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
9 * Portions Copyright (c) 2006-2008 Sine Nomine Associates
15 Institution: The Information Technology Center, Carnegie-Mellon University
22 #include <afs/afssyscalls.h>
25 #define VolumeWriteable(vp) (V_type(vp)==readwriteVolume)
26 #define VolumeWriteable2(vol) (vol.type == readwriteVolume)
27 typedef bit32 FileOffset; /* Offset in this file */
28 #define Date afs_uint32
29 #include "daemon_com.h"
33 /** turn this on if you suspect a volume package locking bug */
34 #define VOL_LOCK_DEBUG 1
38 #define VOL_LOCK_ASSERT_HELD \
39 opr_Assert(vol_glock_holder == pthread_self())
40 #define VOL_LOCK_ASSERT_UNHELD \
41 opr_Assert(vol_glock_holder == 0)
42 #define _VOL_LOCK_SET_HELD \
43 vol_glock_holder = pthread_self()
44 #define _VOL_LOCK_SET_UNHELD \
46 #define VOL_LOCK_DBG_CV_WAIT_END \
48 VOL_LOCK_ASSERT_UNHELD; \
51 #define VOL_LOCK_DBG_CV_WAIT_BEGIN \
53 VOL_LOCK_ASSERT_HELD; \
54 _VOL_LOCK_SET_UNHELD; \
57 #define VOL_LOCK_ASSERT_HELD
58 #define VOL_LOCK_ASSERT_UNHELD
59 #define VOL_LOCK_DBG_CV_WAIT_BEGIN
60 #define VOL_LOCK_DBG_CV_WAIT_END
64 #ifdef AFS_PTHREAD_ENV
66 extern pthread_mutex_t vol_glock_mutex;
67 extern pthread_mutex_t vol_trans_mutex;
68 extern pthread_cond_t vol_put_volume_cond;
69 extern pthread_cond_t vol_sleep_cond;
70 extern pthread_cond_t vol_vinit_cond;
71 extern ih_init_params vol_io_params;
72 extern int vol_attach_threads;
74 extern pthread_t vol_glock_holder;
77 opr_mutex_enter(&vol_glock_mutex); \
78 VOL_LOCK_ASSERT_UNHELD; \
83 VOL_LOCK_ASSERT_HELD; \
84 _VOL_LOCK_SET_UNHELD; \
85 opr_mutex_exit(&vol_glock_mutex); \
87 #define VOL_CV_WAIT(cv) \
89 VOL_LOCK_DBG_CV_WAIT_BEGIN; \
90 opr_cv_wait((cv), &vol_glock_mutex); \
91 VOL_LOCK_DBG_CV_WAIT_END; \
93 #else /* !VOL_LOCK_DEBUG */
94 #define VOL_LOCK opr_mutex_enter(&vol_glock_mutex)
95 #define VOL_UNLOCK opr_mutex_exit(&vol_glock_mutex)
96 #define VOL_CV_WAIT(cv) opr_cv_wait((cv), &vol_glock_mutex)
97 #endif /* !VOL_LOCK_DEBUG */
99 #define VSALVSYNC_LOCK opr_mutex_enter(&vol_salvsync_mutex)
100 #define VSALVSYNC_UNLOCK opr_mutex_exit(&vol_salvsync_mutex)
101 #define VTRANS_LOCK opr_mutex_enter(&vol_trans_mutex)
102 #define VTRANS_UNLOCK opr_mutex_exit(&vol_trans_mutex)
103 #else /* AFS_PTHREAD_ENV */
106 #define VSALVSYNC_LOCK
107 #define VSALVSYNC_UNLOCK
109 #define VTRANS_UNLOCK
110 #endif /* AFS_PTHREAD_ENV */
113 * volume package program type enumeration.
116 fileServer = 1, /**< the fileserver process */
117 volumeUtility = 2, /**< any miscellaneous volume utility */
118 salvager = 3, /**< standalone whole-partition salvager */
119 salvageServer = 4, /**< dafs online salvager */
120 debugUtility = 5, /**< fssync-debug or similar utility */
121 volumeServer = 6, /**< the volserver process */
122 volumeSalvager = 7 /**< the standalone single-volume salvager */
124 extern ProgramType programType; /* The type of program using the package */
126 /* Some initialization parameters for the volume package */
127 /* Add new initialization parameters here */
128 extern int (*V_BreakVolumeCallbacks) (VolumeId);
129 extern int (*vol_PollProc) (void);
131 #define DOPOLL ((vol_PollProc)? (*vol_PollProc)() : 0)
133 #ifdef AFS_DEMAND_ATTACH_FS
135 * variable error return code based upon programType and DAFS presence
137 #define DAFS_VSALVAGE ((programType == fileServer) ? VSALVAGING : VSALVAGE)
139 #define DAFS_VSALVAGE (VSALVAGE)
142 struct versionStamp { /* Version stamp for critical volume files */
143 bit32 magic; /* Magic number */
144 bit32 version; /* Version number of this file, or software
145 * that created this file */
148 #ifdef AFS_DEMAND_ATTACH_FS
150 * demand attach volume state enumeration.
