blob: b3865c84f54ce5b6c18a0b18ac3bf0e039842783 [file] [log] [blame]
Al Virof466c6f2012-03-17 01:16:43 -04001/*
2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
3 */
4
5#include <linux/reiserfs_fs.h>
6
7#include <linux/slab.h>
8#include <linux/interrupt.h>
9#include <linux/sched.h>
10#include <linux/workqueue.h>
11#include <asm/unaligned.h>
12#include <linux/bitops.h>
13#include <linux/proc_fs.h>
14#include <linux/buffer_head.h>
15#include <linux/reiserfs_fs_i.h>
16#include <linux/reiserfs_fs_sb.h>
17
18/* the 32 bit compat definitions with int argument */
19#define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
20#define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
21#define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
22#define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
23#define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
24
25/*
26 * Locking primitives. The write lock is a per superblock
27 * special mutex that has properties close to the Big Kernel Lock
28 * which was used in the previous locking scheme.
29 */
30void reiserfs_write_lock(struct super_block *s);
31void reiserfs_write_unlock(struct super_block *s);
32int reiserfs_write_lock_once(struct super_block *s);
33void reiserfs_write_unlock_once(struct super_block *s, int lock_depth);
34
35#ifdef CONFIG_REISERFS_CHECK
36void reiserfs_lock_check_recursive(struct super_block *s);
37#else
38static inline void reiserfs_lock_check_recursive(struct super_block *s) { }
39#endif
40
41/*
42 * Several mutexes depend on the write lock.
43 * However sometimes we want to relax the write lock while we hold
44 * these mutexes, according to the release/reacquire on schedule()
45 * properties of the Bkl that were used.
46 * Reiserfs performances and locking were based on this scheme.
47 * Now that the write lock is a mutex and not the bkl anymore, doing so
48 * may result in a deadlock:
49 *
50 * A acquire write_lock
51 * A acquire j_commit_mutex
52 * A release write_lock and wait for something
53 * B acquire write_lock
54 * B can't acquire j_commit_mutex and sleep
55 * A can't acquire write lock anymore
56 * deadlock
57 *
58 * What we do here is avoiding such deadlock by playing the same game
59 * than the Bkl: if we can't acquire a mutex that depends on the write lock,
60 * we release the write lock, wait a bit and then retry.
61 *
62 * The mutexes concerned by this hack are:
63 * - The commit mutex of a journal list
64 * - The flush mutex
65 * - The journal lock
66 * - The inode mutex
67 */
68static inline void reiserfs_mutex_lock_safe(struct mutex *m,
69 struct super_block *s)
70{
71 reiserfs_lock_check_recursive(s);
72 reiserfs_write_unlock(s);
73 mutex_lock(m);
74 reiserfs_write_lock(s);
75}
76
77static inline void
78reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,
79 struct super_block *s)
80{
81 reiserfs_lock_check_recursive(s);
82 reiserfs_write_unlock(s);
83 mutex_lock_nested(m, subclass);
84 reiserfs_write_lock(s);
85}
86
87static inline void
88reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)
89{
90 reiserfs_lock_check_recursive(s);
91 reiserfs_write_unlock(s);
92 down_read(sem);
93 reiserfs_write_lock(s);
94}
95
96/*
97 * When we schedule, we usually want to also release the write lock,
98 * according to the previous bkl based locking scheme of reiserfs.
99 */
100static inline void reiserfs_cond_resched(struct super_block *s)
101{
102 if (need_resched()) {
103 reiserfs_write_unlock(s);
104 schedule();
105 reiserfs_write_lock(s);
106 }
107}
108
109struct fid;
110
111/* in reading the #defines, it may help to understand that they employ
112 the following abbreviations:
113
114 B = Buffer
115 I = Item header
116 H = Height within the tree (should be changed to LEV)
117 N = Number of the item in the node
118 STAT = stat data
119 DEH = Directory Entry Header
120 EC = Entry Count
121 E = Entry number
122 UL = Unsigned Long
123 BLKH = BLocK Header
124 UNFM = UNForMatted node
125 DC = Disk Child
126 P = Path
127
128 These #defines are named by concatenating these abbreviations,
129 where first comes the arguments, and last comes the return value,
130 of the macro.
131
132*/
133
134#define USE_INODE_GENERATION_COUNTER
135
136#define REISERFS_PREALLOCATE
137#define DISPLACE_NEW_PACKING_LOCALITIES
138#define PREALLOCATION_SIZE 9
139
140/* n must be power of 2 */
141#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
142
143// to be ok for alpha and others we have to align structures to 8 byte
144// boundary.
145// FIXME: do not change 4 by anything else: there is code which relies on that
146#define ROUND_UP(x) _ROUND_UP(x,8LL)
147
148/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
149** messages.
150*/
151#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
152
153void __reiserfs_warning(struct super_block *s, const char *id,
154 const char *func, const char *fmt, ...);
155#define reiserfs_warning(s, id, fmt, args...) \
156 __reiserfs_warning(s, id, __func__, fmt, ##args)
157/* assertions handling */
158
159/** always check a condition and panic if it's false. */
160#define __RASSERT(cond, scond, format, args...) \
161do { \
162 if (!(cond)) \
163 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
164 __FILE__ ":%i:%s: " format "\n", \
165 in_interrupt() ? -1 : task_pid_nr(current), \
166 __LINE__, __func__ , ##args); \
167} while (0)
168
169#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
170
171#if defined( CONFIG_REISERFS_CHECK )
172#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
173#else
174#define RFALSE( cond, format, args... ) do {;} while( 0 )
175#endif
176
177#define CONSTF __attribute_const__
178/*
179 * Disk Data Structures
180 */
181
182/***************************************************************************/
183/* SUPER BLOCK */
184/***************************************************************************/
185
186/*
187 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
188 * the version in RAM is part of a larger structure containing fields never written to disk.
189 */
190#define UNSET_HASH 0 // read_super will guess about, what hash names
191 // in directories were sorted with
192#define TEA_HASH 1
193#define YURA_HASH 2
194#define R5_HASH 3
195#define DEFAULT_HASH R5_HASH
196
197struct journal_params {
198 __le32 jp_journal_1st_block; /* where does journal start from on its
199 * device */
200 __le32 jp_journal_dev; /* journal device st_rdev */
201 __le32 jp_journal_size; /* size of the journal */
202 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
203 __le32 jp_journal_magic; /* random value made on fs creation (this
204 * was sb_journal_block_count) */
205 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
206 * trans */
207 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
208 * commit be */
209 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
210 * be */
211};
212
213/* this is the super from 3.5.X, where X >= 10 */
214struct reiserfs_super_block_v1 {
215 __le32 s_block_count; /* blocks count */
216 __le32 s_free_blocks; /* free blocks count */
217 __le32 s_root_block; /* root block number */
218 struct journal_params s_journal;
219 __le16 s_blocksize; /* block size */
220 __le16 s_oid_maxsize; /* max size of object id array, see
221 * get_objectid() commentary */
222 __le16 s_oid_cursize; /* current size of object id array */
223 __le16 s_umount_state; /* this is set to 1 when filesystem was
224 * umounted, to 2 - when not */
225 char s_magic[10]; /* reiserfs magic string indicates that
226 * file system is reiserfs:
227 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
228 __le16 s_fs_state; /* it is set to used by fsck to mark which
229 * phase of rebuilding is done */
230 __le32 s_hash_function_code; /* indicate, what hash function is being use
231 * to sort names in a directory*/
232 __le16 s_tree_height; /* height of disk tree */
233 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
234 * each block of file system */
235 __le16 s_version; /* this field is only reliable on filesystem
236 * with non-standard journal */
237 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
238 * device, we need to keep after
239 * making fs with non-standard journal */
240} __attribute__ ((__packed__));
241
242#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
243
244/* this is the on disk super block */
245struct reiserfs_super_block {
246 struct reiserfs_super_block_v1 s_v1;
247 __le32 s_inode_generation;
248 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
249 unsigned char s_uuid[16]; /* filesystem unique identifier */
250 unsigned char s_label[16]; /* filesystem volume label */
251 __le16 s_mnt_count; /* Count of mounts since last fsck */
252 __le16 s_max_mnt_count; /* Maximum mounts before check */
253 __le32 s_lastcheck; /* Timestamp of last fsck */
254 __le32 s_check_interval; /* Interval between checks */
255 char s_unused[76]; /* zero filled by mkreiserfs and
256 * reiserfs_convert_objectid_map_v1()
257 * so any additions must be updated
258 * there as well. */
259} __attribute__ ((__packed__));
260
261#define SB_SIZE (sizeof(struct reiserfs_super_block))
262
263#define REISERFS_VERSION_1 0
264#define REISERFS_VERSION_2 2
265
266// on-disk super block fields converted to cpu form
267#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
268#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
269#define SB_BLOCKSIZE(s) \
270 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
271#define SB_BLOCK_COUNT(s) \
272 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
273#define SB_FREE_BLOCKS(s) \
274 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
275#define SB_REISERFS_MAGIC(s) \
276 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
277#define SB_ROOT_BLOCK(s) \
278 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
279#define SB_TREE_HEIGHT(s) \
280 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
281#define SB_REISERFS_STATE(s) \
282 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
283#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
284#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
285
286#define PUT_SB_BLOCK_COUNT(s, val) \
287 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
288#define PUT_SB_FREE_BLOCKS(s, val) \
289 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
290#define PUT_SB_ROOT_BLOCK(s, val) \
291 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
292#define PUT_SB_TREE_HEIGHT(s, val) \
293 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
294#define PUT_SB_REISERFS_STATE(s, val) \
295 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
296#define PUT_SB_VERSION(s, val) \
297 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
298#define PUT_SB_BMAP_NR(s, val) \
299 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
300
301#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
302#define SB_ONDISK_JOURNAL_SIZE(s) \
303 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
304#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
305 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
306#define SB_ONDISK_JOURNAL_DEVICE(s) \
307 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
308#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
309 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
310
311#define is_block_in_log_or_reserved_area(s, block) \
312 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
313 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
314 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
315 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
316
317int is_reiserfs_3_5(struct reiserfs_super_block *rs);
318int is_reiserfs_3_6(struct reiserfs_super_block *rs);
319int is_reiserfs_jr(struct reiserfs_super_block *rs);
320
321/* ReiserFS leaves the first 64k unused, so that partition labels have
322 enough space. If someone wants to write a fancy bootloader that
323 needs more than 64k, let us know, and this will be increased in size.
