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 /* return early and quietly for VNOVOL; don't continue the attempt to delete. */
1175 if (code == VNOVOL) {
1180 EGOTO2(dfail, code, "%sFailed to start transaction on %u\n",
1184 code = AFSVolGetStatus(aconn, ttid, volstatus);
1185 EGOTO2(dfail, code, "%sCould not get timestamp from volume %u\n",
1190 AFSVolSetFlags(aconn, ttid,
1191 VTDeleteOnSalvage | VTOutOfService);
1193 EGOTO2(dfail, code, "%sCould not set flags on volume %u \n",
1197 VPRINT1("%sSetting volume forwarding pointer ...", prefix);
1198 AFSVolSetForwarding(aconn, ttid, atoserver);
1202 code = AFSVolDeleteVolume(aconn, ttid);
1203 EGOTO2(dfail, code, "%sCould not delete volume %u\n", prefix, avolid);
1207 code = AFSVolEndTrans(aconn, ttid, &rcode);
1212 fprintf(STDERR, "%sCould not end transaction on %s volume %lu \n",
1213 prefix, typestring, (unsigned long)avolid);
1219 if (beverbose && !error)
1225 DoVolClone(struct rx_connection *aconn, afs_uint32 avolid,
1226 afs_int32 apart, int type, afs_uint32 cloneid,
1227 char *typestring, char *pname, char *vname, char *suffix,
1228 struct volser_status *volstatus, afs_int32 *transPtr)
1231 afs_int32 ttid = 0, btid = 0;
1232 afs_int32 code = 0, rcode = 0;
1233 afs_int32 error = 0;
1234 int cloneexists = 1;
1236 /* Test to see if the clone volume exists by trying to create
1237 * a transaction on the clone volume. We've assumed the clone exists.
1239 code = AFSVolTransCreate_retry(aconn, cloneid, apart, ITOffline, &btid);
1241 if (code != VNOVOL) {
1242 EPRINT2(code, "Could not reach the %s volume %lu\n",
1243 typestring, (unsigned long)cloneid);
1247 cloneexists = 0; /* clone volume does not exist */
1250 code = AFSVolEndTrans(aconn, btid, &rcode);
1252 if (code || rcode) {
1254 "Could not end transaction on the previous %s volume %lu\n",
1255 typestring, (unsigned long)cloneid);
1256 error = (code ? code : rcode);
1261 /* Now go ahead and try to clone the RW volume.
1262 * First start a transaction on the RW volume
1264 code = AFSVolTransCreate_retry(aconn, avolid, apart, ITBusy, &ttid);
1266 fprintf(STDERR, "Could not start a transaction on the volume %lu\n",
1267 (unsigned long)avolid);
1272 /* Clone or reclone the volume, depending on whether the clone
1273 * volume exists or not
1276 VPRINT2("Re-cloning %s volume %u ...", typestring, cloneid);
1278 code = AFSVolReClone(aconn, ttid, cloneid);
1280 EPRINT2(code, "Could not re-clone %s volume %lu\n",
1281 typestring, (unsigned long)cloneid);
1286 VPRINT2("Creating a new %s clone %u ...", typestring, cloneid);
1289 strcpy(cname, pname);
1290 strcat(cname, suffix);
1293 code = AFSVolClone(aconn, ttid, 0, type, vname?vname:cname,
1296 fprintf(STDERR, "Failed to clone the volume %lu\n",
1297 (unsigned long)avolid);
1306 VPRINT1("Getting status of parent volume %u...", avolid);
1307 code = AFSVolGetStatus(aconn, ttid, volstatus);
1309 fprintf(STDERR, "Failed to get the status of the parent volume %lu\n",
1310 (unsigned long)avolid);
1319 code = AFSVolEndTrans(aconn, ttid, &rcode);
1320 if (code || rcode) {
1321 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
1322 (unsigned long)avolid);
1324 error = (code ? code : rcode);
1329 code = AFSVolEndTrans(aconn, btid, &rcode);
1330 if (code || rcode) {
1332 "Could not end transaction on the %s volume %lu\n",
1333 typestring, (unsigned long)cloneid);
1335 error = (code ? code : rcode);
1341 /* Move volume <afromvol> on <afromserver> <afrompart> to <atoserver>
1342 * <atopart>. The operation is almost idempotent. The following
1343 * flags are recognized:
1345 * RV_NOCLONE - don't use a copy clone
1349 UV_MoveVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
1350 afs_uint32 atoserver, afs_int32 atopart, int flags)
1352 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
1353 * be changing during the move */
1354 struct rx_connection * volatile toconn;
1355 struct rx_connection * volatile fromconn;
1356 afs_int32 volatile fromtid;
1357 afs_int32 volatile totid;
1358 afs_int32 volatile clonetid;
1359 afs_uint32 volatile newVol;
1360 afs_uint32 volatile volid;
1361 afs_uint32 volatile backupId;
1362 int volatile islocked;
1367 char tmpName[VOLSER_MAXVOLNAME + 1];
1372 struct restoreCookie cookie;
1373 afs_int32 vcode, code;
1374 struct volser_status tstatus;
1375 struct destServer destination;
1377 struct nvldbentry entry, storeEntry;
1380 char in, lf; /* for test code */
1384 #ifdef ENABLE_BUGFIX_1165
1385 volEntries volumeInfo;
1386 struct volintInfo *infop = 0;
1390 fromconn = (struct rx_connection *)0;
1391 toconn = (struct rx_connection *)0;
1401 /* support control-c processing */
1404 (void)signal(SIGINT, sigint_handler);
1408 "\nThere are three tests points - verifies all code paths through recovery.\n");
1409 fprintf(STDOUT, "First test point - operation not started.\n");
1410 fprintf(STDOUT, "...test here (y, n)? ");
1412 fscanf(stdin, "%c", &in);
1413 fscanf(stdin, "%c", &lf); /* toss away */
1415 fprintf(STDOUT, "type control-c\n");
1417 fprintf(stdout, ".");
1422 /* or drop through */
1425 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
1426 EGOTO1(mfail, vcode,
1427 "Could not fetch the entry for the volume %u from the VLDB \n",
1430 if (entry.volumeId[RWVOL] != afromvol) {
1431 fprintf(STDERR, "Only RW volume can be moved\n");
1435 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_MOVE);
1436 EGOTO1(mfail, vcode, "Could not lock entry for volume %u \n", afromvol);
1439 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
1440 EGOTO1(mfail, vcode,
1441 "Could not fetch the entry for the volume %u from the VLDB \n",
1444 backupId = entry.volumeId[BACKVOL];
1445 MapHostToNetwork(&entry);
1447 if (!Lp_Match(afromserver, afrompart, &entry)) {
1448 /* the from server and partition do not exist in the vldb entry corresponding to volid */
1449 if (!Lp_Match(atoserver, atopart, &entry)) {
1450 /* the to server and partition do not exist in the vldb entry corresponding to volid */
1451 fprintf(STDERR, "The volume %lu is not on the specified site. \n",
1452 (unsigned long)afromvol);
1453 fprintf(STDERR, "The current site is :");
1454 for (i = 0; i < entry.nServers; i++) {
1455 if (entry.serverFlags[i] == ITSRWVOL) {
1457 MapPartIdIntoName(entry.serverPartition[i], pname);
1458 fprintf(STDERR, " server %s partition %s \n",
1459 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
1460 hostutil_GetNameByINet(entry.serverNumber[i]),
1465 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1466 (LOCKREL_OPCODE | LOCKREL_AFSID |
1467 LOCKREL_TIMESTAMP));
1468 EGOTO1(mfail, vcode,
1469 " Could not release lock on the VLDB entry for the volume %u \n",
1472 return VOLSERVOLMOVED;
1475 /* delete the volume afromvol on src_server */
1476 /* from-info does not exist but to-info does =>
1477 * we have already done the move, but the volume
1478 * may still be existing physically on from fileserver
1480 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1483 code = DoVolDelete(fromconn, afromvol, afrompart,
1484 "leftover", 0, NULL, NULL);
1485 if (code && code != VNOVOL) {
1490 code = DoVolDelete(fromconn, backupId, afrompart,
1491 "leftover backup", 0, NULL, NULL);
1492 if (code && code != VNOVOL) {
1502 /* From-info matches the vldb info about volid,
1503 * its ok start the move operation, the backup volume
1504 * on the old site is deleted in the process
1506 if (afrompart == atopart) {
1507 same = VLDB_IsSameAddrs(afromserver, atoserver, &error);
1508 EGOTO2(mfail, error,
1509 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
1510 afromserver, error);
1513 EGOTO1(mfail, VOLSERVOLMOVED,
1514 "Warning: Moving volume %u to its home partition ignored!\n",
1520 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
1521 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1522 totid = 0; /* initialize to uncreated */
1525 * clone the read/write volume locally.
1528 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1529 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1531 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
1535 if (!(flags & RV_NOCLONE)) {
1536 /* Get a clone id */
1537 VPRINT1("Allocating new volume id for clone of volume %u ...",
1540 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
1542 EGOTO1(mfail, vcode,
1543 "Could not get an ID for the clone of volume %u from the VLDB\n",
1547 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
1548 VPRINT1("Cloning source volume %u ...", afromvol);
1549 strcpy(vname, "move-clone-temp");
1551 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname, &tmpVol);
1553 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
1558 /* lookup the name of the volume we just cloned */
1560 code = AFSVolGetName(fromconn, fromtid, &volName);
1561 EGOTO1(mfail, code, "Failed to get the name of the volume %u\n",
1564 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
1566 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1571 "Failed to end the transaction on the source volume %u\n",
1576 * Create the destination volume
1579 if (!(flags & RV_NOCLONE)) {
1580 /* All of this is to get the fromDate */
1581 VPRINT1("Starting transaction on the cloned volume %u ...", newVol);
1584 AFSVolTransCreate_retry(fromconn, newVol, afrompart, ITOffline,
1588 "Failed to start a transaction on the cloned volume%u\n",
1592 VPRINT1("Setting flags on cloned volume %u ...", newVol);
1594 AFSVolSetFlags(fromconn, clonetid,
1595 VTDeleteOnSalvage | VTOutOfService); /*redundant */
1596 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
1600 /* remember time from which we've dumped the volume */
1601 VPRINT1("Getting status of cloned volume %u ...", newVol);
1602 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
1604 "Failed to get the status of the cloned volume %u\n",
1608 fromDate = CLOCKADJ(tstatus.creationDate);
1610 /* With RV_NOCLONE, just do a full copy from the source */
1615 #ifdef ENABLE_BUGFIX_1165
1617 * Get the internal volume state from the source volume. We'll use such info (i.e. dayUse)
1618 * to copy it to the new volume (via AFSSetInfo later on) so that when we move volumes we
1619 * don't use this information...
1621 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
1622 volumeInfo.volEntries_len = 0;
1623 code = AFSVolListOneVolume(fromconn, afrompart, afromvol, &volumeInfo);
1625 "Failed to get the volint Info of the cloned volume %u\n",
1628 infop = (volintInfo *) volumeInfo.volEntries_val;
1629 infop->maxquota = -1; /* Else it will replace the default quota */
1630 infop->creationDate = -1; /* Else it will use the source creation date */
1631 infop->updateDate = -1; /* Else it will use the source update date */
1634 /* create a volume on the target machine */
1636 code = DoVolDelete(toconn, volid, atopart,
1637 "pre-existing destination", 0, NULL, NULL);
1638 if (code && code != VNOVOL) {
1643 VPRINT1("Creating the destination volume %u ...", volid);
1647 AFSVolCreateVolume(toconn, atopart, volName, volser_RW, volid, &tmpVol,
1651 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
1655 strncpy(tmpName, volName, VOLSER_OLDMAXVOLNAME);
1659 VPRINT1("Setting volume flags on destination volume %u ...", volid);
1661 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
1663 "Failed to set the flags on the destination volume %u\n", volid);
1667 * Now dump the clone to the new volume
1670 destination.destHost = ntohl(atoserver);
1671 destination.destPort = AFSCONF_VOLUMEPORT;
1672 destination.destSSID = 1;
1674 strncpy(cookie.name, tmpName, VOLSER_OLDMAXVOLNAME);
1675 cookie.type = RWVOL;
1676 cookie.parent = entry.volumeId[RWVOL];
1679 if (!(flags & RV_NOCLONE)) {
1680 /* Copy the clone to the new volume */
1681 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
1684 AFSVolForward(fromconn, clonetid, 0, &destination, totid,
1686 EGOTO1(mfail, code, "Failed to move data for the volume %u\n", volid);
1689 VPRINT1("Ending transaction on cloned volume %u ...", newVol);
1690 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
1695 "Failed to end the transaction on the cloned volume %u\n",
1701 * reattach to the main-line volume, and incrementally dump it.
1704 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1706 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1709 "Failed to create a transaction on the source volume %u\n",
1713 /* now do the incremental */
1715 ("Doing the%s dump from source to destination for volume %u ... ",
1716 (flags & RV_NOCLONE) ? "" : " incremental",
1719 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
1722 "Failed to do the%s dump from rw volume on old site to rw volume on newsite\n",
1723 (flags & RV_NOCLONE) ? "" : " incremental");
1726 /* now adjust the flags so that the new volume becomes official */
1727 VPRINT1("Setting volume flags on old source volume %u ...", afromvol);
1728 code = AFSVolSetFlags(fromconn, fromtid, VTOutOfService);
1730 "Failed to set the flags to make old source volume offline\n");
1733 VPRINT1("Setting volume flags on new source volume %u ...", afromvol);
1734 code = AFSVolSetFlags(toconn, totid, 0);
1736 "Failed to set the flags to make new source volume online\n");
1739 #ifdef ENABLE_BUGFIX_1165
1740 VPRINT1("Setting volume status on destination volume %u ...", volid);
1741 code = AFSVolSetInfo(toconn, totid, infop);
1743 "Failed to set volume status on the destination volume %u\n",
1748 /* put new volume online */
1749 VPRINT1("Ending transaction on destination volume %u ...", afromvol);
1750 code = AFSVolEndTrans(toconn, totid, &rcode);
1755 "Failed to end the transaction on the volume %u on the new site\n",
1759 Lp_SetRWValue(&entry, afromserver, afrompart, atoserver, atopart);
1760 MapNetworkToHost(&entry, &storeEntry);
1761 storeEntry.flags &= ~BACK_EXISTS;
1765 "Second test point - operation in progress but not complete.\n");
1766 fprintf(STDOUT, "...test here (y, n)? ");
1768 fscanf(stdin, "%c", &in);
1769 fscanf(stdin, "%c", &lf); /* toss away */
1771 fprintf(STDOUT, "type control-c\n");
1773 fprintf(stdout, ".");
1778 /* or drop through */
1781 VPRINT1("Releasing lock on VLDB entry for volume %u ...", afromvol);
1783 VLDB_ReplaceEntry(afromvol, -1, &storeEntry,
1784 (LOCKREL_OPCODE | LOCKREL_AFSID |
1785 LOCKREL_TIMESTAMP));
1788 " Could not release the lock on the VLDB entry for the volume %s %lu \n",
1789 storeEntry.name, (unsigned long)afromvol);
1798 "Third test point - operation complete but no cleanup.\n");
1799 fprintf(STDOUT, "...test here (y, n)? ");
1801 fscanf(stdin, "%c", &in);
1802 fscanf(stdin, "%c", &lf); /* toss away */
1804 fprintf(STDOUT, "type control-c\n");
1806 fprintf(stdout, ".");
1811 /* or drop through */
1814 /* This is tricky. File server is very stupid, and if you mark the volume
1815 * as VTOutOfService, it may mark the *good* instance (if you're moving
1816 * between partitions on the same machine) as out of service. Since
1817 * we're cleaning this code up in DEcorum, we're just going to kludge around
1818 * it for now by removing this call. */
1819 /* already out of service, just zap it now */
1821 AFSVolSetFlags(fromconn, fromtid, VTDeleteOnSalvage | VTOutOfService);
1824 "Failed to set the flags to make the old source volume offline\n");
1828 if (atoserver != afromserver) {
1829 /* set forwarding pointer for moved volumes */
1830 VPRINT1("Setting forwarding pointer for volume %u ...", afromvol);
1831 code = AFSVolSetForwarding(fromconn, fromtid, atoserver);
1833 "Failed to set the forwarding pointer for the volume %u\n",
1838 VPRINT1("Deleting old volume %u on source ...", afromvol);
1839 code = AFSVolDeleteVolume(fromconn, fromtid); /* zap original volume */
1840 EGOTO1(mfail, code, "Failed to delete the old volume %u on source\n",
1844 VPRINT1("Ending transaction on old volume %u on the source ...",
1846 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1851 "Failed to end the transaction on the old volume %u on the source\n",
1855 code = DoVolDelete(fromconn, backupId, afrompart,
1856 "source backup", 0, NULL, NULL);
1857 if (code && code != VNOVOL) {
1862 code = 0; /* no backup volume? that's okay */
1865 if (!(flags & RV_NOCLONE)) {
1866 code = DoVolDelete(fromconn, newVol, afrompart,
1867 "cloned", 0, NULL, NULL);
1878 fprintf(STDOUT, "Fourth test point - operation complete.\n");
1879 fprintf(STDOUT, "...test here (y, n)? ");
1881 fscanf(stdin, "%c", &in);
1882 fscanf(stdin, "%c", &lf); /* toss away */
1884 fprintf(STDOUT, "type control-c\n");
1886 fprintf(stdout, ".");
1891 /* or drop through */
1894 /* normal cleanup code */
1896 if (entry.flags & RO_EXISTS)
1897 fprintf(STDERR, "WARNING : readOnly copies still exist \n");
1900 VPRINT1("Cleanup: Releasing VLDB lock on volume %u ...", afromvol);
1902 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1903 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
1907 " Could not release the lock on the VLDB entry for the volume %lu \n",
1908 (unsigned long)afromvol);
1916 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
1918 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1919 if (code || rcode) {
1922 "Could not end transaction on the source volume %lu\n",
1923 (unsigned long)afromvol);
1925 error = (code ? code : rcode);
1931 VPRINT1("Cleanup: Ending transaction on clone volume %u ...", newVol);
1932 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
1933 if (code || rcode) {
1936 "Could not end transaction on the source's clone volume %lu\n",
1937 (unsigned long)newVol);
1939 error = (code ? code : rcode);
1945 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
1947 code = AFSVolEndTrans(toconn, totid, &rcode);
1951 "Could not end transaction on destination volume %lu\n",
1952 (unsigned long)afromvol);
1954 error = (code ? code : rcode);
1960 #ifdef ENABLE_BUGFIX_1165
1965 rx_DestroyConnection(fromconn);
1967 rx_DestroyConnection(toconn);
1968 PrintError("", error);
1971 /* come here only when the sky falls */
1976 "vos move: operation interrupted, cleanup in progress...\n");
1977 fprintf(STDOUT, "clear transaction contexts\n");
1981 /* unlock VLDB entry */
1983 VPRINT1("Recovery: Releasing VLDB lock on volume %u ...", afromvol);
1984 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1985 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
1991 VPRINT("Recovery: Ending transaction on clone volume ...");
1992 AFSVolEndTrans(fromconn, clonetid, &rcode);
1996 VPRINT("Recovery: Ending transaction on destination volume ...");
1997 AFSVolEndTrans(toconn, totid, &rcode);
2000 if (fromtid) { /* put it on-line */
2001 VPRINT("Recovery: Setting volume flags on source volume ...");
2002 AFSVolSetFlags(fromconn, fromtid, 0);
2005 VPRINT("Recovery: Ending transaction on source volume ...");
2006 AFSVolEndTrans(fromconn, fromtid, &rcode);
2010 VPRINT("Recovery: Accessing VLDB.\n");
2011 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2013 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2017 MapHostToNetwork(&entry);
2019 /* Delete either the volume on the source location or the target location.
2020 * If the vldb entry still points to the source location, then we know the
2021 * volume move didn't finish so we remove the volume from the target
2022 * location. Otherwise, we remove the volume from the source location.
