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 DoVolClone(struct rx_connection *aconn, afs_uint32 avolid,
176 afs_int32 apart, int type, afs_uint32 cloneid,
177 char *typestring, char *pname, char *vname, char *suffix,
178 struct volser_status *volstatus, afs_int32 *transPtr);
179 static int DoVolDelete(struct rx_connection *aconn, afs_uint32 avolid,
180 afs_int32 apart, char *typestring, afs_uint32 atoserver,
181 struct volser_status *volstatus, char *pprefix);
182 static afs_int32 CheckVolume(volintInfo * volumeinfo, afs_uint32 aserver,
183 afs_int32 apart, afs_int32 * modentry,
184 afs_uint32 * maxvolid, struct nvldbentry *aentry);
185 static afs_int32 VolumeExists(afs_uint32 server, afs_int32 partition,
186 afs_uint32 volumeid);
187 static afs_int32 CheckVldbRWBK(struct nvldbentry * entry,
188 afs_int32 * modified);
189 static afs_int32 CheckVldbRO(struct nvldbentry *entry, afs_int32 * modified);
190 static afs_int32 CheckVldb(struct nvldbentry *entry, afs_int32 * modified,
192 static void dump_sig_handler(int x);
193 static int sortVolumes(const void *a, const void *b);
196 /*map the partition <partId> into partition name <partName>*/
198 MapPartIdIntoName(afs_int32 partId, char *partName)
200 if (partId < 26) { /* what if partId > = 26 ? */
201 strcpy(partName, "/vicep");
202 partName[6] = partId + 'a';
205 } else if (partId < VOLMAXPARTS) {
206 strcpy(partName, "/vicep");
208 partName[6] = 'a' + (partId / 26);
209 partName[7] = 'a' + (partId % 26);
216 PrintError(char *msg, afs_int32 errcode)
218 fprintf(STDERR, "%s", msg);
219 /*replace by a big switch statement */
224 fprintf(STDERR, "Possible communication failure\n");
227 fprintf(STDERR, "Volume needs to be salvaged\n");
230 fprintf(STDERR, "Bad vnode number quoted\n");
234 "Volume not attached, does not exist, or not on line\n");
237 fprintf(STDERR, "Volume already exists\n");
240 fprintf(STDERR, "Volume is not in service\n");
243 fprintf(STDERR, "Volume is off line\n");
246 fprintf(STDERR, "Volume is already on line\n");
249 fprintf(STDERR, "Partition is full\n");
252 fprintf(STDERR, "Volume max quota exceeded\n");
255 fprintf(STDERR, "Volume temporarily unavailable\n");
258 fprintf(STDERR, "Volume has moved to another server\n");
261 fprintf(STDERR, "VLDB: volume Id exists in the vldb\n");
264 fprintf(STDERR, "VLDB: a read terminated too early\n");
267 fprintf(STDERR, "VLDB: volume entry exists in the vldb\n");
270 fprintf(STDERR, "VLDB: internal creation failure\n");
273 fprintf(STDERR, "VLDB: no such entry\n");
276 fprintf(STDERR, "VLDB: vldb database is empty\n");
279 fprintf(STDERR, "VLDB: entry is deleted (soft delete)\n");
282 fprintf(STDERR, "VLDB: volume name is illegal\n");
285 fprintf(STDERR, "VLDB: index was out of range\n");
288 fprintf(STDERR, "VLDB: bad volume type\n");
291 fprintf(STDERR, "VLDB: illegal server number (not within limits)\n");
293 case VL_BADPARTITION:
294 fprintf(STDERR, "VLDB: bad partition number\n");
297 fprintf(STDERR, "VLDB: run out of space for replication sites\n");
300 fprintf(STDERR, "VLDB: no such repsite server exists\n");
302 case VL_DUPREPSERVER:
303 fprintf(STDERR, "VLDB: replication site server already exists\n");
306 fprintf(STDERR, "VLDB: parent r/w entry not found\n");
309 fprintf(STDERR, "VLDB: illegal reference count number\n");
311 case VL_SIZEEXCEEDED:
312 fprintf(STDERR, "VLDB: vldb size for attributes exceeded\n");
315 fprintf(STDERR, "VLDB: bad incoming vldb entry\n");
317 case VL_BADVOLIDBUMP:
318 fprintf(STDERR, "VLDB: illegal max volid increment\n");
320 case VL_IDALREADYHASHED:
321 fprintf(STDERR, "VLDB: (RO/BACK) Id already hashed\n");
324 fprintf(STDERR, "VLDB: vldb entry is already locked\n");
327 fprintf(STDERR, "VLDB: bad volume operation code\n");
329 case VL_BADRELLOCKTYPE:
330 fprintf(STDERR, "VLDB: bad release lock type\n");
333 fprintf(STDERR, "VLDB: status report: last release was aborted\n");
335 case VL_BADSERVERFLAG:
336 fprintf(STDERR, "VLDB: invalid replication site server flag\n");
339 fprintf(STDERR, "VLDB: no permission access for call\n");
341 case VOLSERREAD_DUMPERROR:
343 "VOLSER: Problems encountered in reading the dump file !\n");
345 case VOLSERDUMPERROR:
346 fprintf(STDERR, "VOLSER: Problems encountered in doing the dump !\n");
348 case VOLSERATTACH_ERROR:
349 fprintf(STDERR, "VOLSER: Could not attach the volume\n");
351 case VOLSERDETACH_ERROR:
352 fprintf(STDERR, "VOLSER: Could not detach the volume\n");
354 case VOLSERILLEGAL_PARTITION:
355 fprintf(STDERR, "VOLSER: encountered illegal partition number\n");
357 case VOLSERBAD_ACCESS:
358 fprintf(STDERR, "VOLSER: permission denied, not a super user\n");
360 case VOLSERVLDB_ERROR:
361 fprintf(STDERR, "VOLSER: error detected in the VLDB\n");
364 fprintf(STDERR, "VOLSER: error in volume name\n");
367 fprintf(STDERR, "VOLSER: volume has moved\n");
370 fprintf(STDERR, "VOLSER: illegal operation\n");
372 case VOLSERBADRELEASE:
373 fprintf(STDERR, "VOLSER: release could not be completed\n");
376 fprintf(STDERR, "VOLSER: volume is busy\n");
378 case VOLSERNO_MEMORY:
379 fprintf(STDERR, "VOLSER: volume server is out of memory\n");
383 "VOLSER: no such volume - location specified incorrectly or volume does not exist\n");
385 case VOLSERMULTIRWVOL:
387 "VOLSER: multiple RW volumes with same ID, one of which should be deleted\n");
391 "VOLSER: not all entries were successfully processed\n");
395 initialize_RXK_error_table();
396 initialize_KTC_error_table();
397 initialize_ACFG_error_table();
398 initialize_VL_error_table();
400 fprintf(STDERR, "%s: %s\n", afs_error_table_name(errcode),
401 afs_error_message(errcode));
408 void init_volintInfo(struct volintInfo *vinfo) {
409 memset(vinfo, 0, sizeof(struct volintInfo));
411 vinfo->maxquota = -1;
413 vinfo->creationDate = -1;
414 vinfo->updateDate = -1;
422 static struct rx_securityClass *uvclass = 0;
423 static int uvindex = -1;
424 /* called by VLDBClient_Init to set the security module to be used in the RPC */
426 UV_SetSecurity(struct rx_securityClass *as, afs_int32 aindex)
433 /* bind to volser on <port> <aserver> */
434 /* takes server address in network order, port in host order. dumb */
435 struct rx_connection *
436 UV_Bind(afs_uint32 aserver, afs_int32 port)
438 struct rx_connection *tc;
440 tc = rx_NewConnection(aserver, htons(port), VOLSERVICE_ID, uvclass,
446 AFSVolCreateVolume_retry(struct rx_connection *z_conn,
447 afs_int32 partition, char *name, afs_int32 type,
448 afs_int32 parent, afs_uint32 *volid, afs_int32 *trans)
453 code = AFSVolCreateVolume(z_conn, partition, name, type, parent,
455 if (code != VOLSERVOLBUSY)
458 #ifdef AFS_PTHREAD_ENV
461 IOMGR_Sleep(3-retries);
468 AFSVolTransCreate_retry(struct rx_connection *z_conn,
469 afs_int32 volume, afs_int32 partition,
470 afs_int32 flags, afs_int32 * trans)
475 code = AFSVolTransCreate(z_conn, volume, partition, flags, trans);
476 if (code != VOLSERVOLBUSY)
479 #ifdef AFS_PTHREAD_ENV
482 IOMGR_Sleep(3-retries);
489 /* if <okvol> is allright(indicated by beibg able to
490 * start a transaction, delete the <delvol> */
492 CheckAndDeleteVolume(struct rx_connection *aconn, afs_int32 apart,
493 afs_uint32 okvol, afs_uint32 delvol)
495 afs_int32 error, code, tid, rcode;
500 code = AFSVolTransCreate_retry(aconn, delvol, apart, ITOffline, &tid);
503 code = AFSVolDeleteVolume(aconn, tid);
506 code = AFSVolEndTrans(aconn, tid, &rcode);
513 code = AFSVolTransCreate_retry(aconn, okvol, apart, ITOffline, &tid);
515 code = AFSVolEndTrans(aconn, tid, &rcode);
520 code = AFSVolTransCreate_retry(aconn, delvol, apart, ITOffline, &tid);
523 code = AFSVolDeleteVolume(aconn, tid);
526 code = AFSVolEndTrans(aconn, tid, &rcode);
539 /* called by EmuerateEntry, show vldb entry in a reasonable format */
541 SubEnumerateEntry(struct nvldbentry *entry)
549 fprintf(STDOUT, " readWriteID %-10u ", entry->volumeId[RWVOL]);
550 if (entry->flags & VLF_RWEXISTS)
551 fprintf(STDOUT, " valid \n");
553 fprintf(STDOUT, " invalid \n");
554 fprintf(STDOUT, " readOnlyID %-10u ", entry->volumeId[ROVOL]);
555 if (entry->flags & VLF_ROEXISTS)
556 fprintf(STDOUT, " valid \n");
558 fprintf(STDOUT, " invalid \n");
559 fprintf(STDOUT, " backUpID %-10u ", entry->volumeId[BACKVOL]);
560 if (entry->flags & VLF_BACKEXISTS)
561 fprintf(STDOUT, " valid \n");
563 fprintf(STDOUT, " invalid \n");
564 if ((entry->cloneId != 0) && (entry->flags & VLF_ROEXISTS))
565 fprintf(STDOUT, " releaseClone %-10u \n", entry->cloneId);
567 if (entry->flags & VLF_RWEXISTS)
568 fprintf(STDOUT, " RWrite: %-10u", entry->volumeId[RWVOL]);
569 if (entry->flags & VLF_ROEXISTS)
570 fprintf(STDOUT, " ROnly: %-10u", entry->volumeId[ROVOL]);
571 if (entry->flags & VLF_BACKEXISTS)
572 fprintf(STDOUT, " Backup: %-10u", entry->volumeId[BACKVOL]);
573 if ((entry->cloneId != 0) && (entry->flags & VLF_ROEXISTS))
574 fprintf(STDOUT, " RClone: %-10lu", (unsigned long)entry->cloneId);
575 fprintf(STDOUT, "\n");
577 fprintf(STDOUT, " number of sites -> %lu\n",
578 (unsigned long)entry->nServers);
579 for (i = 0; i < entry->nServers; i++) {
580 if (entry->serverFlags[i] & VLSF_NEWREPSITE)
583 for (i = 0; i < entry->nServers; i++) {
584 MapPartIdIntoName(entry->serverPartition[i], pname);
585 fprintf(STDOUT, " server %s partition %s ",
586 noresolve ? afs_inet_ntoa_r(entry->serverNumber[i], hoststr) :
587 hostutil_GetNameByINet(entry->serverNumber[i]), pname);
588 if (entry->serverFlags[i] & VLSF_RWVOL)
589 fprintf(STDOUT, "RW Site ");
591 fprintf(STDOUT, "RO Site ");
593 if (entry->serverFlags[i] & VLSF_NEWREPSITE)
594 fprintf(STDOUT," -- New release");
596 if (!(entry->serverFlags[i] & VLSF_RWVOL))
597 fprintf(STDOUT," -- Old release");
599 if (entry->serverFlags[i] & VLSF_DONTUSE)
600 fprintf(STDOUT, " -- Not released");
602 fprintf(STDOUT, "\n");
609 /*enumerate the vldb entry corresponding to <entry> */
611 EnumerateEntry(struct nvldbentry *entry)
614 fprintf(STDOUT, "\n");
615 fprintf(STDOUT, "%s \n", entry->name);
616 SubEnumerateEntry(entry);
620 /* forcibly remove a volume. Very dangerous call */
622 UV_NukeVolume(afs_uint32 server, afs_int32 partid, afs_uint32 volid)
624 struct rx_connection *tconn;
627 tconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
629 code = AFSVolNukeVolume(tconn, partid, volid);
630 rx_DestroyConnection(tconn);
636 /* like df. Return usage of <pname> on <server> in <partition> */
638 UV_PartitionInfo64(afs_uint32 server, char *pname,
639 struct diskPartition64 *partition)
641 struct rx_connection *aconn;
644 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
645 code = AFSVolPartitionInfo64(aconn, pname, partition);
646 if (code == RXGEN_OPCODE) {
647 struct diskPartition *dpp = malloc(sizeof(struct diskPartition));
648 code = AFSVolPartitionInfo(aconn, pname, dpp);
650 strncpy(partition->name, dpp->name, 32);
651 strncpy(partition->devName, dpp->devName, 32);
652 partition->lock_fd = dpp->lock_fd;
653 partition->free = dpp->free;
654 partition->minFree = dpp->minFree;
659 fprintf(STDERR, "Could not get information on partition %s\n", pname);
660 PrintError("", code);
663 rx_DestroyConnection(aconn);
667 /* old interface to create volumes */
669 UV_CreateVolume(afs_uint32 aserver, afs_int32 apart, char *aname,
674 code = UV_CreateVolume2(aserver, apart, aname, 5000, 0, 0, 0, 0, anewid);
678 /* less old interface to create volumes */
680 UV_CreateVolume2(afs_uint32 aserver, afs_int32 apart, char *aname,
681 afs_int32 aquota, afs_int32 aspare1, afs_int32 aspare2,
682 afs_int32 aspare3, afs_int32 aspare4, afs_uint32 * anewid)
684 afs_uint32 roid = 0, bkid = 0;
685 return UV_CreateVolume3(aserver, apart, aname, aquota, aspare1, aspare2,
686 aspare3, aspare4, anewid, &roid, &bkid);
690 * Create a volume on the given server and partition
692 * @param aserver server to create volume on
693 * @param spart partition to create volume on
694 * @param aname name of new volume
695 * @param aquota quota for new volume
696 * @param anewid contains the desired volume id for the new volume. If
697 * *anewid == 0, a new id will be chosen, and will be placed
698 * in *anewid when UV_CreateVolume3 returns.
699 * @param aroid contains the desired RO volume id. If NULL, the RO id entry
700 * will be unset. If *aroid == 0, an id will be chosen, and
701 * will be placed in *anewid when UV_CreateVolume3 returns.
702 * @param abkid same as aroid, except for the BK volume id instead of the
704 * @return 0 on success, error code otherwise.
707 UV_CreateVolume3(afs_uint32 aserver, afs_int32 apart, char *aname,
708 afs_int32 aquota, afs_int32 aspare1, afs_int32 aspare2,
709 afs_int32 aspare3, afs_int32 aspare4, afs_uint32 * anewid,
710 afs_uint32 * aroid, afs_uint32 * abkid)
712 struct rx_connection *aconn;
716 afs_int32 rcode, vcode;
718 struct nvldbentry entry, storeEntry; /*the new vldb entry */
719 struct volintInfo tstatus;
724 memset(&storeEntry, 0, sizeof(struct nvldbentry));
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 = VLF_RWEXISTS; /* this records that rw volume exists */
799 entry.serverFlags[0] = VLSF_RWVOL; /*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 memset(&storeEntry, 0, sizeof(struct nvldbentry));
858 aconn = (struct rx_connection *)0;
861 /* set up the vldb entry for this volume */
862 strncpy(entry.name, aname, VOLSER_OLDMAXVOLNAME);
864 entry.serverNumber[0] = aserver; /* this should have another
865 * level of indirection later */
866 entry.serverPartition[0] = apart; /* this should also have
867 * another indirection level */
868 entry.flags = VLF_RWEXISTS; /* this records that rw volume exists */
869 entry.serverFlags[0] = VLSF_RWVOL; /*this rep site has rw vol */
870 entry.volumeId[RWVOL] = aid;
872 entry.volumeId[ROVOL] = anewid + 1; /* rw,ro, bk id are related in the default case */
873 entry.volumeId[BACKVOL] = *anewid + 2;
875 entry.volumeId[ROVOL] = 0;
876 entry.volumeId[BACKVOL] = 0;
879 /*map into right byte order, before passing to xdr, the stuff has to be in host
880 * byte order. Xdr converts it into network order */
881 MapNetworkToHost(&entry, &storeEntry);
882 /* create the vldb entry */
883 vcode = VLDB_CreateEntry(&storeEntry);
886 "Could not create a VLDB entry for the volume %s %lu\n",
887 aname, (unsigned long)aid);
891 VPRINT2("Created the VLDB entry for the volume %s %u\n", aname, aid);
895 rx_DestroyConnection(aconn);
896 PrintError("", error);
900 /* Delete the volume <volid>on <aserver> <apart>
901 * the physical entry gets removed from the vldb only if the ref count
905 UV_DeleteVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid)
907 struct rx_connection *aconn = (struct rx_connection *)0;
909 afs_int32 code, rcode;
911 struct nvldbentry entry, storeEntry;
913 afs_int32 avoltype = -1, vtype;
914 int notondisk = 0, notinvldb = 0;
916 memset(&storeEntry, 0, sizeof(struct nvldbentry));
918 /* Find and read bhe VLDB entry for this volume */
919 code = ubik_VL_SetLock(cstruct, 0, avolid, avoltype, VLOP_DELETE);
921 if (code != VL_NOENT) {
922 EGOTO1(error_exit, code,
923 "Could not lock VLDB entry for the volume %u\n", avolid);
929 code = VLDB_GetEntryByID(avolid, avoltype, &entry);
930 EGOTO1(error_exit, code, "Could not fetch VLDB entry for volume %u\n",
932 MapHostToNetwork(&entry);
935 EnumerateEntry(&entry);
938 /* Whether volume is in the VLDB or not. Delete the volume on disk */
939 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
941 code = DoVolDelete(aconn, avolid, apart, "the", 0, NULL, NULL);
951 /* Now update the VLDB entry.
952 * But first, verify we have a VLDB entry.
953 * Whether volume is on disk or not. Delete the volume in VLDB.
958 if (avolid == entry.volumeId[BACKVOL]) {
959 /* Its a backup volume, modify the VLDB entry. Check that the
960 * backup volume is on the server/partition we asked to delete.
962 if (!(entry.flags & VLF_BACKEXISTS) || !Lp_Match(aserver, apart, &entry)) {
963 notinvldb = 2; /* Not on this server and partition */
967 VPRINT1("Marking the backup volume %u deleted in the VLDB\n", avolid);
969 entry.flags &= ~VLF_BACKEXISTS;
973 else if (avolid == entry.volumeId[ROVOL]) {
974 /* Its a read-only volume, modify the VLDB entry. Check that the
975 * readonly volume is on the server/partition we asked to delete.
976 * If flags does not have RO_EIXSTS set, then this may mean the RO
977 * hasn't been released (and could exist in VLDB).
979 if (!Lp_ROMatch(aserver, apart, &entry)) {
980 notinvldb = 2; /* Not found on this server and partition */
986 "Marking the readonly volume %lu deleted in the VLDB\n",
987 (unsigned long)avolid);
989 Lp_SetROValue(&entry, aserver, apart, 0, 0); /* delete the site */
991 if (!Lp_ROMatch(0, 0, &entry))
992 entry.flags &= ~VLF_ROEXISTS; /* This was the last ro volume */
996 else if (avolid == entry.volumeId[RWVOL]) {
997 /* It's a rw volume, delete the backup volume, modify the VLDB entry.
998 * Check that the readwrite volumes is on the server/partition we
1001 if (!(entry.flags & VLF_RWEXISTS) || !Lp_Match(aserver, apart, &entry)) {
1002 notinvldb = 2; /* Not found on this server and partition */
1006 if (entry.volumeId[BACKVOL]) {
1007 /* Delete backup if it exists */
1008 code = DoVolDelete(aconn, entry.volumeId[BACKVOL], apart,
1009 "the backup", 0, NULL, NULL);
1010 if (code && code != VNOVOL) {
1018 "Marking the readwrite volume %lu%s deleted in the VLDB\n",
1019 (unsigned long)avolid,
1021 flags & VLF_BACKEXISTS) ? ", and its backup volume," :
1024 Lp_SetRWValue(&entry, aserver, apart, 0L, 0L);
1026 entry.flags &= ~(VLF_BACKEXISTS | VLF_RWEXISTS);
1029 if (entry.flags & VLF_ROEXISTS)
1030 fprintf(STDERR, "WARNING: ReadOnly copy(s) may still exist\n");
1034 notinvldb = 2; /* Not found on this server and partition */
1038 /* Either delete or replace the VLDB entry */
1039 if ((entry.nServers <= 0) || !(entry.flags & (VLF_ROEXISTS | VLF_RWEXISTS))) {
1042 "Last reference to the VLDB entry for %lu - deleting entry\n",
1043 (unsigned long)avolid);
1044 code = ubik_VL_DeleteEntry(cstruct, 0, avolid, vtype);
1045 EGOTO1(error_exit, code,
1046 "Could not delete the VLDB entry for the volume %u \n",
1049 MapNetworkToHost(&entry, &storeEntry);
1051 VLDB_ReplaceEntry(avolid, vtype, &storeEntry,
1052 (LOCKREL_OPCODE | LOCKREL_AFSID |
1053 LOCKREL_TIMESTAMP));
1054 EGOTO1(error_exit, code,
1055 "Could not update the VLDB entry for the volume %u \n",
1062 EPRINT(error, "\n");
1064 if (notondisk && notinvldb) {
1065 EPRINT2(VOLSERNOVOL, "Volume %u does not exist %s\n", avolid,
1066 ((notinvldb == 2) ? "on server and partition" : ""));
1068 error = VOLSERNOVOL;
1069 } else if (notondisk) {
1071 "WARNING: Volume %lu did not exist on the partition\n",
1072 (unsigned long)avolid);
1073 } else if (notinvldb) {
1074 fprintf(STDERR, "WARNING: Volume %lu does not exist in VLDB %s\n",
1075 (unsigned long)avolid,
1076 ((notinvldb == 2) ? "on server and partition" : ""));
1080 code = AFSVolEndTrans(aconn, ttid, &rcode);
1081 code = (code ? code : rcode);
1083 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
1084 (unsigned long)avolid);
1085 PrintError("", code);
1093 ubik_VL_ReleaseLock(cstruct, 0, avolid, -1,
1094 (LOCKREL_OPCODE | LOCKREL_AFSID |
1095 LOCKREL_TIMESTAMP));
1098 "Could not release the lock on the VLDB entry for the volume %u \n",
1106 rx_DestroyConnection(aconn);
1110 /* add recovery to UV_MoveVolume */
1112 #define TESTC 0 /* set to test recovery code, clear for production */
1118 do_interrupt(void * unused)
1121 #if !defined(AFS_PTHREAD_ENV) && !defined(AFS_NT40_ENV)
1122 /* Avoid UNIX LWP from getting confused that our stack has suddenly
1123 * changed. This will avoid some sanity checks, but until a better way
1124 * is found, the only alternative is always crashing and burning on at
1125 * least the stack-overflow check. */
1126 lwp_cpptr->stack = NULL;
1131 fprintf(STDOUT, "\nSIGINT handler: vos move operation in progress\n");
1133 "WARNING: may leave AFS storage and metadata in indeterminate state\n");
1134 fprintf(STDOUT, "enter second control-c to exit\n");
1142 sigint_handler(int x)
1144 #ifdef AFS_PTHREAD_ENV
1147 IOMGR_SoftSig(do_interrupt, 0);
1149 (void)signal(SIGINT, sigint_handler);
1153 DoVolDelete(struct rx_connection *aconn, afs_uint32 avolid,
1154 afs_int32 apart, char *ptypestring, afs_uint32 atoserver,
1155 struct volser_status *volstatus, char *pprefix)
1157 afs_int32 ttid = 0, code, rcode, error = 0;
1158 char *prefix, *typestring;
1167 typestring = ptypestring;
1173 VPRINT3("%sDeleting %s volume %u ...", prefix, typestring, avolid);
1176 AFSVolTransCreate_retry(aconn, avolid, apart, ITOffline, &ttid);
1178 /* return early and quietly for VNOVOL; don't continue the attempt to delete. */
1179 if (code == VNOVOL) {
1184 EGOTO2(dfail, code, "%sFailed to start transaction on %u\n",
1188 code = AFSVolGetStatus(aconn, ttid, volstatus);
1189 EGOTO2(dfail, code, "%sCould not get timestamp from volume %u\n",
1194 AFSVolSetFlags(aconn, ttid,
1195 VTDeleteOnSalvage | VTOutOfService);
1197 EGOTO2(dfail, code, "%sCould not set flags on volume %u \n",
1201 VPRINT1("%sSetting volume forwarding pointer ...", prefix);
1202 AFSVolSetForwarding(aconn, ttid, atoserver);
1206 code = AFSVolDeleteVolume(aconn, ttid);
1207 EGOTO2(dfail, code, "%sCould not delete volume %u\n", prefix, avolid);
1211 code = AFSVolEndTrans(aconn, ttid, &rcode);
1216 fprintf(STDERR, "%sCould not end transaction on %s volume %lu \n",
1217 prefix, typestring, (unsigned long)avolid);
1223 if (beverbose && !error)
1229 DoVolClone(struct rx_connection *aconn, afs_uint32 avolid,
1230 afs_int32 apart, int type, afs_uint32 cloneid,
1231 char *typestring, char *pname, char *vname, char *suffix,
1232 struct volser_status *volstatus, afs_int32 *transPtr)
1235 afs_int32 ttid = 0, btid = 0;
1236 afs_int32 code = 0, rcode = 0;
1237 afs_int32 error = 0;
1238 int cloneexists = 1;
1240 /* Test to see if the clone volume exists by trying to create
1241 * a transaction on the clone volume. We've assumed the clone exists.
