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 /* Convert volume from RO to RW; adjust the VLDB entry to match.
1342 * The nvldbentry passed to us has already been MapHostToNetwork'd
1347 UV_ConvertRO(afs_uint32 server, afs_uint32 partition, afs_uint32 volid,
1348 struct nvldbentry *entry)
1350 afs_int32 code, i, same;
1351 struct nvldbentry checkEntry, storeEntry;
1353 afs_int32 rwindex = 0;
1354 afs_uint32 rwserver = 0;
1355 afs_int32 roindex = 0;
1356 afs_uint32 roserver = 0;
1357 struct rx_connection *aconn;
1360 ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL,
1364 "Unable to lock volume %lu, code %d\n",
1365 (unsigned long)entry->volumeId[RWVOL],vcode);
1366 PrintError("", vcode);
1370 /* make sure the VLDB entry hasn't changed since we started */
1371 memset(&checkEntry, 0, sizeof(checkEntry));
1372 vcode = VLDB_GetEntryByID(volid, -1, &checkEntry);
1375 "Could not fetch the entry for volume %lu from VLDB\n",
1376 (unsigned long)volid);
1377 PrintError("convertROtoRW ", vcode);
1382 MapHostToNetwork(&checkEntry);
1383 entry->flags &= ~VLOP_ALLOPERS; /* clear any stale lock operation flags */
1384 entry->flags |= VLOP_MOVE; /* set to match SetLock operation above */
1385 if (memcmp(entry, &checkEntry, sizeof(*entry)) != 0) {
1387 "VLDB entry for volume %lu has changed; please reissue the command.\n",
1388 (unsigned long)volid);
1393 /* extract information from the original entry */
1394 for (i = 0; i < entry->nServers; i++) {
1395 if (entry->serverFlags[i] & ITSRWVOL) {
1397 rwserver = entry->serverNumber[i];
1398 /* rwpartition = entry->serverPartition[i]; */
1401 } else if ((entry->serverFlags[i] & ITSROVOL) && !roserver) {
1402 same = VLDB_IsSameAddrs(server, entry->serverNumber[i], &code);
1405 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
1412 roserver = entry->serverNumber[i];
1413 /* ropartition = entry->serverPartition[i]; */
1420 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
1421 code = AFSVolConvertROtoRWvolume(aconn, partition, volid);
1424 "Converting RO volume %lu to RW volume failed with code %d\n",
1425 (unsigned long)volid, code);
1426 PrintError("convertROtoRW ", code);
1429 /* Update the VLDB to match what we did on disk as much as possible. */
1430 /* If the converted RO was in the VLDB, make it look like the new RW. */
1432 entry->serverFlags[roindex] = ITSRWVOL;
1434 /* Add a new site entry for the newly created RW. It's possible
1435 * (but unlikely) that we are already at MAXNSERVERS and that this
1436 * new site will invalidate the whole VLDB entry; however,
1437 * VLDB_ReplaceEntry will detect this and return VL_BADSERVER,
1438 * so we need no extra guard logic here.
1440 afs_int32 newrwindex = entry->nServers;
1441 (entry->nServers)++;
1442 entry->serverNumber[newrwindex] = server;
1443 entry->serverPartition[newrwindex] = partition;
1444 entry->serverFlags[newrwindex] = ITSRWVOL;
1446 entry->flags |= RW_EXISTS;
1447 entry->flags &= ~BACK_EXISTS;
1449 /* if the old RW was in the VLDB, remove it by decrementing the number */
1450 /* of servers, replacing the RW entry with the last entry, and zeroing */
1451 /* out the last entry. */
1453 (entry->nServers)--;
1454 if (rwindex != entry->nServers) {
1455 entry->serverNumber[rwindex] = entry->serverNumber[entry->nServers];
1456 entry->serverPartition[rwindex] =
1457 entry->serverPartition[entry->nServers];
1458 entry->serverFlags[rwindex] = entry->serverFlags[entry->nServers];
1459 entry->serverNumber[entry->nServers] = 0;
1460 entry->serverPartition[entry->nServers] = 0;
1461 entry->serverFlags[entry->nServers] = 0;
1464 entry->flags &= ~RO_EXISTS;
1465 for (i = 0; i < entry->nServers; i++) {
1466 if (entry->serverFlags[i] & ITSROVOL) {
1467 if (!(entry->serverFlags[i] & (RO_DONTUSE | NEW_REPSITE)))
1468 entry->flags |= RO_EXISTS;
1471 MapNetworkToHost(entry, &storeEntry);
1473 VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry,
1474 (LOCKREL_OPCODE | LOCKREL_AFSID |
1475 LOCKREL_TIMESTAMP));
1478 "Warning: volume converted, but vldb update failed with code %d!\n",
1483 vcode = UV_LockRelease(entry->volumeId[RWVOL]);
1486 "Unable to unlock volume %lu, code %d\n",
1487 (unsigned long)entry->volumeId[RWVOL],vcode);
1488 PrintError("", vcode);
1494 /* Move volume <afromvol> on <afromserver> <afrompart> to <atoserver>
1495 * <atopart>. The operation is almost idempotent. The following
1496 * flags are recognized:
1498 * RV_NOCLONE - don't use a copy clone
1502 UV_MoveVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
1503 afs_uint32 atoserver, afs_int32 atopart, int flags)
1505 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
1506 * be changing during the move */
1507 struct rx_connection * volatile toconn;
1508 struct rx_connection * volatile fromconn;
1509 afs_int32 volatile fromtid;
1510 afs_int32 volatile totid;
1511 afs_int32 volatile clonetid;
1512 afs_uint32 volatile newVol;
1513 afs_uint32 volatile volid;
1514 afs_uint32 volatile backupId;
1515 int volatile islocked;
1520 char tmpName[VOLSER_MAXVOLNAME + 1];
1525 struct restoreCookie cookie;
1526 afs_int32 vcode, code;
1527 struct volser_status tstatus;
1528 struct destServer destination;
1530 struct nvldbentry entry, storeEntry;
1533 char in, lf; /* for test code */
1537 #ifdef ENABLE_BUGFIX_1165
1538 volEntries volumeInfo;
1539 struct volintInfo *infop = 0;
1543 fromconn = (struct rx_connection *)0;
1544 toconn = (struct rx_connection *)0;
1554 /* support control-c processing */
1557 (void)signal(SIGINT, sigint_handler);
1561 "\nThere are three tests points - verifies all code paths through recovery.\n");
1562 fprintf(STDOUT, "First test point - operation not started.\n");
1563 fprintf(STDOUT, "...test here (y, n)? ");
1565 fscanf(stdin, "%c", &in);
1566 fscanf(stdin, "%c", &lf); /* toss away */
1568 fprintf(STDOUT, "type control-c\n");
1570 fprintf(stdout, ".");
1575 /* or drop through */
1578 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
1579 EGOTO1(mfail, vcode,
1580 "Could not fetch the entry for the volume %u from the VLDB \n",
1583 if (entry.volumeId[RWVOL] != afromvol) {
1584 fprintf(STDERR, "Only RW volume can be moved\n");
1588 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_MOVE);
1589 EGOTO1(mfail, vcode, "Could not lock entry for volume %u \n", afromvol);
1592 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
1593 EGOTO1(mfail, vcode,
1594 "Could not fetch the entry for the volume %u from the VLDB \n",
1597 backupId = entry.volumeId[BACKVOL];
1598 MapHostToNetwork(&entry);
1600 if (!Lp_Match(afromserver, afrompart, &entry)) {
1601 /* the from server and partition do not exist in the vldb entry corresponding to volid */
1602 if (!Lp_Match(atoserver, atopart, &entry)) {
1603 /* the to server and partition do not exist in the vldb entry corresponding to volid */
1604 fprintf(STDERR, "The volume %lu is not on the specified site. \n",
1605 (unsigned long)afromvol);
1606 fprintf(STDERR, "The current site is :");
1607 for (i = 0; i < entry.nServers; i++) {
1608 if (entry.serverFlags[i] == ITSRWVOL) {
1610 MapPartIdIntoName(entry.serverPartition[i], pname);
1611 fprintf(STDERR, " server %s partition %s \n",
1612 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
1613 hostutil_GetNameByINet(entry.serverNumber[i]),
1618 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1619 (LOCKREL_OPCODE | LOCKREL_AFSID |
1620 LOCKREL_TIMESTAMP));
1621 EGOTO1(mfail, vcode,
1622 " Could not release lock on the VLDB entry for the volume %u \n",
1625 return VOLSERVOLMOVED;
1628 /* delete the volume afromvol on src_server */
1629 /* from-info does not exist but to-info does =>
1630 * we have already done the move, but the volume
1631 * may still be existing physically on from fileserver
1633 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1636 code = DoVolDelete(fromconn, afromvol, afrompart,
1637 "leftover", 0, NULL, NULL);
1638 if (code && code != VNOVOL) {
1643 code = DoVolDelete(fromconn, backupId, afrompart,
1644 "leftover backup", 0, NULL, NULL);
1645 if (code && code != VNOVOL) {
1655 /* From-info matches the vldb info about volid,
1656 * its ok start the move operation, the backup volume
1657 * on the old site is deleted in the process
1659 if (afrompart == atopart) {
1660 same = VLDB_IsSameAddrs(afromserver, atoserver, &error);
1661 EGOTO2(mfail, error,
1662 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
1663 afromserver, error);
1666 EGOTO1(mfail, VOLSERVOLMOVED,
1667 "Warning: Moving volume %u to its home partition ignored!\n",
1673 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
1674 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1675 totid = 0; /* initialize to uncreated */
1678 * clone the read/write volume locally.