152 * @note values must be contiguous in order for VIsValidState() to work correctly
155 VOL_STATE_UNATTACHED = 0, /**< volume is unattached */
156 VOL_STATE_PREATTACHED = 1, /**< volume has been pre-attached */
157 VOL_STATE_ATTACHING = 2, /**< volume is transitioning to fully attached */
158 VOL_STATE_ATTACHED = 3, /**< volume has been fully attached */
159 VOL_STATE_UPDATING = 4, /**< volume is updating on-disk structures */
160 VOL_STATE_GET_BITMAP = 5, /**< volume is getting bitmap entries */
161 VOL_STATE_HDR_LOADING = 6, /**< volume is loading disk header */
162 VOL_STATE_HDR_ATTACHING = 7, /**< volume is getting a header from the LRU */
163 VOL_STATE_SHUTTING_DOWN = 8, /**< volume is shutting down */
164 VOL_STATE_GOING_OFFLINE = 9, /**< volume is going offline */
165 VOL_STATE_OFFLINING = 10, /**< volume is transitioning to offline */
166 VOL_STATE_DETACHING = 11, /**< volume is transitioning to detached */
167 VOL_STATE_SALVSYNC_REQ = 12, /**< volume is blocked on a salvsync request */
168 VOL_STATE_SALVAGING = 13, /**< volume is being salvaged */
169 VOL_STATE_ERROR = 14, /**< volume is in an error state */
170 VOL_STATE_VNODE_ALLOC = 15, /**< volume is busy allocating a new vnode */
171 VOL_STATE_VNODE_GET = 16, /**< volume is busy getting vnode disk data */
172 VOL_STATE_VNODE_CLOSE = 17, /**< volume is busy closing vnodes */
173 VOL_STATE_VNODE_RELEASE = 18, /**< volume is busy releasing vnodes */
174 VOL_STATE_VLRU_ADD = 19, /**< volume is busy being added to a VLRU queue */
175 VOL_STATE_DELETED = 20, /**< volume has been deleted by the volserver */
176 VOL_STATE_SALVAGE_REQ = 21, /**< volume has been requested to be salvaged,
177 * but is waiting for other users to go away
178 * so it can be offlined */
179 VOL_STATE_SCANNING_RXCALLS = 22, /**< volume is scanning vp->rx_call_list
180 * to interrupt RX calls */
181 /* please add new states directly above this line */
182 VOL_STATE_FREED = 23, /**< debugging aid */
183 VOL_STATE_COUNT = 24 /**< total number of valid states */
187 * V_attachFlags bits.
190 VOL_HDR_ATTACHED = 0x1, /**< volume header is attached to Volume struct */
191 VOL_HDR_LOADED = 0x2, /**< volume header contents are valid */
192 VOL_HDR_IN_LRU = 0x4, /**< volume header is in LRU */
193 VOL_IN_HASH = 0x8, /**< volume is in hash table */
194 VOL_ON_VBYP_LIST = 0x10, /**< volume is on VByP list */
195 VOL_IS_BUSY = 0x20, /**< volume is not to be free()d */
196 VOL_ON_VLRU = 0x40, /**< volume is on the VLRU */
197 VOL_HDR_DONTSALV = 0x80, /**< volume header DONTSALVAGE flag is set */
198 VOL_LOCKED = 0x100 /**< volume is disk-locked (@see VLockVolumeNB) */
201 /* VPrintExtendedCacheStats flags */
202 #define VOL_STATS_PER_CHAIN 0x1 /**< compute simple per-chain stats */
203 #define VOL_STATS_PER_CHAIN2 0x2 /**< compute per-chain stats that require scanning
204 * every element of the chain */
206 /* VLRU_SetOptions options */
207 #define VLRU_SET_THRESH 1
208 #define VLRU_SET_INTERVAL 2
209 #define VLRU_SET_MAX 3
210 #define VLRU_SET_ENABLED 4
216 VLRU_QUEUE_NEW = 0, /**< LRU queue for new volumes */
217 VLRU_QUEUE_MID = 1, /**< survivor generation */
218 VLRU_QUEUE_OLD = 2, /**< old generation */
219 VLRU_QUEUE_CANDIDATE = 3, /**< soft detach candidate pool */
220 VLRU_QUEUE_HELD = 4, /* volumes which are not allowed
221 * to be soft detached */
222 VLRU_QUEUE_INVALID = 5 /**< invalid queue id */
225 /* default scanner timing parameters */
226 #define VLRU_DEFAULT_OFFLINE_THRESH (60*60*2) /* 2 hours */
227 #define VLRU_DEFAULT_OFFLINE_INTERVAL (60*2) /* 2 minutes */
228 #define VLRU_DEFAULT_OFFLINE_MAX 8 /* 8 volumes */
232 * DAFS thread-specific options structure
234 typedef struct VThreadOptions {
235 int disallow_salvsync; /**< whether or not salvsync calls are allowed
236 * on this thread (deadlock prevention). */
238 extern pthread_key_t VThread_key;
239 extern VThreadOptions_t VThread_defaults;
241 #endif /* AFS_DEMAND_ATTACH_FS */
243 typedef struct VolumePackageOptions {
244 afs_uint32 nLargeVnodes; /**< size of large vnode cache */
245 afs_uint32 nSmallVnodes; /**< size of small vnode cache */
246 afs_uint32 volcache; /**< size of volume header cache */
248 afs_int32 canScheduleSalvage; /**< can we schedule salvages? (DAFS) */
249 /* (if 'no', we will just error out if we
251 afs_int32 canUseFSSYNC; /**< can we use the FSSYNC channel? */
252 afs_int32 canUseSALVSYNC; /**< can we use the SALVSYNC channel? (DAFS) */
253 afs_int32 unsafe_attach; /**< can we bypass checking the inUse vol
254 * header on attach? */
255 void (*interrupt_rxcall) (struct rx_call *call, afs_int32 error);
256 /**< callback to interrupt RX calls accessing
257 * a going-offline volume */
258 afs_int32 offline_timeout; /**< how long (in seconds) to wait before
259 * interrupting RX calls accessing a
260 * going-offline volume. -1 disables,
261 * 0 means immediately. */
262 afs_int32 offline_shutdown_timeout;
263 /**< how long (in seconds) to wait before
264 * interrupting RX calls accessing a
265 * going-offline volume during shutdown.
266 * -1 disables, 0 means immediately.
267 * Note that the timeout time is calculated
268 * once, when we encounter the first going-
269 * offline volume during shutdown. So if we
270 * encounter multiple going-offline volumes
271 * during shutdown, we will still only wait
272 * for this amount of time in total, not e.g.
273 * for each going-offline volume encountered. */
274 afs_int32 usage_threshold; /*< number of accesses before writing volume header */
275 afs_int32 usage_rate_limit; /*< minimum number of seconds before writing volume
276 * header, after usage_threshold is exceeded */
277 } VolumePackageOptions;
279 /* Magic numbers and version stamps for each type of file */
280 #define VOLUMEHEADERMAGIC ((bit32)0x88a1bb3c)
281 #define VOLUMEINFOMAGIC ((bit32)0x78a1b2c5)
282 #define SMALLINDEXMAGIC 0x99776655
283 #define LARGEINDEXMAGIC 0x88664433
284 #define MOUNTMAGIC 0x9a8b7c6d
285 #define ACLMAGIC 0x88877712
286 #define LINKTABLEMAGIC 0x99877712
288 #define VOLUMEHEADERVERSION 1
289 #define VOLUMEINFOVERSION 1
290 #define SMALLINDEXVERSION 1
291 #define LARGEINDEXVERSION 1
292 #define MOUNTVERSION 1
294 #define LINKTABLEVERSION 1
298 * Define various indices and counts used in keeping volume-level statistics.
