]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/linux/reiserfs_fs.h
reiserfs: rename p_s_inode to inode
[net-next-2.6.git] / include / linux / reiserfs_fs.h
CommitLineData
1da177e4
LT
1/*
2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
3 */
4
5 /* this file has an amazingly stupid
bd4c625c
LT
6 name, yura please fix it to be
7 reiserfs.h, and merge all the rest
8 of our .h files that are in this
9 directory into it. */
1da177e4
LT
10
11#ifndef _LINUX_REISER_FS_H
12#define _LINUX_REISER_FS_H
13
14#include <linux/types.h>
e18fa700
JG
15#include <linux/magic.h>
16
1da177e4
LT
17#ifdef __KERNEL__
18#include <linux/slab.h>
19#include <linux/interrupt.h>
20#include <linux/sched.h>
21#include <linux/workqueue.h>
22#include <asm/unaligned.h>
23#include <linux/bitops.h>
24#include <linux/proc_fs.h>
25#include <linux/smp_lock.h>
26#include <linux/buffer_head.h>
27#include <linux/reiserfs_fs_i.h>
28#include <linux/reiserfs_fs_sb.h>
29#endif
30
be55caf1
CH
31struct fid;
32
1da177e4
LT
33/*
34 * include/linux/reiser_fs.h
35 *
36 * Reiser File System constants and structures
37 *
38 */
39
40/* in reading the #defines, it may help to understand that they employ
41 the following abbreviations:
42
43 B = Buffer
44 I = Item header
45 H = Height within the tree (should be changed to LEV)
46 N = Number of the item in the node
47 STAT = stat data
48 DEH = Directory Entry Header
49 EC = Entry Count
50 E = Entry number
51 UL = Unsigned Long
52 BLKH = BLocK Header
53 UNFM = UNForMatted node
54 DC = Disk Child
55 P = Path
56
57 These #defines are named by concatenating these abbreviations,
58 where first comes the arguments, and last comes the return value,
59 of the macro.
60
61*/
62
63#define USE_INODE_GENERATION_COUNTER
64
65#define REISERFS_PREALLOCATE
66#define DISPLACE_NEW_PACKING_LOCALITIES
67#define PREALLOCATION_SIZE 9
68
69/* n must be power of 2 */
70#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
71
72// to be ok for alpha and others we have to align structures to 8 byte
73// boundary.
74// FIXME: do not change 4 by anything else: there is code which relies on that
75#define ROUND_UP(x) _ROUND_UP(x,8LL)
76
77/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
78** messages.
79*/
bd4c625c 80#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
1da177e4 81
45b03d5e
JM
82void __reiserfs_warning(struct super_block *s, const char *id,
83 const char *func, const char *fmt, ...);
84#define reiserfs_warning(s, id, fmt, args...) \
85 __reiserfs_warning(s, id, __func__, fmt, ##args)
1da177e4
LT
86/* assertions handling */
87
88/** always check a condition and panic if it's false. */
c3a9c210
JM
89#define __RASSERT(cond, scond, format, args...) \
90do { \
91 if (!(cond)) \
92 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
93 __FILE__ ":%i:%s: " format "\n", \
94 in_interrupt() ? -1 : task_pid_nr(current), \
95 __LINE__, __func__ , ##args); \
96} while (0)
1da177e4 97
2d954d06
AV
98#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
99
1da177e4 100#if defined( CONFIG_REISERFS_CHECK )
2d954d06 101#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
1da177e4
LT
102#else
103#define RFALSE( cond, format, args... ) do {;} while( 0 )
104#endif
105
106#define CONSTF __attribute_const__
107/*
108 * Disk Data Structures
109 */
110
111/***************************************************************************/
112/* SUPER BLOCK */
113/***************************************************************************/
114
115/*
116 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
117 * the version in RAM is part of a larger structure containing fields never written to disk.
118 */
bd4c625c
LT
119#define UNSET_HASH 0 // read_super will guess about, what hash names
120 // in directories were sorted with
1da177e4
LT
121#define TEA_HASH 1
122#define YURA_HASH 2
123#define R5_HASH 3
124#define DEFAULT_HASH R5_HASH
125
1da177e4 126struct journal_params {
bd4c625c
LT
127 __le32 jp_journal_1st_block; /* where does journal start from on its
128 * device */
129 __le32 jp_journal_dev; /* journal device st_rdev */
130 __le32 jp_journal_size; /* size of the journal */
131 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
132 __le32 jp_journal_magic; /* random value made on fs creation (this
133 * was sb_journal_block_count) */
134 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
135 * trans */
136 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
137 * commit be */
138 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
139 * be */
1da177e4
LT
140};
141
142/* this is the super from 3.5.X, where X >= 10 */
bd4c625c
LT
143struct reiserfs_super_block_v1 {
144 __le32 s_block_count; /* blocks count */
145 __le32 s_free_blocks; /* free blocks count */
146 __le32 s_root_block; /* root block number */
147 struct journal_params s_journal;
148 __le16 s_blocksize; /* block size */
149 __le16 s_oid_maxsize; /* max size of object id array, see
150 * get_objectid() commentary */
151 __le16 s_oid_cursize; /* current size of object id array */
152 __le16 s_umount_state; /* this is set to 1 when filesystem was
153 * umounted, to 2 - when not */
154 char s_magic[10]; /* reiserfs magic string indicates that
155 * file system is reiserfs:
156 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
157 __le16 s_fs_state; /* it is set to used by fsck to mark which
158 * phase of rebuilding is done */
159 __le32 s_hash_function_code; /* indicate, what hash function is being use
160 * to sort names in a directory*/
161 __le16 s_tree_height; /* height of disk tree */
162 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
163 * each block of file system */
164 __le16 s_version; /* this field is only reliable on filesystem
165 * with non-standard journal */
166 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
167 * device, we need to keep after
168 * making fs with non-standard journal */
1da177e4
LT
169} __attribute__ ((__packed__));
170
171#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
172
173/* this is the on disk super block */
bd4c625c
LT
174struct reiserfs_super_block {
175 struct reiserfs_super_block_v1 s_v1;
176 __le32 s_inode_generation;
177 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
178 unsigned char s_uuid[16]; /* filesystem unique identifier */
179 unsigned char s_label[16]; /* filesystem volume label */
702d21c6
JM
180 __le16 s_mnt_count; /* Count of mounts since last fsck */
181 __le16 s_max_mnt_count; /* Maximum mounts before check */
182 __le32 s_lastcheck; /* Timestamp of last fsck */
183 __le32 s_check_interval; /* Interval between checks */
184 char s_unused[76]; /* zero filled by mkreiserfs and
bd4c625c
LT
185 * reiserfs_convert_objectid_map_v1()
186 * so any additions must be updated
187 * there as well. */
188} __attribute__ ((__packed__));
1da177e4
LT
189
190#define SB_SIZE (sizeof(struct reiserfs_super_block))
191
192#define REISERFS_VERSION_1 0
193#define REISERFS_VERSION_2 2
194
1da177e4
LT
195// on-disk super block fields converted to cpu form
196#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
197#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
198#define SB_BLOCKSIZE(s) \
199 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
200#define SB_BLOCK_COUNT(s) \
201 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
202#define SB_FREE_BLOCKS(s) \
203 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
204#define SB_REISERFS_MAGIC(s) \
205 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
206#define SB_ROOT_BLOCK(s) \
207 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
208#define SB_TREE_HEIGHT(s) \
209 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
210#define SB_REISERFS_STATE(s) \
211 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
212#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
213#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
214
215#define PUT_SB_BLOCK_COUNT(s, val) \
216 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
217#define PUT_SB_FREE_BLOCKS(s, val) \
218 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
219#define PUT_SB_ROOT_BLOCK(s, val) \
220 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
221#define PUT_SB_TREE_HEIGHT(s, val) \
222 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
223#define PUT_SB_REISERFS_STATE(s, val) \
bd4c625c 224 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1da177e4
LT
225#define PUT_SB_VERSION(s, val) \
226 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
227#define PUT_SB_BMAP_NR(s, val) \
228 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
229
1da177e4
LT
230#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
231#define SB_ONDISK_JOURNAL_SIZE(s) \
232 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
233#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
234 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
235#define SB_ONDISK_JOURNAL_DEVICE(s) \
236 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
237#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
b8cc936f 238 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1da177e4
LT
239
240#define is_block_in_log_or_reserved_area(s, block) \
241 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
242 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
243 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
bd4c625c 244 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1da177e4 245
bd4c625c
LT
246int is_reiserfs_3_5(struct reiserfs_super_block *rs);
247int is_reiserfs_3_6(struct reiserfs_super_block *rs);
248int is_reiserfs_jr(struct reiserfs_super_block *rs);
1da177e4
LT
249
250/* ReiserFS leaves the first 64k unused, so that partition labels have
251 enough space. If someone wants to write a fancy bootloader that
252 needs more than 64k, let us know, and this will be increased in size.
