Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
[deliverable/linux.git] / include / linux / reiserfs_fs.h
CommitLineData
1da177e4
LT
1/*
2 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details
3 */
4
5 /* this file has an amazingly stupid
bd4c625c
LT
6 name, yura please fix it to be
7 reiserfs.h, and merge all the rest
8 of our .h files that are in this
9 directory into it. */
1da177e4
LT
10
11#ifndef _LINUX_REISER_FS_H
12#define _LINUX_REISER_FS_H
13
14#include <linux/types.h>
e18fa700
JG
15#include <linux/magic.h>
16
1da177e4
LT
17#ifdef __KERNEL__
18#include <linux/slab.h>
19#include <linux/interrupt.h>
20#include <linux/sched.h>
21#include <linux/workqueue.h>
22#include <asm/unaligned.h>
23#include <linux/bitops.h>
24#include <linux/proc_fs.h>
25#include <linux/smp_lock.h>
26#include <linux/buffer_head.h>
27#include <linux/reiserfs_fs_i.h>
28#include <linux/reiserfs_fs_sb.h>
29#endif
30
31/*
32 * include/linux/reiser_fs.h
33 *
34 * Reiser File System constants and structures
35 *
36 */
37
750e1c18
JSR
38/* ioctl's command */
39#define REISERFS_IOC_UNPACK _IOW(0xCD,1,long)
40/* define following flags to be the same as in ext2, so that chattr(1),
41 lsattr(1) will work with us. */
42#define REISERFS_IOC_GETFLAGS FS_IOC_GETFLAGS
43#define REISERFS_IOC_SETFLAGS FS_IOC_SETFLAGS
44#define REISERFS_IOC_GETVERSION FS_IOC_GETVERSION
45#define REISERFS_IOC_SETVERSION FS_IOC_SETVERSION
46
47#ifdef __KERNEL__
48/* the 32 bit compat definitions with int argument */
49#define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int)
50#define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS
51#define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS
52#define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION
53#define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION
54
55/* Locking primitives */
56/* Right now we are still falling back to (un)lock_kernel, but eventually that
57 would evolve into real per-fs locks */
58#define reiserfs_write_lock( sb ) lock_kernel()
59#define reiserfs_write_unlock( sb ) unlock_kernel()
60
750e1c18
JSR
61struct fid;
62
1da177e4
LT
63/* in reading the #defines, it may help to understand that they employ
64 the following abbreviations:
65
66 B = Buffer
67 I = Item header
68 H = Height within the tree (should be changed to LEV)
69 N = Number of the item in the node
70 STAT = stat data
71 DEH = Directory Entry Header
72 EC = Entry Count
73 E = Entry number
74 UL = Unsigned Long
75 BLKH = BLocK Header
76 UNFM = UNForMatted node
77 DC = Disk Child
78 P = Path
79
80 These #defines are named by concatenating these abbreviations,
81 where first comes the arguments, and last comes the return value,
82 of the macro.
83
84*/
85
86#define USE_INODE_GENERATION_COUNTER
87
88#define REISERFS_PREALLOCATE
89#define DISPLACE_NEW_PACKING_LOCALITIES
90#define PREALLOCATION_SIZE 9
91
92/* n must be power of 2 */
93#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))
94
95// to be ok for alpha and others we have to align structures to 8 byte
96// boundary.
97// FIXME: do not change 4 by anything else: there is code which relies on that
98#define ROUND_UP(x) _ROUND_UP(x,8LL)
99
100/* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug
101** messages.
102*/
bd4c625c 103#define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */
1da177e4 104
45b03d5e
JM
105void __reiserfs_warning(struct super_block *s, const char *id,
106 const char *func, const char *fmt, ...);
107#define reiserfs_warning(s, id, fmt, args...) \
108 __reiserfs_warning(s, id, __func__, fmt, ##args)
1da177e4
LT
109/* assertions handling */
110
111/** always check a condition and panic if it's false. */
c3a9c210
JM
112#define __RASSERT(cond, scond, format, args...) \
113do { \
114 if (!(cond)) \
115 reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \
116 __FILE__ ":%i:%s: " format "\n", \
117 in_interrupt() ? -1 : task_pid_nr(current), \
118 __LINE__, __func__ , ##args); \
119} while (0)
1da177e4 120
2d954d06
AV
121#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)
122
1da177e4 123#if defined( CONFIG_REISERFS_CHECK )
2d954d06 124#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)
1da177e4
LT
125#else
126#define RFALSE( cond, format, args... ) do {;} while( 0 )
127#endif
128
129#define CONSTF __attribute_const__
130/*
131 * Disk Data Structures
132 */
133
134/***************************************************************************/
135/* SUPER BLOCK */
136/***************************************************************************/
137
138/*
139 * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs
140 * the version in RAM is part of a larger structure containing fields never written to disk.
141 */
bd4c625c
LT
142#define UNSET_HASH 0 // read_super will guess about, what hash names
143 // in directories were sorted with
1da177e4
LT
144#define TEA_HASH 1
145#define YURA_HASH 2
146#define R5_HASH 3
147#define DEFAULT_HASH R5_HASH
148
1da177e4 149struct journal_params {
bd4c625c
LT
150 __le32 jp_journal_1st_block; /* where does journal start from on its
151 * device */
152 __le32 jp_journal_dev; /* journal device st_rdev */
153 __le32 jp_journal_size; /* size of the journal */
154 __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */
155 __le32 jp_journal_magic; /* random value made on fs creation (this
156 * was sb_journal_block_count) */
157 __le32 jp_journal_max_batch; /* max number of blocks to batch into a
158 * trans */
159 __le32 jp_journal_max_commit_age; /* in seconds, how old can an async
160 * commit be */
161 __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction
162 * be */
1da177e4
LT
163};
164
165/* this is the super from 3.5.X, where X >= 10 */
bd4c625c
LT
166struct reiserfs_super_block_v1 {
167 __le32 s_block_count; /* blocks count */
168 __le32 s_free_blocks; /* free blocks count */
169 __le32 s_root_block; /* root block number */
170 struct journal_params s_journal;
171 __le16 s_blocksize; /* block size */
172 __le16 s_oid_maxsize; /* max size of object id array, see
173 * get_objectid() commentary */
174 __le16 s_oid_cursize; /* current size of object id array */
175 __le16 s_umount_state; /* this is set to 1 when filesystem was
176 * umounted, to 2 - when not */
177 char s_magic[10]; /* reiserfs magic string indicates that
178 * file system is reiserfs:
179 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */
180 __le16 s_fs_state; /* it is set to used by fsck to mark which
181 * phase of rebuilding is done */
182 __le32 s_hash_function_code; /* indicate, what hash function is being use
183 * to sort names in a directory*/
184 __le16 s_tree_height; /* height of disk tree */
185 __le16 s_bmap_nr; /* amount of bitmap blocks needed to address
186 * each block of file system */
187 __le16 s_version; /* this field is only reliable on filesystem
188 * with non-standard journal */
189 __le16 s_reserved_for_journal; /* size in blocks of journal area on main
190 * device, we need to keep after
191 * making fs with non-standard journal */
1da177e4
LT
192} __attribute__ ((__packed__));
193
194#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))
195
196/* this is the on disk super block */
bd4c625c
LT
197struct reiserfs_super_block {
198 struct reiserfs_super_block_v1 s_v1;
199 __le32 s_inode_generation;
200 __le32 s_flags; /* Right now used only by inode-attributes, if enabled */
201 unsigned char s_uuid[16]; /* filesystem unique identifier */
202 unsigned char s_label[16]; /* filesystem volume label */
702d21c6
JM
203 __le16 s_mnt_count; /* Count of mounts since last fsck */
204 __le16 s_max_mnt_count; /* Maximum mounts before check */
205 __le32 s_lastcheck; /* Timestamp of last fsck */
206 __le32 s_check_interval; /* Interval between checks */
207 char s_unused[76]; /* zero filled by mkreiserfs and
bd4c625c
LT
208 * reiserfs_convert_objectid_map_v1()
209 * so any additions must be updated
210 * there as well. */
211} __attribute__ ((__packed__));
1da177e4
LT
212
213#define SB_SIZE (sizeof(struct reiserfs_super_block))
214
215#define REISERFS_VERSION_1 0
216#define REISERFS_VERSION_2 2
217
1da177e4
LT
218// on-disk super block fields converted to cpu form
219#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)
220#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))
221#define SB_BLOCKSIZE(s) \
222 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))
223#define SB_BLOCK_COUNT(s) \
224 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))
225#define SB_FREE_BLOCKS(s) \
226 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))
227#define SB_REISERFS_MAGIC(s) \
228 (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)
229#define SB_ROOT_BLOCK(s) \
230 le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))
231#define SB_TREE_HEIGHT(s) \
232 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))
233#define SB_REISERFS_STATE(s) \
234 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))
235#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))
236#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))
237
238#define PUT_SB_BLOCK_COUNT(s, val) \
239 do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)
240#define PUT_SB_FREE_BLOCKS(s, val) \
241 do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)
242#define PUT_SB_ROOT_BLOCK(s, val) \
243 do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)
244#define PUT_SB_TREE_HEIGHT(s, val) \
245 do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)
246#define PUT_SB_REISERFS_STATE(s, val) \
bd4c625c 247 do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)
1da177e4
LT
248#define PUT_SB_VERSION(s, val) \
249 do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)
250#define PUT_SB_BMAP_NR(s, val) \
251 do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)
252
1da177e4
LT
253#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)
254#define SB_ONDISK_JOURNAL_SIZE(s) \
255 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))
256#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \
257 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))
258#define SB_ONDISK_JOURNAL_DEVICE(s) \
259 le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
260#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
b8cc936f 261 le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
1da177e4
LT
262
263#define is_block_in_log_or_reserved_area(s, block) \
264 block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
265 && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \
266 ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \
bd4c625c 267 SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))
1da177e4 268
bd4c625c
LT
269int is_reiserfs_3_5(struct reiserfs_super_block *rs);
270int is_reiserfs_3_6(struct reiserfs_super_block *rs);
271int is_reiserfs_jr(struct reiserfs_super_block *rs);
1da177e4
LT
272
273/* ReiserFS leaves the first 64k unused, so that partition labels have
274 enough space. If someone wants to write a fancy bootloader that
275 needs more than 64k, let us know, and this will be increased in size.