300 #define VOL_STATS_NUM_RWINFO_FIELDS 4
302 #define VOL_STATS_SAME_NET 0 /*Within same site (total) */
303 #define VOL_STATS_SAME_NET_AUTH 1 /*Within same site (authenticated);
304 * (must be 1 more than above) */
305 #define VOL_STATS_DIFF_NET 2 /*From external site (total) */
306 #define VOL_STATS_DIFF_NET_AUTH 3 /*From external site (authenticated)
307 * (must be 1 more than above) */
309 #define VOL_STATS_NUM_TIME_RANGES 6
311 #define VOL_STATS_TIME_CAP_0 60 /*60 seconds */
312 #define VOL_STATS_TIME_CAP_1 600 /*10 minutes, in seconds */
313 #define VOL_STATS_TIME_CAP_2 3600 /*1 hour, in seconds */
314 #define VOL_STATS_TIME_CAP_3 86400 /*1 day, in seconds */
315 #define VOL_STATS_TIME_CAP_4 604800 /*1 week, in seconds */
317 #define VOL_STATS_NUM_TIME_FIELDS 6
319 #define VOL_STATS_TIME_IDX_0 0 /*0 secs to 60 secs */
320 #define VOL_STATS_TIME_IDX_1 1 /*1 min to 10 mins */
321 #define VOL_STATS_TIME_IDX_2 2 /*10 mins to 60 mins */
322 #define VOL_STATS_TIME_IDX_3 3 /*1 hr to 24 hrs */
323 #define VOL_STATS_TIME_IDX_4 4 /*1 day to 7 days */
324 #define VOL_STATS_TIME_IDX_5 5 /*Greater than 1 week */
326 /* Volume header. This is the contents of the named file representing
327 * the volume. Read-only by the file server!
329 typedef struct VolumeHeader {
330 struct versionStamp stamp; /* Must be first field */
331 VolumeId id; /* Volume number */
332 VolumeId parent; /* Read-write volume number (or this volume
333 * number if this is a read-write volume) */
335 Inode smallVnodeIndex;
336 Inode largeVnodeIndex;
338 Inode volumeMountTable;
343 typedef struct VolumeDiskHeader {
344 struct versionStamp stamp; /* Must be first field */
345 VolumeId id; /* Volume number */
346 VolumeId parent; /* Read-write volume number (or this volume
347 * number if this is a read-write volume) */
348 afs_int32 volumeInfo_lo;
349 afs_int32 smallVnodeIndex_lo;
350 afs_int32 largeVnodeIndex_lo;
351 afs_int32 volumeAcl_lo;
352 afs_int32 volumeMountTable_lo;
353 afs_int32 volumeInfo_hi;
354 afs_int32 smallVnodeIndex_hi;
355 afs_int32 largeVnodeIndex_hi;
356 afs_int32 volumeAcl_hi;
357 afs_int32 volumeMountTable_hi;
358 afs_int32 linkTable_lo;
359 afs_int32 linkTable_hi;
360 /* If you add fields, add them before here and reduce the size of array */
362 } VolumeDiskHeader_t;
364 /* A vnode index file header */
365 struct IndexFileHeader {
366 struct versionStamp stamp;
370 /******************************************************************************/
371 /* Volume Data which is stored on disk and can also be maintained in memory. */
372 /******************************************************************************/
373 typedef struct VolumeDiskData {
374 struct versionStamp stamp; /* Must be first field */
375 VolumeId id; /* Volume id--unique over all systems */
376 #define VNAMESIZE 32 /* including 0 byte */
377 char name[VNAMESIZE]; /* Unofficial name for the volume */
378 byte inUse; /* Volume is being used (perhaps it is online),
379 * or the system crashed while it was used */
380 byte inService; /* Volume in service, not necessarily on line
381 * This bit is set by an operator/system
382 * programmer. Manually taking a volume offline
383 * always clears the inService bit. Taking
384 * it out of service also takes it offline */
385 byte blessed; /* Volume is administratively blessed with
386 * the ability to go on line. Set by a system
387 * administrator. Clearing this bit will
388 * take the volume offline */
389 byte needsSalvaged; /* Volume needs salvaged--an unrecoverable
390 * error occured to the volume. Note: a volume
391 * may still require salvage even if this
392 * flag isn't set--e.g. if a system crash
393 * occurred while the volume was on line. */
394 bit32 uniquifier; /* Next vnode uniquifier for this volume */
396 VolumeId parentId; /* Id of parent, if type==readonly */
397 VolumeId cloneId; /* Latest read-only clone, if type==readwrite,
398 * 0 if the volume has never been cloned. Note: the
399 * indicated volume does not necessarily exist (it
400 * may have been deleted since cloning). */
401 VolumeId backupId; /* Latest backup copy of this read write volume */
402 VolumeId restoredFromId; /* The id in the dump this volume was restored from--used simply
403 * to make sure that an incremental dump is not restored on top
404 * of something inappropriate: Note: this field itself is NEVER
406 byte needsCallback; /* Set by the salvager if anything was changed
407 * about the volume. Note: this is not set by
408 * clone/makebackups when setting the copy-on-write
409 * flag in directories; this flag is not seen by
411 #define DESTROY_ME 0xD3
412 byte destroyMe; /* If this is set to DESTROY_ME, then the salvager should destroy
413 * this volume; it is bogus (left over from an aborted volume move,
414 * for example). Note: if this flag is on, then inService should
415 * be OFF--only the salvager checks this flag */
416 #define DONT_SALVAGE 0xE5
417 byte dontSalvage; /* If this is on, then don't bother salvaging this volume */
423 /* Administrative stuff */
424 int maxquota; /* Quota maximum, 1K blocks */
425 int minquota; /* Quota minimum, 1K blocks */
426 int maxfiles; /* Maximum number of files (i.e. inodes) */
427 bit32 accountNumber; /* Uninterpreted account number */
428 bit32 owner; /* The person administratively responsible
430 int reserved2[8]; /* Other administrative constraints */
432 /* Resource usage & statistics */
433 int filecount; /* Actual number of files */
434 int diskused; /* Actual disk space used, 1K blocks */
435 int dayUse; /* Metric for today's usage of this volume so far */
436 int weekUse[7]; /* Usage of the volume for the last week.
