blob: 32148625fc2f0b025a9070fc93650c2fd84903dc [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
3 */
4
5 /* this file has an amazingly stupid
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. */
10
11
12#ifndef _LINUX_REISER_FS_H
13#define _LINUX_REISER_FS_H
14
15#include <linux/types.h>
16#ifdef __KERNEL__
17#include <linux/slab.h>
18#include <linux/interrupt.h>
19#include <linux/sched.h>
20#include <linux/workqueue.h>
21#include <asm/unaligned.h>
22#include <linux/bitops.h>
23#include <linux/proc_fs.h>
24#include <linux/smp_lock.h>
25#include <linux/buffer_head.h>
26#include <linux/reiserfs_fs_i.h>
27#include <linux/reiserfs_fs_sb.h>
28#endif
29
30/*
31 * include/linux/reiser_fs.h
32 *
33 * Reiser File System constants and structures
34 *
35 */
36
37/* in reading the #defines, it may help to understand that they employ
38 the following abbreviations:
39
40 B = Buffer
41 I = Item header
42 H = Height within the tree (should be changed to LEV)
43 N = Number of the item in the node
44 STAT = stat data
45 DEH = Directory Entry Header
46 EC = Entry Count
47 E = Entry number
48 UL = Unsigned Long
49 BLKH = BLocK Header
50 UNFM = UNForMatted node
51 DC = Disk Child
52 P = Path
53
54 These #defines are named by concatenating these abbreviations,
55 where first comes the arguments, and last comes the return value,
56 of the macro.
57
58*/
59
60#define USE_INODE_GENERATION_COUNTER
61
62#define REISERFS_PREALLOCATE
63#define DISPLACE_NEW_PACKING_LOCALITIES
64#define PREALLOCATION_SIZE 9
65
66/* n must be power of 2 */
67#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
68
69// to be ok for alpha and others we have to align structures to 8 byte
70// boundary.
71// FIXME: do not change 4 by anything else: there is code which relies on that
72#define ROUND_UP(x) _ROUND_UP(x,8LL)
73
74/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
75** messages.
76*/
77#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
78
79void reiserfs_warning (struct super_block *s, const char * fmt, ...);
80/* assertions handling */
81
82/** always check a condition and panic if it's false. */
83#define RASSERT( cond, format, args... ) \
84if( !( cond ) ) \
85 reiserfs_panic( NULL, "reiserfs[%i]: assertion " #cond " failed at " \
86 __FILE__ ":%i:%s: " format "\n", \
87 in_interrupt() ? -1 : current -> pid, __LINE__ , __FUNCTION__ , ##args )
88
89#if defined( CONFIG_REISERFS_CHECK )
90#define RFALSE( cond, format, args... ) RASSERT( !( cond ), format, ##args )
91#else
92#define RFALSE( cond, format, args... ) do {;} while( 0 )
93#endif
94
95#define CONSTF __attribute_const__
96/*
97 * Disk Data Structures
98 */
99
100/***************************************************************************/
101/* SUPER BLOCK */
102/***************************************************************************/
103
104/*
105 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
106 * the version in RAM is part of a larger structure containing fields never written to disk.
107 */
108#define UNSET_HASH 0 // read_super will guess about, what hash names
109 // in directories were sorted with
110#define TEA_HASH 1
111#define YURA_HASH 2
112#define R5_HASH 3
113#define DEFAULT_HASH R5_HASH
114
115
116struct journal_params {
Al Viro3e8962b2005-05-01 08:59:18 -0700117 __le32 jp_journal_1st_block; /* where does journal start from on its
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 * device */
Al Viro3e8962b2005-05-01 08:59:18 -0700119 __le32 jp_journal_dev; /* journal device st_rdev */
120 __le32 jp_journal_size; /* size of the journal */
121 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
122 __le32 jp_journal_magic; /* random value made on fs creation (this
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 * was sb_journal_block_count) */
Al Viro3e8962b2005-05-01 08:59:18 -0700124 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 * trans */
Al Viro3e8962b2005-05-01 08:59:18 -0700126 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 * commit be */
Al Viro3e8962b2005-05-01 08:59:18 -0700128 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 * be */
130};
131
132/* this is the super from 3.5.X, where X >= 10 */
133struct reiserfs_super_block_v1
134{
Al Viro3e8962b2005-05-01 08:59:18 -0700135 __le32 s_block_count; /* blocks count */
136 __le32 s_free_blocks; /* free blocks count */
137 __le32 s_root_block; /* root block number */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138 struct journal_params s_journal;
Al Viro3e8962b2005-05-01 08:59:18 -0700139 __le16 s_blocksize; /* block size */
140 __le16 s_oid_maxsize; /* max size of object id array, see
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 * get_objectid() commentary */
Al Viro3e8962b2005-05-01 08:59:18 -0700142 __le16 s_oid_cursize; /* current size of object id array */
143 __le16 s_umount_state; /* this is set to 1 when filesystem was
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 * umounted, to 2 - when not */
145 char s_magic[10]; /* reiserfs magic string indicates that
146 * file system is reiserfs:
147 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
Al Viro3e8962b2005-05-01 08:59:18 -0700148 __le16 s_fs_state; /* it is set to used by fsck to mark which
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 * phase of rebuilding is done */
Al Viro3e8962b2005-05-01 08:59:18 -0700150 __le32 s_hash_function_code; /* indicate, what hash function is being use
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 * to sort names in a directory*/
Al Viro3e8962b2005-05-01 08:59:18 -0700152 __le16 s_tree_height; /* height of disk tree */
153 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 * each block of file system */
Al Viro3e8962b2005-05-01 08:59:18 -0700155 __le16 s_version; /* this field is only reliable on filesystem
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 * with non-standard journal */
Al Viro3e8962b2005-05-01 08:59:18 -0700157 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 * device, we need to keep after
159 * making fs with non-standard journal */
160} __attribute__ ((__packed__));
161
162#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
163
164/* this is the on disk super block */
165struct reiserfs_super_block
166{
167 struct reiserfs_super_block_v1 s_v1;
Al Viro3e8962b2005-05-01 08:59:18 -0700168 __le32 s_inode_generation;
169 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 unsigned char s_uuid[16]; /* filesystem unique identifier */
171 unsigned char s_label[16]; /* filesystem volume label */
172 char s_unused[88] ; /* zero filled by mkreiserfs and
173 * reiserfs_convert_objectid_map_v1()
174 * so any additions must be updated
175 * there as well. */
176} __attribute__ ((__packed__));
177
178#define SB_SIZE (sizeof(struct reiserfs_super_block))
179
180#define REISERFS_VERSION_1 0
181#define REISERFS_VERSION_2 2
182
183
184// on-disk super block fields converted to cpu form
185#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
186#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
187#define SB_BLOCKSIZE(s) \
188 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
189#define SB_BLOCK_COUNT(s) \
190 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
191#define SB_FREE_BLOCKS(s) \
192 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
193#define SB_REISERFS_MAGIC(s) \
194 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
195#define SB_ROOT_BLOCK(s) \
196 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
197#define SB_TREE_HEIGHT(s) \
198 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
199#define SB_REISERFS_STATE(s) \
200 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
201#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
202#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
203
204#define PUT_SB_BLOCK_COUNT(s, val) \
205 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
206#define PUT_SB_FREE_BLOCKS(s, val) \
207 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
208#define PUT_SB_ROOT_BLOCK(s, val) \
209 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
210#define PUT_SB_TREE_HEIGHT(s, val) \
211 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
212#define PUT_SB_REISERFS_STATE(s, val) \
213 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
214#define PUT_SB_VERSION(s, val) \
215 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
216#define PUT_SB_BMAP_NR(s, val) \
217 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
218
219
220#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
221#define SB_ONDISK_JOURNAL_SIZE(s) \
222 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
223#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
224 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
225#define SB_ONDISK_JOURNAL_DEVICE(s) \
226 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
227#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
Al Virob8cc9362005-05-01 08:59:18 -0700228 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
230#define is_block_in_log_or_reserved_area(s, block) \
231 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
232 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
233 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
234 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
235
236
237
238 /* used by gcc */
239#define REISERFS_SUPER_MAGIC 0x52654973
240 /* used by file system utilities that
241 look at the superblock, etc. */
242#define REISERFS_SUPER_MAGIC_STRING "ReIsErFs"
243#define REISER2FS_SUPER_MAGIC_STRING "ReIsEr2Fs"
244#define REISER2FS_JR_SUPER_MAGIC_STRING "ReIsEr3Fs"
245
246int is_reiserfs_3_5 (struct reiserfs_super_block * rs);
247int is_reiserfs_3_6 (struct reiserfs_super_block * rs);
248int is_reiserfs_jr (struct reiserfs_super_block * rs);
249
250/* ReiserFS leaves the first 64k unused, so that partition labels have
251 enough space. If someone wants to write a fancy bootloader that
252 needs more than 64k, let us know, and this will be increased in size.