1681 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1682 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1684 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
1688 if (!(flags & RV_NOCLONE)) {
1689 /* Get a clone id */
1690 VPRINT1("Allocating new volume id for clone of volume %u ...",
1693 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
1695 EGOTO1(mfail, vcode,
1696 "Could not get an ID for the clone of volume %u from the VLDB\n",
1700 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
1701 VPRINT1("Cloning source volume %u ...", afromvol);
1702 strcpy(vname, "move-clone-temp");
1704 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname, &tmpVol);
1706 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
1711 /* lookup the name of the volume we just cloned */
1713 code = AFSVolGetName(fromconn, fromtid, &volName);
1714 EGOTO1(mfail, code, "Failed to get the name of the volume %u\n",
1717 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
1719 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1724 "Failed to end the transaction on the source volume %u\n",
1729 * Create the destination volume
1732 if (!(flags & RV_NOCLONE)) {
1733 /* All of this is to get the fromDate */
1734 VPRINT1("Starting transaction on the cloned volume %u ...", newVol);
1737 AFSVolTransCreate_retry(fromconn, newVol, afrompart, ITOffline,
1741 "Failed to start a transaction on the cloned volume%u\n",
1745 VPRINT1("Setting flags on cloned volume %u ...", newVol);
1747 AFSVolSetFlags(fromconn, clonetid,
1748 VTDeleteOnSalvage | VTOutOfService); /*redundant */
1749 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
1753 /* remember time from which we've dumped the volume */
1754 VPRINT1("Getting status of cloned volume %u ...", newVol);
1755 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
1757 "Failed to get the status of the cloned volume %u\n",
1761 fromDate = CLOCKADJ(tstatus.creationDate);
1763 /* With RV_NOCLONE, just do a full copy from the source */
1768 #ifdef ENABLE_BUGFIX_1165
1770 * Get the internal volume state from the source volume. We'll use such info (i.e. dayUse)
1771 * to copy it to the new volume (via AFSSetInfo later on) so that when we move volumes we
1772 * don't use this information...
1774 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
1775 volumeInfo.volEntries_len = 0;
1776 code = AFSVolListOneVolume(fromconn, afrompart, afromvol, &volumeInfo);
1778 "Failed to get the volint Info of the cloned volume %u\n",
1781 infop = (volintInfo *) volumeInfo.volEntries_val;
1782 infop->maxquota = -1; /* Else it will replace the default quota */
1783 infop->creationDate = -1; /* Else it will use the source creation date */
1784 infop->updateDate = -1; /* Else it will use the source update date */
1787 /* create a volume on the target machine */
1789 code = DoVolDelete(toconn, volid, atopart,
1790 "pre-existing destination", 0, NULL, NULL);
1791 if (code && code != VNOVOL) {
1796 VPRINT1("Creating the destination volume %u ...", volid);
1800 AFSVolCreateVolume(toconn, atopart, volName, volser_RW, volid, &tmpVol,
1804 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
1808 strncpy(tmpName, volName, VOLSER_OLDMAXVOLNAME);
1812 VPRINT1("Setting volume flags on destination volume %u ...", volid);
1814 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
1816 "Failed to set the flags on the destination volume %u\n", volid);
1820 * Now dump the clone to the new volume
1823 destination.destHost = ntohl(atoserver);
1824 destination.destPort = AFSCONF_VOLUMEPORT;
1825 destination.destSSID = 1;
1827 strncpy(cookie.name, tmpName, VOLSER_OLDMAXVOLNAME);
1828 cookie.type = RWVOL;
1829 cookie.parent = entry.volumeId[RWVOL];
1832 if (!(flags & RV_NOCLONE)) {
1833 /* Copy the clone to the new volume */
1834 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
1837 AFSVolForward(fromconn, clonetid, 0, &destination, totid,
1839 EGOTO1(mfail, code, "Failed to move data for the volume %u\n", volid);
1842 VPRINT1("Ending transaction on cloned volume %u ...", newVol);
1843 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
1848 "Failed to end the transaction on the cloned volume %u\n",
1854 * reattach to the main-line volume, and incrementally dump it.
1857 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1859 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1862 "Failed to create a transaction on the source volume %u\n",
1866 /* now do the incremental */
1868 ("Doing the%s dump from source to destination for volume %u ... ",
1869 (flags & RV_NOCLONE) ? "" : " incremental",
1872 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
1875 "Failed to do the%s dump from rw volume on old site to rw volume on newsite\n",
1876 (flags & RV_NOCLONE) ? "" : " incremental");
1879 /* now adjust the flags so that the new volume becomes official */
1880 VPRINT1("Setting volume flags on old source volume %u ...", afromvol);
1881 code = AFSVolSetFlags(fromconn, fromtid, VTOutOfService);
1883 "Failed to set the flags to make old source volume offline\n");
1886 VPRINT1("Setting volume flags on new source volume %u ...", afromvol);
1887 code = AFSVolSetFlags(toconn, totid, 0);
1889 "Failed to set the flags to make new source volume online\n");
1892 #ifdef ENABLE_BUGFIX_1165
1893 VPRINT1("Setting volume status on destination volume %u ...", volid);
1894 code = AFSVolSetInfo(toconn, totid, infop);
1896 "Failed to set volume status on the destination volume %u\n",
1901 /* put new volume online */
1902 VPRINT1("Ending transaction on destination volume %u ...", afromvol);
1903 code = AFSVolEndTrans(toconn, totid, &rcode);
1908 "Failed to end the transaction on the volume %u on the new site\n",
1912 Lp_SetRWValue(&entry, afromserver, afrompart, atoserver, atopart);
1913 MapNetworkToHost(&entry, &storeEntry);
1914 storeEntry.flags &= ~BACK_EXISTS;
1918 "Second test point - operation in progress but not complete.\n");
1919 fprintf(STDOUT, "...test here (y, n)? ");
1921 fscanf(stdin, "%c", &in);
1922 fscanf(stdin, "%c", &lf); /* toss away */
1924 fprintf(STDOUT, "type control-c\n");
1926 fprintf(stdout, ".");
1931 /* or drop through */
1934 VPRINT1("Releasing lock on VLDB entry for volume %u ...", afromvol);
1936 VLDB_ReplaceEntry(afromvol, -1, &storeEntry,
1937 (LOCKREL_OPCODE | LOCKREL_AFSID |
1938 LOCKREL_TIMESTAMP));
1941 " Could not release the lock on the VLDB entry for the volume %s %lu \n",
1942 storeEntry.name, (unsigned long)afromvol);
1951 "Third test point - operation complete but no cleanup.\n");
1952 fprintf(STDOUT, "...test here (y, n)? ");
1954 fscanf(stdin, "%c", &in);
1955 fscanf(stdin, "%c", &lf); /* toss away */
1957 fprintf(STDOUT, "type control-c\n");
1959 fprintf(stdout, ".");
1964 /* or drop through */
1967 /* This is tricky. File server is very stupid, and if you mark the volume
1968 * as VTOutOfService, it may mark the *good* instance (if you're moving
1969 * between partitions on the same machine) as out of service. Since
1970 * we're cleaning this code up in DEcorum, we're just going to kludge around
1971 * it for now by removing this call. */
1972 /* already out of service, just zap it now */
1974 AFSVolSetFlags(fromconn, fromtid, VTDeleteOnSalvage | VTOutOfService);
1977 "Failed to set the flags to make the old source volume offline\n");
1981 if (atoserver != afromserver) {
1982 /* set forwarding pointer for moved volumes */
1983 VPRINT1("Setting forwarding pointer for volume %u ...", afromvol);
1984 code = AFSVolSetForwarding(fromconn, fromtid, atoserver);
1986 "Failed to set the forwarding pointer for the volume %u\n",
1991 VPRINT1("Deleting old volume %u on source ...", afromvol);
1992 code = AFSVolDeleteVolume(fromconn, fromtid); /* zap original volume */
1993 EGOTO1(mfail, code, "Failed to delete the old volume %u on source\n",
1997 VPRINT1("Ending transaction on old volume %u on the source ...",
1999 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2004 "Failed to end the transaction on the old volume %u on the source\n",
2008 code = DoVolDelete(fromconn, backupId, afrompart,
2009 "source backup", 0, NULL, NULL);
2010 if (code && code != VNOVOL) {
2015 code = 0; /* no backup volume? that's okay */
2018 if (!(flags & RV_NOCLONE)) {
2019 code = DoVolDelete(fromconn, newVol, afrompart,
2020 "cloned", 0, NULL, NULL);
2022 if (code == VNOVOL) {
2023 EPRINT1(code, "Failed to start transaction on %u\n", newVol);
2034 fprintf(STDOUT, "Fourth test point - operation complete.\n");
2035 fprintf(STDOUT, "...test here (y, n)? ");
2037 fscanf(stdin, "%c", &in);
2038 fscanf(stdin, "%c", &lf); /* toss away */
2040 fprintf(STDOUT, "type control-c\n");
2042 fprintf(stdout, ".");
2047 /* or drop through */
2050 /* normal cleanup code */
2052 if (entry.flags & RO_EXISTS)
2053 fprintf(STDERR, "WARNING : readOnly copies still exist \n");
2056 VPRINT1("Cleanup: Releasing VLDB lock on volume %u ...", afromvol);
2058 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2059 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2063 " Could not release the lock on the VLDB entry for the volume %lu \n",
2064 (unsigned long)afromvol);
2072 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
2074 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2075 if (code || rcode) {
2078 "Could not end transaction on the source volume %lu\n",
2079 (unsigned long)afromvol);
2081 error = (code ? code : rcode);
2087 VPRINT1("Cleanup: Ending transaction on clone volume %u ...", newVol);
2088 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2089 if (code || rcode) {
2092 "Could not end transaction on the source's clone volume %lu\n",
2093 (unsigned long)newVol);
2095 error = (code ? code : rcode);
2101 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
2103 code = AFSVolEndTrans(toconn, totid, &rcode);
2107 "Could not end transaction on destination volume %lu\n",
2108 (unsigned long)afromvol);
2110 error = (code ? code : rcode);
2116 #ifdef ENABLE_BUGFIX_1165
2121 rx_DestroyConnection(fromconn);
2123 rx_DestroyConnection(toconn);
2124 PrintError("", error);
2127 /* come here only when the sky falls */
2132 "vos move: operation interrupted, cleanup in progress...\n");
2133 fprintf(STDOUT, "clear transaction contexts\n");
2137 /* unlock VLDB entry */
2139 VPRINT1("Recovery: Releasing VLDB lock on volume %u ...", afromvol);
2140 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2141 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2147 VPRINT("Recovery: Ending transaction on clone volume ...");
2148 AFSVolEndTrans(fromconn, clonetid, &rcode);
2152 VPRINT("Recovery: Ending transaction on destination volume ...");
2153 AFSVolEndTrans(toconn, totid, &rcode);
2156 if (fromtid) { /* put it on-line */
2157 VPRINT("Recovery: Setting volume flags on source volume ...");
2158 AFSVolSetFlags(fromconn, fromtid, 0);
2161 VPRINT("Recovery: Ending transaction on source volume ...");
2162 AFSVolEndTrans(fromconn, fromtid, &rcode);
2166 VPRINT("Recovery: Accessing VLDB.\n");
2167 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2169 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2173 MapHostToNetwork(&entry);
2175 /* Delete either the volume on the source location or the target location.
