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
13 Institution: The Information Technology Center, Carnegie-Mellon University
20 #include <afs/afssyscalls.h>
23 #define VolumeWriteable(vp) (V_type(vp)==readwriteVolume)
24 #define VolumeWriteable2(vol) (vol.type == readwriteVolume)
25 typedef bit32 FileOffset; /* Offset in this file */
26 #define Date afs_uint32
28 #ifdef AFS_PTHREAD_ENV
31 extern pthread_mutex_t vol_glock_mutex;
32 extern pthread_mutex_t vol_attach_mutex;
33 extern pthread_cond_t vol_put_volume_cond;
34 extern pthread_cond_t vol_sleep_cond;
35 #define VATTACH_LOCK \
36 assert(pthread_mutex_lock(&vol_attach_mutex) == 0);
37 #define VATTACH_UNLOCK \
38 assert(pthread_mutex_unlock(&vol_attach_mutex) == 0);
40 assert(pthread_mutex_lock(&vol_glock_mutex) == 0);
42 assert(pthread_mutex_unlock(&vol_glock_mutex) == 0);
43 #else /* AFS_PTHREAD_ENV */
45 #define VATTACH_UNLOCK
48 #endif /* AFS_PTHREAD_ENV */
50 typedef enum { fileServer, volumeUtility, salvager } ProgramType;
51 extern ProgramType programType; /* The type of program using the package */
53 /* Some initialization parameters for the volume package */
54 /* Add new initialization parameters here */
55 extern int (*V_BreakVolumeCallbacks) ();
56 extern int (*vol_PollProc) ();
57 #define DOPOLL ((vol_PollProc)? (*vol_PollProc)() : 0)
59 struct versionStamp { /* Version stamp for critical volume files */
60 bit32 magic; /* Magic number */
61 bit32 version; /* Version number of this file, or software
62 * that created this file */
65 /* Magic numbers and version stamps for each type of file */
66 #define VOLUMEHEADERMAGIC ((bit32)0x88a1bb3c)
67 #define VOLUMEINFOMAGIC ((bit32)0x78a1b2c5)
68 #define SMALLINDEXMAGIC 0x99776655
69 #define LARGEINDEXMAGIC 0x88664433
70 #define MOUNTMAGIC 0x9a8b7c6d
71 #define ACLMAGIC 0x88877712
72 #define LINKTABLEMAGIC 0x99877712
74 #define VOLUMEHEADERVERSION 1
75 #define VOLUMEINFOVERSION 1
76 #define SMALLINDEXVERSION 1
77 #define LARGEINDEXVERSION 1
78 #define MOUNTVERSION 1
80 #define LINKTABLEVERSION 1
83 * Define whether we are keeping detailed statistics on volume dealings.
85 #define OPENAFS_VOL_STATS 1
89 * Define various indices and counts used in keeping volume-level statistics.
91 #define VOL_STATS_NUM_RWINFO_FIELDS 4
93 #define VOL_STATS_SAME_NET 0 /*Within same site (total) */
94 #define VOL_STATS_SAME_NET_AUTH 1 /*Within same site (authenticated);
95 * (must be 1 more than above) */
96 #define VOL_STATS_DIFF_NET 2 /*From external site (total) */
97 #define VOL_STATS_DIFF_NET_AUTH 3 /*From external site (authenticated)
98 * (must be 1 more than above) */
100 #define VOL_STATS_NUM_TIME_RANGES 6
102 #define VOL_STATS_TIME_CAP_0 60 /*60 seconds */
103 #define VOL_STATS_TIME_CAP_1 600 /*10 minutes, in seconds */
104 #define VOL_STATS_TIME_CAP_2 3600 /*1 hour, in seconds */
105 #define VOL_STATS_TIME_CAP_3 86400 /*1 day, in seconds */
106 #define VOL_STATS_TIME_CAP_4 604800 /*1 week, in seconds */
108 #define VOL_STATS_NUM_TIME_FIELDS 6
110 #define VOL_STATS_TIME_IDX_0 0 /*0 secs to 60 secs */
111 #define VOL_STATS_TIME_IDX_1 1 /*1 min to 10 mins */
112 #define VOL_STATS_TIME_IDX_2 2 /*10 mins to 60 mins */
113 #define VOL_STATS_TIME_IDX_3 3 /*1 hr to 24 hrs */
114 #define VOL_STATS_TIME_IDX_4 4 /*1 day to 7 days */
115 #define VOL_STATS_TIME_IDX_5 5 /*Greater than 1 week */
116 #endif /* OPENAFS_VOL_STATS */
118 /* Volume header. This is the contents of the named file representing
119 * the volume. Read-only by the file server!
