]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/reiserfs_fs.h
Use helpers to obtain task pid in printks
[mirror_ubuntu-artful-kernel.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
31/*
32 * include/linux/reiser_fs.h
33 *
34 * Reiser File System constants and structures
35 *
36 */
37
38/* in reading the #defines, it may help to understand that they employ
39 the following abbreviations:
40
41 B = Buffer
42 I = Item header
43 H = Height within the tree (should be changed to LEV)
44 N = Number of the item in the node
45 STAT = stat data
46 DEH = Directory Entry Header
47 EC = Entry Count
48 E = Entry number
49 UL = Unsigned Long
50 BLKH = BLocK Header
51 UNFM = UNForMatted node
52 DC = Disk Child
53 P = Path
54
55 These #defines are named by concatenating these abbreviations,
56 where first comes the arguments, and last comes the return value,
57 of the macro.
58
59*/
60
61#define USE_INODE_GENERATION_COUNTER
62
63#define REISERFS_PREALLOCATE
64#define DISPLACE_NEW_PACKING_LOCALITIES
65#define PREALLOCATION_SIZE 9
66
67/* n must be power of 2 */
68#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
69
70// to be ok for alpha and others we have to align structures to 8 byte
71// boundary.
72// FIXME: do not change 4 by anything else: there is code which relies on that
73#define ROUND_UP(x) _ROUND_UP(x,8LL)
74
75/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
76** messages.
77*/
bd4c625c 78#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
1da177e4 79
bd4c625c 80void reiserfs_warning(struct super_block *s, const char *fmt, ...);
1da177e4
LT
81/* assertions handling */
82
83/** always check a condition and panic if it's false. */
2d954d06 84#define __RASSERT( cond, scond, format, args... ) \
1da177e4 85if( !( cond ) ) \
2d954d06 86 reiserfs_panic( NULL, "reiserfs[%i]: assertion " scond " failed at " \
1da177e4 87 __FILE__ ":%i:%s: " format "\n", \
ba25f9dc 88 in_interrupt() ? -1 : task_pid_nr(current), __LINE__ , __FUNCTION__ , ##args )
1da177e4 89
2d954d06
AV
90#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
91
1da177e4 92#if defined( CONFIG_REISERFS_CHECK )
2d954d06 93#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
1da177e4
LT
94#else
95#define RFALSE( cond, format, args... ) do {;} while( 0 )
96#endif
97
98#define CONSTF __attribute_const__
99/*
100 * Disk Data Structures
101 */
102
103/***************************************************************************/
104/* SUPER BLOCK */
105/***************************************************************************/
106
107/*
108 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
109 * the version in RAM is part of a larger structure containing fields never written to disk.
110 */
bd4c625c
LT
111#define UNSET_HASH 0 // read_super will guess about, what hash names
112 // in directories were sorted with
1da177e4
LT
113#define TEA_HASH 1
114#define YURA_HASH 2
115#define R5_HASH 3
116#define DEFAULT_HASH R5_HASH
117
1da177e4 118struct journal_params {
bd4c625c
LT
119 __le32 jp_journal_1st_block; /* where does journal start from on its
120 * device */
121 __le32 jp_journal_dev; /* journal device st_rdev */
122 __le32 jp_journal_size; /* size of the journal */
123 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
124 __le32 jp_journal_magic; /* random value made on fs creation (this
125 * was sb_journal_block_count) */
126 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
127 * trans */
128 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
129 * commit be */
130 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
131 * be */
1da177e4
LT
132};
133
134/* this is the super from 3.5.X, where X >= 10 */
bd4c625c
LT
135struct reiserfs_super_block_v1 {
136 __le32 s_block_count; /* blocks count */
137 __le32 s_free_blocks; /* free blocks count */
138 __le32 s_root_block; /* root block number */
139 struct journal_params s_journal;
140 __le16 s_blocksize; /* block size */
141 __le16 s_oid_maxsize; /* max size of object id array, see
142 * get_objectid() commentary */
143 __le16 s_oid_cursize; /* current size of object id array */
144 __le16 s_umount_state; /* this is set to 1 when filesystem was
145 * umounted, to 2 - when not */
146 char s_magic[10]; /* reiserfs magic string indicates that
147 * file system is reiserfs:
148 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
149 __le16 s_fs_state; /* it is set to used by fsck to mark which
150 * phase of rebuilding is done */
151 __le32 s_hash_function_code; /* indicate, what hash function is being use
152 * to sort names in a directory*/
153 __le16 s_tree_height; /* height of disk tree */
154 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
155 * each block of file system */
156 __le16 s_version; /* this field is only reliable on filesystem
157 * with non-standard journal */
158 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
159 * device, we need to keep after
160 * making fs with non-standard journal */
1da177e4
LT
161} __attribute__ ((__packed__));
162
163#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
164
165/* this is the on disk super block */
bd4c625c
LT
166struct reiserfs_super_block {
167 struct reiserfs_super_block_v1 s_v1;
168 __le32 s_inode_generation;
169 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
170 unsigned char s_uuid[16]; /* filesystem unique identifier */
171 unsigned char s_label[16]; /* filesystem volume label */
172 char s_unused[88]; /* zero filled by mkreiserfs and
173 * reiserfs_convert_objectid_map_v1()
174 * so any additions must be updated
175 * there as well. */
176} __attribute__ ((__packed__));
1da177e4
LT
177
178#define SB_SIZE (sizeof(struct reiserfs_super_block))
179
180#define REISERFS_VERSION_1 0
181#define REISERFS_VERSION_2 2
182
1da177e4
LT
183// on-disk super block fields converted to cpu form
184#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
185#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
186#define SB_BLOCKSIZE(s) \
187 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
188#define SB_BLOCK_COUNT(s) \
189 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
190#define SB_FREE_BLOCKS(s) \
191 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
192#define SB_REISERFS_MAGIC(s) \
193 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
194#define SB_ROOT_BLOCK(s) \
195 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
196#define SB_TREE_HEIGHT(s) \
197 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
198#define SB_REISERFS_STATE(s) \
199 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
200#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
201#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
202
203#define PUT_SB_BLOCK_COUNT(s, val) \
204 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
205#define PUT_SB_FREE_BLOCKS(s, val) \
206 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
207#define PUT_SB_ROOT_BLOCK(s, val) \
208 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
209#define PUT_SB_TREE_HEIGHT(s, val) \
210 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
211#define PUT_SB_REISERFS_STATE(s, val) \
bd4c625c 212 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1da177e4
LT
213#define PUT_SB_VERSION(s, val) \
214 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
215#define PUT_SB_BMAP_NR(s, val) \
216 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
217
1da177e4
LT
218#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
219#define SB_ONDISK_JOURNAL_SIZE(s) \
220 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
221#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
222 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
223#define SB_ONDISK_JOURNAL_DEVICE(s) \
224 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
225#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
b8cc936f 226 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1da177e4
LT
227
228#define is_block_in_log_or_reserved_area(s, block) \
229 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
230 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
231 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
bd4c625c 232 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1da177e4 233
bd4c625c
LT
234int is_reiserfs_3_5(struct reiserfs_super_block *rs);
235int is_reiserfs_3_6(struct reiserfs_super_block *rs);
236int is_reiserfs_jr(struct reiserfs_super_block *rs);
1da177e4
LT
237
238/* ReiserFS leaves the first 64k unused, so that partition labels have
239 enough space. If someone wants to write a fancy bootloader that
240 needs more than 64k, let us know, and this will be increased in size.
241 This number must be larger than than the largest block size on any
242 platform, or code will break. -Hans */
243#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
244#define REISERFS_FIRST_BLOCK unused_define
245#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
246
247/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
248#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
249
250// reiserfs internal error code (used by search_by_key adn fix_nodes))
251#define CARRY_ON 0
252#define REPEAT_SEARCH -1
253#define IO_ERROR -2
254#define NO_DISK_SPACE -3
255#define NO_BALANCING_NEEDED (-4)
256#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
257#define QUOTA_EXCEEDED -6
258
259typedef __u32 b_blocknr_t;
3e8962be 260typedef __le32 unp_t;
1da177e4
LT
261
262struct unfm_nodeinfo {
bd4c625c
LT
263 unp_t unfm_nodenum;
264 unsigned short unfm_freespace;
1da177e4
LT
265};
266
267/* there are two formats of keys: 3.5 and 3.6
268 */
269#define KEY_FORMAT_3_5 0
270#define KEY_FORMAT_3_6 1
271
272/* there are two stat datas */
273#define STAT_DATA_V1 0
274#define STAT_DATA_V2 1
275
1da177e4
LT
276static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
277{
278 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
279}
280
281static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
282{
283 return sb->s_fs_info;
284}
285
cb680c1b
JM
286/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
287 * which overflows on large file systems. */
288static inline u32 reiserfs_bmap_count(struct super_block *sb)
289{
290 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
291}
292
293static inline int bmap_would_wrap(unsigned bmap_nr)
294{
295 return bmap_nr > ((1LL << 16) - 1);
296}
297
1da177e4
LT
298/** this says about version of key of all items (but stat data) the
299 object consists of */
300#define get_inode_item_key_version( inode ) \
301 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
302
303#define set_inode_item_key_version( inode, version ) \
304 ({ if((version)==KEY_FORMAT_3_6) \
305 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
306 else \
307 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
308
309#define get_inode_sd_version(inode) \
310 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
311
312#define set_inode_sd_version(inode, version) \
313 ({ if((version)==STAT_DATA_V2) \
314 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
315 else \
316 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
317
318/* This is an aggressive tail suppression policy, I am hoping it
319 improves our benchmarks. The principle behind it is that percentage
320 space saving is what matters, not absolute space saving. This is
321 non-intuitive, but it helps to understand it if you consider that the
322 cost to access 4 blocks is not much more than the cost to access 1
323 block, if you have to do a seek and rotate. A tail risks a
324 non-linear disk access that is significant as a percentage of total
325 time cost for a 4 block file and saves an amount of space that is
326 less significant as a percentage of space, or so goes the hypothesis.