253 This number must be larger than than the largest block size on any
254 platform, or code will break. -Hans */
255#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
256#define REISERFS_FIRST_BLOCK unused_define
257#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
258
259/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
260#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
261
262// reiserfs internal error code (used by search_by_key adn fix_nodes))
263#define CARRY_ON 0
264#define REPEAT_SEARCH -1
265#define IO_ERROR -2
266#define NO_DISK_SPACE -3
267#define NO_BALANCING_NEEDED (-4)
268#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
269#define QUOTA_EXCEEDED -6
270
271typedef __u32 b_blocknr_t;
3e8962be 272typedef __le32 unp_t;
1da177e4
LT
273
274struct unfm_nodeinfo {
bd4c625c
LT
275 unp_t unfm_nodenum;
276 unsigned short unfm_freespace;
1da177e4
LT
277};
278
279/* there are two formats of keys: 3.5 and 3.6
280 */
281#define KEY_FORMAT_3_5 0
282#define KEY_FORMAT_3_6 1
283
284/* there are two stat datas */
285#define STAT_DATA_V1 0
286#define STAT_DATA_V2 1
287
1da177e4
LT
288static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
289{
290 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
291}
292
293static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
294{
295 return sb->s_fs_info;
296}
297
cb680c1b
JM
298/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
299 * which overflows on large file systems. */
13d8bcd2 300static inline __u32 reiserfs_bmap_count(struct super_block *sb)
cb680c1b
JM
301{
302 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
303}
304
305static inline int bmap_would_wrap(unsigned bmap_nr)
306{
307 return bmap_nr > ((1LL << 16) - 1);
308}
309
1da177e4
LT
310/** this says about version of key of all items (but stat data) the
311 object consists of */
312#define get_inode_item_key_version( inode ) \
313 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
314
315#define set_inode_item_key_version( inode, version ) \
316 ({ if((version)==KEY_FORMAT_3_6) \
317 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
318 else \
319 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
320
321#define get_inode_sd_version(inode) \
322 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
323
324#define set_inode_sd_version(inode, version) \
325 ({ if((version)==STAT_DATA_V2) \
326 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
327 else \
328 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
329
330/* This is an aggressive tail suppression policy, I am hoping it
331 improves our benchmarks. The principle behind it is that percentage
332 space saving is what matters, not absolute space saving. This is
333 non-intuitive, but it helps to understand it if you consider that the
334 cost to access 4 blocks is not much more than the cost to access 1
335 block, if you have to do a seek and rotate. A tail risks a
336 non-linear disk access that is significant as a percentage of total
337 time cost for a 4 block file and saves an amount of space that is
338 less significant as a percentage of space, or so goes the hypothesis.
339 -Hans */
340#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
341(\
342 (!(n_tail_size)) || \
343 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
344 ( (n_file_size) >= (n_block_size) * 4 ) || \
345 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
346 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
347 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
348 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
349 ( ( (n_file_size) >= (n_block_size) ) && \
350 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
351)
352
353/* Another strategy for tails, this one means only create a tail if all the
354 file would fit into one DIRECT item.
355 Primary intention for this one is to increase performance by decreasing
356 seeking.
bd4c625c 357*/
1da177e4
LT
358#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
359(\
360 (!(n_tail_size)) || \
361 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
362)
363
1da177e4
LT
364/*
365 * values for s_umount_state field
366 */
367#define REISERFS_VALID_FS 1
368#define REISERFS_ERROR_FS 2
369
370//
371// there are 5 item types currently
372//
373#define TYPE_STAT_DATA 0
374#define TYPE_INDIRECT 1
375#define TYPE_DIRECT 2
bd4c625c
LT
376#define TYPE_DIRENTRY 3
377#define TYPE_MAXTYPE 3
378#define TYPE_ANY 15 // FIXME: comment is required
1da177e4
LT
379
380/***************************************************************************/
381/* KEY & ITEM HEAD */
382/***************************************************************************/
383
384//
385// directories use this key as well as old files
386//
387struct offset_v1 {
bd4c625c
LT
388 __le32 k_offset;
389 __le32 k_uniqueness;
1da177e4
LT
390} __attribute__ ((__packed__));
391
392struct offset_v2 {
f8e08a84 393 __le64 v;
1da177e4
LT
394} __attribute__ ((__packed__));
395
bd4c625c 396static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1da177e4 397{
f8e08a84 398 __u8 type = le64_to_cpu(v2->v) >> 60;
bd4c625c 399 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1da177e4 400}
bd4c625c
LT
401
402static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1da177e4 403{
bd4c625c
LT
404 v2->v =
405 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1da177e4 406}
bd4c625c
LT
407
408static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1da177e4 409{
bd4c625c 410 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1da177e4
LT
411}
412
bd4c625c
LT
413static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
414{
415 offset &= (~0ULL >> 4);
416 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1da177e4 417}
1da177e4
LT
418
419/* Key of an item determines its location in the S+tree, and
420 is composed of 4 components */
421struct reiserfs_key {
bd4c625c
LT
422 __le32 k_dir_id; /* packing locality: by default parent
423 directory object id */
424 __le32 k_objectid; /* object identifier */
425 union {
426 struct offset_v1 k_offset_v1;
427 struct offset_v2 k_offset_v2;
428 } __attribute__ ((__packed__)) u;
1da177e4
LT
429} __attribute__ ((__packed__));
430
6a3a16f2 431struct in_core_key {
bd4c625c
LT
432 __u32 k_dir_id; /* packing locality: by default parent
433 directory object id */
434 __u32 k_objectid; /* object identifier */
435 __u64 k_offset;
436 __u8 k_type;
6b9f5829 437};
1da177e4
LT
438
439struct cpu_key {
bd4c625c
LT
440 struct in_core_key on_disk_key;
441 int version;
442 int key_length; /* 3 in all cases but direct2indirect and
443 indirect2direct conversion */
1da177e4
LT
444};
445
446/* Our function for comparing keys can compare keys of different
447 lengths. It takes as a parameter the length of the keys it is to
448 compare. These defines are used in determining what is to be passed
449 to it as that parameter. */
450#define REISERFS_FULL_KEY_LEN 4
451#define REISERFS_SHORT_KEY_LEN 2
452
453/* The result of the key compare */
454#define FIRST_GREATER 1
455#define SECOND_GREATER -1
456#define KEYS_IDENTICAL 0
457#define KEY_FOUND 1
458#define KEY_NOT_FOUND 0
459
460#define KEY_SIZE (sizeof(struct reiserfs_key))
461#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
462
463/* return values for search_by_key and clones */
464#define ITEM_FOUND 1
465#define ITEM_NOT_FOUND 0
466#define ENTRY_FOUND 1
467#define ENTRY_NOT_FOUND 0
468#define DIRECTORY_NOT_FOUND -1
469#define REGULAR_FILE_FOUND -2
470#define DIRECTORY_FOUND -3
471#define BYTE_FOUND 1
472#define BYTE_NOT_FOUND 0
473#define FILE_NOT_FOUND -1
474
475#define POSITION_FOUND 1
476#define POSITION_NOT_FOUND 0
477
478// return values for reiserfs_find_entry and search_by_entry_key
479#define NAME_FOUND 1
480#define NAME_NOT_FOUND 0
481#define GOTO_PREVIOUS_ITEM 2
482#define NAME_FOUND_INVISIBLE 3
483
484/* Everything in the filesystem is stored as a set of items. The
485 item head contains the key of the item, its free space (for
486 indirect items) and specifies the location of the item itself
487 within the block. */
488
bd4c625c 489struct item_head {
1da177e4
LT
490 /* Everything in the tree is found by searching for it based on
491 * its key.*/
492 struct reiserfs_key ih_key;
493 union {
494 /* The free space in the last unformatted node of an
495 indirect item if this is an indirect item. This
496 equals 0xFFFF iff this is a direct item or stat data
497 item. Note that the key, not this field, is used to
498 determine the item type, and thus which field this
499 union contains. */
3e8962be 500 __le16 ih_free_space_reserved;
1da177e4
LT
501 /* Iff this is a directory item, this field equals the
502 number of directory entries in the directory item. */
3e8962be 503 __le16 ih_entry_count;
1da177e4 504 } __attribute__ ((__packed__)) u;
bd4c625c
LT
505 __le16 ih_item_len; /* total size of the item body */
506 __le16 ih_item_location; /* an offset to the item body
507 * within the block */
508 __le16 ih_version; /* 0 for all old items, 2 for new
509 ones. Highest bit is set by fsck
510 temporary, cleaned after all
511 done */
1da177e4
LT
512} __attribute__ ((__packed__));
513/* size of item header */
514#define IH_SIZE (sizeof(struct item_head))
515
516#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
517#define ih_version(ih) le16_to_cpu((ih)->ih_version)
518#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
519#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
520#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
521
522#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
523#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
524#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
525#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
526#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
527
1da177e4
LT
528#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
529
530#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
531#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
532
533/* these operate on indirect items, where you've got an array of ints
534** at a possibly unaligned location. These are a noop on ia32
535**
536** p is the array of __u32, i is the index into the array, v is the value
537** to store there.
538*/
8b5ac31e
HH
539#define get_block_num(p, i) get_unaligned_le32((p) + (i))
540#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1da177e4
LT
541
542//
543// in old version uniqueness field shows key type
544//
545#define V1_SD_UNIQUENESS 0
546#define V1_INDIRECT_UNIQUENESS 0xfffffffe
547#define V1_DIRECT_UNIQUENESS 0xffffffff
548#define V1_DIRENTRY_UNIQUENESS 500
bd4c625c 549#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
1da177e4
LT
550
551//
552// here are conversion routines
553//
bd4c625c
LT
554static inline int uniqueness2type(__u32 uniqueness) CONSTF;
555static inline int uniqueness2type(__u32 uniqueness)
1da177e4 556{
bd4c625c
LT
557 switch ((int)uniqueness) {
558 case V1_SD_UNIQUENESS:
559 return TYPE_STAT_DATA;
560 case V1_INDIRECT_UNIQUENESS:
561 return TYPE_INDIRECT;
562 case V1_DIRECT_UNIQUENESS:
563 return TYPE_DIRECT;
564 case V1_DIRENTRY_UNIQUENESS:
565 return TYPE_DIRENTRY;
1da177e4 566 case V1_ANY_UNIQUENESS:
fd7cb031 567 default:
bd4c625c
LT
568 return TYPE_ANY;
569 }
1da177e4
LT
570}
571
bd4c625c
LT
572static inline __u32 type2uniqueness(int type) CONSTF;
573static inline __u32 type2uniqueness(int type)
1da177e4 574{
bd4c625c
LT
575 switch (type) {
576 case TYPE_STAT_DATA:
577 return V1_SD_UNIQUENESS;
578 case TYPE_INDIRECT:
579 return V1_INDIRECT_UNIQUENESS;
580 case TYPE_DIRECT:
581 return V1_DIRECT_UNIQUENESS;
582 case TYPE_DIRENTRY:
583 return V1_DIRENTRY_UNIQUENESS;
1da177e4 584 case TYPE_ANY:
fd7cb031 585 default:
bd4c625c
LT
586 return V1_ANY_UNIQUENESS;
587 }
1da177e4
LT
588}
589
590//
591// key is pointer to on disk key which is stored in le, result is cpu,
592// there is no way to get version of object from key, so, provide
593// version to these defines
594//
bd4c625c
LT
595static inline loff_t le_key_k_offset(int version,
596 const struct reiserfs_key *key)
1da177e4 597{
bd4c625c
LT
598 return (version == KEY_FORMAT_3_5) ?
