]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/linux/reiserfs_fs.h
[PATCH] mutex subsystem, semaphore to mutex: VFS, ->i_sem
[net-next-2.6.git] / include / linux / reiserfs_fs.h
CommitLineData
1da177e4
LT
1/*
2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
3 */
4
5 /* this file has an amazingly stupid
bd4c625c
LT
6 name, yura please fix it to be
7 reiserfs.h, and merge all the rest
8 of our .h files that are in this
9 directory into it. */
1da177e4
LT
10
11#ifndef _LINUX_REISER_FS_H
12#define _LINUX_REISER_FS_H
13
14#include <linux/types.h>
15#ifdef __KERNEL__
16#include <linux/slab.h>
17#include <linux/interrupt.h>
18#include <linux/sched.h>
19#include <linux/workqueue.h>
20#include <asm/unaligned.h>
21#include <linux/bitops.h>
22#include <linux/proc_fs.h>
23#include <linux/smp_lock.h>
24#include <linux/buffer_head.h>
25#include <linux/reiserfs_fs_i.h>
26#include <linux/reiserfs_fs_sb.h>
27#endif
28
29/*
30 * include/linux/reiser_fs.h
31 *
32 * Reiser File System constants and structures
33 *
34 */
35
36/* in reading the #defines, it may help to understand that they employ
37 the following abbreviations:
38
39 B = Buffer
40 I = Item header
41 H = Height within the tree (should be changed to LEV)
42 N = Number of the item in the node
43 STAT = stat data
44 DEH = Directory Entry Header
45 EC = Entry Count
46 E = Entry number
47 UL = Unsigned Long
48 BLKH = BLocK Header
49 UNFM = UNForMatted node
50 DC = Disk Child
51 P = Path
52
53 These #defines are named by concatenating these abbreviations,
54 where first comes the arguments, and last comes the return value,
55 of the macro.
56
57*/
58
59#define USE_INODE_GENERATION_COUNTER
60
61#define REISERFS_PREALLOCATE
62#define DISPLACE_NEW_PACKING_LOCALITIES
63#define PREALLOCATION_SIZE 9
64
65/* n must be power of 2 */
66#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
67
68// to be ok for alpha and others we have to align structures to 8 byte
69// boundary.
70// FIXME: do not change 4 by anything else: there is code which relies on that
71#define ROUND_UP(x) _ROUND_UP(x,8LL)
72
73/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
74** messages.
75*/
bd4c625c 76#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
1da177e4 77
bd4c625c 78void reiserfs_warning(struct super_block *s, const char *fmt, ...);
1da177e4
LT
79/* assertions handling */
80
81/** always check a condition and panic if it's false. */
82#define RASSERT( cond, format, args... ) \
83if( !( cond ) ) \
84 reiserfs_panic( NULL, "reiserfs[%i]: assertion " #cond " failed at " \
85 __FILE__ ":%i:%s: " format "\n", \
86 in_interrupt() ? -1 : current -> pid, __LINE__ , __FUNCTION__ , ##args )
87
88#if defined( CONFIG_REISERFS_CHECK )
89#define RFALSE( cond, format, args... ) RASSERT( !( cond ), format, ##args )
90#else
91#define RFALSE( cond, format, args... ) do {;} while( 0 )
92#endif
93
94#define CONSTF __attribute_const__
95/*
96 * Disk Data Structures
97 */
98
99/***************************************************************************/
100/* SUPER BLOCK */
101/***************************************************************************/
102
103/*
104 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
105 * the version in RAM is part of a larger structure containing fields never written to disk.
106 */
bd4c625c
LT
107#define UNSET_HASH 0 // read_super will guess about, what hash names
108 // in directories were sorted with
1da177e4
LT
109#define TEA_HASH 1
110#define YURA_HASH 2
111#define R5_HASH 3
112#define DEFAULT_HASH R5_HASH
113
1da177e4 114struct journal_params {
bd4c625c
LT
115 __le32 jp_journal_1st_block; /* where does journal start from on its
116 * device */
117 __le32 jp_journal_dev; /* journal device st_rdev */
118 __le32 jp_journal_size; /* size of the journal */
119 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
120 __le32 jp_journal_magic; /* random value made on fs creation (this
121 * was sb_journal_block_count) */
122 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
123 * trans */
124 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
125 * commit be */
126 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
127 * be */
1da177e4
LT
128};
129
130/* this is the super from 3.5.X, where X >= 10 */
bd4c625c
LT
131struct reiserfs_super_block_v1 {
132 __le32 s_block_count; /* blocks count */
133 __le32 s_free_blocks; /* free blocks count */
134 __le32 s_root_block; /* root block number */
135 struct journal_params s_journal;
136 __le16 s_blocksize; /* block size */
137 __le16 s_oid_maxsize; /* max size of object id array, see
138 * get_objectid() commentary */
139 __le16 s_oid_cursize; /* current size of object id array */
140 __le16 s_umount_state; /* this is set to 1 when filesystem was
141 * umounted, to 2 - when not */
142 char s_magic[10]; /* reiserfs magic string indicates that
143 * file system is reiserfs:
144 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
145 __le16 s_fs_state; /* it is set to used by fsck to mark which
146 * phase of rebuilding is done */
147 __le32 s_hash_function_code; /* indicate, what hash function is being use
148 * to sort names in a directory*/
149 __le16 s_tree_height; /* height of disk tree */
150 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
151 * each block of file system */
152 __le16 s_version; /* this field is only reliable on filesystem
153 * with non-standard journal */
154 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
155 * device, we need to keep after
156 * making fs with non-standard journal */
1da177e4
LT
157} __attribute__ ((__packed__));
158
159#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
160
161/* this is the on disk super block */
bd4c625c
LT
162struct reiserfs_super_block {
163 struct reiserfs_super_block_v1 s_v1;
164 __le32 s_inode_generation;
165 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
166 unsigned char s_uuid[16]; /* filesystem unique identifier */
167 unsigned char s_label[16]; /* filesystem volume label */
168 char s_unused[88]; /* zero filled by mkreiserfs and
169 * reiserfs_convert_objectid_map_v1()
170 * so any additions must be updated
171 * there as well. */
172} __attribute__ ((__packed__));
1da177e4
LT
173
174#define SB_SIZE (sizeof(struct reiserfs_super_block))
175
176#define REISERFS_VERSION_1 0
177#define REISERFS_VERSION_2 2
178
1da177e4
LT
179// on-disk super block fields converted to cpu form
180#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
181#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
182#define SB_BLOCKSIZE(s) \
183 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
184#define SB_BLOCK_COUNT(s) \
185 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
186#define SB_FREE_BLOCKS(s) \
187 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
188#define SB_REISERFS_MAGIC(s) \
189 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
190#define SB_ROOT_BLOCK(s) \
191 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
192#define SB_TREE_HEIGHT(s) \
193 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
194#define SB_REISERFS_STATE(s) \
195 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
196#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
197#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
198
199#define PUT_SB_BLOCK_COUNT(s, val) \
200 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
201#define PUT_SB_FREE_BLOCKS(s, val) \
202 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
203#define PUT_SB_ROOT_BLOCK(s, val) \
204 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
205#define PUT_SB_TREE_HEIGHT(s, val) \
206 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
207#define PUT_SB_REISERFS_STATE(s, val) \
bd4c625c 208 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1da177e4
LT
209#define PUT_SB_VERSION(s, val) \
210 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
211#define PUT_SB_BMAP_NR(s, val) \
212 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
213
1da177e4
LT
214#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
215#define SB_ONDISK_JOURNAL_SIZE(s) \
216 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
217#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
218 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
219#define SB_ONDISK_JOURNAL_DEVICE(s) \
220 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
221#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
b8cc936f 222 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1da177e4
LT
223
224#define is_block_in_log_or_reserved_area(s, block) \
225 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
226 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
227 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
bd4c625c 228 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1da177e4
LT
229
230 /* used by gcc */
231#define REISERFS_SUPER_MAGIC 0x52654973
232 /* used by file system utilities that
bd4c625c 233 look at the superblock, etc. */
1da177e4
LT
234#define REISERFS_SUPER_MAGIC_STRING "ReIsErFs"
235#define REISER2FS_SUPER_MAGIC_STRING "ReIsEr2Fs"
236#define REISER2FS_JR_SUPER_MAGIC_STRING "ReIsEr3Fs"
237
bd4c625c
LT
238int is_reiserfs_3_5(struct reiserfs_super_block *rs);
239int is_reiserfs_3_6(struct reiserfs_super_block *rs);
240int is_reiserfs_jr(struct reiserfs_super_block *rs);
1da177e4
LT
241
242/* ReiserFS leaves the first 64k unused, so that partition labels have
243 enough space. If someone wants to write a fancy bootloader that
244 needs more than 64k, let us know, and this will be increased in size.
