2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
5 #include <linux/reiserfs_fs.h>
7 #include <linux/slab.h>
8 #include <linux/interrupt.h>
9 #include <linux/sched.h>
10 #include <linux/bug.h>
11 #include <linux/workqueue.h>
12 #include <asm/unaligned.h>
13 #include <linux/bitops.h>
14 #include <linux/proc_fs.h>
15 #include <linux/buffer_head.h>
17 /* the 32 bit compat definitions with int argument */
18 #define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
19 #define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
20 #define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
21 #define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
22 #define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
24 struct reiserfs_journal_list;
26 /** bitmasks for i_flags field in reiserfs-specific part of inode */
28 /** this says what format of key do all items (but stat data) of
29 an object have. If this is set, that format is 3.6 otherwise
31 i_item_key_version_mask = 0x0001,
32 /** If this is unset, object has 3.5 stat data, otherwise, it has
33 3.6 stat data with 64bit size, 32bit nlink etc. */
34 i_stat_data_version_mask = 0x0002,
35 /** file might need tail packing on close */
36 i_pack_on_close_mask = 0x0004,
37 /** don't pack tail of file */
38 i_nopack_mask = 0x0008,
39 /** If those is set, "safe link" was created for this file during
40 truncate or unlink. Safe link is used to avoid leakage of disk
41 space on crash with some files open, but unlinked. */
42 i_link_saved_unlink_mask = 0x0010,
43 i_link_saved_truncate_mask = 0x0020,
44 i_has_xattr_dir = 0x0040,
46 } reiserfs_inode_flags;
48 struct reiserfs_inode_info {
49 __u32 i_key[4]; /* key is still 4 32 bit integers */
50 /** transient inode flags that are never stored on disk. Bitmasks
51 for this field are defined above. */
54 __u32 i_first_direct_byte; // offset of first byte stored in direct item.
56 /* copy of persistent inode flags read from sd_attrs. */
59 int i_prealloc_block; /* first unused block of a sequence of unused blocks */
60 int i_prealloc_count; /* length of that sequence */
61 struct list_head i_prealloc_list; /* per-transaction list of inodes which
62 * have preallocated blocks */
64 unsigned new_packing_locality:1; /* new_packig_locality is created; new blocks
65 * for the contents of this directory should be
68 /* we use these for fsync or O_SYNC to decide which transaction
69 ** needs to be committed in order for this inode to be properly
71 unsigned int i_trans_id;
72 struct reiserfs_journal_list *i_jl;
74 struct mutex tailpack;
75 #ifdef CONFIG_REISERFS_FS_XATTR
76 struct rw_semaphore i_xattr_sem;
78 struct inode vfs_inode;
82 reiserfs_attrs_cleared = 0x00000001,
83 } reiserfs_super_block_flags;
85 /* struct reiserfs_super_block accessors/mutators
86 * since this is a disk structure, it will always be in
87 * little endian format. */
88 #define sb_block_count(sbp) (le32_to_cpu((sbp)->s_v1.s_block_count))
89 #define set_sb_block_count(sbp,v) ((sbp)->s_v1.s_block_count = cpu_to_le32(v))
90 #define sb_free_blocks(sbp) (le32_to_cpu((sbp)->s_v1.s_free_blocks))
91 #define set_sb_free_blocks(sbp,v) ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))
92 #define sb_root_block(sbp) (le32_to_cpu((sbp)->s_v1.s_root_block))
93 #define set_sb_root_block(sbp,v) ((sbp)->s_v1.s_root_block = cpu_to_le32(v))
95 #define sb_jp_journal_1st_block(sbp) \
96 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))
97 #define set_sb_jp_journal_1st_block(sbp,v) \
98 ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))
99 #define sb_jp_journal_dev(sbp) \
100 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))
101 #define set_sb_jp_journal_dev(sbp,v) \
102 ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))
103 #define sb_jp_journal_size(sbp) \
104 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))
105 #define set_sb_jp_journal_size(sbp,v) \
106 ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))
107 #define sb_jp_journal_trans_max(sbp) \
108 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))
109 #define set_sb_jp_journal_trans_max(sbp,v) \
110 ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))
111 #define sb_jp_journal_magic(sbp) \
112 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))
113 #define set_sb_jp_journal_magic(sbp,v) \
114 ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))
115 #define sb_jp_journal_max_batch(sbp) \
116 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))
117 #define set_sb_jp_journal_max_batch(sbp,v) \
118 ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))
119 #define sb_jp_jourmal_max_commit_age(sbp) \
120 (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))
121 #define set_sb_jp_journal_max_commit_age(sbp,v) \
122 ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))
124 #define sb_blocksize(sbp) (le16_to_cpu((sbp)->s_v1.s_blocksize))
125 #define set_sb_blocksize(sbp,v) ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))
126 #define sb_oid_maxsize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))
127 #define set_sb_oid_maxsize(sbp,v) ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))
128 #define sb_oid_cursize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_cursize))
129 #define set_sb_oid_cursize(sbp,v) ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))
130 #define sb_umount_state(sbp) (le16_to_cpu((sbp)->s_v1.s_umount_state))
131 #define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))
132 #define sb_fs_state(sbp) (le16_to_cpu((sbp)->s_v1.s_fs_state))
133 #define set_sb_fs_state(sbp,v) ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))
134 #define sb_hash_function_code(sbp) \
135 (le32_to_cpu((sbp)->s_v1.s_hash_function_code))
136 #define set_sb_hash_function_code(sbp,v) \
137 ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))
138 #define sb_tree_height(sbp) (le16_to_cpu((sbp)->s_v1.s_tree_height))
139 #define set_sb_tree_height(sbp,v) ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))
140 #define sb_bmap_nr(sbp) (le16_to_cpu((sbp)->s_v1.s_bmap_nr))
141 #define set_sb_bmap_nr(sbp,v) ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))
142 #define sb_version(sbp) (le16_to_cpu((sbp)->s_v1.s_version))
143 #define set_sb_version(sbp,v) ((sbp)->s_v1.s_version = cpu_to_le16(v))
145 #define sb_mnt_count(sbp) (le16_to_cpu((sbp)->s_mnt_count))
146 #define set_sb_mnt_count(sbp, v) ((sbp)->s_mnt_count = cpu_to_le16(v))
148 #define sb_reserved_for_journal(sbp) \
149 (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))
150 #define set_sb_reserved_for_journal(sbp,v) \
151 ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))
155 /* These all interelate for performance.
157 ** If the journal block count is smaller than n transactions, you lose speed.
158 ** I don't know what n is yet, I'm guessing 8-16.
160 ** typical transaction size depends on the application, how often fsync is
161 ** called, and how many metadata blocks you dirty in a 30 second period.
162 ** The more small files (<16k) you use, the larger your transactions will
165 ** If your journal fills faster than dirty buffers get flushed to disk, it must flush them before allowing the journal
166 ** to wrap, which slows things down. If you need high speed meta data updates, the journal should be big enough
167 ** to prevent wrapping before dirty meta blocks get to disk.
169 ** If the batch max is smaller than the transaction max, you'll waste space at the end of the journal
170 ** because journal_end sets the next transaction to start at 0 if the next transaction has any chance of wrapping.
172 ** The large the batch max age, the better the speed, and the more meta data changes you'll lose after a crash.
176 /* don't mess with these for a while */
177 /* we have a node size define somewhere in reiserfs_fs.h. -Hans */
178 #define JOURNAL_BLOCK_SIZE 4096 /* BUG gotta get rid of this */
179 #define JOURNAL_MAX_CNODE 1500 /* max cnodes to allocate. */
180 #define JOURNAL_HASH_SIZE 8192
181 #define JOURNAL_NUM_BITMAPS 5 /* number of copies of the bitmaps to have floating. Must be >= 2 */
183 /* One of these for every block in every transaction
184 ** Each one is in two hash tables. First, a hash of the current transaction, and after journal_end, a
185 ** hash of all the in memory transactions.
186 ** next and prev are used by the current transaction (journal_hash).
187 ** hnext and hprev are used by journal_list_hash. If a block is in more than one transaction, the journal_list_hash
188 ** links it in multiple times. This allows flush_journal_list to remove just the cnode belonging
189 ** to a given transaction.
191 struct reiserfs_journal_cnode {
192 struct buffer_head *bh; /* real buffer head */
193 struct super_block *sb; /* dev of real buffer head */
194 __u32 blocknr; /* block number of real buffer head, == 0 when buffer on disk */
196 struct reiserfs_journal_list *jlist; /* journal list this cnode lives in */
197 struct reiserfs_journal_cnode *next; /* next in transaction list */
198 struct reiserfs_journal_cnode *prev; /* prev in transaction list */
199 struct reiserfs_journal_cnode *hprev; /* prev in hash list */
200 struct reiserfs_journal_cnode *hnext; /* next in hash list */
203 struct reiserfs_bitmap_node {
206 struct list_head list;
209 struct reiserfs_list_bitmap {
210 struct reiserfs_journal_list *journal_list;
211 struct reiserfs_bitmap_node **bitmaps;
215 ** one of these for each transaction. The most important part here is the j_realblock.
