xref: /openbmc/linux/fs/reiserfs/reiserfs.h (revision d2999e1b)
1 /*
2  * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for
3  * licensing and copyright details
4  */
5 
6 #include <linux/reiserfs_fs.h>
7 
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/sched.h>
11 #include <linux/bug.h>
12 #include <linux/workqueue.h>
13 #include <asm/unaligned.h>
14 #include <linux/bitops.h>
15 #include <linux/proc_fs.h>
16 #include <linux/buffer_head.h>
17 
18 /* the 32 bit compat definitions with int argument */
19 #define REISERFS_IOC32_UNPACK		_IOW(0xCD, 1, int)
20 #define REISERFS_IOC32_GETFLAGS		FS_IOC32_GETFLAGS
21 #define REISERFS_IOC32_SETFLAGS		FS_IOC32_SETFLAGS
22 #define REISERFS_IOC32_GETVERSION	FS_IOC32_GETVERSION
23 #define REISERFS_IOC32_SETVERSION	FS_IOC32_SETVERSION
24 
25 struct reiserfs_journal_list;
26 
27 /* bitmasks for i_flags field in reiserfs-specific part of inode */
28 typedef enum {
29 	/*
30 	 * this says what format of key do all items (but stat data) of
31 	 * an object have.  If this is set, that format is 3.6 otherwise - 3.5
32 	 */
33 	i_item_key_version_mask = 0x0001,
34 
35 	/*
36 	 * If this is unset, object has 3.5 stat data, otherwise,
37 	 * it has 3.6 stat data with 64bit size, 32bit nlink etc.
38 	 */
39 	i_stat_data_version_mask = 0x0002,
40 
41 	/* file might need tail packing on close */
42 	i_pack_on_close_mask = 0x0004,
43 
44 	/* don't pack tail of file */
45 	i_nopack_mask = 0x0008,
46 
47 	/*
48 	 * If either of these are set, "safe link" was created for this
49 	 * file during truncate or unlink. Safe link is used to avoid
50 	 * leakage of disk space on crash with some files open, but unlinked.
51 	 */
52 	i_link_saved_unlink_mask = 0x0010,
53 	i_link_saved_truncate_mask = 0x0020,
54 
55 	i_has_xattr_dir = 0x0040,
56 	i_data_log = 0x0080,
57 } reiserfs_inode_flags;
58 
59 struct reiserfs_inode_info {
60 	__u32 i_key[4];		/* key is still 4 32 bit integers */
61 
62 	/*
63 	 * transient inode flags that are never stored on disk. Bitmasks
64 	 * for this field are defined above.
65 	 */
66 	__u32 i_flags;
67 
68 	/* offset of first byte stored in direct item. */
69 	__u32 i_first_direct_byte;
70 
71 	/* copy of persistent inode flags read from sd_attrs. */
72 	__u32 i_attrs;
73 
74 	/* first unused block of a sequence of unused blocks */
75 	int i_prealloc_block;
76 	int i_prealloc_count;	/* length of that sequence */
77 
78 	/* per-transaction list of inodes which  have preallocated blocks */
79 	struct list_head i_prealloc_list;
80 
81 	/*
82 	 * new_packing_locality is created; new blocks for the contents
83 	 * of this directory should be displaced
84 	 */
85 	unsigned new_packing_locality:1;
86 
87 	/*
88 	 * we use these for fsync or O_SYNC to decide which transaction
89 	 * needs to be committed in order for this inode to be properly
90 	 * flushed
91 	 */
92 	unsigned int i_trans_id;
93 
94 	struct reiserfs_journal_list *i_jl;
95 	atomic_t openers;
96 	struct mutex tailpack;
97 #ifdef CONFIG_REISERFS_FS_XATTR
98 	struct rw_semaphore i_xattr_sem;
99 #endif
100 	struct inode vfs_inode;
101 };
102 
103 typedef enum {
104 	reiserfs_attrs_cleared = 0x00000001,
105 } reiserfs_super_block_flags;
106 
107 /*
108  * struct reiserfs_super_block accessors/mutators since this is a disk
109  * structure, it will always be in little endian format.
110  */
111 #define sb_block_count(sbp)         (le32_to_cpu((sbp)->s_v1.s_block_count))
112 #define set_sb_block_count(sbp,v)   ((sbp)->s_v1.s_block_count = cpu_to_le32(v))
113 #define sb_free_blocks(sbp)         (le32_to_cpu((sbp)->s_v1.s_free_blocks))
114 #define set_sb_free_blocks(sbp,v)   ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))
115 #define sb_root_block(sbp)          (le32_to_cpu((sbp)->s_v1.s_root_block))
116 #define set_sb_root_block(sbp,v)    ((sbp)->s_v1.s_root_block = cpu_to_le32(v))
117 
118 #define sb_jp_journal_1st_block(sbp)  \
119               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))
120 #define set_sb_jp_journal_1st_block(sbp,v) \
121               ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))
122 #define sb_jp_journal_dev(sbp) \
123               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))
124 #define set_sb_jp_journal_dev(sbp,v) \
125               ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))
126 #define sb_jp_journal_size(sbp) \
127               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))
128 #define set_sb_jp_journal_size(sbp,v) \
129               ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))
130 #define sb_jp_journal_trans_max(sbp) \
131               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))
132 #define set_sb_jp_journal_trans_max(sbp,v) \
133               ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))
134 #define sb_jp_journal_magic(sbp) \
135               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))
136 #define set_sb_jp_journal_magic(sbp,v) \
137               ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))
138 #define sb_jp_journal_max_batch(sbp) \
139               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))
140 #define set_sb_jp_journal_max_batch(sbp,v) \
141               ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))
142 #define sb_jp_jourmal_max_commit_age(sbp) \
143               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))
144 #define set_sb_jp_journal_max_commit_age(sbp,v) \
145               ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))
146 
147 #define sb_blocksize(sbp)          (le16_to_cpu((sbp)->s_v1.s_blocksize))
148 #define set_sb_blocksize(sbp,v)    ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))
149 #define sb_oid_maxsize(sbp)        (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))
150 #define set_sb_oid_maxsize(sbp,v)  ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))
151 #define sb_oid_cursize(sbp)        (le16_to_cpu((sbp)->s_v1.s_oid_cursize))
152 #define set_sb_oid_cursize(sbp,v)  ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))
153 #define sb_umount_state(sbp)       (le16_to_cpu((sbp)->s_v1.s_umount_state))
154 #define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))
155 #define sb_fs_state(sbp)           (le16_to_cpu((sbp)->s_v1.s_fs_state))
156 #define set_sb_fs_state(sbp,v)     ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))
157 #define sb_hash_function_code(sbp) \
158               (le32_to_cpu((sbp)->s_v1.s_hash_function_code))
159 #define set_sb_hash_function_code(sbp,v) \
160               ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))
161 #define sb_tree_height(sbp)        (le16_to_cpu((sbp)->s_v1.s_tree_height))
162 #define set_sb_tree_height(sbp,v)  ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))
163 #define sb_bmap_nr(sbp)            (le16_to_cpu((sbp)->s_v1.s_bmap_nr))
164 #define set_sb_bmap_nr(sbp,v)      ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))
165 #define sb_version(sbp)            (le16_to_cpu((sbp)->s_v1.s_version))
166 #define set_sb_version(sbp,v)      ((sbp)->s_v1.s_version = cpu_to_le16(v))
167 
168 #define sb_mnt_count(sbp)	   (le16_to_cpu((sbp)->s_mnt_count))
169 #define set_sb_mnt_count(sbp, v)   ((sbp)->s_mnt_count = cpu_to_le16(v))
170 
171 #define sb_reserved_for_journal(sbp) \
172               (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))
173 #define set_sb_reserved_for_journal(sbp,v) \
174               ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))
175 
176 /* LOGGING -- */
177 
178 /*
179  * These all interelate for performance.
180  *
181  * If the journal block count is smaller than n transactions, you lose speed.
182  * I don't know what n is yet, I'm guessing 8-16.
183  *
184  * typical transaction size depends on the application, how often fsync is
185  * called, and how many metadata blocks you dirty in a 30 second period.
186  * The more small files (<16k) you use, the larger your transactions will
187  * be.
188  *
189  * If your journal fills faster than dirty buffers get flushed to disk, it
190  * must flush them before allowing the journal to wrap, which slows things
191  * down.  If you need high speed meta data updates, the journal should be
192  * big enough to prevent wrapping before dirty meta blocks get to disk.
193  *
194  * If the batch max is smaller than the transaction max, you'll waste space
195  * at the end of the journal because journal_end sets the next transaction
196  * to start at 0 if the next transaction has any chance of wrapping.
197  *
198  * The large the batch max age, the better the speed, and the more meta
199  * data changes you'll lose after a crash.
200  */
201 
202 /* don't mess with these for a while */
203 /* we have a node size define somewhere in reiserfs_fs.h. -Hans */
204 #define JOURNAL_BLOCK_SIZE  4096	/* BUG gotta get rid of this */
205 #define JOURNAL_MAX_CNODE   1500	/* max cnodes to allocate. */
206 #define JOURNAL_HASH_SIZE 8192
207 
208 /* number of copies of the bitmaps to have floating.  Must be >= 2 */
209 #define JOURNAL_NUM_BITMAPS 5
210 
211 /*
212  * One of these for every block in every transaction
213  * Each one is in two hash tables.  First, a hash of the current transaction,
214  * and after journal_end, a hash of all the in memory transactions.
215  * next and prev are used by the current transaction (journal_hash).
216  * hnext and hprev are used by journal_list_hash.  If a block is in more
217  * than one transaction, the journal_list_hash links it in multiple times.
218  * This allows flush_journal_list to remove just the cnode belonging to a
219  * given transaction.
220  */
221 struct reiserfs_journal_cnode {
222 	struct buffer_head *bh;	/* real buffer head */
223 	struct super_block *sb;	/* dev of real buffer head */
224 
225 	/* block number of real buffer head, == 0 when buffer on disk */
226 	__u32 blocknr;
227 
228 	unsigned long state;
229 
230 	/* journal list this cnode lives in */
231 	struct reiserfs_journal_list *jlist;
232 
233 	struct reiserfs_journal_cnode *next;	/* next in transaction list */
234 	struct reiserfs_journal_cnode *prev;	/* prev in transaction list */
235 	struct reiserfs_journal_cnode *hprev;	/* prev in hash list */
236 	struct reiserfs_journal_cnode *hnext;	/* next in hash list */
237 };
238 
239 struct reiserfs_bitmap_node {
240 	int id;
241 	char *data;
242 	struct list_head list;
243 };
244 
245 struct reiserfs_list_bitmap {
246 	struct reiserfs_journal_list *journal_list;
247 	struct reiserfs_bitmap_node **bitmaps;
248 };
249 
250 /*
251  * one of these for each transaction.  The most important part here is the
252  * j_realblock.  this list of cnodes is used to hash all the blocks in all
253  * the commits, to mark all the real buffer heads dirty once all the commits
254  * hit the disk, and to make sure every real block in a transaction is on
255  * disk before allowing the log area to be overwritten
256  */
257 struct reiserfs_journal_list {
258 	unsigned long j_start;
259 	unsigned long j_state;
260 	unsigned long j_len;
261 	atomic_t j_nonzerolen;
262 	atomic_t j_commit_left;
263 
264 	/* all commits older than this on disk */
265 	atomic_t j_older_commits_done;
266 
267 	struct mutex j_commit_mutex;
268 	unsigned int j_trans_id;
269 	time_t j_timestamp;
270 	struct reiserfs_list_bitmap *j_list_bitmap;
271 	struct buffer_head *j_commit_bh;	/* commit buffer head */
272 	struct reiserfs_journal_cnode *j_realblock;
273 	struct reiserfs_journal_cnode *j_freedlist;	/* list of buffers that were freed during this trans.  free each of these on flush */
274 	/* time ordered list of all active transactions */
275 	struct list_head j_list;
276 
277 	/*
278 	 * time ordered list of all transactions we haven't tried
279 	 * to flush yet
280 	 */
281 	struct list_head j_working_list;
282 
283 	/* list of tail conversion targets in need of flush before commit */
284 	struct list_head j_tail_bh_list;
285 
286 	/* list of data=ordered buffers in need of flush before commit */
287 	struct list_head j_bh_list;
288 	int j_refcount;
289 };
290 
291 struct reiserfs_journal {
292 	struct buffer_head **j_ap_blocks;	/* journal blocks on disk */
293 	/* newest journal block */
294 	struct reiserfs_journal_cnode *j_last;
295 
296 	/* oldest journal block.  start here for traverse */
297 	struct reiserfs_journal_cnode *j_first;
298 
299 	struct block_device *j_dev_bd;
300 	fmode_t j_dev_mode;
301 
302 	/* first block on s_dev of reserved area journal */
303 	int j_1st_reserved_block;
304 
305 	unsigned long j_state;
306 	unsigned int j_trans_id;
307 	unsigned long j_mount_id;
308 
309 	/* start of current waiting commit (index into j_ap_blocks) */
310 	unsigned long j_start;
311 	unsigned long j_len;	/* length of current waiting commit */
312 
313 	/* number of buffers requested by journal_begin() */
314 	unsigned long j_len_alloc;
315 
316 	atomic_t j_wcount;	/* count of writers for current commit */
317 
318 	/* batch count. allows turning X transactions into 1 */
319 	unsigned long j_bcount;
320 
321 	/* first unflushed transactions offset */
322 	unsigned long j_first_unflushed_offset;
323 
324 	/* last fully flushed journal timestamp */
325 	unsigned j_last_flush_trans_id;
326 
327 	struct buffer_head *j_header_bh;
328 
329 	time_t j_trans_start_time;	/* time this transaction started */
330 	struct mutex j_mutex;
331 	struct mutex j_flush_mutex;
332 
333 	/* wait for current transaction to finish before starting new one */
334 	wait_queue_head_t j_join_wait;
335 
336 	atomic_t j_jlock;		/* lock for j_join_wait */
337 	int j_list_bitmap_index;	/* number of next list bitmap to use */
338 
339 	/* no more journal begins allowed. MUST sleep on j_join_wait */
340 	int j_must_wait;
341 
342 	/* next journal_end will flush all journal list */
343 	int j_next_full_flush;
344 
345 	/* next journal_end will flush all async commits */
346 	int j_next_async_flush;
347 
348 	int j_cnode_used;	/* number of cnodes on the used list */
349 	int j_cnode_free;	/* number of cnodes on the free list */
350 
351 	/* max number of blocks in a transaction.  */
352 	unsigned int j_trans_max;
353 
354 	/* max number of blocks to batch into a trans */
355 	unsigned int j_max_batch;
356 
357 	/* in seconds, how old can an async commit be */
358 	unsigned int j_max_commit_age;
359 
360 	/* in seconds, how old can a transaction be */
361 	unsigned int j_max_trans_age;
362 
363 	/* the default for the max commit age */
364 	unsigned int j_default_max_commit_age;
365 
366 	struct reiserfs_journal_cnode *j_cnode_free_list;
367 
368 	/* orig pointer returned from vmalloc */
369 	struct reiserfs_journal_cnode *j_cnode_free_orig;
370 
371 	struct reiserfs_journal_list *j_current_jl;
372 	int j_free_bitmap_nodes;
373 	int j_used_bitmap_nodes;
374 
375 	int j_num_lists;	/* total number of active transactions */
376 	int j_num_work_lists;	/* number that need attention from kreiserfsd */
377 
378 	/* debugging to make sure things are flushed in order */
379 	unsigned int j_last_flush_id;
380 
381 	/* debugging to make sure things are committed in order */
382 	unsigned int j_last_commit_id;
383 
384 	struct list_head j_bitmap_nodes;
385 	struct list_head j_dirty_buffers;
386 	spinlock_t j_dirty_buffers_lock;	/* protects j_dirty_buffers */
387 
388 	/* list of all active transactions */
389 	struct list_head j_journal_list;
390 
391 	/* lists that haven't been touched by writeback attempts */
392 	struct list_head j_working_list;
393 
394 	/* hash table for real buffer heads in current trans */
395 	struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE];
396 
397 	/* hash table for all the real buffer heads in all the transactions */
398 	struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE];
399 
400 	/* array of bitmaps to record the deleted blocks */
401 	struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS];
402 
403 	/* list of inodes which have preallocated blocks */
404 	struct list_head j_prealloc_list;
405 	int j_persistent_trans;
406 	unsigned long j_max_trans_size;
407 	unsigned long j_max_batch_size;
408 
409 	int j_errno;
410 
411 	/* when flushing ordered buffers, throttle new ordered writers */
412 	struct delayed_work j_work;
413 	struct super_block *j_work_sb;
414 	atomic_t j_async_throttle;
415 };
416 
417 enum journal_state_bits {
418 	J_WRITERS_BLOCKED = 1,	/* set when new writers not allowed */
419 	J_WRITERS_QUEUED,    /* set when log is full due to too many writers */
420 	J_ABORTED,           /* set when log is aborted */
421 };
422 
423 /* ick.  magic string to find desc blocks in the journal */
424 #define JOURNAL_DESC_MAGIC "ReIsErLB"
425 
426 typedef __u32(*hashf_t) (const signed char *, int);
427 
428 struct reiserfs_bitmap_info {
429 	__u32 free_count;
430 };
431 
432 struct proc_dir_entry;
433 
434 #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
435 typedef unsigned long int stat_cnt_t;
436 typedef struct reiserfs_proc_info_data {
437 	spinlock_t lock;
438 	int exiting;
439 	int max_hash_collisions;
440 
441 	stat_cnt_t breads;
442 	stat_cnt_t bread_miss;
443 	stat_cnt_t search_by_key;
444 	stat_cnt_t search_by_key_fs_changed;
445 	stat_cnt_t search_by_key_restarted;
446 
447 	stat_cnt_t insert_item_restarted;
448 	stat_cnt_t paste_into_item_restarted;
449 	stat_cnt_t cut_from_item_restarted;
450 	stat_cnt_t delete_solid_item_restarted;
451 	stat_cnt_t delete_item_restarted;
452 
453 	stat_cnt_t leaked_oid;
454 	stat_cnt_t leaves_removable;
455 
456 	/*
457 	 * balances per level.