121 typedef struct VolumeHeader {
122 struct versionStamp stamp; /* Must be first field */
123 VolumeId id; /* Volume number */
124 VolumeId parent; /* Read-write volume number (or this volume
125 * number if this is a read-write volume) */
127 Inode smallVnodeIndex;
128 Inode largeVnodeIndex;
130 Inode volumeMountTable;
135 typedef struct VolumeDiskHeader {
136 struct versionStamp stamp; /* Must be first field */
137 VolumeId id; /* Volume number */
138 VolumeId parent; /* Read-write volume number (or this volume
139 * number if this is a read-write volume) */
140 afs_int32 volumeInfo_lo;
141 afs_int32 smallVnodeIndex_lo;
142 afs_int32 largeVnodeIndex_lo;
143 afs_int32 volumeAcl_lo;
144 afs_int32 volumeMountTable_lo;
145 afs_int32 volumeInfo_hi;
146 afs_int32 smallVnodeIndex_hi;
147 afs_int32 largeVnodeIndex_hi;
148 afs_int32 volumeAcl_hi;
149 afs_int32 volumeMountTable_hi;
150 afs_int32 linkTable_lo;
151 afs_int32 linkTable_hi;
152 /* If you add fields, add them before here and reduce the size of array */
154 } VolumeDiskHeader_t;
156 /* A vnode index file header */
157 struct IndexFileHeader {
158 struct versionStamp stamp;
162 /******************************************************************************/
163 /* Volume Data which is stored on disk and can also be maintained in memory. */
164 /******************************************************************************/
165 typedef struct VolumeDiskData {
166 struct versionStamp stamp; /* Must be first field */
167 VolumeId id; /* Volume id--unique over all systems */
168 #define VNAMESIZE 32 /* including 0 byte */
169 char name[VNAMESIZE]; /* Unofficial name for the volume */
170 byte inUse; /* Volume is being used (perhaps it is online),
171 * or the system crashed while it was used */
172 byte inService; /* Volume in service, not necessarily on line
173 * This bit is set by an operator/system
174 * programmer. Manually taking a volume offline
175 * always clears the inService bit. Taking
176 * it out of service also takes it offline */
177 byte blessed; /* Volume is administratively blessed with
178 * the ability to go on line. Set by a system
179 * administrator. Clearing this bit will
180 * take the volume offline */
181 byte needsSalvaged; /* Volume needs salvaged--an unrecoverable
182 * error occured to the volume. Note: a volume
183 * may still require salvage even if this
184 * flag isn't set--e.g. if a system crash
185 * occurred while the volume was on line. */
186 bit32 uniquifier; /* Next vnode uniquifier for this volume */
188 VolId parentId; /* Id of parent, if type==readonly */
189 VolId cloneId; /* Latest read-only clone, if type==readwrite,
190 * 0 if the volume has never been cloned. Note: the
191 * indicated volume does not necessarily exist (it
192 * may have been deleted since cloning). */
193 VolId backupId; /* Latest backup copy of this read write volume */
194 VolId restoredFromId; /* The id in the dump this volume was restored from--used simply
195 * to make sure that an incremental dump is not restored on top
196 * of something inappropriate: Note: this field itself is NEVER
198 byte needsCallback; /* Set by the salvager if anything was changed
199 * about the volume. Note: this is not set by
200 * clone/makebackups when setting the copy-on-write
201 * flag in directories; this flag is not seen by
203 #define DESTROY_ME 0xD3
204 byte destroyMe; /* If this is set to DESTROY_ME, then the salvager should destroy
