2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for
3 * licensing and copyright details
6 #include <linux/reiserfs_fs.h>
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/sched.h>
11 #include <linux/bug.h>
12 #include <linux/workqueue.h>
13 #include <asm/unaligned.h>
14 #include <linux/bitops.h>
15 #include <linux/proc_fs.h>
16 #include <linux/buffer_head.h>
18 /* the 32 bit compat definitions with int argument */
19 #define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
20 #define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
21 #define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
22 #define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
23 #define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
25 struct reiserfs_journal_list;
27 /* bitmasks for i_flags field in reiserfs-specific part of inode */
30 * this says what format of key do all items (but stat data) of
31 * an object have. If this is set, that format is 3.6 otherwise - 3.5
33 i_item_key_version_mask = 0x0001,
36 * If this is unset, object has 3.5 stat data, otherwise,
37 * it has 3.6 stat data with 64bit size, 32bit nlink etc.
39 i_stat_data_version_mask = 0x0002,
41 /* file might need tail packing on close */
42 i_pack_on_close_mask = 0x0004,
44 /* don't pack tail of file */
45 i_nopack_mask = 0x0008,
48 * If either of these are set, "safe link" was created for this
49 * file during truncate or unlink. Safe link is used to avoid
50 * leakage of disk space on crash with some files open, but unlinked.
52 i_link_saved_unlink_mask = 0x0010,
53 i_link_saved_truncate_mask = 0x0020,
55 i_has_xattr_dir = 0x0040,
57 } reiserfs_inode_flags;
59 struct reiserfs_inode_info {
60 __u32 i_key[4]; /* key is still 4 32 bit integers */
63 * transient inode flags that are never stored on disk. Bitmasks
64 * for this field are defined above.
68 /* offset of first byte stored in direct item. */
69 __u32 i_first_direct_byte;
71 /* copy of persistent inode flags read from sd_attrs. */
74 /* first unused block of a sequence of unused blocks */
76 int i_prealloc_count; /* length of that sequence */
78 /* per-transaction list of inodes which have preallocated blocks */
79 struct list_head i_prealloc_list;
82 * new_packing_locality is created; new blocks for the contents
83 * of this directory should be displaced
85 unsigned new_packing_locality:1;
88 * we use these for fsync or O_SYNC to decide which transaction
89 * needs to be committed in order for this inode to be properly
92 unsigned int i_trans_id;
94 struct reiserfs_journal_list *i_jl;
96 struct mutex tailpack;
97 #ifdef CONFIG_REISERFS_FS_XATTR
98 struct rw_semaphore i_xattr_sem;
101 struct dquot *i_dquot[MAXQUOTAS];
104 struct inode vfs_inode;
108 reiserfs_attrs_cleared = 0x00000001,
109 } reiserfs_super_block_flags;
112 * struct reiserfs_super_block accessors/mutators since this is a disk
113 * structure, it will always be in little endian format.
115 #define sb_block_count(sbp) (le32_to_cpu((sbp)->s_v1.s_block_count))
116 #define set_sb_block_count(sbp,v) ((sbp)->s_v1.s_block_count = cpu_to_le32(v))
117 #define sb_free_blocks(sbp) (le32_to_cpu((sbp)->s_v1.s_free_blocks))
118 #define set_sb_free_blocks(sbp,v) ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))
119 #define sb_root_block(sbp) (le32_to_cpu((sbp)->s_v1.s_root_block))
120 #define set_sb_root_block(sbp,v) ((sbp)->s_v1.s_root_block = cpu_to_le32(v))
122 #define sb_jp_journal_1st_block(sbp) \
123 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))
124 #define set_sb_jp_journal_1st_block(sbp,v) \
125 ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))
126 #define sb_jp_journal_dev(sbp) \
127 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))
128 #define set_sb_jp_journal_dev(sbp,v) \
129 ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))
130 #define sb_jp_journal_size(sbp) \
131 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))
132 #define set_sb_jp_journal_size(sbp,v) \
133 ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))
134 #define sb_jp_journal_trans_max(sbp) \
135 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))
136 #define set_sb_jp_journal_trans_max(sbp,v) \
137 ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))
138 #define sb_jp_journal_magic(sbp) \
139 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))
140 #define set_sb_jp_journal_magic(sbp,v) \
141 ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))
142 #define sb_jp_journal_max_batch(sbp) \
143 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))
144 #define set_sb_jp_journal_max_batch(sbp,v) \
145 ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))
146 #define sb_jp_jourmal_max_commit_age(sbp) \
147 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))
148 #define set_sb_jp_journal_max_commit_age(sbp,v) \
149 ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))
151 #define sb_blocksize(sbp) (le16_to_cpu((sbp)->s_v1.s_blocksize))
152 #define set_sb_blocksize(sbp,v) ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))
153 #define sb_oid_maxsize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))
154 #define set_sb_oid_maxsize(sbp,v) ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))
155 #define sb_oid_cursize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_cursize))
156 #define set_sb_oid_cursize(sbp,v) ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))
157 #define sb_umount_state(sbp) (le16_to_cpu((sbp)->s_v1.s_umount_state))
158 #define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))
159 #define sb_fs_state(sbp) (le16_to_cpu((sbp)->s_v1.s_fs_state))
160 #define set_sb_fs_state(sbp,v) ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))
161 #define sb_hash_function_code(sbp) \
162 (le32_to_cpu((sbp)->s_v1.s_hash_function_code))
163 #define set_sb_hash_function_code(sbp,v) \
164 ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))
165 #define sb_tree_height(sbp) (le16_to_cpu((sbp)->s_v1.s_tree_height))
166 #define set_sb_tree_height(sbp,v) ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))
167 #define sb_bmap_nr(sbp) (le16_to_cpu((sbp)->s_v1.s_bmap_nr))
168 #define set_sb_bmap_nr(sbp,v) ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))
169 #define sb_version(sbp) (le16_to_cpu((sbp)->s_v1.s_version))
170 #define set_sb_version(sbp,v) ((sbp)->s_v1.s_version = cpu_to_le16(v))
172 #define sb_mnt_count(sbp) (le16_to_cpu((sbp)->s_mnt_count))
173 #define set_sb_mnt_count(sbp, v) ((sbp)->s_mnt_count = cpu_to_le16(v))
175 #define sb_reserved_for_journal(sbp) \
176 (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))
177 #define set_sb_reserved_for_journal(sbp,v) \
178 ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))
183 * These all interelate for performance.
185 * If the journal block count is smaller than n transactions, you lose speed.
186 * I don't know what n is yet, I'm guessing 8-16.
188 * typical transaction size depends on the application, how often fsync is
189 * called, and how many metadata blocks you dirty in a 30 second period.
190 * The more small files (<16k) you use, the larger your transactions will
193 * If your journal fills faster than dirty buffers get flushed to disk, it
194 * must flush them before allowing the journal to wrap, which slows things
195 * down. If you need high speed meta data updates, the journal should be
196 * big enough to prevent wrapping before dirty meta blocks get to disk.
198 * If the batch max is smaller than the transaction max, you'll waste space
199 * at the end of the journal because journal_end sets the next transaction
200 * to start at 0 if the next transaction has any chance of wrapping.
202 * The large the batch max age, the better the speed, and the more meta
203 * data changes you'll lose after a crash.
206 /* don't mess with these for a while */
207 /* we have a node size define somewhere in reiserfs_fs.h. -Hans */
208 #define JOURNAL_BLOCK_SIZE 4096 /* BUG gotta get rid of this */
209 #define JOURNAL_MAX_CNODE 1500 /* max cnodes to allocate. */
210 #define JOURNAL_HASH_SIZE 8192
212 /* number of copies of the bitmaps to have floating. Must be >= 2 */
213 #define JOURNAL_NUM_BITMAPS 5
216 * One of these for every block in every transaction
217 * Each one is in two hash tables. First, a hash of the current transaction,
218 * and after journal_end, a hash of all the in memory transactions.
219 * next and prev are used by the current transaction (journal_hash).
220 * hnext and hprev are used by journal_list_hash. If a block is in more
221 * than one transaction, the journal_list_hash links it in multiple times.
222 * This allows flush_journal_list to remove just the cnode belonging to a
225 struct reiserfs_journal_cnode {
226 struct buffer_head *bh; /* real buffer head */
227 struct super_block *sb; /* dev of real buffer head */
229 /* block number of real buffer head, == 0 when buffer on disk */
234 /* journal list this cnode lives in */
235 struct reiserfs_journal_list *jlist;
237 struct reiserfs_journal_cnode *next; /* next in transaction list */
238 struct reiserfs_journal_cnode *prev; /* prev in transaction list */
239 struct reiserfs_journal_cnode *hprev; /* prev in hash list */
240 struct reiserfs_journal_cnode *hnext; /* next in hash list */
243 struct reiserfs_bitmap_node {
246 struct list_head list;
249 struct reiserfs_list_bitmap {
250 struct reiserfs_journal_list *journal_list;
251 struct reiserfs_bitmap_node **bitmaps;
255 * one of these for each transaction. The most important part here is the
256 * j_realblock. this list of cnodes is used to hash all the blocks in all
257 * the commits, to mark all the real buffer heads dirty once all the commits
258 * hit the disk, and to make sure every real block in a transaction is on
259 * disk before allowing the log area to be overwritten
261 struct reiserfs_journal_list {
262 unsigned long j_start;
263 unsigned long j_state;
265 atomic_t j_nonzerolen;
266 atomic_t j_commit_left;
268 /* all commits older than this on disk */
269 atomic_t j_older_commits_done;
271 struct mutex j_commit_mutex;
272 unsigned int j_trans_id;
274 struct reiserfs_list_bitmap *j_list_bitmap;
275 struct buffer_head *j_commit_bh; /* commit buffer head */
276 struct reiserfs_journal_cnode *j_realblock;
277 struct reiserfs_journal_cnode *j_freedlist; /* list of buffers that were freed during this trans. free each of these on flush */
278 /* time ordered list of all active transactions */
279 struct list_head j_list;
282 * time ordered list of all transactions we haven't tried
285 struct list_head j_working_list;
287 /* list of tail conversion targets in need of flush before commit */
288 struct list_head j_tail_bh_list;
290 /* list of data=ordered buffers in need of flush before commit */
291 struct list_head j_bh_list;
295 struct reiserfs_journal {
296 struct buffer_head **j_ap_blocks; /* journal blocks on disk */
297 /* newest journal block */
298 struct reiserfs_journal_cnode *j_last;
300 /* oldest journal block. start here for traverse */
301 struct reiserfs_journal_cnode *j_first;
303 struct block_device *j_dev_bd;
306 /* first block on s_dev of reserved area journal */
307 int j_1st_reserved_block;
309 unsigned long j_state;
310 unsigned int j_trans_id;
311 unsigned long j_mount_id;
313 /* start of current waiting commit (index into j_ap_blocks) */
314 unsigned long j_start;
315 unsigned long j_len; /* length of current waiting commit */
317 /* number of buffers requested by journal_begin() */
318 unsigned long j_len_alloc;
320 atomic_t j_wcount; /* count of writers for current commit */
322 /* batch count. allows turning X transactions into 1 */
323 unsigned long j_bcount;
325 /* first unflushed transactions offset */
326 unsigned long j_first_unflushed_offset;
328 /* last fully flushed journal timestamp */
329 unsigned j_last_flush_trans_id;
331 struct buffer_head *j_header_bh;
333 time_t j_trans_start_time; /* time this transaction started */
334 struct mutex j_mutex;
335 struct mutex j_flush_mutex;
337 /* wait for current transaction to finish before starting new one */
338 wait_queue_head_t j_join_wait;
340 atomic_t j_jlock; /* lock for j_join_wait */
341 int j_list_bitmap_index; /* number of next list bitmap to use */
343 /* no more journal begins allowed. MUST sleep on j_join_wait */
346 /* next journal_end will flush all journal list */
347 int j_next_full_flush;
349 /* next journal_end will flush all async commits */
350 int j_next_async_flush;
352 int j_cnode_used; /* number of cnodes on the used list */
353 int j_cnode_free; /* number of cnodes on the free list */
355 /* max number of blocks in a transaction. */
356 unsigned int j_trans_max;
358 /* max number of blocks to batch into a trans */
359 unsigned int j_max_batch;
361 /* in seconds, how old can an async commit be */
362 unsigned int j_max_commit_age;
364 /* in seconds, how old can a transaction be */
365 unsigned int j_max_trans_age;
367 /* the default for the max commit age */
368 unsigned int j_default_max_commit_age;
370 struct reiserfs_journal_cnode *j_cnode_free_list;
372 /* orig pointer returned from vmalloc */
373 struct reiserfs_journal_cnode *j_cnode_free_orig;
375 struct reiserfs_journal_list *j_current_jl;
376 int j_free_bitmap_nodes;
377 int j_used_bitmap_nodes;
379 int j_num_lists; /* total number of active transactions */
380 int j_num_work_lists; /* number that need attention from kreiserfsd */
382 /* debugging to make sure things are flushed in order */
383 unsigned int j_last_flush_id;
385 /* debugging to make sure things are committed in order */
386 unsigned int j_last_commit_id;
388 struct list_head j_bitmap_nodes;
389 struct list_head j_dirty_buffers;
390 spinlock_t j_dirty_buffers_lock; /* protects j_dirty_buffers */
392 /* list of all active transactions */
393 struct list_head j_journal_list;
395 /* lists that haven't been touched by writeback attempts */
396 struct list_head j_working_list;
398 /* hash table for real buffer heads in current trans */
399 struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE];
401 /* hash table for all the real buffer heads in all the transactions */
402 struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE];
404 /* array of bitmaps to record the deleted blocks */
405 struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS];
407 /* list of inodes which have preallocated blocks */
408 struct list_head j_prealloc_list;
409 int j_persistent_trans;
410 unsigned long j_max_trans_size;
411 unsigned long j_max_batch_size;
415 /* when flushing ordered buffers, throttle new ordered writers */
416 struct delayed_work j_work;
417 struct super_block *j_work_sb;
418 atomic_t j_async_throttle;
421 enum journal_state_bits {
422 J_WRITERS_BLOCKED = 1, /* set when new writers not allowed */
423 J_WRITERS_QUEUED, /* set when log is full due to too many writers */
424 J_ABORTED, /* set when log is aborted */
427 /* ick. magic string to find desc blocks in the journal */
428 #define JOURNAL_DESC_MAGIC "ReIsErLB"
430 typedef __u32(*hashf_t) (const signed char *, int);
432 struct reiserfs_bitmap_info {
436 struct proc_dir_entry;
438 #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
439 typedef unsigned long int stat_cnt_t;
440 typedef struct reiserfs_proc_info_data {
443 int max_hash_collisions;
446 stat_cnt_t bread_miss;
447 stat_cnt_t search_by_key;
448 stat_cnt_t search_by_key_fs_changed;
449 stat_cnt_t search_by_key_restarted;
451 stat_cnt_t insert_item_restarted;
452 stat_cnt_t paste_into_item_restarted;
453 stat_cnt_t cut_from_item_restarted;
454 stat_cnt_t delete_solid_item_restarted;
455 stat_cnt_t delete_item_restarted;
457 stat_cnt_t leaked_oid;
458 stat_cnt_t leaves_removable;
461 * balances per level.