253 This number must be larger than than the largest block size on any
254 platform, or code will break. -Hans */
255#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
256#define REISERFS_FIRST_BLOCK unused_define
257#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
258
259/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
260#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
261
262// reiserfs internal error code (used by search_by_key adn fix_nodes))
263#define CARRY_ON 0
264#define REPEAT_SEARCH -1
265#define IO_ERROR -2
266#define NO_DISK_SPACE -3
267#define NO_BALANCING_NEEDED (-4)
268#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
269#define QUOTA_EXCEEDED -6
270
271typedef __u32 b_blocknr_t;
Al Viro3e8962b2005-05-01 08:59:18 -0700272typedef __le32 unp_t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273
274struct unfm_nodeinfo {
275 unp_t unfm_nodenum;
276 unsigned short unfm_freespace;
277};
278
279/* there are two formats of keys: 3.5 and 3.6
280 */
281#define KEY_FORMAT_3_5 0
282#define KEY_FORMAT_3_6 1
283
284/* there are two stat datas */
285#define STAT_DATA_V1 0
286#define STAT_DATA_V2 1
287
288
289static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
290{
291 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
292}
293
294static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
295{
296 return sb->s_fs_info;
297}
298
299/** this says about version of key of all items (but stat data) the
300 object consists of */
301#define get_inode_item_key_version( inode ) \
302 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
303
304#define set_inode_item_key_version( inode, version ) \
305 ({ if((version)==KEY_FORMAT_3_6) \
306 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
307 else \
308 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
309
310#define get_inode_sd_version(inode) \
311 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
312
313#define set_inode_sd_version(inode, version) \
314 ({ if((version)==STAT_DATA_V2) \
315 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
316 else \
317 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
318
319/* This is an aggressive tail suppression policy, I am hoping it
320 improves our benchmarks. The principle behind it is that percentage
321 space saving is what matters, not absolute space saving. This is
322 non-intuitive, but it helps to understand it if you consider that the
323 cost to access 4 blocks is not much more than the cost to access 1
324 block, if you have to do a seek and rotate. A tail risks a
325 non-linear disk access that is significant as a percentage of total
326 time cost for a 4 block file and saves an amount of space that is
327 less significant as a percentage of space, or so goes the hypothesis.
328 -Hans */
329#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
330(\
331 (!(n_tail_size)) || \
332 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
333 ( (n_file_size) >= (n_block_size) * 4 ) || \
334 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
335 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
336 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
337 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
338 ( ( (n_file_size) >= (n_block_size) ) && \
339 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
340)
341
342/* Another strategy for tails, this one means only create a tail if all the
343 file would fit into one DIRECT item.
344 Primary intention for this one is to increase performance by decreasing
345 seeking.
346*/
347#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
348(\
349 (!(n_tail_size)) || \
350 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
351)
352
353
354
355/*
356 * values for s_umount_state field
357 */
358#define REISERFS_VALID_FS 1
359#define REISERFS_ERROR_FS 2
360
361//
362// there are 5 item types currently
363//
364#define TYPE_STAT_DATA 0
365#define TYPE_INDIRECT 1
366#define TYPE_DIRECT 2
367#define TYPE_DIRENTRY 3
368#define TYPE_MAXTYPE 3
369#define TYPE_ANY 15 // FIXME: comment is required
370
371/***************************************************************************/
372/* KEY & ITEM HEAD */
373/***************************************************************************/
374
375//
376// directories use this key as well as old files
377//
378struct offset_v1 {
Al Viro3e8962b2005-05-01 08:59:18 -0700379 __le32 k_offset;
380 __le32 k_uniqueness;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381} __attribute__ ((__packed__));
382
383struct offset_v2 {
Al Virof8e08a82005-05-01 08:59:19 -0700384 __le64 v;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385} __attribute__ ((__packed__));
386
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387static inline __u16 offset_v2_k_type( const struct offset_v2 *v2 )
388{
Al Virof8e08a82005-05-01 08:59:19 -0700389 __u8 type = le64_to_cpu(v2->v) >> 60;
390 return (type <= TYPE_MAXTYPE)?type:TYPE_ANY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391}
392
393static inline void set_offset_v2_k_type( struct offset_v2 *v2, int type )
394{
Al Virof8e08a82005-05-01 08:59:19 -0700395 v2->v = (v2->v & cpu_to_le64(~0ULL>>4)) | cpu_to_le64((__u64)type<<60);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396}
397
398static inline loff_t offset_v2_k_offset( const struct offset_v2 *v2 )
399{
Al Virof8e08a82005-05-01 08:59:19 -0700400 return le64_to_cpu(v2->v) & (~0ULL>>4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401}
402
403static inline void set_offset_v2_k_offset( struct offset_v2 *v2, loff_t offset ){
Al Virof8e08a82005-05-01 08:59:19 -0700404 offset &= (~0ULL>>4);
405 v2->v = (v2->v & cpu_to_le64(15ULL<<60)) | cpu_to_le64(offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
408/* Key of an item determines its location in the S+tree, and
409 is composed of 4 components */
410struct reiserfs_key {
Al Viro3e8962b2005-05-01 08:59:18 -0700411 __le32 k_dir_id; /* packing locality: by default parent
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 directory object id */
Al Viro3e8962b2005-05-01 08:59:18 -0700413 __le32 k_objectid; /* object identifier */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 union {
415 struct offset_v1 k_offset_v1;
416 struct offset_v2 k_offset_v2;
417 } __attribute__ ((__packed__)) u;
418} __attribute__ ((__packed__));
419
Al Viro6a3a16f2005-05-01 08:59:17 -0700420struct in_core_key {
421 __u32 k_dir_id; /* packing locality: by default parent
422 directory object id */
423 __u32 k_objectid; /* object identifier */
Al Viro6b9f5822005-05-01 08:59:19 -0700424 __u64 k_offset;
425 __u8 k_type;
426};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
428struct cpu_key {
Al Viro6a3a16f2005-05-01 08:59:17 -0700429 struct in_core_key on_disk_key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 int version;
431 int key_length; /* 3 in all cases but direct2indirect and
432 indirect2direct conversion */
433};
434
435/* Our function for comparing keys can compare keys of different
436 lengths. It takes as a parameter the length of the keys it is to
437 compare. These defines are used in determining what is to be passed
438 to it as that parameter. */
439#define REISERFS_FULL_KEY_LEN 4
440#define REISERFS_SHORT_KEY_LEN 2
441
442/* The result of the key compare */
443#define FIRST_GREATER 1
444#define SECOND_GREATER -1
445#define KEYS_IDENTICAL 0
446#define KEY_FOUND 1
447#define KEY_NOT_FOUND 0
448
449#define KEY_SIZE (sizeof(struct reiserfs_key))
450#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
451
452/* return values for search_by_key and clones */
453#define ITEM_FOUND 1
454#define ITEM_NOT_FOUND 0
455#define ENTRY_FOUND 1
456#define ENTRY_NOT_FOUND 0
457#define DIRECTORY_NOT_FOUND -1
458#define REGULAR_FILE_FOUND -2
459#define DIRECTORY_FOUND -3
460#define BYTE_FOUND 1
461#define BYTE_NOT_FOUND 0
462#define FILE_NOT_FOUND -1
463
464#define POSITION_FOUND 1
465#define POSITION_NOT_FOUND 0
466
467// return values for reiserfs_find_entry and search_by_entry_key
468#define NAME_FOUND 1
469#define NAME_NOT_FOUND 0
470#define GOTO_PREVIOUS_ITEM 2
471#define NAME_FOUND_INVISIBLE 3
472
473/* Everything in the filesystem is stored as a set of items. The
474 item head contains the key of the item, its free space (for
475 indirect items) and specifies the location of the item itself
476 within the block. */
477
478struct item_head
479{
480 /* Everything in the tree is found by searching for it based on
481 * its key.*/
482 struct reiserfs_key ih_key;
483 union {
484 /* The free space in the last unformatted node of an
485 indirect item if this is an indirect item. This
486 equals 0xFFFF iff this is a direct item or stat data
487 item. Note that the key, not this field, is used to
488 determine the item type, and thus which field this
489 union contains. */
Al Viro3e8962b2005-05-01 08:59:18 -0700490 __le16 ih_free_space_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 /* Iff this is a directory item, this field equals the
492 number of directory entries in the directory item. */
Al Viro3e8962b2005-05-01 08:59:18 -0700493 __le16 ih_entry_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 } __attribute__ ((__packed__)) u;
Al Viro3e8962b2005-05-01 08:59:18 -0700495 __le16 ih_item_len; /* total size of the item body */
496 __le16 ih_item_location; /* an offset to the item body
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 * within the block */
Al Viro3e8962b2005-05-01 08:59:18 -0700498 __le16 ih_version; /* 0 for all old items, 2 for new
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 ones. Highest bit is set by fsck
500 temporary, cleaned after all
501 done */
502} __attribute__ ((__packed__));
503/* size of item header */
504#define IH_SIZE (sizeof(struct item_head))
505
506#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
507#define ih_version(ih) le16_to_cpu((ih)->ih_version)
508#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
509#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
510#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
511
512#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
513#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
514#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
515#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
516#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
517
518
519#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
520
521#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
522#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
523
524/* these operate on indirect items, where you've got an array of ints
525** at a possibly unaligned location. These are a noop on ia32
526**
527** p is the array of __u32, i is the index into the array, v is the value
528** to store there.
529*/
530#define get_block_num(p, i) le32_to_cpu(get_unaligned((p) + (i)))
531#define put_block_num(p, i, v) put_unaligned(cpu_to_le32(v), (p) + (i))
532
533//
534// in old version uniqueness field shows key type
535//
536#define V1_SD_UNIQUENESS 0
537#define V1_INDIRECT_UNIQUENESS 0xfffffffe
538#define V1_DIRECT_UNIQUENESS 0xffffffff
539#define V1_DIRENTRY_UNIQUENESS 500
540#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
541
542//
543// here are conversion routines
544//
545static inline int uniqueness2type (__u32 uniqueness) CONSTF;
546static inline int uniqueness2type (__u32 uniqueness)
547{
548 switch ((int)uniqueness) {
549 case V1_SD_UNIQUENESS: return TYPE_STAT_DATA;
550 case V1_INDIRECT_UNIQUENESS: return TYPE_INDIRECT;
551 case V1_DIRECT_UNIQUENESS: return TYPE_DIRECT;
552 case V1_DIRENTRY_UNIQUENESS: return TYPE_DIRENTRY;
553 default:
554 reiserfs_warning (NULL, "vs-500: unknown uniqueness %d",
555 uniqueness);
556 case V1_ANY_UNIQUENESS:
557 return TYPE_ANY;
558 }
559}
560
561static inline __u32 type2uniqueness (int type) CONSTF;
562static inline __u32 type2uniqueness (int type)
563{
564 switch (type) {
565 case TYPE_STAT_DATA: return V1_SD_UNIQUENESS;
566 case TYPE_INDIRECT: return V1_INDIRECT_UNIQUENESS;
567 case TYPE_DIRECT: return V1_DIRECT_UNIQUENESS;
568 case TYPE_DIRENTRY: return V1_DIRENTRY_UNIQUENESS;
569 default:
570 reiserfs_warning (NULL, "vs-501: unknown type %d", type);
571 case TYPE_ANY:
572 return V1_ANY_UNIQUENESS;
573 }
574}
575
576//
577// key is pointer to on disk key which is stored in le, result is cpu,
578// there is no way to get version of object from key, so, provide
579// version to these defines
580//
581static inline loff_t le_key_k_offset (int version, const struct reiserfs_key * key)
582{
583 return (version == KEY_FORMAT_3_5) ?