458 	 * Use explicit 5 as MAX_HEIGHT is not visible yet.
459 	 */
460 	stat_cnt_t balance_at[5];	/* XXX */
461 	/* sbk == search_by_key */
462 	stat_cnt_t sbk_read_at[5];	/* XXX */
463 	stat_cnt_t sbk_fs_changed[5];
464 	stat_cnt_t sbk_restarted[5];
465 	stat_cnt_t items_at[5];	/* XXX */
466 	stat_cnt_t free_at[5];	/* XXX */
467 	stat_cnt_t can_node_be_removed[5];	/* XXX */
468 	long int lnum[5];	/* XXX */
469 	long int rnum[5];	/* XXX */
470 	long int lbytes[5];	/* XXX */
471 	long int rbytes[5];	/* XXX */
472 	stat_cnt_t get_neighbors[5];
473 	stat_cnt_t get_neighbors_restart[5];
474 	stat_cnt_t need_l_neighbor[5];
475 	stat_cnt_t need_r_neighbor[5];
476 
477 	stat_cnt_t free_block;
478 	struct __scan_bitmap_stats {
479 		stat_cnt_t call;
480 		stat_cnt_t wait;
481 		stat_cnt_t bmap;
482 		stat_cnt_t retry;
483 		stat_cnt_t in_journal_hint;
484 		stat_cnt_t in_journal_nohint;
485 		stat_cnt_t stolen;
486 	} scan_bitmap;
487 	struct __journal_stats {
488 		stat_cnt_t in_journal;
489 		stat_cnt_t in_journal_bitmap;
490 		stat_cnt_t in_journal_reusable;
491 		stat_cnt_t lock_journal;
492 		stat_cnt_t lock_journal_wait;
493 		stat_cnt_t journal_being;
494 		stat_cnt_t journal_relock_writers;
495 		stat_cnt_t journal_relock_wcount;
496 		stat_cnt_t mark_dirty;
497 		stat_cnt_t mark_dirty_already;
498 		stat_cnt_t mark_dirty_notjournal;
499 		stat_cnt_t restore_prepared;
500 		stat_cnt_t prepare;
501 		stat_cnt_t prepare_retry;
502 	} journal;
503 } reiserfs_proc_info_data_t;
504 #else
505 typedef struct reiserfs_proc_info_data {
506 } reiserfs_proc_info_data_t;
507 #endif
508 
509 /* reiserfs union of in-core super block data */
510 struct reiserfs_sb_info {
511 	/* Buffer containing the super block */
512 	struct buffer_head *s_sbh;
513 
514 	/* Pointer to the on-disk super block in the buffer */
515 	struct reiserfs_super_block *s_rs;
516 	struct reiserfs_bitmap_info *s_ap_bitmap;
517 
518 	/* pointer to journal information */
519 	struct reiserfs_journal *s_journal;
520 
521 	unsigned short s_mount_state;	/* reiserfs state (valid, invalid) */
522 
523 	/* Serialize writers access, replace the old bkl */
524 	struct mutex lock;
525 
526 	/* Owner of the lock (can be recursive) */
527 	struct task_struct *lock_owner;
528 
529 	/* Depth of the lock, start from -1 like the bkl */
530 	int lock_depth;
531 
532 	struct workqueue_struct *commit_wq;
533 
534 	/* Comment? -Hans */
535 	void (*end_io_handler) (struct buffer_head *, int);
536 
537 	/*
538 	 * pointer to function which is used to sort names in directory.
539 	 * Set on mount
540 	 */
541 	hashf_t s_hash_function;
542 
543 	/* reiserfs's mount options are set here */
544 	unsigned long s_mount_opt;
545 
546 	/* This is a structure that describes block allocator options */
547 	struct {
548 		/* Bitfield for enable/disable kind of options */
549 		unsigned long bits;
550 
551 		/*
552 		 * size started from which we consider file
553 		 * to be a large one (in blocks)
554 		 */
555 		unsigned long large_file_size;
556 
557 		int border;	/* percentage of disk, border takes */
558 
559 		/*
560 		 * Minimal file size (in blocks) starting
561 		 * from which we do preallocations
562 		 */
563 		int preallocmin;
564 
565 		/*
566 		 * Number of blocks we try to prealloc when file
567 		 * reaches preallocmin size (in blocks) or prealloc_list
568 		 is empty.
569 		 */
570 		int preallocsize;
571 	} s_alloc_options;
572 
573 	/* Comment? -Hans */
574 	wait_queue_head_t s_wait;
575 	/* increased by one every time the  tree gets re-balanced */
576 	atomic_t s_generation_counter;
577 
578 	/* File system properties. Currently holds on-disk FS format */
579 	unsigned long s_properties;
580 
581 	/* session statistics */
582 	int s_disk_reads;
583 	int s_disk_writes;
584 	int s_fix_nodes;
585 	int s_do_balance;
586 	int s_unneeded_left_neighbor;
587 	int s_good_search_by_key_reada;
588 	int s_bmaps;
589 	int s_bmaps_without_search;
590 	int s_direct2indirect;
591 	int s_indirect2direct;
592 
593 	/*
594 	 * set up when it's ok for reiserfs_read_inode2() to read from
595 	 * disk inode with nlink==0. Currently this is only used during
596 	 * finish_unfinished() processing at mount time
597 	 */
598 	int s_is_unlinked_ok;
599 
600 	reiserfs_proc_info_data_t s_proc_info_data;
601 	struct proc_dir_entry *procdir;
602 
603 	/* amount of blocks reserved for further allocations */
604 	int reserved_blocks;
605 
606 
607 	/* this lock on now only used to protect reserved_blocks variable */
608 	spinlock_t bitmap_lock;
609 	struct dentry *priv_root;	/* root of /.reiserfs_priv */
610 	struct dentry *xattr_root;	/* root of /.reiserfs_priv/xattrs */
611 	int j_errno;
612 
613 	int work_queued;              /* non-zero delayed work is queued */
614 	struct delayed_work old_work; /* old transactions flush delayed work */
615 	spinlock_t old_work_lock;     /* protects old_work and work_queued */
616 
617 #ifdef CONFIG_QUOTA
618 	char *s_qf_names[MAXQUOTAS];
619 	int s_jquota_fmt;
620 #endif
621 	char *s_jdev;		/* Stored jdev for mount option showing */
622 #ifdef CONFIG_REISERFS_CHECK
623 
624 	/*
625 	 * Detects whether more than one copy of tb exists per superblock
626 	 * as a means of checking whether do_balance is executing
627 	 * concurrently against another tree reader/writer on a same
628 	 * mount point.
629 	 */
630 	struct tree_balance *cur_tb;
631 #endif
632 };
633 
634 /* Definitions of reiserfs on-disk properties: */
635 #define REISERFS_3_5 0
636 #define REISERFS_3_6 1
637 #define REISERFS_OLD_FORMAT 2
638 
639 /* Mount options */
640 enum reiserfs_mount_options {
641 	/* large tails will be created in a session */
642 	REISERFS_LARGETAIL,
643 	/*
644 	 * small (for files less than block size) tails will
645 	 * be created in a session
646 	 */
647 	REISERFS_SMALLTAIL,
648 
649 	/* replay journal and return 0. Use by fsck */
650 	REPLAYONLY,
651 
652 	/*
653 	 * -o conv: causes conversion of old format super block to the
654 	 * new format. If not specified - old partition will be dealt
655 	 * with in a manner of 3.5.x
656 	 */
657 	REISERFS_CONVERT,
658 
659 	/*
660 	 * -o hash={tea, rupasov, r5, detect} is meant for properly mounting
661 	 * reiserfs disks from 3.5.19 or earlier.  99% of the time, this
662 	 * option is not required.  If the normal autodection code can't
663 	 * determine which hash to use (because both hashes had the same
664 	 * value for a file) use this option to force a specific hash.
665 	 * It won't allow you to override the existing hash on the FS, so
666 	 * if you have a tea hash disk, and mount with -o hash=rupasov,
667 	 * the mount will fail.
668 	 */
669 	FORCE_TEA_HASH,		/* try to force tea hash on mount */
670 	FORCE_RUPASOV_HASH,	/* try to force rupasov hash on mount */
671 	FORCE_R5_HASH,		/* try to force rupasov hash on mount */
672 	FORCE_HASH_DETECT,	/* try to detect hash function on mount */
673 
674 	REISERFS_DATA_LOG,
675 	REISERFS_DATA_ORDERED,
676 	REISERFS_DATA_WRITEBACK,
677 
678 	/*
679 	 * used for testing experimental features, makes benchmarking new
680 	 * features with and without more convenient, should never be used by
681 	 * users in any code shipped to users (ideally)
682 	 */
683 
684 	REISERFS_NO_BORDER,
685 	REISERFS_NO_UNHASHED_RELOCATION,
686 	REISERFS_HASHED_RELOCATION,
687 	REISERFS_ATTRS,
688 	REISERFS_XATTRS_USER,
689 	REISERFS_POSIXACL,
690 	REISERFS_EXPOSE_PRIVROOT,
691 	REISERFS_BARRIER_NONE,
692 	REISERFS_BARRIER_FLUSH,
693 
694 	/* Actions on error */
695 	REISERFS_ERROR_PANIC,
696 	REISERFS_ERROR_RO,
697 	REISERFS_ERROR_CONTINUE,
698 
699 	REISERFS_USRQUOTA,	/* User quota option specified */
700 	REISERFS_GRPQUOTA,	/* Group quota option specified */
701 
702 	REISERFS_TEST1,
703 	REISERFS_TEST2,
704 	REISERFS_TEST3,
705 	REISERFS_TEST4,
706 	REISERFS_UNSUPPORTED_OPT,
707 };
708 
709 #define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))
710 #define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))
711 #define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))
712 #define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))
713 #define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))
714 #define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))
715 #define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))
716 #define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))
717 
718 #define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))
719 #define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))
720 #define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))
721 #define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))
722 #define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))
723 #define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))
724 #define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))
725 #define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))
726 #define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))
727 #define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))
728 #define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))
729 #define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))
730 #define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))
731 #define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))
732 #define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))
733 
734 #define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))
735 #define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))
736 
737 void reiserfs_file_buffer(struct buffer_head *bh, int list);
738 extern struct file_system_type reiserfs_fs_type;
739 int reiserfs_resize(struct super_block *, unsigned long);
740 
741 #define CARRY_ON                0
742 #define SCHEDULE_OCCURRED       1
743 
744 #define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)
745 #define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)
746 #define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)
747 #define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)
748 #define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)
749 
750 #define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)
751 
752 #define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
753 static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
754 						*journal)
755 {
756 	return test_bit(J_ABORTED, &journal->j_state);
757 }
758 
759 /*
760  * Locking primitives. The write lock is a per superblock
761  * special mutex that has properties close to the Big Kernel Lock
762  * which was used in the previous locking scheme.
763  */
764 void reiserfs_write_lock(struct super_block *s);
765 void reiserfs_write_unlock(struct super_block *s);
766 int __must_check reiserfs_write_unlock_nested(struct super_block *s);
767 void reiserfs_write_lock_nested(struct super_block *s, int depth);
768 
769 #ifdef CONFIG_REISERFS_CHECK
770 void reiserfs_lock_check_recursive(struct super_block *s);
771 #else
772 static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
773 #endif
774 
775 /*
776  * Several mutexes depend on the write lock.
777  * However sometimes we want to relax the write lock while we hold
778  * these mutexes, according to the release/reacquire on schedule()
779  * properties of the Bkl that were used.
780  * Reiserfs performances and locking were based on this scheme.
781  * Now that the write lock is a mutex and not the bkl anymore, doing so
782  * may result in a deadlock:
783  *
784  * A acquire write_lock
785  * A acquire j_commit_mutex
786  * A release write_lock and wait for something
787  * B acquire write_lock
788  * B can't acquire j_commit_mutex and sleep
789  * A can't acquire write lock anymore
790  * deadlock
791  *
792  * What we do here is avoiding such deadlock by playing the same game
793  * than the Bkl: if we can't acquire a mutex that depends on the write lock,
794  * we release the write lock, wait a bit and then retry.
795  *
796  * The mutexes concerned by this hack are:
797  * - The commit mutex of a journal list
798  * - The flush mutex
799  * - The journal lock
800  * - The inode mutex
801  */
802 static inline void reiserfs_mutex_lock_safe(struct mutex *m,
803 					    struct super_block *s)
804 {
805 	int depth;
806 
807 	depth = reiserfs_write_unlock_nested(s);
808 	mutex_lock(m);
809 	reiserfs_write_lock_nested(s, depth);
810 }
811 
812 static inline void
813 reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
814 				struct super_block *s)
815 {
816 	int depth;
817 
818 	depth = reiserfs_write_unlock_nested(s);
819 	mutex_lock_nested(m, subclass);
820 	reiserfs_write_lock_nested(s, depth);
821 }
822 
823 static inline void
824 reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
825 {
826        int depth;
827        depth = reiserfs_write_unlock_nested(s);
828        down_read(sem);
829        reiserfs_write_lock_nested(s, depth);
830 }
831 
832 /*
833  * When we schedule, we usually want to also release the write lock,
834  * according to the previous bkl based locking scheme of reiserfs.