324 This number must be larger than than the largest block size on any
325 platform, or code will break. -Hans */
326#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
327#define REISERFS_FIRST_BLOCK unused_define
328#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
329
330/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
331#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
332
333/* reiserfs internal error code (used by search_by_key and fix_nodes)) */
334#define CARRY_ON 0
335#define REPEAT_SEARCH -1
336#define IO_ERROR -2
337#define NO_DISK_SPACE -3
338#define NO_BALANCING_NEEDED (-4)
339#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
340#define QUOTA_EXCEEDED -6
341
342typedef __u32 b_blocknr_t;
343typedef __le32 unp_t;
344
345struct unfm_nodeinfo {
346 unp_t unfm_nodenum;
347 unsigned short unfm_freespace;
348};
349
350/* there are two formats of keys: 3.5 and 3.6
351 */
352#define KEY_FORMAT_3_5 0
353#define KEY_FORMAT_3_6 1
354
355/* there are two stat datas */
356#define STAT_DATA_V1 0
357#define STAT_DATA_V2 1
358
359static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
360{
361 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
362}
363
364static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
365{
366 return sb->s_fs_info;
367}
368
369/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
370 * which overflows on large file systems. */
371static inline __u32 reiserfs_bmap_count(struct super_block *sb)
372{
373 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
374}
375
376static inline int bmap_would_wrap(unsigned bmap_nr)
377{
378 return bmap_nr > ((1LL << 16) - 1);
379}
380
381/** this says about version of key of all items (but stat data) the
382 object consists of */
383#define get_inode_item_key_version( inode ) \
384 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
385
386#define set_inode_item_key_version( inode, version ) \
387 ({ if((version)==KEY_FORMAT_3_6) \
388 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
389 else \
390 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
391
392#define get_inode_sd_version(inode) \
393 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
394
395#define set_inode_sd_version(inode, version) \
396 ({ if((version)==STAT_DATA_V2) \
397 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
398 else \
399 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
400
401/* This is an aggressive tail suppression policy, I am hoping it
402 improves our benchmarks. The principle behind it is that percentage
403 space saving is what matters, not absolute space saving. This is
404 non-intuitive, but it helps to understand it if you consider that the
405 cost to access 4 blocks is not much more than the cost to access 1
406 block, if you have to do a seek and rotate. A tail risks a
407 non-linear disk access that is significant as a percentage of total
408 time cost for a 4 block file and saves an amount of space that is
409 less significant as a percentage of space, or so goes the hypothesis.
410 -Hans */
411#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
412(\
413 (!(n_tail_size)) || \
414 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
415 ( (n_file_size) >= (n_block_size) * 4 ) || \
416 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
417 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
418 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
419 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
420 ( ( (n_file_size) >= (n_block_size) ) && \
421 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
422)
423
424/* Another strategy for tails, this one means only create a tail if all the
425 file would fit into one DIRECT item.
426 Primary intention for this one is to increase performance by decreasing
427 seeking.
428*/
429#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
430(\
431 (!(n_tail_size)) || \
432 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
433)
434
435/*
436 * values for s_umount_state field
437 */
438#define REISERFS_VALID_FS 1
439#define REISERFS_ERROR_FS 2
440
441//
442// there are 5 item types currently
443//
444#define TYPE_STAT_DATA 0
445#define TYPE_INDIRECT 1
446#define TYPE_DIRECT 2
447#define TYPE_DIRENTRY 3
448#define TYPE_MAXTYPE 3
449#define TYPE_ANY 15 // FIXME: comment is required
450
451/***************************************************************************/
452/* KEY & ITEM HEAD */
453/***************************************************************************/
454
455//
456// directories use this key as well as old files
457//
458struct offset_v1 {
459 __le32 k_offset;
460 __le32 k_uniqueness;
461} __attribute__ ((__packed__));
462
463struct offset_v2 {
464 __le64 v;
465} __attribute__ ((__packed__));
466
467static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
468{
469 __u8 type = le64_to_cpu(v2->v) >> 60;
470 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
471}
472
473static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
474{
475 v2->v =
476 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
477}
478
479static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
480{
481 return le64_to_cpu(v2->v) & (~0ULL >> 4);
482}
483
484static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
485{
486 offset &= (~0ULL >> 4);
487 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
488}
489
490/* Key of an item determines its location in the S+tree, and
491 is composed of 4 components */
492struct reiserfs_key {
493 __le32 k_dir_id; /* packing locality: by default parent
494 directory object id */
495 __le32 k_objectid; /* object identifier */
496 union {
497 struct offset_v1 k_offset_v1;
498 struct offset_v2 k_offset_v2;
499 } __attribute__ ((__packed__)) u;
500} __attribute__ ((__packed__));
501
502struct in_core_key {
503 __u32 k_dir_id; /* packing locality: by default parent
504 directory object id */
505 __u32 k_objectid; /* object identifier */
506 __u64 k_offset;
507 __u8 k_type;
508};
509
510struct cpu_key {
511 struct in_core_key on_disk_key;
512 int version;
513 int key_length; /* 3 in all cases but direct2indirect and
514 indirect2direct conversion */
515};
516
517/* Our function for comparing keys can compare keys of different
518 lengths. It takes as a parameter the length of the keys it is to
519 compare. These defines are used in determining what is to be passed
520 to it as that parameter. */
521#define REISERFS_FULL_KEY_LEN 4
522#define REISERFS_SHORT_KEY_LEN 2
523
524/* The result of the key compare */
525#define FIRST_GREATER 1
526#define SECOND_GREATER -1
527#define KEYS_IDENTICAL 0
528#define KEY_FOUND 1
529#define KEY_NOT_FOUND 0
530
531#define KEY_SIZE (sizeof(struct reiserfs_key))
532#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
533
534/* return values for search_by_key and clones */
535#define ITEM_FOUND 1
536#define ITEM_NOT_FOUND 0
537#define ENTRY_FOUND 1
538#define ENTRY_NOT_FOUND 0
539#define DIRECTORY_NOT_FOUND -1
540#define REGULAR_FILE_FOUND -2
541#define DIRECTORY_FOUND -3
542#define BYTE_FOUND 1
543#define BYTE_NOT_FOUND 0
544#define FILE_NOT_FOUND -1
545
546#define POSITION_FOUND 1
547#define POSITION_NOT_FOUND 0
548
549// return values for reiserfs_find_entry and search_by_entry_key
550#define NAME_FOUND 1
551#define NAME_NOT_FOUND 0
552#define GOTO_PREVIOUS_ITEM 2
553#define NAME_FOUND_INVISIBLE 3
554
555/* Everything in the filesystem is stored as a set of items. The
556 item head contains the key of the item, its free space (for
557 indirect items) and specifies the location of the item itself
558 within the block. */
559
560struct item_head {
561 /* Everything in the tree is found by searching for it based on
562 * its key.*/
563 struct reiserfs_key ih_key;
564 union {
565 /* The free space in the last unformatted node of an
566 indirect item if this is an indirect item. This
567 equals 0xFFFF iff this is a direct item or stat data
568 item. Note that the key, not this field, is used to
569 determine the item type, and thus which field this
570 union contains. */
571 __le16 ih_free_space_reserved;
572 /* Iff this is a directory item, this field equals the
573 number of directory entries in the directory item. */
574 __le16 ih_entry_count;
575 } __attribute__ ((__packed__)) u;
576 __le16 ih_item_len; /* total size of the item body */
577 __le16 ih_item_location; /* an offset to the item body
578 * within the block */
579 __le16 ih_version; /* 0 for all old items, 2 for new
580 ones. Highest bit is set by fsck
581 temporary, cleaned after all
582 done */
583} __attribute__ ((__packed__));
584/* size of item header */
585#define IH_SIZE (sizeof(struct item_head))
586
587#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
588#define ih_version(ih) le16_to_cpu((ih)->ih_version)
589#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
590#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
591#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
592
593#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
594#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
595#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
596#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
597#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
598
599#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
600
601#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
602#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
603
604/* these operate on indirect items, where you've got an array of ints
605** at a possibly unaligned location. These are a noop on ia32
606**
607** p is the array of __u32, i is the index into the array, v is the value
608** to store there.