276 This number must be larger than than the largest block size on any
277 platform, or code will break. -Hans */
278#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)
279#define REISERFS_FIRST_BLOCK unused_define
280#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES
281
282/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */
283#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)
284
285// reiserfs internal error code (used by search_by_key adn fix_nodes))
286#define CARRY_ON 0
287#define REPEAT_SEARCH -1
288#define IO_ERROR -2
289#define NO_DISK_SPACE -3
290#define NO_BALANCING_NEEDED (-4)
291#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)
292#define QUOTA_EXCEEDED -6
293
294typedef __u32 b_blocknr_t;
3e8962be 295typedef __le32 unp_t;
1da177e4
LT
296
297struct unfm_nodeinfo {
bd4c625c
LT
298 unp_t unfm_nodenum;
299 unsigned short unfm_freespace;
1da177e4
LT
300};
301
302/* there are two formats of keys: 3.5 and 3.6
303 */
304#define KEY_FORMAT_3_5 0
305#define KEY_FORMAT_3_6 1
306
307/* there are two stat datas */
308#define STAT_DATA_V1 0
309#define STAT_DATA_V2 1
310
1da177e4
LT
311static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)
312{
313 return container_of(inode, struct reiserfs_inode_info, vfs_inode);
314}
315
316static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)
317{
318 return sb->s_fs_info;
319}
320
cb680c1b
JM
321/* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16
322 * which overflows on large file systems. */
13d8bcd2 323static inline __u32 reiserfs_bmap_count(struct super_block *sb)
cb680c1b
JM
324{
325 return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;
326}
327
328static inline int bmap_would_wrap(unsigned bmap_nr)
329{
330 return bmap_nr > ((1LL << 16) - 1);
331}
332
1da177e4
LT
333/** this says about version of key of all items (but stat data) the
334 object consists of */
335#define get_inode_item_key_version( inode ) \
336 ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)
337
338#define set_inode_item_key_version( inode, version ) \
339 ({ if((version)==KEY_FORMAT_3_6) \
340 REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \
341 else \
342 REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })
343
344#define get_inode_sd_version(inode) \
345 ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)
346
347#define set_inode_sd_version(inode, version) \
348 ({ if((version)==STAT_DATA_V2) \
349 REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \
350 else \
351 REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })
352
353/* This is an aggressive tail suppression policy, I am hoping it
354 improves our benchmarks. The principle behind it is that percentage
355 space saving is what matters, not absolute space saving. This is
356 non-intuitive, but it helps to understand it if you consider that the
357 cost to access 4 blocks is not much more than the cost to access 1
358 block, if you have to do a seek and rotate. A tail risks a
359 non-linear disk access that is significant as a percentage of total
360 time cost for a 4 block file and saves an amount of space that is
361 less significant as a percentage of space, or so goes the hypothesis.
362 -Hans */
363#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \
364(\
365 (!(n_tail_size)) || \
366 (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \
367 ( (n_file_size) >= (n_block_size) * 4 ) || \
368 ( ( (n_file_size) >= (n_block_size) * 3 ) && \
369 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \
370 ( ( (n_file_size) >= (n_block_size) * 2 ) && \
371 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \
372 ( ( (n_file_size) >= (n_block_size) ) && \
373 ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \
374)
375
376/* Another strategy for tails, this one means only create a tail if all the
377 file would fit into one DIRECT item.
378 Primary intention for this one is to increase performance by decreasing
379 seeking.
bd4c625c 380*/
1da177e4
LT
381#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \
382(\
383 (!(n_tail_size)) || \
384 (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \
385)
386
1da177e4
LT
387/*
388 * values for s_umount_state field
389 */
390#define REISERFS_VALID_FS 1
391#define REISERFS_ERROR_FS 2
392
393//
394// there are 5 item types currently
395//
396#define TYPE_STAT_DATA 0
397#define TYPE_INDIRECT 1
398#define TYPE_DIRECT 2
bd4c625c
LT
399#define TYPE_DIRENTRY 3
400#define TYPE_MAXTYPE 3
401#define TYPE_ANY 15 // FIXME: comment is required
1da177e4
LT
402
403/***************************************************************************/
404/* KEY & ITEM HEAD */
405/***************************************************************************/
406
407//
408// directories use this key as well as old files
409//
410struct offset_v1 {
bd4c625c
LT
411 __le32 k_offset;
412 __le32 k_uniqueness;
1da177e4
LT
413} __attribute__ ((__packed__));
414
415struct offset_v2 {
f8e08a84 416 __le64 v;
1da177e4
LT
417} __attribute__ ((__packed__));
418
bd4c625c 419static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)
1da177e4 420{
f8e08a84 421 __u8 type = le64_to_cpu(v2->v) >> 60;
bd4c625c 422 return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;
1da177e4 423}
bd4c625c
LT
424
425static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)
1da177e4 426{
bd4c625c
LT
427 v2->v =
428 (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);
1da177e4 429}
bd4c625c
LT
430
431static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)
1da177e4 432{
bd4c625c 433 return le64_to_cpu(v2->v) & (~0ULL >> 4);
1da177e4
LT
434}
435
bd4c625c
LT
436static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)
437{
438 offset &= (~0ULL >> 4);
439 v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);
1da177e4 440}
1da177e4
LT
441
442/* Key of an item determines its location in the S+tree, and
443 is composed of 4 components */
444struct reiserfs_key {
bd4c625c
LT
445 __le32 k_dir_id; /* packing locality: by default parent
446 directory object id */
447 __le32 k_objectid; /* object identifier */
448 union {
449 struct offset_v1 k_offset_v1;
450 struct offset_v2 k_offset_v2;
451 } __attribute__ ((__packed__)) u;
1da177e4
LT
452} __attribute__ ((__packed__));
453
6a3a16f2 454struct in_core_key {
bd4c625c
LT
455 __u32 k_dir_id; /* packing locality: by default parent
456 directory object id */
457 __u32 k_objectid; /* object identifier */
458 __u64 k_offset;
459 __u8 k_type;
6b9f5829 460};
1da177e4
LT
461
462struct cpu_key {
bd4c625c
LT
463 struct in_core_key on_disk_key;
464 int version;
465 int key_length; /* 3 in all cases but direct2indirect and
466 indirect2direct conversion */
1da177e4
LT
467};
468
469/* Our function for comparing keys can compare keys of different
470 lengths. It takes as a parameter the length of the keys it is to
471 compare. These defines are used in determining what is to be passed
472 to it as that parameter. */
473#define REISERFS_FULL_KEY_LEN 4
474#define REISERFS_SHORT_KEY_LEN 2
475
476/* The result of the key compare */
477#define FIRST_GREATER 1
478#define SECOND_GREATER -1
479#define KEYS_IDENTICAL 0
480#define KEY_FOUND 1
481#define KEY_NOT_FOUND 0
482
483#define KEY_SIZE (sizeof(struct reiserfs_key))