437 * weekUse[0] is for most recent complete 24 hour period
438 * of measurement; week[6] is 7 days ago */
439 Date dayUseDate; /* Date the dayUse statistics refer to; the week use stats
440 * are the preceding 7 days */
441 unsigned int volUpdateCounter; /*incremented at every update of volume*/
442 int reserved3[10]; /* Other stats here */
444 /* Server supplied dates */
445 Date creationDate; /* Creation date for a read/write
446 * volume; cloning date for original copy of
447 * a readonly volume (replicated volumes have
448 * the same creation date) */
449 Date accessDate; /* Last access time by a user, large granularity */
450 Date updateDate; /* Last modification by user or salvager */
451 Date expirationDate; /* 0 if it never expires */
452 Date backupDate; /* last time a backup clone was taken */
454 /* Time that this copy of this volume was made. NEVER backed up. This field is only
455 * set when the copy is created */
458 bit32 stat_initialized; /*Are the stat fields below set up? */
463 char offlineMessage[VMSGSIZE]; /* Why the volume is offline */
464 #define VOL_STATS_BYTES 128
466 * Keep per-volume aggregate statistics on type and distance of access,
467 * along with authorship info.
469 bit32 stat_reads[VOL_STATS_NUM_RWINFO_FIELDS];
470 bit32 stat_writes[VOL_STATS_NUM_RWINFO_FIELDS];
471 bit32 stat_fileSameAuthor[VOL_STATS_NUM_TIME_FIELDS];
472 bit32 stat_fileDiffAuthor[VOL_STATS_NUM_TIME_FIELDS];
473 bit32 stat_dirSameAuthor[VOL_STATS_NUM_TIME_FIELDS];
474 bit32 stat_dirDiffAuthor[VOL_STATS_NUM_TIME_FIELDS];
479 /**************************************/
480 /* Memory resident volume information */
481 /**************************************/
484 * global volume package stats.
486 typedef struct VolPkgStats {
487 #ifdef AFS_DEMAND_ATTACH_FS
490 * extended volume package statistics
494 afs_uint32 state_levels[VOL_STATE_COUNT]; /**< volume state transition counters */
497 afs_uint64 hash_looks; /**< number of hash chain element traversals */
498 afs_uint64 hash_reorders; /**< number of hash chain reorders */
499 afs_uint64 salvages; /**< online salvages since fileserver start */
500 afs_uint64 vol_ops; /**< volume operations since fileserver start */
501 #endif /* AFS_DEMAND_ATTACH_FS */
503 afs_uint64 hdr_loads; /**< header loads from disk */
504 afs_uint64 hdr_gets; /**< header pulls out of LRU */
505 afs_uint64 attaches; /**< volume attaches since fileserver start */
506 afs_uint64 soft_detaches; /**< soft detach ops since fileserver start */
508 /* configuration parameters */
509 afs_uint32 hdr_cache_size; /**< size of volume header cache */
511 extern VolPkgStats VStats;
514 * volume header cache supporting structures
516 struct volume_hdr_LRU_stats {
522 struct volume_hdr_LRU_t {
524 struct volume_hdr_LRU_stats stats;
526 extern struct volume_hdr_LRU_t volume_hdr_LRU;
529 * volume hash chain supporting structures
531 typedef struct VolumeHashChainHead {
532 struct rx_queue queue;
534 /* someday we could put a per-chain lock here... */
535 #ifdef AFS_DEMAND_ATTACH_FS
539 /* per-chain statistics */
544 pthread_cond_t chain_busy_cv;
545 #endif /* AFS_DEMAND_ATTACH_FS */
546 } VolumeHashChainHead;
548 typedef struct VolumeHashTable {
551 VolumeHashChainHead * Table;
553 extern VolumeHashTable_t VolumeHashTable;
555 struct VolumeHashChainStats {
556 afs_int32 table_size;
558 #ifdef AFS_DEMAND_ATTACH_FS
559 afs_int32 chain_cacheCheck;
560 afs_int32 chain_busy;
561 afs_uint64 chain_looks;
562 afs_uint64 chain_gets;
563 afs_uint64 chain_reorders;
568 #ifdef AFS_DEMAND_ATTACH_FS
570 * DAFS extended per-volume statistics.
572 * @note this data lives across the entire
573 * lifetime of the fileserver process
575 typedef struct VolumeStats {
577 afs_uint64 hash_lookups; /**< hash table lookups */
578 afs_uint64 hash_short_circuits; /**< short circuited hash lookups (due to cacheCheck) */
579 afs_uint64 hdr_loads; /**< header loads from disk */
580 afs_uint64 hdr_gets; /**< header pulls out of LRU */
581 afs_uint16 attaches; /**< attaches of this volume since fileserver start */
582 afs_uint16 soft_detaches; /**< soft detaches of this volume */
583 afs_uint16 salvages; /**< online salvages since fileserver start */
584 afs_uint16 vol_ops; /**< volume operations since fileserver start */
587 afs_uint32 last_attach; /**< unix timestamp of last VAttach */
588 afs_uint32 last_get; /**< unix timestamp of last VGet/VHold */
589 afs_uint32 last_promote; /**< unix timestamp of last VLRU promote/demote */
590 afs_uint32 last_hdr_get; /**< unix timestamp of last GetVolumeHeader() */
591 afs_uint32 last_hdr_load; /**< unix timestamp of last LoadVolumeHeader() */
592 afs_uint32 last_salvage; /**< unix timestamp of last initiation of an online salvage */
593 afs_uint32 last_salvage_req; /**< unix timestamp of last SALVSYNC request */
594 afs_uint32 last_vol_op; /**< unix timestamp of last volume operation */
598 #define SALVAGE_PRIO_UPDATE_INTERVAL 3 /**< number of seconds between prio updates */
599 #define SALVAGE_COUNT_MAX 16 /**< number of online salvages we
600 * allow before moving the volume
601 * into a permanent error state
603 * once this threshold is reached,
604 * the operator will have to manually
605 * issue a 'bos salvage' to bring
606 * the volume back online
610 * DAFS online salvager state.