599 le32_to_cpu(key->u.k_offset_v1.k_offset) :
600 offset_v2_k_offset(&(key->u.k_offset_v2));
1da177e4
LT
601}
602
bd4c625c 603static inline loff_t le_ih_k_offset(const struct item_head *ih)
1da177e4 604{
bd4c625c 605 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1da177e4
LT
606}
607
bd4c625c 608static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1da177e4 609{
bd4c625c
LT
610 return (version == KEY_FORMAT_3_5) ?
611 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
612 offset_v2_k_type(&(key->u.k_offset_v2));
1da177e4
LT
613}
614
bd4c625c 615static inline loff_t le_ih_k_type(const struct item_head *ih)
1da177e4 616{
bd4c625c 617 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1da177e4
LT
618}
619
bd4c625c
LT
620static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
621 loff_t offset)
1da177e4 622{
bd4c625c
LT
623 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
624 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
1da177e4
LT
625}
626
bd4c625c 627static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1da177e4 628{
bd4c625c 629 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1da177e4
LT
630}
631
bd4c625c
LT
632static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
633 int type)
1da177e4 634{
bd4c625c
LT
635 (version == KEY_FORMAT_3_5) ?
636 (void)(key->u.k_offset_v1.k_uniqueness =
637 cpu_to_le32(type2uniqueness(type)))
638 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
1da177e4 639}
bd4c625c 640static inline void set_le_ih_k_type(struct item_head *ih, int type)
1da177e4 641{
bd4c625c 642 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1da177e4
LT
643}
644
1da177e4
LT
645#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
646#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
647#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
648#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
649
650//
651// item header has version.
652//
653#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
654#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
655#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
656#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
657
1da177e4
LT
658//
659// key is pointer to cpu key, result is cpu
660//
bd4c625c 661static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1da177e4 662{
bd4c625c 663 return key->on_disk_key.k_offset;
1da177e4
LT
664}
665
bd4c625c 666static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1da177e4 667{
bd4c625c 668 return key->on_disk_key.k_type;
1da177e4
LT
669}
670
bd4c625c 671static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1da177e4 672{
6b9f5829 673 key->on_disk_key.k_offset = offset;
1da177e4
LT
674}
675
bd4c625c 676static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1da177e4 677{
6b9f5829 678 key->on_disk_key.k_type = type;
1da177e4
LT
679}
680
bd4c625c 681static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1da177e4 682{
bd4c625c 683 key->on_disk_key.k_offset--;
1da177e4
LT
684}
685
1da177e4
LT
686#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
687#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
688#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
689#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
690
1da177e4
LT
691/* are these used ? */
692#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
693#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
694#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
695#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
696
1da177e4
LT
697#define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
698 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
699 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
700
bd4c625c 701/* maximal length of item */
1da177e4
LT
702#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
703#define MIN_ITEM_LEN 1
704
1da177e4
LT
705/* object identifier for root dir */
706#define REISERFS_ROOT_OBJECTID 2
707#define REISERFS_ROOT_PARENT_OBJECTID 1
708extern struct reiserfs_key root_key;
709
1da177e4
LT
710/*
711 * Picture represents a leaf of the S+tree
712 * ______________________________________________________
713 * | | Array of | | |
714 * |Block | Object-Item | F r e e | Objects- |
715 * | head | Headers | S p a c e | Items |
716 * |______|_______________|___________________|___________|
717 */
718
719/* Header of a disk block. More precisely, header of a formatted leaf
720 or internal node, and not the header of an unformatted node. */
bd4c625c
LT
721struct block_head {
722 __le16 blk_level; /* Level of a block in the tree. */
723 __le16 blk_nr_item; /* Number of keys/items in a block. */
724 __le16 blk_free_space; /* Block free space in bytes. */
725 __le16 blk_reserved;
726 /* dump this in v4/planA */
727 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
1da177e4
LT
728};
729
730#define BLKH_SIZE (sizeof(struct block_head))
731#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
732#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
733#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
734#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
735#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
736#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
737#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
738#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
739#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
740#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
741
742/*
743 * values for blk_level field of the struct block_head
744 */
745
bd4c625c
LT
746#define FREE_LEVEL 0 /* when node gets removed from the tree its
747 blk_level is set to FREE_LEVEL. It is then
748 used to see whether the node is still in the
749 tree */
1da177e4 750
bd4c625c 751#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1da177e4
LT
752
753/* Given the buffer head of a formatted node, resolve to the block head of that node. */
ad31a4fc 754#define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1da177e4 755/* Number of items that are in buffer. */
ad31a4fc
JM
756#define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
757#define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
758#define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1da177e4 759
ad31a4fc
JM
760#define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
761#define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
762#define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1da177e4 763
1da177e4 764/* Get right delimiting key. -- little endian */
ad31a4fc 765#define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1da177e4
LT
766
767/* Does the buffer contain a disk leaf. */
ad31a4fc 768#define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1da177e4
LT
769
770/* Does the buffer contain a disk internal node */
ad31a4fc
JM
771#define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
772 && B_LEVEL(bh) <= MAX_HEIGHT)
1da177e4 773
1da177e4
LT
774/***************************************************************************/
775/* STAT DATA */
776/***************************************************************************/
777
1da177e4
LT
778//
779// old stat data is 32 bytes long. We are going to distinguish new one by
780// different size
781//
bd4c625c
LT
782struct stat_data_v1 {
783 __le16 sd_mode; /* file type, permissions */
784 __le16 sd_nlink; /* number of hard links */
785 __le16 sd_uid; /* owner */
786 __le16 sd_gid; /* group */
787 __le32 sd_size; /* file size */
788 __le32 sd_atime; /* time of last access */
789 __le32 sd_mtime; /* time file was last modified */
790 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
791 union {
792 __le32 sd_rdev;
793 __le32 sd_blocks; /* number of blocks file uses */
794 } __attribute__ ((__packed__)) u;
795 __le32 sd_first_direct_byte; /* first byte of file which is stored
796 in a direct item: except that if it
797 equals 1 it is a symlink and if it
798 equals ~(__u32)0 there is no
799 direct item. The existence of this
800 field really grates on me. Let's
801 replace it with a macro based on
802 sd_size and our tail suppression
803 policy. Someday. -Hans */
1da177e4
LT
804} __attribute__ ((__packed__));
805
806#define SD_V1_SIZE (sizeof(struct stat_data_v1))
807#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
808#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
809#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
810#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
811#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
812#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
813#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
814#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
815#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
816#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
817#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
818#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
819#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
820#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
821#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
822#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
823#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
824#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
825#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
826#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
827#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
828#define sd_v1_first_direct_byte(sdp) \
829 (le32_to_cpu((sdp)->sd_first_direct_byte))
830#define set_sd_v1_first_direct_byte(sdp,v) \
831 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
832
1da177e4
LT
833/* inode flags stored in sd_attrs (nee sd_reserved) */
834
835/* we want common flags to have the same values as in ext2,
836 so chattr(1) will work without problems */
36695673
DH
837#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
838#define REISERFS_APPEND_FL FS_APPEND_FL
839#define REISERFS_SYNC_FL FS_SYNC_FL
840#define REISERFS_NOATIME_FL FS_NOATIME_FL
841#define REISERFS_NODUMP_FL FS_NODUMP_FL
842#define REISERFS_SECRM_FL FS_SECRM_FL
843#define REISERFS_UNRM_FL FS_UNRM_FL
844#define REISERFS_COMPR_FL FS_COMPR_FL
845#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1da177e4
LT
846
847/* persistent flags that file inherits from the parent directory */
848#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
849 REISERFS_SYNC_FL | \
850 REISERFS_NOATIME_FL | \
851 REISERFS_NODUMP_FL | \
852 REISERFS_SECRM_FL | \
853 REISERFS_COMPR_FL | \
854 REISERFS_NOTAIL_FL )
855
856/* Stat Data on disk (reiserfs version of UFS disk inode minus the
857 address blocks) */
858struct stat_data {
bd4c625c
LT
859 __le16 sd_mode; /* file type, permissions */
860 __le16 sd_attrs; /* persistent inode flags */
861 __le32 sd_nlink; /* number of hard links */
862 __le64 sd_size; /* file size */
863 __le32 sd_uid; /* owner */
864 __le32 sd_gid; /* group */
865 __le32 sd_atime; /* time of last access */
866 __le32 sd_mtime; /* time file was last modified */
867 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
868 __le32 sd_blocks;
869 union {
870 __le32 sd_rdev;
871 __le32 sd_generation;
872 //__le32 sd_first_direct_byte;
873 /* first byte of file which is stored in a
874 direct item: except that if it equals 1
875 it is a symlink and if it equals
876 ~(__u32)0 there is no direct item. The
877 existence of this field really grates
878 on me. Let's replace it with a macro
879 based on sd_size and our tail
880 suppression policy? */
881 } __attribute__ ((__packed__)) u;
1da177e4
LT
882} __attribute__ ((__packed__));
883//
884// this is 44 bytes long
885//
886#define SD_SIZE (sizeof(struct stat_data))