484#define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32))
485
486/* return values for search_by_key and clones */
487#define ITEM_FOUND 1
488#define ITEM_NOT_FOUND 0
489#define ENTRY_FOUND 1
490#define ENTRY_NOT_FOUND 0
491#define DIRECTORY_NOT_FOUND -1
492#define REGULAR_FILE_FOUND -2
493#define DIRECTORY_FOUND -3
494#define BYTE_FOUND 1
495#define BYTE_NOT_FOUND 0
496#define FILE_NOT_FOUND -1
497
498#define POSITION_FOUND 1
499#define POSITION_NOT_FOUND 0
500
501// return values for reiserfs_find_entry and search_by_entry_key
502#define NAME_FOUND 1
503#define NAME_NOT_FOUND 0
504#define GOTO_PREVIOUS_ITEM 2
505#define NAME_FOUND_INVISIBLE 3
506
507/* Everything in the filesystem is stored as a set of items. The
508 item head contains the key of the item, its free space (for
509 indirect items) and specifies the location of the item itself
510 within the block. */
511
bd4c625c 512struct item_head {
1da177e4
LT
513 /* Everything in the tree is found by searching for it based on
514 * its key.*/
515 struct reiserfs_key ih_key;
516 union {
517 /* The free space in the last unformatted node of an
518 indirect item if this is an indirect item. This
519 equals 0xFFFF iff this is a direct item or stat data
520 item. Note that the key, not this field, is used to
521 determine the item type, and thus which field this
522 union contains. */
3e8962be 523 __le16 ih_free_space_reserved;
1da177e4
LT
524 /* Iff this is a directory item, this field equals the
525 number of directory entries in the directory item. */
3e8962be 526 __le16 ih_entry_count;
1da177e4 527 } __attribute__ ((__packed__)) u;
bd4c625c
LT
528 __le16 ih_item_len; /* total size of the item body */
529 __le16 ih_item_location; /* an offset to the item body
530 * within the block */
531 __le16 ih_version; /* 0 for all old items, 2 for new
532 ones. Highest bit is set by fsck
533 temporary, cleaned after all
534 done */
1da177e4
LT
535} __attribute__ ((__packed__));
536/* size of item header */
537#define IH_SIZE (sizeof(struct item_head))
538
539#define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved)
540#define ih_version(ih) le16_to_cpu((ih)->ih_version)
541#define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count)
542#define ih_location(ih) le16_to_cpu((ih)->ih_item_location)
543#define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len)
544
545#define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)
546#define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0)
547#define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)
548#define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)
549#define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)
550
1da177e4
LT
551#define unreachable_item(ih) (ih_version(ih) & (1 << 15))
552
553#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))
554#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))
555
556/* these operate on indirect items, where you've got an array of ints
557** at a possibly unaligned location. These are a noop on ia32
558**
559** p is the array of __u32, i is the index into the array, v is the value
560** to store there.
561*/
8b5ac31e
HH
562#define get_block_num(p, i) get_unaligned_le32((p) + (i))
563#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))
1da177e4
LT
564
565//
566// in old version uniqueness field shows key type
567//
568#define V1_SD_UNIQUENESS 0
569#define V1_INDIRECT_UNIQUENESS 0xfffffffe
570#define V1_DIRECT_UNIQUENESS 0xffffffff
571#define V1_DIRENTRY_UNIQUENESS 500
bd4c625c 572#define V1_ANY_UNIQUENESS 555 // FIXME: comment is required
1da177e4
LT
573
574//
575// here are conversion routines
576//
bd4c625c
LT
577static inline int uniqueness2type(__u32 uniqueness) CONSTF;
578static inline int uniqueness2type(__u32 uniqueness)
1da177e4 579{
bd4c625c
LT
580 switch ((int)uniqueness) {
581 case V1_SD_UNIQUENESS:
582 return TYPE_STAT_DATA;
583 case V1_INDIRECT_UNIQUENESS:
584 return TYPE_INDIRECT;
585 case V1_DIRECT_UNIQUENESS:
586 return TYPE_DIRECT;
587 case V1_DIRENTRY_UNIQUENESS:
588 return TYPE_DIRENTRY;
1da177e4 589 case V1_ANY_UNIQUENESS:
fd7cb031 590 default:
bd4c625c
LT
591 return TYPE_ANY;
592 }
1da177e4
LT
593}
594
bd4c625c
LT
595static inline __u32 type2uniqueness(int type) CONSTF;
596static inline __u32 type2uniqueness(int type)
1da177e4 597{
bd4c625c
LT
598 switch (type) {
599 case TYPE_STAT_DATA:
600 return V1_SD_UNIQUENESS;
601 case TYPE_INDIRECT:
602 return V1_INDIRECT_UNIQUENESS;
603 case TYPE_DIRECT:
604 return V1_DIRECT_UNIQUENESS;
605 case TYPE_DIRENTRY:
606 return V1_DIRENTRY_UNIQUENESS;
1da177e4 607 case TYPE_ANY:
fd7cb031 608 default:
bd4c625c
LT
609 return V1_ANY_UNIQUENESS;
610 }
1da177e4
LT
611}
612
613//
614// key is pointer to on disk key which is stored in le, result is cpu,
615// there is no way to get version of object from key, so, provide
616// version to these defines
617//
bd4c625c
LT
618static inline loff_t le_key_k_offset(int version,
619 const struct reiserfs_key *key)
1da177e4 620{
bd4c625c
LT
621 return (version == KEY_FORMAT_3_5) ?
622 le32_to_cpu(key->u.k_offset_v1.k_offset) :
623 offset_v2_k_offset(&(key->u.k_offset_v2));
1da177e4
LT
624}
625
bd4c625c 626static inline loff_t le_ih_k_offset(const struct item_head *ih)
1da177e4 627{
bd4c625c 628 return le_key_k_offset(ih_version(ih), &(ih->ih_key));
1da177e4
LT
629}
630
bd4c625c 631static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)
1da177e4 632{
bd4c625c
LT
633 return (version == KEY_FORMAT_3_5) ?
634 uniqueness2type(le32_to_cpu(key->u.k_offset_v1.k_uniqueness)) :
635 offset_v2_k_type(&(key->u.k_offset_v2));
1da177e4
LT
636}
637
bd4c625c 638static inline loff_t le_ih_k_type(const struct item_head *ih)
1da177e4 639{
bd4c625c 640 return le_key_k_type(ih_version(ih), &(ih->ih_key));
1da177e4
LT
641}
642
bd4c625c
LT
643static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,
644 loff_t offset)
1da177e4 645{
bd4c625c
LT
646 (version == KEY_FORMAT_3_5) ? (void)(key->u.k_offset_v1.k_offset = cpu_to_le32(offset)) : /* jdm check */
647 (void)(set_offset_v2_k_offset(&(key->u.k_offset_v2), offset));
1da177e4
LT
648}
649
bd4c625c 650static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)
1da177e4 651{
bd4c625c 652 set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);
1da177e4
LT
653}
654
bd4c625c
LT
655static inline void set_le_key_k_type(int version, struct reiserfs_key *key,
656 int type)
1da177e4 657{
bd4c625c
LT
658 (version == KEY_FORMAT_3_5) ?
659 (void)(key->u.k_offset_v1.k_uniqueness =
660 cpu_to_le32(type2uniqueness(type)))
661 : (void)(set_offset_v2_k_type(&(key->u.k_offset_v2), type));
1da177e4 662}
bd4c625c 663static inline void set_le_ih_k_type(struct item_head *ih, int type)
1da177e4 664{
bd4c625c 665 set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);
1da177e4
LT
666}
667
1da177e4
LT
668#define is_direntry_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRENTRY)
669#define is_direct_le_key(version,key) (le_key_k_type (version, key) == TYPE_DIRECT)
670#define is_indirect_le_key(version,key) (le_key_k_type (version, key) == TYPE_INDIRECT)
671#define is_statdata_le_key(version,key) (le_key_k_type (version, key) == TYPE_STAT_DATA)
672
673//
674// item header has version.
