1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for
4  * licensing and copyright details
5  */
6 
7 #include <linux/reiserfs_fs.h>
8 
9 #include <linux/slab.h>
10 #include <linux/interrupt.h>
11 #include <linux/sched.h>
12 #include <linux/bug.h>
13 #include <linux/workqueue.h>
14 #include <asm/unaligned.h>
15 #include <linux/bitops.h>
16 #include <linux/proc_fs.h>
17 #include <linux/buffer_head.h>
18 
19 /* the 32 bit compat definitions with int argument */
20 #define REISERFS_IOC32_UNPACK		_IOW(0xCD, 1, int)
21 #define REISERFS_IOC32_GETFLAGS		FS_IOC32_GETFLAGS
22 #define REISERFS_IOC32_SETFLAGS		FS_IOC32_SETFLAGS
23 #define REISERFS_IOC32_GETVERSION	FS_IOC32_GETVERSION
24 #define REISERFS_IOC32_SETVERSION	FS_IOC32_SETVERSION
25 
26 struct reiserfs_journal_list;
27 
28 /* bitmasks for i_flags field in reiserfs-specific part of inode */
29 typedef enum {
30 	/*
31 	 * this says what format of key do all items (but stat data) of
32 	 * an object have.  If this is set, that format is 3.6 otherwise - 3.5
33 	 */
34 	i_item_key_version_mask = 0x0001,
35 
36 	/*
37 	 * If this is unset, object has 3.5 stat data, otherwise,
38 	 * it has 3.6 stat data with 64bit size, 32bit nlink etc.
39 	 */
40 	i_stat_data_version_mask = 0x0002,
41 
42 	/* file might need tail packing on close */
43 	i_pack_on_close_mask = 0x0004,
44 
45 	/* don't pack tail of file */
46 	i_nopack_mask = 0x0008,
47 
48 	/*
49 	 * If either of these are set, "safe link" was created for this
50 	 * file during truncate or unlink. Safe link is used to avoid
51 	 * leakage of disk space on crash with some files open, but unlinked.
52 	 */
53 	i_link_saved_unlink_mask = 0x0010,
54 	i_link_saved_truncate_mask = 0x0020,
55 
56 	i_has_xattr_dir = 0x0040,
57 	i_data_log = 0x0080,
58 } reiserfs_inode_flags;
59 
60 struct reiserfs_inode_info {
61 	__u32 i_key[4];		/* key is still 4 32 bit integers */
62 
63 	/*
64 	 * transient inode flags that are never stored on disk. Bitmasks
65 	 * for this field are defined above.
66 	 */
67 	__u32 i_flags;
68 
69 	/* offset of first byte stored in direct item. */
70 	__u32 i_first_direct_byte;
71 
72 	/* copy of persistent inode flags read from sd_attrs. */
73 	__u32 i_attrs;
74 
75 	/* first unused block of a sequence of unused blocks */
76 	int i_prealloc_block;
77 	int i_prealloc_count;	/* length of that sequence */
78 
79 	/* per-transaction list of inodes which  have preallocated blocks */
80 	struct list_head i_prealloc_list;
81 
82 	/*
83 	 * new_packing_locality is created; new blocks for the contents
84 	 * of this directory should be displaced
85 	 */
86 	unsigned new_packing_locality:1;
87 
88 	/*
89 	 * we use these for fsync or O_SYNC to decide which transaction
90 	 * needs to be committed in order for this inode to be properly
91 	 * flushed
92 	 */
93 	unsigned int i_trans_id;
94 
95 	struct reiserfs_journal_list *i_jl;
96 	atomic_t openers;
97 	struct mutex tailpack;
98 #ifdef CONFIG_REISERFS_FS_XATTR
99 	struct rw_semaphore i_xattr_sem;
100 #endif
101 #ifdef CONFIG_QUOTA
102 	struct dquot *i_dquot[MAXQUOTAS];
103 #endif
104 
105 	struct inode vfs_inode;
106 };
107 
108 typedef enum {
109 	reiserfs_attrs_cleared = 0x00000001,
110 } reiserfs_super_block_flags;
111 
112 /*
113  * struct reiserfs_super_block accessors/mutators since this is a disk
114  * structure, it will always be in little endian format.
115  */
116 #define sb_block_count(sbp)         (le32_to_cpu((sbp)->s_v1.s_block_count))
117 #define set_sb_block_count(sbp,v)   ((sbp)->s_v1.s_block_count = cpu_to_le32(v))
118 #define sb_free_blocks(sbp)         (le32_to_cpu((sbp)->s_v1.s_free_blocks))
119 #define set_sb_free_blocks(sbp,v)   ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))
120 #define sb_root_block(sbp)          (le32_to_cpu((sbp)->s_v1.s_root_block))
121 #define set_sb_root_block(sbp,v)    ((sbp)->s_v1.s_root_block = cpu_to_le32(v))
122 
123 #define sb_jp_journal_1st_block(sbp)  \
124               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))
125 #define set_sb_jp_journal_1st_block(sbp,v) \
126               ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))
127 #define sb_jp_journal_dev(sbp) \
128               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))
129 #define set_sb_jp_journal_dev(sbp,v) \
130               ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))
131 #define sb_jp_journal_size(sbp) \
132               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))
133 #define set_sb_jp_journal_size(sbp,v) \
134               ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))
135 #define sb_jp_journal_trans_max(sbp) \
136               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))
137 #define set_sb_jp_journal_trans_max(sbp,v) \
138               ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))
139 #define sb_jp_journal_magic(sbp) \
140               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))
141 #define set_sb_jp_journal_magic(sbp,v) \
142               ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))
143 #define sb_jp_journal_max_batch(sbp) \
144               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))
145 #define set_sb_jp_journal_max_batch(sbp,v) \
146               ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))
147 #define sb_jp_jourmal_max_commit_age(sbp) \
148               (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))
149 #define set_sb_jp_journal_max_commit_age(sbp,v) \
150               ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))
151 
152 #define sb_blocksize(sbp)          (le16_to_cpu((sbp)->s_v1.s_blocksize))
153 #define set_sb_blocksize(sbp,v)    ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))
154 #define sb_oid_maxsize(sbp)        (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))
155 #define set_sb_oid_maxsize(sbp,v)  ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))
156 #define sb_oid_cursize(sbp)        (le16_to_cpu((sbp)->s_v1.s_oid_cursize))
157 #define set_sb_oid_cursize(sbp,v)  ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))
158 #define sb_umount_state(sbp)       (le16_to_cpu((sbp)->s_v1.s_umount_state))
159 #define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))
160 #define sb_fs_state(sbp)           (le16_to_cpu((sbp)->s_v1.s_fs_state))
161 #define set_sb_fs_state(sbp,v)     ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))
162 #define sb_hash_function_code(sbp) \
163               (le32_to_cpu((sbp)->s_v1.s_hash_function_code))
164 #define set_sb_hash_function_code(sbp,v) \
165               ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))
166 #define sb_tree_height(sbp)        (le16_to_cpu((sbp)->s_v1.s_tree_height))
167 #define set_sb_tree_height(sbp,v)  ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))
168 #define sb_bmap_nr(sbp)            (le16_to_cpu((sbp)->s_v1.s_bmap_nr))
169 #define set_sb_bmap_nr(sbp,v)      ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))
170 #define sb_version(sbp)            (le16_to_cpu((sbp)->s_v1.s_version))
171 #define set_sb_version(sbp,v)      ((sbp)->s_v1.s_version = cpu_to_le16(v))
172 
173 #define sb_mnt_count(sbp)	   (le16_to_cpu((sbp)->s_mnt_count))
174 #define set_sb_mnt_count(sbp, v)   ((sbp)->s_mnt_count = cpu_to_le16(v))
175 
176 #define sb_reserved_for_journal(sbp) \
177               (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))
178 #define set_sb_reserved_for_journal(sbp,v) \
179               ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))
180 
181 /* LOGGING -- */
182 
183 /*
184  * These all interelate for performance.
185  *
186  * If the journal block count is smaller than n transactions, you lose speed.
187  * I don't know what n is yet, I'm guessing 8-16.
188  *
189  * typical transaction size depends on the application, how often fsync is
190  * called, and how many metadata blocks you dirty in a 30 second period.
191  * The more small files (<16k) you use, the larger your transactions will
192  * be.
193  *
194  * If your journal fills faster than dirty buffers get flushed to disk, it
195  * must flush them before allowing the journal to wrap, which slows things
196  * down.  If you need high speed meta data updates, the journal should be
197  * big enough to prevent wrapping before dirty meta blocks get to disk.
198  *
199  * If the batch max is smaller than the transaction max, you'll waste space
200  * at the end of the journal because journal_end sets the next transaction
201  * to start at 0 if the next transaction has any chance of wrapping.
202  *
203  * The large the batch max age, the better the speed, and the more meta
204  * data changes you'll lose after a crash.
205  */
206 
207 /* don't mess with these for a while */
208 /* we have a node size define somewhere in reiserfs_fs.h. -Hans */
209 #define JOURNAL_BLOCK_SIZE  4096	/* BUG gotta get rid of this */
210 #define JOURNAL_MAX_CNODE   1500	/* max cnodes to allocate. */
211 #define JOURNAL_HASH_SIZE 8192
212 
213 /* number of copies of the bitmaps to have floating.  Must be >= 2 */
214 #define JOURNAL_NUM_BITMAPS 5
215 
216 /*
217  * One of these for every block in every transaction
218  * Each one is in two hash tables.  First, a hash of the current transaction,
219  * and after journal_end, a hash of all the in memory transactions.
220  * next and prev are used by the current transaction (journal_hash).
221  * hnext and hprev are used by journal_list_hash.  If a block is in more
222  * than one transaction, the journal_list_hash links it in multiple times.
223  * This allows flush_journal_list to remove just the cnode belonging to a
224  * given transaction.
225  */
226 struct reiserfs_journal_cnode {
227 	struct buffer_head *bh;	/* real buffer head */
228 	struct super_block *sb;	/* dev of real buffer head */
229 
230 	/* block number of real buffer head, == 0 when buffer on disk */
231 	__u32 blocknr;
232 
233 	unsigned long state;
234 
235 	/* journal list this cnode lives in */
236 	struct reiserfs_journal_list *jlist;
237 
238 	struct reiserfs_journal_cnode *next;	/* next in transaction list */
239 	struct reiserfs_journal_cnode *prev;	/* prev in transaction list */
240 	struct reiserfs_journal_cnode *hprev;	/* prev in hash list */
241 	struct reiserfs_journal_cnode *hnext;	/* next in hash list */
242 };
243 
244 struct reiserfs_bitmap_node {
245 	int id;
246 	char *data;
247 	struct list_head list;
248 };
249 
250 struct reiserfs_list_bitmap {
251 	struct reiserfs_journal_list *journal_list;
252 	struct reiserfs_bitmap_node **bitmaps;
253 };
254 
255 /*
256  * one of these for each transaction.  The most important part here is the
257  * j_realblock.  this list of cnodes is used to hash all the blocks in all
258  * the commits, to mark all the real buffer heads dirty once all the commits
259  * hit the disk, and to make sure every real block in a transaction is on
260  * disk before allowing the log area to be overwritten
261  */
262 struct reiserfs_journal_list {
263 	unsigned long j_start;
264 	unsigned long j_state;
265 	unsigned long j_len;
266 	atomic_t j_nonzerolen;
267 	atomic_t j_commit_left;
268 
269 	/* all commits older than this on disk */
270 	atomic_t j_older_commits_done;
271 
272 	struct mutex j_commit_mutex;
273 	unsigned int j_trans_id;
274 	time64_t j_timestamp; /* write-only but useful for crash dump analysis */
275 	struct reiserfs_list_bitmap *j_list_bitmap;
276 	struct buffer_head *j_commit_bh;	/* commit buffer head */
277 	struct reiserfs_journal_cnode *j_realblock;
278 	struct reiserfs_journal_cnode *j_freedlist;	/* list of buffers that were freed during this trans.  free each of these on flush */
279 	/* time ordered list of all active transactions */
280 	struct list_head j_list;
281 
282 	/*
283 	 * time ordered list of all transactions we haven't tried
284 	 * to flush yet
285 	 */
286 	struct list_head j_working_list;
287 
288 	/* list of tail conversion targets in need of flush before commit */
289 	struct list_head j_tail_bh_list;
290 
291 	/* list of data=ordered buffers in need of flush before commit */
292 	struct list_head j_bh_list;
293 	int j_refcount;
294 };
295 
296 struct reiserfs_journal {
297 	struct buffer_head **j_ap_blocks;	/* journal blocks on disk */
298 	/* newest journal block */
299 	struct reiserfs_journal_cnode *j_last;
300 
301 	/* oldest journal block.  start here for traverse */
302 	struct reiserfs_journal_cnode *j_first;
303 
304 	struct block_device *j_dev_bd;
305 	fmode_t j_dev_mode;
306 
307 	/* first block on s_dev of reserved area journal */
308 	int j_1st_reserved_block;
309 
310 	unsigned long j_state;
311 	unsigned int j_trans_id;
312 	unsigned long j_mount_id;
313 
314 	/* start of current waiting commit (index into j_ap_blocks) */
315 	unsigned long j_start;
316 	unsigned long j_len;	/* length of current waiting commit */
317 
318 	/* number of buffers requested by journal_begin() */
319 	unsigned long j_len_alloc;
320 
321 	atomic_t j_wcount;	/* count of writers for current commit */
322 
323 	/* batch count. allows turning X transactions into 1 */
324 	unsigned long j_bcount;
325 
326 	/* first unflushed transactions offset */
327 	unsigned long j_first_unflushed_offset;
328 
329 	/* last fully flushed journal timestamp */
330 	unsigned j_last_flush_trans_id;
331 
332 	struct buffer_head *j_header_bh;
333 
334 	time64_t j_trans_start_time;	/* time this transaction started */
335 	struct mutex j_mutex;
336 	struct mutex j_flush_mutex;
337 
338 	/* wait for current transaction to finish before starting new one */
339 	wait_queue_head_t j_join_wait;
340 
341 	atomic_t j_jlock;		/* lock for j_join_wait */
342 	int j_list_bitmap_index;	/* number of next list bitmap to use */
343 
344 	/* no more journal begins allowed. MUST sleep on j_join_wait */
345 	int j_must_wait;
346 
347 	/* next journal_end will flush all journal list */
348 	int j_next_full_flush;
349 
350 	/* next journal_end will flush all async commits */
351 	int j_next_async_flush;
352 
353 	int j_cnode_used;	/* number of cnodes on the used list */
354 	int j_cnode_free;	/* number of cnodes on the free list */
355 
356 	/* max number of blocks in a transaction.  */
357 	unsigned int j_trans_max;
358 
359 	/* max number of blocks to batch into a trans */
360 	unsigned int j_max_batch;
361 
362 	/* in seconds, how old can an async commit be */
363 	unsigned int j_max_commit_age;
364 
365 	/* in seconds, how old can a transaction be */
366 	unsigned int j_max_trans_age;
367 
368 	/* the default for the max commit age */
369 	unsigned int j_default_max_commit_age;
370 
371 	struct reiserfs_journal_cnode *j_cnode_free_list;
372 
373 	/* orig pointer returned from vmalloc */
374 	struct reiserfs_journal_cnode *j_cnode_free_orig;
375 
376 	struct reiserfs_journal_list *j_current_jl;
377 	int j_free_bitmap_nodes;
378 	int j_used_bitmap_nodes;
379 
380 	int j_num_lists;	/* total number of active transactions */
381 	int j_num_work_lists;	/* number that need attention from kreiserfsd */
382 
383 	/* debugging to make sure things are flushed in order */
384 	unsigned int j_last_flush_id;
385 
386 	/* debugging to make sure things are committed in order */
387 	unsigned int j_last_commit_id;
388 
389 	struct list_head j_bitmap_nodes;
390 	struct list_head j_dirty_buffers;
391 	spinlock_t j_dirty_buffers_lock;	/* protects j_dirty_buffers */
392 
393 	/* list of all active transactions */
394 	struct list_head j_journal_list;
395 
396 	/* lists that haven't been touched by writeback attempts */
397 	struct list_head j_working_list;
398 
399 	/* hash table for real buffer heads in current trans */
400 	struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE];
401 
402 	/* hash table for all the real buffer heads in all the transactions */
403 	struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE];
404 
405 	/* array of bitmaps to record the deleted blocks */
406 	struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS];
407 
408 	/* list of inodes which have preallocated blocks */
409 	struct list_head j_prealloc_list;
410 	int j_persistent_trans;
411 	unsigned long j_max_trans_size;
412 	unsigned long j_max_batch_size;
413 
414 	int j_errno;
415 
416 	/* when flushing ordered buffers, throttle new ordered writers */
417 	struct delayed_work j_work;
418 	struct super_block *j_work_sb;
419 	atomic_t j_async_throttle;
420 };
421 
422 enum journal_state_bits {
423 	J_WRITERS_BLOCKED = 1,	/* set when new writers not allowed */
424 	J_WRITERS_QUEUED,    /* set when log is full due to too many writers */
425 	J_ABORTED,           /* set when log is aborted */
426 };
427 
428 /* ick.  magic string to find desc blocks in the journal */
429 #define JOURNAL_DESC_MAGIC "ReIsErLB"
430 
431 typedef __u32(*hashf_t) (const signed char *, int);
432 
433 struct reiserfs_bitmap_info {
434 	__u32 free_count;
435 };
436 
437 struct proc_dir_entry;
438 
439 #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
440 typedef unsigned long int stat_cnt_t;
441 typedef struct reiserfs_proc_info_data {
442 	spinlock_t lock;
443 	int exiting;
444 	int max_hash_collisions;
445 
446 	stat_cnt_t breads;
447 	stat_cnt_t bread_miss;
448 	stat_cnt_t search_by_key;
449 	stat_cnt_t search_by_key_fs_changed;
450 	stat_cnt_t search_by_key_restarted;
451 
452 	stat_cnt_t insert_item_restarted;
453 	stat_cnt_t paste_into_item_restarted;
454 	stat_cnt_t cut_from_item_restarted;
455 	stat_cnt_t delete_solid_item_restarted;
456 	stat_cnt_t delete_item_restarted;
457 
458 	stat_cnt_t leaked_oid;
459 	stat_cnt_t leaves_removable;
460 
461 	/*
462 	 * balances per level.