612 typedef struct VolumeOnlineSalvage {
613 afs_uint32 prio; /**< number of VGetVolume's since salvage requested */
614 int reason; /**< reason for requesting online salvage */
615 byte requested; /**< flag specifying that salvage should be scheduled */
616 byte scheduled; /**< flag specifying whether online salvage scheduled */
617 byte scheduling; /**< if nonzero, this volume has entered
618 * VCheckSalvage(), so if we recurse into
619 * VCheckSalvage() with this set, exit immediately
620 * to avoid recursing forever */
621 byte reserved[1]; /**< padding */
622 } VolumeOnlineSalvage;
625 * DAFS Volume LRU state.
627 typedef struct VolumeVLRUState {
628 struct rx_queue lru; /**< VLRU queue for this generation */
629 VLRUQueueName idx; /**< VLRU generation index */
631 #endif /* AFS_DEMAND_ATTACH_FS */
634 * node for a volume's rx_call_list.
638 struct rx_call *call;
641 typedef struct Volume {
642 struct rx_queue q; /* Volume hash chain pointers */
643 VolumeId hashid; /* Volume number -- for hash table lookup */
644 struct volHeader *header; /* Cached disk data */
645 Device device; /* Unix device for the volume */
646 struct DiskPartition64
647 *partition; /* Information about the Unix partition */
649 IHandle_t *handle; /* Unix inode holding this index */
650 byte *bitmap; /* Index bitmap */
651 afs_uint32 bitmapSize; /* length of bitmap, in bytes */
652 afs_uint32 bitmapOffset; /* Which byte address of the first long to
653 * start search from in bitmap */
654 } vnodeIndex[nVNODECLASSES];
655 IHandle_t *linkHandle;
656 Unique nextVnodeUnique; /* Derived originally from volume uniquifier.
657 * This is the actual next version number to
658 * assign; the uniquifier is bumped by 200 and
659 * and written to disk every 200 file creates
660 * If the volume is shutdown gracefully, the
661 * uniquifier should be rewritten with the
662 * value nextVnodeVersion */
663 IHandle_t *diskDataHandle; /* Unix inode holding general volume info */
664 byte shuttingDown; /* This volume is going to be detached */
665 byte goingOffline; /* This volume is going offline */
666 bit32 cacheCheck; /* Online sequence number to be used to invalidate vnode cache entries
667 * that stayed around while a volume was offline */
668 short nUsers; /* Number of users of this volume header */
669 #define VOL_PUTBACK 1
670 #define VOL_PUTBACK_DELETE 2
671 byte needsPutBack; /* For a volume utility, this flag is set to VOL_PUTBACK if we
672 * need to give the volume back when we detach it. The server has
673 * certain modes where it doesn't detach the volume, and
674 * if we give it back spuriously, the server aborts. If set to
675 * VOL_PUTBACK_DELETE, it indicates that we need to tell the
676 * fileserver that the volume is gone entirely, instead of just
677 * giving the volume back to the fileserver. This field
678 * is meaningless on the file server */
679 byte specialStatus; /* An error code to return on VGetVolume: the
680 * volume is unavailable for the reason quoted,
681 * currently VBUSY or VMOVED */
682 afs_uint32 checkoutMode; /* for volume utilities, mode number for current checkout */
683 afs_uint32 updateTime; /* Time that this volume was put on the updated
684 * volume list--the list of volumes that will be
685 * salvaged should the file server crash */
686 struct rx_queue vnode_list; /**< linked list of cached vnodes for this volume */
687 struct rx_queue rx_call_list; /**< linked list of split RX calls using this
688 * volume (fileserver only) */
689 #ifdef AFS_DEMAND_ATTACH_FS
690 VolState attach_state; /* what stage of attachment has been completed */
691 afs_uint32 attach_flags; /* flags related to attachment state */
692 pthread_cond_t attach_cv; /* state change condition variable */
693 short nWaiters; /* volume package internal ref count */
694 int chainCacheCheck; /* Volume hash chain cache check */
695 struct rx_queue vol_list; /* per-partition volume list (VByPList) */
697 VolumeOnlineSalvage salvage; /* online salvager state */
698 VolumeStats stats; /* per-volume statistics */
699 VolumeVLRUState vlru; /* state specific to the VLRU */
700 FSSYNC_VolOp_info * pending_vol_op; /* fssync command info for any pending vol ops */
701 #endif /* AFS_DEMAND_ATTACH_FS */
702 int usage_bumps_outstanding; /**< to rate limit the usage update i/o by accesses */
703 int usage_bumps_next_write; /**< to rate limit the usage update i/o by time */
708 VolumeDiskData diskstuff; /* General volume info read from disk */
709 Volume *back; /* back pointer to current volume structure */
712 /* These macros are used to export fields within the volume header. This was added
713 to facilitate changing the actual representation */
715 #define V_device(vp) ((vp)->device)
716 #define V_partition(vp) ((vp)->partition)
717 #define V_diskDataHandle(vp) ((vp)->diskDataHandle)
718 #define V_vnodeIndex(vp) ((vp)->vnodeIndex)
719 #define V_nextVnodeUnique(vp) ((vp)->nextVnodeUnique)
720 #define V_linkHandle(vp) ((vp)->linkHandle)
721 #define V_checkoutMode(vp) ((vp)->checkoutMode)
722 #ifdef AFS_DEMAND_ATTACH_FS
723 #define V_attachState(vp) ((vp)->attach_state)
