREL_AFSID |
2776 LOCKREL_TIMESTAMP));
2779 "Could not unlock the VLDB entry for the volume %lu \n",
2780 (unsigned long)avolid);
2788 rx_DestroyConnection(aconn);
2790 PrintError("", error);
2794 /* Make a new clone of volume <avolid> on <aserver> and <apart>
2795 * using volume ID <acloneid>, or a new ID allocated from the VLDB.
2796 * The new volume is named by <aname>, or by appending ".clone" to
2797 * the existing name if <aname> is NULL. The following flags are
2800 * RV_RDONLY - target volume is RO
2801 * RV_OFFLINE - leave target volume offline
2805 UV_CloneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid,
2806 afs_uint32 acloneid, char *aname, int flags)
2808 struct rx_connection *aconn = (struct rx_connection *)0;
2809 afs_int32 ttid = 0, btid = 0;
2810 afs_int32 code = 0, rcode = 0;
2811 char vname[VOLSER_MAXVOLNAME + 1];
2812 afs_int32 error = 0;
2813 volEntries volumeInfo;
2816 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
2819 volumeInfo.volEntries_val = (volintInfo *) 0;
2820 volumeInfo.volEntries_len = 0;
2821 code = AFSVolListOneVolume(aconn, apart, avolid, &volumeInfo);
2823 fprintf(stderr, "Could not get info for volume %lu\n",
2824 (unsigned long)avolid);
2828 strncpy(vname, volumeInfo.volEntries_val[0].name,
2829 VOLSER_OLDMAXVOLNAME - 7);
2830 vname[VOLSER_OLDMAXVOLNAME - 7] = 0;
2831 strcat(vname, ".clone");
2833 if (volumeInfo.volEntries_val)
2834 free(volumeInfo.volEntries_val);
2838 /* Get a clone id */
2839 VPRINT1("Allocating new volume id for clone of volume %u ...",
2841 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &acloneid);
2843 "Could not get an ID for the clone of volume %u from the VLDB\n",
2848 if (flags & RV_RWONLY)
2849 type = readwriteVolume;
2850 else if (flags & RV_RDONLY)
2851 type = readonlyVolume;
2853 type = backupVolume;
2855 code = DoVolClone(aconn, avolid, apart, type, acloneid, "clone",
2856 NULL, ".clone", NULL, NULL, NULL);
2862 /* Now go back to the backup volume and bring it on line */
2863 if (!(flags & RV_OFFLINE)) {
2864 code = AFSVolTransCreate_retry(aconn, acloneid, apart, ITOffline, &btid);
2867 "Failed to start a transaction on the clone volume %lu\n",
2868 (unsigned long)acloneid);
2873 code = AFSVolSetFlags(aconn, btid, 0);
2875 fprintf(STDERR, "Could not mark the clone volume %lu on line \n",
2876 (unsigned long)acloneid);
2881 code = AFSVolEndTrans(aconn, btid, &rcode);
2883 if (code || rcode) {
2885 "Failed to end the transaction on the clone volume %lu\n",
2886 (unsigned long)acloneid);
2887 error = (code ? code : rcode);
2896 code = AFSVolEndTrans(aconn, ttid, &rcode);
2897 if (code || rcode) {
2898 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
2899 (unsigned long)avolid);
2901 error = (code ? code : rcode);
2906 code = AFSVolEndTrans(aconn, btid, &rcode);
2907 if (code || rcode) {
2909 "Could not end transaction on the clone volume %lu\n",
2910 (unsigned long)acloneid);
2912 error = (code ? code : rcode);
2917 rx_DestroyConnection(aconn);
2919 PrintError("", error);
2923 #define ONERROR(ec, ep, es) do { \
2925 fprintf(STDERR, (es), (ep)); \
2930 #define ONERROR0(ec, es) do { \
2932 fprintf(STDERR, (es)); \
2937 #define ERROREXIT(ec) do { \
2942 /* Get a "transaction" on this replica. Create the volume
2943 * if necessary. Return the time from which a dump should
2944 * be made (0 if it's a new volume)
2947 GetTrans(struct nvldbentry *vldbEntryPtr, afs_int32 index,
2948 struct rx_connection **connPtr, afs_int32 * transPtr,
2949 afs_uint32 * crtimePtr, afs_uint32 * uptimePtr,
2950 afs_int32 *origflags, afs_uint32 tmpVolId)
2953 struct volser_status tstatus;
2958 *connPtr = (struct rx_connection *)0;
2963 /* get connection to the replication site */
2964 *connPtr = UV_Bind(vldbEntryPtr->serverNumber[index], AFSCONF_VOLUMEPORT);
2966 goto fail; /* server is down */
2968 volid = vldbEntryPtr->volumeId[ROVOL];
2972 AFSVolTransCreate_retry(*connPtr, volid,
2973 vldbEntryPtr->serverPartition[index], ITOffline,
2976 if (!code && (origflags[index] & RO_DONTUSE)) {
2977 /* If RO_DONTUSE is set, this is supposed to be an entirely new
2978 * site. Don't trust any data on it, since it is possible we
2979 * have encountered some temporary volume from some other
2980 * incomplete volume operation. It is difficult to detect if
2981 * that has happened vs if this is a legit volume, so just
2982 * delete it to be safe. */
2984 VPRINT1("Deleting extant RO_DONTUSE site on %s...",
2985 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
2986 serverNumber[index], hoststr) :
2987 hostutil_GetNameByINet(vldbEntryPtr->
2988 serverNumber[index]));
2990 code = AFSVolDeleteVolume(*connPtr, *transPtr);
2992 PrintError("Failed to delete RO_DONTUSE site: ", code);
2996 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3002 PrintError("Failed to end transaction on RO_DONTUSE site: ",
3009 /* emulate what TransCreate would have returned, so we try to
3010 * create the volume below */
3015 /* If the volume does not exist, create it */
3016 if (!volid || code) {
3020 if (volid && (code != VNOVOL)) {
3021 PrintError("Failed to start a transaction on the RO volume.\n",
3026 strcpy(volname, vldbEntryPtr->name);
3028 strcat(volname, ".roclone");
3030 strcat(volname, ".readonly");
3034 "Creating new volume %lu on replication site %s: ",
3035 tmpVolId?(unsigned long)tmpVolId:(unsigned long)volid,
3036 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3037 serverNumber[index], hoststr) :
3038 hostutil_GetNameByINet(vldbEntryPtr->
3039 serverNumber[index]));
3044 AFSVolCreateVolume(*connPtr, vldbEntryPtr->serverPartition[index],
3046 vldbEntryPtr->volumeId[RWVOL],
3047 tmpVolId?&tmpVolId:&volid,
3050 PrintError("Failed to create the ro volume: ", code);
3053 vldbEntryPtr->volumeId[ROVOL] = volid;
3057 /* The following is a bit redundant, since create sets these flags by default */
3059 AFSVolSetFlags(*connPtr, *transPtr,
3060 VTDeleteOnSalvage | VTOutOfService);
3062 PrintError("Failed to set flags on the ro volume: ", code);
3067 /* Otherwise, the transaction did succeed, so get the creation date of the
3068 * latest RO volume on the replication site
3071 VPRINT2("Updating existing ro volume %u on %s ...\n", volid,
3072 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3073 serverNumber[index], hoststr) :
3074 hostutil_GetNameByINet(vldbEntryPtr->serverNumber[index]));
3076 code = AFSVolGetStatus(*connPtr, *transPtr, &tstatus);
3078 PrintError("Failed to get status of volume on destination: ",
3083 code = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3088 code = DoVolClone(*connPtr, volid,
3089 vldbEntryPtr->serverPartition[index],
3090 readonlyVolume, tmpVolId, "temporary",
3091 vldbEntryPtr->name, NULL, ".roclone", NULL,
3096 *crtimePtr = CLOCKADJ(tstatus.creationDate);
3097 *uptimePtr = CLOCKADJ(tstatus.updateDate);
3104 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3108 if (tcode && tcode != ENOENT)
3109 PrintError("Could not end transaction on a ro volume: ", tcode);
3116 SimulateForwardMultiple(struct rx_connection *fromconn, afs_int32 fromtid,
3117 afs_int32 fromdate, manyDests * tr, afs_int32 flags,
3118 void *cookie, manyResults * results)
3122 for (i = 0; i < tr->manyDests_len; i++) {
3123 results->manyResults_val[i] =
3124 AFSVolForward(fromconn, fromtid, fromdate,
3125 &(tr->manyDests_val[i].server),
3126 tr->manyDests_val[i].trans, cookie);
3132 * Check if a trans has timed out, and recreate it if necessary.
3134 * @param[in] aconn RX connection to the relevant server
3135 * @param[inout] atid Transaction ID to check; if we recreated the trans,
3136 * contains the new trans ID on success
3137 * @param[in] apart Partition for the transaction
3138 * @param[in] astat The status of the original transaction
3140 * @return operation status
3141 * @retval 0 existing transaction is still valid, or we managed to recreate
3142 * the trans successfully
3143 * @retval nonzero Fatal error; bail out
3146 CheckTrans(struct rx_connection *aconn, afs_int32 *atid, afs_int32 apart,
3147 struct volser_status *astat)
3149 struct volser_status new_status;
3152 memset(&new_status, 0, sizeof(new_status));
3153 code = AFSVolGetStatus(aconn, *atid, &new_status);
3155 if (code == ENOENT) {
3157 VPRINT1("Old transaction on cloned volume %lu timed out, "
3158 "restarting transaction\n", (long unsigned) astat->volID);
3159 code = AFSVolTransCreate_retry(aconn, astat->volID, apart,
3162 PrintError("Failed to recreate cloned RO volume transaction\n",
3167 memset(&new_status, 0, sizeof(new_status));
3168 code = AFSVolGetStatus(aconn, *atid, &new_status);
3170 PrintError("Failed to get status on recreated transaction\n",
3175 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3176 PrintError("Recreated transaction on cloned RO volume, but "
3177 "the volume has changed!\n", 0);
3181 PrintError("Unable to get status of current cloned RO transaction\n",
3186 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3188 PrintError("Internal error: current GetStatus does not match "
3189 "original GetStatus?\n", 0);
3198 PutTrans(afs_int32 *vldbindex, struct replica *replicas,
3199 struct rx_connection **toconns, struct release *times,
3202 afs_int32 s, code = 0, rcode = 0;
3203 /* End the transactions and destroy the connections */
3204 for (s = 0; s < volcount; s++) {
3205 if (replicas[s].trans) {
3206 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
3208 replicas[s].trans = 0;
3212 if ((s == 0) || (code != ENOENT)) {
3213 PrintError("Could not end transaction on a ro volume: ",
3217 ("Transaction timed out on a ro volume. Will retry.\n",
3219 if (times[s].vldbEntryIndex < *vldbindex)
3220 *vldbindex = times[s].vldbEntryIndex;
3225 rx_DestroyConnection(toconns[s]);
3231 DoVolOnline(struct nvldbentry *vldbEntryPtr, afs_uint32 avolid, int index,
3232 char *vname, struct rx_connection *connPtr)
3234 afs_int32 code = 0, rcode = 0, onlinetid = 0;
3237 AFSVolTransCreate_retry(connPtr, avolid,
3238 vldbEntryPtr->serverPartition[index],
3242 EPRINT(code, "Could not create transaction on readonly...\n");
3245 code = AFSVolSetFlags(connPtr, onlinetid, 0);
3247 EPRINT(code, "Could not set flags on readonly...\n");
3252 AFSVolSetIdsTypes(connPtr, onlinetid, vname,
3253 ROVOL, vldbEntryPtr->volumeId[RWVOL],
3256 EPRINT(code, "Could not set ids on readonly...\n");
3259 code = AFSVolEndTrans(connPtr, onlinetid, &rcode);
3265 /* UV_ReleaseVolume()
3266 * Release volume <afromvol> on <afromserver> <afrompart> to all
3267 * its RO sites (full release). Unless the previous release was
3268 * incomplete: in which case we bring the remaining incomplete
3269 * volumes up to date with the volumes that were released
3271 * forceflag: Performs a full release.
3273 * Will create a clone from the RW, then dump the clone out to
3274 * the remaining replicas. If there is more than 1 RO sites,
3275 * ensure that the VLDB says at least one RO is available all
3276 * the time: Influences when we write back the VLDB entry.
3280 UV_ReleaseVolume(afs_uint32 afromvol, afs_uint32 afromserver,
3281 afs_int32 afrompart, int forceflag, int stayUp)
3285 afs_int32 vcode, rcode, tcode;
3286 afs_uint32 cloneVolId = 0, roVolId;
3287 struct replica *replicas = 0;
3288 struct nvldbentry entry, storeEntry;
3289 int i, volcount = 0, m, fullrelease, vldbindex;
3291 struct restoreCookie cookie;
3292 struct rx_connection **toconns = 0;
3293 struct release *times = 0;
3295 struct rx_connection *fromconn = (struct rx_connection *)0;
3296 afs_int32 error = 0;
3298 afs_int32 clonetid = 0, onlinetid;
3299 afs_int32 fromtid = 0;
3300 afs_uint32 fromdate = 0;
3301 afs_uint32 thisdate;
3305 manyResults results;
3306 int rwindex, roindex, roclone, roexists;
3307 afs_uint32 rwcrdate = 0, rwupdate = 0;
3308 afs_uint32 clcrdate;
3312 } remembertime[NMAXNSERVERS];
3313 int releasecount = 0;
3314 struct volser_status volstatus;
3316 afs_int32 origflags[NMAXNSERVERS];
3317 struct volser_status orig_status;
3318 int notreleased = 0;
3319 int tried_justnewsites = 0;
3320 int justnewsites = 0; /* are we just trying to release to new RO sites? */
3322 memset(remembertime, 0, sizeof(remembertime));
3323 memset(&results, 0, sizeof(results));
3324 memset(origflags, 0, sizeof(origflags));
3326 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_RELEASE);
3327 if (vcode != VL_RERELEASE)
3328 ONERROR(vcode, afromvol,
3329 "Could not lock the VLDB entry for the volume %u.\n");
3332 /* Get the vldb entry in readable format */
3333 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
3334 ONERROR(vcode, afromvol,
3335 "Could not fetch the entry for the volume %u from the VLDB.\n");
3336 MapHostToNetwork(&entry);
3339 EnumerateEntry(&entry);
3341 if (!ISNAMEVALID(entry.name))
3342 ONERROR(VOLSERBADOP, entry.name,
3343 "Volume name %s is too long, rename before releasing.\n");
3344 if (entry.volumeId[RWVOL] != afromvol)
3345 ONERROR(VOLSERBADOP, afromvol,
3346 "The volume %u being released is not a read-write volume.\n");
3347 if (entry.nServers <= 1)
3348 ONERROR(VOLSERBADOP, afromvol,
3349 "Volume %u has no replicas - release operation is meaningless!\n");
3350 if (strlen(entry.name) > (VOLSER_OLDMAXVOLNAME - 10))
3351 ONERROR(VOLSERBADOP, entry.name,
3352 "RO volume name %s exceeds (VOLSER_OLDMAXVOLNAME - 10) character limit\n");
3354 /* roclone is true if one of the RO volumes is on the same
3355 * partition as the RW volume. In this case, we make the RO volume
3356 * on the same partition a clone instead of a complete copy.
