blob: d445b682ce00a92eb89f6c47e29547504dff3ea5 [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 {
384#ifdef __LITTLE_ENDIAN
385 /* little endian version */
386 __u64 k_offset:60;
387 __u64 k_type: 4;
388#else
389 /* big endian version */
390 __u64 k_type: 4;
391 __u64 k_offset:60;
392#endif
393} __attribute__ ((__packed__));
394
395#ifndef __LITTLE_ENDIAN
396typedef union {
397 struct offset_v2 offset_v2;
398 __u64 linear;
399} __attribute__ ((__packed__)) offset_v2_esafe_overlay;
400
401static inline __u16 offset_v2_k_type( const struct offset_v2 *v2 )
402{
403 offset_v2_esafe_overlay tmp = *(const offset_v2_esafe_overlay *)v2;
404 tmp.linear = le64_to_cpu( tmp.linear );
405 return (tmp.offset_v2.k_type <= TYPE_MAXTYPE)?tmp.offset_v2.k_type:TYPE_ANY;
406}
407
408static inline void set_offset_v2_k_type( struct offset_v2 *v2, int type )
409{
410 offset_v2_esafe_overlay *tmp = (offset_v2_esafe_overlay *)v2;
411 tmp->linear = le64_to_cpu(tmp->linear);
412 tmp->offset_v2.k_type = type;
413 tmp->linear = cpu_to_le64(tmp->linear);
414}
415
416static inline loff_t offset_v2_k_offset( const struct offset_v2 *v2 )
417{
418 offset_v2_esafe_overlay tmp = *(const offset_v2_esafe_overlay *)v2;
419 tmp.linear = le64_to_cpu( tmp.linear );
420 return tmp.offset_v2.k_offset;
421}
422
423static inline void set_offset_v2_k_offset( struct offset_v2 *v2, loff_t offset ){
424 offset_v2_esafe_overlay *tmp = (offset_v2_esafe_overlay *)v2;
425 tmp->linear = le64_to_cpu(tmp->linear);
426 tmp->offset_v2.k_offset = offset;
427 tmp->linear = cpu_to_le64(tmp->linear);
428}
429#else
430# define offset_v2_k_type(v2) ((v2)->k_type)
431# define set_offset_v2_k_type(v2,val) (offset_v2_k_type(v2) = (val))
432# define offset_v2_k_offset(v2) ((v2)->k_offset)
433# define set_offset_v2_k_offset(v2,val) (offset_v2_k_offset(v2) = (val))
434#endif
435
436/* Key of an item determines its location in the S+tree, and
437 is composed of 4 components */
438struct reiserfs_key {
Al Viro3e8962b2005-05-01 08:59:18 -0700439 __le32 k_dir_id; /* packing locality: by default parent
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 directory object id */
Al Viro3e8962b2005-05-01 08:59:18 -0700441 __le32 k_objectid; /* object identifier */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 union {
443 struct offset_v1 k_offset_v1;
444 struct offset_v2 k_offset_v2;
445 } __attribute__ ((__packed__)) u;
446} __attribute__ ((__packed__));
447
Al Viro6a3a16f2005-05-01 08:59:17 -0700448struct in_core_key {
449 __u32 k_dir_id; /* packing locality: by default parent
450 directory object id */
451 __u32 k_objectid; /* object identifier */
Al Viro6b9f5822005-05-01 08:59:19 -0700452 __u64 k_offset;
453 __u8 k_type;
454};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
456struct cpu_key {
Al Viro6a3a16f2005-05-01 08:59:17 -0700457 struct in_core_key on_disk_key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 int version;
459 int key_length; /* 3 in all cases but direct2indirect and
460 indirect2direct conversion */
461};
462
463/* Our function for comparing keys can compare keys of different
464 lengths. It takes as a parameter the length of the keys it is to
465 compare. These defines are used in determining what is to be passed
466 to it as that parameter. */
467#define REISERFS_FULL_KEY_LEN 4
468#define REISERFS_SHORT_KEY_LEN 2
469
470/* The result of the key compare */
471#define FIRST_GREATER 1
472#define SECOND_GREATER -1
473#define KEYS_IDENTICAL 0
474#define KEY_FOUND 1
475#define KEY_NOT_FOUND 0
476
477#define KEY_SIZE (sizeof(struct reiserfs_key))
478#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
479
480/* return values for search_by_key and clones */
481#define ITEM_FOUND 1
482#define ITEM_NOT_FOUND 0
483#define ENTRY_FOUND 1
484#define ENTRY_NOT_FOUND 0
485#define DIRECTORY_NOT_FOUND -1
486#define REGULAR_FILE_FOUND -2
487#define DIRECTORY_FOUND -3
488#define BYTE_FOUND 1
489#define BYTE_NOT_FOUND 0
490#define FILE_NOT_FOUND -1
491
492#define POSITION_FOUND 1
493#define POSITION_NOT_FOUND 0
494
495// return values for reiserfs_find_entry and search_by_entry_key
496#define NAME_FOUND 1
497#define NAME_NOT_FOUND 0
498#define GOTO_PREVIOUS_ITEM 2
499#define NAME_FOUND_INVISIBLE 3
500
501/* Everything in the filesystem is stored as a set of items. The
502 item head contains the key of the item, its free space (for
503 indirect items) and specifies the location of the item itself
504 within the block. */
505
506struct item_head
507{
508 /* Everything in the tree is found by searching for it based on
509 * its key.*/
510 struct reiserfs_key ih_key;
511 union {
512 /* The free space in the last unformatted node of an
513 indirect item if this is an indirect item. This
514 equals 0xFFFF iff this is a direct item or stat data
515 item. Note that the key, not this field, is used to
516 determine the item type, and thus which field this
517 union contains. */
Al Viro3e8962b2005-05-01 08:59:18 -0700518 __le16 ih_free_space_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 /* Iff this is a directory item, this field equals the
520 number of directory entries in the directory item. */
Al Viro3e8962b2005-05-01 08:59:18 -0700521 __le16 ih_entry_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 } __attribute__ ((__packed__)) u;
Al Viro3e8962b2005-05-01 08:59:18 -0700523 __le16 ih_item_len; /* total size of the item body */
524 __le16 ih_item_location; /* an offset to the item body
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 * within the block */
Al Viro3e8962b2005-05-01 08:59:18 -0700526 __le16 ih_version; /* 0 for all old items, 2 for new
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 ones. Highest bit is set by fsck
528 temporary, cleaned after all
529 done */
530} __attribute__ ((__packed__));
531/* size of item header */
532#define IH_SIZE (sizeof(struct item_head))
533
534#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
535#define ih_version(ih) le16_to_cpu((ih)->ih_version)
536#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
537#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
538#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
539
540#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
541#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
542#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
543#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
544#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
545
546
547#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
548
549#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
550#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
551
552/* these operate on indirect items, where you've got an array of ints
553** at a possibly unaligned location. These are a noop on ia32
554**
555** p is the array of __u32, i is the index into the array, v is the value
556** to store there.
557*/
558#define get_block_num(p, i) le32_to_cpu(get_unaligned((p) + (i)))
559#define put_block_num(p, i, v) put_unaligned(cpu_to_le32(v), (p) + (i))
560
561//
562// in old version uniqueness field shows key type
563//
564#define V1_SD_UNIQUENESS 0
565#define V1_INDIRECT_UNIQUENESS 0xfffffffe
566#define V1_DIRECT_UNIQUENESS 0xffffffff
567#define V1_DIRENTRY_UNIQUENESS 500
568#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
569
570//
571// here are conversion routines
572//
573static inline int uniqueness2type (__u32 uniqueness) CONSTF;
574static inline int uniqueness2type (__u32 uniqueness)
575{
576 switch ((int)uniqueness) {
577 case V1_SD_UNIQUENESS: return TYPE_STAT_DATA;
578 case V1_INDIRECT_UNIQUENESS: return TYPE_INDIRECT;
579 case V1_DIRECT_UNIQUENESS: return TYPE_DIRECT;
580 case V1_DIRENTRY_UNIQUENESS: return TYPE_DIRENTRY;
581 default:
582 reiserfs_warning (NULL, "vs-500: unknown uniqueness %d",
583 uniqueness);
584 case V1_ANY_UNIQUENESS:
585 return TYPE_ANY;
586 }
587}
588
589static inline __u32 type2uniqueness (int type) CONSTF;
590static inline __u32 type2uniqueness (int type)
591{
592 switch (type) {
593 case TYPE_STAT_DATA: return V1_SD_UNIQUENESS;
594 case TYPE_INDIRECT: return V1_INDIRECT_UNIQUENESS;
595 case TYPE_DIRECT: return V1_DIRECT_UNIQUENESS;
596 case TYPE_DIRENTRY: return V1_DIRENTRY_UNIQUENESS;
597 default:
598 reiserfs_warning (NULL, "vs-501: unknown type %d", type);
599 case TYPE_ANY:
600 return V1_ANY_UNIQUENESS;
601 }
602}
603
604//
605// key is pointer to on disk key which is stored in le, result is cpu,
606// there is no way to get version of object from key, so, provide
607// version to these defines
608//
609static inline loff_t le_key_k_offset (int version, const struct reiserfs_key * key)
610{
611 return (version == KEY_FORMAT_3_5) ?