609*/
610#define get_block_num(p, i) get_unaligned_le32((p) + (i))
611#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
612
613//
614// in old version uniqueness field shows key type
615//
616#define V1_SD_UNIQUENESS 0
617#define V1_INDIRECT_UNIQUENESS 0xfffffffe
618#define V1_DIRECT_UNIQUENESS 0xffffffff
619#define V1_DIRENTRY_UNIQUENESS 500
620#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
621
622//
623// here are conversion routines
624//
625static inline int uniqueness2type(__u32 uniqueness) CONSTF;
626static inline int uniqueness2type(__u32 uniqueness)
627{
628 switch ((int)uniqueness) {
629 case V1_SD_UNIQUENESS:
630 return TYPE_STAT_DATA;
631 case V1_INDIRECT_UNIQUENESS:
632 return TYPE_INDIRECT;
633 case V1_DIRECT_UNIQUENESS:
634 return TYPE_DIRECT;
635 case V1_DIRENTRY_UNIQUENESS:
636 return TYPE_DIRENTRY;
637 case V1_ANY_UNIQUENESS:
638 default:
639 return TYPE_ANY;
640 }
641}
642
643static inline __u32 type2uniqueness(int type) CONSTF;
644static inline __u32 type2uniqueness(int type)
645{
646 switch (type) {
647 case TYPE_STAT_DATA:
648 return V1_SD_UNIQUENESS;
649 case TYPE_INDIRECT:
650 return V1_INDIRECT_UNIQUENESS;
651 case TYPE_DIRECT:
652 return V1_DIRECT_UNIQUENESS;
653 case TYPE_DIRENTRY:
654 return V1_DIRENTRY_UNIQUENESS;
655 case TYPE_ANY:
656 default:
657 return V1_ANY_UNIQUENESS;
658 }
659}
660
661//
662// key is pointer to on disk key which is stored in le, result is cpu,
663// there is no way to get version of object from key, so, provide
664// version to these defines
665//
666static inline loff_t le_key_k_offset(int version,
667 const struct reiserfs_key *key)
668{
669 return (version == KEY_FORMAT_3_5) ?
670 le32_to_cpu(key->u.k_offset_v1.k_offset) :
671 offset_v2_k_offset(&(key->u.k_offset_v2));
672}
673
674static inline loff_t le_ih_k_offset(const struct item_head *ih)
675{
676 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
677}
678
679static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
680{
681 return (version == KEY_FORMAT_3_5) ?
682 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
683 offset_v2_k_type(&(key->u.k_offset_v2));
684}
685
686static inline loff_t le_ih_k_type(const struct item_head *ih)
687{
688 return le_key_k_type(ih_version(ih), &(ih->ih_key));
689}
690
691static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
692 loff_t offset)
693{
694 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
695 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
696}
697
698static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
699{
700 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
701}
702
703static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
704 int type)
705{
706 (version == KEY_FORMAT_3_5) ?
707 (void)(key->u.k_offset_v1.k_uniqueness =
708 cpu_to_le32(type2uniqueness(type)))
709 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
710}
711
712static inline void set_le_ih_k_type(struct item_head *ih, int type)
713{
714 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
715}
716
717static inline int is_direntry_le_key(int version, struct reiserfs_key *key)
718{
719 return le_key_k_type(version, key) == TYPE_DIRENTRY;
720}
721
722static inline int is_direct_le_key(int version, struct reiserfs_key *key)
723{
724 return le_key_k_type(version, key) == TYPE_DIRECT;
725}
726
727static inline int is_indirect_le_key(int version, struct reiserfs_key *key)
728{
729 return le_key_k_type(version, key) == TYPE_INDIRECT;
730}
731
732static inline int is_statdata_le_key(int version, struct reiserfs_key *key)
733{
734 return le_key_k_type(version, key) == TYPE_STAT_DATA;
735}
736
737//
738// item header has version.
739//
740static inline int is_direntry_le_ih(struct item_head *ih)
741{
742 return is_direntry_le_key(ih_version(ih), &ih->ih_key);
743}
744
745static inline int is_direct_le_ih(struct item_head *ih)
746{
747 return is_direct_le_key(ih_version(ih), &ih->ih_key);
748}
749
750static inline int is_indirect_le_ih(struct item_head *ih)
751{
752 return is_indirect_le_key(ih_version(ih), &ih->ih_key);
753}
754
755static inline int is_statdata_le_ih(struct item_head *ih)
756{
757 return is_statdata_le_key(ih_version(ih), &ih->ih_key);
758}
759
760//
761// key is pointer to cpu key, result is cpu
762//
763static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
764{
765 return key->on_disk_key.k_offset;
766}
767
768static inline loff_t cpu_key_k_type(const struct cpu_key *key)
769{
770 return key->on_disk_key.k_type;
771}
772
773static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
774{
775 key->on_disk_key.k_offset = offset;
776}
777
778static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
779{
780 key->on_disk_key.k_type = type;
781}
782
783static inline void cpu_key_k_offset_dec(struct cpu_key *key)
784{
785 key->on_disk_key.k_offset--;
786}
787
788#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
789#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
790#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
791#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
792
793/* are these used ? */
794#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
795#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
796#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
797#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
798
799#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
800 (!COMP_SHORT_KEYS(ih, key) && \
801 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
802
803/* maximal length of item */
804#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
805#define MIN_ITEM_LEN 1
806
807/* object identifier for root dir */
808#define REISERFS_ROOT_OBJECTID 2
809#define REISERFS_ROOT_PARENT_OBJECTID 1
810
811extern struct reiserfs_key root_key;
812
813/*
814 * Picture represents a leaf of the S+tree
815 * ______________________________________________________
816 * | | Array of | | |
817 * |Block | Object-Item | F r e e | Objects- |
818 * | head | Headers | S p a c e | Items |
819 * |______|_______________|___________________|___________|
820 */
821
822/* Header of a disk block. More precisely, header of a formatted leaf
823 or internal node, and not the header of an unformatted node. */
824struct block_head {
825 __le16 blk_level; /* Level of a block in the tree. */
826 __le16 blk_nr_item; /* Number of keys/items in a block. */
827 __le16 blk_free_space; /* Block free space in bytes. */
828 __le16 blk_reserved;
829 /* dump this in v4/planA */
830 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
831};
832
833#define BLKH_SIZE (sizeof(struct block_head))
834#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
835#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
836#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
837#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
838#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
839#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
840#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
841#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
842#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
843#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
844
845/*
846 * values for blk_level field of the struct block_head
847 */
848
849#define FREE_LEVEL 0 /* when node gets removed from the tree its
850 blk_level is set to FREE_LEVEL. It is then
851 used to see whether the node is still in the
852 tree */
853
854#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
855
856/* Given the buffer head of a formatted node, resolve to the block head of that node. */
857#define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
858/* Number of items that are in buffer. */
859#define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
860#define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
861#define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
862
863#define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
864#define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
865#define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
866
867/* Get right delimiting key. -- little endian */
868#define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
869
870/* Does the buffer contain a disk leaf. */
871#define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
872
873/* Does the buffer contain a disk internal node */
874#define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
875 && B_LEVEL(bh) <= MAX_HEIGHT)
876
877/***************************************************************************/
878/* STAT DATA */
879/***************************************************************************/
880
881//
882// old stat data is 32 bytes long. We are going to distinguish new one by
883// different size
884//
885struct stat_data_v1 {
886 __le16 sd_mode; /* file type, permissions */
887 __le16 sd_nlink; /* number of hard links */
888 __le16 sd_uid; /* owner */
889 __le16 sd_gid; /* group */
890 __le32 sd_size; /* file size */
891 __le32 sd_atime; /* time of last access */
892 __le32 sd_mtime; /* time file was last modified */
893 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
894 union {
895 __le32 sd_rdev;
896 __le32 sd_blocks; /* number of blocks file uses */
897 } __attribute__ ((__packed__)) u;
898 __le32 sd_first_direct_byte; /* first byte of file which is stored
899 in a direct item: except that if it
900 equals 1 it is a symlink and if it
901 equals ~(__u32)0 there is no
902 direct item. The existence of this
903 field really grates on me. Let's
904 replace it with a macro based on
905 sd_size and our tail suppression
906 policy. Someday. -Hans */
907} __attribute__ ((__packed__));
908
909#define SD_V1_SIZE (sizeof(struct stat_data_v1))
910#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
911#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
912#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
913#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
914#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
915#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
916#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
917#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
918#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
919#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
920#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
921#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
922#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
923#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
924#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
925#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
926#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
927#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
928#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
929#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
930#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
931#define sd_v1_first_direct_byte(sdp) \
932 (le32_to_cpu((sdp)->sd_first_direct_byte))
933#define set_sd_v1_first_direct_byte(sdp,v) \
934 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
935
936/* inode flags stored in sd_attrs (nee sd_reserved) */
937
938/* we want common flags to have the same values as in ext2,
939 so chattr(1) will work without problems */
940#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
941#define REISERFS_APPEND_FL FS_APPEND_FL
942#define REISERFS_SYNC_FL FS_SYNC_FL
943#define REISERFS_NOATIME_FL FS_NOATIME_FL
944#define REISERFS_NODUMP_FL FS_NODUMP_FL
945#define REISERFS_SECRM_FL FS_SECRM_FL
946#define REISERFS_UNRM_FL FS_UNRM_FL
947#define REISERFS_COMPR_FL FS_COMPR_FL
948#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
949
950/* persistent flags that file inherits from the parent directory */
951#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
952 REISERFS_SYNC_FL | \
953 REISERFS_NOATIME_FL | \
954 REISERFS_NODUMP_FL | \
955 REISERFS_SECRM_FL | \
956 REISERFS_COMPR_FL | \
957 REISERFS_NOTAIL_FL )
958
959/* Stat Data on disk (reiserfs version of UFS disk inode minus the
960 address blocks) */
961struct stat_data {
962 __le16 sd_mode; /* file type, permissions */
963 __le16 sd_attrs; /* persistent inode flags */
964 __le32 sd_nlink; /* number of hard links */
965 __le64 sd_size; /* file size */
966 __le32 sd_uid; /* owner */
967 __le32 sd_gid; /* group */
968 __le32 sd_atime; /* time of last access */
969 __le32 sd_mtime; /* time file was last modified */
970 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
971 __le32 sd_blocks;
972 union {
973 __le32 sd_rdev;
974 __le32 sd_generation;
975 //__le32 sd_first_direct_byte;
976 /* first byte of file which is stored in a
977 direct item: except that if it equals 1
978 it is a symlink and if it equals