327 -Hans */
328#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
329(\
330 (!(n_tail_size)) || \
331 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
332 ( (n_file_size) >= (n_block_size) * 4 ) || \
333 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
334 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
335 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
336 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
337 ( ( (n_file_size) >= (n_block_size) ) && \
338 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
339)
340
341/* Another strategy for tails, this one means only create a tail if all the
342 file would fit into one DIRECT item.
343 Primary intention for this one is to increase performance by decreasing
344 seeking.
bd4c625c 345*/
1da177e4
LT
346#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
347(\
348 (!(n_tail_size)) || \
349 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
350)
351
1da177e4
LT
352/*
353 * values for s_umount_state field
354 */
355#define REISERFS_VALID_FS 1
356#define REISERFS_ERROR_FS 2
357
358//
359// there are 5 item types currently
360//
361#define TYPE_STAT_DATA 0
362#define TYPE_INDIRECT 1
363#define TYPE_DIRECT 2
bd4c625c
LT
364#define TYPE_DIRENTRY 3
365#define TYPE_MAXTYPE 3
366#define TYPE_ANY 15 // FIXME: comment is required
1da177e4
LT
367
368/***************************************************************************/
369/* KEY & ITEM HEAD */
370/***************************************************************************/
371
372//
373// directories use this key as well as old files
374//
375struct offset_v1 {
bd4c625c
LT
376 __le32 k_offset;
377 __le32 k_uniqueness;
1da177e4
LT
378} __attribute__ ((__packed__));
379
380struct offset_v2 {
f8e08a84 381 __le64 v;
1da177e4
LT
382} __attribute__ ((__packed__));
383
bd4c625c 384static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1da177e4 385{
f8e08a84 386 __u8 type = le64_to_cpu(v2->v) >> 60;
bd4c625c 387 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1da177e4 388}
bd4c625c
LT
389
390static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1da177e4 391{
bd4c625c
LT
392 v2->v =
393 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1da177e4 394}
bd4c625c
LT
395
396static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1da177e4 397{
bd4c625c 398 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1da177e4
LT
399}
400
bd4c625c
LT
401static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
402{
403 offset &= (~0ULL >> 4);
404 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1da177e4 405}
1da177e4
LT
406
407/* Key of an item determines its location in the S+tree, and
408 is composed of 4 components */
409struct reiserfs_key {
bd4c625c
LT
410 __le32 k_dir_id; /* packing locality: by default parent
411 directory object id */
412 __le32 k_objectid; /* object identifier */
413 union {
414 struct offset_v1 k_offset_v1;
415 struct offset_v2 k_offset_v2;
416 } __attribute__ ((__packed__)) u;
1da177e4
LT
417} __attribute__ ((__packed__));
418
6a3a16f2 419struct in_core_key {
bd4c625c
LT
420 __u32 k_dir_id; /* packing locality: by default parent
421 directory object id */
422 __u32 k_objectid; /* object identifier */
423 __u64 k_offset;
424 __u8 k_type;
6b9f5829 425};
1da177e4
LT
426
427struct cpu_key {
bd4c625c
LT
428 struct in_core_key on_disk_key;
429 int version;
430 int key_length; /* 3 in all cases but direct2indirect and
431 indirect2direct conversion */
1da177e4
LT
432};
433
434/* Our function for comparing keys can compare keys of different
435 lengths. It takes as a parameter the length of the keys it is to
436 compare. These defines are used in determining what is to be passed
437 to it as that parameter. */
438#define REISERFS_FULL_KEY_LEN 4
439#define REISERFS_SHORT_KEY_LEN 2
440
441/* The result of the key compare */
442#define FIRST_GREATER 1
443#define SECOND_GREATER -1
444#define KEYS_IDENTICAL 0
445#define KEY_FOUND 1
446#define KEY_NOT_FOUND 0
447
448#define KEY_SIZE (sizeof(struct reiserfs_key))
449#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
450
451/* return values for search_by_key and clones */
452#define ITEM_FOUND 1
453#define ITEM_NOT_FOUND 0
454#define ENTRY_FOUND 1
455#define ENTRY_NOT_FOUND 0
456#define DIRECTORY_NOT_FOUND -1
457#define REGULAR_FILE_FOUND -2
458#define DIRECTORY_FOUND -3
459#define BYTE_FOUND 1
460#define BYTE_NOT_FOUND 0
461#define FILE_NOT_FOUND -1
462
463#define POSITION_FOUND 1
464#define POSITION_NOT_FOUND 0
465
466// return values for reiserfs_find_entry and search_by_entry_key
467#define NAME_FOUND 1
468#define NAME_NOT_FOUND 0
469#define GOTO_PREVIOUS_ITEM 2
470#define NAME_FOUND_INVISIBLE 3
471
472/* Everything in the filesystem is stored as a set of items. The
473 item head contains the key of the item, its free space (for
474 indirect items) and specifies the location of the item itself
475 within the block. */
476
bd4c625c 477struct item_head {
1da177e4
LT
478 /* Everything in the tree is found by searching for it based on
479 * its key.*/
480 struct reiserfs_key ih_key;
481 union {
482 /* The free space in the last unformatted node of an
483 indirect item if this is an indirect item. This
484 equals 0xFFFF iff this is a direct item or stat data
485 item. Note that the key, not this field, is used to
486 determine the item type, and thus which field this
487 union contains. */
3e8962be 488 __le16 ih_free_space_reserved;
1da177e4
LT
489 /* Iff this is a directory item, this field equals the
490 number of directory entries in the directory item. */
3e8962be 491 __le16 ih_entry_count;
1da177e4 492 } __attribute__ ((__packed__)) u;
bd4c625c
LT
493 __le16 ih_item_len; /* total size of the item body */
494 __le16 ih_item_location; /* an offset to the item body
495 * within the block */
496 __le16 ih_version; /* 0 for all old items, 2 for new
497 ones. Highest bit is set by fsck
498 temporary, cleaned after all
499 done */
1da177e4
LT
500} __attribute__ ((__packed__));
501/* size of item header */
502#define IH_SIZE (sizeof(struct item_head))
503
504#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
505#define ih_version(ih) le16_to_cpu((ih)->ih_version)
506#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
507#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
508#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
509
510#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
511#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
512#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
513#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
514#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
515
1da177e4
LT
516#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
517
518#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
519#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
520
521/* these operate on indirect items, where you've got an array of ints
522** at a possibly unaligned location. These are a noop on ia32
523**
524** p is the array of __u32, i is the index into the array, v is the value
525** to store there.
526*/
527#define get_block_num(p, i) le32_to_cpu(get_unaligned((p) + (i)))
528#define put_block_num(p, i, v) put_unaligned(cpu_to_le32(v), (p) + (i))
529
530//
531// in old version uniqueness field shows key type
532//
533#define V1_SD_UNIQUENESS 0
534#define V1_INDIRECT_UNIQUENESS 0xfffffffe
535#define V1_DIRECT_UNIQUENESS 0xffffffff
536#define V1_DIRENTRY_UNIQUENESS 500
bd4c625c 537#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
1da177e4
LT
538
539//
540// here are conversion routines
541//
bd4c625c
LT
542static inline int uniqueness2type(__u32 uniqueness) CONSTF;
543static inline int uniqueness2type(__u32 uniqueness)
1da177e4 544{
bd4c625c
LT
545 switch ((int)uniqueness) {
546 case V1_SD_UNIQUENESS:
547 return TYPE_STAT_DATA;
548 case V1_INDIRECT_UNIQUENESS:
549 return TYPE_INDIRECT;
550 case V1_DIRECT_UNIQUENESS:
551 return TYPE_DIRECT;
552 case V1_DIRENTRY_UNIQUENESS:
553 return TYPE_DIRENTRY;
554 default:
555 reiserfs_warning(NULL, "vs-500: unknown uniqueness %d",
556 uniqueness);
1da177e4 557 case V1_ANY_UNIQUENESS:
bd4c625c
LT
558 return TYPE_ANY;
559 }
1da177e4
LT
560}
561
bd4c625c
LT
562static inline __u32 type2uniqueness(int type) CONSTF;
563static inline __u32 type2uniqueness(int type)
1da177e4 564{
bd4c625c
LT
565 switch (type) {
566 case TYPE_STAT_DATA:
567 return V1_SD_UNIQUENESS;
568 case TYPE_INDIRECT:
569 return V1_INDIRECT_UNIQUENESS;
570 case TYPE_DIRECT:
571 return V1_DIRECT_UNIQUENESS;
572 case TYPE_DIRENTRY:
573 return V1_DIRENTRY_UNIQUENESS;
574 default:
575 reiserfs_warning(NULL, "vs-501: unknown type %d", type);
1da177e4 576 case TYPE_ANY:
bd4c625c
LT
577 return V1_ANY_UNIQUENESS;
578 }
1da177e4
LT
579}
580
581//
582// key is pointer to on disk key which is stored in le, result is cpu,
583// there is no way to get version of object from key, so, provide
584// version to these defines
585//
bd4c625c
LT
586static inline loff_t le_key_k_offset(int version,
587 const struct reiserfs_key *key)
1da177e4 588{
bd4c625c
LT
589 return (version == KEY_FORMAT_3_5) ?