612 le32_to_cpu( key->u.k_offset_v1.k_offset ) :
613 offset_v2_k_offset( &(key->u.k_offset_v2) );
614}
615
616static inline loff_t le_ih_k_offset (const struct item_head * ih)
617{
618 return le_key_k_offset (ih_version (ih), &(ih->ih_key));
619}
620
621static inline loff_t le_key_k_type (int version, const struct reiserfs_key * key)
622{
623 return (version == KEY_FORMAT_3_5) ?
624 uniqueness2type( le32_to_cpu( key->u.k_offset_v1.k_uniqueness)) :
625 offset_v2_k_type( &(key->u.k_offset_v2) );
626}
627
628static inline loff_t le_ih_k_type (const struct item_head * ih)
629{
630 return le_key_k_type (ih_version (ih), &(ih->ih_key));
631}
632
633
634static inline void set_le_key_k_offset (int version, struct reiserfs_key * key, loff_t offset)
635{
636 (version == KEY_FORMAT_3_5) ?
637 (void)(key->u.k_offset_v1.k_offset = cpu_to_le32 (offset)) : /* jdm check */
638 (void)(set_offset_v2_k_offset( &(key->u.k_offset_v2), offset ));
639}
640
641
642static inline void set_le_ih_k_offset (struct item_head * ih, loff_t offset)
643{
644 set_le_key_k_offset (ih_version (ih), &(ih->ih_key), offset);
645}
646
647
648static inline void set_le_key_k_type (int version, struct reiserfs_key * key, int type)
649{
650 (version == KEY_FORMAT_3_5) ?
651 (void)(key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type2uniqueness(type))):
652 (void)(set_offset_v2_k_type( &(key->u.k_offset_v2), type ));
653}
654static inline void set_le_ih_k_type (struct item_head * ih, int type)
655{
656 set_le_key_k_type (ih_version (ih), &(ih->ih_key), type);
657}
658
659
660#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
661#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
662#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
663#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
664
665//
666// item header has version.
667//
668#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
669#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
670#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
671#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
672
673
674
675//
676// key is pointer to cpu key, result is cpu
677//
678static inline loff_t cpu_key_k_offset (const struct cpu_key * key)
679{
Al Viro6b9f5822005-05-01 08:59:19 -0700680 return key->on_disk_key.k_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681}
682
683static inline loff_t cpu_key_k_type (const struct cpu_key * key)
684{
Al Viro6b9f5822005-05-01 08:59:19 -0700685 return key->on_disk_key.k_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686}
687
688static inline void set_cpu_key_k_offset (struct cpu_key * key, loff_t offset)
689{
Al Viro6b9f5822005-05-01 08:59:19 -0700690 key->on_disk_key.k_offset = offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691}
692
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693static inline void set_cpu_key_k_type (struct cpu_key * key, int type)
694{
Al Viro6b9f5822005-05-01 08:59:19 -0700695 key->on_disk_key.k_type = type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696}
697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698static inline void cpu_key_k_offset_dec (struct cpu_key * key)
699{
Al Viro6b9f5822005-05-01 08:59:19 -0700700 key->on_disk_key.k_offset --;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701}
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
704#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
705#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
706#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
707
708
709/* are these used ? */
710#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
711#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
712#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
713#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
714
715
716
717
718
719#define I_K_KEY_IN_ITEM(p_s_ih, p_s_key, n_blocksize) \
720 ( ! COMP_SHORT_KEYS(p_s_ih, p_s_key) && \
721 I_OFF_BYTE_IN_ITEM(p_s_ih, k_offset (p_s_key), n_blocksize) )
722
723/* maximal length of item */
724#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
725#define MIN_ITEM_LEN 1
726
727
728/* object identifier for root dir */
729#define REISERFS_ROOT_OBJECTID 2
730#define REISERFS_ROOT_PARENT_OBJECTID 1
731extern struct reiserfs_key root_key;
732
733
734
735
736/*
737 * Picture represents a leaf of the S+tree
738 * ______________________________________________________
739 * | | Array of | | |
740 * |Block | Object-Item | F r e e | Objects- |
741 * | head | Headers | S p a c e | Items |
742 * |______|_______________|___________________|___________|
743 */
744
745/* Header of a disk block. More precisely, header of a formatted leaf
746 or internal node, and not the header of an unformatted node. */
747struct block_head {
Al Viro3e8962b2005-05-01 08:59:18 -0700748 __le16 blk_level; /* Level of a block in the tree. */
749 __le16 blk_nr_item; /* Number of keys/items in a block. */
750 __le16 blk_free_space; /* Block free space in bytes. */
751 __le16 blk_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 /* dump this in v4/planA */
753 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
754};
755
756#define BLKH_SIZE (sizeof(struct block_head))
757#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
758#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
759#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
760#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
761#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
762#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
763#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
764#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
765#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
766#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
767
768/*
769 * values for blk_level field of the struct block_head
770 */
771
772#define FREE_LEVEL 0 /* when node gets removed from the tree its
773 blk_level is set to FREE_LEVEL. It is then
774 used to see whether the node is still in the
775 tree */
776
777#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level.*/
778
779/* Given the buffer head of a formatted node, resolve to the block head of that node. */
780#define B_BLK_HEAD(p_s_bh) ((struct block_head *)((p_s_bh)->b_data))
781/* Number of items that are in buffer. */
782#define B_NR_ITEMS(p_s_bh) (blkh_nr_item(B_BLK_HEAD(p_s_bh)))
783#define B_LEVEL(p_s_bh) (blkh_level(B_BLK_HEAD(p_s_bh)))
784#define B_FREE_SPACE(p_s_bh) (blkh_free_space(B_BLK_HEAD(p_s_bh)))
785
786#define PUT_B_NR_ITEMS(p_s_bh,val) do { set_blkh_nr_item(B_BLK_HEAD(p_s_bh),val); } while (0)
787#define PUT_B_LEVEL(p_s_bh,val) do { set_blkh_level(B_BLK_HEAD(p_s_bh),val); } while (0)
788#define PUT_B_FREE_SPACE(p_s_bh,val) do { set_blkh_free_space(B_BLK_HEAD(p_s_bh),val); } while (0)
789
790
791/* Get right delimiting key. -- little endian */
792#define B_PRIGHT_DELIM_KEY(p_s_bh) (&(blk_right_delim_key(B_BLK_HEAD(p_s_bh))
793
794/* Does the buffer contain a disk leaf. */
795#define B_IS_ITEMS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) == DISK_LEAF_NODE_LEVEL)
796
797/* Does the buffer contain a disk internal node */
798#define B_IS_KEYS_LEVEL(p_s_bh) (B_LEVEL(p_s_bh) > DISK_LEAF_NODE_LEVEL \
799 && B_LEVEL(p_s_bh) <= MAX_HEIGHT)
800
801
802
803
804/***************************************************************************/
805/* STAT DATA */
806/***************************************************************************/
807
808
809//
810// old stat data is 32 bytes long. We are going to distinguish new one by
811// different size
812//
813struct stat_data_v1
814{
Al Viro3e8962b2005-05-01 08:59:18 -0700815 __le16 sd_mode; /* file type, permissions */
816 __le16 sd_nlink; /* number of hard links */
817 __le16 sd_uid; /* owner */
818 __le16 sd_gid; /* group */
819 __le32 sd_size; /* file size */
820 __le32 sd_atime; /* time of last access */
821 __le32 sd_mtime; /* time file was last modified */
822 __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 -0700823 union {
Al Viro3e8962b2005-05-01 08:59:18 -0700824 __le32 sd_rdev;
825 __le32 sd_blocks; /* number of blocks file uses */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 } __attribute__ ((__packed__)) u;
Al Viro3e8962b2005-05-01 08:59:18 -0700827 __le32 sd_first_direct_byte; /* first byte of file which is stored
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 in a direct item: except that if it
829 equals 1 it is a symlink and if it
830 equals ~(__u32)0 there is no
831 direct item. The existence of this
832 field really grates on me. Let's
833 replace it with a macro based on
834 sd_size and our tail suppression
835 policy. Someday. -Hans */
836} __attribute__ ((__packed__));
837
838#define SD_V1_SIZE (sizeof(struct stat_data_v1))
839#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
840#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
841#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
842#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
843#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
844#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
845#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
846#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
847#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
848#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
849#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