584 le32_to_cpu( key->u.k_offset_v1.k_offset ) :
585 offset_v2_k_offset( &(key->u.k_offset_v2) );
586}
587
588static inline loff_t le_ih_k_offset (const struct item_head * ih)
589{
590 return le_key_k_offset (ih_version (ih), &(ih->ih_key));
591}
592
593static inline loff_t le_key_k_type (int version, const struct reiserfs_key * key)
594{
595 return (version == KEY_FORMAT_3_5) ?
596 uniqueness2type( le32_to_cpu( key->u.k_offset_v1.k_uniqueness)) :
597 offset_v2_k_type( &(key->u.k_offset_v2) );
598}
599
600static inline loff_t le_ih_k_type (const struct item_head * ih)
601{
602 return le_key_k_type (ih_version (ih), &(ih->ih_key));
603}
604
605
606static inline void set_le_key_k_offset (int version, struct reiserfs_key * key, loff_t offset)
607{
608 (version == KEY_FORMAT_3_5) ?
609 (void)(key->u.k_offset_v1.k_offset = cpu_to_le32 (offset)) : /* jdm check */
610 (void)(set_offset_v2_k_offset( &(key->u.k_offset_v2), offset ));
611}
612
613
614static inline void set_le_ih_k_offset (struct item_head * ih, loff_t offset)
615{
616 set_le_key_k_offset (ih_version (ih), &(ih->ih_key), offset);
617}
618
619
620static inline void set_le_key_k_type (int version, struct reiserfs_key * key, int type)
621{
622 (version == KEY_FORMAT_3_5) ?
623 (void)(key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type2uniqueness(type))):
624 (void)(set_offset_v2_k_type( &(key->u.k_offset_v2), type ));
625}
626static inline void set_le_ih_k_type (struct item_head * ih, int type)
627{
628 set_le_key_k_type (ih_version (ih), &(ih->ih_key), type);
629}
630
631
632#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
633#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
634#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
635#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
636
637//
638// item header has version.
639//
640#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
641#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
642#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
643#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
644
645
646
647//
648// key is pointer to cpu key, result is cpu
649//
650static inline loff_t cpu_key_k_offset (const struct cpu_key * key)
651{
Al Viro6b9f5822005-05-01 08:59:19 -0700652 return key->on_disk_key.k_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653}
654
655static inline loff_t cpu_key_k_type (const struct cpu_key * key)
656{
Al Viro6b9f5822005-05-01 08:59:19 -0700657 return key->on_disk_key.k_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658}
659
660static inline void set_cpu_key_k_offset (struct cpu_key * key, loff_t offset)
661{
Al Viro6b9f5822005-05-01 08:59:19 -0700662 key->on_disk_key.k_offset = offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663}
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665static inline void set_cpu_key_k_type (struct cpu_key * key, int type)
666{
Al Viro6b9f5822005-05-01 08:59:19 -0700667 key->on_disk_key.k_type = type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668}
669
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670static inline void cpu_key_k_offset_dec (struct cpu_key * key)
671{
Al Viro6b9f5822005-05-01 08:59:19 -0700672 key->on_disk_key.k_offset --;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673}
674
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
676#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
677#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
678#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
679
680
681/* are these used ? */
682#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
683#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
684#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
685#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
686
687
688
689
690
691#define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
692 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
693 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
694
695/* maximal length of item */
696#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
697#define MIN_ITEM_LEN 1
698
699
700/* object identifier for root dir */
701#define REISERFS_ROOT_OBJECTID 2
702#define REISERFS_ROOT_PARENT_OBJECTID 1
703extern struct reiserfs_key root_key;
704
705
706
707
708/*
709 * Picture represents a leaf of the S+tree
710 * ______________________________________________________
711 * | | Array of | | |
712 * |Block | Object-Item | F r e e | Objects- |
713 * | head | Headers | S p a c e | Items |
714 * |______|_______________|___________________|___________|
715 */
716
717/* Header of a disk block. More precisely, header of a formatted leaf
718 or internal node, and not the header of an unformatted node. */
719struct block_head {
Al Viro3e8962b2005-05-01 08:59:18 -0700720 __le16 blk_level; /* Level of a block in the tree. */
721 __le16 blk_nr_item; /* Number of keys/items in a block. */
722 __le16 blk_free_space; /* Block free space in bytes. */
723 __le16 blk_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 /* dump this in v4/planA */
725 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
726};
727
728#define BLKH_SIZE (sizeof(struct block_head))
729#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
730#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
731#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
732#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
733#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
734#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
735#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
736#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
737#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
738#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
739
740/*
741 * values for blk_level field of the struct block_head
742 */
743
744#define FREE_LEVEL 0 /* when node gets removed from the tree its
745 blk_level is set to FREE_LEVEL. It is then
746 used to see whether the node is still in the
747 tree */
748
749#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level.*/
750
751/* Given the buffer head of a formatted node, resolve to the block head of that node. */
752#define B_BLK_HEAD(p_s_bh) ((struct block_head *)((p_s_bh)->b_data))
753/* Number of items that are in buffer. */
754#define B_NR_ITEMS(p_s_bh) (blkh_nr_item(B_BLK_HEAD(p_s_bh)))
755#define B_LEVEL(p_s_bh) (blkh_level(B_BLK_HEAD(p_s_bh)))
756#define B_FREE_SPACE(p_s_bh) (blkh_free_space(B_BLK_HEAD(p_s_bh)))
757
758#define PUT_B_NR_ITEMS(p_s_bh,val) do { set_blkh_nr_item(B_BLK_HEAD(p_s_bh),val); } while (0)
759#define PUT_B_LEVEL(p_s_bh,val) do { set_blkh_level(B_BLK_HEAD(p_s_bh),val); } while (0)
760#define PUT_B_FREE_SPACE(p_s_bh,val) do { set_blkh_free_space(B_BLK_HEAD(p_s_bh),val); } while (0)
761
762
763/* Get right delimiting key. -- little endian */
764#define B_PRIGHT_DELIM_KEY(p_s_bh) (&(blk_right_delim_key(B_BLK_HEAD(p_s_bh))
765
766/* Does the buffer contain a disk leaf. */
767#define B_IS_ITEMS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) == DISK_LEAF_NODE_LEVEL)
768
769/* Does the buffer contain a disk internal node */
770#define B_IS_KEYS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) > DISK_LEAF_NODE_LEVEL \
771 && B_LEVEL(p_s_bh) <= MAX_HEIGHT)
772
773
774
775
776/***************************************************************************/
777/* STAT DATA */
778/***************************************************************************/
779
780
781//
782// old stat data is 32 bytes long. We are going to distinguish new one by
783// different size
784//
785struct stat_data_v1
786{
Al Viro3e8962b2005-05-01 08:59:18 -0700787 __le16 sd_mode; /* file type, permissions */
788 __le16 sd_nlink; /* number of hard links */
789 __le16 sd_uid; /* owner */
790 __le16 sd_gid; /* group */
791 __le32 sd_size; /* file size */
792 __le32 sd_atime; /* time of last access */
793 __le32 sd_mtime; /* time file was last modified */
794 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 union {
Al Viro3e8962b2005-05-01 08:59:18 -0700796 __le32 sd_rdev;
797 __le32 sd_blocks; /* number of blocks file uses */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 } __attribute__ ((__packed__)) u;
Al Viro3e8962b2005-05-01 08:59:18 -0700799 __le32 sd_first_direct_byte; /* first byte of file which is stored
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 in a direct item: except that if it
801 equals 1 it is a symlink and if it
802 equals ~(__u32)0 there is no
803 direct item. The existence of this
804 field really grates on me. Let's
805 replace it with a macro based on
806 sd_size and our tail suppression
807 policy. Someday. -Hans */
808} __attribute__ ((__packed__));
809
810#define SD_V1_SIZE (sizeof(struct stat_data_v1))
811#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
812#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
813#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
814#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
815#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
816#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
817#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
818#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
819#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
820#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
821#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
822#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
823#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
824#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
825#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
826#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
827#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
828#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
829#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
830#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
831#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
832#define sd_v1_first_direct_byte(sdp) \
833 (le32_to_cpu((sdp)->sd_first_direct_byte))
834#define set_sd_v1_first_direct_byte(sdp,v) \
835 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
836
837#include <linux/ext2_fs.h>
838
839/* inode flags stored in sd_attrs (nee sd_reserved) */
840
841/* we want common flags to have the same values as in ext2,
842 so chattr(1) will work without problems */
843#define REISERFS_IMMUTABLE_FL EXT2_IMMUTABLE_FL
844#define REISERFS_APPEND_FL EXT2_APPEND_FL
845#define REISERFS_SYNC_FL EXT2_SYNC_FL
846#define REISERFS_NOATIME_FL EXT2_NOATIME_FL
847#define REISERFS_NODUMP_FL EXT2_NODUMP_FL
848#define REISERFS_SECRM_FL EXT2_SECRM_FL
849#define REISERFS_UNRM_FL EXT2_UNRM_FL
850#define REISERFS_COMPR_FL EXT2_COMPR_FL
851#define REISERFS_NOTAIL_FL EXT2_NOTAIL_FL
852
853/* persistent flags that file inherits from the parent directory */
854#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
855 REISERFS_SYNC_FL | \
856 REISERFS_NOATIME_FL | \
857 REISERFS_NODUMP_FL | \
858 REISERFS_SECRM_FL | \
859 REISERFS_COMPR_FL | \
860 REISERFS_NOTAIL_FL )
861
862/* Stat Data on disk (reiserfs version of UFS disk inode minus the
863 address blocks) */
864struct stat_data {
Al Viro3e8962b2005-05-01 08:59:18 -0700865 __le16 sd_mode; /* file type, permissions */
866 __le16 sd_attrs; /* persistent inode flags */
867 __le32 sd_nlink; /* number of hard links */
868 __le64 sd_size; /* file size */
869 __le32 sd_uid; /* owner */
870 __le32 sd_gid; /* group */
871 __le32 sd_atime; /* time of last access */
872 __le32 sd_mtime; /* time file was last modified */
873 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
874 __le32 sd_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 union {
Al Viro3e8962b2005-05-01 08:59:18 -0700876 __le32 sd_rdev;
877 __le32 sd_generation;
878 //__le32 sd_first_direct_byte;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 /* first byte of file which is stored in a
880 direct item: except that if it equals 1
881 it is a symlink and if it equals
882 ~(__u32)0 there is no direct item. The
883 existence of this field really grates
884 on me. Let's replace it with a macro
885 based on sd_size and our tail
886 suppression policy? */
887 } __attribute__ ((__packed__)) u;
888} __attribute__ ((__packed__));
889//
890// this is 44 bytes long
891//
892#define SD_SIZE (sizeof(struct stat_data))