724 #define V_attachFlags(vp) ((vp)->attach_flags)
725 #define V_attachCV(vp) ((vp)->attach_cv)
726 #endif /* AFS_DEMAND_ATTACH_FS */
728 /* N.B. V_id must be this, rather than vp->id, or some programs will break, probably */
729 #define V_stamp(vp) ((vp)->header->diskstuff.stamp)
730 #define V_id(vp) ((vp)->header->diskstuff.id)
731 #define V_name(vp) ((vp)->header->diskstuff.name)
732 #define V_inUse(vp) ((vp)->header->diskstuff.inUse)
733 #define V_inService(vp) ((vp)->header->diskstuff.inService)
734 #define V_blessed(vp) ((vp)->header->diskstuff.blessed)
735 #define V_needsSalvaged(vp) ((vp)->header->diskstuff.needsSalvaged)
736 #define V_uniquifier(vp) ((vp)->header->diskstuff.uniquifier)
737 #define V_type(vp) ((vp)->header->diskstuff.type)
738 #define V_parentId(vp) ((vp)->header->diskstuff.parentId)
739 #define V_cloneId(vp) ((vp)->header->diskstuff.cloneId)
740 #define V_backupId(vp) ((vp)->header->diskstuff.backupId)
741 #define V_restoredFromId(vp) ((vp)->header->diskstuff.restoredFromId)
742 #define V_needsCallback(vp) ((vp)->header->diskstuff.needsCallback)
743 #define V_destroyMe(vp) ((vp)->header->diskstuff.destroyMe)
744 #define V_dontSalvage(vp) ((vp)->header->diskstuff.dontSalvage)
745 #define V_maxquota(vp) ((vp)->header->diskstuff.maxquota)
746 #define V_minquota(vp) ((vp)->header->diskstuff.minquota)
747 #define V_maxfiles(vp) ((vp)->header->diskstuff.maxfiles)
748 #define V_accountNumber(vp) ((vp)->header->diskstuff.accountNumber)
749 #define V_owner(vp) ((vp)->header->diskstuff.owner)
750 #define V_filecount(vp) ((vp)->header->diskstuff.filecount)
751 #define V_diskused(vp) ((vp)->header->diskstuff.diskused)
752 #define V_dayUse(vp) ((vp)->header->diskstuff.dayUse)
753 #define V_weekUse(vp) ((vp)->header->diskstuff.weekUse)
754 #define V_dayUseDate(vp) ((vp)->header->diskstuff.dayUseDate)
755 #define V_creationDate(vp) ((vp)->header->diskstuff.creationDate)
756 #define V_accessDate(vp) ((vp)->header->diskstuff.accessDate)
757 #define V_updateDate(vp) ((vp)->header->diskstuff.updateDate)
758 #define V_expirationDate(vp) ((vp)->header->diskstuff.expirationDate)
759 #define V_backupDate(vp) ((vp)->header->diskstuff.backupDate)
760 #define V_copyDate(vp) ((vp)->header->diskstuff.copyDate)
761 #define V_offlineMessage(vp) ((vp)->header->diskstuff.offlineMessage)
762 #define V_disk(vp) ((vp)->header->diskstuff)
763 #define V_motd(vp) ((vp)->header->diskstuff.motd)
764 #define V_stat_initialized(vp) ((vp)->header->diskstuff.stat_initialized)
765 #define V_stat_area(vp) (((vp)->header->diskstuff.stat_reads))
766 #define V_stat_reads(vp, idx) (((vp)->header->diskstuff.stat_reads)[idx])
767 #define V_stat_writes(vp, idx) (((vp)->header->diskstuff.stat_writes)[idx])
768 #define V_stat_fileSameAuthor(vp, idx) (((vp)->header->diskstuff.stat_fileSameAuthor)[idx])
769 #define V_stat_fileDiffAuthor(vp, idx) (((vp)->header->diskstuff.stat_fileDiffAuthor)[idx])
770 #define V_stat_dirSameAuthor(vp, idx) (((vp)->header->diskstuff.stat_dirSameAuthor)[idx])
771 #define V_stat_dirDiffAuthor(vp, idx) (((vp)->header->diskstuff.stat_dirDiffAuthor)[idx])
772 #define V_volUpdateCounter(vp) ((vp)->header->diskstuff.volUpdateCounter)
774 /* File offset computations. The offset values in the volume header are
775 computed with these macros -- when the file is written only!! */
776 #define VOLUME_MOUNT_TABLE_OFFSET(Volume) (sizeof (VolumeDiskData))
777 #define VOLUME_BITMAP_OFFSET(Volume) \
778 (sizeof (VolumeDiskData) + (Volume)->disk.mountTableSize)
781 extern char *VSalvageMessage; /* Canonical message when a volume is forced
783 extern Volume *VGetVolume(Error * ec, Error * client_ec, VolumeId volumeId);
784 extern Volume *VGetVolumeWithCall(Error * ec, Error * client_ec, VolumeId volumeId,
785 const struct timespec *ts, struct VCallByVol *cbv);
786 extern Volume *VGetVolume_r(Error * ec, VolumeId volumeId);
787 extern void VPutVolume(Volume *);
788 extern void VPutVolumeWithCall(Volume *vp, struct VCallByVol *cbv);
789 extern void VPutVolume_r(Volume *);
790 extern void VOffline(Volume * vp, char *message);
791 extern void VOffline_r(Volume * vp, char *message);
792 extern int VConnectFS(void);
793 extern int VConnectFS_r(void);
794 extern void VDisconnectFS(void);
795 extern void VDisconnectFS_r(void);
796 extern int VChildProcReconnectFS(void);
797 extern Volume *VAttachVolume(Error * ec, VolumeId volumeId, int mode);
798 extern Volume *VAttachVolume_r(Error * ec, VolumeId volumeId, int mode);
799 extern Volume *VCreateVolume(Error * ec, char *partname, VolumeId volumeId,
801 extern Volume *VCreateVolume_r(Error * ec, char *partname, VolumeId volumeId,
803 extern void VGrowBitmap(struct vnodeIndex *index);
804 extern int VAllocBitmapEntry(Error * ec, Volume * vp,
805 struct vnodeIndex *index);
806 extern int VAllocBitmapEntry_r(Error * ec, Volume * vp,
807 struct vnodeIndex *index, int flags);
808 extern void VFreeBitMapEntry(Error * ec, Volume *vp, struct vnodeIndex *index,
810 extern void VFreeBitMapEntry_r(Error * ec, Volume *vp, struct vnodeIndex *index,
811 unsigned bitNumber, int flags);
812 extern int VolumeNumber(char *name);