850#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
851#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
852#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
853#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
854#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
855#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
856#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
857#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
858#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
859#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
860#define sd_v1_first_direct_byte(sdp) \
861 (le32_to_cpu((sdp)->sd_first_direct_byte))
862#define set_sd_v1_first_direct_byte(sdp,v) \
863 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
864
865#include <linux/ext2_fs.h>
866
867/* inode flags stored in sd_attrs (nee sd_reserved) */
868
869/* we want common flags to have the same values as in ext2,
870 so chattr(1) will work without problems */
871#define REISERFS_IMMUTABLE_FL EXT2_IMMUTABLE_FL
872#define REISERFS_APPEND_FL EXT2_APPEND_FL
873#define REISERFS_SYNC_FL EXT2_SYNC_FL
874#define REISERFS_NOATIME_FL EXT2_NOATIME_FL
875#define REISERFS_NODUMP_FL EXT2_NODUMP_FL
876#define REISERFS_SECRM_FL EXT2_SECRM_FL
877#define REISERFS_UNRM_FL EXT2_UNRM_FL
878#define REISERFS_COMPR_FL EXT2_COMPR_FL
879#define REISERFS_NOTAIL_FL EXT2_NOTAIL_FL
880
881/* persistent flags that file inherits from the parent directory */
882#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
883 REISERFS_SYNC_FL | \
884 REISERFS_NOATIME_FL | \
885 REISERFS_NODUMP_FL | \
886 REISERFS_SECRM_FL | \
887 REISERFS_COMPR_FL | \
888 REISERFS_NOTAIL_FL )
889
890/* Stat Data on disk (reiserfs version of UFS disk inode minus the
891 address blocks) */
892struct stat_data {
Al Viro3e8962b2005-05-01 08:59:18 -0700893 __le16 sd_mode; /* file type, permissions */
894 __le16 sd_attrs; /* persistent inode flags */
895 __le32 sd_nlink; /* number of hard links */
896 __le64 sd_size; /* file size */
897 __le32 sd_uid; /* owner */
898 __le32 sd_gid; /* group */
899 __le32 sd_atime; /* time of last access */
900 __le32 sd_mtime; /* time file was last modified */
901 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
902 __le32 sd_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 union {
Al Viro3e8962b2005-05-01 08:59:18 -0700904 __le32 sd_rdev;
905 __le32 sd_generation;
906 //__le32 sd_first_direct_byte;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 /* first byte of file which is stored in a
908 direct item: except that if it equals 1
909 it is a symlink and if it equals
910 ~(__u32)0 there is no direct item. The
911 existence of this field really grates
912 on me. Let's replace it with a macro
913 based on sd_size and our tail
914 suppression policy? */
915 } __attribute__ ((__packed__)) u;
916} __attribute__ ((__packed__));
917//
918// this is 44 bytes long
919//
920#define SD_SIZE (sizeof(struct stat_data))
921#define SD_V2_SIZE SD_SIZE
922#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
923#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
924#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
925/* sd_reserved */
926/* set_sd_reserved */
927#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
928#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
929#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
930#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
931#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
932#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
933#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
934#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
935#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
936#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
937#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
938#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
939#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
940#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
941#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
942#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
943#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
944#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
945#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
946#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
947#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
948#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
949
950
951/***************************************************************************/
952/* DIRECTORY STRUCTURE */
953/***************************************************************************/
954/*
955 Picture represents the structure of directory items
956 ________________________________________________
957 | Array of | | | | | |
958 | directory |N-1| N-2 | .... | 1st |0th|
959 | entry headers | | | | | |
960 |_______________|___|_____|________|_______|___|
961 <---- directory entries ------>
962
963 First directory item has k_offset component 1. We store "." and ".."
964 in one item, always, we never split "." and ".." into differing
965 items. This makes, among other things, the code for removing
966 directories simpler. */
967#define SD_OFFSET 0
968#define SD_UNIQUENESS 0
969#define DOT_OFFSET 1
970#define DOT_DOT_OFFSET 2
971#define DIRENTRY_UNIQUENESS 500
972
973/* */
974#define FIRST_ITEM_OFFSET 1
975
976/*
977 Q: How to get key of object pointed to by entry from entry?
978
979 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
980 of object, entry points to */
981
982/* NOT IMPLEMENTED:
983 Directory will someday contain stat data of object */
984
985
986
987struct reiserfs_de_head
988{
Al Viro3e8962b2005-05-01 08:59:18 -0700989 __le32 deh_offset; /* third component of the directory entry key */
990 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 by directory entry */
Al Viro3e8962b2005-05-01 08:59:18 -0700992 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
993 __le16 deh_location; /* offset of name in the whole item */
994 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 entry is hidden (unlinked) */
996} __attribute__ ((__packed__));
997#define DEH_SIZE sizeof(struct reiserfs_de_head)
998#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
999#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
1000#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
1001#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
1002#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
1003
1004#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
1005#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
1006#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
1007#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
1008#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
1009
1010/* empty directory contains two entries "." and ".." and their headers */
1011#define EMPTY_DIR_SIZE \
1012(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
1013
1014/* old format directories have this size when empty */
1015#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1016
1017#define DEH_Statdata 0 /* not used now */
1018#define DEH_Visible 2
1019
1020/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1021#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1022# define ADDR_UNALIGNED_BITS (3)
1023#endif
1024
1025/* These are only used to manipulate deh_state.
1026 * Because of this, we'll use the ext2_ bit routines,
1027 * since they are little endian */
1028#ifdef ADDR_UNALIGNED_BITS
1029
1030# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1031# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1032
1033# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1034# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1035# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1036
1037#else
1038
1039# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
1040# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
1041# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
1042
1043#endif
1044
1045#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1046#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1047#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1048#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1049
1050#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1051#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1052#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1053
Al Viro3e8962b2005-05-01 08:59:18 -07001054extern void make_empty_dir_item_v1 (char * body, __le32 dirid, __le32 objid,
1055 __le32 par_dirid, __le32 par_objid);
1056extern void make_empty_dir_item (char * body, __le32 dirid, __le32 objid,
1057 __le32 par_dirid, __le32 par_objid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
1059/* array of the entry headers */
1060 /* get item body */
1061#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1062#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1063
1064/* length of the directory entry in directory item. This define
1065 calculates length of i-th directory entry using directory entry
1066 locations from dir entry head. When it calculates length of 0-th
1067 directory entry, it uses length of whole item in place of entry
1068 location of the non-existent following entry in the calculation.