3359 roindex = Lp_ROMatch(afromserver, afrompart, &entry) - 1;
3360 roclone = ((roindex == -1) ? 0 : 1);
3361 rwindex = Lp_GetRwIndex(&entry);
3363 ONERROR0(VOLSERNOVOL, "There is no RW volume \n");
3365 /* Make sure we have a RO volume id to work with */
3366 if (entry.volumeId[ROVOL] == INVALID_BID) {
3367 /* need to get a new RO volume id */
3368 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &roVolId);
3369 ONERROR(vcode, entry.name, "Cant allocate ID for RO volume of %s\n");
3371 entry.volumeId[ROVOL] = roVolId;
3372 MapNetworkToHost(&entry, &storeEntry);
3373 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3374 ONERROR(vcode, entry.name, "Could not update vldb entry for %s.\n");
3377 /* Will we be completing a previously unfinished release. -force overrides */
3378 for (s = 0, m = 0, fullrelease=0, i=0; (i<entry.nServers); i++) {
3379 if (entry.serverFlags[i] & ITSROVOL) {
3381 if (entry.serverFlags[i] & NEW_REPSITE) s++;
3382 if (entry.serverFlags[i] & RO_DONTUSE) notreleased++;
3384 origflags[i] = entry.serverFlags[i];
3386 if ((forceflag && !fullrelease) || (s == m) || (s == 0))
3389 if (!forceflag && (s == m || s == 0)) {
3390 if (notreleased && notreleased != m) {
3391 /* we have some new unreleased sites. try to just release to those,
3392 * if the RW has not changed */
3397 /* Determine which volume id to use and see if it exists */
3400 || (entry.cloneId == 0)) ? entry.volumeId[ROVOL] : entry.cloneId);
3401 code = VolumeExists(afromserver, afrompart, cloneVolId);
3402 roexists = ((code == ENODEV) ? 0 : 1);
3404 /* For stayUp case, if roclone is the only site, bypass special handling */
3405 if (stayUp && roclone) {
3409 for (e = 0; (e < entry.nServers) && !error; e++) {
3410 if ((entry.serverFlags[e] & ITSROVOL)) {
3411 if (!(VLDB_IsSameAddrs(entry.serverNumber[e], afromserver,
3416 if (e >= entry.nServers)
3420 /* If we had a previous release to complete, do so, else: */
3421 if (stayUp && (cloneVolId == entry.volumeId[ROVOL])) {
3422 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &cloneVolId);
3423 ONERROR(code, afromvol,
3424 "Cannot get temporary clone id for volume %u\n");
3427 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
3429 ONERROR(-1, afromserver,
3430 "Cannot establish connection with server 0x%x\n");
3434 fullrelease = 1; /* Do a full release if RO clone does not exist */
3436 /* Begin transaction on RW and mark it busy while we query it */
3437 code = AFSVolTransCreate_retry(
3438 fromconn, afromvol, afrompart, ITBusy, &fromtid
3440 ONERROR(code, afromvol,
3441 "Failed to start transaction on RW volume %u\n");
3443 /* Query the creation date for the RW */
3444 code = AFSVolGetStatus(fromconn, fromtid, &volstatus);
3445 ONERROR(code, afromvol,
3446 "Failed to get the status of RW volume %u\n");
3447 rwcrdate = volstatus.creationDate;
3449 /* End transaction on RW */
3450 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
3452 ONERROR((code ? code : rcode), afromvol,
3453 "Failed to end transaction on RW volume %u\n");
3455 /* Begin transaction on clone and mark it busy while we query it */
3456 code = AFSVolTransCreate_retry(
3457 fromconn, cloneVolId, afrompart, ITBusy, &clonetid
3459 ONERROR(code, cloneVolId,
3460 "Failed to start transaction on RW clone %u\n");
3462 /* Query the creation date for the clone */
3463 code = AFSVolGetStatus(fromconn, clonetid, &volstatus);
3464 ONERROR(code, cloneVolId,
3465 "Failed to get the status of RW clone %u\n");
3466 clcrdate = volstatus.creationDate;
3468 /* End transaction on clone */
3469 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
3471 ONERROR((code ? code : rcode), cloneVolId,
3472 "Failed to end transaction on RW clone %u\n");
3474 if (rwcrdate > clcrdate)
3475 fullrelease = 2;/* Do a full release if RO clone older than RW */
3479 if (fullrelease != 1) {
3480 /* in case the RW has changed, and just to be safe */
3485 switch (fullrelease) {
3487 fprintf(STDOUT, "RW %lu changed, doing a complete release\n",
3488 (unsigned long)afromvol);
3491 fprintf(STDOUT, "This is a complete release of volume %lu\n",
3492 (unsigned long)afromvol);
3494 tried_justnewsites = 1;
3495 fprintf(STDOUT, "There are new RO sites; we will try to "
3496 "only release to new sites\n");
3500 fprintf(STDOUT, "This is a completion of a previous release\n");
3506 afs_int32 oldest = 0;
3507 /* If the RO clone exists, then if the clone is a temporary
3508 * clone, delete it. Or if the RO clone is marked RO_DONTUSE
3509 * (it was recently added), then also delete it. We do not
3510 * want to "reclone" a temporary RO clone.
3513 code = VolumeExists(afromserver, afrompart, cloneVolId);
3515 code = DoVolDelete(fromconn, cloneVolId, afrompart, "previous clone", 0,
3517 if (code && (code != VNOVOL))
3522 /* clean up any previous tmp clone before starting if staying up */
3524 && (!roclone || (entry.serverFlags[roindex] & RO_DONTUSE))) {
3525 code = DoVolDelete(fromconn,
3526 stayUp ? entry.volumeId[ROVOL] : cloneVolId,
3527 afrompart, "the", 0, NULL, NULL);
3528 if (code && (code != VNOVOL))
3534 VPRINT("Querying old RO sites for update times...");
3535 for (vldbindex = 0; vldbindex < entry.nServers; vldbindex++) {
3536 volEntries volumeInfo;
3537 struct rx_connection *conn;
3540 if (!(entry.serverFlags[vldbindex] & ITSROVOL)) {
3543 if ((entry.serverFlags[vldbindex] & RO_DONTUSE)) {
3546 conn = UV_Bind(entry.serverNumber[vldbindex], AFSCONF_VOLUMEPORT);
3548 fprintf(STDERR, "Cannot establish connection to server %s\n",
3549 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3553 volumeInfo.volEntries_val = NULL;
3554 volumeInfo.volEntries_len = 0;
3555 code = AFSVolListOneVolume(conn, entry.serverPartition[vldbindex],
3556 entry.volumeId[ROVOL],
3559 fprintf(STDERR, "Could not fetch information about RO vol %lu from server %s\n",
3560 (unsigned long)entry.volumeId[ROVOL],
3561 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3562 PrintError("", code);
3564 rx_DestroyConnection(conn);
3568 crdate = CLOCKADJ(volumeInfo.volEntries_val[0].creationDate);
3570 if (oldest == 0 || crdate < oldest) {
3574 rx_DestroyConnection(conn);
3575 free(volumeInfo.volEntries_val);
3576 volumeInfo.volEntries_val = NULL;
3577 volumeInfo.volEntries_len = 0;
3582 volEntries volumeInfo;
3583 volumeInfo.volEntries_val = NULL;
3584 volumeInfo.volEntries_len = 0;
3585 code = AFSVolListOneVolume(fromconn, afrompart, afromvol,
3588 fprintf(STDERR, "Could not fetch information about RW vol %lu from server %s\n",
3589 (unsigned long)afromvol,
3590 hostutil_GetNameByINet(afromserver));
3591 PrintError("", code);
3594 rwupdate = volumeInfo.volEntries_val[0].updateDate;
3596 free(volumeInfo.volEntries_val);
3597 volumeInfo.volEntries_val = NULL;
3598 volumeInfo.volEntries_len = 0;
3601 if (justnewsites && oldest <= rwupdate) {
3602 /* RW has changed */
3606 /* Mark all the ROs in the VLDB entry as RO_DONTUSE. We don't
3607 * write this entry out to the vlserver until after the first
3608 * RO volume is released (temp RO clones don't count).
3610 * If 'justnewsites' is set, we're only updating sites that have
3611 * RO_DONTUSE set, so set NEW_REPSITE for all of the others.
3613 for (i = 0; i < entry.nServers; i++) {
3615 if ((entry.serverFlags[i] & RO_DONTUSE)) {
3616 entry.serverFlags[i] &= ~NEW_REPSITE;
3618 entry.serverFlags[i] |= NEW_REPSITE;
3621 entry.serverFlags[i] &= ~NEW_REPSITE;
3622 entry.serverFlags[i] |= RO_DONTUSE;
3625 entry.serverFlags[rwindex] |= NEW_REPSITE;
3626 entry.serverFlags[rwindex] &= ~RO_DONTUSE;
3629 if (justnewsites && roexists) {
3630 /* if 'justnewsites' and 'roexists' are set, we don't need to do
3631 * anything with the RO clone, so skip the reclone */
3634 } else if (fullrelease) {
3637 strcpy(vname, entry.name);
3639 strcat(vname, ".roclone");
3641 strcat(vname, ".readonly");
3643 strcpy(vname, "readonly-clone-temp");
3646 code = DoVolClone(fromconn, afromvol, afrompart, readonlyVolume,
3647 cloneVolId, (roclone && !stayUp)?"permanent RO":
3648 "temporary RO", NULL, vname, NULL, &volstatus, NULL);
3654 if (justnewsites && rwupdate != volstatus.updateDate) {
3656 /* reset the serverFlags as if 'justnewsites' had never been set */
3657 for (i = 0; i < entry.nServers; i++) {
3658 entry.serverFlags[i] &= ~NEW_REPSITE;
3659 entry.serverFlags[i] |= RO_DONTUSE;
3661 entry.serverFlags[rwindex] |= NEW_REPSITE;
3662 entry.serverFlags[rwindex] &= ~RO_DONTUSE;
3665 rwcrdate = volstatus.creationDate;
3667 /* Remember clone volume ID in case we fail or are interrupted */
3668 entry.cloneId = cloneVolId;
3670 if (roclone && !stayUp) {
3671 /* Bring the RO clone online - though not if it's a temporary clone */
3672 VPRINT1("Starting transaction on RO clone volume %u...",
3675 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITOffline,
3677 ONERROR(code, cloneVolId,
3678 "Failed to start transaction on volume %u\n");
3681 VPRINT1("Setting volume flags for volume %u...", cloneVolId);
3682 tcode = AFSVolSetFlags(fromconn, onlinetid, 0);
3685 VPRINT1("Ending transaction on volume %u...", cloneVolId);
3686 code = AFSVolEndTrans(fromconn, onlinetid, &rcode);
3687 ONERROR((code ? code : rcode), cloneVolId,
3688 "Failed to end transaction on RO clone %u\n");
3691 ONERROR(tcode, cloneVolId, "Could not bring volume %u on line\n");
3693 /* Sleep so that a client searching for an online volume won't
3694 * find the clone offline and then the next RO offline while the
3695 * release brings the clone online and the next RO offline (race).
3696 * There is a fix in the 3.4 client that does not need this sleep
3697 * anymore, but we don't know what clients we have.
3699 if (entry.nServers > 2 && !justnewsites)
3702 /* Mark the RO clone in the VLDB as a good site (already released) */
3703 entry.serverFlags[roindex] |= NEW_REPSITE;
3704 entry.serverFlags[roindex] &= ~RO_DONTUSE;
3705 entry.flags |= RO_EXISTS;
3709 /* Write out the VLDB entry only if the clone is not a temporary
3710 * clone. If we did this to a temporary clone then we would end
3711 * up marking all the ROs as "old release" making the ROs
3712 * temporarily unavailable.
3714 MapNetworkToHost(&entry, &storeEntry);
3715 VPRINT1("Replacing VLDB entry for %s...", entry.name);
3716 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3717 ONERROR(vcode, entry.name,
3718 "Could not update vldb entry for %s.\n");
3724 VPRINT("RW vol has not changed; only releasing to new RO sites\n");
3725 /* act like this is a completion of a previous release */
3727 } else if (tried_justnewsites) {
3728 VPRINT("RW vol has changed; releasing to all sites\n");
3731 /* Now we will release from the clone to the remaining RO replicas.
3732 * The first 2 ROs (counting the non-temporary RO clone) are released
3733 * individually: releasecount. This is to reduce the race condition
3734 * of clients trying to find an on-line RO volume. The remaining ROs
3735 * are released in parallel but no more than half the number of ROs
3736 * (rounded up) at a time: nservers.
3739 strcpy(vname, entry.name);
3741 strcat(vname, ".roclone");
3743 strcat(vname, ".readonly");
3744 memset(&cookie, 0, sizeof(cookie));
3745 strncpy(cookie.name, vname, VOLSER_OLDMAXVOLNAME);
3746 cookie.type = ROVOL;
3747 cookie.parent = entry.volumeId[RWVOL];
3750 /* how many to do at once, excluding clone */
3751 if (stayUp || justnewsites)
3752 nservers = entry.nServers; /* can do all, none offline */
3754 nservers = entry.nServers / 2;
3755 replicas = calloc(nservers + 1, sizeof(struct replica));
3756 times = calloc(nservers + 1, sizeof(struct release));
3757 toconns = calloc(nservers + 1, sizeof(struct rx_connection *));
3758 results.manyResults_val = calloc(nservers + 1, sizeof(afs_int32));
3759 if (!replicas || !times || !results.manyResults_val || !toconns)
3761 "Failed to create transaction on the release clone\n");
3763 /* Create a transaction on the cloned volume */
3764 VPRINT1("Starting transaction on cloned volume %u...", cloneVolId);
3766 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITBusy, &fromtid);
3768 memset(&orig_status, 0, sizeof(orig_status));
3769 code = AFSVolGetStatus(fromconn, fromtid, &orig_status);
3771 if (!fullrelease && code)
3772 ONERROR(VOLSERNOVOL, afromvol,
3773 "Old clone is inaccessible. Try vos release -f %u.\n");
3774 ONERROR0(code, "Failed to create transaction on the release clone\n");
3777 /* if we have a clone, treat this as done, for now */
3778 if (stayUp && !fullrelease) {
3779 entry.serverFlags[roindex] |= NEW_REPSITE;
3780 entry.serverFlags[roindex] &= ~RO_DONTUSE;
3781 entry.flags |= RO_EXISTS;
3786 /* For each index in the VLDB */
3787 for (vldbindex = 0; vldbindex < entry.nServers;) {
3788 /* Get a transaction on the replicas. Pick replicas which have an old release. */
3790 ((volcount < nservers) && (vldbindex < entry.nServers));
3792 if (!stayUp && !justnewsites) {
3793 /* The first two RO volumes will be released individually.
3794 * The rest are then released in parallel. This is a hack
3795 * for clients not recognizing right away when a RO volume
3796 * comes back on-line.
3798 if ((volcount == 1) && (releasecount < 2))
3802 if (vldbindex == roindex)
3803 continue; /* the clone */
3804 if ((entry.serverFlags[vldbindex] & NEW_REPSITE)
3805 && !(entry.serverFlags[vldbindex] & RO_DONTUSE))
3807 if (!(entry.serverFlags[vldbindex] & ITSROVOL))
3808 continue; /* not a RO vol */
3811 /* Get a Transaction on this replica. Get a new connection if
3812 * necessary. Create the volume if necessary. Return the
3813 * time from which the dump should be made (0 if it's a new
3814 * volume). Each volume might have a different time.
3816 replicas[volcount].server.destHost =
3817 ntohl(entry.serverNumber[vldbindex]);
3818 replicas[volcount].server.destPort = AFSCONF_VOLUMEPORT;
3819 replicas[volcount].server.destSSID = 1;
3820 times[volcount].vldbEntryIndex = vldbindex;
3823 GetTrans(&entry, vldbindex, &(toconns[volcount]),
3824 &(replicas[volcount].trans),
3825 &(times[volcount].crtime),
3826 &(times[volcount].uptime),
3827 origflags, stayUp?cloneVolId:0);
3831 /* Thisdate is the date from which we want to pick up all changes */
3832 if (forceflag || !fullrelease
3833 || (rwcrdate > times[volcount].crtime)) {
3834 /* If the forceflag is set, then we want to do a full dump.
3835 * If it's not a full release, we can't be sure that the creation
3836 * date is good (so we also do a full dump).
3837 * If the RW volume was replaced (its creation date is newer than
3838 * the last release), then we can't be sure what has changed (so
3839 * we do a full dump).
3842 } else if (remembertime[vldbindex].validtime) {
3843 /* Trans was prev ended. Use the time from the prev trans
3844 * because, prev trans may have created the volume. In which
3845 * case time[volcount].time would be now instead of 0.