893#define SD_V2_SIZE SD_SIZE
894#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
895#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
896#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
897/* sd_reserved */
898/* set_sd_reserved */
899#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
900#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
901#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
902#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
903#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
904#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
905#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
906#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
907#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
908#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
909#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
910#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
911#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
912#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
913#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
914#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
915#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
916#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
917#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
918#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
919#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
920#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
921
922
923/***************************************************************************/
924/* DIRECTORY STRUCTURE */
925/***************************************************************************/
926/*
927 Picture represents the structure of directory items
928 ________________________________________________
929 | Array of | | | | | |
930 | directory |N-1| N-2 | .... | 1st |0th|
931 | entry headers | | | | | |
932 |_______________|___|_____|________|_______|___|
933 <---- directory entries ------>
934
935 First directory item has k_offset component 1. We store "." and ".."
936 in one item, always, we never split "." and ".." into differing
937 items. This makes, among other things, the code for removing
938 directories simpler. */
939#define SD_OFFSET 0
940#define SD_UNIQUENESS 0
941#define DOT_OFFSET 1
942#define DOT_DOT_OFFSET 2
943#define DIRENTRY_UNIQUENESS 500
944
945/* */
946#define FIRST_ITEM_OFFSET 1
947
948/*
949 Q: How to get key of object pointed to by entry from entry?
950
951 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
952 of object, entry points to */
953
954/* NOT IMPLEMENTED:
955 Directory will someday contain stat data of object */
956
957
958
959struct reiserfs_de_head
960{
Al Viro3e8962b2005-05-01 08:59:18 -0700961 __le32 deh_offset; /* third component of the directory entry key */
962 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 by directory entry */
Al Viro3e8962b2005-05-01 08:59:18 -0700964 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
965 __le16 deh_location; /* offset of name in the whole item */
966 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 entry is hidden (unlinked) */
968} __attribute__ ((__packed__));
969#define DEH_SIZE sizeof(struct reiserfs_de_head)
970#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
971#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
972#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
973#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
974#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
975
976#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
977#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
978#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
979#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
980#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
981
982/* empty directory contains two entries "." and ".." and their headers */
983#define EMPTY_DIR_SIZE \
984(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
985
986/* old format directories have this size when empty */
987#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
988
989#define DEH_Statdata 0 /* not used now */
990#define DEH_Visible 2
991
992/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
993#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
994# define ADDR_UNALIGNED_BITS (3)
995#endif
996
997/* These are only used to manipulate deh_state.
998 * Because of this, we'll use the ext2_ bit routines,
999 * since they are little endian */
1000#ifdef ADDR_UNALIGNED_BITS
1001
1002# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1003# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1004
1005# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1006# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1007# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1008
1009#else
1010
1011# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
1012# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
1013# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
1014
1015#endif
1016
1017#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1018#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1019#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1020#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1021
1022#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1023#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1024#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1025
Al Viro3e8962b2005-05-01 08:59:18 -07001026extern void make_empty_dir_item_v1 (char * body, __le32 dirid, __le32 objid,
1027 __le32 par_dirid, __le32 par_objid);
1028extern void make_empty_dir_item (char * body, __le32 dirid, __le32 objid,
1029 __le32 par_dirid, __le32 par_objid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
1031/* array of the entry headers */
1032 /* get item body */
1033#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1034#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1035
1036/* length of the directory entry in directory item. This define
1037 calculates length of i-th directory entry using directory entry
1038 locations from dir entry head. When it calculates length of 0-th
1039 directory entry, it uses length of whole item in place of entry
1040 location of the non-existent following entry in the calculation.
1041 See picture above.*/
1042/*
1043#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1044((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1045*/
1046static inline int entry_length (const struct buffer_head * bh,
1047 const struct item_head * ih, int pos_in_item)
1048{
1049 struct reiserfs_de_head * deh;
1050
1051 deh = B_I_DEH (bh, ih) + pos_in_item;
1052 if (pos_in_item)
1053 return deh_location(deh-1) - deh_location(deh);
1054
1055 return ih_item_len(ih) - deh_location(deh);
1056}
1057
1058
1059
1060/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1061#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1062
1063
1064/* name by bh, ih and entry_num */
1065#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))))
1066
1067// two entries per block (at least)
1068#define REISERFS_MAX_NAME(block_size) 255
1069
1070
1071/* this structure is used for operations on directory entries. It is
1072 not a disk structure. */
1073/* When reiserfs_find_entry or search_by_entry_key find directory
1074 entry, they return filled reiserfs_dir_entry structure */
1075struct reiserfs_dir_entry
1076{
1077 struct buffer_head * de_bh;
1078 int de_item_num;
1079 struct item_head * de_ih;
1080 int de_entry_num;
1081 struct reiserfs_de_head * de_deh;
1082 int de_entrylen;
1083 int de_namelen;
1084 char * de_name;
1085 char * de_gen_number_bit_string;
1086
1087 __u32 de_dir_id;
1088 __u32 de_objectid;
1089
1090 struct cpu_key de_entry_key;
1091};
1092
1093/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1094
1095/* pointer to file name, stored in entry */
1096#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1097
1098/* length of name */
1099#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1100(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1101
1102
1103
1104/* hash value occupies bits from 7 up to 30 */
1105#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1106/* generation number occupies 7 bits starting from 0 up to 6 */
1107#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1108#define MAX_GENERATION_NUMBER 127
1109
1110#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1111
1112
1113/*
1114 * Picture represents an internal node of the reiserfs tree
1115 * ______________________________________________________
1116 * | | Array of | Array of | Free |
1117 * |block | keys | pointers | space |
1118 * | head | N | N+1 | |
1119 * |______|_______________|___________________|___________|
1120 */
1121
1122/***************************************************************************/
1123/* DISK CHILD */
1124/***************************************************************************/
1125/* Disk child pointer: The pointer from an internal node of the tree
1126 to a node that is on disk. */
1127struct disk_child {
Al Viro3e8962b2005-05-01 08:59:18 -07001128 __le32 dc_block_number; /* Disk child's block number. */
1129 __le16 dc_size; /* Disk child's used space. */
1130 __le16 dc_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131};
1132
1133#define DC_SIZE (sizeof(struct disk_child))
1134#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1135#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1136#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1137#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1138
1139/* Get disk child by buffer header and position in the tree node. */
1140#define B_N_CHILD(p_s_bh,n_pos) ((struct disk_child *)\
1141((p_s_bh)->b_data+BLKH_SIZE+B_NR_ITEMS(p_s_bh)*KEY_SIZE+DC_SIZE*(n_pos)))
1142
1143/* Get disk child number by buffer header and position in the tree node. */
1144#define B_N_CHILD_NUM(p_s_bh,n_pos) (dc_block_number(B_N_CHILD(p_s_bh,n_pos)))
1145#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 ))
1146
1147 /* maximal value of field child_size in structure disk_child */
1148 /* child size is the combined size of all items and their headers */
1149#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1150
1151/* amount of used space in buffer (not including block head) */
1152#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1153
1154/* max and min number of keys in internal node */
1155#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1156#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1157
1158/***************************************************************************/
1159/* PATH STRUCTURES AND DEFINES */
1160/***************************************************************************/
1161
1162
1163/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1164 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1165 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1166 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1167 position of the block_number of the next node if it is looking through an internal node. If it
1168 is looking through a leaf node bin_search will find the position of the item which has key either
1169 equal to given key, or which is the maximal key less than the given key. */
1170
1171struct path_element {
1172 struct buffer_head * pe_buffer; /* Pointer to the buffer at the path in the tree. */
1173 int pe_position; /* Position in the tree node which is placed in the */
1174 /* buffer above. */
1175};
1176
1177#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1178#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1179#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1180
1181#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1182#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1183
1184
1185
1186/* We need to keep track of who the ancestors of nodes are. When we
1187 perform a search we record which nodes were visited while
1188 descending the tree looking for the node we searched for. This list
1189 of nodes is called the path. This information is used while
1190 performing balancing. Note that this path information may become
1191 invalid, and this means we must check it when using it to see if it
1192 is still valid. You'll need to read search_by_key and the comments
1193 in it, especially about decrement_counters_in_path(), to understand
1194 this structure.
1195
1196Paths make the code so much harder to work with and debug.... An
1197enormous number of bugs are due to them, and trying to write or modify
1198code that uses them just makes my head hurt. They are based on an
1199excessive effort to avoid disturbing the precious VFS code.:-( The
1200gods only know how we are going to SMP the code that uses them.