979 ~(__u32)0 there is no direct item. The
980 existence of this field really grates
981 on me. Let's replace it with a macro
982 based on sd_size and our tail
983 suppression policy? */
984 } __attribute__ ((__packed__)) u;
985} __attribute__ ((__packed__));
986//
987// this is 44 bytes long
988//
989#define SD_SIZE (sizeof(struct stat_data))
990#define SD_V2_SIZE SD_SIZE
991#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
992#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
993#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
994/* sd_reserved */
995/* set_sd_reserved */
996#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
997#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
998#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
999#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
1000#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
1001#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
1002#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
1003#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
1004#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
1005#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
1006#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
1007#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
1008#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
1009#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
1010#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
1011#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
1012#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
1013#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
1014#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
1015#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
1016#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
1017#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
1018
1019/***************************************************************************/
1020/* DIRECTORY STRUCTURE */
1021/***************************************************************************/
1022/*
1023 Picture represents the structure of directory items
1024 ________________________________________________
1025 | Array of | | | | | |
1026 | directory |N-1| N-2 | .... | 1st |0th|
1027 | entry headers | | | | | |
1028 |_______________|___|_____|________|_______|___|
1029 <---- directory entries ------>
1030
1031 First directory item has k_offset component 1. We store "." and ".."
1032 in one item, always, we never split "." and ".." into differing
1033 items. This makes, among other things, the code for removing
1034 directories simpler. */
1035#define SD_OFFSET 0
1036#define SD_UNIQUENESS 0
1037#define DOT_OFFSET 1
1038#define DOT_DOT_OFFSET 2
1039#define DIRENTRY_UNIQUENESS 500
1040
1041/* */
1042#define FIRST_ITEM_OFFSET 1
1043
1044/*
1045 Q: How to get key of object pointed to by entry from entry?
1046
1047 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
1048 of object, entry points to */
1049
1050/* NOT IMPLEMENTED:
1051 Directory will someday contain stat data of object */
1052
1053struct reiserfs_de_head {
1054 __le32 deh_offset; /* third component of the directory entry key */
1055 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
1056 by directory entry */
1057 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
1058 __le16 deh_location; /* offset of name in the whole item */
1059 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
1060 entry is hidden (unlinked) */
1061} __attribute__ ((__packed__));
1062#define DEH_SIZE sizeof(struct reiserfs_de_head)
1063#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
1064#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1065#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1066#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1067#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1068
1069#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1070#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1071#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1072#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1073#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1074
1075/* empty directory contains two entries "." and ".." and their headers */
1076#define EMPTY_DIR_SIZE \
1077(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1078
1079/* old format directories have this size when empty */
1080#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1081
1082#define DEH_Statdata 0 /* not used now */
1083#define DEH_Visible 2
1084
1085/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1086#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1087# define ADDR_UNALIGNED_BITS (3)
1088#endif
1089
1090/* These are only used to manipulate deh_state.
1091 * Because of this, we'll use the ext2_ bit routines,
1092 * since they are little endian */
1093#ifdef ADDR_UNALIGNED_BITS
1094
1095# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1096# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1097
1098# define set_bit_unaligned(nr, addr) \
1099 __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1100# define clear_bit_unaligned(nr, addr) \
1101 __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1102# define test_bit_unaligned(nr, addr) \
1103 test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))
1104
1105#else
1106
1107# define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr)
1108# define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr)
1109# define test_bit_unaligned(nr, addr) test_bit_le(nr, addr)
1110
1111#endif
1112
1113#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1114#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1115#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1116#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1117
1118#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1119#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1120#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1121
1122extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1123 __le32 par_dirid, __le32 par_objid);
1124extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1125 __le32 par_dirid, __le32 par_objid);
1126
1127/* array of the entry headers */
1128 /* get item body */
1129#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1130#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1131
1132/* length of the directory entry in directory item. This define
1133 calculates length of i-th directory entry using directory entry
1134 locations from dir entry head. When it calculates length of 0-th
1135 directory entry, it uses length of whole item in place of entry
1136 location of the non-existent following entry in the calculation.
1137 See picture above.*/
1138/*
1139#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1140((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1141*/
1142static inline int entry_length(const struct buffer_head *bh,
1143 const struct item_head *ih, int pos_in_item)
1144{
1145 struct reiserfs_de_head *deh;
1146
1147 deh = B_I_DEH(bh, ih) + pos_in_item;
1148 if (pos_in_item)
1149 return deh_location(deh - 1) - deh_location(deh);
1150
1151 return ih_item_len(ih) - deh_location(deh);
1152}
1153
1154/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1155#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1156
1157/* name by bh, ih and entry_num */
1158#define B_I_E_NAME(bh,ih,entry_num) ((char *)(bh->b_data + ih_location(ih) + deh_location(B_I_DEH(bh,ih)+(entry_num))))
1159
1160// two entries per block (at least)
1161#define REISERFS_MAX_NAME(block_size) 255
1162
1163/* this structure is used for operations on directory entries. It is
1164 not a disk structure. */
1165/* When reiserfs_find_entry or search_by_entry_key find directory
1166 entry, they return filled reiserfs_dir_entry structure */
1167struct reiserfs_dir_entry {
1168 struct buffer_head *de_bh;
1169 int de_item_num;
1170 struct item_head *de_ih;
1171 int de_entry_num;
1172 struct reiserfs_de_head *de_deh;
1173 int de_entrylen;
1174 int de_namelen;
1175 char *de_name;
1176 unsigned long *de_gen_number_bit_string;
1177
1178 __u32 de_dir_id;
1179 __u32 de_objectid;
1180
1181 struct cpu_key de_entry_key;
1182};
1183
1184/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1185
1186/* pointer to file name, stored in entry */
1187#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1188
1189/* length of name */
1190#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1191(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1192
1193/* hash value occupies bits from 7 up to 30 */
1194#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1195/* generation number occupies 7 bits starting from 0 up to 6 */
1196#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1197#define MAX_GENERATION_NUMBER 127
1198
1199#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1200
1201/*
1202 * Picture represents an internal node of the reiserfs tree
1203 * ______________________________________________________
1204 * | | Array of | Array of | Free |
1205 * |block | keys | pointers | space |
1206 * | head | N | N+1 | |
1207 * |______|_______________|___________________|___________|
1208 */
1209
1210/***************************************************************************/
1211/* DISK CHILD */
1212/***************************************************************************/
1213/* Disk child pointer: The pointer from an internal node of the tree
1214 to a node that is on disk. */
1215struct disk_child {
1216 __le32 dc_block_number; /* Disk child's block number. */
1217 __le16 dc_size; /* Disk child's used space. */
1218 __le16 dc_reserved;
1219};
1220
1221#define DC_SIZE (sizeof(struct disk_child))
1222#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1223#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1224#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1225#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1226
1227/* Get disk child by buffer header and position in the tree node. */
1228#define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
1229((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
1230
1231/* Get disk child number by buffer header and position in the tree node. */
1232#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
1233#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
1234 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
1235
1236 /* maximal value of field child_size in structure disk_child */
1237 /* child size is the combined size of all items and their headers */
1238#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1239
1240/* amount of used space in buffer (not including block head) */
1241#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1242
1243/* max and min number of keys in internal node */
1244#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1245#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1246
1247/***************************************************************************/
1248/* PATH STRUCTURES AND DEFINES */
1249/***************************************************************************/
1250
1251/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1252 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1253 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1254 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1255 position of the block_number of the next node if it is looking through an internal node. If it
1256 is looking through a leaf node bin_search will find the position of the item which has key either
1257 equal to given key, or which is the maximal key less than the given key. */
1258
1259struct path_element {
1260 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1261 int pe_position; /* Position in the tree node which is placed in the */
1262 /* buffer above. */
1263};
1264
1265#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1266#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1267#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1268
1269#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1270#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1271
1272/* We need to keep track of who the ancestors of nodes are. When we
1273 perform a search we record which nodes were visited while
1274 descending the tree looking for the node we searched for. This list
1275 of nodes is called the path. This information is used while
1276 performing balancing. Note that this path information may become
1277 invalid, and this means we must check it when using it to see if it
1278 is still valid. You'll need to read search_by_key and the comments
1279 in it, especially about decrement_counters_in_path(), to understand
1280 this structure.