835  */
836 static inline void reiserfs_cond_resched(struct super_block *s)
837 {
838 	if (need_resched()) {
839 		int depth;
840 
841 		depth = reiserfs_write_unlock_nested(s);
842 		schedule();
843 		reiserfs_write_lock_nested(s, depth);
844 	}
845 }
846 
847 struct fid;
848 
849 /*
850  * in reading the #defines, it may help to understand that they employ
851  *  the following abbreviations:
852  *
853  *  B = Buffer
854  *  I = Item header
855  *  H = Height within the tree (should be changed to LEV)
856  *  N = Number of the item in the node
857  *  STAT = stat data
858  *  DEH = Directory Entry Header
859  *  EC = Entry Count
860  *  E = Entry number
861  *  UL = Unsigned Long
862  *  BLKH = BLocK Header
863  *  UNFM = UNForMatted node
864  *  DC = Disk Child
865  *  P = Path
866  *
867  *  These #defines are named by concatenating these abbreviations,
868  *  where first comes the arguments, and last comes the return value,
869  *  of the macro.
870  */
871 
872 #define USE_INODE_GENERATION_COUNTER
873 
874 #define REISERFS_PREALLOCATE
875 #define DISPLACE_NEW_PACKING_LOCALITIES
876 #define PREALLOCATION_SIZE 9
877 
878 /* n must be power of 2 */
879 #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
880 
881 /*
882  * to be ok for alpha and others we have to align structures to 8 byte
883  * boundary.
884  * FIXME: do not change 4 by anything else: there is code which relies on that
885  */
886 #define ROUND_UP(x) _ROUND_UP(x,8LL)
887 
888 /*
889  * debug levels.  Right now, CONFIG_REISERFS_CHECK means print all debug
890  * messages.
891  */
892 #define REISERFS_DEBUG_CODE 5	/* extra messages to help find/debug errors */
893 
894 void __reiserfs_warning(struct super_block *s, const char *id,
895 			 const char *func, const char *fmt, ...);
896 #define reiserfs_warning(s, id, fmt, args...) \
897 	 __reiserfs_warning(s, id, __func__, fmt, ##args)
898 /* assertions handling */
899 
900 /* always check a condition and panic if it's false. */
901 #define __RASSERT(cond, scond, format, args...)			\
902 do {									\
903 	if (!(cond))							\
904 		reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
905 			       __FILE__ ":%i:%s: " format "\n",		\
906 			       in_interrupt() ? -1 : task_pid_nr(current), \
907 			       __LINE__, __func__ , ##args);		\
908 } while (0)
909 
910 #define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
911 
912 #if defined( CONFIG_REISERFS_CHECK )
913 #define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
914 #else
915 #define RFALSE( cond, format, args... ) do {;} while( 0 )
916 #endif
917 
918 #define CONSTF __attribute_const__
919 /*
920  * Disk Data Structures
921  */
922 
923 /***************************************************************************
924  *                             SUPER BLOCK                                 *
925  ***************************************************************************/
926 
927 /*
928  * Structure of super block on disk, a version of which in RAM is often
929  * accessed as REISERFS_SB(s)->s_rs. The version in RAM is part of a larger
930  * structure containing fields never written to disk.
931  */
932 #define UNSET_HASH 0	/* Detect hash on disk */
933 #define TEA_HASH  1
934 #define YURA_HASH 2
935 #define R5_HASH   3
936 #define DEFAULT_HASH R5_HASH
937 
938 struct journal_params {
939 	/* where does journal start from on its * device */
940 	__le32 jp_journal_1st_block;
941 
942 	/* journal device st_rdev */
943 	__le32 jp_journal_dev;
944 
945 	/* size of the journal */
946 	__le32 jp_journal_size;
947 
948 	/* max number of blocks in a transaction. */
949 	__le32 jp_journal_trans_max;
950 
951 	/*
952 	 * random value made on fs creation
953 	 * (this was sb_journal_block_count)
954 	 */
955 	__le32 jp_journal_magic;
956 
957 	/* max number of blocks to batch into a trans */
958 	__le32 jp_journal_max_batch;
959 
960 	/* in seconds, how old can an async  commit be */
961 	__le32 jp_journal_max_commit_age;
962 
963 	/* in seconds, how old can a transaction be */
964 	__le32 jp_journal_max_trans_age;
965 };
966 
967 /* this is the super from 3.5.X, where X >= 10 */
968 struct reiserfs_super_block_v1 {
969 	__le32 s_block_count;	/* blocks count         */
970 	__le32 s_free_blocks;	/* free blocks count    */
971 	__le32 s_root_block;	/* root block number    */
972 	struct journal_params s_journal;
973 	__le16 s_blocksize;	/* block size */
974 
975 	/* max size of object id array, see get_objectid() commentary  */
976 	__le16 s_oid_maxsize;
977 	__le16 s_oid_cursize;	/* current size of object id array */
978 
979 	/* this is set to 1 when filesystem was umounted, to 2 - when not */
980 	__le16 s_umount_state;
981 
982 	/*
983 	 * reiserfs magic string indicates that file system is reiserfs:
984 	 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs"
985 	 */
986 	char s_magic[10];
987 
988 	/*
989 	 * it is set to used by fsck to mark which
990 	 * phase of rebuilding is done
991 	 */
992 	__le16 s_fs_state;
993 	/*
994 	 * indicate, what hash function is being use
995 	 * to sort names in a directory
996 	 */
997 	__le32 s_hash_function_code;
998 	__le16 s_tree_height;	/* height of disk tree */
999 
1000 	/*
1001 	 * amount of bitmap blocks needed to address
1002 	 * each block of file system
1003 	 */
1004 	__le16 s_bmap_nr;
1005 
1006 	/*
1007 	 * this field is only reliable on filesystem with non-standard journal
1008 	 */
1009 	__le16 s_version;
1010 
1011 	/*
1012 	 * size in blocks of journal area on main device, we need to
1013 	 * keep after making fs with non-standard journal
1014 	 */
1015 	__le16 s_reserved_for_journal;
1016 } __attribute__ ((__packed__));
1017 
1018 #define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
1019 
1020 /* this is the on disk super block */
1021 struct reiserfs_super_block {
1022 	struct reiserfs_super_block_v1 s_v1;
1023 	__le32 s_inode_generation;
1024 
1025 	/* Right now used only by inode-attributes, if enabled */
1026 	__le32 s_flags;
1027 
1028 	unsigned char s_uuid[16];	/* filesystem unique identifier */
1029 	unsigned char s_label[16];	/* filesystem volume label */
1030 	__le16 s_mnt_count;		/* Count of mounts since last fsck */
1031 	__le16 s_max_mnt_count;		/* Maximum mounts before check */
1032 	__le32 s_lastcheck;		/* Timestamp of last fsck */
1033 	__le32 s_check_interval;	/* Interval between checks */
1034 
1035 	/*
1036 	 * zero filled by mkreiserfs and reiserfs_convert_objectid_map_v1()
1037 	 * so any additions must be updated there as well. */
1038 	char s_unused[76];
1039 } __attribute__ ((__packed__));
1040 
1041 #define SB_SIZE (sizeof(struct reiserfs_super_block))
1042 
1043 #define REISERFS_VERSION_1 0
1044 #define REISERFS_VERSION_2 2
1045 
1046 /* on-disk super block fields converted to cpu form */
1047 #define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
1048 #define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
1049 #define SB_BLOCKSIZE(s) \
1050         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
1051 #define SB_BLOCK_COUNT(s) \
1052         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
1053 #define SB_FREE_BLOCKS(s) \
1054         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
1055 #define SB_REISERFS_MAGIC(s) \
1056         (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
1057 #define SB_ROOT_BLOCK(s) \
1058         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
1059 #define SB_TREE_HEIGHT(s) \
1060         le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
1061 #define SB_REISERFS_STATE(s) \
1062         le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
1063 #define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
1064 #define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
1065 
1066 #define PUT_SB_BLOCK_COUNT(s, val) \
1067    do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
1068 #define PUT_SB_FREE_BLOCKS(s, val) \
1069    do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
1070 #define PUT_SB_ROOT_BLOCK(s, val) \
1071    do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
1072 #define PUT_SB_TREE_HEIGHT(s, val) \
1073    do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
1074 #define PUT_SB_REISERFS_STATE(s, val) \
1075    do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1076 #define PUT_SB_VERSION(s, val) \
1077    do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
1078 #define PUT_SB_BMAP_NR(s, val) \
1079    do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
1080 
1081 #define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
1082 #define SB_ONDISK_JOURNAL_SIZE(s) \
1083          le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
1084 #define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
1085          le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
1086 #define SB_ONDISK_JOURNAL_DEVICE(s) \
1087          le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
1088 #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
1089          le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1090 
1091 #define is_block_in_log_or_reserved_area(s, block) \
1092          block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
1093          && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) +  \
1094          ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
1095          SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1096 
1097 int is_reiserfs_3_5(struct reiserfs_super_block *rs);
1098 int is_reiserfs_3_6(struct reiserfs_super_block *rs);
1099 int is_reiserfs_jr(struct reiserfs_super_block *rs);
1100 
1101 /*
1102  * ReiserFS leaves the first 64k unused, so that partition labels have
1103  * enough space.  If someone wants to write a fancy bootloader that
1104  * needs more than 64k, let us know, and this will be increased in size.
1105  * This number must be larger than than the largest block size on any
1106  * platform, or code will break.  -Hans
1107  */
1108 #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
1109 #define REISERFS_FIRST_BLOCK unused_define
1110 #define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
1111 
1112 /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
1113 #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
1114 
1115 /* reiserfs internal error code (used by search_by_key and fix_nodes)) */
1116 #define CARRY_ON      0
1117 #define REPEAT_SEARCH -1
1118 #define IO_ERROR      -2
1119 #define NO_DISK_SPACE -3
1120 #define NO_BALANCING_NEEDED  (-4)
1121 #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
1122 #define QUOTA_EXCEEDED -6
1123 
1124 typedef __u32 b_blocknr_t;
1125 typedef __le32 unp_t;
1126 
1127 struct unfm_nodeinfo {
1128 	unp_t unfm_nodenum;
1129 	unsigned short unfm_freespace;
1130 };
1131 
1132 /* there are two formats of keys: 3.5 and 3.6 */
1133 #define KEY_FORMAT_3_5 0
1134 #define KEY_FORMAT_3_6 1
1135 
1136 /* there are two stat datas */
1137 #define STAT_DATA_V1 0
1138 #define STAT_DATA_V2 1
1139 
1140 static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
1141 {
1142 	return container_of(inode, struct reiserfs_inode_info, vfs_inode);
1143 }
1144 
1145 static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
1146 {
1147 	return sb->s_fs_info;
1148 }
1149 
1150 /*
1151  * Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
1152  * which overflows on large file systems.
1153  */
1154 static inline __u32 reiserfs_bmap_count(struct super_block *sb)
1155 {
1156 	return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
1157 }
1158 
1159 static inline int bmap_would_wrap(unsigned bmap_nr)
1160 {
1161 	return bmap_nr > ((1LL << 16) - 1);
1162 }
1163 
1164 /*
1165  * this says about version of key of all items (but stat data) the
1166  * object consists of
1167  */
1168 #define get_inode_item_key_version( inode )                                    \
1169     ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
1170 
1171 #define set_inode_item_key_version( inode, version )                           \
1172          ({ if((version)==KEY_FORMAT_3_6)                                      \
1173                 REISERFS_I(inode)->i_flags |= i_item_key_version_mask;      \
1174             else                                                               \
1175                 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
1176 
1177 #define get_inode_sd_version(inode)                                            \
1178     ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
1179 
1180 #define set_inode_sd_version(inode, version)                                   \
1181          ({ if((version)==STAT_DATA_V2)                                        \
1182                 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask;     \
1183             else                                                               \
1184                 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1185 
1186 /*
1187  * This is an aggressive tail suppression policy, I am hoping it
1188  * improves our benchmarks. The principle behind it is that percentage
1189  * space saving is what matters, not absolute space saving.  This is
1190  * non-intuitive, but it helps to understand it if you consider that the
1191  * cost to access 4 blocks is not much more than the cost to access 1
1192  * block, if you have to do a seek and rotate.  A tail risks a
1193  * non-linear disk access that is significant as a percentage of total
1194  * time cost for a 4 block file and saves an amount of space that is
1195  * less significant as a percentage of space, or so goes the hypothesis.
1196  * -Hans
1197  */
1198 #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1199 (\
1200   (!(n_tail_size)) || \
1201   (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1202    ( (n_file_size) >= (n_block_size) * 4 ) || \
1203    ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1204      ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1205    ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1206      ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1207    ( ( (n_file_size) >= (n_block_size) ) && \
1208      ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1209 )
1210 
1211 /*
1212  * Another strategy for tails, this one means only create a tail if all the
1213  * file would fit into one DIRECT item.
1214  * Primary intention for this one is to increase performance by decreasing
1215  * seeking.
1216 */
1217 #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1218 (\
1219   (!(n_tail_size)) || \
1220   (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1221 )
1222 
1223 /*
1224  * values for s_umount_state field
1225  */
1226 #define REISERFS_VALID_FS    1
1227 #define REISERFS_ERROR_FS    2
1228 
1229 /*
1230  * there are 5 item types currently
1231  */
1232 #define TYPE_STAT_DATA 0
1233 #define TYPE_INDIRECT 1
1234 #define TYPE_DIRECT 2
1235 #define TYPE_DIRENTRY 3
1236 #define TYPE_MAXTYPE 3
1237 #define TYPE_ANY 15		/* FIXME: comment is required */
1238 
1239 /***************************************************************************
1240  *                       KEY & ITEM HEAD                                   *
1241  ***************************************************************************/
1242 
1243 /* * directories use this key as well as old files */
1244 struct offset_v1 {
1245 	__le32 k_offset;
1246 	__le32 k_uniqueness;
1247 } __attribute__ ((__packed__));
1248 
1249 struct offset_v2 {
1250 	__le64 v;
1251 } __attribute__ ((__packed__));
1252 
1253 static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1254 {
1255 	__u8 type = le64_to_cpu(v2->v) >> 60;
1256 	return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1257 }
1258 
1259 static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1260 {
1261 	v2->v =
1262 	    (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1263 }
1264 
1265 static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1266 {
1267 	return le64_to_cpu(v2->v) & (~0ULL >> 4);
1268 }
1269 
1270 static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1271 {
1272 	offset &= (~0ULL >> 4);
1273 	v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1274 }
1275 
1276 /*
1277  * Key of an item determines its location in the S+tree, and
1278  * is composed of 4 components
1279  */
1280 struct reiserfs_key {
1281 	/* packing locality: by default parent directory object id */
1282 	__le32 k_dir_id;
1283 
1284 	__le32 k_objectid;	/* object identifier */
1285 	union {
1286 		struct offset_v1 k_offset_v1;
1287 		struct offset_v2 k_offset_v2;
1288 	} __attribute__ ((__packed__)) u;
1289 } __attribute__ ((__packed__));
1290 
1291 struct in_core_key {
1292 	/* packing locality: by default parent directory object id */
1293 	__u32 k_dir_id;
1294 	__u32 k_objectid;	/* object identifier */
1295 	__u64 k_offset;
1296 	__u8 k_type;
1297 };
1298 
1299 struct cpu_key {
1300 	struct in_core_key on_disk_key;
1301 	int version;
1302 	/* 3 in all cases but direct2indirect and indirect2direct conversion */
1303 	int key_length;
1304 };
1305 
1306 /*
1307  * Our function for comparing keys can compare keys of different
1308  * lengths.  It takes as a parameter the length of the keys it is to
1309  * compare.  These defines are used in determining what is to be passed
1310  * to it as that parameter.