245 This number must be larger than than the largest block size on any
246 platform, or code will break. -Hans */
247#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
248#define REISERFS_FIRST_BLOCK unused_define
249#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
250
251/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
252#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
253
254// reiserfs internal error code (used by search_by_key adn fix_nodes))
255#define CARRY_ON 0
256#define REPEAT_SEARCH -1
257#define IO_ERROR -2
258#define NO_DISK_SPACE -3
259#define NO_BALANCING_NEEDED (-4)
260#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
261#define QUOTA_EXCEEDED -6
262
263typedef __u32 b_blocknr_t;
3e8962be 264typedef __le32 unp_t;
1da177e4
LT
265
266struct unfm_nodeinfo {
bd4c625c
LT
267 unp_t unfm_nodenum;
268 unsigned short unfm_freespace;
1da177e4
LT
269};
270
271/* there are two formats of keys: 3.5 and 3.6
272 */
273#define KEY_FORMAT_3_5 0
274#define KEY_FORMAT_3_6 1
275
276/* there are two stat datas */
277#define STAT_DATA_V1 0
278#define STAT_DATA_V2 1
279
1da177e4
LT
280static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
281{
282 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
283}
284
285static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
286{
287 return sb->s_fs_info;
288}
289
290/** this says about version of key of all items (but stat data) the
291 object consists of */
292#define get_inode_item_key_version( inode ) \
293 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
294
295#define set_inode_item_key_version( inode, version ) \
296 ({ if((version)==KEY_FORMAT_3_6) \
297 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
298 else \
299 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
300
301#define get_inode_sd_version(inode) \
302 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
303
304#define set_inode_sd_version(inode, version) \
305 ({ if((version)==STAT_DATA_V2) \
306 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
307 else \
308 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
309
310/* This is an aggressive tail suppression policy, I am hoping it
311 improves our benchmarks. The principle behind it is that percentage
312 space saving is what matters, not absolute space saving. This is
313 non-intuitive, but it helps to understand it if you consider that the
314 cost to access 4 blocks is not much more than the cost to access 1
315 block, if you have to do a seek and rotate. A tail risks a
316 non-linear disk access that is significant as a percentage of total
317 time cost for a 4 block file and saves an amount of space that is
318 less significant as a percentage of space, or so goes the hypothesis.
319 -Hans */
320#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
321(\
322 (!(n_tail_size)) || \
323 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
324 ( (n_file_size) >= (n_block_size) * 4 ) || \
325 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
326 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
327 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
328 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
329 ( ( (n_file_size) >= (n_block_size) ) && \
330 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
331)
332
333/* Another strategy for tails, this one means only create a tail if all the
334 file would fit into one DIRECT item.
335 Primary intention for this one is to increase performance by decreasing
336 seeking.
bd4c625c 337*/
1da177e4
LT
338#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
339(\
340 (!(n_tail_size)) || \
341 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
342)
343
1da177e4
LT
344/*
345 * values for s_umount_state field
346 */
347#define REISERFS_VALID_FS 1
348#define REISERFS_ERROR_FS 2
349
350//
351// there are 5 item types currently
352//
353#define TYPE_STAT_DATA 0
354#define TYPE_INDIRECT 1
355#define TYPE_DIRECT 2
bd4c625c
LT
356#define TYPE_DIRENTRY 3
357#define TYPE_MAXTYPE 3
358#define TYPE_ANY 15 // FIXME: comment is required
1da177e4
LT
359
360/***************************************************************************/
361/* KEY & ITEM HEAD */
362/***************************************************************************/
363
364//
365// directories use this key as well as old files
366//
367struct offset_v1 {
bd4c625c
LT
368 __le32 k_offset;
369 __le32 k_uniqueness;
1da177e4
LT
370} __attribute__ ((__packed__));
371
372struct offset_v2 {
f8e08a84 373 __le64 v;
1da177e4
LT
374} __attribute__ ((__packed__));
375
bd4c625c 376static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1da177e4 377{
f8e08a84 378 __u8 type = le64_to_cpu(v2->v) >> 60;
bd4c625c 379 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1da177e4 380}
bd4c625c
LT
381
382static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1da177e4 383{
bd4c625c
LT
384 v2->v =
385 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1da177e4 386}
bd4c625c
LT
387
388static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1da177e4 389{
bd4c625c 390 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1da177e4
LT
391}
392
bd4c625c
LT
393static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
394{
395 offset &= (~0ULL >> 4);
396 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1da177e4 397}
1da177e4
LT
398
399/* Key of an item determines its location in the S+tree, and
400 is composed of 4 components */
401struct reiserfs_key {
bd4c625c
LT
402 __le32 k_dir_id; /* packing locality: by default parent
403 directory object id */
404 __le32 k_objectid; /* object identifier */
405 union {
406 struct offset_v1 k_offset_v1;
407 struct offset_v2 k_offset_v2;
408 } __attribute__ ((__packed__)) u;
1da177e4
LT
409} __attribute__ ((__packed__));
410
6a3a16f2 411struct in_core_key {
bd4c625c
LT
412 __u32 k_dir_id; /* packing locality: by default parent
413 directory object id */
414 __u32 k_objectid; /* object identifier */
415 __u64 k_offset;
416 __u8 k_type;
6b9f5829 417};
1da177e4
LT
418
419struct cpu_key {
bd4c625c
LT
420 struct in_core_key on_disk_key;
421 int version;
422 int key_length; /* 3 in all cases but direct2indirect and
423 indirect2direct conversion */
1da177e4
LT
424};
425
426/* Our function for comparing keys can compare keys of different
427 lengths. It takes as a parameter the length of the keys it is to
428 compare. These defines are used in determining what is to be passed
429 to it as that parameter. */
430#define REISERFS_FULL_KEY_LEN 4
431#define REISERFS_SHORT_KEY_LEN 2
432
433/* The result of the key compare */
434#define FIRST_GREATER 1
435#define SECOND_GREATER -1
436#define KEYS_IDENTICAL 0
437#define KEY_FOUND 1
438#define KEY_NOT_FOUND 0
439
440#define KEY_SIZE (sizeof(struct reiserfs_key))
441#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
442
443/* return values for search_by_key and clones */
444#define ITEM_FOUND 1
445#define ITEM_NOT_FOUND 0
446#define ENTRY_FOUND 1
447#define ENTRY_NOT_FOUND 0
448#define DIRECTORY_NOT_FOUND -1
449#define REGULAR_FILE_FOUND -2
450#define DIRECTORY_FOUND -3
451#define BYTE_FOUND 1
452#define BYTE_NOT_FOUND 0
453#define FILE_NOT_FOUND -1
454
455#define POSITION_FOUND 1
456#define POSITION_NOT_FOUND 0
457
458// return values for reiserfs_find_entry and search_by_entry_key
459#define NAME_FOUND 1
460#define NAME_NOT_FOUND 0
461#define GOTO_PREVIOUS_ITEM 2
462#define NAME_FOUND_INVISIBLE 3
463
464/* Everything in the filesystem is stored as a set of items. The
465 item head contains the key of the item, its free space (for
466 indirect items) and specifies the location of the item itself
467 within the block. */
468
bd4c625c 469struct item_head {
1da177e4
LT
470 /* Everything in the tree is found by searching for it based on
471 * its key.*/
472 struct reiserfs_key ih_key;
473 union {
474 /* The free space in the last unformatted node of an
475 indirect item if this is an indirect item. This
476 equals 0xFFFF iff this is a direct item or stat data
477 item. Note that the key, not this field, is used to
478 determine the item type, and thus which field this
479 union contains. */
3e8962be 480 __le16 ih_free_space_reserved;
1da177e4
LT
481 /* Iff this is a directory item, this field equals the
482 number of directory entries in the directory item. */
3e8962be 483 __le16 ih_entry_count;
1da177e4 484 } __attribute__ ((__packed__)) u;
bd4c625c
LT
485 __le16 ih_item_len; /* total size of the item body */
486 __le16 ih_item_location; /* an offset to the item body
487 * within the block */
488 __le16 ih_version; /* 0 for all old items, 2 for new
489 ones. Highest bit is set by fsck
490 temporary, cleaned after all
491 done */
1da177e4
LT
492} __attribute__ ((__packed__));
493/* size of item header */
494#define IH_SIZE (sizeof(struct item_head))
495
496#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
497#define ih_version(ih) le16_to_cpu((ih)->ih_version)
498#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
499#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
500#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
501
502#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
503#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
504#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
505#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
506#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
507
1da177e4
LT
508#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
509
510#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
511#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
512
513/* these operate on indirect items, where you've got an array of ints
514** at a possibly unaligned location. These are a noop on ia32
515**
516** p is the array of __u32, i is the index into the array, v is the value
517** to store there.
518*/
519#define get_block_num(p, i) le32_to_cpu(get_unaligned((p) + (i)))
520#define put_block_num(p, i, v) put_unaligned(cpu_to_le32(v), (p) + (i))
521
522//
523// in old version uniqueness field shows key type
524//
525#define V1_SD_UNIQUENESS 0
526#define V1_INDIRECT_UNIQUENESS 0xfffffffe
527#define V1_DIRECT_UNIQUENESS 0xffffffff
528#define V1_DIRENTRY_UNIQUENESS 500
bd4c625c 529#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
1da177e4
LT
530
531//
532// here are conversion routines
533//
bd4c625c
LT
534static inline int uniqueness2type(__u32 uniqueness) CONSTF;
535static inline int uniqueness2type(__u32 uniqueness)
1da177e4 536{
bd4c625c
LT
537 switch ((int)uniqueness) {
538 case V1_SD_UNIQUENESS:
539 return TYPE_STAT_DATA;
540 case V1_INDIRECT_UNIQUENESS:
541 return TYPE_INDIRECT;
542 case V1_DIRECT_UNIQUENESS:
543 return TYPE_DIRECT;
544 case V1_DIRENTRY_UNIQUENESS:
545 return TYPE_DIRENTRY;
546 default:
547 reiserfs_warning(NULL, "vs-500: unknown uniqueness %d",
548 uniqueness);
1da177e4 549 case V1_ANY_UNIQUENESS:
bd4c625c
LT
550 return TYPE_ANY;
551 }
1da177e4
LT
552}
553
bd4c625c
LT
554static inline __u32 type2uniqueness(int type) CONSTF;
555static inline __u32 type2uniqueness(int type)
1da177e4 556{
bd4c625c
LT
557 switch (type) {
558 case TYPE_STAT_DATA:
559 return V1_SD_UNIQUENESS;
560 case TYPE_INDIRECT:
561 return V1_INDIRECT_UNIQUENESS;
562 case TYPE_DIRECT:
563 return V1_DIRECT_UNIQUENESS;
564 case TYPE_DIRENTRY:
565 return V1_DIRENTRY_UNIQUENESS;
566 default:
567 reiserfs_warning(NULL, "vs-501: unknown type %d", type);
1da177e4 568 case TYPE_ANY:
bd4c625c
LT
569 return V1_ANY_UNIQUENESS;
570 }
1da177e4
LT
571}
572
573//
574// key is pointer to on disk key which is stored in le, result is cpu,
575// there is no way to get version of object from key, so, provide
576// version to these defines
577//
bd4c625c
LT
578static inline loff_t le_key_k_offset(int version,
579 const struct reiserfs_key *key)
1da177e4 580{
bd4c625c
LT
581 return (version == KEY_FORMAT_3_5) ?