1281
1282Paths make the code so much harder to work with and debug.... An
1283enormous number of bugs are due to them, and trying to write or modify
1284code that uses them just makes my head hurt. They are based on an
1285excessive effort to avoid disturbing the precious VFS code.:-( The
1286gods only know how we are going to SMP the code that uses them.
1287znodes are the way! */
1288
1289#define PATH_READA 0x1 /* do read ahead */
1290#define PATH_READA_BACK 0x2 /* read backwards */
1291
1292struct treepath {
1293 int path_length; /* Length of the array above. */
1294 int reada;
1295 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1296 int pos_in_item;
1297};
1298
1299#define pos_in_item(path) ((path)->pos_in_item)
1300
1301#define INITIALIZE_PATH(var) \
1302struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1303
1304/* Get path element by path and path position. */
1305#define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
1306
1307/* Get buffer header at the path by path and path position. */
1308#define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
1309
1310/* Get position in the element at the path by path and path position. */
1311#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
1312
1313#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
1314 /* you know, to the person who didn't
1315 write this the macro name does not
1316 at first suggest what it does.
1317 Maybe POSITION_FROM_PATH_END? Or
1318 maybe we should just focus on
1319 dumping paths... -Hans */
1320#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
1321
1322#define PATH_PITEM_HEAD(path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path))
1323
1324/* in do_balance leaf has h == 0 in contrast with path structure,
1325 where root has level == 0. That is why we need these defines */
1326#define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h)) /* tb->S[h] */
1327#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1328#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1329#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1330
1331#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
1332
1333#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1334#define get_ih(path) PATH_PITEM_HEAD(path)
1335#define get_item_pos(path) PATH_LAST_POSITION(path)
1336#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1337#define item_moved(ih,path) comp_items(ih, path)
1338#define path_changed(ih,path) comp_items (ih, path)
1339
1340/***************************************************************************/
1341/* MISC */
1342/***************************************************************************/
1343
1344/* Size of pointer to the unformatted node. */
1345#define UNFM_P_SIZE (sizeof(unp_t))
1346#define UNFM_P_SHIFT 2
1347
1348// in in-core inode key is stored on le form
1349#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1350
1351#define MAX_UL_INT 0xffffffff
1352#define MAX_INT 0x7ffffff
1353#define MAX_US_INT 0xffff
1354
1355// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1356#define U32_MAX (~(__u32)0)
1357
1358static inline loff_t max_reiserfs_offset(struct inode *inode)
1359{
1360 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1361 return (loff_t) U32_MAX;
1362
1363 return (loff_t) ((~(__u64) 0) >> 4);
1364}
1365
1366/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1367#define MAX_KEY_OBJECTID MAX_UL_INT
1368
1369#define MAX_B_NUM MAX_UL_INT
1370#define MAX_FC_NUM MAX_US_INT
1371
1372/* the purpose is to detect overflow of an unsigned short */
1373#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1374
1375/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
1376#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1377#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1378
1379#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1380#define get_generation(s) atomic_read (&fs_generation(s))
1381#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1382#define __fs_changed(gen,s) (gen != get_generation (s))
1383#define fs_changed(gen,s) \
1384({ \
1385 reiserfs_cond_resched(s); \
1386 __fs_changed(gen, s); \
1387})
1388
1389/***************************************************************************/
1390/* FIXATE NODES */
1391/***************************************************************************/
1392
1393#define VI_TYPE_LEFT_MERGEABLE 1
1394#define VI_TYPE_RIGHT_MERGEABLE 2
1395
1396/* To make any changes in the tree we always first find node, that
1397 contains item to be changed/deleted or place to insert a new
1398 item. We call this node S. To do balancing we need to decide what
1399 we will shift to left/right neighbor, or to a new node, where new
1400 item will be etc. To make this analysis simpler we build virtual
1401 node. Virtual node is an array of items, that will replace items of
1402 node S. (For instance if we are going to delete an item, virtual
1403 node does not contain it). Virtual node keeps information about
1404 item sizes and types, mergeability of first and last items, sizes
1405 of all entries in directory item. We use this array of items when
1406 calculating what we can shift to neighbors and how many nodes we
1407 have to have if we do not any shiftings, if we shift to left/right
1408 neighbor or to both. */
1409struct virtual_item {
1410 int vi_index; // index in the array of item operations
1411 unsigned short vi_type; // left/right mergeability
1412 unsigned short vi_item_len; /* length of item that it will have after balancing */
1413 struct item_head *vi_ih;
1414 const char *vi_item; // body of item (old or new)
1415 const void *vi_new_data; // 0 always but paste mode
1416 void *vi_uarea; // item specific area
1417};
1418
1419struct virtual_node {
1420 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
1421 unsigned short vn_nr_item; /* number of items in virtual node */
1422 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1423 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1424 short vn_affected_item_num;
1425 short vn_pos_in_item;
1426 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
1427 const void *vn_data;
1428 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1429};
1430
1431/* used by directory items when creating virtual nodes */
1432struct direntry_uarea {
1433 int flags;
1434 __u16 entry_count;
1435 __u16 entry_sizes[1];
1436} __attribute__ ((__packed__));
1437
1438/***************************************************************************/
1439/* TREE BALANCE */
1440/***************************************************************************/
1441
1442/* This temporary structure is used in tree balance algorithms, and
1443 constructed as we go to the extent that its various parts are
1444 needed. It contains arrays of nodes that can potentially be
1445 involved in the balancing of node S, and parameters that define how
1446 each of the nodes must be balanced. Note that in these algorithms
1447 for balancing the worst case is to need to balance the current node
1448 S and the left and right neighbors and all of their parents plus
1449 create a new node. We implement S1 balancing for the leaf nodes
1450 and S0 balancing for the internal nodes (S1 and S0 are defined in
1451 our papers.)*/
1452
1453#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1454
1455/* maximum number of FEB blocknrs on a single level */
1456#define MAX_AMOUNT_NEEDED 2
1457
1458/* someday somebody will prefix every field in this struct with tb_ */
1459struct tree_balance {
1460 int tb_mode;
1461 int need_balance_dirty;
1462 struct super_block *tb_sb;
1463 struct reiserfs_transaction_handle *transaction_handle;
1464 struct treepath *tb_path;
1465 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1466 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
1467 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1468 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1469 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1470 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1471
1472 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1473 cur_blknum. */
1474 struct buffer_head *used[MAX_FEB_SIZE];
1475 struct buffer_head *thrown[MAX_FEB_SIZE];
1476 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1477 shifted to the left in order to balance the
1478 current node; for leaves includes item that
1479 will be partially shifted; for internal
1480 nodes, it is the number of child pointers
1481 rather than items. It includes the new item
1482 being created. The code sometimes subtracts
1483 one to get the number of wholly shifted
1484 items for other purposes. */
1485 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1486 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1487 S[h] to its item number within the node CFL[h] */
1488 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1489 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1490 S[h]. A negative value means removing. */
1491 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1492 balancing on the level h of the tree. If 0 then S is
1493 being deleted, if 1 then S is remaining and no new nodes
1494 are being created, if 2 or 3 then 1 or 2 new nodes is
1495 being created */
1496
1497 /* fields that are used only for balancing leaves of the tree */
1498 int cur_blknum; /* number of empty blocks having been already allocated */
1499 int s0num; /* number of items that fall into left most node when S[0] splits */
1500 int s1num; /* number of items that fall into first new node when S[0] splits */
1501 int s2num; /* number of items that fall into second new node when S[0] splits */
1502 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1503 /* most liquid item that cannot be shifted from S[0] entirely */
1504 /* if -1 then nothing will be partially shifted */
1505 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1506 /* most liquid item that cannot be shifted from S[0] entirely */
1507 /* if -1 then nothing will be partially shifted */
1508 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1509 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1510 int s2bytes;
1511 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1512 char *vn_buf; /* kmalloced memory. Used to create
1513 virtual node and keep map of
1514 dirtied bitmap blocks */
1515 int vn_buf_size; /* size of the vn_buf */
1516 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
1517
1518 int fs_gen; /* saved value of `reiserfs_generation' counter
1519 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1520#ifdef DISPLACE_NEW_PACKING_LOCALITIES
1521 struct in_core_key key; /* key pointer, to pass to block allocator or
1522 another low-level subsystem */
1523#endif
1524};
1525
1526/* These are modes of balancing */
1527
1528/* When inserting an item. */
1529#define M_INSERT 'i'
1530/* When inserting into (directories only) or appending onto an already
1531 existent item. */
1532#define M_PASTE 'p'
1533/* When deleting an item. */
1534#define M_DELETE 'd'
1535/* When truncating an item or removing an entry from a (directory) item. */
1536#define M_CUT 'c'
1537