463 	 * Use explicit 5 as MAX_HEIGHT is not visible yet.
464 	 */
465 	stat_cnt_t balance_at[5];	/* XXX */
466 	/* sbk == search_by_key */
467 	stat_cnt_t sbk_read_at[5];	/* XXX */
468 	stat_cnt_t sbk_fs_changed[5];
469 	stat_cnt_t sbk_restarted[5];
470 	stat_cnt_t items_at[5];	/* XXX */
471 	stat_cnt_t free_at[5];	/* XXX */
472 	stat_cnt_t can_node_be_removed[5];	/* XXX */
473 	long int lnum[5];	/* XXX */
474 	long int rnum[5];	/* XXX */
475 	long int lbytes[5];	/* XXX */
476 	long int rbytes[5];	/* XXX */
477 	stat_cnt_t get_neighbors[5];
478 	stat_cnt_t get_neighbors_restart[5];
479 	stat_cnt_t need_l_neighbor[5];
480 	stat_cnt_t need_r_neighbor[5];
481 
482 	stat_cnt_t free_block;
483 	struct __scan_bitmap_stats {
484 		stat_cnt_t call;
485 		stat_cnt_t wait;
486 		stat_cnt_t bmap;
487 		stat_cnt_t retry;
488 		stat_cnt_t in_journal_hint;
489 		stat_cnt_t in_journal_nohint;
490 		stat_cnt_t stolen;
491 	} scan_bitmap;
492 	struct __journal_stats {
493 		stat_cnt_t in_journal;
494 		stat_cnt_t in_journal_bitmap;
495 		stat_cnt_t in_journal_reusable;
496 		stat_cnt_t lock_journal;
497 		stat_cnt_t lock_journal_wait;
498 		stat_cnt_t journal_being;
499 		stat_cnt_t journal_relock_writers;
500 		stat_cnt_t journal_relock_wcount;
501 		stat_cnt_t mark_dirty;
502 		stat_cnt_t mark_dirty_already;
503 		stat_cnt_t mark_dirty_notjournal;
504 		stat_cnt_t restore_prepared;
505 		stat_cnt_t prepare;
506 		stat_cnt_t prepare_retry;
507 	} journal;
508 } reiserfs_proc_info_data_t;
509 #else
510 typedef struct reiserfs_proc_info_data {
511 } reiserfs_proc_info_data_t;
512 #endif
513 
514 /* Number of quota types we support */
515 #define REISERFS_MAXQUOTAS 2
516 
517 /* reiserfs union of in-core super block data */
518 struct reiserfs_sb_info {
519 	/* Buffer containing the super block */
520 	struct buffer_head *s_sbh;
521 
522 	/* Pointer to the on-disk super block in the buffer */
523 	struct reiserfs_super_block *s_rs;
524 	struct reiserfs_bitmap_info *s_ap_bitmap;
525 
526 	/* pointer to journal information */
527 	struct reiserfs_journal *s_journal;
528 
529 	unsigned short s_mount_state;	/* reiserfs state (valid, invalid) */
530 
531 	/* Serialize writers access, replace the old bkl */
532 	struct mutex lock;
533 
534 	/* Owner of the lock (can be recursive) */
535 	struct task_struct *lock_owner;
536 
537 	/* Depth of the lock, start from -1 like the bkl */
538 	int lock_depth;
539 
540 	struct workqueue_struct *commit_wq;
541 
542 	/* Comment? -Hans */
543 	void (*end_io_handler) (struct buffer_head *, int);
544 
545 	/*
546 	 * pointer to function which is used to sort names in directory.
547 	 * Set on mount
548 	 */
549 	hashf_t s_hash_function;
550 
551 	/* reiserfs's mount options are set here */
552 	unsigned long s_mount_opt;
553 
554 	/* This is a structure that describes block allocator options */
555 	struct {
556 		/* Bitfield for enable/disable kind of options */
557 		unsigned long bits;
558 
559 		/*
560 		 * size started from which we consider file
561 		 * to be a large one (in blocks)
562 		 */
563 		unsigned long large_file_size;
564 
565 		int border;	/* percentage of disk, border takes */
566 
567 		/*
568 		 * Minimal file size (in blocks) starting
569 		 * from which we do preallocations
570 		 */
571 		int preallocmin;
572 
573 		/*
574 		 * Number of blocks we try to prealloc when file
575 		 * reaches preallocmin size (in blocks) or prealloc_list
576 		 is empty.
577 		 */
578 		int preallocsize;
579 	} s_alloc_options;
580 
581 	/* Comment? -Hans */
582 	wait_queue_head_t s_wait;
583 	/* increased by one every time the  tree gets re-balanced */
584 	atomic_t s_generation_counter;
585 
586 	/* File system properties. Currently holds on-disk FS format */
587 	unsigned long s_properties;
588 
589 	/* session statistics */
590 	int s_disk_reads;
591 	int s_disk_writes;
592 	int s_fix_nodes;
593 	int s_do_balance;
594 	int s_unneeded_left_neighbor;
595 	int s_good_search_by_key_reada;
596 	int s_bmaps;
597 	int s_bmaps_without_search;
598 	int s_direct2indirect;
599 	int s_indirect2direct;
600 
601 	/*
602 	 * set up when it's ok for reiserfs_read_inode2() to read from
603 	 * disk inode with nlink==0. Currently this is only used during
604 	 * finish_unfinished() processing at mount time
605 	 */
606 	int s_is_unlinked_ok;
607 
608 	reiserfs_proc_info_data_t s_proc_info_data;
609 	struct proc_dir_entry *procdir;
610 
611 	/* amount of blocks reserved for further allocations */
612 	int reserved_blocks;
613 
614 
615 	/* this lock on now only used to protect reserved_blocks variable */
616 	spinlock_t bitmap_lock;
617 	struct dentry *priv_root;	/* root of /.reiserfs_priv */
618 	struct dentry *xattr_root;	/* root of /.reiserfs_priv/xattrs */
619 	int j_errno;
620 
621 	int work_queued;              /* non-zero delayed work is queued */
622 	struct delayed_work old_work; /* old transactions flush delayed work */
623 	spinlock_t old_work_lock;     /* protects old_work and work_queued */
624 
625 #ifdef CONFIG_QUOTA
626 	char *s_qf_names[REISERFS_MAXQUOTAS];
627 	int s_jquota_fmt;
628 #endif
629 	char *s_jdev;		/* Stored jdev for mount option showing */
630 #ifdef CONFIG_REISERFS_CHECK
631 
632 	/*
633 	 * Detects whether more than one copy of tb exists per superblock
634 	 * as a means of checking whether do_balance is executing
635 	 * concurrently against another tree reader/writer on a same
636 	 * mount point.
637 	 */
638 	struct tree_balance *cur_tb;
639 #endif
640 };
641 
642 /* Definitions of reiserfs on-disk properties: */
643 #define REISERFS_3_5 0
644 #define REISERFS_3_6 1
645 #define REISERFS_OLD_FORMAT 2
646 
647 /* Mount options */
648 enum reiserfs_mount_options {
649 	/* large tails will be created in a session */
650 	REISERFS_LARGETAIL,
651 	/*
652 	 * small (for files less than block size) tails will
653 	 * be created in a session
654 	 */
655 	REISERFS_SMALLTAIL,
656 
657 	/* replay journal and return 0. Use by fsck */
658 	REPLAYONLY,
659 
660 	/*
661 	 * -o conv: causes conversion of old format super block to the
662 	 * new format. If not specified - old partition will be dealt
663 	 * with in a manner of 3.5.x
664 	 */
665 	REISERFS_CONVERT,
666 
667 	/*
668 	 * -o hash={tea, rupasov, r5, detect} is meant for properly mounting
669 	 * reiserfs disks from 3.5.19 or earlier.  99% of the time, this
670 	 * option is not required.  If the normal autodection code can't
671 	 * determine which hash to use (because both hashes had the same
672 	 * value for a file) use this option to force a specific hash.
673 	 * It won't allow you to override the existing hash on the FS, so
674 	 * if you have a tea hash disk, and mount with -o hash=rupasov,
675 	 * the mount will fail.
676 	 */
677 	FORCE_TEA_HASH,		/* try to force tea hash on mount */
678 	FORCE_RUPASOV_HASH,	/* try to force rupasov hash on mount */
679 	FORCE_R5_HASH,		/* try to force rupasov hash on mount */
680 	FORCE_HASH_DETECT,	/* try to detect hash function on mount */
681 
682 	REISERFS_DATA_LOG,
683 	REISERFS_DATA_ORDERED,
684 	REISERFS_DATA_WRITEBACK,
685 
686 	/*
687 	 * used for testing experimental features, makes benchmarking new
688 	 * features with and without more convenient, should never be used by
689 	 * users in any code shipped to users (ideally)
690 	 */
691 
692 	REISERFS_NO_BORDER,
693 	REISERFS_NO_UNHASHED_RELOCATION,
694 	REISERFS_HASHED_RELOCATION,
695 	REISERFS_ATTRS,
696 	REISERFS_XATTRS_USER,
697 	REISERFS_POSIXACL,
698 	REISERFS_EXPOSE_PRIVROOT,
699 	REISERFS_BARRIER_NONE,
700 	REISERFS_BARRIER_FLUSH,
701 
702 	/* Actions on error */
703 	REISERFS_ERROR_PANIC,
704 	REISERFS_ERROR_RO,
705 	REISERFS_ERROR_CONTINUE,
706 
707 	REISERFS_USRQUOTA,	/* User quota option specified */
708 	REISERFS_GRPQUOTA,	/* Group quota option specified */
709 
710 	REISERFS_TEST1,
711 	REISERFS_TEST2,
712 	REISERFS_TEST3,
713 	REISERFS_TEST4,
714 	REISERFS_UNSUPPORTED_OPT,
715 };
716 
717 #define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))
718 #define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))
719 #define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))
720 #define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))
721 #define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))
722 #define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))
723 #define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))
724 #define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))
725 
726 #define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))
727 #define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))
728 #define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))
729 #define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))
730 #define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))
731 #define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))
732 #define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))
733 #define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))
734 #define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))
735 #define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))
736 #define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))
737 #define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))
738 #define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))
739 #define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))
740 #define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))
741 
742 #define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))
743 #define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))
744 
745 void reiserfs_file_buffer(struct buffer_head *bh, int list);
746 extern struct file_system_type reiserfs_fs_type;
747 int reiserfs_resize(struct super_block *, unsigned long);
748 
749 #define CARRY_ON                0
750 #define SCHEDULE_OCCURRED       1
751 
752 #define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)
753 #define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)
754 #define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)
755 #define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)
756 #define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)
757 
758 #define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)
759 
760 #define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))
__reiserfs_is_journal_aborted(struct reiserfs_journal * journal)761 static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal
762 						*journal)
763 {
764 	return test_bit(J_ABORTED, &journal->j_state);
765 }
766 
767 /*
768  * Locking primitives. The write lock is a per superblock
769  * special mutex that has properties close to the Big Kernel Lock
770  * which was used in the previous locking scheme.
771  */
772 void reiserfs_write_lock(struct super_block *s);
773 void reiserfs_write_unlock(struct super_block *s);
774 int __must_check reiserfs_write_unlock_nested(struct super_block *s);
775 void reiserfs_write_lock_nested(struct super_block *s, int depth);
776 
777 #ifdef CONFIG_REISERFS_CHECK
778 void reiserfs_lock_check_recursive(struct super_block *s);
779 #else
reiserfs_lock_check_recursive(struct super_block * s)780 static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
781 #endif
782 
783 /*
784  * Several mutexes depend on the write lock.
785  * However sometimes we want to relax the write lock while we hold
786  * these mutexes, according to the release/reacquire on schedule()
787  * properties of the Bkl that were used.
788  * Reiserfs performances and locking were based on this scheme.
789  * Now that the write lock is a mutex and not the bkl anymore, doing so
790  * may result in a deadlock:
791  *
792  * A acquire write_lock
793  * A acquire j_commit_mutex
794  * A release write_lock and wait for something
795  * B acquire write_lock
796  * B can't acquire j_commit_mutex and sleep
797  * A can't acquire write lock anymore
798  * deadlock
799  *
800  * What we do here is avoiding such deadlock by playing the same game
801  * than the Bkl: if we can't acquire a mutex that depends on the write lock,
802  * we release the write lock, wait a bit and then retry.
803  *
804  * The mutexes concerned by this hack are:
805  * - The commit mutex of a journal list
806  * - The flush mutex
807  * - The journal lock
808  * - The inode mutex
809  */
reiserfs_mutex_lock_safe(struct mutex * m,struct super_block * s)810 static inline void reiserfs_mutex_lock_safe(struct mutex *m,
811 					    struct super_block *s)
812 {
813 	int depth;
814 
815 	depth = reiserfs_write_unlock_nested(s);
816 	mutex_lock(m);
817 	reiserfs_write_lock_nested(s, depth);
818 }
819 
820 static inline void
reiserfs_mutex_lock_nested_safe(struct mutex * m,unsigned int subclass,struct super_block * s)821 reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
822 				struct super_block *s)
823 {
824 	int depth;
825 
826 	depth = reiserfs_write_unlock_nested(s);
827 	mutex_lock_nested(m, subclass);
828 	reiserfs_write_lock_nested(s, depth);
829 }
830 
831 static inline void
reiserfs_down_read_safe(struct rw_semaphore * sem,struct super_block * s)832 reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
833 {
834        int depth;
835        depth = reiserfs_write_unlock_nested(s);
836        down_read(sem);
837        reiserfs_write_lock_nested(s, depth);
838 }
839 
840 /*
841  * When we schedule, we usually want to also release the write lock,
842  * according to the previous bkl based locking scheme of reiserfs.
843  */
reiserfs_cond_resched(struct super_block * s)844 static inline void reiserfs_cond_resched(struct super_block *s)
845 {
846 	if (need_resched()) {
847 		int depth;
848 
849 		depth = reiserfs_write_unlock_nested(s);
850 		schedule();
851 		reiserfs_write_lock_nested(s, depth);
852 	}
853 }
854 
855 struct fid;
856 
857 /*
858  * in reading the #defines, it may help to understand that they employ
859  *  the following abbreviations:
860  *
861  *  B = Buffer
862  *  I = Item header
863  *  H = Height within the tree (should be changed to LEV)
864  *  N = Number of the item in the node
865  *  STAT = stat data
866  *  DEH = Directory Entry Header
867  *  EC = Entry Count
868  *  E = Entry number
869  *  UL = Unsigned Long
870  *  BLKH = BLocK Header
871  *  UNFM = UNForMatted node
872  *  DC = Disk Child
873  *  P = Path
874  *
875  *  These #defines are named by concatenating these abbreviations,
876  *  where first comes the arguments, and last comes the return value,
877  *  of the macro.
878  */
879 
880 #define USE_INODE_GENERATION_COUNTER
881 
882 #define REISERFS_PREALLOCATE
883 #define DISPLACE_NEW_PACKING_LOCALITIES
884 #define PREALLOCATION_SIZE 9
885 
886 /* n must be power of 2 */
887 #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
888 
889 /*
890  * to be ok for alpha and others we have to align structures to 8 byte
891  * boundary.