216 ** this list of cnodes is used to hash all the blocks in all the commits, to mark all the
217 ** real buffer heads dirty once all the commits hit the disk,
218 ** and to make sure every real block in a transaction is on disk before allowing the log area
219 ** to be overwritten */
220 struct reiserfs_journal_list {
221 unsigned long j_start;
222 unsigned long j_state;
224 atomic_t j_nonzerolen;
225 atomic_t j_commit_left;
226 atomic_t j_older_commits_done; /* all commits older than this on disk */
227 struct mutex j_commit_mutex;
228 unsigned int j_trans_id;
230 struct reiserfs_list_bitmap *j_list_bitmap;
231 struct buffer_head *j_commit_bh; /* commit buffer head */
232 struct reiserfs_journal_cnode *j_realblock;
233 struct reiserfs_journal_cnode *j_freedlist; /* list of buffers that were freed during this trans. free each of these on flush */
234 /* time ordered list of all active transactions */
235 struct list_head j_list;
237 /* time ordered list of all transactions we haven't tried to flush yet */
238 struct list_head j_working_list;
240 /* list of tail conversion targets in need of flush before commit */
241 struct list_head j_tail_bh_list;
242 /* list of data=ordered buffers in need of flush before commit */
243 struct list_head j_bh_list;
247 struct reiserfs_journal {
248 struct buffer_head **j_ap_blocks; /* journal blocks on disk */
249 struct reiserfs_journal_cnode *j_last; /* newest journal block */
250 struct reiserfs_journal_cnode *j_first; /* oldest journal block. start here for traverse */
252 struct block_device *j_dev_bd;
254 int j_1st_reserved_block; /* first block on s_dev of reserved area journal */
256 unsigned long j_state;
257 unsigned int j_trans_id;
258 unsigned long j_mount_id;
259 unsigned long j_start; /* start of current waiting commit (index into j_ap_blocks) */
260 unsigned long j_len; /* length of current waiting commit */
261 unsigned long j_len_alloc; /* number of buffers requested by journal_begin() */
262 atomic_t j_wcount; /* count of writers for current commit */
263 unsigned long j_bcount; /* batch count. allows turning X transactions into 1 */
264 unsigned long j_first_unflushed_offset; /* first unflushed transactions offset */
265 unsigned j_last_flush_trans_id; /* last fully flushed journal timestamp */
266 struct buffer_head *j_header_bh;
268 time_t j_trans_start_time; /* time this transaction started */
269 struct mutex j_mutex;
270 struct mutex j_flush_mutex;
271 wait_queue_head_t j_join_wait; /* wait for current transaction to finish before starting new one */
272 atomic_t j_jlock; /* lock for j_join_wait */
273 int j_list_bitmap_index; /* number of next list bitmap to use */
274 int j_must_wait; /* no more journal begins allowed. MUST sleep on j_join_wait */
275 int j_next_full_flush; /* next journal_end will flush all journal list */
276 int j_next_async_flush; /* next journal_end will flush all async commits */
278 int j_cnode_used; /* number of cnodes on the used list */
279 int j_cnode_free; /* number of cnodes on the free list */
281 unsigned int j_trans_max; /* max number of blocks in a transaction. */
282 unsigned int j_max_batch; /* max number of blocks to batch into a trans */
283 unsigned int j_max_commit_age; /* in seconds, how old can an async commit be */
284 unsigned int j_max_trans_age; /* in seconds, how old can a transaction be */
285 unsigned int j_default_max_commit_age; /* the default for the max commit age */
287 struct reiserfs_journal_cnode *j_cnode_free_list;
288 struct reiserfs_journal_cnode *j_cnode_free_orig; /* orig pointer returned from vmalloc */
290 struct reiserfs_journal_list *j_current_jl;
291 int j_free_bitmap_nodes;
292 int j_used_bitmap_nodes;
294 int j_num_lists; /* total number of active transactions */
295 int j_num_work_lists; /* number that need attention from kreiserfsd */
297 /* debugging to make sure things are flushed in order */
298 unsigned int j_last_flush_id;
300 /* debugging to make sure things are committed in order */
301 unsigned int j_last_commit_id;
303 struct list_head j_bitmap_nodes;
304 struct list_head j_dirty_buffers;
305 spinlock_t j_dirty_buffers_lock; /* protects j_dirty_buffers */
307 /* list of all active transactions */
308 struct list_head j_journal_list;
309 /* lists that haven't been touched by writeback attempts */
310 struct list_head j_working_list;
312 struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS]; /* array of bitmaps to record the deleted blocks */
313 struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE]; /* hash table for real buffer heads in current trans */
314 struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE]; /* hash table for all the real buffer heads in all
316 struct list_head j_prealloc_list; /* list of inodes which have preallocated blocks */
317 int j_persistent_trans;
318 unsigned long j_max_trans_size;
319 unsigned long j_max_batch_size;
323 /* when flushing ordered buffers, throttle new ordered writers */
324 struct delayed_work j_work;
325 struct super_block *j_work_sb;
326 atomic_t j_async_throttle;
329 enum journal_state_bits {
330 J_WRITERS_BLOCKED = 1, /* set when new writers not allowed */
331 J_WRITERS_QUEUED, /* set when log is full due to too many writers */
332 J_ABORTED, /* set when log is aborted */
335 #define JOURNAL_DESC_MAGIC "ReIsErLB" /* ick. magic string to find desc blocks in the journal */
337 typedef __u32(*hashf_t) (const signed char *, int);
339 struct reiserfs_bitmap_info {
343 struct proc_dir_entry;
345 #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
346 typedef unsigned long int stat_cnt_t;
347 typedef struct reiserfs_proc_info_data {
350 int max_hash_collisions;
353 stat_cnt_t bread_miss;
354 stat_cnt_t search_by_key;
355 stat_cnt_t search_by_key_fs_changed;
356 stat_cnt_t search_by_key_restarted;
358 stat_cnt_t insert_item_restarted;
359 stat_cnt_t paste_into_item_restarted;
360 stat_cnt_t cut_from_item_restarted;
361 stat_cnt_t delete_solid_item_restarted;
362 stat_cnt_t delete_item_restarted;
364 stat_cnt_t leaked_oid;
365 stat_cnt_t leaves_removable;
367 /* balances per level. Use explicit 5 as MAX_HEIGHT is not visible yet. */
368 stat_cnt_t balance_at[5]; /* XXX */
369 /* sbk == search_by_key */
370 stat_cnt_t sbk_read_at[5]; /* XXX */
371 stat_cnt_t sbk_fs_changed[5];
372 stat_cnt_t sbk_restarted[5];
373 stat_cnt_t items_at[5]; /* XXX */
374 stat_cnt_t free_at[5]; /* XXX */
375 stat_cnt_t can_node_be_removed[5]; /* XXX */
376 long int lnum[5]; /* XXX */
377 long int rnum[5]; /* XXX */
378 long int lbytes[5]; /* XXX */
379 long int rbytes[5]; /* XXX */
380 stat_cnt_t get_neighbors[5];
381 stat_cnt_t get_neighbors_restart[5];
382 stat_cnt_t need_l_neighbor[5];
383 stat_cnt_t need_r_neighbor[5];
385 stat_cnt_t free_block;
386 struct __scan_bitmap_stats {
391 stat_cnt_t in_journal_hint;
392 stat_cnt_t in_journal_nohint;
395 struct __journal_stats {
396 stat_cnt_t in_journal;
397 stat_cnt_t in_journal_bitmap;
398 stat_cnt_t in_journal_reusable;
399 stat_cnt_t lock_journal;
400 stat_cnt_t lock_journal_wait;
401 stat_cnt_t journal_being;
402 stat_cnt_t journal_relock_writers;
403 stat_cnt_t journal_relock_wcount;
404 stat_cnt_t mark_dirty;
405 stat_cnt_t mark_dirty_already;
406 stat_cnt_t mark_dirty_notjournal;
407 stat_cnt_t restore_prepared;
409 stat_cnt_t prepare_retry;
411 } reiserfs_proc_info_data_t;
413 typedef struct reiserfs_proc_info_data {
414 } reiserfs_proc_info_data_t;
417 /* reiserfs union of in-core super block data */
418 struct reiserfs_sb_info {
419 struct buffer_head *s_sbh; /* Buffer containing the super block */
420 /* both the comment and the choice of
421 name are unclear for s_rs -Hans */
422 struct reiserfs_super_block *s_rs; /* Pointer to the super block in the buffer */
423 struct reiserfs_bitmap_info *s_ap_bitmap;
424 struct reiserfs_journal *s_journal; /* pointer to journal information */
425 unsigned short s_mount_state; /* reiserfs state (valid, invalid) */
427 /* Serialize writers access, replace the old bkl */
429 /* Owner of the lock (can be recursive) */
430 struct task_struct *lock_owner;
431 /* Depth of the lock, start from -1 like the bkl */
435 void (*end_io_handler) (struct buffer_head *, int);
436 hashf_t s_hash_function; /* pointer to function which is used
437 to sort names in directory. Set on
439 unsigned long s_mount_opt; /* reiserfs's mount options are set
440 here (currently - NOTAIL, NOLOG,
443 struct { /* This is a structure that describes block allocator options */
444 unsigned long bits; /* Bitfield for enable/disable kind of options */
445 unsigned long large_file_size; /* size started from which we consider file to be a large one(in blocks) */
446 int border; /* percentage of disk, border takes */
447 int preallocmin; /* Minimal file size (in blocks) starting from which we do preallocations */
448 int preallocsize; /* Number of blocks we try to prealloc when file
449 reaches preallocmin size (in blocks) or
450 prealloc_list is empty. */
454 wait_queue_head_t s_wait;
455 /* To be obsoleted soon by per buffer seals.. -Hans */
456 atomic_t s_generation_counter; // increased by one every time the
457 // tree gets re-balanced
458 unsigned long s_properties; /* File system properties. Currently holds
461 /* session statistics */
466 int s_unneeded_left_neighbor;
467 int s_good_search_by_key_reada;
469 int s_bmaps_without_search;
470 int s_direct2indirect;
471 int s_indirect2direct;
472 /* set up when it's ok for reiserfs_read_inode2() to read from
473 disk inode with nlink==0. Currently this is only used during
474 finish_unfinished() processing at mount time */
475 int s_is_unlinked_ok;
476 reiserfs_proc_info_data_t s_proc_info_data;
477 struct proc_dir_entry *procdir;
478 int reserved_blocks; /* amount of blocks reserved for further allocations */
479 spinlock_t bitmap_lock; /* this lock on now only used to protect reserved_blocks variable */
480 struct dentry *priv_root; /* root of /.reiserfs_priv */
481 struct dentry *xattr_root; /* root of /.reiserfs_priv/xattrs */
484 int work_queued; /* non-zero delayed work is queued */
485 struct delayed_work old_work; /* old transactions flush delayed work */
486 spinlock_t old_work_lock; /* protects old_work and work_queued */
489 char *s_qf_names[MAXQUOTAS];
492 char *s_jdev; /* Stored jdev for mount option showing */
493 #ifdef CONFIG_REISERFS_CHECK
495 struct tree_balance *cur_tb; /*
496 * Detects whether more than one
497 * copy of tb exists per superblock
498 * as a means of checking whether
499 * do_balance is executing concurrently
500 * against another tree reader/writer
501 * on a same mount point.
506 /* Definitions of reiserfs on-disk properties: */
507 #define REISERFS_3_5 0
508 #define REISERFS_3_6 1
509 #define REISERFS_OLD_FORMAT 2
511 enum reiserfs_mount_options {
513 REISERFS_LARGETAIL, /* large tails will be created in a session */
514 REISERFS_SMALLTAIL, /* small (for files less than block size) tails will be created in a session */
515 REPLAYONLY, /* replay journal and return 0. Use by fsck */
516 REISERFS_CONVERT, /* -o conv: causes conversion of old
517 format super block to the new
518 format. If not specified - old
519 partition will be dealt with in a
522 /* -o hash={tea, rupasov, r5, detect} is meant for properly mounting
523 ** reiserfs disks from 3.5.19 or earlier. 99% of the time, this option
524 ** is not required. If the normal autodection code can't determine which
525 ** hash to use (because both hashes had the same value for a file)
526 ** use this option to force a specific hash. It won't allow you to override
527 ** the existing hash on the FS, so if you have a tea hash disk, and mount
528 ** with -o hash=rupasov, the mount will fail.