590 le32_to_cpu(key->u.k_offset_v1.k_offset) :
591 offset_v2_k_offset(&(key->u.k_offset_v2));
1da177e4
LT
592}
593
bd4c625c 594static inline loff_t le_ih_k_offset(const struct item_head *ih)
1da177e4 595{
bd4c625c 596 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1da177e4
LT
597}
598
bd4c625c 599static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1da177e4 600{
bd4c625c
LT
601 return (version == KEY_FORMAT_3_5) ?
602 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
603 offset_v2_k_type(&(key->u.k_offset_v2));
1da177e4
LT
604}
605
bd4c625c 606static inline loff_t le_ih_k_type(const struct item_head *ih)
1da177e4 607{
bd4c625c 608 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1da177e4
LT
609}
610
bd4c625c
LT
611static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
612 loff_t offset)
1da177e4 613{
bd4c625c
LT
614 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
615 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
1da177e4
LT
616}
617
bd4c625c 618static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1da177e4 619{
bd4c625c 620 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1da177e4
LT
621}
622
bd4c625c
LT
623static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
624 int type)
1da177e4 625{
bd4c625c
LT
626 (version == KEY_FORMAT_3_5) ?
627 (void)(key->u.k_offset_v1.k_uniqueness =
628 cpu_to_le32(type2uniqueness(type)))
629 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
1da177e4 630}
bd4c625c 631static inline void set_le_ih_k_type(struct item_head *ih, int type)
1da177e4 632{
bd4c625c 633 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1da177e4
LT
634}
635
1da177e4
LT
636#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
637#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
638#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
639#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
640
641//
642// item header has version.
643//
644#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
645#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
646#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
647#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
648
1da177e4
LT
649//
650// key is pointer to cpu key, result is cpu
651//
bd4c625c 652static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1da177e4 653{
bd4c625c 654 return key->on_disk_key.k_offset;
1da177e4
LT
655}
656
bd4c625c 657static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1da177e4 658{
bd4c625c 659 return key->on_disk_key.k_type;
1da177e4
LT
660}
661
bd4c625c 662static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1da177e4 663{
6b9f5829 664 key->on_disk_key.k_offset = offset;
1da177e4
LT
665}
666
bd4c625c 667static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1da177e4 668{
6b9f5829 669 key->on_disk_key.k_type = type;
1da177e4
LT
670}
671
bd4c625c 672static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1da177e4 673{
bd4c625c 674 key->on_disk_key.k_offset--;
1da177e4
LT
675}
676
1da177e4
LT
677#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
678#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
679#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
680#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
681
1da177e4
LT
682/* are these used ? */
683#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
684#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
685#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
686#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
687
1da177e4
LT
688#define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
689 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
690 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
691
bd4c625c 692/* maximal length of item */
1da177e4
LT
693#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
694#define MIN_ITEM_LEN 1
695
1da177e4
LT
696/* object identifier for root dir */
697#define REISERFS_ROOT_OBJECTID 2
698#define REISERFS_ROOT_PARENT_OBJECTID 1
699extern struct reiserfs_key root_key;
700
1da177e4
LT
701/*
702 * Picture represents a leaf of the S+tree
703 * ______________________________________________________
704 * | | Array of | | |
705 * |Block | Object-Item | F r e e | Objects- |
706 * | head | Headers | S p a c e | Items |
707 * |______|_______________|___________________|___________|
708 */
709
710/* Header of a disk block. More precisely, header of a formatted leaf
711 or internal node, and not the header of an unformatted node. */
bd4c625c
LT
712struct block_head {
713 __le16 blk_level; /* Level of a block in the tree. */
714 __le16 blk_nr_item; /* Number of keys/items in a block. */
715 __le16 blk_free_space; /* Block free space in bytes. */
716 __le16 blk_reserved;
717 /* dump this in v4/planA */
718 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
1da177e4
LT
719};
720
721#define BLKH_SIZE (sizeof(struct block_head))
722#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
723#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
724#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
725#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
726#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
727#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
728#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
729#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
730#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
731#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
732
733/*
734 * values for blk_level field of the struct block_head
735 */
736
bd4c625c
LT
737#define FREE_LEVEL 0 /* when node gets removed from the tree its
738 blk_level is set to FREE_LEVEL. It is then
739 used to see whether the node is still in the
740 tree */
1da177e4 741
bd4c625c 742#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1da177e4
LT
743
744/* Given the buffer head of a formatted node, resolve to the block head of that node. */
745#define B_BLK_HEAD(p_s_bh) ((struct block_head *)((p_s_bh)->b_data))
746/* Number of items that are in buffer. */
747#define B_NR_ITEMS(p_s_bh) (blkh_nr_item(B_BLK_HEAD(p_s_bh)))
748#define B_LEVEL(p_s_bh) (blkh_level(B_BLK_HEAD(p_s_bh)))
749#define B_FREE_SPACE(p_s_bh) (blkh_free_space(B_BLK_HEAD(p_s_bh)))
750
751#define PUT_B_NR_ITEMS(p_s_bh,val) do { set_blkh_nr_item(B_BLK_HEAD(p_s_bh),val); } while (0)
752#define PUT_B_LEVEL(p_s_bh,val) do { set_blkh_level(B_BLK_HEAD(p_s_bh),val); } while (0)
753#define PUT_B_FREE_SPACE(p_s_bh,val) do { set_blkh_free_space(B_BLK_HEAD(p_s_bh),val); } while (0)
754
1da177e4 755/* Get right delimiting key. -- little endian */
5296c7be 756#define B_PRIGHT_DELIM_KEY(p_s_bh) (&(blk_right_delim_key(B_BLK_HEAD(p_s_bh))))
1da177e4
LT
757
758/* Does the buffer contain a disk leaf. */
759#define B_IS_ITEMS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) == DISK_LEAF_NODE_LEVEL)
760
761/* Does the buffer contain a disk internal node */
762#define B_IS_KEYS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) > DISK_LEAF_NODE_LEVEL \
763 && B_LEVEL(p_s_bh) <= MAX_HEIGHT)
764
1da177e4
LT
765/***************************************************************************/
766/* STAT DATA */
767/***************************************************************************/
768
1da177e4
LT
769//
770// old stat data is 32 bytes long. We are going to distinguish new one by
771// different size
772//
bd4c625c
LT
773struct stat_data_v1 {
774 __le16 sd_mode; /* file type, permissions */
775 __le16 sd_nlink; /* number of hard links */
776 __le16 sd_uid; /* owner */
777 __le16 sd_gid; /* group */
778 __le32 sd_size; /* file size */
779 __le32 sd_atime; /* time of last access */
780 __le32 sd_mtime; /* time file was last modified */
781 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
782 union {
783 __le32 sd_rdev;
784 __le32 sd_blocks; /* number of blocks file uses */
785 } __attribute__ ((__packed__)) u;
786 __le32 sd_first_direct_byte; /* first byte of file which is stored
787 in a direct item: except that if it
788 equals 1 it is a symlink and if it
789 equals ~(__u32)0 there is no
790 direct item. The existence of this
791 field really grates on me. Let's
792 replace it with a macro based on
793 sd_size and our tail suppression
794 policy. Someday. -Hans */
1da177e4
LT
795} __attribute__ ((__packed__));
796
797#define SD_V1_SIZE (sizeof(struct stat_data_v1))
798#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
799#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
800#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
801#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
802#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
803#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
804#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
805#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
806#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
807#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
808#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
809#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
810#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
811#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
812#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
813#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
814#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
815#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
816#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
817#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
818#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
819#define sd_v1_first_direct_byte(sdp) \
820 (le32_to_cpu((sdp)->sd_first_direct_byte))
821#define set_sd_v1_first_direct_byte(sdp,v) \
822 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
823
1da177e4
LT
824/* inode flags stored in sd_attrs (nee sd_reserved) */
825
826/* we want common flags to have the same values as in ext2,
827 so chattr(1) will work without problems */
36695673
DH
828#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
829#define REISERFS_APPEND_FL FS_APPEND_FL
830#define REISERFS_SYNC_FL FS_SYNC_FL
831#define REISERFS_NOATIME_FL FS_NOATIME_FL
832#define REISERFS_NODUMP_FL FS_NODUMP_FL
833#define REISERFS_SECRM_FL FS_SECRM_FL
834#define REISERFS_UNRM_FL FS_UNRM_FL
835#define REISERFS_COMPR_FL FS_COMPR_FL
836#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1da177e4
LT
837
838/* persistent flags that file inherits from the parent directory */
839#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
840 REISERFS_SYNC_FL | \
841 REISERFS_NOATIME_FL | \
842 REISERFS_NODUMP_FL | \
843 REISERFS_SECRM_FL | \
844 REISERFS_COMPR_FL | \
845 REISERFS_NOTAIL_FL )
846
847/* Stat Data on disk (reiserfs version of UFS disk inode minus the
848 address blocks) */
849struct stat_data {
bd4c625c
LT
850 __le16 sd_mode; /* file type, permissions */
851 __le16 sd_attrs; /* persistent inode flags */
852 __le32 sd_nlink; /* number of hard links */
853 __le64 sd_size; /* file size */
854 __le32 sd_uid; /* owner */
855 __le32 sd_gid; /* group */
856 __le32 sd_atime; /* time of last access */
857 __le32 sd_mtime; /* time file was last modified */
858 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
859 __le32 sd_blocks;
860 union {
861 __le32 sd_rdev;
862 __le32 sd_generation;
863 //__le32 sd_first_direct_byte;
864 /* first byte of file which is stored in a
865 direct item: except that if it equals 1
866 it is a symlink and if it equals
867 ~(__u32)0 there is no direct item. The
868 existence of this field really grates
869 on me. Let's replace it with a macro
870 based on sd_size and our tail
871 suppression policy? */
872 } __attribute__ ((__packed__)) u;
1da177e4
LT
873} __attribute__ ((__packed__));
874//
875// this is 44 bytes long
876//
877#define SD_SIZE (sizeof(struct stat_data))