1311  */
1312 #define REISERFS_FULL_KEY_LEN     4
1313 #define REISERFS_SHORT_KEY_LEN    2
1314 
1315 /* The result of the key compare */
1316 #define FIRST_GREATER 1
1317 #define SECOND_GREATER -1
1318 #define KEYS_IDENTICAL 0
1319 #define KEY_FOUND 1
1320 #define KEY_NOT_FOUND 0
1321 
1322 #define KEY_SIZE (sizeof(struct reiserfs_key))
1323 #define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
1324 
1325 /* return values for search_by_key and clones */
1326 #define ITEM_FOUND 1
1327 #define ITEM_NOT_FOUND 0
1328 #define ENTRY_FOUND 1
1329 #define ENTRY_NOT_FOUND 0
1330 #define DIRECTORY_NOT_FOUND -1
1331 #define REGULAR_FILE_FOUND -2
1332 #define DIRECTORY_FOUND -3
1333 #define BYTE_FOUND 1
1334 #define BYTE_NOT_FOUND 0
1335 #define FILE_NOT_FOUND -1
1336 
1337 #define POSITION_FOUND 1
1338 #define POSITION_NOT_FOUND 0
1339 
1340 /* return values for reiserfs_find_entry and search_by_entry_key */
1341 #define NAME_FOUND 1
1342 #define NAME_NOT_FOUND 0
1343 #define GOTO_PREVIOUS_ITEM 2
1344 #define NAME_FOUND_INVISIBLE 3
1345 
1346 /*
1347  * Everything in the filesystem is stored as a set of items.  The
1348  * item head contains the key of the item, its free space (for
1349  * indirect items) and specifies the location of the item itself
1350  * within the block.
1351  */
1352 
1353 struct item_head {
1354 	/*
1355 	 * Everything in the tree is found by searching for it based on
1356 	 * its key.
1357 	 */
1358 	struct reiserfs_key ih_key;
1359 	union {
1360 		/*
1361 		 * The free space in the last unformatted node of an
1362 		 * indirect item if this is an indirect item.  This
1363 		 * equals 0xFFFF iff this is a direct item or stat data
1364 		 * item. Note that the key, not this field, is used to
1365 		 * determine the item type, and thus which field this
1366 		 * union contains.
1367 		 */
1368 		__le16 ih_free_space_reserved;
1369 
1370 		/*
1371 		 * Iff this is a directory item, this field equals the
1372 		 * number of directory entries in the directory item.
1373 		 */
1374 		__le16 ih_entry_count;
1375 	} __attribute__ ((__packed__)) u;
1376 	__le16 ih_item_len;	/* total size of the item body */
1377 
1378 	/* an offset to the item body within the block */
1379 	__le16 ih_item_location;
1380 
1381 	/*
1382 	 * 0 for all old items, 2 for new ones. Highest bit is set by fsck
1383 	 * temporary, cleaned after all done
1384 	 */
1385 	__le16 ih_version;
1386 } __attribute__ ((__packed__));
1387 /* size of item header     */
1388 #define IH_SIZE (sizeof(struct item_head))
1389 
1390 #define ih_free_space(ih)            le16_to_cpu((ih)->u.ih_free_space_reserved)
1391 #define ih_version(ih)               le16_to_cpu((ih)->ih_version)
1392 #define ih_entry_count(ih)           le16_to_cpu((ih)->u.ih_entry_count)
1393 #define ih_location(ih)              le16_to_cpu((ih)->ih_item_location)
1394 #define ih_item_len(ih)              le16_to_cpu((ih)->ih_item_len)
1395 
1396 #define put_ih_free_space(ih, val)   do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1397 #define put_ih_version(ih, val)      do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1398 #define put_ih_entry_count(ih, val)  do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1399 #define put_ih_location(ih, val)     do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1400 #define put_ih_item_len(ih, val)     do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1401 
1402 #define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1403 
1404 #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1405 #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1406 
1407 /*
1408  * these operate on indirect items, where you've got an array of ints
1409  * at a possibly unaligned location.  These are a noop on ia32
1410  *
1411  * p is the array of __u32, i is the index into the array, v is the value
1412  * to store there.
1413  */
1414 #define get_block_num(p, i) get_unaligned_le32((p) + (i))
1415 #define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1416 
1417 /* * in old version uniqueness field shows key type */
1418 #define V1_SD_UNIQUENESS 0
1419 #define V1_INDIRECT_UNIQUENESS 0xfffffffe
1420 #define V1_DIRECT_UNIQUENESS 0xffffffff
1421 #define V1_DIRENTRY_UNIQUENESS 500
1422 #define V1_ANY_UNIQUENESS 555	/* FIXME: comment is required */
1423 
1424 /* here are conversion routines */
1425 static inline int uniqueness2type(__u32 uniqueness) CONSTF;
1426 static inline int uniqueness2type(__u32 uniqueness)
1427 {
1428 	switch ((int)uniqueness) {
1429 	case V1_SD_UNIQUENESS:
1430 		return TYPE_STAT_DATA;
1431 	case V1_INDIRECT_UNIQUENESS:
1432 		return TYPE_INDIRECT;
1433 	case V1_DIRECT_UNIQUENESS:
1434 		return TYPE_DIRECT;
1435 	case V1_DIRENTRY_UNIQUENESS:
1436 		return TYPE_DIRENTRY;
1437 	case V1_ANY_UNIQUENESS:
1438 	default:
1439 		return TYPE_ANY;
1440 	}
1441 }
1442 
1443 static inline __u32 type2uniqueness(int type) CONSTF;
1444 static inline __u32 type2uniqueness(int type)
1445 {
1446 	switch (type) {
1447 	case TYPE_STAT_DATA:
1448 		return V1_SD_UNIQUENESS;
1449 	case TYPE_INDIRECT:
1450 		return V1_INDIRECT_UNIQUENESS;
1451 	case TYPE_DIRECT:
1452 		return V1_DIRECT_UNIQUENESS;
1453 	case TYPE_DIRENTRY:
1454 		return V1_DIRENTRY_UNIQUENESS;
1455 	case TYPE_ANY:
1456 	default:
1457 		return V1_ANY_UNIQUENESS;
1458 	}
1459 }
1460 
1461 /*
1462  * key is pointer to on disk key which is stored in le, result is cpu,
1463  * there is no way to get version of object from key, so, provide
1464  * version to these defines
1465  */
1466 static inline loff_t le_key_k_offset(int version,
1467 				     const struct reiserfs_key *key)
1468 {
1469 	return (version == KEY_FORMAT_3_5) ?
1470 	    le32_to_cpu(key->u.k_offset_v1.k_offset) :
1471 	    offset_v2_k_offset(&(key->u.k_offset_v2));
1472 }
1473 
1474 static inline loff_t le_ih_k_offset(const struct item_head *ih)
1475 {
1476 	return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1477 }
1478 
1479 static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1480 {
1481 	if (version == KEY_FORMAT_3_5) {
1482 		loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness);
1483 		return uniqueness2type(val);
1484 	} else
1485 		return offset_v2_k_type(&(key->u.k_offset_v2));
1486 }
1487 
1488 static inline loff_t le_ih_k_type(const struct item_head *ih)
1489 {
1490 	return le_key_k_type(ih_version(ih), &(ih->ih_key));
1491 }
1492 
1493 static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1494 				       loff_t offset)
1495 {
1496 	if (version == KEY_FORMAT_3_5)
1497 		key->u.k_offset_v1.k_offset = cpu_to_le32(offset);
1498 	else
1499 		set_offset_v2_k_offset(&key->u.k_offset_v2, offset);
1500 }
1501 
1502 static inline void add_le_key_k_offset(int version, struct reiserfs_key *key,
1503 				       loff_t offset)
1504 {
1505 	set_le_key_k_offset(version, key,
1506 			    le_key_k_offset(version, key) + offset);
1507 }
1508 
1509 static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset)
1510 {
1511 	add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1512 }
1513 
1514 static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1515 {
1516 	set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1517 }
1518 
1519 static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1520 				     int type)
1521 {
1522 	if (version == KEY_FORMAT_3_5) {
1523 		type = type2uniqueness(type);
1524 		key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type);
1525 	} else
1526 	       set_offset_v2_k_type(&key->u.k_offset_v2, type);
1527 }
1528 
1529 static inline void set_le_ih_k_type(struct item_head *ih, int type)
1530 {
1531 	set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1532 }
1533 
1534 static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1535 {
1536 	return le_key_k_type(version, key) == TYPE_DIRENTRY;
1537 }
1538 
1539 static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1540 {
1541 	return le_key_k_type(version, key) == TYPE_DIRECT;
1542 }
1543 
1544 static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1545 {
1546 	return le_key_k_type(version, key) == TYPE_INDIRECT;
1547 }
1548 
1549 static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1550 {
1551 	return le_key_k_type(version, key) == TYPE_STAT_DATA;
1552 }
1553 
1554 /* item header has version.  */
1555 static inline int is_direntry_le_ih(struct item_head *ih)
1556 {
1557 	return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1558 }
1559 
1560 static inline int is_direct_le_ih(struct item_head *ih)
1561 {
1562 	return is_direct_le_key(ih_version(ih), &ih->ih_key);
1563 }
1564 
1565 static inline int is_indirect_le_ih(struct item_head *ih)
1566 {
1567 	return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1568 }
1569 
1570 static inline int is_statdata_le_ih(struct item_head *ih)
1571 {
1572 	return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1573 }
1574 
1575 /* key is pointer to cpu key, result is cpu */
1576 static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1577 {
1578 	return key->on_disk_key.k_offset;
1579 }
1580 
1581 static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1582 {
1583 	return key->on_disk_key.k_type;
1584 }
1585 
1586 static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1587 {
1588 	key->on_disk_key.k_offset = offset;
1589 }
1590 
1591 static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1592 {
1593 	key->on_disk_key.k_type = type;
1594 }
1595 
1596 static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1597 {
1598 	key->on_disk_key.k_offset--;
1599 }
1600 
1601 #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1602 #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1603 #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1604 #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1605 
1606 /* are these used ? */
1607 #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1608 #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1609 #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1610 #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1611 
1612 #define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1613     (!COMP_SHORT_KEYS(ih, key) && \
1614 	  I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1615 
1616 /* maximal length of item */
1617 #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1618 #define MIN_ITEM_LEN 1
1619 
1620 /* object identifier for root dir */
1621 #define REISERFS_ROOT_OBJECTID 2
1622 #define REISERFS_ROOT_PARENT_OBJECTID 1
1623 
1624 extern struct reiserfs_key root_key;
1625 
1626 /*
1627  * Picture represents a leaf of the S+tree
1628  *  ______________________________________________________
1629  * |      |  Array of     |                   |           |
1630  * |Block |  Object-Item  |      F r e e      |  Objects- |
1631  * | head |  Headers      |     S p a c e     |   Items   |
1632  * |______|_______________|___________________|___________|
1633  */
1634 
1635 /*
1636  * Header of a disk block.  More precisely, header of a formatted leaf
1637  * or internal node, and not the header of an unformatted node.
1638  */
1639 struct block_head {
1640 	__le16 blk_level;	/* Level of a block in the tree. */
1641 	__le16 blk_nr_item;	/* Number of keys/items in a block. */
1642 	__le16 blk_free_space;	/* Block free space in bytes. */
1643 	__le16 blk_reserved;
1644 	/* dump this in v4/planA */
1645 
1646 	/* kept only for compatibility */
1647 	struct reiserfs_key blk_right_delim_key;
1648 };
1649 
1650 #define BLKH_SIZE                     (sizeof(struct block_head))
1651 #define blkh_level(p_blkh)            (le16_to_cpu((p_blkh)->blk_level))
1652 #define blkh_nr_item(p_blkh)          (le16_to_cpu((p_blkh)->blk_nr_item))
1653 #define blkh_free_space(p_blkh)       (le16_to_cpu((p_blkh)->blk_free_space))
1654 #define blkh_reserved(p_blkh)         (le16_to_cpu((p_blkh)->blk_reserved))
1655 #define set_blkh_level(p_blkh,val)    ((p_blkh)->blk_level = cpu_to_le16(val))
1656 #define set_blkh_nr_item(p_blkh,val)  ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1657 #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1658 #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1659 #define blkh_right_delim_key(p_blkh)  ((p_blkh)->blk_right_delim_key)
1660 #define set_blkh_right_delim_key(p_blkh,val)  ((p_blkh)->blk_right_delim_key = val)
1661 
1662 /* values for blk_level field of the struct block_head */
1663 
1664 /*
1665  * When node gets removed from the tree its blk_level is set to FREE_LEVEL.
1666  * It is then  used to see whether the node is still in the tree
1667  */
1668 #define FREE_LEVEL 0
1669 
1670 #define DISK_LEAF_NODE_LEVEL  1	/* Leaf node level. */
1671 
1672 /*
1673  * Given the buffer head of a formatted node, resolve to the
1674  * block head of that node.
1675  */
1676 #define B_BLK_HEAD(bh)			((struct block_head *)((bh)->b_data))
1677 /* Number of items that are in buffer. */
1678 #define B_NR_ITEMS(bh)			(blkh_nr_item(B_BLK_HEAD(bh)))
1679 #define B_LEVEL(bh)			(blkh_level(B_BLK_HEAD(bh)))
1680 #define B_FREE_SPACE(bh)		(blkh_free_space(B_BLK_HEAD(bh)))
1681 
1682 #define PUT_B_NR_ITEMS(bh, val)		do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1683 #define PUT_B_LEVEL(bh, val)		do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1684 #define PUT_B_FREE_SPACE(bh, val)	do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1685 
1686 /* Get right delimiting key. -- little endian */
1687 #define B_PRIGHT_DELIM_KEY(bh)		(&(blk_right_delim_key(B_BLK_HEAD(bh))))
1688 
1689 /* Does the buffer contain a disk leaf. */
1690 #define B_IS_ITEMS_LEVEL(bh)		(B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1691 
1692 /* Does the buffer contain a disk internal node */
1693 #define B_IS_KEYS_LEVEL(bh)      (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1694 					    && B_LEVEL(bh) <= MAX_HEIGHT)
1695 
1696 /***************************************************************************
1697  *                             STAT DATA                                   *
1698  ***************************************************************************/
1699 
1700 /*
1701  * old stat data is 32 bytes long. We are going to distinguish new one by
1702  * different size
1703 */
1704 struct stat_data_v1 {
1705 	__le16 sd_mode;		/* file type, permissions */
1706 	__le16 sd_nlink;	/* number of hard links */
1707 	__le16 sd_uid;		/* owner */
1708 	__le16 sd_gid;		/* group */
1709 	__le32 sd_size;		/* file size */
1710 	__le32 sd_atime;	/* time of last access */
1711 	__le32 sd_mtime;	/* time file was last modified  */
1712 
1713 	/*
1714 	 * time inode (stat data) was last changed
1715 	 * (except changes to sd_atime and sd_mtime)
1716 	 */
1717 	__le32 sd_ctime;
1718 	union {
1719 		__le32 sd_rdev;
1720 		__le32 sd_blocks;	/* number of blocks file uses */
1721 	} __attribute__ ((__packed__)) u;
1722 
1723 	/*
1724 	 * first byte of file which is stored in a direct item: except that if
1725 	 * it equals 1 it is a symlink and if it equals ~(__u32)0 there is no
1726 	 * direct item.  The existence of this field really grates on me.