1538/* used when balancing on leaf level skipped (in reiserfsck) */
1539#define M_INTERNAL 'n'
1540
1541/* When further balancing is not needed, then do_balance does not need
1542 to be called. */
1543#define M_SKIP_BALANCING 's'
1544#define M_CONVERT 'v'
1545
1546/* modes of leaf_move_items */
1547#define LEAF_FROM_S_TO_L 0
1548#define LEAF_FROM_S_TO_R 1
1549#define LEAF_FROM_R_TO_L 2
1550#define LEAF_FROM_L_TO_R 3
1551#define LEAF_FROM_S_TO_SNEW 4
1552
1553#define FIRST_TO_LAST 0
1554#define LAST_TO_FIRST 1
1555
1556/* used in do_balance for passing parent of node information that has
1557 been gotten from tb struct */
1558struct buffer_info {
1559 struct tree_balance *tb;
1560 struct buffer_head *bi_bh;
1561 struct buffer_head *bi_parent;
1562 int bi_position;
1563};
1564
1565static inline struct super_block *sb_from_tb(struct tree_balance *tb)
1566{
1567 return tb ? tb->tb_sb : NULL;
1568}
1569
1570static inline struct super_block *sb_from_bi(struct buffer_info *bi)
1571{
1572 return bi ? sb_from_tb(bi->tb) : NULL;
1573}
1574
1575/* there are 4 types of items: stat data, directory item, indirect, direct.
1576+-------------------+------------+--------------+------------+
1577| | k_offset | k_uniqueness | mergeable? |
1578+-------------------+------------+--------------+------------+
1579| stat data | 0 | 0 | no |
1580+-------------------+------------+--------------+------------+
1581| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1582| non 1st directory | hash value | | yes |
1583| item | | | |
1584+-------------------+------------+--------------+------------+
1585| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1586+-------------------+------------+--------------+------------+
1587| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1588+-------------------+------------+--------------+------------+
1589*/
1590
1591struct item_operations {
1592 int (*bytes_number) (struct item_head * ih, int block_size);
1593 void (*decrement_key) (struct cpu_key *);
1594 int (*is_left_mergeable) (struct reiserfs_key * ih,
1595 unsigned long bsize);
1596 void (*print_item) (struct item_head *, char *item);
1597 void (*check_item) (struct item_head *, char *item);
1598
1599 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1600 int is_affected, int insert_size);
1601 int (*check_left) (struct virtual_item * vi, int free,
1602 int start_skip, int end_skip);
1603 int (*check_right) (struct virtual_item * vi, int free);
1604 int (*part_size) (struct virtual_item * vi, int from, int to);
1605 int (*unit_num) (struct virtual_item * vi);
1606 void (*print_vi) (struct virtual_item * vi);
1607};
1608
1609extern struct item_operations *item_ops[TYPE_ANY + 1];
1610
1611#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1612#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1613#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1614#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1615#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)
1616#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1617#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1618#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1619#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1620#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1621
1622#define COMP_SHORT_KEYS comp_short_keys
1623
1624/* number of blocks pointed to by the indirect item */
1625#define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
1626
1627/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1628#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1629
1630/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1631
1632/* get the item header */
1633#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1634
1635/* get key */
1636#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1637
1638/* get the key */
1639#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1640
1641/* get item body */
1642#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1643
1644/* get the stat data by the buffer header and the item order */
1645#define B_N_STAT_DATA(bh,nr) \
1646( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1647
1648 /* following defines use reiserfs buffer header and item header */
1649
1650/* get stat-data */
1651#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1652
1653// this is 3976 for size==4096
1654#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1655
1656/* indirect items consist of entries which contain blocknrs, pos
1657 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1658 blocknr contained by the entry pos points to */
1659#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1660#define PUT_B_I_POS_UNFM_POINTER(bh,ih,pos, val) do {*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)) = cpu_to_le32(val); } while (0)
1661
1662struct reiserfs_iget_args {
1663 __u32 objectid;
1664 __u32 dirid;
1665};
1666
1667/***************************************************************************/
1668/* FUNCTION DECLARATIONS */
1669/***************************************************************************/
1670
1671#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1672
1673#define journal_trans_half(blocksize) \
1674 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1675
1676/* journal.c see journal.c for all the comments here */
1677
1678/* first block written in a commit. */
1679struct reiserfs_journal_desc {
1680 __le32 j_trans_id; /* id of commit */
1681 __le32 j_len; /* length of commit. len +1 is the commit block */
1682 __le32 j_mount_id; /* mount id of this trans */
1683 __le32 j_realblock[1]; /* real locations for each block */
1684};
1685
1686#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1687#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1688#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1689
1690#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1691#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1692#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1693
1694/* last block written in a commit */
1695struct reiserfs_journal_commit {
1696 __le32 j_trans_id; /* must match j_trans_id from the desc block */
1697 __le32 j_len; /* ditto */
1698 __le32 j_realblock[1]; /* real locations for each block */
1699};
1700
1701#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1702#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1703#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1704
1705#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1706#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1707
1708/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1709** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1710** and this transaction does not need to be replayed.
1711*/
1712struct reiserfs_journal_header {
1713 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
1714 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
1715 __le32 j_mount_id;
1716 /* 12 */ struct journal_params jh_journal;
1717};
1718
1719/* biggest tunable defines are right here */
1720#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1721#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1722#define JOURNAL_TRANS_MIN_DEFAULT 256
1723#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1724#define JOURNAL_MIN_RATIO 2
1725#define JOURNAL_MAX_COMMIT_AGE 30
1726#define JOURNAL_MAX_TRANS_AGE 30
1727#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1728#define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
1729 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
1730 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
1731
1732#ifdef CONFIG_QUOTA
1733#define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))
1734/* We need to update data and inode (atime) */
1735#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)
1736/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1737#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
1738(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1739/* same as with INIT */
1740#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \
1741(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
1742#else
1743#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1744#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1745#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
1746#endif
1747
1748/* both of these can be as low as 1, or as high as you want. The min is the
1749** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1750** as needed, and released when transactions are committed. On release, if
1751** the current number of nodes is > max, the node is freed, otherwise,
1752** it is put on a free list for faster use later.
1753*/
1754#define REISERFS_MIN_BITMAP_NODES 10
1755#define REISERFS_MAX_BITMAP_NODES 100
1756
1757#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1758#define JBH_HASH_MASK 8191
1759
1760#define _jhashfn(sb,block) \
1761 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1762 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1763#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1764
1765// We need these to make journal.c code more readable
1766#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1767#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1768#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1769
1770enum reiserfs_bh_state_bits {
1771 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1772 BH_JDirty_wait,
1773 BH_JNew, /* disk block was taken off free list before
1774 * being in a finished transaction, or
1775 * written to disk. Can be reused immed. */
1776 BH_JPrepared,
1777 BH_JRestore_dirty,
1778 BH_JTest, // debugging only will go away
1779};
1780
1781BUFFER_FNS(JDirty, journaled);
1782TAS_BUFFER_FNS(JDirty, journaled);
1783BUFFER_FNS(JDirty_wait, journal_dirty);
1784TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1785BUFFER_FNS(JNew, journal_new);
1786TAS_BUFFER_FNS(JNew, journal_new);
1787BUFFER_FNS(JPrepared, journal_prepared);
1788TAS_BUFFER_FNS(JPrepared, journal_prepared);
1789BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1790TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1791BUFFER_FNS(JTest, journal_test);
1792TAS_BUFFER_FNS(JTest, journal_test);
1793
1794/*
1795** transaction handle which is passed around for all journal calls
1796*/
1797struct reiserfs_transaction_handle {
1798 struct super_block *t_super; /* super for this FS when journal_begin was
1799 called. saves calls to reiserfs_get_super
1800 also used by nested transactions to make
1801 sure they are nesting on the right FS
1802 _must_ be first in the handle
1803 */
1804 int t_refcount;
1805 int t_blocks_logged; /* number of blocks this writer has logged */
1806 int t_blocks_allocated; /* number of blocks this writer allocated */
1807 unsigned int t_trans_id; /* sanity check, equals the current trans id */
1808 void *t_handle_save; /* save existing current->journal_info */
1809 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1810 should be displaced from others */
1811 struct list_head t_list;
1812};
1813
1814/* used to keep track of ordered and tail writes, attached to the buffer
1815 * head through b_journal_head.