462 * Use explicit 5 as MAX_HEIGHT is not visible yet.
464 stat_cnt_t balance_at[5]; /* XXX */
465 /* sbk == search_by_key */
466 stat_cnt_t sbk_read_at[5]; /* XXX */
467 stat_cnt_t sbk_fs_changed[5];
468 stat_cnt_t sbk_restarted[5];
469 stat_cnt_t items_at[5]; /* XXX */
470 stat_cnt_t free_at[5]; /* XXX */
471 stat_cnt_t can_node_be_removed[5]; /* XXX */
472 long int lnum[5]; /* XXX */
473 long int rnum[5]; /* XXX */
474 long int lbytes[5]; /* XXX */
475 long int rbytes[5]; /* XXX */
476 stat_cnt_t get_neighbors[5];
477 stat_cnt_t get_neighbors_restart[5];
478 stat_cnt_t need_l_neighbor[5];
479 stat_cnt_t need_r_neighbor[5];
481 stat_cnt_t free_block;
482 struct __scan_bitmap_stats {
487 stat_cnt_t in_journal_hint;
488 stat_cnt_t in_journal_nohint;
491 struct __journal_stats {
492 stat_cnt_t in_journal;
493 stat_cnt_t in_journal_bitmap;
494 stat_cnt_t in_journal_reusable;
495 stat_cnt_t lock_journal;
496 stat_cnt_t lock_journal_wait;
497 stat_cnt_t journal_being;
498 stat_cnt_t journal_relock_writers;
499 stat_cnt_t journal_relock_wcount;
500 stat_cnt_t mark_dirty;
501 stat_cnt_t mark_dirty_already;
502 stat_cnt_t mark_dirty_notjournal;
503 stat_cnt_t restore_prepared;
505 stat_cnt_t prepare_retry;
507 } reiserfs_proc_info_data_t;
509 typedef struct reiserfs_proc_info_data {
510 } reiserfs_proc_info_data_t;
513 /* Number of quota types we support */
514 #define REISERFS_MAXQUOTAS 2
516 /* reiserfs union of in-core super block data */
517 struct reiserfs_sb_info {
518 /* Buffer containing the super block */
519 struct buffer_head *s_sbh;
521 /* Pointer to the on-disk super block in the buffer */
522 struct reiserfs_super_block *s_rs;
523 struct reiserfs_bitmap_info *s_ap_bitmap;
525 /* pointer to journal information */
526 struct reiserfs_journal *s_journal;
528 unsigned short s_mount_state; /* reiserfs state (valid, invalid) */
530 /* Serialize writers access, replace the old bkl */
533 /* Owner of the lock (can be recursive) */
534 struct task_struct *lock_owner;
536 /* Depth of the lock, start from -1 like the bkl */
539 struct workqueue_struct *commit_wq;
542 void (*end_io_handler) (struct buffer_head *, int);
545 * pointer to function which is used to sort names in directory.
548 hashf_t s_hash_function;
550 /* reiserfs's mount options are set here */
551 unsigned long s_mount_opt;
553 /* This is a structure that describes block allocator options */
555 /* Bitfield for enable/disable kind of options */
559 * size started from which we consider file
560 * to be a large one (in blocks)
562 unsigned long large_file_size;
564 int border; /* percentage of disk, border takes */
567 * Minimal file size (in blocks) starting
568 * from which we do preallocations
573 * Number of blocks we try to prealloc when file
574 * reaches preallocmin size (in blocks) or prealloc_list
581 wait_queue_head_t s_wait;
582 /* increased by one every time the tree gets re-balanced */
583 atomic_t s_generation_counter;
585 /* File system properties. Currently holds on-disk FS format */
586 unsigned long s_properties;
588 /* session statistics */
593 int s_unneeded_left_neighbor;
594 int s_good_search_by_key_reada;
596 int s_bmaps_without_search;
597 int s_direct2indirect;
598 int s_indirect2direct;
601 * set up when it's ok for reiserfs_read_inode2() to read from
602 * disk inode with nlink==0. Currently this is only used during
603 * finish_unfinished() processing at mount time
605 int s_is_unlinked_ok;
607 reiserfs_proc_info_data_t s_proc_info_data;
608 struct proc_dir_entry *procdir;
610 /* amount of blocks reserved for further allocations */
614 /* this lock on now only used to protect reserved_blocks variable */
615 spinlock_t bitmap_lock;
616 struct dentry *priv_root; /* root of /.reiserfs_priv */
617 struct dentry *xattr_root; /* root of /.reiserfs_priv/xattrs */
620 int work_queued; /* non-zero delayed work is queued */
621 struct delayed_work old_work; /* old transactions flush delayed work */
622 spinlock_t old_work_lock; /* protects old_work and work_queued */
625 char *s_qf_names[REISERFS_MAXQUOTAS];
628 char *s_jdev; /* Stored jdev for mount option showing */
629 #ifdef CONFIG_REISERFS_CHECK
632 * Detects whether more than one copy of tb exists per superblock
633 * as a means of checking whether do_balance is executing
634 * concurrently against another tree reader/writer on a same
637 struct tree_balance *cur_tb;
641 /* Definitions of reiserfs on-disk properties: */
642 #define REISERFS_3_5 0
643 #define REISERFS_3_6 1
644 #define REISERFS_OLD_FORMAT 2
647 enum reiserfs_mount_options {
648 /* large tails will be created in a session */
651 * small (for files less than block size) tails will
652 * be created in a session
656 /* replay journal and return 0. Use by fsck */
660 * -o conv: causes conversion of old format super block to the
661 * new format. If not specified - old partition will be dealt
662 * with in a manner of 3.5.x
667 * -o hash={tea, rupasov, r5, detect} is meant for properly mounting
668 * reiserfs disks from 3.5.19 or earlier. 99% of the time, this
669 * option is not required. If the normal autodection code can't
670 * determine which hash to use (because both hashes had the same
671 * value for a file) use this option to force a specific hash.
672 * It won't allow you to override the existing hash on the FS, so
673 * if you have a tea hash disk, and mount with -o hash=rupasov,
674 * the mount will fail.
676 FORCE_TEA_HASH, /* try to force tea hash on mount */
677 FORCE_RUPASOV_HASH, /* try to force rupasov hash on mount */
678 FORCE_R5_HASH, /* try to force rupasov hash on mount */
679 FORCE_HASH_DETECT, /* try to detect hash function on mount */
682 REISERFS_DATA_ORDERED,
683 REISERFS_DATA_WRITEBACK,
686 * used for testing experimental features, makes benchmarking new
687 * features with and without more convenient, should never be used by
688 * users in any code shipped to users (ideally)
692 REISERFS_NO_UNHASHED_RELOCATION,
693 REISERFS_HASHED_RELOCATION,
695 REISERFS_XATTRS_USER,
697 REISERFS_EXPOSE_PRIVROOT,
698 REISERFS_BARRIER_NONE,
699 REISERFS_BARRIER_FLUSH,
701 /* Actions on error */
702 REISERFS_ERROR_PANIC,
704 REISERFS_ERROR_CONTINUE,
706 REISERFS_USRQUOTA, /* User quota option specified */
707 REISERFS_GRPQUOTA, /* Group quota option specified */
713 REISERFS_UNSUPPORTED_OPT,
716 #define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))
717 #define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))
718 #define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))
719 #define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))
720 #define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))
721 #define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))
722 #define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))
723 #define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))
725 #define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))
726 #define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))
727 #define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))
728 #define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))
729 #define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))
730 #define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))
731 #define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))
732 #define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))
733 #define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))
734 #define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))
735 #define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))
736 #define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))
737 #define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))
738 #define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))
739 #define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))
741 #define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))
742 #define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))
744 void reiserfs_file_buffer(struct buffer_head *bh, int list);
745 extern struct file_system_type reiserfs_fs_type;
746 int reiserfs_resize(struct super_block *, unsigned long);
749 #define SCHEDULE_OCCURRED 1
751 #define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)
752 #define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)
753 #define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)
754 #define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)
755 #define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)
757 #define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)
759 #define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
760 static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
763 return test_bit(J_ABORTED, &journal->j_state);
767 * Locking primitives. The write lock is a per superblock
768 * special mutex that has properties close to the Big Kernel Lock
769 * which was used in the previous locking scheme.
771 void reiserfs_write_lock(struct super_block *s);
772 void reiserfs_write_unlock(struct super_block *s);
773 int __must_check reiserfs_write_unlock_nested(struct super_block *s);
774 void reiserfs_write_lock_nested(struct super_block *s, int depth);
776 #ifdef CONFIG_REISERFS_CHECK
777 void reiserfs_lock_check_recursive(struct super_block *s);
779 static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
783 * Several mutexes depend on the write lock.
784 * However sometimes we want to relax the write lock while we hold
785 * these mutexes, according to the release/reacquire on schedule()
786 * properties of the Bkl that were used.