2024 if (Lp_Match(afromserver, afrompart, &entry)) { /* didn't move - delete target volume */
2027 "move incomplete - attempt cleanup of target partition - no guarantee\n");
2031 if (volid && toconn)
2032 DoVolDelete(toconn, volid, atopart,
2033 "destination", 0, NULL, "Recovery:");
2035 /* put source volume on-line */
2037 VPRINT1("Recovery: Creating transaction on source volume %u ...",
2041 AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2047 VPRINT1("Recovery: Setting flags on source volume %u ...",
2049 AFSVolSetFlags(fromconn, fromtid, 0);
2053 ("Recovery: Ending transaction on source volume %u ...",
2055 AFSVolEndTrans(fromconn, fromtid, &rcode);
2059 ("\nRecovery: Unable to start transaction on source volume %u.\n",
2063 } else { /* yep, move complete */
2066 "move complete - attempt cleanup of source partition - no guarantee\n");
2070 /* delete backup volume */
2072 DoVolDelete(fromconn, backupId, afrompart,
2073 "backup", 0, NULL, "Recovery:");
2075 DoVolDelete(fromconn, afromvol, afrompart, "source",
2076 (atoserver != afromserver)?atoserver:0,
2081 /* common cleanup - delete local clone */
2083 DoVolDelete(fromconn, newVol, afrompart,
2084 "clone", 0, NULL, "Recovery:");
2086 /* unlock VLDB entry */
2088 VPRINT1("Recovery: Releasing lock on VLDB entry for volume %u ...",
2090 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2091 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2094 done: /* routine cleanup */
2097 #ifdef ENABLE_BUGFIX_1165
2102 rx_DestroyConnection(fromconn);
2104 rx_DestroyConnection(toconn);
2107 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2115 UV_MoveVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2116 afs_uint32 atoserver, afs_int32 atopart)
2118 return UV_MoveVolume2(afromvol, afromserver, afrompart,
2119 atoserver, atopart, 0);
2123 /* Copy volume <afromvol> from <afromserver> <afrompart> to <atoserver>
2124 * <atopart>. The new volume is named by <atovolname>. The new volume
2125 * has ID <atovolid> if that is nonzero; otherwise a new ID is allocated
2126 * from the VLDB. the following flags are supported:
2128 * RV_RDONLY - target volume is RO
2129 * RV_OFFLINE - leave target volume offline
2130 * RV_CPINCR - do incremental dump if target exists
2131 * RV_NOVLDB - don't create/update VLDB entry
2132 * RV_NOCLONE - don't use a copy clone
2135 UV_CopyVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2136 char *atovolname, afs_uint32 atoserver, afs_int32 atopart,
2137 afs_uint32 atovolid, int flags)
2139 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
2140 * be changing during the copy */
2142 afs_int32 volatile clonetid;
2143 afs_int32 volatile totid;
2144 afs_int32 volatile fromtid;
2145 struct rx_connection * volatile fromconn;
2146 struct rx_connection * volatile toconn;
2147 afs_uint32 volatile cloneVol;
2151 afs_int32 fromDate, cloneFromDate;
2152 struct restoreCookie cookie;
2153 afs_int32 vcode, code;
2156 struct volser_status tstatus;
2157 struct destServer destination;
2158 struct nvldbentry entry, newentry, storeEntry;
2163 fromconn = (struct rx_connection *)0;
2164 toconn = (struct rx_connection *)0;
2172 /* support control-c processing */
2175 (void)signal(SIGINT, sigint_handler);
2177 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2178 EGOTO1(mfail, vcode,
2179 "Could not fetch the entry for the volume %u from the VLDB \n",
2181 MapHostToNetwork(&entry);
2184 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
2185 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
2186 fromtid = totid = 0; /* initialize to uncreated */
2189 * clone the read/write volume locally.
2193 if (!(flags & RV_NOCLONE)) {
2194 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2196 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2199 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
2203 /* Get a clone id */
2204 VPRINT1("Allocating new volume id for clone of volume %u ...",
2208 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
2210 EGOTO1(mfail, vcode,
2211 "Could not get an ID for the clone of volume %u from the VLDB\n",
2219 /* Get a new volume id */
2220 VPRINT1("Allocating new volume id for copy of volume %u ...", afromvol);
2222 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &newVol);
2223 EGOTO1(mfail, vcode,
2224 "Could not get an ID for the copy of volume %u from the VLDB\n",
2229 if (!(flags & RV_NOCLONE)) {
2230 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
2231 VPRINT1("Cloning source volume %u ...", afromvol);
2232 strcpy(vname, "copy-clone-temp");
2235 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname,
2238 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
2242 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
2244 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2249 "Failed to end the transaction on the source volume %u\n",
2255 * Create the destination volume
2258 if (!(flags & RV_NOCLONE)) {
2259 VPRINT1("Starting transaction on the cloned volume %u ...", cloneVol);
2262 AFSVolTransCreate_retry(fromconn, cloneVol, afrompart, ITOffline,
2266 "Failed to start a transaction on the cloned volume%u\n",
2270 VPRINT1("Setting flags on cloned volume %u ...", cloneVol);
2272 AFSVolSetFlags(fromconn, clonetid,
2273 VTDeleteOnSalvage | VTOutOfService); /*redundant */
2274 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
2278 /* remember time from which we've dumped the volume */
2279 VPRINT1("Getting status of cloned volume %u ...", cloneVol);
2280 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
2282 "Failed to get the status of the cloned volume %u\n",
2286 fromDate = CLOCKADJ(tstatus.creationDate);
2291 /* create a volume on the target machine */
2294 code = AFSVolTransCreate_retry(toconn, newVol, atopart, ITOffline, &tmp);
2297 if ((flags & RV_CPINCR)) {
2298 VPRINT1("Getting status of pre-existing volume %u ...", newVol);
2299 code = AFSVolGetStatus(toconn, totid, &tstatus);
2301 "Failed to get the status of the pre-existing volume %u\n",
2305 /* Using the update date should be OK here, but add some fudge */
2306 cloneFromDate = CLOCKADJ(tstatus.updateDate);
2307 if ((flags & RV_NOCLONE))
2308 fromDate = cloneFromDate;
2310 /* XXX We should check that the source volume's creationDate is
2311 * XXX not newer than the existing target volume, and if not,
2312 * XXX throw away the existing target and do a full dump. */
2317 /* Delete the existing volume.
2318 * While we are deleting the volume in these steps, the transaction
2319 * we started against the cloned volume (clonetid above) will be
2320 * sitting idle. It will get cleaned up after 600 seconds
2322 VPRINT1("Deleting pre-existing volume %u on destination ...", newVol);
2323 code = AFSVolDeleteVolume(toconn, totid);
2325 "Could not delete the pre-existing volume %u on destination\n",
2330 ("Ending transaction on pre-existing volume %u on destination ...",
2332 code = AFSVolEndTrans(toconn, totid, &rcode);
2337 "Could not end the transaction on pre-existing volume %u on destination\n",
2342 VPRINT1("Creating the destination volume %u ...", newVol);
2345 AFSVolCreateVolume(toconn, atopart, atovolname,
2346 (flags & RV_RDONLY) ? volser_RO : volser_RW,
2347 newVol, &newVol, &tmp);
2349 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
2353 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2355 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
2357 "Failed to set the flags on the destination volume %u\n", newVol);
2362 destination.destHost = ntohl(atoserver);
2363 destination.destPort = AFSCONF_VOLUMEPORT;
2364 destination.destSSID = 1;
2366 strncpy(cookie.name, atovolname, VOLSER_OLDMAXVOLNAME);
2367 cookie.type = (flags & RV_RDONLY) ? ROVOL : RWVOL;
2372 * Now dump the clone to the new volume
2375 if (!(flags & RV_NOCLONE)) {
2376 /* XXX probably should have some code here that checks to see if
2377 * XXX we are copying to same server and partition - if so, just
2378 * XXX use a clone to save disk space */
2380 /* Copy the clone to the new volume */
2381 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
2384 AFSVolForward(fromconn, clonetid, cloneFromDate, &destination,
2386 EGOTO1(mfail, code, "Failed to move data for the volume %u\n",
2390 VPRINT1("Ending transaction on cloned volume %u ...", cloneVol);
2391 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2396 "Failed to end the transaction on the cloned volume %u\n",
2402 * reattach to the main-line volume, and incrementally dump it.
2405 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2407 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
2410 "Failed to create a transaction on the source volume %u\n",
2414 /* now do the incremental */
2416 ("Doing the%s dump from source to destination for volume %u ... ",
2417 (flags & RV_NOCLONE) ? "" : " incremental",
2420 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
2423 "Failed to do the%s dump from old site to new site\n",
2424 (flags & RV_NOCLONE) ? "" : " incremental");
2427 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2428 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
2429 code = AFSVolSetFlags(toconn, totid, volflag);
2431 "Failed to set the flags to make destination volume online\n");
2434 /* put new volume online */
2435 VPRINT1("Ending transaction on destination volume %u ...", newVol);
2436 code = AFSVolEndTrans(toconn, totid, &rcode);
2441 "Failed to end the transaction on the destination volume %u\n",
2445 VPRINT1("Ending transaction on source volume %u ...", afromvol);
2446 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2451 "Failed to end the transaction on the source volume %u\n",
2457 if (!(flags & RV_NOCLONE)) {
2458 code = DoVolDelete(fromconn, cloneVol, afrompart,
2459 "cloned", 0, NULL, NULL);
2466 if (!(flags & RV_NOVLDB)) {
2467 /* create the vldb entry for the copied volume */
2468 strncpy(newentry.name, atovolname, VOLSER_OLDMAXVOLNAME);
2469 newentry.nServers = 1;
2470 newentry.serverNumber[0] = atoserver;
2471 newentry.serverPartition[0] = atopart;
2472 newentry.flags = (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
2473 newentry.serverFlags[0] = (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
2474 newentry.volumeId[RWVOL] = newVol;
2475 newentry.volumeId[ROVOL] = (flags & RV_RDONLY) ? newVol : 0;
2476 newentry.volumeId[BACKVOL] = 0;
2477 newentry.cloneId = 0;
2478 /*map into right byte order, before passing to xdr, the stuff has to be in host
2479 * byte order. Xdr converts it into network order */
2480 MapNetworkToHost(&newentry, &storeEntry);
2481 /* create the vldb entry */
2482 vcode = VLDB_CreateEntry(&storeEntry);
2485 "Could not create a VLDB entry for the volume %s %lu\n",
2486 atovolname, (unsigned long)newVol);
2487 /*destroy the created volume */
2488 VPRINT1("Deleting the newly created volume %u\n", newVol);
2489 AFSVolDeleteVolume(toconn, totid);
2493 VPRINT2("Created the VLDB entry for the volume %s %u\n", atovolname,
2497 /* normal cleanup code */
2500 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
2502 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2503 if (code || rcode) {
2506 "Could not end transaction on the source volume %lu\n",
2507 (unsigned long)afromvol);
2509 error = (code ? code : rcode);
2515 VPRINT1("Cleanup: Ending transaction on clone volume %u ...",
2517 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2518 if (code || rcode) {
2521 "Could not end transaction on the source's clone volume %lu\n",
2522 (unsigned long)cloneVol);
2524 error = (code ? code : rcode);
2530 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
2532 code = AFSVolEndTrans(toconn, totid, &rcode);
2536 "Could not end transaction on destination volume %lu\n",
2537 (unsigned long)newVol);
2539 error = (code ? code : rcode);
2544 rx_DestroyConnection(fromconn);
2546 rx_DestroyConnection(toconn);
2547 PrintError("", error);
2550 /* come here only when the sky falls */
2555 "vos copy: operation interrupted, cleanup in progress...\n");
2556 fprintf(STDOUT, "clear transaction contexts\n");
2561 VPRINT("Recovery: Ending transaction on clone volume ...");
2562 AFSVolEndTrans(fromconn, clonetid, &rcode);
2566 VPRINT("Recovery: Ending transaction on destination volume ...");
2567 AFSVolEndTrans(toconn, totid, &rcode);
2570 if (fromtid) { /* put it on-line */
2571 VPRINT("Recovery: Ending transaction on source volume ...");
2572 AFSVolEndTrans(fromconn, fromtid, &rcode);
2576 VPRINT("Recovery: Accessing VLDB.\n");
2577 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2579 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2583 MapHostToNetwork(&entry);
2585 /* common cleanup - delete local clone */
2587 DoVolDelete(fromconn, cloneVol, afrompart,
2588 "clone", 0, NULL, "Recovery:");
2590 done: /* routine cleanup */
2592 rx_DestroyConnection(fromconn);
2594 rx_DestroyConnection(toconn);
2597 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2605 UV_CopyVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2606 char *atovolname, afs_uint32 atoserver, afs_int32 atopart)
2608 return UV_CopyVolume2(afromvol, afromserver, afrompart,
2609 atovolname, atoserver, atopart, 0, 0);
2614 /* Make a new backup of volume <avolid> on <aserver> and <apart>
2615 * if one already exists, update it
2619 UV_BackupVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid)
2621 struct rx_connection *aconn = (struct rx_connection *)0;
2622 afs_int32 ttid = 0, btid = 0;
2623 afs_uint32 backupID;
2624 afs_int32 code = 0, rcode = 0;
2625 struct nvldbentry entry, storeEntry;
2626 afs_int32 error = 0;
2627 int vldblocked = 0, vldbmod = 0;
2629 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
2631 /* the calls to VLDB will succeed only if avolid is a RW volume,
2632 * since we are following the RW hash chain for searching */
2633 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2636 "Could not fetch the entry for the volume %lu from the VLDB \n",
2637 (unsigned long)avolid);
2641 MapHostToNetwork(&entry);
2643 /* These operations require the VLDB be locked since it means the VLDB
2644 * will change or the vldb is already locked.
2646 if (!(entry.flags & BACK_EXISTS) || /* backup volume doesnt exist */
2647 (entry.flags & VLOP_ALLOPERS) || /* vldb lock already held */
2648 (entry.volumeId[BACKVOL] == INVALID_BID)) { /* no assigned backup volume id */
2650 code = ubik_VL_SetLock(cstruct, 0, avolid, RWVOL, VLOP_BACKUP);
2653 "Could not lock the VLDB entry for the volume %lu\n",
2654 (unsigned long)avolid);
2660 /* Reread the vldb entry */
2661 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2664 "Could not fetch the entry for the volume %lu from the VLDB \n",
2665 (unsigned long)avolid);
2669 MapHostToNetwork(&entry);
2672 if (!ISNAMEVALID(entry.name)) {
2673 fprintf(STDERR, "Name of the volume %s exceeds the size limit\n",
2675 error = VOLSERBADNAME;
2679 backupID = entry.volumeId[BACKVOL];
2680 if (backupID == INVALID_BID) {
2681 /* Get a backup volume id from the VLDB and update the vldb
2684 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &backupID);
2687 "Could not allocate ID for the backup volume of %lu from the VLDB\n",
2688 (unsigned long)avolid);
2692 entry.volumeId[BACKVOL] = backupID;
2696 code = DoVolClone(aconn, avolid, apart, backupVolume, backupID, "backup",
2697 entry.name, NULL, ".backup", NULL, NULL);
2703 /* Mark vldb as backup exists */
2704 if (!(entry.flags & BACK_EXISTS)) {
2705 entry.flags |= BACK_EXISTS;
2709 /* Now go back to the backup volume and bring it on line */
2710 code = AFSVolTransCreate_retry(aconn, backupID, apart, ITOffline, &btid);
2713 "Failed to start a transaction on the backup volume %lu\n",
2714 (unsigned long)backupID);
2719 code = AFSVolSetFlags(aconn, btid, 0);
2721 fprintf(STDERR, "Could not mark the backup volume %lu on line \n",
2722 (unsigned long)backupID);
2727 code = AFSVolEndTrans(aconn, btid, &rcode);
2729 if (code || rcode) {
2731 "Failed to end the transaction on the backup volume %lu\n",
2732 (unsigned long)backupID);
2733 error = (code ? code : rcode);
2739 /* Will update the vldb below */
2743 code = AFSVolEndTrans(aconn, ttid, &rcode);
2744 if (code || rcode) {
2745 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
2746 (unsigned long)avolid);
2748 error = (code ? code : rcode);
2753 code = AFSVolEndTrans(aconn, btid, &rcode);
2754 if (code || rcode) {
2756 "Could not end transaction the backup volume %lu\n",
2757 (unsigned long)backupID);
2759 error = (code ? code : rcode);
2763 /* Now update the vldb - if modified */
2766 MapNetworkToHost(&entry, &storeEntry);
2768 VLDB_ReplaceEntry(avolid, RWVOL, &storeEntry,
2769 (LOCKREL_OPCODE | LOCKREL_AFSID |
2770 LOCKREL_TIMESTAMP));
2773 "Could not update the VLDB entry for the volume %lu \n",
2774 (unsigned long)avolid);
2780 ubik_VL_ReleaseLock(cstruct, 0, avolid, RWVOL,
2781 (LOCKREL_OPCODE | LOCKREL_AFSID |
2782 LOCKREL_TIMESTAMP));
2785 "Could not unlock the VLDB entry for the volume %lu \n",
2786 (unsigned long)avolid);
2794 rx_DestroyConnection(aconn);
2796 PrintError("", error);
2800 /* Make a new clone of volume <avolid> on <aserver> and <apart>
2801 * using volume ID <acloneid>, or a new ID allocated from the VLDB.
2802 * The new volume is named by <aname>, or by appending ".clone" to
2803 * the existing name if <aname> is NULL. The following flags are
2806 * RV_RDONLY - target volume is RO
2807 * RV_OFFLINE - leave target volume offline
2811 UV_CloneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid,
2812 afs_uint32 acloneid, char *aname, int flags)
2814 struct rx_connection *aconn = (struct rx_connection *)0;
2815 afs_int32 ttid = 0, btid = 0;
2816 afs_int32 code = 0, rcode = 0;
2817 char vname[VOLSER_MAXVOLNAME + 1];
2818 afs_int32 error = 0;
2819 volEntries volumeInfo;
2822 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
2825 volumeInfo.volEntries_val = (volintInfo *) 0;
2826 volumeInfo.volEntries_len = 0;
2827 code = AFSVolListOneVolume(aconn, apart, avolid, &volumeInfo);
2829 fprintf(stderr, "Could not get info for volume %lu\n",
2830 (unsigned long)avolid);
2834 strncpy(vname, volumeInfo.volEntries_val[0].name,
2835 VOLSER_OLDMAXVOLNAME - 7);
2836 vname[VOLSER_OLDMAXVOLNAME - 7] = 0;
2837 strcat(vname, ".clone");
2839 if (volumeInfo.volEntries_val)
2840 free(volumeInfo.volEntries_val);
2844 /* Get a clone id */
2845 VPRINT1("Allocating new volume id for clone of volume %u ...",
2847 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &acloneid);
2849 "Could not get an ID for the clone of volume %u from the VLDB\n",
2854 if (flags & RV_RWONLY)
2855 type = readwriteVolume;
2856 else if (flags & RV_RDONLY)
2857 type = readonlyVolume;
2859 type = backupVolume;
2861 code = DoVolClone(aconn, avolid, apart, type, acloneid, "clone",
2862 NULL, ".clone", NULL, NULL, NULL);
2868 /* Now go back to the backup volume and bring it on line */
2869 if (!(flags & RV_OFFLINE)) {
2870 code = AFSVolTransCreate_retry(aconn, acloneid, apart, ITOffline, &btid);
2873 "Failed to start a transaction on the clone volume %lu\n",
2874 (unsigned long)acloneid);
2879 code = AFSVolSetFlags(aconn, btid, 0);
2881 fprintf(STDERR, "Could not mark the clone volume %lu on line \n",
2882 (unsigned long)acloneid);
2887 code = AFSVolEndTrans(aconn, btid, &rcode);
2889 if (code || rcode) {
2891 "Failed to end the transaction on the clone volume %lu\n",
2892 (unsigned long)acloneid);
2893 error = (code ? code : rcode);
2902 code = AFSVolEndTrans(aconn, ttid, &rcode);
2903 if (code || rcode) {
2904 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
2905 (unsigned long)avolid);
2907 error = (code ? code : rcode);
2912 code = AFSVolEndTrans(aconn, btid, &rcode);
2913 if (code || rcode) {
2915 "Could not end transaction on the clone volume %lu\n",
2916 (unsigned long)acloneid);
2918 error = (code ? code : rcode);
2923 rx_DestroyConnection(aconn);
2925 PrintError("", error);
2929 #define ONERROR(ec, ep, es) do { \
2931 fprintf(STDERR, (es), (ep)); \
2936 #define ONERROR0(ec, es) do { \
2938 fprintf(STDERR, (es)); \
2943 #define ERROREXIT(ec) do { \
2948 /* Get a "transaction" on this replica. Create the volume
2949 * if necessary. Return the time from which a dump should
2950 * be made (0 if it's a new volume)
2953 GetTrans(struct nvldbentry *vldbEntryPtr, afs_int32 index,
2954 struct rx_connection **connPtr, afs_int32 * transPtr,
2955 afs_uint32 * crtimePtr, afs_uint32 * uptimePtr,
2956 afs_int32 *origflags, afs_uint32 tmpVolId)
2959 struct volser_status tstatus;
2964 *connPtr = (struct rx_connection *)0;
2969 /* get connection to the replication site */
2970 *connPtr = UV_Bind(vldbEntryPtr->serverNumber[index], AFSCONF_VOLUMEPORT);
2972 goto fail; /* server is down */
2974 volid = vldbEntryPtr->volumeId[ROVOL];
2978 AFSVolTransCreate_retry(*connPtr, volid,
2979 vldbEntryPtr->serverPartition[index], ITOffline,
2982 if (!code && (origflags[index] & RO_DONTUSE)) {
2983 /* If RO_DONTUSE is set, this is supposed to be an entirely new
2984 * site. Don't trust any data on it, since it is possible we
2985 * have encountered some temporary volume from some other
2986 * incomplete volume operation. It is difficult to detect if
2987 * that has happened vs if this is a legit volume, so just
2988 * delete it to be safe. */
2990 VPRINT1("Deleting extant RO_DONTUSE site on %s...",
2991 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
2992 serverNumber[index], hoststr) :
2993 hostutil_GetNameByINet(vldbEntryPtr->
2994 serverNumber[index]));
2996 code = AFSVolDeleteVolume(*connPtr, *transPtr);
2998 PrintError("Failed to delete RO_DONTUSE site: ", code);
3002 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3008 PrintError("Failed to end transaction on RO_DONTUSE site: ",
3015 /* emulate what TransCreate would have returned, so we try to
3016 * create the volume below */
3021 /* If the volume does not exist, create it */
3022 if (!volid || code) {
3023 char volname[VL_MAXNAMELEN];
3026 if (volid && (code != VNOVOL)) {
3027 PrintError("Failed to start a transaction on the RO volume.\n",
3032 strlcpy(volname, vldbEntryPtr->name, sizeof(volname));
3034 if (strlcat(volname,
3035 tmpVolId?".roclone":".readonly",
3036 sizeof(volname)) >= sizeof(volname)) {
3038 PrintError("Volume name is too long\n", code);
3043 strcat(volname, ".roclone");
3045 strcat(volname, ".readonly");
3049 "Creating new volume %lu on replication site %s: ",
3050 tmpVolId?(unsigned long)tmpVolId:(unsigned long)volid,
3051 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3052 serverNumber[index], hoststr) :
3053 hostutil_GetNameByINet(vldbEntryPtr->
3054 serverNumber[index]));
3059 AFSVolCreateVolume(*connPtr, vldbEntryPtr->serverPartition[index],
3061 vldbEntryPtr->volumeId[RWVOL],
3062 tmpVolId?&tmpVolId:&volid,
3065 PrintError("Failed to create the ro volume: ", code);
3068 vldbEntryPtr->volumeId[ROVOL] = volid;
3072 /* The following is a bit redundant, since create sets these flags by default */
3074 AFSVolSetFlags(*connPtr, *transPtr,
3075 VTDeleteOnSalvage | VTOutOfService);
3077 PrintError("Failed to set flags on the ro volume: ", code);
3082 /* Otherwise, the transaction did succeed, so get the creation date of the
3083 * latest RO volume on the replication site
3086 VPRINT2("Updating existing ro volume %u on %s ...\n", volid,
3087 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3088 serverNumber[index], hoststr) :
3089 hostutil_GetNameByINet(vldbEntryPtr->serverNumber[index]));
3091 code = AFSVolGetStatus(*connPtr, *transPtr, &tstatus);
3093 PrintError("Failed to get status of volume on destination: ",
3098 code = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3103 code = DoVolClone(*connPtr, volid,
3104 vldbEntryPtr->serverPartition[index],
3105 readonlyVolume, tmpVolId, "temporary",
3106 vldbEntryPtr->name, NULL, ".roclone", NULL,
3111 *crtimePtr = CLOCKADJ(tstatus.creationDate);
3112 *uptimePtr = CLOCKADJ(tstatus.updateDate);
3119 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3123 if (tcode && tcode != ENOENT)
3124 PrintError("Could not end transaction on a ro volume: ", tcode);
3131 SimulateForwardMultiple(struct rx_connection *fromconn, afs_int32 fromtid,
3132 afs_int32 fromdate, manyDests * tr, afs_int32 flags,
3133 void *cookie, manyResults * results)
3137 for (i = 0; i < tr->manyDests_len; i++) {
3138 results->manyResults_val[i] =
3139 AFSVolForward(fromconn, fromtid, fromdate,
3140 &(tr->manyDests_val[i].server),
3141 tr->manyDests_val[i].trans, cookie);
3147 * Check if a trans has timed out, and recreate it if necessary.