892  * FIXME: do not change 4 by anything else: there is code which relies on that
893  */
894 #define ROUND_UP(x) _ROUND_UP(x,8LL)
895 
896 /*
897  * debug levels.  Right now, CONFIG_REISERFS_CHECK means print all debug
898  * messages.
899  */
900 #define REISERFS_DEBUG_CODE 5	/* extra messages to help find/debug errors */
901 
902 void __reiserfs_warning(struct super_block *s, const char *id,
903 			 const char *func, const char *fmt, ...);
904 #define reiserfs_warning(s, id, fmt, args...) \
905 	 __reiserfs_warning(s, id, __func__, fmt, ##args)
906 /* assertions handling */
907 
908 /* always check a condition and panic if it's false. */
909 #define __RASSERT(cond, scond, format, args...)			\
910 do {									\
911 	if (!(cond))							\
912 		reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
913 			       __FILE__ ":%i:%s: " format "\n",		\
914 			       __LINE__, __func__ , ##args);		\
915 } while (0)
916 
917 #define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
918 
919 #if defined( CONFIG_REISERFS_CHECK )
920 #define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
921 #else
922 #define RFALSE( cond, format, args... ) do {;} while( 0 )
923 #endif
924 
925 #define CONSTF __attribute_const__
926 /*
927  * Disk Data Structures
928  */
929 
930 /***************************************************************************
931  *                             SUPER BLOCK                                 *
932  ***************************************************************************/
933 
934 /*
935  * Structure of super block on disk, a version of which in RAM is often
936  * accessed as REISERFS_SB(s)->s_rs. The version in RAM is part of a larger
937  * structure containing fields never written to disk.
938  */
939 #define UNSET_HASH 0	/* Detect hash on disk */
940 #define TEA_HASH  1
941 #define YURA_HASH 2
942 #define R5_HASH   3
943 #define DEFAULT_HASH R5_HASH
944 
945 struct journal_params {
946 	/* where does journal start from on its * device */
947 	__le32 jp_journal_1st_block;
948 
949 	/* journal device st_rdev */
950 	__le32 jp_journal_dev;
951 
952 	/* size of the journal */
953 	__le32 jp_journal_size;
954 
955 	/* max number of blocks in a transaction. */
956 	__le32 jp_journal_trans_max;
957 
958 	/*
959 	 * random value made on fs creation
960 	 * (this was sb_journal_block_count)
961 	 */
962 	__le32 jp_journal_magic;
963 
964 	/* max number of blocks to batch into a trans */
965 	__le32 jp_journal_max_batch;
966 
967 	/* in seconds, how old can an async  commit be */
968 	__le32 jp_journal_max_commit_age;
969 
970 	/* in seconds, how old can a transaction be */
971 	__le32 jp_journal_max_trans_age;
972 };
973 
974 /* this is the super from 3.5.X, where X >= 10 */
975 struct reiserfs_super_block_v1 {
976 	__le32 s_block_count;	/* blocks count         */
977 	__le32 s_free_blocks;	/* free blocks count    */
978 	__le32 s_root_block;	/* root block number    */
979 	struct journal_params s_journal;
980 	__le16 s_blocksize;	/* block size */
981 
982 	/* max size of object id array, see get_objectid() commentary  */
983 	__le16 s_oid_maxsize;
984 	__le16 s_oid_cursize;	/* current size of object id array */
985 
986 	/* this is set to 1 when filesystem was umounted, to 2 - when not */
987 	__le16 s_umount_state;
988 
989 	/*
990 	 * reiserfs magic string indicates that file system is reiserfs:
991 	 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs"
992 	 */
993 	char s_magic[10];
994 
995 	/*
996 	 * it is set to used by fsck to mark which
997 	 * phase of rebuilding is done
998 	 */
999 	__le16 s_fs_state;
1000 	/*
1001 	 * indicate, what hash function is being use
1002 	 * to sort names in a directory
1003 	 */
1004 	__le32 s_hash_function_code;
1005 	__le16 s_tree_height;	/* height of disk tree */
1006 
1007 	/*
1008 	 * amount of bitmap blocks needed to address
1009 	 * each block of file system
1010 	 */
1011 	__le16 s_bmap_nr;
1012 
1013 	/*
1014 	 * this field is only reliable on filesystem with non-standard journal
1015 	 */
1016 	__le16 s_version;
1017 
1018 	/*
1019 	 * size in blocks of journal area on main device, we need to
1020 	 * keep after making fs with non-standard journal
1021 	 */
1022 	__le16 s_reserved_for_journal;
1023 } __attribute__ ((__packed__));
1024 
1025 #define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
1026 
1027 /* this is the on disk super block */
1028 struct reiserfs_super_block {
1029 	struct reiserfs_super_block_v1 s_v1;
1030 	__le32 s_inode_generation;
1031 
1032 	/* Right now used only by inode-attributes, if enabled */
1033 	__le32 s_flags;
1034 
1035 	unsigned char s_uuid[16];	/* filesystem unique identifier */
1036 	unsigned char s_label[16];	/* filesystem volume label */
1037 	__le16 s_mnt_count;		/* Count of mounts since last fsck */
1038 	__le16 s_max_mnt_count;		/* Maximum mounts before check */
1039 	__le32 s_lastcheck;		/* Timestamp of last fsck */
1040 	__le32 s_check_interval;	/* Interval between checks */
1041 
1042 	/*
1043 	 * zero filled by mkreiserfs and reiserfs_convert_objectid_map_v1()
1044 	 * so any additions must be updated there as well. */
1045 	char s_unused[76];
1046 } __attribute__ ((__packed__));
1047 
1048 #define SB_SIZE (sizeof(struct reiserfs_super_block))
1049 
1050 #define REISERFS_VERSION_1 0
1051 #define REISERFS_VERSION_2 2
1052 
1053 /* on-disk super block fields converted to cpu form */
1054 #define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
1055 #define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
1056 #define SB_BLOCKSIZE(s) \
1057         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
1058 #define SB_BLOCK_COUNT(s) \
1059         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
1060 #define SB_FREE_BLOCKS(s) \
1061         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
1062 #define SB_REISERFS_MAGIC(s) \
1063         (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
1064 #define SB_ROOT_BLOCK(s) \
1065         le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
1066 #define SB_TREE_HEIGHT(s) \
1067         le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
1068 #define SB_REISERFS_STATE(s) \
1069         le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
1070 #define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
1071 #define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
1072 
1073 #define PUT_SB_BLOCK_COUNT(s, val) \
1074    do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
1075 #define PUT_SB_FREE_BLOCKS(s, val) \
1076    do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
1077 #define PUT_SB_ROOT_BLOCK(s, val) \
1078    do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
1079 #define PUT_SB_TREE_HEIGHT(s, val) \
1080    do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
1081 #define PUT_SB_REISERFS_STATE(s, val) \
1082    do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1083 #define PUT_SB_VERSION(s, val) \
1084    do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
1085 #define PUT_SB_BMAP_NR(s, val) \
1086    do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
1087 
1088 #define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
1089 #define SB_ONDISK_JOURNAL_SIZE(s) \
1090          le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
1091 #define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
1092          le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
1093 #define SB_ONDISK_JOURNAL_DEVICE(s) \
1094          le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
1095 #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
1096          le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1097 
1098 #define is_block_in_log_or_reserved_area(s, block) \
1099          block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
1100          && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) +  \
1101          ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
1102          SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1103 
1104 int is_reiserfs_3_5(struct reiserfs_super_block *rs);
1105 int is_reiserfs_3_6(struct reiserfs_super_block *rs);
1106 int is_reiserfs_jr(struct reiserfs_super_block *rs);
1107 
1108 /*
1109  * ReiserFS leaves the first 64k unused, so that partition labels have
1110  * enough space.  If someone wants to write a fancy bootloader that
1111  * needs more than 64k, let us know, and this will be increased in size.
1112  * This number must be larger than than the largest block size on any
1113  * platform, or code will break.  -Hans
1114  */
1115 #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
1116 #define REISERFS_FIRST_BLOCK unused_define
1117 #define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
1118 
1119 /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
1120 #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
1121 
1122 /* reiserfs internal error code (used by search_by_key and fix_nodes)) */
1123 #define CARRY_ON      0
1124 #define REPEAT_SEARCH -1
1125 #define IO_ERROR      -2
1126 #define NO_DISK_SPACE -3
1127 #define NO_BALANCING_NEEDED  (-4)
1128 #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
1129 #define QUOTA_EXCEEDED -6
1130 
1131 typedef __u32 b_blocknr_t;
1132 typedef __le32 unp_t;
1133 
1134 struct unfm_nodeinfo {
1135 	unp_t unfm_nodenum;
1136 	unsigned short unfm_freespace;
1137 };
1138 
1139 /* there are two formats of keys: 3.5 and 3.6 */
1140 #define KEY_FORMAT_3_5 0
1141 #define KEY_FORMAT_3_6 1
1142 
1143 /* there are two stat datas */
1144 #define STAT_DATA_V1 0
1145 #define STAT_DATA_V2 1
1146 
REISERFS_I(const struct inode * inode)1147 static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
1148 {
1149 	return container_of(inode, struct reiserfs_inode_info, vfs_inode);
1150 }
1151 
REISERFS_SB(const struct super_block * sb)1152 static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
1153 {
1154 	return sb->s_fs_info;
1155 }
1156 
1157 /*
1158  * Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
1159  * which overflows on large file systems.
1160  */
reiserfs_bmap_count(struct super_block * sb)1161 static inline __u32 reiserfs_bmap_count(struct super_block *sb)
1162 {
1163 	return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
1164 }
1165 
bmap_would_wrap(unsigned bmap_nr)1166 static inline int bmap_would_wrap(unsigned bmap_nr)
1167 {
1168 	return bmap_nr > ((1LL << 16) - 1);
1169 }
1170 
1171 extern const struct xattr_handler *reiserfs_xattr_handlers[];
1172 
1173 /*
1174  * this says about version of key of all items (but stat data) the
1175  * object consists of
1176  */
1177 #define get_inode_item_key_version( inode )                                    \
1178     ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
1179 
1180 #define set_inode_item_key_version( inode, version )                           \
1181          ({ if((version)==KEY_FORMAT_3_6)                                      \
1182                 REISERFS_I(inode)->i_flags |= i_item_key_version_mask;      \
1183             else                                                               \
1184                 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
1185 
1186 #define get_inode_sd_version(inode)                                            \
1187     ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
1188 
1189 #define set_inode_sd_version(inode, version)                                   \
1190          ({ if((version)==STAT_DATA_V2)                                        \
1191                 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask;     \
1192             else                                                               \
1193                 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
1194 
1195 /*
1196  * This is an aggressive tail suppression policy, I am hoping it
1197  * improves our benchmarks. The principle behind it is that percentage
1198  * space saving is what matters, not absolute space saving.  This is
1199  * non-intuitive, but it helps to understand it if you consider that the
1200  * cost to access 4 blocks is not much more than the cost to access 1
1201  * block, if you have to do a seek and rotate.  A tail risks a
1202  * non-linear disk access that is significant as a percentage of total
1203  * time cost for a 4 block file and saves an amount of space that is
1204  * less significant as a percentage of space, or so goes the hypothesis.
1205  * -Hans
1206  */
1207 #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
1208 (\
1209   (!(n_tail_size)) || \
1210   (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
1211    ( (n_file_size) >= (n_block_size) * 4 ) || \
1212    ( ( (n_file_size) >= (n_block_size) * 3 ) && \
1213      ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
1214    ( ( (n_file_size) >= (n_block_size) * 2 ) && \
1215      ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
1216    ( ( (n_file_size) >= (n_block_size) ) && \
1217      ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
1218 )
1219 
1220 /*
1221  * Another strategy for tails, this one means only create a tail if all the
1222  * file would fit into one DIRECT item.
1223  * Primary intention for this one is to increase performance by decreasing
1224  * seeking.
1225 */
1226 #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
1227 (\
1228   (!(n_tail_size)) || \
1229   (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
1230 )
1231 
1232 /*
1233  * values for s_umount_state field
1234  */
1235 #define REISERFS_VALID_FS    1
1236 #define REISERFS_ERROR_FS    2
1237 
1238 /*
1239  * there are 5 item types currently
1240  */
1241 #define TYPE_STAT_DATA 0
1242 #define TYPE_INDIRECT 1
1243 #define TYPE_DIRECT 2
1244 #define TYPE_DIRENTRY 3
1245 #define TYPE_MAXTYPE 3
1246 #define TYPE_ANY 15		/* FIXME: comment is required */
1247 
1248 /***************************************************************************
1249  *                       KEY & ITEM HEAD                                   *
1250  ***************************************************************************/
1251 
1252 /* * directories use this key as well as old files */
1253 struct offset_v1 {
1254 	__le32 k_offset;
1255 	__le32 k_uniqueness;
1256 } __attribute__ ((__packed__));
1257 
1258 struct offset_v2 {
1259 	__le64 v;
1260 } __attribute__ ((__packed__));
1261 
offset_v2_k_type(const struct offset_v2 * v2)1262 static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1263 {
1264 	__u8 type = le64_to_cpu(v2->v) >> 60;
1265 	return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1266 }
1267 
set_offset_v2_k_type(struct offset_v2 * v2,int type)1268 static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1269 {
1270 	v2->v =
1271 	    (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1272 }
1273 
offset_v2_k_offset(const struct offset_v2 * v2)1274 static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1275 {
1276 	return le64_to_cpu(v2->v) & (~0ULL >> 4);
1277 }
1278 
set_offset_v2_k_offset(struct offset_v2 * v2,loff_t offset)1279 static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
1280 {
1281 	offset &= (~0ULL >> 4);
1282 	v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1283 }
1284 
1285 /*
1286  * Key of an item determines its location in the S+tree, and
1287  * is composed of 4 components
1288  */
1289 struct reiserfs_key {
1290 	/* packing locality: by default parent directory object id */
1291 	__le32 k_dir_id;
1292 
1293 	__le32 k_objectid;	/* object identifier */
1294 	union {
1295 		struct offset_v1 k_offset_v1;
1296 		struct offset_v2 k_offset_v2;
1297 	} __attribute__ ((__packed__)) u;
1298 } __attribute__ ((__packed__));
1299 
1300 struct in_core_key {
1301 	/* packing locality: by default parent directory object id */
1302 	__u32 k_dir_id;
1303 	__u32 k_objectid;	/* object identifier */
1304 	__u64 k_offset;
1305 	__u8 k_type;
1306 };
1307 
1308 struct cpu_key {
1309 	struct in_core_key on_disk_key;
1310 	int version;
1311 	/* 3 in all cases but direct2indirect and indirect2direct conversion */
1312 	int key_length;
1313 };
1314 
1315 /*
1316  * Our function for comparing keys can compare keys of different
1317  * lengths.  It takes as a parameter the length of the keys it is to
1318  * compare.  These defines are used in determining what is to be passed
1319  * to it as that parameter.
1320  */
1321 #define REISERFS_FULL_KEY_LEN     4
1322 #define REISERFS_SHORT_KEY_LEN    2
1323 
1324 /* The result of the key compare */
1325 #define FIRST_GREATER 1
1326 #define SECOND_GREATER -1
1327 #define KEYS_IDENTICAL 0
1328 #define KEY_FOUND 1
1329 #define KEY_NOT_FOUND 0
1330 
1331 #define KEY_SIZE (sizeof(struct reiserfs_key))
1332 
1333 /* return values for search_by_key and clones */
1334 #define ITEM_FOUND 1
1335 #define ITEM_NOT_FOUND 0
1336 #define ENTRY_FOUND 1
1337 #define ENTRY_NOT_FOUND 0
1338 #define DIRECTORY_NOT_FOUND -1
1339 #define REGULAR_FILE_FOUND -2
1340 #define DIRECTORY_FOUND -3
1341 #define BYTE_FOUND 1
1342 #define BYTE_NOT_FOUND 0
1343 #define FILE_NOT_FOUND -1
1344 
1345 #define POSITION_FOUND 1
1346 #define POSITION_NOT_FOUND 0
1347 
1348 /* return values for reiserfs_find_entry and search_by_entry_key */
1349 #define NAME_FOUND 1
1350 #define NAME_NOT_FOUND 0
1351 #define GOTO_PREVIOUS_ITEM 2
1352 #define NAME_FOUND_INVISIBLE 3
1353 
1354 /*
1355  * Everything in the filesystem is stored as a set of items.  The
1356  * item head contains the key of the item, its free space (for
1357  * indirect items) and specifies the location of the item itself
1358  * within the block.