887#define SD_V2_SIZE SD_SIZE
888#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
889#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
890#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
891/* sd_reserved */
892/* set_sd_reserved */
893#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
894#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
895#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
896#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
897#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
898#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
899#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
900#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
901#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
902#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
903#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
904#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
905#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
906#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
907#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
908#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
909#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
910#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
911#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
912#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
913#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
914#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
915
1da177e4
LT
916/***************************************************************************/
917/* DIRECTORY STRUCTURE */
918/***************************************************************************/
919/*
920 Picture represents the structure of directory items
921 ________________________________________________
922 | Array of | | | | | |
923 | directory |N-1| N-2 | .... | 1st |0th|
924 | entry headers | | | | | |
925 |_______________|___|_____|________|_______|___|
926 <---- directory entries ------>
927
928 First directory item has k_offset component 1. We store "." and ".."
929 in one item, always, we never split "." and ".." into differing
930 items. This makes, among other things, the code for removing
931 directories simpler. */
932#define SD_OFFSET 0
933#define SD_UNIQUENESS 0
934#define DOT_OFFSET 1
935#define DOT_DOT_OFFSET 2
936#define DIRENTRY_UNIQUENESS 500
937
938/* */
939#define FIRST_ITEM_OFFSET 1
940
941/*
942 Q: How to get key of object pointed to by entry from entry?
943
944 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
945 of object, entry points to */
946
947/* NOT IMPLEMENTED:
948 Directory will someday contain stat data of object */
949
bd4c625c
LT
950struct reiserfs_de_head {
951 __le32 deh_offset; /* third component of the directory entry key */
952 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
953 by directory entry */
954 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
955 __le16 deh_location; /* offset of name in the whole item */
956 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
957 entry is hidden (unlinked) */
1da177e4
LT
958} __attribute__ ((__packed__));
959#define DEH_SIZE sizeof(struct reiserfs_de_head)
960#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
961#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
962#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
963#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
964#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
965
966#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
967#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
968#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
969#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
970#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
971
972/* empty directory contains two entries "." and ".." and their headers */
973#define EMPTY_DIR_SIZE \
974(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
975
976/* old format directories have this size when empty */
977#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
978
bd4c625c 979#define DEH_Statdata 0 /* not used now */
1da177e4
LT
980#define DEH_Visible 2
981
982/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
983#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
984# define ADDR_UNALIGNED_BITS (3)
985#endif
986
987/* These are only used to manipulate deh_state.
988 * Because of this, we'll use the ext2_ bit routines,
989 * since they are little endian */
990#ifdef ADDR_UNALIGNED_BITS
991
992# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
993# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
994
995# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
996# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
997# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
998
999#else
1000
1001# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
1002# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
1003# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
1004
1005#endif
1006
1007#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1008#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1009#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1010#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1011
1012#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1013#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1014#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1015
bd4c625c
LT
1016extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1017 __le32 par_dirid, __le32 par_objid);
1018extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1019 __le32 par_dirid, __le32 par_objid);
1da177e4
LT
1020
1021/* array of the entry headers */
1022 /* get item body */
1023#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1024#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1025
1026/* length of the directory entry in directory item. This define
1027 calculates length of i-th directory entry using directory entry
1028 locations from dir entry head. When it calculates length of 0-th
1029 directory entry, it uses length of whole item in place of entry
1030 location of the non-existent following entry in the calculation.
1031 See picture above.*/
1032/*
1033#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1034((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1035*/
bd4c625c
LT
1036static inline int entry_length(const struct buffer_head *bh,
1037 const struct item_head *ih, int pos_in_item)
1da177e4 1038{
bd4c625c 1039 struct reiserfs_de_head *deh;
1da177e4 1040
bd4c625c
LT
1041 deh = B_I_DEH(bh, ih) + pos_in_item;
1042 if (pos_in_item)
1043 return deh_location(deh - 1) - deh_location(deh);
1da177e4 1044
bd4c625c 1045 return ih_item_len(ih) - deh_location(deh);
1da177e4
LT
1046}
1047
1da177e4
LT
1048/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1049#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1050
1da177e4
LT
1051/* name by bh, ih and entry_num */
1052#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))))
1053
1054// two entries per block (at least)
1055#define REISERFS_MAX_NAME(block_size) 255
1056
1da177e4
LT
1057/* this structure is used for operations on directory entries. It is
1058 not a disk structure. */
1059/* When reiserfs_find_entry or search_by_entry_key find directory
1060 entry, they return filled reiserfs_dir_entry structure */
bd4c625c
LT
1061struct reiserfs_dir_entry {
1062 struct buffer_head *de_bh;
1063 int de_item_num;
1064 struct item_head *de_ih;
1065 int de_entry_num;
1066 struct reiserfs_de_head *de_deh;
1067 int de_entrylen;
1068 int de_namelen;
1069 char *de_name;
3af1efe8 1070 unsigned long *de_gen_number_bit_string;
bd4c625c
LT
1071
1072 __u32 de_dir_id;
1073 __u32 de_objectid;
1074
1075 struct cpu_key de_entry_key;
1da177e4 1076};
bd4c625c 1077
1da177e4
LT
1078/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1079
1080/* pointer to file name, stored in entry */
1081#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1082
1083/* length of name */
1084#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1085(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1086
1da177e4
LT
1087/* hash value occupies bits from 7 up to 30 */
1088#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1089/* generation number occupies 7 bits starting from 0 up to 6 */
1090#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1091#define MAX_GENERATION_NUMBER 127
1092
1093#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1094
1da177e4
LT
1095/*
1096 * Picture represents an internal node of the reiserfs tree
1097 * ______________________________________________________
1098 * | | Array of | Array of | Free |
1099 * |block | keys | pointers | space |
1100 * | head | N | N+1 | |
1101 * |______|_______________|___________________|___________|
1102 */
1103
1104/***************************************************************************/
1105/* DISK CHILD */
1106/***************************************************************************/
1107/* Disk child pointer: The pointer from an internal node of the tree
1108 to a node that is on disk. */
1109struct disk_child {
bd4c625c
LT
1110 __le32 dc_block_number; /* Disk child's block number. */
1111 __le16 dc_size; /* Disk child's used space. */
1112 __le16 dc_reserved;
1da177e4
LT
1113};
1114
1115#define DC_SIZE (sizeof(struct disk_child))
1116#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1117#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1118#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1119#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1120
1121/* Get disk child by buffer header and position in the tree node. */
ad31a4fc
JM
1122#define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
1123((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
1da177e4
LT
1124
1125/* Get disk child number by buffer header and position in the tree node. */
ad31a4fc
JM
1126#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
1127#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
1128 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
1da177e4 1129
bd4c625c 1130 /* maximal value of field child_size in structure disk_child */
1da177e4
LT
1131 /* child size is the combined size of all items and their headers */
1132#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1133
1134/* amount of used space in buffer (not including block head) */
1135#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1136
1137/* max and min number of keys in internal node */
1138#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1139#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1140
1141/***************************************************************************/
1142/* PATH STRUCTURES AND DEFINES */
1143/***************************************************************************/
1144
1da177e4
LT
1145/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1146 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1147 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1148 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1149 position of the block_number of the next node if it is looking through an internal node. If it
1150 is looking through a leaf node bin_search will find the position of the item which has key either
1151 equal to given key, or which is the maximal key less than the given key. */
1152
bd4c625c
LT
1153struct path_element {
1154 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1155 int pe_position; /* Position in the tree node which is placed in the */
1156 /* buffer above. */
1da177e4
LT
1157};
1158
bd4c625c
LT
1159#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1160#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1161#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1da177e4 1162
bd4c625c
LT
1163#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1164#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1da177e4
LT
1165
1166/* We need to keep track of who the ancestors of nodes are. When we
1167 perform a search we record which nodes were visited while
1168 descending the tree looking for the node we searched for. This list
1169 of nodes is called the path. This information is used while
1170 performing balancing. Note that this path information may become
1171 invalid, and this means we must check it when using it to see if it
1172 is still valid. You'll need to read search_by_key and the comments
1173 in it, especially about decrement_counters_in_path(), to understand
1174 this structure.
1175
1176Paths make the code so much harder to work with and debug.... An
1177enormous number of bugs are due to them, and trying to write or modify
1178code that uses them just makes my head hurt. They are based on an
1179excessive effort to avoid disturbing the precious VFS code.:-( The
1180gods only know how we are going to SMP the code that uses them.
1181znodes are the way! */
1182
bd4c625c
LT
1183#define PATH_READA 0x1 /* do read ahead */
1184#define PATH_READA_BACK 0x2 /* read backwards */
1da177e4 1185
fec6d055 1186struct treepath {
bd4c625c
LT
1187 int path_length; /* Length of the array above. */
1188 int reada;
1189 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1190 int pos_in_item;
1da177e4
LT
1191};
1192
1193#define pos_in_item(path) ((path)->pos_in_item)
1194
1195#define INITIALIZE_PATH(var) \
fec6d055 1196struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1da177e4
LT
1197
1198/* Get path element by path and path position. */
1199#define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1200
1201/* Get buffer header at the path by path and path position. */
1202#define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1203
1204/* Get position in the element at the path by path and path position. */
1205#define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1206
1da177e4
LT
1207#define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1208 /* you know, to the person who didn't
bd4c625c
LT
1209 write this the macro name does not
1210 at first suggest what it does.