1069 See picture above.*/
1070/*
1071#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1072((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1073*/
1074static inline int entry_length (const struct buffer_head * bh,
1075 const struct item_head * ih, int pos_in_item)
1076{
1077 struct reiserfs_de_head * deh;
1078
1079 deh = B_I_DEH (bh, ih) + pos_in_item;
1080 if (pos_in_item)
1081 return deh_location(deh-1) - deh_location(deh);
1082
1083 return ih_item_len(ih) - deh_location(deh);
1084}
1085
1086
1087
1088/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1089#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1090
1091
1092/* name by bh, ih and entry_num */
1093#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))))
1094
1095// two entries per block (at least)
1096#define REISERFS_MAX_NAME(block_size) 255
1097
1098
1099/* this structure is used for operations on directory entries. It is
1100 not a disk structure. */
1101/* When reiserfs_find_entry or search_by_entry_key find directory
1102 entry, they return filled reiserfs_dir_entry structure */
1103struct reiserfs_dir_entry
1104{
1105 struct buffer_head * de_bh;
1106 int de_item_num;
1107 struct item_head * de_ih;
1108 int de_entry_num;
1109 struct reiserfs_de_head * de_deh;
1110 int de_entrylen;
1111 int de_namelen;
1112 char * de_name;
1113 char * de_gen_number_bit_string;
1114
1115 __u32 de_dir_id;
1116 __u32 de_objectid;
1117
1118 struct cpu_key de_entry_key;
1119};
1120
1121/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1122
1123/* pointer to file name, stored in entry */
1124#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1125
1126/* length of name */
1127#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1128(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1129
1130
1131
1132/* hash value occupies bits from 7 up to 30 */
1133#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1134/* generation number occupies 7 bits starting from 0 up to 6 */
1135#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1136#define MAX_GENERATION_NUMBER 127
1137
1138#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1139
1140
1141/*
1142 * Picture represents an internal node of the reiserfs tree
1143 * ______________________________________________________
1144 * | | Array of | Array of | Free |
1145 * |block | keys | pointers | space |
1146 * | head | N | N+1 | |
1147 * |______|_______________|___________________|___________|
1148 */
1149
1150/***************************************************************************/
1151/* DISK CHILD */
1152/***************************************************************************/
1153/* Disk child pointer: The pointer from an internal node of the tree
1154 to a node that is on disk. */
1155struct disk_child {
Al Viro3e8962b2005-05-01 08:59:18 -07001156 __le32 dc_block_number; /* Disk child's block number. */
1157 __le16 dc_size; /* Disk child's used space. */
1158 __le16 dc_reserved;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159};
1160
1161#define DC_SIZE (sizeof(struct disk_child))
1162#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1163#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1164#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1165#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1166
1167/* Get disk child by buffer header and position in the tree node. */
1168#define B_N_CHILD(p_s_bh,n_pos) ((struct disk_child *)\
1169((p_s_bh)->b_data+BLKH_SIZE+B_NR_ITEMS(p_s_bh)*KEY_SIZE+DC_SIZE*(n_pos)))
1170
1171/* Get disk child number by buffer header and position in the tree node. */
1172#define B_N_CHILD_NUM(p_s_bh,n_pos) (dc_block_number(B_N_CHILD(p_s_bh,n_pos)))
1173#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 ))
1174
1175 /* maximal value of field child_size in structure disk_child */
1176 /* child size is the combined size of all items and their headers */
1177#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1178
1179/* amount of used space in buffer (not including block head) */
1180#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1181
1182/* max and min number of keys in internal node */
1183#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1184#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1185
1186/***************************************************************************/
1187/* PATH STRUCTURES AND DEFINES */
1188/***************************************************************************/
1189
1190
1191/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1192 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1193 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1194 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1195 position of the block_number of the next node if it is looking through an internal node. If it
1196 is looking through a leaf node bin_search will find the position of the item which has key either
1197 equal to given key, or which is the maximal key less than the given key. */
1198
1199struct path_element {
1200 struct buffer_head * pe_buffer; /* Pointer to the buffer at the path in the tree. */
1201 int pe_position; /* Position in the tree node which is placed in the */
1202 /* buffer above. */
1203};
1204
1205#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1206#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1207#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1208
1209#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1210#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1211
1212
1213
1214/* We need to keep track of who the ancestors of nodes are. When we
1215 perform a search we record which nodes were visited while
1216 descending the tree looking for the node we searched for. This list
1217 of nodes is called the path. This information is used while
1218 performing balancing. Note that this path information may become
1219 invalid, and this means we must check it when using it to see if it
1220 is still valid. You'll need to read search_by_key and the comments
1221 in it, especially about decrement_counters_in_path(), to understand
1222 this structure.
1223
1224Paths make the code so much harder to work with and debug.... An
1225enormous number of bugs are due to them, and trying to write or modify
1226code that uses them just makes my head hurt. They are based on an
1227excessive effort to avoid disturbing the precious VFS code.:-( The
1228gods only know how we are going to SMP the code that uses them.
1229znodes are the way! */
1230
1231#define PATH_READA 0x1 /* do read ahead */
1232#define PATH_READA_BACK 0x2 /* read backwards */
1233
1234struct path {
1235 int path_length; /* Length of the array above. */
1236 int reada;
1237 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1238 int pos_in_item;
1239};
1240
1241#define pos_in_item(path) ((path)->pos_in_item)
1242
1243#define INITIALIZE_PATH(var) \
1244struct path var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1245
1246/* Get path element by path and path position. */
1247#define PATH_OFFSET_PELEMENT(p_s_path,n_offset) ((p_s_path)->path_elements +(n_offset))
1248
1249/* Get buffer header at the path by path and path position. */
1250#define PATH_OFFSET_PBUFFER(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_buffer)
1251
1252/* Get position in the element at the path by path and path position. */
1253#define PATH_OFFSET_POSITION(p_s_path,n_offset) (PATH_OFFSET_PELEMENT(p_s_path,n_offset)->pe_position)
1254
1255
1256#define PATH_PLAST_BUFFER(p_s_path) (PATH_OFFSET_PBUFFER((p_s_path), (p_s_path)->path_length))
1257 /* you know, to the person who didn't
1258 write this the macro name does not
1259 at first suggest what it does.
1260 Maybe POSITION_FROM_PATH_END? Or
1261 maybe we should just focus on
1262 dumping paths... -Hans */
1263#define PATH_LAST_POSITION(p_s_path) (PATH_OFFSET_POSITION((p_s_path), (p_s_path)->path_length))
1264
1265
1266#define PATH_PITEM_HEAD(p_s_path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_path),PATH_LAST_POSITION(p_s_path))
1267
1268/* in do_balance leaf has h == 0 in contrast with path structure,
1269 where root has level == 0. That is why we need these defines */
1270#define PATH_H_PBUFFER(p_s_path, h) PATH_OFFSET_PBUFFER (p_s_path, p_s_path->path_length - (h)) /* tb->S[h] */
1271#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1272#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1273#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1274
1275#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
1276
1277#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1278#define get_ih(path) PATH_PITEM_HEAD(path)
1279#define get_item_pos(path) PATH_LAST_POSITION(path)
1280#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1281#define item_moved(ih,path) comp_items(ih, path)
1282#define path_changed(ih,path) comp_items (ih, path)
1283
1284
1285/***************************************************************************/
1286/* MISC */
1287/***************************************************************************/
1288
1289/* Size of pointer to the unformatted node. */
1290#define UNFM_P_SIZE (sizeof(unp_t))
1291#define UNFM_P_SHIFT 2
1292
1293// in in-core inode key is stored on le form
1294#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1295
1296#define MAX_UL_INT 0xffffffff
1297#define MAX_INT 0x7ffffff
1298#define MAX_US_INT 0xffff
1299
1300// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1301#define U32_MAX (~(__u32)0)
1302
1303static inline loff_t max_reiserfs_offset (struct inode * inode)
1304{
1305 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1306 return (loff_t)U32_MAX;
1307
1308 return (loff_t)((~(__u64)0) >> 4);
1309}
1310
1311
1312/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1313#define MAX_KEY_OBJECTID MAX_UL_INT
1314
1315
1316#define MAX_B_NUM MAX_UL_INT
1317#define MAX_FC_NUM MAX_US_INT
1318
1319
1320/* the purpose is to detect overflow of an unsigned short */
1321#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1322
1323
1324/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
1325#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1326#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1327
1328#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1329#define get_generation(s) atomic_read (&fs_generation(s))
1330#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1331#define __fs_changed(gen,s) (gen != get_generation (s))
1332#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1333
1334
1335/***************************************************************************/
1336/* FIXATE NODES */
1337/***************************************************************************/
1338
1339#define VI_TYPE_LEFT_MERGEABLE 1
1340#define VI_TYPE_RIGHT_MERGEABLE 2
1341
1342/* To make any changes in the tree we always first find node, that
1343 contains item to be changed/deleted or place to insert a new
1344 item. We call this node S. To do balancing we need to decide what
1345 we will shift to left/right neighbor, or to a new node, where new
1346 item will be etc. To make this analysis simpler we build virtual
1347 node. Virtual node is an array of items, that will replace items of
1348 node S. (For instance if we are going to delete an item, virtual
1349 node does not contain it). Virtual node keeps information about
1350 item sizes and types, mergeability of first and last items, sizes
1351 of all entries in directory item. We use this array of items when
1352 calculating what we can shift to neighbors and how many nodes we
1353 have to have if we do not any shiftings, if we shift to left/right
1354 neighbor or to both. */
1355struct virtual_item
1356{
1357 int vi_index; // index in the array of item operations
1358 unsigned short vi_type; // left/right mergeability
1359 unsigned short vi_item_len; /* length of item that it will have after balancing */
1360 struct item_head * vi_ih;
1361 const char * vi_item; // body of item (old or new)
1362 const void * vi_new_data; // 0 always but paste mode
1363 void * vi_uarea; // item specific area
1364};
1365
1366
1367struct virtual_node
1368{
1369 char * vn_free_ptr; /* this is a pointer to the free space in the buffer */
1370 unsigned short vn_nr_item; /* number of items in virtual node */