2176 * If the vldb entry still points to the source location, then we know the
2177 * volume move didn't finish so we remove the volume from the target
2178 * location. Otherwise, we remove the volume from the source location.
2180 if (Lp_Match(afromserver, afrompart, &entry)) { /* didn't move - delete target volume */
2183 "move incomplete - attempt cleanup of target partition - no guarantee\n");
2187 if (volid && toconn) {
2188 code = DoVolDelete(toconn, volid, atopart,
2189 "destination", 0, NULL, "Recovery:");
2190 if (code == VNOVOL) {
2191 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", volid);
2195 /* put source volume on-line */
2197 VPRINT1("Recovery: Creating transaction on source volume %u ...",
2201 AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2207 VPRINT1("Recovery: Setting flags on source volume %u ...",
2209 AFSVolSetFlags(fromconn, fromtid, 0);
2213 ("Recovery: Ending transaction on source volume %u ...",
2215 AFSVolEndTrans(fromconn, fromtid, &rcode);
2219 ("\nRecovery: Unable to start transaction on source volume %u.\n",
2223 } else { /* yep, move complete */
2226 "move complete - attempt cleanup of source partition - no guarantee\n");
2230 /* delete backup volume */
2232 code = DoVolDelete(fromconn, backupId, afrompart,
2233 "backup", 0, NULL, "Recovery:");
2234 if (code == VNOVOL) {
2235 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", backupId);
2238 code = DoVolDelete(fromconn, afromvol, afrompart, "source",
2239 (atoserver != afromserver)?atoserver:0,
2241 if (code == VNOVOL) {
2242 EPRINT1(code, "Failed to start transaction on %u\n", afromvol);
2247 /* common cleanup - delete local clone */
2249 code = DoVolDelete(fromconn, newVol, afrompart,
2250 "clone", 0, NULL, "Recovery:");
2251 if (code == VNOVOL) {
2252 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", newVol);
2256 /* unlock VLDB entry */
2258 VPRINT1("Recovery: Releasing lock on VLDB entry for volume %u ...",
2260 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2261 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2264 done: /* routine cleanup */
2267 #ifdef ENABLE_BUGFIX_1165
2272 rx_DestroyConnection(fromconn);
2274 rx_DestroyConnection(toconn);
2277 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2285 UV_MoveVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2286 afs_uint32 atoserver, afs_int32 atopart)
2288 return UV_MoveVolume2(afromvol, afromserver, afrompart,
2289 atoserver, atopart, 0);
2293 /* Copy volume <afromvol> from <afromserver> <afrompart> to <atoserver>
2294 * <atopart>. The new volume is named by <atovolname>. The new volume
2295 * has ID <atovolid> if that is nonzero; otherwise a new ID is allocated
2296 * from the VLDB. the following flags are supported:
2298 * RV_RDONLY - target volume is RO
2299 * RV_OFFLINE - leave target volume offline
2300 * RV_CPINCR - do incremental dump if target exists
2301 * RV_NOVLDB - don't create/update VLDB entry
2302 * RV_NOCLONE - don't use a copy clone
2305 UV_CopyVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2306 char *atovolname, afs_uint32 atoserver, afs_int32 atopart,
2307 afs_uint32 atovolid, int flags)
2309 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
2310 * be changing during the copy */
2312 afs_int32 volatile clonetid;
2313 afs_int32 volatile totid;
2314 afs_int32 volatile fromtid;
2315 struct rx_connection * volatile fromconn;
2316 struct rx_connection * volatile toconn;
2317 afs_uint32 volatile cloneVol;
2321 afs_int32 fromDate, cloneFromDate;
2322 struct restoreCookie cookie;
2323 afs_int32 vcode, code;
2326 struct volser_status tstatus;
2327 struct destServer destination;
2328 struct nvldbentry entry, newentry, storeEntry;
2333 fromconn = (struct rx_connection *)0;
2334 toconn = (struct rx_connection *)0;
2342 /* support control-c processing */
2345 (void)signal(SIGINT, sigint_handler);
2347 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2348 EGOTO1(mfail, vcode,
2349 "Could not fetch the entry for the volume %u from the VLDB \n",
2351 MapHostToNetwork(&entry);
2354 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
2355 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
2356 fromtid = totid = 0; /* initialize to uncreated */
2359 * clone the read/write volume locally.
2363 if (!(flags & RV_NOCLONE)) {
2364 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2366 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2369 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
2373 /* Get a clone id */
2374 VPRINT1("Allocating new volume id for clone of volume %u ...",
2378 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
2380 EGOTO1(mfail, vcode,
2381 "Could not get an ID for the clone of volume %u from the VLDB\n",
2389 /* Get a new volume id */
2390 VPRINT1("Allocating new volume id for copy of volume %u ...", afromvol);
2392 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &newVol);
2393 EGOTO1(mfail, vcode,
2394 "Could not get an ID for the copy of volume %u from the VLDB\n",
2399 if (!(flags & RV_NOCLONE)) {
2400 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
2401 VPRINT1("Cloning source volume %u ...", afromvol);
2402 strcpy(vname, "copy-clone-temp");
2405 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname,
2408 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
2412 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
2414 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2419 "Failed to end the transaction on the source volume %u\n",
2425 * Create the destination volume
2428 if (!(flags & RV_NOCLONE)) {
2429 VPRINT1("Starting transaction on the cloned volume %u ...", cloneVol);
2432 AFSVolTransCreate_retry(fromconn, cloneVol, afrompart, ITOffline,
2436 "Failed to start a transaction on the cloned volume%u\n",
2440 VPRINT1("Setting flags on cloned volume %u ...", cloneVol);
2442 AFSVolSetFlags(fromconn, clonetid,
2443 VTDeleteOnSalvage | VTOutOfService); /*redundant */
2444 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
2448 /* remember time from which we've dumped the volume */
2449 VPRINT1("Getting status of cloned volume %u ...", cloneVol);
2450 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
2452 "Failed to get the status of the cloned volume %u\n",
2456 fromDate = CLOCKADJ(tstatus.creationDate);
2461 /* create a volume on the target machine */
2464 code = AFSVolTransCreate_retry(toconn, newVol, atopart, ITOffline, &tmp);
2467 if ((flags & RV_CPINCR)) {
2468 VPRINT1("Getting status of pre-existing volume %u ...", newVol);
2469 code = AFSVolGetStatus(toconn, totid, &tstatus);
2471 "Failed to get the status of the pre-existing volume %u\n",
2475 /* Using the update date should be OK here, but add some fudge */
2476 cloneFromDate = CLOCKADJ(tstatus.updateDate);
2477 if ((flags & RV_NOCLONE))
2478 fromDate = cloneFromDate;
2480 /* XXX We should check that the source volume's creationDate is
2481 * XXX not newer than the existing target volume, and if not,
2482 * XXX throw away the existing target and do a full dump. */
2487 /* Delete the existing volume.