582 le32_to_cpu(key->u.k_offset_v1.k_offset) :
583 offset_v2_k_offset(&(key->u.k_offset_v2));
1da177e4
LT
584}
585
bd4c625c 586static inline loff_t le_ih_k_offset(const struct item_head *ih)
1da177e4 587{
bd4c625c 588 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1da177e4
LT
589}
590
bd4c625c 591static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1da177e4 592{
bd4c625c
LT
593 return (version == KEY_FORMAT_3_5) ?
594 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
595 offset_v2_k_type(&(key->u.k_offset_v2));
1da177e4
LT
596}
597
bd4c625c 598static inline loff_t le_ih_k_type(const struct item_head *ih)
1da177e4 599{
bd4c625c 600 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1da177e4
LT
601}
602
bd4c625c
LT
603static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
604 loff_t offset)
1da177e4 605{
bd4c625c
LT
606 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
607 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
1da177e4
LT
608}
609
bd4c625c 610static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1da177e4 611{
bd4c625c 612 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1da177e4
LT
613}
614
bd4c625c
LT
615static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
616 int type)
1da177e4 617{
bd4c625c
LT
618 (version == KEY_FORMAT_3_5) ?
619 (void)(key->u.k_offset_v1.k_uniqueness =
620 cpu_to_le32(type2uniqueness(type)))
621 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
1da177e4 622}
bd4c625c 623static inline void set_le_ih_k_type(struct item_head *ih, int type)
1da177e4 624{
bd4c625c 625 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1da177e4
LT
626}
627
1da177e4
LT
628#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
629#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
630#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
631#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
632
633//
634// item header has version.
635//
636#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
637#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
638#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
639#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
640
1da177e4
LT
641//
642// key is pointer to cpu key, result is cpu
643//
bd4c625c 644static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1da177e4 645{
bd4c625c 646 return key->on_disk_key.k_offset;
1da177e4
LT
647}
648
bd4c625c 649static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1da177e4 650{
bd4c625c 651 return key->on_disk_key.k_type;
1da177e4
LT
652}
653
bd4c625c 654static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1da177e4 655{
6b9f5829 656 key->on_disk_key.k_offset = offset;
1da177e4
LT
657}
658
bd4c625c 659static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1da177e4 660{
6b9f5829 661 key->on_disk_key.k_type = type;
1da177e4
LT
662}
663
bd4c625c 664static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1da177e4 665{
bd4c625c 666 key->on_disk_key.k_offset--;
1da177e4
LT
667}
668
1da177e4
LT
669#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
670#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
671#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
672#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
673
1da177e4
LT
674/* are these used ? */
675#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
676#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
677#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
678#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
679
1da177e4
LT
680#define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
681 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
682 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
683
bd4c625c 684/* maximal length of item */
1da177e4
LT
685#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
686#define MIN_ITEM_LEN 1
687
1da177e4
LT
688/* object identifier for root dir */
689#define REISERFS_ROOT_OBJECTID 2
690#define REISERFS_ROOT_PARENT_OBJECTID 1
691extern struct reiserfs_key root_key;
692
1da177e4
LT
693/*
694 * Picture represents a leaf of the S+tree
695 * ______________________________________________________
696 * | | Array of | | |
697 * |Block | Object-Item | F r e e | Objects- |
698 * | head | Headers | S p a c e | Items |
699 * |______|_______________|___________________|___________|
700 */
701
702/* Header of a disk block. More precisely, header of a formatted leaf
703 or internal node, and not the header of an unformatted node. */
bd4c625c
LT
704struct block_head {
705 __le16 blk_level; /* Level of a block in the tree. */
706 __le16 blk_nr_item; /* Number of keys/items in a block. */
707 __le16 blk_free_space; /* Block free space in bytes. */
708 __le16 blk_reserved;
709 /* dump this in v4/planA */
710 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
1da177e4
LT
711};
712
713#define BLKH_SIZE (sizeof(struct block_head))
714#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
715#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
716#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
717#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
718#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
719#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
720#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
721#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
722#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
723#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
724
725/*
726 * values for blk_level field of the struct block_head
727 */
728
bd4c625c
LT
729#define FREE_LEVEL 0 /* when node gets removed from the tree its
730 blk_level is set to FREE_LEVEL. It is then
731 used to see whether the node is still in the
732 tree */
1da177e4 733
bd4c625c 734#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1da177e4
LT
735
736/* Given the buffer head of a formatted node, resolve to the block head of that node. */
737#define B_BLK_HEAD(p_s_bh) ((struct block_head *)((p_s_bh)->b_data))
738/* Number of items that are in buffer. */
739#define B_NR_ITEMS(p_s_bh) (blkh_nr_item(B_BLK_HEAD(p_s_bh)))
740#define B_LEVEL(p_s_bh) (blkh_level(B_BLK_HEAD(p_s_bh)))
741#define B_FREE_SPACE(p_s_bh) (blkh_free_space(B_BLK_HEAD(p_s_bh)))
742
743#define PUT_B_NR_ITEMS(p_s_bh,val) do { set_blkh_nr_item(B_BLK_HEAD(p_s_bh),val); } while (0)
744#define PUT_B_LEVEL(p_s_bh,val) do { set_blkh_level(B_BLK_HEAD(p_s_bh),val); } while (0)
745#define PUT_B_FREE_SPACE(p_s_bh,val) do { set_blkh_free_space(B_BLK_HEAD(p_s_bh),val); } while (0)
746
1da177e4
LT
747/* Get right delimiting key. -- little endian */
748#define B_PRIGHT_DELIM_KEY(p_s_bh) (&(blk_right_delim_key(B_BLK_HEAD(p_s_bh))
749
750/* Does the buffer contain a disk leaf. */
751#define B_IS_ITEMS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) == DISK_LEAF_NODE_LEVEL)
752
753/* Does the buffer contain a disk internal node */
754#define B_IS_KEYS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) > DISK_LEAF_NODE_LEVEL \
755 && B_LEVEL(p_s_bh) <= MAX_HEIGHT)
756
1da177e4
LT
757/***************************************************************************/
758/* STAT DATA */
759/***************************************************************************/
760
1da177e4
LT
761//
762// old stat data is 32 bytes long. We are going to distinguish new one by
763// different size
764//
bd4c625c
LT
765struct stat_data_v1 {
766 __le16 sd_mode; /* file type, permissions */
767 __le16 sd_nlink; /* number of hard links */
768 __le16 sd_uid; /* owner */
769 __le16 sd_gid; /* group */
770 __le32 sd_size; /* file size */
771 __le32 sd_atime; /* time of last access */
772 __le32 sd_mtime; /* time file was last modified */
773 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
774 union {
775 __le32 sd_rdev;
776 __le32 sd_blocks; /* number of blocks file uses */
777 } __attribute__ ((__packed__)) u;
778 __le32 sd_first_direct_byte; /* first byte of file which is stored
779 in a direct item: except that if it
780 equals 1 it is a symlink and if it
781 equals ~(__u32)0 there is no
782 direct item. The existence of this
783 field really grates on me. Let's
784 replace it with a macro based on
785 sd_size and our tail suppression
786 policy. Someday. -Hans */
1da177e4
LT
787} __attribute__ ((__packed__));
788
789#define SD_V1_SIZE (sizeof(struct stat_data_v1))
790#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
791#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
792#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
793#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
794#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
795#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
796#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
797#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
798#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
799#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
800#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
801#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
802#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
803#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
804#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
805#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
806#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
807#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
808#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
809#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
810#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
811#define sd_v1_first_direct_byte(sdp) \
812 (le32_to_cpu((sdp)->sd_first_direct_byte))
813#define set_sd_v1_first_direct_byte(sdp,v) \
814 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
815
816#include <linux/ext2_fs.h>
817
818/* inode flags stored in sd_attrs (nee sd_reserved) */
819
820/* we want common flags to have the same values as in ext2,
821 so chattr(1) will work without problems */
822#define REISERFS_IMMUTABLE_FL EXT2_IMMUTABLE_FL
823#define REISERFS_APPEND_FL EXT2_APPEND_FL
824#define REISERFS_SYNC_FL EXT2_SYNC_FL
825#define REISERFS_NOATIME_FL EXT2_NOATIME_FL
826#define REISERFS_NODUMP_FL EXT2_NODUMP_FL
827#define REISERFS_SECRM_FL EXT2_SECRM_FL
828#define REISERFS_UNRM_FL EXT2_UNRM_FL
829#define REISERFS_COMPR_FL EXT2_COMPR_FL
830#define REISERFS_NOTAIL_FL EXT2_NOTAIL_FL
831
832/* persistent flags that file inherits from the parent directory */
833#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
834 REISERFS_SYNC_FL | \
835 REISERFS_NOATIME_FL | \
836 REISERFS_NODUMP_FL | \
837 REISERFS_SECRM_FL | \
838 REISERFS_COMPR_FL | \
839 REISERFS_NOTAIL_FL )
840
841/* Stat Data on disk (reiserfs version of UFS disk inode minus the
842 address blocks) */
843struct stat_data {
bd4c625c
LT
844 __le16 sd_mode; /* file type, permissions */
845 __le16 sd_attrs; /* persistent inode flags */
846 __le32 sd_nlink; /* number of hard links */
847 __le64 sd_size; /* file size */
848 __le32 sd_uid; /* owner */
849 __le32 sd_gid; /* group */
850 __le32 sd_atime; /* time of last access */
851 __le32 sd_mtime; /* time file was last modified */
852 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
853 __le32 sd_blocks;
854 union {
855 __le32 sd_rdev;
856 __le32 sd_generation;
857 //__le32 sd_first_direct_byte;
858 /* first byte of file which is stored in a
859 direct item: except that if it equals 1
860 it is a symlink and if it equals
861 ~(__u32)0 there is no direct item. The
862 existence of this field really grates
863 on me. Let's replace it with a macro
864 based on sd_size and our tail
865 suppression policy? */
866 } __attribute__ ((__packed__)) u;
1da177e4
LT
867} __attribute__ ((__packed__));
868//
869// this is 44 bytes long
870//
871#define SD_SIZE (sizeof(struct stat_data))