813 extern char *VolumeExternalName(VolumeId volumeId);
814 extern int VolumeExternalName_r(VolumeId volumeId, char *name, size_t len);
815 extern Volume *VAttachVolumeByName(Error * ec, char *partition, char *name,
817 extern Volume *VAttachVolumeByName_r(Error * ec, char *partition, char *name,
819 extern void VShutdown(void);
820 extern void VSetTranquil(void);
821 extern void VUpdateVolume(Error * ec, Volume * vp);
822 extern void VUpdateVolume_r(Error * ec, Volume * vp, int flags);
823 extern void VAddToVolumeUpdateList(Error * ec, Volume * vp);
824 extern void VAddToVolumeUpdateList_r(Error * ec, Volume * vp);
825 extern void VDetachVolume(Error * ec, Volume * vp);
826 extern void VDetachVolume_r(Error * ec, Volume * vp);
827 extern void VForceOffline(Volume * vp);
828 extern void VForceOffline_r(Volume * vp, int flags);
829 extern void VBumpVolumeUsage(Volume * vp);
830 extern void VBumpVolumeUsage_r(Volume * vp);
831 extern void VSetDiskUsage(void);
832 extern void VPrintCacheStats(void);
833 extern void VReleaseVnodeFiles_r(Volume * vp);
834 extern void VCloseVnodeFiles_r(Volume * vp);
835 extern struct DiskPartition64 *VGetPartition(char *name, int abortp);
836 extern struct DiskPartition64 *VGetPartition_r(char *name, int abortp);
837 extern void VOptDefaults(ProgramType pt, VolumePackageOptions * opts);
838 extern int VInitVolumePackage2(ProgramType pt, VolumePackageOptions * opts);
839 extern int VInitAttachVolumes(ProgramType pt);
840 extern void DiskToVolumeHeader(VolumeHeader_t * h, VolumeDiskHeader_t * dh);
841 extern void VolumeHeaderToDisk(VolumeDiskHeader_t * dh, VolumeHeader_t * h);
842 extern void AssignVolumeName(VolumeDiskData * vol, char *name, char *ext);
843 extern void VTakeOffline_r(Volume * vp);
844 extern void VTakeOffline(Volume * vp);
845 extern Volume * VLookupVolume_r(Error * ec, VolumeId volumeId, Volume * hint);
846 extern void VGetVolumePath(Error * ec, VolumeId volumeId, char **partitionp,
848 extern char *vol_DevName(dev_t adev, char *wpath);
849 extern afs_int32 VIsGoingOffline(struct Volume *vp);
852 extern void VLockFileInit(struct VLockFile *lf, const char *path);
853 extern void VLockFileReinit(struct VLockFile *lf);
854 extern int VLockFileLock(struct VLockFile *lf, afs_uint32 offset,
855 int locktype, int nonblock);
856 extern void VLockFileUnlock(struct VLockFile *lf, afs_uint32 offset);
858 extern int VSetVolHashSize(int logsize);
860 #ifdef AFS_DEMAND_ATTACH_FS
861 extern Volume *VPreAttachVolumeByName(Error * ec, char *partition, char *name);
862 extern Volume *VPreAttachVolumeByName_r(Error * ec, char *partition, char *name);
863 extern Volume *VPreAttachVolumeById_r(Error * ec, char * partition,
865 extern Volume *VPreAttachVolumeByVp_r(Error * ec, struct DiskPartition64 * partp,
866 Volume * vp, VolumeId volume_id);
867 extern Volume *VGetVolumeByVp_r(Error * ec, Volume * vp);
868 extern int VShutdownByPartition_r(struct DiskPartition64 * dp);
869 extern int VShutdownVolume_r(Volume * vp);
870 extern int VConnectSALV(void);
871 extern int VConnectSALV_r(void);
872 extern int VReconnectSALV(void);
873 extern int VReconnectSALV_r(void);
874 extern int VDisconnectSALV(void);
875 extern int VDisconnectSALV_r(void);
876 extern void VPrintExtendedCacheStats(int flags);
877 extern void VPrintExtendedCacheStats_r(int flags);
878 extern void VLRU_SetOptions(int option, afs_uint32 val);
879 extern int VRequestSalvage_r(Error * ec, Volume * vp, int reason, int flags);
880 extern int VUpdateSalvagePriority_r(Volume * vp);
881 extern int VRegisterVolOp_r(Volume * vp, FSSYNC_VolOp_info * vopinfo);
882 extern int VDeregisterVolOp_r(Volume * vp);
883 extern void VCancelReservation_r(Volume * vp);
884 extern int VChildProcReconnectFS_r(void);
885 extern void VOfflineForVolOp_r(Error *ec, Volume *vp, char *message);
886 #endif /* AFS_DEMAND_ATTACH_FS */
888 #ifdef AFS_DEMAND_ATTACH_FS
890 extern void VDiskLockInit(struct VDiskLock *dl, struct VLockFile *lf,
892 extern int VGetDiskLock(struct VDiskLock *dl, int locktype, int nonblock);
893 extern void VReleaseDiskLock(struct VDiskLock *dl, int locktype);
894 #endif /* AFS_DEMAND_ATTACH_FS */
895 extern int VVolOpLeaveOnline_r(Volume * vp, FSSYNC_VolOp_info * vopinfo);
896 extern int VVolOpLeaveOnlineNoHeader_r(Volume * vp, FSSYNC_VolOp_info * vopinfo);
897 extern int VVolOpSetVBusy_r(Volume * vp, FSSYNC_VolOp_info * vopinfo);
899 extern void VPurgeVolume(Error * ec, Volume * vp);
901 extern afs_int32 VCanScheduleSalvage(void);
902 extern afs_int32 VCanUseFSSYNC(void);
903 extern afs_int32 VCanUseSALVSYNC(void);
904 extern afs_int32 VCanUnsafeAttach(void);
905 extern afs_int32 VReadVolumeDiskHeader(VolumeId volid,
906 struct DiskPartition64 * dp,
907 VolumeDiskHeader_t * hdr);
908 extern afs_int32 VWriteVolumeDiskHeader(VolumeDiskHeader_t * hdr,
909 struct DiskPartition64 * dp);
910 extern afs_int32 VCreateVolumeDiskHeader(VolumeDiskHeader_t * hdr,
911 struct DiskPartition64 * dp);
912 extern afs_int32 VDestroyVolumeDiskHeader(struct DiskPartition64 * dp,
913 VolumeId volid, VolumeId parent);
916 * VWalkVolumeHeaders header callback.