878#define SD_V2_SIZE SD_SIZE
879#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
880#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
881#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
882/* sd_reserved */
883/* set_sd_reserved */
884#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
885#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
886#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
887#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
888#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
889#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
890#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
891#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
892#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
893#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
894#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
895#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
896#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
897#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
898#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
899#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
900#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
901#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
902#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
903#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
904#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
905#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
906
1da177e4
LT
907/***************************************************************************/
908/* DIRECTORY STRUCTURE */
909/***************************************************************************/
910/*
911 Picture represents the structure of directory items
912 ________________________________________________
913 | Array of | | | | | |
914 | directory |N-1| N-2 | .... | 1st |0th|
915 | entry headers | | | | | |
916 |_______________|___|_____|________|_______|___|
917 <---- directory entries ------>
918
919 First directory item has k_offset component 1. We store "." and ".."
920 in one item, always, we never split "." and ".." into differing
921 items. This makes, among other things, the code for removing
922 directories simpler. */
923#define SD_OFFSET 0
924#define SD_UNIQUENESS 0
925#define DOT_OFFSET 1
926#define DOT_DOT_OFFSET 2
927#define DIRENTRY_UNIQUENESS 500
928
929/* */
930#define FIRST_ITEM_OFFSET 1
931
932/*
933 Q: How to get key of object pointed to by entry from entry?
934
935 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
936 of object, entry points to */
937
938/* NOT IMPLEMENTED:
939 Directory will someday contain stat data of object */
940
bd4c625c
LT
941struct reiserfs_de_head {
942 __le32 deh_offset; /* third component of the directory entry key */
943 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
944 by directory entry */
945 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
946 __le16 deh_location; /* offset of name in the whole item */
947 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
948 entry is hidden (unlinked) */
1da177e4
LT
949} __attribute__ ((__packed__));
950#define DEH_SIZE sizeof(struct reiserfs_de_head)
951#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
952#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
953#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
954#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
955#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
956
957#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
958#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
959#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
960#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
961#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
962
963/* empty directory contains two entries "." and ".." and their headers */
964#define EMPTY_DIR_SIZE \
965(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
966
967/* old format directories have this size when empty */
968#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
969
bd4c625c 970#define DEH_Statdata 0 /* not used now */
1da177e4
LT
971#define DEH_Visible 2
972
973/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
974#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
975# define ADDR_UNALIGNED_BITS (3)
976#endif
977
978/* These are only used to manipulate deh_state.
979 * Because of this, we'll use the ext2_ bit routines,
980 * since they are little endian */
981#ifdef ADDR_UNALIGNED_BITS
982
983# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
984# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
985
986# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
987# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
988# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
989
990#else
991
992# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
993# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
994# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
995
996#endif
997
998#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
999#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1000#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1001#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1002
1003#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1004#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1005#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1006
bd4c625c
LT
1007extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1008 __le32 par_dirid, __le32 par_objid);
1009extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1010 __le32 par_dirid, __le32 par_objid);
1da177e4
LT
1011
1012/* array of the entry headers */
1013 /* get item body */
1014#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1015#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1016
1017/* length of the directory entry in directory item. This define
1018 calculates length of i-th directory entry using directory entry
1019 locations from dir entry head. When it calculates length of 0-th
1020 directory entry, it uses length of whole item in place of entry
1021 location of the non-existent following entry in the calculation.
1022 See picture above.*/
1023/*
1024#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1025((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1026*/
bd4c625c
LT
1027static inline int entry_length(const struct buffer_head *bh,
1028 const struct item_head *ih, int pos_in_item)
1da177e4 1029{
bd4c625c 1030 struct reiserfs_de_head *deh;
1da177e4 1031
bd4c625c
LT
1032 deh = B_I_DEH(bh, ih) + pos_in_item;
1033 if (pos_in_item)
1034 return deh_location(deh - 1) - deh_location(deh);
1da177e4 1035
bd4c625c 1036 return ih_item_len(ih) - deh_location(deh);
1da177e4
LT
1037}
1038
1da177e4
LT
1039/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1040#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1041
1da177e4
LT
1042/* name by bh, ih and entry_num */
1043#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))))
1044
1045// two entries per block (at least)
1046#define REISERFS_MAX_NAME(block_size) 255
1047
1da177e4
LT
1048/* this structure is used for operations on directory entries. It is
1049 not a disk structure. */
1050/* When reiserfs_find_entry or search_by_entry_key find directory
1051 entry, they return filled reiserfs_dir_entry structure */
bd4c625c
LT
1052struct reiserfs_dir_entry {
1053 struct buffer_head *de_bh;
1054 int de_item_num;
1055 struct item_head *de_ih;
1056 int de_entry_num;
1057 struct reiserfs_de_head *de_deh;
1058 int de_entrylen;
1059 int de_namelen;
1060 char *de_name;
3af1efe8 1061 unsigned long *de_gen_number_bit_string;
bd4c625c
LT
1062
1063 __u32 de_dir_id;
1064 __u32 de_objectid;
1065
1066 struct cpu_key de_entry_key;
1da177e4 1067};
bd4c625c 1068
1da177e4
LT
1069/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1070
1071/* pointer to file name, stored in entry */
1072#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1073
1074/* length of name */
1075#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1076(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1077
1da177e4
LT
1078/* hash value occupies bits from 7 up to 30 */
1079#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1080/* generation number occupies 7 bits starting from 0 up to 6 */
1081#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1082#define MAX_GENERATION_NUMBER 127
1083
1084#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1085
1da177e4
LT
1086/*
1087 * Picture represents an internal node of the reiserfs tree
1088 * ______________________________________________________
1089 * | | Array of | Array of | Free |
1090 * |block | keys | pointers | space |
1091 * | head | N | N+1 | |
1092 * |______|_______________|___________________|___________|
1093 */
1094
1095/***************************************************************************/
1096/* DISK CHILD */
1097/***************************************************************************/
1098/* Disk child pointer: The pointer from an internal node of the tree
1099 to a node that is on disk. */
1100struct disk_child {
bd4c625c
LT
1101 __le32 dc_block_number; /* Disk child's block number. */
1102 __le16 dc_size; /* Disk child's used space. */
1103 __le16 dc_reserved;
1da177e4
LT
1104};
1105
1106#define DC_SIZE (sizeof(struct disk_child))
1107#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1108#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1109#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1110#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1111
1112/* Get disk child by buffer header and position in the tree node. */
1113#define B_N_CHILD(p_s_bh,n_pos) ((struct disk_child *)\
1114((p_s_bh)->b_data+BLKH_SIZE+B_NR_ITEMS(p_s_bh)*KEY_SIZE+DC_SIZE*(n_pos)))
1115
1116/* Get disk child number by buffer header and position in the tree node. */
1117#define B_N_CHILD_NUM(p_s_bh,n_pos) (dc_block_number(B_N_CHILD(p_s_bh,n_pos)))
1118#define PUT_B_N_CHILD_NUM(p_s_bh,n_pos, val) (put_dc_block_number(B_N_CHILD(p_s_bh,n_pos), val ))
1119
bd4c625c 1120 /* maximal value of field child_size in structure disk_child */
1da177e4
LT
1121 /* child size is the combined size of all items and their headers */
1122#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1123
1124/* amount of used space in buffer (not including block head) */
1125#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1126
1127/* max and min number of keys in internal node */
1128#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1129#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1130
1131/***************************************************************************/
1132/* PATH STRUCTURES AND DEFINES */
1133/***************************************************************************/
1134
1da177e4
LT
1135/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1136 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1137 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1138 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1139 position of the block_number of the next node if it is looking through an internal node. If it
1140 is looking through a leaf node bin_search will find the position of the item which has key either
1141 equal to given key, or which is the maximal key less than the given key. */
1142
bd4c625c
LT
1143struct path_element {
1144 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1145 int pe_position; /* Position in the tree node which is placed in the */
1146 /* buffer above. */
1da177e4
LT
1147};
1148
bd4c625c
LT
1149#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1150#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1151#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1da177e4 1152
bd4c625c
LT
1153#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1154#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1da177e4
LT
1155
1156/* We need to keep track of who the ancestors of nodes are. When we
1157 perform a search we record which nodes were visited while
1158 descending the tree looking for the node we searched for. This list
1159 of nodes is called the path. This information is used while
1160 performing balancing. Note that this path information may become
1161 invalid, and this means we must check it when using it to see if it
1162 is still valid. You'll need to read search_by_key and the comments
1163 in it, especially about decrement_counters_in_path(), to understand
1164 this structure.
1165
1166Paths make the code so much harder to work with and debug.... An
1167enormous number of bugs are due to them, and trying to write or modify
1168code that uses them just makes my head hurt. They are based on an
1169excessive effort to avoid disturbing the precious VFS code.:-( The
1170gods only know how we are going to SMP the code that uses them.
1171znodes are the way! */
1172
bd4c625c
LT
1173#define PATH_READA 0x1 /* do read ahead */
1174#define PATH_READA_BACK 0x2 /* read backwards */
1da177e4 1175
fec6d055 1176struct treepath {
bd4c625c
LT
1177 int path_length; /* Length of the array above. */
1178 int reada;
1179 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1180 int pos_in_item;
1da177e4
LT
1181};
1182
1183#define pos_in_item(path) ((path)->pos_in_item)
1184
1185#define INITIALIZE_PATH(var) \
fec6d055 1186struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1da177e4
LT
1187
1188/* Get path element by path and path position. */
1189#define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1190
1191/* Get buffer header at the path by path and path position. */
1192#define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1193
1194/* Get position in the element at the path by path and path position. */
1195#define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1196
1da177e4
LT
1197#define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1198 /* you know, to the person who didn't
bd4c625c
LT
1199 write this the macro name does not
1200 at first suggest what it does.