3149 * @param[in] aconn RX connection to the relevant server
3150 * @param[inout] atid Transaction ID to check; if we recreated the trans,
3151 * contains the new trans ID on success
3152 * @param[in] apart Partition for the transaction
3153 * @param[in] astat The status of the original transaction
3155 * @return operation status
3156 * @retval 0 existing transaction is still valid, or we managed to recreate
3157 * the trans successfully
3158 * @retval nonzero Fatal error; bail out
3161 CheckTrans(struct rx_connection *aconn, afs_int32 *atid, afs_int32 apart,
3162 struct volser_status *astat)
3164 struct volser_status new_status;
3167 memset(&new_status, 0, sizeof(new_status));
3168 code = AFSVolGetStatus(aconn, *atid, &new_status);
3170 if (code == ENOENT) {
3172 VPRINT1("Old transaction on cloned volume %lu timed out, "
3173 "restarting transaction\n", (long unsigned) astat->volID);
3174 code = AFSVolTransCreate_retry(aconn, astat->volID, apart,
3177 PrintError("Failed to recreate cloned RO volume transaction\n",
3182 memset(&new_status, 0, sizeof(new_status));
3183 code = AFSVolGetStatus(aconn, *atid, &new_status);
3185 PrintError("Failed to get status on recreated transaction\n",
3190 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3191 PrintError("Recreated transaction on cloned RO volume, but "
3192 "the volume has changed!\n", 0);
3196 PrintError("Unable to get status of current cloned RO transaction\n",
3201 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3203 PrintError("Internal error: current GetStatus does not match "
3204 "original GetStatus?\n", 0);
3213 PutTrans(afs_int32 *vldbindex, struct replica *replicas,
3214 struct rx_connection **toconns, struct release *times,
3217 afs_int32 s, code = 0, rcode = 0;
3218 /* End the transactions and destroy the connections */
3219 for (s = 0; s < volcount; s++) {
3220 if (replicas[s].trans) {
3221 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
3223 replicas[s].trans = 0;
3227 if ((s == 0) || (code != ENOENT)) {
3228 PrintError("Could not end transaction on a ro volume: ",
3232 ("Transaction timed out on a ro volume. Will retry.\n",
3234 if (times[s].vldbEntryIndex < *vldbindex)
3235 *vldbindex = times[s].vldbEntryIndex;
3240 rx_DestroyConnection(toconns[s]);
3246 DoVolOnline(struct nvldbentry *vldbEntryPtr, afs_uint32 avolid, int index,
3247 char *vname, struct rx_connection *connPtr)
3249 afs_int32 code = 0, rcode = 0, onlinetid = 0;
3252 AFSVolTransCreate_retry(connPtr, avolid,
3253 vldbEntryPtr->serverPartition[index],
3257 EPRINT(code, "Could not create transaction on readonly...\n");
3260 code = AFSVolSetFlags(connPtr, onlinetid, 0);
3262 EPRINT(code, "Could not set flags on readonly...\n");
3267 AFSVolSetIdsTypes(connPtr, onlinetid, vname,
3268 ROVOL, vldbEntryPtr->volumeId[RWVOL],
3271 EPRINT(code, "Could not set ids on readonly...\n");
3274 code = AFSVolEndTrans(connPtr, onlinetid, &rcode);
3280 /* UV_ReleaseVolume()
3281 * Release volume <afromvol> on <afromserver> <afrompart> to all
3282 * its RO sites (full release). Unless the previous release was
3283 * incomplete: in which case we bring the remaining incomplete
3284 * volumes up to date with the volumes that were released
3286 * forceflag: Performs a full release.
3288 * Will create a clone from the RW, then dump the clone out to
3289 * the remaining replicas. If there is more than 1 RO sites,
3290 * ensure that the VLDB says at least one RO is available all
3291 * the time: Influences when we write back the VLDB entry.
3295 UV_ReleaseVolume(afs_uint32 afromvol, afs_uint32 afromserver,
3296 afs_int32 afrompart, int forceflag, int stayUp)
3300 afs_int32 vcode, rcode, tcode;
3301 afs_uint32 cloneVolId = 0, roVolId;
3302 struct replica *replicas = 0;
3303 struct nvldbentry entry, storeEntry;
3304 int i, volcount = 0, m, fullrelease, vldbindex;
3306 struct restoreCookie cookie;
3307 struct rx_connection **toconns = 0;
3308 struct release *times = 0;
3310 struct rx_connection *fromconn = (struct rx_connection *)0;
3311 afs_int32 error = 0;
3313 afs_int32 clonetid = 0, onlinetid;
3314 afs_int32 fromtid = 0;
3315 afs_uint32 fromdate = 0;
3316 afs_uint32 thisdate;
3320 manyResults results;
3321 int rwindex, roindex, roclone, roexists;
3322 afs_uint32 rwcrdate = 0, rwupdate = 0;
3323 afs_uint32 clcrdate;
3327 } remembertime[NMAXNSERVERS];
3328 int releasecount = 0;
3329 struct volser_status volstatus;
3331 afs_int32 origflags[NMAXNSERVERS];
3332 struct volser_status orig_status;
3333 int notreleased = 0;
3334 int tried_justnewsites = 0;
3335 int justnewsites = 0; /* are we just trying to release to new RO sites? */
3337 memset(remembertime, 0, sizeof(remembertime));
3338 memset(&results, 0, sizeof(results));
3339 memset(origflags, 0, sizeof(origflags));
3341 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_RELEASE);
3342 if (vcode != VL_RERELEASE)
3343 ONERROR(vcode, afromvol,
3344 "Could not lock the VLDB entry for the volume %u.\n");
3347 /* Get the vldb entry in readable format */
3348 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
3349 ONERROR(vcode, afromvol,
3350 "Could not fetch the entry for the volume %u from the VLDB.\n");
3351 MapHostToNetwork(&entry);
3354 EnumerateEntry(&entry);
3356 if (!ISNAMEVALID(entry.name))
3357 ONERROR(VOLSERBADOP, entry.name,
3358 "Volume name %s is too long, rename before releasing.\n");
3359 if (entry.volumeId[RWVOL] != afromvol)
3360 ONERROR(VOLSERBADOP, afromvol,
3361 "The volume %u being released is not a read-write volume.\n");
3362 if (entry.nServers <= 1)
3363 ONERROR(VOLSERBADOP, afromvol,
3364 "Volume %u has no replicas - release operation is meaningless!\n");
3365 if (strlen(entry.name) > (VOLSER_OLDMAXVOLNAME - 10))
3366 ONERROR(VOLSERBADOP, entry.name,
3367 "RO volume name %s exceeds (VOLSER_OLDMAXVOLNAME - 10) character limit\n");
3369 /* roclone is true if one of the RO volumes is on the same
3370 * partition as the RW volume. In this case, we make the RO volume
3371 * on the same partition a clone instead of a complete copy.
3374 roindex = Lp_ROMatch(afromserver, afrompart, &entry) - 1;
3375 roclone = ((roindex == -1) ? 0 : 1);
3376 rwindex = Lp_GetRwIndex(&entry);
3378 ONERROR0(VOLSERNOVOL, "There is no RW volume \n");
3380 /* Make sure we have a RO volume id to work with */
3381 if (entry.volumeId[ROVOL] == INVALID_BID) {
3382 /* need to get a new RO volume id */
3383 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &roVolId);
3384 ONERROR(vcode, entry.name, "Cant allocate ID for RO volume of %s\n");
3386 entry.volumeId[ROVOL] = roVolId;
3387 MapNetworkToHost(&entry, &storeEntry);
3388 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3389 ONERROR(vcode, entry.name, "Could not update vldb entry for %s.\n");
3392 /* Will we be completing a previously unfinished release. -force overrides */
3393 for (s = 0, m = 0, fullrelease=0, i=0; (i<entry.nServers); i++) {
3394 if (entry.serverFlags[i] & ITSROVOL) {
3396 if (entry.serverFlags[i] & NEW_REPSITE) s++;
3397 if (entry.serverFlags[i] & RO_DONTUSE) notreleased++;
3399 origflags[i] = entry.serverFlags[i];
3401 if ((forceflag && !fullrelease) || (s == m) || (s == 0))
3404 if (!forceflag && (s == m || s == 0)) {
3405 if (notreleased && notreleased != m) {
3406 /* we have some new unreleased sites. try to just release to those,
3407 * if the RW has not changed */
3412 /* Determine which volume id to use and see if it exists */
3415 || (entry.cloneId == 0)) ? entry.volumeId[ROVOL] : entry.cloneId);
3416 code = VolumeExists(afromserver, afrompart, cloneVolId);
3417 roexists = ((code == ENODEV) ? 0 : 1);
3419 /* For stayUp case, if roclone is the only site, bypass special handling */
3420 if (stayUp && roclone) {
3424 for (e = 0; (e < entry.nServers) && !error; e++) {
3425 if ((entry.serverFlags[e] & ITSROVOL)) {
3426 if (!(VLDB_IsSameAddrs(entry.serverNumber[e], afromserver,
3431 if (e >= entry.nServers)
3435 /* If we had a previous release to complete, do so, else: */
3436 if (stayUp && (cloneVolId == entry.volumeId[ROVOL])) {
3437 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &cloneVolId);
3438 ONERROR(code, afromvol,
3439 "Cannot get temporary clone id for volume %u\n");
3442 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
3444 ONERROR(-1, afromserver,
3445 "Cannot establish connection with server 0x%x\n");
3449 fullrelease = 1; /* Do a full release if RO clone does not exist */
3451 /* Begin transaction on RW and mark it busy while we query it */
3452 code = AFSVolTransCreate_retry(
3453 fromconn, afromvol, afrompart, ITBusy, &fromtid
3455 ONERROR(code, afromvol,
3456 "Failed to start transaction on RW volume %u\n");
3458 /* Query the creation date for the RW */
3459 code = AFSVolGetStatus(fromconn, fromtid, &volstatus);
3460 ONERROR(code, afromvol,
3461 "Failed to get the status of RW volume %u\n");
3462 rwcrdate = volstatus.creationDate;
3464 /* End transaction on RW */
3465 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
3467 ONERROR((code ? code : rcode), afromvol,
3468 "Failed to end transaction on RW volume %u\n");
3470 /* Begin transaction on clone and mark it busy while we query it */
3471 code = AFSVolTransCreate_retry(
3472 fromconn, cloneVolId, afrompart, ITBusy, &clonetid
3474 ONERROR(code, cloneVolId,
3475 "Failed to start transaction on RW clone %u\n");
3477 /* Query the creation date for the clone */
3478 code = AFSVolGetStatus(fromconn, clonetid, &volstatus);
3479 ONERROR(code, cloneVolId,
3480 "Failed to get the status of RW clone %u\n");
3481 clcrdate = volstatus.creationDate;
3483 /* End transaction on clone */
3484 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
3486 ONERROR((code ? code : rcode), cloneVolId,
3487 "Failed to end transaction on RW clone %u\n");
3489 if (rwcrdate > clcrdate)
3490 fullrelease = 2;/* Do a full release if RO clone older than RW */
3494 if (fullrelease != 1) {
3495 /* in case the RW has changed, and just to be safe */
3500 switch (fullrelease) {
3502 fprintf(STDOUT, "RW %lu changed, doing a complete release\n",
3503 (unsigned long)afromvol);
3506 fprintf(STDOUT, "This is a complete release of volume %lu\n",
3507 (unsigned long)afromvol);
3509 tried_justnewsites = 1;
3510 fprintf(STDOUT, "There are new RO sites; we will try to "
3511 "only release to new sites\n");
3515 fprintf(STDOUT, "This is a completion of a previous release\n");
3521 afs_int32 oldest = 0;
3522 /* If the RO clone exists, then if the clone is a temporary
3523 * clone, delete it. Or if the RO clone is marked RO_DONTUSE
3524 * (it was recently added), then also delete it. We do not
3525 * want to "reclone" a temporary RO clone.
3528 code = VolumeExists(afromserver, afrompart, cloneVolId);
3530 code = DoVolDelete(fromconn, cloneVolId, afrompart, "previous clone", 0,
3532 if (code && (code != VNOVOL))
3537 /* clean up any previous tmp clone before starting if staying up */
3539 && (!roclone || (entry.serverFlags[roindex] & RO_DONTUSE))) {
3540 code = DoVolDelete(fromconn,
3541 stayUp ? entry.volumeId[ROVOL] : cloneVolId,
3542 afrompart, "the", 0, NULL, NULL);
3543 if (code && (code != VNOVOL))
3549 VPRINT("Querying old RO sites for update times...");
3550 for (vldbindex = 0; vldbindex < entry.nServers; vldbindex++) {
3551 volEntries volumeInfo;
3552 struct rx_connection *conn;
3555 if (!(entry.serverFlags[vldbindex] & ITSROVOL)) {
3558 if ((entry.serverFlags[vldbindex] & RO_DONTUSE)) {
3561 conn = UV_Bind(entry.serverNumber[vldbindex], AFSCONF_VOLUMEPORT);
3563 fprintf(STDERR, "Cannot establish connection to server %s\n",
3564 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3568 volumeInfo.volEntries_val = NULL;
3569 volumeInfo.volEntries_len = 0;
3570 code = AFSVolListOneVolume(conn, entry.serverPartition[vldbindex],
3571 entry.volumeId[ROVOL],
3574 fprintf(STDERR, "Could not fetch information about RO vol %lu from server %s\n",
3575 (unsigned long)entry.volumeId[ROVOL],
3576 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3577 PrintError("", code);
3579 rx_DestroyConnection(conn);
3583 crdate = CLOCKADJ(volumeInfo.volEntries_val[0].creationDate);
3585 if (oldest == 0 || crdate < oldest) {
3589 rx_DestroyConnection(conn);
3590 free(volumeInfo.volEntries_val);
3591 volumeInfo.volEntries_val = NULL;
3592 volumeInfo.volEntries_len = 0;
3597 volEntries volumeInfo;
3598 volumeInfo.volEntries_val = NULL;
3599 volumeInfo.volEntries_len = 0;
3600 code = AFSVolListOneVolume(fromconn, afrompart, afromvol,
3603 fprintf(STDERR, "Could not fetch information about RW vol %lu from server %s\n",
3604 (unsigned long)afromvol,
3605 hostutil_GetNameByINet(afromserver));
3606 PrintError("", code);
3609 rwupdate = volumeInfo.volEntries_val[0].updateDate;
3611 free(volumeInfo.volEntries_val);
3612 volumeInfo.volEntries_val = NULL;
3613 volumeInfo.volEntries_len = 0;
3616 if (justnewsites && oldest <= rwupdate) {
3617 /* RW has changed */
3621 /* Mark all the ROs in the VLDB entry as RO_DONTUSE. We don't
3622 * write this entry out to the vlserver until after the first
3623 * RO volume is released (temp RO clones don't count).
3625 * If 'justnewsites' is set, we're only updating sites that have
3626 * RO_DONTUSE set, so set NEW_REPSITE for all of the others.
3628 for (i = 0; i < entry.nServers; i++) {
3630 if ((entry.serverFlags[i] & RO_DONTUSE)) {
3631 entry.serverFlags[i] &= ~NEW_REPSITE;
3633 entry.serverFlags[i] |= NEW_REPSITE;
3636 entry.serverFlags[i] &= ~NEW_REPSITE;
3637 entry.serverFlags[i] |= RO_DONTUSE;
3640 entry.serverFlags[rwindex] |= NEW_REPSITE;
3641 entry.serverFlags[rwindex] &= ~RO_DONTUSE;
3644 if (justnewsites && roexists) {
3645 /* if 'justnewsites' and 'roexists' are set, we don't need to do
3646 * anything with the RO clone, so skip the reclone */
3649 } else if (fullrelease) {
3652 strcpy(vname, entry.name);
3654 strcat(vname, ".roclone");
3656 strcat(vname, ".readonly");
3658 strcpy(vname, "readonly-clone-temp");
3661 code = DoVolClone(fromconn, afromvol, afrompart, readonlyVolume,
3662 cloneVolId, (roclone && !stayUp)?"permanent RO":
3663 "temporary RO", NULL, vname, NULL, &volstatus, NULL);
3669 if (justnewsites && rwupdate != volstatus.updateDate) {
3671 /* reset the serverFlags as if 'justnewsites' had never been set */
3672 for (i = 0; i < entry.nServers; i++) {
3673 entry.serverFlags[i] &= ~NEW_REPSITE;
3674 entry.serverFlags[i] |= RO_DONTUSE;
3676 entry.serverFlags[rwindex] |= NEW_REPSITE;
3677 entry.serverFlags[rwindex] &= ~RO_DONTUSE;
3680 rwcrdate = volstatus.creationDate;
3682 /* Remember clone volume ID in case we fail or are interrupted */
3683 entry.cloneId = cloneVolId;
3685 if (roclone && !stayUp) {
3686 /* Bring the RO clone online - though not if it's a temporary clone */
3687 VPRINT1("Starting transaction on RO clone volume %u...",
3690 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITOffline,
3692 ONERROR(code, cloneVolId,
3693 "Failed to start transaction on volume %u\n");
3696 VPRINT1("Setting volume flags for volume %u...", cloneVolId);
3697 tcode = AFSVolSetFlags(fromconn, onlinetid, 0);
3700 VPRINT1("Ending transaction on volume %u...", cloneVolId);
3701 code = AFSVolEndTrans(fromconn, onlinetid, &rcode);
3702 ONERROR((code ? code : rcode), cloneVolId,
3703 "Failed to end transaction on RO clone %u\n");
3706 ONERROR(tcode, cloneVolId, "Could not bring volume %u on line\n");
3708 /* Sleep so that a client searching for an online volume won't
3709 * find the clone offline and then the next RO offline while the
3710 * release brings the clone online and the next RO offline (race).