2488 * While we are deleting the volume in these steps, the transaction
2489 * we started against the cloned volume (clonetid above) will be
2490 * sitting idle. It will get cleaned up after 600 seconds
2492 VPRINT1("Deleting pre-existing volume %u on destination ...", newVol);
2493 code = AFSVolDeleteVolume(toconn, totid);
2495 "Could not delete the pre-existing volume %u on destination\n",
2500 ("Ending transaction on pre-existing volume %u on destination ...",
2502 code = AFSVolEndTrans(toconn, totid, &rcode);
2507 "Could not end the transaction on pre-existing volume %u on destination\n",
2512 VPRINT1("Creating the destination volume %u ...", newVol);
2515 AFSVolCreateVolume(toconn, atopart, atovolname,
2516 (flags & RV_RDONLY) ? volser_RO : volser_RW,
2517 newVol, &newVol, &tmp);
2519 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
2523 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2525 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
2527 "Failed to set the flags on the destination volume %u\n", newVol);
2532 destination.destHost = ntohl(atoserver);
2533 destination.destPort = AFSCONF_VOLUMEPORT;
2534 destination.destSSID = 1;
2536 strncpy(cookie.name, atovolname, VOLSER_OLDMAXVOLNAME);
2537 cookie.type = (flags & RV_RDONLY) ? ROVOL : RWVOL;
2542 * Now dump the clone to the new volume
2545 if (!(flags & RV_NOCLONE)) {
2546 /* XXX probably should have some code here that checks to see if
2547 * XXX we are copying to same server and partition - if so, just
2548 * XXX use a clone to save disk space */
2550 /* Copy the clone to the new volume */
2551 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
2554 AFSVolForward(fromconn, clonetid, cloneFromDate, &destination,
2556 EGOTO1(mfail, code, "Failed to move data for the volume %u\n",
2560 VPRINT1("Ending transaction on cloned volume %u ...", cloneVol);
2561 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2566 "Failed to end the transaction on the cloned volume %u\n",
2572 * reattach to the main-line volume, and incrementally dump it.
2575 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2577 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
2580 "Failed to create a transaction on the source volume %u\n",
2584 /* now do the incremental */
2586 ("Doing the%s dump from source to destination for volume %u ... ",
2587 (flags & RV_NOCLONE) ? "" : " incremental",
2590 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
2593 "Failed to do the%s dump from old site to new site\n",
2594 (flags & RV_NOCLONE) ? "" : " incremental");
2597 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2598 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
2599 code = AFSVolSetFlags(toconn, totid, volflag);
2601 "Failed to set the flags to make destination volume online\n");
2604 /* put new volume online */
2605 VPRINT1("Ending transaction on destination volume %u ...", newVol);
2606 code = AFSVolEndTrans(toconn, totid, &rcode);
2611 "Failed to end the transaction on the destination volume %u\n",
2615 VPRINT1("Ending transaction on source volume %u ...", afromvol);
2616 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2621 "Failed to end the transaction on the source volume %u\n",
2627 if (!(flags & RV_NOCLONE)) {
2628 code = DoVolDelete(fromconn, cloneVol, afrompart,
2629 "cloned", 0, NULL, NULL);
2631 if (code == VNOVOL) {
2632 EPRINT1(code, "Failed to start transaction on %u\n", cloneVol);
2639 if (!(flags & RV_NOVLDB)) {
2640 /* create the vldb entry for the copied volume */
2641 strncpy(newentry.name, atovolname, VOLSER_OLDMAXVOLNAME);
2642 newentry.nServers = 1;
2643 newentry.serverNumber[0] = atoserver;
2644 newentry.serverPartition[0] = atopart;
2645 newentry.flags = (flags & RV_RDONLY) ? RO_EXISTS : RW_EXISTS;
2646 newentry.serverFlags[0] = (flags & RV_RDONLY) ? ITSROVOL : ITSRWVOL;
2647 newentry.volumeId[RWVOL] = newVol;
2648 newentry.volumeId[ROVOL] = (flags & RV_RDONLY) ? newVol : 0;
2649 newentry.volumeId[BACKVOL] = 0;
2650 newentry.cloneId = 0;
2651 /*map into right byte order, before passing to xdr, the stuff has to be in host
2652 * byte order. Xdr converts it into network order */
2653 MapNetworkToHost(&newentry, &storeEntry);
2654 /* create the vldb entry */
2655 vcode = VLDB_CreateEntry(&storeEntry);
2658 "Could not create a VLDB entry for the volume %s %lu\n",
2659 atovolname, (unsigned long)newVol);
2660 /*destroy the created volume */
2661 VPRINT1("Deleting the newly created volume %u\n", newVol);
2662 AFSVolDeleteVolume(toconn, totid);
2666 VPRINT2("Created the VLDB entry for the volume %s %u\n", atovolname,
2670 /* normal cleanup code */
2673 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
2675 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2676 if (code || rcode) {
2679 "Could not end transaction on the source volume %lu\n",
2680 (unsigned long)afromvol);
2682 error = (code ? code : rcode);
2688 VPRINT1("Cleanup: Ending transaction on clone volume %u ...",
2690 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2691 if (code || rcode) {
2694 "Could not end transaction on the source's clone volume %lu\n",
2695 (unsigned long)cloneVol);
2697 error = (code ? code : rcode);
2703 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
2705 code = AFSVolEndTrans(toconn, totid, &rcode);
2709 "Could not end transaction on destination volume %lu\n",
2710 (unsigned long)newVol);
2712 error = (code ? code : rcode);
2717 rx_DestroyConnection(fromconn);
2719 rx_DestroyConnection(toconn);
2720 PrintError("", error);
2723 /* come here only when the sky falls */
2728 "vos copy: operation interrupted, cleanup in progress...\n");
2729 fprintf(STDOUT, "clear transaction contexts\n");
2734 VPRINT("Recovery: Ending transaction on clone volume ...");
2735 AFSVolEndTrans(fromconn, clonetid, &rcode);
2739 VPRINT("Recovery: Ending transaction on destination volume ...");
2740 AFSVolEndTrans(toconn, totid, &rcode);
2743 if (fromtid) { /* put it on-line */
2744 VPRINT("Recovery: Ending transaction on source volume ...");
2745 AFSVolEndTrans(fromconn, fromtid, &rcode);
2749 VPRINT("Recovery: Accessing VLDB.\n");
2750 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2752 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2756 MapHostToNetwork(&entry);
2758 /* common cleanup - delete local clone */
2760 code = DoVolDelete(fromconn, cloneVol, afrompart,
2761 "clone", 0, NULL, "Recovery:");
2762 if (code == VNOVOL) {
2763 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", cloneVol);
2767 done: /* routine cleanup */
2769 rx_DestroyConnection(fromconn);
2771 rx_DestroyConnection(toconn);
2774 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2782 UV_CopyVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2783 char *atovolname, afs_uint32 atoserver, afs_int32 atopart)
2785 return UV_CopyVolume2(afromvol, afromserver, afrompart,
2786 atovolname, atoserver, atopart, 0, 0);
2791 /* Make a new backup of volume <avolid> on <aserver> and <apart>
2792 * if one already exists, update it
2796 UV_BackupVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid)
2798 struct rx_connection *aconn = (struct rx_connection *)0;
2799 afs_int32 ttid = 0, btid = 0;
2800 afs_uint32 backupID;
2801 afs_int32 code = 0, rcode = 0;
2802 struct nvldbentry entry, storeEntry;
2803 afs_int32 error = 0;
2804 int vldblocked = 0, vldbmod = 0;
2806 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
2808 /* the calls to VLDB will succeed only if avolid is a RW volume,
2809 * since we are following the RW hash chain for searching */
2810 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2813 "Could not fetch the entry for the volume %lu from the VLDB \n",
2814 (unsigned long)avolid);
2818 MapHostToNetwork(&entry);
2820 /* These operations require the VLDB be locked since it means the VLDB
2821 * will change or the vldb is already locked.