918 * @param[in] dp disk partition
919 * @param[in] name full path to the .vol header file
920 * @param[in] hdr the header data that was read from the .vol header
921 * @param[in] last 1 if this is the last attempt to read the vol header, 0
922 * otherwise. DAFS VWalkVolumeHeaders will retry reading the
923 * header once, if a non-fatal error occurs when reading the
924 * header, or if this function returns a positive error code.
925 * So, if there is a problem, this function will be called
926 * first with last=0, then with last=1, then the error function
927 * callback will be called. For non-DAFS, this is always 1.
928 * @param[in] rock the rock passed to VWalkVolumeHeaders
930 * @return operation status
932 * @retval negative a fatal error that should stop the walk immediately
933 * @retval positive an error with the volume header was encountered; the walk
934 * should continue, but the error function should be called on this
937 * @see VWalkVolumeHeaders
939 typedef int (*VWalkVolFunc)(struct DiskPartition64 *dp, const char *name,
940 struct VolumeDiskHeader *hdr, int last,
943 * VWalkVolumeHeaders error callback.
945 * This is called from VWalkVolumeHeaders when an invalid or otherwise
946 * problematic volume header is encountered. It is typically implemented as a
947 * wrapper to unlink the .vol file.
949 * @param[in] dp disk partition
950 * @param[in] name full path to the .vol header file
951 * @param[in] hdr header read in from the .vol file, or NULL if it could not
953 * @param[in] rock rock passed to VWalkVolumeHeaders
955 * @see VWalkVolumeHeaders
957 typedef void (*VWalkErrFunc)(struct DiskPartition64 *dp, const char *name,
958 struct VolumeDiskHeader *hdr, void *rock);
959 extern int VWalkVolumeHeaders(struct DiskPartition64 *dp, const char *partpath,
960 VWalkVolFunc volfunc, VWalkErrFunc errfunc,
963 /* Naive formula relating number of file size to number of 1K blocks in file */
964 /* Note: we charge 1 block for 0 length files so the user can't store
965 an inifite number of them; for most files, we give him the inode, vnode,
966 and indirect block overhead, for FREE! */
967 #define nBlocks(bytes) ((afs_sfsize_t)((bytes) == 0? 1: (((afs_sfsize_t)(bytes))+1023)/1024))
969 /* Client process id -- file server sends a Check volumes signal back to the client at this pid */
970 #define CLIENTPID "/vice/vol/clientpid"
972 /* Modes of attachment, for VAttachVolume[ByName] to convey to the file server */
973 #define V_READONLY 1 /* Absolutely no updates will be done to the volume */
974 #define V_CLONE 2 /* Cloning the volume: if it is read/write, then directory
975 * version numbers will change. Header will be updated. If
976 * the volume is read-only, the file server may continue to
977 * server it; it may also continue to server it in read/write
978 * mode if the writes are deferred */
979 #define V_VOLUPD 3 /* General update or volume purge is possible. Volume must
981 #define V_DUMP 4 /* A dump of the volume is requested; the volume can be served
982 * read-only during this time */
983 #define V_SECRETLY 5 /* Secret attach of the volume. This is used to attach a volume
984 * which the file server doesn't know about--and which it shouldn't
985 * know about yet, since the volume has just been created and
986 * is somewhat bogus. Required to make sure that a file server
987 * never knows about more than one copy of the same volume--when
988 * a volume is moved from one partition to another on a single
990 #define V_PEEK 6 /* "Peek" at the volume without telling the fileserver. This is
991 * similar to V_SECRETLY, but read-only. It is used in cases where
992 * not impacting fileserver performance is more important than
993 * getting the most recent data. */
997 /* VUpdateVolume_r flags */
998 #define VOL_UPDATE_WAIT 0x1 /* for demand attach, wait for other exclusive ops to end */
999 #define VOL_UPDATE_NOFORCEOFF 0x2 /* don't force offline on failure. this is to prevent
1000 * infinite recursion between vupdate and vforceoff */
1002 /* VForceOffline_r flags */
1003 #define VOL_FORCEOFF_NOUPDATE 0x1 /* don't force update on forceoff. this is to prevent
1004 * infinite recursion between vupdate and vforceoff */
1006 /* VSyncVolume_r flags */
1007 #define VOL_SYNC_WAIT 0x1 /* for demand attach, wait for other exclusive ops to end */
1009 /* VAllocBitmapEntry_r flags */
1010 #define VOL_ALLOC_BITMAP_WAIT 0x1 /* for demand attach, wait for other exclusive ops to end */
1012 /* VFreeBitMapEntry_r flags */
1013 #define VOL_FREE_BITMAP_WAIT 0x1 /* for demand attach, wait for other exclusive ops to end */
1015 /* VRequestSalvage_r flags */
1016 #define VOL_SALVAGE_NO_OFFLINE 0x1 /* we do not need to wait to offline the volume; it has
1017 * not been fully attached */
1019 #define VOLUME_BITMAP_GROWSIZE 16 /* bytes, => 128vnodes */
1020 /* Must be a multiple of 4 (1 word) !! */
1022 #if defined(NEARINODE_HINT)
1023 #define V_pref(vp,nearInode) nearInodeHash(V_id(vp),(nearInode)); (nearInode) %= V_partition(vp)->f_files
1025 #define V_pref(vp,nearInode) nearInode = 0
1026 #endif /* NEARINODE_HINT */
1028 hdr_static_inline(unsigned long)
1029 afs_printable_VolumeId_lu(VolumeId d) { return (unsigned long) (d); }
1031 hdr_static_inline(unsigned int)
1032 afs_printable_VnodeId_u(VnodeId d) { return (unsigned int) d; }
1034 #endif /* __volume_h */