1201znodes are the way! */
1202
1203#define PATH_READA 0x1 /* do read ahead */
1204#define PATH_READA_BACK 0x2 /* read backwards */
1205
1206struct path {
1207 int path_length; /* Length of the array above. */
1208 int reada;
1209 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1210 int pos_in_item;
1211};
1212
1213#define pos_in_item(path) ((path)->pos_in_item)
1214
1215#define INITIALIZE_PATH(var) \
1216struct path var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1217
1218/* Get path element by path and path position. */
1219#define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1220
1221/* Get buffer header at the path by path and path position. */
1222#define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1223
1224/* Get position in the element at the path by path and path position. */
1225#define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1226
1227
1228#define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1229 /* you know, to the person who didn't
1230 write this the macro name does not
1231 at first suggest what it does.
1232 Maybe POSITION_FROM_PATH_END? Or
1233 maybe we should just focus on
1234 dumping paths... -Hans */
1235#define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1236
1237
1238#define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1239
1240/* in do_balance leaf has h == 0 in contrast with path structure,
1241 where root has level == 0. That is why we need these defines */
1242#define PATH_H_PBUFFER(p_s_path, h) PATH_OFFSET_PBUFFER (p_s_path, p_s_path->path_length - (h)) /* tb->S[h] */
1243#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1244#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1245#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1246
1247#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1248
1249#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1250#define get_ih(path) PATH_PITEM_HEAD(path)
1251#define get_item_pos(path) PATH_LAST_POSITION(path)
1252#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1253#define item_moved(ih,path) comp_items(ih, path)
1254#define path_changed(ih,path) comp_items (ih, path)
1255
1256
1257/***************************************************************************/
1258/* MISC */
1259/***************************************************************************/
1260
1261/* Size of pointer to the unformatted node. */
1262#define UNFM_P_SIZE (sizeof(unp_t))
1263#define UNFM_P_SHIFT 2
1264
1265// in in-core inode key is stored on le form
1266#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1267
1268#define MAX_UL_INT 0xffffffff
1269#define MAX_INT 0x7ffffff
1270#define MAX_US_INT 0xffff
1271
1272// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1273#define U32_MAX (~(__u32)0)
1274
1275static inline loff_t max_reiserfs_offset (struct inode * inode)
1276{
1277 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1278 return (loff_t)U32_MAX;
1279
1280 return (loff_t)((~(__u64)0) >> 4);
1281}
1282
1283
1284/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1285#define MAX_KEY_OBJECTID MAX_UL_INT
1286
1287
1288#define MAX_B_NUM MAX_UL_INT
1289#define MAX_FC_NUM MAX_US_INT
1290
1291
1292/* the purpose is to detect overflow of an unsigned short */
1293#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1294
1295
1296/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
1297#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1298#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1299
1300#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1301#define get_generation(s) atomic_read (&fs_generation(s))
1302#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1303#define __fs_changed(gen,s) (gen != get_generation (s))
1304#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1305
1306
1307/***************************************************************************/
1308/* FIXATE NODES */
1309/***************************************************************************/
1310
1311#define VI_TYPE_LEFT_MERGEABLE 1
1312#define VI_TYPE_RIGHT_MERGEABLE 2
1313
1314/* To make any changes in the tree we always first find node, that
1315 contains item to be changed/deleted or place to insert a new
1316 item. We call this node S. To do balancing we need to decide what
1317 we will shift to left/right neighbor, or to a new node, where new
1318 item will be etc. To make this analysis simpler we build virtual
1319 node. Virtual node is an array of items, that will replace items of
1320 node S. (For instance if we are going to delete an item, virtual
1321 node does not contain it). Virtual node keeps information about
1322 item sizes and types, mergeability of first and last items, sizes
1323 of all entries in directory item. We use this array of items when
1324 calculating what we can shift to neighbors and how many nodes we
1325 have to have if we do not any shiftings, if we shift to left/right
1326 neighbor or to both. */
1327struct virtual_item
1328{
1329 int vi_index; // index in the array of item operations
1330 unsigned short vi_type; // left/right mergeability
1331 unsigned short vi_item_len; /* length of item that it will have after balancing */
1332 struct item_head * vi_ih;
1333 const char * vi_item; // body of item (old or new)
1334 const void * vi_new_data; // 0 always but paste mode
1335 void * vi_uarea; // item specific area
1336};
1337
1338
1339struct virtual_node
1340{
1341 char * vn_free_ptr; /* this is a pointer to the free space in the buffer */
1342 unsigned short vn_nr_item; /* number of items in virtual node */
1343 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1344 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1345 short vn_affected_item_num;
1346 short vn_pos_in_item;
1347 struct item_head * vn_ins_ih; /* item header of inserted item, 0 for other modes */
1348 const void * vn_data;
1349 struct virtual_item * vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1350};
1351
1352/* used by directory items when creating virtual nodes */
1353struct direntry_uarea {
1354 int flags;
1355 __u16 entry_count;
1356 __u16 entry_sizes[1];
1357} __attribute__ ((__packed__)) ;
1358
1359
1360/***************************************************************************/
1361/* TREE BALANCE */
1362/***************************************************************************/
1363
1364/* This temporary structure is used in tree balance algorithms, and
1365 constructed as we go to the extent that its various parts are
1366 needed. It contains arrays of nodes that can potentially be
1367 involved in the balancing of node S, and parameters that define how
1368 each of the nodes must be balanced. Note that in these algorithms
1369 for balancing the worst case is to need to balance the current node
1370 S and the left and right neighbors and all of their parents plus
1371 create a new node. We implement S1 balancing for the leaf nodes
1372 and S0 balancing for the internal nodes (S1 and S0 are defined in
1373 our papers.)*/
1374
1375#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1376
1377/* maximum number of FEB blocknrs on a single level */
1378#define MAX_AMOUNT_NEEDED 2
1379
1380/* someday somebody will prefix every field in this struct with tb_ */
1381struct tree_balance
1382{
1383 int tb_mode;
1384 int need_balance_dirty;
1385 struct super_block * tb_sb;
1386 struct reiserfs_transaction_handle *transaction_handle ;
1387 struct path * tb_path;
1388 struct buffer_head * L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1389 struct buffer_head * R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path*/
1390 struct buffer_head * FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1391 struct buffer_head * FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1392 struct buffer_head * CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1393 struct buffer_head * CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1394
1395 struct buffer_head * FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1396 cur_blknum. */
1397 struct buffer_head * used[MAX_FEB_SIZE];
1398 struct buffer_head * thrown[MAX_FEB_SIZE];
1399 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1400 shifted to the left in order to balance the
1401 current node; for leaves includes item that
1402 will be partially shifted; for internal
1403 nodes, it is the number of child pointers
1404 rather than items. It includes the new item
1405 being created. The code sometimes subtracts
1406 one to get the number of wholly shifted
1407 items for other purposes. */
1408 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1409 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1410 S[h] to its item number within the node CFL[h] */
1411 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1412 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1413 S[h]. A negative value means removing. */
1414 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1415 balancing on the level h of the tree. If 0 then S is
1416 being deleted, if 1 then S is remaining and no new nodes
1417 are being created, if 2 or 3 then 1 or 2 new nodes is
1418 being created */
1419
1420 /* fields that are used only for balancing leaves of the tree */
1421 int cur_blknum; /* number of empty blocks having been already allocated */
1422 int s0num; /* number of items that fall into left most node when S[0] splits */
1423 int s1num; /* number of items that fall into first new node when S[0] splits */
1424 int s2num; /* number of items that fall into second new node when S[0] splits */
1425 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1426 /* most liquid item that cannot be shifted from S[0] entirely */
1427 /* if -1 then nothing will be partially shifted */
1428 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1429 /* most liquid item that cannot be shifted from S[0] entirely */
1430 /* if -1 then nothing will be partially shifted */
1431 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1432 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1433 int s2bytes;
1434 struct buffer_head * buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1435 char * vn_buf; /* kmalloced memory. Used to create
1436 virtual node and keep map of
1437 dirtied bitmap blocks */
1438 int vn_buf_size; /* size of the vn_buf */
1439 struct virtual_node * tb_vn; /* VN starts after bitmap of bitmap blocks */
1440
1441 int fs_gen; /* saved value of `reiserfs_generation' counter
1442 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1443#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Al Viro6a3a16f2005-05-01 08:59:17 -07001444 struct in_core_key key; /* key pointer, to pass to block allocator or
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 another low-level subsystem */
1446#endif
1447} ;
1448
1449/* These are modes of balancing */
1450
1451/* When inserting an item. */
1452#define M_INSERT 'i'
1453/* When inserting into (directories only) or appending onto an already
1454 existant item. */
1455#define M_PASTE 'p'
1456/* When deleting an item. */
1457#define M_DELETE 'd'
1458/* When truncating an item or removing an entry from a (directory) item. */
1459#define M_CUT 'c'
1460
1461/* used when balancing on leaf level skipped (in reiserfsck) */
1462#define M_INTERNAL 'n'
1463
1464/* When further balancing is not needed, then do_balance does not need
1465 to be called. */
1466#define M_SKIP_BALANCING 's'
1467#define M_CONVERT 'v'
1468
1469/* modes of leaf_move_items */
1470#define LEAF_FROM_S_TO_L 0
1471#define LEAF_FROM_S_TO_R 1
1472#define LEAF_FROM_R_TO_L 2
1473#define LEAF_FROM_L_TO_R 3
1474#define LEAF_FROM_S_TO_SNEW 4
1475
1476#define FIRST_TO_LAST 0
1477#define LAST_TO_FIRST 1
1478
1479/* used in do_balance for passing parent of node information that has
1480 been gotten from tb struct */
1481struct buffer_info {
1482 struct tree_balance * tb;
1483 struct buffer_head * bi_bh;
1484 struct buffer_head * bi_parent;
1485 int bi_position;
1486};
1487
1488
1489/* there are 4 types of items: stat data, directory item, indirect, direct.