675//
676#define is_direntry_le_ih(ih) is_direntry_le_key (ih_version (ih), &((ih)->ih_key))
677#define is_direct_le_ih(ih) is_direct_le_key (ih_version (ih), &((ih)->ih_key))
678#define is_indirect_le_ih(ih) is_indirect_le_key (ih_version(ih), &((ih)->ih_key))
679#define is_statdata_le_ih(ih) is_statdata_le_key (ih_version (ih), &((ih)->ih_key))
680
1da177e4
LT
681//
682// key is pointer to cpu key, result is cpu
683//
bd4c625c 684static inline loff_t cpu_key_k_offset(const struct cpu_key *key)
1da177e4 685{
bd4c625c 686 return key->on_disk_key.k_offset;
1da177e4
LT
687}
688
bd4c625c 689static inline loff_t cpu_key_k_type(const struct cpu_key *key)
1da177e4 690{
bd4c625c 691 return key->on_disk_key.k_type;
1da177e4
LT
692}
693
bd4c625c 694static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)
1da177e4 695{
6b9f5829 696 key->on_disk_key.k_offset = offset;
1da177e4
LT
697}
698
bd4c625c 699static inline void set_cpu_key_k_type(struct cpu_key *key, int type)
1da177e4 700{
6b9f5829 701 key->on_disk_key.k_type = type;
1da177e4
LT
702}
703
bd4c625c 704static inline void cpu_key_k_offset_dec(struct cpu_key *key)
1da177e4 705{
bd4c625c 706 key->on_disk_key.k_offset--;
1da177e4
LT
707}
708
1da177e4
LT
709#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)
710#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)
711#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)
712#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)
713
1da177e4
LT
714/* are these used ? */
715#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))
716#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))
717#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))
718#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))
719
d68caa95
JM
720#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \
721 (!COMP_SHORT_KEYS(ih, key) && \
722 I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))
1da177e4 723
bd4c625c 724/* maximal length of item */
1da177e4
LT
725#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)
726#define MIN_ITEM_LEN 1
727
1da177e4
LT
728/* object identifier for root dir */
729#define REISERFS_ROOT_OBJECTID 2
730#define REISERFS_ROOT_PARENT_OBJECTID 1
750e1c18 731
1da177e4
LT
732extern struct reiserfs_key root_key;
733
1da177e4
LT
734/*
735 * Picture represents a leaf of the S+tree
736 * ______________________________________________________
737 * | | Array of | | |
738 * |Block | Object-Item | F r e e | Objects- |
739 * | head | Headers | S p a c e | Items |
740 * |______|_______________|___________________|___________|
741 */
742
743/* Header of a disk block. More precisely, header of a formatted leaf
744 or internal node, and not the header of an unformatted node. */
bd4c625c
LT
745struct block_head {
746 __le16 blk_level; /* Level of a block in the tree. */
747 __le16 blk_nr_item; /* Number of keys/items in a block. */
748 __le16 blk_free_space; /* Block free space in bytes. */
749 __le16 blk_reserved;
750 /* dump this in v4/planA */
751 struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */
1da177e4
LT
752};
753
754#define BLKH_SIZE (sizeof(struct block_head))
755#define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level))
756#define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item))
757#define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space))
758#define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved))
759#define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val))
760#define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val))
761#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))
762#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))
763#define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key)
764#define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val)
765
766/*
767 * values for blk_level field of the struct block_head
768 */
769
bd4c625c
LT
770#define FREE_LEVEL 0 /* when node gets removed from the tree its
771 blk_level is set to FREE_LEVEL. It is then
772 used to see whether the node is still in the
773 tree */
1da177e4 774
bd4c625c 775#define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */
1da177e4
LT
776
777/* Given the buffer head of a formatted node, resolve to the block head of that node. */
ad31a4fc 778#define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data))
1da177e4 779/* Number of items that are in buffer. */
ad31a4fc
JM
780#define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh)))
781#define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh)))
782#define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh)))
1da177e4 783
ad31a4fc
JM
784#define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)
785#define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)
786#define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)
1da177e4 787
1da177e4 788/* Get right delimiting key. -- little endian */
ad31a4fc 789#define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh))))
1da177e4
LT
790
791/* Does the buffer contain a disk leaf. */
ad31a4fc 792#define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)
1da177e4
LT
793
794/* Does the buffer contain a disk internal node */
ad31a4fc
JM
795#define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \
796 && B_LEVEL(bh) <= MAX_HEIGHT)
1da177e4 797
1da177e4
LT
798/***************************************************************************/
799/* STAT DATA */
800/***************************************************************************/
801
1da177e4
LT
802//
803// old stat data is 32 bytes long. We are going to distinguish new one by
804// different size
805//
bd4c625c
LT
806struct stat_data_v1 {
807 __le16 sd_mode; /* file type, permissions */
808 __le16 sd_nlink; /* number of hard links */
809 __le16 sd_uid; /* owner */
810 __le16 sd_gid; /* group */
811 __le32 sd_size; /* file size */
812 __le32 sd_atime; /* time of last access */
813 __le32 sd_mtime; /* time file was last modified */
814 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
815 union {
816 __le32 sd_rdev;
817 __le32 sd_blocks; /* number of blocks file uses */
818 } __attribute__ ((__packed__)) u;
819 __le32 sd_first_direct_byte; /* first byte of file which is stored
820 in a direct item: except that if it
821 equals 1 it is a symlink and if it
822 equals ~(__u32)0 there is no
823 direct item. The existence of this
824 field really grates on me. Let's
825 replace it with a macro based on
826 sd_size and our tail suppression
827 policy. Someday. -Hans */
1da177e4
LT
828} __attribute__ ((__packed__));
829
830#define SD_V1_SIZE (sizeof(struct stat_data_v1))
831#define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5)
832#define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
833#define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
834#define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink))
835#define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v))
836#define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid))
837#define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v))
838#define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid))
839#define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v))
840#define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size))
841#define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v))
842#define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
843#define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
844#define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
845#define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
846#define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
847#define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
848#define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
849#define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
850#define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks))
851#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))
852#define sd_v1_first_direct_byte(sdp) \
853 (le32_to_cpu((sdp)->sd_first_direct_byte))
854#define set_sd_v1_first_direct_byte(sdp,v) \
855 ((sdp)->sd_first_direct_byte = cpu_to_le32(v))
856
1da177e4
LT
857/* inode flags stored in sd_attrs (nee sd_reserved) */
858
859/* we want common flags to have the same values as in ext2,
860 so chattr(1) will work without problems */
36695673
DH
861#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL
862#define REISERFS_APPEND_FL FS_APPEND_FL
863#define REISERFS_SYNC_FL FS_SYNC_FL
864#define REISERFS_NOATIME_FL FS_NOATIME_FL
865#define REISERFS_NODUMP_FL FS_NODUMP_FL
866#define REISERFS_SECRM_FL FS_SECRM_FL
867#define REISERFS_UNRM_FL FS_UNRM_FL
868#define REISERFS_COMPR_FL FS_COMPR_FL
869#define REISERFS_NOTAIL_FL FS_NOTAIL_FL
1da177e4
LT
870
871/* persistent flags that file inherits from the parent directory */
872#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \
873 REISERFS_SYNC_FL | \
874 REISERFS_NOATIME_FL | \
875 REISERFS_NODUMP_FL | \
876 REISERFS_SECRM_FL | \
877 REISERFS_COMPR_FL | \
878 REISERFS_NOTAIL_FL )
879
880/* Stat Data on disk (reiserfs version of UFS disk inode minus the
881 address blocks) */
882struct stat_data {
bd4c625c
LT
883 __le16 sd_mode; /* file type, permissions */
884 __le16 sd_attrs; /* persistent inode flags */
885 __le32 sd_nlink; /* number of hard links */
886 __le64 sd_size; /* file size */
887 __le32 sd_uid; /* owner */
888 __le32 sd_gid; /* group */
889 __le32 sd_atime; /* time of last access */
890 __le32 sd_mtime; /* time file was last modified */
891 __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */
892 __le32 sd_blocks;
893 union {
894 __le32 sd_rdev;
895 __le32 sd_generation;
896 //__le32 sd_first_direct_byte;
897 /* first byte of file which is stored in a
898 direct item: except that if it equals 1
899 it is a symlink and if it equals
900 ~(__u32)0 there is no direct item. The
901 existence of this field really grates
902 on me. Let's replace it with a macro
903 based on sd_size and our tail
904 suppression policy? */
905 } __attribute__ ((__packed__)) u;
1da177e4
LT
906} __attribute__ ((__packed__));
907//
908// this is 44 bytes long
909//
910#define SD_SIZE (sizeof(struct stat_data))
911#define SD_V2_SIZE SD_SIZE
912#define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6)
913#define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode))
914#define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v))
915/* sd_reserved */
916/* set_sd_reserved */
917#define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink))
918#define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v))
919#define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size))
920#define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v))
921#define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid))
922#define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v))
923#define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid))
924#define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v))
925#define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime))
926#define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v))
927#define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime))
928#define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v))
929#define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime))
930#define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v))
931#define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks))
932#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))
933#define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev))
934#define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v))
935#define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation))
936#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))
937#define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs))
938#define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v))