1243 code = AFSVolTransCreate_retry(aconn, cloneid, apart, ITOffline, &btid);
1245 if (code != VNOVOL) {
1246 EPRINT2(code, "Could not reach the %s volume %lu\n",
1247 typestring, (unsigned long)cloneid);
1251 cloneexists = 0; /* clone volume does not exist */
1254 code = AFSVolEndTrans(aconn, btid, &rcode);
1256 if (code || rcode) {
1258 "Could not end transaction on the previous %s volume %lu\n",
1259 typestring, (unsigned long)cloneid);
1260 error = (code ? code : rcode);
1265 /* Now go ahead and try to clone the RW volume.
1266 * First start a transaction on the RW volume
1268 code = AFSVolTransCreate_retry(aconn, avolid, apart, ITBusy, &ttid);
1270 fprintf(STDERR, "Could not start a transaction on the volume %lu\n",
1271 (unsigned long)avolid);
1276 /* Clone or reclone the volume, depending on whether the clone
1277 * volume exists or not
1280 VPRINT2("Re-cloning %s volume %u ...", typestring, cloneid);
1282 code = AFSVolReClone(aconn, ttid, cloneid);
1284 EPRINT2(code, "Could not re-clone %s volume %lu\n",
1285 typestring, (unsigned long)cloneid);
1290 VPRINT2("Creating a new %s clone %u ...", typestring, cloneid);
1293 strcpy(cname, pname);
1294 strcat(cname, suffix);
1297 code = AFSVolClone(aconn, ttid, 0, type, vname?vname:cname,
1300 fprintf(STDERR, "Failed to clone the volume %lu\n",
1301 (unsigned long)avolid);
1310 VPRINT1("Getting status of parent volume %u...", avolid);
1311 code = AFSVolGetStatus(aconn, ttid, volstatus);
1313 fprintf(STDERR, "Failed to get the status of the parent volume %lu\n",
1314 (unsigned long)avolid);
1323 code = AFSVolEndTrans(aconn, ttid, &rcode);
1324 if (code || rcode) {
1325 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
1326 (unsigned long)avolid);
1328 error = (code ? code : rcode);
1333 code = AFSVolEndTrans(aconn, btid, &rcode);
1334 if (code || rcode) {
1336 "Could not end transaction on the %s volume %lu\n",
1337 typestring, (unsigned long)cloneid);
1339 error = (code ? code : rcode);
1345 /* Convert volume from RO to RW; adjust the VLDB entry to match.
1346 * The nvldbentry passed to us has already been MapHostToNetwork'd
1351 UV_ConvertRO(afs_uint32 server, afs_uint32 partition, afs_uint32 volid,
1352 struct nvldbentry *entry)
1354 afs_int32 code, i, same;
1355 struct nvldbentry checkEntry, storeEntry;
1357 afs_int32 rwindex = 0;
1358 afs_uint32 rwserver = 0;
1359 afs_int32 roindex = 0;
1360 afs_uint32 roserver = 0;
1361 struct rx_connection *aconn;
1363 memset(&storeEntry, 0, sizeof(struct nvldbentry));
1366 ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL,
1370 "Unable to lock volume %lu, code %d\n",
1371 (unsigned long)entry->volumeId[RWVOL],vcode);
1372 PrintError("", vcode);
1376 /* make sure the VLDB entry hasn't changed since we started */
1377 memset(&checkEntry, 0, sizeof(checkEntry));
1378 vcode = VLDB_GetEntryByID(volid, -1, &checkEntry);
1381 "Could not fetch the entry for volume %lu from VLDB\n",
1382 (unsigned long)volid);
1383 PrintError("convertROtoRW ", vcode);
1388 MapHostToNetwork(&checkEntry);
1389 entry->flags &= ~VLOP_ALLOPERS; /* clear any stale lock operation flags */
1390 entry->flags |= VLOP_MOVE; /* set to match SetLock operation above */
1391 if (memcmp(entry, &checkEntry, sizeof(*entry)) != 0) {
1393 "VLDB entry for volume %lu has changed; please reissue the command.\n",
1394 (unsigned long)volid);
1399 /* extract information from the original entry */
1400 for (i = 0; i < entry->nServers; i++) {
1401 if (entry->serverFlags[i] & VLSF_RWVOL) {
1403 rwserver = entry->serverNumber[i];
1404 /* rwpartition = entry->serverPartition[i]; */
1407 } else if ((entry->serverFlags[i] & VLSF_ROVOL) && !roserver) {
1408 same = VLDB_IsSameAddrs(server, entry->serverNumber[i], &code);
1411 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
1418 roserver = entry->serverNumber[i];
1419 /* ropartition = entry->serverPartition[i]; */
1426 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
1427 code = AFSVolConvertROtoRWvolume(aconn, partition, volid);
1430 "Converting RO volume %lu to RW volume failed with code %d\n",
1431 (unsigned long)volid, code);
1432 PrintError("convertROtoRW ", code);
1435 /* Update the VLDB to match what we did on disk as much as possible. */
1436 /* If the converted RO was in the VLDB, make it look like the new RW. */
1438 entry->serverFlags[roindex] = VLSF_RWVOL;
1440 /* Add a new site entry for the newly created RW. It's possible
1441 * (but unlikely) that we are already at MAXNSERVERS and that this
1442 * new site will invalidate the whole VLDB entry; however,
1443 * VLDB_ReplaceEntry will detect this and return VL_BADSERVER,
1444 * so we need no extra guard logic here.
1446 afs_int32 newrwindex = entry->nServers;
1447 (entry->nServers)++;
1448 entry->serverNumber[newrwindex] = server;
1449 entry->serverPartition[newrwindex] = partition;
1450 entry->serverFlags[newrwindex] = VLSF_RWVOL;
1452 entry->flags |= VLF_RWEXISTS;
1453 entry->flags &= ~VLF_BACKEXISTS;
1455 /* if the old RW was in the VLDB, remove it by decrementing the number */
1456 /* of servers, replacing the RW entry with the last entry, and zeroing */
1457 /* out the last entry. */
1459 (entry->nServers)--;
1460 if (rwindex != entry->nServers) {
1461 entry->serverNumber[rwindex] = entry->serverNumber[entry->nServers];
1462 entry->serverPartition[rwindex] =
1463 entry->serverPartition[entry->nServers];
1464 entry->serverFlags[rwindex] = entry->serverFlags[entry->nServers];
1465 entry->serverNumber[entry->nServers] = 0;
1466 entry->serverPartition[entry->nServers] = 0;
1467 entry->serverFlags[entry->nServers] = 0;
1470 entry->flags &= ~VLF_ROEXISTS;
1471 for (i = 0; i < entry->nServers; i++) {
1472 if (entry->serverFlags[i] & VLSF_ROVOL) {
1473 if (!(entry->serverFlags[i] & (VLSF_DONTUSE | VLSF_NEWREPSITE)))
1474 entry->flags |= VLF_ROEXISTS;
1477 MapNetworkToHost(entry, &storeEntry);
1479 VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry,
1480 (LOCKREL_OPCODE | LOCKREL_AFSID |
1481 LOCKREL_TIMESTAMP));
1484 "Warning: volume converted, but vldb update failed with code %d!\n",
1489 vcode = UV_LockRelease(entry->volumeId[RWVOL]);
1492 "Unable to unlock volume %lu, code %d\n",
1493 (unsigned long)entry->volumeId[RWVOL],vcode);
1494 PrintError("", vcode);
1500 /* Move volume <afromvol> on <afromserver> <afrompart> to <atoserver>
1501 * <atopart>. The operation is almost idempotent. The following
1502 * flags are recognized:
1504 * RV_NOCLONE - don't use a copy clone
1508 UV_MoveVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
1509 afs_uint32 atoserver, afs_int32 atopart, int flags)
1511 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
1512 * be changing during the move */
1513 struct rx_connection * volatile toconn;
1514 struct rx_connection * volatile fromconn;
1515 afs_int32 volatile fromtid;
1516 afs_int32 volatile totid;
1517 afs_int32 volatile clonetid;
1518 afs_uint32 volatile newVol;
1519 afs_uint32 volatile volid;
1520 afs_uint32 volatile backupId;
1521 int volatile islocked;
1526 char tmpName[VOLSER_MAXVOLNAME + 1];
1531 struct restoreCookie cookie;
1532 afs_int32 vcode, code;
1533 struct volser_status tstatus;
1534 struct destServer destination;
1536 struct nvldbentry entry, storeEntry;
1539 char in, lf; /* for test code */
1543 #ifdef ENABLE_BUGFIX_1165
1544 volEntries volumeInfo;
1545 struct volintInfo *infop = 0;
1549 fromconn = (struct rx_connection *)0;
1550 toconn = (struct rx_connection *)0;
1560 /* support control-c processing */
1563 (void)signal(SIGINT, sigint_handler);
1567 "\nThere are three tests points - verifies all code paths through recovery.\n");
1568 fprintf(STDOUT, "First test point - operation not started.\n");
1569 fprintf(STDOUT, "...test here (y, n)? ");
1571 if (fscanf(stdin, "%c", &in) < 1)
1573 if (fscanf(stdin, "%c", &lf) < 0) {
1574 /* toss away; don't care */
1577 fprintf(STDOUT, "type control-c\n");
1579 fprintf(stdout, ".");
1584 /* or drop through */
1587 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
1588 EGOTO1(mfail, vcode,
1589 "Could not fetch the entry for the volume %u from the VLDB \n",
1592 if (entry.volumeId[RWVOL] != afromvol) {
1593 fprintf(STDERR, "Only RW volume can be moved\n");
1597 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_MOVE);
1598 EGOTO1(mfail, vcode, "Could not lock entry for volume %u \n", afromvol);
1601 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
1602 EGOTO1(mfail, vcode,
1603 "Could not fetch the entry for the volume %u from the VLDB \n",
1606 backupId = entry.volumeId[BACKVOL];
1607 MapHostToNetwork(&entry);
1609 if (!Lp_Match(afromserver, afrompart, &entry)) {
1610 /* the from server and partition do not exist in the vldb entry corresponding to volid */
1611 if (!Lp_Match(atoserver, atopart, &entry)) {
1612 /* the to server and partition do not exist in the vldb entry corresponding to volid */
1613 fprintf(STDERR, "The volume %lu is not on the specified site. \n",
1614 (unsigned long)afromvol);
1615 fprintf(STDERR, "The current site is :");
1616 for (i = 0; i < entry.nServers; i++) {
1617 if (entry.serverFlags[i] == VLSF_RWVOL) {
1619 MapPartIdIntoName(entry.serverPartition[i], pname);
1620 fprintf(STDERR, " server %s partition %s \n",
1621 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
1622 hostutil_GetNameByINet(entry.serverNumber[i]),
1627 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
1628 (LOCKREL_OPCODE | LOCKREL_AFSID |
1629 LOCKREL_TIMESTAMP));
1630 EGOTO1(mfail, vcode,
1631 " Could not release lock on the VLDB entry for the volume %u \n",
1634 return VOLSERVOLMOVED;
1637 /* delete the volume afromvol on src_server */
1638 /* from-info does not exist but to-info does =>
1639 * we have already done the move, but the volume
1640 * may still be existing physically on from fileserver
1642 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1645 code = DoVolDelete(fromconn, afromvol, afrompart,
1646 "leftover", 0, NULL, NULL);
1647 if (code && code != VNOVOL) {
1652 code = DoVolDelete(fromconn, backupId, afrompart,
1653 "leftover backup", 0, NULL, NULL);
1654 if (code && code != VNOVOL) {
1664 /* From-info matches the vldb info about volid,
1665 * its ok start the move operation, the backup volume
1666 * on the old site is deleted in the process
1668 if (afrompart == atopart) {
1669 same = VLDB_IsSameAddrs(afromserver, atoserver, &error);
1670 EGOTO2(mfail, error,
1671 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
1672 afromserver, error);
1675 EGOTO1(mfail, VOLSERVOLMOVED,
1676 "Warning: Moving volume %u to its home partition ignored!\n",
1682 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
1683 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
1684 totid = 0; /* initialize to uncreated */
1687 * clone the read/write volume locally.
1690 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1691 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1693 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
1697 if (!(flags & RV_NOCLONE)) {
1698 /* Get a clone id */
1699 VPRINT1("Allocating new volume id for clone of volume %u ...",
1702 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
1704 EGOTO1(mfail, vcode,
1705 "Could not get an ID for the clone of volume %u from the VLDB\n",
1709 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
1710 VPRINT1("Cloning source volume %u ...", afromvol);
1711 strcpy(vname, "move-clone-temp");
1713 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname, &tmpVol);
1715 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
1720 /* lookup the name of the volume we just cloned */
1722 code = AFSVolGetName(fromconn, fromtid, &volName);
1723 EGOTO1(mfail, code, "Failed to get the name of the volume %u\n",
1726 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
1728 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
1733 "Failed to end the transaction on the source volume %u\n",
1738 * Create the destination volume
1741 if (!(flags & RV_NOCLONE)) {
1742 /* All of this is to get the fromDate */
1743 VPRINT1("Starting transaction on the cloned volume %u ...", newVol);
1746 AFSVolTransCreate_retry(fromconn, newVol, afrompart, ITOffline,
1750 "Failed to start a transaction on the cloned volume%u\n",
1754 VPRINT1("Setting flags on cloned volume %u ...", newVol);
1756 AFSVolSetFlags(fromconn, clonetid,
1757 VTDeleteOnSalvage | VTOutOfService); /*redundant */
1758 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
1762 /* remember time from which we've dumped the volume */
1763 VPRINT1("Getting status of cloned volume %u ...", newVol);
1764 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
1766 "Failed to get the status of the cloned volume %u\n",
1770 fromDate = CLOCKADJ(tstatus.creationDate);
1772 /* With RV_NOCLONE, just do a full copy from the source */
1777 #ifdef ENABLE_BUGFIX_1165
1779 * Get the internal volume state from the source volume. We'll use such info (i.e. dayUse)
1780 * to copy it to the new volume (via AFSSetInfo later on) so that when we move volumes we
1781 * don't use this information...
1783 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
1784 volumeInfo.volEntries_len = 0;
1785 code = AFSVolListOneVolume(fromconn, afrompart, afromvol, &volumeInfo);
1787 "Failed to get the volint Info of the cloned volume %u\n",
1790 infop = (volintInfo *) volumeInfo.volEntries_val;
1791 infop->maxquota = -1; /* Else it will replace the default quota */
1792 infop->creationDate = -1; /* Else it will use the source creation date */
1793 infop->updateDate = -1; /* Else it will use the source update date */
1796 /* create a volume on the target machine */
1798 code = DoVolDelete(toconn, volid, atopart,
1799 "pre-existing destination", 0, NULL, NULL);
1800 if (code && code != VNOVOL) {
1805 VPRINT1("Creating the destination volume %u ...", volid);
1809 AFSVolCreateVolume(toconn, atopart, volName, volser_RW, volid, &tmpVol,
1813 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
1817 strncpy(tmpName, volName, VOLSER_OLDMAXVOLNAME);
1821 VPRINT1("Setting volume flags on destination volume %u ...", volid);
1823 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
1825 "Failed to set the flags on the destination volume %u\n", volid);
1829 * Now dump the clone to the new volume
1832 destination.destHost = ntohl(atoserver);
1833 destination.destPort = AFSCONF_VOLUMEPORT;
1834 destination.destSSID = 1;
1836 strncpy(cookie.name, tmpName, VOLSER_OLDMAXVOLNAME);
1837 cookie.type = RWVOL;
1838 cookie.parent = entry.volumeId[RWVOL];
1841 if (!(flags & RV_NOCLONE)) {
1842 /* Copy the clone to the new volume */
1843 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
1846 AFSVolForward(fromconn, clonetid, 0, &destination, totid,
1848 EGOTO1(mfail, code, "Failed to move data for the volume %u\n", volid);
1851 VPRINT1("Ending transaction on cloned volume %u ...", newVol);
1852 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
1857 "Failed to end the transaction on the cloned volume %u\n",
1863 * reattach to the main-line volume, and incrementally dump it.
1866 VPRINT1("Starting transaction on source volume %u ...", afromvol);
1868 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
1871 "Failed to create a transaction on the source volume %u\n",
1875 /* now do the incremental */
1877 ("Doing the%s dump from source to destination for volume %u ... ",
1878 (flags & RV_NOCLONE) ? "" : " incremental",
1881 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
1884 "Failed to do the%s dump from rw volume on old site to rw volume on newsite\n",
1885 (flags & RV_NOCLONE) ? "" : " incremental");
1888 /* now adjust the flags so that the new volume becomes official */
1889 VPRINT1("Setting volume flags on old source volume %u ...", afromvol);
1890 code = AFSVolSetFlags(fromconn, fromtid, VTOutOfService);
1892 "Failed to set the flags to make old source volume offline\n");
1895 VPRINT1("Setting volume flags on new source volume %u ...", afromvol);
1896 code = AFSVolSetFlags(toconn, totid, 0);
1898 "Failed to set the flags to make new source volume online\n");
1901 #ifdef ENABLE_BUGFIX_1165
1902 VPRINT1("Setting volume status on destination volume %u ...", volid);
1903 code = AFSVolSetInfo(toconn, totid, infop);
1905 "Failed to set volume status on the destination volume %u\n",
1910 /* put new volume online */
1911 VPRINT1("Ending transaction on destination volume %u ...", afromvol);
1912 code = AFSVolEndTrans(toconn, totid, &rcode);
1917 "Failed to end the transaction on the volume %u on the new site\n",
1921 Lp_SetRWValue(&entry, afromserver, afrompart, atoserver, atopart);
1922 MapNetworkToHost(&entry, &storeEntry);
1923 storeEntry.flags &= ~VLF_BACKEXISTS;
1927 "Second test point - operation in progress but not complete.\n");
1928 fprintf(STDOUT, "...test here (y, n)? ");
1930 if (fscanf(stdin, "%c", &in) < 1)
1932 if (fscanf(stdin, "%c", &lf) < 0) {
1933 /* toss away, don't care */
1936 fprintf(STDOUT, "type control-c\n");
1938 fprintf(stdout, ".");
1943 /* or drop through */
1946 VPRINT1("Releasing lock on VLDB entry for volume %u ...", afromvol);
1948 VLDB_ReplaceEntry(afromvol, -1, &storeEntry,
1949 (LOCKREL_OPCODE | LOCKREL_AFSID |
1950 LOCKREL_TIMESTAMP));
1953 " Could not release the lock on the VLDB entry for the volume %s %lu \n",
1954 storeEntry.name, (unsigned long)afromvol);
1963 "Third test point - operation complete but no cleanup.\n");
1964 fprintf(STDOUT, "...test here (y, n)? ");
1966 if (fscanf(stdin, "%c", &in) < 1)
1968 if (fscanf(stdin, "%c", &lf) < 0) {
1969 /* toss away; don't care */
1972 fprintf(STDOUT, "type control-c\n");
1974 fprintf(stdout, ".");
1979 /* or drop through */
1982 /* This is tricky. File server is very stupid, and if you mark the volume
1983 * as VTOutOfService, it may mark the *good* instance (if you're moving
1984 * between partitions on the same machine) as out of service. Since
1985 * we're cleaning this code up in DEcorum, we're just going to kludge around
1986 * it for now by removing this call. */
1987 /* already out of service, just zap it now */
1989 AFSVolSetFlags(fromconn, fromtid, VTDeleteOnSalvage | VTOutOfService);
1992 "Failed to set the flags to make the old source volume offline\n");
1996 if (atoserver != afromserver) {
1997 /* set forwarding pointer for moved volumes */
1998 VPRINT1("Setting forwarding pointer for volume %u ...", afromvol);
1999 code = AFSVolSetForwarding(fromconn, fromtid, atoserver);
2001 "Failed to set the forwarding pointer for the volume %u\n",
2006 VPRINT1("Deleting old volume %u on source ...", afromvol);
2007 code = AFSVolDeleteVolume(fromconn, fromtid); /* zap original volume */
2008 EGOTO1(mfail, code, "Failed to delete the old volume %u on source\n",
2012 VPRINT1("Ending transaction on old volume %u on the source ...",
2014 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2019 "Failed to end the transaction on the old volume %u on the source\n",
2023 code = DoVolDelete(fromconn, backupId, afrompart,
2024 "source backup", 0, NULL, NULL);
2025 if (code && code != VNOVOL) {
2030 code = 0; /* no backup volume? that's okay */
2033 if (!(flags & RV_NOCLONE)) {
2034 code = DoVolDelete(fromconn, newVol, afrompart,
2035 "cloned", 0, NULL, NULL);
2037 if (code == VNOVOL) {
2038 EPRINT1(code, "Failed to start transaction on %u\n", newVol);
2049 fprintf(STDOUT, "Fourth test point - operation complete.\n");
2050 fprintf(STDOUT, "...test here (y, n)? ");
2052 if (fscanf(stdin, "%c", &in) < 1)
2054 if (fscanf(stdin, "%c", &lf) < 0) { /* toss away */
2058 fprintf(STDOUT, "type control-c\n");
2060 fprintf(stdout, ".");
2065 /* or drop through */
2068 /* normal cleanup code */
2070 if (entry.flags & VLF_ROEXISTS)
2071 fprintf(STDERR, "WARNING : readOnly copies still exist \n");
2074 VPRINT1("Cleanup: Releasing VLDB lock on volume %u ...", afromvol);
2076 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2077 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2081 " Could not release the lock on the VLDB entry for the volume %lu \n",
2082 (unsigned long)afromvol);
2090 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
2092 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2093 if (code || rcode) {
2096 "Could not end transaction on the source volume %lu\n",
2097 (unsigned long)afromvol);
2099 error = (code ? code : rcode);
2105 VPRINT1("Cleanup: Ending transaction on clone volume %u ...", newVol);
2106 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2107 if (code || rcode) {
2110 "Could not end transaction on the source's clone volume %lu\n",
2111 (unsigned long)newVol);
2113 error = (code ? code : rcode);
2119 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
2121 code = AFSVolEndTrans(toconn, totid, &rcode);
2125 "Could not end transaction on destination volume %lu\n",
2126 (unsigned long)afromvol);
2128 error = (code ? code : rcode);
2134 #ifdef ENABLE_BUGFIX_1165
2139 rx_DestroyConnection(fromconn);
2141 rx_DestroyConnection(toconn);
2142 PrintError("", error);
2145 /* come here only when the sky falls */
2150 "vos move: operation interrupted, cleanup in progress...\n");
2151 fprintf(STDOUT, "clear transaction contexts\n");
2155 /* unlock VLDB entry */
2157 VPRINT1("Recovery: Releasing VLDB lock on volume %u ...", afromvol);
2158 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2159 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2165 VPRINT("Recovery: Ending transaction on clone volume ...");
2166 AFSVolEndTrans(fromconn, clonetid, &rcode);
2170 VPRINT("Recovery: Ending transaction on destination volume ...");
2171 AFSVolEndTrans(toconn, totid, &rcode);
2174 if (fromtid) { /* put it on-line */
2175 VPRINT("Recovery: Setting volume flags on source volume ...");
2176 AFSVolSetFlags(fromconn, fromtid, 0);
2179 VPRINT("Recovery: Ending transaction on source volume ...");
2180 AFSVolEndTrans(fromconn, fromtid, &rcode);
2184 VPRINT("Recovery: Accessing VLDB.\n");
2185 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2187 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2191 MapHostToNetwork(&entry);
2193 /* Delete either the volume on the source location or the target location.