530 FORCE_TEA_HASH, /* try to force tea hash on mount */
531 FORCE_RUPASOV_HASH, /* try to force rupasov hash on mount */
532 FORCE_R5_HASH, /* try to force rupasov hash on mount */
533 FORCE_HASH_DETECT, /* try to detect hash function on mount */
536 REISERFS_DATA_ORDERED,
537 REISERFS_DATA_WRITEBACK,
539 /* used for testing experimental features, makes benchmarking new
540 features with and without more convenient, should never be used by
541 users in any code shipped to users (ideally) */
544 REISERFS_NO_UNHASHED_RELOCATION,
545 REISERFS_HASHED_RELOCATION,
547 REISERFS_XATTRS_USER,
549 REISERFS_EXPOSE_PRIVROOT,
550 REISERFS_BARRIER_NONE,
551 REISERFS_BARRIER_FLUSH,
553 /* Actions on error */
554 REISERFS_ERROR_PANIC,
556 REISERFS_ERROR_CONTINUE,
558 REISERFS_USRQUOTA, /* User quota option specified */
559 REISERFS_GRPQUOTA, /* Group quota option specified */
565 REISERFS_UNSUPPORTED_OPT,
568 #define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))
569 #define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))
570 #define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))
571 #define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))
572 #define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))
573 #define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))
574 #define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))
575 #define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))
577 #define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))
578 #define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))
579 #define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))
580 #define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))
581 #define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))
582 #define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))
583 #define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))
584 #define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))
585 #define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))
586 #define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))
587 #define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))
588 #define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))
589 #define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))
590 #define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))
591 #define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))
593 #define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))
594 #define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))
596 void reiserfs_file_buffer(struct buffer_head *bh, int list);
597 extern struct file_system_type reiserfs_fs_type;
598 int reiserfs_resize(struct super_block *, unsigned long);
601 #define SCHEDULE_OCCURRED 1
603 #define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)
604 #define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)
605 #define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)
606 #define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)
607 #define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)
609 #define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)
611 /* A safe version of the "bdevname", which returns the "s_id" field of
612 * a superblock or else "Null superblock" if the super block is NULL.
614 static inline char *reiserfs_bdevname(struct super_block *s)
616 return (s == NULL) ? "Null superblock" : s->s_id;
619 #define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
620 static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
623 return test_bit(J_ABORTED, &journal->j_state);
627 * Locking primitives. The write lock is a per superblock
628 * special mutex that has properties close to the Big Kernel Lock
629 * which was used in the previous locking scheme.
631 void reiserfs_write_lock(struct super_block *s);
632 void reiserfs_write_unlock(struct super_block *s);
633 int __must_check reiserfs_write_unlock_nested(struct super_block *s);
634 void reiserfs_write_lock_nested(struct super_block *s, int depth);
636 #ifdef CONFIG_REISERFS_CHECK
637 void reiserfs_lock_check_recursive(struct super_block *s);
639 static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
643 * Several mutexes depend on the write lock.
644 * However sometimes we want to relax the write lock while we hold
645 * these mutexes, according to the release/reacquire on schedule()
646 * properties of the Bkl that were used.
647 * Reiserfs performances and locking were based on this scheme.
648 * Now that the write lock is a mutex and not the bkl anymore, doing so
649 * may result in a deadlock:
651 * A acquire write_lock
652 * A acquire j_commit_mutex
653 * A release write_lock and wait for something
654 * B acquire write_lock
655 * B can't acquire j_commit_mutex and sleep
656 * A can't acquire write lock anymore
659 * What we do here is avoiding such deadlock by playing the same game
660 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
661 * we release the write lock, wait a bit and then retry.
663 * The mutexes concerned by this hack are:
664 * - The commit mutex of a journal list
669 static inline void reiserfs_mutex_lock_safe(struct mutex *m,
670 struct super_block *s)
674 depth = reiserfs_write_unlock_nested(s);
676 reiserfs_write_lock_nested(s, depth);
680 reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
681 struct super_block *s)
685 depth = reiserfs_write_unlock_nested(s);
686 mutex_lock_nested(m, subclass);
687 reiserfs_write_lock_nested(s, depth);
691 reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
694 depth = reiserfs_write_unlock_nested(s);
696 reiserfs_write_lock_nested(s, depth);
700 * When we schedule, we usually want to also release the write lock,
701 * according to the previous bkl based locking scheme of reiserfs.
703 static inline void reiserfs_cond_resched(struct super_block *s)
705 if (need_resched()) {
708 depth = reiserfs_write_unlock_nested(s);
710 reiserfs_write_lock_nested(s, depth);
716 /* in reading the #defines, it may help to understand that they employ
717 the following abbreviations:
721 H = Height within the tree (should be changed to LEV)
722 N = Number of the item in the node
724 DEH = Directory Entry Header
729 UNFM = UNForMatted node
733 These #defines are named by concatenating these abbreviations,
734 where first comes the arguments, and last comes the return value,
739 #define USE_INODE_GENERATION_COUNTER
741 #define REISERFS_PREALLOCATE
742 #define DISPLACE_NEW_PACKING_LOCALITIES
743 #define PREALLOCATION_SIZE 9
745 /* n must be power of 2 */
746 #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
748 // to be ok for alpha and others we have to align structures to 8 byte
750 // FIXME: do not change 4 by anything else: there is code which relies on that
751 #define ROUND_UP(x) _ROUND_UP(x,8LL)
753 /* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
756 #define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
758 void __reiserfs_warning(struct super_block *s, const char *id,
759 const char *func, const char *fmt, ...);
760 #define reiserfs_warning(s, id, fmt, args...) \
761 __reiserfs_warning(s, id, __func__, fmt, ##args)
762 /* assertions handling */
764 /** always check a condition and panic if it's false. */
765 #define __RASSERT(cond, scond, format, args...) \
768 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
769 __FILE__ ":%i:%s: " format "\n", \
770 in_interrupt() ? -1 : task_pid_nr(current), \
771 __LINE__, __func__ , ##args); \
774 #define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
776 #if defined( CONFIG_REISERFS_CHECK )
777 #define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
779 #define RFALSE( cond, format, args... ) do {;} while( 0 )
782 #define CONSTF __attribute_const__
784 * Disk Data Structures
787 /***************************************************************************/
789 /***************************************************************************/
792 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
793 * the version in RAM is part of a larger structure containing fields never written to disk.
795 #define UNSET_HASH 0 // read_super will guess about, what hash names
796 // in directories were sorted with
800 #define DEFAULT_HASH R5_HASH
802 struct journal_params {
803 __le32 jp_journal_1st_block; /* where does journal start from on its
805 __le32 jp_journal_dev; /* journal device st_rdev */
806 __le32 jp_journal_size; /* size of the journal */
807 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
808 __le32 jp_journal_magic; /* random value made on fs creation (this
809 * was sb_journal_block_count) */
810 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
812 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
814 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
818 /* this is the super from 3.5.X, where X >= 10 */
819 struct reiserfs_super_block_v1 {
820 __le32 s_block_count; /* blocks count */
821 __le32 s_free_blocks; /* free blocks count */
822 __le32 s_root_block; /* root block number */
823 struct journal_params s_journal;
824 __le16 s_blocksize; /* block size */
825 __le16 s_oid_maxsize; /* max size of object id array, see
826 * get_objectid() commentary */
827 __le16 s_oid_cursize; /* current size of object id array */
828 __le16 s_umount_state; /* this is set to 1 when filesystem was
829 * umounted, to 2 - when not */
830 char s_magic[10]; /* reiserfs magic string indicates that
831 * file system is reiserfs:
832 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
833 __le16 s_fs_state; /* it is set to used by fsck to mark which
834 * phase of rebuilding is done */
835 __le32 s_hash_function_code; /* indicate, what hash function is being use
836 * to sort names in a directory*/
837 __le16 s_tree_height; /* height of disk tree */
838 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
839 * each block of file system */
840 __le16 s_version; /* this field is only reliable on filesystem
841 * with non-standard journal */
842 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
843 * device, we need to keep after
844 * making fs with non-standard journal */
845 } __attribute__ ((__packed__));
847 #define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
849 /* this is the on disk super block */
850 struct reiserfs_super_block {
851 struct reiserfs_super_block_v1 s_v1;
852 __le32 s_inode_generation;
853 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
854 unsigned char s_uuid[16]; /* filesystem unique identifier */
855 unsigned char s_label[16]; /* filesystem volume label */
856 __le16 s_mnt_count; /* Count of mounts since last fsck */
857 __le16 s_max_mnt_count; /* Maximum mounts before check */
858 __le32 s_lastcheck; /* Timestamp of last fsck */
859 __le32 s_check_interval; /* Interval between checks */
860 char s_unused[76]; /* zero filled by mkreiserfs and
861 * reiserfs_convert_objectid_map_v1()
862 * so any additions must be updated
864 } __attribute__ ((__packed__));
866 #define SB_SIZE (sizeof(struct reiserfs_super_block))
868 #define REISERFS_VERSION_1 0
869 #define REISERFS_VERSION_2 2
871 // on-disk super block fields converted to cpu form
872 #define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
873 #define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
874 #define SB_BLOCKSIZE(s) \
875 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
876 #define SB_BLOCK_COUNT(s) \
877 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
878 #define SB_FREE_BLOCKS(s) \
879 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
880 #define SB_REISERFS_MAGIC(s) \
881 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
882 #define SB_ROOT_BLOCK(s) \
883 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
884 #define SB_TREE_HEIGHT(s) \
885 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
886 #define SB_REISERFS_STATE(s) \
887 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
888 #define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
889 #define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
891 #define PUT_SB_BLOCK_COUNT(s, val) \
892 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
893 #define PUT_SB_FREE_BLOCKS(s, val) \
894 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
895 #define PUT_SB_ROOT_BLOCK(s, val) \
896 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
897 #define PUT_SB_TREE_HEIGHT(s, val) \
898 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
899 #define PUT_SB_REISERFS_STATE(s, val) \
900 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
901 #define PUT_SB_VERSION(s, val) \
902 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
903 #define PUT_SB_BMAP_NR(s, val) \
904 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
906 #define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
907 #define SB_ONDISK_JOURNAL_SIZE(s) \
908 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
909 #define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
910 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
911 #define SB_ONDISK_JOURNAL_DEVICE(s) \
912 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
913 #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
914 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
916 #define is_block_in_log_or_reserved_area(s, block) \
917 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
918 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
919 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
920 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
922 int is_reiserfs_3_5(struct reiserfs_super_block *rs);
923 int is_reiserfs_3_6(struct reiserfs_super_block *rs);
924 int is_reiserfs_jr(struct reiserfs_super_block *rs);
926 /* ReiserFS leaves the first 64k unused, so that partition labels have
927 enough space. If someone wants to write a fancy bootloader that
928 needs more than 64k, let us know, and this will be increased in size.