787 * Reiserfs performances and locking were based on this scheme.
788 * Now that the write lock is a mutex and not the bkl anymore, doing so
789 * may result in a deadlock:
791 * A acquire write_lock
792 * A acquire j_commit_mutex
793 * A release write_lock and wait for something
794 * B acquire write_lock
795 * B can't acquire j_commit_mutex and sleep
796 * A can't acquire write lock anymore
799 * What we do here is avoiding such deadlock by playing the same game
800 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
801 * we release the write lock, wait a bit and then retry.
803 * The mutexes concerned by this hack are:
804 * - The commit mutex of a journal list
809 static inline void reiserfs_mutex_lock_safe(struct mutex *m,
810 struct super_block *s)
814 depth = reiserfs_write_unlock_nested(s);
816 reiserfs_write_lock_nested(s, depth);
820 reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
821 struct super_block *s)
825 depth = reiserfs_write_unlock_nested(s);
826 mutex_lock_nested(m, subclass);
827 reiserfs_write_lock_nested(s, depth);
831 reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
834 depth = reiserfs_write_unlock_nested(s);
836 reiserfs_write_lock_nested(s, depth);
840 * When we schedule, we usually want to also release the write lock,
841 * according to the previous bkl based locking scheme of reiserfs.
843 static inline void reiserfs_cond_resched(struct super_block *s)
845 if (need_resched()) {
848 depth = reiserfs_write_unlock_nested(s);
850 reiserfs_write_lock_nested(s, depth);
857 * in reading the #defines, it may help to understand that they employ
858 * the following abbreviations:
862 * H = Height within the tree (should be changed to LEV)
863 * N = Number of the item in the node
865 * DEH = Directory Entry Header
869 * BLKH = BLocK Header
870 * UNFM = UNForMatted node
874 * These #defines are named by concatenating these abbreviations,
875 * where first comes the arguments, and last comes the return value,
879 #define USE_INODE_GENERATION_COUNTER
881 #define REISERFS_PREALLOCATE
882 #define DISPLACE_NEW_PACKING_LOCALITIES
883 #define PREALLOCATION_SIZE 9
885 /* n must be power of 2 */
886 #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
889 * to be ok for alpha and others we have to align structures to 8 byte
891 * FIXME: do not change 4 by anything else: there is code which relies on that
893 #define ROUND_UP(x) _ROUND_UP(x,8LL)
896 * debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
899 #define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
901 void __reiserfs_warning(struct super_block *s, const char *id,
902 const char *func, const char *fmt, ...);
903 #define reiserfs_warning(s, id, fmt, args...) \
904 __reiserfs_warning(s, id, __func__, fmt, ##args)
905 /* assertions handling */
907 /* always check a condition and panic if it's false. */
908 #define __RASSERT(cond, scond, format, args...) \
911 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
912 __FILE__ ":%i:%s: " format "\n", \
913 in_interrupt() ? -1 : task_pid_nr(current), \
914 __LINE__, __func__ , ##args); \
917 #define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
919 #if defined( CONFIG_REISERFS_CHECK )
920 #define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
922 #define RFALSE( cond, format, args... ) do {;} while( 0 )
925 #define CONSTF __attribute_const__
927 * Disk Data Structures
930 /***************************************************************************
932 ***************************************************************************/
935 * Structure of super block on disk, a version of which in RAM is often
936 * accessed as REISERFS_SB(s)->s_rs. The version in RAM is part of a larger
937 * structure containing fields never written to disk.
939 #define UNSET_HASH 0 /* Detect hash on disk */
943 #define DEFAULT_HASH R5_HASH
945 struct journal_params {
946 /* where does journal start from on its * device */
947 __le32 jp_journal_1st_block;
949 /* journal device st_rdev */
950 __le32 jp_journal_dev;
952 /* size of the journal */
953 __le32 jp_journal_size;
955 /* max number of blocks in a transaction. */
956 __le32 jp_journal_trans_max;
959 * random value made on fs creation
960 * (this was sb_journal_block_count)
962 __le32 jp_journal_magic;
964 /* max number of blocks to batch into a trans */
965 __le32 jp_journal_max_batch;
967 /* in seconds, how old can an async commit be */
968 __le32 jp_journal_max_commit_age;
970 /* in seconds, how old can a transaction be */
971 __le32 jp_journal_max_trans_age;
974 /* this is the super from 3.5.X, where X >= 10 */
975 struct reiserfs_super_block_v1 {
976 __le32 s_block_count; /* blocks count */
977 __le32 s_free_blocks; /* free blocks count */
978 __le32 s_root_block; /* root block number */
979 struct journal_params s_journal;
980 __le16 s_blocksize; /* block size */
982 /* max size of object id array, see get_objectid() commentary */
983 __le16 s_oid_maxsize;
984 __le16 s_oid_cursize; /* current size of object id array */
986 /* this is set to 1 when filesystem was umounted, to 2 - when not */
987 __le16 s_umount_state;
990 * reiserfs magic string indicates that file system is reiserfs:
991 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs"
996 * it is set to used by fsck to mark which
997 * phase of rebuilding is done
1001 * indicate, what hash function is being use
1002 * to sort names in a directory
1004 __le32 s_hash_function_code;
1005 __le16 s_tree_height; /* height of disk tree */
1008 * amount of bitmap blocks needed to address
1009 * each block of file system
1014 * this field is only reliable on filesystem with non-standard journal
1019 * size in blocks of journal area on main device, we need to
1020 * keep after making fs with non-standard journal
1022 __le16 s_reserved_for_journal;
1023 } __attribute__ ((__packed__));
1025 #define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
1027 /* this is the on disk super block */
1028 struct reiserfs_super_block {
1029 struct reiserfs_super_block_v1 s_v1;
1030 __le32 s_inode_generation;
1032 /* Right now used only by inode-attributes, if enabled */
1035 unsigned char s_uuid[16]; /* filesystem unique identifier */
1036 unsigned char s_label[16]; /* filesystem volume label */
1037 __le16 s_mnt_count; /* Count of mounts since last fsck */
1038 __le16 s_max_mnt_count; /* Maximum mounts before check */
1039 __le32 s_lastcheck; /* Timestamp of last fsck */
1040 __le32 s_check_interval; /* Interval between checks */
1043 * zero filled by mkreiserfs and reiserfs_convert_objectid_map_v1()
1044 * so any additions must be updated there as well. */
1046 } __attribute__ ((__packed__));
1048 #define SB_SIZE (sizeof(struct reiserfs_super_block))
1050 #define REISERFS_VERSION_1 0
1051 #define REISERFS_VERSION_2 2
1053 /* on-disk super block fields converted to cpu form */
1054 #define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
1055 #define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
1056 #define SB_BLOCKSIZE(s) \
1057 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
1058 #define SB_BLOCK_COUNT(s) \
1059 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
1060 #define SB_FREE_BLOCKS(s) \
1061 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
1062 #define SB_REISERFS_MAGIC(s) \
1063 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
1064 #define SB_ROOT_BLOCK(s) \
1065 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
1066 #define SB_TREE_HEIGHT(s) \
1067 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
1068 #define SB_REISERFS_STATE(s) \
1069 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
1070 #define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
1071 #define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
1073 #define PUT_SB_BLOCK_COUNT(s, val) \
1074 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
1075 #define PUT_SB_FREE_BLOCKS(s, val) \
1076 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
1077 #define PUT_SB_ROOT_BLOCK(s, val) \
1078 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
1079 #define PUT_SB_TREE_HEIGHT(s, val) \
1080 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
1081 #define PUT_SB_REISERFS_STATE(s, val) \
1082 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1083 #define PUT_SB_VERSION(s, val) \
1084 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
1085 #define PUT_SB_BMAP_NR(s, val) \
1086 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
1088 #define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
1089 #define SB_ONDISK_JOURNAL_SIZE(s) \
1090 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
1091 #define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
1092 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
1093 #define SB_ONDISK_JOURNAL_DEVICE(s) \
1094 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
1095 #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
1096 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1098 #define is_block_in_log_or_reserved_area(s, block) \
1099 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
1100 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
1101 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
1102 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1104 int is_reiserfs_3_5(struct reiserfs_super_block *rs);
1105 int is_reiserfs_3_6(struct reiserfs_super_block *rs);
1106 int is_reiserfs_jr(struct reiserfs_super_block *rs);
1109 * ReiserFS leaves the first 64k unused, so that partition labels have
1110 * enough space. If someone wants to write a fancy bootloader that
1111 * needs more than 64k, let us know, and this will be increased in size.
1112 * This number must be larger than than the largest block size on any
1113 * platform, or code will break. -Hans
1115 #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
1116 #define REISERFS_FIRST_BLOCK unused_define
1117 #define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
1119 /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
1120 #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
1122 /* reiserfs internal error code (used by search_by_key and fix_nodes)) */
1124 #define REPEAT_SEARCH -1
1126 #define NO_DISK_SPACE -3
1127 #define NO_BALANCING_NEEDED (-4)
1128 #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
1129 #define QUOTA_EXCEEDED -6
1131 typedef __u32 b_blocknr_t;
1132 typedef __le32 unp_t;
1134 struct unfm_nodeinfo {
1136 unsigned short unfm_freespace;
1139 /* there are two formats of keys: 3.5 and 3.6 */
1140 #define KEY_FORMAT_3_5 0
1141 #define KEY_FORMAT_3_6 1
1143 /* there are two stat datas */
1144 #define STAT_DATA_V1 0
1145 #define STAT_DATA_V2 1
1147 static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
1149 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
1152 static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
1154 return sb->s_fs_info;
1158 * Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
1159 * which overflows on large file systems.
1161 static inline __u32 reiserfs_bmap_count(struct super_block *sb)
1163 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
1166 static inline int bmap_would_wrap(unsigned bmap_nr)
1168 return bmap_nr > ((1LL << 16) - 1);
1172 * this says about version of key of all items (but stat data) the
1173 * object consists of
1175 #define get_inode_item_key_version( inode ) \
1176 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
1178 #define set_inode_item_key_version( inode, version ) \
1179 ({ if((version)==KEY_FORMAT_3_6) \
1180 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
1182 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
1184 #define get_inode_sd_version(inode) \
1185 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
1187 #define set_inode_sd_version(inode, version) \
1188 ({ if((version)==STAT_DATA_V2) \
1189 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
1191 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1194 * This is an aggressive tail suppression policy, I am hoping it
1195 * improves our benchmarks. The principle behind it is that percentage
1196 * space saving is what matters, not absolute space saving. This is
1197 * non-intuitive, but it helps to understand it if you consider that the
1198 * cost to access 4 blocks is not much more than the cost to access 1
1199 * block, if you have to do a seek and rotate. A tail risks a
1200 * non-linear disk access that is significant as a percentage of total
1201 * time cost for a 4 block file and saves an amount of space that is
1202 * less significant as a percentage of space, or so goes the hypothesis.
1205 #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1207 (!(n_tail_size)) || \
1208 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1209 ( (n_file_size) >= (n_block_size) * 4 ) || \
1210 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1211 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1212 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1213 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1214 ( ( (n_file_size) >= (n_block_size) ) && \
1215 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1219 * Another strategy for tails, this one means only create a tail if all the
1220 * file would fit into one DIRECT item.