2194 * If the vldb entry still points to the source location, then we know the
2195 * volume move didn't finish so we remove the volume from the target
2196 * location. Otherwise, we remove the volume from the source location.
2198 if (Lp_Match(afromserver, afrompart, &entry)) { /* didn't move - delete target volume */
2201 "move incomplete - attempt cleanup of target partition - no guarantee\n");
2205 if (volid && toconn) {
2206 code = DoVolDelete(toconn, volid, atopart,
2207 "destination", 0, NULL, "Recovery:");
2208 if (code == VNOVOL) {
2209 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", volid);
2213 /* put source volume on-line */
2215 VPRINT1("Recovery: Creating transaction on source volume %u ...",
2219 AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2225 VPRINT1("Recovery: Setting flags on source volume %u ...",
2227 AFSVolSetFlags(fromconn, fromtid, 0);
2231 ("Recovery: Ending transaction on source volume %u ...",
2233 AFSVolEndTrans(fromconn, fromtid, &rcode);
2237 ("\nRecovery: Unable to start transaction on source volume %u.\n",
2241 } else { /* yep, move complete */
2244 "move complete - attempt cleanup of source partition - no guarantee\n");
2248 /* delete backup volume */
2250 code = DoVolDelete(fromconn, backupId, afrompart,
2251 "backup", 0, NULL, "Recovery:");
2252 if (code == VNOVOL) {
2253 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", backupId);
2256 code = DoVolDelete(fromconn, afromvol, afrompart, "source",
2257 (atoserver != afromserver)?atoserver:0,
2259 if (code == VNOVOL) {
2260 EPRINT1(code, "Failed to start transaction on %u\n", afromvol);
2265 /* common cleanup - delete local clone */
2267 code = DoVolDelete(fromconn, newVol, afrompart,
2268 "clone", 0, NULL, "Recovery:");
2269 if (code == VNOVOL) {
2270 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", newVol);
2274 /* unlock VLDB entry */
2276 VPRINT1("Recovery: Releasing lock on VLDB entry for volume %u ...",
2278 ubik_VL_ReleaseLock(cstruct, 0, afromvol, -1,
2279 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
2282 done: /* routine cleanup */
2285 #ifdef ENABLE_BUGFIX_1165
2290 rx_DestroyConnection(fromconn);
2292 rx_DestroyConnection(toconn);
2295 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2303 UV_MoveVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2304 afs_uint32 atoserver, afs_int32 atopart)
2306 return UV_MoveVolume2(afromvol, afromserver, afrompart,
2307 atoserver, atopart, 0);
2311 /* Copy volume <afromvol> from <afromserver> <afrompart> to <atoserver>
2312 * <atopart>. The new volume is named by <atovolname>. The new volume
2313 * has ID <atovolid> if that is nonzero; otherwise a new ID is allocated
2314 * from the VLDB. the following flags are supported:
2316 * RV_RDONLY - target volume is RO
2317 * RV_OFFLINE - leave target volume offline
2318 * RV_CPINCR - do incremental dump if target exists
2319 * RV_NOVLDB - don't create/update VLDB entry
2320 * RV_NOCLONE - don't use a copy clone
2323 UV_CopyVolume2(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2324 char *atovolname, afs_uint32 atoserver, afs_int32 atopart,
2325 afs_uint32 atovolid, int flags)
2327 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
2328 * be changing during the copy */
2330 afs_int32 volatile clonetid;
2331 afs_int32 volatile totid;
2332 afs_int32 volatile fromtid;
2333 struct rx_connection * volatile fromconn;
2334 struct rx_connection * volatile toconn;
2335 afs_uint32 volatile cloneVol;
2339 afs_int32 fromDate, cloneFromDate;
2340 struct restoreCookie cookie;
2341 afs_int32 vcode, code;
2344 struct volser_status tstatus;
2345 struct destServer destination;
2346 struct nvldbentry entry, newentry, storeEntry;
2351 fromconn = (struct rx_connection *)0;
2352 toconn = (struct rx_connection *)0;
2360 /* support control-c processing */
2363 (void)signal(SIGINT, sigint_handler);
2365 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2366 EGOTO1(mfail, vcode,
2367 "Could not fetch the entry for the volume %u from the VLDB \n",
2369 MapHostToNetwork(&entry);
2372 toconn = UV_Bind(atoserver, AFSCONF_VOLUMEPORT); /* get connections to the servers */
2373 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
2374 fromtid = totid = 0; /* initialize to uncreated */
2377 * clone the read/write volume locally.
2381 if (!(flags & RV_NOCLONE)) {
2382 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2384 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy,
2387 EGOTO1(mfail, code, "Failed to create transaction on the volume %u\n",
2391 /* Get a clone id */
2392 VPRINT1("Allocating new volume id for clone of volume %u ...",
2396 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
2398 EGOTO1(mfail, vcode,
2399 "Could not get an ID for the clone of volume %u from the VLDB\n",
2407 /* Get a new volume id */
2408 VPRINT1("Allocating new volume id for copy of volume %u ...", afromvol);
2410 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &newVol);
2411 EGOTO1(mfail, vcode,
2412 "Could not get an ID for the copy of volume %u from the VLDB\n",
2417 if (!(flags & RV_NOCLONE)) {
2418 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
2419 VPRINT1("Cloning source volume %u ...", afromvol);
2420 strcpy(vname, "copy-clone-temp");
2423 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, vname,
2426 EGOTO1(mfail, code, "Failed to clone the source volume %u\n",
2430 VPRINT1("Ending the transaction on the source volume %u ...", afromvol);
2432 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2437 "Failed to end the transaction on the source volume %u\n",
2443 * Create the destination volume
2446 if (!(flags & RV_NOCLONE)) {
2447 VPRINT1("Starting transaction on the cloned volume %u ...", cloneVol);
2450 AFSVolTransCreate_retry(fromconn, cloneVol, afrompart, ITOffline,
2454 "Failed to start a transaction on the cloned volume%u\n",
2458 VPRINT1("Setting flags on cloned volume %u ...", cloneVol);
2460 AFSVolSetFlags(fromconn, clonetid,
2461 VTDeleteOnSalvage | VTOutOfService); /*redundant */
2462 EGOTO1(mfail, code, "Could not set flags on the cloned volume %u\n",
2466 /* remember time from which we've dumped the volume */
2467 VPRINT1("Getting status of cloned volume %u ...", cloneVol);
2468 code = AFSVolGetStatus(fromconn, clonetid, &tstatus);
2470 "Failed to get the status of the cloned volume %u\n",
2474 fromDate = CLOCKADJ(tstatus.creationDate);
2479 /* create a volume on the target machine */
2482 code = AFSVolTransCreate_retry(toconn, newVol, atopart, ITOffline, &tmp);
2485 if ((flags & RV_CPINCR)) {
2486 VPRINT1("Getting status of pre-existing volume %u ...", newVol);
2487 code = AFSVolGetStatus(toconn, totid, &tstatus);
2489 "Failed to get the status of the pre-existing volume %u\n",
2493 /* Using the update date should be OK here, but add some fudge */
2494 cloneFromDate = CLOCKADJ(tstatus.updateDate);
2495 if ((flags & RV_NOCLONE))
2496 fromDate = cloneFromDate;
2498 /* XXX We should check that the source volume's creationDate is
2499 * XXX not newer than the existing target volume, and if not,
2500 * XXX throw away the existing target and do a full dump. */
2505 /* Delete the existing volume.
2506 * While we are deleting the volume in these steps, the transaction
2507 * we started against the cloned volume (clonetid above) will be
2508 * sitting idle. It will get cleaned up after 600 seconds
2510 VPRINT1("Deleting pre-existing volume %u on destination ...", newVol);
2511 code = AFSVolDeleteVolume(toconn, totid);
2513 "Could not delete the pre-existing volume %u on destination\n",
2518 ("Ending transaction on pre-existing volume %u on destination ...",
2520 code = AFSVolEndTrans(toconn, totid, &rcode);
2525 "Could not end the transaction on pre-existing volume %u on destination\n",
2530 VPRINT1("Creating the destination volume %u ...", newVol);
2533 AFSVolCreateVolume(toconn, atopart, atovolname,
2534 (flags & RV_RDONLY) ? volser_RO : volser_RW,
2535 newVol, &newVol, &tmp);
2537 EGOTO1(mfail, code, "Failed to create the destination volume %u\n",
2541 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2543 AFSVolSetFlags(toconn, totid, (VTDeleteOnSalvage | VTOutOfService));
2545 "Failed to set the flags on the destination volume %u\n", newVol);
2550 destination.destHost = ntohl(atoserver);
2551 destination.destPort = AFSCONF_VOLUMEPORT;
2552 destination.destSSID = 1;
2554 strncpy(cookie.name, atovolname, VOLSER_OLDMAXVOLNAME);
2555 cookie.type = (flags & RV_RDONLY) ? ROVOL : RWVOL;
2560 * Now dump the clone to the new volume
2563 if (!(flags & RV_NOCLONE)) {
2564 /* XXX probably should have some code here that checks to see if
2565 * XXX we are copying to same server and partition - if so, just
2566 * XXX use a clone to save disk space */
2568 /* Copy the clone to the new volume */
2569 VPRINT2("Dumping from clone %u on source to volume %u on destination ...",
2572 AFSVolForward(fromconn, clonetid, cloneFromDate, &destination,
2574 EGOTO1(mfail, code, "Failed to move data for the volume %u\n",
2578 VPRINT1("Ending transaction on cloned volume %u ...", cloneVol);
2579 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2584 "Failed to end the transaction on the cloned volume %u\n",
2590 * reattach to the main-line volume, and incrementally dump it.
2593 VPRINT1("Starting transaction on source volume %u ...", afromvol);
2595 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
2598 "Failed to create a transaction on the source volume %u\n",
2602 /* now do the incremental */
2604 ("Doing the%s dump from source to destination for volume %u ... ",
2605 (flags & RV_NOCLONE) ? "" : " incremental",
2608 AFSVolForward(fromconn, fromtid, fromDate, &destination, totid,
2611 "Failed to do the%s dump from old site to new site\n",
2612 (flags & RV_NOCLONE) ? "" : " incremental");
2615 VPRINT1("Setting volume flags on destination volume %u ...", newVol);
2616 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
2617 code = AFSVolSetFlags(toconn, totid, volflag);
2619 "Failed to set the flags to make destination volume online\n");
2622 /* put new volume online */
2623 VPRINT1("Ending transaction on destination volume %u ...", newVol);
2624 code = AFSVolEndTrans(toconn, totid, &rcode);
2629 "Failed to end the transaction on the destination volume %u\n",
2633 VPRINT1("Ending transaction on source volume %u ...", afromvol);
2634 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2639 "Failed to end the transaction on the source volume %u\n",
2645 if (!(flags & RV_NOCLONE)) {
2646 code = DoVolDelete(fromconn, cloneVol, afrompart,
2647 "cloned", 0, NULL, NULL);
2649 if (code == VNOVOL) {
2650 EPRINT1(code, "Failed to start transaction on %u\n", cloneVol);
2657 if (!(flags & RV_NOVLDB)) {
2658 /* create the vldb entry for the copied volume */
2659 strncpy(newentry.name, atovolname, VOLSER_OLDMAXVOLNAME);
2660 newentry.nServers = 1;
2661 newentry.serverNumber[0] = atoserver;
2662 newentry.serverPartition[0] = atopart;
2663 newentry.flags = (flags & RV_RDONLY) ? VLF_ROEXISTS : VLF_RWEXISTS;
2664 newentry.serverFlags[0] = (flags & RV_RDONLY) ? VLSF_ROVOL : VLSF_RWVOL;
2665 newentry.volumeId[RWVOL] = newVol;
2666 newentry.volumeId[ROVOL] = (flags & RV_RDONLY) ? newVol : 0;
2667 newentry.volumeId[BACKVOL] = 0;
2668 newentry.cloneId = 0;
2669 /*map into right byte order, before passing to xdr, the stuff has to be in host
2670 * byte order. Xdr converts it into network order */
2671 MapNetworkToHost(&newentry, &storeEntry);
2672 /* create the vldb entry */
2673 vcode = VLDB_CreateEntry(&storeEntry);
2676 "Could not create a VLDB entry for the volume %s %lu\n",
2677 atovolname, (unsigned long)newVol);
2678 /*destroy the created volume */
2679 VPRINT1("Deleting the newly created volume %u\n", newVol);
2680 AFSVolDeleteVolume(toconn, totid);
2684 VPRINT2("Created the VLDB entry for the volume %s %u\n", atovolname,
2688 /* normal cleanup code */
2691 VPRINT1("Cleanup: Ending transaction on source volume %u ...",
2693 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
2694 if (code || rcode) {
2697 "Could not end transaction on the source volume %lu\n",
2698 (unsigned long)afromvol);
2700 error = (code ? code : rcode);
2706 VPRINT1("Cleanup: Ending transaction on clone volume %u ...",
2708 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
2709 if (code || rcode) {
2712 "Could not end transaction on the source's clone volume %lu\n",
2713 (unsigned long)cloneVol);
2715 error = (code ? code : rcode);
2721 VPRINT1("Cleanup: Ending transaction on destination volume %u ...",
2723 code = AFSVolEndTrans(toconn, totid, &rcode);
2727 "Could not end transaction on destination volume %lu\n",
2728 (unsigned long)newVol);
2730 error = (code ? code : rcode);
2735 rx_DestroyConnection(fromconn);
2737 rx_DestroyConnection(toconn);
2738 PrintError("", error);
2741 /* come here only when the sky falls */
2746 "vos copy: operation interrupted, cleanup in progress...\n");
2747 fprintf(STDOUT, "clear transaction contexts\n");
2752 VPRINT("Recovery: Ending transaction on clone volume ...");
2753 AFSVolEndTrans(fromconn, clonetid, &rcode);
2757 VPRINT("Recovery: Ending transaction on destination volume ...");
2758 AFSVolEndTrans(toconn, totid, &rcode);
2761 if (fromtid) { /* put it on-line */
2762 VPRINT("Recovery: Ending transaction on source volume ...");
2763 AFSVolEndTrans(fromconn, fromtid, &rcode);
2767 VPRINT("Recovery: Accessing VLDB.\n");
2768 vcode = VLDB_GetEntryByID(afromvol, -1, &entry);
2770 fprintf(STDOUT, "FATAL: VLDB access error: abort cleanup\n");
2774 MapHostToNetwork(&entry);
2776 /* common cleanup - delete local clone */
2778 code = DoVolDelete(fromconn, cloneVol, afrompart,
2779 "clone", 0, NULL, "Recovery:");
2780 if (code == VNOVOL) {
2781 EPRINT1(code, "Recovery: Failed to start transaction on %u\n", cloneVol);
2785 done: /* routine cleanup */
2787 rx_DestroyConnection(fromconn);
2789 rx_DestroyConnection(toconn);
2792 fprintf(STDOUT, "cleanup complete - user verify desired result\n");
2800 UV_CopyVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
2801 char *atovolname, afs_uint32 atoserver, afs_int32 atopart)
2803 return UV_CopyVolume2(afromvol, afromserver, afrompart,
2804 atovolname, atoserver, atopart, 0, 0);
2809 /* Make a new backup of volume <avolid> on <aserver> and <apart>
2810 * if one already exists, update it
2814 UV_BackupVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid)
2816 struct rx_connection *aconn = (struct rx_connection *)0;
2817 afs_int32 ttid = 0, btid = 0;
2818 afs_uint32 backupID;
2819 afs_int32 code = 0, rcode = 0;
2820 struct nvldbentry entry, storeEntry;
2821 afs_int32 error = 0;
2822 int vldblocked = 0, vldbmod = 0;
2824 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
2826 /* the calls to VLDB will succeed only if avolid is a RW volume,
2827 * since we are following the RW hash chain for searching */
2828 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2831 "Could not fetch the entry for the volume %lu from the VLDB \n",
2832 (unsigned long)avolid);
2836 MapHostToNetwork(&entry);
2838 /* These operations require the VLDB be locked since it means the VLDB
2839 * will change or the vldb is already locked.
2841 if (!(entry.flags & VLF_BACKEXISTS) || /* backup volume doesnt exist */
2842 (entry.flags & VLOP_ALLOPERS) || /* vldb lock already held */
2843 (entry.volumeId[BACKVOL] == INVALID_BID)) { /* no assigned backup volume id */
2845 code = ubik_VL_SetLock(cstruct, 0, avolid, RWVOL, VLOP_BACKUP);
2848 "Could not lock the VLDB entry for the volume %lu\n",
2849 (unsigned long)avolid);
2855 /* Reread the vldb entry */
2856 code = VLDB_GetEntryByID(avolid, RWVOL, &entry);
2859 "Could not fetch the entry for the volume %lu from the VLDB \n",
2860 (unsigned long)avolid);
2864 MapHostToNetwork(&entry);
2867 if (!ISNAMEVALID(entry.name)) {
2868 fprintf(STDERR, "Name of the volume %s exceeds the size limit\n",
2870 error = VOLSERBADNAME;
2874 backupID = entry.volumeId[BACKVOL];
2875 if (backupID == INVALID_BID) {
2876 /* Get a backup volume id from the VLDB and update the vldb
2879 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &backupID);
2882 "Could not allocate ID for the backup volume of %lu from the VLDB\n",
2883 (unsigned long)avolid);
2887 entry.volumeId[BACKVOL] = backupID;
2891 code = DoVolClone(aconn, avolid, apart, backupVolume, backupID, "backup",
2892 entry.name, NULL, ".backup", NULL, NULL);
2898 /* Mark vldb as backup exists */
2899 if (!(entry.flags & VLF_BACKEXISTS)) {
2900 entry.flags |= VLF_BACKEXISTS;
2904 /* Now go back to the backup volume and bring it on line */
2905 code = AFSVolTransCreate_retry(aconn, backupID, apart, ITOffline, &btid);
2908 "Failed to start a transaction on the backup volume %lu\n",
2909 (unsigned long)backupID);
2914 code = AFSVolSetFlags(aconn, btid, 0);
2916 fprintf(STDERR, "Could not mark the backup volume %lu on line \n",
2917 (unsigned long)backupID);
2922 code = AFSVolEndTrans(aconn, btid, &rcode);
2924 if (code || rcode) {
2926 "Failed to end the transaction on the backup volume %lu\n",
2927 (unsigned long)backupID);
2928 error = (code ? code : rcode);
2932 /* Will update the vldb below */
2936 code = AFSVolEndTrans(aconn, ttid, &rcode);
2937 if (code || rcode) {
2938 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
2939 (unsigned long)avolid);
2941 error = (code ? code : rcode);
2946 code = AFSVolEndTrans(aconn, btid, &rcode);
2947 if (code || rcode) {
2949 "Could not end transaction the backup volume %lu\n",
2950 (unsigned long)backupID);
2952 error = (code ? code : rcode);
2956 /* Now update the vldb - if modified */
2959 MapNetworkToHost(&entry, &storeEntry);
2961 VLDB_ReplaceEntry(avolid, RWVOL, &storeEntry,
2962 (LOCKREL_OPCODE | LOCKREL_AFSID |
2963 LOCKREL_TIMESTAMP));
2966 "Could not update the VLDB entry for the volume %lu \n",
2967 (unsigned long)avolid);
2973 ubik_VL_ReleaseLock(cstruct, 0, avolid, RWVOL,
2974 (LOCKREL_OPCODE | LOCKREL_AFSID |
2975 LOCKREL_TIMESTAMP));
2978 "Could not unlock the VLDB entry for the volume %lu \n",
2979 (unsigned long)avolid);
2987 rx_DestroyConnection(aconn);
2989 PrintError("", error);
2993 /* Make a new clone of volume <avolid> on <aserver> and <apart>
2994 * using volume ID <acloneid>, or a new ID allocated from the VLDB.
2995 * The new volume is named by <aname>, or by appending ".clone" to
2996 * the existing name if <aname> is NULL. The following flags are
2999 * RV_RDONLY - target volume is RO
3000 * RV_OFFLINE - leave target volume offline
3004 UV_CloneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 avolid,
3005 afs_uint32 acloneid, char *aname, int flags)
3007 struct rx_connection *aconn = (struct rx_connection *)0;
3008 afs_int32 ttid = 0, btid = 0;
3009 afs_int32 code = 0, rcode = 0;
3010 char vname[VOLSER_MAXVOLNAME + 1];
3011 afs_int32 error = 0;
3012 volEntries volumeInfo;
3015 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
3018 volumeInfo.volEntries_val = (volintInfo *) 0;
3019 volumeInfo.volEntries_len = 0;
3020 code = AFSVolListOneVolume(aconn, apart, avolid, &volumeInfo);
3022 fprintf(stderr, "Could not get info for volume %lu\n",
3023 (unsigned long)avolid);
3027 strncpy(vname, volumeInfo.volEntries_val[0].name,
3028 VOLSER_OLDMAXVOLNAME - 7);
3029 vname[VOLSER_OLDMAXVOLNAME - 7] = 0;
3030 strcat(vname, ".clone");
3032 if (volumeInfo.volEntries_val)
3033 free(volumeInfo.volEntries_val);
3037 /* Get a clone id */
3038 VPRINT1("Allocating new volume id for clone of volume %u ...",
3040 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &acloneid);
3042 "Could not get an ID for the clone of volume %u from the VLDB\n",
3047 if (flags & RV_RWONLY)
3048 type = readwriteVolume;
3049 else if (flags & RV_RDONLY)
3050 type = readonlyVolume;
3052 type = backupVolume;
3054 code = DoVolClone(aconn, avolid, apart, type, acloneid, "clone",
3055 NULL, aname, NULL, NULL, NULL);
3061 /* Now go back to the backup volume and bring it on line */
3062 if (!(flags & RV_OFFLINE)) {
3063 code = AFSVolTransCreate_retry(aconn, acloneid, apart, ITOffline, &btid);
3066 "Failed to start a transaction on the clone volume %lu\n",
3067 (unsigned long)acloneid);
3072 code = AFSVolSetFlags(aconn, btid, 0);
3074 fprintf(STDERR, "Could not mark the clone volume %lu on line \n",
3075 (unsigned long)acloneid);
3080 code = AFSVolEndTrans(aconn, btid, &rcode);
3082 if (code || rcode) {
3084 "Failed to end the transaction on the clone volume %lu\n",
3085 (unsigned long)acloneid);
3086 error = (code ? code : rcode);
3093 code = AFSVolEndTrans(aconn, ttid, &rcode);
3094 if (code || rcode) {
3095 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
3096 (unsigned long)avolid);
3098 error = (code ? code : rcode);
3103 code = AFSVolEndTrans(aconn, btid, &rcode);
3104 if (code || rcode) {
3106 "Could not end transaction on the clone volume %lu\n",
3107 (unsigned long)acloneid);
3109 error = (code ? code : rcode);
3114 rx_DestroyConnection(aconn);
3116 PrintError("", error);
3120 #define ONERROR(ec, ep, es) do { \
3122 fprintf(STDERR, (es), (ep)); \
3127 #define ONERROR0(ec, es) do { \
3129 fprintf(STDERR, (es)); \
3134 #define ERROREXIT(ec) do { \
3139 /* Get a "transaction" on this replica. Create the volume
3140 * if necessary. Return the time from which a dump should
3141 * be made (0 if it's a new volume)
3144 GetTrans(struct nvldbentry *vldbEntryPtr, afs_int32 index,
3145 struct rx_connection **connPtr, afs_int32 * transPtr,
3146 afs_uint32 * crtimePtr, afs_uint32 * uptimePtr,
3147 afs_int32 *origflags, afs_uint32 tmpVolId)
3150 struct volser_status tstatus;
3155 *connPtr = (struct rx_connection *)0;
3160 /* get connection to the replication site */
3161 *connPtr = UV_Bind(vldbEntryPtr->serverNumber[index], AFSCONF_VOLUMEPORT);
3163 goto fail; /* server is down */
3165 volid = vldbEntryPtr->volumeId[ROVOL];
3169 AFSVolTransCreate_retry(*connPtr, volid,
3170 vldbEntryPtr->serverPartition[index], ITOffline,
3173 if (!code && (origflags[index] & VLSF_DONTUSE)) {
3174 /* If RO_DONTUSE is set, this is supposed to be an entirely new
3175 * site. Don't trust any data on it, since it is possible we
3176 * have encountered some temporary volume from some other
3177 * incomplete volume operation. It is difficult to detect if
3178 * that has happened vs if this is a legit volume, so just
3179 * delete it to be safe. */
3181 VPRINT1("Deleting extant RO_DONTUSE site on %s...",
3182 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3183 serverNumber[index], hoststr) :
3184 hostutil_GetNameByINet(vldbEntryPtr->
3185 serverNumber[index]));
3187 code = AFSVolDeleteVolume(*connPtr, *transPtr);
3189 PrintError("Failed to delete RO_DONTUSE site: ", code);
3193 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3199 PrintError("Failed to end transaction on RO_DONTUSE site: ",
3206 /* emulate what TransCreate would have returned, so we try to
3207 * create the volume below */
3212 /* If the volume does not exist, create it */
3213 if (!volid || code) {
3214 char volname[VL_MAXNAMELEN];
3217 if (volid && (code != VNOVOL)) {
3218 PrintError("Failed to start a transaction on the RO volume.\n",
3223 strlcpy(volname, vldbEntryPtr->name, sizeof(volname));
3225 if (strlcat(volname,
3226 tmpVolId?".roclone":".readonly",
3227 sizeof(volname)) >= sizeof(volname)) {
3229 PrintError("Volume name is too long\n", code);
3235 "Creating new volume %lu on replication site %s: ",
3236 tmpVolId?(unsigned long)tmpVolId:(unsigned long)volid,
3237 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3238 serverNumber[index], hoststr) :
3239 hostutil_GetNameByINet(vldbEntryPtr->
3240 serverNumber[index]));
3245 AFSVolCreateVolume(*connPtr, vldbEntryPtr->serverPartition[index],
3247 vldbEntryPtr->volumeId[RWVOL],
3248 tmpVolId?&tmpVolId:&volid,
3251 PrintError("Failed to create the ro volume: ", code);
3254 vldbEntryPtr->volumeId[ROVOL] = volid;
3258 /* The following is a bit redundant, since create sets these flags by default */
3260 AFSVolSetFlags(*connPtr, *transPtr,
3261 VTDeleteOnSalvage | VTOutOfService);
3263 PrintError("Failed to set flags on the ro volume: ", code);
3268 /* Otherwise, the transaction did succeed, so get the creation date of the
3269 * latest RO volume on the replication site
3272 VPRINT2("Updating existing ro volume %u on %s ...\n", volid,
3273 noresolve ? afs_inet_ntoa_r(vldbEntryPtr->
3274 serverNumber[index], hoststr) :
3275 hostutil_GetNameByINet(vldbEntryPtr->serverNumber[index]));
3277 code = AFSVolGetStatus(*connPtr, *transPtr, &tstatus);
3279 PrintError("Failed to get status of volume on destination: ",
3284 code = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3289 code = DoVolClone(*connPtr, volid,
3290 vldbEntryPtr->serverPartition[index],
3291 readonlyVolume, tmpVolId, "temporary",
3292 vldbEntryPtr->name, NULL, ".roclone", NULL,
3297 *crtimePtr = CLOCKADJ(tstatus.creationDate);
3298 *uptimePtr = CLOCKADJ(tstatus.updateDate);
3305 tcode = AFSVolEndTrans(*connPtr, *transPtr, &rcode);
3309 if (tcode && tcode != ENOENT)
3310 PrintError("Could not end transaction on a ro volume: ", tcode);
3317 SimulateForwardMultiple(struct rx_connection *fromconn, afs_int32 fromtid,
3318 afs_int32 fromdate, manyDests * tr, afs_int32 flags,
3319 void *cookie, manyResults * results)
3323 for (i = 0; i < tr->manyDests_len; i++) {
3324 results->manyResults_val[i] =
3325 AFSVolForward(fromconn, fromtid, fromdate,
3326 &(tr->manyDests_val[i].server),
3327 tr->manyDests_val[i].trans, cookie);
3333 * Check if a trans has timed out, and recreate it if necessary.