939
1da177e4
LT
940/***************************************************************************/
941/* DIRECTORY STRUCTURE */
942/***************************************************************************/
943/*
944 Picture represents the structure of directory items
945 ________________________________________________
946 | Array of | | | | | |
947 | directory |N-1| N-2 | .... | 1st |0th|
948 | entry headers | | | | | |
949 |_______________|___|_____|________|_______|___|
950 <---- directory entries ------>
951
952 First directory item has k_offset component 1. We store "." and ".."
953 in one item, always, we never split "." and ".." into differing
954 items. This makes, among other things, the code for removing
955 directories simpler. */
956#define SD_OFFSET 0
957#define SD_UNIQUENESS 0
958#define DOT_OFFSET 1
959#define DOT_DOT_OFFSET 2
960#define DIRENTRY_UNIQUENESS 500
961
962/* */
963#define FIRST_ITEM_OFFSET 1
964
965/*
966 Q: How to get key of object pointed to by entry from entry?
967
968 A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key
969 of object, entry points to */
970
971/* NOT IMPLEMENTED:
972 Directory will someday contain stat data of object */
973
bd4c625c
LT
974struct reiserfs_de_head {
975 __le32 deh_offset; /* third component of the directory entry key */
976 __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced
977 by directory entry */
978 __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */
979 __le16 deh_location; /* offset of name in the whole item */
980 __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether
981 entry is hidden (unlinked) */
1da177e4
LT
982} __attribute__ ((__packed__));
983#define DEH_SIZE sizeof(struct reiserfs_de_head)
984#define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset))
985#define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id))
986#define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid))
987#define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location))
988#define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state))
989
990#define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v)))
991#define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v)))
992#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))
993#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))
994#define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v)))
995
996/* empty directory contains two entries "." and ".." and their headers */
997#define EMPTY_DIR_SIZE \
998(DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen ("..")))
999
1000/* old format directories have this size when empty */
1001#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)
1002
bd4c625c 1003#define DEH_Statdata 0 /* not used now */
1da177e4
LT
1004#define DEH_Visible 2
1005
1006/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */
1007#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)
1008# define ADDR_UNALIGNED_BITS (3)
1009#endif
1010
1011/* These are only used to manipulate deh_state.
1012 * Because of this, we'll use the ext2_ bit routines,
1013 * since they are little endian */
1014#ifdef ADDR_UNALIGNED_BITS
1015
1016# define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))
1017# define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)
1018
1019# define set_bit_unaligned(nr, addr) ext2_set_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1020# define clear_bit_unaligned(nr, addr) ext2_clear_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1021# define test_bit_unaligned(nr, addr) ext2_test_bit((nr) + unaligned_offset(addr), aligned_address(addr))
1022
1023#else
1024
1025# define set_bit_unaligned(nr, addr) ext2_set_bit(nr, addr)
1026# define clear_bit_unaligned(nr, addr) ext2_clear_bit(nr, addr)
1027# define test_bit_unaligned(nr, addr) ext2_test_bit(nr, addr)
1028
1029#endif
1030
1031#define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1032#define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1033#define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1034#define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1035
1036#define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))
1037#define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1038#define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))
1039
bd4c625c
LT
1040extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,
1041 __le32 par_dirid, __le32 par_objid);
1042extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,
1043 __le32 par_dirid, __le32 par_objid);
1da177e4
LT
1044
1045/* array of the entry headers */
1046 /* get item body */
1047#define B_I_PITEM(bh,ih) ( (bh)->b_data + ih_location(ih) )
1048#define B_I_DEH(bh,ih) ((struct reiserfs_de_head *)(B_I_PITEM(bh,ih)))
1049
1050/* length of the directory entry in directory item. This define
1051 calculates length of i-th directory entry using directory entry
1052 locations from dir entry head. When it calculates length of 0-th
1053 directory entry, it uses length of whole item in place of entry
1054 location of the non-existent following entry in the calculation.
1055 See picture above.*/
1056/*
1057#define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \
1058((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh))))
1059*/
bd4c625c
LT
1060static inline int entry_length(const struct buffer_head *bh,
1061 const struct item_head *ih, int pos_in_item)
1da177e4 1062{
bd4c625c 1063 struct reiserfs_de_head *deh;
1da177e4 1064
bd4c625c
LT
1065 deh = B_I_DEH(bh, ih) + pos_in_item;
1066 if (pos_in_item)
1067 return deh_location(deh - 1) - deh_location(deh);
1da177e4 1068
bd4c625c 1069 return ih_item_len(ih) - deh_location(deh);
1da177e4
LT
1070}
1071
1da177e4
LT
1072/* number of entries in the directory item, depends on ENTRY_COUNT being at the start of directory dynamic data. */
1073#define I_ENTRY_COUNT(ih) (ih_entry_count((ih)))
1074
1da177e4
LT
1075/* name by bh, ih and entry_num */
1076#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))))
1077
1078// two entries per block (at least)
1079#define REISERFS_MAX_NAME(block_size) 255
1080
1da177e4
LT
1081/* this structure is used for operations on directory entries. It is
1082 not a disk structure. */
1083/* When reiserfs_find_entry or search_by_entry_key find directory
1084 entry, they return filled reiserfs_dir_entry structure */
bd4c625c
LT
1085struct reiserfs_dir_entry {
1086 struct buffer_head *de_bh;
1087 int de_item_num;
1088 struct item_head *de_ih;
1089 int de_entry_num;
1090 struct reiserfs_de_head *de_deh;
1091 int de_entrylen;
1092 int de_namelen;
1093 char *de_name;
3af1efe8 1094 unsigned long *de_gen_number_bit_string;
bd4c625c
LT
1095
1096 __u32 de_dir_id;
1097 __u32 de_objectid;
1098
1099 struct cpu_key de_entry_key;
1da177e4 1100};
bd4c625c 1101
1da177e4
LT
1102/* these defines are useful when a particular member of a reiserfs_dir_entry is needed */
1103
1104/* pointer to file name, stored in entry */
1105#define B_I_DEH_ENTRY_FILE_NAME(bh,ih,deh) (B_I_PITEM (bh, ih) + deh_location(deh))
1106
1107/* length of name */
1108#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \
1109(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))
1110
1da177e4
LT
1111/* hash value occupies bits from 7 up to 30 */
1112#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)
1113/* generation number occupies 7 bits starting from 0 up to 6 */
1114#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)
1115#define MAX_GENERATION_NUMBER 127
1116
1117#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))
1118
1da177e4
LT
1119/*
1120 * Picture represents an internal node of the reiserfs tree
1121 * ______________________________________________________
1122 * | | Array of | Array of | Free |
1123 * |block | keys | pointers | space |
1124 * | head | N | N+1 | |
1125 * |______|_______________|___________________|___________|
1126 */
1127
1128/***************************************************************************/
1129/* DISK CHILD */
1130/***************************************************************************/
1131/* Disk child pointer: The pointer from an internal node of the tree
1132 to a node that is on disk. */
1133struct disk_child {
bd4c625c
LT
1134 __le32 dc_block_number; /* Disk child's block number. */
1135 __le16 dc_size; /* Disk child's used space. */
1136 __le16 dc_reserved;
1da177e4
LT
1137};
1138
1139#define DC_SIZE (sizeof(struct disk_child))
1140#define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number))
1141#define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size))
1142#define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)
1143#define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)
1144
1145/* Get disk child by buffer header and position in the tree node. */
ad31a4fc
JM
1146#define B_N_CHILD(bh, n_pos) ((struct disk_child *)\
1147((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))
1da177e4
LT
1148
1149/* Get disk child number by buffer header and position in the tree node. */
ad31a4fc
JM
1150#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))
1151#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \
1152 (put_dc_block_number(B_N_CHILD(bh, n_pos), val))
1da177e4 1153
bd4c625c 1154 /* maximal value of field child_size in structure disk_child */
1da177e4
LT
1155 /* child size is the combined size of all items and their headers */
1156#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))
1157
1158/* amount of used space in buffer (not including block head) */
1159#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))
1160
1161/* max and min number of keys in internal node */
1162#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )
1163#define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2)
1164
1165/***************************************************************************/
1166/* PATH STRUCTURES AND DEFINES */
1167/***************************************************************************/
1168
1da177e4
LT
1169/* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the
1170 key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it
1171 does not find them in the cache it reads them from disk. For each node search_by_key finds using
1172 reiserfs_bread it then uses bin_search to look through that node. bin_search will find the
1173 position of the block_number of the next node if it is looking through an internal node. If it
1174 is looking through a leaf node bin_search will find the position of the item which has key either
1175 equal to given key, or which is the maximal key less than the given key. */
1176
bd4c625c
LT
1177struct path_element {
1178 struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */
1179 int pe_position; /* Position in the tree node which is placed in the */
1180 /* buffer above. */
1da177e4
LT
1181};
1182
bd4c625c
LT
1183#define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */
1184#define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */
1185#define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */
1da177e4 1186
bd4c625c
LT
1187#define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */
1188#define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */
1da177e4
LT
1189
1190/* We need to keep track of who the ancestors of nodes are. When we
1191 perform a search we record which nodes were visited while
1192 descending the tree looking for the node we searched for. This list
1193 of nodes is called the path. This information is used while
1194 performing balancing. Note that this path information may become
1195 invalid, and this means we must check it when using it to see if it
1196 is still valid. You'll need to read search_by_key and the comments
1197 in it, especially about decrement_counters_in_path(), to understand
1198 this structure.