1816 */
1817struct reiserfs_jh {
1818 struct reiserfs_journal_list *jl;
1819 struct buffer_head *bh;
1820 struct list_head list;
1821};
1822
1823void reiserfs_free_jh(struct buffer_head *bh);
1824int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1825int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
1826int journal_mark_dirty(struct reiserfs_transaction_handle *,
1827 struct super_block *, struct buffer_head *bh);
1828
1829static inline int reiserfs_file_data_log(struct inode *inode)
1830{
1831 if (reiserfs_data_log(inode->i_sb) ||
1832 (REISERFS_I(inode)->i_flags & i_data_log))
1833 return 1;
1834 return 0;
1835}
1836
1837static inline int reiserfs_transaction_running(struct super_block *s)
1838{
1839 struct reiserfs_transaction_handle *th = current->journal_info;
1840 if (th && th->t_super == s)
1841 return 1;
1842 if (th && th->t_super == NULL)
1843 BUG();
1844 return 0;
1845}
1846
1847static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
1848{
1849 return th->t_blocks_allocated - th->t_blocks_logged;
1850}
1851
1852struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1853 super_block
1854 *,
1855 int count);
1856int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1857int reiserfs_commit_page(struct inode *inode, struct page *page,
1858 unsigned from, unsigned to);
1859int reiserfs_flush_old_commits(struct super_block *);
1860int reiserfs_commit_for_inode(struct inode *);
1861int reiserfs_inode_needs_commit(struct inode *);
1862void reiserfs_update_inode_transaction(struct inode *);
1863void reiserfs_wait_on_write_block(struct super_block *s);
1864void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1865void reiserfs_allow_writes(struct super_block *s);
1866void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1867int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1868 int wait);
1869void reiserfs_restore_prepared_buffer(struct super_block *,
1870 struct buffer_head *bh);
1871int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1872 unsigned int);
1873int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1874int journal_release_error(struct reiserfs_transaction_handle *,
1875 struct super_block *);
1876int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1877 unsigned long);
1878int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1879 unsigned long);
1880int journal_mark_freed(struct reiserfs_transaction_handle *,
1881 struct super_block *, b_blocknr_t blocknr);
1882int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
1883int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
1884 int bit_nr, int searchall, b_blocknr_t *next);
1885int journal_begin(struct reiserfs_transaction_handle *,
1886 struct super_block *sb, unsigned long);
1887int journal_join_abort(struct reiserfs_transaction_handle *,
1888 struct super_block *sb, unsigned long);
1889void reiserfs_abort_journal(struct super_block *sb, int errno);
1890void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1891int reiserfs_allocate_list_bitmaps(struct super_block *s,
1892 struct reiserfs_list_bitmap *, unsigned int);
1893
1894void add_save_link(struct reiserfs_transaction_handle *th,
1895 struct inode *inode, int truncate);
1896int remove_save_link(struct inode *inode, int truncate);
1897
1898/* objectid.c */
1899__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1900void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1901 __u32 objectid_to_release);
1902int reiserfs_convert_objectid_map_v1(struct super_block *);
1903
1904/* stree.c */
1905int B_IS_IN_TREE(const struct buffer_head *);
1906extern void copy_item_head(struct item_head *to,
1907 const struct item_head *from);
1908
1909// first key is in cpu form, second - le
1910extern int comp_short_keys(const struct reiserfs_key *le_key,
1911 const struct cpu_key *cpu_key);
1912extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
1913
1914// both are in le form
1915extern int comp_le_keys(const struct reiserfs_key *,
1916 const struct reiserfs_key *);
1917extern int comp_short_le_keys(const struct reiserfs_key *,
1918 const struct reiserfs_key *);
1919
1920//
1921// get key version from on disk key - kludge
1922//
1923static inline int le_key_version(const struct reiserfs_key *key)
1924{
1925 int type;
1926
1927 type = offset_v2_k_type(&(key->u.k_offset_v2));
1928 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1929 && type != TYPE_DIRENTRY)
1930 return KEY_FORMAT_3_5;
1931
1932 return KEY_FORMAT_3_6;
1933
1934}
1935
1936static inline void copy_key(struct reiserfs_key *to,
1937 const struct reiserfs_key *from)
1938{
1939 memcpy(to, from, KEY_SIZE);
1940}
1941
1942int comp_items(const struct item_head *stored_ih, const struct treepath *path);
1943const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
1944 const struct super_block *sb);
1945int search_by_key(struct super_block *, const struct cpu_key *,
1946 struct treepath *, int);
1947#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
1948int search_for_position_by_key(struct super_block *sb,
1949 const struct cpu_key *cpu_key,
1950 struct treepath *search_path);
1951extern void decrement_bcount(struct buffer_head *bh);
1952void decrement_counters_in_path(struct treepath *search_path);
1953void pathrelse(struct treepath *search_path);
1954int reiserfs_check_path(struct treepath *p);
1955void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
1956
1957int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
1958 struct treepath *path,
1959 const struct cpu_key *key,
1960 struct item_head *ih,
1961 struct inode *inode, const char *body);
1962
1963int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
1964 struct treepath *path,
1965 const struct cpu_key *key,
1966 struct inode *inode,
1967 const char *body, int paste_size);
1968
1969int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
1970 struct treepath *path,
1971 struct cpu_key *key,
1972 struct inode *inode,
1973 struct page *page, loff_t new_file_size);
1974
1975int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
1976 struct treepath *path,
1977 const struct cpu_key *key,
1978 struct inode *inode, struct buffer_head *un_bh);
1979
1980void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1981 struct inode *inode, struct reiserfs_key *key);
1982int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
1983 struct inode *inode);
1984int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
1985 struct inode *inode, struct page *,
1986 int update_timestamps);
1987
1988#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1989#define file_size(inode) ((inode)->i_size)
1990#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1991
1992#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1993!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 )
1994
1995void padd_item(char *item, int total_length, int length);
1996
1997/* inode.c */
1998/* args for the create parameter of reiserfs_get_block */
1999#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
2000#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
2001#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
2002#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
2003#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
2004#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
2005
2006void reiserfs_read_locked_inode(struct inode *inode,
2007 struct reiserfs_iget_args *args);
2008int reiserfs_find_actor(struct inode *inode, void *p);
2009int reiserfs_init_locked_inode(struct inode *inode, void *p);
2010void reiserfs_evict_inode(struct inode *inode);
2011int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);
2012int reiserfs_get_block(struct inode *inode, sector_t block,
2013 struct buffer_head *bh_result, int create);
2014struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2015 int fh_len, int fh_type);
2016struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
2017 int fh_len, int fh_type);
2018int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
2019 int connectable);
2020
2021int reiserfs_truncate_file(struct inode *, int update_timestamps);
2022void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
2023 int type, int key_length);
2024void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
2025 int version,
2026 loff_t offset, int type, int length, int entry_count);
2027struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
2028
2029struct reiserfs_security_handle;
2030int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
2031 struct inode *dir, umode_t mode,
2032 const char *symname, loff_t i_size,
2033 struct dentry *dentry, struct inode *inode,
2034 struct reiserfs_security_handle *security);
2035
2036void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
2037 struct inode *inode, loff_t size);
2038
2039static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
2040 struct inode *inode)
2041{
2042 reiserfs_update_sd_size(th, inode, inode->i_size);
2043}
2044
2045void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
2046void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
2047int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
2048
2049int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);
2050
2051/* namei.c */
2052void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
2053int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
2054 struct treepath *path, struct reiserfs_dir_entry *de);
2055struct dentry *reiserfs_get_parent(struct dentry *);
2056
2057#ifdef CONFIG_REISERFS_PROC_INFO
2058int reiserfs_proc_info_init(struct super_block *sb);
2059int reiserfs_proc_info_done(struct super_block *sb);
2060int reiserfs_proc_info_global_init(void);
2061int reiserfs_proc_info_global_done(void);
2062
2063#define PROC_EXP( e ) e
2064
2065#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
2066#define PROC_INFO_MAX( sb, field, value ) \
2067 __PINFO( sb ).field = \
2068 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
2069#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
2070#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
2071#define PROC_INFO_BH_STAT( sb, bh, level ) \
2072 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
2073 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
2074 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
2075#else
2076static inline int reiserfs_proc_info_init(struct super_block *sb)
2077{
2078 return 0;
2079}
2080
2081static inline int reiserfs_proc_info_done(struct super_block *sb)
2082{
2083 return 0;
2084}
2085
2086static inline int reiserfs_proc_info_global_init(void)
2087{
2088 return 0;
2089}
2090
2091static inline int reiserfs_proc_info_global_done(void)
2092{
2093 return 0;
2094}
2095
2096#define PROC_EXP( e )
2097#define VOID_V ( ( void ) 0 )
2098#define PROC_INFO_MAX( sb, field, value ) VOID_V
2099#define PROC_INFO_INC( sb, field ) VOID_V
2100#define PROC_INFO_ADD( sb, field, val ) VOID_V
2101#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
2102#endif
2103
2104/* dir.c */
2105extern const struct inode_operations reiserfs_dir_inode_operations;
2106extern const struct inode_operations reiserfs_symlink_inode_operations;
2107extern const struct inode_operations reiserfs_special_inode_operations;
2108extern const struct file_operations reiserfs_dir_operations;
2109int reiserfs_readdir_dentry(struct dentry *, void *, filldir_t, loff_t *);
2110
2111/* tail_conversion.c */
2112int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
2113 struct treepath *, struct buffer_head *, loff_t);
2114int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
2115 struct page *, struct treepath *, const struct cpu_key *,
2116 loff_t, char *);
2117void reiserfs_unmap_buffer(struct buffer_head *);
2118
2119/* file.c */
2120extern const struct inode_operations reiserfs_file_inode_operations;
2121extern const struct file_operations reiserfs_file_operations;
2122extern const struct address_space_operations reiserfs_address_space_operations;
2123
2124/* fix_nodes.c */
2125
2126int fix_nodes(int n_op_mode, struct tree_balance *tb,
2127 struct item_head *ins_ih, const void *);
2128void unfix_nodes(struct tree_balance *);
2129
2130/* prints.c */
2131void __reiserfs_panic(struct super_block *s, const char *id,
2132 const char *function, const char *fmt, ...)