1201 Maybe POSITION_FROM_PATH_END? Or
1202 maybe we should just focus on
1203 dumping paths... -Hans */
1da177e4
LT
1204#define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1205
1da177e4
LT
1206#define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1207
1208/* in do_balance leaf has h == 0 in contrast with path structure,
1209 where root has level == 0. That is why we need these defines */
1210#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
1211#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1212#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1213#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1da177e4
LT
1214
1215#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1216
1217#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1218#define get_ih(path) PATH_PITEM_HEAD(path)
1219#define get_item_pos(path) PATH_LAST_POSITION(path)
1220#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1221#define item_moved(ih,path) comp_items(ih, path)
1222#define path_changed(ih,path) comp_items (ih, path)
1223
1da177e4
LT
1224/***************************************************************************/
1225/* MISC */
1226/***************************************************************************/
1227
1228/* Size of pointer to the unformatted node. */
1229#define UNFM_P_SIZE (sizeof(unp_t))
1230#define UNFM_P_SHIFT 2
1231
1232// in in-core inode key is stored on le form
1233#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1234
1235#define MAX_UL_INT 0xffffffff
1236#define MAX_INT 0x7ffffff
1237#define MAX_US_INT 0xffff
1238
1239// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1240#define U32_MAX (~(__u32)0)
1241
bd4c625c 1242static inline loff_t max_reiserfs_offset(struct inode *inode)
1da177e4 1243{
bd4c625c
LT
1244 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1245 return (loff_t) U32_MAX;
1da177e4 1246
bd4c625c 1247 return (loff_t) ((~(__u64) 0) >> 4);
1da177e4
LT
1248}
1249
1da177e4
LT
1250/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1251#define MAX_KEY_OBJECTID MAX_UL_INT
1252
1da177e4
LT
1253#define MAX_B_NUM MAX_UL_INT
1254#define MAX_FC_NUM MAX_US_INT
1255
1da177e4
LT
1256/* the purpose is to detect overflow of an unsigned short */
1257#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1258
1da177e4 1259/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
bd4c625c
LT
1260#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1261#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1da177e4
LT
1262
1263#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1264#define get_generation(s) atomic_read (&fs_generation(s))
1265#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1266#define __fs_changed(gen,s) (gen != get_generation (s))
1267#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1268
1da177e4
LT
1269/***************************************************************************/
1270/* FIXATE NODES */
1271/***************************************************************************/
1272
1273#define VI_TYPE_LEFT_MERGEABLE 1
1274#define VI_TYPE_RIGHT_MERGEABLE 2
1275
1276/* To make any changes in the tree we always first find node, that
1277 contains item to be changed/deleted or place to insert a new
1278 item. We call this node S. To do balancing we need to decide what
1279 we will shift to left/right neighbor, or to a new node, where new
1280 item will be etc. To make this analysis simpler we build virtual
1281 node. Virtual node is an array of items, that will replace items of
1282 node S. (For instance if we are going to delete an item, virtual
1283 node does not contain it). Virtual node keeps information about
1284 item sizes and types, mergeability of first and last items, sizes
1285 of all entries in directory item. We use this array of items when
1286 calculating what we can shift to neighbors and how many nodes we
1287 have to have if we do not any shiftings, if we shift to left/right
1288 neighbor or to both. */
bd4c625c
LT
1289struct virtual_item {
1290 int vi_index; // index in the array of item operations
1291 unsigned short vi_type; // left/right mergeability
1292 unsigned short vi_item_len; /* length of item that it will have after balancing */
1293 struct item_head *vi_ih;
1294 const char *vi_item; // body of item (old or new)
1295 const void *vi_new_data; // 0 always but paste mode
1296 void *vi_uarea; // item specific area
1da177e4
LT
1297};
1298
bd4c625c
LT
1299struct virtual_node {
1300 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
1301 unsigned short vn_nr_item; /* number of items in virtual node */
1302 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1303 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1304 short vn_affected_item_num;
1305 short vn_pos_in_item;
1306 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
1307 const void *vn_data;
1308 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1da177e4
LT
1309};
1310
1311/* used by directory items when creating virtual nodes */
1312struct direntry_uarea {
bd4c625c
LT
1313 int flags;
1314 __u16 entry_count;
1315 __u16 entry_sizes[1];
1316} __attribute__ ((__packed__));
1da177e4
LT
1317
1318/***************************************************************************/
1319/* TREE BALANCE */
1320/***************************************************************************/
1321
1322/* This temporary structure is used in tree balance algorithms, and
1323 constructed as we go to the extent that its various parts are
1324 needed. It contains arrays of nodes that can potentially be
1325 involved in the balancing of node S, and parameters that define how
1326 each of the nodes must be balanced. Note that in these algorithms
1327 for balancing the worst case is to need to balance the current node
1328 S and the left and right neighbors and all of their parents plus
1329 create a new node. We implement S1 balancing for the leaf nodes
1330 and S0 balancing for the internal nodes (S1 and S0 are defined in
1331 our papers.)*/
1332
1333#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1334
1335/* maximum number of FEB blocknrs on a single level */
1336#define MAX_AMOUNT_NEEDED 2
1337
1338/* someday somebody will prefix every field in this struct with tb_ */
bd4c625c
LT
1339struct tree_balance {
1340 int tb_mode;
1341 int need_balance_dirty;
1342 struct super_block *tb_sb;
1343 struct reiserfs_transaction_handle *transaction_handle;
fec6d055 1344 struct treepath *tb_path;
bd4c625c
LT
1345 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1346 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
1347 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1348 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1349 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1350 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1351
1352 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1353 cur_blknum. */
1354 struct buffer_head *used[MAX_FEB_SIZE];
1355 struct buffer_head *thrown[MAX_FEB_SIZE];
1356 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1357 shifted to the left in order to balance the
1358 current node; for leaves includes item that
1359 will be partially shifted; for internal
1360 nodes, it is the number of child pointers
1361 rather than items. It includes the new item
1362 being created. The code sometimes subtracts
1363 one to get the number of wholly shifted
1364 items for other purposes. */
1365 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1366 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1367 S[h] to its item number within the node CFL[h] */
1368 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1369 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1370 S[h]. A negative value means removing. */
1371 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1372 balancing on the level h of the tree. If 0 then S is
1373 being deleted, if 1 then S is remaining and no new nodes
1374 are being created, if 2 or 3 then 1 or 2 new nodes is
1375 being created */
1376
1377 /* fields that are used only for balancing leaves of the tree */
1378 int cur_blknum; /* number of empty blocks having been already allocated */
1379 int s0num; /* number of items that fall into left most node when S[0] splits */
1380 int s1num; /* number of items that fall into first new node when S[0] splits */
1381 int s2num; /* number of items that fall into second new node when S[0] splits */
1382 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1383 /* most liquid item that cannot be shifted from S[0] entirely */
1384 /* if -1 then nothing will be partially shifted */
1385 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1386 /* most liquid item that cannot be shifted from S[0] entirely */
1387 /* if -1 then nothing will be partially shifted */
1388 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1389 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1390 int s2bytes;
1391 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1392 char *vn_buf; /* kmalloced memory. Used to create
1da177e4
LT
1393 virtual node and keep map of
1394 dirtied bitmap blocks */
bd4c625c
LT
1395 int vn_buf_size; /* size of the vn_buf */
1396 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
1da177e4 1397
bd4c625c
LT
1398 int fs_gen; /* saved value of `reiserfs_generation' counter
1399 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1da177e4 1400#ifdef DISPLACE_NEW_PACKING_LOCALITIES
bd4c625c
LT
1401 struct in_core_key key; /* key pointer, to pass to block allocator or
1402 another low-level subsystem */
1da177e4 1403#endif
bd4c625c 1404};
1da177e4
LT
1405
1406/* These are modes of balancing */
1407
1408/* When inserting an item. */
1409#define M_INSERT 'i'
1410/* When inserting into (directories only) or appending onto an already
1411 existant item. */
1412#define M_PASTE 'p'
1413/* When deleting an item. */
1414#define M_DELETE 'd'
1415/* When truncating an item or removing an entry from a (directory) item. */
1416#define M_CUT 'c'
1417
1418/* used when balancing on leaf level skipped (in reiserfsck) */
1419#define M_INTERNAL 'n'
1420
1421/* When further balancing is not needed, then do_balance does not need
1422 to be called. */
1423#define M_SKIP_BALANCING 's'
1424#define M_CONVERT 'v'
1425
1426/* modes of leaf_move_items */
1427#define LEAF_FROM_S_TO_L 0
1428#define LEAF_FROM_S_TO_R 1
1429#define LEAF_FROM_R_TO_L 2
1430#define LEAF_FROM_L_TO_R 3
1431#define LEAF_FROM_S_TO_SNEW 4
1432
1433#define FIRST_TO_LAST 0
1434#define LAST_TO_FIRST 1
1435
1436/* used in do_balance for passing parent of node information that has
1437 been gotten from tb struct */
1438struct buffer_info {
bd4c625c
LT
1439 struct tree_balance *tb;
1440 struct buffer_head *bi_bh;
1441 struct buffer_head *bi_parent;
1442 int bi_position;
1da177e4
LT
1443};
1444
1da177e4
LT
1445/* there are 4 types of items: stat data, directory item, indirect, direct.