3711 * There is a fix in the 3.4 client that does not need this sleep
3712 * anymore, but we don't know what clients we have.
3714 if (entry.nServers > 2 && !justnewsites)
3717 /* Mark the RO clone in the VLDB as a good site (already released) */
3718 entry.serverFlags[roindex] |= NEW_REPSITE;
3719 entry.serverFlags[roindex] &= ~RO_DONTUSE;
3720 entry.flags |= RO_EXISTS;
3724 /* Write out the VLDB entry only if the clone is not a temporary
3725 * clone. If we did this to a temporary clone then we would end
3726 * up marking all the ROs as "old release" making the ROs
3727 * temporarily unavailable.
3729 MapNetworkToHost(&entry, &storeEntry);
3730 VPRINT1("Replacing VLDB entry for %s...", entry.name);
3731 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3732 ONERROR(vcode, entry.name,
3733 "Could not update vldb entry for %s.\n");
3739 VPRINT("RW vol has not changed; only releasing to new RO sites\n");
3740 /* act like this is a completion of a previous release */
3742 } else if (tried_justnewsites) {
3743 VPRINT("RW vol has changed; releasing to all sites\n");
3746 /* Now we will release from the clone to the remaining RO replicas.
3747 * The first 2 ROs (counting the non-temporary RO clone) are released
3748 * individually: releasecount. This is to reduce the race condition
3749 * of clients trying to find an on-line RO volume. The remaining ROs
3750 * are released in parallel but no more than half the number of ROs
3751 * (rounded up) at a time: nservers.
3754 strcpy(vname, entry.name);
3756 strcat(vname, ".roclone");
3758 strcat(vname, ".readonly");
3759 memset(&cookie, 0, sizeof(cookie));
3760 strncpy(cookie.name, vname, VOLSER_OLDMAXVOLNAME);
3761 cookie.type = ROVOL;
3762 cookie.parent = entry.volumeId[RWVOL];
3765 /* how many to do at once, excluding clone */
3766 if (stayUp || justnewsites)
3767 nservers = entry.nServers; /* can do all, none offline */
3769 nservers = entry.nServers / 2;
3770 replicas = calloc(nservers + 1, sizeof(struct replica));
3771 times = calloc(nservers + 1, sizeof(struct release));
3772 toconns = calloc(nservers + 1, sizeof(struct rx_connection *));
3773 results.manyResults_val = calloc(nservers + 1, sizeof(afs_int32));
3774 if (!replicas || !times || !results.manyResults_val || !toconns)
3776 "Failed to create transaction on the release clone\n");
3778 /* Create a transaction on the cloned volume */
3779 VPRINT1("Starting transaction on cloned volume %u...", cloneVolId);
3781 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITBusy, &fromtid);
3783 memset(&orig_status, 0, sizeof(orig_status));
3784 code = AFSVolGetStatus(fromconn, fromtid, &orig_status);
3786 if (!fullrelease && code)
3787 ONERROR(VOLSERNOVOL, afromvol,
3788 "Old clone is inaccessible. Try vos release -f %u.\n");
3789 ONERROR0(code, "Failed to create transaction on the release clone\n");
3792 /* if we have a clone, treat this as done, for now */
3793 if (stayUp && !fullrelease) {
3794 entry.serverFlags[roindex] |= NEW_REPSITE;
3795 entry.serverFlags[roindex] &= ~RO_DONTUSE;
3796 entry.flags |= RO_EXISTS;
3801 /* For each index in the VLDB */
3802 for (vldbindex = 0; vldbindex < entry.nServers;) {
3803 /* Get a transaction on the replicas. Pick replicas which have an old release. */
3805 ((volcount < nservers) && (vldbindex < entry.nServers));
3807 if (!stayUp && !justnewsites) {
3808 /* The first two RO volumes will be released individually.
3809 * The rest are then released in parallel. This is a hack
3810 * for clients not recognizing right away when a RO volume
3811 * comes back on-line.
3813 if ((volcount == 1) && (releasecount < 2))
3817 if (vldbindex == roindex)
3818 continue; /* the clone */
3819 if ((entry.serverFlags[vldbindex] & NEW_REPSITE)
3820 && !(entry.serverFlags[vldbindex] & RO_DONTUSE))
3822 if (!(entry.serverFlags[vldbindex] & ITSROVOL))
3823 continue; /* not a RO vol */
3826 /* Get a Transaction on this replica. Get a new connection if
3827 * necessary. Create the volume if necessary. Return the
3828 * time from which the dump should be made (0 if it's a new
3829 * volume). Each volume might have a different time.
3831 replicas[volcount].server.destHost =
3832 ntohl(entry.serverNumber[vldbindex]);
3833 replicas[volcount].server.destPort = AFSCONF_VOLUMEPORT;
3834 replicas[volcount].server.destSSID = 1;
3835 times[volcount].vldbEntryIndex = vldbindex;
3838 GetTrans(&entry, vldbindex, &(toconns[volcount]),
3839 &(replicas[volcount].trans),
3840 &(times[volcount].crtime),
3841 &(times[volcount].uptime),
3842 origflags, stayUp?cloneVolId:0);
3846 /* Thisdate is the date from which we want to pick up all changes */
3847 if (forceflag || !fullrelease
3848 || (rwcrdate > times[volcount].crtime)) {
3849 /* If the forceflag is set, then we want to do a full dump.
3850 * If it's not a full release, we can't be sure that the creation
3851 * date is good (so we also do a full dump).
3852 * If the RW volume was replaced (its creation date is newer than
3853 * the last release), then we can't be sure what has changed (so
3854 * we do a full dump).
3857 } else if (remembertime[vldbindex].validtime) {
3858 /* Trans was prev ended. Use the time from the prev trans
3859 * because, prev trans may have created the volume. In which
3860 * case time[volcount].time would be now instead of 0.
3863 (remembertime[vldbindex].uptime < times[volcount].uptime)
3864 ? remembertime[vldbindex].uptime
3865 : times[volcount].uptime;
3867 thisdate = times[volcount].uptime;
3869 remembertime[vldbindex].validtime = 1;
3870 remembertime[vldbindex].uptime = thisdate;
3872 if (volcount == 0) {
3873 fromdate = thisdate;
3875 /* Include this volume if it is within 15 minutes of the earliest */
3877 thisdate) ? (fromdate - thisdate) : (thisdate -
3879 AFSVolEndTrans(toconns[volcount],
3880 replicas[volcount].trans, &rcode);
3881 replicas[volcount].trans = 0;
3884 if (thisdate < fromdate)
3885 fromdate = thisdate;
3892 code = CheckTrans(fromconn, &fromtid, afrompart, &orig_status);
3899 fprintf(STDOUT, "Starting ForwardMulti from %lu to %u on %s",
3900 (unsigned long)cloneVolId, stayUp?
3901 cloneVolId:entry.volumeId[ROVOL],
3902 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[0].
3903 vldbEntryIndex], hoststr) :
3904 hostutil_GetNameByINet(entry.
3905 serverNumber[times[0].
3908 for (s = 1; s < volcount; s++) {
3909 fprintf(STDOUT, " and %s",
3910 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[s].
3911 vldbEntryIndex], hoststr) :
3912 hostutil_GetNameByINet(entry.
3913 serverNumber[times[s].
3918 fprintf(STDOUT, " (full release)");
3921 fprintf(STDOUT, " (as of %.24s)", ctime(&tmv));
3923 fprintf(STDOUT, ".\n");
3927 /* Release the ones we have collected */
3928 tr.manyDests_val = &(replicas[0]);
3929 tr.manyDests_len = results.manyResults_len = volcount;
3931 AFSVolForwardMultiple(fromconn, fromtid, fromdate, &tr,
3932 0 /*spare */ , &cookie, &results);
3933 if (code == RXGEN_OPCODE) { /* RPC Interface Mismatch */
3935 SimulateForwardMultiple(fromconn, fromtid, fromdate, &tr,
3936 0 /*spare */ , &cookie, &results);
3941 PrintError("Release failed: ", code);
3943 for (m = 0; m < volcount; m++) {
3944 if (results.manyResults_val[m]) {
3945 if ((m == 0) || (results.manyResults_val[m] != ENOENT)) {
3946 /* we retry timed out transaction. When it is
3947 * not the first volume and the transaction wasn't found
3948 * (assume it timed out and was garbage collected by volser).
3951 ("Failed to dump volume from clone to a ro site: ",
3952 results.manyResults_val[m]);
3958 AFSVolSetIdsTypes(toconns[m], replicas[m].trans, vname,
3959 ROVOL, entry.volumeId[RWVOL], 0, 0);
3961 if ((m == 0) || (code != ENOENT)) {
3962 PrintError("Failed to set correct names and ids: ",
3968 /* have to clear dest. flags to ensure new vol goes online:
3969 * because the restore (forwarded) operation copied
3970 * the V_inService(=0) flag over to the destination.
3972 code = AFSVolSetFlags(toconns[m], replicas[m].trans, 0);
3974 if ((m == 0) || (code != ENOENT)) {
3975 PrintError("Failed to set flags on ro volume: ",
3981 entry.serverFlags[times[m].vldbEntryIndex] |= NEW_REPSITE;
3982 entry.serverFlags[times[m].vldbEntryIndex] &= ~RO_DONTUSE;
3983 entry.flags |= RO_EXISTS;
3989 PutTrans(&vldbindex, replicas, toconns, times, volcount);
3990 MapNetworkToHost(&entry, &storeEntry);
3991 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3992 ONERROR(vcode, afromvol,
3993 " Could not update VLDB entry for volume %u\n");
3995 } /* for each index in the vldb */
3997 /* for the stayup case, put back at the end */
3999 afs_uint32 tmpVol = entry.volumeId[ROVOL];
4000 strcpy(vname, entry.name);
4001 strcat(vname, ".readonly");
4004 /* have to clear flags to ensure new vol goes online
4006 code = AFSVolSetFlags(fromconn, fromtid, 0);
4007 if (code && (code != ENOENT)) {
4008 PrintError("Failed to set flags on ro volume: ",
4012 VPRINT3("%sloning to permanent RO %u on %s...", roexists?"Re-c":"C", tmpVol,
4014 afs_inet_ntoa_r(entry.serverNumber[roindex],
4016 hostutil_GetNameByINet(entry.serverNumber[roindex]));
4018 code = AFSVolClone(fromconn, fromtid, roexists?tmpVol:0,
4019 readonlyVolume, vname, &tmpVol);
4023 VPRINT("Bringing readonly online...");
4024 code = DoVolOnline(&entry, tmpVol, roindex, vname,
4028 EPRINT(code, "Failed: ");
4029 entry.serverFlags[roindex] &= ~NEW_REPSITE;
4030 entry.serverFlags[roindex] |= RO_DONTUSE;
4032 entry.serverFlags[roindex] |= NEW_REPSITE;
4033 entry.serverFlags[roindex] &= ~RO_DONTUSE;
4034 entry.flags |= RO_EXISTS;
4039 for (s = 0; s < volcount; s++) {
4040 if (replicas[s].trans) {
4041 vldbindex = times[s].vldbEntryIndex;
4043 /* ok, so now we have to end the previous transaction */
4044 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
4049 code = AFSVolTransCreate_retry(toconns[s],
4051 entry.serverPartition[vldbindex],
4053 &(replicas[s].trans));
4055 PrintError("Unable to begin transaction on temporary clone: ", code);
4058 PrintError("Unable to end transaction on temporary clone: ", code);
4061 VPRINT3("%sloning to permanent RO %u on %s...", times[s].crtime?"Re-c":"C",
4063 afs_inet_ntoa_r(htonl(replicas[s].server.destHost),
4065 hostutil_GetNameByINet(htonl(replicas[s].server.destHost)));
4066 if (times[s].crtime)
4067 code = AFSVolClone(toconns[s], replicas[s].trans, tmpVol,
4068 readonlyVolume, vname, &tmpVol);
4070 code = AFSVolClone(toconns[s], replicas[s].trans, 0,
4071 readonlyVolume, vname, &tmpVol);
4074 if (!times[s].crtime) {
4075 entry.serverFlags[vldbindex] |= RO_DONTUSE;
4077 entry.serverFlags[vldbindex] &= ~NEW_REPSITE;
4078 PrintError("Failed: ",
4083 if (entry.serverFlags[vldbindex] != RO_DONTUSE) {
4084 /* bring it online (mark it InService) */
4085 VPRINT1("Bringing readonly online on %s...",
4088 htonl(replicas[s].server.destHost),
4090 hostutil_GetNameByINet(
4091 htonl(replicas[s].server.destHost)));
4093 code = DoVolOnline(&entry, tmpVol, vldbindex, vname,
4095 /* needed to come online for cloning */
4097 /* technically it's still new, just not online */
4098 entry.serverFlags[s] &= ~NEW_REPSITE;
4099 entry.serverFlags[s] |= RO_DONTUSE;
4100 if (code != ENOENT) {
4101 PrintError("Failed to set correct names and ids: ",
4108 VPRINT("Marking temporary clone for deletion...\n");
4109 code = AFSVolSetFlags(toconns[s],
4114 EPRINT(code, "Failed: ");
4118 VPRINT("Ending transaction on temporary clone...\n");
4119 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
4123 PrintError("Failed: ", code);
4126 /* ended successfully */
4127 replicas[s].trans = 0;
4129 VPRINT2("Deleting temporary clone %u on %s...", cloneVolId,
4131 afs_inet_ntoa_r(htonl(replicas[s].server.destHost),
4133 hostutil_GetNameByINet(htonl(replicas[s].server.destHost)));
4134 code = DoVolDelete(toconns[s], cloneVolId,
4135 entry.serverPartition[vldbindex],
4136 NULL, 0, NULL, NULL);
4138 EPRINT(code, "Failed: ");
4145 /* done. put the vldb entry in the success tail case*/
4146 PutTrans(&vldbindex, replicas, toconns, times, volcount);
4149 /* End the transaction on the cloned volume */
4150 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4155 PrintError("Failed to end transaction on rw volume: ", code);
4157 /* Figure out if any volume were not released and say so */
4158 for (failure = 0, i = 0; i < entry.nServers; i++) {
4159 if (!(entry.serverFlags[i] & NEW_REPSITE))
4165 "The volume %lu could not be released to the following %d sites:\n",
4166 (unsigned long)afromvol, failure);
4167 for (i = 0; i < entry.nServers; i++) {
4168 if (!(entry.serverFlags[i] & NEW_REPSITE)) {
4169 MapPartIdIntoName(entry.serverPartition[i], pname);
4170 fprintf(STDERR, "\t%35s %s\n",
4171 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
4172 hostutil_GetNameByINet(entry.serverNumber[i]), pname);
4175 MapNetworkToHost(&entry, &storeEntry);
4177 VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry,
4179 ONERROR(vcode, afromvol,
4180 " Could not update VLDB entry for volume %u\n");
4182 ERROREXIT(VOLSERBADRELEASE);
4186 /* All the ROs were release successfully. Remove the temporary clone */
4187 if (!roclone || stayUp) {
4189 fprintf(STDOUT, "Deleting the releaseClone %lu ...",
4190 (unsigned long)cloneVolId);
4193 code = DoVolDelete(fromconn, cloneVolId, afrompart, NULL, 0, NULL,
4195 ONERROR(code, cloneVolId, "Failed to delete volume %u.\n");
4199 for (i = 0; i < entry.nServers; i++)
4200 entry.serverFlags[i] &= ~NEW_REPSITE;
4202 /* Update the VLDB */
4203 VPRINT("updating VLDB ...");
4205 MapNetworkToHost(&entry, &storeEntry);
4207 VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry,
4208 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4209 ONERROR(vcode, afromvol, " Could not update VLDB entry for volume %u\n");
4214 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
4218 "Failed to end cloning transaction on the RW volume %lu\n",
4219 (unsigned long)afromvol);
4225 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4229 "Failed to end transaction on the release clone %lu\n",
4230 (unsigned long)cloneVolId);
4235 for (i = 0; i < nservers; i++) {
4236 if (replicas && replicas[i].trans) {
4237 code = AFSVolEndTrans(toconns[i], replicas[i].trans, &rcode);
4238 replicas[i].trans = 0;
4241 "Failed to end transaction on ro volume %u at server %s\n",
4242 entry.volumeId[ROVOL],
4243 noresolve ? afs_inet_ntoa_r(htonl(replicas[i].server.
4244 destHost), hoststr) :
4245 hostutil_GetNameByINet(htonl
4246 (replicas[i].server.destHost)));
4251 if (toconns && toconns[i]) {
4252 rx_DestroyConnection(toconns[i]);
4258 ubik_VL_ReleaseLock(cstruct, 0, afromvol, RWVOL,
4259 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4262 "Could not release lock on the VLDB entry for volume %lu\n",
4263 (unsigned long)afromvol);
4269 PrintError("", error);
4272 rx_DestroyConnection(fromconn);
4273 if (results.manyResults_val)
4274 free(results.manyResults_val);
4286 dump_sig_handler(int x)
4288 fprintf(STDERR, "\nSignal handler: vos dump operation\n");
4292 /* Dump the volume <afromvol> on <afromserver> and
4293 * <afrompart> to <afilename> starting from <fromdate>.
4294 * DumpFunction does the real work behind the scenes after
4295 * extracting parameters from the rock
4298 UV_DumpVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
4300 afs_int32(*DumpFunction) (struct rx_call *, void *), void *rock,
4303 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
4304 * be changing during the dump */
4305 struct rx_call * volatile fromcall = NULL;
4306 struct rx_connection * volatile fromconn = NULL;
4307 afs_int32 volatile fromtid = 0;
4309 afs_int32 rxError = 0, rcode = 0;
4310 afs_int32 code, error = 0;
4312 time_t tmv = fromdate;
4316 #ifndef AFS_NT40_ENV
4317 (void)signal(SIGPIPE, dump_sig_handler);
4319 (void)signal(SIGINT, dump_sig_handler);
4322 VEPRINT("Full Dump ...\n");
4324 VEPRINT1("Incremental Dump (as of %.24s)...\n",
4328 /* get connections to the servers */
4329 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
4331 VEPRINT1("Starting transaction on volume %u...", afromvol);
4333 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
4335 EGOTO1(error_exit, code,
4336 "Could not start transaction on the volume %u to be dumped\n",
4340 fromcall = rx_NewCall(fromconn);
4342 VEPRINT1("Starting volume dump on volume %u...", afromvol);
4343 if (flags & VOLDUMPV2_OMITDIRS)
4344 code = StartAFSVolDumpV2(fromcall, fromtid, fromdate, flags);
4346 code = StartAFSVolDump(fromcall, fromtid, fromdate);
4347 EGOTO(error_exit, code, "Could not start the dump process \n");
4350 VEPRINT1("Dumping volume %u...", afromvol);
4351 code = DumpFunction(fromcall, rock);
4352 if (code == RXGEN_OPCODE)
4354 EGOTO(error_exit, code, "Error while dumping volume \n");
4359 code = rx_EndCall(fromcall, rxError);
4360 if (code && code != RXGEN_OPCODE)
4361 fprintf(STDERR, "Error in rx_EndCall\n");
4366 VEPRINT1("Ending transaction on volume %u...", afromvol);
4367 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4368 if (code || rcode) {
4369 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
4370 (unsigned long)afromvol);
4372 error = (code ? code : rcode);
4377 rx_DestroyConnection(fromconn);
4379 if (error != RXGEN_OPCODE)
4380 PrintError("", error);
4384 /* Clone the volume <afromvol> on <afromserver> and
4385 * <afrompart>, and then dump the clone volume to
4386 * <afilename> starting from <fromdate>.