1221 * Primary intention for this one is to increase performance by decreasing
1224 #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1226 (!(n_tail_size)) || \
1227 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1231 * values for s_umount_state field
1233 #define REISERFS_VALID_FS 1
1234 #define REISERFS_ERROR_FS 2
1237 * there are 5 item types currently
1239 #define TYPE_STAT_DATA 0
1240 #define TYPE_INDIRECT 1
1241 #define TYPE_DIRECT 2
1242 #define TYPE_DIRENTRY 3
1243 #define TYPE_MAXTYPE 3
1244 #define TYPE_ANY 15 /* FIXME: comment is required */
1246 /***************************************************************************
1248 ***************************************************************************/
1250 /* * directories use this key as well as old files */
1253 __le32 k_uniqueness;
1254 } __attribute__ ((__packed__));
1258 } __attribute__ ((__packed__));
1260 static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1262 __u8 type = le64_to_cpu(v2->v) >> 60;
1263 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1266 static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1269 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1272 static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1274 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1277 static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1279 offset &= (~0ULL >> 4);
1280 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1284 * Key of an item determines its location in the S+tree, and
1285 * is composed of 4 components
1287 struct reiserfs_key {
1288 /* packing locality: by default parent directory object id */
1291 __le32 k_objectid; /* object identifier */
1293 struct offset_v1 k_offset_v1;
1294 struct offset_v2 k_offset_v2;
1295 } __attribute__ ((__packed__)) u;
1296 } __attribute__ ((__packed__));
1298 struct in_core_key {
1299 /* packing locality: by default parent directory object id */
1301 __u32 k_objectid; /* object identifier */
1307 struct in_core_key on_disk_key;
1309 /* 3 in all cases but direct2indirect and indirect2direct conversion */
1314 * Our function for comparing keys can compare keys of different
1315 * lengths. It takes as a parameter the length of the keys it is to
1316 * compare. These defines are used in determining what is to be passed
1317 * to it as that parameter.
1319 #define REISERFS_FULL_KEY_LEN 4
1320 #define REISERFS_SHORT_KEY_LEN 2
1322 /* The result of the key compare */
1323 #define FIRST_GREATER 1
1324 #define SECOND_GREATER -1
1325 #define KEYS_IDENTICAL 0
1327 #define KEY_NOT_FOUND 0
1329 #define KEY_SIZE (sizeof(struct reiserfs_key))
1330 #define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
1332 /* return values for search_by_key and clones */
1333 #define ITEM_FOUND 1
1334 #define ITEM_NOT_FOUND 0
1335 #define ENTRY_FOUND 1
1336 #define ENTRY_NOT_FOUND 0
1337 #define DIRECTORY_NOT_FOUND -1
1338 #define REGULAR_FILE_FOUND -2
1339 #define DIRECTORY_FOUND -3
1340 #define BYTE_FOUND 1
1341 #define BYTE_NOT_FOUND 0
1342 #define FILE_NOT_FOUND -1
1344 #define POSITION_FOUND 1
1345 #define POSITION_NOT_FOUND 0
1347 /* return values for reiserfs_find_entry and search_by_entry_key */
1348 #define NAME_FOUND 1
1349 #define NAME_NOT_FOUND 0
1350 #define GOTO_PREVIOUS_ITEM 2
1351 #define NAME_FOUND_INVISIBLE 3
1354 * Everything in the filesystem is stored as a set of items. The
1355 * item head contains the key of the item, its free space (for
1356 * indirect items) and specifies the location of the item itself
1362 * Everything in the tree is found by searching for it based on
1365 struct reiserfs_key ih_key;
1368 * The free space in the last unformatted node of an
1369 * indirect item if this is an indirect item. This
1370 * equals 0xFFFF iff this is a direct item or stat data
1371 * item. Note that the key, not this field, is used to
1372 * determine the item type, and thus which field this
1375 __le16 ih_free_space_reserved;
1378 * Iff this is a directory item, this field equals the
1379 * number of directory entries in the directory item.
1381 __le16 ih_entry_count;
1382 } __attribute__ ((__packed__)) u;
1383 __le16 ih_item_len; /* total size of the item body */
1385 /* an offset to the item body within the block */
1386 __le16 ih_item_location;
1389 * 0 for all old items, 2 for new ones. Highest bit is set by fsck
1390 * temporary, cleaned after all done
1393 } __attribute__ ((__packed__));
1394 /* size of item header */
1395 #define IH_SIZE (sizeof(struct item_head))
1397 #define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
1398 #define ih_version(ih) le16_to_cpu((ih)->ih_version)
1399 #define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
1400 #define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
1401 #define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
1403 #define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1404 #define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1405 #define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1406 #define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1407 #define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1409 #define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1411 #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1412 #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1415 * these operate on indirect items, where you've got an array of ints
1416 * at a possibly unaligned location. These are a noop on ia32
1418 * p is the array of __u32, i is the index into the array, v is the value
1421 #define get_block_num(p, i) get_unaligned_le32((p) + (i))
1422 #define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1424 /* * in old version uniqueness field shows key type */
1425 #define V1_SD_UNIQUENESS 0
1426 #define V1_INDIRECT_UNIQUENESS 0xfffffffe
1427 #define V1_DIRECT_UNIQUENESS 0xffffffff
1428 #define V1_DIRENTRY_UNIQUENESS 500
1429 #define V1_ANY_UNIQUENESS 555 /* FIXME: comment is required */
1431 /* here are conversion routines */
1432 static inline int uniqueness2type(__u32 uniqueness) CONSTF;
1433 static inline int uniqueness2type(__u32 uniqueness)
1435 switch ((int)uniqueness) {
1436 case V1_SD_UNIQUENESS:
1437 return TYPE_STAT_DATA;
1438 case V1_INDIRECT_UNIQUENESS:
1439 return TYPE_INDIRECT;
1440 case V1_DIRECT_UNIQUENESS:
1442 case V1_DIRENTRY_UNIQUENESS:
1443 return TYPE_DIRENTRY;
1444 case V1_ANY_UNIQUENESS:
1450 static inline __u32 type2uniqueness(int type) CONSTF;
1451 static inline __u32 type2uniqueness(int type)
1454 case TYPE_STAT_DATA:
1455 return V1_SD_UNIQUENESS;
1457 return V1_INDIRECT_UNIQUENESS;
1459 return V1_DIRECT_UNIQUENESS;
1461 return V1_DIRENTRY_UNIQUENESS;
1464 return V1_ANY_UNIQUENESS;
1469 * key is pointer to on disk key which is stored in le, result is cpu,
1470 * there is no way to get version of object from key, so, provide
1471 * version to these defines
1473 static inline loff_t le_key_k_offset(int version,
1474 const struct reiserfs_key *key)
1476 return (version == KEY_FORMAT_3_5) ?
1477 le32_to_cpu(key->u.k_offset_v1.k_offset) :
1478 offset_v2_k_offset(&(key->u.k_offset_v2));
1481 static inline loff_t le_ih_k_offset(const struct item_head *ih)
1483 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1486 static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1488 if (version == KEY_FORMAT_3_5) {
1489 loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness);
1490 return uniqueness2type(val);
1492 return offset_v2_k_type(&(key->u.k_offset_v2));
1495 static inline loff_t le_ih_k_type(const struct item_head *ih)
1497 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1500 static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1503 if (version == KEY_FORMAT_3_5)
1504 key->u.k_offset_v1.k_offset = cpu_to_le32(offset);
1506 set_offset_v2_k_offset(&key->u.k_offset_v2, offset);
1509 static inline void add_le_key_k_offset(int version, struct reiserfs_key *key,
1512 set_le_key_k_offset(version, key,
1513 le_key_k_offset(version, key) + offset);
1516 static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset)
1518 add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1521 static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1523 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1526 static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1529 if (version == KEY_FORMAT_3_5) {
1530 type = type2uniqueness(type);
1531 key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type);
1533 set_offset_v2_k_type(&key->u.k_offset_v2, type);
1536 static inline void set_le_ih_k_type(struct item_head *ih, int type)
1538 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1541 static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1543 return le_key_k_type(version, key) == TYPE_DIRENTRY;
1546 static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1548 return le_key_k_type(version, key) == TYPE_DIRECT;
1551 static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1553 return le_key_k_type(version, key) == TYPE_INDIRECT;
1556 static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1558 return le_key_k_type(version, key) == TYPE_STAT_DATA;
1561 /* item header has version. */
1562 static inline int is_direntry_le_ih(struct item_head *ih)
1564 return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1567 static inline int is_direct_le_ih(struct item_head *ih)
1569 return is_direct_le_key(ih_version(ih), &ih->ih_key);
1572 static inline int is_indirect_le_ih(struct item_head *ih)
1574 return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1577 static inline int is_statdata_le_ih(struct item_head *ih)
1579 return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1582 /* key is pointer to cpu key, result is cpu */
1583 static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1585 return key->on_disk_key.k_offset;
1588 static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1590 return key->on_disk_key.k_type;
1593 static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1595 key->on_disk_key.k_offset = offset;
1598 static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1600 key->on_disk_key.k_type = type;
1603 static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1605 key->on_disk_key.k_offset--;
1608 #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1609 #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1610 #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1611 #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1613 /* are these used ? */
1614 #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1615 #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1616 #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1617 #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1619 #define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1620 (!COMP_SHORT_KEYS(ih, key) && \
1621 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1623 /* maximal length of item */
1624 #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1625 #define MIN_ITEM_LEN 1
1627 /* object identifier for root dir */
1628 #define REISERFS_ROOT_OBJECTID 2
1629 #define REISERFS_ROOT_PARENT_OBJECTID 1
1631 extern struct reiserfs_key root_key;
1634 * Picture represents a leaf of the S+tree
1635 * ______________________________________________________
1636 * | | Array of | | |
1637 * |Block | Object-Item | F r e e | Objects- |
1638 * | head | Headers | S p a c e | Items |
1639 * |______|_______________|___________________|___________|
1643 * Header of a disk block. More precisely, header of a formatted leaf
1644 * or internal node, and not the header of an unformatted node.
1647 __le16 blk_level; /* Level of a block in the tree. */
1648 __le16 blk_nr_item; /* Number of keys/items in a block. */
1649 __le16 blk_free_space; /* Block free space in bytes. */
1650 __le16 blk_reserved;
1651 /* dump this in v4/planA */
1653 /* kept only for compatibility */
1654 struct reiserfs_key blk_right_delim_key;
1657 #define BLKH_SIZE (sizeof(struct block_head))
1658 #define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
1659 #define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
1660 #define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
1661 #define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
1662 #define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
1663 #define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1664 #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1665 #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1666 #define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
1667 #define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
1669 /* values for blk_level field of the struct block_head */
1672 * When node gets removed from the tree its blk_level is set to FREE_LEVEL.
1673 * It is then used to see whether the node is still in the tree
1675 #define FREE_LEVEL 0
1677 #define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1680 * Given the buffer head of a formatted node, resolve to the
1681 * block head of that node.
1683 #define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1684 /* Number of items that are in buffer. */
1685 #define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
1686 #define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
1687 #define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1689 #define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1690 #define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1691 #define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1693 /* Get right delimiting key. -- little endian */
1694 #define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1696 /* Does the buffer contain a disk leaf. */
1697 #define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1699 /* Does the buffer contain a disk internal node */
1700 #define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1701 && B_LEVEL(bh) <= MAX_HEIGHT)
1703 /***************************************************************************
1705 ***************************************************************************/
1708 * old stat data is 32 bytes long. We are going to distinguish new one by
1711 struct stat_data_v1 {
1712 __le16 sd_mode; /* file type, permissions */
1713 __le16 sd_nlink; /* number of hard links */
1714 __le16 sd_uid; /* owner */
1715 __le16 sd_gid; /* group */
1716 __le32 sd_size; /* file size */
1717 __le32 sd_atime; /* time of last access */
1718 __le32 sd_mtime; /* time file was last modified */
1721 * time inode (stat data) was last changed
1722 * (except changes to sd_atime and sd_mtime)
1727 __le32 sd_blocks; /* number of blocks file uses */
1728 } __attribute__ ((__packed__)) u;
1731 * first byte of file which is stored in a direct item: except that if
1732 * it equals 1 it is a symlink and if it equals ~(__u32)0 there is no
1733 * direct item. The existence of this field really grates on me.