1446+-------------------+------------+--------------+------------+
1447| | k_offset | k_uniqueness | mergeable? |
1448+-------------------+------------+--------------+------------+
1449| stat data | 0 | 0 | no |
1450+-------------------+------------+--------------+------------+
1451| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1452| non 1st directory | hash value | | yes |
1453| item | | | |
1454+-------------------+------------+--------------+------------+
1455| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1456+-------------------+------------+--------------+------------+
1457| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1458+-------------------+------------+--------------+------------+
1459*/
1460
1461struct item_operations {
bd4c625c
LT
1462 int (*bytes_number) (struct item_head * ih, int block_size);
1463 void (*decrement_key) (struct cpu_key *);
1464 int (*is_left_mergeable) (struct reiserfs_key * ih,
1465 unsigned long bsize);
1466 void (*print_item) (struct item_head *, char *item);
1467 void (*check_item) (struct item_head *, char *item);
1468
1469 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1470 int is_affected, int insert_size);
1471 int (*check_left) (struct virtual_item * vi, int free,
1472 int start_skip, int end_skip);
1473 int (*check_right) (struct virtual_item * vi, int free);
1474 int (*part_size) (struct virtual_item * vi, int from, int to);
1475 int (*unit_num) (struct virtual_item * vi);
1476 void (*print_vi) (struct virtual_item * vi);
1da177e4
LT
1477};
1478
bd4c625c 1479extern struct item_operations *item_ops[TYPE_ANY + 1];
1da177e4
LT
1480
1481#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1482#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1483#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1484#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1485#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)
1486#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1487#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1488#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1489#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1490#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1491
1da177e4
LT
1492#define COMP_SHORT_KEYS comp_short_keys
1493
1494/* number of blocks pointed to by the indirect item */
1495#define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1496
1497/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1498#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1499
1500/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1501
bd4c625c 1502/* get the item header */
1da177e4
LT
1503#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1504
1505/* get key */
1506#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1507
1508/* get the key */
1509#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1510
1511/* get item body */
1512#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1513
1514/* get the stat data by the buffer header and the item order */
1515#define B_N_STAT_DATA(bh,nr) \
1516( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1517
1518 /* following defines use reiserfs buffer header and item header */
1519
1520/* get stat-data */
1521#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1522
1523// this is 3976 for size==4096
1524#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1525
1526/* indirect items consist of entries which contain blocknrs, pos
1527 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1528 blocknr contained by the entry pos points to */
1529#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1530#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)
1531
1532struct reiserfs_iget_args {
bd4c625c
LT
1533 __u32 objectid;
1534 __u32 dirid;
1535};
1da177e4
LT
1536
1537/***************************************************************************/
1538/* FUNCTION DECLARATIONS */
1539/***************************************************************************/
1540
1541/*#ifdef __KERNEL__*/
1542#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1543
1544#define journal_trans_half(blocksize) \
1545 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1546
1547/* journal.c see journal.c for all the comments here */
1548
1549/* first block written in a commit. */
1550struct reiserfs_journal_desc {
bd4c625c
LT
1551 __le32 j_trans_id; /* id of commit */
1552 __le32 j_len; /* length of commit. len +1 is the commit block */
1553 __le32 j_mount_id; /* mount id of this trans */
1554 __le32 j_realblock[1]; /* real locations for each block */
1555};
1da177e4
LT
1556
1557#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1558#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1559#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1560
1561#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1562#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1563#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1564
1565/* last block written in a commit */
1566struct reiserfs_journal_commit {
bd4c625c
LT
1567 __le32 j_trans_id; /* must match j_trans_id from the desc block */
1568 __le32 j_len; /* ditto */
1569 __le32 j_realblock[1]; /* real locations for each block */
1570};
1da177e4
LT
1571
1572#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1573#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1574#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1575
1576#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1577#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1578
1579/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1580** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1581** and this transaction does not need to be replayed.
1582*/
1583struct reiserfs_journal_header {
bd4c625c
LT
1584 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
1585 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
1586 __le32 j_mount_id;
1587 /* 12 */ struct journal_params jh_journal;
1588};
1da177e4
LT
1589
1590/* biggest tunable defines are right here */
bd4c625c
LT
1591#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1592#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1da177e4 1593#define JOURNAL_TRANS_MIN_DEFAULT 256
bd4c625c 1594#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1da177e4 1595#define JOURNAL_MIN_RATIO 2
bd4c625c 1596#define JOURNAL_MAX_COMMIT_AGE 30
1da177e4
LT
1597#define JOURNAL_MAX_TRANS_AGE 30
1598#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1599#ifdef CONFIG_QUOTA
556a2a45
JK
1600/* We need to update data and inode (atime) */
1601#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? 2 : 0)
1602/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1603#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1604(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1605/* same as with INIT */
1606#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1607(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
1da177e4 1608#else
556a2a45
JK
1609#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1610#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1611#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
1da177e4
LT
1612#endif
1613
1614/* both of these can be as low as 1, or as high as you want. The min is the
1615** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1616** as needed, and released when transactions are committed. On release, if
1617** the current number of nodes is > max, the node is freed, otherwise,
1618** it is put on a free list for faster use later.
1619*/
bd4c625c
LT
1620#define REISERFS_MIN_BITMAP_NODES 10
1621#define REISERFS_MAX_BITMAP_NODES 100
1da177e4 1622
bd4c625c 1623#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1da177e4
LT
1624#define JBH_HASH_MASK 8191
1625
1626#define _jhashfn(sb,block) \
1627 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1628 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1629#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1630
1631// We need these to make journal.c code more readable
1632#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1633#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1634#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1635
1636enum reiserfs_bh_state_bits {
bd4c625c
LT
1637 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1638 BH_JDirty_wait,
1639 BH_JNew, /* disk block was taken off free list before
1640 * being in a finished transaction, or
1641 * written to disk. Can be reused immed. */
1642 BH_JPrepared,
1643 BH_JRestore_dirty,
1644 BH_JTest, // debugging only will go away
1da177e4
LT
1645};
1646
1647BUFFER_FNS(JDirty, journaled);
1648TAS_BUFFER_FNS(JDirty, journaled);
1649BUFFER_FNS(JDirty_wait, journal_dirty);
1650TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1651BUFFER_FNS(JNew, journal_new);
1652TAS_BUFFER_FNS(JNew, journal_new);
1653BUFFER_FNS(JPrepared, journal_prepared);
1654TAS_BUFFER_FNS(JPrepared, journal_prepared);
1655BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1656TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1657BUFFER_FNS(JTest, journal_test);
1658TAS_BUFFER_FNS(JTest, journal_test);
1659
1660/*
1661** transaction handle which is passed around for all journal calls
1662*/
1663struct reiserfs_transaction_handle {
bd4c625c
LT
1664 struct super_block *t_super; /* super for this FS when journal_begin was
1665 called. saves calls to reiserfs_get_super
1666 also used by nested transactions to make
1667 sure they are nesting on the right FS
1668 _must_ be first in the handle
1669 */
1670 int t_refcount;
1671 int t_blocks_logged; /* number of blocks this writer has logged */
1672 int t_blocks_allocated; /* number of blocks this writer allocated */
1673 unsigned long t_trans_id; /* sanity check, equals the current trans id */
1674 void *t_handle_save; /* save existing current->journal_info */
1675 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1676 should be displaced from others */
1677 struct list_head t_list;
1678};
1da177e4
LT
1679
1680/* used to keep track of ordered and tail writes, attached to the buffer
1681 * head through b_journal_head.