205 * this volume; it is bogus (left over from an aborted volume move,
206 * for example). Note: if this flag is on, then inService should
207 * be OFF--only the salvager checks this flag */
208 #ifdef ALPHA_DUX40_ENV
209 #define DONT_SALVAGE 0xE6
210 #else /* ALPHA_DUX40_ENV */
211 #define DONT_SALVAGE 0xE5
212 #endif /* ALPHA_DUX40_ENV */
213 byte dontSalvage; /* If this is on, then don't bother salvaging this volume */
219 /* Administrative stuff */
220 int maxquota; /* Quota maximum, 1K blocks */
221 int minquota; /* Quota minimum, 1K blocks */
222 int maxfiles; /* Maximum number of files (i.e. inodes) */
223 bit32 accountNumber; /* Uninterpreted account number */
224 bit32 owner; /* The person administratively responsible
226 int reserved2[8]; /* Other administrative constraints */
228 /* Resource usage & statistics */
229 int filecount; /* Actual number of files */
230 int diskused; /* Actual disk space used, 1K blocks */
231 int dayUse; /* Metric for today's usage of this volume so far */
232 int weekUse[7]; /* Usage of the volume for the last week.
233 * weekUse[0] is for most recent complete 24 hour period
234 * of measurement; week[6] is 7 days ago */
235 Date dayUseDate; /* Date the dayUse statistics refer to; the week use stats
236 * are the preceding 7 days */
237 int reserved3[11]; /* Other stats here */
239 /* Server supplied dates */
240 Date creationDate; /* Creation date for a read/write
241 * volume; cloning date for original copy of
242 * a readonly volume (replicated volumes have
243 * the same creation date) */
244 Date accessDate; /* Last access time by a user, large granularity */
245 Date updateDate; /* Last modification by user */
246 Date expirationDate; /* 0 if it never expires */
247 Date backupDate; /* last time a backup clone was taken */
249 /* Time that this copy of this volume was made. NEVER backed up. This field is only
250 * set when the copy is created */
253 #if OPENAFS_VOL_STATS
254 bit32 stat_initialized; /*Are the stat fields below set up? */
258 #endif /* OPENAFS_VOL_STATS */
262 char offlineMessage[VMSGSIZE]; /* Why the volume is offline */
263 #if OPENAFS_VOL_STATS
264 #define VOL_STATS_BYTES 128
266 * Keep per-volume aggregate statistics on type and distance of access,
267 * along with authorship info.
269 bit32 stat_reads[VOL_STATS_NUM_RWINFO_FIELDS];
270 bit32 stat_writes[VOL_STATS_NUM_RWINFO_FIELDS];
271 bit32 stat_fileSameAuthor[VOL_STATS_NUM_TIME_FIELDS];
272 bit32 stat_fileDiffAuthor[VOL_STATS_NUM_TIME_FIELDS];
273 bit32 stat_dirSameAuthor[VOL_STATS_NUM_TIME_FIELDS];
274 bit32 stat_dirDiffAuthor[VOL_STATS_NUM_TIME_FIELDS];
276 char motd[VMSGSIZE]; /* Volume "message of the day" */
277 #endif /* OPENAFS_VOL_STATS */
282 /**************************************/
283 /* Memory resident volume information */
284 /**************************************/
285 typedef struct Volume {
286 struct Volume *hashNext; /* Next in hash resolution table */
287 VolumeId hashid; /* Volume number -- for hash table lookup */
288 struct volHeader *header; /* Cached disk data */
289 Device device; /* Unix device for the volume */
291 *partition; /* Information about the Unix partition */
293 IHandle_t *handle; /* Unix inode holding this index */
294 byte *bitmap; /* Index bitmap */
295 afs_uint32 bitmapSize; /* length of bitmap, in bytes */
296 afs_uint32 bitmapOffset; /* Which byte address of the first long to
297 * start search from in bitmap */
298 } vnodeIndex[nVNODECLASSES];