1734 * Let's replace it with a macro based on sd_size and our tail
1735 * suppression policy. Someday. -Hans
1737 __le32 sd_first_direct_byte;
1738 } __attribute__ ((__packed__));
1740 #define SD_V1_SIZE (sizeof(struct stat_data_v1))
1741 #define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
1742 #define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1743 #define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1744 #define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
1745 #define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
1746 #define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
1747 #define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
1748 #define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
1749 #define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
1750 #define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
1751 #define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
1752 #define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1753 #define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1754 #define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1755 #define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1756 #define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1757 #define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1758 #define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1759 #define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1760 #define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
1761 #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1762 #define sd_v1_first_direct_byte(sdp) \
1763 (le32_to_cpu((sdp)->sd_first_direct_byte))
1764 #define set_sd_v1_first_direct_byte(sdp,v) \
1765 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1767 /* inode flags stored in sd_attrs (nee sd_reserved) */
1770 * we want common flags to have the same values as in ext2,
1771 * so chattr(1) will work without problems
1773 #define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1774 #define REISERFS_APPEND_FL FS_APPEND_FL
1775 #define REISERFS_SYNC_FL FS_SYNC_FL
1776 #define REISERFS_NOATIME_FL FS_NOATIME_FL
1777 #define REISERFS_NODUMP_FL FS_NODUMP_FL
1778 #define REISERFS_SECRM_FL FS_SECRM_FL
1779 #define REISERFS_UNRM_FL FS_UNRM_FL
1780 #define REISERFS_COMPR_FL FS_COMPR_FL
1781 #define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1783 /* persistent flags that file inherits from the parent directory */
1784 #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
1785 REISERFS_SYNC_FL | \
1786 REISERFS_NOATIME_FL | \
1787 REISERFS_NODUMP_FL | \
1788 REISERFS_SECRM_FL | \
1789 REISERFS_COMPR_FL | \
1790 REISERFS_NOTAIL_FL )
1793 * Stat Data on disk (reiserfs version of UFS disk inode minus the
1797 __le16 sd_mode; /* file type, permissions */
1798 __le16 sd_attrs; /* persistent inode flags */
1799 __le32 sd_nlink; /* number of hard links */
1800 __le64 sd_size; /* file size */
1801 __le32 sd_uid; /* owner */
1802 __le32 sd_gid; /* group */
1803 __le32 sd_atime; /* time of last access */
1804 __le32 sd_mtime; /* time file was last modified */
1807 * time inode (stat data) was last changed
1808 * (except changes to sd_atime and sd_mtime)
1814 __le32 sd_generation;
1815 } __attribute__ ((__packed__)) u;
1816 } __attribute__ ((__packed__));
1818 /* this is 44 bytes long */
1819 #define SD_SIZE (sizeof(struct stat_data))
1820 #define SD_V2_SIZE SD_SIZE
1821 #define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
1822 #define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1823 #define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1825 /* set_sd_reserved */
1826 #define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
1827 #define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
1828 #define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
1829 #define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
1830 #define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
1831 #define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
1832 #define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
1833 #define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
1834 #define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1835 #define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1836 #define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1837 #define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1838 #define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1839 #define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1840 #define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
1841 #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1842 #define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1843 #define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1844 #define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
1845 #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1846 #define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
1847 #define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
1849 /***************************************************************************
1850 * DIRECTORY STRUCTURE *
1851 ***************************************************************************/
1853 * Picture represents the structure of directory items
1854 * ________________________________________________
1855 * | Array of | | | | | |
1856 * | directory |N-1| N-2 | .... | 1st |0th|
1857 * | entry headers | | | | | |
1858 * |_______________|___|_____|________|_______|___|
1859 * <---- directory entries ------>
1861 * First directory item has k_offset component 1. We store "." and ".."
1862 * in one item, always, we never split "." and ".." into differing
1863 * items. This makes, among other things, the code for removing
1864 * directories simpler.
1867 #define SD_UNIQUENESS 0
1868 #define DOT_OFFSET 1
1869 #define DOT_DOT_OFFSET 2
1870 #define DIRENTRY_UNIQUENESS 500
1872 #define FIRST_ITEM_OFFSET 1
1875 * Q: How to get key of object pointed to by entry from entry?
1877 * A: Each directory entry has its header. This header has deh_dir_id
1878 * and deh_objectid fields, those are key of object, entry points to
1883 * Directory will someday contain stat data of object
1886 struct reiserfs_de_head {
1887 __le32 deh_offset; /* third component of the directory entry key */
1890 * objectid of the parent directory of the object, that is referenced
1891 * by directory entry
1895 /* objectid of the object, that is referenced by directory entry */
1896 __le32 deh_objectid;
1897 __le16 deh_location; /* offset of name in the whole item */
1900 * whether 1) entry contains stat data (for future), and
1901 * 2) whether entry is hidden (unlinked)
1904 } __attribute__ ((__packed__));
1905 #define DEH_SIZE sizeof(struct reiserfs_de_head)
1906 #define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
1907 #define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1908 #define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1909 #define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1910 #define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1912 #define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1913 #define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1914 #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1915 #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1916 #define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1918 /* empty directory contains two entries "." and ".." and their headers */
1919 #define EMPTY_DIR_SIZE \
1920 (DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1922 /* old format directories have this size when empty */
1923 #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1925 #define DEH_Statdata 0 /* not used now */
1926 #define DEH_Visible 2
1928 /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1929 #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1930 # define ADDR_UNALIGNED_BITS (3)
1934 * These are only used to manipulate deh_state.
1935 * Because of this, we'll use the ext2_ bit routines,
1936 * since they are little endian
1938 #ifdef ADDR_UNALIGNED_BITS
1940 # define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1941 # define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1943 # define set_bit_unaligned(nr, addr) \
1944 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1945 # define clear_bit_unaligned(nr, addr) \
1946 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1947 # define test_bit_unaligned(nr, addr) \
1948 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1952 # define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr)
1953 # define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr)
1954 # define test_bit_unaligned(nr, addr) test_bit_le(nr, addr)
1958 #define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1959 #define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1960 #define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1961 #define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1963 #define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1964 #define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1965 #define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1967 extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1968 __le32 par_dirid, __le32 par_objid);
1969 extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1970 __le32 par_dirid, __le32 par_objid);
1972 /* two entries per block (at least) */
1973 #define REISERFS_MAX_NAME(block_size) 255
1976 * this structure is used for operations on directory entries. It is
1977 * not a disk structure.
1979 * When reiserfs_find_entry or search_by_entry_key find directory
1980 * entry, they return filled reiserfs_dir_entry structure
1982 struct reiserfs_dir_entry {
1983 struct buffer_head *de_bh;
1985 struct item_head *de_ih;
1987 struct reiserfs_de_head *de_deh;
1991 unsigned long *de_gen_number_bit_string;
1996 struct cpu_key de_entry_key;
2000 * these defines are useful when a particular member of
2001 * a reiserfs_dir_entry is needed
2004 /* pointer to file name, stored in entry */
2005 #define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \
2006 (ih_item_body(bh, ih) + deh_location(deh))
2008 /* length of name */
2009 #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
2010 (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
2012 /* hash value occupies bits from 7 up to 30 */
2013 #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
2014 /* generation number occupies 7 bits starting from 0 up to 6 */
2015 #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
2016 #define MAX_GENERATION_NUMBER 127
2018 #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
2021 * Picture represents an internal node of the reiserfs tree
2022 * ______________________________________________________
2023 * | | Array of | Array of | Free |
2024 * |block | keys | pointers | space |
2025 * | head | N | N+1 | |
2026 * |______|_______________|___________________|___________|
2029 /***************************************************************************
2031 ***************************************************************************/
2033 * Disk child pointer:
2034 * The pointer from an internal node of the tree to a node that is on disk.
2037 __le32 dc_block_number; /* Disk child's block number. */
2038 __le16 dc_size; /* Disk child's used space. */
2042 #define DC_SIZE (sizeof(struct disk_child))
2043 #define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
2044 #define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
2045 #define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
2046 #define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
2048 /* Get disk child by buffer header and position in the tree node. */
2049 #define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
2050 ((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
2052 /* Get disk child number by buffer header and position in the tree node. */
2053 #define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
2054 #define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
2055 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
2057 /* maximal value of field child_size in structure disk_child */
2058 /* child size is the combined size of all items and their headers */
2059 #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
2061 /* amount of used space in buffer (not including block head) */
2062 #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
2064 /* max and min number of keys in internal node */
2065 #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
2066 #define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
2068 /***************************************************************************
2069 * PATH STRUCTURES AND DEFINES *
2070 ***************************************************************************/
2073 * search_by_key fills up the path from the root to the leaf as it descends
2074 * the tree looking for the key. It uses reiserfs_bread to try to find
2075 * buffers in the cache given their block number. If it does not find
2076 * them in the cache it reads them from disk. For each node search_by_key
2077 * finds using reiserfs_bread it then uses bin_search to look through that
2078 * node. bin_search will find the position of the block_number of the next
2079 * node if it is looking through an internal node. If it is looking through
2080 * a leaf node bin_search will find the position of the item which has key
2081 * either equal to given key, or which is the maximal key less than the
2085 struct path_element {
2086 /* Pointer to the buffer at the path in the tree. */
2087 struct buffer_head *pe_buffer;
2088 /* Position in the tree node which is placed in the buffer above. */
2093 * maximal height of a tree. don't change this without
2094 * changing JOURNAL_PER_BALANCE_CNT
2096 #define MAX_HEIGHT 5
2098 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
2099 #define EXTENDED_MAX_HEIGHT 7
2101 /* Must be equal to at least 2. */
2102 #define FIRST_PATH_ELEMENT_OFFSET 2
2104 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
2105 #define ILLEGAL_PATH_ELEMENT_OFFSET 1
2107 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
2108 #define MAX_FEB_SIZE 6
2111 * We need to keep track of who the ancestors of nodes are. When we
2112 * perform a search we record which nodes were visited while
2113 * descending the tree looking for the node we searched for. This list
2114 * of nodes is called the path. This information is used while
2115 * performing balancing. Note that this path information may become
2116 * invalid, and this means we must check it when using it to see if it
2117 * is still valid. You'll need to read search_by_key and the comments
2118 * in it, especially about decrement_counters_in_path(), to understand
2121 * Paths make the code so much harder to work with and debug.... An
2122 * enormous number of bugs are due to them, and trying to write or modify
2123 * code that uses them just makes my head hurt. They are based on an
2124 * excessive effort to avoid disturbing the precious VFS code.:-( The
2125 * gods only know how we are going to SMP the code that uses them.
2126 * znodes are the way!
2129 #define PATH_READA 0x1 /* do read ahead */
2130 #define PATH_READA_BACK 0x2 /* read backwards */
2133 int path_length; /* Length of the array above. */
2135 /* Array of the path elements. */
2136 struct path_element path_elements[EXTENDED_MAX_HEIGHT];
2140 #define pos_in_item(path) ((path)->pos_in_item)
2142 #define INITIALIZE_PATH(var) \
2143 struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
2145 /* Get path element by path and path position. */
2146 #define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
2148 /* Get buffer header at the path by path and path position. */
2149 #define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
2151 /* Get position in the element at the path by path and path position. */
2152 #define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
2154 #define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
2157 * you know, to the person who didn't write this the macro name does not
2158 * at first suggest what it does. Maybe POSITION_FROM_PATH_END? Or
2159 * maybe we should just focus on dumping paths... -Hans
2161 #define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
2164 * in do_balance leaf has h == 0 in contrast with path structure,
2165 * where root has level == 0. That is why we need these defines
2169 #define PATH_H_PBUFFER(path, h) \
2170 PATH_OFFSET_PBUFFER(path, path->path_length - (h))
2172 /* tb->F[h] or tb->S[0]->b_parent */
2173 #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER(path, (h) + 1)
2175 #define PATH_H_POSITION(path, h) \
2176 PATH_OFFSET_POSITION(path, path->path_length - (h))
2178 /* tb->S[h]->b_item_order */
2179 #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)
2181 #define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
2183 static inline void *reiserfs_node_data(const struct buffer_head *bh)
2185 return bh->b_data + sizeof(struct block_head);
2188 /* get key from internal node */
2189 static inline struct reiserfs_key *internal_key(struct buffer_head *bh,
2192 struct reiserfs_key *key = reiserfs_node_data(bh);
2194 return &key[item_num];
2197 /* get the item header from leaf node */
2198 static inline struct item_head *item_head(const struct buffer_head *bh,
2201 struct item_head *ih = reiserfs_node_data(bh);
2203 return &ih[item_num];
2206 /* get the key from leaf node */
2207 static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh,
2210 return &item_head(bh, item_num)->ih_key;
2213 static inline void *ih_item_body(const struct buffer_head *bh,
2214 const struct item_head *ih)
2216 return bh->b_data + ih_location(ih);
2219 /* get item body from leaf node */
2220 static inline void *item_body(const struct buffer_head *bh, int item_num)
2222 return ih_item_body(bh, item_head(bh, item_num));
2225 static inline struct item_head *tp_item_head(const struct treepath *path)
2227 return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2230 static inline void *tp_item_body(const struct treepath *path)
2232 return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2235 #define get_last_bh(path) PATH_PLAST_BUFFER(path)
2236 #define get_item_pos(path) PATH_LAST_POSITION(path)
2237 #define item_moved(ih,path) comp_items(ih, path)
2238 #define path_changed(ih,path) comp_items (ih, path)
2240 /* array of the entry headers */
2242 #define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih)))
2245 * length of the directory entry in directory item. This define
2246 * calculates length of i-th directory entry using directory entry
2247 * locations from dir entry head. When it calculates length of 0-th
2248 * directory entry, it uses length of whole item in place of entry
2249 * location of the non-existent following entry in the calculation.