1490+-------------------+------------+--------------+------------+
1491| | k_offset | k_uniqueness | mergeable? |
1492+-------------------+------------+--------------+------------+
1493| stat data | 0 | 0 | no |
1494+-------------------+------------+--------------+------------+
1495| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1496| non 1st directory | hash value | | yes |
1497| item | | | |
1498+-------------------+------------+--------------+------------+
1499| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1500+-------------------+------------+--------------+------------+
1501| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1502+-------------------+------------+--------------+------------+
1503*/
1504
1505struct item_operations {
1506 int (*bytes_number) (struct item_head * ih, int block_size);
1507 void (*decrement_key) (struct cpu_key *);
1508 int (*is_left_mergeable) (struct reiserfs_key * ih, unsigned long bsize);
1509 void (*print_item) (struct item_head *, char * item);
1510 void (*check_item) (struct item_head *, char * item);
1511
1512 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1513 int is_affected, int insert_size);
1514 int (*check_left) (struct virtual_item * vi, int free,
1515 int start_skip, int end_skip);
1516 int (*check_right) (struct virtual_item * vi, int free);
1517 int (*part_size) (struct virtual_item * vi, int from, int to);
1518 int (*unit_num) (struct virtual_item * vi);
1519 void (*print_vi) (struct virtual_item * vi);
1520};
1521
1522
1523extern struct item_operations * item_ops [TYPE_ANY + 1];
1524
1525#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1526#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1527#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1528#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1529#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)
1530#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1531#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1532#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1533#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1534#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1535
1536
1537
1538#define COMP_SHORT_KEYS comp_short_keys
1539
1540/* number of blocks pointed to by the indirect item */
1541#define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1542
1543/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1544#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1545
1546/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1547
1548
1549/* get the item header */
1550#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1551
1552/* get key */
1553#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1554
1555/* get the key */
1556#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1557
1558/* get item body */
1559#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1560
1561/* get the stat data by the buffer header and the item order */
1562#define B_N_STAT_DATA(bh,nr) \
1563( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1564
1565 /* following defines use reiserfs buffer header and item header */
1566
1567/* get stat-data */
1568#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1569
1570// this is 3976 for size==4096
1571#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1572
1573/* indirect items consist of entries which contain blocknrs, pos
1574 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1575 blocknr contained by the entry pos points to */
1576#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1577#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)
1578
1579struct reiserfs_iget_args {
1580 __u32 objectid ;
1581 __u32 dirid ;
1582} ;
1583
1584/***************************************************************************/
1585/* FUNCTION DECLARATIONS */
1586/***************************************************************************/
1587
1588/*#ifdef __KERNEL__*/
1589#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1590
1591#define journal_trans_half(blocksize) \
1592 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1593
1594/* journal.c see journal.c for all the comments here */
1595
1596/* first block written in a commit. */
1597struct reiserfs_journal_desc {
Al Viro3e8962b2005-05-01 08:59:18 -07001598 __le32 j_trans_id ; /* id of commit */
1599 __le32 j_len ; /* length of commit. len +1 is the commit block */
1600 __le32 j_mount_id ; /* mount id of this trans*/
1601 __le32 j_realblock[1] ; /* real locations for each block */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602} ;
1603
1604#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1605#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1606#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1607
1608#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1609#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1610#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1611
1612/* last block written in a commit */
1613struct reiserfs_journal_commit {
Al Viro3e8962b2005-05-01 08:59:18 -07001614 __le32 j_trans_id ; /* must match j_trans_id from the desc block */
1615 __le32 j_len ; /* ditto */
1616 __le32 j_realblock[1] ; /* real locations for each block */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617} ;
1618
1619#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1620#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1621#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1622
1623#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1624#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1625
1626/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1627** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1628** and this transaction does not need to be replayed.
1629*/
1630struct reiserfs_journal_header {
Al Viro3e8962b2005-05-01 08:59:18 -07001631 __le32 j_last_flush_trans_id ; /* id of last fully flushed transaction */
1632 __le32 j_first_unflushed_offset ; /* offset in the log of where to start replay after a crash */
1633 __le32 j_mount_id ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634 /* 12 */ struct journal_params jh_journal;
1635} ;
1636
1637/* biggest tunable defines are right here */
1638#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1639#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1640#define JOURNAL_TRANS_MIN_DEFAULT 256
1641#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1642#define JOURNAL_MIN_RATIO 2
1643#define JOURNAL_MAX_COMMIT_AGE 30
1644#define JOURNAL_MAX_TRANS_AGE 30
1645#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1646#ifdef CONFIG_QUOTA
1647#define REISERFS_QUOTA_TRANS_BLOCKS 2 /* We need to update data and inode (atime) */
1648#define REISERFS_QUOTA_INIT_BLOCKS (DQUOT_MAX_WRITES*(JOURNAL_PER_BALANCE_CNT+2)+1) /* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1649#else
1650#define REISERFS_QUOTA_TRANS_BLOCKS 0
1651#define REISERFS_QUOTA_INIT_BLOCKS 0
1652#endif
1653
1654/* both of these can be as low as 1, or as high as you want. The min is the
1655** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1656** as needed, and released when transactions are committed. On release, if
1657** the current number of nodes is > max, the node is freed, otherwise,
1658** it is put on a free list for faster use later.
1659*/
1660#define REISERFS_MIN_BITMAP_NODES 10
1661#define REISERFS_MAX_BITMAP_NODES 100
1662
1663#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1664#define JBH_HASH_MASK 8191
1665
1666#define _jhashfn(sb,block) \
1667 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1668 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1669#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1670
1671// We need these to make journal.c code more readable
1672#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1673#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1674#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1675
1676enum reiserfs_bh_state_bits {
1677 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1678 BH_JDirty_wait,
1679 BH_JNew, /* disk block was taken off free list before
1680 * being in a finished transaction, or
1681 * written to disk. Can be reused immed. */
1682 BH_JPrepared,
1683 BH_JRestore_dirty,
1684 BH_JTest, // debugging only will go away
1685};
1686
1687BUFFER_FNS(JDirty, journaled);
1688TAS_BUFFER_FNS(JDirty, journaled);
1689BUFFER_FNS(JDirty_wait, journal_dirty);
1690TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1691BUFFER_FNS(JNew, journal_new);
1692TAS_BUFFER_FNS(JNew, journal_new);
1693BUFFER_FNS(JPrepared, journal_prepared);
1694TAS_BUFFER_FNS(JPrepared, journal_prepared);
1695BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1696TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1697BUFFER_FNS(JTest, journal_test);
1698TAS_BUFFER_FNS(JTest, journal_test);
1699
1700/*
1701** transaction handle which is passed around for all journal calls
1702*/
1703struct reiserfs_transaction_handle {
1704 struct super_block *t_super ; /* super for this FS when journal_begin was
1705 called. saves calls to reiserfs_get_super
1706 also used by nested transactions to make
1707 sure they are nesting on the right FS
1708 _must_ be first in the handle
1709 */
1710 int t_refcount;
1711 int t_blocks_logged ; /* number of blocks this writer has logged */
1712 int t_blocks_allocated ; /* number of blocks this writer allocated */
1713 unsigned long t_trans_id ; /* sanity check, equals the current trans id */
1714 void *t_handle_save ; /* save existing current->journal_info */
1715 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1716 should be displaced from others */
1717 struct list_head t_list;
1718} ;
1719
1720/* used to keep track of ordered and tail writes, attached to the buffer
1721 * head through b_journal_head.