1371 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1372 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1373 short vn_affected_item_num;
1374 short vn_pos_in_item;
1375 struct item_head * vn_ins_ih; /* item header of inserted item, 0 for other modes */
1376 const void * vn_data;
1377 struct virtual_item * vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1378};
1379
1380/* used by directory items when creating virtual nodes */
1381struct direntry_uarea {
1382 int flags;
1383 __u16 entry_count;
1384 __u16 entry_sizes[1];
1385} __attribute__ ((__packed__)) ;
1386
1387
1388/***************************************************************************/
1389/* TREE BALANCE */
1390/***************************************************************************/
1391
1392/* This temporary structure is used in tree balance algorithms, and
1393 constructed as we go to the extent that its various parts are
1394 needed. It contains arrays of nodes that can potentially be
1395 involved in the balancing of node S, and parameters that define how
1396 each of the nodes must be balanced. Note that in these algorithms
1397 for balancing the worst case is to need to balance the current node
1398 S and the left and right neighbors and all of their parents plus
1399 create a new node. We implement S1 balancing for the leaf nodes
1400 and S0 balancing for the internal nodes (S1 and S0 are defined in
1401 our papers.)*/
1402
1403#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1404
1405/* maximum number of FEB blocknrs on a single level */
1406#define MAX_AMOUNT_NEEDED 2
1407
1408/* someday somebody will prefix every field in this struct with tb_ */
1409struct tree_balance
1410{
1411 int tb_mode;
1412 int need_balance_dirty;
1413 struct super_block * tb_sb;
1414 struct reiserfs_transaction_handle *transaction_handle ;
1415 struct path * tb_path;
1416 struct buffer_head * L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1417 struct buffer_head * R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path*/
1418 struct buffer_head * FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1419 struct buffer_head * FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1420 struct buffer_head * CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1421 struct buffer_head * CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1422
1423 struct buffer_head * FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1424 cur_blknum. */
1425 struct buffer_head * used[MAX_FEB_SIZE];
1426 struct buffer_head * thrown[MAX_FEB_SIZE];
1427 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1428 shifted to the left in order to balance the
1429 current node; for leaves includes item that
1430 will be partially shifted; for internal
1431 nodes, it is the number of child pointers
1432 rather than items. It includes the new item
1433 being created. The code sometimes subtracts
1434 one to get the number of wholly shifted
1435 items for other purposes. */
1436 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1437 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1438 S[h] to its item number within the node CFL[h] */
1439 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1440 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1441 S[h]. A negative value means removing. */
1442 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1443 balancing on the level h of the tree. If 0 then S is
1444 being deleted, if 1 then S is remaining and no new nodes
1445 are being created, if 2 or 3 then 1 or 2 new nodes is
1446 being created */
1447
1448 /* fields that are used only for balancing leaves of the tree */
1449 int cur_blknum; /* number of empty blocks having been already allocated */
1450 int s0num; /* number of items that fall into left most node when S[0] splits */
1451 int s1num; /* number of items that fall into first new node when S[0] splits */
1452 int s2num; /* number of items that fall into second new node when S[0] splits */
1453 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1454 /* most liquid item that cannot be shifted from S[0] entirely */
1455 /* if -1 then nothing will be partially shifted */
1456 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1457 /* most liquid item that cannot be shifted from S[0] entirely */
1458 /* if -1 then nothing will be partially shifted */
1459 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1460 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1461 int s2bytes;
1462 struct buffer_head * buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1463 char * vn_buf; /* kmalloced memory. Used to create
1464 virtual node and keep map of
1465 dirtied bitmap blocks */
1466 int vn_buf_size; /* size of the vn_buf */
1467 struct virtual_node * tb_vn; /* VN starts after bitmap of bitmap blocks */
1468
1469 int fs_gen; /* saved value of `reiserfs_generation' counter
1470 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1471#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Al Viro6a3a16f2005-05-01 08:59:17 -07001472 struct in_core_key key; /* key pointer, to pass to block allocator or
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 another low-level subsystem */
1474#endif
1475} ;
1476
1477/* These are modes of balancing */
1478
1479/* When inserting an item. */
1480#define M_INSERT 'i'
1481/* When inserting into (directories only) or appending onto an already
1482 existant item. */
1483#define M_PASTE 'p'
1484/* When deleting an item. */
1485#define M_DELETE 'd'
1486/* When truncating an item or removing an entry from a (directory) item. */
1487#define M_CUT 'c'
1488
1489/* used when balancing on leaf level skipped (in reiserfsck) */
1490#define M_INTERNAL 'n'
1491
1492/* When further balancing is not needed, then do_balance does not need
1493 to be called. */
1494#define M_SKIP_BALANCING 's'
1495#define M_CONVERT 'v'
1496
1497/* modes of leaf_move_items */
1498#define LEAF_FROM_S_TO_L 0
1499#define LEAF_FROM_S_TO_R 1
1500#define LEAF_FROM_R_TO_L 2
1501#define LEAF_FROM_L_TO_R 3
1502#define LEAF_FROM_S_TO_SNEW 4
1503
1504#define FIRST_TO_LAST 0
1505#define LAST_TO_FIRST 1
1506
1507/* used in do_balance for passing parent of node information that has
1508 been gotten from tb struct */
1509struct buffer_info {
1510 struct tree_balance * tb;
1511 struct buffer_head * bi_bh;
1512 struct buffer_head * bi_parent;
1513 int bi_position;
1514};
1515
1516
1517/* there are 4 types of items: stat data, directory item, indirect, direct.
1518+-------------------+------------+--------------+------------+
1519| | k_offset | k_uniqueness | mergeable? |
1520+-------------------+------------+--------------+------------+
1521| stat data | 0 | 0 | no |
1522+-------------------+------------+--------------+------------+
1523| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1524| non 1st directory | hash value | | yes |
1525| item | | | |
1526+-------------------+------------+--------------+------------+
1527| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1528+-------------------+------------+--------------+------------+
1529| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1530+-------------------+------------+--------------+------------+
1531*/
1532
1533struct item_operations {
1534 int (*bytes_number) (struct item_head * ih, int block_size);
1535 void (*decrement_key) (struct cpu_key *);
1536 int (*is_left_mergeable) (struct reiserfs_key * ih, unsigned long bsize);
1537 void (*print_item) (struct item_head *, char * item);
1538 void (*check_item) (struct item_head *, char * item);
1539
1540 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1541 int is_affected, int insert_size);
1542 int (*check_left) (struct virtual_item * vi, int free,
1543 int start_skip, int end_skip);
1544 int (*check_right) (struct virtual_item * vi, int free);
1545 int (*part_size) (struct virtual_item * vi, int from, int to);
1546 int (*unit_num) (struct virtual_item * vi);
1547 void (*print_vi) (struct virtual_item * vi);
1548};
1549
1550
1551extern struct item_operations * item_ops [TYPE_ANY + 1];
1552
1553#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1554#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1555#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1556#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1557#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)
1558#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1559#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1560#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1561#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1562#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1563
1564
1565
1566#define COMP_SHORT_KEYS comp_short_keys
1567
1568/* number of blocks pointed to by the indirect item */
1569#define I_UNFM_NUM(p_s_ih) ( ih_item_len(p_s_ih) / UNFM_P_SIZE )
1570
1571/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1572#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1573
1574/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1575
1576
1577/* get the item header */
1578#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1579
1580/* get key */
1581#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1582
1583/* get the key */
1584#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1585
1586/* get item body */
1587#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1588
1589/* get the stat data by the buffer header and the item order */
1590#define B_N_STAT_DATA(bh,nr) \
1591( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1592
1593 /* following defines use reiserfs buffer header and item header */
1594
1595/* get stat-data */
1596#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1597
1598// this is 3976 for size==4096
1599#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1600
1601/* indirect items consist of entries which contain blocknrs, pos
1602 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1603 blocknr contained by the entry pos points to */
1604#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1605#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)
1606
1607struct reiserfs_iget_args {
1608 __u32 objectid ;
1609 __u32 dirid ;
1610} ;
1611
1612/***************************************************************************/
1613/* FUNCTION DECLARATIONS */
1614/***************************************************************************/
1615
1616/*#ifdef __KERNEL__*/
1617#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1618
1619#define journal_trans_half(blocksize) \
1620 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1621
1622/* journal.c see journal.c for all the comments here */
1623
1624/* first block written in a commit. */
1625struct reiserfs_journal_desc {
Al Viro3e8962b2005-05-01 08:59:18 -07001626 __le32 j_trans_id ; /* id of commit */
1627 __le32 j_len ; /* length of commit. len +1 is the commit block */
1628 __le32 j_mount_id ; /* mount id of this trans*/
1629 __le32 j_realblock[1] ; /* real locations for each block */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630} ;
1631
1632#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1633#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1634#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1635
1636#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1637#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1638#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1639
1640/* last block written in a commit */
1641struct reiserfs_journal_commit {
Al Viro3e8962b2005-05-01 08:59:18 -07001642 __le32 j_trans_id ; /* must match j_trans_id from the desc block */
1643 __le32 j_len ; /* ditto */
1644 __le32 j_realblock[1] ; /* real locations for each block */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645} ;
1646
1647#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1648#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1649#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1650
1651#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1652#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1653
1654/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1655** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1656** and this transaction does not need to be replayed.