2823 if (!(entry.flags & BACK_EXISTS) || /* backup volume doesnt exist */
2824 (entry.flags & VLOP_ALLOPERS) || /* vldb lock already held */
2825 (entry.volumeId[BACKVOL] == INVALID_BID)) { /* no assigned backup volume id */
2827 code = ubik_VL_SetLock(cstruct, 0, avolid, RWVOL, VLOP_BACKUP);
2830 "Could not lock the VLDB entry for the volume %lu\n",
2831 (unsigned long)avolid);
2837 /* Reread the vldb entry */
2838 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2841 "Could not fetch the entry for the volume %lu from the VLDB \n",
2842 (unsigned long)avolid);
2846 MapHostToNetwork(&entry);
2849 if (!ISNAMEVALID(entry.name)) {
2850 fprintf(STDERR, "Name of the volume %s exceeds the size limit\n",
2852 error = VOLSERBADNAME;
2856 backupID = entry.volumeId[BACKVOL];
2857 if (backupID == INVALID_BID) {
2858 /* Get a backup volume id from the VLDB and update the vldb
2861 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &backupID);
2864 "Could not allocate ID for the backup volume of %lu from the VLDB\n",
2865 (unsigned long)avolid);
2869 entry.volumeId[BACKVOL] = backupID;
2873 code = DoVolClone(aconn, avolid, apart, backupVolume, backupID, "backup",
2874 entry.name, NULL, ".backup", NULL, NULL);
2880 /* Mark vldb as backup exists */
2881 if (!(entry.flags & BACK_EXISTS)) {
2882 entry.flags |= BACK_EXISTS;
2886 /* Now go back to the backup volume and bring it on line */
2887 code = AFSVolTransCreate_retry(aconn, backupID, apart, ITOffline, &btid);
2890 "Failed to start a transaction on the backup volume %lu\n",
2891 (unsigned long)backupID);
2896 code = AFSVolSetFlags(aconn, btid, 0);
2898 fprintf(STDERR, "Could not mark the backup volume %lu on line \n",
2899 (unsigned long)backupID);
2904 code = AFSVolEndTrans(aconn, btid, &rcode);
2906 if (code || rcode) {
2908 "Failed to end the transaction on the backup volume %lu\n",
2909 (unsigned long)backupID);
2910 error = (code ? code : rcode);
2916 /* Will update the vldb below */
2920 code = AFSVolEndTrans(aconn, ttid, &rcode);
2921 if (code || rcode) {
2922 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
2923 (unsigned long)avolid);
2925 error = (code ? code : rcode);
2930 code = AFSVolEndTrans(aconn, btid, &rcode);
2931 if (code || rcode) {
2933 "Could not end transaction the backup volume %lu\n",
2934 (unsigned long)backupID);
2936 error = (code ? code : rcode);
2940 /* Now update the vldb - if modified */
2943 MapNetworkToHost(&entry, &storeEntry);
2945 VLDB_ReplaceEntry(avolid, RWVOL, &storeEntry,
2946 (LOCKREL_OPCODE | LOCKREL_AFSID |
2947 LOCKREL_TIMESTAMP));
2950 "Could not update the VLDB entry for the volume %lu \n",
2951 (unsigned long)avolid);
2957 ubik_VL_ReleaseLock(cstruct, 0, avolid, RWVOL,
2958 (LOCKREL_OPCODE | LOCKREL_AFSID |
2959 LOCKREL_TIMESTAMP));
2962 "Could not unlock the VLDB entry for the volume %lu \n",
2963 (unsigned long)avolid);
2971 rx_DestroyConnection(aconn);
2973 PrintError("", error);
2977 /* Make a new clone of volume <avolid> on <aserver> and <apart>
2978 * using volume ID <acloneid>, or a new ID allocated from the VLDB.
2979 * The new volume is named by <aname>, or by appending ".clone" to
2980 * the existing name if <aname> is NULL. The following flags are
2983 * RV_RDONLY - target volume is RO
2984 * RV_OFFLINE - leave target volume offline
2988 UV_CloneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid,
2989 afs_uint32 acloneid, char *aname, int flags)
2991 struct rx_connection *aconn = (struct rx_connection *)0;
2992 afs_int32 ttid = 0, btid = 0;
2993 afs_int32 code = 0, rcode = 0;
2994 char vname[VOLSER_MAXVOLNAME + 1];
2995 afs_int32 error = 0;
2996 volEntries volumeInfo;
2999 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
3002 volumeInfo.volEntries_val = (volintInfo *) 0;
3003 volumeInfo.volEntries_len = 0;
3004 code = AFSVolListOneVolume(aconn, apart, avolid, &volumeInfo);
3006 fprintf(stderr, "Could not get info for volume %lu\n",
3007 (unsigned long)avolid);
3011 strncpy(vname, volumeInfo.volEntries_val[0].name,
3012 VOLSER_OLDMAXVOLNAME - 7);
3013 vname[VOLSER_OLDMAXVOLNAME - 7] = 0;
3014 strcat(vname, ".clone");
3016 if (volumeInfo.volEntries_val)
3017 free(volumeInfo.volEntries_val);
3021 /* Get a clone id */
3022 VPRINT1("Allocating new volume id for clone of volume %u ...",
3024 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &acloneid);
3026 "Could not get an ID for the clone of volume %u from the VLDB\n",
3031 if (flags & RV_RWONLY)
3032 type = readwriteVolume;
3033 else if (flags & RV_RDONLY)
3034 type = readonlyVolume;
3036 type = backupVolume;
3038 code = DoVolClone(aconn, avolid, apart, type, acloneid, "clone",
3039 NULL, ".clone", NULL, NULL, NULL);
3045 /* Now go back to the backup volume and bring it on line */
3046 if (!(flags & RV_OFFLINE)) {
3047 code = AFSVolTransCreate_retry(aconn, acloneid, apart, ITOffline, &btid);
3050 "Failed to start a transaction on the clone volume %lu\n",
3051 (unsigned long)acloneid);
3056 code = AFSVolSetFlags(aconn, btid, 0);
3058 fprintf(STDERR, "Could not mark the clone volume %lu on line \n",
3059 (unsigned long)acloneid);
3064 code = AFSVolEndTrans(aconn, btid, &rcode);
3066 if (code || rcode) {
3068 "Failed to end the transaction on the clone volume %lu\n",
3069 (unsigned long)acloneid);
3070 error = (code ? code : rcode);
3079 code = AFSVolEndTrans(aconn, ttid, &rcode);
3080 if (code || rcode) {
3081 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
3082 (unsigned long)avolid);
3084 error = (code ? code : rcode);
3089 code = AFSVolEndTrans(aconn, btid, &rcode);
3090 if (code || rcode) {
3092 "Could not end transaction on the clone volume %lu\n",
3093 (unsigned long)acloneid);
3095 error = (code ? code : rcode);
3100 rx_DestroyConnection(aconn);
3102 PrintError("", error);
3106 #define ONERROR(ec, ep, es) do { \
3108 fprintf(STDERR, (es), (ep)); \
3113 #define ONERROR0(ec, es) do { \
3115 fprintf(STDERR, (es)); \
3120 #define ERROREXIT(ec) do { \
3125 /* Get a "transaction" on this replica. Create the volume
3126 * if necessary. Return the time from which a dump should
3127 * be made (0 if it's a new volume)
3130 GetTrans(struct nvldbentry *vldbEntryPtr, afs_int32 index,
3131 struct rx_connection **connPtr, afs_int32 * transPtr,
3132 afs_uint32 * crtimePtr, afs_uint32 * uptimePtr,
3133 afs_int32 *origflags, afs_uint32 tmpVolId)
3136 struct volser_status tstatus;
3141 *connPtr = (struct rx_connection *)0;
3146 /* get connection to the replication site */
3147 *connPtr = UV_Bind(vldbEntryPtr->serverNumber[index], AFSCONF_VOLUMEPORT);
3149 goto fail; /* server is down */
3151 volid = vldbEntryPtr->volumeId[ROVOL];
3155 AFSVolTransCreate_retry(*connPtr, volid,
3156 vldbEntryPtr->serverPartition[index], ITOffline,
3159 if (!code && (origflags[index] & RO_DONTUSE)) {
3160 /* If RO_DONTUSE is set, this is supposed to be an entirely new
3161 * site. Don't trust any data on it, since it is possible we
3162 * have encountered some temporary volume from some other
3163 * incomplete volume operation. It is difficult to detect if
3164 * that has happened vs if this is a legit volume, so just
3165 * delete it to be safe. */
3167 VPRINT1("Deleting extant RO_DONTUSE site on %s...",
3168 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3169 serverNumber[index], hoststr) :
3170 hostutil_GetNameByINet(vldbEntryPtr->
3171 serverNumber[index]));
3173 code = AFSVolDeleteVolume(*connPtr, *transPtr);
3175 PrintError("Failed to delete RO_DONTUSE site: ", code);
3179 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3185 PrintError("Failed to end transaction on RO_DONTUSE site: ",
3192 /* emulate what TransCreate would have returned, so we try to
3193 * create the volume below */
3198 /* If the volume does not exist, create it */
3199 if (!volid || code) {
3200 char volname[VL_MAXNAMELEN];
3203 if (volid && (code != VNOVOL)) {
3204 PrintError("Failed to start a transaction on the RO volume.\n",
3209 strlcpy(volname, vldbEntryPtr->name, sizeof(volname));
3211 if (strlcat(volname,
3212 tmpVolId?".roclone":".readonly",
3213 sizeof(volname)) >= sizeof(volname)) {
3215 PrintError("Volume name is too long\n", code);
3221 "Creating new volume %lu on replication site %s: ",
3222 tmpVolId?(unsigned long)tmpVolId:(unsigned long)volid,
3223 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3224 serverNumber[index], hoststr) :
3225 hostutil_GetNameByINet(vldbEntryPtr->
3226 serverNumber[index]));
3231 AFSVolCreateVolume(*connPtr, vldbEntryPtr->serverPartition[index],
3233 vldbEntryPtr->volumeId[RWVOL],
3234 tmpVolId?&tmpVolId:&volid,
3237 PrintError("Failed to create the ro volume: ", code);
3240 vldbEntryPtr->volumeId[ROVOL] = volid;
3244 /* The following is a bit redundant, since create sets these flags by default */
3246 AFSVolSetFlags(*connPtr, *transPtr,
3247 VTDeleteOnSalvage | VTOutOfService);
3249 PrintError("Failed to set flags on the ro volume: ", code);
3254 /* Otherwise, the transaction did succeed, so get the creation date of the
3255 * latest RO volume on the replication site
3258 VPRINT2("Updating existing ro volume %u on %s ...\n", volid,
3259 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3260 serverNumber[index], hoststr) :
3261 hostutil_GetNameByINet(vldbEntryPtr->serverNumber[index]));
3263 code = AFSVolGetStatus(*connPtr, *transPtr, &tstatus);
3265 PrintError("Failed to get status of volume on destination: ",
3270 code = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3275 code = DoVolClone(*connPtr, volid,
3276 vldbEntryPtr->serverPartition[index],
3277 readonlyVolume, tmpVolId, "temporary",
3278 vldbEntryPtr->name, NULL, ".roclone", NULL,
3283 *crtimePtr = CLOCKADJ(tstatus.creationDate);
3284 *uptimePtr = CLOCKADJ(tstatus.updateDate);
3291 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3295 if (tcode && tcode != ENOENT)
3296 PrintError("Could not end transaction on a ro volume: ", tcode);
3303 SimulateForwardMultiple(struct rx_connection *fromconn, afs_int32 fromtid,
3304 afs_int32 fromdate, manyDests * tr, afs_int32 flags,
3305 void *cookie, manyResults * results)
3309 for (i = 0; i < tr->manyDests_len; i++) {
3310 results->manyResults_val[i] =
3311 AFSVolForward(fromconn, fromtid, fromdate,
3312 &(tr->manyDests_val[i].server),
3313 tr->manyDests_val[i].trans, cookie);
3319 * Check if a trans has timed out, and recreate it if necessary.