1359  */
1360 
1361 struct item_head {
1362 	/*
1363 	 * Everything in the tree is found by searching for it based on
1364 	 * its key.
1365 	 */
1366 	struct reiserfs_key ih_key;
1367 	union {
1368 		/*
1369 		 * The free space in the last unformatted node of an
1370 		 * indirect item if this is an indirect item.  This
1371 		 * equals 0xFFFF iff this is a direct item or stat data
1372 		 * item. Note that the key, not this field, is used to
1373 		 * determine the item type, and thus which field this
1374 		 * union contains.
1375 		 */
1376 		__le16 ih_free_space_reserved;
1377 
1378 		/*
1379 		 * Iff this is a directory item, this field equals the
1380 		 * number of directory entries in the directory item.
1381 		 */
1382 		__le16 ih_entry_count;
1383 	} __attribute__ ((__packed__)) u;
1384 	__le16 ih_item_len;	/* total size of the item body */
1385 
1386 	/* an offset to the item body within the block */
1387 	__le16 ih_item_location;
1388 
1389 	/*
1390 	 * 0 for all old items, 2 for new ones. Highest bit is set by fsck
1391 	 * temporary, cleaned after all done
1392 	 */
1393 	__le16 ih_version;
1394 } __attribute__ ((__packed__));
1395 /* size of item header     */
1396 #define IH_SIZE (sizeof(struct item_head))
1397 
1398 #define ih_free_space(ih)            le16_to_cpu((ih)->u.ih_free_space_reserved)
1399 #define ih_version(ih)               le16_to_cpu((ih)->ih_version)
1400 #define ih_entry_count(ih)           le16_to_cpu((ih)->u.ih_entry_count)
1401 #define ih_location(ih)              le16_to_cpu((ih)->ih_item_location)
1402 #define ih_item_len(ih)              le16_to_cpu((ih)->ih_item_len)
1403 
1404 #define put_ih_free_space(ih, val)   do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
1405 #define put_ih_version(ih, val)      do { (ih)->ih_version = cpu_to_le16(val); } while (0)
1406 #define put_ih_entry_count(ih, val)  do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
1407 #define put_ih_location(ih, val)     do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
1408 #define put_ih_item_len(ih, val)     do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
1409 
1410 #define unreachable_item(ih) (ih_version(ih) & (1 << 15))
1411 
1412 #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
1413 #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
1414 
1415 /*
1416  * these operate on indirect items, where you've got an array of ints
1417  * at a possibly unaligned location.  These are a noop on ia32
1418  *
1419  * p is the array of __u32, i is the index into the array, v is the value
1420  * to store there.
1421  */
1422 #define get_block_num(p, i) get_unaligned_le32((p) + (i))
1423 #define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1424 
1425 /* * in old version uniqueness field shows key type */
1426 #define V1_SD_UNIQUENESS 0
1427 #define V1_INDIRECT_UNIQUENESS 0xfffffffe
1428 #define V1_DIRECT_UNIQUENESS 0xffffffff
1429 #define V1_DIRENTRY_UNIQUENESS 500
1430 #define V1_ANY_UNIQUENESS 555	/* FIXME: comment is required */
1431 
1432 /* here are conversion routines */
1433 static inline int uniqueness2type(__u32 uniqueness) CONSTF;
uniqueness2type(__u32 uniqueness)1434 static inline int uniqueness2type(__u32 uniqueness)
1435 {
1436 	switch ((int)uniqueness) {
1437 	case V1_SD_UNIQUENESS:
1438 		return TYPE_STAT_DATA;
1439 	case V1_INDIRECT_UNIQUENESS:
1440 		return TYPE_INDIRECT;
1441 	case V1_DIRECT_UNIQUENESS:
1442 		return TYPE_DIRECT;
1443 	case V1_DIRENTRY_UNIQUENESS:
1444 		return TYPE_DIRENTRY;
1445 	case V1_ANY_UNIQUENESS:
1446 	default:
1447 		return TYPE_ANY;
1448 	}
1449 }
1450 
1451 static inline __u32 type2uniqueness(int type) CONSTF;
type2uniqueness(int type)1452 static inline __u32 type2uniqueness(int type)
1453 {
1454 	switch (type) {
1455 	case TYPE_STAT_DATA:
1456 		return V1_SD_UNIQUENESS;
1457 	case TYPE_INDIRECT:
1458 		return V1_INDIRECT_UNIQUENESS;
1459 	case TYPE_DIRECT:
1460 		return V1_DIRECT_UNIQUENESS;
1461 	case TYPE_DIRENTRY:
1462 		return V1_DIRENTRY_UNIQUENESS;
1463 	case TYPE_ANY:
1464 	default:
1465 		return V1_ANY_UNIQUENESS;
1466 	}
1467 }
1468 
1469 /*
1470  * key is pointer to on disk key which is stored in le, result is cpu,
1471  * there is no way to get version of object from key, so, provide
1472  * version to these defines
1473  */
le_key_k_offset(int version,const struct reiserfs_key * key)1474 static inline loff_t le_key_k_offset(int version,
1475 				     const struct reiserfs_key *key)
1476 {
1477 	return (version == KEY_FORMAT_3_5) ?
1478 	    le32_to_cpu(key->u.k_offset_v1.k_offset) :
1479 	    offset_v2_k_offset(&(key->u.k_offset_v2));
1480 }
1481 
le_ih_k_offset(const struct item_head * ih)1482 static inline loff_t le_ih_k_offset(const struct item_head *ih)
1483 {
1484 	return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1485 }
1486 
le_key_k_type(int version,const struct reiserfs_key * key)1487 static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1488 {
1489 	if (version == KEY_FORMAT_3_5) {
1490 		loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness);
1491 		return uniqueness2type(val);
1492 	} else
1493 		return offset_v2_k_type(&(key->u.k_offset_v2));
1494 }
1495 
le_ih_k_type(const struct item_head * ih)1496 static inline loff_t le_ih_k_type(const struct item_head *ih)
1497 {
1498 	return le_key_k_type(ih_version(ih), &(ih->ih_key));
1499 }
1500 
set_le_key_k_offset(int version,struct reiserfs_key * key,loff_t offset)1501 static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
1502 				       loff_t offset)
1503 {
1504 	if (version == KEY_FORMAT_3_5)
1505 		key->u.k_offset_v1.k_offset = cpu_to_le32(offset);
1506 	else
1507 		set_offset_v2_k_offset(&key->u.k_offset_v2, offset);
1508 }
1509 
add_le_key_k_offset(int version,struct reiserfs_key * key,loff_t offset)1510 static inline void add_le_key_k_offset(int version, struct reiserfs_key *key,
1511 				       loff_t offset)
1512 {
1513 	set_le_key_k_offset(version, key,
1514 			    le_key_k_offset(version, key) + offset);
1515 }
1516 
add_le_ih_k_offset(struct item_head * ih,loff_t offset)1517 static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset)
1518 {
1519 	add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1520 }
1521 
set_le_ih_k_offset(struct item_head * ih,loff_t offset)1522 static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1523 {
1524 	set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1525 }
1526 
set_le_key_k_type(int version,struct reiserfs_key * key,int type)1527 static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
1528 				     int type)
1529 {
1530 	if (version == KEY_FORMAT_3_5) {
1531 		type = type2uniqueness(type);
1532 		key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type);
1533 	} else
1534 	       set_offset_v2_k_type(&key->u.k_offset_v2, type);
1535 }
1536 
set_le_ih_k_type(struct item_head * ih,int type)1537 static inline void set_le_ih_k_type(struct item_head *ih, int type)
1538 {
1539 	set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1540 }
1541 
is_direntry_le_key(int version,struct reiserfs_key * key)1542 static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
1543 {
1544 	return le_key_k_type(version, key) == TYPE_DIRENTRY;
1545 }
1546 
is_direct_le_key(int version,struct reiserfs_key * key)1547 static inline int is_direct_le_key(int version, struct reiserfs_key *key)
1548 {
1549 	return le_key_k_type(version, key) == TYPE_DIRECT;
1550 }
1551 
is_indirect_le_key(int version,struct reiserfs_key * key)1552 static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
1553 {
1554 	return le_key_k_type(version, key) == TYPE_INDIRECT;
1555 }
1556 
is_statdata_le_key(int version,struct reiserfs_key * key)1557 static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
1558 {
1559 	return le_key_k_type(version, key) == TYPE_STAT_DATA;
1560 }
1561 
1562 /* item header has version.  */
is_direntry_le_ih(struct item_head * ih)1563 static inline int is_direntry_le_ih(struct item_head *ih)
1564 {
1565 	return is_direntry_le_key(ih_version(ih), &ih->ih_key);
1566 }
1567 
is_direct_le_ih(struct item_head * ih)1568 static inline int is_direct_le_ih(struct item_head *ih)
1569 {
1570 	return is_direct_le_key(ih_version(ih), &ih->ih_key);
1571 }
1572 
is_indirect_le_ih(struct item_head * ih)1573 static inline int is_indirect_le_ih(struct item_head *ih)
1574 {
1575 	return is_indirect_le_key(ih_version(ih), &ih->ih_key);
1576 }
1577 
is_statdata_le_ih(struct item_head * ih)1578 static inline int is_statdata_le_ih(struct item_head *ih)
1579 {
1580 	return is_statdata_le_key(ih_version(ih), &ih->ih_key);
1581 }
1582 
1583 /* key is pointer to cpu key, result is cpu */
cpu_key_k_offset(const struct cpu_key * key)1584 static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1585 {
1586 	return key->on_disk_key.k_offset;
1587 }
1588 
cpu_key_k_type(const struct cpu_key * key)1589 static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1590 {
1591 	return key->on_disk_key.k_type;
1592 }
1593 
set_cpu_key_k_offset(struct cpu_key * key,loff_t offset)1594 static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1595 {
1596 	key->on_disk_key.k_offset = offset;
1597 }
1598 
set_cpu_key_k_type(struct cpu_key * key,int type)1599 static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1600 {
1601 	key->on_disk_key.k_type = type;
1602 }
1603 
cpu_key_k_offset_dec(struct cpu_key * key)1604 static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1605 {
1606 	key->on_disk_key.k_offset--;
1607 }
1608 
1609 #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
1610 #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
1611 #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
1612 #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
1613 
1614 /* are these used ? */
1615 #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
1616 #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
1617 #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
1618 #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
1619 
1620 #define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
1621     (!COMP_SHORT_KEYS(ih, key) && \
1622 	  I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1623 
1624 /* maximal length of item */
1625 #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
1626 #define MIN_ITEM_LEN 1
1627 
1628 /* object identifier for root dir */
1629 #define REISERFS_ROOT_OBJECTID 2
1630 #define REISERFS_ROOT_PARENT_OBJECTID 1
1631 
1632 extern struct reiserfs_key root_key;
1633 
1634 /*
1635  * Picture represents a leaf of the S+tree
1636  *  ______________________________________________________
1637  * |      |  Array of     |                   |           |
1638  * |Block |  Object-Item  |      F r e e      |  Objects- |
1639  * | head |  Headers      |     S p a c e     |   Items   |
1640  * |______|_______________|___________________|___________|
1641  */
1642 
1643 /*
1644  * Header of a disk block.  More precisely, header of a formatted leaf
1645  * or internal node, and not the header of an unformatted node.
1646  */
1647 struct block_head {
1648 	__le16 blk_level;	/* Level of a block in the tree. */
1649 	__le16 blk_nr_item;	/* Number of keys/items in a block. */
1650 	__le16 blk_free_space;	/* Block free space in bytes. */
1651 	__le16 blk_reserved;
1652 	/* dump this in v4/planA */
1653 
1654 	/* kept only for compatibility */
1655 	struct reiserfs_key blk_right_delim_key;
1656 };
1657 
1658 #define BLKH_SIZE                     (sizeof(struct block_head))
1659 #define blkh_level(p_blkh)            (le16_to_cpu((p_blkh)->blk_level))
1660 #define blkh_nr_item(p_blkh)          (le16_to_cpu((p_blkh)->blk_nr_item))
1661 #define blkh_free_space(p_blkh)       (le16_to_cpu((p_blkh)->blk_free_space))
1662 #define blkh_reserved(p_blkh)         (le16_to_cpu((p_blkh)->blk_reserved))
1663 #define set_blkh_level(p_blkh,val)    ((p_blkh)->blk_level = cpu_to_le16(val))
1664 #define set_blkh_nr_item(p_blkh,val)  ((p_blkh)->blk_nr_item = cpu_to_le16(val))
1665 #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
1666 #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
1667 #define blkh_right_delim_key(p_blkh)  ((p_blkh)->blk_right_delim_key)
1668 #define set_blkh_right_delim_key(p_blkh,val)  ((p_blkh)->blk_right_delim_key = val)
1669 
1670 /* values for blk_level field of the struct block_head */
1671 
1672 /*
1673  * When node gets removed from the tree its blk_level is set to FREE_LEVEL.
1674  * It is then  used to see whether the node is still in the tree
1675  */
1676 #define FREE_LEVEL 0
1677 
1678 #define DISK_LEAF_NODE_LEVEL  1	/* Leaf node level. */
1679 
1680 /*
1681  * Given the buffer head of a formatted node, resolve to the
1682  * block head of that node.
1683  */
1684 #define B_BLK_HEAD(bh)			((struct block_head *)((bh)->b_data))
1685 /* Number of items that are in buffer. */
1686 #define B_NR_ITEMS(bh)			(blkh_nr_item(B_BLK_HEAD(bh)))
1687 #define B_LEVEL(bh)			(blkh_level(B_BLK_HEAD(bh)))
1688 #define B_FREE_SPACE(bh)		(blkh_free_space(B_BLK_HEAD(bh)))
1689 
1690 #define PUT_B_NR_ITEMS(bh, val)		do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
1691 #define PUT_B_LEVEL(bh, val)		do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
1692 #define PUT_B_FREE_SPACE(bh, val)	do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1693 
1694 /* Get right delimiting key. -- little endian */
1695 #define B_PRIGHT_DELIM_KEY(bh)		(&(blk_right_delim_key(B_BLK_HEAD(bh))))
1696 
1697 /* Does the buffer contain a disk leaf. */
1698 #define B_IS_ITEMS_LEVEL(bh)		(B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1699 
1700 /* Does the buffer contain a disk internal node */
1701 #define B_IS_KEYS_LEVEL(bh)      (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
1702 					    && B_LEVEL(bh) <= MAX_HEIGHT)
1703 
1704 /***************************************************************************
1705  *                             STAT DATA                                   *
1706  ***************************************************************************/
1707 
1708 /*
1709  * old stat data is 32 bytes long. We are going to distinguish new one by
1710  * different size
1711 */
1712 struct stat_data_v1 {
1713 	__le16 sd_mode;		/* file type, permissions */
1714 	__le16 sd_nlink;	/* number of hard links */
1715 	__le16 sd_uid;		/* owner */
1716 	__le16 sd_gid;		/* group */
1717 	__le32 sd_size;		/* file size */
1718 	__le32 sd_atime;	/* time of last access */
1719 	__le32 sd_mtime;	/* time file was last modified  */
1720 
1721 	/*
1722 	 * time inode (stat data) was last changed
1723 	 * (except changes to sd_atime and sd_mtime)
1724 	 */
1725 	__le32 sd_ctime;
1726 	union {
1727 		__le32 sd_rdev;
1728 		__le32 sd_blocks;	/* number of blocks file uses */
1729 	} __attribute__ ((__packed__)) u;
1730 
1731 	/*
1732 	 * first byte of file which is stored in a direct item: except that if
1733 	 * it equals 1 it is a symlink and if it equals ~(__u32)0 there is no
1734 	 * direct item.  The existence of this field really grates on me.