3335 * @param[in] aconn RX connection to the relevant server
3336 * @param[inout] atid Transaction ID to check; if we recreated the trans,
3337 * contains the new trans ID on success
3338 * @param[in] apart Partition for the transaction
3339 * @param[in] astat The status of the original transaction
3341 * @return operation status
3342 * @retval 0 existing transaction is still valid, or we managed to recreate
3343 * the trans successfully
3344 * @retval nonzero Fatal error; bail out
3347 CheckTrans(struct rx_connection *aconn, afs_int32 *atid, afs_int32 apart,
3348 struct volser_status *astat)
3350 struct volser_status new_status;
3353 memset(&new_status, 0, sizeof(new_status));
3354 code = AFSVolGetStatus(aconn, *atid, &new_status);
3356 if (code == ENOENT) {
3358 VPRINT1("Old transaction on cloned volume %lu timed out, "
3359 "restarting transaction\n", (long unsigned) astat->volID);
3360 code = AFSVolTransCreate_retry(aconn, astat->volID, apart,
3363 PrintError("Failed to recreate cloned RO volume transaction\n",
3368 memset(&new_status, 0, sizeof(new_status));
3369 code = AFSVolGetStatus(aconn, *atid, &new_status);
3371 PrintError("Failed to get status on recreated transaction\n",
3376 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3377 PrintError("Recreated transaction on cloned RO volume, but "
3378 "the volume has changed!\n", 0);
3382 PrintError("Unable to get status of current cloned RO transaction\n",
3387 if (memcmp(&new_status, astat, sizeof(new_status)) != 0) {
3389 PrintError("Internal error: current GetStatus does not match "
3390 "original GetStatus?\n", 0);
3399 PutTrans(afs_int32 *vldbindex, struct replica *replicas,
3400 struct rx_connection **toconns, struct release *times,
3403 afs_int32 s, code = 0, rcode = 0;
3404 /* End the transactions and destroy the connections */
3405 for (s = 0; s < volcount; s++) {
3406 if (replicas[s].trans) {
3407 code = AFSVolEndTrans(toconns[s], replicas[s].trans, &rcode);
3409 replicas[s].trans = 0;
3413 if ((s == 0) || (code != ENOENT)) {
3414 PrintError("Could not end transaction on a ro volume: ",
3418 ("Transaction timed out on a ro volume. Will retry.\n",
3420 if (times[s].vldbEntryIndex < *vldbindex)
3421 *vldbindex = times[s].vldbEntryIndex;
3426 rx_DestroyConnection(toconns[s]);
3432 * Release a volume to read-only sites
3434 * Release volume <afromvol> on <afromserver> <afrompart> to all its RO
3435 * sites (complete release), unless the previous release was incomplete
3436 * or new sites were added without changes to the read-write volume, in
3437 * which case we bring the remaining volumes up to date with the volumes
3438 * that were previously released successfully.
3440 * Will create a clone from the RW, then dump the clone out to
3441 * the remaining replicas. If there is more than 1 RO sites,
3442 * ensure that the VLDB says at least one RO is available all
3443 * the time: Influences when we write back the VLDB entry.
3445 * @param[in] afromvol volume to be released
3446 * @param[in] afromserver server containing afromvol
3447 * @param[in] afrompart partition containing afromvol
3448 * @param[in] flags bitmap of options
3449 * REL_COMPLETE - force a complete release
3450 * REL_FULLDUMPS - force full dumps
3453 UV_ReleaseVolume(afs_uint32 afromvol, afs_uint32 afromserver,
3454 afs_int32 afrompart, int flags)
3458 afs_int32 vcode, rcode, tcode;
3459 afs_uint32 cloneVolId = 0, roVolId;
3460 struct replica *replicas = 0;
3461 struct nvldbentry entry, storeEntry;
3462 int i, volcount = 0, m, vldbindex;
3464 struct restoreCookie cookie;
3465 struct rx_connection **toconns = 0;
3466 struct release *times = 0;
3468 struct rx_connection *fromconn = (struct rx_connection *)0;
3469 afs_int32 error = 0;
3471 afs_int32 clonetid = 0, onlinetid;
3472 afs_int32 fromtid = 0;
3473 afs_uint32 fromdate = 0;
3474 afs_uint32 thisdate;
3478 manyResults results;
3479 int rwindex, roindex, roclone, roexists;
3480 afs_uint32 rwcrdate = 0, rwupdate = 0;
3481 afs_uint32 clcrdate;
3485 } remembertime[NMAXNSERVERS];
3486 int releasecount = 0;
3487 struct volser_status volstatus;
3489 afs_int32 origflags[NMAXNSERVERS];
3490 struct volser_status orig_status;
3491 int notreleased = 0;
3492 int tried_justnewsites = 0;
3493 int justnewsites = 0; /* are we just trying to release to new RO sites? */
3494 int sites = 0; /* number of ro sites */
3495 int new_sites = 0; /* number of ro sites markes as new */
3498 CR_PARTIAL = 0x0000, /**< just new sites added or recover from a previous failed release */
3499 CR_FORCED = 0x0001, /**< complete: forced by caller */
3500 CR_LAST_OK = 0x0002, /**< complete: no sites have been marked as new release */
3501 CR_ALL_NEW = 0x0004, /**< complete: all sites have been marked as new release */
3502 CR_NEW_RW = 0x0008, /**< complete: read-write has changed */
3503 CR_RO_MISSING = 0x0010, /**< complete: ro clone is missing */
3504 } complete_release_t;
3506 complete_release_t complete_release = CR_PARTIAL;
3508 memset(remembertime, 0, sizeof(remembertime));
3509 memset(&results, 0, sizeof(results));
3510 memset(origflags, 0, sizeof(origflags));
3512 vcode = ubik_VL_SetLock(cstruct, 0, afromvol, RWVOL, VLOP_RELEASE);
3513 if (vcode != VL_RERELEASE)
3514 ONERROR(vcode, afromvol,
3515 "Could not lock the VLDB entry for the volume %u.\n");
3518 /* Get the vldb entry in readable format */
3519 vcode = VLDB_GetEntryByID(afromvol, RWVOL, &entry);
3520 ONERROR(vcode, afromvol,
3521 "Could not fetch the entry for the volume %u from the VLDB.\n");
3522 MapHostToNetwork(&entry);
3525 EnumerateEntry(&entry);
3527 if (!ISNAMEVALID(entry.name))
3528 ONERROR(VOLSERBADOP, entry.name,
3529 "Volume name %s is too long, rename before releasing.\n");
3530 if (entry.volumeId[RWVOL] != afromvol)
3531 ONERROR(VOLSERBADOP, afromvol,
3532 "The volume %u being released is not a read-write volume.\n");
3533 if (entry.nServers <= 1)
3534 ONERROR(VOLSERBADOP, afromvol,
3535 "Volume %u has no replicas - release operation is meaningless!\n");
3536 if (strlen(entry.name) > (VOLSER_OLDMAXVOLNAME - 10))
3537 ONERROR(VOLSERBADOP, entry.name,
3538 "RO volume name %s exceeds (VOLSER_OLDMAXVOLNAME - 10) character limit\n");
3540 /* roclone is true if one of the RO volumes is on the same
3541 * partition as the RW volume. In this case, we make the RO volume
3542 * on the same partition a clone instead of a complete copy.
3545 roindex = Lp_ROMatch(afromserver, afrompart, &entry) - 1;
3546 roclone = ((roindex == -1) ? 0 : 1);
3547 rwindex = Lp_GetRwIndex(&entry);
3549 ONERROR0(VOLSERNOVOL, "There is no RW volume \n");
3551 /* Make sure we have a RO volume id to work with */
3552 if (entry.volumeId[ROVOL] == INVALID_BID) {
3553 /* need to get a new RO volume id */
3554 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &roVolId);
3555 ONERROR(vcode, entry.name, "Cant allocate ID for RO volume of %s\n");
3557 entry.volumeId[ROVOL] = roVolId;
3558 MapNetworkToHost(&entry, &storeEntry);
3559 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3560 ONERROR(vcode, entry.name, "Could not update vldb entry for %s.\n");
3564 * Determine if this is to be a complete release or a recovery of a
3565 * previous unfinished release. The previous release is considered to be
3566 * unfinished when the clone was successfully distributed to at least one
3567 * (but not all) of the read-only sites, as indicated by the NEW_REPSITE
3570 * The caller can override the vldb flags check using the -force
3571 * or -force-reclone flag, to force this to be a complete release.
3573 for (i = 0; i < entry.nServers; i++) {
3574 if (entry.serverFlags[i] & VLSF_ROVOL) {
3576 if (entry.serverFlags[i] & VLSF_NEWREPSITE)
3578 if (entry.serverFlags[i] & VLSF_DONTUSE)
3581 origflags[i] = entry.serverFlags[i];
3584 if (flags & REL_COMPLETE) {
3585 complete_release |= CR_FORCED;
3588 if (new_sites == 0) {
3589 complete_release |= CR_LAST_OK;
3590 } else if (new_sites == sites) {
3591 complete_release |= CR_ALL_NEW;
3594 if ((complete_release & (CR_LAST_OK | CR_ALL_NEW))
3595 && !(complete_release & CR_FORCED)) {
3596 if (notreleased && notreleased != sites) {
3597 /* we have some new unreleased sites. try to just release to those,
3598 * if the RW has not changed. The caller can override with -force
3599 * or with -force-reclone. */
3604 /* Determine which volume id to use and see if it exists */
3605 cloneVolId = (complete_release || entry.cloneId == 0)
3606 ? entry.volumeId[ROVOL] : entry.cloneId;
3608 code = VolumeExists(afromserver, afrompart, cloneVolId);
3609 roexists = ((code == ENODEV) ? 0 : 1);
3611 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
3613 ONERROR(-1, afromserver,
3614 "Cannot establish connection with server 0x%x\n");
3616 if (!complete_release) {
3618 complete_release |= CR_RO_MISSING; /* Do a complete release if RO clone does not exist */
3620 /* Begin transaction on RW and mark it busy while we query it */
3621 code = AFSVolTransCreate_retry(
3622 fromconn, afromvol, afrompart, ITBusy, &fromtid
3624 ONERROR(code, afromvol,
3625 "Failed to start transaction on RW volume %u\n");
3627 /* Query the creation date for the RW */
3628 code = AFSVolGetStatus(fromconn, fromtid, &volstatus);
3629 ONERROR(code, afromvol,
3630 "Failed to get the status of RW volume %u\n");
3631 rwcrdate = volstatus.creationDate;
3633 /* End transaction on RW */
3634 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
3636 ONERROR((code ? code : rcode), afromvol,
3637 "Failed to end transaction on RW volume %u\n");
3639 /* Begin transaction on clone and mark it busy while we query it */
3640 code = AFSVolTransCreate_retry(
3641 fromconn, cloneVolId, afrompart, ITBusy, &clonetid
3643 ONERROR(code, cloneVolId,
3644 "Failed to start transaction on RW clone %u\n");
3646 /* Query the creation date for the clone */
3647 code = AFSVolGetStatus(fromconn, clonetid, &volstatus);
3648 ONERROR(code, cloneVolId,
3649 "Failed to get the status of RW clone %u\n");
3650 clcrdate = volstatus.creationDate;
3652 /* End transaction on clone */
3653 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
3655 ONERROR((code ? code : rcode), cloneVolId,
3656 "Failed to end transaction on RW clone %u\n");
3658 if (rwcrdate > clcrdate)
3659 complete_release |= CR_NEW_RW; /* Do a complete release if RO clone older than RW */
3663 if (!complete_release || (complete_release & CR_NEW_RW)) {
3664 /* in case the RW has changed, and just to be safe */
3669 if (!complete_release) {
3671 "This is a recovery of previously failed release\n");
3673 fprintf(STDOUT, "This is a complete release of volume %u", afromvol);
3674 if (complete_release == CR_LAST_OK) {
3676 tried_justnewsites = 1;
3677 fprintf(STDOUT, "\nThere are new RO sites; we will try to "
3678 "only release to new sites");
3682 if (complete_release & CR_FORCED) {
3683 fprintf(STDOUT, "%sforced", sep);
3686 if (complete_release & CR_LAST_OK) {
3687 fprintf(STDOUT, "%slast ok", sep);
3690 if (complete_release & CR_ALL_NEW) {
3691 fprintf(STDOUT, "%sall sites are new", sep);
3694 if (complete_release & CR_NEW_RW) {
3695 fprintf(STDOUT, "%srw %u changed", sep, afromvol);
3698 if (complete_release & CR_RO_MISSING) {
3699 fprintf(STDOUT, "%sro clone missing", sep);
3701 fprintf(STDOUT, ")");
3703 fprintf(STDOUT, "\n");
3707 if (complete_release) {
3708 afs_int32 oldest = 0;
3709 /* If the RO clone exists, then if the clone is a temporary
3710 * clone, delete it. Or if the RO clone is marked RO_DONTUSE
3711 * (it was recently added), then also delete it. We do not
3712 * want to "reclone" a temporary RO clone.
3715 /* clean up any previous tmp clone before starting if staying up */
3717 && (!roclone || (entry.serverFlags[roindex] & VLSF_DONTUSE))) {
3718 code = DoVolDelete(fromconn,
3720 afrompart, "the", 0, NULL, NULL);
3721 if (code && (code != VNOVOL))
3727 VPRINT("Querying old RO sites for update times...");
3728 for (vldbindex = 0; vldbindex < entry.nServers; vldbindex++) {
3729 volEntries volumeInfo;
3730 struct rx_connection *conn;
3733 if (!(entry.serverFlags[vldbindex] & VLSF_ROVOL)) {
3736 if ((entry.serverFlags[vldbindex] & VLSF_DONTUSE)) {
3739 conn = UV_Bind(entry.serverNumber[vldbindex], AFSCONF_VOLUMEPORT);
3741 fprintf(STDERR, "Cannot establish connection to server %s\n",
3742 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3746 volumeInfo.volEntries_val = NULL;
3747 volumeInfo.volEntries_len = 0;
3748 code = AFSVolListOneVolume(conn, entry.serverPartition[vldbindex],
3749 entry.volumeId[ROVOL],
3752 fprintf(STDERR, "Could not fetch information about RO vol %lu from server %s\n",
3753 (unsigned long)entry.volumeId[ROVOL],
3754 hostutil_GetNameByINet(entry.serverNumber[vldbindex]));
3755 PrintError("", code);
3757 rx_DestroyConnection(conn);
3761 crdate = CLOCKADJ(volumeInfo.volEntries_val[0].creationDate);
3763 if (oldest == 0 || crdate < oldest) {
3767 rx_DestroyConnection(conn);
3768 free(volumeInfo.volEntries_val);
3769 volumeInfo.volEntries_val = NULL;
3770 volumeInfo.volEntries_len = 0;
3775 volEntries volumeInfo;
3776 volumeInfo.volEntries_val = NULL;
3777 volumeInfo.volEntries_len = 0;
3778 code = AFSVolListOneVolume(fromconn, afrompart, afromvol,
3781 fprintf(STDERR, "Could not fetch information about RW vol %lu from server %s\n",
3782 (unsigned long)afromvol,
3783 hostutil_GetNameByINet(afromserver));
3784 PrintError("", code);
3787 rwupdate = volumeInfo.volEntries_val[0].updateDate;
3789 free(volumeInfo.volEntries_val);
3790 volumeInfo.volEntries_val = NULL;
3791 volumeInfo.volEntries_len = 0;
3794 if (justnewsites && oldest <= rwupdate) {
3795 /* RW has changed */
3799 /* Mark all the ROs in the VLDB entry as RO_DONTUSE. We don't
3800 * write this entry out to the vlserver until after the first
3801 * RO volume is released (temp RO clones don't count).
3803 * If 'justnewsites' is set, we're only updating sites that have
3804 * RO_DONTUSE set, so set NEW_REPSITE for all of the others.
3806 for (i = 0; i < entry.nServers; i++) {
3808 if ((entry.serverFlags[i] & VLSF_DONTUSE)) {
3809 entry.serverFlags[i] &= ~VLSF_NEWREPSITE;
3811 entry.serverFlags[i] |= VLSF_NEWREPSITE;
3814 entry.serverFlags[i] &= ~VLSF_NEWREPSITE;
3815 entry.serverFlags[i] |= VLSF_DONTUSE;
3818 entry.serverFlags[rwindex] |= VLSF_NEWREPSITE;
3819 entry.serverFlags[rwindex] &= ~VLSF_DONTUSE;
3822 if (justnewsites && roexists) {
3823 /* if 'justnewsites' and 'roexists' are set, we don't need to do
3824 * anything with the RO clone, so skip the reclone */
3827 } else if (complete_release) {
3830 strcpy(vname, entry.name);
3831 strcat(vname, ".readonly");
3833 strcpy(vname, "readonly-clone-temp");
3836 code = DoVolClone(fromconn, afromvol, afrompart, readonlyVolume,
3837 cloneVolId, roclone?"permanent RO":
3838 "temporary RO", NULL, vname, NULL, &volstatus, NULL);
3844 if (justnewsites && rwupdate != volstatus.updateDate) {
3846 /* reset the serverFlags as if 'justnewsites' had never been set */
3847 for (i = 0; i < entry.nServers; i++) {
3848 entry.serverFlags[i] &= ~VLSF_NEWREPSITE;
3849 entry.serverFlags[i] |= VLSF_DONTUSE;
3851 entry.serverFlags[rwindex] |= VLSF_NEWREPSITE;
3852 entry.serverFlags[rwindex] &= ~VLSF_DONTUSE;
3855 rwcrdate = volstatus.creationDate;
3857 /* Remember clone volume ID in case we fail or are interrupted */
3858 entry.cloneId = cloneVolId;
3861 /* Bring the RO clone online - though not if it's a temporary clone */
3862 VPRINT1("Starting transaction on RO clone volume %u...",
3865 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITOffline,
3867 ONERROR(code, cloneVolId,
3868 "Failed to start transaction on volume %u\n");
3871 VPRINT1("Setting volume flags for volume %u...", cloneVolId);
3872 tcode = AFSVolSetFlags(fromconn, onlinetid, 0);
3875 VPRINT1("Ending transaction on volume %u...", cloneVolId);
3876 code = AFSVolEndTrans(fromconn, onlinetid, &rcode);
3877 ONERROR((code ? code : rcode), cloneVolId,
3878 "Failed to end transaction on RO clone %u\n");
3881 ONERROR(tcode, cloneVolId, "Could not bring volume %u on line\n");
3883 /* Sleep so that a client searching for an online volume won't
3884 * find the clone offline and then the next RO offline while the
3885 * release brings the clone online and the next RO offline (race).
3886 * There is a fix in the 3.4 client that does not need this sleep
3887 * anymore, but we don't know what clients we have.
3889 if (entry.nServers > 2 && !justnewsites)
3892 /* Mark the RO clone in the VLDB as a good site (already released) */
3893 entry.serverFlags[roindex] |= VLSF_NEWREPSITE;
3894 entry.serverFlags[roindex] &= ~VLSF_DONTUSE;
3895 entry.flags |= VLF_ROEXISTS;
3899 /* Write out the VLDB entry only if the clone is not a temporary
3900 * clone. If we did this to a temporary clone then we would end
3901 * up marking all the ROs as "old release" making the ROs
3902 * temporarily unavailable.
3904 MapNetworkToHost(&entry, &storeEntry);
3905 VPRINT1("Replacing VLDB entry for %s...", entry.name);
3906 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
3907 ONERROR(vcode, entry.name,
3908 "Could not update vldb entry for %s.\n");
3914 VPRINT("RW vol has not changed; only releasing to new RO sites\n");
3915 /* act like this is a completion of a previous release */
3916 complete_release = CR_PARTIAL;
3917 } else if (tried_justnewsites) {
3918 VPRINT("RW vol has changed; releasing to all sites\n");
3921 /* Now we will release from the clone to the remaining RO replicas.
3922 * The first 2 ROs (counting the non-temporary RO clone) are released
3923 * individually: releasecount. This is to reduce the race condition
3924 * of clients trying to find an on-line RO volume. The remaining ROs
3925 * are released in parallel but no more than half the number of ROs
3926 * (rounded up) at a time: nservers.
3929 strcpy(vname, entry.name);
3930 strcat(vname, ".readonly");
3931 memset(&cookie, 0, sizeof(cookie));
3932 strncpy(cookie.name, vname, VOLSER_OLDMAXVOLNAME);
3933 cookie.type = ROVOL;
3934 cookie.parent = entry.volumeId[RWVOL];
3937 /* how many to do at once, excluding clone */
3939 nservers = entry.nServers; /* can do all, none offline */
3941 nservers = entry.nServers / 2;
3942 replicas = calloc(nservers + 1, sizeof(struct replica));
3943 times = calloc(nservers + 1, sizeof(struct release));
3944 toconns = calloc(nservers + 1, sizeof(struct rx_connection *));
3945 results.manyResults_val = calloc(nservers + 1, sizeof(afs_int32));
3946 if (!replicas || !times || !results.manyResults_val || !toconns)
3948 "Failed to create transaction on the release clone\n");
3950 /* Create a transaction on the cloned volume */
3951 VPRINT1("Starting transaction on cloned volume %u...", cloneVolId);
3953 AFSVolTransCreate_retry(fromconn, cloneVolId, afrompart, ITBusy, &fromtid);
3955 memset(&orig_status, 0, sizeof(orig_status));
3956 code = AFSVolGetStatus(fromconn, fromtid, &orig_status);
3958 if (!complete_release && code)
3959 ONERROR(VOLSERNOVOL, afromvol,
3960 "Old clone is inaccessible. Try vos release -f %u.\n");
3961 ONERROR0(code, "Failed to create transaction on the release clone\n");
3964 /* For each index in the VLDB */
3965 for (vldbindex = 0; vldbindex < entry.nServers;) {
3966 /* Get a transaction on the replicas. Pick replicas which have an old release. */
3968 ((volcount < nservers) && (vldbindex < entry.nServers));
3970 if (!justnewsites) {
3971 /* The first two RO volumes will be released individually.