929 This number must be larger than than the largest block size on any
930 platform, or code will break. -Hans */
931 #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
932 #define REISERFS_FIRST_BLOCK unused_define
933 #define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
935 /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
936 #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
938 /* reiserfs internal error code (used by search_by_key and fix_nodes)) */
940 #define REPEAT_SEARCH -1
942 #define NO_DISK_SPACE -3
943 #define NO_BALANCING_NEEDED (-4)
944 #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
945 #define QUOTA_EXCEEDED -6
947 typedef __u32 b_blocknr_t;
948 typedef __le32 unp_t;
950 struct unfm_nodeinfo {
952 unsigned short unfm_freespace;
955 /* there are two formats of keys: 3.5 and 3.6
957 #define KEY_FORMAT_3_5 0
958 #define KEY_FORMAT_3_6 1
960 /* there are two stat datas */
961 #define STAT_DATA_V1 0
962 #define STAT_DATA_V2 1
964 static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
966 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
969 static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
971 return sb->s_fs_info;
974 /* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
975 * which overflows on large file systems. */
976 static inline __u32 reiserfs_bmap_count(struct super_block *sb)
978 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
981 static inline int bmap_would_wrap(unsigned bmap_nr)
983 return bmap_nr > ((1LL << 16) - 1);
986 /** this says about version of key of all items (but stat data) the
987 object consists of */
988 #define get_inode_item_key_version( inode ) \
989 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
991 #define set_inode_item_key_version( inode, version ) \
992 ({ if((version)==KEY_FORMAT_3_6) \
993 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
995 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
997 #define get_inode_sd_version(inode) \
998 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
1000 #define set_inode_sd_version(inode, version) \
1001 ({ if((version)==STAT_DATA_V2) \
1002 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
1004 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1006 /* This is an aggressive tail suppression policy, I am hoping it
1007 improves our benchmarks. The principle behind it is that percentage
1008 space saving is what matters, not absolute space saving. This is
1009 non-intuitive, but it helps to understand it if you consider that the
1010 cost to access 4 blocks is not much more than the cost to access 1
1011 block, if you have to do a seek and rotate. A tail risks a
1012 non-linear disk access that is significant as a percentage of total
1013 time cost for a 4 block file and saves an amount of space that is
1014 less significant as a percentage of space, or so goes the hypothesis.
1016 #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1018 (!(n_tail_size)) || \
1019 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1020 ( (n_file_size) >= (n_block_size) * 4 ) || \
1021 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1022 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1023 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1024 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1025 ( ( (n_file_size) >= (n_block_size) ) && \
1026 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1029 /* Another strategy for tails, this one means only create a tail if all the
1030 file would fit into one DIRECT item.
1031 Primary intention for this one is to increase performance by decreasing
1034 #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1036 (!(n_tail_size)) || \
1037 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1041 * values for s_umount_state field
1043 #define REISERFS_VALID_FS 1
1044 #define REISERFS_ERROR_FS 2
1047 // there are 5 item types currently
1049 #define TYPE_STAT_DATA 0
1050 #define TYPE_INDIRECT 1
1051 #define TYPE_DIRECT 2
1052 #define TYPE_DIRENTRY 3
1053 #define TYPE_MAXTYPE 3
1054 #define TYPE_ANY 15 // FIXME: comment is required
1056 /***************************************************************************/
1057 /* KEY & ITEM HEAD */
1058 /***************************************************************************/
1061 // directories use this key as well as old files
1065 __le32 k_uniqueness;
1066 } __attribute__ ((__packed__));
1070 } __attribute__ ((__packed__));
1072 static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1074 __u8 type = le64_to_cpu(v2->v) >> 60;
1075 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1078 static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1081 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1084 static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1086 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1089 static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1091 offset &= (~0ULL >> 4);
1092 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1095 /* Key of an item determines its location in the S+tree, and
1096 is composed of 4 components */
1097 struct reiserfs_key {
1098 __le32 k_dir_id; /* packing locality: by default parent
1099 directory object id */
1100 __le32 k_objectid; /* object identifier */
1102 struct offset_v1 k_offset_v1;
1103 struct offset_v2 k_offset_v2;
1104 } __attribute__ ((__packed__)) u;
1105 } __attribute__ ((__packed__));
1107 struct in_core_key {
1108 __u32 k_dir_id; /* packing locality: by default parent
1109 directory object id */
1110 __u32 k_objectid; /* object identifier */
1116 struct in_core_key on_disk_key;
1118 int key_length; /* 3 in all cases but direct2indirect and
1119 indirect2direct conversion */
1122 /* Our function for comparing keys can compare keys of different
1123 lengths. It takes as a parameter the length of the keys it is to
1124 compare. These defines are used in determining what is to be passed
1125 to it as that parameter. */
1126 #define REISERFS_FULL_KEY_LEN 4
1127 #define REISERFS_SHORT_KEY_LEN 2
1129 /* The result of the key compare */
1130 #define FIRST_GREATER 1
1131 #define SECOND_GREATER -1
1132 #define KEYS_IDENTICAL 0
1134 #define KEY_NOT_FOUND 0
1136 #define KEY_SIZE (sizeof(struct reiserfs_key))
1137 #define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
1139 /* return values for search_by_key and clones */
1140 #define ITEM_FOUND 1
1141 #define ITEM_NOT_FOUND 0
1142 #define ENTRY_FOUND 1
1143 #define ENTRY_NOT_FOUND 0
1144 #define DIRECTORY_NOT_FOUND -1
1145 #define REGULAR_FILE_FOUND -2
1146 #define DIRECTORY_FOUND -3
1147 #define BYTE_FOUND 1
1148 #define BYTE_NOT_FOUND 0
1149 #define FILE_NOT_FOUND -1
1151 #define POSITION_FOUND 1
1152 #define POSITION_NOT_FOUND 0
1154 // return values for reiserfs_find_entry and search_by_entry_key
1155 #define NAME_FOUND 1
1156 #define NAME_NOT_FOUND 0
1157 #define GOTO_PREVIOUS_ITEM 2
1158 #define NAME_FOUND_INVISIBLE 3
1160 /* Everything in the filesystem is stored as a set of items. The
1161 item head contains the key of the item, its free space (for
1162 indirect items) and specifies the location of the item itself
1163 within the block. */
1166 /* Everything in the tree is found by searching for it based on
1168 struct reiserfs_key ih_key;
1170 /* The free space in the last unformatted node of an
1171 indirect item if this is an indirect item. This
1172 equals 0xFFFF iff this is a direct item or stat data
1173 item. Note that the key, not this field, is used to
1174 determine the item type, and thus which field this
1176 __le16 ih_free_space_reserved;
1177 /* Iff this is a directory item, this field equals the
1178 number of directory entries in the directory item. */
1179 __le16 ih_entry_count;
1180 } __attribute__ ((__packed__)) u;
1181 __le16 ih_item_len; /* total size of the item body */
1182 __le16 ih_item_location; /* an offset to the item body
1183 * within the block */
1184 __le16 ih_version; /* 0 for all old items, 2 for new
1185 ones. Highest bit is set by fsck
1186 temporary, cleaned after all
1188 } __attribute__ ((__packed__));
1189 /* size of item header */
1190 #define IH_SIZE (sizeof(struct item_head))
1192 #define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
1193 #define ih_version(ih) le16_to_cpu((ih)->ih_version)
1194 #define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
1195 #define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
1196 #define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
1198 #define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1199 #define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1200 #define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1201 #define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1202 #define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1204 #define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1206 #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1207 #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1209 /* these operate on indirect items, where you've got an array of ints
1210 ** at a possibly unaligned location. These are a noop on ia32
1212 ** p is the array of __u32, i is the index into the array, v is the value
1215 #define get_block_num(p, i) get_unaligned_le32((p) + (i))
1216 #define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1219 // in old version uniqueness field shows key type
1221 #define V1_SD_UNIQUENESS 0
1222 #define V1_INDIRECT_UNIQUENESS 0xfffffffe
1223 #define V1_DIRECT_UNIQUENESS 0xffffffff
1224 #define V1_DIRENTRY_UNIQUENESS 500
1225 #define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
1228 // here are conversion routines
1230 static inline int uniqueness2type(__u32 uniqueness) CONSTF;
1231 static inline int uniqueness2type(__u32 uniqueness)
1233 switch ((int)uniqueness) {
1234 case V1_SD_UNIQUENESS:
1235 return TYPE_STAT_DATA;
1236 case V1_INDIRECT_UNIQUENESS:
1237 return TYPE_INDIRECT;
1238 case V1_DIRECT_UNIQUENESS:
1240 case V1_DIRENTRY_UNIQUENESS:
1241 return TYPE_DIRENTRY;
1242 case V1_ANY_UNIQUENESS:
1248 static inline __u32 type2uniqueness(int type) CONSTF;
1249 static inline __u32 type2uniqueness(int type)
1252 case TYPE_STAT_DATA:
1253 return V1_SD_UNIQUENESS;
1255 return V1_INDIRECT_UNIQUENESS;
1257 return V1_DIRECT_UNIQUENESS;
1259 return V1_DIRENTRY_UNIQUENESS;
1262 return V1_ANY_UNIQUENESS;
1267 // key is pointer to on disk key which is stored in le, result is cpu,
1268 // there is no way to get version of object from key, so, provide
1269 // version to these defines
1271 static inline loff_t le_key_k_offset(int version,
1272 const struct reiserfs_key *key)
1274 return (version == KEY_FORMAT_3_5) ?
1275 le32_to_cpu(key->u.k_offset_v1.k_offset) :
1276 offset_v2_k_offset(&(key->u.k_offset_v2));
1279 static inline loff_t le_ih_k_offset(const struct item_head *ih)
1281 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1284 static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1286 return (version == KEY_FORMAT_3_5) ?
1287 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
1288 offset_v2_k_type(&(key->u.k_offset_v2));
1291 static inline loff_t le_ih_k_type(const struct item_head *ih)
1293 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1296 static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1299 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
1300 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
1303 static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1305 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1308 static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1311 (version == KEY_FORMAT_3_5) ?
1312 (void)(key->u.k_offset_v1.k_uniqueness =
1313 cpu_to_le32(type2uniqueness(type)))
1314 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
1317 static inline void set_le_ih_k_type(struct item_head *ih, int type)
1319 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1322 static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1324 return le_key_k_type(version, key) == TYPE_DIRENTRY;
1327 static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1329 return le_key_k_type(version, key) == TYPE_DIRECT;
1332 static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1334 return le_key_k_type(version, key) == TYPE_INDIRECT;
1337 static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1339 return le_key_k_type(version, key) == TYPE_STAT_DATA;
1343 // item header has version.