1657*/
1658struct reiserfs_journal_header {
Al Viro3e8962b2005-05-01 08:59:18 -07001659 __le32 j_last_flush_trans_id ; /* id of last fully flushed transaction */
1660 __le32 j_first_unflushed_offset ; /* offset in the log of where to start replay after a crash */
1661 __le32 j_mount_id ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 /* 12 */ struct journal_params jh_journal;
1663} ;
1664
1665/* biggest tunable defines are right here */
1666#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1667#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1668#define JOURNAL_TRANS_MIN_DEFAULT 256
1669#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1670#define JOURNAL_MIN_RATIO 2
1671#define JOURNAL_MAX_COMMIT_AGE 30
1672#define JOURNAL_MAX_TRANS_AGE 30
1673#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
1674#ifdef CONFIG_QUOTA
1675#define REISERFS_QUOTA_TRANS_BLOCKS 2 /* We need to update data and inode (atime) */
1676#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 */
1677#else
1678#define REISERFS_QUOTA_TRANS_BLOCKS 0
1679#define REISERFS_QUOTA_INIT_BLOCKS 0
1680#endif
1681
1682/* both of these can be as low as 1, or as high as you want. The min is the
1683** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1684** as needed, and released when transactions are committed. On release, if
1685** the current number of nodes is > max, the node is freed, otherwise,
1686** it is put on a free list for faster use later.
1687*/
1688#define REISERFS_MIN_BITMAP_NODES 10
1689#define REISERFS_MAX_BITMAP_NODES 100
1690
1691#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1692#define JBH_HASH_MASK 8191
1693
1694#define _jhashfn(sb,block) \
1695 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1696 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1697#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1698
1699// We need these to make journal.c code more readable
1700#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1701#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1702#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1703
1704enum reiserfs_bh_state_bits {
1705 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1706 BH_JDirty_wait,
1707 BH_JNew, /* disk block was taken off free list before
1708 * being in a finished transaction, or
1709 * written to disk. Can be reused immed. */
1710 BH_JPrepared,
1711 BH_JRestore_dirty,
1712 BH_JTest, // debugging only will go away
1713};
1714
1715BUFFER_FNS(JDirty, journaled);
1716TAS_BUFFER_FNS(JDirty, journaled);
1717BUFFER_FNS(JDirty_wait, journal_dirty);
1718TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1719BUFFER_FNS(JNew, journal_new);
1720TAS_BUFFER_FNS(JNew, journal_new);
1721BUFFER_FNS(JPrepared, journal_prepared);
1722TAS_BUFFER_FNS(JPrepared, journal_prepared);
1723BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1724TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1725BUFFER_FNS(JTest, journal_test);
1726TAS_BUFFER_FNS(JTest, journal_test);
1727
1728/*
1729** transaction handle which is passed around for all journal calls
1730*/
1731struct reiserfs_transaction_handle {
1732 struct super_block *t_super ; /* super for this FS when journal_begin was
1733 called. saves calls to reiserfs_get_super
1734 also used by nested transactions to make
1735 sure they are nesting on the right FS
1736 _must_ be first in the handle
1737 */
1738 int t_refcount;
1739 int t_blocks_logged ; /* number of blocks this writer has logged */
1740 int t_blocks_allocated ; /* number of blocks this writer allocated */
1741 unsigned long t_trans_id ; /* sanity check, equals the current trans id */
1742 void *t_handle_save ; /* save existing current->journal_info */
1743 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1744 should be displaced from others */
1745 struct list_head t_list;
1746} ;
1747
1748/* used to keep track of ordered and tail writes, attached to the buffer
1749 * head through b_journal_head.
1750 */
1751struct reiserfs_jh {
1752 struct reiserfs_journal_list *jl;
1753 struct buffer_head *bh;
1754 struct list_head list;
1755};
1756
1757void reiserfs_free_jh(struct buffer_head *bh);
1758int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1759int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
1760int journal_mark_dirty(struct reiserfs_transaction_handle *, struct super_block *, struct buffer_head *bh) ;
1761
1762static inline int
1763reiserfs_file_data_log(struct inode *inode) {
1764 if (reiserfs_data_log(inode->i_sb) ||
1765 (REISERFS_I(inode)->i_flags & i_data_log))
1766 return 1 ;
1767 return 0 ;
1768}
1769
1770static inline int reiserfs_transaction_running(struct super_block *s) {
1771 struct reiserfs_transaction_handle *th = current->journal_info ;
1772 if (th && th->t_super == s)
1773 return 1 ;
1774 if (th && th->t_super == NULL)
1775 BUG();
1776 return 0 ;
1777}
1778
1779int reiserfs_async_progress_wait(struct super_block *s);
1780
1781struct reiserfs_transaction_handle *
1782reiserfs_persistent_transaction(struct super_block *, int count);
1783int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1784int reiserfs_commit_page(struct inode *inode, struct page *page,
1785 unsigned from, unsigned to);
1786int reiserfs_flush_old_commits(struct super_block *);
1787int reiserfs_commit_for_inode(struct inode *) ;
1788int reiserfs_inode_needs_commit(struct inode *) ;
1789void reiserfs_update_inode_transaction(struct inode *) ;
1790void reiserfs_wait_on_write_block(struct super_block *s) ;
1791void reiserfs_block_writes(struct reiserfs_transaction_handle *th) ;
1792void reiserfs_allow_writes(struct super_block *s) ;
1793void reiserfs_check_lock_depth(struct super_block *s, char *caller) ;
1794int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh, int wait) ;
1795void reiserfs_restore_prepared_buffer(struct super_block *, struct buffer_head *bh) ;
1796int journal_init(struct super_block *, const char * j_dev_name, int old_format, unsigned int) ;
1797int journal_release(struct reiserfs_transaction_handle*, struct super_block *) ;
1798int journal_release_error(struct reiserfs_transaction_handle*, struct super_block *) ;
1799int journal_end(struct reiserfs_transaction_handle *, struct super_block *, unsigned long) ;
1800int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *, unsigned long) ;
1801int journal_mark_freed(struct reiserfs_transaction_handle *, struct super_block *, b_blocknr_t blocknr) ;
1802int journal_transaction_should_end(struct reiserfs_transaction_handle *, int) ;
1803int reiserfs_in_journal(struct super_block *p_s_sb, int bmap_nr, int bit_nr, int searchall, b_blocknr_t *next) ;
1804int journal_begin(struct reiserfs_transaction_handle *, struct super_block *p_s_sb, unsigned long) ;
1805int journal_join_abort(struct reiserfs_transaction_handle *, struct super_block *p_s_sb, unsigned long) ;
1806void reiserfs_journal_abort (struct super_block *sb, int errno);
1807void reiserfs_abort (struct super_block *sb, int errno, const char *fmt, ...);
1808int reiserfs_allocate_list_bitmaps(struct super_block *s, struct reiserfs_list_bitmap *, int) ;
1809
1810void add_save_link (struct reiserfs_transaction_handle * th,
1811 struct inode * inode, int truncate);
1812int remove_save_link (struct inode * inode, int truncate);
1813
1814/* objectid.c */
1815__u32 reiserfs_get_unused_objectid (struct reiserfs_transaction_handle *th);
1816void reiserfs_release_objectid (struct reiserfs_transaction_handle *th, __u32 objectid_to_release);
1817int reiserfs_convert_objectid_map_v1(struct super_block *) ;
1818
1819/* stree.c */
1820int B_IS_IN_TREE(const struct buffer_head *);
1821extern void copy_item_head(struct item_head * p_v_to,
1822 const struct item_head * p_v_from);
1823
1824// first key is in cpu form, second - le
1825extern int comp_short_keys (const struct reiserfs_key * le_key,
1826 const struct cpu_key * cpu_key);
1827extern void le_key2cpu_key (struct cpu_key * to, const struct reiserfs_key * from);
1828
1829// both are in le form
1830extern int comp_le_keys (const struct reiserfs_key *, const struct reiserfs_key *);
1831extern int comp_short_le_keys (const struct reiserfs_key *, const struct reiserfs_key *);
1832
1833//
1834// get key version from on disk key - kludge
1835//
1836static inline int le_key_version (const struct reiserfs_key * key)
1837{
1838 int type;
1839
1840 type = offset_v2_k_type( &(key->u.k_offset_v2));
1841 if (type != TYPE_DIRECT && type != TYPE_INDIRECT && type != TYPE_DIRENTRY)
1842 return KEY_FORMAT_3_5;
1843
1844 return KEY_FORMAT_3_6;
1845
1846}
1847
1848
1849static inline void copy_key (struct reiserfs_key *to, const struct reiserfs_key *from)
1850{
1851 memcpy (to, from, KEY_SIZE);
1852}