1199
1200Paths make the code so much harder to work with and debug.... An
1201enormous number of bugs are due to them, and trying to write or modify
1202code that uses them just makes my head hurt. They are based on an
1203excessive effort to avoid disturbing the precious VFS code.:-( The
1204gods only know how we are going to SMP the code that uses them.
1205znodes are the way! */
1206
bd4c625c
LT
1207#define PATH_READA 0x1 /* do read ahead */
1208#define PATH_READA_BACK 0x2 /* read backwards */
1da177e4 1209
fec6d055 1210struct treepath {
bd4c625c
LT
1211 int path_length; /* Length of the array above. */
1212 int reada;
1213 struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */
1214 int pos_in_item;
1da177e4
LT
1215};
1216
1217#define pos_in_item(path) ((path)->pos_in_item)
1218
1219#define INITIALIZE_PATH(var) \
fec6d055 1220struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}
1da177e4
LT
1221
1222/* Get path element by path and path position. */
d68caa95 1223#define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset))
1da177e4
LT
1224
1225/* Get buffer header at the path by path and path position. */
d68caa95 1226#define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)
1da177e4
LT
1227
1228/* Get position in the element at the path by path and path position. */
d68caa95 1229#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)
1da177e4 1230
d68caa95 1231#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))
1da177e4 1232 /* you know, to the person who didn't
bd4c625c
LT
1233 write this the macro name does not
1234 at first suggest what it does.
1235 Maybe POSITION_FROM_PATH_END? Or
1236 maybe we should just focus on
1237 dumping paths... -Hans */
d68caa95 1238#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))
1da177e4 1239
d68caa95 1240#define PATH_PITEM_HEAD(path) B_N_PITEM_HEAD(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path))
1da177e4
LT
1241
1242/* in do_balance leaf has h == 0 in contrast with path structure,
1243 where root has level == 0. That is why we need these defines */
d68caa95 1244#define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h)) /* tb->S[h] */
bd4c625c
LT
1245#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */
1246#define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h))
1247#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */
1da177e4 1248
d68caa95 1249#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))
1da177e4
LT
1250
1251#define get_last_bh(path) PATH_PLAST_BUFFER(path)
1252#define get_ih(path) PATH_PITEM_HEAD(path)
1253#define get_item_pos(path) PATH_LAST_POSITION(path)
1254#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
1255#define item_moved(ih,path) comp_items(ih, path)
1256#define path_changed(ih,path) comp_items (ih, path)
1257
1da177e4
LT
1258/***************************************************************************/
1259/* MISC */
1260/***************************************************************************/
1261
1262/* Size of pointer to the unformatted node. */
1263#define UNFM_P_SIZE (sizeof(unp_t))
1264#define UNFM_P_SHIFT 2
1265
1266// in in-core inode key is stored on le form
1267#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))
1268
1269#define MAX_UL_INT 0xffffffff
1270#define MAX_INT 0x7ffffff
1271#define MAX_US_INT 0xffff
1272
1273// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset
1274#define U32_MAX (~(__u32)0)
1275
bd4c625c 1276static inline loff_t max_reiserfs_offset(struct inode *inode)
1da177e4 1277{
bd4c625c
LT
1278 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)
1279 return (loff_t) U32_MAX;
1da177e4 1280
bd4c625c 1281 return (loff_t) ((~(__u64) 0) >> 4);
1da177e4
LT
1282}
1283
1da177e4
LT
1284/*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/
1285#define MAX_KEY_OBJECTID MAX_UL_INT
1286
1da177e4
LT
1287#define MAX_B_NUM MAX_UL_INT
1288#define MAX_FC_NUM MAX_US_INT
1289
1da177e4
LT
1290/* the purpose is to detect overflow of an unsigned short */
1291#define REISERFS_LINK_MAX (MAX_US_INT - 1000)
1292
1da177e4 1293/* The following defines are used in reiserfs_insert_item and reiserfs_append_item */
bd4c625c
LT
1294#define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */
1295#define REISERFS_USER_MEM 1 /* reiserfs user memory mode */
1da177e4
LT
1296
1297#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)
1298#define get_generation(s) atomic_read (&fs_generation(s))
1299#define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen)
1300#define __fs_changed(gen,s) (gen != get_generation (s))
1301#define fs_changed(gen,s) ({cond_resched(); __fs_changed(gen, s);})
1302
1da177e4
LT
1303/***************************************************************************/
1304/* FIXATE NODES */
1305/***************************************************************************/
1306
1307#define VI_TYPE_LEFT_MERGEABLE 1
1308#define VI_TYPE_RIGHT_MERGEABLE 2
1309
1310/* To make any changes in the tree we always first find node, that
1311 contains item to be changed/deleted or place to insert a new
1312 item. We call this node S. To do balancing we need to decide what
1313 we will shift to left/right neighbor, or to a new node, where new
1314 item will be etc. To make this analysis simpler we build virtual
1315 node. Virtual node is an array of items, that will replace items of
1316 node S. (For instance if we are going to delete an item, virtual
1317 node does not contain it). Virtual node keeps information about
1318 item sizes and types, mergeability of first and last items, sizes
1319 of all entries in directory item. We use this array of items when
1320 calculating what we can shift to neighbors and how many nodes we
1321 have to have if we do not any shiftings, if we shift to left/right
1322 neighbor or to both. */
bd4c625c
LT
1323struct virtual_item {
1324 int vi_index; // index in the array of item operations
1325 unsigned short vi_type; // left/right mergeability
1326 unsigned short vi_item_len; /* length of item that it will have after balancing */
1327 struct item_head *vi_ih;
1328 const char *vi_item; // body of item (old or new)
1329 const void *vi_new_data; // 0 always but paste mode
1330 void *vi_uarea; // item specific area
1da177e4
LT
1331};
1332
bd4c625c
LT
1333struct virtual_node {
1334 char *vn_free_ptr; /* this is a pointer to the free space in the buffer */
1335 unsigned short vn_nr_item; /* number of items in virtual node */
1336 short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */
1337 short vn_mode; /* mode of balancing (paste, insert, delete, cut) */
1338 short vn_affected_item_num;
1339 short vn_pos_in_item;
1340 struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */
1341 const void *vn_data;
1342 struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */
1da177e4
LT
1343};
1344
1345/* used by directory items when creating virtual nodes */
1346struct direntry_uarea {
bd4c625c
LT
1347 int flags;
1348 __u16 entry_count;
1349 __u16 entry_sizes[1];
1350} __attribute__ ((__packed__));
1da177e4
LT
1351
1352/***************************************************************************/
1353/* TREE BALANCE */
1354/***************************************************************************/
1355
1356/* This temporary structure is used in tree balance algorithms, and
1357 constructed as we go to the extent that its various parts are
1358 needed. It contains arrays of nodes that can potentially be
1359 involved in the balancing of node S, and parameters that define how
1360 each of the nodes must be balanced. Note that in these algorithms
1361 for balancing the worst case is to need to balance the current node
1362 S and the left and right neighbors and all of their parents plus
1363 create a new node. We implement S1 balancing for the leaf nodes
1364 and S0 balancing for the internal nodes (S1 and S0 are defined in
1365 our papers.)*/
1366
1367#define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */
1368
1369/* maximum number of FEB blocknrs on a single level */
1370#define MAX_AMOUNT_NEEDED 2
1371
1372/* someday somebody will prefix every field in this struct with tb_ */
bd4c625c
LT
1373struct tree_balance {
1374 int tb_mode;
1375 int need_balance_dirty;
1376 struct super_block *tb_sb;
1377 struct reiserfs_transaction_handle *transaction_handle;
fec6d055 1378 struct treepath *tb_path;
bd4c625c
LT
1379 struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */
1380 struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */
1381 struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */
1382 struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */
1383 struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */
1384 struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */
1385
1386 struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals
1387 cur_blknum. */
1388 struct buffer_head *used[MAX_FEB_SIZE];
1389 struct buffer_head *thrown[MAX_FEB_SIZE];
1390 int lnum[MAX_HEIGHT]; /* array of number of items which must be
1391 shifted to the left in order to balance the
1392 current node; for leaves includes item that
1393 will be partially shifted; for internal
1394 nodes, it is the number of child pointers
1395 rather than items. It includes the new item
1396 being created. The code sometimes subtracts
1397 one to get the number of wholly shifted
1398 items for other purposes. */
1399 int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */
1400 int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and
1401 S[h] to its item number within the node CFL[h] */
1402 int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */
1403 int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from
1404 S[h]. A negative value means removing. */
1405 int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after