2250 * See picture above.
2252 static inline int entry_length(const struct buffer_head *bh,
2253 const struct item_head *ih, int pos_in_item)
2255 struct reiserfs_de_head *deh;
2257 deh = B_I_DEH(bh, ih) + pos_in_item;
2259 return deh_location(deh - 1) - deh_location(deh);
2261 return ih_item_len(ih) - deh_location(deh);
2264 /***************************************************************************
2266 ***************************************************************************/
2268 /* Size of pointer to the unformatted node. */
2269 #define UNFM_P_SIZE (sizeof(unp_t))
2270 #define UNFM_P_SHIFT 2
2272 /* in in-core inode key is stored on le form */
2273 #define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
2275 #define MAX_UL_INT 0xffffffff
2276 #define MAX_INT 0x7ffffff
2277 #define MAX_US_INT 0xffff
2279 // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
2280 static inline loff_t max_reiserfs_offset(struct inode *inode)
2282 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
2283 return (loff_t) U32_MAX;
2285 return (loff_t) ((~(__u64) 0) >> 4);
2288 #define MAX_KEY_OBJECTID MAX_UL_INT
2290 #define MAX_B_NUM MAX_UL_INT
2291 #define MAX_FC_NUM MAX_US_INT
2293 /* the purpose is to detect overflow of an unsigned short */
2294 #define REISERFS_LINK_MAX (MAX_US_INT - 1000)
2297 * The following defines are used in reiserfs_insert_item
2298 * and reiserfs_append_item
2300 #define REISERFS_KERNEL_MEM 0 /* kernel memory mode */
2301 #define REISERFS_USER_MEM 1 /* user memory mode */
2303 #define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
2304 #define get_generation(s) atomic_read (&fs_generation(s))
2305 #define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
2306 #define __fs_changed(gen,s) (gen != get_generation (s))
2307 #define fs_changed(gen,s) \
2309 reiserfs_cond_resched(s); \
2310 __fs_changed(gen, s); \
2313 /***************************************************************************
2315 ***************************************************************************/
2317 #define VI_TYPE_LEFT_MERGEABLE 1
2318 #define VI_TYPE_RIGHT_MERGEABLE 2
2321 * To make any changes in the tree we always first find node, that
2322 * contains item to be changed/deleted or place to insert a new
2323 * item. We call this node S. To do balancing we need to decide what
2324 * we will shift to left/right neighbor, or to a new node, where new
2325 * item will be etc. To make this analysis simpler we build virtual
2326 * node. Virtual node is an array of items, that will replace items of
2327 * node S. (For instance if we are going to delete an item, virtual
2328 * node does not contain it). Virtual node keeps information about
2329 * item sizes and types, mergeability of first and last items, sizes
2330 * of all entries in directory item. We use this array of items when
2331 * calculating what we can shift to neighbors and how many nodes we
2332 * have to have if we do not any shiftings, if we shift to left/right
2333 * neighbor or to both.
2335 struct virtual_item {
2336 int vi_index; /* index in the array of item operations */
2337 unsigned short vi_type; /* left/right mergeability */
2339 /* length of item that it will have after balancing */
2340 unsigned short vi_item_len;
2342 struct item_head *vi_ih;
2343 const char *vi_item; /* body of item (old or new) */
2344 const void *vi_new_data; /* 0 always but paste mode */
2345 void *vi_uarea; /* item specific area */
2348 struct virtual_node {
2349 /* this is a pointer to the free space in the buffer */
2352 unsigned short vn_nr_item; /* number of items in virtual node */
2355 * size of node , that node would have if it has
2356 * unlimited size and no balancing is performed
2360 /* mode of balancing (paste, insert, delete, cut) */
2363 short vn_affected_item_num;
2364 short vn_pos_in_item;
2366 /* item header of inserted item, 0 for other modes */
2367 struct item_head *vn_ins_ih;
2368 const void *vn_data;
2370 /* array of items (including a new one, excluding item to be deleted) */
2371 struct virtual_item *vn_vi;
2374 /* used by directory items when creating virtual nodes */
2375 struct direntry_uarea {
2378 __u16 entry_sizes[1];
2379 } __attribute__ ((__packed__));
2381 /***************************************************************************
2383 ***************************************************************************/
2386 * This temporary structure is used in tree balance algorithms, and
2387 * constructed as we go to the extent that its various parts are
2388 * needed. It contains arrays of nodes that can potentially be
2389 * involved in the balancing of node S, and parameters that define how
2390 * each of the nodes must be balanced. Note that in these algorithms
2391 * for balancing the worst case is to need to balance the current node
2392 * S and the left and right neighbors and all of their parents plus
2393 * create a new node. We implement S1 balancing for the leaf nodes
2394 * and S0 balancing for the internal nodes (S1 and S0 are defined in
2398 /* size of the array of buffers to free at end of do_balance */
2399 #define MAX_FREE_BLOCK 7
2401 /* maximum number of FEB blocknrs on a single level */
2402 #define MAX_AMOUNT_NEEDED 2
2404 /* someday somebody will prefix every field in this struct with tb_ */
2405 struct tree_balance {
2407 int need_balance_dirty;
2408 struct super_block *tb_sb;
2409 struct reiserfs_transaction_handle *transaction_handle;
2410 struct treepath *tb_path;
2412 /* array of left neighbors of nodes in the path */
2413 struct buffer_head *L[MAX_HEIGHT];
2415 /* array of right neighbors of nodes in the path */
2416 struct buffer_head *R[MAX_HEIGHT];
2418 /* array of fathers of the left neighbors */
2419 struct buffer_head *FL[MAX_HEIGHT];
2421 /* array of fathers of the right neighbors */
2422 struct buffer_head *FR[MAX_HEIGHT];
2423 /* array of common parents of center node and its left neighbor */
2424 struct buffer_head *CFL[MAX_HEIGHT];
2426 /* array of common parents of center node and its right neighbor */
2427 struct buffer_head *CFR[MAX_HEIGHT];
2430 * array of empty buffers. Number of buffers in array equals
2433 struct buffer_head *FEB[MAX_FEB_SIZE];
2434 struct buffer_head *used[MAX_FEB_SIZE];
2435 struct buffer_head *thrown[MAX_FEB_SIZE];
2438 * array of number of items which must be shifted to the left in
2439 * order to balance the current node; for leaves includes item that
2440 * will be partially shifted; for internal nodes, it is the number
2441 * of child pointers rather than items. It includes the new item
2442 * being created. The code sometimes subtracts one to get the
2443 * number of wholly shifted items for other purposes.
2445 int lnum[MAX_HEIGHT];
2447 /* substitute right for left in comment above */
2448 int rnum[MAX_HEIGHT];
2451 * array indexed by height h mapping the key delimiting L[h] and
2452 * S[h] to its item number within the node CFL[h]
2454 int lkey[MAX_HEIGHT];
2456 /* substitute r for l in comment above */
2457 int rkey[MAX_HEIGHT];
2460 * the number of bytes by we are trying to add or remove from
2461 * S[h]. A negative value means removing.
2463 int insert_size[MAX_HEIGHT];
2466 * number of nodes that will replace node S[h] after balancing
2467 * on the level h of the tree. If 0 then S is being deleted,
2468 * if 1 then S is remaining and no new nodes are being created,
2469 * if 2 or 3 then 1 or 2 new nodes is being created
2471 int blknum[MAX_HEIGHT];
2473 /* fields that are used only for balancing leaves of the tree */
2475 /* number of empty blocks having been already allocated */
2478 /* number of items that fall into left most node when S[0] splits */
2482 * number of bytes which can flow to the left neighbor from the left
2483 * most liquid item that cannot be shifted from S[0] entirely
2484 * if -1 then nothing will be partially shifted
2489 * number of bytes which will flow to the right neighbor from the right
2490 * most liquid item that cannot be shifted from S[0] entirely
2491 * if -1 then nothing will be partially shifted
2497 * index into the array of item headers in
2498 * S[0] of the affected item
2502 /* new nodes allocated to hold what could not fit into S */
2503 struct buffer_head *S_new[2];
2506 * number of items that will be placed into nodes in S_new
2512 * number of bytes which flow to nodes in S_new when S[0] splits
2513 * note: if S[0] splits into 3 nodes, then items do not need to be cut
2521 * buffers which are to be freed after do_balance finishes
2524 struct buffer_head *buf_to_free[MAX_FREE_BLOCK];
2527 * kmalloced memory. Used to create virtual node and keep
2528 * map of dirtied bitmap blocks
2532 int vn_buf_size; /* size of the vn_buf */
2534 /* VN starts after bitmap of bitmap blocks */
2535 struct virtual_node *tb_vn;
2538 * saved value of `reiserfs_generation' counter see
2539 * FILESYSTEM_CHANGED() macro in reiserfs_fs.h
2543 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2545 * key pointer, to pass to block allocator or
2546 * another low-level subsystem
2548 struct in_core_key key;
2552 /* These are modes of balancing */
2554 /* When inserting an item. */
2555 #define M_INSERT 'i'
2557 * When inserting into (directories only) or appending onto an already
2561 /* When deleting an item. */
2562 #define M_DELETE 'd'
2563 /* When truncating an item or removing an entry from a (directory) item. */
2566 /* used when balancing on leaf level skipped (in reiserfsck) */
2567 #define M_INTERNAL 'n'
2570 * When further balancing is not needed, then do_balance does not need
2573 #define M_SKIP_BALANCING 's'
2574 #define M_CONVERT 'v'
2576 /* modes of leaf_move_items */
2577 #define LEAF_FROM_S_TO_L 0
2578 #define LEAF_FROM_S_TO_R 1
2579 #define LEAF_FROM_R_TO_L 2
2580 #define LEAF_FROM_L_TO_R 3
2581 #define LEAF_FROM_S_TO_SNEW 4
2583 #define FIRST_TO_LAST 0
2584 #define LAST_TO_FIRST 1
2587 * used in do_balance for passing parent of node information that has
2588 * been gotten from tb struct
2590 struct buffer_info {
2591 struct tree_balance *tb;
2592 struct buffer_head *bi_bh;
2593 struct buffer_head *bi_parent;
2597 static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2599 return tb ? tb->tb_sb : NULL;
2602 static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2604 return bi ? sb_from_tb(bi->tb) : NULL;
2608 * there are 4 types of items: stat data, directory item, indirect, direct.