3848 (remembertime[vldbindex].uptime < times[volcount].uptime)
3849 ? remembertime[vldbindex].uptime
3850 : times[volcount].uptime;
3852 thisdate = times[volcount].uptime;
3854 remembertime[vldbindex].validtime = 1;
3855 remembertime[vldbindex].uptime = thisdate;
3857 if (volcount == 0) {
3858 fromdate = thisdate;
3860 /* Include this volume if it is within 15 minutes of the earliest */
3862 thisdate) ? (fromdate - thisdate) : (thisdate -
3864 AFSVolEndTrans(toconns[volcount],
3865 replicas[volcount].trans, &rcode);
3866 replicas[volcount].trans = 0;
3869 if (thisdate < fromdate)
3870 fromdate = thisdate;
3877 code = CheckTrans(fromconn, &fromtid, afrompart, &orig_status);
3884 fprintf(STDOUT, "Starting ForwardMulti from %lu to %u on %s",
3885 (unsigned long)cloneVolId, stayUp?
3886 cloneVolId:entry.volumeId[ROVOL],
3887 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[0].
3888 vldbEntryIndex], hoststr) :
3889 hostutil_GetNameByINet(entry.
3890 serverNumber[times[0].
3893 for (s = 1; s < volcount; s++) {
3894 fprintf(STDOUT, " and %s",
3895 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[s].
3896 vldbEntryIndex], hoststr) :
3897 hostutil_GetNameByINet(entry.
3898 serverNumber[times[s].
3903 fprintf(STDOUT, " (full release)");
3906 fprintf(STDOUT, " (as of %.24s)", ctime(&tmv));
3908 fprintf(STDOUT, ".\n");
3912 /* Release the ones we have collected */
3913 tr.manyDests_val = &(replicas[0]);
3914 tr.manyDests_len = results.manyResults_len = volcount;
3916 AFSVolForwardMultiple(fromconn, fromtid, fromdate, &tr,
3917 0 /*spare */ , &cookie, &results);
3918 if (code == RXGEN_OPCODE) { /* RPC Interface Mismatch */
3920 SimulateForwardMultiple(fromconn, fromtid, fromdate, &tr,
3921 0 /*spare */ , &cookie, &results);
3926 PrintError("Release failed: ", code);
3928 for (m = 0; m < volcount; m++) {
3929 if (results.manyResults_val[m]) {
3930 if ((m == 0) || (results.manyResults_val[m] != ENOENT)) {
3931 /* we retry timed out transaction. When it is
3932 * not the first volume and the transaction wasn't found
3933 * (assume it timed out and was garbage collected by volser).
3936 ("Failed to dump volume from clone to a ro site: ",
3937 results.manyResults_val[m]);
3943 AFSVolSetIdsTypes(toconns[m], replicas[m].trans, vname,
3944 ROVOL, entry.volumeId[RWVOL], 0, 0);
3946 if ((m == 0) || (code != ENOENT)) {
3947 PrintError("Failed to set correct names and ids: ",
3953 /* have to clear dest. flags to ensure new vol goes online:
3954 * because the restore (forwarded) operation copied
3955 * the V_inService(=0) flag over to the destination.
3957 code = AFSVolSetFlags(toconns[m], replicas[m].trans, 0);
3959 if ((m == 0) || (code != ENOENT)) {
3960 PrintError("Failed to set flags on ro volume: ",
3966 entry.serverFlags[times[m].vldbEntryIndex] |= NEW_REPSITE;
3967 entry.serverFlags[times[m].vldbEntryIndex] &= ~RO_DONTUSE;
3968 entry.flags |= RO_EXISTS;
3974 PutTrans(&vldbindex, replicas, toconns, times, volcount);
3975 MapNetworkToHost(&entry, &storeEntry);
3976 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3977 ONERROR(vcode, afromvol,
3978 " Could not update VLDB entry for volume %u\n");
3980 } /* for each index in the vldb */
3982 /* for the stayup case, put back at the end */
3984 afs_uint32 tmpVol = entry.volumeId[ROVOL];
3985 strcpy(vname, entry.name);
3986 strcat(vname, ".readonly");
3989 /* have to clear flags to ensure new vol goes online
3991 code = AFSVolSetFlags(fromconn, fromtid, 0);
3992 if (code && (code != ENOENT)) {
3993 PrintError("Failed to set flags on ro volume: ",
3997 VPRINT3("%sloning to permanent RO %u on %s...", roexists?"Re-c":"C", tmpVol,
3999 afs_inet_ntoa_r(entry.serverNumber[roindex],
4001 hostutil_GetNameByINet(entry.serverNumber[roindex]));
4003 code = AFSVolClone(fromconn, fromtid, roexists?tmpVol:0,
4004 readonlyVolume, vname, &tmpVol);
4008 VPRINT("Bringing readonly online...");
4009 code = DoVolOnline(&entry, tmpVol, roindex, vname,
4013 EPRINT(code, "Failed: ");
4014 entry.serverFlags[roindex] &= ~NEW_REPSITE;
4015 entry.serverFlags[roindex] |= RO_DONTUSE;
4017 entry.serverFlags[roindex] |= NEW_REPSITE;
4018 entry.serverFlags[roindex] &= ~RO_DONTUSE;
4019 entry.flags |= RO_EXISTS;
4024 for (s = 0; s < volcount; s++) {
4025 if (replicas[s].trans) {
4026 vldbindex = times[s].vldbEntryIndex;
4028 /* ok, so now we have to end the previous transaction */
4029 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
4034 code = AFSVolTransCreate_retry(toconns[s],
4036 entry.serverPartition[vldbindex],
4038 &(replicas[s].trans));
4040 PrintError("Unable to begin transaction on temporary clone: ", code);
4043 PrintError("Unable to end transaction on temporary clone: ", code);
4046 VPRINT3("%sloning to permanent RO %u on %s...", times[s].crtime?"Re-c":"C",
4048 afs_inet_ntoa_r(htonl(replicas[s].server.destHost),
4050 hostutil_GetNameByINet(htonl(replicas[s].server.destHost)));
4051 if (times[s].crtime)
4052 code = AFSVolClone(toconns[s], replicas[s].trans, tmpVol,
4053 readonlyVolume, vname, &tmpVol);
4055 code = AFSVolClone(toconns[s], replicas[s].trans, 0,
4056 readonlyVolume, vname, &tmpVol);
4059 if (!times[s].crtime) {
4060 entry.serverFlags[vldbindex] |= RO_DONTUSE;
4062 entry.serverFlags[vldbindex] &= ~NEW_REPSITE;
4063 PrintError("Failed: ",
4068 if (entry.serverFlags[vldbindex] != RO_DONTUSE) {
4069 /* bring it online (mark it InService) */
4070 VPRINT1("Bringing readonly online on %s...",
4073 htonl(replicas[s].server.destHost),
4075 hostutil_GetNameByINet(
4076 htonl(replicas[s].server.destHost)));
4078 code = DoVolOnline(&entry, tmpVol, vldbindex, vname,
4080 /* needed to come online for cloning */
4082 /* technically it's still new, just not online */
4083 entry.serverFlags[s] &= ~NEW_REPSITE;
4084 entry.serverFlags[s] |= RO_DONTUSE;
4085 if (code != ENOENT) {
4086 PrintError("Failed to set correct names and ids: ",
4093 VPRINT("Marking temporary clone for deletion...\n");
4094 code = AFSVolSetFlags(toconns[s],
4099 EPRINT(code, "Failed: ");
4103 VPRINT("Ending transaction on temporary clone...\n");
4104 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
4108 PrintError("Failed: ", code);
4111 /* ended successfully */
4112 replicas[s].trans = 0;
4114 VPRINT2("Deleting temporary clone %u on %s...", cloneVolId,
4116 afs_inet_ntoa_r(htonl(replicas[s].server.destHost),
4118 hostutil_GetNameByINet(htonl(replicas[s].server.destHost)));
4119 code = DoVolDelete(toconns[s], cloneVolId,
4120 entry.serverPartition[vldbindex],
4121 NULL, 0, NULL, NULL);
4123 EPRINT(code, "Failed: ");
4130 /* done. put the vldb entry in the success tail case*/
4131 PutTrans(&vldbindex, replicas, toconns, times, volcount);
4134 /* End the transaction on the cloned volume */
4135 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4140 PrintError("Failed to end transaction on rw volume: ", code);
4142 /* Figure out if any volume were not released and say so */
4143 for (failure = 0, i = 0; i < entry.nServers; i++) {
4144 if (!(entry.serverFlags[i] & NEW_REPSITE))
4150 "The volume %lu could not be released to the following %d sites:\n",
4151 (unsigned long)afromvol, failure);
4152 for (i = 0; i < entry.nServers; i++) {
4153 if (!(entry.serverFlags[i] & NEW_REPSITE)) {
4154 MapPartIdIntoName(entry.serverPartition[i], pname);
4155 fprintf(STDERR, "\t%35s %s\n",
4156 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
4157 hostutil_GetNameByINet(entry.serverNumber[i]), pname);
4160 MapNetworkToHost(&entry, &storeEntry);
4162 VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry,
4164 ONERROR(vcode, afromvol,
4165 " Could not update VLDB entry for volume %u\n");
4167 ERROREXIT(VOLSERBADRELEASE);
4171 /* All the ROs were release successfully. Remove the temporary clone */
4172 if (!roclone || stayUp) {
4174 fprintf(STDOUT, "Deleting the releaseClone %lu ...",
4175 (unsigned long)cloneVolId);
4178 code = DoVolDelete(fromconn, cloneVolId, afrompart, NULL, 0, NULL,
4180 ONERROR(code, cloneVolId, "Failed to delete volume %u.\n");
4184 for (i = 0; i < entry.nServers; i++)
4185 entry.serverFlags[i] &= ~NEW_REPSITE;
4187 /* Update the VLDB */
4188 VPRINT("updating VLDB ...");
4190 MapNetworkToHost(&entry, &storeEntry);
4192 VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry,
4193 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4194 ONERROR(vcode, afromvol, " Could not update VLDB entry for volume %u\n");
4199 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
4203 "Failed to end cloning transaction on the RW volume %lu\n",
4204 (unsigned long)afromvol);
4210 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4214 "Failed to end transaction on the release clone %lu\n",
4215 (unsigned long)cloneVolId);
4220 for (i = 0; i < nservers; i++) {
4221 if (replicas && replicas[i].trans) {
4222 code = AFSVolEndTrans(toconns[i], replicas[i].trans, &rcode);
4223 replicas[i].trans = 0;
4226 "Failed to end transaction on ro volume %u at server %s\n",
4227 entry.volumeId[ROVOL],
4228 noresolve ? afs_inet_ntoa_r(htonl(replicas[i].server.
4229 destHost), hoststr) :
4230 hostutil_GetNameByINet(htonl
4231 (replicas[i].server.destHost)));
4236 if (toconns && toconns[i]) {
4237 rx_DestroyConnection(toconns[i]);
4243 ubik_VL_ReleaseLock(cstruct, 0, afromvol, RWVOL,
4244 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4247 "Could not release lock on the VLDB entry for volume %lu\n",
4248 (unsigned long)afromvol);
4254 PrintError("", error);
4257 rx_DestroyConnection(fromconn);
4258 if (results.manyResults_val)
4259 free(results.manyResults_val);
4271 dump_sig_handler(int x)
4273 fprintf(STDERR, "\nSignal handler: vos dump operation\n");
4277 /* Dump the volume <afromvol> on <afromserver> and
4278 * <afrompart> to <afilename> starting from <fromdate>.
4279 * DumpFunction does the real work behind the scenes after
4280 * extracting parameters from the rock
4283 UV_DumpVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
4285 afs_int32(*DumpFunction) (struct rx_call *, void *), void *rock,
4288 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
4289 * be changing during the dump */
4290 struct rx_call * volatile fromcall = NULL;
4291 struct rx_connection * volatile fromconn = NULL;
4292 afs_int32 volatile fromtid = 0;
4294 afs_int32 rxError = 0, rcode = 0;
4295 afs_int32 code, error = 0;
4297 time_t tmv = fromdate;
4301 #ifndef AFS_NT40_ENV
4302 (void)signal(SIGPIPE, dump_sig_handler);
4304 (void)signal(SIGINT, dump_sig_handler);
4307 VEPRINT("Full Dump ...\n");
4309 VEPRINT1("Incremental Dump (as of %.24s)...\n",
4313 /* get connections to the servers */
4314 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
4316 VEPRINT1("Starting transaction on volume %u...", afromvol);
4318 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
4320 EGOTO1(error_exit, code,
4321 "Could not start transaction on the volume %u to be dumped\n",
4325 fromcall = rx_NewCall(fromconn);
4327 VEPRINT1("Starting volume dump on volume %u...", afromvol);
4328 if (flags & VOLDUMPV2_OMITDIRS)
4329 code = StartAFSVolDumpV2(fromcall, fromtid, fromdate, flags);
4331 code = StartAFSVolDump(fromcall, fromtid, fromdate);
4332 EGOTO(error_exit, code, "Could not start the dump process \n");
4335 VEPRINT1("Dumping volume %u...", afromvol);
4336 code = DumpFunction(fromcall, rock);
4337 if (code == RXGEN_OPCODE)
4339 EGOTO(error_exit, code, "Error while dumping volume \n");
4344 code = rx_EndCall(fromcall, rxError);
4345 if (code && code != RXGEN_OPCODE)
4346 fprintf(STDERR, "Error in rx_EndCall\n");
4351 VEPRINT1("Ending transaction on volume %u...", afromvol);
4352 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4353 if (code || rcode) {
4354 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
4355 (unsigned long)afromvol);
4357 error = (code ? code : rcode);
4362 rx_DestroyConnection(fromconn);
4364 if (error != RXGEN_OPCODE)
4365 PrintError("", error);
4369 /* Clone the volume <afromvol> on <afromserver> and
4370 * <afrompart>, and then dump the clone volume to
4371 * <afilename> starting from <fromdate>.