1345 static inline int is_direntry_le_ih(struct item_head *ih)
1347 return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1350 static inline int is_direct_le_ih(struct item_head *ih)
1352 return is_direct_le_key(ih_version(ih), &ih->ih_key);
1355 static inline int is_indirect_le_ih(struct item_head *ih)
1357 return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1360 static inline int is_statdata_le_ih(struct item_head *ih)
1362 return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1366 // key is pointer to cpu key, result is cpu
1368 static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1370 return key->on_disk_key.k_offset;
1373 static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1375 return key->on_disk_key.k_type;
1378 static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1380 key->on_disk_key.k_offset = offset;
1383 static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1385 key->on_disk_key.k_type = type;
1388 static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1390 key->on_disk_key.k_offset--;
1393 #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1394 #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1395 #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1396 #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1398 /* are these used ? */
1399 #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1400 #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1401 #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1402 #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1404 #define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1405 (!COMP_SHORT_KEYS(ih, key) && \
1406 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1408 /* maximal length of item */
1409 #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1410 #define MIN_ITEM_LEN 1
1412 /* object identifier for root dir */
1413 #define REISERFS_ROOT_OBJECTID 2
1414 #define REISERFS_ROOT_PARENT_OBJECTID 1
1416 extern struct reiserfs_key root_key;
1419 * Picture represents a leaf of the S+tree
1420 * ______________________________________________________
1421 * | | Array of | | |
1422 * |Block | Object-Item | F r e e | Objects- |
1423 * | head | Headers | S p a c e | Items |
1424 * |______|_______________|___________________|___________|
1427 /* Header of a disk block. More precisely, header of a formatted leaf
1428 or internal node, and not the header of an unformatted node. */
1430 __le16 blk_level; /* Level of a block in the tree. */
1431 __le16 blk_nr_item; /* Number of keys/items in a block. */
1432 __le16 blk_free_space; /* Block free space in bytes. */
1433 __le16 blk_reserved;
1434 /* dump this in v4/planA */
1435 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
1438 #define BLKH_SIZE (sizeof(struct block_head))
1439 #define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
1440 #define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
1441 #define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
1442 #define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
1443 #define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
1444 #define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1445 #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1446 #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1447 #define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
1448 #define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
1451 * values for blk_level field of the struct block_head
1454 #define FREE_LEVEL 0 /* when node gets removed from the tree its
1455 blk_level is set to FREE_LEVEL. It is then
1456 used to see whether the node is still in the
1459 #define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1461 /* Given the buffer head of a formatted node, resolve to the block head of that node. */
1462 #define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1463 /* Number of items that are in buffer. */
1464 #define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
1465 #define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
1466 #define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1468 #define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1469 #define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1470 #define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1472 /* Get right delimiting key. -- little endian */
1473 #define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1475 /* Does the buffer contain a disk leaf. */
1476 #define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1478 /* Does the buffer contain a disk internal node */
1479 #define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1480 && B_LEVEL(bh) <= MAX_HEIGHT)
1482 /***************************************************************************/
1484 /***************************************************************************/
1487 // old stat data is 32 bytes long. We are going to distinguish new one by
1490 struct stat_data_v1 {
1491 __le16 sd_mode; /* file type, permissions */
1492 __le16 sd_nlink; /* number of hard links */
1493 __le16 sd_uid; /* owner */
1494 __le16 sd_gid; /* group */
1495 __le32 sd_size; /* file size */
1496 __le32 sd_atime; /* time of last access */
1497 __le32 sd_mtime; /* time file was last modified */
1498 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
1501 __le32 sd_blocks; /* number of blocks file uses */
1502 } __attribute__ ((__packed__)) u;
1503 __le32 sd_first_direct_byte; /* first byte of file which is stored
1504 in a direct item: except that if it
1505 equals 1 it is a symlink and if it
1506 equals ~(__u32)0 there is no
1507 direct item. The existence of this
1508 field really grates on me. Let's
1509 replace it with a macro based on
1510 sd_size and our tail suppression
1511 policy. Someday. -Hans */
1512 } __attribute__ ((__packed__));
1514 #define SD_V1_SIZE (sizeof(struct stat_data_v1))
1515 #define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
1516 #define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1517 #define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1518 #define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
1519 #define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
1520 #define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
1521 #define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
1522 #define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
1523 #define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
1524 #define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
1525 #define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
1526 #define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1527 #define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1528 #define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1529 #define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1530 #define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1531 #define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1532 #define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1533 #define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1534 #define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
1535 #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1536 #define sd_v1_first_direct_byte(sdp) \
1537 (le32_to_cpu((sdp)->sd_first_direct_byte))
1538 #define set_sd_v1_first_direct_byte(sdp,v) \
1539 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1541 /* inode flags stored in sd_attrs (nee sd_reserved) */
1543 /* we want common flags to have the same values as in ext2,
1544 so chattr(1) will work without problems */
1545 #define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1546 #define REISERFS_APPEND_FL FS_APPEND_FL
1547 #define REISERFS_SYNC_FL FS_SYNC_FL
1548 #define REISERFS_NOATIME_FL FS_NOATIME_FL
1549 #define REISERFS_NODUMP_FL FS_NODUMP_FL
1550 #define REISERFS_SECRM_FL FS_SECRM_FL
1551 #define REISERFS_UNRM_FL FS_UNRM_FL
1552 #define REISERFS_COMPR_FL FS_COMPR_FL
1553 #define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1555 /* persistent flags that file inherits from the parent directory */
1556 #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
1557 REISERFS_SYNC_FL | \
1558 REISERFS_NOATIME_FL | \
1559 REISERFS_NODUMP_FL | \
1560 REISERFS_SECRM_FL | \
1561 REISERFS_COMPR_FL | \
1562 REISERFS_NOTAIL_FL )
1564 /* Stat Data on disk (reiserfs version of UFS disk inode minus the
1567 __le16 sd_mode; /* file type, permissions */
1568 __le16 sd_attrs; /* persistent inode flags */
1569 __le32 sd_nlink; /* number of hard links */
1570 __le64 sd_size; /* file size */
1571 __le32 sd_uid; /* owner */
1572 __le32 sd_gid; /* group */
1573 __le32 sd_atime; /* time of last access */
1574 __le32 sd_mtime; /* time file was last modified */
1575 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
1579 __le32 sd_generation;
1580 //__le32 sd_first_direct_byte;
1581 /* first byte of file which is stored in a
1582 direct item: except that if it equals 1
1583 it is a symlink and if it equals
1584 ~(__u32)0 there is no direct item. The
1585 existence of this field really grates
1586 on me. Let's replace it with a macro
1587 based on sd_size and our tail
1588 suppression policy? */
1589 } __attribute__ ((__packed__)) u;
1590 } __attribute__ ((__packed__));
1592 // this is 44 bytes long
1594 #define SD_SIZE (sizeof(struct stat_data))
1595 #define SD_V2_SIZE SD_SIZE
1596 #define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
1597 #define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
1598 #define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
1600 /* set_sd_reserved */
1601 #define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
1602 #define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
1603 #define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
1604 #define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
1605 #define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
1606 #define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
1607 #define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
1608 #define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
1609 #define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1610 #define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1611 #define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1612 #define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1613 #define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1614 #define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1615 #define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
1616 #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1617 #define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1618 #define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1619 #define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
1620 #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1621 #define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
1622 #define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
1624 /***************************************************************************/
1625 /* DIRECTORY STRUCTURE */
1626 /***************************************************************************/
1628 Picture represents the structure of directory items
1629 ________________________________________________
1630 | Array of | | | | | |
1631 | directory |N-1| N-2 | .... | 1st |0th|
1632 | entry headers | | | | | |
1633 |_______________|___|_____|________|_______|___|
1634 <---- directory entries ------>
1636 First directory item has k_offset component 1. We store "." and ".."
1637 in one item, always, we never split "." and ".." into differing
1638 items. This makes, among other things, the code for removing
1639 directories simpler. */
1641 #define SD_UNIQUENESS 0
1642 #define DOT_OFFSET 1
1643 #define DOT_DOT_OFFSET 2
1644 #define DIRENTRY_UNIQUENESS 500
1647 #define FIRST_ITEM_OFFSET 1
1650 Q: How to get key of object pointed to by entry from entry?
1652 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
1653 of object, entry points to */
1656 Directory will someday contain stat data of object */
1658 struct reiserfs_de_head {
1659 __le32 deh_offset; /* third component of the directory entry key */
1660 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
1661 by directory entry */
1662 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
1663 __le16 deh_location; /* offset of name in the whole item */
1664 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
1665 entry is hidden (unlinked) */
1666 } __attribute__ ((__packed__));
1667 #define DEH_SIZE sizeof(struct reiserfs_de_head)
1668 #define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
1669 #define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1670 #define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1671 #define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1672 #define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1674 #define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1675 #define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1676 #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1677 #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1678 #define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1680 /* empty directory contains two entries "." and ".." and their headers */
1681 #define EMPTY_DIR_SIZE \
1682 (DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1684 /* old format directories have this size when empty */
1685 #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1687 #define DEH_Statdata 0 /* not used now */
1688 #define DEH_Visible 2
1690 /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1691 #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1692 # define ADDR_UNALIGNED_BITS (3)
1695 /* These are only used to manipulate deh_state.
1696 * Because of this, we'll use the ext2_ bit routines,
1697 * since they are little endian */
1698 #ifdef ADDR_UNALIGNED_BITS
1700 # define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1701 # define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1703 # define set_bit_unaligned(nr, addr) \
1704 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1705 # define clear_bit_unaligned(nr, addr) \
1706 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1707 # define test_bit_unaligned(nr, addr) \
1708 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1712 # define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr)
1713 # define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr)
1714 # define test_bit_unaligned(nr, addr) test_bit_le(nr, addr)
1718 #define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1719 #define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1720 #define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1721 #define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1723 #define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1724 #define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1725 #define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1727 extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1728 __le32 par_dirid, __le32 par_objid);
1729 extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1730 __le32 par_dirid, __le32 par_objid);
1732 /* array of the entry headers */
1734 #define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1735 #define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1737 /* length of the directory entry in directory item. This define
1738 calculates length of i-th directory entry using directory entry
1739 locations from dir entry head. When it calculates length of 0-th
1740 directory entry, it uses length of whole item in place of entry
1741 location of the non-existent following entry in the calculation.
1742 See picture above.*/
1744 #define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1745 ((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1747 static inline int entry_length(const struct buffer_head *bh,
1748 const struct item_head *ih, int pos_in_item)
1750 struct reiserfs_de_head *deh;
1752 deh = B_I_DEH(bh, ih) + pos_in_item;
1754 return deh_location(deh - 1) - deh_location(deh);
1756 return ih_item_len(ih) - deh_location(deh);
1759 /* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1760 #define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1762 /* name by bh, ih and entry_num */
1763 #define B_I_E_NAME(bh,ih,entry_num) ((char *)(bh->b_data + ih_location(ih) + deh_location(B_I_DEH(bh,ih)+(entry_num))))
1765 // two entries per block (at least)
1766 #define REISERFS_MAX_NAME(block_size) 255
1768 /* this structure is used for operations on directory entries. It is
1769 not a disk structure. */
1770 /* When reiserfs_find_entry or search_by_entry_key find directory
1771 entry, they return filled reiserfs_dir_entry structure */
1772 struct reiserfs_dir_entry {
1773 struct buffer_head *de_bh;
1775 struct item_head *de_ih;
1777 struct reiserfs_de_head *de_deh;
1781 unsigned long *de_gen_number_bit_string;
1786 struct cpu_key de_entry_key;
1789 /* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1791 /* pointer to file name, stored in entry */
1792 #define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1794 /* length of name */
1795 #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1796 (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1798 /* hash value occupies bits from 7 up to 30 */
1799 #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1800 /* generation number occupies 7 bits starting from 0 up to 6 */
1801 #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1802 #define MAX_GENERATION_NUMBER 127
1804 #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1807 * Picture represents an internal node of the reiserfs tree
1808 * ______________________________________________________
1809 * | | Array of | Array of | Free |
1810 * |block | keys | pointers | space |
1811 * | head | N | N+1 | |
1812 * |______|_______________|___________________|___________|
1815 /***************************************************************************/
1817 /***************************************************************************/
1818 /* Disk child pointer: The pointer from an internal node of the tree
1819 to a node that is on disk. */
1821 __le32 dc_block_number; /* Disk child's block number. */
1822 __le16 dc_size; /* Disk child's used space. */
1826 #define DC_SIZE (sizeof(struct disk_child))
1827 #define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1828 #define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1829 #define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1830 #define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1832 /* Get disk child by buffer header and position in the tree node. */
1833 #define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
1834 ((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
1836 /* Get disk child number by buffer header and position in the tree node. */
1837 #define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
1838 #define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
1839 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
1841 /* maximal value of field child_size in structure disk_child */
1842 /* child size is the combined size of all items and their headers */
1843 #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1845 /* amount of used space in buffer (not including block head) */
1846 #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1848 /* max and min number of keys in internal node */
1849 #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1850 #define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1852 /***************************************************************************/
1853 /* PATH STRUCTURES AND DEFINES */
1854 /***************************************************************************/
1856 /* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1857 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1858 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1859 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1860 position of the block_number of the next node if it is looking through an internal node. If it
1861 is looking through a leaf node bin_search will find the position of the item which has key either
1862 equal to given key, or which is the maximal key less than the given key. */
1864 struct path_element {
1865 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1866 int pe_position; /* Position in the tree node which is placed in the */
1870 #define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1871 #define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1872 #define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1874 #define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1875 #define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1877 /* We need to keep track of who the ancestors of nodes are. When we
1878 perform a search we record which nodes were visited while
1879 descending the tree looking for the node we searched for. This list
1880 of nodes is called the path. This information is used while
1881 performing balancing. Note that this path information may become
1882 invalid, and this means we must check it when using it to see if it
1883 is still valid. You'll need to read search_by_key and the comments
1884 in it, especially about decrement_counters_in_path(), to understand
1887 Paths make the code so much harder to work with and debug.... An
1888 enormous number of bugs are due to them, and trying to write or modify
1889 code that uses them just makes my head hurt. They are based on an
1890 excessive effort to avoid disturbing the precious VFS code.:-( The
1891 gods only know how we are going to SMP the code that uses them.