3321 * @param[in] aconn RX connection to the relevant server
3322 * @param[inout] atid Transaction ID to check; if we recreated the trans,
3323 * contains the new trans ID on success
3324 * @param[in] apart Partition for the transaction
3325 * @param[in] astat The status of the original transaction
3327 * @return operation status
3328 * @retval 0 existing transaction is still valid, or we managed to recreate
3329 * the trans successfully
3330 * @retval nonzero Fatal error; bail out
3333 CheckTrans(struct rx_connection *aconn, afs_int32 *atid, afs_int32 apart,
3334 struct volser_status *astat)
3336 struct volser_status new_status;
3339 memset(&new_status, 0, sizeof(new_status));
3340 code = AFSVolGetStatus(aconn, *atid, &new_status);
3342 if (code == ENOENT) {
3344 VPRINT1("Old transaction on cloned volume %lu timed out, "
3345 "restarting transaction\n", (long unsigned) astat->volID);
3346 code = AFSVolTransCreate_retry(aconn, astat->volID, apart,
3349 PrintError("Failed to recreate cloned RO volume transaction\n",
3354 memset(&new_status, 0, sizeof(new_status));
3355 code = AFSVolGetStatus(aconn, *atid, &new_status);
3357 PrintError("Failed to get status on recreated transaction\n",
3362 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3363 PrintError("Recreated transaction on cloned RO volume, but "
3364 "the volume has changed!\n", 0);
3368 PrintError("Unable to get status of current cloned RO transaction\n",
3373 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3375 PrintError("Internal error: current GetStatus does not match "
3376 "original GetStatus?\n", 0);
3385 PutTrans(afs_int32 *vldbindex, struct replica *replicas,
3386 struct rx_connection **toconns, struct release *times,
3389 afs_int32 s, code = 0, rcode = 0;
3390 /* End the transactions and destroy the connections */
3391 for (s = 0; s < volcount; s++) {
3392 if (replicas[s].trans) {
3393 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
3395 replicas[s].trans = 0;
3399 if ((s == 0) || (code != ENOENT)) {
3400 PrintError("Could not end transaction on a ro volume: ",
3404 ("Transaction timed out on a ro volume. Will retry.\n",
3406 if (times[s].vldbEntryIndex < *vldbindex)
3407 *vldbindex = times[s].vldbEntryIndex;
3412 rx_DestroyConnection(toconns[s]);
3418 DoVolOnline(struct nvldbentry *vldbEntryPtr, afs_uint32 avolid, int index,
3419 char *vname, struct rx_connection *connPtr)
3421 afs_int32 code = 0, rcode = 0, onlinetid = 0;
3424 AFSVolTransCreate_retry(connPtr, avolid,
3425 vldbEntryPtr->serverPartition[index],
3429 EPRINT(code, "Could not create transaction on readonly...\n");
3432 code = AFSVolSetFlags(connPtr, onlinetid, 0);
3434 EPRINT(code, "Could not set flags on readonly...\n");
3439 AFSVolSetIdsTypes(connPtr, onlinetid, vname,
3440 ROVOL, vldbEntryPtr->volumeId[RWVOL],
3443 EPRINT(code, "Could not set ids on readonly...\n");
3446 code = AFSVolEndTrans(connPtr, onlinetid, &rcode);
3452 /* UV_ReleaseVolume()
3453 * Release volume <afromvol> on <afromserver> <afrompart> to all
3454 * its RO sites (full release). Unless the previous release was
3455 * incomplete: in which case we bring the remaining incomplete
3456 * volumes up to date with the volumes that were released
3458 * forceflag: Performs a full release.
3460 * Will create a clone from the RW, then dump the clone out to
3461 * the remaining replicas. If there is more than 1 RO sites,
3462 * ensure that the VLDB says at least one RO is available all
3463 * the time: Influences when we write back the VLDB entry.
3467 UV_ReleaseVolume(afs_uint32 afromvol, afs_uint32 afromserver,
3468 afs_int32 afrompart, int forceflag, int stayUp)
3472 afs_int32 vcode, rcode, tcode;
3473 afs_uint32 cloneVolId = 0, roVolId;
3474 struct replica *replicas = 0;
3475 struct nvldbentry entry, storeEntry;
3476 int i, volcount = 0, m, vldbindex;
3478 struct restoreCookie cookie;
3479 struct rx_connection **toconns = 0;
3480 struct release *times = 0;
3482 struct rx_connection *fromconn = (struct rx_connection *)0;
3483 afs_int32 error = 0;
3485 afs_int32 clonetid = 0, onlinetid;
3486 afs_int32 fromtid = 0;
3487 afs_uint32 fromdate = 0;
3488 afs_uint32 thisdate;
3492 manyResults results;
3493 int rwindex, roindex, roclone, roexists;
3494 afs_uint32 rwcrdate = 0, rwupdate = 0;
3495 afs_uint32 clcrdate;
3499 } remembertime[NMAXNSERVERS];
3500 int releasecount = 0;
3501 struct volser_status volstatus;
3503 afs_int32 origflags[NMAXNSERVERS];
3504 struct volser_status orig_status;
3505 int notreleased = 0;
3506 int tried_justnewsites = 0;
3507 int justnewsites = 0; /* are we just trying to release to new RO sites? */
3508 int sites = 0; /* number of ro sites */
3509 int new_sites = 0; /* number of ro sites markes as new */
3512 CR_RECOVER = 0x0000, /**< not complete: a recovery from a previous failed release */
3513 CR_FORCED = 0x0001, /**< complete: forced by caller */
3514 CR_LAST_OK = 0x0002, /**< complete: no sites have been marked as new release */
3515 CR_ALL_NEW = 0x0004, /**< complete: all sites have been marked as new release */
3516 CR_NEW_RW = 0x0008, /**< complete: read-write has changed */
3517 CR_RO_MISSING = 0x0010, /**< complete: ro clone is missing */
3518 } complete_release_t;
3520 complete_release_t complete_release = CR_RECOVER;
3522 memset(remembertime, 0, sizeof(remembertime));
3523 memset(&results, 0, sizeof(results));
3524 memset(origflags, 0, sizeof(origflags));
3526 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_RELEASE);
3527 if (vcode != VL_RERELEASE)
3528 ONERROR(vcode, afromvol,
3529 "Could not lock the VLDB entry for the volume %u.\n");
3532 /* Get the vldb entry in readable format */
3533 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
3534 ONERROR(vcode, afromvol,
3535 "Could not fetch the entry for the volume %u from the VLDB.\n");
3536 MapHostToNetwork(&entry);
3539 EnumerateEntry(&entry);
3541 if (!ISNAMEVALID(entry.name))
3542 ONERROR(VOLSERBADOP, entry.name,
3543 "Volume name %s is too long, rename before releasing.\n");
3544 if (entry.volumeId[RWVOL] != afromvol)
3545 ONERROR(VOLSERBADOP, afromvol,
3546 "The volume %u being released is not a read-write volume.\n");
3547 if (entry.nServers <= 1)
3548 ONERROR(VOLSERBADOP, afromvol,
3549 "Volume %u has no replicas - release operation is meaningless!\n");
3550 if (strlen(entry.name) > (VOLSER_OLDMAXVOLNAME - 10))
3551 ONERROR(VOLSERBADOP, entry.name,
3552 "RO volume name %s exceeds (VOLSER_OLDMAXVOLNAME - 10) character limit\n");
3554 /* roclone is true if one of the RO volumes is on the same
3555 * partition as the RW volume. In this case, we make the RO volume
3556 * on the same partition a clone instead of a complete copy.
3559 roindex = Lp_ROMatch(afromserver, afrompart, &entry) - 1;
3560 roclone = ((roindex == -1) ? 0 : 1);
3561 rwindex = Lp_GetRwIndex(&entry);
3563 ONERROR0(VOLSERNOVOL, "There is no RW volume \n");
3565 /* Make sure we have a RO volume id to work with */
3566 if (entry.volumeId[ROVOL] == INVALID_BID) {
3567 /* need to get a new RO volume id */
3568 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &roVolId);
3569 ONERROR(vcode, entry.name, "Cant allocate ID for RO volume of %s\n");
3571 entry.volumeId[ROVOL] = roVolId;
3572 MapNetworkToHost(&entry, &storeEntry);
3573 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3574 ONERROR(vcode, entry.name, "Could not update vldb entry for %s.\n");
3578 * Determine if this is to be a complete release or a recovery of a
3579 * previous unfinished release. The previous release is considered to be
3580 * unfinished when the clone was successfully distributed to at least one
3581 * (but not all) of the read-only sites, as indicated by the NEW_REPSITE
3584 * The caller can override the vldb flags check using the -force
3585 * flag, to force this to be a complete release.