4387 * DumpFunction does the real work behind the scenes after
4388 * extracting parameters from the rock
4391 UV_DumpClonedVolume(afs_uint32 afromvol, afs_uint32 afromserver,
4392 afs_int32 afrompart, afs_int32 fromdate,
4393 afs_int32(*DumpFunction) (struct rx_call *, void *),
4394 void *rock, afs_int32 flags)
4396 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
4397 * be changing during the dump */
4398 struct rx_connection * volatile fromconn = NULL;
4399 struct rx_call * volatile fromcall = NULL;
4400 afs_int32 volatile clonetid = 0;
4401 afs_uint32 volatile clonevol = 0;
4404 afs_int32 fromtid = 0, rxError = 0, rcode = 0;
4405 afs_int32 code = 0, error = 0;
4408 time_t tmv = fromdate;
4412 #ifndef AFS_NT40_ENV
4413 (void)signal(SIGPIPE, dump_sig_handler);
4415 (void)signal(SIGINT, dump_sig_handler);
4418 VEPRINT("Full Dump ...\n");
4420 VEPRINT1("Incremental Dump (as of %.24s)...\n",
4424 /* get connections to the servers */
4425 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
4427 VEPRINT1("Starting transaction on volume %u...", afromvol);
4428 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &fromtid);
4429 EGOTO1(error_exit, code,
4430 "Could not start transaction on the volume %u to be dumped\n",
4434 /* Get a clone id */
4435 VEPRINT1("Allocating new volume id for clone of volume %u ...", afromvol);
4437 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
4439 EGOTO1(error_exit, code,
4440 "Could not get an ID for the clone of volume %u from the VLDB\n",
4444 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
4445 VEPRINT2("Cloning source volume %u to clone volume %u...", afromvol,
4447 strcpy(vname, "dump-clone-temp");
4450 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname, &tmpVol);
4452 EGOTO1(error_exit, code, "Failed to clone the source volume %u\n",
4456 VEPRINT1("Ending the transaction on the volume %u ...", afromvol);
4458 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4462 EGOTO1(error_exit, code,
4463 "Failed to end the transaction on the volume %u\n", afromvol);
4467 VEPRINT1("Starting transaction on the cloned volume %u ...", clonevol);
4470 AFSVolTransCreate_retry(fromconn, clonevol, afrompart, ITOffline,
4473 EGOTO1(error_exit, code,
4474 "Failed to start a transaction on the cloned volume%u\n",
4478 VEPRINT1("Setting flags on cloned volume %u ...", clonevol);
4479 code = AFSVolSetFlags(fromconn, clonetid, VTDeleteOnSalvage | VTOutOfService); /*redundant */
4480 EGOTO1(error_exit, code, "Could not set falgs on the cloned volume %u\n",
4485 fromcall = rx_NewCall(fromconn);
4487 VEPRINT1("Starting volume dump from cloned volume %u...", clonevol);
4488 if (flags & VOLDUMPV2_OMITDIRS)
4489 code = StartAFSVolDumpV2(fromcall, clonetid, fromdate, flags);
4491 code = StartAFSVolDump(fromcall, clonetid, fromdate);
4492 EGOTO(error_exit, code, "Could not start the dump process \n");
4495 VEPRINT1("Dumping volume %u...", afromvol);
4496 code = DumpFunction(fromcall, rock);
4497 EGOTO(error_exit, code, "Error while dumping volume \n");
4501 /* now delete the clone */
4502 VEPRINT1("Deleting the cloned volume %u ...", clonevol);
4503 code = AFSVolDeleteVolume(fromconn, clonetid);
4505 fprintf(STDERR, "Failed to delete the cloned volume %lu\n",
4506 (unsigned long)clonevol);
4512 code = rx_EndCall(fromcall, rxError);
4514 fprintf(STDERR, "Error in rx_EndCall\n");
4520 VEPRINT1("Ending transaction on cloned volume %u...", clonevol);
4521 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
4522 if (code || rcode) {
4524 "Could not end transaction on the cloned volume %lu\n",
4525 (unsigned long)clonevol);
4527 error = (code ? code : rcode);
4532 rx_DestroyConnection(fromconn);
4534 PrintError("", error);
4541 * Restore a volume <tovolid> <tovolname> on <toserver> <topart> from
4542 * the dump file <afilename>. WriteData does all the real work
4543 * after extracting params from the rock
4546 UV_RestoreVolume2(afs_uint32 toserver, afs_int32 topart, afs_uint32 tovolid,
4547 afs_uint32 toparentid, char tovolname[], int flags,
4548 afs_int32(*WriteData) (struct rx_call *, void *),
4551 struct rx_connection *toconn, *tempconn;
4552 struct rx_call *tocall;
4553 afs_int32 totid, code, rcode, vcode, terror = 0;
4554 afs_int32 rxError = 0;
4555 struct volser_status tstatus;
4556 struct volintInfo vinfo;
4558 char tovolreal[VOLSER_OLDMAXVOLNAME];
4560 afs_int32 temptid, pparentid;
4561 struct nvldbentry entry, storeEntry;
4564 struct restoreCookie cookie;
4566 afs_int32 volflag, voltype, volsertype;
4567 afs_int32 oldCreateDate, oldUpdateDate, newCreateDate, newUpdateDate;
4568 int index, same, errcode;
4572 memset(&cookie, 0, sizeof(cookie));
4576 tocall = (struct rx_call *)0;
4577 tempconn = (struct rx_connection *)0;
4581 if (flags & RV_RDONLY) {
4583 volsertype = volser_RO;
4586 volsertype = volser_RW;
4590 pparentid = toparentid;
4591 toconn = UV_Bind(toserver, AFSCONF_VOLUMEPORT);
4592 if (pvolid == 0) { /*alot a new id if needed */
4593 vcode = VLDB_GetEntryByName(tovolname, &entry);
4594 if (vcode == VL_NOENT) {
4595 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &pvolid);
4597 fprintf(STDERR, "Could not get an Id for the volume %s\n",
4603 } else if (flags & RV_RDONLY) {
4604 if (entry.flags & RW_EXISTS) {
4606 "Entry for ReadWrite volume %s already exists!\n",
4608 error = VOLSERBADOP;
4611 if (!entry.volumeId[ROVOL]) {
4613 "Existing entry for volume %s has no ReadOnly ID\n",
4615 error = VOLSERBADOP;
4618 pvolid = entry.volumeId[ROVOL];
4619 pparentid = entry.volumeId[RWVOL];
4621 pvolid = entry.volumeId[RWVOL];
4622 pparentid = entry.volumeId[RWVOL];
4625 if (!pparentid) pparentid = pvolid;
4626 /* at this point we have a volume id to use/reuse for the volume to be restored */
4627 strncpy(tovolreal, tovolname, VOLSER_OLDMAXVOLNAME);
4629 if (strlen(tovolname) > (VOLSER_OLDMAXVOLNAME - 1)) {
4630 EGOTO1(refail, VOLSERBADOP,
4631 "The volume name %s exceeds the maximum limit of (VOLSER_OLDMAXVOLNAME -1 ) bytes\n",
4634 if ((pparentid != pvolid) && (flags & RV_RDONLY)) {
4635 if (strlen(tovolname) > (VOLSER_OLDMAXVOLNAME - 10)) {
4636 EGOTO1(refail, VOLSERBADOP,
4637 "The volume name %s exceeds the maximum limit of (VOLSER_OLDMAXVOLNAME -1 ) bytes\n", tovolname);
4639 snprintf(tovolreal, VOLSER_OLDMAXVOLNAME, "%s.readonly", tovolname);
4642 MapPartIdIntoName(topart, partName);
4643 fprintf(STDOUT, "Restoring volume %s Id %lu on server %s partition %s ..",
4644 tovolreal, (unsigned long)pvolid,
4645 noresolve ? afs_inet_ntoa_r(toserver, hoststr) :
4646 hostutil_GetNameByINet(toserver), partName);
4649 AFSVolCreateVolume(toconn, topart, tovolreal, volsertype, pparentid, &pvolid,
4652 if (flags & RV_FULLRST) { /* full restore: delete then create anew */
4653 code = DoVolDelete(toconn, pvolid, topart, "the previous", 0,
4655 if (code && code != VNOVOL) {
4661 AFSVolCreateVolume(toconn, topart, tovolreal, volsertype, pparentid,
4663 EGOTO1(refail, code, "Could not create new volume %u\n", pvolid);
4666 AFSVolTransCreate_retry(toconn, pvolid, topart, ITOffline, &totid);
4667 EGOTO1(refail, code, "Failed to start transaction on %u\n",
4670 code = AFSVolGetStatus(toconn, totid, &tstatus);
4671 EGOTO1(refail, code, "Could not get timestamp from volume %u\n",
4675 oldCreateDate = tstatus.creationDate;
4676 oldUpdateDate = tstatus.updateDate;
4682 cookie.parent = pparentid;
4683 cookie.type = voltype;
4685 strncpy(cookie.name, tovolreal, VOLSER_OLDMAXVOLNAME);
4687 tocall = rx_NewCall(toconn);
4688 terror = StartAFSVolRestore(tocall, totid, 1, &cookie);
4690 fprintf(STDERR, "Volume restore Failed \n");
4694 code = WriteData(tocall, rock);
4696 fprintf(STDERR, "Could not transmit data\n");
4700 terror = rx_EndCall(tocall, rxError);
4701 tocall = (struct rx_call *)0;
4703 fprintf(STDERR, "rx_EndCall Failed \n");
4707 code = AFSVolGetStatus(toconn, totid, &tstatus);
4710 "Could not get status information about the volume %lu\n",
4711 (unsigned long)pvolid);
4715 code = AFSVolSetIdsTypes(toconn, totid, tovolreal, voltype, pparentid, 0, 0);
4717 fprintf(STDERR, "Could not set the right type and ID on %lu\n",
4718 (unsigned long)pvolid);
4723 if (flags & RV_CRDUMP)
4724 newCreateDate = tstatus.creationDate;
4725 else if (flags & RV_CRKEEP && oldCreateDate != 0)
4726 newCreateDate = oldCreateDate;
4728 newCreateDate = time(0);
4729 if (flags & RV_LUDUMP)
4730 newUpdateDate = tstatus.updateDate;
4731 else if (flags & RV_LUKEEP)
4732 newUpdateDate = oldUpdateDate;
4734 newUpdateDate = time(0);
4735 code = AFSVolSetDate(toconn,totid, newCreateDate);
4737 fprintf(STDERR, "Could not set the 'creation' date on %u\n", pvolid);
4742 init_volintInfo(&vinfo);
4743 vinfo.creationDate = newCreateDate;
4744 vinfo.updateDate = newUpdateDate;
4745 code = AFSVolSetInfo(toconn, totid, &vinfo);
4747 fprintf(STDERR, "Could not set the 'last updated' date on %u\n",
4753 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
4754 code = AFSVolSetFlags(toconn, totid, volflag);
4756 fprintf(STDERR, "Could not mark %lu online\n", (unsigned long)pvolid);
4761 /* It isn't handled right in refail */
4762 code = AFSVolEndTrans(toconn, totid, &rcode);
4767 fprintf(STDERR, "Could not end transaction on %lu\n",
4768 (unsigned long)pvolid);
4773 fprintf(STDOUT, " done\n");
4775 if (!reuseID || (flags & RV_FULLRST)) {
4776 /* Volume was restored on the file server, update the
4777 * VLDB to reflect the change.
4779 vcode = VLDB_GetEntryByID(pvolid, voltype, &entry);
4780 if (vcode && vcode != VL_NOENT && vcode != VL_ENTDELETED) {
4782 "Could not fetch the entry for volume number %lu from VLDB \n",
4783 (unsigned long)pvolid);
4788 MapHostToNetwork(&entry);
4789 if (vcode == VL_NOENT) { /* it doesnot exist already */
4790 /*make the vldb return this indication specifically */
4791 VPRINT("------- Creating a new VLDB entry ------- \n");
4792 strcpy(entry.name, tovolname);
4794 entry.serverNumber[0] = toserver; /*should be indirect */
4795 entry.serverPartition[0] = topart;
4796 entry.serverFlags[0] = (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
4797 entry.flags = (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
4798 if (flags & RV_RDONLY)
4799 entry.volumeId[ROVOL] = pvolid;
4800 else if (tstatus.cloneID != 0) {
4801 entry.volumeId[ROVOL] = tstatus.cloneID; /*this should come from status info on the volume if non zero */
4803 entry.volumeId[ROVOL] = INVALID_BID;
4804 entry.volumeId[RWVOL] = pparentid;
4806 if (tstatus.backupID != 0) {
4807 entry.volumeId[BACKVOL] = tstatus.backupID;
4808 /*this should come from status info on the volume if non zero */
4810 entry.volumeId[BACKVOL] = INVALID_BID;
4811 MapNetworkToHost(&entry, &storeEntry);
4812 vcode = VLDB_CreateEntry(&storeEntry);
4815 "Could not create the VLDB entry for volume number %lu \n",
4816 (unsigned long)pvolid);
4822 EnumerateEntry(&entry);
4823 } else { /*update the existing entry */
4825 fprintf(STDOUT, "Updating the existing VLDB entry\n");
4826 fprintf(STDOUT, "------- Old entry -------\n");
4827 EnumerateEntry(&entry);
4828 fprintf(STDOUT, "------- New entry -------\n");
4831 ubik_VL_SetLock(cstruct, 0, pvolid, voltype,
4835 "Could not lock the entry for volume number %lu \n",
4836 (unsigned long)pvolid);
4841 strcpy(entry.name, tovolname);
4843 /* Update the vlentry with the new information */
4844 if (flags & RV_RDONLY)
4845 index = Lp_ROMatch(toserver, topart, &entry) - 1;
4847 index = Lp_GetRwIndex(&entry);
4849 /* Add the new site for the volume being restored */
4850 entry.serverNumber[entry.nServers] = toserver;
4851 entry.serverPartition[entry.nServers] = topart;
4852 entry.serverFlags[entry.nServers] =
4853 (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
4856 /* This volume should be deleted on the old site
4857 * if its different from new site.
4860 VLDB_IsSameAddrs(toserver, entry.serverNumber[index],
4864 "Failed to get info about server's %d address(es) from vlserver (err=%d)\n",
4866 if ((!errcode && !same)
4867 || (entry.serverPartition[index] != topart)) {
4868 if (flags & RV_NODEL) {
4870 ("Not deleting the previous volume %u on server %s, ...",
4872 noresolve ? afs_inet_ntoa_r(entry.serverNumber[index], hoststr) :
4873 hostutil_GetNameByINet(entry.serverNumber[index]));
4876 UV_Bind(entry.serverNumber[index],
4877 AFSCONF_VOLUMEPORT);
4879 MapPartIdIntoName(entry.serverPartition[index],
4882 ("Deleting the previous volume %u on server %s, partition %s ...",
4884 noresolve ? afs_inet_ntoa_r(entry.serverNumber[index], hoststr) :
4885 hostutil_GetNameByINet(entry.serverNumber[index]),
4887 code = DoVolDelete(tempconn, pvolid,
4888 entry.serverPartition[index],
4889 "the", 0, NULL, NULL);
4890 if (code && code != VNOVOL) {
4894 MapPartIdIntoName(entry.serverPartition[index],
4898 entry.serverNumber[index] = toserver;
4899 entry.serverPartition[index] = topart;
4902 entry.flags |= (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
4903 MapNetworkToHost(&entry, &storeEntry);
4905 VLDB_ReplaceEntry(pvolid, voltype, &storeEntry,
4906 LOCKREL_OPCODE | LOCKREL_AFSID |
4910 "Could not update the entry for volume number %lu \n",
4911 (unsigned long)pvolid);
4917 EnumerateEntry(&entry);
4924 code = rx_EndCall(tocall, rxError);
4930 ubik_VL_ReleaseLock(cstruct, 0, pvolid, voltype,
4931 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4934 "Could not release lock on the VLDB entry for the volume %lu\n",
4935 (unsigned long)pvolid);
4941 code = AFSVolEndTrans(toconn, totid, &rcode);
4945 fprintf(STDERR, "Could not end transaction on the volume %lu \n",
4946 (unsigned long)pvolid);
4952 code = AFSVolEndTrans(toconn, temptid, &rcode);
4956 fprintf(STDERR, "Could not end transaction on the volume %lu \n",
4957 (unsigned long)pvolid);
4963 rx_DestroyConnection(tempconn);
4965 rx_DestroyConnection(toconn);
4966 PrintError("", error);
4971 UV_RestoreVolume(afs_uint32 toserver, afs_int32 topart, afs_uint32 tovolid,
4972 char tovolname[], int flags,
4973 afs_int32(*WriteData) (struct rx_call *, void *),
4976 return UV_RestoreVolume2(toserver, topart, tovolid, 0, tovolname, flags,
4981 /*unlocks the vldb entry associated with <volid> */
4983 UV_LockRelease(afs_uint32 volid)
4987 VPRINT("Binding to the VLDB server\n");
4989 ubik_VL_ReleaseLock(cstruct, 0, volid, -1,
4990 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4993 "Could not unlock the entry for volume number %lu in VLDB \n",
4994 (unsigned long)volid);
4995 PrintError("", vcode);
4998 VPRINT("VLDB updated\n");
5003 /* old interface to add rosites */
5005 UV_AddSite(afs_uint32 server, afs_int32 part, afs_uint32 volid,
5008 return UV_AddSite2(server, part, volid, 0, valid);
5011 /*adds <server> and <part> as a readonly replication site for <volid>
5014 UV_AddSite2(afs_uint32 server, afs_int32 part, afs_uint32 volid,
5015 afs_uint32 rovolid, afs_int32 valid)
5017 int j, nro = 0, islocked = 0;
5018 struct nvldbentry entry, storeEntry, entry2;
5019 afs_int32 vcode, error = 0;
5022 error = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5025 " Could not lock the VLDB entry for the volume %lu \n",
5026 (unsigned long)volid);
5031 error = VLDB_GetEntryByID(volid, RWVOL, &entry);
5034 "Could not fetch the VLDB entry for volume number %lu \n",
5035 (unsigned long)volid);
5039 if (!ISNAMEVALID(entry.name)) {
5041 "Volume name %s is too long, rename before adding site\n",
5043 error = VOLSERBADOP;
5046 MapHostToNetwork(&entry);
5048 /* See if it's too many entries */
5049 if (entry.nServers >= NMAXNSERVERS) {
5050 fprintf(STDERR, "Total number of entries will exceed %u\n",
5052 error = VOLSERBADOP;
5056 /* See if it's on the same server */
5057 for (j = 0; j < entry.nServers; j++) {
5058 if (entry.serverFlags[j] & ITSROVOL) {
5060 if (VLDB_IsSameAddrs(server, entry.serverNumber[j], &error)) {
5063 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
5066 MapPartIdIntoName(entry.serverPartition[j], apartName);
5068 "RO already exists on partition %s. Multiple ROs on a single server aren't allowed\n",
5070 error = VOLSERBADOP;
5077 /* See if it's too many RO sites - leave one for the RW */
5078 if (nro >= NMAXNSERVERS - 1) {
5079 fprintf(STDERR, "Total number of sites will exceed %u\n",
5081 error = VOLSERBADOP;
5085 /* if rovolid == 0, we leave the RO volume id alone. If the volume doesn't
5086 * have an RO volid at this point (i.e. entry.volumeId[ROVOL] ==
5087 * INVALID_BID) and we leave it alone, it gets an RO volid at release-time.