1722 */
1723struct reiserfs_jh {
1724 struct reiserfs_journal_list *jl;
1725 struct buffer_head *bh;
1726 struct list_head list;
1727};
1728
1729void reiserfs_free_jh(struct buffer_head *bh);
1730int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1731int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
1732int journal_mark_dirty(struct reiserfs_transaction_handle *, struct super_block *, struct buffer_head *bh) ;
1733
1734static inline int
1735reiserfs_file_data_log(struct inode *inode) {
1736 if (reiserfs_data_log(inode->i_sb) ||
1737 (REISERFS_I(inode)->i_flags & i_data_log))
1738 return 1 ;
1739 return 0 ;
1740}
1741
1742static inline int reiserfs_transaction_running(struct super_block *s) {
1743 struct reiserfs_transaction_handle *th = current->journal_info ;
1744 if (th && th->t_super == s)
1745 return 1 ;
1746 if (th && th->t_super == NULL)
1747 BUG();
1748 return 0 ;
1749}
1750
1751int reiserfs_async_progress_wait(struct super_block *s);
1752
1753struct reiserfs_transaction_handle *
1754reiserfs_persistent_transaction(struct super_block *, int count);
1755int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1756int reiserfs_commit_page(struct inode *inode, struct page *page,
1757 unsigned from, unsigned to);
1758int reiserfs_flush_old_commits(struct super_block *);
1759int reiserfs_commit_for_inode(struct inode *) ;
1760int reiserfs_inode_needs_commit(struct inode *) ;
1761void reiserfs_update_inode_transaction(struct inode *) ;
1762void reiserfs_wait_on_write_block(struct super_block *s) ;
1763void reiserfs_block_writes(struct reiserfs_transaction_handle *th) ;
1764void reiserfs_allow_writes(struct super_block *s) ;
1765void reiserfs_check_lock_depth(struct super_block *s, char *caller) ;
1766int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh, int wait) ;
1767void reiserfs_restore_prepared_buffer(struct super_block *, struct buffer_head *bh) ;
1768int journal_init(struct super_block *, const char * j_dev_name, int old_format, unsigned int) ;
1769int journal_release(struct reiserfs_transaction_handle*, struct super_block *) ;
1770int journal_release_error(struct reiserfs_transaction_handle*, struct super_block *) ;
1771int journal_end(struct reiserfs_transaction_handle *, struct super_block *, unsigned long) ;
1772int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *, unsigned long) ;
1773int journal_mark_freed(struct reiserfs_transaction_handle *, struct super_block *, b_blocknr_t blocknr) ;
1774int journal_transaction_should_end(struct reiserfs_transaction_handle *, int) ;
1775int reiserfs_in_journal(struct super_block *p_s_sb, int bmap_nr, int bit_nr, int searchall, b_blocknr_t *next) ;
1776int journal_begin(struct reiserfs_transaction_handle *, struct super_block *p_s_sb, unsigned long) ;
1777int journal_join_abort(struct reiserfs_transaction_handle *, struct super_block *p_s_sb, unsigned long) ;
1778void reiserfs_journal_abort (struct super_block *sb, int errno);
1779void reiserfs_abort (struct super_block *sb, int errno, const char *fmt, ...);
1780int reiserfs_allocate_list_bitmaps(struct super_block *s, struct reiserfs_list_bitmap *, int) ;
1781
1782void add_save_link (struct reiserfs_transaction_handle * th,
1783 struct inode * inode, int truncate);
1784int remove_save_link (struct inode * inode, int truncate);
1785
1786/* objectid.c */
1787__u32 reiserfs_get_unused_objectid (struct reiserfs_transaction_handle *th);
1788void reiserfs_release_objectid (struct reiserfs_transaction_handle *th, __u32 objectid_to_release);
1789int reiserfs_convert_objectid_map_v1(struct super_block *) ;
1790
1791/* stree.c */
1792int B_IS_IN_TREE(const struct buffer_head *);
1793extern void copy_item_head(struct item_head * p_v_to,
1794 const struct item_head * p_v_from);
1795
1796// first key is in cpu form, second - le
1797extern int comp_short_keys (const struct reiserfs_key * le_key,
1798 const struct cpu_key * cpu_key);
1799extern void le_key2cpu_key (struct cpu_key * to, const struct reiserfs_key * from);
1800
1801// both are in le form
1802extern int comp_le_keys (const struct reiserfs_key *, const struct reiserfs_key *);
1803extern int comp_short_le_keys (const struct reiserfs_key *, const struct reiserfs_key *);
1804
1805//
1806// get key version from on disk key - kludge
1807//
1808static inline int le_key_version (const struct reiserfs_key * key)
1809{
1810 int type;
1811
1812 type = offset_v2_k_type( &(key->u.k_offset_v2));
1813 if (type != TYPE_DIRECT && type != TYPE_INDIRECT && type != TYPE_DIRENTRY)
1814 return KEY_FORMAT_3_5;
1815
1816 return KEY_FORMAT_3_6;
1817
1818}
1819
1820
1821static inline void copy_key (struct reiserfs_key *to, const struct reiserfs_key *from)
1822{
1823 memcpy (to, from, KEY_SIZE);
1824}
1825
1826
1827int comp_items (const struct item_head * stored_ih, const struct path * p_s_path);
1828const struct reiserfs_key * get_rkey (const struct path * p_s_chk_path,
1829 const struct super_block * p_s_sb);
1830int search_by_key (struct super_block *, const struct cpu_key *,
1831 struct path *, int);
1832#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
1833int search_for_position_by_key (struct super_block * p_s_sb,
1834 const struct cpu_key * p_s_cpu_key,
1835 struct path * p_s_search_path);
1836extern void decrement_bcount (struct buffer_head * p_s_bh);
1837void decrement_counters_in_path (struct path * p_s_search_path);
1838void pathrelse (struct path * p_s_search_path);
1839int reiserfs_check_path(struct path *p) ;
1840void pathrelse_and_restore (struct super_block *s, struct path * p_s_search_path);
1841
1842int reiserfs_insert_item (struct reiserfs_transaction_handle *th,
1843 struct path * path,
1844 const struct cpu_key * key,
1845 struct item_head * ih,
1846 struct inode *inode, const char * body);
1847
1848int reiserfs_paste_into_item (struct reiserfs_transaction_handle *th,
1849 struct path * path,
1850 const struct cpu_key * key,
1851 struct inode *inode,
1852 const char * body, int paste_size);
1853
1854int reiserfs_cut_from_item (struct reiserfs_transaction_handle *th,
1855 struct path * path,
1856 struct cpu_key * key,
1857 struct inode * inode,
1858 struct page *page,
1859 loff_t new_file_size);
1860
1861int reiserfs_delete_item (struct reiserfs_transaction_handle *th,
1862 struct path * path,
1863 const struct cpu_key * key,
1864 struct inode * inode,
1865 struct buffer_head * p_s_un_bh);
1866
1867void reiserfs_delete_solid_item (struct reiserfs_transaction_handle *th,
1868 struct inode *inode, struct reiserfs_key * key);
1869int reiserfs_delete_object (struct reiserfs_transaction_handle *th, struct inode * p_s_inode);
1870int reiserfs_do_truncate (struct reiserfs_transaction_handle *th,
1871 struct inode * p_s_inode, struct page *,
1872 int update_timestamps);
1873
1874#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1875#define file_size(inode) ((inode)->i_size)
1876#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1877
1878#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1879!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 )
1880
1881void padd_item (char * item, int total_length, int length);
1882
1883/* inode.c */
1884/* args for the create parameter of reiserfs_get_block */
1885#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1886#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1887#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1888#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
1889#define GET_BLOCK_NO_ISEM 8 /* i_sem is not held, don't preallocate */
1890#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
1891
1892int restart_transaction(struct reiserfs_transaction_handle *th, struct inode *inode, struct path *path);
1893void reiserfs_read_locked_inode(struct inode * inode, struct reiserfs_iget_args *args) ;
1894int reiserfs_find_actor(struct inode * inode, void *p) ;
1895int reiserfs_init_locked_inode(struct inode * inode, void *p) ;
1896void reiserfs_delete_inode (struct inode * inode);
1897int reiserfs_write_inode (struct inode * inode, int) ;
1898int reiserfs_get_block (struct inode * inode, sector_t block, struct buffer_head * bh_result, int create);
1899struct dentry *reiserfs_get_dentry(struct super_block *, void *) ;
1900struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 *data,
1901 int len, int fhtype,
1902 int (*acceptable)(void *contect, struct dentry *de),
1903 void *context) ;
1904int reiserfs_encode_fh( struct dentry *dentry, __u32 *data, int *lenp,
1905 int connectable );
1906
1907int reiserfs_truncate_file(struct inode *, int update_timestamps) ;
1908void make_cpu_key (struct cpu_key * cpu_key, struct inode * inode, loff_t offset,
1909 int type, int key_length);
1910void make_le_item_head (struct item_head * ih, const struct cpu_key * key,
1911 int version,
1912 loff_t offset, int type, int length, int entry_count);
1913struct inode * reiserfs_iget (struct super_block * s,
1914 const struct cpu_key * key);
1915
1916
1917int reiserfs_new_inode (struct reiserfs_transaction_handle *th,
1918 struct inode * dir, int mode,
1919 const char * symname, loff_t i_size,
1920 struct dentry *dentry, struct inode *inode);
1921
1922void reiserfs_update_sd_size (struct reiserfs_transaction_handle *th,
1923 struct inode * inode, loff_t size);
1924
1925static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
1926 struct inode *inode)
1927{
1928 reiserfs_update_sd_size(th, inode, inode->i_size) ;
1929}
1930
1931void sd_attrs_to_i_attrs( __u16 sd_attrs, struct inode *inode );
1932void i_attrs_to_sd_attrs( struct inode *inode, __u16 *sd_attrs );
1933int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1934
1935/* namei.c */
1936void set_de_name_and_namelen (struct reiserfs_dir_entry * de);
1937int search_by_entry_key (struct super_block * sb, const struct cpu_key * key,
1938 struct path * path,
1939 struct reiserfs_dir_entry * de);
1940struct dentry *reiserfs_get_parent(struct dentry *) ;
1941/* procfs.c */
1942
1943#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1944#define REISERFS_PROC_INFO
1945#else
1946#undef REISERFS_PROC_INFO
1947#endif
1948
1949int reiserfs_proc_info_init( struct super_block *sb );
1950int reiserfs_proc_info_done( struct super_block *sb );
1951struct proc_dir_entry *reiserfs_proc_register_global( char *name,
1952 read_proc_t *func );
1953void reiserfs_proc_unregister_global( const char *name );
1954int reiserfs_proc_info_global_init( void );
1955int reiserfs_proc_info_global_done( void );
1956int reiserfs_global_version_in_proc( char *buffer, char **start, off_t offset,
1957 int count, int *eof, void *data );
1958
1959#if defined( REISERFS_PROC_INFO )
1960
1961#define PROC_EXP( e ) e
1962
1963#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1964#define PROC_INFO_MAX( sb, field, value ) \
1965 __PINFO( sb ).field = \
1966 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1967#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1968#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1969#define PROC_INFO_BH_STAT( sb, bh, level ) \