3587 for (i = 0; i < entry.nServers; i++) {
3588 if (entry.serverFlags[i] & ITSROVOL) {
3590 if (entry.serverFlags[i] & NEW_REPSITE)
3592 if (entry.serverFlags[i] & RO_DONTUSE)
3595 origflags[i] = entry.serverFlags[i];
3599 complete_release |= CR_FORCED;
3602 if (new_sites == 0) {
3603 complete_release |= CR_LAST_OK;
3604 } else if (new_sites == sites) {
3605 complete_release |= CR_ALL_NEW;
3608 if ((complete_release & (CR_LAST_OK | CR_ALL_NEW))
3609 && !(complete_release & CR_FORCED)) {
3610 if (notreleased && notreleased != sites) {
3611 /* we have some new unreleased sites. try to just release to those,
3612 * if the RW has not changed. The caller can override with -force. */
3617 /* Determine which volume id to use and see if it exists */
3618 cloneVolId = (complete_release || entry.cloneId == 0)
3619 ? entry.volumeId[ROVOL] : entry.cloneId;
3621 code = VolumeExists(afromserver, afrompart, cloneVolId);
3622 roexists = ((code == ENODEV) ? 0 : 1);
3624 /* For stayUp case, if roclone is the only site, bypass special handling */
3625 if (stayUp && roclone) {
3629 for (e = 0; (e < entry.nServers) && !error; e++) {
3630 if ((entry.serverFlags[e] & ITSROVOL)) {
3631 if (!(VLDB_IsSameAddrs(entry.serverNumber[e], afromserver,
3636 if (e >= entry.nServers)
3640 /* If we had a previous release to complete, do so, else: */
3641 if (stayUp && (cloneVolId == entry.volumeId[ROVOL])) {
3642 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &cloneVolId);
3643 ONERROR(code, afromvol,
3644 "Cannot get temporary clone id for volume %u\n");
3647 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
3649 ONERROR(-1, afromserver,
3650 "Cannot establish connection with server 0x%x\n");
3652 if (!complete_release) {
3654 complete_release |= CR_RO_MISSING; /* Do a complete release if RO clone does not exist */
3656 /* Begin transaction on RW and mark it busy while we query it */
3657 code = AFSVolTransCreate_retry(
3658 fromconn, afromvol, afrompart, ITBusy, &fromtid
3660 ONERROR(code, afromvol,
3661 "Failed to start transaction on RW volume %u\n");
3663 /* Query the creation date for the RW */
3664 code = AFSVolGetStatus(fromconn, fromtid, &volstatus);
3665 ONERROR(code, afromvol,
3666 "Failed to get the status of RW volume %u\n");
3667 rwcrdate = volstatus.creationDate;
3669 /* End transaction on RW */
3670 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
3672 ONERROR((code ? code : rcode), afromvol,
3673 "Failed to end transaction on RW volume %u\n");
3675 /* Begin transaction on clone and mark it busy while we query it */
3676 code = AFSVolTransCreate_retry(
3677 fromconn, cloneVolId, afrompart, ITBusy, &clonetid
3679 ONERROR(code, cloneVolId,
3680 "Failed to start transaction on RW clone %u\n");
3682 /* Query the creation date for the clone */
3683 code = AFSVolGetStatus(fromconn, clonetid, &volstatus);
3684 ONERROR(code, cloneVolId,
3685 "Failed to get the status of RW clone %u\n");
3686 clcrdate = volstatus.creationDate;
3688 /* End transaction on clone */
3689 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
3691 ONERROR((code ? code : rcode), cloneVolId,
3692 "Failed to end transaction on RW clone %u\n");
3694 if (rwcrdate > clcrdate)
3695 complete_release |= CR_NEW_RW; /* Do a complete release if RO clone older than RW */
3699 if (!complete_release || (complete_release & CR_NEW_RW)) {
3700 /* in case the RW has changed, and just to be safe */
3705 if (!complete_release) {
3707 "This is a recovery of previously failed release\n");
3709 fprintf(STDOUT, "This is a complete release of volume %u", afromvol);
3710 /* Give the reasons for a complete release, except if only CR_LAST_OK. */
3711 if (complete_release != CR_LAST_OK) {
3713 if (complete_release & CR_FORCED) {
3714 fprintf(STDOUT, "%sforced", sep);
3717 if (complete_release & CR_LAST_OK) {
3718 fprintf(STDOUT, "%slast ok", sep);
3721 if (complete_release & CR_ALL_NEW) {
3722 fprintf(STDOUT, "%sall sites are new", sep);
3725 if (complete_release & CR_NEW_RW) {
3726 fprintf(STDOUT, "%srw %u changed", sep, afromvol);
3729 if (complete_release & CR_RO_MISSING) {
3730 fprintf(STDOUT, "%sro clone missing", sep);
3732 fprintf(STDOUT, ")");
3734 fprintf(STDOUT, "\n");
3736 tried_justnewsites = 1;
3737 fprintf(STDOUT, "There are new RO sites; we will try to "
3738 "only release to new sites\n");
3743 if (complete_release) {
3744 afs_int32 oldest = 0;
3745 /* If the RO clone exists, then if the clone is a temporary
3746 * clone, delete it. Or if the RO clone is marked RO_DONTUSE
3747 * (it was recently added), then also delete it. We do not
3748 * want to "reclone" a temporary RO clone.
3751 code = VolumeExists(afromserver, afrompart, cloneVolId);
3753 code = DoVolDelete(fromconn, cloneVolId, afrompart, "previous clone", 0,
3755 if (code && (code != VNOVOL))
3760 /* clean up any previous tmp clone before starting if staying up */
3762 && (!roclone || (entry.serverFlags[roindex] & RO_DONTUSE))) {
3763 code = DoVolDelete(fromconn,
3764 stayUp ? entry.volumeId[ROVOL] : cloneVolId,
3765 afrompart, "the", 0, NULL, NULL);
3766 if (code && (code != VNOVOL))
3772 VPRINT("Querying old RO sites for update times...");
3773 for (vldbindex = 0; vldbindex < entry.nServers; vldbindex++) {
3774 volEntries volumeInfo;
3775 struct rx_connection *conn;
3778 if (!(entry.serverFlags[vldbindex] & ITSROVOL)) {
3781 if ((entry.serverFlags[vldbindex] & RO_DONTUSE)) {
3784 conn = UV_Bind(entry.serverNumber[vldbindex], AFSCONF_VOLUMEPORT);
3786 fprintf(STDERR, "Cannot establish connection to server %s\n",
3787 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3791 volumeInfo.volEntries_val = NULL;
3792 volumeInfo.volEntries_len = 0;
3793 code = AFSVolListOneVolume(conn, entry.serverPartition[vldbindex],
3794 entry.volumeId[ROVOL],
3797 fprintf(STDERR, "Could not fetch information about RO vol %lu from server %s\n",
3798 (unsigned long)entry.volumeId[ROVOL],
3799 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3800 PrintError("", code);
3802 rx_DestroyConnection(conn);
3806 crdate = CLOCKADJ(volumeInfo.volEntries_val[0].creationDate);
3808 if (oldest == 0 || crdate < oldest) {
3812 rx_DestroyConnection(conn);
3813 free(volumeInfo.volEntries_val);
3814 volumeInfo.volEntries_val = NULL;
3815 volumeInfo.volEntries_len = 0;
3820 volEntries volumeInfo;
3821 volumeInfo.volEntries_val = NULL;
3822 volumeInfo.volEntries_len = 0;
3823 code = AFSVolListOneVolume(fromconn, afrompart, afromvol,
3826 fprintf(STDERR, "Could not fetch information about RW vol %lu from server %s\n",
3827 (unsigned long)afromvol,
3828 hostutil_GetNameByINet(afromserver));
3829 PrintError("", code);
3832 rwupdate = volumeInfo.volEntries_val[0].updateDate;
3834 free(volumeInfo.volEntries_val);
3835 volumeInfo.volEntries_val = NULL;
3836 volumeInfo.volEntries_len = 0;
3839 if (justnewsites && oldest <= rwupdate) {
3840 /* RW has changed */
3844 /* Mark all the ROs in the VLDB entry as RO_DONTUSE. We don't
3845 * write this entry out to the vlserver until after the first
3846 * RO volume is released (temp RO clones don't count).
3848 * If 'justnewsites' is set, we're only updating sites that have
3849 * RO_DONTUSE set, so set NEW_REPSITE for all of the others.