1211 Maybe POSITION_FROM_PATH_END? Or
1212 maybe we should just focus on
1213 dumping paths... -Hans */
1da177e4
LT
1214#define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1215
1da177e4
LT
1216#define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1217
1218/* in do_balance leaf has h == 0 in contrast with path structure,
1219 where root has level == 0. That is why we need these defines */
1220#define PATH_H_PBUFFER(p_s_path, h) PATH_OFFSET_PBUFFER (p_s_path, p_s_path->path_length - (h)) /* tb->S[h] */
bd4c625c
LT
1221#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1222#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1223#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1da177e4
LT
1224
1225#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1226
1227#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1228#define get_ih(path) PATH_PITEM_HEAD(path)
1229#define get_item_pos(path) PATH_LAST_POSITION(path)
1230#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1231#define item_moved(ih,path) comp_items(ih, path)
1232#define path_changed(ih,path) comp_items (ih, path)
1233
1da177e4
LT
1234/***************************************************************************/
1235/* MISC */
1236/***************************************************************************/
1237
1238/* Size of pointer to the unformatted node. */
1239#define UNFM_P_SIZE (sizeof(unp_t))
1240#define UNFM_P_SHIFT 2
1241
1242// in in-core inode key is stored on le form
1243#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1244
1245#define MAX_UL_INT 0xffffffff
1246#define MAX_INT 0x7ffffff
1247#define MAX_US_INT 0xffff
1248
1249// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1250#define U32_MAX (~(__u32)0)
1251
bd4c625c 1252static inline loff_t max_reiserfs_offset(struct inode *inode)
1da177e4 1253{
bd4c625c
LT
1254 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1255 return (loff_t) U32_MAX;
1da177e4 1256
bd4c625c 1257 return (loff_t) ((~(__u64) 0) >> 4);
1da177e4
LT
1258}
1259
1da177e4
LT
1260/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1261#define MAX_KEY_OBJECTID MAX_UL_INT
1262
1da177e4
LT
1263#define MAX_B_NUM MAX_UL_INT
1264#define MAX_FC_NUM MAX_US_INT
1265
1da177e4
LT
1266/* the purpose is to detect overflow of an unsigned short */
1267#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1268
1da177e4 1269/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
bd4c625c
LT
1270#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1271#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1da177e4
LT
1272
1273#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1274#define get_generation(s) atomic_read (&fs_generation(s))
1275#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1276#define __fs_changed(gen,s) (gen != get_generation (s))
1277#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1278
1da177e4
LT
1279/***************************************************************************/
1280/* FIXATE NODES */
1281/***************************************************************************/
1282
1283#define VI_TYPE_LEFT_MERGEABLE 1
1284#define VI_TYPE_RIGHT_MERGEABLE 2
1285
1286/* To make any changes in the tree we always first find node, that
1287 contains item to be changed/deleted or place to insert a new
1288 item. We call this node S. To do balancing we need to decide what
1289 we will shift to left/right neighbor, or to a new node, where new
1290 item will be etc. To make this analysis simpler we build virtual
1291 node. Virtual node is an array of items, that will replace items of
1292 node S. (For instance if we are going to delete an item, virtual
1293 node does not contain it). Virtual node keeps information about
1294 item sizes and types, mergeability of first and last items, sizes
1295 of all entries in directory item. We use this array of items when
1296 calculating what we can shift to neighbors and how many nodes we
1297 have to have if we do not any shiftings, if we shift to left/right
1298 neighbor or to both. */
bd4c625c
LT
1299struct virtual_item {
1300 int vi_index; // index in the array of item operations
1301 unsigned short vi_type; // left/right mergeability
1302 unsigned short vi_item_len; /* length of item that it will have after balancing */
1303 struct item_head *vi_ih;
1304 const char *vi_item; // body of item (old or new)
1305 const void *vi_new_data; // 0 always but paste mode
1306 void *vi_uarea; // item specific area
1da177e4
LT
1307};
1308
bd4c625c
LT
1309struct virtual_node {
1310 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
1311 unsigned short vn_nr_item; /* number of items in virtual node */
1312 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1313 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1314 short vn_affected_item_num;
1315 short vn_pos_in_item;
1316 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
1317 const void *vn_data;
1318 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1da177e4
LT
1319};
1320
1321/* used by directory items when creating virtual nodes */
1322struct direntry_uarea {
bd4c625c
LT
1323 int flags;
1324 __u16 entry_count;
1325 __u16 entry_sizes[1];
1326} __attribute__ ((__packed__));
1da177e4
LT
1327
1328/***************************************************************************/
1329/* TREE BALANCE */
1330/***************************************************************************/
1331
1332/* This temporary structure is used in tree balance algorithms, and
1333 constructed as we go to the extent that its various parts are
1334 needed. It contains arrays of nodes that can potentially be
1335 involved in the balancing of node S, and parameters that define how
1336 each of the nodes must be balanced. Note that in these algorithms
1337 for balancing the worst case is to need to balance the current node
1338 S and the left and right neighbors and all of their parents plus
1339 create a new node. We implement S1 balancing for the leaf nodes
1340 and S0 balancing for the internal nodes (S1 and S0 are defined in
1341 our papers.)*/
1342
1343#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1344
1345/* maximum number of FEB blocknrs on a single level */
1346#define MAX_AMOUNT_NEEDED 2
1347
1348/* someday somebody will prefix every field in this struct with tb_ */
bd4c625c
LT
1349struct tree_balance {
1350 int tb_mode;
1351 int need_balance_dirty;
1352 struct super_block *tb_sb;
1353 struct reiserfs_transaction_handle *transaction_handle;
fec6d055 1354 struct treepath *tb_path;
bd4c625c
LT
1355 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1356 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
1357 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1358 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1359 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1360 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1361
1362 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1363 cur_blknum. */
1364 struct buffer_head *used[MAX_FEB_SIZE];
1365 struct buffer_head *thrown[MAX_FEB_SIZE];
1366 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1367 shifted to the left in order to balance the
1368 current node; for leaves includes item that
1369 will be partially shifted; for internal
1370 nodes, it is the number of child pointers
1371 rather than items. It includes the new item
1372 being created. The code sometimes subtracts
1373 one to get the number of wholly shifted
1374 items for other purposes. */
1375 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1376 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1377 S[h] to its item number within the node CFL[h] */
1378 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1379 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1380 S[h]. A negative value means removing. */
1381 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1382 balancing on the level h of the tree. If 0 then S is
1383 being deleted, if 1 then S is remaining and no new nodes
1384 are being created, if 2 or 3 then 1 or 2 new nodes is
1385 being created */
1386
1387 /* fields that are used only for balancing leaves of the tree */
1388 int cur_blknum; /* number of empty blocks having been already allocated */
1389 int s0num; /* number of items that fall into left most node when S[0] splits */
1390 int s1num; /* number of items that fall into first new node when S[0] splits */
1391 int s2num; /* number of items that fall into second new node when S[0] splits */
1392 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1393 /* most liquid item that cannot be shifted from S[0] entirely */
1394 /* if -1 then nothing will be partially shifted */
1395 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1396 /* most liquid item that cannot be shifted from S[0] entirely */
1397 /* if -1 then nothing will be partially shifted */
1398 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1399 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1400 int s2bytes;
1401 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1402 char *vn_buf; /* kmalloced memory. Used to create
1da177e4
LT
1403 virtual node and keep map of
1404 dirtied bitmap blocks */
bd4c625c
LT
1405 int vn_buf_size; /* size of the vn_buf */
1406 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
1da177e4 1407
bd4c625c
LT
1408 int fs_gen; /* saved value of `reiserfs_generation' counter
1409 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1da177e4 1410#ifdef DISPLACE_NEW_PACKING_LOCALITIES
bd4c625c
LT
1411 struct in_core_key key; /* key pointer, to pass to block allocator or
1412 another low-level subsystem */
1da177e4 1413#endif
bd4c625c 1414};
1da177e4
LT
1415
1416/* These are modes of balancing */
1417
1418/* When inserting an item. */
1419#define M_INSERT 'i'
1420/* When inserting into (directories only) or appending onto an already
1421 existant item. */
1422#define M_PASTE 'p'
1423/* When deleting an item. */
1424#define M_DELETE 'd'
1425/* When truncating an item or removing an entry from a (directory) item. */
1426#define M_CUT 'c'
1427
1428/* used when balancing on leaf level skipped (in reiserfsck) */
1429#define M_INTERNAL 'n'
1430
1431/* When further balancing is not needed, then do_balance does not need
1432 to be called. */
1433#define M_SKIP_BALANCING 's'
1434#define M_CONVERT 'v'
1435
1436/* modes of leaf_move_items */
1437#define LEAF_FROM_S_TO_L 0
1438#define LEAF_FROM_S_TO_R 1
1439#define LEAF_FROM_R_TO_L 2
1440#define LEAF_FROM_L_TO_R 3
1441#define LEAF_FROM_S_TO_SNEW 4
1442
1443#define FIRST_TO_LAST 0
1444#define LAST_TO_FIRST 1
1445
1446/* used in do_balance for passing parent of node information that has
1447 been gotten from tb struct */
1448struct buffer_info {
bd4c625c
LT
1449 struct tree_balance *tb;
1450 struct buffer_head *bi_bh;
1451 struct buffer_head *bi_parent;
1452 int bi_position;
1da177e4
LT
1453};
1454
c3a9c210
JM
1455static inline struct super_block *sb_from_tb(struct tree_balance *tb)
1456{
1457 return tb ? tb->tb_sb : NULL;
1458}
1459
1460static inline struct super_block *sb_from_bi(struct buffer_info *bi)
1461{
1462 return bi ? sb_from_tb(bi->tb) : NULL;
1463}
1464
1da177e4
LT
1465/* there are 4 types of items: stat data, directory item, indirect, direct.