2133 __attribute__ ((noreturn));
2134#define reiserfs_panic(s, id, fmt, args...) \
2135 __reiserfs_panic(s, id, __func__, fmt, ##args)
2136void __reiserfs_error(struct super_block *s, const char *id,
2137 const char *function, const char *fmt, ...);
2138#define reiserfs_error(s, id, fmt, args...) \
2139 __reiserfs_error(s, id, __func__, fmt, ##args)
2140void reiserfs_info(struct super_block *s, const char *fmt, ...);
2141void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2142void print_indirect_item(struct buffer_head *bh, int item_num);
2143void store_print_tb(struct tree_balance *tb);
2144void print_cur_tb(char *mes);
2145void print_de(struct reiserfs_dir_entry *de);
2146void print_bi(struct buffer_info *bi, char *mes);
2147#define PRINT_LEAF_ITEMS 1 /* print all items */
2148#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2149#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2150void print_block(struct buffer_head *bh, ...);
2151void print_bmap(struct super_block *s, int silent);
2152void print_bmap_block(int i, char *data, int size, int silent);
2153/*void print_super_block (struct super_block * s, char * mes);*/
2154void print_objectid_map(struct super_block *s);
2155void print_block_head(struct buffer_head *bh, char *mes);
2156void check_leaf(struct buffer_head *bh);
2157void check_internal(struct buffer_head *bh);
2158void print_statistics(struct super_block *s);
2159char *reiserfs_hashname(int code);
2160
2161/* lbalance.c */
2162int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2163 int mov_bytes, struct buffer_head *Snew);
2164int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2165int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2166void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2167 int del_num, int del_bytes);
2168void leaf_insert_into_buf(struct buffer_info *bi, int before,
2169 struct item_head *inserted_item_ih,
2170 const char *inserted_item_body, int zeros_number);
2171void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2172 int pos_in_item, int paste_size, const char *body,
2173 int zeros_number);
2174void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2175 int pos_in_item, int cut_size);
2176void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
2177 int new_entry_count, struct reiserfs_de_head *new_dehs,
2178 const char *records, int paste_size);
2179/* ibalance.c */
2180int balance_internal(struct tree_balance *, int, int, struct item_head *,
2181 struct buffer_head **);
2182
2183/* do_balance.c */
2184void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2185 struct buffer_head *bh, int flag);
2186#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2187#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2188
2189void do_balance(struct tree_balance *tb, struct item_head *ih,
2190 const char *body, int flag);
2191void reiserfs_invalidate_buffer(struct tree_balance *tb,
2192 struct buffer_head *bh);
2193
2194int get_left_neighbor_position(struct tree_balance *tb, int h);
2195int get_right_neighbor_position(struct tree_balance *tb, int h);
2196void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2197 struct buffer_head *, int);
2198void make_empty_node(struct buffer_info *);
2199struct buffer_head *get_FEB(struct tree_balance *);
2200
2201/* bitmap.c */
2202
2203/* structure contains hints for block allocator, and it is a container for
2204 * arguments, such as node, search path, transaction_handle, etc. */
2205struct __reiserfs_blocknr_hint {
2206 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
2207 sector_t block; /* file offset, in blocks */
2208 struct in_core_key key;
2209 struct treepath *path; /* search path, used by allocator to deternine search_start by
2210 * various ways */
2211 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2212 * bitmap blocks changes */
2213 b_blocknr_t beg, end;
2214 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
2215 * between different block allocator procedures
2216 * (determine_search_start() and others) */
2217 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2218 * function that do actual allocation */
2219
2220 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
2221 * formatted/unformatted blocks with/without preallocation */
2222 unsigned preallocate:1;
2223};
2224
2225typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2226
2227int reiserfs_parse_alloc_options(struct super_block *, char *);
2228void reiserfs_init_alloc_options(struct super_block *s);
2229
2230/*
2231 * given a directory, this will tell you what packing locality
2232 * to use for a new object underneat it. The locality is returned
2233 * in disk byte order (le).
2234 */
2235__le32 reiserfs_choose_packing(struct inode *dir);
2236
2237int reiserfs_init_bitmap_cache(struct super_block *sb);
2238void reiserfs_free_bitmap_cache(struct super_block *sb);
2239void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2240struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
2241int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2242void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2243 b_blocknr_t, int for_unformatted);
2244int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2245 int);
2246static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
2247 b_blocknr_t * new_blocknrs,
2248 int amount_needed)
2249{
2250 reiserfs_blocknr_hint_t hint = {
2251 .th = tb->transaction_handle,
2252 .path = tb->tb_path,
2253 .inode = NULL,
2254 .key = tb->key,
2255 .block = 0,
2256 .formatted_node = 1
2257 };
2258 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2259 0);
2260}
2261
2262static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
2263 *th, struct inode *inode,
2264 b_blocknr_t * new_blocknrs,
2265 struct treepath *path,
2266 sector_t block)
2267{
2268 reiserfs_blocknr_hint_t hint = {
2269 .th = th,
2270 .path = path,
2271 .inode = inode,
2272 .block = block,
2273 .formatted_node = 0,
2274 .preallocate = 0
2275 };
2276 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2277}
2278
2279#ifdef REISERFS_PREALLOCATE
2280static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
2281 *th, struct inode *inode,
2282 b_blocknr_t * new_blocknrs,
2283 struct treepath *path,
2284 sector_t block)
2285{
2286 reiserfs_blocknr_hint_t hint = {
2287 .th = th,
2288 .path = path,
2289 .inode = inode,
2290 .block = block,
2291 .formatted_node = 0,
2292 .preallocate = 1
2293 };
2294 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2295}
2296
2297void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2298 struct inode *inode);
2299void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
2300#endif
2301
2302/* hashes.c */
2303__u32 keyed_hash(const signed char *msg, int len);
2304__u32 yura_hash(const signed char *msg, int len);
2305__u32 r5_hash(const signed char *msg, int len);
2306
2307#define reiserfs_set_le_bit __set_bit_le
2308#define reiserfs_test_and_set_le_bit __test_and_set_bit_le
2309#define reiserfs_clear_le_bit __clear_bit_le
2310#define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le
2311#define reiserfs_test_le_bit test_bit_le
2312#define reiserfs_find_next_zero_le_bit find_next_zero_bit_le
2313
2314/* sometimes reiserfs_truncate may require to allocate few new blocks
2315 to perform indirect2direct conversion. People probably used to
2316 think, that truncate should work without problems on a filesystem
2317 without free disk space. They may complain that they can not
2318 truncate due to lack of free disk space. This spare space allows us
2319 to not worry about it. 500 is probably too much, but it should be
2320 absolutely safe */
2321#define SPARE_SPACE 500
2322
2323/* prototypes from ioctl.c */
2324long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
2325long reiserfs_compat_ioctl(struct file *filp,
2326 unsigned int cmd, unsigned long arg);
2327int reiserfs_unpack(struct inode *inode, struct file *filp);