1406 balancing on the level h of the tree. If 0 then S is
1407 being deleted, if 1 then S is remaining and no new nodes
1408 are being created, if 2 or 3 then 1 or 2 new nodes is
1409 being created */
1410
1411 /* fields that are used only for balancing leaves of the tree */
1412 int cur_blknum; /* number of empty blocks having been already allocated */
1413 int s0num; /* number of items that fall into left most node when S[0] splits */
1414 int s1num; /* number of items that fall into first new node when S[0] splits */
1415 int s2num; /* number of items that fall into second new node when S[0] splits */
1416 int lbytes; /* number of bytes which can flow to the left neighbor from the left */
1417 /* most liquid item that cannot be shifted from S[0] entirely */
1418 /* if -1 then nothing will be partially shifted */
1419 int rbytes; /* number of bytes which will flow to the right neighbor from the right */
1420 /* most liquid item that cannot be shifted from S[0] entirely */
1421 /* if -1 then nothing will be partially shifted */
1422 int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */
1423 /* note: if S[0] splits into 3 nodes, then items do not need to be cut */
1424 int s2bytes;
1425 struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */
1426 char *vn_buf; /* kmalloced memory. Used to create
1da177e4
LT
1427 virtual node and keep map of
1428 dirtied bitmap blocks */
bd4c625c
LT
1429 int vn_buf_size; /* size of the vn_buf */
1430 struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */
1da177e4 1431
bd4c625c
LT
1432 int fs_gen; /* saved value of `reiserfs_generation' counter
1433 see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */
1da177e4 1434#ifdef DISPLACE_NEW_PACKING_LOCALITIES
bd4c625c
LT
1435 struct in_core_key key; /* key pointer, to pass to block allocator or
1436 another low-level subsystem */
1da177e4 1437#endif
bd4c625c 1438};
1da177e4
LT
1439
1440/* These are modes of balancing */
1441
1442/* When inserting an item. */
1443#define M_INSERT 'i'
1444/* When inserting into (directories only) or appending onto an already
1445 existant item. */
1446#define M_PASTE 'p'
1447/* When deleting an item. */
1448#define M_DELETE 'd'
1449/* When truncating an item or removing an entry from a (directory) item. */
1450#define M_CUT 'c'
1451
1452/* used when balancing on leaf level skipped (in reiserfsck) */
1453#define M_INTERNAL 'n'
1454
1455/* When further balancing is not needed, then do_balance does not need
1456 to be called. */
1457#define M_SKIP_BALANCING 's'
1458#define M_CONVERT 'v'
1459
1460/* modes of leaf_move_items */
1461#define LEAF_FROM_S_TO_L 0
1462#define LEAF_FROM_S_TO_R 1
1463#define LEAF_FROM_R_TO_L 2
1464#define LEAF_FROM_L_TO_R 3
1465#define LEAF_FROM_S_TO_SNEW 4
1466
1467#define FIRST_TO_LAST 0
1468#define LAST_TO_FIRST 1
1469
1470/* used in do_balance for passing parent of node information that has
1471 been gotten from tb struct */
1472struct buffer_info {
bd4c625c
LT
1473 struct tree_balance *tb;
1474 struct buffer_head *bi_bh;
1475 struct buffer_head *bi_parent;
1476 int bi_position;
1da177e4
LT
1477};
1478
c3a9c210
JM
1479static inline struct super_block *sb_from_tb(struct tree_balance *tb)
1480{
1481 return tb ? tb->tb_sb : NULL;
1482}
1483
1484static inline struct super_block *sb_from_bi(struct buffer_info *bi)
1485{
1486 return bi ? sb_from_tb(bi->tb) : NULL;
1487}
1488
1da177e4
LT
1489/* there are 4 types of items: stat data, directory item, indirect, direct.
1490+-------------------+------------+--------------+------------+
1491| | k_offset | k_uniqueness | mergeable? |
1492+-------------------+------------+--------------+------------+
1493| stat data | 0 | 0 | no |
1494+-------------------+------------+--------------+------------+
1495| 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no |
1496| non 1st directory | hash value | | yes |
1497| item | | | |
1498+-------------------+------------+--------------+------------+
1499| indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object
1500+-------------------+------------+--------------+------------+
1501| direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object
1502+-------------------+------------+--------------+------------+
1503*/
1504
1505struct item_operations {
bd4c625c
LT
1506 int (*bytes_number) (struct item_head * ih, int block_size);
1507 void (*decrement_key) (struct cpu_key *);
1508 int (*is_left_mergeable) (struct reiserfs_key * ih,
1509 unsigned long bsize);
1510 void (*print_item) (struct item_head *, char *item);
1511 void (*check_item) (struct item_head *, char *item);
1512
1513 int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,
1514 int is_affected, int insert_size);
1515 int (*check_left) (struct virtual_item * vi, int free,
1516 int start_skip, int end_skip);
1517 int (*check_right) (struct virtual_item * vi, int free);
1518 int (*part_size) (struct virtual_item * vi, int from, int to);
1519 int (*unit_num) (struct virtual_item * vi);
1520 void (*print_vi) (struct virtual_item * vi);
1da177e4
LT
1521};
1522
bd4c625c 1523extern struct item_operations *item_ops[TYPE_ANY + 1];
1da177e4
LT
1524
1525#define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)
1526#define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)
1527#define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item)
1528#define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item)
1529#define op_create_vi(vn,vi,is_affected,insert_size) item_ops[le_ih_k_type ((vi)->vi_ih)]->create_vi (vn,vi,is_affected,insert_size)
1530#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)
1531#define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free)
1532#define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to)
1533#define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi)
1534#define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi)
1535
1da177e4
LT
1536#define COMP_SHORT_KEYS comp_short_keys
1537
1538/* number of blocks pointed to by the indirect item */
d68caa95 1539#define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE)
1da177e4
LT
1540
1541/* the used space within the unformatted node corresponding to pos within the item pointed to by ih */
1542#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))
1543
1544/* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */
1545
bd4c625c 1546/* get the item header */
1da177e4
LT
1547#define B_N_PITEM_HEAD(bh,item_num) ( (struct item_head * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1548
1549/* get key */
1550#define B_N_PDELIM_KEY(bh,item_num) ( (struct reiserfs_key * )((bh)->b_data + BLKH_SIZE) + (item_num) )
1551
1552/* get the key */
1553#define B_N_PKEY(bh,item_num) ( &(B_N_PITEM_HEAD(bh,item_num)->ih_key) )
1554
1555/* get item body */
1556#define B_N_PITEM(bh,item_num) ( (bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(item_num))))
1557
1558/* get the stat data by the buffer header and the item order */
1559#define B_N_STAT_DATA(bh,nr) \
1560( (struct stat_data *)((bh)->b_data + ih_location(B_N_PITEM_HEAD((bh),(nr))) ) )
1561
1562 /* following defines use reiserfs buffer header and item header */
1563
1564/* get stat-data */
1565#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )
1566
1567// this is 3976 for size==4096
1568#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)
1569
1570/* indirect items consist of entries which contain blocknrs, pos
1571 indicates which entry, and B_I_POS_UNFM_POINTER resolves to the
1572 blocknr contained by the entry pos points to */
1573#define B_I_POS_UNFM_POINTER(bh,ih,pos) le32_to_cpu(*(((unp_t *)B_I_PITEM(bh,ih)) + (pos)))
1574#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)
1575
1576struct reiserfs_iget_args {
bd4c625c
LT
1577 __u32 objectid;
1578 __u32 dirid;
1579};
1da177e4
LT
1580
1581/***************************************************************************/
1582/* FUNCTION DECLARATIONS */
1583/***************************************************************************/
1584
1da177e4
LT
1585#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)
1586
1587#define journal_trans_half(blocksize) \
1588 ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32))
1589
1590/* journal.c see journal.c for all the comments here */
1591
1592/* first block written in a commit. */
1593struct reiserfs_journal_desc {
bd4c625c
LT
1594 __le32 j_trans_id; /* id of commit */
1595 __le32 j_len; /* length of commit. len +1 is the commit block */
1596 __le32 j_mount_id; /* mount id of this trans */
1597 __le32 j_realblock[1]; /* real locations for each block */
1598};
1da177e4
LT
1599
1600#define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id)
1601#define get_desc_trans_len(d) le32_to_cpu((d)->j_len)
1602#define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id)
1603
1604#define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)
1605#define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0)
1606#define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)
1607
1608/* last block written in a commit */
1609struct reiserfs_journal_commit {
bd4c625c
LT
1610 __le32 j_trans_id; /* must match j_trans_id from the desc block */
1611 __le32 j_len; /* ditto */
1612 __le32 j_realblock[1]; /* real locations for each block */
1613};
1da177e4
LT
1614
1615#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)
1616#define get_commit_trans_len(c) le32_to_cpu((c)->j_len)
1617#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)
1618
1619#define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)
1620#define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0)
1621
1622/* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the
1623** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk,
1624** and this transaction does not need to be replayed.