1727 	 * Let's replace it with a macro based on sd_size and our tail
1728 	 * suppression policy.  Someday.  -Hans
1729 	 */
1730 	__le32 sd_first_direct_byte;
1731 } __attribute__ ((__packed__));
1732 
1733 #define SD_V1_SIZE              (sizeof(struct stat_data_v1))
1734 #define stat_data_v1(ih)        (ih_version (ih) == KEY_FORMAT_3_5)
1735 #define sd_v1_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))
1736 #define set_sd_v1_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))
1737 #define sd_v1_nlink(sdp)        (le16_to_cpu((sdp)->sd_nlink))
1738 #define set_sd_v1_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le16(v))
1739 #define sd_v1_uid(sdp)          (le16_to_cpu((sdp)->sd_uid))
1740 #define set_sd_v1_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le16(v))
1741 #define sd_v1_gid(sdp)          (le16_to_cpu((sdp)->sd_gid))
1742 #define set_sd_v1_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le16(v))
1743 #define sd_v1_size(sdp)         (le32_to_cpu((sdp)->sd_size))
1744 #define set_sd_v1_size(sdp,v)   ((sdp)->sd_size = cpu_to_le32(v))
1745 #define sd_v1_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))
1746 #define set_sd_v1_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))
1747 #define sd_v1_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))
1748 #define set_sd_v1_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))
1749 #define sd_v1_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))
1750 #define set_sd_v1_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))
1751 #define sd_v1_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))
1752 #define set_sd_v1_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))
1753 #define sd_v1_blocks(sdp)       (le32_to_cpu((sdp)->u.sd_blocks))
1754 #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1755 #define sd_v1_first_direct_byte(sdp) \
1756                                 (le32_to_cpu((sdp)->sd_first_direct_byte))
1757 #define set_sd_v1_first_direct_byte(sdp,v) \
1758                                 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1759 
1760 /* inode flags stored in sd_attrs (nee sd_reserved) */
1761 
1762 /*
1763  * we want common flags to have the same values as in ext2,
1764  * so chattr(1) will work without problems
1765  */
1766 #define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1767 #define REISERFS_APPEND_FL    FS_APPEND_FL
1768 #define REISERFS_SYNC_FL      FS_SYNC_FL
1769 #define REISERFS_NOATIME_FL   FS_NOATIME_FL
1770 #define REISERFS_NODUMP_FL    FS_NODUMP_FL
1771 #define REISERFS_SECRM_FL     FS_SECRM_FL
1772 #define REISERFS_UNRM_FL      FS_UNRM_FL
1773 #define REISERFS_COMPR_FL     FS_COMPR_FL
1774 #define REISERFS_NOTAIL_FL    FS_NOTAIL_FL
1775 
1776 /* persistent flags that file inherits from the parent directory */
1777 #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL |	\
1778 				REISERFS_SYNC_FL |	\
1779 				REISERFS_NOATIME_FL |	\
1780 				REISERFS_NODUMP_FL |	\
1781 				REISERFS_SECRM_FL |	\
1782 				REISERFS_COMPR_FL |	\
1783 				REISERFS_NOTAIL_FL )
1784 
1785 /*
1786  * Stat Data on disk (reiserfs version of UFS disk inode minus the
1787  * address blocks)
1788  */
1789 struct stat_data {
1790 	__le16 sd_mode;		/* file type, permissions */
1791 	__le16 sd_attrs;	/* persistent inode flags */
1792 	__le32 sd_nlink;	/* number of hard links */
1793 	__le64 sd_size;		/* file size */
1794 	__le32 sd_uid;		/* owner */
1795 	__le32 sd_gid;		/* group */
1796 	__le32 sd_atime;	/* time of last access */
1797 	__le32 sd_mtime;	/* time file was last modified  */
1798 
1799 	/*
1800 	 * time inode (stat data) was last changed
1801 	 * (except changes to sd_atime and sd_mtime)
1802 	 */
1803 	__le32 sd_ctime;
1804 	__le32 sd_blocks;
1805 	union {
1806 		__le32 sd_rdev;
1807 		__le32 sd_generation;
1808 	} __attribute__ ((__packed__)) u;
1809 } __attribute__ ((__packed__));
1810 
1811 /* this is 44 bytes long */
1812 #define SD_SIZE (sizeof(struct stat_data))
1813 #define SD_V2_SIZE              SD_SIZE
1814 #define stat_data_v2(ih)        (ih_version (ih) == KEY_FORMAT_3_6)
1815 #define sd_v2_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))
1816 #define set_sd_v2_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))
1817 /* sd_reserved */
1818 /* set_sd_reserved */
1819 #define sd_v2_nlink(sdp)        (le32_to_cpu((sdp)->sd_nlink))
1820 #define set_sd_v2_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le32(v))
1821 #define sd_v2_size(sdp)         (le64_to_cpu((sdp)->sd_size))
1822 #define set_sd_v2_size(sdp,v)   ((sdp)->sd_size = cpu_to_le64(v))
1823 #define sd_v2_uid(sdp)          (le32_to_cpu((sdp)->sd_uid))
1824 #define set_sd_v2_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le32(v))
1825 #define sd_v2_gid(sdp)          (le32_to_cpu((sdp)->sd_gid))
1826 #define set_sd_v2_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le32(v))
1827 #define sd_v2_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))
1828 #define set_sd_v2_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))
1829 #define sd_v2_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))
1830 #define set_sd_v2_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))
1831 #define sd_v2_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))
1832 #define set_sd_v2_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))
1833 #define sd_v2_blocks(sdp)       (le32_to_cpu((sdp)->sd_blocks))
1834 #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1835 #define sd_v2_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))
1836 #define set_sd_v2_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))
1837 #define sd_v2_generation(sdp)   (le32_to_cpu((sdp)->u.sd_generation))
1838 #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1839 #define sd_v2_attrs(sdp)         (le16_to_cpu((sdp)->sd_attrs))
1840 #define set_sd_v2_attrs(sdp,v)   ((sdp)->sd_attrs = cpu_to_le16(v))
1841 
1842 /***************************************************************************
1843  *                      DIRECTORY STRUCTURE                                *
1844  ***************************************************************************/
1845 /*
1846  * Picture represents the structure of directory items
1847  * ________________________________________________
1848  * |  Array of     |   |     |        |       |   |
1849  * | directory     |N-1| N-2 | ....   |   1st |0th|
1850  * | entry headers |   |     |        |       |   |
1851  * |_______________|___|_____|________|_______|___|
1852  *                  <----   directory entries         ------>
1853  *
1854  * First directory item has k_offset component 1. We store "." and ".."
1855  * in one item, always, we never split "." and ".." into differing
1856  * items.  This makes, among other things, the code for removing
1857  * directories simpler.
1858  */
1859 #define SD_OFFSET  0
1860 #define SD_UNIQUENESS 0
1861 #define DOT_OFFSET 1
1862 #define DOT_DOT_OFFSET 2
1863 #define DIRENTRY_UNIQUENESS 500
1864 
1865 #define FIRST_ITEM_OFFSET 1
1866 
1867 /*
1868  * Q: How to get key of object pointed to by entry from entry?
1869  *
1870  * A: Each directory entry has its header. This header has deh_dir_id
1871  *    and deh_objectid fields, those are key of object, entry points to
1872  */
1873 
1874 /*
1875  * NOT IMPLEMENTED:
1876  * Directory will someday contain stat data of object
1877  */
1878 
1879 struct reiserfs_de_head {
1880 	__le32 deh_offset;	/* third component of the directory entry key */
1881 
1882 	/*
1883 	 * objectid of the parent directory of the object, that is referenced
1884 	 * by directory entry
1885 	 */
1886 	__le32 deh_dir_id;
1887 
1888 	/* objectid of the object, that is referenced by directory entry */
1889 	__le32 deh_objectid;
1890 	__le16 deh_location;	/* offset of name in the whole item */
1891 
1892 	/*
1893 	 * whether 1) entry contains stat data (for future), and
1894 	 * 2) whether entry is hidden (unlinked)
1895 	 */
1896 	__le16 deh_state;
1897 } __attribute__ ((__packed__));
1898 #define DEH_SIZE                  sizeof(struct reiserfs_de_head)
1899 #define deh_offset(p_deh)         (le32_to_cpu((p_deh)->deh_offset))
1900 #define deh_dir_id(p_deh)         (le32_to_cpu((p_deh)->deh_dir_id))
1901 #define deh_objectid(p_deh)       (le32_to_cpu((p_deh)->deh_objectid))
1902 #define deh_location(p_deh)       (le16_to_cpu((p_deh)->deh_location))
1903 #define deh_state(p_deh)          (le16_to_cpu((p_deh)->deh_state))
1904 
1905 #define put_deh_offset(p_deh,v)   ((p_deh)->deh_offset = cpu_to_le32((v)))
1906 #define put_deh_dir_id(p_deh,v)   ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1907 #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1908 #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1909 #define put_deh_state(p_deh,v)    ((p_deh)->deh_state = cpu_to_le16((v)))
1910 
1911 /* empty directory contains two entries "." and ".." and their headers */
1912 #define EMPTY_DIR_SIZE \
1913 (DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1914 
1915 /* old format directories have this size when empty */
1916 #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1917 
1918 #define DEH_Statdata 0		/* not used now */
1919 #define DEH_Visible 2
1920 
1921 /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1922 #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1923 #   define ADDR_UNALIGNED_BITS  (3)
1924 #endif
1925 
1926 /*
1927  * These are only used to manipulate deh_state.
1928  * Because of this, we'll use the ext2_ bit routines,
1929  * since they are little endian
1930  */
1931 #ifdef ADDR_UNALIGNED_BITS
1932 
1933 #   define aligned_address(addr)           ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1934 #   define unaligned_offset(addr)          (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1935 
1936 #   define set_bit_unaligned(nr, addr)	\
1937 	__test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1938 #   define clear_bit_unaligned(nr, addr)	\
1939 	__test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1940 #   define test_bit_unaligned(nr, addr)	\
1941 	test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1942 
1943 #else
1944 
1945 #   define set_bit_unaligned(nr, addr)	__test_and_set_bit_le(nr, addr)
1946 #   define clear_bit_unaligned(nr, addr)	__test_and_clear_bit_le(nr, addr)
1947 #   define test_bit_unaligned(nr, addr)	test_bit_le(nr, addr)
1948 
1949 #endif
1950 
1951 #define mark_de_with_sd(deh)        set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1952 #define mark_de_without_sd(deh)     clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1953 #define mark_de_visible(deh)	    set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1954 #define mark_de_hidden(deh)	    clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1955 
1956 #define de_with_sd(deh)		    test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1957 #define de_visible(deh)	    	    test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1958 #define de_hidden(deh)	    	    !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1959 
1960 extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1961 				   __le32 par_dirid, __le32 par_objid);
1962 extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1963 				__le32 par_dirid, __le32 par_objid);
1964 
1965 /* two entries per block (at least) */
1966 #define REISERFS_MAX_NAME(block_size) 255
1967 
1968 /*
1969  * this structure is used for operations on directory entries. It is
1970  * not a disk structure.
1971  *
1972  * When reiserfs_find_entry or search_by_entry_key find directory
1973  * entry, they return filled reiserfs_dir_entry structure
1974  */
1975 struct reiserfs_dir_entry {
1976 	struct buffer_head *de_bh;
1977 	int de_item_num;
1978 	struct item_head *de_ih;
1979 	int de_entry_num;
1980 	struct reiserfs_de_head *de_deh;
1981 	int de_entrylen;
1982 	int de_namelen;
1983 	char *de_name;
1984 	unsigned long *de_gen_number_bit_string;
1985 
1986 	__u32 de_dir_id;
1987 	__u32 de_objectid;
1988 
1989 	struct cpu_key de_entry_key;
1990 };
1991 
1992 /*
1993  * these defines are useful when a particular member of
1994  * a reiserfs_dir_entry is needed
1995  */
1996 
1997 /* pointer to file name, stored in entry */
1998 #define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \
1999 				(ih_item_body(bh, ih) + deh_location(deh))
2000 
2001 /* length of name */
2002 #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
2003 (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
2004 
2005 /* hash value occupies bits from 7 up to 30 */
2006 #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
2007 /* generation number occupies 7 bits starting from 0 up to 6 */
2008 #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
2009 #define MAX_GENERATION_NUMBER  127
2010 
2011 #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
2012 
2013 /*
2014  * Picture represents an internal node of the reiserfs tree
2015  *  ______________________________________________________
2016  * |      |  Array of     |  Array of         |  Free     |
2017  * |block |    keys       |  pointers         | space     |
2018  * | head |      N        |      N+1          |           |
2019  * |______|_______________|___________________|___________|
2020  */
2021 
2022 /***************************************************************************
2023  *                      DISK CHILD                                         *
2024  ***************************************************************************/
2025 /*
2026  * Disk child pointer:
2027  * The pointer from an internal node of the tree to a node that is on disk.
2028  */
2029 struct disk_child {
2030 	__le32 dc_block_number;	/* Disk child's block number. */
2031 	__le16 dc_size;		/* Disk child's used space.   */
2032 	__le16 dc_reserved;
2033 };
2034 
2035 #define DC_SIZE (sizeof(struct disk_child))
2036 #define dc_block_number(dc_p)	(le32_to_cpu((dc_p)->dc_block_number))
2037 #define dc_size(dc_p)		(le16_to_cpu((dc_p)->dc_size))
2038 #define put_dc_block_number(dc_p, val)   do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
2039 #define put_dc_size(dc_p, val)   do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
2040 
2041 /* Get disk child by buffer header and position in the tree node. */
2042 #define B_N_CHILD(bh, n_pos)  ((struct disk_child *)\
2043 ((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
2044 
2045 /* Get disk child number by buffer header and position in the tree node. */
2046 #define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
2047 #define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
2048 				(put_dc_block_number(B_N_CHILD(bh, n_pos), val))
2049 
2050  /* maximal value of field child_size in structure disk_child */
2051  /* child size is the combined size of all items and their headers */
2052 #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
2053 
2054 /* amount of used space in buffer (not including block head) */
2055 #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
2056 
2057 /* max and min number of keys in internal node */
2058 #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
2059 #define MIN_NR_KEY(bh)    (MAX_NR_KEY(bh)/2)
2060 
2061 /***************************************************************************
2062  *                      PATH STRUCTURES AND DEFINES                        *
2063  ***************************************************************************/
2064 
2065 /*
2066  * search_by_key fills up the path from the root to the leaf as it descends
2067  * the tree looking for the key.  It uses reiserfs_bread to try to find
2068  * buffers in the cache given their block number.  If it does not find
2069  * them in the cache it reads them from disk.  For each node search_by_key
2070  * finds using reiserfs_bread it then uses bin_search to look through that
2071  * node.  bin_search will find the position of the block_number of the next
2072  * node if it is looking through an internal node.  If it is looking through
2073  * a leaf node bin_search will find the position of the item which has key
2074  * either equal to given key, or which is the maximal key less than the
2075  * given key.
2076  */
2077 
2078 struct path_element {
2079 	/* Pointer to the buffer at the path in the tree. */
2080 	struct buffer_head *pe_buffer;
2081 	/* Position in the tree node which is placed in the buffer above. */
2082 	int pe_position;
2083 };
2084 
2085 /*
2086  * maximal height of a tree. don't change this without
2087  * changing JOURNAL_PER_BALANCE_CNT
2088  */
2089 #define MAX_HEIGHT 5
2090 
2091 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
2092 #define EXTENDED_MAX_HEIGHT         7
2093 
2094 /* Must be equal to at least 2. */
2095 #define FIRST_PATH_ELEMENT_OFFSET   2
2096 
2097 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
2098 #define ILLEGAL_PATH_ELEMENT_OFFSET 1
2099 
2100 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
2101 #define MAX_FEB_SIZE 6
2102 
2103 /*
2104  * We need to keep track of who the ancestors of nodes are.  When we
2105  * perform a search we record which nodes were visited while
2106  * descending the tree looking for the node we searched for. This list
2107  * of nodes is called the path.  This information is used while
2108  * performing balancing.  Note that this path information may become
2109  * invalid, and this means we must check it when using it to see if it
2110  * is still valid. You'll need to read search_by_key and the comments
2111  * in it, especially about decrement_counters_in_path(), to understand
2112  * this structure.
2113  *
2114  * Paths make the code so much harder to work with and debug.... An
2115  * enormous number of bugs are due to them, and trying to write or modify
2116  * code that uses them just makes my head hurt.  They are based on an
2117  * excessive effort to avoid disturbing the precious VFS code.:-( The
2118  * gods only know how we are going to SMP the code that uses them.
2119  * znodes are the way!