1970 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1971 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1972 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1973#else
1974#define PROC_EXP( e )
1975#define VOID_V ( ( void ) 0 )
1976#define PROC_INFO_MAX( sb, field, value ) VOID_V
1977#define PROC_INFO_INC( sb, field ) VOID_V
1978#define PROC_INFO_ADD( sb, field, val ) VOID_V
1979#define PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level ) VOID_V
1980#endif
1981
1982/* dir.c */
1983extern struct inode_operations reiserfs_dir_inode_operations;
1984extern struct inode_operations reiserfs_symlink_inode_operations;
1985extern struct inode_operations reiserfs_special_inode_operations;
1986extern struct file_operations reiserfs_dir_operations;
1987
1988/* tail_conversion.c */
1989int direct2indirect (struct reiserfs_transaction_handle *, struct inode *, struct path *, struct buffer_head *, loff_t);
1990int indirect2direct (struct reiserfs_transaction_handle *, struct inode *, struct page *, struct path *, const struct cpu_key *, loff_t, char *);
1991void reiserfs_unmap_buffer(struct buffer_head *) ;
1992
1993
1994/* file.c */
1995extern struct inode_operations reiserfs_file_inode_operations;
1996extern struct file_operations reiserfs_file_operations;
1997extern struct address_space_operations reiserfs_address_space_operations ;
1998
1999/* fix_nodes.c */
2000#ifdef CONFIG_REISERFS_CHECK
2001void * reiserfs_kmalloc (size_t size, int flags, struct super_block * s);
2002void reiserfs_kfree (const void * vp, size_t size, struct super_block * s);
2003#else
2004static inline void *reiserfs_kmalloc(size_t size, int flags,
2005 struct super_block *s)
2006{
2007 return kmalloc(size, flags);
2008}
2009
2010static inline void reiserfs_kfree(const void *vp, size_t size,
2011 struct super_block *s)
2012{
2013 kfree(vp);
2014}
2015#endif
2016
2017int fix_nodes (int n_op_mode, struct tree_balance * p_s_tb,
2018 struct item_head * p_s_ins_ih, const void *);
2019void unfix_nodes (struct tree_balance *);
2020
2021
2022/* prints.c */
2023void reiserfs_panic (struct super_block * s, const char * fmt, ...) __attribute__ ( ( noreturn ) );
2024void reiserfs_info (struct super_block *s, const char * fmt, ...);
2025void reiserfs_debug (struct super_block *s, int level, const char * fmt, ...);
2026void print_indirect_item (struct buffer_head * bh, int item_num);
2027void store_print_tb (struct tree_balance * tb);
2028void print_cur_tb (char * mes);
2029void print_de (struct reiserfs_dir_entry * de);
2030void print_bi (struct buffer_info * bi, char * mes);
2031#define PRINT_LEAF_ITEMS 1 /* print all items */
2032#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2033#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2034void print_block (struct buffer_head * bh, ...);
2035void print_bmap (struct super_block * s, int silent);
2036void print_bmap_block (int i, char * data, int size, int silent);
2037/*void print_super_block (struct super_block * s, char * mes);*/
2038void print_objectid_map (struct super_block * s);
2039void print_block_head (struct buffer_head * bh, char * mes);
2040void check_leaf (struct buffer_head * bh);
2041void check_internal (struct buffer_head * bh);
2042void print_statistics (struct super_block * s);
2043char * reiserfs_hashname(int code);
2044
2045/* lbalance.c */
2046int leaf_move_items (int shift_mode, struct tree_balance * tb, int mov_num, int mov_bytes, struct buffer_head * Snew);
2047int leaf_shift_left (struct tree_balance * tb, int shift_num, int shift_bytes);
2048int leaf_shift_right (struct tree_balance * tb, int shift_num, int shift_bytes);
2049void leaf_delete_items (struct buffer_info * cur_bi, int last_first, int first, int del_num, int del_bytes);
2050void leaf_insert_into_buf (struct buffer_info * bi, int before,
2051 struct item_head * inserted_item_ih, const char * inserted_item_body, int zeros_number);
2052void leaf_paste_in_buffer (struct buffer_info * bi, int pasted_item_num,
2053 int pos_in_item, int paste_size, const char * body, int zeros_number);
2054void leaf_cut_from_buffer (struct buffer_info * bi, int cut_item_num, int pos_in_item,
2055 int cut_size);
2056void leaf_paste_entries (struct buffer_head * bh, int item_num, int before,
2057 int new_entry_count, struct reiserfs_de_head * new_dehs, const char * records, int paste_size);
2058/* ibalance.c */
2059int balance_internal (struct tree_balance * , int, int, struct item_head * ,
2060 struct buffer_head **);
2061
2062/* do_balance.c */
2063void do_balance_mark_leaf_dirty (struct tree_balance * tb,
2064 struct buffer_head * bh, int flag);
2065#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2066#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2067
2068void do_balance (struct tree_balance * tb, struct item_head * ih,
2069 const char * body, int flag);
2070void reiserfs_invalidate_buffer (struct tree_balance * tb, struct buffer_head * bh);
2071
2072int get_left_neighbor_position (struct tree_balance * tb, int h);
2073int get_right_neighbor_position (struct tree_balance * tb, int h);
2074void replace_key (struct tree_balance * tb, struct buffer_head *, int, struct buffer_head *, int);
2075void make_empty_node (struct buffer_info *);
2076struct buffer_head * get_FEB (struct tree_balance *);
2077
2078/* bitmap.c */
2079
2080/* structure contains hints for block allocator, and it is a container for
2081 * arguments, such as node, search path, transaction_handle, etc. */
2082 struct __reiserfs_blocknr_hint {
2083 struct inode * inode; /* inode passed to allocator, if we allocate unf. nodes */
2084 long block; /* file offset, in blocks */
Al Viro6a3a16f2005-05-01 08:59:17 -07002085 struct in_core_key key;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 struct path * path; /* search path, used by allocator to deternine search_start by
2087 * various ways */
2088 struct reiserfs_transaction_handle * th; /* transaction handle is needed to log super blocks and
2089 * bitmap blocks changes */
2090 b_blocknr_t beg, end;
2091 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
2092 * between different block allocator procedures
2093 * (determine_search_start() and others) */
2094 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2095 * function that do actual allocation */
2096
2097 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
2098 * formatted/unformatted blocks with/without preallocation */
2099 unsigned preallocate:1;
2100};
2101
2102typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2103
2104int reiserfs_parse_alloc_options (struct super_block *, char *);
2105void reiserfs_init_alloc_options (struct super_block *s);
2106
2107/*
2108 * given a directory, this will tell you what packing locality
2109 * to use for a new object underneat it. The locality is returned
2110 * in disk byte order (le).
2111 */
Al Viro3e8962b2005-05-01 08:59:18 -07002112__le32 reiserfs_choose_packing(struct inode *dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113
2114int is_reusable (struct super_block * s, b_blocknr_t block, int bit_value);
2115void reiserfs_free_block (struct reiserfs_transaction_handle *th, struct inode *, b_blocknr_t, int for_unformatted);
2116int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t * , int, int);
2117extern inline int reiserfs_new_form_blocknrs (struct tree_balance * tb,
2118 b_blocknr_t *new_blocknrs, int amount_needed)
2119{
2120 reiserfs_blocknr_hint_t hint = {
2121 .th = tb->transaction_handle,
2122 .path = tb->tb_path,
2123 .inode = NULL,
2124 .key = tb->key,
2125 .block = 0,
2126 .formatted_node = 1
2127 };
2128 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed, 0);
2129}
2130
2131extern inline int reiserfs_new_unf_blocknrs (struct reiserfs_transaction_handle *th,
2132 struct inode *inode,
2133 b_blocknr_t *new_blocknrs,
2134 struct path * path, long block)
2135{
2136 reiserfs_blocknr_hint_t hint = {
2137 .th = th,
2138 .path = path,
2139 .inode = inode,
2140 .block = block,
2141 .formatted_node = 0,
2142 .preallocate = 0
2143 };
2144 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2145}
2146
2147#ifdef REISERFS_PREALLOCATE
2148extern inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle *th,
2149 struct inode * inode,
2150 b_blocknr_t *new_blocknrs,
2151 struct path * path, long block)
2152{
2153 reiserfs_blocknr_hint_t hint = {
2154 .th = th,
2155 .path = path,
2156 .inode = inode,
2157 .block = block,
2158 .formatted_node = 0,
2159 .preallocate = 1
2160 };
2161 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2162}
2163
2164void reiserfs_discard_prealloc (struct reiserfs_transaction_handle *th,
2165 struct inode * inode);
2166void reiserfs_discard_all_prealloc (struct reiserfs_transaction_handle *th);
2167#endif
2168void reiserfs_claim_blocks_to_be_allocated( struct super_block *sb, int blocks);
2169void reiserfs_release_claimed_blocks( struct super_block *sb, int blocks);
2170int reiserfs_can_fit_pages(struct super_block *sb);
2171
2172/* hashes.c */
2173__u32 keyed_hash (const signed char *msg, int len);
2174__u32 yura_hash (const signed char *msg, int len);
2175__u32 r5_hash (const signed char *msg, int len);
2176
2177/* the ext2 bit routines adjust for big or little endian as
2178** appropriate for the arch, so in our laziness we use them rather
2179** than using the bit routines they call more directly. These
2180** routines must be used when changing on disk bitmaps. */
2181#define reiserfs_test_and_set_le_bit ext2_set_bit
2182#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2183#define reiserfs_test_le_bit ext2_test_bit
2184#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2185
2186/* sometimes reiserfs_truncate may require to allocate few new blocks
2187 to perform indirect2direct conversion. People probably used to
2188 think, that truncate should work without problems on a filesystem
2189 without free disk space. They may complain that they can not
2190 truncate due to lack of free disk space. This spare space allows us
2191 to not worry about it. 500 is probably too much, but it should be
2192 absolutely safe */
2193#define SPARE_SPACE 500
2194
2195
2196/* prototypes from ioctl.c */
2197int reiserfs_ioctl (struct inode * inode, struct file * filp,
2198 unsigned int cmd, unsigned long arg);
2199
2200/* ioctl's command */
2201#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
2202/* define following flags to be the same as in ext2, so that chattr(1),
2203 lsattr(1) will work with us. */
2204#define REISERFS_IOC_GETFLAGS EXT2_IOC_GETFLAGS
2205#define REISERFS_IOC_SETFLAGS EXT2_IOC_SETFLAGS
2206#define REISERFS_IOC_GETVERSION EXT2_IOC_GETVERSION
2207#define REISERFS_IOC_SETVERSION EXT2_IOC_SETVERSION
2208
2209/* Locking primitives */
2210/* Right now we are still falling back to (un)lock_kernel, but eventually that
2211 would evolve into real per-fs locks */
2212#define reiserfs_write_lock( sb ) lock_kernel()
2213#define reiserfs_write_unlock( sb ) unlock_kernel()
2214
2215/* xattr stuff */
2216#define REISERFS_XATTR_DIR_SEM(s) (REISERFS_SB(s)->xattr_dir_sem)
2217
2218#endif /* _LINUX_REISER_FS_H */
2219
2220