872#define SD_V2_SIZE SD_SIZE
873#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
874#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
875#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
876/* sd_reserved */
877/* set_sd_reserved */
878#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
879#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
880#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
881#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
882#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
883#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
884#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
885#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
886#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
887#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
888#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
889#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
890#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
891#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
892#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
893#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
894#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
895#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
896#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
897#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
898#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
899#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
900
1da177e4
LT
901/***************************************************************************/
902/* DIRECTORY STRUCTURE */
903/***************************************************************************/
904/*
905 Picture represents the structure of directory items
906 ________________________________________________
907 | Array of | | | | | |
908 | directory |N-1| N-2 | .... | 1st |0th|
909 | entry headers | | | | | |
910 |_______________|___|_____|________|_______|___|
911 <---- directory entries ------>
912
913 First directory item has k_offset component 1. We store "." and ".."
914 in one item, always, we never split "." and ".." into differing
915 items. This makes, among other things, the code for removing
916 directories simpler. */
917#define SD_OFFSET 0
918#define SD_UNIQUENESS 0
919#define DOT_OFFSET 1
920#define DOT_DOT_OFFSET 2
921#define DIRENTRY_UNIQUENESS 500
922
923/* */
924#define FIRST_ITEM_OFFSET 1
925
926/*
927 Q: How to get key of object pointed to by entry from entry?
928
929 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
930 of object, entry points to */
931
932/* NOT IMPLEMENTED:
933 Directory will someday contain stat data of object */
934
bd4c625c
LT
935struct reiserfs_de_head {
936 __le32 deh_offset; /* third component of the directory entry key */
937 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
938 by directory entry */
939 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
940 __le16 deh_location; /* offset of name in the whole item */
941 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
942 entry is hidden (unlinked) */
1da177e4
LT
943} __attribute__ ((__packed__));
944#define DEH_SIZE sizeof(struct reiserfs_de_head)
945#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
946#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
947#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
948#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
949#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
950
951#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
952#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
953#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
954#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
955#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
956
957/* empty directory contains two entries "." and ".." and their headers */
958#define EMPTY_DIR_SIZE \
959(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
960
961/* old format directories have this size when empty */
962#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
963
bd4c625c 964#define DEH_Statdata 0 /* not used now */
1da177e4
LT
965#define DEH_Visible 2
966
967/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
968#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
969# define ADDR_UNALIGNED_BITS (3)
970#endif
971
972/* These are only used to manipulate deh_state.
973 * Because of this, we'll use the ext2_ bit routines,
974 * since they are little endian */
975#ifdef ADDR_UNALIGNED_BITS
976
977# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
978# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
979
980# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
981# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
982# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
983
984#else
985
986# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
987# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
988# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
989
990#endif
991
992#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
993#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
994#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
995#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
996
997#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
998#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
999#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1000
bd4c625c
LT
1001extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1002 __le32 par_dirid, __le32 par_objid);
1003extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1004 __le32 par_dirid, __le32 par_objid);
1da177e4
LT
1005
1006/* array of the entry headers */
1007 /* get item body */
1008#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1009#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1010
1011/* length of the directory entry in directory item. This define
1012 calculates length of i-th directory entry using directory entry
1013 locations from dir entry head. When it calculates length of 0-th
1014 directory entry, it uses length of whole item in place of entry
1015 location of the non-existent following entry in the calculation.
1016 See picture above.*/
1017/*
1018#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1019((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1020*/
bd4c625c
LT
1021static inline int entry_length(const struct buffer_head *bh,
1022 const struct item_head *ih, int pos_in_item)
1da177e4 1023{
bd4c625c 1024 struct reiserfs_de_head *deh;
1da177e4 1025
bd4c625c
LT
1026 deh = B_I_DEH(bh, ih) + pos_in_item;
1027 if (pos_in_item)
1028 return deh_location(deh - 1) - deh_location(deh);
1da177e4 1029
bd4c625c 1030 return ih_item_len(ih) - deh_location(deh);
1da177e4
LT
1031}
1032
1da177e4
LT
1033/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1034#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1035
1da177e4
LT
1036/* name by bh, ih and entry_num */
1037#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))))
1038
1039// two entries per block (at least)
1040#define REISERFS_MAX_NAME(block_size) 255
1041
1da177e4
LT
1042/* this structure is used for operations on directory entries. It is
1043 not a disk structure. */
1044/* When reiserfs_find_entry or search_by_entry_key find directory
1045 entry, they return filled reiserfs_dir_entry structure */
bd4c625c
LT
1046struct reiserfs_dir_entry {
1047 struct buffer_head *de_bh;
1048 int de_item_num;
1049 struct item_head *de_ih;
1050 int de_entry_num;
1051 struct reiserfs_de_head *de_deh;
1052 int de_entrylen;
1053 int de_namelen;
1054 char *de_name;
1055 char *de_gen_number_bit_string;
1056
1057 __u32 de_dir_id;
1058 __u32 de_objectid;
1059
1060 struct cpu_key de_entry_key;
1da177e4 1061};
bd4c625c 1062
1da177e4
LT
1063/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1064
1065/* pointer to file name, stored in entry */
1066#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1067
1068/* length of name */
1069#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1070(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1071
1da177e4
LT
1072/* hash value occupies bits from 7 up to 30 */
1073#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1074/* generation number occupies 7 bits starting from 0 up to 6 */
1075#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1076#define MAX_GENERATION_NUMBER 127
1077
1078#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1079
1da177e4
LT
1080/*
1081 * Picture represents an internal node of the reiserfs tree
1082 * ______________________________________________________
1083 * | | Array of | Array of | Free |
1084 * |block | keys | pointers | space |
1085 * | head | N | N+1 | |
1086 * |______|_______________|___________________|___________|
1087 */
1088
1089/***************************************************************************/
1090/* DISK CHILD */
1091/***************************************************************************/
1092/* Disk child pointer: The pointer from an internal node of the tree
1093 to a node that is on disk. */
1094struct disk_child {
bd4c625c
LT
1095 __le32 dc_block_number; /* Disk child's block number. */
1096 __le16 dc_size; /* Disk child's used space. */
1097 __le16 dc_reserved;
1da177e4
LT
1098};
1099
1100#define DC_SIZE (sizeof(struct disk_child))
1101#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1102#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1103#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1104#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1105
1106/* Get disk child by buffer header and position in the tree node. */
1107#define B_N_CHILD(p_s_bh,n_pos) ((struct disk_child *)\
1108((p_s_bh)->b_data+BLKH_SIZE+B_NR_ITEMS(p_s_bh)*KEY_SIZE+DC_SIZE*(n_pos)))
1109
1110/* Get disk child number by buffer header and position in the tree node. */
1111#define B_N_CHILD_NUM(p_s_bh,n_pos) (dc_block_number(B_N_CHILD(p_s_bh,n_pos)))
1112#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 ))
1113
bd4c625c 1114 /* maximal value of field child_size in structure disk_child */
1da177e4
LT
1115 /* child size is the combined size of all items and their headers */
1116#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1117
1118/* amount of used space in buffer (not including block head) */
1119#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1120
1121/* max and min number of keys in internal node */
1122#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1123#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1124
1125/***************************************************************************/
1126/* PATH STRUCTURES AND DEFINES */
1127/***************************************************************************/
1128
1da177e4
LT
1129/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1130 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1131 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1132 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1133 position of the block_number of the next node if it is looking through an internal node. If it
1134 is looking through a leaf node bin_search will find the position of the item which has key either
1135 equal to given key, or which is the maximal key less than the given key. */
1136
bd4c625c
LT
1137struct path_element {
1138 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1139 int pe_position; /* Position in the tree node which is placed in the */
1140 /* buffer above. */
1da177e4
LT
1141};
1142
bd4c625c
LT
1143#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1144#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1145#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1da177e4 1146
bd4c625c
LT
1147#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1148#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1da177e4
LT
1149
1150/* We need to keep track of who the ancestors of nodes are. When we
1151 perform a search we record which nodes were visited while
1152 descending the tree looking for the node we searched for. This list
1153 of nodes is called the path. This information is used while
1154 performing balancing. Note that this path information may become
1155 invalid, and this means we must check it when using it to see if it
1156 is still valid. You'll need to read search_by_key and the comments
1157 in it, especially about decrement_counters_in_path(), to understand
1158 this structure.
1159
1160Paths make the code so much harder to work with and debug.... An
1161enormous number of bugs are due to them, and trying to write or modify
1162code that uses them just makes my head hurt. They are based on an
1163excessive effort to avoid disturbing the precious VFS code.:-( The
1164gods only know how we are going to SMP the code that uses them.