1682 */
1683struct reiserfs_jh {
bd4c625c
LT
1684 struct reiserfs_journal_list *jl;
1685 struct buffer_head *bh;
1686 struct list_head list;
1da177e4
LT
1687};
1688
1689void reiserfs_free_jh(struct buffer_head *bh);
1690int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1691int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
bd4c625c
LT
1692int journal_mark_dirty(struct reiserfs_transaction_handle *,
1693 struct super_block *, struct buffer_head *bh);
1694
1695static inline int reiserfs_file_data_log(struct inode *inode)
1696{
1697 if (reiserfs_data_log(inode->i_sb) ||
1698 (REISERFS_I(inode)->i_flags & i_data_log))
1699 return 1;
1700 return 0;
1da177e4
LT
1701}
1702
bd4c625c
LT
1703static inline int reiserfs_transaction_running(struct super_block *s)
1704{
1705 struct reiserfs_transaction_handle *th = current->journal_info;
1706 if (th && th->t_super == s)
1707 return 1;
1708 if (th && th->t_super == NULL)
1709 BUG();
1710 return 0;
1da177e4
LT
1711}
1712
23f9e0f8
AZ
1713static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
1714{
1715 return th->t_blocks_allocated - th->t_blocks_logged;
1716}
1717
bd4c625c
LT
1718struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1719 super_block
1720 *,
1721 int count);
1da177e4
LT
1722int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1723int reiserfs_commit_page(struct inode *inode, struct page *page,
bd4c625c 1724 unsigned from, unsigned to);
1da177e4 1725int reiserfs_flush_old_commits(struct super_block *);
bd4c625c
LT
1726int reiserfs_commit_for_inode(struct inode *);
1727int reiserfs_inode_needs_commit(struct inode *);
1728void reiserfs_update_inode_transaction(struct inode *);
1729void reiserfs_wait_on_write_block(struct super_block *s);
1730void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1731void reiserfs_allow_writes(struct super_block *s);
1732void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1733int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1734 int wait);
1735void reiserfs_restore_prepared_buffer(struct super_block *,
1736 struct buffer_head *bh);
1737int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1738 unsigned int);
1739int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1740int journal_release_error(struct reiserfs_transaction_handle *,
1741 struct super_block *);
1742int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1743 unsigned long);
1744int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1745 unsigned long);
1746int journal_mark_freed(struct reiserfs_transaction_handle *,
1747 struct super_block *, b_blocknr_t blocknr);
1748int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
3ee16670
JM
1749int reiserfs_in_journal(struct super_block *p_s_sb, unsigned int bmap_nr,
1750 int bit_nr, int searchall, b_blocknr_t *next);
bd4c625c
LT
1751int journal_begin(struct reiserfs_transaction_handle *,
1752 struct super_block *p_s_sb, unsigned long);
1753int journal_join_abort(struct reiserfs_transaction_handle *,
1754 struct super_block *p_s_sb, unsigned long);
1755void reiserfs_journal_abort(struct super_block *sb, int errno);
1756void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1757int reiserfs_allocate_list_bitmaps(struct super_block *s,
3ee16670 1758 struct reiserfs_list_bitmap *, unsigned int);
bd4c625c
LT
1759
1760void add_save_link(struct reiserfs_transaction_handle *th,
1761 struct inode *inode, int truncate);
1762int remove_save_link(struct inode *inode, int truncate);
1da177e4
LT
1763
1764/* objectid.c */
bd4c625c
LT
1765__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1766void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1767 __u32 objectid_to_release);
1768int reiserfs_convert_objectid_map_v1(struct super_block *);
1da177e4
LT
1769
1770/* stree.c */
1771int B_IS_IN_TREE(const struct buffer_head *);
bd4c625c
LT
1772extern void copy_item_head(struct item_head *p_v_to,
1773 const struct item_head *p_v_from);
1da177e4
LT
1774
1775// first key is in cpu form, second - le
bd4c625c
LT
1776extern int comp_short_keys(const struct reiserfs_key *le_key,
1777 const struct cpu_key *cpu_key);
1778extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
1da177e4
LT
1779
1780// both are in le form
bd4c625c
LT
1781extern int comp_le_keys(const struct reiserfs_key *,
1782 const struct reiserfs_key *);
1783extern int comp_short_le_keys(const struct reiserfs_key *,
1784 const struct reiserfs_key *);
1da177e4
LT
1785
1786//
1787// get key version from on disk key - kludge
1788//
bd4c625c 1789static inline int le_key_version(const struct reiserfs_key *key)
1da177e4 1790{
bd4c625c 1791 int type;
1da177e4 1792
bd4c625c
LT
1793 type = offset_v2_k_type(&(key->u.k_offset_v2));
1794 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1795 && type != TYPE_DIRENTRY)
1796 return KEY_FORMAT_3_5;
1797
1798 return KEY_FORMAT_3_6;
1da177e4 1799
1da177e4
LT
1800}
1801
bd4c625c
LT
1802static inline void copy_key(struct reiserfs_key *to,
1803 const struct reiserfs_key *from)
1804{
1805 memcpy(to, from, KEY_SIZE);
1806}
1da177e4 1807
fec6d055
JJS
1808int comp_items(const struct item_head *stored_ih, const struct treepath *p_s_path);
1809const struct reiserfs_key *get_rkey(const struct treepath *p_s_chk_path,
bd4c625c
LT
1810 const struct super_block *p_s_sb);
1811int search_by_key(struct super_block *, const struct cpu_key *,
fec6d055 1812 struct treepath *, int);
1da177e4 1813#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
bd4c625c
LT
1814int search_for_position_by_key(struct super_block *p_s_sb,
1815 const struct cpu_key *p_s_cpu_key,
fec6d055 1816 struct treepath *p_s_search_path);
bd4c625c 1817extern void decrement_bcount(struct buffer_head *p_s_bh);
fec6d055
JJS
1818void decrement_counters_in_path(struct treepath *p_s_search_path);
1819void pathrelse(struct treepath *p_s_search_path);
1820int reiserfs_check_path(struct treepath *p);
1821void pathrelse_and_restore(struct super_block *s, struct treepath *p_s_search_path);
bd4c625c
LT
1822
1823int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
fec6d055 1824 struct treepath *path,
bd4c625c
LT
1825 const struct cpu_key *key,
1826 struct item_head *ih,
1827 struct inode *inode, const char *body);
1828
1829int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
fec6d055 1830 struct treepath *path,
bd4c625c
LT
1831 const struct cpu_key *key,
1832 struct inode *inode,
1833 const char *body, int paste_size);
1834
1835int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
fec6d055 1836 struct treepath *path,
bd4c625c
LT
1837 struct cpu_key *key,
1838 struct inode *inode,
1839 struct page *page, loff_t new_file_size);
1840
1841int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
fec6d055 1842 struct treepath *path,
bd4c625c
LT
1843 const struct cpu_key *key,
1844 struct inode *inode, struct buffer_head *p_s_un_bh);
1845
1846void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1847 struct inode *inode, struct reiserfs_key *key);
1848int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
1849 struct inode *p_s_inode);
1850int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
1851 struct inode *p_s_inode, struct page *,
1852 int update_timestamps);
1da177e4
LT
1853
1854#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1855#define file_size(inode) ((inode)->i_size)
1856#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1857
1858#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1859!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 )
1860
bd4c625c 1861void padd_item(char *item, int total_length, int length);
1da177e4
LT
1862
1863/* inode.c */
1864/* args for the create parameter of reiserfs_get_block */
bd4c625c
LT
1865#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1866#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1867#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1868#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
1b1dcc1b 1869#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
bd4c625c
LT
1870#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
1871
bd4c625c
LT
1872void reiserfs_read_locked_inode(struct inode *inode,
1873 struct reiserfs_iget_args *args);
1874int reiserfs_find_actor(struct inode *inode, void *p);
1875int reiserfs_init_locked_inode(struct inode *inode, void *p);
1876void reiserfs_delete_inode(struct inode *inode);
1877int reiserfs_write_inode(struct inode *inode, int);
1878int reiserfs_get_block(struct inode *inode, sector_t block,
1879 struct buffer_head *bh_result, int create);
1880struct dentry *reiserfs_get_dentry(struct super_block *, void *);
1881struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 * data,
1882 int len, int fhtype,
1883 int (*acceptable) (void *contect,
1884 struct dentry * de),
1885 void *context);
1886int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1887 int connectable);
1888
1889int reiserfs_truncate_file(struct inode *, int update_timestamps);
1890void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
1891 int type, int key_length);
1892void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
1893 int version,
1894 loff_t offset, int type, int length, int entry_count);
1895struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
1896
1897int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1898 struct inode *dir, int mode,
1899 const char *symname, loff_t i_size,
1900 struct dentry *dentry, struct inode *inode);
1901
1902void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1903 struct inode *inode, loff_t size);
1da177e4
LT
1904
1905static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
bd4c625c 1906 struct inode *inode)
1da177e4 1907{
bd4c625c 1908 reiserfs_update_sd_size(th, inode, inode->i_size);
1da177e4
LT
1909}
1910
bd4c625c
LT
1911void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
1912void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
1da177e4
LT
1913int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1914
1915/* namei.c */
bd4c625c
LT
1916void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
1917int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
fec6d055 1918 struct treepath *path, struct reiserfs_dir_entry *de);
bd4c625c 1919struct dentry *reiserfs_get_parent(struct dentry *);
1da177e4
LT
1920/* procfs.c */
1921
1922#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1923#define REISERFS_PROC_INFO
1924#else
1925#undef REISERFS_PROC_INFO
1926#endif
1927
bd4c625c
LT
1928int reiserfs_proc_info_init(struct super_block *sb);
1929int reiserfs_proc_info_done(struct super_block *sb);
1930struct proc_dir_entry *reiserfs_proc_register_global(char *name,
1931 read_proc_t * func);
1932void reiserfs_proc_unregister_global(const char *name);
1933int reiserfs_proc_info_global_init(void);
1934int reiserfs_proc_info_global_done(void);
1935int reiserfs_global_version_in_proc(char *buffer, char **start, off_t offset,
1936 int count, int *eof, void *data);
1da177e4
LT
1937
1938#if defined( REISERFS_PROC_INFO )
1939
1940#define PROC_EXP( e ) e
1941
1942#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1943#define PROC_INFO_MAX( sb, field, value ) \
1944 __PINFO( sb ).field = \
1945 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1946#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1947#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1948#define PROC_INFO_BH_STAT( sb, bh, level ) \
1949 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1950 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1951 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1952#else
1953#define PROC_EXP( e )
1954#define VOID_V ( ( void ) 0 )
1955#define PROC_INFO_MAX( sb, field, value ) VOID_V
1956#define PROC_INFO_INC( sb, field ) VOID_V
1957#define PROC_INFO_ADD( sb, field, val ) VOID_V
1958#define PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level ) VOID_V
1959#endif
1960
1961/* dir.c */
c5ef1c42
AV
1962extern const struct inode_operations reiserfs_dir_inode_operations;
1963extern const struct inode_operations reiserfs_symlink_inode_operations;
1964extern const struct inode_operations reiserfs_special_inode_operations;
4b6f5d20 1965extern const struct file_operations reiserfs_dir_operations;
1da177e4
LT
1966
1967/* tail_conversion.c */
bd4c625c 1968int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
fec6d055 1969 struct treepath *, struct buffer_head *, loff_t);
bd4c625c 1970int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
fec6d055 1971 struct page *, struct treepath *, const struct cpu_key *,
bd4c625c
LT
1972 loff_t, char *);
1973void reiserfs_unmap_buffer(struct buffer_head *);
1da177e4
LT
1974
1975/* file.c */
c5ef1c42 1976extern const struct inode_operations reiserfs_file_inode_operations;
4b6f5d20 1977extern const struct file_operations reiserfs_file_operations;
f5e54d6e 1978extern const struct address_space_operations reiserfs_address_space_operations;
1da177e4
LT
1979
1980/* fix_nodes.c */
1da177e4 1981
bd4c625c
LT
1982int fix_nodes(int n_op_mode, struct tree_balance *p_s_tb,
1983 struct item_head *p_s_ins_ih, const void *);
1984void unfix_nodes(struct tree_balance *);
1da177e4
LT
1985
1986/* prints.c */
bd4c625c
LT
1987void reiserfs_panic(struct super_block *s, const char *fmt, ...)