1735 	 * Let's replace it with a macro based on sd_size and our tail
1736 	 * suppression policy.  Someday.  -Hans
1737 	 */
1738 	__le32 sd_first_direct_byte;
1739 } __attribute__ ((__packed__));
1740 
1741 #define SD_V1_SIZE              (sizeof(struct stat_data_v1))
1742 #define stat_data_v1(ih)        (ih_version (ih) == KEY_FORMAT_3_5)
1743 #define sd_v1_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))
1744 #define set_sd_v1_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))
1745 #define sd_v1_nlink(sdp)        (le16_to_cpu((sdp)->sd_nlink))
1746 #define set_sd_v1_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le16(v))
1747 #define sd_v1_uid(sdp)          (le16_to_cpu((sdp)->sd_uid))
1748 #define set_sd_v1_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le16(v))
1749 #define sd_v1_gid(sdp)          (le16_to_cpu((sdp)->sd_gid))
1750 #define set_sd_v1_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le16(v))
1751 #define sd_v1_size(sdp)         (le32_to_cpu((sdp)->sd_size))
1752 #define set_sd_v1_size(sdp,v)   ((sdp)->sd_size = cpu_to_le32(v))
1753 #define sd_v1_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))
1754 #define set_sd_v1_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))
1755 #define sd_v1_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))
1756 #define set_sd_v1_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))
1757 #define sd_v1_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))
1758 #define set_sd_v1_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))
1759 #define sd_v1_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))
1760 #define set_sd_v1_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))
1761 #define sd_v1_blocks(sdp)       (le32_to_cpu((sdp)->u.sd_blocks))
1762 #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
1763 #define sd_v1_first_direct_byte(sdp) \
1764                                 (le32_to_cpu((sdp)->sd_first_direct_byte))
1765 #define set_sd_v1_first_direct_byte(sdp,v) \
1766                                 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
1767 
1768 /* inode flags stored in sd_attrs (nee sd_reserved) */
1769 
1770 /*
1771  * we want common flags to have the same values as in ext2,
1772  * so chattr(1) will work without problems
1773  */
1774 #define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
1775 #define REISERFS_APPEND_FL    FS_APPEND_FL
1776 #define REISERFS_SYNC_FL      FS_SYNC_FL
1777 #define REISERFS_NOATIME_FL   FS_NOATIME_FL
1778 #define REISERFS_NODUMP_FL    FS_NODUMP_FL
1779 #define REISERFS_SECRM_FL     FS_SECRM_FL
1780 #define REISERFS_UNRM_FL      FS_UNRM_FL
1781 #define REISERFS_COMPR_FL     FS_COMPR_FL
1782 #define REISERFS_NOTAIL_FL    FS_NOTAIL_FL
1783 
1784 /* persistent flags that file inherits from the parent directory */
1785 #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL |	\
1786 				REISERFS_SYNC_FL |	\
1787 				REISERFS_NOATIME_FL |	\
1788 				REISERFS_NODUMP_FL |	\
1789 				REISERFS_SECRM_FL |	\
1790 				REISERFS_COMPR_FL |	\
1791 				REISERFS_NOTAIL_FL )
1792 
1793 /*
1794  * Stat Data on disk (reiserfs version of UFS disk inode minus the
1795  * address blocks)
1796  */
1797 struct stat_data {
1798 	__le16 sd_mode;		/* file type, permissions */
1799 	__le16 sd_attrs;	/* persistent inode flags */
1800 	__le32 sd_nlink;	/* number of hard links */
1801 	__le64 sd_size;		/* file size */
1802 	__le32 sd_uid;		/* owner */
1803 	__le32 sd_gid;		/* group */
1804 	__le32 sd_atime;	/* time of last access */
1805 	__le32 sd_mtime;	/* time file was last modified  */
1806 
1807 	/*
1808 	 * time inode (stat data) was last changed
1809 	 * (except changes to sd_atime and sd_mtime)
1810 	 */
1811 	__le32 sd_ctime;
1812 	__le32 sd_blocks;
1813 	union {
1814 		__le32 sd_rdev;
1815 		__le32 sd_generation;
1816 	} __attribute__ ((__packed__)) u;
1817 } __attribute__ ((__packed__));
1818 
1819 /* this is 44 bytes long */
1820 #define SD_SIZE (sizeof(struct stat_data))
1821 #define SD_V2_SIZE              SD_SIZE
1822 #define stat_data_v2(ih)        (ih_version (ih) == KEY_FORMAT_3_6)
1823 #define sd_v2_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))
1824 #define set_sd_v2_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))
1825 /* sd_reserved */
1826 /* set_sd_reserved */
1827 #define sd_v2_nlink(sdp)        (le32_to_cpu((sdp)->sd_nlink))
1828 #define set_sd_v2_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le32(v))
1829 #define sd_v2_size(sdp)         (le64_to_cpu((sdp)->sd_size))
1830 #define set_sd_v2_size(sdp,v)   ((sdp)->sd_size = cpu_to_le64(v))
1831 #define sd_v2_uid(sdp)          (le32_to_cpu((sdp)->sd_uid))
1832 #define set_sd_v2_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le32(v))
1833 #define sd_v2_gid(sdp)          (le32_to_cpu((sdp)->sd_gid))
1834 #define set_sd_v2_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le32(v))
1835 #define sd_v2_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))
1836 #define set_sd_v2_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))
1837 #define sd_v2_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))
1838 #define set_sd_v2_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))
1839 #define sd_v2_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))
1840 #define set_sd_v2_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))
1841 #define sd_v2_blocks(sdp)       (le32_to_cpu((sdp)->sd_blocks))
1842 #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1843 #define sd_v2_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))
1844 #define set_sd_v2_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))
1845 #define sd_v2_generation(sdp)   (le32_to_cpu((sdp)->u.sd_generation))
1846 #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1847 #define sd_v2_attrs(sdp)         (le16_to_cpu((sdp)->sd_attrs))
1848 #define set_sd_v2_attrs(sdp,v)   ((sdp)->sd_attrs = cpu_to_le16(v))
1849 
1850 /***************************************************************************
1851  *                      DIRECTORY STRUCTURE                                *
1852  ***************************************************************************/
1853 /*
1854  * Picture represents the structure of directory items
1855  * ________________________________________________
1856  * |  Array of     |   |     |        |       |   |
1857  * | directory     |N-1| N-2 | ....   |   1st |0th|
1858  * | entry headers |   |     |        |       |   |
1859  * |_______________|___|_____|________|_______|___|
1860  *                  <----   directory entries         ------>
1861  *
1862  * First directory item has k_offset component 1. We store "." and ".."
1863  * in one item, always, we never split "." and ".." into differing
1864  * items.  This makes, among other things, the code for removing
1865  * directories simpler.
1866  */
1867 #define SD_OFFSET  0
1868 #define SD_UNIQUENESS 0
1869 #define DOT_OFFSET 1
1870 #define DOT_DOT_OFFSET 2
1871 #define DIRENTRY_UNIQUENESS 500
1872 
1873 #define FIRST_ITEM_OFFSET 1
1874 
1875 /*
1876  * Q: How to get key of object pointed to by entry from entry?
1877  *
1878  * A: Each directory entry has its header. This header has deh_dir_id
1879  *    and deh_objectid fields, those are key of object, entry points to
1880  */
1881 
1882 /*
1883  * NOT IMPLEMENTED:
1884  * Directory will someday contain stat data of object
1885  */
1886 
1887 struct reiserfs_de_head {
1888 	__le32 deh_offset;	/* third component of the directory entry key */
1889 
1890 	/*
1891 	 * objectid of the parent directory of the object, that is referenced
1892 	 * by directory entry
1893 	 */
1894 	__le32 deh_dir_id;
1895 
1896 	/* objectid of the object, that is referenced by directory entry */
1897 	__le32 deh_objectid;
1898 	__le16 deh_location;	/* offset of name in the whole item */
1899 
1900 	/*
1901 	 * whether 1) entry contains stat data (for future), and
1902 	 * 2) whether entry is hidden (unlinked)
1903 	 */
1904 	__le16 deh_state;
1905 } __attribute__ ((__packed__));
1906 #define DEH_SIZE                  sizeof(struct reiserfs_de_head)
1907 #define deh_offset(p_deh)         (le32_to_cpu((p_deh)->deh_offset))
1908 #define deh_dir_id(p_deh)         (le32_to_cpu((p_deh)->deh_dir_id))
1909 #define deh_objectid(p_deh)       (le32_to_cpu((p_deh)->deh_objectid))
1910 #define deh_location(p_deh)       (le16_to_cpu((p_deh)->deh_location))
1911 #define deh_state(p_deh)          (le16_to_cpu((p_deh)->deh_state))
1912 
1913 #define put_deh_offset(p_deh,v)   ((p_deh)->deh_offset = cpu_to_le32((v)))
1914 #define put_deh_dir_id(p_deh,v)   ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1915 #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1916 #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1917 #define put_deh_state(p_deh,v)    ((p_deh)->deh_state = cpu_to_le16((v)))
1918 
1919 /* empty directory contains two entries "." and ".." and their headers */
1920 #define EMPTY_DIR_SIZE \
1921 (DEH_SIZE * 2 + ROUND_UP (sizeof(".") - 1) + ROUND_UP (sizeof("..") - 1))
1922 
1923 /* old format directories have this size when empty */
1924 #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1925 
1926 #define DEH_Statdata 0		/* not used now */
1927 #define DEH_Visible 2
1928 
1929 /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1930 #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1931 #   define ADDR_UNALIGNED_BITS  (3)
1932 #endif
1933 
1934 /*
1935  * These are only used to manipulate deh_state.
1936  * Because of this, we'll use the ext2_ bit routines,
1937  * since they are little endian
1938  */
1939 #ifdef ADDR_UNALIGNED_BITS
1940 
1941 #   define aligned_address(addr)           ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1942 #   define unaligned_offset(addr)          (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1943 
1944 #   define set_bit_unaligned(nr, addr)	\
1945 	__test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1946 #   define clear_bit_unaligned(nr, addr)	\
1947 	__test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1948 #   define test_bit_unaligned(nr, addr)	\
1949 	test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1950 
1951 #else
1952 
1953 #   define set_bit_unaligned(nr, addr)	__test_and_set_bit_le(nr, addr)
1954 #   define clear_bit_unaligned(nr, addr)	__test_and_clear_bit_le(nr, addr)
1955 #   define test_bit_unaligned(nr, addr)	test_bit_le(nr, addr)
1956 
1957 #endif
1958 
1959 #define mark_de_with_sd(deh)        set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1960 #define mark_de_without_sd(deh)     clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1961 #define mark_de_visible(deh)	    set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1962 #define mark_de_hidden(deh)	    clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1963 
1964 #define de_with_sd(deh)		    test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1965 #define de_visible(deh)	    	    test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1966 #define de_hidden(deh)	    	    !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1967 
1968 extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1969 				   __le32 par_dirid, __le32 par_objid);
1970 extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1971 				__le32 par_dirid, __le32 par_objid);
1972 
1973 /* two entries per block (at least) */
1974 #define REISERFS_MAX_NAME(block_size) 255
1975 
1976 /*
1977  * this structure is used for operations on directory entries. It is
1978  * not a disk structure.
1979  *
1980  * When reiserfs_find_entry or search_by_entry_key find directory
1981  * entry, they return filled reiserfs_dir_entry structure
1982  */
1983 struct reiserfs_dir_entry {
1984 	struct buffer_head *de_bh;
1985 	int de_item_num;
1986 	struct item_head *de_ih;
1987 	int de_entry_num;
1988 	struct reiserfs_de_head *de_deh;
1989 	int de_entrylen;
1990 	int de_namelen;
1991 	char *de_name;
1992 	unsigned long *de_gen_number_bit_string;
1993 
1994 	__u32 de_dir_id;
1995 	__u32 de_objectid;
1996 
1997 	struct cpu_key de_entry_key;
1998 };
1999 
2000 /*
2001  * these defines are useful when a particular member of
2002  * a reiserfs_dir_entry is needed
2003  */
2004 
2005 /* pointer to file name, stored in entry */
2006 #define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \
2007 				(ih_item_body(bh, ih) + deh_location(deh))
2008 
2009 /* length of name */
2010 #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
2011 (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
2012 
2013 /* hash value occupies bits from 7 up to 30 */
2014 #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
2015 /* generation number occupies 7 bits starting from 0 up to 6 */
2016 #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
2017 #define MAX_GENERATION_NUMBER  127
2018 
2019 #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
2020 
2021 /*
2022  * Picture represents an internal node of the reiserfs tree
2023  *  ______________________________________________________
2024  * |      |  Array of     |  Array of         |  Free     |
2025  * |block |    keys       |  pointers         | space     |
2026  * | head |      N        |      N+1          |           |
2027  * |______|_______________|___________________|___________|
2028  */
2029 
2030 /***************************************************************************
2031  *                      DISK CHILD                                         *
2032  ***************************************************************************/
2033 /*
2034  * Disk child pointer:
2035  * The pointer from an internal node of the tree to a node that is on disk.
2036  */
2037 struct disk_child {
2038 	__le32 dc_block_number;	/* Disk child's block number. */
2039 	__le16 dc_size;		/* Disk child's used space.   */
2040 	__le16 dc_reserved;
2041 };
2042 
2043 #define DC_SIZE (sizeof(struct disk_child))
2044 #define dc_block_number(dc_p)	(le32_to_cpu((dc_p)->dc_block_number))
2045 #define dc_size(dc_p)		(le16_to_cpu((dc_p)->dc_size))
2046 #define put_dc_block_number(dc_p, val)   do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
2047 #define put_dc_size(dc_p, val)   do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
2048 
2049 /* Get disk child by buffer header and position in the tree node. */
2050 #define B_N_CHILD(bh, n_pos)  ((struct disk_child *)\
2051 ((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
2052 
2053 /* Get disk child number by buffer header and position in the tree node. */
2054 #define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
2055 #define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
2056 				(put_dc_block_number(B_N_CHILD(bh, n_pos), val))
2057 
2058  /* maximal value of field child_size in structure disk_child */
2059  /* child size is the combined size of all items and their headers */
2060 #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
2061 
2062 /* amount of used space in buffer (not including block head) */
2063 #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
2064 
2065 /* max and min number of keys in internal node */
2066 #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
2067 #define MIN_NR_KEY(bh)    (MAX_NR_KEY(bh)/2)
2068 
2069 /***************************************************************************
2070  *                      PATH STRUCTURES AND DEFINES                        *
2071  ***************************************************************************/
2072 
2073 /*
2074  * search_by_key fills up the path from the root to the leaf as it descends
2075  * the tree looking for the key.  It uses reiserfs_bread to try to find
2076  * buffers in the cache given their block number.  If it does not find
2077  * them in the cache it reads them from disk.  For each node search_by_key
2078  * finds using reiserfs_bread it then uses bin_search to look through that
2079  * node.  bin_search will find the position of the block_number of the next
2080  * node if it is looking through an internal node.  If it is looking through
2081  * a leaf node bin_search will find the position of the item which has key
2082  * either equal to given key, or which is the maximal key less than the
2083  * given key.
2084  */
2085 
2086 struct path_element {
2087 	/* Pointer to the buffer at the path in the tree. */
2088 	struct buffer_head *pe_buffer;
2089 	/* Position in the tree node which is placed in the buffer above. */
2090 	int pe_position;
2091 };
2092 
2093 /*
2094  * maximal height of a tree. don't change this without
2095  * changing JOURNAL_PER_BALANCE_CNT
2096  */
2097 #define MAX_HEIGHT 5
2098 
2099 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
2100 #define EXTENDED_MAX_HEIGHT         7
2101 
2102 /* Must be equal to at least 2. */
2103 #define FIRST_PATH_ELEMENT_OFFSET   2
2104 
2105 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
2106 #define ILLEGAL_PATH_ELEMENT_OFFSET 1
2107 
2108 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
2109 #define MAX_FEB_SIZE 6
2110 
2111 /*
2112  * We need to keep track of who the ancestors of nodes are.  When we
2113  * perform a search we record which nodes were visited while
2114  * descending the tree looking for the node we searched for. This list
2115  * of nodes is called the path.  This information is used while
2116  * performing balancing.  Note that this path information may become
2117  * invalid, and this means we must check it when using it to see if it
2118  * is still valid. You'll need to read search_by_key and the comments
2119  * in it, especially about decrement_counters_in_path(), to understand
2120  * this structure.
2121  *
2122  * Paths make the code so much harder to work with and debug.... An
2123  * enormous number of bugs are due to them, and trying to write or modify
2124  * code that uses them just makes my head hurt.  They are based on an
2125  * excessive effort to avoid disturbing the precious VFS code.:-( The
2126  * gods only know how we are going to SMP the code that uses them.
2127  * znodes are the way!