5090 if (entry.volumeId[ROVOL] == INVALID_BID) {
5091 vcode = VLDB_GetEntryByID(rovolid, -1, &entry2);
5093 fprintf(STDERR, "Volume ID %d already exists\n", rovolid);
5096 VPRINT1("Using RO volume id %d.\n", rovolid);
5097 entry.volumeId[ROVOL] = rovolid;
5099 fprintf(STDERR, "Ignoring given RO id %d, since volume already has RO id %d\n",
5100 rovolid, entry.volumeId[ROVOL]);
5104 VPRINT("Adding a new site ...");
5105 entry.serverNumber[entry.nServers] = server;
5106 entry.serverPartition[entry.nServers] = part;
5108 entry.serverFlags[entry.nServers] = (ITSROVOL | RO_DONTUSE);
5110 entry.serverFlags[entry.nServers] = (ITSROVOL);
5114 MapNetworkToHost(&entry, &storeEntry);
5116 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5117 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5119 fprintf(STDERR, "Could not update entry for volume %lu \n",
5120 (unsigned long)volid);
5129 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5130 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5133 "Could not release lock on volume entry for %lu \n",
5134 (unsigned long)volid);
5135 PrintError("", vcode);
5139 PrintError("", error);
5143 /*removes <server> <part> as read only site for <volid> from the vldb */
5145 UV_RemoveSite(afs_uint32 server, afs_int32 part, afs_uint32 volid)
5148 struct nvldbentry entry, storeEntry;
5150 vcode = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5152 fprintf(STDERR, " Could not lock the VLDB entry for volume %lu \n",
5153 (unsigned long)volid);
5154 PrintError("", vcode);
5157 vcode = VLDB_GetEntryByID(volid, RWVOL, &entry);
5160 "Could not fetch the entry for volume number %lu from VLDB \n",
5161 (unsigned long)volid);
5162 PrintError("", vcode);
5165 MapHostToNetwork(&entry);
5166 if (!Lp_ROMatch(server, part, &entry)) {
5167 /*this site doesnot exist */
5168 fprintf(STDERR, "This site is not a replication site \n");
5170 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5171 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5173 fprintf(STDERR, "Could not update entry for volume %lu \n",
5174 (unsigned long)volid);
5175 PrintError("", vcode);
5176 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5177 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5181 } else { /*remove the rep site */
5182 Lp_SetROValue(&entry, server, part, 0, 0);
5184 if ((entry.nServers == 1) && (entry.flags & RW_EXISTS))
5185 entry.flags &= ~RO_EXISTS;
5186 if (entry.nServers < 1) { /*this is the last ref */
5187 VPRINT1("Deleting the VLDB entry for %u ...", volid);
5189 vcode = ubik_VL_DeleteEntry(cstruct, 0, volid, ROVOL);
5192 "Could not delete VLDB entry for volume %lu \n",
5193 (unsigned long)volid);
5194 PrintError("", vcode);
5199 MapNetworkToHost(&entry, &storeEntry);
5200 fprintf(STDOUT, "Deleting the replication site for volume %lu ...",
5201 (unsigned long)volid);
5204 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5205 LOCKREL_OPCODE | LOCKREL_AFSID |
5209 "Could not release lock on volume entry for %lu \n",
5210 (unsigned long)volid);
5211 PrintError("", vcode);
5212 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5213 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5221 /*sets <server> <part> as read/write site for <volid> in the vldb */
5223 UV_ChangeLocation(afs_uint32 server, afs_int32 part, afs_uint32 volid)
5226 struct nvldbentry entry, storeEntry;
5229 vcode = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5231 fprintf(STDERR, " Could not lock the VLDB entry for volume %lu \n",
5232 (unsigned long)volid);
5233 PrintError("", vcode);
5236 vcode = VLDB_GetEntryByID(volid, RWVOL, &entry);
5239 "Could not fetch the entry for volume number %lu from VLDB \n",
5240 (unsigned long)volid);
5241 PrintError("", vcode);
5244 MapHostToNetwork(&entry);
5245 index = Lp_GetRwIndex(&entry);
5247 /* no RW site exists */
5248 fprintf(STDERR, "No existing RW site for volume %lu",
5249 (unsigned long)volid);
5251 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5252 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5255 "Could not release lock on entry for volume %lu \n",
5256 (unsigned long)volid);
5257 PrintError("", vcode);
5261 } else { /* change the RW site */
5262 entry.serverNumber[index] = server;
5263 entry.serverPartition[index] = part;
5264 MapNetworkToHost(&entry, &storeEntry);
5266 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5267 LOCKREL_OPCODE | LOCKREL_AFSID |
5270 fprintf(STDERR, "Could not update entry for volume %lu \n",
5271 (unsigned long)volid);
5272 PrintError("", vcode);
5273 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5274 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5282 /*list all the partitions on <aserver> */
5284 UV_ListPartitions(afs_uint32 aserver, struct partList *ptrPartList,
5287 struct rx_connection *aconn;
5288 struct pIDs partIds;
5289 struct partEntries partEnts;
5293 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5295 partEnts.partEntries_len = 0;
5296 partEnts.partEntries_val = NULL;
5297 code = AFSVolXListPartitions(aconn, &partEnts); /* this is available only on new servers */
5298 if (code == RXGEN_OPCODE) {
5299 for (i = 0; i < 26; i++) /* try old interface */
5300 partIds.partIds[i] = -1;
5301 code = AFSVolListPartitions(aconn, &partIds);
5303 for (i = 0; i < 26; i++) {
5304 if ((partIds.partIds[i]) != -1) {
5305 ptrPartList->partId[j] = partIds.partIds[i];
5306 ptrPartList->partFlags[j] = PARTVALID;
5309 ptrPartList->partFlags[i] = 0;
5314 *cntp = partEnts.partEntries_len;
5315 if (*cntp > VOLMAXPARTS) {
5317 "Warning: number of partitions on the server too high %d (process only %d)\n",
5318 *cntp, VOLMAXPARTS);
5319 *cntp = VOLMAXPARTS;
5321 for (i = 0; i < *cntp; i++) {
5322 ptrPartList->partId[i] = partEnts.partEntries_val[i];
5323 ptrPartList->partFlags[i] = PARTVALID;
5325 free(partEnts.partEntries_val);
5331 "Could not fetch the list of partitions from the server\n");
5332 PrintError("", code);
5334 rx_DestroyConnection(aconn);
5339 /*zap the list of volumes specified by volPtrArray (the volCloneId field).
5340 This is used by the backup system */
5342 UV_ZapVolumeClones(afs_uint32 aserver, afs_int32 apart,
5343 struct volDescription *volPtr, afs_int32 arraySize)
5345 struct rx_connection *aconn;
5346 struct volDescription *curPtr;
5349 afs_int32 success = 1;
5351 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5353 for (curPtr = volPtr; curPos < arraySize; curPtr++) {
5354 if (curPtr->volFlags & CLONEVALID) {
5355 curPtr->volFlags &= ~CLONEZAPPED;
5358 code = DoVolDelete(aconn, curPtr->volCloneId, apart,
5359 "clone", 0, NULL, NULL);
5364 curPtr->volFlags |= CLONEZAPPED;
5366 fprintf(STDERR, "Could not zap volume %lu\n",
5367 (unsigned long)curPtr->volCloneId);
5369 VPRINT2("Clone of %s %u deleted\n", curPtr->volName,
5370 curPtr->volCloneId);
5375 rx_DestroyConnection(aconn);
5379 /*return a list of clones of the volumes specified by volPtrArray. Used by the
5382 UV_GenerateVolumeClones(afs_uint32 aserver, afs_int32 apart,
5383 struct volDescription *volPtr, afs_int32 arraySize)
5385 struct rx_connection *aconn;
5386 struct volDescription *curPtr;
5389 afs_int32 rcode = 0;
5391 int reuseCloneId = 0;
5392 afs_uint32 curCloneId = 0;
5393 char cloneName[256]; /*max vol name */
5395 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5397 if ((volPtr->volFlags & REUSECLONEID) && (volPtr->volFlags & ENTRYVALID))
5399 else { /*get a bunch of id's from vldb */
5401 ubik_VL_GetNewVolumeId(cstruct, 0, arraySize, &curCloneId);
5403 fprintf(STDERR, "Could not get ID's for the clone from VLDB\n");
5404 PrintError("", code);
5409 for (curPtr = volPtr; curPos < arraySize; curPtr++) {
5410 if (curPtr->volFlags & ENTRYVALID) {
5412 curPtr->volFlags |= CLONEVALID;
5413 /*make a clone of curParentId and record as curPtr->volCloneId */
5415 AFSVolTransCreate_retry(aconn, curPtr->volId, apart, ITOffline,
5418 VPRINT2("Clone for volume %s %u failed \n", curPtr->volName,
5421 curPtr->volFlags &= ~CLONEVALID; /*cant clone */
5425 if (strlen(curPtr->volName) < (VOLSER_OLDMAXVOLNAME - 9)) {
5426 strcpy(cloneName, curPtr->volName);
5427 strcat(cloneName, "-tmpClone-");
5429 strcpy(cloneName, "-tmpClone");
5431 curPtr->volCloneId = curCloneId;
5436 AFSVolClone(aconn, tid, 0, readonlyVolume, cloneName,
5437 &(curPtr->volCloneId));
5439 curPtr->volFlags &= ~CLONEVALID;
5441 fprintf(STDERR, "Could not clone %s due to error %lu\n",
5442 curPtr->volName, (unsigned long)code);
5443 code = AFSVolEndTrans(aconn, tid, &rcode);
5445 fprintf(STDERR, "WARNING: could not end transaction\n");
5448 VPRINT2("********** Cloned %s temporary %u\n", cloneName,
5449 curPtr->volCloneId);
5450 code = AFSVolEndTrans(aconn, tid, &rcode);
5451 if (code || rcode) {
5452 curPtr->volFlags &= ~CLONEVALID;
5461 rx_DestroyConnection(aconn);
5466 /*list all the volumes on <aserver> and <apart>. If all = 1, then all the
5467 * relevant fields of the volume are also returned. This is a heavy weight operation.*/
5469 UV_ListVolumes(afs_uint32 aserver, afs_int32 apart, int all,
5470 struct volintInfo **resultPtr, afs_int32 * size)
5472 struct rx_connection *aconn;
5474 volEntries volumeInfo;
5478 *resultPtr = (volintInfo *) 0;
5479 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
5480 volumeInfo.volEntries_len = 0;
5482 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5483 code = AFSVolListVolumes(aconn, apart, all, &volumeInfo);
5486 "Could not fetch the list of volumes from the server\n");
5488 *resultPtr = volumeInfo.volEntries_val;
5489 *size = volumeInfo.volEntries_len;
5493 rx_DestroyConnection(aconn);
5494 PrintError("", code);
5498 /*------------------------------------------------------------------------
5499 * EXPORTED UV_XListVolumes
5502 * List the extended information for all the volumes on a particular
5503 * File Server and partition. We may either return the volume's ID
5504 * or all of its extended information.
5507 * a_serverID : Address of the File Server for which we want
5508 * extended volume info.
5509 * a_partID : Partition for which we want the extended
5511 * a_all : If non-zero, fetch ALL the volume info,
5512 * otherwise just the volume ID.
5513 * a_resultPP : Ptr to the address of the area containing
5514 * the returned volume info.
5515 * a_numEntsInResultP : Ptr for the value we set for the number of
5520 * Otherise, the return value of AFSVolXListVolumes.
5523 * This routine is closely related to UV_ListVolumes, which returns
5524 * only the standard level of detail on AFS volumes. It is a
5525 * heavyweight operation, zipping through all the volume entries for
5526 * a given server/partition.
5530 *------------------------------------------------------------------------*/
5533 UV_XListVolumes(afs_uint32 a_serverID, afs_int32 a_partID, int a_all,
5534 struct volintXInfo **a_resultPP,
5535 afs_int32 * a_numEntsInResultP)
5537 struct rx_connection *rxConnP; /*Ptr to the Rx connection involved */
5538 afs_int32 code; /*Error code to return */
5539 volXEntries volumeXInfo; /*Area for returned extended vol info */
5542 * Set up our error code and the area for returned extended volume info.
5543 * We set the val field to a null pointer as a hint for the stub to
5546 *a_numEntsInResultP = 0;
5547 *a_resultPP = (volintXInfo *) 0;
5548 volumeXInfo.volXEntries_val = (volintXInfo *) 0;
5549 volumeXInfo.volXEntries_len = 0;
5552 * Bind to the Volume Server port on the File Server machine in question,
5555 rxConnP = UV_Bind(a_serverID, AFSCONF_VOLUMEPORT);
5556 code = AFSVolXListVolumes(rxConnP, a_partID, a_all, &volumeXInfo);
5558 fprintf(STDERR, "[UV_XListVolumes] Couldn't fetch volume list\n");
5561 * We got the info; pull out the pointer to where the results lie
5562 * and how many entries are there.
5564 *a_resultPP = volumeXInfo.volXEntries_val;
5565 *a_numEntsInResultP = volumeXInfo.volXEntries_len;
5569 * If we got an Rx connection, throw it away.
5572 rx_DestroyConnection(rxConnP);
5574 PrintError("", code);
5576 } /*UV_XListVolumes */
5578 /* get all the information about volume <volid> on <aserver> and <apart> */
5580 UV_ListOneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 volid,
5581 struct volintInfo **resultPtr)
5583 struct rx_connection *aconn;
5585 volEntries volumeInfo;
5587 *resultPtr = (volintInfo *) 0;
5588 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
5589 volumeInfo.volEntries_len = 0;
5591 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5592 code = AFSVolListOneVolume(aconn, apart, volid, &volumeInfo);
5595 "Could not fetch the information about volume %lu from the server\n",
5596 (unsigned long)volid);
5598 *resultPtr = volumeInfo.volEntries_val;
5603 rx_DestroyConnection(aconn);
5604 PrintError("", code);
5608 /*------------------------------------------------------------------------
5609 * EXPORTED UV_XListOneVolume
5612 * List the extended information for a volume on a particular File
5613 * Server and partition.
5616 * a_serverID : Address of the File Server for which we want
5617 * extended volume info.
5618 * a_partID : Partition for which we want the extended
5620 * a_volID : Volume ID for which we want the info.
5621 * a_resultPP : Ptr to the address of the area containing
5622 * the returned volume info.
5626 * Otherise, the return value of AFSVolXListOneVolume.
5629 * This routine is closely related to UV_ListOneVolume, which returns
5630 * only the standard level of detail on the chosen AFS volume.
5634 *------------------------------------------------------------------------*/
5637 UV_XListOneVolume(afs_uint32 a_serverID, afs_int32 a_partID, afs_uint32 a_volID,
5638 struct volintXInfo **a_resultPP)
5640 struct rx_connection *rxConnP; /*Rx connection to Volume Server */
5641 afs_int32 code; /*Error code */
5642 volXEntries volumeXInfo; /*Area for returned info */
5645 * Set up our error code, and the area we're in which we are returning
5646 * the info. Setting the val field to a null pointer tells the stub
5647 * to allocate space for us.
5649 *a_resultPP = (volintXInfo *) 0;
5650 volumeXInfo.volXEntries_val = (volintXInfo *) 0;
5651 volumeXInfo.volXEntries_len = 0;
5654 * Bind to the Volume Server port on the File Server machine in question,
5657 rxConnP = UV_Bind(a_serverID, AFSCONF_VOLUMEPORT);
5658 code = AFSVolXListOneVolume(rxConnP, a_partID, a_volID, &volumeXInfo);
5661 "[UV_XListOneVolume] Couldn't fetch the volume information\n");
5664 * We got the info; pull out the pointer to where the results lie.
5666 *a_resultPP = volumeXInfo.volXEntries_val;
5669 * If we got an Rx connection, throw it away.
5672 rx_DestroyConnection(rxConnP);
5674 PrintError("", code);
5679 * Given a volume we read from a partition, check if it is
5680 * represented in the VLDB correctly.
5682 * The VLDB is looked up by the RW volume id (not its name).
5683 * The RW contains the true name of the volume (BK and RO set
5684 * the name in the VLDB only on creation of the VLDB entry).
5685 * We want rules strict enough that when we check all volumes
5686 * on one partition, it does not need to be done again. IE:
5687 * two volumes on different partitions won't constantly
5688 * change a VLDB entry away from what the other set.
5689 * For RW and BK volumes, we will always check the VLDB to see
5690 * if the two exist on the server/partition. May seem redundant,
5691 * but this is an easy check of the VLDB. IE: if the VLDB entry
5692 * says the BK exists but no BK volume is there, we will detect
5693 * this when we check the RW volume.
5694 * VLDB entries are locked only when a change needs to be done.
5695 * Output changed to look a lot like the "vos syncserv" otuput.
5698 CheckVolume(volintInfo * volumeinfo, afs_uint32 aserver, afs_int32 apart,
5699 afs_int32 * modentry, afs_uint32 * maxvolid,
5700 struct nvldbentry *aentry)
5704 afs_int32 code, error = 0;
5705 struct nvldbentry entry, storeEntry;
5707 int pass = 0, createentry, addvolume, modified, mod, doit = 1;
5717 ((volumeinfo->type ==
5718 RWVOL) ? volumeinfo->volid : volumeinfo->parentID);
5721 /* Check to see if the VLDB is ok without locking it (pass 1).
5722 * If it will change, then lock the VLDB entry, read it again,
5723 * then make the changes to it (pass 2).
5726 code = ubik_VL_SetLock(cstruct, 0, rwvolid, RWVOL, VLOP_DELETE);
5728 fprintf(STDERR, "Could not lock VLDB entry for %lu\n",
5729 (unsigned long)rwvolid);
5734 createentry = 0; /* Do we need to create a VLDB entry */
5735 addvolume = 0; /* Add this volume to the VLDB entry */
5736 modified = 0; /* The VLDB entry was modified */
5739 memcpy(&entry, aentry, sizeof(entry));
5741 /* Read the entry from VLDB by its RW volume id */
5742 code = VLDB_GetEntryByID(rwvolid, RWVOL, &entry);
5744 if (code != VL_NOENT) {
5746 "Could not retrieve the VLDB entry for volume %lu \n",
5747 (unsigned long)rwvolid);
5751 memset(&entry, 0, sizeof(entry));
5752 vsu_ExtractName(entry.name, volumeinfo->name); /* Store name of RW */
5756 MapHostToNetwork(&entry);
5760 if (verbose && (pass == 1)) {
5761 fprintf(STDOUT, "_______________________________\n");
5762 fprintf(STDOUT, "\n-- status before -- \n");
5764 fprintf(STDOUT, "\n**does not exist**\n");
5766 if ((entry.flags & RW_EXISTS) || (entry.flags & RO_EXISTS)
5767 || (entry.flags & BACK_EXISTS))
5768 EnumerateEntry(&entry);
5770 fprintf(STDOUT, "\n");
5773 if (volumeinfo->type == RWVOL) { /* RW volume exists */
5779 /* Check existence of RW and BK volumes */
5780 code = CheckVldbRWBK(&entry, &mod);
5786 idx = Lp_GetRwIndex(&entry);
5787 if (idx == -1) { /* RW index not found in the VLDB entry */
5788 idx = entry.nServers; /* put it into next index */
5791 } else { /* RW index found in the VLDB entry. */
5792 /* Verify if this volume's location matches where the VLDB says it is */
5793 if (!Lp_Match(aserver, apart, &entry)) {
5794 if (entry.flags & RW_EXISTS) {
5795 /* The RW volume exists elsewhere - report this one a duplicate */
5797 MapPartIdIntoName(apart, pname);
5799 "*** Warning: Orphaned RW volume %lu exists on %s %s\n",
5800 (unsigned long)rwvolid,
5802 afs_inet_ntoa_r(aserver, hoststr) :
5803 hostutil_GetNameByINet(aserver), pname);
5804 MapPartIdIntoName(entry.serverPartition[idx],
5807 " VLDB reports RW volume %lu exists on %s %s\n",
5808 (unsigned long)rwvolid,
5810 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5811 hostutil_GetNameByINet(entry.
5816 /* The RW volume does not exist - have VLDB point to this one */
5819 /* Check for orphaned BK volume on old partition */
5820 if (entry.flags & BACK_EXISTS) {
5822 MapPartIdIntoName(entry.serverPartition[idx],
5825 "*** Warning: Orphaned BK volume %u exists on %s %s\n",
5826 entry.volumeId[BACKVOL],
5828 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5829 hostutil_GetNameByINet(entry.
5832 MapPartIdIntoName(apart, pname);
5834 " VLDB reports its RW volume %lu exists on %s %s\n",
5835 (unsigned long)rwvolid,
5837 afs_inet_ntoa_r(aserver, hoststr) :
5838 hostutil_GetNameByINet(aserver),
5844 /* Volume location matches the VLDB location */
5845 if ((volumeinfo->backupID && !entry.volumeId[BACKVOL])
5846 || (volumeinfo->cloneID && !entry.volumeId[ROVOL])
5849 (entry.name, volumeinfo->name,
5850 VOLSER_OLDMAXVOLNAME) != 0)) {
5858 entry.flags |= RW_EXISTS;
5859 entry.volumeId[RWVOL] = rwvolid;
5860 if (!entry.volumeId[BACKVOL])
5861 entry.volumeId[BACKVOL] = volumeinfo->backupID;
5862 if (!entry.volumeId[ROVOL])
5863 entry.volumeId[ROVOL] = volumeinfo->cloneID;
5865 entry.serverFlags[idx] = ITSRWVOL;
5866 entry.serverNumber[idx] = aserver;
5867 entry.serverPartition[idx] = apart;
5868 strncpy(entry.name, volumeinfo->name, VOLSER_OLDMAXVOLNAME);
5872 /* One last check - to update BK if need to */
5873 code = CheckVldbRWBK(&entry, &mod);
5881 else if (volumeinfo->type == BACKVOL) { /* A BK volume */
5887 /* Check existence of RW and BK volumes */
5888 code = CheckVldbRWBK(&entry, &mod);
5894 idx = Lp_GetRwIndex(&entry);
5895 if (idx == -1) { /* RW index not found in the VLDB entry */
5896 idx = entry.nServers; /* Put it into next index */
5899 } else { /* RW index found in the VLDB entry */
5900 /* Verify if this volume's location matches where the VLDB says it is */
5901 if (!Lp_Match(aserver, apart, &entry)) {
5902 /* VLDB says RW and/or BK is elsewhere - report this BK volume orphaned */
5904 MapPartIdIntoName(apart, pname);
5906 "*** Warning: Orphaned BK volume %lu exists on %s %s\n",
5907 (unsigned long)volumeinfo->volid,
5909 afs_inet_ntoa_r(aserver, hoststr) :
5910 hostutil_GetNameByINet(aserver), pname);
5911 MapPartIdIntoName(entry.serverPartition[idx], pname);
5913 " VLDB reports its RW/BK volume %lu exists on %s %s\n",
5914 (unsigned long)rwvolid,
5916 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5917 hostutil_GetNameByINet(entry.