1892 znodes are the way! */
1894 #define PATH_READA 0x1 /* do read ahead */
1895 #define PATH_READA_BACK 0x2 /* read backwards */
1898 int path_length; /* Length of the array above. */
1900 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1904 #define pos_in_item(path) ((path)->pos_in_item)
1906 #define INITIALIZE_PATH(var) \
1907 struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1909 /* Get path element by path and path position. */
1910 #define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
1912 /* Get buffer header at the path by path and path position. */
1913 #define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
1915 /* Get position in the element at the path by path and path position. */
1916 #define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
1918 #define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
1919 /* you know, to the person who didn't
1920 write this the macro name does not
1921 at first suggest what it does.
1922 Maybe POSITION_FROM_PATH_END? Or
1923 maybe we should just focus on
1924 dumping paths... -Hans */
1925 #define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
1927 #define PATH_PITEM_HEAD(path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path))
1929 /* in do_balance leaf has h == 0 in contrast with path structure,
1930 where root has level == 0. That is why we need these defines */
1931 #define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h)) /* tb->S[h] */
1932 #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1933 #define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1934 #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1936 #define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
1938 #define get_last_bh(path) PATH_PLAST_BUFFER(path)
1939 #define get_ih(path) PATH_PITEM_HEAD(path)
1940 #define get_item_pos(path) PATH_LAST_POSITION(path)
1941 #define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1942 #define item_moved(ih,path) comp_items(ih, path)
1943 #define path_changed(ih,path) comp_items (ih, path)
1945 /***************************************************************************/
1947 /***************************************************************************/
1949 /* Size of pointer to the unformatted node. */
1950 #define UNFM_P_SIZE (sizeof(unp_t))
1951 #define UNFM_P_SHIFT 2
1953 // in in-core inode key is stored on le form
1954 #define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1956 #define MAX_UL_INT 0xffffffff
1957 #define MAX_INT 0x7ffffff
1958 #define MAX_US_INT 0xffff
1960 // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1962 #define U32_MAX (~(__u32)0)
1963 #endif /* !U32_MAX */
1965 static inline loff_t max_reiserfs_offset(struct inode *inode)
1967 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1968 return (loff_t) U32_MAX;
1970 return (loff_t) ((~(__u64) 0) >> 4);
1973 /*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1974 #define MAX_KEY_OBJECTID MAX_UL_INT
1976 #define MAX_B_NUM MAX_UL_INT
1977 #define MAX_FC_NUM MAX_US_INT
1979 /* the purpose is to detect overflow of an unsigned short */
1980 #define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1982 /* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
1983 #define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1984 #define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1986 #define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1987 #define get_generation(s) atomic_read (&fs_generation(s))
1988 #define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1989 #define __fs_changed(gen,s) (gen != get_generation (s))
1990 #define fs_changed(gen,s) \
1992 reiserfs_cond_resched(s); \
1993 __fs_changed(gen, s); \
1996 /***************************************************************************/
1998 /***************************************************************************/
2000 #define VI_TYPE_LEFT_MERGEABLE 1
2001 #define VI_TYPE_RIGHT_MERGEABLE 2
2003 /* To make any changes in the tree we always first find node, that
2004 contains item to be changed/deleted or place to insert a new
2005 item. We call this node S. To do balancing we need to decide what
2006 we will shift to left/right neighbor, or to a new node, where new
2007 item will be etc. To make this analysis simpler we build virtual
2008 node. Virtual node is an array of items, that will replace items of
2009 node S. (For instance if we are going to delete an item, virtual
2010 node does not contain it). Virtual node keeps information about
2011 item sizes and types, mergeability of first and last items, sizes
2012 of all entries in directory item. We use this array of items when
2013 calculating what we can shift to neighbors and how many nodes we
2014 have to have if we do not any shiftings, if we shift to left/right
2015 neighbor or to both. */
2016 struct virtual_item {
2017 int vi_index; // index in the array of item operations
2018 unsigned short vi_type; // left/right mergeability
2019 unsigned short vi_item_len; /* length of item that it will have after balancing */
2020 struct item_head *vi_ih;
2021 const char *vi_item; // body of item (old or new)
2022 const void *vi_new_data; // 0 always but paste mode
2023 void *vi_uarea; // item specific area
2026 struct virtual_node {
2027 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
2028 unsigned short vn_nr_item; /* number of items in virtual node */
2029 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
2030 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
2031 short vn_affected_item_num;
2032 short vn_pos_in_item;
2033 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
2034 const void *vn_data;
2035 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
2038 /* used by directory items when creating virtual nodes */
2039 struct direntry_uarea {
2042 __u16 entry_sizes[1];
2043 } __attribute__ ((__packed__));
2045 /***************************************************************************/
2047 /***************************************************************************/
2049 /* This temporary structure is used in tree balance algorithms, and
2050 constructed as we go to the extent that its various parts are
2051 needed. It contains arrays of nodes that can potentially be
2052 involved in the balancing of node S, and parameters that define how
2053 each of the nodes must be balanced. Note that in these algorithms
2054 for balancing the worst case is to need to balance the current node
2055 S and the left and right neighbors and all of their parents plus
2056 create a new node. We implement S1 balancing for the leaf nodes
2057 and S0 balancing for the internal nodes (S1 and S0 are defined in
2060 #define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
2062 /* maximum number of FEB blocknrs on a single level */
2063 #define MAX_AMOUNT_NEEDED 2
2065 /* someday somebody will prefix every field in this struct with tb_ */
2066 struct tree_balance {
2068 int need_balance_dirty;
2069 struct super_block *tb_sb;
2070 struct reiserfs_transaction_handle *transaction_handle;
2071 struct treepath *tb_path;
2072 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
2073 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
2074 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
2075 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
2076 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
2077 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
2079 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
2081 struct buffer_head *used[MAX_FEB_SIZE];
2082 struct buffer_head *thrown[MAX_FEB_SIZE];
2083 int lnum[MAX_HEIGHT]; /* array of number of items which must be
2084 shifted to the left in order to balance the
2085 current node; for leaves includes item that
2086 will be partially shifted; for internal
2087 nodes, it is the number of child pointers
2088 rather than items. It includes the new item
2089 being created. The code sometimes subtracts
2090 one to get the number of wholly shifted
2091 items for other purposes. */
2092 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
2093 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
2094 S[h] to its item number within the node CFL[h] */
2095 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
2096 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
2097 S[h]. A negative value means removing. */
2098 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
2099 balancing on the level h of the tree. If 0 then S is
2100 being deleted, if 1 then S is remaining and no new nodes
2101 are being created, if 2 or 3 then 1 or 2 new nodes is
2104 /* fields that are used only for balancing leaves of the tree */
2105 int cur_blknum; /* number of empty blocks having been already allocated */
2106 int s0num; /* number of items that fall into left most node when S[0] splits */
2107 int s1num; /* number of items that fall into first new node when S[0] splits */
2108 int s2num; /* number of items that fall into second new node when S[0] splits */
2109 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
2110 /* most liquid item that cannot be shifted from S[0] entirely */
2111 /* if -1 then nothing will be partially shifted */
2112 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
2113 /* most liquid item that cannot be shifted from S[0] entirely */
2114 /* if -1 then nothing will be partially shifted */
2115 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
2116 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
2118 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
2119 char *vn_buf; /* kmalloced memory. Used to create
2120 virtual node and keep map of
2121 dirtied bitmap blocks */
2122 int vn_buf_size; /* size of the vn_buf */
2123 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
2125 int fs_gen; /* saved value of `reiserfs_generation' counter
2126 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
2127 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2128 struct in_core_key key; /* key pointer, to pass to block allocator or
2129 another low-level subsystem */
2133 /* These are modes of balancing */
2135 /* When inserting an item. */
2136 #define M_INSERT 'i'
2137 /* When inserting into (directories only) or appending onto an already
2140 /* When deleting an item. */
2141 #define M_DELETE 'd'
2142 /* When truncating an item or removing an entry from a (directory) item. */
2145 /* used when balancing on leaf level skipped (in reiserfsck) */
2146 #define M_INTERNAL 'n'
2148 /* When further balancing is not needed, then do_balance does not need
2150 #define M_SKIP_BALANCING 's'
2151 #define M_CONVERT 'v'
2153 /* modes of leaf_move_items */
2154 #define LEAF_FROM_S_TO_L 0
2155 #define LEAF_FROM_S_TO_R 1
2156 #define LEAF_FROM_R_TO_L 2
2157 #define LEAF_FROM_L_TO_R 3
2158 #define LEAF_FROM_S_TO_SNEW 4
2160 #define FIRST_TO_LAST 0
2161 #define LAST_TO_FIRST 1
2163 /* used in do_balance for passing parent of node information that has
2164 been gotten from tb struct */
2165 struct buffer_info {
2166 struct tree_balance *tb;
2167 struct buffer_head *bi_bh;
2168 struct buffer_head *bi_parent;
2172 static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2174 return tb ? tb->tb_sb : NULL;
2177 static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2179 return bi ? sb_from_tb(bi->tb) : NULL;
2182 /* there are 4 types of items: stat data, directory item, indirect, direct.