2128  */
2129 
2130 #define PATH_READA	0x1	/* do read ahead */
2131 #define PATH_READA_BACK 0x2	/* read backwards */
2132 
2133 struct treepath {
2134 	int path_length;	/* Length of the array above.   */
2135 	int reada;
2136 	/* Array of the path elements.  */
2137 	struct path_element path_elements[EXTENDED_MAX_HEIGHT];
2138 	int pos_in_item;
2139 };
2140 
2141 #define pos_in_item(path) ((path)->pos_in_item)
2142 
2143 #define INITIALIZE_PATH(var) \
2144 struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
2145 
2146 /* Get path element by path and path position. */
2147 #define PATH_OFFSET_PELEMENT(path, n_offset)  ((path)->path_elements + (n_offset))
2148 
2149 /* Get buffer header at the path by path and path position. */
2150 #define PATH_OFFSET_PBUFFER(path, n_offset)   (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
2151 
2152 /* Get position in the element at the path by path and path position. */
2153 #define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
2154 
2155 #define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
2156 
2157 /*
2158  * you know, to the person who didn't write this the macro name does not
2159  * at first suggest what it does.  Maybe POSITION_FROM_PATH_END? Or
2160  * maybe we should just focus on dumping paths... -Hans
2161  */
2162 #define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
2163 
2164 /*
2165  * in do_balance leaf has h == 0 in contrast with path structure,
2166  * where root has level == 0. That is why we need these defines
2167  */
2168 
2169 /* tb->S[h] */
2170 #define PATH_H_PBUFFER(path, h) \
2171 			PATH_OFFSET_PBUFFER(path, path->path_length - (h))
2172 
2173 /* tb->F[h] or tb->S[0]->b_parent */
2174 #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER(path, (h) + 1)
2175 
2176 #define PATH_H_POSITION(path, h) \
2177 			PATH_OFFSET_POSITION(path, path->path_length - (h))
2178 
2179 /* tb->S[h]->b_item_order */
2180 #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)
2181 
2182 #define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
2183 
reiserfs_node_data(const struct buffer_head * bh)2184 static inline void *reiserfs_node_data(const struct buffer_head *bh)
2185 {
2186 	return bh->b_data + sizeof(struct block_head);
2187 }
2188 
2189 /* get key from internal node */
internal_key(struct buffer_head * bh,int item_num)2190 static inline struct reiserfs_key *internal_key(struct buffer_head *bh,
2191 						int item_num)
2192 {
2193 	struct reiserfs_key *key = reiserfs_node_data(bh);
2194 
2195 	return &key[item_num];
2196 }
2197 
2198 /* get the item header from leaf node */
item_head(const struct buffer_head * bh,int item_num)2199 static inline struct item_head *item_head(const struct buffer_head *bh,
2200 					  int item_num)
2201 {
2202 	struct item_head *ih = reiserfs_node_data(bh);
2203 
2204 	return &ih[item_num];
2205 }
2206 
2207 /* get the key from leaf node */
leaf_key(const struct buffer_head * bh,int item_num)2208 static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh,
2209 					    int item_num)
2210 {
2211 	return &item_head(bh, item_num)->ih_key;
2212 }
2213 
ih_item_body(const struct buffer_head * bh,const struct item_head * ih)2214 static inline void *ih_item_body(const struct buffer_head *bh,
2215 				 const struct item_head *ih)
2216 {
2217 	return bh->b_data + ih_location(ih);
2218 }
2219 
2220 /* get item body from leaf node */
item_body(const struct buffer_head * bh,int item_num)2221 static inline void *item_body(const struct buffer_head *bh, int item_num)
2222 {
2223 	return ih_item_body(bh, item_head(bh, item_num));
2224 }
2225 
tp_item_head(const struct treepath * path)2226 static inline struct item_head *tp_item_head(const struct treepath *path)
2227 {
2228 	return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2229 }
2230 
tp_item_body(const struct treepath * path)2231 static inline void *tp_item_body(const struct treepath *path)
2232 {
2233 	return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));
2234 }
2235 
2236 #define get_last_bh(path) PATH_PLAST_BUFFER(path)
2237 #define get_item_pos(path) PATH_LAST_POSITION(path)
2238 #define item_moved(ih,path) comp_items(ih, path)
2239 #define path_changed(ih,path) comp_items (ih, path)
2240 
2241 /* array of the entry headers */
2242  /* get item body */
2243 #define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih)))
2244 
2245 /*
2246  * length of the directory entry in directory item. This define
2247  * calculates length of i-th directory entry using directory entry
2248  * locations from dir entry head. When it calculates length of 0-th
2249  * directory entry, it uses length of whole item in place of entry
2250  * location of the non-existent following entry in the calculation.
2251  * See picture above.
2252  */
entry_length(const struct buffer_head * bh,const struct item_head * ih,int pos_in_item)2253 static inline int entry_length(const struct buffer_head *bh,
2254 			       const struct item_head *ih, int pos_in_item)
2255 {
2256 	struct reiserfs_de_head *deh;
2257 
2258 	deh = B_I_DEH(bh, ih) + pos_in_item;
2259 	if (pos_in_item)
2260 		return deh_location(deh - 1) - deh_location(deh);
2261 
2262 	return ih_item_len(ih) - deh_location(deh);
2263 }
2264 
2265 /***************************************************************************
2266  *                       MISC                                              *
2267  ***************************************************************************/
2268 
2269 /* Size of pointer to the unformatted node. */
2270 #define UNFM_P_SIZE (sizeof(unp_t))
2271 #define UNFM_P_SHIFT 2
2272 
2273 /* in in-core inode key is stored on le form */
2274 #define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
2275 
2276 #define MAX_UL_INT 0xffffffff
2277 #define MAX_INT    0x7ffffff
2278 #define MAX_US_INT 0xffff
2279 
2280 // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
max_reiserfs_offset(struct inode * inode)2281 static inline loff_t max_reiserfs_offset(struct inode *inode)
2282 {
2283 	if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
2284 		return (loff_t) U32_MAX;
2285 
2286 	return (loff_t) ((~(__u64) 0) >> 4);
2287 }
2288 
2289 #define MAX_KEY_OBJECTID	MAX_UL_INT
2290 
2291 #define MAX_B_NUM  MAX_UL_INT
2292 #define MAX_FC_NUM MAX_US_INT
2293 
2294 /* the purpose is to detect overflow of an unsigned short */
2295 #define REISERFS_LINK_MAX (MAX_US_INT - 1000)
2296 
2297 /*
2298  * The following defines are used in reiserfs_insert_item
2299  * and reiserfs_append_item
2300  */
2301 #define REISERFS_KERNEL_MEM		0	/* kernel memory mode */
2302 #define REISERFS_USER_MEM		1	/* user memory mode */
2303 
2304 #define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
2305 #define get_generation(s) atomic_read (&fs_generation(s))
2306 #define FILESYSTEM_CHANGED_TB(tb)  (get_generation((tb)->tb_sb) != (tb)->fs_gen)
2307 #define __fs_changed(gen,s) (gen != get_generation (s))
2308 #define fs_changed(gen,s)		\
2309 ({					\
2310 	reiserfs_cond_resched(s);	\
2311 	__fs_changed(gen, s);		\
2312 })
2313 
2314 /***************************************************************************
2315  *                  FIXATE NODES                                           *
2316  ***************************************************************************/
2317 
2318 #define VI_TYPE_LEFT_MERGEABLE 1
2319 #define VI_TYPE_RIGHT_MERGEABLE 2
2320 
2321 /*
2322  * To make any changes in the tree we always first find node, that
2323  * contains item to be changed/deleted or place to insert a new
2324  * item. We call this node S. To do balancing we need to decide what
2325  * we will shift to left/right neighbor, or to a new node, where new
2326  * item will be etc. To make this analysis simpler we build virtual
2327  * node. Virtual node is an array of items, that will replace items of
2328  * node S. (For instance if we are going to delete an item, virtual
2329  * node does not contain it). Virtual node keeps information about
2330  * item sizes and types, mergeability of first and last items, sizes
2331  * of all entries in directory item. We use this array of items when
2332  * calculating what we can shift to neighbors and how many nodes we
2333  * have to have if we do not any shiftings, if we shift to left/right
2334  * neighbor or to both.
2335  */
2336 struct virtual_item {
2337 	int vi_index;		/* index in the array of item operations */
2338 	unsigned short vi_type;	/* left/right mergeability */
2339 
2340 	/* length of item that it will have after balancing */
2341 	unsigned short vi_item_len;
2342 
2343 	struct item_head *vi_ih;
2344 	const char *vi_item;	/* body of item (old or new) */
2345 	const void *vi_new_data;	/* 0 always but paste mode */
2346 	void *vi_uarea;		/* item specific area */
2347 };
2348 
2349 struct virtual_node {
2350 	/* this is a pointer to the free space in the buffer */
2351 	char *vn_free_ptr;
2352 
2353 	unsigned short vn_nr_item;	/* number of items in virtual node */
2354 
2355 	/*
2356 	 * size of node , that node would have if it has
2357 	 * unlimited size and no balancing is performed
2358 	 */
2359 	short vn_size;
2360 
2361 	/* mode of balancing (paste, insert, delete, cut) */
2362 	short vn_mode;
2363 
2364 	short vn_affected_item_num;
2365 	short vn_pos_in_item;
2366 
2367 	/* item header of inserted item, 0 for other modes */
2368 	struct item_head *vn_ins_ih;
2369 	const void *vn_data;
2370 
2371 	/* array of items (including a new one, excluding item to be deleted) */
2372 	struct virtual_item *vn_vi;
2373 };
2374 
2375 /* used by directory items when creating virtual nodes */
2376 struct direntry_uarea {
2377 	int flags;
2378 	__u16 entry_count;
2379 	__u16 entry_sizes[1];
2380 } __attribute__ ((__packed__));
2381 
2382 /***************************************************************************
2383  *                  TREE BALANCE                                           *
2384  ***************************************************************************/
2385 
2386 /*
2387  * This temporary structure is used in tree balance algorithms, and
2388  * constructed as we go to the extent that its various parts are
2389  * needed.  It contains arrays of nodes that can potentially be
2390  * involved in the balancing of node S, and parameters that define how
2391  * each of the nodes must be balanced.  Note that in these algorithms
2392  * for balancing the worst case is to need to balance the current node
2393  * S and the left and right neighbors and all of their parents plus
2394  * create a new node.  We implement S1 balancing for the leaf nodes
2395  * and S0 balancing for the internal nodes (S1 and S0 are defined in
2396  * our papers.)
2397  */
2398 
2399 /* size of the array of buffers to free at end of do_balance */
2400 #define MAX_FREE_BLOCK 7
2401 
2402 /* maximum number of FEB blocknrs on a single level */
2403 #define MAX_AMOUNT_NEEDED 2
2404 
2405 /* someday somebody will prefix every field in this struct with tb_ */
2406 struct tree_balance {
2407 	int tb_mode;
2408 	int need_balance_dirty;
2409 	struct super_block *tb_sb;
2410 	struct reiserfs_transaction_handle *transaction_handle;
2411 	struct treepath *tb_path;
2412 
2413 	/* array of left neighbors of nodes in the path */
2414 	struct buffer_head *L[MAX_HEIGHT];
2415 
2416 	/* array of right neighbors of nodes in the path */
2417 	struct buffer_head *R[MAX_HEIGHT];
2418 
2419 	/* array of fathers of the left neighbors */
2420 	struct buffer_head *FL[MAX_HEIGHT];
2421 
2422 	/* array of fathers of the right neighbors */
2423 	struct buffer_head *FR[MAX_HEIGHT];
2424 	/* array of common parents of center node and its left neighbor */
2425 	struct buffer_head *CFL[MAX_HEIGHT];
2426 
2427 	/* array of common parents of center node and its right neighbor */
2428 	struct buffer_head *CFR[MAX_HEIGHT];
2429 
2430 	/*
2431 	 * array of empty buffers. Number of buffers in array equals
2432 	 * cur_blknum.
2433 	 */
2434 	struct buffer_head *FEB[MAX_FEB_SIZE];
2435 	struct buffer_head *used[MAX_FEB_SIZE];
2436 	struct buffer_head *thrown[MAX_FEB_SIZE];
2437 
2438 	/*
2439 	 * array of number of items which must be shifted to the left in
2440 	 * order to balance the current node; for leaves includes item that
2441 	 * will be partially shifted; for internal nodes, it is the number
2442 	 * of child pointers rather than items. It includes the new item
2443 	 * being created. The code sometimes subtracts one to get the
2444 	 * number of wholly shifted items for other purposes.
2445 	 */
2446 	int lnum[MAX_HEIGHT];
2447 
2448 	/* substitute right for left in comment above */
2449 	int rnum[MAX_HEIGHT];
2450 
2451 	/*
2452 	 * array indexed by height h mapping the key delimiting L[h] and
2453 	 * S[h] to its item number within the node CFL[h]
2454 	 */
2455 	int lkey[MAX_HEIGHT];
2456 
2457 	/* substitute r for l in comment above */
2458 	int rkey[MAX_HEIGHT];
2459 
2460 	/*
2461 	 * the number of bytes by we are trying to add or remove from
2462 	 * S[h]. A negative value means removing.
2463 	 */
2464 	int insert_size[MAX_HEIGHT];
2465 
2466 	/*
2467 	 * number of nodes that will replace node S[h] after balancing
2468 	 * on the level h of the tree.  If 0 then S is being deleted,
2469 	 * if 1 then S is remaining and no new nodes are being created,
2470 	 * if 2 or 3 then 1 or 2 new nodes is being created
2471 	 */
2472 	int blknum[MAX_HEIGHT];
2473 
2474 	/* fields that are used only for balancing leaves of the tree */
2475 
2476 	/* number of empty blocks having been already allocated */
2477 	int cur_blknum;
2478 
2479 	/* number of items that fall into left most node when S[0] splits */
2480 	int s0num;
2481 
2482 	/*
2483 	 * number of bytes which can flow to the left neighbor from the left
2484 	 * most liquid item that cannot be shifted from S[0] entirely
2485 	 * if -1 then nothing will be partially shifted
2486 	 */
2487 	int lbytes;
2488 
2489 	/*
2490 	 * number of bytes which will flow to the right neighbor from the right
2491 	 * most liquid item that cannot be shifted from S[0] entirely
2492 	 * if -1 then nothing will be partially shifted
2493 	 */
2494 	int rbytes;
2495 
2496 
2497 	/*
2498 	 * index into the array of item headers in
2499 	 * S[0] of the affected item
2500 	 */
2501 	int item_pos;
2502 
2503 	/* new nodes allocated to hold what could not fit into S */
2504 	struct buffer_head *S_new[2];
2505 
2506 	/*
2507 	 * number of items that will be placed into nodes in S_new
2508 	 * when S[0] splits
2509 	 */
2510 	int snum[2];
2511 
2512 	/*
2513 	 * number of bytes which flow to nodes in S_new when S[0] splits
2514 	 * note: if S[0] splits into 3 nodes, then items do not need to be cut
2515 	 */
2516 	int sbytes[2];
2517 
2518 	int pos_in_item;
2519 	int zeroes_num;
2520 
2521 	/*
2522 	 * buffers which are to be freed after do_balance finishes
2523 	 * by unfix_nodes
2524 	 */
2525 	struct buffer_head *buf_to_free[MAX_FREE_BLOCK];
2526 
2527 	/*
2528 	 * kmalloced memory. Used to create virtual node and keep
2529 	 * map of dirtied bitmap blocks
2530 	 */
2531 	char *vn_buf;
2532 
2533 	int vn_buf_size;	/* size of the vn_buf */
2534 
2535 	/* VN starts after bitmap of bitmap blocks */
2536 	struct virtual_node *tb_vn;
2537 
2538 	/*
2539 	 * saved value of `reiserfs_generation' counter see
2540 	 * FILESYSTEM_CHANGED() macro in reiserfs_fs.h
2541 	 */
2542 	int fs_gen;
2543 
2544 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2545 	/*
2546 	 * key pointer, to pass to block allocator or
2547 	 * another low-level subsystem
2548 	 */
2549 	struct in_core_key key;
2550 #endif
2551 };
2552 
2553 /* These are modes of balancing */
2554 
2555 /* When inserting an item. */
2556 #define M_INSERT	'i'
2557 /*
2558  * When inserting into (directories only) or appending onto an already
2559  * existent item.
2560  */
2561 #define M_PASTE		'p'
2562 /* When deleting an item. */
2563 #define M_DELETE	'd'
2564 /* When truncating an item or removing an entry from a (directory) item. */
2565 #define M_CUT		'c'
2566 
2567 /* used when balancing on leaf level skipped (in reiserfsck) */
2568 #define M_INTERNAL	'n'
2569 
2570 /*
2571  * When further balancing is not needed, then do_balance does not need
2572  * to be called.