5922 if (volumeinfo->volid != entry.volumeId[BACKVOL]) {
5923 if (!(entry.flags & BACK_EXISTS)) {
5925 } else if (volumeinfo->volid >
5926 entry.volumeId[BACKVOL]) {
5930 MapPartIdIntoName(entry.serverPartition[idx],
5933 "*** Warning: Orphaned BK volume %u exists on %s %s\n",
5934 entry.volumeId[BACKVOL],
5936 afs_inet_ntoa_r(aserver, hoststr) :
5937 hostutil_GetNameByINet(aserver),
5940 " VLDB reports its BK volume ID is %lu\n",
5941 (unsigned long)volumeinfo->volid);
5945 MapPartIdIntoName(entry.serverPartition[idx],
5948 "*** Warning: Orphaned BK volume %lu exists on %s %s\n",
5949 (unsigned long)volumeinfo->volid,
5951 afs_inet_ntoa_r(aserver, hoststr) :
5952 hostutil_GetNameByINet(aserver),
5955 " VLDB reports its BK volume ID is %u\n",
5956 entry.volumeId[BACKVOL]);
5959 } else if (!entry.volumeId[BACKVOL]) {
5966 entry.flags |= BACK_EXISTS;
5967 entry.volumeId[RWVOL] = rwvolid;
5968 entry.volumeId[BACKVOL] = volumeinfo->volid;
5970 entry.serverNumber[idx] = aserver;
5971 entry.serverPartition[idx] = apart;
5972 entry.serverFlags[idx] = ITSBACKVOL;
5978 else if (volumeinfo->type == ROVOL) { /* A RO volume */
5979 if (volumeinfo->volid == entry.volumeId[ROVOL]) {
5980 /* This is a quick check to see if the RO entry exists in the
5981 * VLDB so we avoid the CheckVldbRO() call (which checks if each
5982 * RO volume listed in the VLDB exists).
5984 idx = Lp_ROMatch(aserver, apart, &entry) - 1;
5986 idx = entry.nServers;
5990 if (!(entry.flags & RO_EXISTS)) {
5995 /* Before we correct the VLDB entry, make sure all the
5996 * ROs listed in the VLDB exist.
5998 code = CheckVldbRO(&entry, &mod);
6004 if (!(entry.flags & RO_EXISTS)) {
6005 /* No RO exists in the VLDB entry - add this one */
6006 idx = entry.nServers;
6009 } else if (volumeinfo->volid > entry.volumeId[ROVOL]) {
6010 /* The volume headers's RO ID does not match that in the VLDB entry,
6011 * and the vol hdr's ID is greater (implies more recent). So delete
6012 * all the RO volumes listed in VLDB entry and add this volume.
6014 for (j = 0; j < entry.nServers; j++) {
6015 if (entry.serverFlags[j] & ITSROVOL) {
6016 /* Verify this volume exists and print message we are orphaning it */
6018 MapPartIdIntoName(apart, pname);
6020 "*** Warning: Orphaned RO volume %u exists on %s %s\n",
6021 entry.volumeId[ROVOL],
6023 afs_inet_ntoa_r(entry.serverNumber[j], hoststr) :
6024 hostutil_GetNameByINet(entry.
6028 " VLDB reports its RO volume ID is %lu\n",
6029 (unsigned long)volumeinfo->volid);
6032 Lp_SetRWValue(&entry, entry.serverNumber[idx],
6033 entry.serverPartition[idx], 0L, 0L);
6040 idx = entry.nServers;
6043 } else if (volumeinfo->volid < entry.volumeId[ROVOL]) {
6044 /* The volume headers's RO ID does not match that in the VLDB entry,
6045 * and the vol hdr's ID is lower (implies its older). So orphan it.
6048 MapPartIdIntoName(apart, pname);
6050 "*** Warning: Orphaned RO volume %lu exists on %s %s\n",
6051 (unsigned long)volumeinfo->volid,
6053 afs_inet_ntoa_r(aserver, hoststr) :
6054 hostutil_GetNameByINet(aserver), pname);
6056 " VLDB reports its RO volume ID is %u\n",
6057 entry.volumeId[ROVOL]);
6060 /* The RO volume ID in the volume header match that in the VLDB entry,
6061 * and there exist RO volumes in the VLDB entry. See if any of them
6062 * are this one. If not, then we add it.
6064 idx = Lp_ROMatch(aserver, apart, &entry) - 1;
6066 idx = entry.nServers;
6074 entry.flags |= RO_EXISTS;
6075 entry.volumeId[RWVOL] = rwvolid;
6076 entry.volumeId[ROVOL] = volumeinfo->volid;
6078 entry.serverNumber[idx] = aserver;
6079 entry.serverPartition[idx] = apart;
6080 entry.serverFlags[idx] = ITSROVOL;
6086 /* Remember largest volume id */
6087 if (entry.volumeId[ROVOL] > *maxvolid)
6088 *maxvolid = entry.volumeId[ROVOL];
6089 if (entry.volumeId[BACKVOL] > *maxvolid)
6090 *maxvolid = entry.volumeId[BACKVOL];
6091 if (entry.volumeId[RWVOL] > *maxvolid)
6092 *maxvolid = entry.volumeId[RWVOL];
6094 if (modified && doit) {
6095 MapNetworkToHost(&entry, &storeEntry);
6098 code = VLDB_CreateEntry(&storeEntry);
6101 "Could not create a VLDB entry for the volume %lu\n",
6102 (unsigned long)rwvolid);
6109 VLDB_ReplaceEntry(rwvolid, RWVOL, &storeEntry,
6110 LOCKREL_OPCODE | LOCKREL_AFSID |
6113 fprintf(STDERR, "Could not update entry for %lu\n",
6114 (unsigned long)rwvolid);
6118 } else if (pass == 2) {
6120 ubik_VL_ReleaseLock(cstruct, 0, rwvolid, RWVOL,
6121 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
6123 PrintError("Could not unlock VLDB entry ", code);
6127 if (modified && modentry) {
6132 memcpy(aentry, &entry, sizeof(entry));
6136 fprintf(STDOUT, "-- status after --\n");
6138 EnumerateEntry(&entry);
6140 fprintf(STDOUT, "\n**no change**\n");
6144 VPRINT("\n_______________________________\n");
6149 sortVolumes(const void *a, const void *b)
6151 volintInfo *v1 = (volintInfo *) a;
6152 volintInfo *v2 = (volintInfo *) b;
6153 afs_uint32 rwvolid1, rwvolid2;
6155 rwvolid1 = ((v1->type == RWVOL) ? v1->volid : v1->parentID);
6156 rwvolid2 = ((v2->type == RWVOL) ? v2->volid : v2->parentID);
6158 if (rwvolid1 > rwvolid2)
6159 return -1; /* lower RW id goes first */
6160 if (rwvolid1 < rwvolid2)
6163 if (v1->type == RWVOL)
6164 return -1; /* RW vols go first */
6165 if (v2->type == RWVOL)
6168 if ((v1->type == BACKVOL) && (v2->type == ROVOL))
6169 return -1; /* BK vols next */
6170 if ((v1->type == ROVOL) && (v2->type == BACKVOL))
6173 if (v1->volid < v2->volid)
6174 return 1; /* larger volids first */
6175 if (v1->volid > v2->volid)
6181 * Synchronise <aserver> <apart>(if flags = 1) <avolid>.
6182 * Synchronize an individual volume against a sever and partition.
6183 * Checks the VLDB entry (similar to syncserv) as well as checks
6184 * if the volume exists on specified servers (similar to syncvldb).
6187 UV_SyncVolume(afs_uint32 aserver, afs_int32 apart, char *avolname, int flags)
6189 struct rx_connection *aconn = 0;
6190 afs_int32 j, k, code, vcode, error = 0;
6192 afs_int32 mod, modified = 0, deleted = 0;
6193 struct nvldbentry vldbentry;
6194 afs_uint32 volumeid = 0;
6195 volEntries volumeInfo;
6196 struct partList PartList;
6198 afs_uint32 maxvolid = 0;
6200 volumeInfo.volEntries_val = (volintInfo *) 0;
6201 volumeInfo.volEntries_len = 0;
6203 /* Turn verbose logging off and do our own verbose logging */
6204 /* tverbose must be set before we call ERROR_EXIT() */
6211 if (!aserver && (flags & 1)) {
6212 /* fprintf(STDERR,"Partition option requires a server option\n"); */
6216 /* Read the VLDB entry */
6217 vcode = VLDB_GetEntryByName(avolname, &vldbentry);
6218 if (vcode && (vcode != VL_NOENT)) {
6219 fprintf(STDERR, "Could not access the VLDB for volume %s\n",
6222 } else if (!vcode) {
6223 MapHostToNetwork(&vldbentry);
6227 fprintf(STDOUT, "Processing VLDB entry %s ...\n", avolname);
6228 fprintf(STDOUT, "_______________________________\n");
6229 fprintf(STDOUT, "\n-- status before -- \n");
6231 fprintf(STDOUT, "\n**does not exist**\n");
6233 if ((vldbentry.flags & RW_EXISTS) || (vldbentry.flags & RO_EXISTS)
6234 || (vldbentry.flags & BACK_EXISTS))
6235 EnumerateEntry(&vldbentry);
6237 fprintf(STDOUT, "\n");
6240 /* Verify that all of the VLDB entries exist on the repective servers
6241 * and partitions (this does not require that avolname be a volume ID).
6242 * Equivalent to a syncserv.
6245 /* Tell CheckVldb not to update if appropriate */
6250 code = CheckVldb(&vldbentry, &mod, &deleted);
6252 fprintf(STDERR, "Could not process VLDB entry for volume %s\n",
6260 /* If aserver is given, we will search for the desired volume on it */
6262 /* Generate array of partitions on the server that we will check */
6264 code = UV_ListPartitions(aserver, &PartList, &pcnt);
6267 "Could not fetch the list of partitions from the server\n");
6271 PartList.partId[0] = apart;
6275 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6277 /* If a volume ID were given, search for it on each partition */
6278 if ((volumeid = atol(avolname))) {
6279 for (j = 0; j < pcnt; j++) {
6281 AFSVolListOneVolume(aconn, PartList.partId[j], volumeid,
6284 if (code != ENODEV) {
6285 fprintf(STDERR, "Could not query server\n");
6293 /* Found one, sync it with VLDB entry */
6295 CheckVolume(volumeInfo.volEntries_val, aserver,
6296 PartList.partId[j], &mod, &maxvolid, &vldbentry);
6303 if (volumeInfo.volEntries_val)
6304 free(volumeInfo.volEntries_val);
6305 volumeInfo.volEntries_val = (volintInfo *) 0;
6306 volumeInfo.volEntries_len = 0;
6310 /* Check to see if the RW, BK, and RO IDs exist on any
6311 * partitions. We get the volume IDs from the VLDB.
6313 for (j = 0; j < MAXTYPES; j++) { /* for RW, RO, and BK IDs */
6314 if (vldbentry.volumeId[j] == 0)
6317 for (k = 0; k < pcnt; k++) { /* For each partition */
6318 volumeInfo.volEntries_val = (volintInfo *) 0;
6319 volumeInfo.volEntries_len = 0;
6321 AFSVolListOneVolume(aconn, PartList.partId[k],
6322 vldbentry.volumeId[j], &volumeInfo);
6324 if (code != ENODEV) {
6325 fprintf(STDERR, "Could not query server\n");
6333 /* Found one, sync it with VLDB entry */
6335 CheckVolume(volumeInfo.volEntries_val, aserver,
6336 PartList.partId[k], &mod, &maxvolid, &vldbentry);
6343 if (volumeInfo.volEntries_val)
6344 free(volumeInfo.volEntries_val);
6345 volumeInfo.volEntries_val = (volintInfo *) 0;
6346 volumeInfo.volEntries_len = 0;
6352 /* If verbose output, print a summary of what changed */
6354 fprintf(STDOUT, "-- status after --\n");
6356 fprintf(STDOUT, "\n**entry deleted**\n");
6357 } else if (modified) {
6358 EnumerateEntry(&vldbentry);
6360 fprintf(STDOUT, "\n**no change**\n");
6362 fprintf(STDOUT, "\n_______________________________\n");
6366 /* Now check if the maxvolid is larger than that stored in the VLDB */
6368 afs_uint32 maxvldbid = 0;
6369 code = ubik_VL_GetNewVolumeId(cstruct, 0, 0, &maxvldbid);
6372 "Could not get the highest allocated volume id from the VLDB\n");
6375 } else if (maxvolid > maxvldbid) {
6377 id = maxvolid - maxvldbid + 1;
6378 code = ubik_VL_GetNewVolumeId(cstruct, 0, id, &nid);
6381 "Error in increasing highest allocated volume id in VLDB\n");
6391 fprintf(STDOUT, "...error encountered");
6393 fprintf(STDOUT, "...done entry\n");
6396 rx_DestroyConnection(aconn);
6397 if (volumeInfo.volEntries_val)
6398 free(volumeInfo.volEntries_val);
6400 PrintError("", error);
6405 * Synchronise vldb with the file server <aserver> and,
6406 * optionally, <apart>.
6409 UV_SyncVldb(afs_uint32 aserver, afs_int32 apart, int flags, int force)
6411 struct rx_connection *aconn;
6412 afs_int32 code, error = 0;
6415 volEntries volumeInfo;
6416 struct partList PartList;
6420 afs_int32 failures = 0, modifications = 0, tentries = 0;
6422 afs_uint32 maxvolid = 0;
6425 volumeInfo.volEntries_val = (volintInfo *) 0;
6426 volumeInfo.volEntries_len = 0;
6428 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6430 /* Generate array of partitions to check */
6432 code = UV_ListPartitions(aserver, &PartList, &pcnt);
6435 "Could not fetch the list of partitions from the server\n");
6439 PartList.partId[0] = apart;
6443 VPRINT("Processing volume entries ...\n");
6445 /* Step through the array of partitions */
6446 for (i = 0; i < pcnt; i++) {
6447 apart = PartList.partId[i];
6448 MapPartIdIntoName(apart, pname);
6450 volumeInfo.volEntries_val = (volintInfo *) 0;
6451 volumeInfo.volEntries_len = 0;
6452 code = AFSVolListVolumes(aconn, apart, 1, &volumeInfo);
6455 "Could not fetch the list of volumes from the server\n");
6459 /* May want to sort the entries: RW, BK (high to low), RO (high to low) */
6460 qsort((char *)volumeInfo.volEntries_val, volumeInfo.volEntries_len,
6461 sizeof(volintInfo), sortVolumes);
6464 for (vi = volumeInfo.volEntries_val, j = 0;
6465 j < volumeInfo.volEntries_len; j++, vi++) {
6473 "Processing volume entry %d: %s (%lu) on server %s %s...\n",
6474 j + 1, vi->name, (unsigned long)vi->volid,
6476 afs_inet_ntoa_r(aserver, hoststr) :
6477 hostutil_GetNameByINet(aserver), pname);
6485 code = CheckVolume(vi, aserver, apart, &modified, &maxvolid, NULL);
6487 PrintError("", code);
6490 } else if (modified) {
6496 fprintf(STDOUT, "...error encountered\n\n");
6498 fprintf(STDOUT, "...done entry %d\n\n", j + 1);
6505 "Could not process entries on server %s partition %s\n",
6507 afs_inet_ntoa_r(aserver, hoststr) :
6508 hostutil_GetNameByINet(aserver), pname);
6510 if (volumeInfo.volEntries_val) {
6511 free(volumeInfo.volEntries_val);
6512 volumeInfo.volEntries_val = 0;
6515 } /* thru all partitions */
6518 VPRINT3("Total entries: %u, Failed to process %d, Would change %d\n",
6519 tentries, failures, modifications);
6521 VPRINT3("Total entries: %u, Failed to process %d, Changed %d\n",
6522 tentries, failures, modifications);
6526 /* Now check if the maxvolid is larger than that stored in the VLDB */
6528 afs_uint32 maxvldbid = 0;
6529 code = ubik_VL_GetNewVolumeId(cstruct, 0, 0, &maxvldbid);
6532 "Could not get the highest allocated volume id from the VLDB\n");
6535 } else if (maxvolid > maxvldbid) {
6537 id = maxvolid - maxvldbid + 1;
6538 code = ubik_VL_GetNewVolumeId(cstruct, 0, id, &nid);
6541 "Error in increasing highest allocated volume id in VLDB\n");
6549 rx_DestroyConnection(aconn);
6550 if (volumeInfo.volEntries_val)
6551 free(volumeInfo.volEntries_val);
6552 PrintError("", error);
6557 * Determine if a volume exists on a server and partition.
6558 * Try creating a transaction on the volume. If we can,
6559 * the volume exists, if not, then return the error code.
6560 * Some error codes mean the volume is unavailable but
6561 * still exists - so we catch these error codes.
6564 VolumeExists(afs_uint32 server, afs_int32 partition, afs_uint32 volumeid)
6566 struct rx_connection *conn = (struct rx_connection *)0;
6567 afs_int32 code = -1;
6568 volEntries volumeInfo;
6570 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
6572 volumeInfo.volEntries_val = (volintInfo *) 0;
6573 volumeInfo.volEntries_len = 0;
6574 code = AFSVolListOneVolume(conn, partition, volumeid, &volumeInfo);
6575 if (volumeInfo.volEntries_val)
6576 free(volumeInfo.volEntries_val);
6577 if (code == VOLSERILLEGAL_PARTITION)
6579 rx_DestroyConnection(conn);
6588 CheckVldbRWBK(struct nvldbentry * entry, afs_int32 * modified)
6592 afs_int32 code, error = 0;
6598 idx = Lp_GetRwIndex(entry);
6600 /* Check to see if the RW volume exists and set the RW_EXISTS
6603 if (idx == -1) { /* Did not find a RW entry */
6604 if (entry->flags & RW_EXISTS) { /* ... yet entry says RW exists */
6605 entry->flags &= ~RW_EXISTS; /* ... so say RW does not exist */
6610 VolumeExists(entry->serverNumber[idx],
6611 entry->serverPartition[idx], entry->volumeId[RWVOL]);
6612 if (code == 0) { /* RW volume exists */
6613 if (!(entry->flags & RW_EXISTS)) { /* ... yet entry says RW does not exist */
6614 entry->flags |= RW_EXISTS; /* ... so say RW does exist */
6617 } else if (code == ENODEV) { /* RW volume does not exist */
6618 if (entry->flags & RW_EXISTS) { /* ... yet entry says RW exists */
6619 entry->flags &= ~RW_EXISTS; /* ... so say RW does not exist */
6623 /* If VLDB says it didn't exist, then ignore error */
6624 if (entry->flags & RW_EXISTS) {
6625 MapPartIdIntoName(entry->serverPartition[idx], pname);
6627 "Transaction call failed for RW volume %u on server %s %s\n",
6628 entry->volumeId[RWVOL],
6630 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6631 hostutil_GetNameByINet(entry->serverNumber[idx]),
6638 /* Check to see if the BK volume exists and set the BACK_EXISTS
6639 * flag accordingly. idx already ponts to the RW entry.
6641 if (idx == -1) { /* Did not find a RW entry */
6642 if (entry->flags & BACK_EXISTS) { /* ... yet entry says BK exists */
6643 entry->flags &= ~BACK_EXISTS; /* ... so say BK does not exist */
6646 } else { /* Found a RW entry */
6648 VolumeExists(entry->serverNumber[idx],
6649 entry->serverPartition[idx],
6650 entry->volumeId[BACKVOL]);
6651 if (code == 0) { /* BK volume exists */
6652 if (!(entry->flags & BACK_EXISTS)) { /* ... yet entry says BK does not exist */
6653 entry->flags |= BACK_EXISTS; /* ... so say BK does exist */
6656 } else if (code == ENODEV) { /* BK volume does not exist */
6657 if (entry->flags & BACK_EXISTS) { /* ... yet entry says BK exists */
6658 entry->flags &= ~BACK_EXISTS; /* ... so say BK does not exist */
6662 /* If VLDB says it didn't exist, then ignore error */
6663 if (entry->flags & BACK_EXISTS) {
6664 MapPartIdIntoName(entry->serverPartition[idx], pname);
6666 "Transaction call failed for BK volume %u on server %s %s\n",
6667 entry->volumeId[BACKVOL],
6669 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6670 hostutil_GetNameByINet(entry->serverNumber[idx]),
6677 /* If there is an idx but the BK and RW volumes no
6678 * longer exist, then remove the RW entry.