4372 * DumpFunction does the real work behind the scenes after
4373 * extracting parameters from the rock
4376 UV_DumpClonedVolume(afs_uint32 afromvol, afs_uint32 afromserver,
4377 afs_int32 afrompart, afs_int32 fromdate,
4378 afs_int32(*DumpFunction) (struct rx_call *, void *),
4379 void *rock, afs_int32 flags)
4381 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
4382 * be changing during the dump */
4383 struct rx_connection * volatile fromconn = NULL;
4384 struct rx_call * volatile fromcall = NULL;
4385 afs_int32 volatile clonetid = 0;
4386 afs_uint32 volatile clonevol = 0;
4389 afs_int32 fromtid = 0, rxError = 0, rcode = 0;
4390 afs_int32 code = 0, error = 0;
4393 time_t tmv = fromdate;
4397 #ifndef AFS_NT40_ENV
4398 (void)signal(SIGPIPE, dump_sig_handler);
4400 (void)signal(SIGINT, dump_sig_handler);
4403 VEPRINT("Full Dump ...\n");
4405 VEPRINT1("Incremental Dump (as of %.24s)...\n",
4409 /* get connections to the servers */
4410 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
4412 VEPRINT1("Starting transaction on volume %u...", afromvol);
4413 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &fromtid);
4414 EGOTO1(error_exit, code,
4415 "Could not start transaction on the volume %u to be dumped\n",
4419 /* Get a clone id */
4420 VEPRINT1("Allocating new volume id for clone of volume %u ...", afromvol);
4422 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
4424 EGOTO1(error_exit, code,
4425 "Could not get an ID for the clone of volume %u from the VLDB\n",
4429 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
4430 VEPRINT2("Cloning source volume %u to clone volume %u...", afromvol,
4432 strcpy(vname, "dump-clone-temp");
4435 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname, &tmpVol);
4437 EGOTO1(error_exit, code, "Failed to clone the source volume %u\n",
4441 VEPRINT1("Ending the transaction on the volume %u ...", afromvol);
4443 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4447 EGOTO1(error_exit, code,
4448 "Failed to end the transaction on the volume %u\n", afromvol);
4452 VEPRINT1("Starting transaction on the cloned volume %u ...", clonevol);
4455 AFSVolTransCreate_retry(fromconn, clonevol, afrompart, ITOffline,
4458 EGOTO1(error_exit, code,
4459 "Failed to start a transaction on the cloned volume%u\n",
4463 VEPRINT1("Setting flags on cloned volume %u ...", clonevol);
4464 code = AFSVolSetFlags(fromconn, clonetid, VTDeleteOnSalvage | VTOutOfService); /*redundant */
4465 EGOTO1(error_exit, code, "Could not set falgs on the cloned volume %u\n",
4470 fromcall = rx_NewCall(fromconn);
4472 VEPRINT1("Starting volume dump from cloned volume %u...", clonevol);
4473 if (flags & VOLDUMPV2_OMITDIRS)
4474 code = StartAFSVolDumpV2(fromcall, clonetid, fromdate, flags);
4476 code = StartAFSVolDump(fromcall, clonetid, fromdate);
4477 EGOTO(error_exit, code, "Could not start the dump process \n");
4480 VEPRINT1("Dumping volume %u...", afromvol);
4481 code = DumpFunction(fromcall, rock);
4482 EGOTO(error_exit, code, "Error while dumping volume \n");
4486 /* now delete the clone */
4487 VEPRINT1("Deleting the cloned volume %u ...", clonevol);
4488 code = AFSVolDeleteVolume(fromconn, clonetid);
4490 fprintf(STDERR, "Failed to delete the cloned volume %lu\n",
4491 (unsigned long)clonevol);
4497 code = rx_EndCall(fromcall, rxError);
4499 fprintf(STDERR, "Error in rx_EndCall\n");
4505 VEPRINT1("Ending transaction on cloned volume %u...", clonevol);
4506 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
4507 if (code || rcode) {
4509 "Could not end transaction on the cloned volume %lu\n",
4510 (unsigned long)clonevol);
4512 error = (code ? code : rcode);
4517 rx_DestroyConnection(fromconn);
4519 PrintError("", error);
4526 * Restore a volume <tovolid> <tovolname> on <toserver> <topart> from
4527 * the dump file <afilename>. WriteData does all the real work
4528 * after extracting params from the rock
4531 UV_RestoreVolume2(afs_uint32 toserver, afs_int32 topart, afs_uint32 tovolid,
4532 afs_uint32 toparentid, char tovolname[], int flags,
4533 afs_int32(*WriteData) (struct rx_call *, void *),
4536 struct rx_connection *toconn, *tempconn;
4537 struct rx_call *tocall;
4538 afs_int32 totid, code, rcode, vcode, terror = 0;
4539 afs_int32 rxError = 0;
4540 struct volser_status tstatus;
4541 struct volintInfo vinfo;
4543 char tovolreal[VOLSER_OLDMAXVOLNAME];
4545 afs_int32 temptid, pparentid;
4546 struct nvldbentry entry, storeEntry;
4549 struct restoreCookie cookie;
4551 afs_int32 volflag, voltype, volsertype;
4552 afs_int32 oldCreateDate, oldUpdateDate, newCreateDate, newUpdateDate;
4553 int index, same, errcode;
4557 memset(&cookie, 0, sizeof(cookie));
4561 tocall = (struct rx_call *)0;
4562 tempconn = (struct rx_connection *)0;
4566 if (flags & RV_RDONLY) {
4568 volsertype = volser_RO;
4571 volsertype = volser_RW;
4575 pparentid = toparentid;
4576 toconn = UV_Bind(toserver, AFSCONF_VOLUMEPORT);
4577 if (pvolid == 0) { /*alot a new id if needed */
4578 vcode = VLDB_GetEntryByName(tovolname, &entry);
4579 if (vcode == VL_NOENT) {
4580 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &pvolid);
4582 fprintf(STDERR, "Could not get an Id for the volume %s\n",
4588 } else if (flags & RV_RDONLY) {
4589 if (entry.flags & RW_EXISTS) {
4591 "Entry for ReadWrite volume %s already exists!\n",
4593 error = VOLSERBADOP;
4596 if (!entry.volumeId[ROVOL]) {
4598 "Existing entry for volume %s has no ReadOnly ID\n",
4600 error = VOLSERBADOP;
4603 pvolid = entry.volumeId[ROVOL];
4604 pparentid = entry.volumeId[RWVOL];
4606 pvolid = entry.volumeId[RWVOL];
4607 pparentid = entry.volumeId[RWVOL];
4610 if (!pparentid) pparentid = pvolid;
4611 /* at this point we have a volume id to use/reuse for the volume to be restored */
4612 strncpy(tovolreal, tovolname, VOLSER_OLDMAXVOLNAME);
4614 if (strlen(tovolname) > (VOLSER_OLDMAXVOLNAME - 1)) {
4615 EGOTO1(refail, VOLSERBADOP,
4616 "The volume name %s exceeds the maximum limit of (VOLSER_OLDMAXVOLNAME -1 ) bytes\n",
4619 if ((pparentid != pvolid) && (flags & RV_RDONLY)) {
4620 if (strlen(tovolname) > (VOLSER_OLDMAXVOLNAME - 10)) {
4621 EGOTO1(refail, VOLSERBADOP,
4622 "The volume name %s exceeds the maximum limit of (VOLSER_OLDMAXVOLNAME -1 ) bytes\n", tovolname);
4624 snprintf(tovolreal, VOLSER_OLDMAXVOLNAME, "%s.readonly", tovolname);
4627 MapPartIdIntoName(topart, partName);
4628 fprintf(STDOUT, "Restoring volume %s Id %lu on server %s partition %s ..",
4629 tovolreal, (unsigned long)pvolid,
4630 noresolve ? afs_inet_ntoa_r(toserver, hoststr) :
4631 hostutil_GetNameByINet(toserver), partName);
4634 AFSVolCreateVolume(toconn, topart, tovolreal, volsertype, pparentid, &pvolid,
4637 if (flags & RV_FULLRST) { /* full restore: delete then create anew */
4638 code = DoVolDelete(toconn, pvolid, topart, "the previous", 0,
4640 if (code && code != VNOVOL) {
4646 AFSVolCreateVolume(toconn, topart, tovolreal, volsertype, pparentid,
4648 EGOTO1(refail, code, "Could not create new volume %u\n", pvolid);
4651 AFSVolTransCreate_retry(toconn, pvolid, topart, ITOffline, &totid);
4652 EGOTO1(refail, code, "Failed to start transaction on %u\n",
4655 code = AFSVolGetStatus(toconn, totid, &tstatus);
4656 EGOTO1(refail, code, "Could not get timestamp from volume %u\n",
4660 oldCreateDate = tstatus.creationDate;
4661 oldUpdateDate = tstatus.updateDate;
4667 cookie.parent = pparentid;
4668 cookie.type = voltype;
4670 strncpy(cookie.name, tovolreal, VOLSER_OLDMAXVOLNAME);
4672 tocall = rx_NewCall(toconn);
4673 terror = StartAFSVolRestore(tocall, totid, 1, &cookie);
4675 fprintf(STDERR, "Volume restore Failed \n");
4679 code = WriteData(tocall, rock);
4681 fprintf(STDERR, "Could not transmit data\n");
4685 terror = rx_EndCall(tocall, rxError);
4686 tocall = (struct rx_call *)0;
4688 fprintf(STDERR, "rx_EndCall Failed \n");
4692 code = AFSVolGetStatus(toconn, totid, &tstatus);
4695 "Could not get status information about the volume %lu\n",
4696 (unsigned long)pvolid);
4700 code = AFSVolSetIdsTypes(toconn, totid, tovolreal, voltype, pparentid, 0, 0);
4702 fprintf(STDERR, "Could not set the right type and ID on %lu\n",
4703 (unsigned long)pvolid);
4708 if (flags & RV_CRDUMP)
4709 newCreateDate = tstatus.creationDate;
4710 else if (flags & RV_CRKEEP && oldCreateDate != 0)
4711 newCreateDate = oldCreateDate;
4713 newCreateDate = time(0);
4714 if (flags & RV_LUDUMP)
4715 newUpdateDate = tstatus.updateDate;
4716 else if (flags & RV_LUKEEP)
4717 newUpdateDate = oldUpdateDate;
4719 newUpdateDate = time(0);
4720 code = AFSVolSetDate(toconn,totid, newCreateDate);
4722 fprintf(STDERR, "Could not set the 'creation' date on %u\n", pvolid);
4727 init_volintInfo(&vinfo);
4728 vinfo.creationDate = newCreateDate;
4729 vinfo.updateDate = newUpdateDate;
4730 code = AFSVolSetInfo(toconn, totid, &vinfo);
4732 fprintf(STDERR, "Could not set the 'last updated' date on %u\n",
4738 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
4739 code = AFSVolSetFlags(toconn, totid, volflag);
4741 fprintf(STDERR, "Could not mark %lu online\n", (unsigned long)pvolid);
4746 /* It isn't handled right in refail */
4747 code = AFSVolEndTrans(toconn, totid, &rcode);
4752 fprintf(STDERR, "Could not end transaction on %lu\n",
4753 (unsigned long)pvolid);
4758 fprintf(STDOUT, " done\n");
4760 if (!reuseID || (flags & RV_FULLRST)) {
4761 /* Volume was restored on the file server, update the
4762 * VLDB to reflect the change.
4764 vcode = VLDB_GetEntryByID(pvolid, voltype, &entry);
4765 if (vcode && vcode != VL_NOENT && vcode != VL_ENTDELETED) {
4767 "Could not fetch the entry for volume number %lu from VLDB \n",
4768 (unsigned long)pvolid);
4773 MapHostToNetwork(&entry);
4774 if (vcode == VL_NOENT) { /* it doesnot exist already */
4775 /*make the vldb return this indication specifically */
4776 VPRINT("------- Creating a new VLDB entry ------- \n");
4777 strcpy(entry.name, tovolname);
4779 entry.serverNumber[0] = toserver; /*should be indirect */
4780 entry.serverPartition[0] = topart;
4781 entry.serverFlags[0] = (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
4782 entry.flags = (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
4783 if (flags & RV_RDONLY)
4784 entry.volumeId[ROVOL] = pvolid;
4785 else if (tstatus.cloneID != 0) {
4786 entry.volumeId[ROVOL] = tstatus.cloneID; /*this should come from status info on the volume if non zero */
4788 entry.volumeId[ROVOL] = INVALID_BID;
4789 entry.volumeId[RWVOL] = pparentid;
4791 if (tstatus.backupID != 0) {
4792 entry.volumeId[BACKVOL] = tstatus.backupID;
4793 /*this should come from status info on the volume if non zero */
4795 entry.volumeId[BACKVOL] = INVALID_BID;
4796 MapNetworkToHost(&entry, &storeEntry);
4797 vcode = VLDB_CreateEntry(&storeEntry);
4800 "Could not create the VLDB entry for volume number %lu \n",
4801 (unsigned long)pvolid);
4807 EnumerateEntry(&entry);
4808 } else { /*update the existing entry */
4810 fprintf(STDOUT, "Updating the existing VLDB entry\n");
4811 fprintf(STDOUT, "------- Old entry -------\n");
4812 EnumerateEntry(&entry);
4813 fprintf(STDOUT, "------- New entry -------\n");
4816 ubik_VL_SetLock(cstruct, 0, pvolid, voltype,
4820 "Could not lock the entry for volume number %lu \n",
4821 (unsigned long)pvolid);
4826 strcpy(entry.name, tovolname);
4828 /* Update the vlentry with the new information */
4829 if (flags & RV_RDONLY)
4830 index = Lp_ROMatch(toserver, topart, &entry) - 1;
4832 index = Lp_GetRwIndex(&entry);
4834 /* Add the new site for the volume being restored */
4835 entry.serverNumber[entry.nServers] = toserver;
4836 entry.serverPartition[entry.nServers] = topart;
4837 entry.serverFlags[entry.nServers] =
4838 (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
4841 /* This volume should be deleted on the old site
4842 * if its different from new site.
4845 VLDB_IsSameAddrs(toserver, entry.serverNumber[index],
4849 "Failed to get info about server's %d address(es) from vlserver (err=%d)\n",
4851 if ((!errcode && !same)
4852 || (entry.serverPartition[index] != topart)) {
4853 if (flags & RV_NODEL) {
4855 ("Not deleting the previous volume %u on server %s, ...",
4857 noresolve ? afs_inet_ntoa_r(entry.serverNumber[index], hoststr) :
4858 hostutil_GetNameByINet(entry.serverNumber[index]));
4861 UV_Bind(entry.serverNumber[index],
4862 AFSCONF_VOLUMEPORT);
4864 MapPartIdIntoName(entry.serverPartition[index],
4867 ("Deleting the previous volume %u on server %s, partition %s ...",
4869 noresolve ? afs_inet_ntoa_r(entry.serverNumber[index], hoststr) :
4870 hostutil_GetNameByINet(entry.serverNumber[index]),
4872 code = DoVolDelete(tempconn, pvolid,
4873 entry.serverPartition[index],
4874 "the", 0, NULL, NULL);
4875 if (code && code != VNOVOL) {
4879 MapPartIdIntoName(entry.serverPartition[index],
4883 entry.serverNumber[index] = toserver;
4884 entry.serverPartition[index] = topart;
4887 entry.flags |= (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
4888 MapNetworkToHost(&entry, &storeEntry);
4890 VLDB_ReplaceEntry(pvolid, voltype, &storeEntry,
4891 LOCKREL_OPCODE | LOCKREL_AFSID |
4895 "Could not update the entry for volume number %lu \n",
4896 (unsigned long)pvolid);
4902 EnumerateEntry(&entry);
4909 code = rx_EndCall(tocall, rxError);
4915 ubik_VL_ReleaseLock(cstruct, 0, pvolid, voltype,
4916 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4919 "Could not release lock on the VLDB entry for the volume %lu\n",
4920 (unsigned long)pvolid);
4926 code = AFSVolEndTrans(toconn, totid, &rcode);
4930 fprintf(STDERR, "Could not end transaction on the volume %lu \n",
4931 (unsigned long)pvolid);
4937 code = AFSVolEndTrans(toconn, temptid, &rcode);
4941 fprintf(STDERR, "Could not end transaction on the volume %lu \n",
4942 (unsigned long)pvolid);
4948 rx_DestroyConnection(tempconn);
4950 rx_DestroyConnection(toconn);
4951 PrintError("", error);
4956 UV_RestoreVolume(afs_uint32 toserver, afs_int32 topart, afs_uint32 tovolid,
4957 char tovolname[], int flags,
4958 afs_int32(*WriteData) (struct rx_call *, void *),
4961 return UV_RestoreVolume2(toserver, topart, tovolid, 0, tovolname, flags,
4966 /*unlocks the vldb entry associated with <volid> */
4968 UV_LockRelease(afs_uint32 volid)
4972 VPRINT("Binding to the VLDB server\n");
4974 ubik_VL_ReleaseLock(cstruct, 0, volid, -1,
4975 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4978 "Could not unlock the entry for volume number %lu in VLDB \n",
4979 (unsigned long)volid);
4980 PrintError("", vcode);
4983 VPRINT("VLDB updated\n");
4988 /* old interface to add rosites */
4990 UV_AddSite(afs_uint32 server, afs_int32 part, afs_uint32 volid,
4993 return UV_AddSite2(server, part, volid, 0, valid);
4996 /*adds <server> and <part> as a readonly replication site for <volid>
4999 UV_AddSite2(afs_uint32 server, afs_int32 part, afs_uint32 volid,
5000 afs_uint32 rovolid, afs_int32 valid)
5002 int j, nro = 0, islocked = 0;
5003 struct nvldbentry entry, storeEntry, entry2;
5004 afs_int32 vcode, error = 0;
5007 error = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5010 " Could not lock the VLDB entry for the volume %lu \n",
5011 (unsigned long)volid);
5016 error = VLDB_GetEntryByID(volid, RWVOL, &entry);
5019 "Could not fetch the VLDB entry for volume number %lu \n",
5020 (unsigned long)volid);
5024 if (!ISNAMEVALID(entry.name)) {
5026 "Volume name %s is too long, rename before adding site\n",
5028 error = VOLSERBADOP;
5031 MapHostToNetwork(&entry);
5033 /* See if it's too many entries */
5034 if (entry.nServers >= NMAXNSERVERS) {
5035 fprintf(STDERR, "Total number of entries will exceed %u\n",
5037 error = VOLSERBADOP;
5041 /* See if it's on the same server */
5042 for (j = 0; j < entry.nServers; j++) {
5043 if (entry.serverFlags[j] & ITSROVOL) {
5045 if (VLDB_IsSameAddrs(server, entry.serverNumber[j], &error)) {
5048 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
5051 MapPartIdIntoName(entry.serverPartition[j], apartName);
5053 "RO already exists on partition %s. Multiple ROs on a single server aren't allowed\n",
5055 error = VOLSERBADOP;
5062 /* See if it's too many RO sites - leave one for the RW */
5063 if (nro >= NMAXNSERVERS - 1) {
5064 fprintf(STDERR, "Total number of sites will exceed %u\n",
5066 error = VOLSERBADOP;
5070 /* if rovolid == 0, we leave the RO volume id alone. If the volume doesn't
5071 * have an RO volid at this point (i.e. entry.volumeId[ROVOL] ==
5072 * INVALID_BID) and we leave it alone, it gets an RO volid at release-time.