2609 * +-------------------+------------+--------------+------------+
2610 * | | k_offset | k_uniqueness | mergeable? |
2611 * +-------------------+------------+--------------+------------+
2612 * | stat data | 0 | 0 | no |
2613 * +-------------------+------------+--------------+------------+
2614 * | 1st directory item| DOT_OFFSET | DIRENTRY_ .. | no |
2615 * | non 1st directory | hash value | UNIQUENESS | yes |
2617 * +-------------------+------------+--------------+------------+
2618 * | indirect item | offset + 1 |TYPE_INDIRECT | [1] |
2619 * +-------------------+------------+--------------+------------+
2620 * | direct item | offset + 1 |TYPE_DIRECT | [2] |
2621 * +-------------------+------------+--------------+------------+
2623 * [1] if this is not the first indirect item of the object
2624 * [2] if this is not the first direct item of the object
2627 struct item_operations {
2628 int (*bytes_number) (struct item_head * ih, int block_size);
2629 void (*decrement_key) (struct cpu_key *);
2630 int (*is_left_mergeable) (struct reiserfs_key * ih,
2631 unsigned long bsize);
2632 void (*print_item) (struct item_head *, char *item);
2633 void (*check_item) (struct item_head *, char *item);
2635 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2636 int is_affected, int insert_size);
2637 int (*check_left) (struct virtual_item * vi, int free,
2638 int start_skip, int end_skip);
2639 int (*check_right) (struct virtual_item * vi, int free);
2640 int (*part_size) (struct virtual_item * vi, int from, int to);
2641 int (*unit_num) (struct virtual_item * vi);
2642 void (*print_vi) (struct virtual_item * vi);
2645 extern struct item_operations *item_ops[TYPE_ANY + 1];
2647 #define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2648 #define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2649 #define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2650 #define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2651 #define op_create_vi(vn,vi,is_affected,insert_size) item_ops[le_ih_k_type ((vi)->vi_ih)]->create_vi (vn,vi,is_affected,insert_size)
2652 #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2653 #define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
2654 #define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
2655 #define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
2656 #define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
2658 #define COMP_SHORT_KEYS comp_short_keys
2660 /* number of blocks pointed to by the indirect item */
2661 #define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
2664 * the used space within the unformatted node corresponding
2665 * to pos within the item pointed to by ih
2667 #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2670 * number of bytes contained by the direct item or the
2671 * unformatted nodes the indirect item points to
2674 /* following defines use reiserfs buffer header and item header */
2677 #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2679 /* this is 3976 for size==4096 */
2680 #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2683 * indirect items consist of entries which contain blocknrs, pos
2684 * indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2685 * blocknr contained by the entry pos points to
2687 #define B_I_POS_UNFM_POINTER(bh, ih, pos) \
2688 le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos)))
2689 #define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val) \
2690 (*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val))
2692 struct reiserfs_iget_args {
2697 /***************************************************************************
2698 * FUNCTION DECLARATIONS *
2699 ***************************************************************************/
2701 #define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2703 #define journal_trans_half(blocksize) \
2704 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2706 /* journal.c see journal.c for all the comments here */
2708 /* first block written in a commit. */
2709 struct reiserfs_journal_desc {
2710 __le32 j_trans_id; /* id of commit */
2712 /* length of commit. len +1 is the commit block */
2715 __le32 j_mount_id; /* mount id of this trans */
2716 __le32 j_realblock[1]; /* real locations for each block */
2719 #define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
2720 #define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
2721 #define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
2723 #define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2724 #define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
2725 #define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2727 /* last block written in a commit */
2728 struct reiserfs_journal_commit {
2729 __le32 j_trans_id; /* must match j_trans_id from the desc block */
2730 __le32 j_len; /* ditto */
2731 __le32 j_realblock[1]; /* real locations for each block */
2734 #define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2735 #define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
2736 #define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2738 #define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2739 #define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
2742 * this header block gets written whenever a transaction is considered
2743 * fully flushed, and is more recent than the last fully flushed transaction.
2744 * fully flushed means all the log blocks and all the real blocks are on
2745 * disk, and this transaction does not need to be replayed.
2747 struct reiserfs_journal_header {
2748 /* id of last fully flushed transaction */
2749 __le32 j_last_flush_trans_id;
2751 /* offset in the log of where to start replay after a crash */
2752 __le32 j_first_unflushed_offset;
2755 /* 12 */ struct journal_params jh_journal;
2758 /* biggest tunable defines are right here */
2759 #define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
2761 /* biggest possible single transaction, don't change for now (8/3/99) */
2762 #define JOURNAL_TRANS_MAX_DEFAULT 1024
2763 #define JOURNAL_TRANS_MIN_DEFAULT 256
2766 * max blocks to batch into one transaction,
2767 * don't make this any bigger than 900
2769 #define JOURNAL_MAX_BATCH_DEFAULT 900
2770 #define JOURNAL_MIN_RATIO 2
2771 #define JOURNAL_MAX_COMMIT_AGE 30
2772 #define JOURNAL_MAX_TRANS_AGE 30
2773 #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2774 #define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
2775 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2776 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2779 #define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2780 /* We need to update data and inode (atime) */
2781 #define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2782 /* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2783 #define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2784 (DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2785 /* same as with INIT */
2786 #define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2787 (DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2789 #define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2790 #define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2791 #define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2795 * both of these can be as low as 1, or as high as you want. The min is the
2796 * number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2797 * as needed, and released when transactions are committed. On release, if
2798 * the current number of nodes is > max, the node is freed, otherwise,
2799 * it is put on a free list for faster use later.
2801 #define REISERFS_MIN_BITMAP_NODES 10
2802 #define REISERFS_MAX_BITMAP_NODES 100
2804 /* these are based on journal hash size of 8192 */
2805 #define JBH_HASH_SHIFT 13
2806 #define JBH_HASH_MASK 8191
2808 #define _jhashfn(sb,block) \
2809 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2810 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2811 #define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2813 /* We need these to make journal.c code more readable */
2814 #define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2815 #define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2816 #define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2818 enum reiserfs_bh_state_bits {
2819 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
2822 * disk block was taken off free list before being in a
2823 * finished transaction, or written to disk. Can be reused immed.
2828 BH_JTest, /* debugging only will go away */
2831 BUFFER_FNS(JDirty, journaled);
2832 TAS_BUFFER_FNS(JDirty, journaled);
2833 BUFFER_FNS(JDirty_wait, journal_dirty);
2834 TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2835 BUFFER_FNS(JNew, journal_new);
2836 TAS_BUFFER_FNS(JNew, journal_new);
2837 BUFFER_FNS(JPrepared, journal_prepared);
2838 TAS_BUFFER_FNS(JPrepared, journal_prepared);
2839 BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2840 TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2841 BUFFER_FNS(JTest, journal_test);
2842 TAS_BUFFER_FNS(JTest, journal_test);
2844 /* transaction handle which is passed around for all journal calls */
2845 struct reiserfs_transaction_handle {
2847 * super for this FS when journal_begin was called. saves calls to
2848 * reiserfs_get_super also used by nested transactions to make
2849 * sure they are nesting on the right FS _must_ be first
2852 struct super_block *t_super;
2855 int t_blocks_logged; /* number of blocks this writer has logged */
2856 int t_blocks_allocated; /* number of blocks this writer allocated */
2858 /* sanity check, equals the current trans id */
2859 unsigned int t_trans_id;
2861 void *t_handle_save; /* save existing current->journal_info */
2864 * if new block allocation occurres, that block
2865 * should be displaced from others
2867 unsigned displace_new_blocks:1;
2869 struct list_head t_list;
2873 * used to keep track of ordered and tail writes, attached to the buffer
2874 * head through b_journal_head.
2876 struct reiserfs_jh {
2877 struct reiserfs_journal_list *jl;
2878 struct buffer_head *bh;
2879 struct list_head list;
2882 void reiserfs_free_jh(struct buffer_head *bh);
2883 int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2884 int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2885 int journal_mark_dirty(struct reiserfs_transaction_handle *,
2886 struct buffer_head *bh);
2888 static inline int reiserfs_file_data_log(struct inode *inode)
2890 if (reiserfs_data_log(inode->i_sb) ||
2891 (REISERFS_I(inode)->i_flags & i_data_log))
2896 static inline int reiserfs_transaction_running(struct super_block *s)
2898 struct reiserfs_transaction_handle *th = current->journal_info;
2899 if (th && th->t_super == s)
2901 if (th && th->t_super == NULL)
2906 static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2908 return th->t_blocks_allocated - th->t_blocks_logged;
2911 struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2915 int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
2916 void reiserfs_vfs_truncate_file(struct inode *inode);
2917 int reiserfs_commit_page(struct inode *inode, struct page *page,
2918 unsigned from, unsigned to);
2919 void reiserfs_flush_old_commits(struct super_block *);
2920 int reiserfs_commit_for_inode(struct inode *);
2921 int reiserfs_inode_needs_commit(struct inode *);
2922 void reiserfs_update_inode_transaction(struct inode *);
2923 void reiserfs_wait_on_write_block(struct super_block *s);
2924 void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2925 void reiserfs_allow_writes(struct super_block *s);
2926 void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2927 int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2929 void reiserfs_restore_prepared_buffer(struct super_block *,
2930 struct buffer_head *bh);
2931 int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2933 int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2934 int journal_release_error(struct reiserfs_transaction_handle *,
2935 struct super_block *);
2936 int journal_end(struct reiserfs_transaction_handle *);
2937 int journal_end_sync(struct reiserfs_transaction_handle *);
2938 int journal_mark_freed(struct reiserfs_transaction_handle *,
2939 struct super_block *, b_blocknr_t blocknr);
2940 int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2941 int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2942 int bit_nr, int searchall, b_blocknr_t *next);
2943 int journal_begin(struct reiserfs_transaction_handle *,
2944 struct super_block *sb, unsigned long);
2945 int journal_join_abort(struct reiserfs_transaction_handle *,
2946 struct super_block *sb);
2947 void reiserfs_abort_journal(struct super_block *sb, int errno);
2948 void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2949 int reiserfs_allocate_list_bitmaps(struct super_block *s,
2950 struct reiserfs_list_bitmap *, unsigned int);
2952 void reiserfs_schedule_old_flush(struct super_block *s);
2953 void add_save_link(struct reiserfs_transaction_handle *th,
2954 struct inode *inode, int truncate);
2955 int remove_save_link(struct inode *inode, int truncate);
2958 __u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2959 void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2960 __u32 objectid_to_release);
2961 int reiserfs_convert_objectid_map_v1(struct super_block *);
2964 int B_IS_IN_TREE(const struct buffer_head *);
2965 extern void copy_item_head(struct item_head *to,
2966 const struct item_head *from);
2968 /* first key is in cpu form, second - le */
2969 extern int comp_short_keys(const struct reiserfs_key *le_key,
2970 const struct cpu_key *cpu_key);
2971 extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2973 /* both are in le form */
2974 extern int comp_le_keys(const struct reiserfs_key *,
2975 const struct reiserfs_key *);
2976 extern int comp_short_le_keys(const struct reiserfs_key *,
2977 const struct reiserfs_key *);
2979 /* * get key version from on disk key - kludge */
2980 static inline int le_key_version(const struct reiserfs_key *key)
2984 type = offset_v2_k_type(&(key->u.k_offset_v2));
2985 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2986 && type != TYPE_DIRENTRY)
2987 return KEY_FORMAT_3_5;
2989 return KEY_FORMAT_3_6;
2993 static inline void copy_key(struct reiserfs_key *to,
2994 const struct reiserfs_key *from)
2996 memcpy(to, from, KEY_SIZE);
2999 int comp_items(const struct item_head *stored_ih, const struct treepath *path);
3000 const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
3001 const struct super_block *sb);
3002 int search_by_key(struct super_block *, const struct cpu_key *,
3003 struct treepath *, int);