6680 if ((idx != -1) && !(entry->flags & RW_EXISTS)
6681 && !(entry->flags & BACK_EXISTS)) {
6682 Lp_SetRWValue(entry, entry->serverNumber[idx],
6683 entry->serverPartition[idx], 0L, 0L);
6690 *modified = modentry;
6695 CheckVldbRO(struct nvldbentry *entry, afs_int32 * modified)
6698 int foundro = 0, modentry = 0;
6699 afs_int32 code, error = 0;
6706 /* Check to see if the RO volumes exist and set the RO_EXISTS
6709 for (idx = 0; idx < entry->nServers; idx++) {
6710 if (!(entry->serverFlags[idx] & ITSROVOL)) {
6711 continue; /* not a RO */
6715 VolumeExists(entry->serverNumber[idx],
6716 entry->serverPartition[idx], entry->volumeId[ROVOL]);
6717 if (code == 0) { /* RO volume exists */
6719 } else if (code == ENODEV) { /* RW volume does not exist */
6720 Lp_SetROValue(entry, entry->serverNumber[idx],
6721 entry->serverPartition[idx], 0L, 0L);
6726 MapPartIdIntoName(entry->serverPartition[idx], pname);
6728 "Transaction call failed for RO %u on server %s %s\n",
6729 entry->volumeId[ROVOL],
6731 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6732 hostutil_GetNameByINet(entry->serverNumber[idx]), pname);
6737 if (foundro) { /* A RO volume exists */
6738 if (!(entry->flags & RO_EXISTS)) { /* ... yet entry says RW does not exist */
6739 entry->flags |= RO_EXISTS; /* ... so say RW does exist */
6742 } else { /* A RO volume does not exist */
6743 if (entry->flags & RO_EXISTS) { /* ... yet entry says RO exists */
6744 entry->flags &= ~RO_EXISTS; /* ... so say RO does not exist */
6751 *modified = modentry;
6756 * Ensure that <entry> matches with the info on file servers
6759 CheckVldb(struct nvldbentry * entry, afs_int32 * modified, afs_int32 * deleted)
6761 afs_int32 code, error = 0;
6762 struct nvldbentry storeEntry;
6763 int islocked = 0, mod, modentry, delentry = 0;
6764 int pass = 0, doit=1;
6772 fprintf(STDOUT, "_______________________________\n");
6773 fprintf(STDOUT, "\n-- status before -- \n");
6774 if ((entry->flags & RW_EXISTS) || (entry->flags & RO_EXISTS)
6775 || (entry->flags & BACK_EXISTS))
6776 EnumerateEntry(entry);
6777 fprintf(STDOUT, "\n");
6780 if (strlen(entry->name) > (VOLSER_OLDMAXVOLNAME - 10)) {
6781 fprintf(STDERR, "Volume name %s exceeds limit of %d characters\n",
6782 entry->name, VOLSER_OLDMAXVOLNAME - 10);
6786 /* Check to see if the VLDB is ok without locking it (pass 1).
6787 * If it will change, then lock the VLDB entry, read it again,
6788 * then make the changes to it (pass 2).
6792 ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL,
6795 fprintf(STDERR, "Could not lock VLDB entry for %u \n",
6796 entry->volumeId[RWVOL]);
6801 code = VLDB_GetEntryByID(entry->volumeId[RWVOL], RWVOL, entry);
6803 fprintf(STDERR, "Could not read VLDB entry for volume %s\n",
6807 MapHostToNetwork(entry);
6813 /* Check if the RW and BK entries are ok */
6814 code = CheckVldbRWBK(entry, &mod);
6817 if (mod && (pass == 1) && doit)
6822 /* Check if the RO volumes entries are ok */
6823 code = CheckVldbRO(entry, &mod);
6826 if (mod && (pass == 1) && doit)
6831 /* The VLDB entry has been updated. If it as been modified, then
6832 * write the entry back out the the VLDB.
6834 if (modentry && doit) {
6838 if (!(entry->flags & RW_EXISTS) && !(entry->flags & BACK_EXISTS)
6839 && !(entry->flags & RO_EXISTS) && doit) {
6840 /* The RW, BK, nor RO volumes do not exist. Delete the VLDB entry */
6842 ubik_VL_DeleteEntry(cstruct, 0, entry->volumeId[RWVOL],
6846 "Could not delete VLDB entry for volume %u \n",
6847 entry->volumeId[RWVOL]);
6852 /* Replace old entry with our new one */
6853 MapNetworkToHost(entry, &storeEntry);
6855 VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry,
6856 (LOCKREL_OPCODE | LOCKREL_AFSID |
6857 LOCKREL_TIMESTAMP));
6859 fprintf(STDERR, "Could not update VLDB entry for volume %u\n",
6860 entry->volumeId[RWVOL]);
6867 if (modified && modentry) {
6870 if (deleted && delentry) {
6875 fprintf(STDOUT, "-- status after --\n");
6877 fprintf(STDOUT, "\n**entry deleted**\n");
6879 EnumerateEntry(entry);
6881 fprintf(STDOUT, "\n**no change**\n");
6885 VPRINT("\n_______________________________\n");
6889 ubik_VL_ReleaseLock(cstruct, 0, entry->volumeId[RWVOL],
6891 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
6894 "Could not release lock on VLDB entry for volume %u\n",
6895 entry->volumeId[RWVOL]);
6904 * Synchronise <aserver> <apart>(if flags = 1) with the VLDB.
6907 UV_SyncServer(afs_uint32 aserver, afs_int32 apart, int flags, int force)
6909 struct rx_connection *aconn;
6910 afs_int32 code, error = 0;
6911 afs_int32 nentries, tentries = 0;
6912 struct VldbListByAttributes attributes;
6913 nbulkentries arrayEntries;
6914 afs_int32 failures = 0, modified, modifications = 0;
6915 struct nvldbentry *vlentry;
6916 afs_int32 si, nsi, j;
6921 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6923 /* Set up attributes to search VLDB */
6924 attributes.server = ntohl(aserver);
6925 attributes.Mask = VLLIST_SERVER;
6927 attributes.partition = apart;
6928 attributes.Mask |= VLLIST_PARTITION;
6931 VPRINT("Processing VLDB entries ...\n");
6933 /* While we need to collect more VLDB entries */
6934 for (si = 0; si != -1; si = nsi) {
6935 memset(&arrayEntries, 0, sizeof(arrayEntries));
6937 /* Collect set of VLDB entries */
6939 VLDB_ListAttributesN2(&attributes, 0, si, &nentries,
6940 &arrayEntries, &nsi);
6941 if (code == RXGEN_OPCODE) {
6942 code = VLDB_ListAttributes(&attributes, &nentries, &arrayEntries);
6946 fprintf(STDERR, "Could not access the VLDB for attributes\n");
6949 tentries += nentries;
6951 for (j = 0; j < nentries; j++) {
6952 vlentry = &arrayEntries.nbulkentries_val[j];
6953 MapHostToNetwork(vlentry);
6955 VPRINT1("Processing VLDB entry %d ...\n", j + 1);
6957 /* Tell CheckVldb not to update if appropriate */
6962 code = CheckVldb(vlentry, &modified, NULL);
6964 PrintError("", code);
6966 "Could not process VLDB entry for volume %s\n",
6969 } else if (modified) {
6975 fprintf(STDOUT, "...error encountered\n\n");
6977 fprintf(STDOUT, "...done entry %d\n\n", j + 1);
6982 if (arrayEntries.nbulkentries_val) {
6983 free(arrayEntries.nbulkentries_val);
6984 arrayEntries.nbulkentries_val = 0;
6989 VPRINT3("Total entries: %u, Failed to process %d, Would change %d\n",
6990 tentries, failures, modifications);
6992 VPRINT3("Total entries: %u, Failed to process %d, Changed %d\n",
6993 tentries, failures, modifications);
6998 rx_DestroyConnection(aconn);
6999 if (arrayEntries.nbulkentries_val)
7000 free(arrayEntries.nbulkentries_val);
7003 error = VOLSERFAILEDOP;
7007 /*rename volume <oldname> to <newname>, changing the names of the related
7008 *readonly and backup volumes. This operation is also idempotent.
7009 *salvager is capable of recovering from rename operation stopping halfway.
7010 *to recover run syncserver on the affected machines,it will force renaming to completion. name clashes should have been detected before calling this proc */
7012 UV_RenameVolume(struct nvldbentry *entry, char oldname[], char newname[])
7014 struct nvldbentry storeEntry;
7015 afs_int32 vcode, code, rcode, error;
7017 char nameBuffer[256];
7019 struct rx_connection *aconn;
7024 aconn = (struct rx_connection *)0;
7028 vcode = ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL, VLOP_ADDSITE); /*last param is dummy */
7031 " Could not lock the VLDB entry for the volume %u \n",
7032 entry->volumeId[RWVOL]);
7037 strncpy(entry->name, newname, VOLSER_OLDMAXVOLNAME);
7038 MapNetworkToHost(entry, &storeEntry);
7039 vcode = VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry, 0);
7041 fprintf(STDERR, "Could not update VLDB entry for %u\n",
7042 entry->volumeId[RWVOL]);
7046 VPRINT1("Recorded the new name %s in VLDB\n", newname);
7047 /*at this stage the intent to rename is recorded in the vldb, as far as the vldb
7048 * is concerned, oldname is lost */
7049 if (entry->flags & RW_EXISTS) {
7050 index = Lp_GetRwIndex(entry);
7051 if (index == -1) { /* there is a serious discrepancy */
7053 "There is a serious discrepancy in VLDB entry for volume %u\n",
7054 entry->volumeId[RWVOL]);
7055 fprintf(STDERR, "try building VLDB from scratch\n");
7056 error = VOLSERVLDB_ERROR;
7059 aconn = UV_Bind(entry->serverNumber[index], AFSCONF_VOLUMEPORT);
7061 AFSVolTransCreate_retry(aconn, entry->volumeId[RWVOL],
7062 entry->serverPartition[index], ITOffline, &tid);
7063 if (code) { /*volume doesnot exist */
7065 "Could not start transaction on the rw volume %u\n",
7066 entry->volumeId[RWVOL]);
7069 } else { /*volume exists, process it */
7072 AFSVolSetIdsTypes(aconn, tid, newname, RWVOL,
7073 entry->volumeId[RWVOL],
7074 entry->volumeId[ROVOL],
7075 entry->volumeId[BACKVOL]);
7077 VPRINT2("Renamed rw volume %s to %s\n", oldname, newname);
7078 code = AFSVolEndTrans(aconn, tid, &rcode);
7082 "Could not end transaction on volume %s %u\n",
7083 entry->name, entry->volumeId[RWVOL]);
7088 fprintf(STDERR, "Could not set parameters on volume %s %u\n",
7089 entry->name, entry->volumeId[RWVOL]);
7095 rx_DestroyConnection(aconn);
7096 aconn = (struct rx_connection *)0;
7098 /*end rw volume processing */
7099 if (entry->flags & BACK_EXISTS) { /*process the backup volume */
7100 index = Lp_GetRwIndex(entry);
7101 if (index == -1) { /* there is a serious discrepancy */
7103 "There is a serious discrepancy in the VLDB entry for the backup volume %u\n",
7104 entry->volumeId[BACKVOL]);
7105 fprintf(STDERR, "try building VLDB from scratch\n");
7106 error = VOLSERVLDB_ERROR;
7109 aconn = UV_Bind(entry->serverNumber[index], AFSCONF_VOLUMEPORT);
7111 AFSVolTransCreate_retry(aconn, entry->volumeId[BACKVOL],
7112 entry->serverPartition[index], ITOffline, &tid);
7113 if (code) { /*volume doesnot exist */
7115 "Could not start transaction on the backup volume %u\n",
7116 entry->volumeId[BACKVOL]);
7119 } else { /*volume exists, process it */
7120 if (strlen(newname) > (VOLSER_OLDMAXVOLNAME - 8)) {
7122 "Volume name %s.backup exceeds the limit of %u characters\n",
7123 newname, VOLSER_OLDMAXVOLNAME);
7127 strcpy(nameBuffer, newname);
7128 strcat(nameBuffer, ".backup");
7131 AFSVolSetIdsTypes(aconn, tid, nameBuffer, BACKVOL,
7132 entry->volumeId[RWVOL], 0, 0);
7134 VPRINT1("Renamed backup volume to %s \n", nameBuffer);
7135 code = AFSVolEndTrans(aconn, tid, &rcode);
7139 "Could not end transaction on the backup volume %u\n",
7140 entry->volumeId[BACKVOL]);
7146 "Could not set parameters on the backup volume %u\n",
7147 entry->volumeId[BACKVOL]);
7152 } /* end backup processing */
7154 rx_DestroyConnection(aconn);
7155 aconn = (struct rx_connection *)0;
7156 if (entry->flags & RO_EXISTS) { /*process the ro volumes */
7157 for (i = 0; i < entry->nServers; i++) {
7158 if (entry->serverFlags[i] & ITSROVOL) {
7159 aconn = UV_Bind(entry->serverNumber[i], AFSCONF_VOLUMEPORT);
7161 AFSVolTransCreate_retry(aconn, entry->volumeId[ROVOL],
7162 entry->serverPartition[i], ITOffline,
7164 if (code) { /*volume doesnot exist */
7166 "Could not start transaction on the ro volume %u\n",
7167 entry->volumeId[ROVOL]);
7170 } else { /*volume exists, process it */
7171 strcpy(nameBuffer, newname);
7172 strcat(nameBuffer, ".readonly");
7173 if (strlen(nameBuffer) > (VOLSER_OLDMAXVOLNAME - 1)) {
7175 "Volume name %s exceeds the limit of %u characters\n",
7176 nameBuffer, VOLSER_OLDMAXVOLNAME);
7181 AFSVolSetIdsTypes(aconn, tid, nameBuffer, ROVOL,
7182 entry->volumeId[RWVOL], 0, 0);
7184 VPRINT2("Renamed RO volume %s on host %s\n",
7187 afs_inet_ntoa_r(entry->serverNumber[i], hoststr) :
7188 hostutil_GetNameByINet(entry->
7190 code = AFSVolEndTrans(aconn, tid, &rcode);
7194 "Could not end transaction on volume %u\n",
7195 entry->volumeId[ROVOL]);
7201 "Could not set parameters on the ro volume %u\n",
7202 entry->volumeId[ROVOL]);
7208 rx_DestroyConnection(aconn);
7209 aconn = (struct rx_connection *)0;
7216 ubik_VL_ReleaseLock(cstruct, 0, entry->volumeId[RWVOL],
7218 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
7221 "Could not unlock the VLDB entry for the volume %s %u\n",
7222 entry->name, entry->volumeId[RWVOL]);
7228 code = AFSVolEndTrans(aconn, tid, &rcode);
7232 fprintf(STDERR, "Failed to end transaction on a volume \n");
7238 rx_DestroyConnection(aconn);
7239 PrintError("", error);
7244 /*report on all the active transactions on volser */
7246 UV_VolserStatus(afs_uint32 server, transDebugInfo ** rpntr, afs_int32 * rcount)
7248 struct rx_connection *aconn;
7249 transDebugEntries transInfo;
7252 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7253 transInfo.transDebugEntries_val = (transDebugInfo *) 0;
7254 transInfo.transDebugEntries_len = 0;
7255 code = AFSVolMonitor(aconn, &transInfo);
7258 "Could not access status information about the server\n");
7259 PrintError("", code);
7260 if (transInfo.transDebugEntries_val)
7261 free(transInfo.transDebugEntries_val);
7263 rx_DestroyConnection(aconn);
7266 *rcount = transInfo.transDebugEntries_len;
7267 *rpntr = transInfo.transDebugEntries_val;
7269 rx_DestroyConnection(aconn);
7276 /*delete the volume without interacting with the vldb */
7278 UV_VolumeZap(afs_uint32 server, afs_int32 part, afs_uint32 volid)
7281 struct rx_connection *aconn;
7283 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7284 error = DoVolDelete(aconn, volid, part,
7285 "the", 0, NULL, NULL);
7287 PrintError("", error);
7289 rx_DestroyConnection(aconn);
7294 UV_SetVolume(afs_uint32 server, afs_int32 partition, afs_uint32 volid,
7295 afs_int32 transflag, afs_int32 setflag, int sleeptime)
7297 struct rx_connection *conn = 0;
7299 afs_int32 code, error = 0, rcode;
7301 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7303 fprintf(STDERR, "SetVolumeStatus: Bind Failed");
7307 code = AFSVolTransCreate_retry(conn, volid, partition, transflag, &tid);
7309 fprintf(STDERR, "SetVolumeStatus: TransCreate Failed\n");
7313 code = AFSVolSetFlags(conn, tid, setflag);
7315 fprintf(STDERR, "SetVolumeStatus: SetFlags Failed\n");
7320 #ifdef AFS_PTHREAD_ENV
7323 IOMGR_Sleep(sleeptime);
7330 code = AFSVolEndTrans(conn, tid, &rcode);
7331 if (code || rcode) {
7332 fprintf(STDERR, "SetVolumeStatus: EndTrans Failed\n");
7334 error = (code ? code : rcode);
7339 rx_DestroyConnection(conn);
7344 UV_SetVolumeInfo(afs_uint32 server, afs_int32 partition, afs_uint32 volid,
7347 struct rx_connection *conn = 0;
7349 afs_int32 code, error = 0, rcode;
7351 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7353 fprintf(STDERR, "SetVolumeInfo: Bind Failed");
7357 code = AFSVolTransCreate_retry(conn, volid, partition, ITOffline, &tid);
7359 fprintf(STDERR, "SetVolumeInfo: TransCreate Failed\n");
7363 code = AFSVolSetInfo(conn, tid, infop);
7365 fprintf(STDERR, "SetVolumeInfo: SetInfo Failed\n");
7372 code = AFSVolEndTrans(conn, tid, &rcode);
7373 if (code || rcode) {
7374 fprintf(STDERR, "SetVolumeInfo: EndTrans Failed\n");
7376 error = (code ? code : rcode);
7381 rx_DestroyConnection(conn);
7386 UV_GetSize(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
7387 afs_int32 fromdate, struct volintSize *vol_size)
7389 struct rx_connection *aconn = (struct rx_connection *)0;
7390 afs_int32 tid = 0, rcode = 0;
7391 afs_int32 code, error = 0;
7394 /* get connections to the servers */
7395 aconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
7397 VPRINT1("Starting transaction on volume %u...", afromvol);
7398 code = AFSVolTransCreate_retry(aconn, afromvol, afrompart, ITBusy, &tid);
7399 EGOTO1(error_exit, code,
7400 "Could not start transaction on the volume %u to be measured\n",
7404 VPRINT1("Getting size of volume on volume %u...", afromvol);
7405 code = AFSVolGetSize(aconn, tid, fromdate, vol_size);
7406 EGOTO(error_exit, code, "Could not start the measurement process \n");
7411 VPRINT1("Ending transaction on volume %u...", afromvol);
7412 code = AFSVolEndTrans(aconn, tid, &rcode);
7413 if (code || rcode) {
7414 fprintf(STDERR, "Could not end transaction on the volume %u\n",
7416 fprintf(STDERR, "error codes: %d and %d\n", code, rcode);
7418 error = (code ? code : rcode);
7423 rx_DestroyConnection(aconn);
7425 PrintError("", error);
7429 /*maps the host addresses in <old > (present in network byte order) to
7430 that in< new> (present in host byte order )*/
7432 MapNetworkToHost(struct nvldbentry *old, struct nvldbentry *new)
7436 /*copy all the fields */
7437 strcpy(new->name, old->name);
7438 /* new->volumeType = old->volumeType;*/
7439 new->nServers = old->nServers;
7440 count = old->nServers;
7441 if (count < NMAXNSERVERS)
7443 for (i = 0; i < count; i++) {
7444 new->serverNumber[i] = ntohl(old->serverNumber[i]);
7445 new->serverPartition[i] = old->serverPartition[i];
7446 new->serverFlags[i] = old->serverFlags[i];
7448 new->volumeId[RWVOL] = old->volumeId[RWVOL];
7449 new->volumeId[ROVOL] = old->volumeId[ROVOL];
7450 new->volumeId[BACKVOL] = old->volumeId[BACKVOL];
7451 new->cloneId = old->cloneId;
7452 new->flags = old->flags;
7455 /*maps the host entries in <entry> which are present in host byte order to network byte order */
7457 MapHostToNetwork(struct nvldbentry *entry)
7461 count = entry->nServers;
7462 if (count < NMAXNSERVERS)
7464 for (i = 0; i < count; i++) {
7465 entry->serverNumber[i] = htonl(entry->serverNumber[i]);