299 IHandle_t *linkHandle;
300 Unique nextVnodeUnique; /* Derived originally from volume uniquifier.
301 * This is the actual next version number to
302 * assign; the uniquifier is bumped by 200 and
303 * and written to disk every 200 file creates
304 * If the volume is shutdown gracefully, the
305 * uniquifier should be rewritten with the
306 * value nextVnodeVersion */
307 IHandle_t *diskDataHandle; /* Unix inode holding general volume info */
308 bit16 vnodeHashOffset; /* Computed by HashOffset function in vnode.h.
309 * Assigned to the volume when initialized.
310 * Added to vnode number for hash table index */
311 byte shuttingDown; /* This volume is going to be detached */
312 byte goingOffline; /* This volume is going offline */
313 bit32 cacheCheck; /* Online sequence number to be used to invalidate vnode cache entries
314 * that stayed around while a volume was offline */
315 short nUsers; /* Number of users of this volume header */
316 byte needsPutBack; /* For a volume utility, this flag is set if we need
317 * to give the volume back when we detach it. The server has
318 * certain modes where it doesn't detach the volume, and
319 * if we give it back spuriously, the server aborts. This field
320 * is meaningless on the file server */
321 byte specialStatus; /* An error code to return on VGetVolume: the
322 * volume is unavailable for the reason quoted,
323 * currently VBUSY or VMOVED */
324 afs_uint32 updateTime; /* Time that this volume was put on the updated
325 * volume list--the list of volumes that will be
326 * salvaged should the file server crash */
330 struct volHeader *prev, *next; /* LRU pointers */
331 VolumeDiskData diskstuff; /* General volume info read from disk */
332 Volume *back; /* back pointer to current volume structure */
335 /* These macros are used to export fields within the volume header. This was added
336 to facilitate changing the actual representation */
338 #define V_device(vp) ((vp)->device)
339 #define V_partition(vp) ((vp)->partition)
340 #define V_diskDataHandle(vp) ((vp)->diskDataHandle)
341 #define V_vnodeIndex(vp) ((vp)->vnodeIndex)
342 #define V_nextVnodeUnique(vp) ((vp)->nextVnodeUnique)
343 #define V_linkHandle(vp) ((vp)->linkHandle)
345 /* N.B. V_id must be this, rather than vp->id, or some programs will break, probably */
346 #define V_stamp(vp) ((vp)->header->diskstuff.stamp)
347 #define V_id(vp) ((vp)->header->diskstuff.id)
348 #define V_name(vp) ((vp)->header->diskstuff.name)
349 #define V_inUse(vp) ((vp)->header->diskstuff.inUse)
350 #define V_inService(vp) ((vp)->header->diskstuff.inService)
351 #define V_blessed(vp) ((vp)->header->diskstuff.blessed)
352 #define V_needsSalvaged(vp) ((vp)->header->diskstuff.needsSalvaged)
353 #define V_uniquifier(vp) ((vp)->header->diskstuff.uniquifier)
354 #define V_type(vp) ((vp)->header->diskstuff.type)
355 #define V_parentId(vp) ((vp)->header->diskstuff.parentId)
356 #define V_cloneId(vp) ((vp)->header->diskstuff.cloneId)
357 #define V_backupId(vp) ((vp)->header->diskstuff.backupId)
358 #define V_restoredFromId(vp) ((vp)->header->diskstuff.restoredFromId)
359 #define V_needsCallback(vp) ((vp)->header->diskstuff.needsCallback)
360 #define V_destroyMe(vp) ((vp)->header->diskstuff.destroyMe)
361 #define V_dontSalvage(vp) ((vp)->header->diskstuff.dontSalvage)
362 #define V_maxquota(vp) ((vp)->header->diskstuff.maxquota)