1165znodes are the way! */
1166
bd4c625c
LT
1167#define PATH_READA 0x1 /* do read ahead */
1168#define PATH_READA_BACK 0x2 /* read backwards */
1da177e4 1169
bd4c625c
LT
1170struct path {
1171 int path_length; /* Length of the array above. */
1172 int reada;
1173 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1174 int pos_in_item;
1da177e4
LT
1175};
1176
1177#define pos_in_item(path) ((path)->pos_in_item)
1178
1179#define INITIALIZE_PATH(var) \
1180struct path var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1181
1182/* Get path element by path and path position. */
1183#define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1184
1185/* Get buffer header at the path by path and path position. */
1186#define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1187
1188/* Get position in the element at the path by path and path position. */
1189#define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1190
1da177e4
LT
1191#define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1192 /* you know, to the person who didn't
bd4c625c
LT
1193 write this the macro name does not
1194 at first suggest what it does.
1195 Maybe POSITION_FROM_PATH_END? Or
1196 maybe we should just focus on
1197 dumping paths... -Hans */
1da177e4
LT
1198#define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1199
1da177e4
LT
1200#define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1201
1202/* in do_balance leaf has h == 0 in contrast with path structure,
1203 where root has level == 0. That is why we need these defines */
1204#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
1205#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1206#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1207#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1da177e4
LT
1208
1209#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1210
1211#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1212#define get_ih(path) PATH_PITEM_HEAD(path)
1213#define get_item_pos(path) PATH_LAST_POSITION(path)
1214#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1215#define item_moved(ih,path) comp_items(ih, path)
1216#define path_changed(ih,path) comp_items (ih, path)
1217
1da177e4
LT
1218/***************************************************************************/
1219/* MISC */
1220/***************************************************************************/
1221
1222/* Size of pointer to the unformatted node. */
1223#define UNFM_P_SIZE (sizeof(unp_t))
1224#define UNFM_P_SHIFT 2
1225
1226// in in-core inode key is stored on le form
1227#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1228
1229#define MAX_UL_INT 0xffffffff
1230#define MAX_INT 0x7ffffff
1231#define MAX_US_INT 0xffff
1232
1233// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1234#define U32_MAX (~(__u32)0)
1235
bd4c625c 1236static inline loff_t max_reiserfs_offset(struct inode *inode)
1da177e4 1237{
bd4c625c
LT
1238 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1239 return (loff_t) U32_MAX;
1da177e4 1240
bd4c625c 1241 return (loff_t) ((~(__u64) 0) >> 4);
1da177e4
LT
1242}
1243
1da177e4
LT
1244/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1245#define MAX_KEY_OBJECTID MAX_UL_INT
1246
1da177e4
LT
1247#define MAX_B_NUM MAX_UL_INT
1248#define MAX_FC_NUM MAX_US_INT
1249
1da177e4
LT
1250/* the purpose is to detect overflow of an unsigned short */
1251#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1252
1da177e4 1253/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
bd4c625c
LT
1254#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1255#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1da177e4
LT
1256
1257#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1258#define get_generation(s) atomic_read (&fs_generation(s))
1259#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1260#define __fs_changed(gen,s) (gen != get_generation (s))
1261#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1262
1da177e4
LT
1263/***************************************************************************/
1264/* FIXATE NODES */
1265/***************************************************************************/
1266
1267#define VI_TYPE_LEFT_MERGEABLE 1
1268#define VI_TYPE_RIGHT_MERGEABLE 2
1269
1270/* To make any changes in the tree we always first find node, that
1271 contains item to be changed/deleted or place to insert a new
1272 item. We call this node S. To do balancing we need to decide what
1273 we will shift to left/right neighbor, or to a new node, where new
1274 item will be etc. To make this analysis simpler we build virtual
1275 node. Virtual node is an array of items, that will replace items of
1276 node S. (For instance if we are going to delete an item, virtual
1277 node does not contain it). Virtual node keeps information about
1278 item sizes and types, mergeability of first and last items, sizes
1279 of all entries in directory item. We use this array of items when
1280 calculating what we can shift to neighbors and how many nodes we
1281 have to have if we do not any shiftings, if we shift to left/right
1282 neighbor or to both. */
bd4c625c
LT
1283struct virtual_item {
1284 int vi_index; // index in the array of item operations
1285 unsigned short vi_type; // left/right mergeability
1286 unsigned short vi_item_len; /* length of item that it will have after balancing */
1287 struct item_head *vi_ih;
1288 const char *vi_item; // body of item (old or new)
1289 const void *vi_new_data; // 0 always but paste mode
1290 void *vi_uarea; // item specific area
1da177e4
LT
1291};
1292
bd4c625c
LT
1293struct virtual_node {
1294 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
1295 unsigned short vn_nr_item; /* number of items in virtual node */
1296 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1297 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1298 short vn_affected_item_num;
1299 short vn_pos_in_item;
1300 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
1301 const void *vn_data;
1302 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1da177e4
LT
1303};
1304
1305/* used by directory items when creating virtual nodes */
1306struct direntry_uarea {
bd4c625c
LT
1307 int flags;
1308 __u16 entry_count;
1309 __u16 entry_sizes[1];
1310} __attribute__ ((__packed__));
1da177e4
LT
1311
1312/***************************************************************************/
1313/* TREE BALANCE */
1314/***************************************************************************/
1315
1316/* This temporary structure is used in tree balance algorithms, and
1317 constructed as we go to the extent that its various parts are
1318 needed. It contains arrays of nodes that can potentially be
1319 involved in the balancing of node S, and parameters that define how
1320 each of the nodes must be balanced. Note that in these algorithms
1321 for balancing the worst case is to need to balance the current node
1322 S and the left and right neighbors and all of their parents plus
1323 create a new node. We implement S1 balancing for the leaf nodes
1324 and S0 balancing for the internal nodes (S1 and S0 are defined in
1325 our papers.)*/
1326
1327#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1328
1329/* maximum number of FEB blocknrs on a single level */
1330#define MAX_AMOUNT_NEEDED 2
1331
1332/* someday somebody will prefix every field in this struct with tb_ */
bd4c625c
LT
1333struct tree_balance {
1334 int tb_mode;
1335 int need_balance_dirty;
1336 struct super_block *tb_sb;
1337 struct reiserfs_transaction_handle *transaction_handle;
1338 struct path *tb_path;
1339 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1340 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
1341 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1342 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1343 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1344 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1345
1346 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1347 cur_blknum. */
1348 struct buffer_head *used[MAX_FEB_SIZE];
1349 struct buffer_head *thrown[MAX_FEB_SIZE];
1350 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1351 shifted to the left in order to balance the
1352 current node; for leaves includes item that
1353 will be partially shifted; for internal
1354 nodes, it is the number of child pointers
1355 rather than items. It includes the new item
1356 being created. The code sometimes subtracts
1357 one to get the number of wholly shifted
1358 items for other purposes. */
1359 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1360 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1361 S[h] to its item number within the node CFL[h] */
1362 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1363 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1364 S[h]. A negative value means removing. */
1365 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1366 balancing on the level h of the tree. If 0 then S is
1367 being deleted, if 1 then S is remaining and no new nodes
1368 are being created, if 2 or 3 then 1 or 2 new nodes is
1369 being created */
1370
1371 /* fields that are used only for balancing leaves of the tree */
1372 int cur_blknum; /* number of empty blocks having been already allocated */
1373 int s0num; /* number of items that fall into left most node when S[0] splits */
1374 int s1num; /* number of items that fall into first new node when S[0] splits */
1375 int s2num; /* number of items that fall into second new node when S[0] splits */
1376 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1377 /* most liquid item that cannot be shifted from S[0] entirely */
1378 /* if -1 then nothing will be partially shifted */
1379 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1380 /* most liquid item that cannot be shifted from S[0] entirely */
1381 /* if -1 then nothing will be partially shifted */
1382 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1383 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1384 int s2bytes;
1385 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1386 char *vn_buf; /* kmalloced memory. Used to create
1da177e4
LT
1387 virtual node and keep map of
1388 dirtied bitmap blocks */
bd4c625c
LT
1389 int vn_buf_size; /* size of the vn_buf */
1390 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
1da177e4 1391
bd4c625c
LT
1392 int fs_gen; /* saved value of `reiserfs_generation' counter
1393 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1da177e4 1394#ifdef DISPLACE_NEW_PACKING_LOCALITIES
bd4c625c
LT
1395 struct in_core_key key; /* key pointer, to pass to block allocator or
1396 another low-level subsystem */
1da177e4 1397#endif
bd4c625c 1398};
1da177e4
LT
1399
1400/* These are modes of balancing */
1401
1402/* When inserting an item. */
1403#define M_INSERT 'i'
1404/* When inserting into (directories only) or appending onto an already
1405 existant item. */
1406#define M_PASTE 'p'
1407/* When deleting an item. */
1408#define M_DELETE 'd'
1409/* When truncating an item or removing an entry from a (directory) item. */
1410#define M_CUT 'c'
1411
1412/* used when balancing on leaf level skipped (in reiserfsck) */
1413#define M_INTERNAL 'n'
1414
1415/* When further balancing is not needed, then do_balance does not need
1416 to be called. */
1417#define M_SKIP_BALANCING 's'
1418#define M_CONVERT 'v'
1419
1420/* modes of leaf_move_items */
1421#define LEAF_FROM_S_TO_L 0
1422#define LEAF_FROM_S_TO_R 1
1423#define LEAF_FROM_R_TO_L 2
1424#define LEAF_FROM_L_TO_R 3
1425#define LEAF_FROM_S_TO_SNEW 4
1426
1427#define FIRST_TO_LAST 0
1428#define LAST_TO_FIRST 1
1429
1430/* used in do_balance for passing parent of node information that has
1431 been gotten from tb struct */
1432struct buffer_info {
bd4c625c
LT
1433 struct tree_balance *tb;
1434 struct buffer_head *bi_bh;
1435 struct buffer_head *bi_parent;
1436 int bi_position;
1da177e4
LT
1437};
1438
1da177e4
LT
1439/* there are 4 types of items: stat data, directory item, indirect, direct.