3972 * The rest are then released in parallel. This is a hack
3973 * for clients not recognizing right away when a RO volume
3974 * comes back on-line.
3976 if ((volcount == 1) && (releasecount < 2))
3980 if (vldbindex == roindex)
3981 continue; /* the clone */
3982 if ((entry.serverFlags[vldbindex] & VLSF_NEWREPSITE)
3983 && !(entry.serverFlags[vldbindex] & VLSF_DONTUSE))
3985 if (!(entry.serverFlags[vldbindex] & VLSF_ROVOL))
3986 continue; /* not a RO vol */
3989 /* Get a Transaction on this replica. Get a new connection if
3990 * necessary. Create the volume if necessary. Return the
3991 * time from which the dump should be made (0 if it's a new
3992 * volume). Each volume might have a different time.
3994 replicas[volcount].server.destHost =
3995 ntohl(entry.serverNumber[vldbindex]);
3996 replicas[volcount].server.destPort = AFSCONF_VOLUMEPORT;
3997 replicas[volcount].server.destSSID = 1;
3998 times[volcount].vldbEntryIndex = vldbindex;
4001 GetTrans(&entry, vldbindex, &(toconns[volcount]),
4002 &(replicas[volcount].trans),
4003 &(times[volcount].crtime),
4004 &(times[volcount].uptime),
4009 /* Thisdate is the date from which we want to pick up all changes */
4010 if (flags & REL_FULLDUMPS) {
4011 /* Do a full dump when forced by the caller. */
4012 VPRINT("This will be a full dump: forced\n");
4014 } else if (complete_release == CR_PARTIAL) {
4016 VPRINT("This will be a full dump: read-only volume needs be created for new site\n");
4019 * We cannot be sure the creation date is good since the previous
4020 * release failed, so do a full dump.
4022 VPRINT("This will be a full dump: previous release failed\n");
4025 } else if (times[volcount].crtime == 0) {
4026 /* A full dump is needed for a new read-only volume. */
4028 ("This will be a full dump: read-only volume needs to be created\n");
4030 } else if ((rwcrdate > times[volcount].crtime)) {
4031 /* If the RW volume was replaced (its creation date is newer than
4032 * the last release), then we can't be sure what has changed (so
4033 * we do a full dump).
4036 ("This will be a full dump: read-write volume was replaced\n");
4038 } else if (remembertime[vldbindex].validtime) {
4039 /* Trans was prev ended. Use the time from the prev trans
4040 * because, prev trans may have created the volume. In which
4041 * case time[volcount].time would be now instead of 0.
4044 (remembertime[vldbindex].uptime < times[volcount].uptime)
4045 ? remembertime[vldbindex].uptime
4046 : times[volcount].uptime;
4048 thisdate = times[volcount].uptime;
4050 remembertime[vldbindex].validtime = 1;
4051 remembertime[vldbindex].uptime = thisdate;
4053 if (volcount == 0) {
4054 fromdate = thisdate;
4056 /* Include this volume if it is within 15 minutes of the earliest */
4058 thisdate) ? (fromdate - thisdate) : (thisdate -
4060 AFSVolEndTrans(toconns[volcount],
4061 replicas[volcount].trans, &rcode);
4062 replicas[volcount].trans = 0;
4065 if (thisdate < fromdate)
4066 fromdate = thisdate;
4073 code = CheckTrans(fromconn, &fromtid, afrompart, &orig_status);
4080 fprintf(STDOUT, "Starting ForwardMulti from %lu to %u on %s",
4081 (unsigned long)cloneVolId,
4082 entry.volumeId[ROVOL],
4083 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[0].
4084 vldbEntryIndex], hoststr) :
4085 hostutil_GetNameByINet(entry.
4086 serverNumber[times[0].
4089 for (s = 1; s < volcount; s++) {
4090 fprintf(STDOUT, " and %s",
4091 noresolve ? afs_inet_ntoa_r(entry.serverNumber[times[s].
4092 vldbEntryIndex], hoststr) :
4093 hostutil_GetNameByINet(entry.
4094 serverNumber[times[s].
4099 fprintf(STDOUT, " (entire volume)");
4102 fprintf(STDOUT, " (as of %.24s)", ctime(&tmv));
4104 fprintf(STDOUT, ".\n");
4108 /* Release the ones we have collected */
4109 tr.manyDests_val = &(replicas[0]);
4110 tr.manyDests_len = results.manyResults_len = volcount;
4112 AFSVolForwardMultiple(fromconn, fromtid, fromdate, &tr,
4113 0 /*spare */ , &cookie, &results);
4114 if (code == RXGEN_OPCODE) { /* RPC Interface Mismatch */
4116 SimulateForwardMultiple(fromconn, fromtid, fromdate, &tr,
4117 0 /*spare */ , &cookie, &results);
4122 PrintError("Release failed: ", code);
4124 for (m = 0; m < volcount; m++) {
4125 if (results.manyResults_val[m]) {
4126 if ((m == 0) || (results.manyResults_val[m] != ENOENT)) {
4127 /* we retry timed out transaction. When it is
4128 * not the first volume and the transaction wasn't found
4129 * (assume it timed out and was garbage collected by volser).
4132 ("Failed to dump volume from clone to a ro site: ",
4133 results.manyResults_val[m]);
4139 AFSVolSetIdsTypes(toconns[m], replicas[m].trans, vname,
4140 ROVOL, entry.volumeId[RWVOL], 0, 0);
4142 if ((m == 0) || (code != ENOENT)) {
4143 PrintError("Failed to set correct names and ids: ",
4149 /* have to clear dest. flags to ensure new vol goes online:
4150 * because the restore (forwarded) operation copied
4151 * the V_inService(=0) flag over to the destination.
4153 code = AFSVolSetFlags(toconns[m], replicas[m].trans, 0);
4155 if ((m == 0) || (code != ENOENT)) {
4156 PrintError("Failed to set flags on ro volume: ",
4162 entry.serverFlags[times[m].vldbEntryIndex] |= VLSF_NEWREPSITE;
4163 entry.serverFlags[times[m].vldbEntryIndex] &= ~VLSF_DONTUSE;
4164 entry.flags |= VLF_ROEXISTS;
4169 PutTrans(&vldbindex, replicas, toconns, times, volcount);
4170 MapNetworkToHost(&entry, &storeEntry);
4171 vcode = VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry, 0);
4172 ONERROR(vcode, afromvol,
4173 " Could not update VLDB entry for volume %u\n");
4174 } /* for each index in the vldb */
4176 /* End the transaction on the cloned volume */
4177 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4182 PrintError("Failed to end transaction on rw volume: ", code);
4184 /* Figure out if any volume were not released and say so */
4185 for (failure = 0, i = 0; i < entry.nServers; i++) {
4186 if (!(entry.serverFlags[i] & VLSF_NEWREPSITE))
4192 "The volume %lu could not be released to the following %d sites:\n",
4193 (unsigned long)afromvol, failure);
4194 for (i = 0; i < entry.nServers; i++) {
4195 if (!(entry.serverFlags[i] & VLSF_NEWREPSITE)) {
4196 MapPartIdIntoName(entry.serverPartition[i], pname);
4197 fprintf(STDERR, "\t%35s %s\n",
4198 noresolve ? afs_inet_ntoa_r(entry.serverNumber[i], hoststr) :
4199 hostutil_GetNameByINet(entry.serverNumber[i]), pname);
4202 MapNetworkToHost(&entry, &storeEntry);
4204 VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry,
4206 ONERROR(vcode, afromvol,
4207 " Could not update VLDB entry for volume %u\n");
4209 ERROREXIT(VOLSERBADRELEASE);
4213 /* All the ROs were release successfully. Remove the temporary clone */
4216 fprintf(STDOUT, "Deleting the releaseClone %lu ...",
4217 (unsigned long)cloneVolId);
4220 code = DoVolDelete(fromconn, cloneVolId, afrompart, NULL, 0, NULL,
4222 ONERROR(code, cloneVolId, "Failed to delete volume %u.\n");
4226 for (i = 0; i < entry.nServers; i++)
4227 entry.serverFlags[i] &= ~VLSF_NEWREPSITE;
4229 /* Update the VLDB */
4230 VPRINT("updating VLDB ...");
4232 MapNetworkToHost(&entry, &storeEntry);
4234 VLDB_ReplaceEntry(afromvol, RWVOL, &storeEntry,
4235 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4236 ONERROR(vcode, afromvol, " Could not update VLDB entry for volume %u\n");
4241 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
4245 "Failed to end cloning transaction on the RW volume %lu\n",
4246 (unsigned long)afromvol);
4252 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4256 "Failed to end transaction on the release clone %lu\n",
4257 (unsigned long)cloneVolId);
4262 for (i = 0; i < nservers; i++) {
4263 if (replicas && replicas[i].trans) {
4264 code = AFSVolEndTrans(toconns[i], replicas[i].trans, &rcode);
4265 replicas[i].trans = 0;
4268 "Failed to end transaction on ro volume %u at server %s\n",
4269 entry.volumeId[ROVOL],
4270 noresolve ? afs_inet_ntoa_r(htonl(replicas[i].server.
4271 destHost), hoststr) :
4272 hostutil_GetNameByINet(htonl
4273 (replicas[i].server.destHost)));
4278 if (toconns && toconns[i]) {
4279 rx_DestroyConnection(toconns[i]);
4285 ubik_VL_ReleaseLock(cstruct, 0, afromvol, RWVOL,
4286 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4289 "Could not release lock on the VLDB entry for volume %lu\n",
4290 (unsigned long)afromvol);
4296 PrintError("", error);
4299 rx_DestroyConnection(fromconn);
4300 if (results.manyResults_val)
4301 free(results.manyResults_val);
4313 dump_sig_handler(int x)
4315 fprintf(STDERR, "\nSignal handler: vos dump operation\n");
4319 /* Dump the volume <afromvol> on <afromserver> and
4320 * <afrompart> to <afilename> starting from <fromdate>.
4321 * DumpFunction does the real work behind the scenes after
4322 * extracting parameters from the rock
4325 UV_DumpVolume(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
4327 afs_int32(*DumpFunction) (struct rx_call *, void *), void *rock,
4330 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
4331 * be changing during the dump */
4332 struct rx_call * volatile fromcall = NULL;
4333 struct rx_connection * volatile fromconn = NULL;
4334 afs_int32 volatile fromtid = 0;
4336 afs_int32 rcode = 0;
4337 afs_int32 code, error = 0;
4339 time_t tmv = fromdate;
4343 #ifndef AFS_NT40_ENV
4344 (void)signal(SIGPIPE, dump_sig_handler);
4346 (void)signal(SIGINT, dump_sig_handler);
4349 VEPRINT("Full Dump ...\n");
4351 VEPRINT1("Incremental Dump (as of %.24s)...\n",
4355 /* get connections to the servers */
4356 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
4358 VEPRINT1("Starting transaction on volume %u...", afromvol);
4360 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &tmp);
4362 EGOTO1(error_exit, code,
4363 "Could not start transaction on the volume %u to be dumped\n",
4367 fromcall = rx_NewCall(fromconn);
4369 VEPRINT1("Starting volume dump on volume %u...", afromvol);
4370 if (flags & VOLDUMPV2_OMITDIRS)
4371 code = StartAFSVolDumpV2(fromcall, fromtid, fromdate, flags);
4373 code = StartAFSVolDump(fromcall, fromtid, fromdate);
4374 EGOTO(error_exit, code, "Could not start the dump process \n");
4377 VEPRINT1("Dumping volume %u...", afromvol);
4378 code = DumpFunction(fromcall, rock);
4379 if (code == RXGEN_OPCODE)
4381 EGOTO(error_exit, code, "Error while dumping volume \n");
4386 code = rx_EndCall(fromcall, 0);
4387 if (code && code != RXGEN_OPCODE)
4388 fprintf(STDERR, "Error in rx_EndCall\n");
4393 VEPRINT1("Ending transaction on volume %u...", afromvol);
4394 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4395 if (code || rcode) {
4396 fprintf(STDERR, "Could not end transaction on the volume %lu\n",
4397 (unsigned long)afromvol);
4399 error = (code ? code : rcode);
4404 rx_DestroyConnection(fromconn);
4406 if (error != RXGEN_OPCODE)
4407 PrintError("", error);
4411 /* Clone the volume <afromvol> on <afromserver> and
4412 * <afrompart>, and then dump the clone volume to
4413 * <afilename> starting from <fromdate>.
4414 * DumpFunction does the real work behind the scenes after
4415 * extracting parameters from the rock
4418 UV_DumpClonedVolume(afs_uint32 afromvol, afs_uint32 afromserver,
4419 afs_int32 afrompart, afs_int32 fromdate,
4420 afs_int32(*DumpFunction) (struct rx_call *, void *),
4421 void *rock, afs_int32 flags)
4423 /* declare stuff 'volatile' that may be used from setjmp/longjmp and may
4424 * be changing during the dump */
4425 struct rx_connection * volatile fromconn = NULL;
4426 struct rx_call * volatile fromcall = NULL;
4427 afs_int32 volatile clonetid = 0;
4428 afs_uint32 volatile clonevol = 0;
4431 afs_int32 fromtid = 0, rcode = 0;
4432 afs_int32 code = 0, error = 0;
4434 time_t tmv = fromdate;
4435 char *volName = NULL;
4439 #ifndef AFS_NT40_ENV
4440 (void)signal(SIGPIPE, dump_sig_handler);
4442 (void)signal(SIGINT, dump_sig_handler);
4445 VEPRINT("Full Dump ...\n");
4447 VEPRINT1("Incremental Dump (as of %.24s)...\n",
4451 /* get connections to the servers */
4452 fromconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
4454 VEPRINT1("Starting transaction on volume %u...", afromvol);
4455 code = AFSVolTransCreate_retry(fromconn, afromvol, afrompart, ITBusy, &fromtid);
4456 EGOTO1(error_exit, code,
4457 "Could not start transaction on the volume %u to be dumped\n",
4461 VEPRINT1("Getting the name for volume %u ...", afromvol);
4462 code = AFSVolGetName(fromconn, fromtid, &volName);
4463 EGOTO1(error_exit, code,
4464 "Failed to get the name of the volume %u\n",afromvol);
4467 /* Get a clone id */
4468 VEPRINT1("Allocating new volume id for clone of volume %u ...", afromvol);
4470 code = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &tmpVol);
4472 EGOTO1(error_exit, code,
4473 "Could not get an ID for the clone of volume %u from the VLDB\n",
4477 /* Do the clone. Default flags on clone are set to delete on salvage and out of service */
4478 VEPRINT2("Cloning source volume %u to clone volume %u...", afromvol,
4482 AFSVolClone(fromconn, fromtid, 0, readonlyVolume, volName, &tmpVol);
4484 EGOTO1(error_exit, code, "Failed to clone the source volume %u\n",
4488 VEPRINT1("Ending the transaction on the volume %u ...", afromvol);
4490 code = AFSVolEndTrans(fromconn, fromtid, &rcode);
4494 EGOTO1(error_exit, code,
4495 "Failed to end the transaction on the volume %u\n", afromvol);
4499 VEPRINT1("Starting transaction on the cloned volume %u ...", clonevol);
4502 AFSVolTransCreate_retry(fromconn, clonevol, afrompart, ITOffline,
4505 EGOTO1(error_exit, code,
4506 "Failed to start a transaction on the cloned volume%u\n",
4510 VEPRINT1("Setting flags on cloned volume %u ...", clonevol);
4511 code = AFSVolSetFlags(fromconn, clonetid, VTDeleteOnSalvage | VTOutOfService); /*redundant */
4512 EGOTO1(error_exit, code, "Could not set falgs on the cloned volume %u\n",
4517 fromcall = rx_NewCall(fromconn);
4519 VEPRINT1("Starting volume dump from cloned volume %u...", clonevol);
4520 if (flags & VOLDUMPV2_OMITDIRS)
4521 code = StartAFSVolDumpV2(fromcall, clonetid, fromdate, flags);
4523 code = StartAFSVolDump(fromcall, clonetid, fromdate);
4524 EGOTO(error_exit, code, "Could not start the dump process \n");
4527 VEPRINT1("Dumping volume %u...", afromvol);
4528 code = DumpFunction(fromcall, rock);
4529 EGOTO(error_exit, code, "Error while dumping volume \n");
4536 /* now delete the clone */
4537 VEPRINT1("Deleting the cloned volume %u ...", clonevol);
4538 code = AFSVolDeleteVolume(fromconn, clonetid);
4540 fprintf(STDERR, "Failed to delete the cloned volume %lu\n",
4541 (unsigned long)clonevol);
4547 code = rx_EndCall(fromcall, 0);
4549 fprintf(STDERR, "Error in rx_EndCall\n");
4555 VEPRINT1("Ending transaction on cloned volume %u...", clonevol);
4556 code = AFSVolEndTrans(fromconn, clonetid, &rcode);
4557 if (code || rcode) {
4559 "Could not end transaction on the cloned volume %lu\n",
4560 (unsigned long)clonevol);
4562 error = (code ? code : rcode);
4567 rx_DestroyConnection(fromconn);
4569 PrintError("", error);
4576 * Restore a volume <tovolid> <tovolname> on <toserver> <topart> from
4577 * the dump file <afilename>. WriteData does all the real work
4578 * after extracting params from the rock
4581 UV_RestoreVolume2(afs_uint32 toserver, afs_int32 topart, afs_uint32 tovolid,
4582 afs_uint32 toparentid, char tovolname[], int flags,
4583 afs_int32(*WriteData) (struct rx_call *, void *),
4586 struct rx_connection *toconn, *tempconn;
4587 struct rx_call *tocall;
4588 afs_int32 totid, code, rcode, vcode, terror = 0;
4589 struct volser_status tstatus;
4590 struct volintInfo vinfo;
4592 char tovolreal[VOLSER_OLDMAXVOLNAME];
4594 afs_int32 temptid, pparentid;
4595 struct nvldbentry entry, storeEntry;
4598 struct restoreCookie cookie;
4600 afs_int32 volflag, voltype, volsertype;
4601 afs_int32 oldCreateDate, oldUpdateDate, newCreateDate, newUpdateDate;
4602 VolumeId oldCloneId = 0;
4603 VolumeId oldBackupId = 0;
4604 int index, same, errcode;
4608 memset(&cookie, 0, sizeof(cookie));
4612 tocall = (struct rx_call *)0;
4613 tempconn = (struct rx_connection *)0;
4617 if (flags & RV_RDONLY) {
4619 volsertype = volser_RO;
4622 volsertype = volser_RW;
4626 pparentid = toparentid;
4627 toconn = UV_Bind(toserver, AFSCONF_VOLUMEPORT);
4628 if (pvolid == 0) { /*alot a new id if needed */
4629 vcode = VLDB_GetEntryByName(tovolname, &entry);
4630 if (vcode == VL_NOENT) {
4631 vcode = ubik_VL_GetNewVolumeId(cstruct, 0, 1, &pvolid);
4633 fprintf(STDERR, "Could not get an Id for the volume %s\n",
4639 } else if (flags & RV_RDONLY) {
4640 if (entry.flags & VLF_RWEXISTS) {
4642 "Entry for ReadWrite volume %s already exists!\n",
4644 error = VOLSERBADOP;
4647 if (!entry.volumeId[ROVOL]) {
4649 "Existing entry for volume %s has no ReadOnly ID\n",
4651 error = VOLSERBADOP;
4654 pvolid = entry.volumeId[ROVOL];
4655 pparentid = entry.volumeId[RWVOL];
4657 pvolid = entry.volumeId[RWVOL];
4658 pparentid = entry.volumeId[RWVOL];
4661 if (!pparentid) pparentid = pvolid;
4662 /* at this point we have a volume id to use/reuse for the volume to be restored */
4663 strncpy(tovolreal, tovolname, VOLSER_OLDMAXVOLNAME);
4665 if (strlen(tovolname) > (VOLSER_OLDMAXVOLNAME - 1)) {
4666 EGOTO1(refail, VOLSERBADOP,
4667 "The volume name %s exceeds the maximum limit of (VOLSER_OLDMAXVOLNAME -1 ) bytes\n",
4670 if ((pparentid != pvolid) && (flags & RV_RDONLY)) {
4671 if (strlen(tovolname) > (VOLSER_OLDMAXVOLNAME - 10)) {
4672 EGOTO1(refail, VOLSERBADOP,
4673 "The volume name %s exceeds the maximum limit of (VOLSER_OLDMAXVOLNAME -1 ) bytes\n", tovolname);
4675 snprintf(tovolreal, VOLSER_OLDMAXVOLNAME, "%s.readonly", tovolname);
4678 MapPartIdIntoName(topart, partName);
4679 fprintf(STDOUT, "Restoring volume %s Id %lu on server %s partition %s ..",
4680 tovolreal, (unsigned long)pvolid,
4681 noresolve ? afs_inet_ntoa_r(toserver, hoststr) :
4682 hostutil_GetNameByINet(toserver), partName);
4685 AFSVolCreateVolume(toconn, topart, tovolreal, volsertype, pparentid, &pvolid,
4688 if (flags & RV_FULLRST) { /* full restore: delete then create anew */
4689 code = DoVolDelete(toconn, pvolid, topart, "the previous", 0,
4691 if (code && code != VNOVOL) {
4697 AFSVolCreateVolume(toconn, topart, tovolreal, volsertype, pparentid,
4699 EGOTO1(refail, code, "Could not create new volume %u\n", pvolid);
4702 AFSVolTransCreate_retry(toconn, pvolid, topart, ITOffline, &totid);
4703 EGOTO1(refail, code, "Failed to start transaction on %u\n",
4706 code = AFSVolGetStatus(toconn, totid, &tstatus);
4707 EGOTO1(refail, code, "Could not get timestamp from volume %u\n",
4711 oldCreateDate = tstatus.creationDate;
4712 oldUpdateDate = tstatus.updateDate;
4713 oldCloneId = tstatus.cloneID;
4714 oldBackupId = tstatus.backupID;
4720 cookie.parent = pparentid;
4721 cookie.type = voltype;
4723 strncpy(cookie.name, tovolreal, VOLSER_OLDMAXVOLNAME);
4725 tocall = rx_NewCall(toconn);
4726 terror = StartAFSVolRestore(tocall, totid, 1, &cookie);
4728 fprintf(STDERR, "Volume restore Failed \n");
4732 code = WriteData(tocall, rock);
4734 fprintf(STDERR, "Could not transmit data\n");
4738 terror = rx_EndCall(tocall, 0);
4739 tocall = (struct rx_call *)0;
4741 fprintf(STDERR, "rx_EndCall Failed \n");
4745 code = AFSVolGetStatus(toconn, totid, &tstatus);
4748 "Could not get status information about the volume %lu\n",
4749 (unsigned long)pvolid);
4753 code = AFSVolSetIdsTypes(toconn, totid, tovolreal, voltype, pparentid,
4754 oldCloneId, oldBackupId);
4756 fprintf(STDERR, "Could not set the right type and IDs on %lu\n",
4757 (unsigned long)pvolid);
4762 if (flags & RV_CRDUMP)
4763 newCreateDate = tstatus.creationDate;
4764 else if (flags & RV_CRKEEP && oldCreateDate != 0)
4765 newCreateDate = oldCreateDate;
4767 newCreateDate = time(0);
4768 if (flags & RV_LUDUMP)
4769 newUpdateDate = tstatus.updateDate;
4770 else if (flags & RV_LUKEEP)
4771 newUpdateDate = oldUpdateDate;
4773 newUpdateDate = time(0);
4774 code = AFSVolSetDate(toconn,totid, newCreateDate);
4776 fprintf(STDERR, "Could not set the 'creation' date on %u\n", pvolid);
4781 init_volintInfo(&vinfo);
4782 vinfo.creationDate = newCreateDate;
4783 vinfo.updateDate = newUpdateDate;
4784 code = AFSVolSetInfo(toconn, totid, &vinfo);
4786 fprintf(STDERR, "Could not set the 'last updated' date on %u\n",
4792 volflag = ((flags & RV_OFFLINE) ? VTOutOfService : 0); /* off or on-line */
4793 code = AFSVolSetFlags(toconn, totid, volflag);
4795 fprintf(STDERR, "Could not mark %lu online\n", (unsigned long)pvolid);
4800 /* It isn't handled right in refail */
4801 code = AFSVolEndTrans(toconn, totid, &rcode);
4806 fprintf(STDERR, "Could not end transaction on %lu\n",
4807 (unsigned long)pvolid);
4812 fprintf(STDOUT, " done\n");
4814 if (!reuseID || (flags & RV_FULLRST)) {
4815 /* Volume was restored on the file server, update the
4816 * VLDB to reflect the change.