2183 +-------------------+------------+--------------+------------+
2184 | | k_offset | k_uniqueness | mergeable? |
2185 +-------------------+------------+--------------+------------+
2186 | stat data | 0 | 0 | no |
2187 +-------------------+------------+--------------+------------+
2188 | 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
2189 | non 1st directory | hash value | | yes |
2191 +-------------------+------------+--------------+------------+
2192 | indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
2193 +-------------------+------------+--------------+------------+
2194 | direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
2195 +-------------------+------------+--------------+------------+
2198 struct item_operations {
2199 int (*bytes_number) (struct item_head * ih, int block_size);
2200 void (*decrement_key) (struct cpu_key *);
2201 int (*is_left_mergeable) (struct reiserfs_key * ih,
2202 unsigned long bsize);
2203 void (*print_item) (struct item_head *, char *item);
2204 void (*check_item) (struct item_head *, char *item);
2206 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2207 int is_affected, int insert_size);
2208 int (*check_left) (struct virtual_item * vi, int free,
2209 int start_skip, int end_skip);
2210 int (*check_right) (struct virtual_item * vi, int free);
2211 int (*part_size) (struct virtual_item * vi, int from, int to);
2212 int (*unit_num) (struct virtual_item * vi);
2213 void (*print_vi) (struct virtual_item * vi);
2216 extern struct item_operations *item_ops[TYPE_ANY + 1];
2218 #define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2219 #define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2220 #define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2221 #define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2222 #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)
2223 #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2224 #define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
2225 #define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
2226 #define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
2227 #define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
2229 #define COMP_SHORT_KEYS comp_short_keys
2231 /* number of blocks pointed to by the indirect item */
2232 #define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
2234 /* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
2235 #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2237 /* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
2239 /* get the item header */
2240 #define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
2243 #define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
2246 #define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
2249 #define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
2251 /* get the stat data by the buffer header and the item order */
2252 #define B_N_STAT_DATA(bh,nr) \
2253 ( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
2255 /* following defines use reiserfs buffer header and item header */
2258 #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2260 // this is 3976 for size==4096
2261 #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2263 /* indirect items consist of entries which contain blocknrs, pos
2264 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2265 blocknr contained by the entry pos points to */
2266 #define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
2267 #define PUT_B_I_POS_UNFM_POINTER(bh,ih,pos, val) do {*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)) = cpu_to_le32(val); } while (0)
2269 struct reiserfs_iget_args {
2274 /***************************************************************************/
2275 /* FUNCTION DECLARATIONS */
2276 /***************************************************************************/
2278 #define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2280 #define journal_trans_half(blocksize) \
2281 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2283 /* journal.c see journal.c for all the comments here */
2285 /* first block written in a commit. */
2286 struct reiserfs_journal_desc {
2287 __le32 j_trans_id; /* id of commit */
2288 __le32 j_len; /* length of commit. len +1 is the commit block */
2289 __le32 j_mount_id; /* mount id of this trans */
2290 __le32 j_realblock[1]; /* real locations for each block */
2293 #define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
2294 #define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
2295 #define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
2297 #define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2298 #define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
2299 #define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2301 /* last block written in a commit */
2302 struct reiserfs_journal_commit {
2303 __le32 j_trans_id; /* must match j_trans_id from the desc block */
2304 __le32 j_len; /* ditto */
2305 __le32 j_realblock[1]; /* real locations for each block */
2308 #define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2309 #define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
2310 #define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2312 #define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2313 #define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
2315 /* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
2316 ** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
2317 ** and this transaction does not need to be replayed.
2319 struct reiserfs_journal_header {
2320 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
2321 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
2323 /* 12 */ struct journal_params jh_journal;
2326 /* biggest tunable defines are right here */
2327 #define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
2328 #define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
2329 #define JOURNAL_TRANS_MIN_DEFAULT 256
2330 #define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
2331 #define JOURNAL_MIN_RATIO 2
2332 #define JOURNAL_MAX_COMMIT_AGE 30
2333 #define JOURNAL_MAX_TRANS_AGE 30
2334 #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2335 #define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
2336 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2337 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2340 #define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2341 /* We need to update data and inode (atime) */
2342 #define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2343 /* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2344 #define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2345 (DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2346 /* same as with INIT */
2347 #define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2348 (DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2350 #define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2351 #define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2352 #define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2355 /* both of these can be as low as 1, or as high as you want. The min is the
2356 ** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2357 ** as needed, and released when transactions are committed. On release, if
2358 ** the current number of nodes is > max, the node is freed, otherwise,
2359 ** it is put on a free list for faster use later.
2361 #define REISERFS_MIN_BITMAP_NODES 10
2362 #define REISERFS_MAX_BITMAP_NODES 100
2364 #define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
2365 #define JBH_HASH_MASK 8191
2367 #define _jhashfn(sb,block) \
2368 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2369 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2370 #define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2372 // We need these to make journal.c code more readable
2373 #define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2374 #define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2375 #define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2377 enum reiserfs_bh_state_bits {
2378 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
2380 BH_JNew, /* disk block was taken off free list before
2381 * being in a finished transaction, or
2382 * written to disk. Can be reused immed. */
2385 BH_JTest, // debugging only will go away
2388 BUFFER_FNS(JDirty, journaled);
2389 TAS_BUFFER_FNS(JDirty, journaled);
2390 BUFFER_FNS(JDirty_wait, journal_dirty);
2391 TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2392 BUFFER_FNS(JNew, journal_new);
2393 TAS_BUFFER_FNS(JNew, journal_new);
2394 BUFFER_FNS(JPrepared, journal_prepared);
2395 TAS_BUFFER_FNS(JPrepared, journal_prepared);
2396 BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2397 TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2398 BUFFER_FNS(JTest, journal_test);
2399 TAS_BUFFER_FNS(JTest, journal_test);
2402 ** transaction handle which is passed around for all journal calls
2404 struct reiserfs_transaction_handle {
2405 struct super_block *t_super; /* super for this FS when journal_begin was
2406 called. saves calls to reiserfs_get_super
2407 also used by nested transactions to make
2408 sure they are nesting on the right FS
2409 _must_ be first in the handle
2412 int t_blocks_logged; /* number of blocks this writer has logged */
2413 int t_blocks_allocated; /* number of blocks this writer allocated */
2414 unsigned int t_trans_id; /* sanity check, equals the current trans id */
2415 void *t_handle_save; /* save existing current->journal_info */
2416 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
2417 should be displaced from others */
2418 struct list_head t_list;
2421 /* used to keep track of ordered and tail writes, attached to the buffer
2422 * head through b_journal_head.
2424 struct reiserfs_jh {
2425 struct reiserfs_journal_list *jl;
2426 struct buffer_head *bh;
2427 struct list_head list;
2430 void reiserfs_free_jh(struct buffer_head *bh);
2431 int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2432 int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2433 int journal_mark_dirty(struct reiserfs_transaction_handle *,
2434 struct super_block *, struct buffer_head *bh);
2436 static inline int reiserfs_file_data_log(struct inode *inode)
2438 if (reiserfs_data_log(inode->i_sb) ||
2439 (REISERFS_I(inode)->i_flags & i_data_log))
2444 static inline int reiserfs_transaction_running(struct super_block *s)
2446 struct reiserfs_transaction_handle *th = current->journal_info;
2447 if (th && th->t_super == s)
2449 if (th && th->t_super == NULL)
2454 static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2456 return th->t_blocks_allocated - th->t_blocks_logged;
2459 struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2463 int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
2464 void reiserfs_vfs_truncate_file(struct inode *inode);
2465 int reiserfs_commit_page(struct inode *inode, struct page *page,
2466 unsigned from, unsigned to);
2467 void reiserfs_flush_old_commits(struct super_block *);
2468 int reiserfs_commit_for_inode(struct inode *);
2469 int reiserfs_inode_needs_commit(struct inode *);
2470 void reiserfs_update_inode_transaction(struct inode *);
2471 void reiserfs_wait_on_write_block(struct super_block *s);
2472 void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2473 void reiserfs_allow_writes(struct super_block *s);
2474 void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2475 int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2477 void reiserfs_restore_prepared_buffer(struct super_block *,
2478 struct buffer_head *bh);
2479 int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2481 int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2482 int journal_release_error(struct reiserfs_transaction_handle *,
2483 struct super_block *);
2484 int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
2486 int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
2488 int journal_mark_freed(struct reiserfs_transaction_handle *,
2489 struct super_block *, b_blocknr_t blocknr);
2490 int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2491 int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2492 int bit_nr, int searchall, b_blocknr_t *next);
2493 int journal_begin(struct reiserfs_transaction_handle *,
2494 struct super_block *sb, unsigned long);
2495 int journal_join_abort(struct reiserfs_transaction_handle *,
2496 struct super_block *sb, unsigned long);
2497 void reiserfs_abort_journal(struct super_block *sb, int errno);
2498 void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2499 int reiserfs_allocate_list_bitmaps(struct super_block *s,
2500 struct reiserfs_list_bitmap *, unsigned int);
2502 void reiserfs_schedule_old_flush(struct super_block *s);
2503 void add_save_link(struct reiserfs_transaction_handle *th,
2504 struct inode *inode, int truncate);
2505 int remove_save_link(struct inode *inode, int truncate);
2508 __u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2509 void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2510 __u32 objectid_to_release);
2511 int reiserfs_convert_objectid_map_v1(struct super_block *);
2514 int B_IS_IN_TREE(const struct buffer_head *);
2515 extern void copy_item_head(struct item_head *to,
2516 const struct item_head *from);
2518 // first key is in cpu form, second - le
2519 extern int comp_short_keys(const struct reiserfs_key *le_key,
2520 const struct cpu_key *cpu_key);
2521 extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2523 // both are in le form
2524 extern int comp_le_keys(const struct reiserfs_key *,
2525 const struct reiserfs_key *);
2526 extern int comp_short_le_keys(const struct reiserfs_key *,
2527 const struct reiserfs_key *);
2530 // get key version from on disk key - kludge
2532 static inline int le_key_version(const struct reiserfs_key *key)
2536 type = offset_v2_k_type(&(key->u.k_offset_v2));
2537 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2538 && type != TYPE_DIRENTRY)
2539 return KEY_FORMAT_3_5;
2541 return KEY_FORMAT_3_6;
2545 static inline void copy_key(struct reiserfs_key *to,
2546 const struct reiserfs_key *from)
2548 memcpy(to, from, KEY_SIZE);
2551 int comp_items(const struct item_head *stored_ih, const struct treepath *path);
2552 const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