1440+-------------------+------------+--------------+------------+
1441| | k_offset | k_uniqueness | mergeable? |
1442+-------------------+------------+--------------+------------+
1443| stat data | 0 | 0 | no |
1444+-------------------+------------+--------------+------------+
1445| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1446| non 1st directory | hash value | | yes |
1447| item | | | |
1448+-------------------+------------+--------------+------------+
1449| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1450+-------------------+------------+--------------+------------+
1451| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1452+-------------------+------------+--------------+------------+
1453*/
1454
1455struct item_operations {
bd4c625c
LT
1456 int (*bytes_number) (struct item_head * ih, int block_size);
1457 void (*decrement_key) (struct cpu_key *);
1458 int (*is_left_mergeable) (struct reiserfs_key * ih,
1459 unsigned long bsize);
1460 void (*print_item) (struct item_head *, char *item);
1461 void (*check_item) (struct item_head *, char *item);
1462
1463 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1464 int is_affected, int insert_size);
1465 int (*check_left) (struct virtual_item * vi, int free,
1466 int start_skip, int end_skip);
1467 int (*check_right) (struct virtual_item * vi, int free);
1468 int (*part_size) (struct virtual_item * vi, int from, int to);
1469 int (*unit_num) (struct virtual_item * vi);
1470 void (*print_vi) (struct virtual_item * vi);
1da177e4
LT
1471};
1472
bd4c625c 1473extern struct item_operations *item_ops[TYPE_ANY + 1];
1da177e4
LT
1474
1475#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1476#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1477#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1478#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1479#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)
1480#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1481#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1482#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1483#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1484#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1485
1da177e4
LT
1486#define COMP_SHORT_KEYS comp_short_keys
1487
1488/* number of blocks pointed to by the indirect item */
1489#define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1490
1491/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1492#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1493
1494/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1495
bd4c625c 1496/* get the item header */
1da177e4
LT
1497#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1498
1499/* get key */
1500#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1501
1502/* get the key */
1503#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1504
1505/* get item body */
1506#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1507
1508/* get the stat data by the buffer header and the item order */
1509#define B_N_STAT_DATA(bh,nr) \
1510( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1511
1512 /* following defines use reiserfs buffer header and item header */
1513
1514/* get stat-data */
1515#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1516
1517// this is 3976 for size==4096
1518#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1519
1520/* indirect items consist of entries which contain blocknrs, pos
1521 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1522 blocknr contained by the entry pos points to */
1523#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1524#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)
1525
1526struct reiserfs_iget_args {
bd4c625c
LT
1527 __u32 objectid;
1528 __u32 dirid;
1529};
1da177e4
LT
1530
1531/***************************************************************************/
1532/* FUNCTION DECLARATIONS */
1533/***************************************************************************/
1534
1535/*#ifdef __KERNEL__*/
1536#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1537
1538#define journal_trans_half(blocksize) \
1539 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1540
1541/* journal.c see journal.c for all the comments here */
1542
1543/* first block written in a commit. */
1544struct reiserfs_journal_desc {
bd4c625c
LT
1545 __le32 j_trans_id; /* id of commit */
1546 __le32 j_len; /* length of commit. len +1 is the commit block */
1547 __le32 j_mount_id; /* mount id of this trans */
1548 __le32 j_realblock[1]; /* real locations for each block */
1549};
1da177e4
LT
1550
1551#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1552#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1553#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1554
1555#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1556#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1557#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1558
1559/* last block written in a commit */
1560struct reiserfs_journal_commit {
bd4c625c
LT
1561 __le32 j_trans_id; /* must match j_trans_id from the desc block */
1562 __le32 j_len; /* ditto */
1563 __le32 j_realblock[1]; /* real locations for each block */
1564};
1da177e4
LT
1565
1566#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1567#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1568#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1569
1570#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1571#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1572
1573/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1574** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1575** and this transaction does not need to be replayed.
1576*/
1577struct reiserfs_journal_header {
bd4c625c
LT
1578 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
1579 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
1580 __le32 j_mount_id;
1581 /* 12 */ struct journal_params jh_journal;
1582};
1da177e4
LT
1583
1584/* biggest tunable defines are right here */
bd4c625c
LT
1585#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1586#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1da177e4 1587#define JOURNAL_TRANS_MIN_DEFAULT 256
bd4c625c 1588#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1da177e4 1589#define JOURNAL_MIN_RATIO 2
bd4c625c 1590#define JOURNAL_MAX_COMMIT_AGE 30
1da177e4
LT
1591#define JOURNAL_MAX_TRANS_AGE 30
1592#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1593#ifdef CONFIG_QUOTA
556a2a45
JK
1594/* We need to update data and inode (atime) */
1595#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? 2 : 0)
1596/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1597#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1598(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1599/* same as with INIT */
1600#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1601(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
1da177e4 1602#else
556a2a45
JK
1603#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1604#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1605#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
1da177e4
LT
1606#endif
1607
1608/* both of these can be as low as 1, or as high as you want. The min is the
1609** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1610** as needed, and released when transactions are committed. On release, if
1611** the current number of nodes is > max, the node is freed, otherwise,
1612** it is put on a free list for faster use later.
1613*/
bd4c625c
LT
1614#define REISERFS_MIN_BITMAP_NODES 10
1615#define REISERFS_MAX_BITMAP_NODES 100
1da177e4 1616
bd4c625c 1617#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1da177e4
LT
1618#define JBH_HASH_MASK 8191
1619
1620#define _jhashfn(sb,block) \
1621 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1622 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1623#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1624
1625// We need these to make journal.c code more readable
1626#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1627#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1628#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1629
1630enum reiserfs_bh_state_bits {
bd4c625c
LT
1631 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1632 BH_JDirty_wait,
1633 BH_JNew, /* disk block was taken off free list before
1634 * being in a finished transaction, or
1635 * written to disk. Can be reused immed. */
1636 BH_JPrepared,
1637 BH_JRestore_dirty,
1638 BH_JTest, // debugging only will go away
1da177e4
LT
1639};
1640
1641BUFFER_FNS(JDirty, journaled);
1642TAS_BUFFER_FNS(JDirty, journaled);
1643BUFFER_FNS(JDirty_wait, journal_dirty);
1644TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1645BUFFER_FNS(JNew, journal_new);
1646TAS_BUFFER_FNS(JNew, journal_new);
1647BUFFER_FNS(JPrepared, journal_prepared);
1648TAS_BUFFER_FNS(JPrepared, journal_prepared);
1649BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1650TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1651BUFFER_FNS(JTest, journal_test);
1652TAS_BUFFER_FNS(JTest, journal_test);
1653
1654/*
1655** transaction handle which is passed around for all journal calls
1656*/
1657struct reiserfs_transaction_handle {
bd4c625c
LT
1658 struct super_block *t_super; /* super for this FS when journal_begin was
1659 called. saves calls to reiserfs_get_super
1660 also used by nested transactions to make
1661 sure they are nesting on the right FS
1662 _must_ be first in the handle
1663 */
1664 int t_refcount;
1665 int t_blocks_logged; /* number of blocks this writer has logged */
1666 int t_blocks_allocated; /* number of blocks this writer allocated */
1667 unsigned long t_trans_id; /* sanity check, equals the current trans id */
1668 void *t_handle_save; /* save existing current->journal_info */
1669 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1670 should be displaced from others */
1671 struct list_head t_list;
1672};
1da177e4
LT
1673
1674/* used to keep track of ordered and tail writes, attached to the buffer
1675 * head through b_journal_head.