2573  */
2574 #define M_SKIP_BALANCING		's'
2575 #define M_CONVERT	'v'
2576 
2577 /* modes of leaf_move_items */
2578 #define LEAF_FROM_S_TO_L 0
2579 #define LEAF_FROM_S_TO_R 1
2580 #define LEAF_FROM_R_TO_L 2
2581 #define LEAF_FROM_L_TO_R 3
2582 #define LEAF_FROM_S_TO_SNEW 4
2583 
2584 #define FIRST_TO_LAST 0
2585 #define LAST_TO_FIRST 1
2586 
2587 /*
2588  * used in do_balance for passing parent of node information that has
2589  * been gotten from tb struct
2590  */
2591 struct buffer_info {
2592 	struct tree_balance *tb;
2593 	struct buffer_head *bi_bh;
2594 	struct buffer_head *bi_parent;
2595 	int bi_position;
2596 };
2597 
sb_from_tb(struct tree_balance * tb)2598 static inline struct super_block *sb_from_tb(struct tree_balance *tb)
2599 {
2600 	return tb ? tb->tb_sb : NULL;
2601 }
2602 
sb_from_bi(struct buffer_info * bi)2603 static inline struct super_block *sb_from_bi(struct buffer_info *bi)
2604 {
2605 	return bi ? sb_from_tb(bi->tb) : NULL;
2606 }
2607 
2608 /*
2609  * there are 4 types of items: stat data, directory item, indirect, direct.
2610  * +-------------------+------------+--------------+------------+
2611  * |                   |  k_offset  | k_uniqueness | mergeable? |
2612  * +-------------------+------------+--------------+------------+
2613  * |     stat data     |     0      |      0       |   no       |
2614  * +-------------------+------------+--------------+------------+
2615  * | 1st directory item| DOT_OFFSET | DIRENTRY_ .. |   no       |
2616  * | non 1st directory | hash value | UNIQUENESS   |   yes      |
2617  * |     item          |            |              |            |
2618  * +-------------------+------------+--------------+------------+
2619  * | indirect item     | offset + 1 |TYPE_INDIRECT |    [1]	|
2620  * +-------------------+------------+--------------+------------+
2621  * | direct item       | offset + 1 |TYPE_DIRECT   |    [2]     |
2622  * +-------------------+------------+--------------+------------+
2623  *
2624  * [1] if this is not the first indirect item of the object
2625  * [2] if this is not the first direct item of the object
2626 */
2627 
2628 struct item_operations {
2629 	int (*bytes_number) (struct item_head * ih, int block_size);
2630 	void (*decrement_key) (struct cpu_key *);
2631 	int (*is_left_mergeable) (struct reiserfs_key * ih,
2632 				  unsigned long bsize);
2633 	void (*print_item) (struct item_head *, char *item);
2634 	void (*check_item) (struct item_head *, char *item);
2635 
2636 	int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
2637 			  int is_affected, int insert_size);
2638 	int (*check_left) (struct virtual_item * vi, int free,
2639 			   int start_skip, int end_skip);
2640 	int (*check_right) (struct virtual_item * vi, int free);
2641 	int (*part_size) (struct virtual_item * vi, int from, int to);
2642 	int (*unit_num) (struct virtual_item * vi);
2643 	void (*print_vi) (struct virtual_item * vi);
2644 };
2645 
2646 extern struct item_operations *item_ops[TYPE_ANY + 1];
2647 
2648 #define op_bytes_number(ih,bsize)                    item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
2649 #define op_is_left_mergeable(key,bsize)              item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
2650 #define op_print_item(ih,item)                       item_ops[le_ih_k_type (ih)]->print_item (ih, item)
2651 #define op_check_item(ih,item)                       item_ops[le_ih_k_type (ih)]->check_item (ih, item)
2652 #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)
2653 #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
2654 #define op_check_right(vi,free)                      item_ops[(vi)->vi_index]->check_right (vi, free)
2655 #define op_part_size(vi,from,to)                     item_ops[(vi)->vi_index]->part_size (vi, from, to)
2656 #define op_unit_num(vi)				     item_ops[(vi)->vi_index]->unit_num (vi)
2657 #define op_print_vi(vi)                              item_ops[(vi)->vi_index]->print_vi (vi)
2658 
2659 #define COMP_SHORT_KEYS comp_short_keys
2660 
2661 /* number of blocks pointed to by the indirect item */
2662 #define I_UNFM_NUM(ih)	(ih_item_len(ih) / UNFM_P_SIZE)
2663 
2664 /*
2665  * the used space within the unformatted node corresponding
2666  * to pos within the item pointed to by ih
2667  */
2668 #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
2669 
2670 /*
2671  * number of bytes contained by the direct item or the
2672  * unformatted nodes the indirect item points to
2673  */
2674 
2675 /* following defines use reiserfs buffer header and item header */
2676 
2677 /* get stat-data */
2678 #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
2679 
2680 /* this is 3976 for size==4096 */
2681 #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
2682 
2683 /*
2684  * indirect items consist of entries which contain blocknrs, pos
2685  * indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
2686  * blocknr contained by the entry pos points to
2687  */
2688 #define B_I_POS_UNFM_POINTER(bh, ih, pos)				\
2689 	le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos)))
2690 #define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val)			\
2691 	(*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val))
2692 
2693 struct reiserfs_iget_args {
2694 	__u32 objectid;
2695 	__u32 dirid;
2696 };
2697 
2698 /***************************************************************************
2699  *                    FUNCTION DECLARATIONS                                *
2700  ***************************************************************************/
2701 
2702 #define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
2703 
2704 #define journal_trans_half(blocksize) \
2705 	((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
2706 
2707 /* journal.c see journal.c for all the comments here */
2708 
2709 /* first block written in a commit.  */
2710 struct reiserfs_journal_desc {
2711 	__le32 j_trans_id;	/* id of commit */
2712 
2713 	/* length of commit. len +1 is the commit block */
2714 	__le32 j_len;
2715 
2716 	__le32 j_mount_id;	/* mount id of this trans */
2717 	__le32 j_realblock[1];	/* real locations for each block */
2718 };
2719 
2720 #define get_desc_trans_id(d)   le32_to_cpu((d)->j_trans_id)
2721 #define get_desc_trans_len(d)  le32_to_cpu((d)->j_len)
2722 #define get_desc_mount_id(d)   le32_to_cpu((d)->j_mount_id)
2723 
2724 #define set_desc_trans_id(d,val)       do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
2725 #define set_desc_trans_len(d,val)      do { (d)->j_len = cpu_to_le32 (val); } while (0)
2726 #define set_desc_mount_id(d,val)       do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
2727 
2728 /* last block written in a commit */
2729 struct reiserfs_journal_commit {
2730 	__le32 j_trans_id;	/* must match j_trans_id from the desc block */
2731 	__le32 j_len;		/* ditto */
2732 	__le32 j_realblock[1];	/* real locations for each block */
2733 };
2734 
2735 #define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
2736 #define get_commit_trans_len(c)        le32_to_cpu((c)->j_len)
2737 #define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
2738 
2739 #define set_commit_trans_id(c,val)     do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
2740 #define set_commit_trans_len(c,val)    do { (c)->j_len = cpu_to_le32 (val); } while (0)
2741 
2742 /*
2743  * this header block gets written whenever a transaction is considered
2744  * fully flushed, and is more recent than the last fully flushed transaction.
2745  * fully flushed means all the log blocks and all the real blocks are on
2746  * disk, and this transaction does not need to be replayed.
2747  */
2748 struct reiserfs_journal_header {
2749 	/* id of last fully flushed transaction */
2750 	__le32 j_last_flush_trans_id;
2751 
2752 	/* offset in the log of where to start replay after a crash */
2753 	__le32 j_first_unflushed_offset;
2754 
2755 	__le32 j_mount_id;
2756 	/* 12 */ struct journal_params jh_journal;
2757 };
2758 
2759 /* biggest tunable defines are right here */
2760 #define JOURNAL_BLOCK_COUNT 8192	/* number of blocks in the journal */
2761 
2762 /* biggest possible single transaction, don't change for now (8/3/99) */
2763 #define JOURNAL_TRANS_MAX_DEFAULT 1024
2764 #define JOURNAL_TRANS_MIN_DEFAULT 256
2765 
2766 /*
2767  * max blocks to batch into one transaction,
2768  * don't make this any bigger than 900
2769  */
2770 #define JOURNAL_MAX_BATCH_DEFAULT   900
2771 #define JOURNAL_MIN_RATIO 2
2772 #define JOURNAL_MAX_COMMIT_AGE 30
2773 #define JOURNAL_MAX_TRANS_AGE 30
2774 #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
2775 #define JOURNAL_BLOCKS_PER_OBJECT(sb)  (JOURNAL_PER_BALANCE_CNT * 3 + \
2776 					 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
2777 					      REISERFS_QUOTA_TRANS_BLOCKS(sb)))
2778 
2779 #ifdef CONFIG_QUOTA
2780 #define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
2781 /* We need to update data and inode (atime) */
2782 #define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
2783 /* 1 balancing, 1 bitmap, 1 data per write + stat data update */
2784 #define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2785 (DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
2786 /* same as with INIT */
2787 #define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
2788 (DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
2789 #else
2790 #define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
2791 #define REISERFS_QUOTA_INIT_BLOCKS(s) 0
2792 #define REISERFS_QUOTA_DEL_BLOCKS(s) 0
2793 #endif
2794 
2795 /*
2796  * both of these can be as low as 1, or as high as you want.  The min is the
2797  * number of 4k bitmap nodes preallocated on mount. New nodes are allocated
2798  * as needed, and released when transactions are committed.  On release, if
2799  * the current number of nodes is > max, the node is freed, otherwise,
2800  * it is put on a free list for faster use later.
2801 */
2802 #define REISERFS_MIN_BITMAP_NODES 10
2803 #define REISERFS_MAX_BITMAP_NODES 100
2804 
2805 /* these are based on journal hash size of 8192 */
2806 #define JBH_HASH_SHIFT 13
2807 #define JBH_HASH_MASK 8191
2808 
2809 #define _jhashfn(sb,block)	\
2810 	(((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
2811 	 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
2812 #define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
2813 
2814 /* We need these to make journal.c code more readable */
2815 #define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2816 #define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2817 #define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
2818 
2819 enum reiserfs_bh_state_bits {
2820 	BH_JDirty = BH_PrivateStart,	/* buffer is in current transaction */
2821 	BH_JDirty_wait,
2822 	/*
2823 	 * disk block was taken off free list before being in a
2824 	 * finished transaction, or written to disk. Can be reused immed.
2825 	 */
2826 	BH_JNew,
2827 	BH_JPrepared,
2828 	BH_JRestore_dirty,
2829 	BH_JTest,		/* debugging only will go away */
2830 };
2831 
2832 BUFFER_FNS(JDirty, journaled);
2833 TAS_BUFFER_FNS(JDirty, journaled);
2834 BUFFER_FNS(JDirty_wait, journal_dirty);
2835 TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
2836 BUFFER_FNS(JNew, journal_new);
2837 TAS_BUFFER_FNS(JNew, journal_new);
2838 BUFFER_FNS(JPrepared, journal_prepared);
2839 TAS_BUFFER_FNS(JPrepared, journal_prepared);
2840 BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2841 TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
2842 BUFFER_FNS(JTest, journal_test);
2843 TAS_BUFFER_FNS(JTest, journal_test);
2844 
2845 /* transaction handle which is passed around for all journal calls */
2846 struct reiserfs_transaction_handle {
2847 	/*
2848 	 * super for this FS when journal_begin was called. saves calls to
2849 	 * reiserfs_get_super also used by nested transactions to make
2850 	 * sure they are nesting on the right FS _must_ be first
2851 	 * in the handle
2852 	 */
2853 	struct super_block *t_super;
2854 
2855 	int t_refcount;
2856 	int t_blocks_logged;	/* number of blocks this writer has logged */
2857 	int t_blocks_allocated;	/* number of blocks this writer allocated */
2858 
2859 	/* sanity check, equals the current trans id */
2860 	unsigned int t_trans_id;
2861 
2862 	void *t_handle_save;	/* save existing current->journal_info */
2863 
2864 	/*
2865 	 * if new block allocation occurres, that block
2866 	 * should be displaced from others
2867 	 */
2868 	unsigned displace_new_blocks:1;
2869 
2870 	struct list_head t_list;
2871 };
2872 
2873 /*
2874  * used to keep track of ordered and tail writes, attached to the buffer
2875  * head through b_journal_head.