4818 vcode = VLDB_GetEntryByID(pvolid, voltype, &entry);
4819 if (vcode && vcode != VL_NOENT && vcode != VL_ENTDELETED) {
4821 "Could not fetch the entry for volume number %lu from VLDB \n",
4822 (unsigned long)pvolid);
4827 MapHostToNetwork(&entry);
4828 if (vcode == VL_NOENT) { /* it doesnot exist already */
4829 /*make the vldb return this indication specifically */
4830 VPRINT("------- Creating a new VLDB entry ------- \n");
4831 strcpy(entry.name, tovolname);
4833 entry.serverNumber[0] = toserver; /*should be indirect */
4834 entry.serverPartition[0] = topart;
4835 entry.serverFlags[0] = (flags & RV_RDONLY) ? VLSF_ROVOL : VLSF_RWVOL;
4836 entry.flags = (flags & RV_RDONLY) ? VLF_ROEXISTS : VLF_RWEXISTS;
4837 if (flags & RV_RDONLY)
4838 entry.volumeId[ROVOL] = pvolid;
4839 else if (tstatus.cloneID != 0) {
4840 entry.volumeId[ROVOL] = tstatus.cloneID; /*this should come from status info on the volume if non zero */
4842 entry.volumeId[ROVOL] = INVALID_BID;
4843 entry.volumeId[RWVOL] = pparentid;
4845 if (tstatus.backupID != 0) {
4846 entry.volumeId[BACKVOL] = tstatus.backupID;
4847 /*this should come from status info on the volume if non zero */
4849 entry.volumeId[BACKVOL] = INVALID_BID;
4850 MapNetworkToHost(&entry, &storeEntry);
4851 vcode = VLDB_CreateEntry(&storeEntry);
4854 "Could not create the VLDB entry for volume number %lu \n",
4855 (unsigned long)pvolid);
4861 EnumerateEntry(&entry);
4862 } else { /*update the existing entry */
4864 fprintf(STDOUT, "Updating the existing VLDB entry\n");
4865 fprintf(STDOUT, "------- Old entry -------\n");
4866 EnumerateEntry(&entry);
4867 fprintf(STDOUT, "------- New entry -------\n");
4870 ubik_VL_SetLock(cstruct, 0, pvolid, voltype,
4874 "Could not lock the entry for volume number %lu \n",
4875 (unsigned long)pvolid);
4880 strcpy(entry.name, tovolname);
4882 /* Update the vlentry with the new information */
4883 if (flags & RV_RDONLY)
4884 index = Lp_ROMatch(toserver, topart, &entry) - 1;
4886 index = Lp_GetRwIndex(&entry);
4888 /* Add the new site for the volume being restored */
4889 entry.serverNumber[entry.nServers] = toserver;
4890 entry.serverPartition[entry.nServers] = topart;
4891 entry.serverFlags[entry.nServers] =
4892 (flags & RV_RDONLY) ? VLSF_ROVOL : VLSF_RWVOL;
4895 /* This volume should be deleted on the old site
4896 * if its different from new site.
4899 VLDB_IsSameAddrs(toserver, entry.serverNumber[index],
4903 "Failed to get info about server's %d address(es) from vlserver (err=%d)\n",
4905 if ((!errcode && !same)
4906 || (entry.serverPartition[index] != topart)) {
4907 if (flags & RV_NODEL) {
4909 ("Not deleting the previous volume %u on server %s, ...",
4911 noresolve ? afs_inet_ntoa_r(entry.serverNumber[index], hoststr) :
4912 hostutil_GetNameByINet(entry.serverNumber[index]));
4915 UV_Bind(entry.serverNumber[index],
4916 AFSCONF_VOLUMEPORT);
4918 MapPartIdIntoName(entry.serverPartition[index],
4921 ("Deleting the previous volume %u on server %s, partition %s ...",
4923 noresolve ? afs_inet_ntoa_r(entry.serverNumber[index], hoststr) :
4924 hostutil_GetNameByINet(entry.serverNumber[index]),
4926 code = DoVolDelete(tempconn, pvolid,
4927 entry.serverPartition[index],
4928 "the", 0, NULL, NULL);
4929 if (code && code != VNOVOL) {
4933 MapPartIdIntoName(entry.serverPartition[index],
4937 entry.serverNumber[index] = toserver;
4938 entry.serverPartition[index] = topart;
4941 entry.flags |= (flags & RV_RDONLY) ? VLF_ROEXISTS : VLF_RWEXISTS;
4942 MapNetworkToHost(&entry, &storeEntry);
4944 VLDB_ReplaceEntry(pvolid, voltype, &storeEntry,
4945 LOCKREL_OPCODE | LOCKREL_AFSID |
4949 "Could not update the entry for volume number %lu \n",
4950 (unsigned long)pvolid);
4956 EnumerateEntry(&entry);
4963 code = rx_EndCall(tocall, 0);
4969 ubik_VL_ReleaseLock(cstruct, 0, pvolid, voltype,
4970 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
4973 "Could not release lock on the VLDB entry for the volume %lu\n",
4974 (unsigned long)pvolid);
4980 code = AFSVolEndTrans(toconn, totid, &rcode);
4984 fprintf(STDERR, "Could not end transaction on the volume %lu \n",
4985 (unsigned long)pvolid);
4991 code = AFSVolEndTrans(toconn, temptid, &rcode);
4995 fprintf(STDERR, "Could not end transaction on the volume %lu \n",
4996 (unsigned long)pvolid);
5002 rx_DestroyConnection(tempconn);
5004 rx_DestroyConnection(toconn);
5005 PrintError("", error);
5010 UV_RestoreVolume(afs_uint32 toserver, afs_int32 topart, afs_uint32 tovolid,
5011 char tovolname[], int flags,
5012 afs_int32(*WriteData) (struct rx_call *, void *),
5015 return UV_RestoreVolume2(toserver, topart, tovolid, 0, tovolname, flags,
5020 /*unlocks the vldb entry associated with <volid> */
5022 UV_LockRelease(afs_uint32 volid)
5026 VPRINT("Binding to the VLDB server\n");
5028 ubik_VL_ReleaseLock(cstruct, 0, volid, -1,
5029 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5032 "Could not unlock the entry for volume number %lu in VLDB \n",
5033 (unsigned long)volid);
5034 PrintError("", vcode);
5037 VPRINT("VLDB updated\n");
5042 /* old interface to add rosites */
5044 UV_AddSite(afs_uint32 server, afs_int32 part, afs_uint32 volid,
5047 return UV_AddSite2(server, part, volid, 0, valid);
5050 /*adds <server> and <part> as a readonly replication site for <volid>
5053 UV_AddSite2(afs_uint32 server, afs_int32 part, afs_uint32 volid,
5054 afs_uint32 rovolid, afs_int32 valid)
5056 int j, nro = 0, islocked = 0;
5057 struct nvldbentry entry, storeEntry, entry2;
5058 afs_int32 vcode, error = 0;
5061 error = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5064 " Could not lock the VLDB entry for the volume %lu \n",
5065 (unsigned long)volid);
5070 error = VLDB_GetEntryByID(volid, RWVOL, &entry);
5073 "Could not fetch the VLDB entry for volume number %lu \n",
5074 (unsigned long)volid);
5078 if (!ISNAMEVALID(entry.name)) {
5080 "Volume name %s is too long, rename before adding site\n",
5082 error = VOLSERBADOP;
5085 MapHostToNetwork(&entry);
5087 /* See if it's too many entries */
5088 if (entry.nServers >= NMAXNSERVERS) {
5089 fprintf(STDERR, "Total number of entries will exceed %u\n",
5091 error = VOLSERBADOP;
5095 /* See if it's on the same server */
5096 for (j = 0; j < entry.nServers; j++) {
5097 if (entry.serverFlags[j] & VLSF_ROVOL) {
5099 if (VLDB_IsSameAddrs(server, entry.serverNumber[j], &error)) {
5102 "Failed to get info about server's %d address(es) from vlserver (err=%d); aborting call!\n",
5105 MapPartIdIntoName(entry.serverPartition[j], apartName);
5107 "RO already exists on partition %s. Multiple ROs on a single server aren't allowed\n",
5109 error = VOLSERBADOP;
5116 /* See if it's too many RO sites - leave one for the RW */
5117 if (nro >= NMAXNSERVERS - 1) {
5118 fprintf(STDERR, "Total number of sites will exceed %u\n",
5120 error = VOLSERBADOP;
5124 /* if rovolid == 0, we leave the RO volume id alone. If the volume doesn't
5125 * have an RO volid at this point (i.e. entry.volumeId[ROVOL] ==
5126 * INVALID_BID) and we leave it alone, it gets an RO volid at release-time.
5129 if (entry.volumeId[ROVOL] == INVALID_BID) {
5130 vcode = VLDB_GetEntryByID(rovolid, -1, &entry2);
5132 fprintf(STDERR, "Volume ID %d already exists\n", rovolid);
5135 VPRINT1("Using RO volume id %d.\n", rovolid);
5136 entry.volumeId[ROVOL] = rovolid;
5138 fprintf(STDERR, "Ignoring given RO id %d, since volume already has RO id %d\n",
5139 rovolid, entry.volumeId[ROVOL]);
5143 VPRINT("Adding a new site ...");
5144 entry.serverNumber[entry.nServers] = server;
5145 entry.serverPartition[entry.nServers] = part;
5147 entry.serverFlags[entry.nServers] = (VLSF_ROVOL | VLSF_DONTUSE);
5149 entry.serverFlags[entry.nServers] = (VLSF_ROVOL);
5153 MapNetworkToHost(&entry, &storeEntry);
5155 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5156 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5158 fprintf(STDERR, "Could not update entry for volume %lu \n",
5159 (unsigned long)volid);
5168 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5169 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5172 "Could not release lock on volume entry for %lu \n",
5173 (unsigned long)volid);
5174 PrintError("", vcode);
5178 PrintError("", error);
5182 /*removes <server> <part> as read only site for <volid> from the vldb */
5184 UV_RemoveSite(afs_uint32 server, afs_int32 part, afs_uint32 volid)
5187 struct nvldbentry entry, storeEntry;
5189 vcode = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5191 fprintf(STDERR, " Could not lock the VLDB entry for volume %lu \n",
5192 (unsigned long)volid);
5193 PrintError("", vcode);
5196 vcode = VLDB_GetEntryByID(volid, RWVOL, &entry);
5199 "Could not fetch the entry for volume number %lu from VLDB \n",
5200 (unsigned long)volid);
5201 PrintError("", vcode);
5204 MapHostToNetwork(&entry);
5205 if (!Lp_ROMatch(server, part, &entry)) {
5206 /*this site doesnot exist */
5207 fprintf(STDERR, "This site is not a replication site \n");
5209 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5210 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5212 fprintf(STDERR, "Could not update entry for volume %lu \n",
5213 (unsigned long)volid);
5214 PrintError("", vcode);
5215 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5216 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5220 } else { /*remove the rep site */
5221 Lp_SetROValue(&entry, server, part, 0, 0);
5223 if ((entry.nServers == 1) && (entry.flags & VLF_RWEXISTS))
5224 entry.flags &= ~VLF_ROEXISTS;
5225 if (entry.nServers < 1) { /*this is the last ref */
5226 VPRINT1("Deleting the VLDB entry for %u ...", volid);
5228 vcode = ubik_VL_DeleteEntry(cstruct, 0, volid, ROVOL);
5231 "Could not delete VLDB entry for volume %lu \n",
5232 (unsigned long)volid);
5233 PrintError("", vcode);
5238 MapNetworkToHost(&entry, &storeEntry);
5239 fprintf(STDOUT, "Deleting the replication site for volume %lu ...",
5240 (unsigned long)volid);
5243 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5244 LOCKREL_OPCODE | LOCKREL_AFSID |
5248 "Could not release lock on volume entry for %lu \n",
5249 (unsigned long)volid);
5250 PrintError("", vcode);
5251 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5252 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5260 /*sets <server> <part> as read/write site for <volid> in the vldb */
5262 UV_ChangeLocation(afs_uint32 server, afs_int32 part, afs_uint32 volid)
5265 struct nvldbentry entry, storeEntry;
5268 vcode = ubik_VL_SetLock(cstruct, 0, volid, RWVOL, VLOP_ADDSITE);
5270 fprintf(STDERR, " Could not lock the VLDB entry for volume %lu \n",
5271 (unsigned long)volid);
5272 PrintError("", vcode);
5275 vcode = VLDB_GetEntryByID(volid, RWVOL, &entry);
5278 "Could not fetch the entry for volume number %lu from VLDB \n",
5279 (unsigned long)volid);
5280 PrintError("", vcode);
5283 MapHostToNetwork(&entry);
5284 index = Lp_GetRwIndex(&entry);
5286 /* no RW site exists */
5287 fprintf(STDERR, "No existing RW site for volume %lu",
5288 (unsigned long)volid);
5290 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5291 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5294 "Could not release lock on entry for volume %lu \n",
5295 (unsigned long)volid);
5296 PrintError("", vcode);
5300 } else { /* change the RW site */
5301 entry.serverNumber[index] = server;
5302 entry.serverPartition[index] = part;
5303 MapNetworkToHost(&entry, &storeEntry);
5305 VLDB_ReplaceEntry(volid, RWVOL, &storeEntry,
5306 LOCKREL_OPCODE | LOCKREL_AFSID |
5309 fprintf(STDERR, "Could not update entry for volume %lu \n",
5310 (unsigned long)volid);
5311 PrintError("", vcode);
5312 ubik_VL_ReleaseLock(cstruct, 0, volid, RWVOL,
5313 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
5321 /*list all the partitions on <aserver> */
5323 UV_ListPartitions(afs_uint32 aserver, struct partList *ptrPartList,
5326 struct rx_connection *aconn;
5327 struct pIDs partIds;
5328 struct partEntries partEnts;
5332 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5334 partEnts.partEntries_len = 0;
5335 partEnts.partEntries_val = NULL;
5336 code = AFSVolXListPartitions(aconn, &partEnts); /* this is available only on new servers */
5337 if (code == RXGEN_OPCODE) {
5338 for (i = 0; i < 26; i++) /* try old interface */
5339 partIds.partIds[i] = -1;
5340 code = AFSVolListPartitions(aconn, &partIds);
5342 for (i = 0; i < 26; i++) {
5343 if ((partIds.partIds[i]) != -1) {
5344 ptrPartList->partId[j] = partIds.partIds[i];
5345 ptrPartList->partFlags[j] = PARTVALID;
5348 ptrPartList->partFlags[i] = 0;
5353 *cntp = partEnts.partEntries_len;
5354 if (*cntp > VOLMAXPARTS) {
5356 "Warning: number of partitions on the server too high %d (process only %d)\n",
5357 *cntp, VOLMAXPARTS);
5358 *cntp = VOLMAXPARTS;
5360 for (i = 0; i < *cntp; i++) {
5361 ptrPartList->partId[i] = partEnts.partEntries_val[i];
5362 ptrPartList->partFlags[i] = PARTVALID;
5364 free(partEnts.partEntries_val);
5370 "Could not fetch the list of partitions from the server\n");
5371 PrintError("", code);
5373 rx_DestroyConnection(aconn);
5378 /*zap the list of volumes specified by volPtrArray (the volCloneId field).
5379 This is used by the backup system */
5381 UV_ZapVolumeClones(afs_uint32 aserver, afs_int32 apart,
5382 struct volDescription *volPtr, afs_int32 arraySize)
5384 struct rx_connection *aconn;
5385 struct volDescription *curPtr;
5388 afs_int32 success = 1;
5390 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5392 for (curPtr = volPtr; curPos < arraySize; curPtr++) {
5393 if (curPtr->volFlags & CLONEVALID) {
5394 curPtr->volFlags &= ~CLONEZAPPED;
5397 code = DoVolDelete(aconn, curPtr->volCloneId, apart,
5398 "clone", 0, NULL, NULL);
5403 curPtr->volFlags |= CLONEZAPPED;
5405 fprintf(STDERR, "Could not zap volume %lu\n",
5406 (unsigned long)curPtr->volCloneId);
5408 VPRINT2("Clone of %s %u deleted\n", curPtr->volName,
5409 curPtr->volCloneId);
5414 rx_DestroyConnection(aconn);
5418 /*return a list of clones of the volumes specified by volPtrArray. Used by the
5421 UV_GenerateVolumeClones(afs_uint32 aserver, afs_int32 apart,
5422 struct volDescription *volPtr, afs_int32 arraySize)
5424 struct rx_connection *aconn;
5425 struct volDescription *curPtr;
5428 afs_int32 rcode = 0;
5430 int reuseCloneId = 0;
5431 afs_uint32 curCloneId = 0;
5432 char cloneName[256]; /*max vol name */
5434 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5436 if ((volPtr->volFlags & REUSECLONEID) && (volPtr->volFlags & ENTRYVALID))
5438 else { /*get a bunch of id's from vldb */
5440 ubik_VL_GetNewVolumeId(cstruct, 0, arraySize, &curCloneId);
5442 fprintf(STDERR, "Could not get ID's for the clone from VLDB\n");
5443 PrintError("", code);
5448 for (curPtr = volPtr; curPos < arraySize; curPtr++) {
5449 if (curPtr->volFlags & ENTRYVALID) {
5451 curPtr->volFlags |= CLONEVALID;
5452 /*make a clone of curParentId and record as curPtr->volCloneId */
5454 AFSVolTransCreate_retry(aconn, curPtr->volId, apart, ITOffline,
5457 VPRINT2("Clone for volume %s %u failed \n", curPtr->volName,
5460 curPtr->volFlags &= ~CLONEVALID; /*cant clone */
5464 if (strlen(curPtr->volName) < (VOLSER_OLDMAXVOLNAME - 9)) {
5465 strcpy(cloneName, curPtr->volName);
5466 strcat(cloneName, "-tmpClone-");
5468 strcpy(cloneName, "-tmpClone");
5470 curPtr->volCloneId = curCloneId;
5475 AFSVolClone(aconn, tid, 0, readonlyVolume, cloneName,
5476 &(curPtr->volCloneId));
5478 curPtr->volFlags &= ~CLONEVALID;
5480 fprintf(STDERR, "Could not clone %s due to error %lu\n",
5481 curPtr->volName, (unsigned long)code);
5482 code = AFSVolEndTrans(aconn, tid, &rcode);
5484 fprintf(STDERR, "WARNING: could not end transaction\n");
5487 VPRINT2("********** Cloned %s temporary %u\n", cloneName,
5488 curPtr->volCloneId);
5489 code = AFSVolEndTrans(aconn, tid, &rcode);
5490 if (code || rcode) {
5491 curPtr->volFlags &= ~CLONEVALID;
5500 rx_DestroyConnection(aconn);
5505 /*list all the volumes on <aserver> and <apart>. If all = 1, then all the
5506 * relevant fields of the volume are also returned. This is a heavy weight operation.*/
5508 UV_ListVolumes(afs_uint32 aserver, afs_int32 apart, int all,
5509 struct volintInfo **resultPtr, afs_int32 * size)
5511 struct rx_connection *aconn;
5513 volEntries volumeInfo;
5516 *resultPtr = (volintInfo *) 0;
5517 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
5518 volumeInfo.volEntries_len = 0;
5520 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5521 code = AFSVolListVolumes(aconn, apart, all, &volumeInfo);
5524 "Could not fetch the list of volumes from the server\n");
5526 *resultPtr = volumeInfo.volEntries_val;
5527 *size = volumeInfo.volEntries_len;
5531 rx_DestroyConnection(aconn);
5532 PrintError("", code);
5536 /*------------------------------------------------------------------------
5537 * EXPORTED UV_XListVolumes
5540 * List the extended information for all the volumes on a particular
5541 * File Server and partition. We may either return the volume's ID
5542 * or all of its extended information.
5545 * a_serverID : Address of the File Server for which we want
5546 * extended volume info.
5547 * a_partID : Partition for which we want the extended
5549 * a_all : If non-zero, fetch ALL the volume info,
5550 * otherwise just the volume ID.
5551 * a_resultPP : Ptr to the address of the area containing
5552 * the returned volume info.
5553 * a_numEntsInResultP : Ptr for the value we set for the number of
5558 * Otherise, the return value of AFSVolXListVolumes.
5561 * This routine is closely related to UV_ListVolumes, which returns
5562 * only the standard level of detail on AFS volumes. It is a
5563 * heavyweight operation, zipping through all the volume entries for
5564 * a given server/partition.
5568 *------------------------------------------------------------------------*/
5571 UV_XListVolumes(afs_uint32 a_serverID, afs_int32 a_partID, int a_all,
5572 struct volintXInfo **a_resultPP,
5573 afs_int32 * a_numEntsInResultP)
5575 struct rx_connection *rxConnP; /*Ptr to the Rx connection involved */
5576 afs_int32 code; /*Error code to return */
5577 volXEntries volumeXInfo; /*Area for returned extended vol info */
5580 * Set up our error code and the area for returned extended volume info.
5581 * We set the val field to a null pointer as a hint for the stub to
5584 *a_numEntsInResultP = 0;
5585 *a_resultPP = (volintXInfo *) 0;
5586 volumeXInfo.volXEntries_val = (volintXInfo *) 0;
5587 volumeXInfo.volXEntries_len = 0;
5590 * Bind to the Volume Server port on the File Server machine in question,
5593 rxConnP = UV_Bind(a_serverID, AFSCONF_VOLUMEPORT);
5594 code = AFSVolXListVolumes(rxConnP, a_partID, a_all, &volumeXInfo);
5596 fprintf(STDERR, "[UV_XListVolumes] Couldn't fetch volume list\n");
5599 * We got the info; pull out the pointer to where the results lie
5600 * and how many entries are there.
5602 *a_resultPP = volumeXInfo.volXEntries_val;
5603 *a_numEntsInResultP = volumeXInfo.volXEntries_len;
5607 * If we got an Rx connection, throw it away.
5610 rx_DestroyConnection(rxConnP);
5612 PrintError("", code);
5614 } /*UV_XListVolumes */
5616 /* get all the information about volume <volid> on <aserver> and <apart> */
5618 UV_ListOneVolume(afs_uint32 aserver, afs_int32 apart, afs_uint32 volid,
5619 struct volintInfo **resultPtr)
5621 struct rx_connection *aconn;
5623 volEntries volumeInfo;
5625 *resultPtr = (volintInfo *) 0;
5626 volumeInfo.volEntries_val = (volintInfo *) 0; /*this hints the stub to allocate space */
5627 volumeInfo.volEntries_len = 0;
5629 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
5630 code = AFSVolListOneVolume(aconn, apart, volid, &volumeInfo);
5633 "Could not fetch the information about volume %lu from the server\n",
5634 (unsigned long)volid);
5636 *resultPtr = volumeInfo.volEntries_val;
5641 rx_DestroyConnection(aconn);
5642 PrintError("", code);
5646 /*------------------------------------------------------------------------
5647 * EXPORTED UV_XListOneVolume
5650 * List the extended information for a volume on a particular File
5651 * Server and partition.
5654 * a_serverID : Address of the File Server for which we want
5655 * extended volume info.
5656 * a_partID : Partition for which we want the extended
5658 * a_volID : Volume ID for which we want the info.
5659 * a_resultPP : Ptr to the address of the area containing
5660 * the returned volume info.
5664 * Otherise, the return value of AFSVolXListOneVolume.
5667 * This routine is closely related to UV_ListOneVolume, which returns
5668 * only the standard level of detail on the chosen AFS volume.
5672 *------------------------------------------------------------------------*/
5675 UV_XListOneVolume(afs_uint32 a_serverID, afs_int32 a_partID, afs_uint32 a_volID,
5676 struct volintXInfo **a_resultPP)
5678 struct rx_connection *rxConnP; /*Rx connection to Volume Server */
5679 afs_int32 code; /*Error code */
5680 volXEntries volumeXInfo; /*Area for returned info */
5683 * Set up our error code, and the area we're in which we are returning
5684 * the info. Setting the val field to a null pointer tells the stub
5685 * to allocate space for us.
5687 *a_resultPP = (volintXInfo *) 0;
5688 volumeXInfo.volXEntries_val = (volintXInfo *) 0;
5689 volumeXInfo.volXEntries_len = 0;
5692 * Bind to the Volume Server port on the File Server machine in question,
5695 rxConnP = UV_Bind(a_serverID, AFSCONF_VOLUMEPORT);
5696 code = AFSVolXListOneVolume(rxConnP, a_partID, a_volID, &volumeXInfo);
5699 "[UV_XListOneVolume] Couldn't fetch the volume information\n");
5702 * We got the info; pull out the pointer to where the results lie.