1988 __attribute__ ((noreturn));
1989void reiserfs_info(struct super_block *s, const char *fmt, ...);
1990void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
1991void print_indirect_item(struct buffer_head *bh, int item_num);
1992void store_print_tb(struct tree_balance *tb);
1993void print_cur_tb(char *mes);
1994void print_de(struct reiserfs_dir_entry *de);
1995void print_bi(struct buffer_info *bi, char *mes);
1996#define PRINT_LEAF_ITEMS 1 /* print all items */
1997#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
1998#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
1999void print_block(struct buffer_head *bh, ...);
2000void print_bmap(struct super_block *s, int silent);
2001void print_bmap_block(int i, char *data, int size, int silent);
1da177e4 2002/*void print_super_block (struct super_block * s, char * mes);*/
bd4c625c
LT
2003void print_objectid_map(struct super_block *s);
2004void print_block_head(struct buffer_head *bh, char *mes);
2005void check_leaf(struct buffer_head *bh);
2006void check_internal(struct buffer_head *bh);
2007void print_statistics(struct super_block *s);
2008char *reiserfs_hashname(int code);
1da177e4
LT
2009
2010/* lbalance.c */
bd4c625c
LT
2011int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2012 int mov_bytes, struct buffer_head *Snew);
2013int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2014int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2015void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2016 int del_num, int del_bytes);
2017void leaf_insert_into_buf(struct buffer_info *bi, int before,
2018 struct item_head *inserted_item_ih,
2019 const char *inserted_item_body, int zeros_number);
2020void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2021 int pos_in_item, int paste_size, const char *body,
2022 int zeros_number);
2023void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2024 int pos_in_item, int cut_size);
2025void leaf_paste_entries(struct buffer_head *bh, int item_num, int before,
2026 int new_entry_count, struct reiserfs_de_head *new_dehs,
2027 const char *records, int paste_size);
1da177e4 2028/* ibalance.c */
bd4c625c
LT
2029int balance_internal(struct tree_balance *, int, int, struct item_head *,
2030 struct buffer_head **);
1da177e4
LT
2031
2032/* do_balance.c */
bd4c625c
LT
2033void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2034 struct buffer_head *bh, int flag);
1da177e4
LT
2035#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2036#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2037
bd4c625c
LT
2038void do_balance(struct tree_balance *tb, struct item_head *ih,
2039 const char *body, int flag);
2040void reiserfs_invalidate_buffer(struct tree_balance *tb,
2041 struct buffer_head *bh);
1da177e4 2042
bd4c625c
LT
2043int get_left_neighbor_position(struct tree_balance *tb, int h);
2044int get_right_neighbor_position(struct tree_balance *tb, int h);
2045void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2046 struct buffer_head *, int);
2047void make_empty_node(struct buffer_info *);
2048struct buffer_head *get_FEB(struct tree_balance *);
1da177e4
LT
2049
2050/* bitmap.c */
2051
2052/* structure contains hints for block allocator, and it is a container for
2053 * arguments, such as node, search path, transaction_handle, etc. */
bd4c625c
LT
2054struct __reiserfs_blocknr_hint {
2055 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
3ee16670 2056 sector_t block; /* file offset, in blocks */
bd4c625c 2057 struct in_core_key key;
fec6d055 2058 struct treepath *path; /* search path, used by allocator to deternine search_start by
bd4c625c
LT
2059 * various ways */
2060 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2061 * bitmap blocks changes */
2062 b_blocknr_t beg, end;
2063 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
1da177e4
LT
2064 * between different block allocator procedures
2065 * (determine_search_start() and others) */
bd4c625c
LT
2066 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2067 * function that do actual allocation */
1da177e4 2068
bd4c625c 2069 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
1da177e4 2070 * formatted/unformatted blocks with/without preallocation */
bd4c625c 2071 unsigned preallocate:1;
1da177e4
LT
2072};
2073
2074typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2075
bd4c625c
LT
2076int reiserfs_parse_alloc_options(struct super_block *, char *);
2077void reiserfs_init_alloc_options(struct super_block *s);
1da177e4
LT
2078
2079/*
2080 * given a directory, this will tell you what packing locality
2081 * to use for a new object underneat it. The locality is returned
2082 * in disk byte order (le).
2083 */
3e8962be 2084__le32 reiserfs_choose_packing(struct inode *dir);
1da177e4 2085
6f01046b
JM
2086int reiserfs_init_bitmap_cache(struct super_block *sb);
2087void reiserfs_free_bitmap_cache(struct super_block *sb);
2088void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2089struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
bd4c625c
LT
2090int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2091void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2092 b_blocknr_t, int for_unformatted);
2093int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2094 int);
9adeb1b4 2095static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
bd4c625c
LT
2096 b_blocknr_t * new_blocknrs,
2097 int amount_needed)
1da177e4 2098{
bd4c625c
LT
2099 reiserfs_blocknr_hint_t hint = {
2100 .th = tb->transaction_handle,
2101 .path = tb->tb_path,
2102 .inode = NULL,
2103 .key = tb->key,
2104 .block = 0,
2105 .formatted_node = 1
2106 };
2107 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2108 0);
1da177e4
LT
2109}
2110
9adeb1b4 2111static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
bd4c625c
LT
2112 *th, struct inode *inode,
2113 b_blocknr_t * new_blocknrs,
3ee16670
JM
2114 struct treepath *path,
2115 sector_t block)
1da177e4 2116{
bd4c625c
LT
2117 reiserfs_blocknr_hint_t hint = {
2118 .th = th,
2119 .path = path,
2120 .inode = inode,
2121 .block = block,
2122 .formatted_node = 0,
2123 .preallocate = 0
2124 };
2125 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2126}
2127
2128#ifdef REISERFS_PREALLOCATE
9adeb1b4 2129static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
bd4c625c
LT
2130 *th, struct inode *inode,
2131 b_blocknr_t * new_blocknrs,
3ee16670
JM
2132 struct treepath *path,
2133 sector_t block)
1da177e4 2134{
bd4c625c
LT
2135 reiserfs_blocknr_hint_t hint = {
2136 .th = th,
2137 .path = path,
2138 .inode = inode,
2139 .block = block,
2140 .formatted_node = 0,
2141 .preallocate = 1
2142 };
2143 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2144}
2145
bd4c625c
LT
2146void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2147 struct inode *inode);
2148void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
1da177e4 2149#endif
1da177e4
LT
2150
2151/* hashes.c */
bd4c625c
LT
2152__u32 keyed_hash(const signed char *msg, int len);
2153__u32 yura_hash(const signed char *msg, int len);
2154__u32 r5_hash(const signed char *msg, int len);
1da177e4
LT
2155
2156/* the ext2 bit routines adjust for big or little endian as
2157** appropriate for the arch, so in our laziness we use them rather
2158** than using the bit routines they call more directly. These
2159** routines must be used when changing on disk bitmaps. */
2160#define reiserfs_test_and_set_le_bit ext2_set_bit
2161#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2162#define reiserfs_test_le_bit ext2_test_bit
2163#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2164
2165/* sometimes reiserfs_truncate may require to allocate few new blocks
2166 to perform indirect2direct conversion. People probably used to
2167 think, that truncate should work without problems on a filesystem
2168 without free disk space. They may complain that they can not
2169 truncate due to lack of free disk space. This spare space allows us
2170 to not worry about it. 500 is probably too much, but it should be
2171 absolutely safe */
2172#define SPARE_SPACE 500
2173
1da177e4 2174/* prototypes from ioctl.c */
bd4c625c
LT
2175int reiserfs_ioctl(struct inode *inode, struct file *filp,
2176 unsigned int cmd, unsigned long arg);
52b499c4
DH
2177long reiserfs_compat_ioctl(struct file *filp,
2178 unsigned int cmd, unsigned long arg);
bd4c625c 2179
1da177e4
LT
2180/* ioctl's command */
2181#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
2182/* define following flags to be the same as in ext2, so that chattr(1),
2183 lsattr(1) will work with us. */
36695673
DH
2184#define REISERFS_IOC_GETFLAGS FS_IOC_GETFLAGS
2185#define REISERFS_IOC_SETFLAGS FS_IOC_SETFLAGS
2186#define REISERFS_IOC_GETVERSION FS_IOC_GETVERSION
2187#define REISERFS_IOC_SETVERSION FS_IOC_SETVERSION
1da177e4 2188
52b499c4
DH
2189/* the 32 bit compat definitions with int argument */
2190#define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
2191#define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
2192#define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
2193#define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
2194#define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
2195
1da177e4
LT
2196/* Locking primitives */
2197/* Right now we are still falling back to (un)lock_kernel, but eventually that
2198 would evolve into real per-fs locks */
2199#define reiserfs_write_lock( sb ) lock_kernel()
2200#define reiserfs_write_unlock( sb ) unlock_kernel()
bd4c625c 2201
1da177e4
LT
2202/* xattr stuff */
2203#define REISERFS_XATTR_DIR_SEM(s) (REISERFS_SB(s)->xattr_dir_sem)
2204
bd4c625c 2205#endif /* _LINUX_REISER_FS_H */