3851 for (i = 0; i < entry.nServers; i++) {
3853 if ((entry.serverFlags[i] & RO_DONTUSE)) {
3854 entry.serverFlags[i] &= ~NEW_REPSITE;
3856 entry.serverFlags[i] |= NEW_REPSITE;
3859 entry.serverFlags[i] &= ~NEW_REPSITE;
3860 entry.serverFlags[i] |= RO_DONTUSE;
3863 entry.serverFlags[rwindex] |= NEW_REPSITE;
3864 entry.serverFlags[rwindex] &= ~RO_DONTUSE;
3867 if (justnewsites && roexists) {
3868 /* if 'justnewsites' and 'roexists' are set, we don't need to do
3869 * anything with the RO clone, so skip the reclone */
3872 } else if (complete_release) {
3875 strcpy(vname, entry.name);
3877 strcat(vname, ".roclone");
3879 strcat(vname, ".readonly");
3881 strcpy(vname, "readonly-clone-temp");
3884 code = DoVolClone(fromconn, afromvol, afrompart, readonlyVolume,
3885 cloneVolId, (roclone && !stayUp)?"permanent RO":
3886 "temporary RO", NULL, vname, NULL, &volstatus, NULL);
3892 if (justnewsites && rwupdate != volstatus.updateDate) {
3894 /* reset the serverFlags as if 'justnewsites' had never been set */
3895 for (i = 0; i < entry.nServers; i++) {
3896 entry.serverFlags[i] &= ~NEW_REPSITE;
3897 entry.serverFlags[i] |= RO_DONTUSE;
3899 entry.serverFlags[rwindex] |= NEW_REPSITE;
3900 entry.serverFlags[rwindex] &= ~RO_DONTUSE;
3903 rwcrdate = volstatus.creationDate;
3905 /* Remember clone volume ID in case we fail or are interrupted */
3906 entry.cloneId = cloneVolId;
3908 if (roclone && !stayUp) {
3909 /* Bring the RO clone online - though not if it's a temporary clone */
3910 VPRINT1("Starting transaction on RO clone volume %u...",
3913 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITOffline,
3915 ONERROR(code, cloneVolId,
3916 "Failed to start transaction on volume %u\n");
3919 VPRINT1("Setting volume flags for volume %u...", cloneVolId);
3920 tcode = AFSVolSetFlags(fromconn, onlinetid, 0);
3923 VPRINT1("Ending transaction on volume %u...", cloneVolId);
3924 code = AFSVolEndTrans(fromconn, onlinetid, &rcode);
3925 ONERROR((code ? code : rcode), cloneVolId,
3926 "Failed to end transaction on RO clone %u\n");
3929 ONERROR(tcode, cloneVolId, "Could not bring volume %u on line\n");
3931 /* Sleep so that a client searching for an online volume won't
3932 * find the clone offline and then the next RO offline while the
3933 * release brings the clone online and the next RO offline (race).
3934 * There is a fix in the 3.4 client that does not need this sleep
3935 * anymore, but we don't know what clients we have.
3937 if (entry.nServers > 2 && !justnewsites)
3940 /* Mark the RO clone in the VLDB as a good site (already released) */
3941 entry.serverFlags[roindex] |= NEW_REPSITE;
3942 entry.serverFlags[roindex] &= ~RO_DONTUSE;
3943 entry.flags |= RO_EXISTS;
3947 /* Write out the VLDB entry only if the clone is not a temporary
3948 * clone. If we did this to a temporary clone then we would end
3949 * up marking all the ROs as "old release" making the ROs
3950 * temporarily unavailable.
3952 MapNetworkToHost(&entry, &storeEntry);
3953 VPRINT1("Replacing VLDB entry for %s...", entry.name);
3954 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3955 ONERROR(vcode, entry.name,
3956 "Could not update vldb entry for %s.\n");
3962 VPRINT("RW vol has not changed; only releasing to new RO sites\n");
3963 /* act like this is a completion of a previous release */
3964 complete_release = CR_RECOVER;
3965 } else if (tried_justnewsites) {
3966 VPRINT("RW vol has changed; releasing to all sites\n");
3969 /* Now we will release from the clone to the remaining RO replicas.
3970 * The first 2 ROs (counting the non-temporary RO clone) are released
3971 * individually: releasecount. This is to reduce the race condition
3972 * of clients trying to find an on-line RO volume. The remaining ROs
3973 * are released in parallel but no more than half the number of ROs
3974 * (rounded up) at a time: nservers.
3977 strcpy(vname, entry.name);
3979 strcat(vname, ".roclone");
3981 strcat(vname, ".readonly");
3982 memset(&cookie, 0, sizeof(cookie));
3983 strncpy(cookie.name, vname, VOLSER_OLDMAXVOLNAME);
3984 cookie.type = ROVOL;
3985 cookie.parent = entry.volumeId[RWVOL];
3988 /* how many to do at once, excluding clone */
3989 if (stayUp || justnewsites)
3990 nservers = entry.nServers; /* can do all, none offline */
3992 nservers = entry.nServers / 2;
3993 replicas = calloc(nservers + 1, sizeof(struct replica));
3994 times = calloc(nservers + 1, sizeof(struct release));
3995 toconns = calloc(nservers + 1, sizeof(struct rx_connection *));
3996 results.manyResults_val = calloc(nservers + 1, sizeof(afs_int32));
3997 if (!replicas || !times || !results.manyResults_val || !toconns)
3999 "Failed to create transaction on the release clone\n");
4001 /* Create a transaction on the cloned volume */
4002 VPRINT1("Starting transaction on cloned volume %u...", cloneVolId);
4004 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITBusy, &fromtid);
4006 memset(&orig_status, 0, sizeof(orig_status));
4007 code = AFSVolGetStatus(fromconn, fromtid, &orig_status);
4009 if (!complete_release && code)
4010 ONERROR(VOLSERNOVOL, afromvol,
4011 "Old clone is inaccessible. Try vos release -f %u.\n");
4012 ONERROR0(code, "Failed to create transaction on the release clone\n");
4015 /* if we have a clone, treat this as done, for now */
4016 if (stayUp && !complete_release) {
4017 entry.serverFlags[roindex] |= NEW_REPSITE;
4018 entry.serverFlags[roindex] &= ~RO_DONTUSE;
4019 entry.flags |= RO_EXISTS;
4024 /* For each index in the VLDB */
4025 for (vldbindex = 0; vldbindex < entry.nServers;) {
4026 /* Get a transaction on the replicas. Pick replicas which have an old release. */
4028 ((volcount < nservers) && (vldbindex < entry.nServers));
4030 if (!stayUp && !justnewsites) {
4031 /* The first two RO volumes will be released individually.
4032 * The rest are then released in parallel. This is a hack
4033 * for clients not recognizing right away when a RO volume
4034 * comes back on-line.
4036 if ((volcount == 1) && (releasecount < 2))
4040 if (vldbindex == roindex)
4041 continue; /* the clone */
4042 if ((entry.serverFlags[vldbindex] & NEW_REPSITE)
4043 && !(entry.serverFlags[vldbindex] & RO_DONTUSE))
4045 if (!(entry.serverFlags[vldbindex] & ITSROVOL))
4046 continue; /* not a RO vol */
4049 /* Get a Transaction on this replica. Get a new connection if
4050 * necessary. Create the volume if necessary. Return the
4051 * time from which the dump should be made (0 if it's a new
4052 * volume). Each volume might have a different time.
4054 replicas[volcount].server.destHost =
4055 ntohl(entry.serverNumber[vldbindex]);
4056 replicas[volcount].server.destPort = AFSCONF_VOLUMEPORT;
4057 replicas[volcount].server.destSSID = 1;
4058 times[volcount].vldbEntryIndex = vldbindex;
4061 GetTrans(&entry, vldbindex, &(toconns[volcount]),
4062 &(replicas[volcount].trans),
4063 &(times[volcount].crtime),
4064 &(times[volcount].uptime),
4065 origflags, stayUp?cloneVolId:0);
4069 /* Thisdate is the date from which we want to pick up all changes */
4071 /* Do a full dump when forced by the caller. */
4072 VPRINT("This will be a full dump: forced\n");
4074 } else if (!complete_release) {
4075 /* If this release is a recovery of a failed release, we can't be
4076 * sure the creation date is good, so do a full dump.
4078 VPRINT("This will be a full dump: previous release failed\n");
4080 } else if (times[volcount].crtime == 0) {
4081 /* A full dump is needed for a new read-only volume. */
4083 ("This will be a full dump: read-only volume needs to be created\n");
4085 } else if ((rwcrdate > times[volcount].crtime)) {
4086 /* If the RW volume was replaced (its creation date is newer than
4087 * the last release), then we can't be sure what has changed (so
4088 * we do a full dump).
4091 ("This will be a full dump: read-write volume was replaced\n");
4093 } else if (remembertime[vldbindex].validtime) {
4094 /* Trans was prev ended. Use the time from the prev trans
4095 * because, prev trans may have created the volume. In which
4096 * case time[volcount].time would be now instead of 0.
4099 (remembertime[vldbindex].uptime < times[volcount].uptime)
4100 ? remembertime[vldbindex].uptime
4101 : times[volcount].uptime;
4103 thisdate = times[volcount].uptime;
4105 remembertime[vldbindex].validtime = 1;
4106 remembertime[vldbindex].uptime = thisdate;
4108 if (volcount == 0) {
4109 fromdate = thisdate;
4111 /* Include this volume if it is within 15 minutes of the earliest */
4113 thisdate) ? (fromdate - thisdate) : (thisdate -
4115 AFSVolEndTrans(toconns[volcount],
4116 replicas[volcount].trans, &rcode);
4117 replicas[volcount].trans = 0;
4120 if (thisdate < fromdate)
4121 fromdate = thisdate;
4128 code = CheckTrans(fromconn, &fromtid, afrompart, &orig_status);
4135 fprintf(STDOUT, "Starting ForwardMulti from %lu to %u on %s",
4136 (unsigned long)cloneVolId, stayUp?
4137 cloneVolId:entry.volumeId[ROVOL],
4138 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[0].
4139 vldbEntryIndex], hoststr) :
4140 hostutil_GetNameByINet(entry.
4141 serverNumber[times[0].
4144 for (s = 1; s < volcount; s++) {