1466+-------------------+------------+--------------+------------+
1467| | k_offset | k_uniqueness | mergeable? |
1468+-------------------+------------+--------------+------------+
1469| stat data | 0 | 0 | no |
1470+-------------------+------------+--------------+------------+
1471| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1472| non 1st directory | hash value | | yes |
1473| item | | | |
1474+-------------------+------------+--------------+------------+
1475| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1476+-------------------+------------+--------------+------------+
1477| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1478+-------------------+------------+--------------+------------+
1479*/
1480
1481struct item_operations {
bd4c625c
LT
1482 int (*bytes_number) (struct item_head * ih, int block_size);
1483 void (*decrement_key) (struct cpu_key *);
1484 int (*is_left_mergeable) (struct reiserfs_key * ih,
1485 unsigned long bsize);
1486 void (*print_item) (struct item_head *, char *item);
1487 void (*check_item) (struct item_head *, char *item);
1488
1489 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1490 int is_affected, int insert_size);
1491 int (*check_left) (struct virtual_item * vi, int free,
1492 int start_skip, int end_skip);
1493 int (*check_right) (struct virtual_item * vi, int free);
1494 int (*part_size) (struct virtual_item * vi, int from, int to);
1495 int (*unit_num) (struct virtual_item * vi);
1496 void (*print_vi) (struct virtual_item * vi);
1da177e4
LT
1497};
1498
bd4c625c 1499extern struct item_operations *item_ops[TYPE_ANY + 1];
1da177e4
LT
1500
1501#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1502#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1503#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1504#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1505#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)
1506#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1507#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1508#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1509#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1510#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1511
1da177e4
LT
1512#define COMP_SHORT_KEYS comp_short_keys
1513
1514/* number of blocks pointed to by the indirect item */
1515#define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1516
1517/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1518#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1519
1520/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1521
bd4c625c 1522/* get the item header */
1da177e4
LT
1523#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1524
1525/* get key */
1526#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1527
1528/* get the key */
1529#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1530
1531/* get item body */
1532#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1533
1534/* get the stat data by the buffer header and the item order */
1535#define B_N_STAT_DATA(bh,nr) \
1536( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1537
1538 /* following defines use reiserfs buffer header and item header */
1539
1540/* get stat-data */
1541#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1542
1543// this is 3976 for size==4096
1544#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1545
1546/* indirect items consist of entries which contain blocknrs, pos
1547 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1548 blocknr contained by the entry pos points to */
1549#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1550#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)
1551
1552struct reiserfs_iget_args {
bd4c625c
LT
1553 __u32 objectid;
1554 __u32 dirid;
1555};
1da177e4
LT
1556
1557/***************************************************************************/
1558/* FUNCTION DECLARATIONS */
1559/***************************************************************************/
1560
1561/*#ifdef __KERNEL__*/
1562#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1563
1564#define journal_trans_half(blocksize) \
1565 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1566
1567/* journal.c see journal.c for all the comments here */
1568
1569/* first block written in a commit. */
1570struct reiserfs_journal_desc {
bd4c625c
LT
1571 __le32 j_trans_id; /* id of commit */
1572 __le32 j_len; /* length of commit. len +1 is the commit block */
1573 __le32 j_mount_id; /* mount id of this trans */
1574 __le32 j_realblock[1]; /* real locations for each block */
1575};
1da177e4
LT
1576
1577#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1578#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1579#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1580
1581#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1582#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1583#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1584
1585/* last block written in a commit */
1586struct reiserfs_journal_commit {
bd4c625c
LT
1587 __le32 j_trans_id; /* must match j_trans_id from the desc block */
1588 __le32 j_len; /* ditto */
1589 __le32 j_realblock[1]; /* real locations for each block */
1590};
1da177e4
LT
1591
1592#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1593#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1594#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1595
1596#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1597#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1598
1599/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1600** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1601** and this transaction does not need to be replayed.
1602*/
1603struct reiserfs_journal_header {
bd4c625c
LT
1604 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
1605 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
1606 __le32 j_mount_id;
1607 /* 12 */ struct journal_params jh_journal;
1608};
1da177e4
LT
1609
1610/* biggest tunable defines are right here */
bd4c625c
LT
1611#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1612#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1da177e4 1613#define JOURNAL_TRANS_MIN_DEFAULT 256
bd4c625c 1614#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1da177e4 1615#define JOURNAL_MIN_RATIO 2
bd4c625c 1616#define JOURNAL_MAX_COMMIT_AGE 30
1da177e4
LT
1617#define JOURNAL_MAX_TRANS_AGE 30
1618#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
0ab2621e
JM
1619#define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
1620 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
1621 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
1622
1da177e4 1623#ifdef CONFIG_QUOTA
556a2a45
JK
1624/* We need to update data and inode (atime) */
1625#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? 2 : 0)
1626/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1627#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1628(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1629/* same as with INIT */
1630#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1631(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
1da177e4 1632#else
556a2a45
JK
1633#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1634#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1635#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
1da177e4
LT
1636#endif
1637
1638/* both of these can be as low as 1, or as high as you want. The min is the
1639** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1640** as needed, and released when transactions are committed. On release, if
1641** the current number of nodes is > max, the node is freed, otherwise,
1642** it is put on a free list for faster use later.
1643*/
bd4c625c
LT
1644#define REISERFS_MIN_BITMAP_NODES 10
1645#define REISERFS_MAX_BITMAP_NODES 100
1da177e4 1646
bd4c625c 1647#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1da177e4
LT
1648#define JBH_HASH_MASK 8191
1649
1650#define _jhashfn(sb,block) \
1651 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1652 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1653#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1654
1655// We need these to make journal.c code more readable
1656#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1657#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1658#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1659
1660enum reiserfs_bh_state_bits {
bd4c625c
LT
1661 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1662 BH_JDirty_wait,
1663 BH_JNew, /* disk block was taken off free list before
1664 * being in a finished transaction, or
1665 * written to disk. Can be reused immed. */
1666 BH_JPrepared,
1667 BH_JRestore_dirty,
1668 BH_JTest, // debugging only will go away
1da177e4
LT
1669};
1670
1671BUFFER_FNS(JDirty, journaled);
1672TAS_BUFFER_FNS(JDirty, journaled);
1673BUFFER_FNS(JDirty_wait, journal_dirty);
1674TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1675BUFFER_FNS(JNew, journal_new);
1676TAS_BUFFER_FNS(JNew, journal_new);
1677BUFFER_FNS(JPrepared, journal_prepared);
1678TAS_BUFFER_FNS(JPrepared, journal_prepared);
1679BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1680TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1681BUFFER_FNS(JTest, journal_test);
1682TAS_BUFFER_FNS(JTest, journal_test);
1683
1684/*
1685** transaction handle which is passed around for all journal calls
1686*/
1687struct reiserfs_transaction_handle {
bd4c625c
LT
1688 struct super_block *t_super; /* super for this FS when journal_begin was
1689 called. saves calls to reiserfs_get_super
1690 also used by nested transactions to make
1691 sure they are nesting on the right FS
1692 _must_ be first in the handle
1693 */
1694 int t_refcount;
1695 int t_blocks_logged; /* number of blocks this writer has logged */
1696 int t_blocks_allocated; /* number of blocks this writer allocated */
600ed416 1697 unsigned int t_trans_id; /* sanity check, equals the current trans id */
bd4c625c
LT
1698 void *t_handle_save; /* save existing current->journal_info */
1699 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1700 should be displaced from others */
1701 struct list_head t_list;
1702};
1da177e4
LT
1703
1704/* used to keep track of ordered and tail writes, attached to the buffer
1705 * head through b_journal_head.