363 #define V_minquota(vp) ((vp)->header->diskstuff.minquota)
364 #define V_maxfiles(vp) ((vp)->header->diskstuff.maxfiles)
365 #define V_accountNumber(vp) ((vp)->header->diskstuff.accountNumber)
366 #define V_owner(vp) ((vp)->header->diskstuff.owner)
367 #define V_filecount(vp) ((vp)->header->diskstuff.filecount)
368 #define V_diskused(vp) ((vp)->header->diskstuff.diskused)
369 #define V_dayUse(vp) ((vp)->header->diskstuff.dayUse)
370 #define V_weekUse(vp) ((vp)->header->diskstuff.weekUse)
371 #define V_dayUseDate(vp) ((vp)->header->diskstuff.dayUseDate)
372 #define V_creationDate(vp) ((vp)->header->diskstuff.creationDate)
373 #define V_accessDate(vp) ((vp)->header->diskstuff.accessDate)
374 #define V_updateDate(vp) ((vp)->header->diskstuff.updateDate)
375 #define V_expirationDate(vp) ((vp)->header->diskstuff.expirationDate)
376 #define V_backupDate(vp) ((vp)->header->diskstuff.backupDate)
377 #define V_copyDate(vp) ((vp)->header->diskstuff.copyDate)
378 #define V_offlineMessage(vp) ((vp)->header->diskstuff.offlineMessage)
379 #define V_disk(vp) ((vp)->header->diskstuff)
380 #define V_motd(vp) ((vp)->header->diskstuff.motd)
381 #if OPENAFS_VOL_STATS
382 #define V_stat_initialized(vp) ((vp)->header->diskstuff.stat_initialized)
383 #define V_stat_area(vp) (((vp)->header->diskstuff.stat_reads))
384 #define V_stat_reads(vp, idx) (((vp)->header->diskstuff.stat_reads)[idx])
385 #define V_stat_writes(vp, idx) (((vp)->header->diskstuff.stat_writes)[idx])
386 #define V_stat_fileSameAuthor(vp, idx) (((vp)->header->diskstuff.stat_fileSameAuthor)[idx])
387 #define V_stat_fileDiffAuthor(vp, idx) (((vp)->header->diskstuff.stat_fileDiffAuthor)[idx])
388 #define V_stat_dirSameAuthor(vp, idx) (((vp)->header->diskstuff.stat_dirSameAuthor)[idx])
389 #define V_stat_dirDiffAuthor(vp, idx) (((vp)->header->diskstuff.stat_dirDiffAuthor)[idx])
390 #endif /* OPENAFS_VOL_STATS */
392 /* File offset computations. The offset values in the volume header are
393 computed with these macros -- when the file is written only!! */
394 #define VOLUME_MOUNT_TABLE_OFFSET(Volume) (sizeof (VolumeDiskData))
395 #define VOLUME_BITMAP_OFFSET(Volume) \
396 (sizeof (VolumeDiskData) + (Volume)->disk.mountTableSize)
399 extern char *VSalvageMessage; /* Canonical message when a volume is forced
401 extern Volume *VGetVolume(Error * ec, VolId volumeId);
402 extern Volume *VGetVolume_r(Error * ec, VolId volumeId);
403 extern void VPutVolume(Volume *);
404 extern void VPutVolume_r(Volume *);
405 extern void VOffline(Volume * vp, char *message);
406 extern void VOffline_r(Volume * vp, char *message);
407 extern int VConnectFS(void);
408 extern int VConnectFS_r(void);
409 extern Volume *VAttachVolume(Error * ec, VolumeId volumeId, int mode);
410 extern Volume *VAttachVolume_r(Error * ec, VolumeId volumeId, int mode);
411 extern Volume *VCreateVolume(Error * ec, char *partname, VolId volumeId,
413 extern Volume *VCreateVolume_r(Error * ec, char *partname, VolId volumeId,
415 extern VnodeId VAllocBitmapEntry(Error * ec, Volume * vp,
416 struct vnodeIndex *index);
417 extern VnodeId VAllocBitmapEntry_r(Error * ec, Volume * vp,
418 struct vnodeIndex *index);
419 extern void VFreeBitMapEntry(Error * ec, register struct vnodeIndex *index,
421 extern void VFreeBitMapEntry_r(Error * ec, register struct vnodeIndex *index,
423 extern int VolumeNumber(char *name);