5075 if (entry.volumeId[ROVOL] == INVALID_BID) {
5076 vcode = VLDB_GetEntryByID(rovolid, -1, &entry2);
5078 fprintf(STDERR, "Volume ID %d already exists\n", rovolid);
5081 VPRINT1("Using RO volume id %d.\n", rovolid);
5082 entry.volumeId[ROVOL] = rovolid;
5084 fprintf(STDERR, "Ignoring given RO id %d, since volume already has RO id %d\n",
5085 rovolid, entry.volumeId[ROVOL]);
5089 VPRINT("Adding a new site ...");
5090 entry.serverNumber[entry.nServers] = server;
5091 entry.serverPartition[entry.nServers] = part;
5093 entry.serverFlags[entry.nServers] = (ITSROVOL | RO_DONTUSE);
5095 entry.serverFlags[entry.nServers] = (ITSROVOL);
5099 MapNetworkToHost(&entry, &storeEntry);
5101 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5102 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5104 fprintf(STDERR, "Could not update entry for volume %lu \n",
5105 (unsigned long)volid);
5114 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5115 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5118 "Could not release lock on volume entry for %lu \n",
5119 (unsigned long)volid);
5120 PrintError("", vcode);
5124 PrintError("", error);
5128 /*removes <server> <part> as read only site for <volid> from the vldb */
5130 UV_RemoveSite(afs_uint32 server, afs_int32 part, afs_uint32 volid)
5133 struct nvldbentry entry, storeEntry;
5135 vcode = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5137 fprintf(STDERR, " Could not lock the VLDB entry for volume %lu \n",
5138 (unsigned long)volid);
5139 PrintError("", vcode);
5142 vcode = VLDB_GetEntryByID(volid, RWVOL, &entry);
5145 "Could not fetch the entry for volume number %lu from VLDB \n",
5146 (unsigned long)volid);
5147 PrintError("", vcode);
5150 MapHostToNetwork(&entry);
5151 if (!Lp_ROMatch(server, part, &entry)) {
5152 /*this site doesnot exist */
5153 fprintf(STDERR, "This site is not a replication site \n");
5155 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5156 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5158 fprintf(STDERR, "Could not update entry for volume %lu \n",
5159 (unsigned long)volid);
5160 PrintError("", vcode);
5161 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5162 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5166 } else { /*remove the rep site */
5167 Lp_SetROValue(&entry, server, part, 0, 0);
5169 if ((entry.nServers == 1) && (entry.flags & RW_EXISTS))
5170 entry.flags &= ~RO_EXISTS;
5171 if (entry.nServers < 1) { /*this is the last ref */
5172 VPRINT1("Deleting the VLDB entry for %u ...", volid);
5174 vcode = ubik_VL_DeleteEntry(cstruct, 0, volid, ROVOL);
5177 "Could not delete VLDB entry for volume %lu \n",
5178 (unsigned long)volid);
5179 PrintError("", vcode);
5184 MapNetworkToHost(&entry, &storeEntry);
5185 fprintf(STDOUT, "Deleting the replication site for volume %lu ...",
5186 (unsigned long)volid);
5189 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5190 LOCKREL_OPCODE | LOCKREL_AFSID |
5194 "Could not release lock on volume entry for %lu \n",
5195 (unsigned long)volid);
5196 PrintError("", vcode);
5197 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5198 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5206 /*sets <server> <part> as read/write site for <volid> in the vldb */
5208 UV_ChangeLocation(afs_uint32 server, afs_int32 part, afs_uint32 volid)
5211 struct nvldbentry entry, storeEntry;
5214 vcode = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5216 fprintf(STDERR, " Could not lock the VLDB entry for volume %lu \n",
5217 (unsigned long)volid);
5218 PrintError("", vcode);
5221 vcode = VLDB_GetEntryByID(volid, RWVOL, &entry);
5224 "Could not fetch the entry for volume number %lu from VLDB \n",
5225 (unsigned long)volid);
5226 PrintError("", vcode);
5229 MapHostToNetwork(&entry);
5230 index = Lp_GetRwIndex(&entry);
5232 /* no RW site exists */
5233 fprintf(STDERR, "No existing RW site for volume %lu",
5234 (unsigned long)volid);
5236 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5237 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5240 "Could not release lock on entry for volume %lu \n",
5241 (unsigned long)volid);
5242 PrintError("", vcode);
5246 } else { /* change the RW site */
5247 entry.serverNumber[index] = server;
5248 entry.serverPartition[index] = part;
5249 MapNetworkToHost(&entry, &storeEntry);
5251 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5252 LOCKREL_OPCODE | LOCKREL_AFSID |
5255 fprintf(STDERR, "Could not update entry for volume %lu \n",
5256 (unsigned long)volid);
5257 PrintError("", vcode);
5258 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5259 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5267 /*list all the partitions on <aserver> */
5269 UV_ListPartitions(afs_uint32 aserver, struct partList *ptrPartList,
5272 struct rx_connection *aconn;
5273 struct pIDs partIds;
5274 struct partEntries partEnts;
5278 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5280 partEnts.partEntries_len = 0;
5281 partEnts.partEntries_val = NULL;
5282 code = AFSVolXListPartitions(aconn, &partEnts); /* this is available only on new servers */
5283 if (code == RXGEN_OPCODE) {
5284 for (i = 0; i < 26; i++) /* try old interface */
5285 partIds.partIds[i] = -1;
5286 code = AFSVolListPartitions(aconn, &partIds);
5288 for (i = 0; i < 26; i++) {
5289 if ((partIds.partIds[i]) != -1) {
5290 ptrPartList->partId[j] = partIds.partIds[i];
5291 ptrPartList->partFlags[j] = PARTVALID;
5294 ptrPartList->partFlags[i] = 0;
5299 *cntp = partEnts.partEntries_len;
5300 if (*cntp > VOLMAXPARTS) {
5302 "Warning: number of partitions on the server too high %d (process only %d)\n",
5303 *cntp, VOLMAXPARTS);
5304 *cntp = VOLMAXPARTS;
5306 for (i = 0; i < *cntp; i++) {
5307 ptrPartList->partId[i] = partEnts.partEntries_val[i];
5308 ptrPartList->partFlags[i] = PARTVALID;
5310 free(partEnts.partEntries_val);
5316 "Could not fetch the list of partitions from the server\n");
5317 PrintError("", code);
5319 rx_DestroyConnection(aconn);
5324 /*zap the list of volumes specified by volPtrArray (the volCloneId field).
5325 This is used by the backup system */
5327 UV_ZapVolumeClones(afs_uint32 aserver, afs_int32 apart,
5328 struct volDescription *volPtr, afs_int32 arraySize)
5330 struct rx_connection *aconn;
5331 struct volDescription *curPtr;
5334 afs_int32 success = 1;
5336 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5338 for (curPtr = volPtr; curPos < arraySize; curPtr++) {
5339 if (curPtr->volFlags & CLONEVALID) {
5340 curPtr->volFlags &= ~CLONEZAPPED;
5343 code = DoVolDelete(aconn, curPtr->volCloneId, apart,
5344 "clone", 0, NULL, NULL);
5349 curPtr->volFlags |= CLONEZAPPED;
5351 fprintf(STDERR, "Could not zap volume %lu\n",
5352 (unsigned long)curPtr->volCloneId);
5354 VPRINT2("Clone of %s %u deleted\n", curPtr->volName,
5355 curPtr->volCloneId);
5360 rx_DestroyConnection(aconn);
5364 /*return a list of clones of the volumes specified by volPtrArray. Used by the
5367 UV_GenerateVolumeClones(afs_uint32 aserver, afs_int32 apart,
5368 struct volDescription *volPtr, afs_int32 arraySize)
5370 struct rx_connection *aconn;
5371 struct volDescription *curPtr;
5374 afs_int32 rcode = 0;
5376 int reuseCloneId = 0;
5377 afs_uint32 curCloneId = 0;
5378 char cloneName[256]; /*max vol name */
5380 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5382 if ((volPtr->volFlags & REUSECLONEID) && (volPtr->volFlags & ENTRYVALID))
5384 else { /*get a bunch of id's from vldb */
5386 ubik_VL_GetNewVolumeId(cstruct, 0, arraySize, &curCloneId);
5388 fprintf(STDERR, "Could not get ID's for the clone from VLDB\n");
5389 PrintError("", code);
5394 for (curPtr = volPtr; curPos < arraySize; curPtr++) {
5395 if (curPtr->volFlags & ENTRYVALID) {
5397 curPtr->volFlags |= CLONEVALID;
5398 /*make a clone of curParentId and record as curPtr->volCloneId */
5400 AFSVolTransCreate_retry(aconn, curPtr->volId, apart, ITOffline,
5403 VPRINT2("Clone for volume %s %u failed \n", curPtr->volName,
5406 curPtr->volFlags &= ~CLONEVALID; /*cant clone */
5410 if (strlen(curPtr->volName) < (VOLSER_OLDMAXVOLNAME - 9)) {
5411 strcpy(cloneName, curPtr->volName);
5412 strcat(cloneName, "-tmpClone-");
5414 strcpy(cloneName, "-tmpClone");
5416 curPtr->volCloneId = curCloneId;
5421 AFSVolClone(aconn, tid, 0, readonlyVolume, cloneName,
5422 &(curPtr->volCloneId));
5424 curPtr->volFlags &= ~CLONEVALID;
5426 fprintf(STDERR, "Could not clone %s due to error %lu\n",
5427 curPtr->volName, (unsigned long)code);
5428 code = AFSVolEndTrans(aconn, tid, &rcode);
5430 fprintf(STDERR, "WARNING: could not end transaction\n");
5433 VPRINT2("********** Cloned %s temporary %u\n", cloneName,
5434 curPtr->volCloneId);
5435 code = AFSVolEndTrans(aconn, tid, &rcode);
5436 if (code || rcode) {
5437 curPtr->volFlags &= ~CLONEVALID;
5446 rx_DestroyConnection(aconn);
5451 /*list all the volumes on <aserver> and <apart>. If all = 1, then all the
5452 * relevant fields of the volume are also returned. This is a heavy weight operation.*/
5454 UV_ListVolumes(afs_uint32 aserver, afs_int32 apart, int all,
5455 struct volintInfo **resultPtr, afs_int32 * size)
5457 struct rx_connection *aconn;
5459 volEntries volumeInfo;
5463 *resultPtr = (volintInfo *) 0;
5464 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
5465 volumeInfo.volEntries_len = 0;
5467 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5468 code = AFSVolListVolumes(aconn, apart, all, &volumeInfo);
5471 "Could not fetch the list of volumes from the server\n");
5473 *resultPtr = volumeInfo.volEntries_val;
5474 *size = volumeInfo.volEntries_len;
5478 rx_DestroyConnection(aconn);
5479 PrintError("", code);
5483 /*------------------------------------------------------------------------
5484 * EXPORTED UV_XListVolumes
5487 * List the extended information for all the volumes on a particular
5488 * File Server and partition. We may either return the volume's ID
5489 * or all of its extended information.
5492 * a_serverID : Address of the File Server for which we want
5493 * extended volume info.
5494 * a_partID : Partition for which we want the extended
5496 * a_all : If non-zero, fetch ALL the volume info,
5497 * otherwise just the volume ID.
5498 * a_resultPP : Ptr to the address of the area containing
5499 * the returned volume info.
5500 * a_numEntsInResultP : Ptr for the value we set for the number of
5505 * Otherise, the return value of AFSVolXListVolumes.
5508 * This routine is closely related to UV_ListVolumes, which returns
5509 * only the standard level of detail on AFS volumes. It is a
5510 * heavyweight operation, zipping through all the volume entries for
5511 * a given server/partition.
5515 *------------------------------------------------------------------------*/
5518 UV_XListVolumes(afs_uint32 a_serverID, afs_int32 a_partID, int a_all,
5519 struct volintXInfo **a_resultPP,
5520 afs_int32 * a_numEntsInResultP)
5522 struct rx_connection *rxConnP; /*Ptr to the Rx connection involved */
5523 afs_int32 code; /*Error code to return */
5524 volXEntries volumeXInfo; /*Area for returned extended vol info */
5527 * Set up our error code and the area for returned extended volume info.
5528 * We set the val field to a null pointer as a hint for the stub to
5531 *a_numEntsInResultP = 0;
5532 *a_resultPP = (volintXInfo *) 0;
5533 volumeXInfo.volXEntries_val = (volintXInfo *) 0;
5534 volumeXInfo.volXEntries_len = 0;
5537 * Bind to the Volume Server port on the File Server machine in question,
5540 rxConnP = UV_Bind(a_serverID, AFSCONF_VOLUMEPORT);
5541 code = AFSVolXListVolumes(rxConnP, a_partID, a_all, &volumeXInfo);
5543 fprintf(STDERR, "[UV_XListVolumes] Couldn't fetch volume list\n");
5546 * We got the info; pull out the pointer to where the results lie
5547 * and how many entries are there.
5549 *a_resultPP = volumeXInfo.volXEntries_val;
5550 *a_numEntsInResultP = volumeXInfo.volXEntries_len;
5554 * If we got an Rx connection, throw it away.
5557 rx_DestroyConnection(rxConnP);
5559 PrintError("", code);
5561 } /*UV_XListVolumes */
5563 /* get all the information about volume <volid> on <aserver> and <apart> */
5565 UV_ListOneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 volid,
5566 struct volintInfo **resultPtr)
5568 struct rx_connection *aconn;
5570 volEntries volumeInfo;
5572 *resultPtr = (volintInfo *) 0;
5573 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
5574 volumeInfo.volEntries_len = 0;
5576 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5577 code = AFSVolListOneVolume(aconn, apart, volid, &volumeInfo);
5580 "Could not fetch the information about volume %lu from the server\n",
5581 (unsigned long)volid);
5583 *resultPtr = volumeInfo.volEntries_val;
5588 rx_DestroyConnection(aconn);
5589 PrintError("", code);
5593 /*------------------------------------------------------------------------
5594 * EXPORTED UV_XListOneVolume
5597 * List the extended information for a volume on a particular File
5598 * Server and partition.
5601 * a_serverID : Address of the File Server for which we want
5602 * extended volume info.
5603 * a_partID : Partition for which we want the extended
5605 * a_volID : Volume ID for which we want the info.
5606 * a_resultPP : Ptr to the address of the area containing
5607 * the returned volume info.
5611 * Otherise, the return value of AFSVolXListOneVolume.
5614 * This routine is closely related to UV_ListOneVolume, which returns
5615 * only the standard level of detail on the chosen AFS volume.
5619 *------------------------------------------------------------------------*/
5622 UV_XListOneVolume(afs_uint32 a_serverID, afs_int32 a_partID, afs_uint32 a_volID,
5623 struct volintXInfo **a_resultPP)
5625 struct rx_connection *rxConnP; /*Rx connection to Volume Server */
5626 afs_int32 code; /*Error code */
5627 volXEntries volumeXInfo; /*Area for returned info */
5630 * Set up our error code, and the area we're in which we are returning
5631 * the info. Setting the val field to a null pointer tells the stub
5632 * to allocate space for us.
5634 *a_resultPP = (volintXInfo *) 0;
5635 volumeXInfo.volXEntries_val = (volintXInfo *) 0;
5636 volumeXInfo.volXEntries_len = 0;
5639 * Bind to the Volume Server port on the File Server machine in question,
5642 rxConnP = UV_Bind(a_serverID, AFSCONF_VOLUMEPORT);
5643 code = AFSVolXListOneVolume(rxConnP, a_partID, a_volID, &volumeXInfo);
5646 "[UV_XListOneVolume] Couldn't fetch the volume information\n");
5649 * We got the info; pull out the pointer to where the results lie.