1706 */
1707struct reiserfs_jh {
bd4c625c
LT
1708 struct reiserfs_journal_list *jl;
1709 struct buffer_head *bh;
1710 struct list_head list;
1da177e4
LT
1711};
1712
1713void reiserfs_free_jh(struct buffer_head *bh);
1714int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1715int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
bd4c625c
LT
1716int journal_mark_dirty(struct reiserfs_transaction_handle *,
1717 struct super_block *, struct buffer_head *bh);
1718
1719static inline int reiserfs_file_data_log(struct inode *inode)
1720{
1721 if (reiserfs_data_log(inode->i_sb) ||
1722 (REISERFS_I(inode)->i_flags & i_data_log))
1723 return 1;
1724 return 0;
1da177e4
LT
1725}
1726
bd4c625c
LT
1727static inline int reiserfs_transaction_running(struct super_block *s)
1728{
1729 struct reiserfs_transaction_handle *th = current->journal_info;
1730 if (th && th->t_super == s)
1731 return 1;
1732 if (th && th->t_super == NULL)
1733 BUG();
1734 return 0;
1da177e4
LT
1735}
1736
23f9e0f8
AZ
1737static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
1738{
1739 return th->t_blocks_allocated - th->t_blocks_logged;
1740}
1741
bd4c625c
LT
1742struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1743 super_block
1744 *,
1745 int count);
1da177e4
LT
1746int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1747int reiserfs_commit_page(struct inode *inode, struct page *page,
bd4c625c 1748 unsigned from, unsigned to);
1da177e4 1749int reiserfs_flush_old_commits(struct super_block *);
bd4c625c
LT
1750int reiserfs_commit_for_inode(struct inode *);
1751int reiserfs_inode_needs_commit(struct inode *);
1752void reiserfs_update_inode_transaction(struct inode *);
1753void reiserfs_wait_on_write_block(struct super_block *s);
1754void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1755void reiserfs_allow_writes(struct super_block *s);
1756void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1757int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1758 int wait);
1759void reiserfs_restore_prepared_buffer(struct super_block *,
1760 struct buffer_head *bh);
1761int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1762 unsigned int);
1763int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1764int journal_release_error(struct reiserfs_transaction_handle *,
1765 struct super_block *);
1766int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1767 unsigned long);
1768int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1769 unsigned long);
1770int journal_mark_freed(struct reiserfs_transaction_handle *,
1771 struct super_block *, b_blocknr_t blocknr);
1772int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
a9dd3643
JM
1773int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
1774 int bit_nr, int searchall, b_blocknr_t *next);
bd4c625c 1775int journal_begin(struct reiserfs_transaction_handle *,
a9dd3643 1776 struct super_block *sb, unsigned long);
bd4c625c 1777int journal_join_abort(struct reiserfs_transaction_handle *,
a9dd3643 1778 struct super_block *sb, unsigned long);
32e8b106 1779void reiserfs_abort_journal(struct super_block *sb, int errno);
bd4c625c
LT
1780void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1781int reiserfs_allocate_list_bitmaps(struct super_block *s,
3ee16670 1782 struct reiserfs_list_bitmap *, unsigned int);
bd4c625c
LT
1783
1784void add_save_link(struct reiserfs_transaction_handle *th,
1785 struct inode *inode, int truncate);
1786int remove_save_link(struct inode *inode, int truncate);
1da177e4
LT
1787
1788/* objectid.c */
bd4c625c
LT
1789__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1790void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1791 __u32 objectid_to_release);
1792int reiserfs_convert_objectid_map_v1(struct super_block *);
1da177e4
LT
1793
1794/* stree.c */
1795int B_IS_IN_TREE(const struct buffer_head *);
bd4c625c
LT
1796extern void copy_item_head(struct item_head *p_v_to,
1797 const struct item_head *p_v_from);
1da177e4
LT
1798
1799// first key is in cpu form, second - le
bd4c625c
LT
1800extern int comp_short_keys(const struct reiserfs_key *le_key,
1801 const struct cpu_key *cpu_key);
1802extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
1da177e4
LT
1803
1804// both are in le form
bd4c625c
LT
1805extern int comp_le_keys(const struct reiserfs_key *,
1806 const struct reiserfs_key *);
1807extern int comp_short_le_keys(const struct reiserfs_key *,
1808 const struct reiserfs_key *);
1da177e4
LT
1809
1810//
1811// get key version from on disk key - kludge
1812//
bd4c625c 1813static inline int le_key_version(const struct reiserfs_key *key)
1da177e4 1814{
bd4c625c 1815 int type;
1da177e4 1816
bd4c625c
LT
1817 type = offset_v2_k_type(&(key->u.k_offset_v2));
1818 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1819 && type != TYPE_DIRENTRY)
1820 return KEY_FORMAT_3_5;
1821
1822 return KEY_FORMAT_3_6;
1da177e4 1823
1da177e4
LT
1824}
1825
bd4c625c
LT
1826static inline void copy_key(struct reiserfs_key *to,
1827 const struct reiserfs_key *from)
1828{
1829 memcpy(to, from, KEY_SIZE);
1830}
1da177e4 1831
fec6d055
JJS
1832int comp_items(const struct item_head *stored_ih, const struct treepath *p_s_path);
1833const struct reiserfs_key *get_rkey(const struct treepath *p_s_chk_path,
a9dd3643 1834 const struct super_block *sb);
bd4c625c 1835int search_by_key(struct super_block *, const struct cpu_key *,
fec6d055 1836 struct treepath *, int);
1da177e4 1837#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
a9dd3643 1838int search_for_position_by_key(struct super_block *sb,
bd4c625c 1839 const struct cpu_key *p_s_cpu_key,
fec6d055 1840 struct treepath *p_s_search_path);
ad31a4fc 1841extern void decrement_bcount(struct buffer_head *bh);
fec6d055
JJS
1842void decrement_counters_in_path(struct treepath *p_s_search_path);
1843void pathrelse(struct treepath *p_s_search_path);
1844int reiserfs_check_path(struct treepath *p);
1845void pathrelse_and_restore(struct super_block *s, struct treepath *p_s_search_path);
bd4c625c
LT
1846
1847int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
fec6d055 1848 struct treepath *path,
bd4c625c
LT
1849 const struct cpu_key *key,
1850 struct item_head *ih,
1851 struct inode *inode, const char *body);
1852
1853int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
fec6d055 1854 struct treepath *path,
bd4c625c
LT
1855 const struct cpu_key *key,
1856 struct inode *inode,
1857 const char *body, int paste_size);
1858
1859int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
fec6d055 1860 struct treepath *path,
bd4c625c
LT
1861 struct cpu_key *key,
1862 struct inode *inode,
1863 struct page *page, loff_t new_file_size);
1864
1865int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
fec6d055 1866 struct treepath *path,
bd4c625c
LT
1867 const struct cpu_key *key,
1868 struct inode *inode, struct buffer_head *p_s_un_bh);
1869
1870void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1871 struct inode *inode, struct reiserfs_key *key);
1872int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
995c762e 1873 struct inode *inode);
bd4c625c 1874int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
995c762e 1875 struct inode *inode, struct page *,
bd4c625c 1876 int update_timestamps);
1da177e4
LT
1877
1878#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1879#define file_size(inode) ((inode)->i_size)
1880#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1881
1882#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1883!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 )
1884
bd4c625c 1885void padd_item(char *item, int total_length, int length);
1da177e4
LT
1886
1887/* inode.c */
1888/* args for the create parameter of reiserfs_get_block */
bd4c625c
LT
1889#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1890#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1891#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1892#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
1b1dcc1b 1893#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
bd4c625c
LT
1894#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
1895
bd4c625c
LT
1896void reiserfs_read_locked_inode(struct inode *inode,
1897 struct reiserfs_iget_args *args);
1898int reiserfs_find_actor(struct inode *inode, void *p);
1899int reiserfs_init_locked_inode(struct inode *inode, void *p);
1900void reiserfs_delete_inode(struct inode *inode);
1901int reiserfs_write_inode(struct inode *inode, int);
1902int reiserfs_get_block(struct inode *inode, sector_t block,
1903 struct buffer_head *bh_result, int create);
be55caf1
CH
1904struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1905 int fh_len, int fh_type);
1906struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1907 int fh_len, int fh_type);
bd4c625c
LT
1908int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1909 int connectable);
1910
1911int reiserfs_truncate_file(struct inode *, int update_timestamps);
1912void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
1913 int type, int key_length);
1914void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
1915 int version,
1916 loff_t offset, int type, int length, int entry_count);
1917struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
1918
57fe60df 1919struct reiserfs_security_handle;
bd4c625c
LT
1920int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1921 struct inode *dir, int mode,
1922 const char *symname, loff_t i_size,
57fe60df
JM
1923 struct dentry *dentry, struct inode *inode,
1924 struct reiserfs_security_handle *security);
bd4c625c
LT
1925
1926void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1927 struct inode *inode, loff_t size);
1da177e4
LT
1928
1929static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
bd4c625c 1930 struct inode *inode)
1da177e4 1931{
bd4c625c 1932 reiserfs_update_sd_size(th, inode, inode->i_size);
1da177e4
LT
1933}
1934
bd4c625c
LT
1935void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
1936void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
1da177e4
LT
1937int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1938
1939/* namei.c */
bd4c625c
LT
1940void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
1941int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
fec6d055 1942 struct treepath *path, struct reiserfs_dir_entry *de);
bd4c625c 1943struct dentry *reiserfs_get_parent(struct dentry *);
1da177e4
LT
1944/* procfs.c */
1945
1946#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1947#define REISERFS_PROC_INFO
1948#else
1949#undef REISERFS_PROC_INFO
1950#endif
1951
bd4c625c
LT
1952int reiserfs_proc_info_init(struct super_block *sb);
1953int reiserfs_proc_info_done(struct super_block *sb);
1954struct proc_dir_entry *reiserfs_proc_register_global(char *name,
1955 read_proc_t * func);
1956void reiserfs_proc_unregister_global(const char *name);
1957int reiserfs_proc_info_global_init(void);
1958int reiserfs_proc_info_global_done(void);
1959int reiserfs_global_version_in_proc(char *buffer, char **start, off_t offset,
1960 int count, int *eof, void *data);
1da177e4
LT
1961
1962#if defined( REISERFS_PROC_INFO )
1963
1964#define PROC_EXP( e ) e
1965
1966#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1967#define PROC_INFO_MAX( sb, field, value ) \
1968 __PINFO( sb ).field = \
1969 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1970#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1971#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1972#define PROC_INFO_BH_STAT( sb, bh, level ) \
1973 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1974 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1975 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1976#else
1977#define PROC_EXP( e )
1978#define VOID_V ( ( void ) 0 )
1979#define PROC_INFO_MAX( sb, field, value ) VOID_V
1980#define PROC_INFO_INC( sb, field ) VOID_V
1981#define PROC_INFO_ADD( sb, field, val ) VOID_V
ad31a4fc 1982#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
1da177e4
LT
1983#endif
1984
1985/* dir.c */
c5ef1c42
AV
1986extern const struct inode_operations reiserfs_dir_inode_operations;
1987extern const struct inode_operations reiserfs_symlink_inode_operations;
1988extern const struct inode_operations reiserfs_special_inode_operations;
4b6f5d20 1989extern const struct file_operations reiserfs_dir_operations;
a41f1a47 1990int reiserfs_readdir_dentry(struct dentry *, void *, filldir_t, loff_t *);
1da177e4
LT
1991
1992/* tail_conversion.c */
bd4c625c 1993int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
fec6d055 1994 struct treepath *, struct buffer_head *, loff_t);
bd4c625c 1995int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
fec6d055 1996 struct page *, struct treepath *, const struct cpu_key *,
bd4c625c
LT
1997 loff_t, char *);
1998void reiserfs_unmap_buffer(struct buffer_head *);
1da177e4
LT
1999
2000/* file.c */
c5ef1c42 2001extern const struct inode_operations reiserfs_file_inode_operations;
4b6f5d20 2002extern const struct file_operations reiserfs_file_operations;
f5e54d6e 2003extern const struct address_space_operations reiserfs_address_space_operations;
1da177e4
LT
2004
2005/* fix_nodes.c */
1da177e4 2006
bd4c625c
LT
2007int fix_nodes(int n_op_mode, struct tree_balance *p_s_tb,
2008 struct item_head *p_s_ins_ih, const void *);
2009void unfix_nodes(struct tree_balance *);
1da177e4
LT
2010
2011/* prints.c */
c3a9c210
JM
2012void __reiserfs_panic(struct super_block *s, const char *id,
2013 const char *function, const char *fmt, ...)