2876  */
2877 struct reiserfs_jh {
2878 	struct reiserfs_journal_list *jl;
2879 	struct buffer_head *bh;
2880 	struct list_head list;
2881 };
2882 
2883 void reiserfs_free_jh(struct buffer_head *bh);
2884 int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
2885 int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
2886 int journal_mark_dirty(struct reiserfs_transaction_handle *,
2887 		       struct buffer_head *bh);
2888 
reiserfs_file_data_log(struct inode * inode)2889 static inline int reiserfs_file_data_log(struct inode *inode)
2890 {
2891 	if (reiserfs_data_log(inode->i_sb) ||
2892 	    (REISERFS_I(inode)->i_flags & i_data_log))
2893 		return 1;
2894 	return 0;
2895 }
2896 
reiserfs_transaction_running(struct super_block * s)2897 static inline int reiserfs_transaction_running(struct super_block *s)
2898 {
2899 	struct reiserfs_transaction_handle *th = current->journal_info;
2900 	if (th && th->t_super == s)
2901 		return 1;
2902 	if (th && th->t_super == NULL)
2903 		BUG();
2904 	return 0;
2905 }
2906 
reiserfs_transaction_free_space(struct reiserfs_transaction_handle * th)2907 static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
2908 {
2909 	return th->t_blocks_allocated - th->t_blocks_logged;
2910 }
2911 
2912 struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
2913 								    super_block
2914 								    *,
2915 								    int count);
2916 int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
2917 void reiserfs_vfs_truncate_file(struct inode *inode);
2918 int reiserfs_commit_page(struct inode *inode, struct page *page,
2919 			 unsigned from, unsigned to);
2920 void reiserfs_flush_old_commits(struct super_block *);
2921 int reiserfs_commit_for_inode(struct inode *);
2922 int reiserfs_inode_needs_commit(struct inode *);
2923 void reiserfs_update_inode_transaction(struct inode *);
2924 void reiserfs_wait_on_write_block(struct super_block *s);
2925 void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
2926 void reiserfs_allow_writes(struct super_block *s);
2927 void reiserfs_check_lock_depth(struct super_block *s, char *caller);
2928 int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
2929 				 int wait);
2930 void reiserfs_restore_prepared_buffer(struct super_block *,
2931 				      struct buffer_head *bh);
2932 int journal_init(struct super_block *, const char *j_dev_name, int old_format,
2933 		 unsigned int);
2934 int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
2935 int journal_release_error(struct reiserfs_transaction_handle *,
2936 			  struct super_block *);
2937 int journal_end(struct reiserfs_transaction_handle *);
2938 int journal_end_sync(struct reiserfs_transaction_handle *);
2939 int journal_mark_freed(struct reiserfs_transaction_handle *,
2940 		       struct super_block *, b_blocknr_t blocknr);
2941 int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
2942 int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
2943 			 int bit_nr, int searchall, b_blocknr_t *next);
2944 int journal_begin(struct reiserfs_transaction_handle *,
2945 		  struct super_block *sb, unsigned long);
2946 int journal_join_abort(struct reiserfs_transaction_handle *,
2947 		       struct super_block *sb);
2948 void reiserfs_abort_journal(struct super_block *sb, int errno);
2949 void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
2950 int reiserfs_allocate_list_bitmaps(struct super_block *s,
2951 				   struct reiserfs_list_bitmap *, unsigned int);
2952 
2953 void reiserfs_schedule_old_flush(struct super_block *s);
2954 void reiserfs_cancel_old_flush(struct super_block *s);
2955 void add_save_link(struct reiserfs_transaction_handle *th,
2956 		   struct inode *inode, int truncate);
2957 int remove_save_link(struct inode *inode, int truncate);
2958 
2959 /* objectid.c */
2960 __u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
2961 void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
2962 			       __u32 objectid_to_release);
2963 int reiserfs_convert_objectid_map_v1(struct super_block *);
2964 
2965 /* stree.c */
2966 int B_IS_IN_TREE(const struct buffer_head *);
2967 extern void copy_item_head(struct item_head *to,
2968 			   const struct item_head *from);
2969 
2970 /* first key is in cpu form, second - le */
2971 extern int comp_short_keys(const struct reiserfs_key *le_key,
2972 			   const struct cpu_key *cpu_key);
2973 extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
2974 
2975 /* both are in le form */
2976 extern int comp_le_keys(const struct reiserfs_key *,
2977 			const struct reiserfs_key *);
2978 extern int comp_short_le_keys(const struct reiserfs_key *,
2979 			      const struct reiserfs_key *);
2980 
2981 /* * get key version from on disk key - kludge */
le_key_version(const struct reiserfs_key * key)2982 static inline int le_key_version(const struct reiserfs_key *key)
2983 {
2984 	int type;
2985 
2986 	type = offset_v2_k_type(&(key->u.k_offset_v2));
2987 	if (type != TYPE_DIRECT && type != TYPE_INDIRECT
2988 	    && type != TYPE_DIRENTRY)
2989 		return KEY_FORMAT_3_5;
2990 
2991 	return KEY_FORMAT_3_6;
2992 
2993 }
2994 
copy_key(struct reiserfs_key * to,const struct reiserfs_key * from)2995 static inline void copy_key(struct reiserfs_key *to,
2996 			    const struct reiserfs_key *from)
2997 {
2998 	memcpy(to, from, KEY_SIZE);
2999 }
3000 
3001 int comp_items(const struct item_head *stored_ih, const struct treepath *path);
3002 const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
3003 				    const struct super_block *sb);
3004 int search_by_key(struct super_block *, const struct cpu_key *,
3005 		  struct treepath *, int);
3006 #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
3007 int search_for_position_by_key(struct super_block *sb,
3008 			       const struct cpu_key *cpu_key,
3009 			       struct treepath *search_path);
3010 extern void decrement_bcount(struct buffer_head *bh);
3011 void decrement_counters_in_path(struct treepath *search_path);
3012 void pathrelse(struct treepath *search_path);
3013 int reiserfs_check_path(struct treepath *p);
3014 void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
3015 
3016 int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
3017 			 struct treepath *path,
3018 			 const struct cpu_key *key,
3019 			 struct item_head *ih,
3020 			 struct inode *inode, const char *body);
3021 
3022 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
3023 			     struct treepath *path,
3024 			     const struct cpu_key *key,
3025 			     struct inode *inode,
3026 			     const char *body, int paste_size);
3027 
3028 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
3029 			   struct treepath *path,
3030 			   struct cpu_key *key,
3031 			   struct inode *inode,
3032 			   struct page *page, loff_t new_file_size);
3033 
3034 int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
3035 			 struct treepath *path,
3036 			 const struct cpu_key *key,
3037 			 struct inode *inode, struct buffer_head *un_bh);
3038 
3039 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
3040 				struct inode *inode, struct reiserfs_key *key);
3041 int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
3042 			   struct inode *inode);
3043 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
3044 			 struct inode *inode, struct page *,
3045 			 int update_timestamps);
3046 
3047 #define i_block_size(inode) ((inode)->i_sb->s_blocksize)
3048 #define file_size(inode) ((inode)->i_size)
3049 #define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
3050 
3051 #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
3052 !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 )
3053 
3054 void padd_item(char *item, int total_length, int length);
3055 
3056 /* inode.c */
3057 /* args for the create parameter of reiserfs_get_block */
3058 #define GET_BLOCK_NO_CREATE 0	 /* don't create new blocks or convert tails */
3059 #define GET_BLOCK_CREATE 1	 /* add anything you need to find block */
3060 #define GET_BLOCK_NO_HOLE 2	 /* return -ENOENT for file holes */
3061 #define GET_BLOCK_READ_DIRECT 4	 /* read the tail if indirect item not found */
3062 #define GET_BLOCK_NO_IMUX     8	 /* i_mutex is not held, don't preallocate */
3063 #define GET_BLOCK_NO_DANGLE   16 /* don't leave any transactions running */
3064 
3065 void reiserfs_read_locked_inode(struct inode *inode,
3066 				struct reiserfs_iget_args *args);
3067 int reiserfs_find_actor(struct inode *inode, void *p);
3068 int reiserfs_init_locked_inode(struct inode *inode, void *p);
3069 void reiserfs_evict_inode(struct inode *inode);
3070 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
3071 int reiserfs_get_block(struct inode *inode, sector_t block,
3072 		       struct buffer_head *bh_result, int create);
3073 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3074 				     int fh_len, int fh_type);
3075 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
3076 				     int fh_len, int fh_type);
3077 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
3078 		       struct inode *parent);
3079 
3080 int reiserfs_truncate_file(struct inode *, int update_timestamps);
3081 void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
3082 		  int type, int key_length);
3083 void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
3084 		       int version,
3085 		       loff_t offset, int type, int length, int entry_count);
3086 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
3087 
3088 struct reiserfs_security_handle;
3089 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
3090 		       struct inode *dir, umode_t mode,
3091 		       const char *symname, loff_t i_size,
3092 		       struct dentry *dentry, struct inode *inode,
3093 		       struct reiserfs_security_handle *security);
3094 
3095 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
3096 			     struct inode *inode, loff_t size);
3097 
reiserfs_update_sd(struct reiserfs_transaction_handle * th,struct inode * inode)3098 static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
3099 				      struct inode *inode)
3100 {
3101 	reiserfs_update_sd_size(th, inode, inode->i_size);
3102 }
3103 
3104 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
3105 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
3106 
3107 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
3108 
3109 /* namei.c */
3110 void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
3111 int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
3112 			struct treepath *path, struct reiserfs_dir_entry *de);
3113 struct dentry *reiserfs_get_parent(struct dentry *);
3114 
3115 #ifdef CONFIG_REISERFS_PROC_INFO
3116 int reiserfs_proc_info_init(struct super_block *sb);
3117 int reiserfs_proc_info_done(struct super_block *sb);
3118 int reiserfs_proc_info_global_init(void);
3119 int reiserfs_proc_info_global_done(void);
3120 
3121 #define PROC_EXP( e )   e
3122 
3123 #define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
3124 #define PROC_INFO_MAX( sb, field, value )								\
3125     __PINFO( sb ).field =												\
3126         max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
3127 #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
3128 #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
3129 #define PROC_INFO_BH_STAT( sb, bh, level )							\
3130     PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] );						\
3131     PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) );	\
3132     PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
3133 #else
reiserfs_proc_info_init(struct super_block * sb)3134 static inline int reiserfs_proc_info_init(struct super_block *sb)
3135 {
3136 	return 0;
3137 }
3138 
reiserfs_proc_info_done(struct super_block * sb)3139 static inline int reiserfs_proc_info_done(struct super_block *sb)
3140 {
3141 	return 0;
3142 }
3143 
reiserfs_proc_info_global_init(void)3144 static inline int reiserfs_proc_info_global_init(void)
3145 {
3146 	return 0;
3147 }
3148 
reiserfs_proc_info_global_done(void)3149 static inline int reiserfs_proc_info_global_done(void)
3150 {
3151 	return 0;
3152 }
3153 
3154 #define PROC_EXP( e )
3155 #define VOID_V ( ( void ) 0 )
3156 #define PROC_INFO_MAX( sb, field, value ) VOID_V
3157 #define PROC_INFO_INC( sb, field ) VOID_V
3158 #define PROC_INFO_ADD( sb, field, val ) VOID_V
3159 #define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
3160 #endif
3161 
3162 /* dir.c */
3163 extern const struct inode_operations reiserfs_dir_inode_operations;
3164 extern const struct inode_operations reiserfs_symlink_inode_operations;
3165 extern const struct inode_operations reiserfs_special_inode_operations;
3166 extern const struct file_operations reiserfs_dir_operations;
3167 int reiserfs_readdir_inode(struct inode *, struct dir_context *);
3168 
3169 /* tail_conversion.c */
3170 int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
3171 		    struct treepath *, struct buffer_head *, loff_t);
3172 int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
3173 		    struct page *, struct treepath *, const struct cpu_key *,
3174 		    loff_t, char *);
3175 void reiserfs_unmap_buffer(struct buffer_head *);
3176 
3177 /* file.c */
3178 extern const struct inode_operations reiserfs_file_inode_operations;
3179 extern const struct file_operations reiserfs_file_operations;
3180 extern const struct address_space_operations reiserfs_address_space_operations;
3181 
3182 /* fix_nodes.c */
3183 
3184 int fix_nodes(int n_op_mode, struct tree_balance *tb,
3185 	      struct item_head *ins_ih, const void *);
3186 void unfix_nodes(struct tree_balance *);
3187 
3188 /* prints.c */
3189 void __reiserfs_panic(struct super_block *s, const char *id,
3190 		      const char *function, const char *fmt, ...)
3191     __attribute__ ((noreturn));
3192 #define reiserfs_panic(s, id, fmt, args...) \
3193 	__reiserfs_panic(s, id, __func__, fmt, ##args)
3194 void __reiserfs_error(struct super_block *s, const char *id,
3195 		      const char *function, const char *fmt, ...);
3196 #define reiserfs_error(s, id, fmt, args...) \
3197 	 __reiserfs_error(s, id, __func__, fmt, ##args)
3198 void reiserfs_info(struct super_block *s, const char *fmt, ...);
3199 void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
3200 void print_indirect_item(struct buffer_head *bh, int item_num);
3201 void store_print_tb(struct tree_balance *tb);
3202 void print_cur_tb(char *mes);
3203 void print_de(struct reiserfs_dir_entry *de);
3204 void print_bi(struct buffer_info *bi, char *mes);
3205 #define PRINT_LEAF_ITEMS 1	/* print all items */
3206 #define PRINT_DIRECTORY_ITEMS 2	/* print directory items */
3207 #define PRINT_DIRECT_ITEMS 4	/* print contents of direct items */
3208 void print_block(struct buffer_head *bh, ...);
3209 void print_bmap(struct super_block *s, int silent);
3210 void print_bmap_block(int i, char *data, int size, int silent);
3211 /*void print_super_block (struct super_block * s, char * mes);*/
3212 void print_objectid_map(struct super_block *s);
3213 void print_block_head(struct buffer_head *bh, char *mes);
3214 void check_leaf(struct buffer_head *bh);
3215 void check_internal(struct buffer_head *bh);
3216 void print_statistics(struct super_block *s);
3217 char *reiserfs_hashname(int code);
3218 
3219 /* lbalance.c */
3220 int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
3221 		    int mov_bytes, struct buffer_head *Snew);
3222 int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
3223 int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
3224 void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
3225 		       int del_num, int del_bytes);
3226 void leaf_insert_into_buf(struct buffer_info *bi, int before,
3227 			  struct item_head * const inserted_item_ih,
3228 			  const char * const inserted_item_body,
3229 			  int zeros_number);
3230 void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
3231 			  int pos_in_item, int paste_size,
3232 			  const char * const body, int zeros_number);
3233 void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
3234 			  int pos_in_item, int cut_size);
3235 void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
3236 			int new_entry_count, struct reiserfs_de_head *new_dehs,
3237 			const char *records, int paste_size);
3238 /* ibalance.c */
3239 int balance_internal(struct tree_balance *, int, int, struct item_head *,
3240 		     struct buffer_head **);
3241 
3242 /* do_balance.c */
3243 void do_balance_mark_leaf_dirty(struct tree_balance *tb,
3244 				struct buffer_head *bh, int flag);
3245 #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
3246 #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
3247 
3248 void do_balance(struct tree_balance *tb, struct item_head *ih,
3249 		const char *body, int flag);
3250 void reiserfs_invalidate_buffer(struct tree_balance *tb,
3251 				struct buffer_head *bh);
3252 
3253 int get_left_neighbor_position(struct tree_balance *tb, int h);
3254 int get_right_neighbor_position(struct tree_balance *tb, int h);
3255 void replace_key(struct tree_balance *tb, struct buffer_head *, int,
3256 		 struct buffer_head *, int);
3257 void make_empty_node(struct buffer_info *);
3258 struct buffer_head *get_FEB(struct tree_balance *);
3259 
3260 /* bitmap.c */
3261 
3262 /*
3263  * structure contains hints for block allocator, and it is a container for
3264  * arguments, such as node, search path, transaction_handle, etc.
3265  */
3266 struct __reiserfs_blocknr_hint {
3267 	/* inode passed to allocator, if we allocate unf. nodes */
3268 	struct inode *inode;
3269 
3270 	sector_t block;		/* file offset, in blocks */
3271 	struct in_core_key key;
3272 
3273 	/*
3274 	 * search path, used by allocator to deternine search_start by
3275 	 * various ways
3276 	 */
3277 	struct treepath *path;
3278 
3279 	/*
3280 	 * transaction handle is needed to log super blocks
3281 	 * and bitmap blocks changes
3282 	 */
3283 	struct reiserfs_transaction_handle *th;
3284 
3285 	b_blocknr_t beg, end;
3286 
3287 	/*
3288 	 * a field used to transfer search start value (block number)
3289 	 * between different block allocator procedures
3290 	 * (determine_search_start() and others)
3291 	 */
3292 	b_blocknr_t search_start;
3293 
3294 	/*
3295 	 * is set in determine_prealloc_size() function,
3296 	 * used by underlayed function that do actual allocation
3297 	 */
3298 	int prealloc_size;
3299 
3300 	/*
3301 	 * the allocator uses different polices for getting disk
3302 	 * space for formatted/unformatted blocks with/without preallocation
3303 	 */
3304 	unsigned formatted_node:1;
3305 	unsigned preallocate:1;
3306 };
3307 
3308 typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
3309 
3310 int reiserfs_parse_alloc_options(struct super_block *, char *);
3311 void reiserfs_init_alloc_options(struct super_block *s);
3312 
3313 /*
3314  * given a directory, this will tell you what packing locality
3315  * to use for a new object underneat it.  The locality is returned
3316  * in disk byte order (le).
3317  */
3318 __le32 reiserfs_choose_packing(struct inode *dir);
3319 
3320 void show_alloc_options(struct seq_file *seq, struct super_block *s);
3321 int reiserfs_init_bitmap_cache(struct super_block *sb);
3322 void reiserfs_free_bitmap_cache(struct super_block *sb);
3323 void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
3324 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
3325 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
3326 void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
3327 			 b_blocknr_t, int for_unformatted);
3328 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
3329 			       int);
reiserfs_new_form_blocknrs(struct tree_balance * tb,b_blocknr_t * new_blocknrs,int amount_needed)3330 static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
3331 					     b_blocknr_t * new_blocknrs,
3332 					     int amount_needed)
3333 {
3334 	reiserfs_blocknr_hint_t hint = {
3335 		.th = tb->transaction_handle,
3336 		.path = tb->tb_path,
3337 		.inode = NULL,
3338 		.key = tb->key,
3339 		.block = 0,
3340 		.formatted_node = 1
3341 	};
3342 	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
3343 					  0);
3344 }
3345 
reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle * th,struct inode * inode,b_blocknr_t * new_blocknrs,struct treepath * path,sector_t block)3346 static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
3347 					    *th, struct inode *inode,
3348 					    b_blocknr_t * new_blocknrs,
3349 					    struct treepath *path,
3350 					    sector_t block)
3351 {
3352 	reiserfs_blocknr_hint_t hint = {
3353 		.th = th,
3354 		.path = path,
3355 		.inode = inode,
3356 		.block = block,
3357 		.formatted_node = 0,
3358 		.preallocate = 0
3359 	};
3360 	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3361 }
3362 
3363 #ifdef REISERFS_PREALLOCATE
reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle * th,struct inode * inode,b_blocknr_t * new_blocknrs,struct treepath * path,sector_t block)3364 static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
3365 					     *th, struct inode *inode,
3366 					     b_blocknr_t * new_blocknrs,
3367 					     struct treepath *path,
3368 					     sector_t block)
3369 {
3370 	reiserfs_blocknr_hint_t hint = {
3371 		.th = th,
3372 		.path = path,
3373 		.inode = inode,
3374 		.block = block,
3375 		.formatted_node = 0,
3376 		.preallocate = 1
3377 	};
3378 	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
3379 }
3380 
3381 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
3382 			       struct inode *inode);
3383 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
3384 #endif
3385 
3386 /* hashes.c */
3387 __u32 keyed_hash(const signed char *msg, int len);
3388 __u32 yura_hash(const signed char *msg, int len);
3389 __u32 r5_hash(const signed char *msg, int len);
3390 
3391 #define reiserfs_set_le_bit		__set_bit_le
3392 #define reiserfs_test_and_set_le_bit	__test_and_set_bit_le
3393 #define reiserfs_clear_le_bit		__clear_bit_le
3394 #define reiserfs_test_and_clear_le_bit	__test_and_clear_bit_le
3395 #define reiserfs_test_le_bit		test_bit_le
3396 #define reiserfs_find_next_zero_le_bit	find_next_zero_bit_le
3397 
3398 /*
3399  * sometimes reiserfs_truncate may require to allocate few new blocks
3400  * to perform indirect2direct conversion. People probably used to
3401  * think, that truncate should work without problems on a filesystem
3402  * without free disk space. They may complain that they can not
3403  * truncate due to lack of free disk space. This spare space allows us
3404  * to not worry about it. 500 is probably too much, but it should be
3405  * absolutely safe
3406  */
3407 #define SPARE_SPACE 500
3408 
3409 /* prototypes from ioctl.c */
3410 long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
3411 long reiserfs_compat_ioctl(struct file *filp,
3412 		   unsigned int cmd, unsigned long arg);
3413 int reiserfs_unpack(struct inode *inode, struct file *filp);
3414