5651 *a_resultPP = volumeXInfo.volXEntries_val;
5654 * If we got an Rx connection, throw it away.
5657 rx_DestroyConnection(rxConnP);
5659 PrintError("", code);
5664 * Given a volume we read from a partition, check if it is
5665 * represented in the VLDB correctly.
5667 * The VLDB is looked up by the RW volume id (not its name).
5668 * The RW contains the true name of the volume (BK and RO set
5669 * the name in the VLDB only on creation of the VLDB entry).
5670 * We want rules strict enough that when we check all volumes
5671 * on one partition, it does not need to be done again. IE:
5672 * two volumes on different partitions won't constantly
5673 * change a VLDB entry away from what the other set.
5674 * For RW and BK volumes, we will always check the VLDB to see
5675 * if the two exist on the server/partition. May seem redundant,
5676 * but this is an easy check of the VLDB. IE: if the VLDB entry
5677 * says the BK exists but no BK volume is there, we will detect
5678 * this when we check the RW volume.
5679 * VLDB entries are locked only when a change needs to be done.
5680 * Output changed to look a lot like the "vos syncserv" otuput.
5683 CheckVolume(volintInfo * volumeinfo, afs_uint32 aserver, afs_int32 apart,
5684 afs_int32 * modentry, afs_uint32 * maxvolid,
5685 struct nvldbentry *aentry)
5689 afs_int32 code, error = 0;
5690 struct nvldbentry entry, storeEntry;
5692 int pass = 0, createentry, addvolume, modified, mod, doit = 1;
5702 ((volumeinfo->type ==
5703 RWVOL) ? volumeinfo->volid : volumeinfo->parentID);
5706 /* Check to see if the VLDB is ok without locking it (pass 1).
5707 * If it will change, then lock the VLDB entry, read it again,
5708 * then make the changes to it (pass 2).
5711 code = ubik_VL_SetLock(cstruct, 0, rwvolid, RWVOL, VLOP_DELETE);
5713 fprintf(STDERR, "Could not lock VLDB entry for %lu\n",
5714 (unsigned long)rwvolid);
5719 createentry = 0; /* Do we need to create a VLDB entry */
5720 addvolume = 0; /* Add this volume to the VLDB entry */
5721 modified = 0; /* The VLDB entry was modified */
5724 memcpy(&entry, aentry, sizeof(entry));
5726 /* Read the entry from VLDB by its RW volume id */
5727 code = VLDB_GetEntryByID(rwvolid, RWVOL, &entry);
5729 if (code != VL_NOENT) {
5731 "Could not retrieve the VLDB entry for volume %lu \n",
5732 (unsigned long)rwvolid);
5736 memset(&entry, 0, sizeof(entry));
5737 vsu_ExtractName(entry.name, volumeinfo->name); /* Store name of RW */
5741 MapHostToNetwork(&entry);
5745 if (verbose && (pass == 1)) {
5746 fprintf(STDOUT, "_______________________________\n");
5747 fprintf(STDOUT, "\n-- status before -- \n");
5749 fprintf(STDOUT, "\n**does not exist**\n");
5751 if ((entry.flags & RW_EXISTS) || (entry.flags & RO_EXISTS)
5752 || (entry.flags & BACK_EXISTS))
5753 EnumerateEntry(&entry);
5755 fprintf(STDOUT, "\n");
5758 if (volumeinfo->type == RWVOL) { /* RW volume exists */
5764 /* Check existence of RW and BK volumes */
5765 code = CheckVldbRWBK(&entry, &mod);
5771 idx = Lp_GetRwIndex(&entry);
5772 if (idx == -1) { /* RW index not found in the VLDB entry */
5773 idx = entry.nServers; /* put it into next index */
5776 } else { /* RW index found in the VLDB entry. */
5777 /* Verify if this volume's location matches where the VLDB says it is */
5778 if (!Lp_Match(aserver, apart, &entry)) {
5779 if (entry.flags & RW_EXISTS) {
5780 /* The RW volume exists elsewhere - report this one a duplicate */
5782 MapPartIdIntoName(apart, pname);
5784 "*** Warning: Orphaned RW volume %lu exists on %s %s\n",
5785 (unsigned long)rwvolid,
5787 afs_inet_ntoa_r(aserver, hoststr) :
5788 hostutil_GetNameByINet(aserver), pname);
5789 MapPartIdIntoName(entry.serverPartition[idx],
5792 " VLDB reports RW volume %lu exists on %s %s\n",
5793 (unsigned long)rwvolid,
5795 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5796 hostutil_GetNameByINet(entry.
5801 /* The RW volume does not exist - have VLDB point to this one */
5804 /* Check for orphaned BK volume on old partition */
5805 if (entry.flags & BACK_EXISTS) {
5807 MapPartIdIntoName(entry.serverPartition[idx],
5810 "*** Warning: Orphaned BK volume %u exists on %s %s\n",
5811 entry.volumeId[BACKVOL],
5813 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5814 hostutil_GetNameByINet(entry.
5817 MapPartIdIntoName(apart, pname);
5819 " VLDB reports its RW volume %lu exists on %s %s\n",
5820 (unsigned long)rwvolid,
5822 afs_inet_ntoa_r(aserver, hoststr) :
5823 hostutil_GetNameByINet(aserver),
5829 /* Volume location matches the VLDB location */
5830 if ((volumeinfo->backupID && !entry.volumeId[BACKVOL])
5831 || (volumeinfo->cloneID && !entry.volumeId[ROVOL])
5834 (entry.name, volumeinfo->name,
5835 VOLSER_OLDMAXVOLNAME) != 0)) {
5843 entry.flags |= RW_EXISTS;
5844 entry.volumeId[RWVOL] = rwvolid;
5845 if (!entry.volumeId[BACKVOL])
5846 entry.volumeId[BACKVOL] = volumeinfo->backupID;
5847 if (!entry.volumeId[ROVOL])
5848 entry.volumeId[ROVOL] = volumeinfo->cloneID;
5850 entry.serverFlags[idx] = ITSRWVOL;
5851 entry.serverNumber[idx] = aserver;
5852 entry.serverPartition[idx] = apart;
5853 strncpy(entry.name, volumeinfo->name, VOLSER_OLDMAXVOLNAME);
5857 /* One last check - to update BK if need to */
5858 code = CheckVldbRWBK(&entry, &mod);
5866 else if (volumeinfo->type == BACKVOL) { /* A BK volume */
5872 /* Check existence of RW and BK volumes */
5873 code = CheckVldbRWBK(&entry, &mod);
5879 idx = Lp_GetRwIndex(&entry);
5880 if (idx == -1) { /* RW index not found in the VLDB entry */
5881 idx = entry.nServers; /* Put it into next index */
5884 } else { /* RW index found in the VLDB entry */
5885 /* Verify if this volume's location matches where the VLDB says it is */
5886 if (!Lp_Match(aserver, apart, &entry)) {
5887 /* VLDB says RW and/or BK is elsewhere - report this BK volume orphaned */
5889 MapPartIdIntoName(apart, pname);
5891 "*** Warning: Orphaned BK volume %lu exists on %s %s\n",
5892 (unsigned long)volumeinfo->volid,
5894 afs_inet_ntoa_r(aserver, hoststr) :
5895 hostutil_GetNameByINet(aserver), pname);
5896 MapPartIdIntoName(entry.serverPartition[idx], pname);
5898 " VLDB reports its RW/BK volume %lu exists on %s %s\n",
5899 (unsigned long)rwvolid,
5901 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5902 hostutil_GetNameByINet(entry.
5907 if (volumeinfo->volid != entry.volumeId[BACKVOL]) {
5908 if (!(entry.flags & BACK_EXISTS)) {
5910 } else if (volumeinfo->volid >
5911 entry.volumeId[BACKVOL]) {
5915 MapPartIdIntoName(entry.serverPartition[idx],
5918 "*** Warning: Orphaned BK volume %u exists on %s %s\n",
5919 entry.volumeId[BACKVOL],
5921 afs_inet_ntoa_r(aserver, hoststr) :
5922 hostutil_GetNameByINet(aserver),
5925 " VLDB reports its BK volume ID is %lu\n",
5926 (unsigned long)volumeinfo->volid);
5930 MapPartIdIntoName(entry.serverPartition[idx],
5933 "*** Warning: Orphaned BK volume %lu exists on %s %s\n",
5934 (unsigned long)volumeinfo->volid,
5936 afs_inet_ntoa_r(aserver, hoststr) :
5937 hostutil_GetNameByINet(aserver),
5940 " VLDB reports its BK volume ID is %u\n",
5941 entry.volumeId[BACKVOL]);
5944 } else if (!entry.volumeId[BACKVOL]) {
5951 entry.flags |= BACK_EXISTS;
5952 entry.volumeId[RWVOL] = rwvolid;
5953 entry.volumeId[BACKVOL] = volumeinfo->volid;
5955 entry.serverNumber[idx] = aserver;
5956 entry.serverPartition[idx] = apart;
5957 entry.serverFlags[idx] = ITSBACKVOL;
5963 else if (volumeinfo->type == ROVOL) { /* A RO volume */
5964 if (volumeinfo->volid == entry.volumeId[ROVOL]) {
5965 /* This is a quick check to see if the RO entry exists in the
5966 * VLDB so we avoid the CheckVldbRO() call (which checks if each
5967 * RO volume listed in the VLDB exists).
5969 idx = Lp_ROMatch(aserver, apart, &entry) - 1;
5971 idx = entry.nServers;
5975 if (!(entry.flags & RO_EXISTS)) {
5980 /* Before we correct the VLDB entry, make sure all the
5981 * ROs listed in the VLDB exist.
5983 code = CheckVldbRO(&entry, &mod);
5989 if (!(entry.flags & RO_EXISTS)) {
5990 /* No RO exists in the VLDB entry - add this one */
5991 idx = entry.nServers;
5994 } else if (volumeinfo->volid > entry.volumeId[ROVOL]) {
5995 /* The volume headers's RO ID does not match that in the VLDB entry,
5996 * and the vol hdr's ID is greater (implies more recent). So delete
5997 * all the RO volumes listed in VLDB entry and add this volume.
5999 for (j = 0; j < entry.nServers; j++) {
6000 if (entry.serverFlags[j] & ITSROVOL) {
6001 /* Verify this volume exists and print message we are orphaning it */
6003 MapPartIdIntoName(apart, pname);
6005 "*** Warning: Orphaned RO volume %u exists on %s %s\n",
6006 entry.volumeId[ROVOL],
6008 afs_inet_ntoa_r(entry.serverNumber[j], hoststr) :
6009 hostutil_GetNameByINet(entry.
6013 " VLDB reports its RO volume ID is %lu\n",
6014 (unsigned long)volumeinfo->volid);
6017 Lp_SetRWValue(&entry, entry.serverNumber[idx],
6018 entry.serverPartition[idx], 0L, 0L);
6025 idx = entry.nServers;
6028 } else if (volumeinfo->volid < entry.volumeId[ROVOL]) {
6029 /* The volume headers's RO ID does not match that in the VLDB entry,
6030 * and the vol hdr's ID is lower (implies its older). So orphan it.
6033 MapPartIdIntoName(apart, pname);
6035 "*** Warning: Orphaned RO volume %lu exists on %s %s\n",
6036 (unsigned long)volumeinfo->volid,
6038 afs_inet_ntoa_r(aserver, hoststr) :
6039 hostutil_GetNameByINet(aserver), pname);
6041 " VLDB reports its RO volume ID is %u\n",
6042 entry.volumeId[ROVOL]);
6045 /* The RO volume ID in the volume header match that in the VLDB entry,
6046 * and there exist RO volumes in the VLDB entry. See if any of them
6047 * are this one. If not, then we add it.
6049 idx = Lp_ROMatch(aserver, apart, &entry) - 1;
6051 idx = entry.nServers;
6059 entry.flags |= RO_EXISTS;
6060 entry.volumeId[RWVOL] = rwvolid;
6061 entry.volumeId[ROVOL] = volumeinfo->volid;
6063 entry.serverNumber[idx] = aserver;
6064 entry.serverPartition[idx] = apart;
6065 entry.serverFlags[idx] = ITSROVOL;
6071 /* Remember largest volume id */
6072 if (entry.volumeId[ROVOL] > *maxvolid)
6073 *maxvolid = entry.volumeId[ROVOL];
6074 if (entry.volumeId[BACKVOL] > *maxvolid)
6075 *maxvolid = entry.volumeId[BACKVOL];
6076 if (entry.volumeId[RWVOL] > *maxvolid)
6077 *maxvolid = entry.volumeId[RWVOL];
6079 if (modified && doit) {
6080 MapNetworkToHost(&entry, &storeEntry);
6083 code = VLDB_CreateEntry(&storeEntry);
6086 "Could not create a VLDB entry for the volume %lu\n",
6087 (unsigned long)rwvolid);
6094 VLDB_ReplaceEntry(rwvolid, RWVOL, &storeEntry,
6095 LOCKREL_OPCODE | LOCKREL_AFSID |
6098 fprintf(STDERR, "Could not update entry for %lu\n",
6099 (unsigned long)rwvolid);
6103 } else if (pass == 2) {
6105 ubik_VL_ReleaseLock(cstruct, 0, rwvolid, RWVOL,
6106 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
6108 PrintError("Could not unlock VLDB entry ", code);
6112 if (modified && modentry) {
6117 memcpy(aentry, &entry, sizeof(entry));
6121 fprintf(STDOUT, "-- status after --\n");
6123 EnumerateEntry(&entry);
6125 fprintf(STDOUT, "\n**no change**\n");
6129 VPRINT("\n_______________________________\n");
6134 sortVolumes(const void *a, const void *b)
6136 volintInfo *v1 = (volintInfo *) a;
6137 volintInfo *v2 = (volintInfo *) b;
6138 afs_uint32 rwvolid1, rwvolid2;
6140 rwvolid1 = ((v1->type == RWVOL) ? v1->volid : v1->parentID);
6141 rwvolid2 = ((v2->type == RWVOL) ? v2->volid : v2->parentID);
6143 if (rwvolid1 > rwvolid2)
6144 return -1; /* lower RW id goes first */
6145 if (rwvolid1 < rwvolid2)
6148 if (v1->type == RWVOL)
6149 return -1; /* RW vols go first */
6150 if (v2->type == RWVOL)
6153 if ((v1->type == BACKVOL) && (v2->type == ROVOL))
6154 return -1; /* BK vols next */
6155 if ((v1->type == ROVOL) && (v2->type == BACKVOL))
6158 if (v1->volid < v2->volid)
6159 return 1; /* larger volids first */
6160 if (v1->volid > v2->volid)
6166 * Synchronise <aserver> <apart>(if flags = 1) <avolid>.
6167 * Synchronize an individual volume against a sever and partition.
6168 * Checks the VLDB entry (similar to syncserv) as well as checks
6169 * if the volume exists on specified servers (similar to syncvldb).
6172 UV_SyncVolume(afs_uint32 aserver, afs_int32 apart, char *avolname, int flags)
6174 struct rx_connection *aconn = 0;
6175 afs_int32 j, k, code, vcode, error = 0;
6177 afs_int32 mod, modified = 0, deleted = 0;
6178 struct nvldbentry vldbentry;
6179 afs_uint32 volumeid = 0;
6180 volEntries volumeInfo;
6181 struct partList PartList;
6183 afs_uint32 maxvolid = 0;
6185 volumeInfo.volEntries_val = (volintInfo *) 0;
6186 volumeInfo.volEntries_len = 0;
6188 /* Turn verbose logging off and do our own verbose logging */
6189 /* tverbose must be set before we call ERROR_EXIT() */
6196 if (!aserver && (flags & 1)) {
6197 /* fprintf(STDERR,"Partition option requires a server option\n"); */
6201 /* Read the VLDB entry */
6202 vcode = VLDB_GetEntryByName(avolname, &vldbentry);
6203 if (vcode && (vcode != VL_NOENT)) {
6204 fprintf(STDERR, "Could not access the VLDB for volume %s\n",
6207 } else if (!vcode) {
6208 MapHostToNetwork(&vldbentry);
6212 fprintf(STDOUT, "Processing VLDB entry %s ...\n", avolname);
6213 fprintf(STDOUT, "_______________________________\n");
6214 fprintf(STDOUT, "\n-- status before -- \n");
6216 fprintf(STDOUT, "\n**does not exist**\n");
6218 if ((vldbentry.flags & RW_EXISTS) || (vldbentry.flags & RO_EXISTS)
6219 || (vldbentry.flags & BACK_EXISTS))
6220 EnumerateEntry(&vldbentry);
6222 fprintf(STDOUT, "\n");
6225 /* Verify that all of the VLDB entries exist on the repective servers
6226 * and partitions (this does not require that avolname be a volume ID).