2553 const struct super_block *sb);
2554 int search_by_key(struct super_block *, const struct cpu_key *,
2555 struct treepath *, int);
2556 #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
2557 int search_for_position_by_key(struct super_block *sb,
2558 const struct cpu_key *cpu_key,
2559 struct treepath *search_path);
2560 extern void decrement_bcount(struct buffer_head *bh);
2561 void decrement_counters_in_path(struct treepath *search_path);
2562 void pathrelse(struct treepath *search_path);
2563 int reiserfs_check_path(struct treepath *p);
2564 void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
2566 int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
2567 struct treepath *path,
2568 const struct cpu_key *key,
2569 struct item_head *ih,
2570 struct inode *inode, const char *body);
2572 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
2573 struct treepath *path,
2574 const struct cpu_key *key,
2575 struct inode *inode,
2576 const char *body, int paste_size);
2578 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
2579 struct treepath *path,
2580 struct cpu_key *key,
2581 struct inode *inode,
2582 struct page *page, loff_t new_file_size);
2584 int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
2585 struct treepath *path,
2586 const struct cpu_key *key,
2587 struct inode *inode, struct buffer_head *un_bh);
2589 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
2590 struct inode *inode, struct reiserfs_key *key);
2591 int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
2592 struct inode *inode);
2593 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
2594 struct inode *inode, struct page *,
2595 int update_timestamps);
2597 #define i_block_size(inode) ((inode)->i_sb->s_blocksize)
2598 #define file_size(inode) ((inode)->i_size)
2599 #define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
2601 #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
2602 !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 )
2604 void padd_item(char *item, int total_length, int length);
2607 /* args for the create parameter of reiserfs_get_block */
2608 #define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
2609 #define GET_BLOCK_CREATE 1 /* add anything you need to find block */
2610 #define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
2611 #define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
2612 #define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
2613 #define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
2615 void reiserfs_read_locked_inode(struct inode *inode,
2616 struct reiserfs_iget_args *args);
2617 int reiserfs_find_actor(struct inode *inode, void *p);
2618 int reiserfs_init_locked_inode(struct inode *inode, void *p);
2619 void reiserfs_evict_inode(struct inode *inode);
2620 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
2621 int reiserfs_get_block(struct inode *inode, sector_t block,
2622 struct buffer_head *bh_result, int create);
2623 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2624 int fh_len, int fh_type);
2625 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
2626 int fh_len, int fh_type);
2627 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
2628 struct inode *parent);
2630 int reiserfs_truncate_file(struct inode *, int update_timestamps);
2631 void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
2632 int type, int key_length);
2633 void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
2635 loff_t offset, int type, int length, int entry_count);
2636 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
2638 struct reiserfs_security_handle;
2639 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
2640 struct inode *dir, umode_t mode,
2641 const char *symname, loff_t i_size,
2642 struct dentry *dentry, struct inode *inode,
2643 struct reiserfs_security_handle *security);
2645 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
2646 struct inode *inode, loff_t size);
2648 static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
2649 struct inode *inode)
2651 reiserfs_update_sd_size(th, inode, inode->i_size);
2654 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
2655 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
2656 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
2658 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
2661 void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
2662 int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
2663 struct treepath *path, struct reiserfs_dir_entry *de);
2664 struct dentry *reiserfs_get_parent(struct dentry *);
2666 #ifdef CONFIG_REISERFS_PROC_INFO
2667 int reiserfs_proc_info_init(struct super_block *sb);
2668 int reiserfs_proc_info_done(struct super_block *sb);
2669 int reiserfs_proc_info_global_init(void);
2670 int reiserfs_proc_info_global_done(void);
2672 #define PROC_EXP( e ) e
2674 #define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
2675 #define PROC_INFO_MAX( sb, field, value ) \
2676 __PINFO( sb ).field = \
2677 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
2678 #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
2679 #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
2680 #define PROC_INFO_BH_STAT( sb, bh, level ) \
2681 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
2682 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
2683 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
2685 static inline int reiserfs_proc_info_init(struct super_block *sb)
2690 static inline int reiserfs_proc_info_done(struct super_block *sb)
2695 static inline int reiserfs_proc_info_global_init(void)
2700 static inline int reiserfs_proc_info_global_done(void)
2705 #define PROC_EXP( e )
2706 #define VOID_V ( ( void ) 0 )
2707 #define PROC_INFO_MAX( sb, field, value ) VOID_V
2708 #define PROC_INFO_INC( sb, field ) VOID_V
2709 #define PROC_INFO_ADD( sb, field, val ) VOID_V
2710 #define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
2714 extern const struct inode_operations reiserfs_dir_inode_operations;
2715 extern const struct inode_operations reiserfs_symlink_inode_operations;
2716 extern const struct inode_operations reiserfs_special_inode_operations;
2717 extern const struct file_operations reiserfs_dir_operations;
2718 int reiserfs_readdir_inode(struct inode *, struct dir_context *);
2720 /* tail_conversion.c */
2721 int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
2722 struct treepath *, struct buffer_head *, loff_t);
2723 int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
2724 struct page *, struct treepath *, const struct cpu_key *,
2726 void reiserfs_unmap_buffer(struct buffer_head *);
2729 extern const struct inode_operations reiserfs_file_inode_operations;
2730 extern const struct file_operations reiserfs_file_operations;
2731 extern const struct address_space_operations reiserfs_address_space_operations;
2735 int fix_nodes(int n_op_mode, struct tree_balance *tb,
2736 struct item_head *ins_ih, const void *);
2737 void unfix_nodes(struct tree_balance *);
2740 void __reiserfs_panic(struct super_block *s, const char *id,
2741 const char *function, const char *fmt, ...)
2742 __attribute__ ((noreturn));
2743 #define reiserfs_panic(s, id, fmt, args...) \
2744 __reiserfs_panic(s, id, __func__, fmt, ##args)
2745 void __reiserfs_error(struct super_block *s, const char *id,
2746 const char *function, const char *fmt, ...);
2747 #define reiserfs_error(s, id, fmt, args...) \
2748 __reiserfs_error(s, id, __func__, fmt, ##args)
2749 void reiserfs_info(struct super_block *s, const char *fmt, ...);
2750 void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2751 void print_indirect_item(struct buffer_head *bh, int item_num);
2752 void store_print_tb(struct tree_balance *tb);
2753 void print_cur_tb(char *mes);
2754 void print_de(struct reiserfs_dir_entry *de);
2755 void print_bi(struct buffer_info *bi, char *mes);
2756 #define PRINT_LEAF_ITEMS 1 /* print all items */
2757 #define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2758 #define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2759 void print_block(struct buffer_head *bh, ...);
2760 void print_bmap(struct super_block *s, int silent);
2761 void print_bmap_block(int i, char *data, int size, int silent);
2762 /*void print_super_block (struct super_block * s, char * mes);*/
2763 void print_objectid_map(struct super_block *s);
2764 void print_block_head(struct buffer_head *bh, char *mes);
2765 void check_leaf(struct buffer_head *bh);
2766 void check_internal(struct buffer_head *bh);
2767 void print_statistics(struct super_block *s);
2768 char *reiserfs_hashname(int code);
2771 int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2772 int mov_bytes, struct buffer_head *Snew);
2773 int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2774 int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2775 void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2776 int del_num, int del_bytes);
2777 void leaf_insert_into_buf(struct buffer_info *bi, int before,
2778 struct item_head *inserted_item_ih,
2779 const char *inserted_item_body, int zeros_number);
2780 void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2781 int pos_in_item, int paste_size, const char *body,
2783 void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2784 int pos_in_item, int cut_size);
2785 void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
2786 int new_entry_count, struct reiserfs_de_head *new_dehs,
2787 const char *records, int paste_size);
2789 int balance_internal(struct tree_balance *, int, int, struct item_head *,
2790 struct buffer_head **);
2793 void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2794 struct buffer_head *bh, int flag);
2795 #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2796 #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2798 void do_balance(struct tree_balance *tb, struct item_head *ih,
2799 const char *body, int flag);
2800 void reiserfs_invalidate_buffer(struct tree_balance *tb,
2801 struct buffer_head *bh);
2803 int get_left_neighbor_position(struct tree_balance *tb, int h);
2804 int get_right_neighbor_position(struct tree_balance *tb, int h);
2805 void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2806 struct buffer_head *, int);
2807 void make_empty_node(struct buffer_info *);
2808 struct buffer_head *get_FEB(struct tree_balance *);
2812 /* structure contains hints for block allocator, and it is a container for
2813 * arguments, such as node, search path, transaction_handle, etc. */
2814 struct __reiserfs_blocknr_hint {
2815 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
2816 sector_t block; /* file offset, in blocks */
2817 struct in_core_key key;
2818 struct treepath *path; /* search path, used by allocator to deternine search_start by
2820 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2821 * bitmap blocks changes */
2822 b_blocknr_t beg, end;
2823 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
2824 * between different block allocator procedures
2825 * (determine_search_start() and others) */
2826 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2827 * function that do actual allocation */
2829 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
2830 * formatted/unformatted blocks with/without preallocation */
2831 unsigned preallocate:1;
2834 typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2836 int reiserfs_parse_alloc_options(struct super_block *, char *);
2837 void reiserfs_init_alloc_options(struct super_block *s);
2840 * given a directory, this will tell you what packing locality
2841 * to use for a new object underneat it. The locality is returned
2842 * in disk byte order (le).
2844 __le32 reiserfs_choose_packing(struct inode *dir);
2846 int reiserfs_init_bitmap_cache(struct super_block *sb);
2847 void reiserfs_free_bitmap_cache(struct super_block *sb);
2848 void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2849 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
2850 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2851 void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2852 b_blocknr_t, int for_unformatted);
2853 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2855 static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
2856 b_blocknr_t * new_blocknrs,
2859 reiserfs_blocknr_hint_t hint = {
2860 .th = tb->transaction_handle,
2861 .path = tb->tb_path,
2867 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2871 static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
2872 *th, struct inode *inode,
2873 b_blocknr_t * new_blocknrs,
2874 struct treepath *path,
2877 reiserfs_blocknr_hint_t hint = {
2882 .formatted_node = 0,
2885 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2888 #ifdef REISERFS_PREALLOCATE
2889 static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
2890 *th, struct inode *inode,
2891 b_blocknr_t * new_blocknrs,
2892 struct treepath *path,
2895 reiserfs_blocknr_hint_t hint = {
2900 .formatted_node = 0,
2903 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2906 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2907 struct inode *inode);
2908 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
2912 __u32 keyed_hash(const signed char *msg, int len);
2913 __u32 yura_hash(const signed char *msg, int len);
2914 __u32 r5_hash(const signed char *msg, int len);
2916 #define reiserfs_set_le_bit __set_bit_le
2917 #define reiserfs_test_and_set_le_bit __test_and_set_bit_le
2918 #define reiserfs_clear_le_bit __clear_bit_le
2919 #define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le
2920 #define reiserfs_test_le_bit test_bit_le
2921 #define reiserfs_find_next_zero_le_bit find_next_zero_bit_le
2923 /* sometimes reiserfs_truncate may require to allocate few new blocks
2924 to perform indirect2direct conversion. People probably used to
2925 think, that truncate should work without problems on a filesystem
2926 without free disk space. They may complain that they can not
2927 truncate due to lack of free disk space. This spare space allows us
2928 to not worry about it. 500 is probably too much, but it should be
2930 #define SPARE_SPACE 500
2932 /* prototypes from ioctl.c */
2933 long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
2934 long reiserfs_compat_ioctl(struct file *filp,
2935 unsigned int cmd, unsigned long arg);
2936 int reiserfs_unpack(struct inode *inode, struct file *filp);
projects / ~andy / linux / blob