1853
1854
1855int comp_items (const struct item_head * stored_ih, const struct path * p_s_path);
1856const struct reiserfs_key * get_rkey (const struct path * p_s_chk_path,
1857 const struct super_block * p_s_sb);
1858int search_by_key (struct super_block *, const struct cpu_key *,
1859 struct path *, int);
1860#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
1861int search_for_position_by_key (struct super_block * p_s_sb,
1862 const struct cpu_key * p_s_cpu_key,
1863 struct path * p_s_search_path);
1864extern void decrement_bcount (struct buffer_head * p_s_bh);
1865void decrement_counters_in_path (struct path * p_s_search_path);
1866void pathrelse (struct path * p_s_search_path);
1867int reiserfs_check_path(struct path *p) ;
1868void pathrelse_and_restore (struct super_block *s, struct path * p_s_search_path);
1869
1870int reiserfs_insert_item (struct reiserfs_transaction_handle *th,
1871 struct path * path,
1872 const struct cpu_key * key,
1873 struct item_head * ih,
1874 struct inode *inode, const char * body);
1875
1876int reiserfs_paste_into_item (struct reiserfs_transaction_handle *th,
1877 struct path * path,
1878 const struct cpu_key * key,
1879 struct inode *inode,
1880 const char * body, int paste_size);
1881
1882int reiserfs_cut_from_item (struct reiserfs_transaction_handle *th,
1883 struct path * path,
1884 struct cpu_key * key,
1885 struct inode * inode,
1886 struct page *page,
1887 loff_t new_file_size);
1888
1889int reiserfs_delete_item (struct reiserfs_transaction_handle *th,
1890 struct path * path,
1891 const struct cpu_key * key,
1892 struct inode * inode,
1893 struct buffer_head * p_s_un_bh);
1894
1895void reiserfs_delete_solid_item (struct reiserfs_transaction_handle *th,
1896 struct inode *inode, struct reiserfs_key * key);
1897int reiserfs_delete_object (struct reiserfs_transaction_handle *th, struct inode * p_s_inode);
1898int reiserfs_do_truncate (struct reiserfs_transaction_handle *th,
1899 struct inode * p_s_inode, struct page *,
1900 int update_timestamps);
1901
1902#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1903#define file_size(inode) ((inode)->i_size)
1904#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1905
1906#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1907!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 )
1908
1909void padd_item (char * item, int total_length, int length);
1910
1911/* inode.c */
1912/* args for the create parameter of reiserfs_get_block */
1913#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1914#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1915#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1916#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
1917#define GET_BLOCK_NO_ISEM 8 /* i_sem is not held, don't preallocate */
1918#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
1919
1920int restart_transaction(struct reiserfs_transaction_handle *th, struct inode *inode, struct path *path);
1921void reiserfs_read_locked_inode(struct inode * inode, struct reiserfs_iget_args *args) ;
1922int reiserfs_find_actor(struct inode * inode, void *p) ;
1923int reiserfs_init_locked_inode(struct inode * inode, void *p) ;
1924void reiserfs_delete_inode (struct inode * inode);
1925int reiserfs_write_inode (struct inode * inode, int) ;
1926int reiserfs_get_block (struct inode * inode, sector_t block, struct buffer_head * bh_result, int create);
1927struct dentry *reiserfs_get_dentry(struct super_block *, void *) ;
1928struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 *data,
1929 int len, int fhtype,
1930 int (*acceptable)(void *contect, struct dentry *de),
1931 void *context) ;
1932int reiserfs_encode_fh( struct dentry *dentry, __u32 *data, int *lenp,
1933 int connectable );
1934
1935int reiserfs_truncate_file(struct inode *, int update_timestamps) ;
1936void make_cpu_key (struct cpu_key * cpu_key, struct inode * inode, loff_t offset,
1937 int type, int key_length);
1938void make_le_item_head (struct item_head * ih, const struct cpu_key * key,
1939 int version,
1940 loff_t offset, int type, int length, int entry_count);
1941struct inode * reiserfs_iget (struct super_block * s,
1942 const struct cpu_key * key);
1943
1944
1945int reiserfs_new_inode (struct reiserfs_transaction_handle *th,
1946 struct inode * dir, int mode,
1947 const char * symname, loff_t i_size,
1948 struct dentry *dentry, struct inode *inode);
1949
1950void reiserfs_update_sd_size (struct reiserfs_transaction_handle *th,
1951 struct inode * inode, loff_t size);
1952
1953static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
1954 struct inode *inode)
1955{
1956 reiserfs_update_sd_size(th, inode, inode->i_size) ;
1957}
1958
1959void sd_attrs_to_i_attrs( __u16 sd_attrs, struct inode *inode );
1960void i_attrs_to_sd_attrs( struct inode *inode, __u16 *sd_attrs );
1961int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1962
1963/* namei.c */
1964void set_de_name_and_namelen (struct reiserfs_dir_entry * de);
1965int search_by_entry_key (struct super_block * sb, const struct cpu_key * key,
1966 struct path * path,
1967 struct reiserfs_dir_entry * de);
1968struct dentry *reiserfs_get_parent(struct dentry *) ;
1969/* procfs.c */
1970
1971#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1972#define REISERFS_PROC_INFO
1973#else
1974#undef REISERFS_PROC_INFO
1975#endif
1976
1977int reiserfs_proc_info_init( struct super_block *sb );
1978int reiserfs_proc_info_done( struct super_block *sb );
1979struct proc_dir_entry *reiserfs_proc_register_global( char *name,
1980 read_proc_t *func );
1981void reiserfs_proc_unregister_global( const char *name );
1982int reiserfs_proc_info_global_init( void );
1983int reiserfs_proc_info_global_done( void );
1984int reiserfs_global_version_in_proc( char *buffer, char **start, off_t offset,
1985 int count, int *eof, void *data );
1986
1987#if defined( REISERFS_PROC_INFO )
1988
1989#define PROC_EXP( e ) e
1990
1991#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1992#define PROC_INFO_MAX( sb, field, value ) \
1993 __PINFO( sb ).field = \
1994 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1995#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1996#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1997#define PROC_INFO_BH_STAT( sb, bh, level ) \
1998 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1999 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
2000 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
2001#else
2002#define PROC_EXP( e )
2003#define VOID_V ( ( void ) 0 )
2004#define PROC_INFO_MAX( sb, field, value ) VOID_V
2005#define PROC_INFO_INC( sb, field ) VOID_V
2006#define PROC_INFO_ADD( sb, field, val ) VOID_V
2007#define PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level ) VOID_V
2008#endif
2009
2010/* dir.c */
2011extern struct inode_operations reiserfs_dir_inode_operations;
2012extern struct inode_operations reiserfs_symlink_inode_operations;
2013extern struct inode_operations reiserfs_special_inode_operations;
2014extern struct file_operations reiserfs_dir_operations;
2015
2016/* tail_conversion.c */
2017int direct2indirect (struct reiserfs_transaction_handle *, struct inode *, struct path *, struct buffer_head *, loff_t);
2018int indirect2direct (struct reiserfs_transaction_handle *, struct inode *, struct page *, struct path *, const struct cpu_key *, loff_t, char *);
2019void reiserfs_unmap_buffer(struct buffer_head *) ;
2020
2021
2022/* file.c */
2023extern struct inode_operations reiserfs_file_inode_operations;
2024extern struct file_operations reiserfs_file_operations;
2025extern struct address_space_operations reiserfs_address_space_operations ;
2026
2027/* fix_nodes.c */
2028#ifdef CONFIG_REISERFS_CHECK
2029void * reiserfs_kmalloc (size_t size, int flags, struct super_block * s);
2030void reiserfs_kfree (const void * vp, size_t size, struct super_block * s);
2031#else
2032static inline void *reiserfs_kmalloc(size_t size, int flags,
2033 struct super_block *s)
2034{
2035 return kmalloc(size, flags);
2036}
2037
2038static inline void reiserfs_kfree(const void *vp, size_t size,
2039 struct super_block *s)
2040{
2041 kfree(vp);
2042}
2043#endif
2044
2045int fix_nodes (int n_op_mode, struct tree_balance * p_s_tb,
2046 struct item_head * p_s_ins_ih, const void *);
2047void unfix_nodes (struct tree_balance *);
2048
2049
2050/* prints.c */
2051void reiserfs_panic (struct super_block * s, const char * fmt, ...) __attribute__ ( ( noreturn ) );
2052void reiserfs_info (struct super_block *s, const char * fmt, ...);
2053void reiserfs_debug (struct super_block *s, int level, const char * fmt, ...);
2054void print_indirect_item (struct buffer_head * bh, int item_num);