bd4c625c 2014 __attribute__ ((noreturn));
c3a9c210
JM
2015#define reiserfs_panic(s, id, fmt, args...) \
2016 __reiserfs_panic(s, id, __func__, fmt, ##args)
1e5e59d4
JM
2017void __reiserfs_error(struct super_block *s, const char *id,
2018 const char *function, const char *fmt, ...);
2019#define reiserfs_error(s, id, fmt, args...) \
2020 __reiserfs_error(s, id, __func__, fmt, ##args)
bd4c625c
LT
2021void reiserfs_info(struct super_block *s, const char *fmt, ...);
2022void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2023void print_indirect_item(struct buffer_head *bh, int item_num);
2024void store_print_tb(struct tree_balance *tb);
2025void print_cur_tb(char *mes);
2026void print_de(struct reiserfs_dir_entry *de);
2027void print_bi(struct buffer_info *bi, char *mes);
2028#define PRINT_LEAF_ITEMS 1 /* print all items */
2029#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2030#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2031void print_block(struct buffer_head *bh, ...);
2032void print_bmap(struct super_block *s, int silent);
2033void print_bmap_block(int i, char *data, int size, int silent);
1da177e4 2034/*void print_super_block (struct super_block * s, char * mes);*/
bd4c625c
LT
2035void print_objectid_map(struct super_block *s);
2036void print_block_head(struct buffer_head *bh, char *mes);
2037void check_leaf(struct buffer_head *bh);
2038void check_internal(struct buffer_head *bh);
2039void print_statistics(struct super_block *s);
2040char *reiserfs_hashname(int code);
1da177e4
LT
2041
2042/* lbalance.c */
bd4c625c
LT
2043int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2044 int mov_bytes, struct buffer_head *Snew);
2045int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2046int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2047void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2048 int del_num, int del_bytes);
2049void leaf_insert_into_buf(struct buffer_info *bi, int before,
2050 struct item_head *inserted_item_ih,
2051 const char *inserted_item_body, int zeros_number);
2052void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2053 int pos_in_item, int paste_size, const char *body,
2054 int zeros_number);
2055void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2056 int pos_in_item, int cut_size);
eba00305 2057void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
bd4c625c
LT
2058 int new_entry_count, struct reiserfs_de_head *new_dehs,
2059 const char *records, int paste_size);
1da177e4 2060/* ibalance.c */
bd4c625c
LT
2061int balance_internal(struct tree_balance *, int, int, struct item_head *,
2062 struct buffer_head **);
1da177e4
LT
2063
2064/* do_balance.c */
bd4c625c
LT
2065void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2066 struct buffer_head *bh, int flag);
1da177e4
LT
2067#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2068#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2069
bd4c625c
LT
2070void do_balance(struct tree_balance *tb, struct item_head *ih,
2071 const char *body, int flag);
2072void reiserfs_invalidate_buffer(struct tree_balance *tb,
2073 struct buffer_head *bh);
1da177e4 2074
bd4c625c
LT
2075int get_left_neighbor_position(struct tree_balance *tb, int h);
2076int get_right_neighbor_position(struct tree_balance *tb, int h);
2077void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2078 struct buffer_head *, int);
2079void make_empty_node(struct buffer_info *);
2080struct buffer_head *get_FEB(struct tree_balance *);
1da177e4
LT
2081
2082/* bitmap.c */
2083
2084/* structure contains hints for block allocator, and it is a container for
2085 * arguments, such as node, search path, transaction_handle, etc. */
bd4c625c
LT
2086struct __reiserfs_blocknr_hint {
2087 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
3ee16670 2088 sector_t block; /* file offset, in blocks */
bd4c625c 2089 struct in_core_key key;
fec6d055 2090 struct treepath *path; /* search path, used by allocator to deternine search_start by
bd4c625c
LT
2091 * various ways */
2092 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2093 * bitmap blocks changes */
2094 b_blocknr_t beg, end;
2095 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
1da177e4
LT
2096 * between different block allocator procedures
2097 * (determine_search_start() and others) */
bd4c625c
LT
2098 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2099 * function that do actual allocation */
1da177e4 2100
bd4c625c 2101 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
1da177e4 2102 * formatted/unformatted blocks with/without preallocation */
bd4c625c 2103 unsigned preallocate:1;
1da177e4
LT
2104};
2105
2106typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2107
bd4c625c
LT
2108int reiserfs_parse_alloc_options(struct super_block *, char *);
2109void reiserfs_init_alloc_options(struct super_block *s);
1da177e4
LT
2110
2111/*
2112 * given a directory, this will tell you what packing locality
2113 * to use for a new object underneat it. The locality is returned
2114 * in disk byte order (le).
2115 */
3e8962be 2116__le32 reiserfs_choose_packing(struct inode *dir);
1da177e4 2117
6f01046b
JM
2118int reiserfs_init_bitmap_cache(struct super_block *sb);
2119void reiserfs_free_bitmap_cache(struct super_block *sb);
2120void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2121struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
bd4c625c
LT
2122int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2123void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2124 b_blocknr_t, int for_unformatted);
2125int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2126 int);
9adeb1b4 2127static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
bd4c625c
LT
2128 b_blocknr_t * new_blocknrs,
2129 int amount_needed)
1da177e4 2130{
bd4c625c
LT
2131 reiserfs_blocknr_hint_t hint = {
2132 .th = tb->transaction_handle,
2133 .path = tb->tb_path,
2134 .inode = NULL,
2135 .key = tb->key,
2136 .block = 0,
2137 .formatted_node = 1
2138 };
2139 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2140 0);
1da177e4
LT
2141}
2142
9adeb1b4 2143static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
bd4c625c
LT
2144 *th, struct inode *inode,
2145 b_blocknr_t * new_blocknrs,
3ee16670
JM
2146 struct treepath *path,
2147 sector_t block)
1da177e4 2148{
bd4c625c
LT
2149 reiserfs_blocknr_hint_t hint = {
2150 .th = th,
2151 .path = path,
2152 .inode = inode,
2153 .block = block,
2154 .formatted_node = 0,
2155 .preallocate = 0
2156 };
2157 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2158}
2159
2160#ifdef REISERFS_PREALLOCATE
9adeb1b4 2161static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
bd4c625c
LT
2162 *th, struct inode *inode,
2163 b_blocknr_t * new_blocknrs,
3ee16670
JM
2164 struct treepath *path,
2165 sector_t block)
1da177e4 2166{
bd4c625c
LT
2167 reiserfs_blocknr_hint_t hint = {
2168 .th = th,
2169 .path = path,
2170 .inode = inode,
2171 .block = block,
2172 .formatted_node = 0,
2173 .preallocate = 1
2174 };
2175 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2176}
2177
bd4c625c
LT
2178void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2179 struct inode *inode);
2180void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
1da177e4 2181#endif
1da177e4
LT
2182
2183/* hashes.c */
bd4c625c
LT
2184__u32 keyed_hash(const signed char *msg, int len);
2185__u32 yura_hash(const signed char *msg, int len);
2186__u32 r5_hash(const signed char *msg, int len);
1da177e4
LT
2187
2188/* the ext2 bit routines adjust for big or little endian as
2189** appropriate for the arch, so in our laziness we use them rather
2190** than using the bit routines they call more directly. These
2191** routines must be used when changing on disk bitmaps. */
2192#define reiserfs_test_and_set_le_bit ext2_set_bit
2193#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2194#define reiserfs_test_le_bit ext2_test_bit
2195#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2196
2197/* sometimes reiserfs_truncate may require to allocate few new blocks
2198 to perform indirect2direct conversion. People probably used to
2199 think, that truncate should work without problems on a filesystem
2200 without free disk space. They may complain that they can not
2201 truncate due to lack of free disk space. This spare space allows us
2202 to not worry about it. 500 is probably too much, but it should be
2203 absolutely safe */
2204#define SPARE_SPACE 500
2205
1da177e4 2206/* prototypes from ioctl.c */
bd4c625c
LT
2207int reiserfs_ioctl(struct inode *inode, struct file *filp,
2208 unsigned int cmd, unsigned long arg);
52b499c4
DH
2209long reiserfs_compat_ioctl(struct file *filp,
2210 unsigned int cmd, unsigned long arg);
d5dee5c3 2211int reiserfs_unpack(struct inode *inode, struct file *filp);
bd4c625c 2212
1da177e4
LT
2213/* ioctl's command */
2214#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
2215/* define following flags to be the same as in ext2, so that chattr(1),
2216 lsattr(1) will work with us. */
36695673
DH
2217#define REISERFS_IOC_GETFLAGS FS_IOC_GETFLAGS
2218#define REISERFS_IOC_SETFLAGS FS_IOC_SETFLAGS
2219#define REISERFS_IOC_GETVERSION FS_IOC_GETVERSION
2220#define REISERFS_IOC_SETVERSION FS_IOC_SETVERSION
1da177e4 2221
52b499c4
DH
2222/* the 32 bit compat definitions with int argument */
2223#define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
2224#define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
2225#define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
2226#define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
2227#define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
2228
1da177e4
LT
2229/* Locking primitives */
2230/* Right now we are still falling back to (un)lock_kernel, but eventually that
2231 would evolve into real per-fs locks */
2232#define reiserfs_write_lock( sb ) lock_kernel()
2233#define reiserfs_write_unlock( sb ) unlock_kernel()
bd4c625c 2234
bd4c625c 2235#endif /* _LINUX_REISER_FS_H */