424 extern char *VolumeExternalName(VolumeId volumeId);
425 extern Volume *VAttachVolumeByName(Error * ec, char *partition, char *name,
427 extern Volume *VAttachVolumeByName_r(Error * ec, char *partition, char *name,
429 extern void VShutdown(void);
430 extern void VUpdateVolume(Error * ec, Volume * vp);
431 extern void VUpdateVolume_r(Error * ec, Volume * vp);
432 extern void VAddToVolumeUpdateList(Error * ec, Volume * vp);
433 extern void VAddToVolumeUpdateList_r(Error * ec, Volume * vp);
434 extern void VDetachVolume(Error * ec, Volume * vp);
435 extern void VDetachVolume_r(Error * ec, Volume * vp);
436 extern void VForceOffline(Volume * vp);
437 extern void VForceOffline_r(Volume * vp);
438 extern void VBumpVolumeUsage(register Volume * vp);
439 extern void VBumpVolumeUsage_r(register Volume * vp);
440 extern void VSetDiskUsage(void);
441 extern void VPrintCacheStats(void);
442 extern void VReleaseVnodeFiles_r(Volume * vp);
443 extern void VCloseVnodeFiles_r(Volume * vp);
444 extern struct DiskPartition *VGetPartition(char *name, int abortp);
445 extern struct DiskPartition *VGetPartition_r(char *name, int abortp);
446 extern int VInitVolumePackage(ProgramType pt, int nLargeVnodes,
447 int nSmallVnodes, int connect, int volcache);
448 extern void DiskToVolumeHeader(VolumeHeader_t * h, VolumeDiskHeader_t * dh);
449 extern void VolumeHeaderToDisk(VolumeDiskHeader_t * dh, VolumeHeader_t * h);
450 extern void VTakeOffline_r(register Volume * vp);
451 extern void VTakeOffline(register Volume * vp);
454 /* Naive formula relating number of file size to number of 1K blocks in file */
455 /* Note: we charge 1 block for 0 length files so the user can't store
456 an inifite number of them; for most files, we give him the inode, vnode,
457 and indirect block overhead, for FREE! */
458 #define nBlocks(bytes) ((afs_sfsize_t)((bytes) == 0? 1: (((afs_sfsize_t)(bytes))+1023)/1024))
460 /* Client process id -- file server sends a Check volumes signal back to the client at this pid */
461 #define CLIENTPID "/vice/vol/clientpid"
463 /* Modes of attachment, for VAttachVolume[ByName] to convey to the file server */
464 #define V_READONLY 1 /* Absolutely no updates will be done to the volume */
465 #define V_CLONE 2 /* Cloning the volume: if it is read/write, then directory
466 * version numbers will change. Header will be updated. If
467 * the volume is read-only, the file server may continue to
468 * server it; it may also continue to server it in read/write
469 * mode if the writes are deferred */
470 #define V_VOLUPD 3 /* General update or volume purge is possible. Volume must
472 #define V_DUMP 4 /* A dump of the volume is requested; the volume can be served
473 * read-only during this time */
474 #define V_SECRETLY 5 /* Secret attach of the volume. This is used to attach a volume
475 * which the file server doesn't know about--and which it shouldn't
476 * know about yet, since the volume has just been created and
477 * is somewhat bogus. Required to make sure that a file server
478 * never knows about more than one copy of the same volume--when
479 * a volume is moved from one partition to another on a single
482 #if defined(NEARINODE_HINT)
483 #define V_pref(vp,nearInode) nearInodeHash(V_id(vp),(nearInode)); (nearInode) %= V_partition(vp)->f_files
485 #define V_pref(vp,nearInode) nearInode = 0
486 #endif /* NEARINODE_HINT */
488 #endif /* __volume_h */