6227 * Equivalent to a syncserv.
6230 /* Tell CheckVldb not to update if appropriate */
6235 code = CheckVldb(&vldbentry, &mod, &deleted);
6237 fprintf(STDERR, "Could not process VLDB entry for volume %s\n",
6245 /* If aserver is given, we will search for the desired volume on it */
6247 /* Generate array of partitions on the server that we will check */
6249 code = UV_ListPartitions(aserver, &PartList, &pcnt);
6252 "Could not fetch the list of partitions from the server\n");
6256 PartList.partId[0] = apart;
6260 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6262 /* If a volume ID were given, search for it on each partition */
6263 if ((volumeid = atol(avolname))) {
6264 for (j = 0; j < pcnt; j++) {
6266 AFSVolListOneVolume(aconn, PartList.partId[j], volumeid,
6269 if (code != ENODEV) {
6270 fprintf(STDERR, "Could not query server\n");
6278 /* Found one, sync it with VLDB entry */
6280 CheckVolume(volumeInfo.volEntries_val, aserver,
6281 PartList.partId[j], &mod, &maxvolid, &vldbentry);
6288 if (volumeInfo.volEntries_val)
6289 free(volumeInfo.volEntries_val);
6290 volumeInfo.volEntries_val = (volintInfo *) 0;
6291 volumeInfo.volEntries_len = 0;
6295 /* Check to see if the RW, BK, and RO IDs exist on any
6296 * partitions. We get the volume IDs from the VLDB.
6298 for (j = 0; j < MAXTYPES; j++) { /* for RW, RO, and BK IDs */
6299 if (vldbentry.volumeId[j] == 0)
6302 for (k = 0; k < pcnt; k++) { /* For each partition */
6303 volumeInfo.volEntries_val = (volintInfo *) 0;
6304 volumeInfo.volEntries_len = 0;
6306 AFSVolListOneVolume(aconn, PartList.partId[k],
6307 vldbentry.volumeId[j], &volumeInfo);
6309 if (code != ENODEV) {
6310 fprintf(STDERR, "Could not query server\n");
6318 /* Found one, sync it with VLDB entry */
6320 CheckVolume(volumeInfo.volEntries_val, aserver,
6321 PartList.partId[k], &mod, &maxvolid, &vldbentry);
6328 if (volumeInfo.volEntries_val)
6329 free(volumeInfo.volEntries_val);
6330 volumeInfo.volEntries_val = (volintInfo *) 0;
6331 volumeInfo.volEntries_len = 0;
6337 /* If verbose output, print a summary of what changed */
6339 fprintf(STDOUT, "-- status after --\n");
6341 fprintf(STDOUT, "\n**entry deleted**\n");
6342 } else if (modified) {
6343 EnumerateEntry(&vldbentry);
6345 fprintf(STDOUT, "\n**no change**\n");
6347 fprintf(STDOUT, "\n_______________________________\n");
6351 /* Now check if the maxvolid is larger than that stored in the VLDB */
6353 afs_uint32 maxvldbid = 0;
6354 code = ubik_VL_GetNewVolumeId(cstruct, 0, 0, &maxvldbid);
6357 "Could not get the highest allocated volume id from the VLDB\n");
6360 } else if (maxvolid > maxvldbid) {
6362 id = maxvolid - maxvldbid + 1;
6363 code = ubik_VL_GetNewVolumeId(cstruct, 0, id, &nid);
6366 "Error in increasing highest allocated volume id in VLDB\n");
6376 fprintf(STDOUT, "...error encountered");
6378 fprintf(STDOUT, "...done entry\n");
6381 rx_DestroyConnection(aconn);
6382 if (volumeInfo.volEntries_val)
6383 free(volumeInfo.volEntries_val);
6385 PrintError("", error);
6390 * Synchronise vldb with the file server <aserver> and,
6391 * optionally, <apart>.
6394 UV_SyncVldb(afs_uint32 aserver, afs_int32 apart, int flags, int force)
6396 struct rx_connection *aconn;
6397 afs_int32 code, error = 0;
6400 volEntries volumeInfo;
6401 struct partList PartList;
6405 afs_int32 failures = 0, modifications = 0, tentries = 0;
6407 afs_uint32 maxvolid = 0;
6410 volumeInfo.volEntries_val = (volintInfo *) 0;
6411 volumeInfo.volEntries_len = 0;
6413 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6415 /* Generate array of partitions to check */
6417 code = UV_ListPartitions(aserver, &PartList, &pcnt);
6420 "Could not fetch the list of partitions from the server\n");
6424 PartList.partId[0] = apart;
6428 VPRINT("Processing volume entries ...\n");
6430 /* Step through the array of partitions */
6431 for (i = 0; i < pcnt; i++) {
6432 apart = PartList.partId[i];
6433 MapPartIdIntoName(apart, pname);
6435 volumeInfo.volEntries_val = (volintInfo *) 0;
6436 volumeInfo.volEntries_len = 0;
6437 code = AFSVolListVolumes(aconn, apart, 1, &volumeInfo);
6440 "Could not fetch the list of volumes from the server\n");
6444 /* May want to sort the entries: RW, BK (high to low), RO (high to low) */
6445 qsort((char *)volumeInfo.volEntries_val, volumeInfo.volEntries_len,
6446 sizeof(volintInfo), sortVolumes);
6449 for (vi = volumeInfo.volEntries_val, j = 0;
6450 j < volumeInfo.volEntries_len; j++, vi++) {
6458 "Processing volume entry %d: %s (%lu) on server %s %s...\n",
6459 j + 1, vi->name, (unsigned long)vi->volid,
6461 afs_inet_ntoa_r(aserver, hoststr) :
6462 hostutil_GetNameByINet(aserver), pname);
6470 code = CheckVolume(vi, aserver, apart, &modified, &maxvolid, NULL);
6472 PrintError("", code);
6475 } else if (modified) {
6481 fprintf(STDOUT, "...error encountered\n\n");
6483 fprintf(STDOUT, "...done entry %d\n\n", j + 1);
6490 "Could not process entries on server %s partition %s\n",
6492 afs_inet_ntoa_r(aserver, hoststr) :
6493 hostutil_GetNameByINet(aserver), pname);
6495 if (volumeInfo.volEntries_val) {
6496 free(volumeInfo.volEntries_val);
6497 volumeInfo.volEntries_val = 0;
6500 } /* thru all partitions */
6503 VPRINT3("Total entries: %u, Failed to process %d, Would change %d\n",
6504 tentries, failures, modifications);
6506 VPRINT3("Total entries: %u, Failed to process %d, Changed %d\n",
6507 tentries, failures, modifications);
6511 /* Now check if the maxvolid is larger than that stored in the VLDB */
6513 afs_uint32 maxvldbid = 0;
6514 code = ubik_VL_GetNewVolumeId(cstruct, 0, 0, &maxvldbid);
6517 "Could not get the highest allocated volume id from the VLDB\n");
6520 } else if (maxvolid > maxvldbid) {
6522 id = maxvolid - maxvldbid + 1;
6523 code = ubik_VL_GetNewVolumeId(cstruct, 0, id, &nid);
6526 "Error in increasing highest allocated volume id in VLDB\n");
6534 rx_DestroyConnection(aconn);
6535 if (volumeInfo.volEntries_val)
6536 free(volumeInfo.volEntries_val);
6537 PrintError("", error);
6542 * Determine if a volume exists on a server and partition.
6543 * Try creating a transaction on the volume. If we can,
6544 * the volume exists, if not, then return the error code.
6545 * Some error codes mean the volume is unavailable but
6546 * still exists - so we catch these error codes.
6549 VolumeExists(afs_uint32 server, afs_int32 partition, afs_uint32 volumeid)
6551 struct rx_connection *conn = (struct rx_connection *)0;
6552 afs_int32 code = -1;
6553 volEntries volumeInfo;
6555 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
6557 volumeInfo.volEntries_val = (volintInfo *) 0;
6558 volumeInfo.volEntries_len = 0;
6559 code = AFSVolListOneVolume(conn, partition, volumeid, &volumeInfo);
6560 if (volumeInfo.volEntries_val)
6561 free(volumeInfo.volEntries_val);
6562 if (code == VOLSERILLEGAL_PARTITION)
6564 rx_DestroyConnection(conn);
6573 CheckVldbRWBK(struct nvldbentry * entry, afs_int32 * modified)
6577 afs_int32 code, error = 0;
6583 idx = Lp_GetRwIndex(entry);
6585 /* Check to see if the RW volume exists and set the RW_EXISTS
6588 if (idx == -1) { /* Did not find a RW entry */
6589 if (entry->flags & RW_EXISTS) { /* ... yet entry says RW exists */
6590 entry->flags &= ~RW_EXISTS; /* ... so say RW does not exist */
6595 VolumeExists(entry->serverNumber[idx],
6596 entry->serverPartition[idx], entry->volumeId[RWVOL]);
6597 if (code == 0) { /* RW volume exists */
6598 if (!(entry->flags & RW_EXISTS)) { /* ... yet entry says RW does not exist */
6599 entry->flags |= RW_EXISTS; /* ... so say RW does exist */
6602 } else if (code == ENODEV) { /* RW volume does not exist */
6603 if (entry->flags & RW_EXISTS) { /* ... yet entry says RW exists */
6604 entry->flags &= ~RW_EXISTS; /* ... so say RW does not exist */
6608 /* If VLDB says it didn't exist, then ignore error */
6609 if (entry->flags & RW_EXISTS) {
6610 MapPartIdIntoName(entry->serverPartition[idx], pname);
6612 "Transaction call failed for RW volume %u on server %s %s\n",
6613 entry->volumeId[RWVOL],
6615 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6616 hostutil_GetNameByINet(entry->serverNumber[idx]),
6623 /* Check to see if the BK volume exists and set the BACK_EXISTS
6624 * flag accordingly. idx already ponts to the RW entry.
6626 if (idx == -1) { /* Did not find a RW entry */
6627 if (entry->flags & BACK_EXISTS) { /* ... yet entry says BK exists */
6628 entry->flags &= ~BACK_EXISTS; /* ... so say BK does not exist */
6631 } else { /* Found a RW entry */
6633 VolumeExists(entry->serverNumber[idx],
6634 entry->serverPartition[idx],
6635 entry->volumeId[BACKVOL]);
6636 if (code == 0) { /* BK volume exists */
6637 if (!(entry->flags & BACK_EXISTS)) { /* ... yet entry says BK does not exist */
6638 entry->flags |= BACK_EXISTS; /* ... so say BK does exist */
6641 } else if (code == ENODEV) { /* BK volume does not exist */
6642 if (entry->flags & BACK_EXISTS) { /* ... yet entry says BK exists */
6643 entry->flags &= ~BACK_EXISTS; /* ... so say BK does not exist */
6647 /* If VLDB says it didn't exist, then ignore error */
6648 if (entry->flags & BACK_EXISTS) {
6649 MapPartIdIntoName(entry->serverPartition[idx], pname);
6651 "Transaction call failed for BK volume %u on server %s %s\n",
6652 entry->volumeId[BACKVOL],
6654 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6655 hostutil_GetNameByINet(entry->serverNumber[idx]),
6662 /* If there is an idx but the BK and RW volumes no
6663 * longer exist, then remove the RW entry.
6665 if ((idx != -1) && !(entry->flags & RW_EXISTS)
6666 && !(entry->flags & BACK_EXISTS)) {
6667 Lp_SetRWValue(entry, entry->serverNumber[idx],
6668 entry->serverPartition[idx], 0L, 0L);
6675 *modified = modentry;
6680 CheckVldbRO(struct nvldbentry *entry, afs_int32 * modified)
6683 int foundro = 0, modentry = 0;
6684 afs_int32 code, error = 0;
6691 /* Check to see if the RO volumes exist and set the RO_EXISTS
6694 for (idx = 0; idx < entry->nServers; idx++) {
6695 if (!(entry->serverFlags[idx] & ITSROVOL)) {
6696 continue; /* not a RO */
6700 VolumeExists(entry->serverNumber[idx],
6701 entry->serverPartition[idx], entry->volumeId[ROVOL]);
6702 if (code == 0) { /* RO volume exists */
6704 } else if (code == ENODEV) { /* RW volume does not exist */
6705 Lp_SetROValue(entry, entry->serverNumber[idx],
6706 entry->serverPartition[idx], 0L, 0L);
6711 MapPartIdIntoName(entry->serverPartition[idx], pname);
6713 "Transaction call failed for RO %u on server %s %s\n",
6714 entry->volumeId[ROVOL],
6716 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6717 hostutil_GetNameByINet(entry->serverNumber[idx]), pname);
6722 if (foundro) { /* A RO volume exists */
6723 if (!(entry->flags & RO_EXISTS)) { /* ... yet entry says RW does not exist */
6724 entry->flags |= RO_EXISTS; /* ... so say RW does exist */
6727 } else { /* A RO volume does not exist */
6728 if (entry->flags & RO_EXISTS) { /* ... yet entry says RO exists */
6729 entry->flags &= ~RO_EXISTS; /* ... so say RO does not exist */
6736 *modified = modentry;
6741 * Ensure that <entry> matches with the info on file servers
6744 CheckVldb(struct nvldbentry * entry, afs_int32 * modified, afs_int32 * deleted)
6746 afs_int32 code, error = 0;
6747 struct nvldbentry storeEntry;
6748 int islocked = 0, mod, modentry, delentry = 0;
6749 int pass = 0, doit=1;
6757 fprintf(STDOUT, "_______________________________\n");
6758 fprintf(STDOUT, "\n-- status before -- \n");
6759 if ((entry->flags & RW_EXISTS) || (entry->flags & RO_EXISTS)
6760 || (entry->flags & BACK_EXISTS))
6761 EnumerateEntry(entry);
6762 fprintf(STDOUT, "\n");
6765 if (strlen(entry->name) > (VOLSER_OLDMAXVOLNAME - 10)) {
6766 fprintf(STDERR, "Volume name %s exceeds limit of %d characters\n",
6767 entry->name, VOLSER_OLDMAXVOLNAME - 10);
6771 /* Check to see if the VLDB is ok without locking it (pass 1).
6772 * If it will change, then lock the VLDB entry, read it again,
6773 * then make the changes to it (pass 2).
6777 ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL,
6780 fprintf(STDERR, "Could not lock VLDB entry for %u \n",
6781 entry->volumeId[RWVOL]);
6786 code = VLDB_GetEntryByID(entry->volumeId[RWVOL], RWVOL, entry);
6788 fprintf(STDERR, "Could not read VLDB entry for volume %s\n",
6792 MapHostToNetwork(entry);
6798 /* Check if the RW and BK entries are ok */
6799 code = CheckVldbRWBK(entry, &mod);
6802 if (mod && (pass == 1) && doit)
6807 /* Check if the RO volumes entries are ok */
6808 code = CheckVldbRO(entry, &mod);
6811 if (mod && (pass == 1) && doit)
6816 /* The VLDB entry has been updated. If it as been modified, then
6817 * write the entry back out the the VLDB.