2055void store_print_tb (struct tree_balance * tb);
2056void print_cur_tb (char * mes);
2057void print_de (struct reiserfs_dir_entry * de);
2058void print_bi (struct buffer_info * bi, char * mes);
2059#define PRINT_LEAF_ITEMS 1 /* print all items */
2060#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2061#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2062void print_block (struct buffer_head * bh, ...);
2063void print_bmap (struct super_block * s, int silent);
2064void print_bmap_block (int i, char * data, int size, int silent);
2065/*void print_super_block (struct super_block * s, char * mes);*/
2066void print_objectid_map (struct super_block * s);
2067void print_block_head (struct buffer_head * bh, char * mes);
2068void check_leaf (struct buffer_head * bh);
2069void check_internal (struct buffer_head * bh);
2070void print_statistics (struct super_block * s);
2071char * reiserfs_hashname(int code);
2072
2073/* lbalance.c */
2074int leaf_move_items (int shift_mode, struct tree_balance * tb, int mov_num, int mov_bytes, struct buffer_head * Snew);
2075int leaf_shift_left (struct tree_balance * tb, int shift_num, int shift_bytes);
2076int leaf_shift_right (struct tree_balance * tb, int shift_num, int shift_bytes);
2077void leaf_delete_items (struct buffer_info * cur_bi, int last_first, int first, int del_num, int del_bytes);
2078void leaf_insert_into_buf (struct buffer_info * bi, int before,
2079 struct item_head * inserted_item_ih, const char * inserted_item_body, int zeros_number);
2080void leaf_paste_in_buffer (struct buffer_info * bi, int pasted_item_num,
2081 int pos_in_item, int paste_size, const char * body, int zeros_number);
2082void leaf_cut_from_buffer (struct buffer_info * bi, int cut_item_num, int pos_in_item,
2083 int cut_size);
2084void leaf_paste_entries (struct buffer_head * bh, int item_num, int before,
2085 int new_entry_count, struct reiserfs_de_head * new_dehs, const char * records, int paste_size);
2086/* ibalance.c */
2087int balance_internal (struct tree_balance * , int, int, struct item_head * ,
2088 struct buffer_head **);
2089
2090/* do_balance.c */
2091void do_balance_mark_leaf_dirty (struct tree_balance * tb,
2092 struct buffer_head * bh, int flag);
2093#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2094#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2095
2096void do_balance (struct tree_balance * tb, struct item_head * ih,
2097 const char * body, int flag);
2098void reiserfs_invalidate_buffer (struct tree_balance * tb, struct buffer_head * bh);
2099
2100int get_left_neighbor_position (struct tree_balance * tb, int h);
2101int get_right_neighbor_position (struct tree_balance * tb, int h);
2102void replace_key (struct tree_balance * tb, struct buffer_head *, int, struct buffer_head *, int);
2103void make_empty_node (struct buffer_info *);
2104struct buffer_head * get_FEB (struct tree_balance *);
2105
2106/* bitmap.c */
2107
2108/* structure contains hints for block allocator, and it is a container for
2109 * arguments, such as node, search path, transaction_handle, etc. */
2110 struct __reiserfs_blocknr_hint {
2111 struct inode * inode; /* inode passed to allocator, if we allocate unf. nodes */
2112 long block; /* file offset, in blocks */
Al Viro6a3a16f2005-05-01 08:59:17 -07002113 struct in_core_key key;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 struct path * path; /* search path, used by allocator to deternine search_start by
2115 * various ways */
2116 struct reiserfs_transaction_handle * th; /* transaction handle is needed to log super blocks and
2117 * bitmap blocks changes */
2118 b_blocknr_t beg, end;
2119 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
2120 * between different block allocator procedures
2121 * (determine_search_start() and others) */
2122 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2123 * function that do actual allocation */
2124
2125 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
2126 * formatted/unformatted blocks with/without preallocation */
2127 unsigned preallocate:1;
2128};
2129
2130typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2131
2132int reiserfs_parse_alloc_options (struct super_block *, char *);
2133void reiserfs_init_alloc_options (struct super_block *s);
2134
2135/*
2136 * given a directory, this will tell you what packing locality
2137 * to use for a new object underneat it. The locality is returned
2138 * in disk byte order (le).
2139 */
Al Viro3e8962b2005-05-01 08:59:18 -07002140__le32 reiserfs_choose_packing(struct inode *dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141
2142int is_reusable (struct super_block * s, b_blocknr_t block, int bit_value);
2143void reiserfs_free_block (struct reiserfs_transaction_handle *th, struct inode *, b_blocknr_t, int for_unformatted);
2144int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t * , int, int);
2145extern inline int reiserfs_new_form_blocknrs (struct tree_balance * tb,
2146 b_blocknr_t *new_blocknrs, int amount_needed)
2147{
2148 reiserfs_blocknr_hint_t hint = {
2149 .th = tb->transaction_handle,
2150 .path = tb->tb_path,
2151 .inode = NULL,
2152 .key = tb->key,
2153 .block = 0,
2154 .formatted_node = 1
2155 };
2156 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed, 0);
2157}
2158
2159extern inline int reiserfs_new_unf_blocknrs (struct reiserfs_transaction_handle *th,
2160 struct inode *inode,
2161 b_blocknr_t *new_blocknrs,
2162 struct path * path, long block)
2163{
2164 reiserfs_blocknr_hint_t hint = {
2165 .th = th,
2166 .path = path,
2167 .inode = inode,
2168 .block = block,
2169 .formatted_node = 0,
2170 .preallocate = 0
2171 };
2172 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2173}
2174
2175#ifdef REISERFS_PREALLOCATE
2176extern inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle *th,
2177 struct inode * inode,
2178 b_blocknr_t *new_blocknrs,
2179 struct path * path, long block)
2180{
2181 reiserfs_blocknr_hint_t hint = {
2182 .th = th,
2183 .path = path,
2184 .inode = inode,
2185 .block = block,
2186 .formatted_node = 0,
2187 .preallocate = 1
2188 };
2189 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
2190}
2191
2192void reiserfs_discard_prealloc (struct reiserfs_transaction_handle *th,
2193 struct inode * inode);
2194void reiserfs_discard_all_prealloc (struct reiserfs_transaction_handle *th);
2195#endif
2196void reiserfs_claim_blocks_to_be_allocated( struct super_block *sb, int blocks);
2197void reiserfs_release_claimed_blocks( struct super_block *sb, int blocks);
2198int reiserfs_can_fit_pages(struct super_block *sb);
2199
2200/* hashes.c */
2201__u32 keyed_hash (const signed char *msg, int len);
2202__u32 yura_hash (const signed char *msg, int len);
2203__u32 r5_hash (const signed char *msg, int len);
2204
2205/* the ext2 bit routines adjust for big or little endian as
2206** appropriate for the arch, so in our laziness we use them rather
2207** than using the bit routines they call more directly. These
2208** routines must be used when changing on disk bitmaps. */
2209#define reiserfs_test_and_set_le_bit ext2_set_bit
2210#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2211#define reiserfs_test_le_bit ext2_test_bit
2212#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2213
2214/* sometimes reiserfs_truncate may require to allocate few new blocks
2215 to perform indirect2direct conversion. People probably used to
2216 think, that truncate should work without problems on a filesystem
2217 without free disk space. They may complain that they can not
2218 truncate due to lack of free disk space. This spare space allows us
2219 to not worry about it. 500 is probably too much, but it should be
2220 absolutely safe */
2221#define SPARE_SPACE 500
2222
2223
2224/* prototypes from ioctl.c */
2225int reiserfs_ioctl (struct inode * inode, struct file * filp,
2226 unsigned int cmd, unsigned long arg);
2227
2228/* ioctl's command */
2229#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
2230/* define following flags to be the same as in ext2, so that chattr(1),
2231 lsattr(1) will work with us. */
2232#define REISERFS_IOC_GETFLAGS EXT2_IOC_GETFLAGS
2233#define REISERFS_IOC_SETFLAGS EXT2_IOC_SETFLAGS
2234#define REISERFS_IOC_GETVERSION EXT2_IOC_GETVERSION
2235#define REISERFS_IOC_SETVERSION EXT2_IOC_SETVERSION
2236
2237/* Locking primitives */
2238/* Right now we are still falling back to (un)lock_kernel, but eventually that
2239 would evolve into real per-fs locks */
2240#define reiserfs_write_lock( sb ) lock_kernel()
2241#define reiserfs_write_unlock( sb ) unlock_kernel()
2242
2243/* xattr stuff */
2244#define REISERFS_XATTR_DIR_SEM(s) (REISERFS_SB(s)->xattr_dir_sem)
2245
2246#endif /* _LINUX_REISER_FS_H */
2247
2248