2120  */
2121 
2122 #define PATH_READA	0x1	/* do read ahead */
2123 #define PATH_READA_BACK 0x2	/* read backwards */
2124 
2125 struct treepath {
2126 	int path_length;	/* Length of the array above.   */
2127 	int reada;
2128 	/* Array of the path elements.  */
2129 	struct path_element path_elements[EXTENDED_MAX_HEIGHT];
2130 	int pos_in_item;
2131 };
2132 
2133 #define pos_in_item(path) ((path)->pos_in_item)
2134 
2135 #define INITIALIZE_PATH(var) \
2136 struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
2137 
2138 /* Get path element by path and path position. */
2139 #define PATH_OFFSET_PELEMENT(path, n_offset)  ((path)->path_elements + (n_offset))
2140 
2141 /* Get buffer header at the path by path and path position. */
2142 #define PATH_OFFSET_PBUFFER(path, n_offset)   (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
2143 
2144 /* Get position in the element at the path by path and path position. */
2145 #define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
2146 
2147 #define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
2148 
2149 /*
2150  * you know, to the person who didn't write this the macro name does not
2151  * at first suggest what it does.  Maybe POSITION_FROM_PATH_END? Or
2152  * maybe we should just focus on dumping paths... -Hans
2153  */
2154 #define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
2155 
2156 /*
2157  * in do_balance leaf has h == 0 in contrast with path structure,
2158  * where root has level == 0. That is why we need these defines
2159  */
2160 
2161 /* tb->S[h] */
2162 #define PATH_H_PBUFFER(path, h) \
2163 			PATH_OFFSET_PBUFFER(path, path->path_length - (h))
2164 
2165 /* tb->F[h] or tb->S[0]->b_parent */
2166 #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER(path, (h) + 1)
2167 
2168 #define PATH_H_POSITION(path, h) \
2169 			PATH_OFFSET_POSITION(path, path->path_length - (h))
2170 
2171 /* tb->S[h]->b_item_order */
2172 #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)
2173 
2174 #define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
2175 
2176 static inline void *reiserfs_node_data(const struct buffer_head *bh)
2177 {
2178 	return bh->b_data + sizeof(struct block_head);
2179 }
2180 
2181 /* get key from internal node */
2182 static inline struct reiserfs_key *internal_key(struct buffer_head *bh,
2183 						int item_num)
2184 {
2185 	struct reiserfs_key *key = reiserfs_node_data(bh);
2186 
2187 	return &key[item_num];
2188 }
2189 
2190 /* get the item header from leaf node */
2191 static inline struct item_head *item_head(const struct buffer_head *bh,
2192 					  int item_num)
2193 {
2194 	struct item_head *ih = reiserfs_node_data(bh);
2195 
2196 	return &ih[item_num];
2197 }
2198 
2199 /* get the key from leaf node */
2200 static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh,
2201 					    int item_num)
2202 {
2203 	return &item_head(bh, item_num)->ih_key;
2204 }
2205 
2206 static inline void *ih_item_body(const struct buffer_head *bh,
2207 				 const struct item_head *ih)
2208 {
2209 	return bh->b_data + ih_location(ih);
2210 }
2211 
2212 /* get item body from leaf node */
2213 static inline void *item_body(const struct buffer_head *bh, int item_num)
2214 {
2215 	return ih_item_body(bh, item_head(bh, item_num));
2216 }
2217 
2218 static inline struct item_head *tp_item_head(const struct treepath *path)
2219 {
2220 	return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2221 }
2222 
2223 static inline void *tp_item_body(const struct treepath *path)
2224 {
2225 	return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2226 }
2227 
2228 #define get_last_bh(path) PATH_PLAST_BUFFER(path)
2229 #define get_item_pos(path) PATH_LAST_POSITION(path)
2230 #define item_moved(ih,path) comp_items(ih, path)
2231 #define path_changed(ih,path) comp_items (ih, path)
2232 
2233 /* array of the entry headers */
2234  /* get item body */
2235 #define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih)))
2236 
2237 /*
2238  * length of the directory entry in directory item. This define
2239  * calculates length of i-th directory entry using directory entry
2240  * locations from dir entry head. When it calculates length of 0-th
2241  * directory entry, it uses length of whole item in place of entry
2242  * location of the non-existent following entry in the calculation.
2243  * See picture above.
2244  */
2245 static inline int entry_length(const struct buffer_head *bh,
2246 			       const struct item_head *ih, int pos_in_item)
2247 {
2248 	struct reiserfs_de_head *deh;
2249 
2250 	deh = B_I_DEH(bh, ih) + pos_in_item;
2251 	if (pos_in_item)
2252 		return deh_location(deh - 1) - deh_location(deh);
2253 
2254 	return ih_item_len(ih) - deh_location(deh);
2255 }
2256 
2257 /***************************************************************************
2258  *                       MISC                                              *
2259  ***************************************************************************/
2260 
2261 /* Size of pointer to the unformatted node. */
2262 #define UNFM_P_SIZE (sizeof(unp_t))
2263 #define UNFM_P_SHIFT 2
2264 
2265 /* in in-core inode key is stored on le form */
2266 #define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
2267 
2268 #define MAX_UL_INT 0xffffffff
2269 #define MAX_INT    0x7ffffff
2270 #define MAX_US_INT 0xffff
2271 
2272 // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
2273 static inline loff_t max_reiserfs_offset(struct inode *inode)
2274 {
2275 	if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
2276 		return (loff_t) U32_MAX;
2277 
2278 	return (loff_t) ((~(__u64) 0) >> 4);
2279 }
2280 
2281 #define MAX_KEY_OBJECTID	MAX_UL_INT
2282 
2283 #define MAX_B_NUM  MAX_UL_INT
2284 #define MAX_FC_NUM MAX_US_INT
2285 
2286 /* the purpose is to detect overflow of an unsigned short */
2287 #define REISERFS_LINK_MAX (MAX_US_INT - 1000)
2288 
2289 /*
2290  * The following defines are used in reiserfs_insert_item
2291  * and reiserfs_append_item
2292  */
2293 #define REISERFS_KERNEL_MEM		0	/* kernel memory mode */
2294 #define REISERFS_USER_MEM		1	/* user memory mode */
2295 
2296 #define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
2297 #define get_generation(s) atomic_read (&fs_generation(s))
2298 #define FILESYSTEM_CHANGED_TB(tb)  (get_generation((tb)->tb_sb) != (tb)->fs_gen)
2299 #define __fs_changed(gen,s) (gen != get_generation (s))
2300 #define fs_changed(gen,s)		\
2301 ({					\
2302 	reiserfs_cond_resched(s);	\
2303 	__fs_changed(gen, s);		\
2304 })
2305 
2306 /***************************************************************************
2307  *                  FIXATE NODES                                           *
2308  ***************************************************************************/
2309 
2310 #define VI_TYPE_LEFT_MERGEABLE 1
2311 #define VI_TYPE_RIGHT_MERGEABLE 2
2312 
2313 /*
2314  * To make any changes in the tree we always first find node, that
2315  * contains item to be changed/deleted or place to insert a new
2316  * item. We call this node S. To do balancing we need to decide what
2317  * we will shift to left/right neighbor, or to a new node, where new
2318  * item will be etc. To make this analysis simpler we build virtual
2319  * node. Virtual node is an array of items, that will replace items of
2320  * node S. (For instance if we are going to delete an item, virtual
2321  * node does not contain it). Virtual node keeps information about
2322  * item sizes and types, mergeability of first and last items, sizes
2323  * of all entries in directory item. We use this array of items when
2324  * calculating what we can shift to neighbors and how many nodes we
2325  * have to have if we do not any shiftings, if we shift to left/right
2326  * neighbor or to both.
2327  */
2328 struct virtual_item {
2329 	int vi_index;		/* index in the array of item operations */
2330 	unsigned short vi_type;	/* left/right mergeability */
2331 
2332 	/* length of item that it will have after balancing */
2333 	unsigned short vi_item_len;
2334 
2335 	struct item_head *vi_ih;
2336 	const char *vi_item;	/* body of item (old or new) */
2337 	const void *vi_new_data;	/* 0 always but paste mode */
2338 	void *vi_uarea;		/* item specific area */
2339 };
2340 
2341 struct virtual_node {
2342 	/* this is a pointer to the free space in the buffer */
2343 	char *vn_free_ptr;
2344 
2345 	unsigned short vn_nr_item;	/* number of items in virtual node */
2346 
2347 	/*
2348 	 * size of node , that node would have if it has
2349 	 * unlimited size and no balancing is performed
2350 	 */
2351 	short vn_size;
2352 
2353 	/* mode of balancing (paste, insert, delete, cut) */
2354 	short vn_mode;
2355 
2356 	short vn_affected_item_num;
2357 	short vn_pos_in_item;
2358 
2359 	/* item header of inserted item, 0 for other modes */
2360 	struct item_head *vn_ins_ih;
2361 	const void *vn_data;
2362 
2363 	/* array of items (including a new one, excluding item to be deleted) */
2364 	struct virtual_item *vn_vi;
2365 };
2366 
2367 /* used by directory items when creating virtual nodes */
2368 struct direntry_uarea {
2369 	int flags;
2370 	__u16 entry_count;
2371 	__u16 entry_sizes[1];
2372 } __attribute__ ((__packed__));
2373 
2374 /***************************************************************************
2375  *                  TREE BALANCE                                           *
2376  ***************************************************************************/
2377 
2378 /*
2379  * This temporary structure is used in tree balance algorithms, and
2380  * constructed as we go to the extent that its various parts are
2381  * needed.  It contains arrays of nodes that can potentially be
2382  * involved in the balancing of node S, and parameters that define how
2383  * each of the nodes must be balanced.  Note that in these algorithms
2384  * for balancing the worst case is to need to balance the current node
2385  * S and the left and right neighbors and all of their parents plus
2386  * create a new node.  We implement S1 balancing for the leaf nodes
2387  * and S0 balancing for the internal nodes (S1 and S0 are defined in
2388  * our papers.)
2389  */
2390 
2391 /* size of the array of buffers to free at end of do_balance */
2392 #define MAX_FREE_BLOCK 7
2393 
2394 /* maximum number of FEB blocknrs on a single level */
2395 #define MAX_AMOUNT_NEEDED 2
2396 
2397 /* someday somebody will prefix every field in this struct with tb_ */
2398 struct tree_balance {
2399 	int tb_mode;
2400 	int need_balance_dirty;
2401 	struct super_block *tb_sb;
2402 	struct reiserfs_transaction_handle *transaction_handle;
2403 	struct treepath *tb_path;
2404 
2405 	/* array of left neighbors of nodes in the path */
2406 	struct buffer_head *L[MAX_HEIGHT];
2407 
2408 	/* array of right neighbors of nodes in the path */
2409 	struct buffer_head *R[MAX_HEIGHT];
2410 
2411 	/* array of fathers of the left neighbors */
2412 	struct buffer_head *FL[MAX_HEIGHT];
2413 
2414 	/* array of fathers of the right neighbors */
2415 	struct buffer_head *FR[MAX_HEIGHT];
2416 	/* array of common parents of center node and its left neighbor */
2417 	struct buffer_head *CFL[MAX_HEIGHT];
2418 
2419 	/* array of common parents of center node and its right neighbor */
2420 	struct buffer_head *CFR[MAX_HEIGHT];
2421 
2422 	/*
2423 	 * array of empty buffers. Number of buffers in array equals
2424 	 * cur_blknum.
2425 	 */
2426 	struct buffer_head *FEB[MAX_FEB_SIZE];
2427 	struct buffer_head *used[MAX_FEB_SIZE];
2428 	struct buffer_head *thrown[MAX_FEB_SIZE];
2429 
2430 	/*
2431 	 * array of number of items which must be shifted to the left in
2432 	 * order to balance the current node; for leaves includes item that
2433 	 * will be partially shifted; for internal nodes, it is the number
2434 	 * of child pointers rather than items. It includes the new item
2435 	 * being created. The code sometimes subtracts one to get the
2436 	 * number of wholly shifted items for other purposes.
2437 	 */
2438 	int lnum[MAX_HEIGHT];
2439 
2440 	/* substitute right for left in comment above */
2441 	int rnum[MAX_HEIGHT];
2442 
2443 	/*
2444 	 * array indexed by height h mapping the key delimiting L[h] and
2445 	 * S[h] to its item number within the node CFL[h]
2446 	 */
2447 	int lkey[MAX_HEIGHT];
2448 
2449 	/* substitute r for l in comment above */
2450 	int rkey[MAX_HEIGHT];
2451 
2452 	/*
2453 	 * the number of bytes by we are trying to add or remove from
2454 	 * S[h]. A negative value means removing.
2455 	 */
2456 	int insert_size[MAX_HEIGHT];
2457 
2458 	/*
2459 	 * number of nodes that will replace node S[h] after balancing
2460 	 * on the level h of the tree.  If 0 then S is being deleted,
2461 	 * if 1 then S is remaining and no new nodes are being created,
2462 	 * if 2 or 3 then 1 or 2 new nodes is being created
2463 	 */
2464 	int blknum[MAX_HEIGHT];
2465 
2466 	/* fields that are used only for balancing leaves of the tree */
2467 
2468 	/* number of empty blocks having been already allocated */
2469 	int cur_blknum;
2470 
2471 	/* number of items that fall into left most node when S[0] splits */
2472 	int s0num;
2473 
2474 	/*
2475 	 * number of bytes which can flow to the left neighbor from the left
2476 	 * most liquid item that cannot be shifted from S[0] entirely
2477 	 * if -1 then nothing will be partially shifted
2478 	 */
2479 	int lbytes;
2480 
2481 	/*
2482 	 * number of bytes which will flow to the right neighbor from the right
2483 	 * most liquid item that cannot be shifted from S[0] entirely
2484 	 * if -1 then nothing will be partially shifted
2485 	 */
2486 	int rbytes;
2487 
2488 
2489 	/*
2490 	 * index into the array of item headers in
2491 	 * S[0] of the affected item
2492 	 */
2493 	int item_pos;
2494 
2495 	/* new nodes allocated to hold what could not fit into S */
2496 	struct buffer_head *S_new[2];
2497 
2498 	/*
2499 	 * number of items that will be placed into nodes in S_new
2500 	 * when S[0] splits
2501 	 */
2502 	int snum[2];
2503 
2504 	/*
2505 	 * number of bytes which flow to nodes in S_new when S[0] splits
2506 	 * note: if S[0] splits into 3 nodes, then items do not need to be cut
2507 	 */
2508 	int sbytes[2];
2509 
2510 	int pos_in_item;
2511 	int zeroes_num;
2512 
2513 	/*
2514 	 * buffers which are to be freed after do_balance finishes
2515 	 * by unfix_nodes
2516 	 */
2517 	struct buffer_head *buf_to_free[MAX_FREE_BLOCK];
2518 
2519 	/*
2520 	 * kmalloced memory. Used to create virtual node and keep
2521 	 * map of dirtied bitmap blocks
2522 	 */
2523 	char *vn_buf;
2524 
2525 	int vn_buf_size;	/* size of the vn_buf */
2526 
2527 	/* VN starts after bitmap of bitmap blocks */
2528 	struct virtual_node *tb_vn;
2529 
2530 	/*
2531 	 * saved value of `reiserfs_generation' counter see
2532 	 * FILESYSTEM_CHANGED() macro in reiserfs_fs.h
2533 	 */
2534 	int fs_gen;
2535 
2536 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2537 	/*
2538 	 * key pointer, to pass to block allocator or
2539 	 * another low-level subsystem
2540 	 */
2541 	struct in_core_key key;
2542 #endif
2543 };
2544 
2545 /* These are modes of balancing */
2546 
2547 /* When inserting an item. */
2548 #define M_INSERT	'i'
2549 /*
2550  * When inserting into (directories only) or appending onto an already
2551  * existent item.
2552  */
2553 #define M_PASTE		'p'
2554 /* When deleting an item. */
2555 #define M_DELETE	'd'
2556 /* When truncating an item or removing an entry from a (directory) item. */
2557 #define M_CUT		'c'
2558 
2559 /* used when balancing on leaf level skipped (in reiserfsck) */
2560 #define M_INTERNAL	'n'
2561 
2562 /*
2563  * When further balancing is not needed, then do_balance does not need
2564  * to be called.