3004 #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
3005 int search_for_position_by_key(struct super_block *sb,
3006 const struct cpu_key *cpu_key,
3007 struct treepath *search_path);
3008 extern void decrement_bcount(struct buffer_head *bh);
3009 void decrement_counters_in_path(struct treepath *search_path);
3010 void pathrelse(struct treepath *search_path);
3011 int reiserfs_check_path(struct treepath *p);
3012 void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
3014 int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
3015 struct treepath *path,
3016 const struct cpu_key *key,
3017 struct item_head *ih,
3018 struct inode *inode, const char *body);
3020 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
3021 struct treepath *path,
3022 const struct cpu_key *key,
3023 struct inode *inode,
3024 const char *body, int paste_size);
3026 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
3027 struct treepath *path,
3028 struct cpu_key *key,
3029 struct inode *inode,
3030 struct page *page, loff_t new_file_size);
3032 int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
3033 struct treepath *path,
3034 const struct cpu_key *key,
3035 struct inode *inode, struct buffer_head *un_bh);
3037 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
3038 struct inode *inode, struct reiserfs_key *key);
3039 int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
3040 struct inode *inode);
3041 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
3042 struct inode *inode, struct page *,
3043 int update_timestamps);
3045 #define i_block_size(inode) ((inode)->i_sb->s_blocksize)
3046 #define file_size(inode) ((inode)->i_size)
3047 #define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
3049 #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
3050 !STORE_TAIL_IN_UNFM_S1(file_size (inode), tail_size(inode), inode->i_sb->s_blocksize):have_small_tails ((inode)->i_sb)?!STORE_TAIL_IN_UNFM_S2(file_size (inode), tail_size(inode), inode->i_sb->s_blocksize):0 )
3052 void padd_item(char *item, int total_length, int length);
3055 /* args for the create parameter of reiserfs_get_block */
3056 #define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
3057 #define GET_BLOCK_CREATE 1 /* add anything you need to find block */
3058 #define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
3059 #define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
3060 #define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
3061 #define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
3063 void reiserfs_read_locked_inode(struct inode *inode,
3064 struct reiserfs_iget_args *args);
3065 int reiserfs_find_actor(struct inode *inode, void *p);
3066 int reiserfs_init_locked_inode(struct inode *inode, void *p);
3067 void reiserfs_evict_inode(struct inode *inode);
3068 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
3069 int reiserfs_get_block(struct inode *inode, sector_t block,
3070 struct buffer_head *bh_result, int create);
3071 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3072 int fh_len, int fh_type);
3073 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
3074 int fh_len, int fh_type);
3075 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
3076 struct inode *parent);
3078 int reiserfs_truncate_file(struct inode *, int update_timestamps);
3079 void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
3080 int type, int key_length);
3081 void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
3083 loff_t offset, int type, int length, int entry_count);
3084 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
3086 struct reiserfs_security_handle;
3087 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
3088 struct inode *dir, umode_t mode,
3089 const char *symname, loff_t i_size,
3090 struct dentry *dentry, struct inode *inode,
3091 struct reiserfs_security_handle *security);
3093 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
3094 struct inode *inode, loff_t size);
3096 static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
3097 struct inode *inode)
3099 reiserfs_update_sd_size(th, inode, inode->i_size);
3102 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
3103 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
3104 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
3106 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
3109 void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
3110 int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
3111 struct treepath *path, struct reiserfs_dir_entry *de);
3112 struct dentry *reiserfs_get_parent(struct dentry *);
3114 #ifdef CONFIG_REISERFS_PROC_INFO
3115 int reiserfs_proc_info_init(struct super_block *sb);
3116 int reiserfs_proc_info_done(struct super_block *sb);
3117 int reiserfs_proc_info_global_init(void);
3118 int reiserfs_proc_info_global_done(void);
3120 #define PROC_EXP( e ) e
3122 #define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
3123 #define PROC_INFO_MAX( sb, field, value ) \
3124 __PINFO( sb ).field = \
3125 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
3126 #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
3127 #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
3128 #define PROC_INFO_BH_STAT( sb, bh, level ) \
3129 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
3130 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
3131 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
3133 static inline int reiserfs_proc_info_init(struct super_block *sb)
3138 static inline int reiserfs_proc_info_done(struct super_block *sb)
3143 static inline int reiserfs_proc_info_global_init(void)
3148 static inline int reiserfs_proc_info_global_done(void)
3153 #define PROC_EXP( e )
3154 #define VOID_V ( ( void ) 0 )
3155 #define PROC_INFO_MAX( sb, field, value ) VOID_V
3156 #define PROC_INFO_INC( sb, field ) VOID_V
3157 #define PROC_INFO_ADD( sb, field, val ) VOID_V
3158 #define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
3162 extern const struct inode_operations reiserfs_dir_inode_operations;
3163 extern const struct inode_operations reiserfs_symlink_inode_operations;
3164 extern const struct inode_operations reiserfs_special_inode_operations;
3165 extern const struct file_operations reiserfs_dir_operations;
3166 int reiserfs_readdir_inode(struct inode *, struct dir_context *);
3168 /* tail_conversion.c */
3169 int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
3170 struct treepath *, struct buffer_head *, loff_t);
3171 int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
3172 struct page *, struct treepath *, const struct cpu_key *,
3174 void reiserfs_unmap_buffer(struct buffer_head *);
3177 extern const struct inode_operations reiserfs_file_inode_operations;
3178 extern const struct file_operations reiserfs_file_operations;
3179 extern const struct address_space_operations reiserfs_address_space_operations;
3183 int fix_nodes(int n_op_mode, struct tree_balance *tb,
3184 struct item_head *ins_ih, const void *);
3185 void unfix_nodes(struct tree_balance *);
3188 void __reiserfs_panic(struct super_block *s, const char *id,
3189 const char *function, const char *fmt, ...)
3190 __attribute__ ((noreturn));
3191 #define reiserfs_panic(s, id, fmt, args...) \
3192 __reiserfs_panic(s, id, __func__, fmt, ##args)
3193 void __reiserfs_error(struct super_block *s, const char *id,
3194 const char *function, const char *fmt, ...);
3195 #define reiserfs_error(s, id, fmt, args...) \
3196 __reiserfs_error(s, id, __func__, fmt, ##args)
3197 void reiserfs_info(struct super_block *s, const char *fmt, ...);
3198 void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
3199 void print_indirect_item(struct buffer_head *bh, int item_num);
3200 void store_print_tb(struct tree_balance *tb);
3201 void print_cur_tb(char *mes);
3202 void print_de(struct reiserfs_dir_entry *de);
3203 void print_bi(struct buffer_info *bi, char *mes);
3204 #define PRINT_LEAF_ITEMS 1 /* print all items */
3205 #define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
3206 #define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
3207 void print_block(struct buffer_head *bh, ...);
3208 void print_bmap(struct super_block *s, int silent);
3209 void print_bmap_block(int i, char *data, int size, int silent);
3210 /*void print_super_block (struct super_block * s, char * mes);*/
3211 void print_objectid_map(struct super_block *s);
3212 void print_block_head(struct buffer_head *bh, char *mes);
3213 void check_leaf(struct buffer_head *bh);
3214 void check_internal(struct buffer_head *bh);
3215 void print_statistics(struct super_block *s);
3216 char *reiserfs_hashname(int code);
3219 int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
3220 int mov_bytes, struct buffer_head *Snew);
3221 int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
3222 int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
3223 void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
3224 int del_num, int del_bytes);
3225 void leaf_insert_into_buf(struct buffer_info *bi, int before,
3226 struct item_head * const inserted_item_ih,
3227 const char * const inserted_item_body,
3229 void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
3230 int pos_in_item, int paste_size,
3231 const char * const body, int zeros_number);
3232 void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
3233 int pos_in_item, int cut_size);
3234 void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
3235 int new_entry_count, struct reiserfs_de_head *new_dehs,
3236 const char *records, int paste_size);
3238 int balance_internal(struct tree_balance *, int, int, struct item_head *,
3239 struct buffer_head **);
3242 void do_balance_mark_leaf_dirty(struct tree_balance *tb,
3243 struct buffer_head *bh, int flag);
3244 #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
3245 #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
3247 void do_balance(struct tree_balance *tb, struct item_head *ih,
3248 const char *body, int flag);
3249 void reiserfs_invalidate_buffer(struct tree_balance *tb,
3250 struct buffer_head *bh);
3252 int get_left_neighbor_position(struct tree_balance *tb, int h);
3253 int get_right_neighbor_position(struct tree_balance *tb, int h);
3254 void replace_key(struct tree_balance *tb, struct buffer_head *, int,
3255 struct buffer_head *, int);
3256 void make_empty_node(struct buffer_info *);
3257 struct buffer_head *get_FEB(struct tree_balance *);
3262 * structure contains hints for block allocator, and it is a container for
3263 * arguments, such as node, search path, transaction_handle, etc.
3265 struct __reiserfs_blocknr_hint {
3266 /* inode passed to allocator, if we allocate unf. nodes */
3267 struct inode *inode;
3269 sector_t block; /* file offset, in blocks */
3270 struct in_core_key key;
3273 * search path, used by allocator to deternine search_start by
3276 struct treepath *path;
3279 * transaction handle is needed to log super blocks
3280 * and bitmap blocks changes
3282 struct reiserfs_transaction_handle *th;
3284 b_blocknr_t beg, end;
3287 * a field used to transfer search start value (block number)
3288 * between different block allocator procedures
3289 * (determine_search_start() and others)
3291 b_blocknr_t search_start;
3294 * is set in determine_prealloc_size() function,
3295 * used by underlayed function that do actual allocation
3300 * the allocator uses different polices for getting disk
3301 * space for formatted/unformatted blocks with/without preallocation
3303 unsigned formatted_node:1;
3304 unsigned preallocate:1;
3307 typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
3309 int reiserfs_parse_alloc_options(struct super_block *, char *);
3310 void reiserfs_init_alloc_options(struct super_block *s);
3313 * given a directory, this will tell you what packing locality
3314 * to use for a new object underneat it. The locality is returned
3315 * in disk byte order (le).
3317 __le32 reiserfs_choose_packing(struct inode *dir);
3319 void show_alloc_options(struct seq_file *seq, struct super_block *s);
3320 int reiserfs_init_bitmap_cache(struct super_block *sb);
3321 void reiserfs_free_bitmap_cache(struct super_block *sb);
3322 void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
3323 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
3324 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
3325 void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
3326 b_blocknr_t, int for_unformatted);
3327 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
3329 static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
3330 b_blocknr_t * new_blocknrs,
3333 reiserfs_blocknr_hint_t hint = {
3334 .th = tb->transaction_handle,
3335 .path = tb->tb_path,
3341 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
3345 static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
3346 *th, struct inode *inode,
3347 b_blocknr_t * new_blocknrs,
3348 struct treepath *path,
3351 reiserfs_blocknr_hint_t hint = {
3356 .formatted_node = 0,
3359 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3362 #ifdef REISERFS_PREALLOCATE
3363 static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
3364 *th, struct inode *inode,
3365 b_blocknr_t * new_blocknrs,
3366 struct treepath *path,
3369 reiserfs_blocknr_hint_t hint = {
3374 .formatted_node = 0,
3377 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3380 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
3381 struct inode *inode);
3382 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
3386 __u32 keyed_hash(const signed char *msg, int len);
3387 __u32 yura_hash(const signed char *msg, int len);
3388 __u32 r5_hash(const signed char *msg, int len);
3390 #define reiserfs_set_le_bit __set_bit_le
3391 #define reiserfs_test_and_set_le_bit __test_and_set_bit_le
3392 #define reiserfs_clear_le_bit __clear_bit_le
3393 #define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le
3394 #define reiserfs_test_le_bit test_bit_le
3395 #define reiserfs_find_next_zero_le_bit find_next_zero_bit_le
3398 * sometimes reiserfs_truncate may require to allocate few new blocks
3399 * to perform indirect2direct conversion. People probably used to
3400 * think, that truncate should work without problems on a filesystem
3401 * without free disk space. They may complain that they can not
3402 * truncate due to lack of free disk space. This spare space allows us
3403 * to not worry about it. 500 is probably too much, but it should be
3406 #define SPARE_SPACE 500
3408 /* prototypes from ioctl.c */
3409 long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
3410 long reiserfs_compat_ioctl(struct file *filp,
3411 unsigned int cmd, unsigned long arg);
3412 int reiserfs_unpack(struct inode *inode, struct file *filp);
git.samba.org / sfrench / cifs-2.6.git / blob