1676 */
1677struct reiserfs_jh {
bd4c625c
LT
1678 struct reiserfs_journal_list *jl;
1679 struct buffer_head *bh;
1680 struct list_head list;
1da177e4
LT
1681};
1682
1683void reiserfs_free_jh(struct buffer_head *bh);
1684int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1685int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
bd4c625c
LT
1686int journal_mark_dirty(struct reiserfs_transaction_handle *,
1687 struct super_block *, struct buffer_head *bh);
1688
1689static inline int reiserfs_file_data_log(struct inode *inode)
1690{
1691 if (reiserfs_data_log(inode->i_sb) ||
1692 (REISERFS_I(inode)->i_flags & i_data_log))
1693 return 1;
1694 return 0;
1da177e4
LT
1695}
1696
bd4c625c
LT
1697static inline int reiserfs_transaction_running(struct super_block *s)
1698{
1699 struct reiserfs_transaction_handle *th = current->journal_info;
1700 if (th && th->t_super == s)
1701 return 1;
1702 if (th && th->t_super == NULL)
1703 BUG();
1704 return 0;
1da177e4
LT
1705}
1706
1707int reiserfs_async_progress_wait(struct super_block *s);
1708
bd4c625c
LT
1709struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1710 super_block
1711 *,
1712 int count);
1da177e4
LT
1713int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1714int reiserfs_commit_page(struct inode *inode, struct page *page,
bd4c625c 1715 unsigned from, unsigned to);
1da177e4 1716int reiserfs_flush_old_commits(struct super_block *);
bd4c625c
LT
1717int reiserfs_commit_for_inode(struct inode *);
1718int reiserfs_inode_needs_commit(struct inode *);
1719void reiserfs_update_inode_transaction(struct inode *);
1720void reiserfs_wait_on_write_block(struct super_block *s);
1721void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1722void reiserfs_allow_writes(struct super_block *s);
1723void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1724int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1725 int wait);
1726void reiserfs_restore_prepared_buffer(struct super_block *,
1727 struct buffer_head *bh);
1728int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1729 unsigned int);
1730int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1731int journal_release_error(struct reiserfs_transaction_handle *,
1732 struct super_block *);
1733int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1734 unsigned long);
1735int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1736 unsigned long);
1737int journal_mark_freed(struct reiserfs_transaction_handle *,
1738 struct super_block *, b_blocknr_t blocknr);
1739int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
1740int reiserfs_in_journal(struct super_block *p_s_sb, int bmap_nr, int bit_nr,
1741 int searchall, b_blocknr_t * next);
1742int journal_begin(struct reiserfs_transaction_handle *,
1743 struct super_block *p_s_sb, unsigned long);
1744int journal_join_abort(struct reiserfs_transaction_handle *,
1745 struct super_block *p_s_sb, unsigned long);
1746void reiserfs_journal_abort(struct super_block *sb, int errno);
1747void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1748int reiserfs_allocate_list_bitmaps(struct super_block *s,
1749 struct reiserfs_list_bitmap *, int);
1750
1751void add_save_link(struct reiserfs_transaction_handle *th,
1752 struct inode *inode, int truncate);
1753int remove_save_link(struct inode *inode, int truncate);
1da177e4
LT
1754
1755/* objectid.c */
bd4c625c
LT
1756__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1757void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1758 __u32 objectid_to_release);
1759int reiserfs_convert_objectid_map_v1(struct super_block *);
1da177e4
LT
1760
1761/* stree.c */
1762int B_IS_IN_TREE(const struct buffer_head *);
bd4c625c
LT
1763extern void copy_item_head(struct item_head *p_v_to,
1764 const struct item_head *p_v_from);
1da177e4
LT
1765
1766// first key is in cpu form, second - le
bd4c625c
LT
1767extern int comp_short_keys(const struct reiserfs_key *le_key,
1768 const struct cpu_key *cpu_key);
1769extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
1da177e4
LT
1770
1771// both are in le form
bd4c625c
LT
1772extern int comp_le_keys(const struct reiserfs_key *,
1773 const struct reiserfs_key *);
1774extern int comp_short_le_keys(const struct reiserfs_key *,
1775 const struct reiserfs_key *);
1da177e4
LT
1776
1777//
1778// get key version from on disk key - kludge
1779//
bd4c625c 1780static inline int le_key_version(const struct reiserfs_key *key)
1da177e4 1781{
bd4c625c 1782 int type;
1da177e4 1783
bd4c625c
LT
1784 type = offset_v2_k_type(&(key->u.k_offset_v2));
1785 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1786 && type != TYPE_DIRENTRY)
1787 return KEY_FORMAT_3_5;
1788
1789 return KEY_FORMAT_3_6;
1da177e4 1790
1da177e4
LT
1791}
1792
bd4c625c
LT
1793static inline void copy_key(struct reiserfs_key *to,
1794 const struct reiserfs_key *from)
1795{
1796 memcpy(to, from, KEY_SIZE);
1797}
1da177e4 1798
bd4c625c
LT
1799int comp_items(const struct item_head *stored_ih, const struct path *p_s_path);
1800const struct reiserfs_key *get_rkey(const struct path *p_s_chk_path,
1801 const struct super_block *p_s_sb);
1802int search_by_key(struct super_block *, const struct cpu_key *,
1803 struct path *, int);
1da177e4 1804#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
bd4c625c
LT
1805int search_for_position_by_key(struct super_block *p_s_sb,
1806 const struct cpu_key *p_s_cpu_key,
1807 struct path *p_s_search_path);
1808extern void decrement_bcount(struct buffer_head *p_s_bh);
1809void decrement_counters_in_path(struct path *p_s_search_path);
1810void pathrelse(struct path *p_s_search_path);
1811int reiserfs_check_path(struct path *p);
1812void pathrelse_and_restore(struct super_block *s, struct path *p_s_search_path);
1813
1814int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
1815 struct path *path,
1816 const struct cpu_key *key,
1817 struct item_head *ih,
1818 struct inode *inode, const char *body);
1819
1820int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
1821 struct path *path,
1822 const struct cpu_key *key,
1823 struct inode *inode,
1824 const char *body, int paste_size);
1825
1826int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
1827 struct path *path,
1828 struct cpu_key *key,
1829 struct inode *inode,
1830 struct page *page, loff_t new_file_size);
1831
1832int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
1833 struct path *path,
1834 const struct cpu_key *key,
1835 struct inode *inode, struct buffer_head *p_s_un_bh);
1836
1837void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1838 struct inode *inode, struct reiserfs_key *key);
1839int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
1840 struct inode *p_s_inode);
1841int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
1842 struct inode *p_s_inode, struct page *,
1843 int update_timestamps);
1da177e4
LT
1844
1845#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1846#define file_size(inode) ((inode)->i_size)
1847#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1848
1849#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1850!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 )
1851
bd4c625c 1852void padd_item(char *item, int total_length, int length);
1da177e4
LT
1853
1854/* inode.c */
1855/* args for the create parameter of reiserfs_get_block */
bd4c625c
LT
1856#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1857#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1858#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1859#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
1b1dcc1b 1860#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
bd4c625c
LT
1861#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
1862
1863int restart_transaction(struct reiserfs_transaction_handle *th,
1864 struct inode *inode, struct path *path);
1865void reiserfs_read_locked_inode(struct inode *inode,
1866 struct reiserfs_iget_args *args);
1867int reiserfs_find_actor(struct inode *inode, void *p);
1868int reiserfs_init_locked_inode(struct inode *inode, void *p);
1869void reiserfs_delete_inode(struct inode *inode);
1870int reiserfs_write_inode(struct inode *inode, int);
1871int reiserfs_get_block(struct inode *inode, sector_t block,
1872 struct buffer_head *bh_result, int create);
1873struct dentry *reiserfs_get_dentry(struct super_block *, void *);
1874struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 * data,
1875 int len, int fhtype,
1876 int (*acceptable) (void *contect,
1877 struct dentry * de),
1878 void *context);
1879int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1880 int connectable);
1881
1882int reiserfs_truncate_file(struct inode *, int update_timestamps);
1883void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
1884 int type, int key_length);
1885void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
1886 int version,
1887 loff_t offset, int type, int length, int entry_count);
1888struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
1889
1890int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1891 struct inode *dir, int mode,
1892 const char *symname, loff_t i_size,
1893 struct dentry *dentry, struct inode *inode);
1894
1895void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1896 struct inode *inode, loff_t size);
1da177e4
LT
1897
1898static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
bd4c625c 1899 struct inode *inode)
1da177e4 1900{
bd4c625c 1901 reiserfs_update_sd_size(th, inode, inode->i_size);
1da177e4
LT
1902}
1903
bd4c625c
LT
1904void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
1905void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
1da177e4
LT
1906int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1907
1908/* namei.c */
bd4c625c
LT
1909void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
1910int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
1911 struct path *path, struct reiserfs_dir_entry *de);
1912struct dentry *reiserfs_get_parent(struct dentry *);
1da177e4
LT
1913/* procfs.c */
1914
1915#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1916#define REISERFS_PROC_INFO
1917#else
1918#undef REISERFS_PROC_INFO
1919#endif
1920
bd4c625c
LT
1921int reiserfs_proc_info_init(struct super_block *sb);
1922int reiserfs_proc_info_done(struct super_block *sb);
1923struct proc_dir_entry *reiserfs_proc_register_global(char *name,
1924 read_proc_t * func);
1925void reiserfs_proc_unregister_global(const char *name);
1926int reiserfs_proc_info_global_init(void);
1927int reiserfs_proc_info_global_done(void);
1928int reiserfs_global_version_in_proc(char *buffer, char **start, off_t offset,
1929 int count, int *eof, void *data);
1da177e4
LT
1930
1931#if defined( REISERFS_PROC_INFO )
1932
1933#define PROC_EXP( e ) e
1934
1935#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1936#define PROC_INFO_MAX( sb, field, value ) \
1937 __PINFO( sb ).field = \
1938 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1939#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1940#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1941#define PROC_INFO_BH_STAT( sb, bh, level ) \
1942 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1943 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1944 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1945#else
1946#define PROC_EXP( e )
1947#define VOID_V ( ( void ) 0 )
1948#define PROC_INFO_MAX( sb, field, value ) VOID_V
1949#define PROC_INFO_INC( sb, field ) VOID_V
1950#define PROC_INFO_ADD( sb, field, val ) VOID_V
1951#define PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level ) VOID_V
1952#endif
1953
1954/* dir.c */
1955extern struct inode_operations reiserfs_dir_inode_operations;
1956extern struct inode_operations reiserfs_symlink_inode_operations;
1957extern struct inode_operations reiserfs_special_inode_operations;
1958extern struct file_operations reiserfs_dir_operations;
1959
1960/* tail_conversion.c */
bd4c625c
LT
1961int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
1962 struct path *, struct buffer_head *, loff_t);
1963int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
1964 struct page *, struct path *, const struct cpu_key *,
1965 loff_t, char *);
1966void reiserfs_unmap_buffer(struct buffer_head *);
1da177e4
LT
1967
1968/* file.c */
1969extern struct inode_operations reiserfs_file_inode_operations;
1970extern struct file_operations reiserfs_file_operations;
bd4c625c 1971extern struct address_space_operations reiserfs_address_space_operations;
1da177e4
LT
1972
1973/* fix_nodes.c */
1974#ifdef CONFIG_REISERFS_CHECK
27496a8c 1975void *reiserfs_kmalloc(size_t size, gfp_t flags, struct super_block *s);
bd4c625c 1976void reiserfs_kfree(const void *vp, size_t size, struct super_block *s);
1da177e4
LT
1977#else
1978static inline void *reiserfs_kmalloc(size_t size, int flags,
bd4c625c 1979 struct super_block *s)
1da177e4
LT
1980{
1981 return kmalloc(size, flags);
1982}
1983
1984static inline void reiserfs_kfree(const void *vp, size_t size,
bd4c625c 1985 struct super_block *s)
1da177e4
LT
1986{
1987 kfree(vp);
1988}
1989#endif
1990
bd4c625c
LT
1991int fix_nodes(int n_op_mode, struct tree_balance *p_s_tb,
1992 struct item_head *p_s_ins_ih, const void *);
1993void unfix_nodes(struct tree_balance *);
1da177e4
LT
1994
1995/* prints.c */
bd4c625c
LT
1996void reiserfs_panic(struct super_block *s, const char *fmt, ...)