5704 *a_resultPP = volumeXInfo.volXEntries_val;
5707 * If we got an Rx connection, throw it away.
5710 rx_DestroyConnection(rxConnP);
5712 PrintError("", code);
5717 * Given a volume we read from a partition, check if it is
5718 * represented in the VLDB correctly.
5720 * The VLDB is looked up by the RW volume id (not its name).
5721 * The RW contains the true name of the volume (BK and RO set
5722 * the name in the VLDB only on creation of the VLDB entry).
5723 * We want rules strict enough that when we check all volumes
5724 * on one partition, it does not need to be done again. IE:
5725 * two volumes on different partitions won't constantly
5726 * change a VLDB entry away from what the other set.
5727 * For RW and BK volumes, we will always check the VLDB to see
5728 * if the two exist on the server/partition. May seem redundant,
5729 * but this is an easy check of the VLDB. IE: if the VLDB entry
5730 * says the BK exists but no BK volume is there, we will detect
5731 * this when we check the RW volume.
5732 * VLDB entries are locked only when a change needs to be done.
5733 * Output changed to look a lot like the "vos syncserv" otuput.
5736 CheckVolume(volintInfo * volumeinfo, afs_uint32 aserver, afs_int32 apart,
5737 afs_int32 * modentry, afs_uint32 * maxvolid,
5738 struct nvldbentry *aentry)
5742 afs_int32 code, error = 0;
5743 struct nvldbentry entry, storeEntry;
5745 int pass = 0, createentry, addvolume, modified, mod, doit = 1;
5755 ((volumeinfo->type ==
5756 RWVOL) ? volumeinfo->volid : volumeinfo->parentID);
5759 /* Check to see if the VLDB is ok without locking it (pass 1).
5760 * If it will change, then lock the VLDB entry, read it again,
5761 * then make the changes to it (pass 2).
5764 code = ubik_VL_SetLock(cstruct, 0, rwvolid, RWVOL, VLOP_DELETE);
5766 fprintf(STDERR, "Could not lock VLDB entry for %lu\n",
5767 (unsigned long)rwvolid);
5772 createentry = 0; /* Do we need to create a VLDB entry */
5773 addvolume = 0; /* Add this volume to the VLDB entry */
5774 modified = 0; /* The VLDB entry was modified */
5777 memcpy(&entry, aentry, sizeof(entry));
5779 /* Read the entry from VLDB by its RW volume id */
5780 code = VLDB_GetEntryByID(rwvolid, RWVOL, &entry);
5782 if (code != VL_NOENT) {
5784 "Could not retrieve the VLDB entry for volume %lu \n",
5785 (unsigned long)rwvolid);
5789 memset(&entry, 0, sizeof(entry));
5790 vsu_ExtractName(entry.name, volumeinfo->name); /* Store name of RW */
5794 MapHostToNetwork(&entry);
5798 if (verbose && (pass == 1)) {
5799 fprintf(STDOUT, "_______________________________\n");
5800 fprintf(STDOUT, "\n-- status before -- \n");
5802 fprintf(STDOUT, "\n**does not exist**\n");
5804 if ((entry.flags & VLF_RWEXISTS) || (entry.flags & VLF_ROEXISTS)
5805 || (entry.flags & VLF_BACKEXISTS))
5806 EnumerateEntry(&entry);
5808 fprintf(STDOUT, "\n");
5811 if (volumeinfo->type == RWVOL) { /* RW volume exists */
5817 /* Check existence of RW and BK volumes */
5818 code = CheckVldbRWBK(&entry, &mod);
5824 idx = Lp_GetRwIndex(&entry);
5825 if (idx == -1) { /* RW index not found in the VLDB entry */
5826 idx = entry.nServers; /* put it into next index */
5829 } else { /* RW index found in the VLDB entry. */
5830 /* Verify if this volume's location matches where the VLDB says it is */
5831 if (!Lp_Match(aserver, apart, &entry)) {
5832 if (entry.flags & VLF_RWEXISTS) {
5833 /* The RW volume exists elsewhere - report this one a duplicate */
5835 MapPartIdIntoName(apart, pname);
5837 "*** Warning: Orphaned RW volume %lu exists on %s %s\n",
5838 (unsigned long)rwvolid,
5840 afs_inet_ntoa_r(aserver, hoststr) :
5841 hostutil_GetNameByINet(aserver), pname);
5842 MapPartIdIntoName(entry.serverPartition[idx],
5845 " VLDB reports RW volume %lu exists on %s %s\n",
5846 (unsigned long)rwvolid,
5848 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5849 hostutil_GetNameByINet(entry.
5854 /* The RW volume does not exist - have VLDB point to this one */
5857 /* Check for orphaned BK volume on old partition */
5858 if (entry.flags & VLF_BACKEXISTS) {
5860 MapPartIdIntoName(entry.serverPartition[idx],
5863 "*** Warning: Orphaned BK volume %u exists on %s %s\n",
5864 entry.volumeId[BACKVOL],
5866 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5867 hostutil_GetNameByINet(entry.
5870 MapPartIdIntoName(apart, pname);
5872 " VLDB reports its RW volume %lu exists on %s %s\n",
5873 (unsigned long)rwvolid,
5875 afs_inet_ntoa_r(aserver, hoststr) :
5876 hostutil_GetNameByINet(aserver),
5882 /* Volume location matches the VLDB location */
5883 if ((volumeinfo->backupID && !entry.volumeId[BACKVOL])
5884 || (volumeinfo->cloneID && !entry.volumeId[ROVOL])
5887 (entry.name, volumeinfo->name,
5888 VOLSER_OLDMAXVOLNAME) != 0)) {
5896 entry.flags |= VLF_RWEXISTS;
5897 entry.volumeId[RWVOL] = rwvolid;
5898 if (!entry.volumeId[BACKVOL])
5899 entry.volumeId[BACKVOL] = volumeinfo->backupID;
5900 if (!entry.volumeId[ROVOL])
5901 entry.volumeId[ROVOL] = volumeinfo->cloneID;
5903 entry.serverFlags[idx] = VLSF_RWVOL;
5904 entry.serverNumber[idx] = aserver;
5905 entry.serverPartition[idx] = apart;
5906 strncpy(entry.name, volumeinfo->name, VOLSER_OLDMAXVOLNAME);
5910 /* One last check - to update BK if need to */
5911 code = CheckVldbRWBK(&entry, &mod);
5919 else if (volumeinfo->type == BACKVOL) { /* A BK volume */
5925 /* Check existence of RW and BK volumes */
5926 code = CheckVldbRWBK(&entry, &mod);
5932 idx = Lp_GetRwIndex(&entry);
5933 if (idx == -1) { /* RW index not found in the VLDB entry */
5934 idx = entry.nServers; /* Put it into next index */
5937 } else { /* RW index found in the VLDB entry */
5938 /* Verify if this volume's location matches where the VLDB says it is */
5939 if (!Lp_Match(aserver, apart, &entry)) {
5940 /* VLDB says RW and/or BK is elsewhere - report this BK volume orphaned */
5942 MapPartIdIntoName(apart, pname);
5944 "*** Warning: Orphaned BK volume %lu exists on %s %s\n",
5945 (unsigned long)volumeinfo->volid,
5947 afs_inet_ntoa_r(aserver, hoststr) :
5948 hostutil_GetNameByINet(aserver), pname);
5949 MapPartIdIntoName(entry.serverPartition[idx], pname);
5951 " VLDB reports its RW/BK volume %lu exists on %s %s\n",
5952 (unsigned long)rwvolid,
5954 afs_inet_ntoa_r(entry.serverNumber[idx], hoststr) :
5955 hostutil_GetNameByINet(entry.
5960 if (volumeinfo->volid != entry.volumeId[BACKVOL]) {
5961 if (!(entry.flags & VLF_BACKEXISTS)) {
5963 } else if (volumeinfo->volid >
5964 entry.volumeId[BACKVOL]) {
5968 MapPartIdIntoName(entry.serverPartition[idx],
5971 "*** Warning: Orphaned BK volume %u exists on %s %s\n",
5972 entry.volumeId[BACKVOL],
5974 afs_inet_ntoa_r(aserver, hoststr) :
5975 hostutil_GetNameByINet(aserver),
5978 " VLDB reports its BK volume ID is %lu\n",
5979 (unsigned long)volumeinfo->volid);
5983 MapPartIdIntoName(entry.serverPartition[idx],
5986 "*** Warning: Orphaned BK volume %lu exists on %s %s\n",
5987 (unsigned long)volumeinfo->volid,
5989 afs_inet_ntoa_r(aserver, hoststr) :
5990 hostutil_GetNameByINet(aserver),
5993 " VLDB reports its BK volume ID is %u\n",
5994 entry.volumeId[BACKVOL]);
5997 } else if (!entry.volumeId[BACKVOL]) {
6004 entry.flags |= VLF_BACKEXISTS;
6005 entry.volumeId[RWVOL] = rwvolid;
6006 entry.volumeId[BACKVOL] = volumeinfo->volid;
6008 entry.serverNumber[idx] = aserver;
6009 entry.serverPartition[idx] = apart;
6010 entry.serverFlags[idx] = VLSF_RWVOL;
6016 else if (volumeinfo->type == ROVOL) { /* A RO volume */
6017 if (volumeinfo->volid == entry.volumeId[ROVOL]) {
6018 /* This is a quick check to see if the RO entry exists in the
6019 * VLDB so we avoid the CheckVldbRO() call (which checks if each
6020 * RO volume listed in the VLDB exists).
6022 idx = Lp_ROMatch(aserver, apart, &entry) - 1;
6024 idx = entry.nServers;
6028 if (!(entry.flags & VLF_ROEXISTS)) {
6033 /* Before we correct the VLDB entry, make sure all the
6034 * ROs listed in the VLDB exist.
6036 code = CheckVldbRO(&entry, &mod);
6042 if (!(entry.flags & VLF_ROEXISTS)) {
6043 /* No RO exists in the VLDB entry - add this one */
6044 idx = entry.nServers;
6047 } else if (volumeinfo->volid > entry.volumeId[ROVOL]) {
6048 /* The volume headers's RO ID does not match that in the VLDB entry,
6049 * and the vol hdr's ID is greater (implies more recent). So delete
6050 * all the RO volumes listed in VLDB entry and add this volume.
6052 for (j = 0; j < entry.nServers; j++) {
6053 if (entry.serverFlags[j] & VLSF_ROVOL) {
6054 /* Verify this volume exists and print message we are orphaning it */
6056 MapPartIdIntoName(apart, pname);
6058 "*** Warning: Orphaned RO volume %u exists on %s %s\n",
6059 entry.volumeId[ROVOL],
6061 afs_inet_ntoa_r(entry.serverNumber[j], hoststr) :
6062 hostutil_GetNameByINet(entry.
6066 " VLDB reports its RO volume ID is %lu\n",
6067 (unsigned long)volumeinfo->volid);
6070 Lp_SetRWValue(&entry, entry.serverNumber[idx],
6071 entry.serverPartition[idx], 0L, 0L);
6078 idx = entry.nServers;
6081 } else if (volumeinfo->volid < entry.volumeId[ROVOL]) {
6082 /* The volume headers's RO ID does not match that in the VLDB entry,
6083 * and the vol hdr's ID is lower (implies its older). So orphan it.
6086 MapPartIdIntoName(apart, pname);
6088 "*** Warning: Orphaned RO volume %lu exists on %s %s\n",
6089 (unsigned long)volumeinfo->volid,
6091 afs_inet_ntoa_r(aserver, hoststr) :
6092 hostutil_GetNameByINet(aserver), pname);
6094 " VLDB reports its RO volume ID is %u\n",
6095 entry.volumeId[ROVOL]);
6098 /* The RO volume ID in the volume header match that in the VLDB entry,
6099 * and there exist RO volumes in the VLDB entry. See if any of them
6100 * are this one. If not, then we add it.
6102 idx = Lp_ROMatch(aserver, apart, &entry) - 1;
6104 idx = entry.nServers;
6112 entry.flags |= VLF_ROEXISTS;
6113 entry.volumeId[RWVOL] = rwvolid;
6114 entry.volumeId[ROVOL] = volumeinfo->volid;
6116 entry.serverNumber[idx] = aserver;
6117 entry.serverPartition[idx] = apart;
6118 entry.serverFlags[idx] = VLSF_ROVOL;
6124 /* Remember largest volume id */
6125 if (entry.volumeId[ROVOL] > *maxvolid)
6126 *maxvolid = entry.volumeId[ROVOL];
6127 if (entry.volumeId[BACKVOL] > *maxvolid)
6128 *maxvolid = entry.volumeId[BACKVOL];
6129 if (entry.volumeId[RWVOL] > *maxvolid)
6130 *maxvolid = entry.volumeId[RWVOL];
6132 if (modified && doit) {
6133 MapNetworkToHost(&entry, &storeEntry);
6136 code = VLDB_CreateEntry(&storeEntry);
6139 "Could not create a VLDB entry for the volume %lu\n",
6140 (unsigned long)rwvolid);
6147 VLDB_ReplaceEntry(rwvolid, RWVOL, &storeEntry,
6148 LOCKREL_OPCODE | LOCKREL_AFSID |
6151 fprintf(STDERR, "Could not update entry for %lu\n",
6152 (unsigned long)rwvolid);
6156 } else if (pass == 2) {
6158 ubik_VL_ReleaseLock(cstruct, 0, rwvolid, RWVOL,
6159 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
6161 PrintError("Could not unlock VLDB entry ", code);
6165 if (modified && modentry) {
6170 memcpy(aentry, &entry, sizeof(entry));
6174 fprintf(STDOUT, "-- status after --\n");
6176 EnumerateEntry(&entry);
6178 fprintf(STDOUT, "\n**no change**\n");
6182 VPRINT("\n_______________________________\n");
6187 sortVolumes(const void *a, const void *b)
6189 volintInfo *v1 = (volintInfo *) a;
6190 volintInfo *v2 = (volintInfo *) b;
6191 afs_uint32 rwvolid1, rwvolid2;
6193 rwvolid1 = ((v1->type == RWVOL) ? v1->volid : v1->parentID);
6194 rwvolid2 = ((v2->type == RWVOL) ? v2->volid : v2->parentID);
6196 if (rwvolid1 > rwvolid2)
6197 return -1; /* lower RW id goes first */
6198 if (rwvolid1 < rwvolid2)
6201 if (v1->type == RWVOL)
6202 return -1; /* RW vols go first */
6203 if (v2->type == RWVOL)
6206 if ((v1->type == BACKVOL) && (v2->type == ROVOL))
6207 return -1; /* BK vols next */
6208 if ((v1->type == ROVOL) && (v2->type == BACKVOL))
6211 if (v1->volid < v2->volid)
6212 return 1; /* larger volids first */
6213 if (v1->volid > v2->volid)
6219 * Synchronise <aserver> <apart>(if flags = 1) <avolid>.
6220 * Synchronize an individual volume against a sever and partition.
6221 * Checks the VLDB entry (similar to syncserv) as well as checks
6222 * if the volume exists on specified servers (similar to syncvldb).
6225 UV_SyncVolume(afs_uint32 aserver, afs_int32 apart, char *avolname, int flags)
6227 struct rx_connection *aconn = 0;
6228 afs_int32 j, k, code, vcode, error = 0;
6230 afs_int32 mod, modified = 0, deleted = 0;
6231 struct nvldbentry vldbentry;
6232 afs_uint32 volumeid = 0;
6233 volEntries volumeInfo;
6234 struct partList PartList;
6236 afs_uint32 maxvolid = 0;
6238 volumeInfo.volEntries_val = (volintInfo *) 0;
6239 volumeInfo.volEntries_len = 0;
6241 /* Turn verbose logging off and do our own verbose logging */
6242 /* tverbose must be set before we call ERROR_EXIT() */
6249 if (!aserver && (flags & 1)) {
6250 /* fprintf(STDERR,"Partition option requires a server option\n"); */
6254 /* Read the VLDB entry */
6255 vcode = VLDB_GetEntryByName(avolname, &vldbentry);
6256 if (vcode && (vcode != VL_NOENT)) {
6257 fprintf(STDERR, "Could not access the VLDB for volume %s\n",
6260 } else if (!vcode) {
6261 MapHostToNetwork(&vldbentry);
6265 fprintf(STDOUT, "Processing VLDB entry %s ...\n", avolname);
6266 fprintf(STDOUT, "_______________________________\n");
6267 fprintf(STDOUT, "\n-- status before -- \n");
6269 fprintf(STDOUT, "\n**does not exist**\n");
6271 if ((vldbentry.flags & VLF_RWEXISTS) || (vldbentry.flags & VLF_ROEXISTS)
6272 || (vldbentry.flags & VLF_BACKEXISTS))
6273 EnumerateEntry(&vldbentry);
6275 fprintf(STDOUT, "\n");
6278 /* Verify that all of the VLDB entries exist on the repective servers
6279 * and partitions (this does not require that avolname be a volume ID).
6280 * Equivalent to a syncserv.
6283 /* Tell CheckVldb not to update if appropriate */
6288 code = CheckVldb(&vldbentry, &mod, &deleted);
6290 fprintf(STDERR, "Could not process VLDB entry for volume %s\n",
6298 /* If aserver is given, we will search for the desired volume on it */
6300 /* Generate array of partitions on the server that we will check */
6302 code = UV_ListPartitions(aserver, &PartList, &pcnt);
6305 "Could not fetch the list of partitions from the server\n");
6309 PartList.partId[0] = apart;
6313 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6315 /* If a volume ID were given, search for it on each partition */
6316 if ((volumeid = atol(avolname))) {
6317 for (j = 0; j < pcnt; j++) {
6319 AFSVolListOneVolume(aconn, PartList.partId[j], volumeid,
6322 if (code != ENODEV) {
6323 fprintf(STDERR, "Could not query server\n");
6331 /* Found one, sync it with VLDB entry */
6333 CheckVolume(volumeInfo.volEntries_val, aserver,
6334 PartList.partId[j], &mod, &maxvolid, &vldbentry);
6341 if (volumeInfo.volEntries_val)
6342 free(volumeInfo.volEntries_val);
6343 volumeInfo.volEntries_val = (volintInfo *) 0;
6344 volumeInfo.volEntries_len = 0;
6348 /* Check to see if the RW, BK, and RO IDs exist on any
6349 * partitions. We get the volume IDs from the VLDB.
6351 for (j = 0; j < MAXTYPES; j++) { /* for RW, RO, and BK IDs */
6352 if (vldbentry.volumeId[j] == 0)
6355 for (k = 0; k < pcnt; k++) { /* For each partition */
6356 volumeInfo.volEntries_val = (volintInfo *) 0;
6357 volumeInfo.volEntries_len = 0;
6359 AFSVolListOneVolume(aconn, PartList.partId[k],
6360 vldbentry.volumeId[j], &volumeInfo);
6362 if (code != ENODEV) {
6363 fprintf(STDERR, "Could not query server\n");
6371 /* Found one, sync it with VLDB entry */
6373 CheckVolume(volumeInfo.volEntries_val, aserver,
6374 PartList.partId[k], &mod, &maxvolid, &vldbentry);
6381 if (volumeInfo.volEntries_val)
6382 free(volumeInfo.volEntries_val);
6383 volumeInfo.volEntries_val = (volintInfo *) 0;
6384 volumeInfo.volEntries_len = 0;
6390 /* If verbose output, print a summary of what changed */
6392 fprintf(STDOUT, "-- status after --\n");
6394 fprintf(STDOUT, "\n**entry deleted**\n");
6395 } else if (modified) {
6396 EnumerateEntry(&vldbentry);
6398 fprintf(STDOUT, "\n**no change**\n");
6400 fprintf(STDOUT, "\n_______________________________\n");
6404 /* Now check if the maxvolid is larger than that stored in the VLDB */
6406 afs_uint32 maxvldbid = 0;
6407 code = ubik_VL_GetNewVolumeId(cstruct, 0, 0, &maxvldbid);
6410 "Could not get the highest allocated volume id from the VLDB\n");
6413 } else if (maxvolid > maxvldbid) {
6415 id = maxvolid - maxvldbid + 1;
6416 code = ubik_VL_GetNewVolumeId(cstruct, 0, id, &nid);
6419 "Error in increasing highest allocated volume id in VLDB\n");
6429 fprintf(STDOUT, "...error encountered");
6431 fprintf(STDOUT, "...done entry\n");
6434 rx_DestroyConnection(aconn);
6435 if (volumeInfo.volEntries_val)
6436 free(volumeInfo.volEntries_val);
6438 PrintError("", error);
6443 * Synchronise vldb with the file server <aserver> and,
6444 * optionally, <apart>.
6447 UV_SyncVldb(afs_uint32 aserver, afs_int32 apart, int flags, int force)
6449 struct rx_connection *aconn;
6450 afs_int32 code, error = 0;
6453 volEntries volumeInfo;
6454 struct partList PartList;
6458 afs_int32 failures = 0, modifications = 0, tentries = 0;
6460 afs_uint32 maxvolid = 0;
6463 volumeInfo.volEntries_val = (volintInfo *) 0;
6464 volumeInfo.volEntries_len = 0;
6466 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6468 /* Generate array of partitions to check */
6470 code = UV_ListPartitions(aserver, &PartList, &pcnt);
6473 "Could not fetch the list of partitions from the server\n");
6477 PartList.partId[0] = apart;
6481 VPRINT("Processing volume entries ...\n");
6483 /* Step through the array of partitions */
6484 for (i = 0; i < pcnt; i++) {
6485 apart = PartList.partId[i];
6486 MapPartIdIntoName(apart, pname);
6488 volumeInfo.volEntries_val = (volintInfo *) 0;
6489 volumeInfo.volEntries_len = 0;
6490 code = AFSVolListVolumes(aconn, apart, 1, &volumeInfo);
6493 "Could not fetch the list of volumes from the server\n");
6497 /* May want to sort the entries: RW, BK (high to low), RO (high to low) */
6498 qsort((char *)volumeInfo.volEntries_val, volumeInfo.volEntries_len,
6499 sizeof(volintInfo), sortVolumes);
6502 for (vi = volumeInfo.volEntries_val, j = 0;
6503 j < volumeInfo.volEntries_len; j++, vi++) {
6511 "Processing volume entry %d: %s (%lu) on server %s %s...\n",
6512 j + 1, vi->name, (unsigned long)vi->volid,
6514 afs_inet_ntoa_r(aserver, hoststr) :
6515 hostutil_GetNameByINet(aserver), pname);
6523 code = CheckVolume(vi, aserver, apart, &modified, &maxvolid, NULL);
6525 PrintError("", code);
6528 } else if (modified) {
6534 fprintf(STDOUT, "...error encountered\n\n");
6536 fprintf(STDOUT, "...done entry %d\n\n", j + 1);
6543 "Could not process entries on server %s partition %s\n",
6545 afs_inet_ntoa_r(aserver, hoststr) :
6546 hostutil_GetNameByINet(aserver), pname);
6548 if (volumeInfo.volEntries_val) {
6549 free(volumeInfo.volEntries_val);
6550 volumeInfo.volEntries_val = 0;
6553 } /* thru all partitions */
6556 VPRINT3("Total entries: %u, Failed to process %d, Would change %d\n",
6557 tentries, failures, modifications);
6559 VPRINT3("Total entries: %u, Failed to process %d, Changed %d\n",
6560 tentries, failures, modifications);
6564 /* Now check if the maxvolid is larger than that stored in the VLDB */
6566 afs_uint32 maxvldbid = 0;
6567 code = ubik_VL_GetNewVolumeId(cstruct, 0, 0, &maxvldbid);
6570 "Could not get the highest allocated volume id from the VLDB\n");
6573 } else if (maxvolid > maxvldbid) {
6575 id = maxvolid - maxvldbid + 1;
6576 code = ubik_VL_GetNewVolumeId(cstruct, 0, id, &nid);
6579 "Error in increasing highest allocated volume id in VLDB\n");
6587 rx_DestroyConnection(aconn);
6588 if (volumeInfo.volEntries_val)
6589 free(volumeInfo.volEntries_val);
6590 PrintError("", error);
6595 * Determine if a volume exists on a server and partition.
6596 * Try creating a transaction on the volume. If we can,
6597 * the volume exists, if not, then return the error code.
6598 * Some error codes mean the volume is unavailable but
6599 * still exists - so we catch these error codes.