2565  */
2566 #define M_SKIP_BALANCING		's'
2567 #define M_CONVERT	'v'
2568 
2569 /* modes of leaf_move_items */
2570 #define LEAF_FROM_S_TO_L 0
2571 #define LEAF_FROM_S_TO_R 1
2572 #define LEAF_FROM_R_TO_L 2
2573 #define LEAF_FROM_L_TO_R 3
2574 #define LEAF_FROM_S_TO_SNEW 4
2575 
2576 #define FIRST_TO_LAST 0
2577 #define LAST_TO_FIRST 1
2578 
2579 /*
2580  * used in do_balance for passing parent of node information that has
2581  * been gotten from tb struct
2582  */
2583 struct buffer_info {
2584 	struct tree_balance *tb;
2585 	struct buffer_head *bi_bh;
2586 	struct buffer_head *bi_parent;
2587 	int bi_position;
2588 };
2589 
2590 static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2591 {
2592 	return tb ? tb->tb_sb : NULL;
2593 }
2594 
2595 static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2596 {
2597 	return bi ? sb_from_tb(bi->tb) : NULL;
2598 }
2599 
2600 /*
2601  * there are 4 types of items: stat data, directory item, indirect, direct.
2602  * +-------------------+------------+--------------+------------+
2603  * |                   |  k_offset  | k_uniqueness | mergeable? |
2604  * +-------------------+------------+--------------+------------+
2605  * |     stat data     |     0      |      0       |   no       |
2606  * +-------------------+------------+--------------+------------+
2607  * | 1st directory item| DOT_OFFSET | DIRENTRY_ .. |   no       |
2608  * | non 1st directory | hash value | UNIQUENESS   |   yes      |
2609  * |     item          |            |              |            |
2610  * +-------------------+------------+--------------+------------+
2611  * | indirect item     | offset + 1 |TYPE_INDIRECT |    [1]	|
2612  * +-------------------+------------+--------------+------------+
2613  * | direct item       | offset + 1 |TYPE_DIRECT   |    [2]     |
2614  * +-------------------+------------+--------------+------------+
2615  *
2616  * [1] if this is not the first indirect item of the object
2617  * [2] if this is not the first direct item of the object
2618 */
2619 
2620 struct item_operations {
2621 	int (*bytes_number) (struct item_head * ih, int block_size);
2622 	void (*decrement_key) (struct cpu_key *);
2623 	int (*is_left_mergeable) (struct reiserfs_key * ih,
2624 				  unsigned long bsize);
2625 	void (*print_item) (struct item_head *, char *item);
2626 	void (*check_item) (struct item_head *, char *item);
2627 
2628 	int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2629 			  int is_affected, int insert_size);
2630 	int (*check_left) (struct virtual_item * vi, int free,
2631 			   int start_skip, int end_skip);
2632 	int (*check_right) (struct virtual_item * vi, int free);
2633 	int (*part_size) (struct virtual_item * vi, int from, int to);
2634 	int (*unit_num) (struct virtual_item * vi);
2635 	void (*print_vi) (struct virtual_item * vi);
2636 };
2637 
2638 extern struct item_operations *item_ops[TYPE_ANY + 1];
2639 
2640 #define op_bytes_number(ih,bsize)                    item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2641 #define op_is_left_mergeable(key,bsize)              item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2642 #define op_print_item(ih,item)                       item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2643 #define op_check_item(ih,item)                       item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2644 #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)
2645 #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2646 #define op_check_right(vi,free)                      item_ops[(vi)->vi_index]->check_right (vi, free)
2647 #define op_part_size(vi,from,to)                     item_ops[(vi)->vi_index]->part_size (vi, from, to)
2648 #define op_unit_num(vi)				     item_ops[(vi)->vi_index]->unit_num (vi)
2649 #define op_print_vi(vi)                              item_ops[(vi)->vi_index]->print_vi (vi)
2650 
2651 #define COMP_SHORT_KEYS comp_short_keys
2652 
2653 /* number of blocks pointed to by the indirect item */
2654 #define I_UNFM_NUM(ih)	(ih_item_len(ih) / UNFM_P_SIZE)
2655 
2656 /*
2657  * the used space within the unformatted node corresponding
2658  * to pos within the item pointed to by ih
2659  */
2660 #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2661 
2662 /*
2663  * number of bytes contained by the direct item or the
2664  * unformatted nodes the indirect item points to
2665  */
2666 
2667 /* following defines use reiserfs buffer header and item header */
2668 
2669 /* get stat-data */
2670 #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2671 
2672 /* this is 3976 for size==4096 */
2673 #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2674 
2675 /*
2676  * indirect items consist of entries which contain blocknrs, pos
2677  * indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2678  * blocknr contained by the entry pos points to
2679  */
2680 #define B_I_POS_UNFM_POINTER(bh, ih, pos)				\
2681 	le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos)))
2682 #define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val)			\
2683 	(*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val))
2684 
2685 struct reiserfs_iget_args {
2686 	__u32 objectid;
2687 	__u32 dirid;
2688 };
2689 
2690 /***************************************************************************
2691  *                    FUNCTION DECLARATIONS                                *
2692  ***************************************************************************/
2693 
2694 #define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2695 
2696 #define journal_trans_half(blocksize) \
2697 	((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2698 
2699 /* journal.c see journal.c for all the comments here */
2700 
2701 /* first block written in a commit.  */
2702 struct reiserfs_journal_desc {
2703 	__le32 j_trans_id;	/* id of commit */
2704 
2705 	/* length of commit. len +1 is the commit block */
2706 	__le32 j_len;
2707 
2708 	__le32 j_mount_id;	/* mount id of this trans */
2709 	__le32 j_realblock[1];	/* real locations for each block */
2710 };
2711 
2712 #define get_desc_trans_id(d)   le32_to_cpu((d)->j_trans_id)
2713 #define get_desc_trans_len(d)  le32_to_cpu((d)->j_len)
2714 #define get_desc_mount_id(d)   le32_to_cpu((d)->j_mount_id)
2715 
2716 #define set_desc_trans_id(d,val)       do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2717 #define set_desc_trans_len(d,val)      do { (d)->j_len = cpu_to_le32 (val); } while (0)
2718 #define set_desc_mount_id(d,val)       do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2719 
2720 /* last block written in a commit */
2721 struct reiserfs_journal_commit {
2722 	__le32 j_trans_id;	/* must match j_trans_id from the desc block */
2723 	__le32 j_len;		/* ditto */
2724 	__le32 j_realblock[1];	/* real locations for each block */
2725 };
2726 
2727 #define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2728 #define get_commit_trans_len(c)        le32_to_cpu((c)->j_len)
2729 #define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2730 
2731 #define set_commit_trans_id(c,val)     do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2732 #define set_commit_trans_len(c,val)    do { (c)->j_len = cpu_to_le32 (val); } while (0)
2733 
2734 /*
2735  * this header block gets written whenever a transaction is considered
2736  * fully flushed, and is more recent than the last fully flushed transaction.
2737  * fully flushed means all the log blocks and all the real blocks are on
2738  * disk, and this transaction does not need to be replayed.
2739  */
2740 struct reiserfs_journal_header {
2741 	/* id of last fully flushed transaction */
2742 	__le32 j_last_flush_trans_id;
2743 
2744 	/* offset in the log of where to start replay after a crash */
2745 	__le32 j_first_unflushed_offset;
2746 
2747 	__le32 j_mount_id;
2748 	/* 12 */ struct journal_params jh_journal;
2749 };
2750 
2751 /* biggest tunable defines are right here */
2752 #define JOURNAL_BLOCK_COUNT 8192	/* number of blocks in the journal */
2753 
2754 /* biggest possible single transaction, don't change for now (8/3/99) */
2755 #define JOURNAL_TRANS_MAX_DEFAULT 1024
2756 #define JOURNAL_TRANS_MIN_DEFAULT 256
2757 
2758 /*
2759  * max blocks to batch into one transaction,
2760  * don't make this any bigger than 900
2761  */
2762 #define JOURNAL_MAX_BATCH_DEFAULT   900
2763 #define JOURNAL_MIN_RATIO 2
2764 #define JOURNAL_MAX_COMMIT_AGE 30
2765 #define JOURNAL_MAX_TRANS_AGE 30
2766 #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2767 #define JOURNAL_BLOCKS_PER_OBJECT(sb)  (JOURNAL_PER_BALANCE_CNT * 3 + \
2768 					 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2769 					      REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2770 
2771 #ifdef CONFIG_QUOTA
2772 #define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2773 /* We need to update data and inode (atime) */
2774 #define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2775 /* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2776 #define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2777 (DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2778 /* same as with INIT */
2779 #define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2780 (DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2781 #else
2782 #define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2783 #define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2784 #define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2785 #endif
2786 
2787 /*
2788  * both of these can be as low as 1, or as high as you want.  The min is the
2789  * number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2790  * as needed, and released when transactions are committed.  On release, if
2791  * the current number of nodes is > max, the node is freed, otherwise,
2792  * it is put on a free list for faster use later.
2793 */
2794 #define REISERFS_MIN_BITMAP_NODES 10
2795 #define REISERFS_MAX_BITMAP_NODES 100
2796 
2797 /* these are based on journal hash size of 8192 */
2798 #define JBH_HASH_SHIFT 13
2799 #define JBH_HASH_MASK 8191
2800 
2801 #define _jhashfn(sb,block)	\
2802 	(((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2803 	 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2804 #define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2805 
2806 /* We need these to make journal.c code more readable */
2807 #define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2808 #define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2809 #define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2810 
2811 enum reiserfs_bh_state_bits {
2812 	BH_JDirty = BH_PrivateStart,	/* buffer is in current transaction */
2813 	BH_JDirty_wait,
2814 	/*
2815 	 * disk block was taken off free list before being in a
2816 	 * finished transaction, or written to disk. Can be reused immed.
2817 	 */
2818 	BH_JNew,
2819 	BH_JPrepared,
2820 	BH_JRestore_dirty,
2821 	BH_JTest,		/* debugging only will go away */
2822 };
2823 
2824 BUFFER_FNS(JDirty, journaled);
2825 TAS_BUFFER_FNS(JDirty, journaled);
2826 BUFFER_FNS(JDirty_wait, journal_dirty);
2827 TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2828 BUFFER_FNS(JNew, journal_new);
2829 TAS_BUFFER_FNS(JNew, journal_new);
2830 BUFFER_FNS(JPrepared, journal_prepared);
2831 TAS_BUFFER_FNS(JPrepared, journal_prepared);
2832 BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2833 TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2834 BUFFER_FNS(JTest, journal_test);
2835 TAS_BUFFER_FNS(JTest, journal_test);
2836 
2837 /* transaction handle which is passed around for all journal calls */
2838 struct reiserfs_transaction_handle {
2839 	/*
2840 	 * super for this FS when journal_begin was called. saves calls to
2841 	 * reiserfs_get_super also used by nested transactions to make
2842 	 * sure they are nesting on the right FS _must_ be first
2843 	 * in the handle
2844 	 */
2845 	struct super_block *t_super;
2846 
2847 	int t_refcount;
2848 	int t_blocks_logged;	/* number of blocks this writer has logged */
2849 	int t_blocks_allocated;	/* number of blocks this writer allocated */
2850 
2851 	/* sanity check, equals the current trans id */
2852 	unsigned int t_trans_id;
2853 
2854 	void *t_handle_save;	/* save existing current->journal_info */
2855 
2856 	/*
2857 	 * if new block allocation occurres, that block
2858 	 * should be displaced from others
2859 	 */
2860 	unsigned displace_new_blocks:1;
2861 
2862 	struct list_head t_list;
2863 };
2864 
2865 /*
2866  * used to keep track of ordered and tail writes, attached to the buffer
2867  * head through b_journal_head.