1997 __attribute__ ((noreturn));
1998void reiserfs_info(struct super_block *s, const char *fmt, ...);
1999void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2000void print_indirect_item(struct buffer_head *bh, int item_num);
2001void store_print_tb(struct tree_balance *tb);
2002void print_cur_tb(char *mes);
2003void print_de(struct reiserfs_dir_entry *de);
2004void print_bi(struct buffer_info *bi, char *mes);
2005#define PRINT_LEAF_ITEMS 1 /* print all items */
2006#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2007#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2008void print_block(struct buffer_head *bh, ...);
2009void print_bmap(struct super_block *s, int silent);
2010void print_bmap_block(int i, char *data, int size, int silent);
1da177e4 2011/*void print_super_block (struct super_block * s, char * mes);*/
bd4c625c
LT
2012void print_objectid_map(struct super_block *s);
2013void print_block_head(struct buffer_head *bh, char *mes);
2014void check_leaf(struct buffer_head *bh);
2015void check_internal(struct buffer_head *bh);
2016void print_statistics(struct super_block *s);
2017char *reiserfs_hashname(int code);
1da177e4
LT
2018
2019/* lbalance.c */
bd4c625c
LT
2020int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2021 int mov_bytes, struct buffer_head *Snew);
2022int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2023int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2024void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2025 int del_num, int del_bytes);
2026void leaf_insert_into_buf(struct buffer_info *bi, int before,
2027 struct item_head *inserted_item_ih,
2028 const char *inserted_item_body, int zeros_number);
2029void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2030 int pos_in_item, int paste_size, const char *body,
2031 int zeros_number);
2032void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2033 int pos_in_item, int cut_size);
2034void leaf_paste_entries(struct buffer_head *bh, int item_num, int before,
2035 int new_entry_count, struct reiserfs_de_head *new_dehs,
2036 const char *records, int paste_size);
1da177e4 2037/* ibalance.c */
bd4c625c
LT
2038int balance_internal(struct tree_balance *, int, int, struct item_head *,
2039 struct buffer_head **);
1da177e4
LT
2040
2041/* do_balance.c */
bd4c625c
LT
2042void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2043 struct buffer_head *bh, int flag);
1da177e4
LT
2044#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2045#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2046
bd4c625c
LT
2047void do_balance(struct tree_balance *tb, struct item_head *ih,
2048 const char *body, int flag);
2049void reiserfs_invalidate_buffer(struct tree_balance *tb,
2050 struct buffer_head *bh);
1da177e4 2051
bd4c625c
LT
2052int get_left_neighbor_position(struct tree_balance *tb, int h);
2053int get_right_neighbor_position(struct tree_balance *tb, int h);
2054void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2055 struct buffer_head *, int);
2056void make_empty_node(struct buffer_info *);
2057struct buffer_head *get_FEB(struct tree_balance *);
1da177e4
LT
2058
2059/* bitmap.c */
2060
2061/* structure contains hints for block allocator, and it is a container for
2062 * arguments, such as node, search path, transaction_handle, etc. */
bd4c625c
LT
2063struct __reiserfs_blocknr_hint {
2064 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
2065 long block; /* file offset, in blocks */
2066 struct in_core_key key;
2067 struct path *path; /* search path, used by allocator to deternine search_start by
2068 * various ways */
2069 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2070 * bitmap blocks changes */
2071 b_blocknr_t beg, end;
2072 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
1da177e4
LT
2073 * between different block allocator procedures
2074 * (determine_search_start() and others) */
bd4c625c
LT
2075 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2076 * function that do actual allocation */
1da177e4 2077
bd4c625c 2078 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
1da177e4 2079 * formatted/unformatted blocks with/without preallocation */
bd4c625c 2080 unsigned preallocate:1;
1da177e4
LT
2081};
2082
2083typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2084
bd4c625c
LT
2085int reiserfs_parse_alloc_options(struct super_block *, char *);
2086void reiserfs_init_alloc_options(struct super_block *s);
1da177e4
LT
2087
2088/*
2089 * given a directory, this will tell you what packing locality
2090 * to use for a new object underneat it. The locality is returned
2091 * in disk byte order (le).
2092 */
3e8962be 2093__le32 reiserfs_choose_packing(struct inode *dir);
1da177e4 2094
bd4c625c
LT
2095int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2096void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2097 b_blocknr_t, int for_unformatted);
2098int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2099 int);
9adeb1b4 2100static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
bd4c625c
LT
2101 b_blocknr_t * new_blocknrs,
2102 int amount_needed)
1da177e4 2103{
bd4c625c
LT
2104 reiserfs_blocknr_hint_t hint = {
2105 .th = tb->transaction_handle,
2106 .path = tb->tb_path,
2107 .inode = NULL,
2108 .key = tb->key,
2109 .block = 0,
2110 .formatted_node = 1
2111 };
2112 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2113 0);
1da177e4
LT
2114}
2115
9adeb1b4 2116static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
bd4c625c
LT
2117 *th, struct inode *inode,
2118 b_blocknr_t * new_blocknrs,
2119 struct path *path, long block)
1da177e4 2120{
bd4c625c
LT
2121 reiserfs_blocknr_hint_t hint = {
2122 .th = th,
2123 .path = path,
2124 .inode = inode,
2125 .block = block,
2126 .formatted_node = 0,
2127 .preallocate = 0
2128 };
2129 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2130}
2131
2132#ifdef REISERFS_PREALLOCATE
9adeb1b4 2133static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
bd4c625c
LT
2134 *th, struct inode *inode,
2135 b_blocknr_t * new_blocknrs,
2136 struct path *path, long block)
1da177e4 2137{
bd4c625c
LT
2138 reiserfs_blocknr_hint_t hint = {
2139 .th = th,
2140 .path = path,
2141 .inode = inode,
2142 .block = block,
2143 .formatted_node = 0,
2144 .preallocate = 1
2145 };
2146 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2147}
2148
bd4c625c
LT
2149void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2150 struct inode *inode);
2151void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
1da177e4 2152#endif
bd4c625c
LT
2153void reiserfs_claim_blocks_to_be_allocated(struct super_block *sb, int blocks);
2154void reiserfs_release_claimed_blocks(struct super_block *sb, int blocks);
1da177e4
LT
2155int reiserfs_can_fit_pages(struct super_block *sb);
2156
2157/* hashes.c */
bd4c625c
LT
2158__u32 keyed_hash(const signed char *msg, int len);
2159__u32 yura_hash(const signed char *msg, int len);
2160__u32 r5_hash(const signed char *msg, int len);
1da177e4
LT
2161
2162/* the ext2 bit routines adjust for big or little endian as
2163** appropriate for the arch, so in our laziness we use them rather
2164** than using the bit routines they call more directly. These
2165** routines must be used when changing on disk bitmaps. */
2166#define reiserfs_test_and_set_le_bit ext2_set_bit
2167#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2168#define reiserfs_test_le_bit ext2_test_bit
2169#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2170
2171/* sometimes reiserfs_truncate may require to allocate few new blocks
2172 to perform indirect2direct conversion. People probably used to
2173 think, that truncate should work without problems on a filesystem
2174 without free disk space. They may complain that they can not
2175 truncate due to lack of free disk space. This spare space allows us
2176 to not worry about it. 500 is probably too much, but it should be
2177 absolutely safe */
2178#define SPARE_SPACE 500
2179
1da177e4 2180/* prototypes from ioctl.c */
bd4c625c
LT
2181int reiserfs_ioctl(struct inode *inode, struct file *filp,
2182 unsigned int cmd, unsigned long arg);
2183
1da177e4
LT
2184/* ioctl's command */
2185#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
2186/* define following flags to be the same as in ext2, so that chattr(1),
2187 lsattr(1) will work with us. */
2188#define REISERFS_IOC_GETFLAGS EXT2_IOC_GETFLAGS
2189#define REISERFS_IOC_SETFLAGS EXT2_IOC_SETFLAGS
2190#define REISERFS_IOC_GETVERSION EXT2_IOC_GETVERSION
2191#define REISERFS_IOC_SETVERSION EXT2_IOC_SETVERSION
2192
2193/* Locking primitives */
2194/* Right now we are still falling back to (un)lock_kernel, but eventually that
2195 would evolve into real per-fs locks */
2196#define reiserfs_write_lock( sb ) lock_kernel()
2197#define reiserfs_write_unlock( sb ) unlock_kernel()
bd4c625c 2198
1da177e4
LT
2199/* xattr stuff */
2200#define REISERFS_XATTR_DIR_SEM(s) (REISERFS_SB(s)->xattr_dir_sem)
2201
bd4c625c 2202#endif /* _LINUX_REISER_FS_H */