6602 VolumeExists(afs_uint32 server, afs_int32 partition, afs_uint32 volumeid)
6604 struct rx_connection *conn = (struct rx_connection *)0;
6605 afs_int32 code = -1;
6606 volEntries volumeInfo;
6608 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
6610 volumeInfo.volEntries_val = (volintInfo *) 0;
6611 volumeInfo.volEntries_len = 0;
6612 code = AFSVolListOneVolume(conn, partition, volumeid, &volumeInfo);
6613 if (volumeInfo.volEntries_val)
6614 free(volumeInfo.volEntries_val);
6615 if (code == VOLSERILLEGAL_PARTITION)
6617 rx_DestroyConnection(conn);
6626 CheckVldbRWBK(struct nvldbentry * entry, afs_int32 * modified)
6630 afs_int32 code, error = 0;
6636 idx = Lp_GetRwIndex(entry);
6638 /* Check to see if the RW volume exists and set the VLF_RWEXISTS
6641 if (idx == -1) { /* Did not find a RW entry */
6642 if (entry->flags & VLF_RWEXISTS) { /* ... yet entry says RW exists */
6643 entry->flags &= ~VLF_RWEXISTS; /* ... so say RW does not exist */
6648 VolumeExists(entry->serverNumber[idx],
6649 entry->serverPartition[idx], entry->volumeId[RWVOL]);
6650 if (code == 0) { /* RW volume exists */
6651 if (!(entry->flags & VLF_RWEXISTS)) { /* ... yet entry says RW does not exist */
6652 entry->flags |= VLF_RWEXISTS; /* ... so say RW does exist */
6655 } else if (code == ENODEV) { /* RW volume does not exist */
6656 if (entry->flags & VLF_RWEXISTS) { /* ... yet entry says RW exists */
6657 entry->flags &= ~VLF_RWEXISTS; /* ... so say RW does not exist */
6661 /* If VLDB says it didn't exist, then ignore error */
6662 if (entry->flags & VLF_RWEXISTS) {
6663 MapPartIdIntoName(entry->serverPartition[idx], pname);
6665 "Transaction call failed for RW volume %u on server %s %s\n",
6666 entry->volumeId[RWVOL],
6668 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6669 hostutil_GetNameByINet(entry->serverNumber[idx]),
6676 /* Check to see if the BK volume exists and set the VLF_BACKEXISTS
6677 * flag accordingly. idx already ponts to the RW entry.
6679 if (idx == -1) { /* Did not find a RW entry */
6680 if (entry->flags & VLF_BACKEXISTS) { /* ... yet entry says BK exists */
6681 entry->flags &= ~VLF_BACKEXISTS; /* ... so say BK does not exist */
6684 } else { /* Found a RW entry */
6686 VolumeExists(entry->serverNumber[idx],
6687 entry->serverPartition[idx],
6688 entry->volumeId[BACKVOL]);
6689 if (code == 0) { /* BK volume exists */
6690 if (!(entry->flags & VLF_BACKEXISTS)) { /* ... yet entry says BK does not exist */
6691 entry->flags |= VLF_BACKEXISTS; /* ... so say BK does exist */
6694 } else if (code == ENODEV) { /* BK volume does not exist */
6695 if (entry->flags & VLF_BACKEXISTS) { /* ... yet entry says BK exists */
6696 entry->flags &= ~VLF_BACKEXISTS; /* ... so say BK does not exist */
6700 /* If VLDB says it didn't exist, then ignore error */
6701 if (entry->flags & VLF_BACKEXISTS) {
6702 MapPartIdIntoName(entry->serverPartition[idx], pname);
6704 "Transaction call failed for BK volume %u on server %s %s\n",
6705 entry->volumeId[BACKVOL],
6707 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6708 hostutil_GetNameByINet(entry->serverNumber[idx]),
6715 /* If there is an idx but the BK and RW volumes no
6716 * longer exist, then remove the RW entry.
6718 if ((idx != -1) && !(entry->flags & VLF_RWEXISTS)
6719 && !(entry->flags & VLF_BACKEXISTS)) {
6720 Lp_SetRWValue(entry, entry->serverNumber[idx],
6721 entry->serverPartition[idx], 0L, 0L);
6728 *modified = modentry;
6733 CheckVldbRO(struct nvldbentry *entry, afs_int32 * modified)
6736 int foundro = 0, modentry = 0;
6737 afs_int32 code, error = 0;
6744 /* Check to see if the RO volumes exist and set the VLF_ROEXISTS
6747 for (idx = 0; idx < entry->nServers; idx++) {
6748 if (!(entry->serverFlags[idx] & VLSF_ROVOL)) {
6749 continue; /* not a RO */
6753 VolumeExists(entry->serverNumber[idx],
6754 entry->serverPartition[idx], entry->volumeId[ROVOL]);
6755 if (code == 0) { /* RO volume exists */
6757 } else if (code == ENODEV) { /* RW volume does not exist */
6758 Lp_SetROValue(entry, entry->serverNumber[idx],
6759 entry->serverPartition[idx], 0L, 0L);
6764 MapPartIdIntoName(entry->serverPartition[idx], pname);
6766 "Transaction call failed for RO %u on server %s %s\n",
6767 entry->volumeId[ROVOL],
6769 afs_inet_ntoa_r(entry->serverNumber[idx], hoststr) :
6770 hostutil_GetNameByINet(entry->serverNumber[idx]), pname);
6775 if (foundro) { /* A RO volume exists */
6776 if (!(entry->flags & VLF_ROEXISTS)) { /* ... yet entry says RW does not exist */
6777 entry->flags |= VLF_ROEXISTS; /* ... so say RW does exist */
6780 } else { /* A RO volume does not exist */
6781 if (entry->flags & VLF_ROEXISTS) { /* ... yet entry says RO exists */
6782 entry->flags &= ~VLF_ROEXISTS; /* ... so say RO does not exist */
6789 *modified = modentry;
6794 * Ensure that <entry> matches with the info on file servers
6797 CheckVldb(struct nvldbentry * entry, afs_int32 * modified, afs_int32 * deleted)
6799 afs_int32 code, error = 0;
6800 struct nvldbentry storeEntry;
6801 int islocked = 0, mod, modentry, delentry = 0;
6802 int pass = 0, doit=1;
6810 fprintf(STDOUT, "_______________________________\n");
6811 fprintf(STDOUT, "\n-- status before -- \n");
6812 if ((entry->flags & VLF_RWEXISTS) || (entry->flags & VLF_ROEXISTS)
6813 || (entry->flags & VLF_BACKEXISTS))
6814 EnumerateEntry(entry);
6815 fprintf(STDOUT, "\n");
6818 if (strlen(entry->name) > (VOLSER_OLDMAXVOLNAME - 10)) {
6819 fprintf(STDERR, "Volume name %s exceeds limit of %d characters\n",
6820 entry->name, VOLSER_OLDMAXVOLNAME - 10);
6824 /* Check to see if the VLDB is ok without locking it (pass 1).
6825 * If it will change, then lock the VLDB entry, read it again,
6826 * then make the changes to it (pass 2).
6830 ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL,
6833 fprintf(STDERR, "Could not lock VLDB entry for %u \n",
6834 entry->volumeId[RWVOL]);
6839 code = VLDB_GetEntryByID(entry->volumeId[RWVOL], RWVOL, entry);
6841 fprintf(STDERR, "Could not read VLDB entry for volume %s\n",
6845 MapHostToNetwork(entry);
6851 /* Check if the RW and BK entries are ok */
6852 code = CheckVldbRWBK(entry, &mod);
6855 if (mod && (pass == 1) && doit)
6860 /* Check if the RO volumes entries are ok */
6861 code = CheckVldbRO(entry, &mod);
6864 if (mod && (pass == 1) && doit)
6869 /* The VLDB entry has been updated. If it as been modified, then
6870 * write the entry back out the the VLDB.
6872 if (modentry && doit) {
6876 if (!(entry->flags & VLF_RWEXISTS) && !(entry->flags & VLF_BACKEXISTS)
6877 && !(entry->flags & VLF_ROEXISTS) && doit) {
6878 /* The RW, BK, nor RO volumes do not exist. Delete the VLDB entry */
6880 ubik_VL_DeleteEntry(cstruct, 0, entry->volumeId[RWVOL],
6884 "Could not delete VLDB entry for volume %u \n",
6885 entry->volumeId[RWVOL]);
6890 /* Replace old entry with our new one */
6891 MapNetworkToHost(entry, &storeEntry);
6893 VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry,
6894 (LOCKREL_OPCODE | LOCKREL_AFSID |
6895 LOCKREL_TIMESTAMP));
6897 fprintf(STDERR, "Could not update VLDB entry for volume %u\n",
6898 entry->volumeId[RWVOL]);
6905 if (modified && modentry) {
6908 if (deleted && delentry) {
6913 fprintf(STDOUT, "-- status after --\n");
6915 fprintf(STDOUT, "\n**entry deleted**\n");
6917 EnumerateEntry(entry);
6919 fprintf(STDOUT, "\n**no change**\n");
6923 VPRINT("\n_______________________________\n");
6927 ubik_VL_ReleaseLock(cstruct, 0, entry->volumeId[RWVOL],
6929 (LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP));
6932 "Could not release lock on VLDB entry for volume %u\n",
6933 entry->volumeId[RWVOL]);
6942 * Synchronise <aserver> <apart>(if flags = 1) with the VLDB.
6945 UV_SyncServer(afs_uint32 aserver, afs_int32 apart, int flags, int force)
6947 struct rx_connection *aconn;
6948 afs_int32 code, error = 0;
6949 afs_int32 nentries, tentries = 0;
6950 struct VldbListByAttributes attributes;
6951 nbulkentries arrayEntries;
6952 afs_int32 failures = 0, modified, modifications = 0;
6953 struct nvldbentry *vlentry;
6954 afs_int32 si, nsi, j;
6959 aconn = UV_Bind(aserver, AFSCONF_VOLUMEPORT);
6961 /* Set up attributes to search VLDB */
6962 memset(&attributes, 0, sizeof(attributes));
6963 attributes.server = ntohl(aserver);
6964 attributes.Mask = VLLIST_SERVER;
6966 attributes.partition = apart;
6967 attributes.Mask |= VLLIST_PARTITION;
6970 VPRINT("Processing VLDB entries ...\n");
6972 /* While we need to collect more VLDB entries */
6973 for (si = 0; si != -1; si = nsi) {
6974 memset(&arrayEntries, 0, sizeof(arrayEntries));
6976 /* Collect set of VLDB entries */
6978 VLDB_ListAttributesN2(&attributes, 0, si, &nentries,
6979 &arrayEntries, &nsi);
6980 if (code == RXGEN_OPCODE) {
6981 code = VLDB_ListAttributes(&attributes, &nentries, &arrayEntries);
6985 fprintf(STDERR, "Could not access the VLDB for attributes\n");
6988 tentries += nentries;
6990 for (j = 0; j < nentries; j++) {
6991 vlentry = &arrayEntries.nbulkentries_val[j];
6992 MapHostToNetwork(vlentry);
6994 VPRINT1("Processing VLDB entry %d ...\n", j + 1);
6996 /* Tell CheckVldb not to update if appropriate */
7001 code = CheckVldb(vlentry, &modified, NULL);
7003 PrintError("", code);
7005 "Could not process VLDB entry for volume %s\n",
7008 } else if (modified) {
7014 fprintf(STDOUT, "...error encountered\n\n");
7016 fprintf(STDOUT, "...done entry %d\n\n", j + 1);
7021 if (arrayEntries.nbulkentries_val) {
7022 free(arrayEntries.nbulkentries_val);
7023 arrayEntries.nbulkentries_val = 0;
7028 VPRINT3("Total entries: %u, Failed to process %d, Would change %d\n",
7029 tentries, failures, modifications);
7031 VPRINT3("Total entries: %u, Failed to process %d, Changed %d\n",
7032 tentries, failures, modifications);
7037 rx_DestroyConnection(aconn);
7038 if (arrayEntries.nbulkentries_val)
7039 free(arrayEntries.nbulkentries_val);
7042 error = VOLSERFAILEDOP;
7046 /*rename volume <oldname> to <newname>, changing the names of the related
7047 *readonly and backup volumes. This operation is also idempotent.
7048 *salvager is capable of recovering from rename operation stopping halfway.
7049 *to recover run syncserver on the affected machines,it will force renaming to completion. name clashes should have been detected before calling this proc */
7051 UV_RenameVolume(struct nvldbentry *entry, char oldname[], char newname[])
7053 struct nvldbentry storeEntry;
7054 afs_int32 vcode, code, rcode, error;
7056 char nameBuffer[256];
7058 struct rx_connection *aconn;
7063 aconn = (struct rx_connection *)0;
7067 vcode = ubik_VL_SetLock(cstruct, 0, entry->volumeId[RWVOL], RWVOL, VLOP_ADDSITE); /*last param is dummy */
7070 " Could not lock the VLDB entry for the volume %u \n",
7071 entry->volumeId[RWVOL]);
7076 strncpy(entry->name, newname, VOLSER_OLDMAXVOLNAME);
7077 MapNetworkToHost(entry, &storeEntry);
7078 vcode = VLDB_ReplaceEntry(entry->volumeId[RWVOL], RWVOL, &storeEntry, 0);
7080 fprintf(STDERR, "Could not update VLDB entry for %u\n",
7081 entry->volumeId[RWVOL]);
7085 VPRINT1("Recorded the new name %s in VLDB\n", newname);
7086 /*at this stage the intent to rename is recorded in the vldb, as far as the vldb
7087 * is concerned, oldname is lost */
7088 if (entry->flags & VLF_RWEXISTS) {
7089 index = Lp_GetRwIndex(entry);
7090 if (index == -1) { /* there is a serious discrepancy */
7092 "There is a serious discrepancy in VLDB entry for volume %u\n",
7093 entry->volumeId[RWVOL]);
7094 fprintf(STDERR, "try building VLDB from scratch\n");
7095 error = VOLSERVLDB_ERROR;
7098 aconn = UV_Bind(entry->serverNumber[index], AFSCONF_VOLUMEPORT);
7100 AFSVolTransCreate_retry(aconn, entry->volumeId[RWVOL],
7101 entry->serverPartition[index], ITOffline, &tid);
7102 if (code) { /*volume doesnot exist */
7104 "Could not start transaction on the rw volume %u\n",
7105 entry->volumeId[RWVOL]);
7108 } else { /*volume exists, process it */
7111 AFSVolSetIdsTypes(aconn, tid, newname, RWVOL,
7112 entry->volumeId[RWVOL],
7113 entry->volumeId[ROVOL],
7114 entry->volumeId[BACKVOL]);
7116 VPRINT2("Renamed rw volume %s to %s\n", oldname, newname);
7117 code = AFSVolEndTrans(aconn, tid, &rcode);
7121 "Could not end transaction on volume %s %u\n",
7122 entry->name, entry->volumeId[RWVOL]);
7127 fprintf(STDERR, "Could not set parameters on volume %s %u\n",
7128 entry->name, entry->volumeId[RWVOL]);
7134 rx_DestroyConnection(aconn);
7135 aconn = (struct rx_connection *)0;
7137 /*end rw volume processing */
7138 if (entry->flags & VLF_BACKEXISTS) { /*process the backup volume */
7139 index = Lp_GetRwIndex(entry);
7140 if (index == -1) { /* there is a serious discrepancy */
7142 "There is a serious discrepancy in the VLDB entry for the backup volume %u\n",
7143 entry->volumeId[BACKVOL]);
7144 fprintf(STDERR, "try building VLDB from scratch\n");
7145 error = VOLSERVLDB_ERROR;
7148 aconn = UV_Bind(entry->serverNumber[index], AFSCONF_VOLUMEPORT);
7150 AFSVolTransCreate_retry(aconn, entry->volumeId[BACKVOL],
7151 entry->serverPartition[index], ITOffline, &tid);
7152 if (code) { /*volume doesnot exist */
7154 "Could not start transaction on the backup volume %u\n",
7155 entry->volumeId[BACKVOL]);
7158 } else { /*volume exists, process it */
7159 if (strlen(newname) > (VOLSER_OLDMAXVOLNAME - 8)) {
7161 "Volume name %s.backup exceeds the limit of %u characters\n",
7162 newname, VOLSER_OLDMAXVOLNAME);
7166 strcpy(nameBuffer, newname);
7167 strcat(nameBuffer, ".backup");
7170 AFSVolSetIdsTypes(aconn, tid, nameBuffer, BACKVOL,
7171 entry->volumeId[RWVOL], 0, 0);
7173 VPRINT1("Renamed backup volume to %s \n", nameBuffer);
7174 code = AFSVolEndTrans(aconn, tid, &rcode);
7178 "Could not end transaction on the backup volume %u\n",
7179 entry->volumeId[BACKVOL]);
7185 "Could not set parameters on the backup volume %u\n",
7186 entry->volumeId[BACKVOL]);
7191 } /* end backup processing */
7193 rx_DestroyConnection(aconn);
7194 aconn = (struct rx_connection *)0;
7195 if (entry->flags & VLF_ROEXISTS) { /*process the ro volumes */
7196 for (i = 0; i < entry->nServers; i++) {
7197 if (entry->serverFlags[i] & VLSF_ROVOL) {
7198 aconn = UV_Bind(entry->serverNumber[i], AFSCONF_VOLUMEPORT);
7200 AFSVolTransCreate_retry(aconn, entry->volumeId[ROVOL],
7201 entry->serverPartition[i], ITOffline,
7203 if (code) { /*volume doesnot exist */
7205 "Could not start transaction on the ro volume %u\n",
7206 entry->volumeId[ROVOL]);
7209 } else { /*volume exists, process it */
7210 strcpy(nameBuffer, newname);
7211 strcat(nameBuffer, ".readonly");
7212 if (strlen(nameBuffer) > (VOLSER_OLDMAXVOLNAME - 1)) {
7214 "Volume name %s exceeds the limit of %u characters\n",
7215 nameBuffer, VOLSER_OLDMAXVOLNAME);
7220 AFSVolSetIdsTypes(aconn, tid, nameBuffer, ROVOL,
7221 entry->volumeId[RWVOL], 0, 0);
7223 VPRINT2("Renamed RO volume %s on host %s\n",
7226 afs_inet_ntoa_r(entry->serverNumber[i], hoststr) :
7227 hostutil_GetNameByINet(entry->
7229 code = AFSVolEndTrans(aconn, tid, &rcode);
7233 "Could not end transaction on volume %u\n",
7234 entry->volumeId[ROVOL]);
7240 "Could not set parameters on the ro volume %u\n",
7241 entry->volumeId[ROVOL]);
7247 rx_DestroyConnection(aconn);
7248 aconn = (struct rx_connection *)0;
7255 ubik_VL_ReleaseLock(cstruct, 0, entry->volumeId[RWVOL],
7257 LOCKREL_OPCODE | LOCKREL_AFSID | LOCKREL_TIMESTAMP);
7260 "Could not unlock the VLDB entry for the volume %s %u\n",
7261 entry->name, entry->volumeId[RWVOL]);
7267 code = AFSVolEndTrans(aconn, tid, &rcode);
7271 fprintf(STDERR, "Failed to end transaction on a volume \n");
7277 rx_DestroyConnection(aconn);
7278 PrintError("", error);
7283 /*report on all the active transactions on volser */
7285 UV_VolserStatus(afs_uint32 server, transDebugInfo ** rpntr, afs_int32 * rcount)
7287 struct rx_connection *aconn;
7288 transDebugEntries transInfo;
7291 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7292 transInfo.transDebugEntries_val = (transDebugInfo *) 0;
7293 transInfo.transDebugEntries_len = 0;
7294 code = AFSVolMonitor(aconn, &transInfo);
7297 "Could not access status information about the server\n");
7298 PrintError("", code);
7299 if (transInfo.transDebugEntries_val)
7300 free(transInfo.transDebugEntries_val);
7302 rx_DestroyConnection(aconn);
7305 *rcount = transInfo.transDebugEntries_len;
7306 *rpntr = transInfo.transDebugEntries_val;
7308 rx_DestroyConnection(aconn);
7315 /*delete the volume without interacting with the vldb */
7317 UV_VolumeZap(afs_uint32 server, afs_int32 part, afs_uint32 volid)
7320 struct rx_connection *aconn;
7322 aconn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7323 error = DoVolDelete(aconn, volid, part,
7324 "the", 0, NULL, NULL);
7325 if (error == VNOVOL) {
7326 EPRINT1(error, "Failed to start transaction on %u\n", volid);
7329 PrintError("", error);
7331 rx_DestroyConnection(aconn);
7336 UV_SetVolume(afs_uint32 server, afs_int32 partition, afs_uint32 volid,
7337 afs_int32 transflag, afs_int32 setflag, int sleeptime)
7339 struct rx_connection *conn = 0;
7341 afs_int32 code, error = 0, rcode;
7343 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7345 fprintf(STDERR, "SetVolumeStatus: Bind Failed");
7349 code = AFSVolTransCreate_retry(conn, volid, partition, transflag, &tid);
7351 fprintf(STDERR, "SetVolumeStatus: TransCreate Failed\n");
7355 code = AFSVolSetFlags(conn, tid, setflag);
7357 fprintf(STDERR, "SetVolumeStatus: SetFlags Failed\n");
7362 #ifdef AFS_PTHREAD_ENV
7365 IOMGR_Sleep(sleeptime);
7372 code = AFSVolEndTrans(conn, tid, &rcode);
7373 if (code || rcode) {
7374 fprintf(STDERR, "SetVolumeStatus: EndTrans Failed\n");
7376 error = (code ? code : rcode);
7381 rx_DestroyConnection(conn);
7386 UV_SetVolumeInfo(afs_uint32 server, afs_int32 partition, afs_uint32 volid,
7389 struct rx_connection *conn = 0;
7391 afs_int32 code, error = 0, rcode;
7393 conn = UV_Bind(server, AFSCONF_VOLUMEPORT);
7395 fprintf(STDERR, "SetVolumeInfo: Bind Failed");
7399 code = AFSVolTransCreate_retry(conn, volid, partition, ITOffline, &tid);
7401 fprintf(STDERR, "SetVolumeInfo: TransCreate Failed\n");
7405 code = AFSVolSetInfo(conn, tid, infop);
7407 fprintf(STDERR, "SetVolumeInfo: SetInfo Failed\n");
7414 code = AFSVolEndTrans(conn, tid, &rcode);
7415 if (code || rcode) {
7416 fprintf(STDERR, "SetVolumeInfo: EndTrans Failed\n");
7418 error = (code ? code : rcode);
7423 rx_DestroyConnection(conn);
7428 UV_GetSize(afs_uint32 afromvol, afs_uint32 afromserver, afs_int32 afrompart,
7429 afs_int32 fromdate, struct volintSize *vol_size)
7431 struct rx_connection *aconn = (struct rx_connection *)0;
7432 afs_int32 tid = 0, rcode = 0;
7433 afs_int32 code, error = 0;
7436 /* get connections to the servers */
7437 aconn = UV_Bind(afromserver, AFSCONF_VOLUMEPORT);
7439 VPRINT1("Starting transaction on volume %u...", afromvol);
7440 code = AFSVolTransCreate_retry(aconn, afromvol, afrompart, ITBusy, &tid);
7441 EGOTO1(error_exit, code,
7442 "Could not start transaction on the volume %u to be measured\n",
7446 VPRINT1("Getting size of volume on volume %u...", afromvol);
7447 code = AFSVolGetSize(aconn, tid, fromdate, vol_size);
7448 EGOTO(error_exit, code, "Could not start the measurement process \n");
7453 VPRINT1("Ending transaction on volume %u...", afromvol);
7454 code = AFSVolEndTrans(aconn, tid, &rcode);
7455 if (code || rcode) {
7456 fprintf(STDERR, "Could not end transaction on the volume %u\n",
7458 fprintf(STDERR, "error codes: %d and %d\n", code, rcode);
7460 error = (code ? code : rcode);
7465 rx_DestroyConnection(aconn);
7467 PrintError("", error);
7471 /*maps the host addresses in <old > (present in network byte order) to
7472 that in< new> (present in host byte order )*/
7474 MapNetworkToHost(struct nvldbentry *old, struct nvldbentry *new)
7478 memset(new, 0, sizeof(struct nvldbentry));
7480 /*copy all the fields */
7481 strcpy(new->name, old->name);
7482 /* new->volumeType = old->volumeType;*/
7483 new->nServers = old->nServers;
7484 count = old->nServers;
7485 if (count < NMAXNSERVERS)
7487 for (i = 0; i < count; i++) {
7488 new->serverNumber[i] = ntohl(old->serverNumber[i]);
7489 new->serverPartition[i] = old->serverPartition[i];
7490 new->serverFlags[i] = old->serverFlags[i];
7492 new->volumeId[RWVOL] = old->volumeId[RWVOL];
7493 new->volumeId[ROVOL] = old->volumeId[ROVOL];
7494 new->volumeId[BACKVOL] = old->volumeId[BACKVOL];
7495 new->cloneId = old->cloneId;
7496 new->flags = old->flags;
7499 /*maps the host entries in <entry> which are present in host byte order to network byte order */
7501 MapHostToNetwork(struct nvldbentry *entry)
7505 count = entry->nServers;
7506 if (count < NMAXNSERVERS)
7508 for (i = 0; i < count; i++) {
7509 entry->serverNumber[i] = htonl(entry->serverNumber[i]);