2868  */
2869 struct reiserfs_jh {
2870 	struct reiserfs_journal_list *jl;
2871 	struct buffer_head *bh;
2872 	struct list_head list;
2873 };
2874 
2875 void reiserfs_free_jh(struct buffer_head *bh);
2876 int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2877 int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2878 int journal_mark_dirty(struct reiserfs_transaction_handle *,
2879 		       struct buffer_head *bh);
2880 
2881 static inline int reiserfs_file_data_log(struct inode *inode)
2882 {
2883 	if (reiserfs_data_log(inode->i_sb) ||
2884 	    (REISERFS_I(inode)->i_flags & i_data_log))
2885 		return 1;
2886 	return 0;
2887 }
2888 
2889 static inline int reiserfs_transaction_running(struct super_block *s)
2890 {
2891 	struct reiserfs_transaction_handle *th = current->journal_info;
2892 	if (th && th->t_super == s)
2893 		return 1;
2894 	if (th && th->t_super == NULL)
2895 		BUG();
2896 	return 0;
2897 }
2898 
2899 static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2900 {
2901 	return th->t_blocks_allocated - th->t_blocks_logged;
2902 }
2903 
2904 struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2905 								    super_block
2906 								    *,
2907 								    int count);
2908 int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
2909 void reiserfs_vfs_truncate_file(struct inode *inode);
2910 int reiserfs_commit_page(struct inode *inode, struct page *page,
2911 			 unsigned from, unsigned to);
2912 void reiserfs_flush_old_commits(struct super_block *);
2913 int reiserfs_commit_for_inode(struct inode *);
2914 int reiserfs_inode_needs_commit(struct inode *);
2915 void reiserfs_update_inode_transaction(struct inode *);
2916 void reiserfs_wait_on_write_block(struct super_block *s);
2917 void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2918 void reiserfs_allow_writes(struct super_block *s);
2919 void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2920 int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2921 				 int wait);
2922 void reiserfs_restore_prepared_buffer(struct super_block *,
2923 				      struct buffer_head *bh);
2924 int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2925 		 unsigned int);
2926 int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2927 int journal_release_error(struct reiserfs_transaction_handle *,
2928 			  struct super_block *);
2929 int journal_end(struct reiserfs_transaction_handle *);
2930 int journal_end_sync(struct reiserfs_transaction_handle *);
2931 int journal_mark_freed(struct reiserfs_transaction_handle *,
2932 		       struct super_block *, b_blocknr_t blocknr);
2933 int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2934 int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2935 			 int bit_nr, int searchall, b_blocknr_t *next);
2936 int journal_begin(struct reiserfs_transaction_handle *,
2937 		  struct super_block *sb, unsigned long);
2938 int journal_join_abort(struct reiserfs_transaction_handle *,
2939 		       struct super_block *sb);
2940 void reiserfs_abort_journal(struct super_block *sb, int errno);
2941 void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2942 int reiserfs_allocate_list_bitmaps(struct super_block *s,
2943 				   struct reiserfs_list_bitmap *, unsigned int);
2944 
2945 void reiserfs_schedule_old_flush(struct super_block *s);
2946 void add_save_link(struct reiserfs_transaction_handle *th,
2947 		   struct inode *inode, int truncate);
2948 int remove_save_link(struct inode *inode, int truncate);
2949 
2950 /* objectid.c */
2951 __u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2952 void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2953 			       __u32 objectid_to_release);
2954 int reiserfs_convert_objectid_map_v1(struct super_block *);
2955 
2956 /* stree.c */
2957 int B_IS_IN_TREE(const struct buffer_head *);
2958 extern void copy_item_head(struct item_head *to,
2959 			   const struct item_head *from);
2960 
2961 /* first key is in cpu form, second - le */
2962 extern int comp_short_keys(const struct reiserfs_key *le_key,
2963 			   const struct cpu_key *cpu_key);
2964 extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2965 
2966 /* both are in le form */
2967 extern int comp_le_keys(const struct reiserfs_key *,
2968 			const struct reiserfs_key *);
2969 extern int comp_short_le_keys(const struct reiserfs_key *,
2970 			      const struct reiserfs_key *);
2971 
2972 /* * get key version from on disk key - kludge */
2973 static inline int le_key_version(const struct reiserfs_key *key)
2974 {
2975 	int type;
2976 
2977 	type = offset_v2_k_type(&(key->u.k_offset_v2));
2978 	if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2979 	    && type != TYPE_DIRENTRY)
2980 		return KEY_FORMAT_3_5;
2981 
2982 	return KEY_FORMAT_3_6;
2983 
2984 }
2985 
2986 static inline void copy_key(struct reiserfs_key *to,
2987 			    const struct reiserfs_key *from)
2988 {
2989 	memcpy(to, from, KEY_SIZE);
2990 }
2991 
2992 int comp_items(const struct item_head *stored_ih, const struct treepath *path);
2993 const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
2994 				    const struct super_block *sb);
2995 int search_by_key(struct super_block *, const struct cpu_key *,
2996 		  struct treepath *, int);
2997 #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
2998 int search_for_position_by_key(struct super_block *sb,
2999 			       const struct cpu_key *cpu_key,
3000 			       struct treepath *search_path);
3001 extern void decrement_bcount(struct buffer_head *bh);
3002 void decrement_counters_in_path(struct treepath *search_path);
3003 void pathrelse(struct treepath *search_path);
3004 int reiserfs_check_path(struct treepath *p);
3005 void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
3006 
3007 int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
3008 			 struct treepath *path,
3009 			 const struct cpu_key *key,
3010 			 struct item_head *ih,
3011 			 struct inode *inode, const char *body);
3012 
3013 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
3014 			     struct treepath *path,
3015 			     const struct cpu_key *key,
3016 			     struct inode *inode,
3017 			     const char *body, int paste_size);
3018 
3019 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
3020 			   struct treepath *path,
3021 			   struct cpu_key *key,
3022 			   struct inode *inode,
3023 			   struct page *page, loff_t new_file_size);
3024 
3025 int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
3026 			 struct treepath *path,
3027 			 const struct cpu_key *key,
3028 			 struct inode *inode, struct buffer_head *un_bh);
3029 
3030 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
3031 				struct inode *inode, struct reiserfs_key *key);
3032 int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
3033 			   struct inode *inode);
3034 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
3035 			 struct inode *inode, struct page *,
3036 			 int update_timestamps);
3037 
3038 #define i_block_size(inode) ((inode)->i_sb->s_blocksize)
3039 #define file_size(inode) ((inode)->i_size)
3040 #define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
3041 
3042 #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
3043 !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 )
3044 
3045 void padd_item(char *item, int total_length, int length);
3046 
3047 /* inode.c */
3048 /* args for the create parameter of reiserfs_get_block */
3049 #define GET_BLOCK_NO_CREATE 0	 /* don't create new blocks or convert tails */
3050 #define GET_BLOCK_CREATE 1	 /* add anything you need to find block */
3051 #define GET_BLOCK_NO_HOLE 2	 /* return -ENOENT for file holes */
3052 #define GET_BLOCK_READ_DIRECT 4	 /* read the tail if indirect item not found */
3053 #define GET_BLOCK_NO_IMUX     8	 /* i_mutex is not held, don't preallocate */
3054 #define GET_BLOCK_NO_DANGLE   16 /* don't leave any transactions running */
3055 
3056 void reiserfs_read_locked_inode(struct inode *inode,
3057 				struct reiserfs_iget_args *args);
3058 int reiserfs_find_actor(struct inode *inode, void *p);
3059 int reiserfs_init_locked_inode(struct inode *inode, void *p);
3060 void reiserfs_evict_inode(struct inode *inode);
3061 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
3062 int reiserfs_get_block(struct inode *inode, sector_t block,
3063 		       struct buffer_head *bh_result, int create);
3064 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3065 				     int fh_len, int fh_type);
3066 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
3067 				     int fh_len, int fh_type);
3068 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
3069 		       struct inode *parent);
3070 
3071 int reiserfs_truncate_file(struct inode *, int update_timestamps);
3072 void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
3073 		  int type, int key_length);
3074 void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
3075 		       int version,
3076 		       loff_t offset, int type, int length, int entry_count);
3077 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
3078 
3079 struct reiserfs_security_handle;
3080 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
3081 		       struct inode *dir, umode_t mode,
3082 		       const char *symname, loff_t i_size,
3083 		       struct dentry *dentry, struct inode *inode,
3084 		       struct reiserfs_security_handle *security);
3085 
3086 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
3087 			     struct inode *inode, loff_t size);
3088 
3089 static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
3090 				      struct inode *inode)
3091 {
3092 	reiserfs_update_sd_size(th, inode, inode->i_size);
3093 }
3094 
3095 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
3096 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
3097 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
3098 
3099 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
3100 
3101 /* namei.c */
3102 void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
3103 int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
3104 			struct treepath *path, struct reiserfs_dir_entry *de);
3105 struct dentry *reiserfs_get_parent(struct dentry *);
3106 
3107 #ifdef CONFIG_REISERFS_PROC_INFO
3108 int reiserfs_proc_info_init(struct super_block *sb);
3109 int reiserfs_proc_info_done(struct super_block *sb);
3110 int reiserfs_proc_info_global_init(void);
3111 int reiserfs_proc_info_global_done(void);
3112 
3113 #define PROC_EXP( e )   e
3114 
3115 #define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
3116 #define PROC_INFO_MAX( sb, field, value )								\
3117     __PINFO( sb ).field =												\
3118         max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
3119 #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
3120 #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
3121 #define PROC_INFO_BH_STAT( sb, bh, level )							\
3122     PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] );						\
3123     PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) );	\
3124     PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
3125 #else
3126 static inline int reiserfs_proc_info_init(struct super_block *sb)
3127 {
3128 	return 0;
3129 }
3130 
3131 static inline int reiserfs_proc_info_done(struct super_block *sb)
3132 {
3133 	return 0;
3134 }
3135 
3136 static inline int reiserfs_proc_info_global_init(void)
3137 {
3138 	return 0;
3139 }
3140 
3141 static inline int reiserfs_proc_info_global_done(void)
3142 {
3143 	return 0;
3144 }
3145 
3146 #define PROC_EXP( e )
3147 #define VOID_V ( ( void ) 0 )
3148 #define PROC_INFO_MAX( sb, field, value ) VOID_V
3149 #define PROC_INFO_INC( sb, field ) VOID_V
3150 #define PROC_INFO_ADD( sb, field, val ) VOID_V
3151 #define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
3152 #endif
3153 
3154 /* dir.c */
3155 extern const struct inode_operations reiserfs_dir_inode_operations;
3156 extern const struct inode_operations reiserfs_symlink_inode_operations;
3157 extern const struct inode_operations reiserfs_special_inode_operations;
3158 extern const struct file_operations reiserfs_dir_operations;
3159 int reiserfs_readdir_inode(struct inode *, struct dir_context *);
3160 
3161 /* tail_conversion.c */
3162 int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
3163 		    struct treepath *, struct buffer_head *, loff_t);
3164 int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
3165 		    struct page *, struct treepath *, const struct cpu_key *,
3166 		    loff_t, char *);
3167 void reiserfs_unmap_buffer(struct buffer_head *);
3168 
3169 /* file.c */
3170 extern const struct inode_operations reiserfs_file_inode_operations;
3171 extern const struct file_operations reiserfs_file_operations;
3172 extern const struct address_space_operations reiserfs_address_space_operations;
3173 
3174 /* fix_nodes.c */
3175 
3176 int fix_nodes(int n_op_mode, struct tree_balance *tb,
3177 	      struct item_head *ins_ih, const void *);
3178 void unfix_nodes(struct tree_balance *);
3179 
3180 /* prints.c */
3181 void __reiserfs_panic(struct super_block *s, const char *id,
3182 		      const char *function, const char *fmt, ...)
3183     __attribute__ ((noreturn));
3184 #define reiserfs_panic(s, id, fmt, args...) \
3185 	__reiserfs_panic(s, id, __func__, fmt, ##args)
3186 void __reiserfs_error(struct super_block *s, const char *id,
3187 		      const char *function, const char *fmt, ...);
3188 #define reiserfs_error(s, id, fmt, args...) \
3189 	 __reiserfs_error(s, id, __func__, fmt, ##args)
3190 void reiserfs_info(struct super_block *s, const char *fmt, ...);
3191 void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
3192 void print_indirect_item(struct buffer_head *bh, int item_num);
3193 void store_print_tb(struct tree_balance *tb);
3194 void print_cur_tb(char *mes);
3195 void print_de(struct reiserfs_dir_entry *de);
3196 void print_bi(struct buffer_info *bi, char *mes);
3197 #define PRINT_LEAF_ITEMS 1	/* print all items */
3198 #define PRINT_DIRECTORY_ITEMS 2	/* print directory items */
3199 #define PRINT_DIRECT_ITEMS 4	/* print contents of direct items */
3200 void print_block(struct buffer_head *bh, ...);
3201 void print_bmap(struct super_block *s, int silent);
3202 void print_bmap_block(int i, char *data, int size, int silent);
3203 /*void print_super_block (struct super_block * s, char * mes);*/
3204 void print_objectid_map(struct super_block *s);
3205 void print_block_head(struct buffer_head *bh, char *mes);
3206 void check_leaf(struct buffer_head *bh);
3207 void check_internal(struct buffer_head *bh);
3208 void print_statistics(struct super_block *s);
3209 char *reiserfs_hashname(int code);
3210 
3211 /* lbalance.c */
3212 int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
3213 		    int mov_bytes, struct buffer_head *Snew);
3214 int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
3215 int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
3216 void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
3217 		       int del_num, int del_bytes);
3218 void leaf_insert_into_buf(struct buffer_info *bi, int before,
3219 			  struct item_head *inserted_item_ih,
3220 			  const char *inserted_item_body, int zeros_number);
3221 void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
3222 			  int pos_in_item, int paste_size, const char *body,
3223 			  int zeros_number);
3224 void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
3225 			  int pos_in_item, int cut_size);
3226 void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
3227 			int new_entry_count, struct reiserfs_de_head *new_dehs,
3228 			const char *records, int paste_size);
3229 /* ibalance.c */
3230 int balance_internal(struct tree_balance *, int, int, struct item_head *,
3231 		     struct buffer_head **);
3232 
3233 /* do_balance.c */
3234 void do_balance_mark_leaf_dirty(struct tree_balance *tb,
3235 				struct buffer_head *bh, int flag);
3236 #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
3237 #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
3238 
3239 void do_balance(struct tree_balance *tb, struct item_head *ih,
3240 		const char *body, int flag);
3241 void reiserfs_invalidate_buffer(struct tree_balance *tb,
3242 				struct buffer_head *bh);
3243 
3244 int get_left_neighbor_position(struct tree_balance *tb, int h);
3245 int get_right_neighbor_position(struct tree_balance *tb, int h);
3246 void replace_key(struct tree_balance *tb, struct buffer_head *, int,
3247 		 struct buffer_head *, int);
3248 void make_empty_node(struct buffer_info *);
3249 struct buffer_head *get_FEB(struct tree_balance *);
3250 
3251 /* bitmap.c */
3252 
3253 /*
3254  * structure contains hints for block allocator, and it is a container for
3255  * arguments, such as node, search path, transaction_handle, etc.
3256  */
3257 struct __reiserfs_blocknr_hint {
3258 	/* inode passed to allocator, if we allocate unf. nodes */
3259 	struct inode *inode;
3260 
3261 	sector_t block;		/* file offset, in blocks */
3262 	struct in_core_key key;
3263 
3264 	/*
3265 	 * search path, used by allocator to deternine search_start by
3266 	 * various ways
3267 	 */
3268 	struct treepath *path;
3269 
3270 	/*
3271 	 * transaction handle is needed to log super blocks
3272 	 * and bitmap blocks changes
3273 	 */
3274 	struct reiserfs_transaction_handle *th;
3275 
3276 	b_blocknr_t beg, end;
3277 
3278 	/*
3279 	 * a field used to transfer search start value (block number)
3280 	 * between different block allocator procedures
3281 	 * (determine_search_start() and others)
3282 	 */
3283 	b_blocknr_t search_start;
3284 
3285 	/*
3286 	 * is set in determine_prealloc_size() function,
3287 	 * used by underlayed function that do actual allocation
3288 	 */
3289 	int prealloc_size;
3290 
3291 	/*
3292 	 * the allocator uses different polices for getting disk
3293 	 * space for formatted/unformatted blocks with/without preallocation
3294 	 */
3295 	unsigned formatted_node:1;
3296 	unsigned preallocate:1;
3297 };
3298 
3299 typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
3300 
3301 int reiserfs_parse_alloc_options(struct super_block *, char *);
3302 void reiserfs_init_alloc_options(struct super_block *s);
3303 
3304 /*
3305  * given a directory, this will tell you what packing locality
3306  * to use for a new object underneat it.  The locality is returned
3307  * in disk byte order (le).
3308  */
3309 __le32 reiserfs_choose_packing(struct inode *dir);
3310 
3311 void show_alloc_options(struct seq_file *seq, struct super_block *s);
3312 int reiserfs_init_bitmap_cache(struct super_block *sb);
3313 void reiserfs_free_bitmap_cache(struct super_block *sb);
3314 void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
3315 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
3316 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
3317 void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
3318 			 b_blocknr_t, int for_unformatted);
3319 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
3320 			       int);
3321 static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
3322 					     b_blocknr_t * new_blocknrs,
3323 					     int amount_needed)
3324 {
3325 	reiserfs_blocknr_hint_t hint = {
3326 		.th = tb->transaction_handle,
3327 		.path = tb->tb_path,
3328 		.inode = NULL,
3329 		.key = tb->key,
3330 		.block = 0,
3331 		.formatted_node = 1
3332 	};
3333 	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
3334 					  0);
3335 }
3336 
3337 static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
3338 					    *th, struct inode *inode,
3339 					    b_blocknr_t * new_blocknrs,
3340 					    struct treepath *path,
3341 					    sector_t block)
3342 {
3343 	reiserfs_blocknr_hint_t hint = {
3344 		.th = th,
3345 		.path = path,
3346 		.inode = inode,
3347 		.block = block,
3348 		.formatted_node = 0,
3349 		.preallocate = 0
3350 	};
3351 	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3352 }
3353 
3354 #ifdef REISERFS_PREALLOCATE
3355 static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
3356 					     *th, struct inode *inode,
3357 					     b_blocknr_t * new_blocknrs,
3358 					     struct treepath *path,
3359 					     sector_t block)
3360 {
3361 	reiserfs_blocknr_hint_t hint = {
3362 		.th = th,
3363 		.path = path,
3364 		.inode = inode,
3365 		.block = block,
3366 		.formatted_node = 0,
3367 		.preallocate = 1
3368 	};
3369 	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3370 }
3371 
3372 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
3373 			       struct inode *inode);
3374 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
3375 #endif
3376 
3377 /* hashes.c */
3378 __u32 keyed_hash(const signed char *msg, int len);
3379 __u32 yura_hash(const signed char *msg, int len);
3380 __u32 r5_hash(const signed char *msg, int len);
3381 
3382 #define reiserfs_set_le_bit		__set_bit_le
3383 #define reiserfs_test_and_set_le_bit	__test_and_set_bit_le
3384 #define reiserfs_clear_le_bit		__clear_bit_le
3385 #define reiserfs_test_and_clear_le_bit	__test_and_clear_bit_le
3386 #define reiserfs_test_le_bit		test_bit_le
3387 #define reiserfs_find_next_zero_le_bit	find_next_zero_bit_le
3388 
3389 /*
3390  * sometimes reiserfs_truncate may require to allocate few new blocks
3391  * to perform indirect2direct conversion. People probably used to
3392  * think, that truncate should work without problems on a filesystem
3393  * without free disk space. They may complain that they can not
3394  * truncate due to lack of free disk space. This spare space allows us
3395  * to not worry about it. 500 is probably too much, but it should be
3396  * absolutely safe
3397  */
3398 #define SPARE_SPACE 500
3399 
3400 /* prototypes from ioctl.c */
3401 long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
3402 long reiserfs_compat_ioctl(struct file *filp,
3403 		   unsigned int cmd, unsigned long arg);
3404 int reiserfs_unpack(struct inode *inode, struct file *filp);
3405