6819 if (modentry && doit) {
6823 if (!(entry->flags & RW_EXISTS) && !(entry->flags & BACK_EXISTS)
6824 && !(entry->flags & RO_EXISTS) && doit) {
6825 /* The RW, BK, nor RO volumes do not exist. Delete the VLDB entry */
6827 ubik_VL_DeleteEntry(cstruct, 0, entry->volumeId[RWVOL],
6831 "Could not delete VLDB entry for volume %u \n",
6832 entry->volumeId[RWVOL]);
6837 /* Replace old entry with our new one */
6838 MapNetworkToHost(entry, &storeEntry);
6840 VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry,
6841 (LOCKREL_OPCODE | LOCKREL_AFSID |
6842 LOCKREL_TIMESTAMP));
6844 fprintf(STDERR, "Could not update VLDB entry for volume %u\n",
6845 entry->volumeId[RWVOL]);
6852 if (modified && modentry) {
6855 if (deleted && delentry) {
6860 fprintf(STDOUT, "-- status after --\n");
6862 fprintf(STDOUT, "\n**entry deleted**\n");
6864 EnumerateEntry(entry);
6866 fprintf(STDOUT, "\n**no change**\n");
6870 VPRINT("\n_______________________________\n");
6874 ubik_VL_ReleaseLock(cstruct, 0, entry->volumeId[RWVOL],
6876 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
6879 "Could not release lock on VLDB entry for volume %u\n",
6880 entry->volumeId[RWVOL]);
6889 * Synchronise <aserver> <apart>(if flags = 1) with the VLDB.
6892 UV_SyncServer(afs_uint32 aserver, afs_int32 apart, int flags, int force)
6894 struct rx_connection *aconn;
6895 afs_int32 code, error = 0;
6896 afs_int32 nentries, tentries = 0;
6897 struct VldbListByAttributes attributes;
6898 nbulkentries arrayEntries;
6899 afs_int32 failures = 0, modified, modifications = 0;
6900 struct nvldbentry *vlentry;
6901 afs_int32 si, nsi, j;
6906 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6908 /* Set up attributes to search VLDB */
6909 attributes.server = ntohl(aserver);
6910 attributes.Mask = VLLIST_SERVER;
6912 attributes.partition = apart;
6913 attributes.Mask |= VLLIST_PARTITION;
6916 VPRINT("Processing VLDB entries ...\n");
6918 /* While we need to collect more VLDB entries */
6919 for (si = 0; si != -1; si = nsi) {
6920 memset(&arrayEntries, 0, sizeof(arrayEntries));
6922 /* Collect set of VLDB entries */
6924 VLDB_ListAttributesN2(&attributes, 0, si, &nentries,
6925 &arrayEntries, &nsi);
6926 if (code == RXGEN_OPCODE) {
6927 code = VLDB_ListAttributes(&attributes, &nentries, &arrayEntries);
6931 fprintf(STDERR, "Could not access the VLDB for attributes\n");
6934 tentries += nentries;
6936 for (j = 0; j < nentries; j++) {
6937 vlentry = &arrayEntries.nbulkentries_val[j];
6938 MapHostToNetwork(vlentry);
6940 VPRINT1("Processing VLDB entry %d ...\n", j + 1);
6942 /* Tell CheckVldb not to update if appropriate */
6947 code = CheckVldb(vlentry, &modified, NULL);
6949 PrintError("", code);
6951 "Could not process VLDB entry for volume %s\n",
6954 } else if (modified) {
6960 fprintf(STDOUT, "...error encountered\n\n");
6962 fprintf(STDOUT, "...done entry %d\n\n", j + 1);
6967 if (arrayEntries.nbulkentries_val) {
6968 free(arrayEntries.nbulkentries_val);
6969 arrayEntries.nbulkentries_val = 0;
6974 VPRINT3("Total entries: %u, Failed to process %d, Would change %d\n",
6975 tentries, failures, modifications);
6977 VPRINT3("Total entries: %u, Failed to process %d, Changed %d\n",
6978 tentries, failures, modifications);
6983 rx_DestroyConnection(aconn);
6984 if (arrayEntries.nbulkentries_val)
6985 free(arrayEntries.nbulkentries_val);
6988 error = VOLSERFAILEDOP;
6992 /*rename volume <oldname> to <newname>, changing the names of the related
6993 *readonly and backup volumes. This operation is also idempotent.
6994 *salvager is capable of recovering from rename operation stopping halfway.
6995 *to recover run syncserver on the affected machines,it will force renaming to completion. name clashes should have been detected before calling this proc */
6997 UV_RenameVolume(struct nvldbentry *entry, char oldname[], char newname[])
6999 struct nvldbentry storeEntry;
7000 afs_int32 vcode, code, rcode, error;
7002 char nameBuffer[256];
7004 struct rx_connection *aconn;
7009 aconn = (struct rx_connection *)0;
7013 vcode = ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL, VLOP_ADDSITE); /*last param is dummy */
7016 " Could not lock the VLDB entry for the volume %u \n",
7017 entry->volumeId[RWVOL]);
7022 strncpy(entry->name, newname, VOLSER_OLDMAXVOLNAME);
7023 MapNetworkToHost(entry, &storeEntry);
7024 vcode = VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry, 0);
7026 fprintf(STDERR, "Could not update VLDB entry for %u\n",
7027 entry->volumeId[RWVOL]);
7031 VPRINT1("Recorded the new name %s in VLDB\n", newname);
7032 /*at this stage the intent to rename is recorded in the vldb, as far as the vldb
7033 * is concerned, oldname is lost */
7034 if (entry->flags & RW_EXISTS) {
7035 index = Lp_GetRwIndex(entry);
7036 if (index == -1) { /* there is a serious discrepancy */
7038 "There is a serious discrepancy in VLDB entry for volume %u\n",
7039 entry->volumeId[RWVOL]);
7040 fprintf(STDERR, "try building VLDB from scratch\n");
7041 error = VOLSERVLDB_ERROR;
7044 aconn = UV_Bind(entry->serverNumber[index], AFSCONF_VOLUMEPORT);
7046 AFSVolTransCreate_retry(aconn, entry->volumeId[RWVOL],
7047 entry->serverPartition[index], ITOffline, &tid);
7048 if (code) { /*volume doesnot exist */
7050 "Could not start transaction on the rw volume %u\n",
7051 entry->volumeId[RWVOL]);
7054 } else { /*volume exists, process it */
7057 AFSVolSetIdsTypes(aconn, tid, newname, RWVOL,
7058 entry->volumeId[RWVOL],
7059 entry->volumeId[ROVOL],
7060 entry->volumeId[BACKVOL]);
7062 VPRINT2("Renamed rw volume %s to %s\n", oldname, newname);
7063 code = AFSVolEndTrans(aconn, tid, &rcode);
7067 "Could not end transaction on volume %s %u\n",
7068 entry->name, entry->volumeId[RWVOL]);
7073 fprintf(STDERR, "Could not set parameters on volume %s %u\n",
7074 entry->name, entry->volumeId[RWVOL]);
7080 rx_DestroyConnection(aconn);
7081 aconn = (struct rx_connection *)0;
7083 /*end rw volume processing */
7084 if (entry->flags & BACK_EXISTS) { /*process the backup volume */
7085 index = Lp_GetRwIndex(entry);
7086 if (index == -1) { /* there is a serious discrepancy */
7088 "There is a serious discrepancy in the VLDB entry for the backup volume %u\n",
7089 entry->volumeId[BACKVOL]);
7090 fprintf(STDERR, "try building VLDB from scratch\n");
7091 error = VOLSERVLDB_ERROR;
7094 aconn = UV_Bind(entry->serverNumber[index], AFSCONF_VOLUMEPORT);
7096 AFSVolTransCreate_retry(aconn, entry->volumeId[BACKVOL],
7097 entry->serverPartition[index], ITOffline, &tid);
7098 if (code) { /*volume doesnot exist */
7100 "Could not start transaction on the backup volume %u\n",
7101 entry->volumeId[BACKVOL]);
7104 } else { /*volume exists, process it */
7105 if (strlen(newname) > (VOLSER_OLDMAXVOLNAME - 8)) {
7107 "Volume name %s.backup exceeds the limit of %u characters\n",
7108 newname, VOLSER_OLDMAXVOLNAME);
7112 strcpy(nameBuffer, newname);
7113 strcat(nameBuffer, ".backup");
7116 AFSVolSetIdsTypes(aconn, tid, nameBuffer, BACKVOL,
7117 entry->volumeId[RWVOL], 0, 0);
7119 VPRINT1("Renamed backup volume to %s \n", nameBuffer);
7120 code = AFSVolEndTrans(aconn, tid, &rcode);
7124 "Could not end transaction on the backup volume %u\n",
7125 entry->volumeId[BACKVOL]);
7131 "Could not set parameters on the backup volume %u\n",
7132 entry->volumeId[BACKVOL]);
7137 } /* end backup processing */
7139 rx_DestroyConnection(aconn);
7140 aconn = (struct rx_connection *)0;
7141 if (entry->flags & RO_EXISTS) { /*process the ro volumes */
7142 for (i = 0; i < entry->nServers; i++) {
7143 if (entry->serverFlags[i] & ITSROVOL) {
7144 aconn = UV_Bind(entry->serverNumber[i], AFSCONF_VOLUMEPORT);
7146 AFSVolTransCreate_retry(aconn, entry->volumeId[ROVOL],
7147 entry->serverPartition[i], ITOffline,
7149 if (code) { /*volume doesnot exist */
7151 "Could not start transaction on the ro volume %u\n",
7152 entry->volumeId[ROVOL]);
7155 } else { /*volume exists, process it */
7156 strcpy(nameBuffer, newname);
7157 strcat(nameBuffer, ".readonly");
7158 if (strlen(nameBuffer) > (VOLSER_OLDMAXVOLNAME - 1)) {
7160 "Volume name %s exceeds the limit of %u characters\n",
7161 nameBuffer, VOLSER_OLDMAXVOLNAME);
7166 AFSVolSetIdsTypes(aconn, tid, nameBuffer, ROVOL,
7167 entry->volumeId[RWVOL], 0, 0);
7169 VPRINT2("Renamed RO volume %s on host %s\n",
7172 afs_inet_ntoa_r(entry->serverNumber[i], hoststr) :
7173 hostutil_GetNameByINet(entry->
7175 code = AFSVolEndTrans(aconn, tid, &rcode);
7179 "Could not end transaction on volume %u\n",
7180 entry->volumeId[ROVOL]);
7186 "Could not set parameters on the ro volume %u\n",
7187 entry->volumeId[ROVOL]);
7193 rx_DestroyConnection(aconn);
7194 aconn = (struct rx_connection *)0;
7201 ubik_VL_ReleaseLock(cstruct, 0, entry->volumeId[RWVOL],
7203 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
7206 "Could not unlock the VLDB entry for the volume %s %u\n",
7207 entry->name, entry->volumeId[RWVOL]);
7213 code = AFSVolEndTrans(aconn, tid, &rcode);
7217 fprintf(STDERR, "Failed to end transaction on a volume \n");
7223 rx_DestroyConnection(aconn);
7224 PrintError("", error);
7229 /*report on all the active transactions on volser */
7231 UV_VolserStatus(afs_uint32 server, transDebugInfo ** rpntr, afs_int32 * rcount)
7233 struct rx_connection *aconn;
7234 transDebugEntries transInfo;
7237 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7238 transInfo.transDebugEntries_val = (transDebugInfo *) 0;
7239 transInfo.transDebugEntries_len = 0;
7240 code = AFSVolMonitor(aconn, &transInfo);
7243 "Could not access status information about the server\n");
7244 PrintError("", code);
7245 if (transInfo.transDebugEntries_val)
7246 free(transInfo.transDebugEntries_val);
7248 rx_DestroyConnection(aconn);
7251 *rcount = transInfo.transDebugEntries_len;
7252 *rpntr = transInfo.transDebugEntries_val;
7254 rx_DestroyConnection(aconn);
7261 /*delete the volume without interacting with the vldb */
7263 UV_VolumeZap(afs_uint32 server, afs_int32 part, afs_uint32 volid)
7266 struct rx_connection *aconn;
7268 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7269 error = DoVolDelete(aconn, volid, part,
7270 "the", 0, NULL, NULL);
7272 PrintError("", error);
7274 rx_DestroyConnection(aconn);
7279 UV_SetVolume(afs_uint32 server, afs_int32 partition, afs_uint32 volid,
7280 afs_int32 transflag, afs_int32 setflag, int sleeptime)
7282 struct rx_connection *conn = 0;
7284 afs_int32 code, error = 0, rcode;
7286 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7288 fprintf(STDERR, "SetVolumeStatus: Bind Failed");
7292 code = AFSVolTransCreate_retry(conn, volid, partition, transflag, &tid);
7294 fprintf(STDERR, "SetVolumeStatus: TransCreate Failed\n");
7298 code = AFSVolSetFlags(conn, tid, setflag);
7300 fprintf(STDERR, "SetVolumeStatus: SetFlags Failed\n");
7305 #ifdef AFS_PTHREAD_ENV
7308 IOMGR_Sleep(sleeptime);
7315 code = AFSVolEndTrans(conn, tid, &rcode);
7316 if (code || rcode) {
7317 fprintf(STDERR, "SetVolumeStatus: EndTrans Failed\n");
7319 error = (code ? code : rcode);
7324 rx_DestroyConnection(conn);
7329 UV_SetVolumeInfo(afs_uint32 server, afs_int32 partition, afs_uint32 volid,
7332 struct rx_connection *conn = 0;
7334 afs_int32 code, error = 0, rcode;
7336 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7338 fprintf(STDERR, "SetVolumeInfo: Bind Failed");
7342 code = AFSVolTransCreate_retry(conn, volid, partition, ITOffline, &tid);
7344 fprintf(STDERR, "SetVolumeInfo: TransCreate Failed\n");
7348 code = AFSVolSetInfo(conn, tid, infop);
7350 fprintf(STDERR, "SetVolumeInfo: SetInfo Failed\n");
7357 code = AFSVolEndTrans(conn, tid, &rcode);
7358 if (code || rcode) {
7359 fprintf(STDERR, "SetVolumeInfo: EndTrans Failed\n");
7361 error = (code ? code : rcode);
7366 rx_DestroyConnection(conn);
7371 UV_GetSize(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
7372 afs_int32 fromdate, struct volintSize *vol_size)
7374 struct rx_connection *aconn = (struct rx_connection *)0;
7375 afs_int32 tid = 0, rcode = 0;
7376 afs_int32 code, error = 0;
7379 /* get connections to the servers */
7380 aconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
7382 VPRINT1("Starting transaction on volume %u...", afromvol);
7383 code = AFSVolTransCreate_retry(aconn, afromvol, afrompart, ITBusy, &tid);
7384 EGOTO1(error_exit, code,
7385 "Could not start transaction on the volume %u to be measured\n",
7389 VPRINT1("Getting size of volume on volume %u...", afromvol);
7390 code = AFSVolGetSize(aconn, tid, fromdate, vol_size);
7391 EGOTO(error_exit, code, "Could not start the measurement process \n");
7396 VPRINT1("Ending transaction on volume %u...", afromvol);
7397 code = AFSVolEndTrans(aconn, tid, &rcode);
7398 if (code || rcode) {
7399 fprintf(STDERR, "Could not end transaction on the volume %u\n",
7401 fprintf(STDERR, "error codes: %d and %d\n", code, rcode);
7403 error = (code ? code : rcode);
7408 rx_DestroyConnection(aconn);
7410 PrintError("", error);
7414 /*maps the host addresses in <old > (present in network byte order) to
7415 that in< new> (present in host byte order )*/
7417 MapNetworkToHost(struct nvldbentry *old, struct nvldbentry *new)
7421 /*copy all the fields */
7422 strcpy(new->name, old->name);
7423 /* new->volumeType = old->volumeType;*/
7424 new->nServers = old->nServers;
7425 count = old->nServers;
7426 if (count < NMAXNSERVERS)
7428 for (i = 0; i < count; i++) {
7429 new->serverNumber[i] = ntohl(old->serverNumber[i]);
7430 new->serverPartition[i] = old->serverPartition[i];
7431 new->serverFlags[i] = old->serverFlags[i];
7433 new->volumeId[RWVOL] = old->volumeId[RWVOL];
7434 new->volumeId[ROVOL] = old->volumeId[ROVOL];
7435 new->volumeId[BACKVOL] = old->volumeId[BACKVOL];
7436 new->cloneId = old->cloneId;
7437 new->flags = old->flags;
7440 /*maps the host entries in <entry> which are present in host byte order to network byte order */
7442 MapHostToNetwork(struct nvldbentry *entry)
7446 count = entry->nServers;
7447 if (count < NMAXNSERVERS)
7449 for (i = 0; i < count; i++) {
7450 entry->serverNumber[i] = htonl(entry->serverNumber[i]);