1625*/
1626struct reiserfs_journal_header {
bd4c625c
LT
1627 __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */
1628 __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */
1629 __le32 j_mount_id;
1630 /* 12 */ struct journal_params jh_journal;
1631};
1da177e4
LT
1632
1633/* biggest tunable defines are right here */
bd4c625c
LT
1634#define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */
1635#define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */
1da177e4 1636#define JOURNAL_TRANS_MIN_DEFAULT 256
bd4c625c 1637#define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */
1da177e4 1638#define JOURNAL_MIN_RATIO 2
bd4c625c 1639#define JOURNAL_MAX_COMMIT_AGE 30
1da177e4
LT
1640#define JOURNAL_MAX_TRANS_AGE 30
1641#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)
0ab2621e
JM
1642#define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \
1643 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \
1644 REISERFS_QUOTA_TRANS_BLOCKS(sb)))
1645
1da177e4 1646#ifdef CONFIG_QUOTA
556a2a45
JK
1647/* We need to update data and inode (atime) */
1648#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? 2 : 0)
1649/* 1 balancing, 1 bitmap, 1 data per write + stat data update */
1650#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1651(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)
1652/* same as with INIT */
1653#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & (1<<REISERFS_QUOTA) ? \
1654(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)
1da177e4 1655#else
556a2a45
JK
1656#define REISERFS_QUOTA_TRANS_BLOCKS(s) 0
1657#define REISERFS_QUOTA_INIT_BLOCKS(s) 0
1658#define REISERFS_QUOTA_DEL_BLOCKS(s) 0
1da177e4
LT
1659#endif
1660
1661/* both of these can be as low as 1, or as high as you want. The min is the
1662** number of 4k bitmap nodes preallocated on mount. New nodes are allocated
1663** as needed, and released when transactions are committed. On release, if
1664** the current number of nodes is > max, the node is freed, otherwise,
1665** it is put on a free list for faster use later.
1666*/
bd4c625c
LT
1667#define REISERFS_MIN_BITMAP_NODES 10
1668#define REISERFS_MAX_BITMAP_NODES 100
1da177e4 1669
bd4c625c 1670#define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */
1da177e4
LT
1671#define JBH_HASH_MASK 8191
1672
1673#define _jhashfn(sb,block) \
1674 (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \
1675 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))
1676#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])
1677
1678// We need these to make journal.c code more readable
1679#define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1680#define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1681#define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize)
1682
1683enum reiserfs_bh_state_bits {
bd4c625c
LT
1684 BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */
1685 BH_JDirty_wait,
1686 BH_JNew, /* disk block was taken off free list before
1687 * being in a finished transaction, or
1688 * written to disk. Can be reused immed. */
1689 BH_JPrepared,
1690 BH_JRestore_dirty,
1691 BH_JTest, // debugging only will go away
1da177e4
LT
1692};
1693
1694BUFFER_FNS(JDirty, journaled);
1695TAS_BUFFER_FNS(JDirty, journaled);
1696BUFFER_FNS(JDirty_wait, journal_dirty);
1697TAS_BUFFER_FNS(JDirty_wait, journal_dirty);
1698BUFFER_FNS(JNew, journal_new);
1699TAS_BUFFER_FNS(JNew, journal_new);
1700BUFFER_FNS(JPrepared, journal_prepared);
1701TAS_BUFFER_FNS(JPrepared, journal_prepared);
1702BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1703TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);
1704BUFFER_FNS(JTest, journal_test);
1705TAS_BUFFER_FNS(JTest, journal_test);
1706
1707/*
1708** transaction handle which is passed around for all journal calls
1709*/
1710struct reiserfs_transaction_handle {
bd4c625c
LT
1711 struct super_block *t_super; /* super for this FS when journal_begin was
1712 called. saves calls to reiserfs_get_super
1713 also used by nested transactions to make
1714 sure they are nesting on the right FS
1715 _must_ be first in the handle
1716 */
1717 int t_refcount;
1718 int t_blocks_logged; /* number of blocks this writer has logged */
1719 int t_blocks_allocated; /* number of blocks this writer allocated */
600ed416 1720 unsigned int t_trans_id; /* sanity check, equals the current trans id */
bd4c625c
LT
1721 void *t_handle_save; /* save existing current->journal_info */
1722 unsigned displace_new_blocks:1; /* if new block allocation occurres, that block
1723 should be displaced from others */
1724 struct list_head t_list;
1725};
1da177e4
LT
1726
1727/* used to keep track of ordered and tail writes, attached to the buffer
1728 * head through b_journal_head.
1729 */
1730struct reiserfs_jh {
bd4c625c
LT
1731 struct reiserfs_journal_list *jl;
1732 struct buffer_head *bh;
1733 struct list_head list;
1da177e4
LT
1734};
1735
1736void reiserfs_free_jh(struct buffer_head *bh);
1737int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);
1738int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);
bd4c625c
LT
1739int journal_mark_dirty(struct reiserfs_transaction_handle *,
1740 struct super_block *, struct buffer_head *bh);
1741
1742static inline int reiserfs_file_data_log(struct inode *inode)
1743{
1744 if (reiserfs_data_log(inode->i_sb) ||
1745 (REISERFS_I(inode)->i_flags & i_data_log))
1746 return 1;
1747 return 0;
1da177e4
LT
1748}
1749
bd4c625c
LT
1750static inline int reiserfs_transaction_running(struct super_block *s)
1751{
1752 struct reiserfs_transaction_handle *th = current->journal_info;
1753 if (th && th->t_super == s)
1754 return 1;
1755 if (th && th->t_super == NULL)
1756 BUG();
1757 return 0;
1da177e4
LT
1758}
1759
23f9e0f8
AZ
1760static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)
1761{
1762 return th->t_blocks_allocated - th->t_blocks_logged;
1763}
1764
bd4c625c
LT
1765struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
1766 super_block
1767 *,
1768 int count);
1da177e4
LT
1769int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);
1770int reiserfs_commit_page(struct inode *inode, struct page *page,
bd4c625c 1771 unsigned from, unsigned to);
1da177e4 1772int reiserfs_flush_old_commits(struct super_block *);
bd4c625c
LT
1773int reiserfs_commit_for_inode(struct inode *);
1774int reiserfs_inode_needs_commit(struct inode *);
1775void reiserfs_update_inode_transaction(struct inode *);
1776void reiserfs_wait_on_write_block(struct super_block *s);
1777void reiserfs_block_writes(struct reiserfs_transaction_handle *th);
1778void reiserfs_allow_writes(struct super_block *s);
1779void reiserfs_check_lock_depth(struct super_block *s, char *caller);
1780int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,
1781 int wait);
1782void reiserfs_restore_prepared_buffer(struct super_block *,
1783 struct buffer_head *bh);
1784int journal_init(struct super_block *, const char *j_dev_name, int old_format,
1785 unsigned int);
1786int journal_release(struct reiserfs_transaction_handle *, struct super_block *);
1787int journal_release_error(struct reiserfs_transaction_handle *,
1788 struct super_block *);
1789int journal_end(struct reiserfs_transaction_handle *, struct super_block *,
1790 unsigned long);
1791int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *,
1792 unsigned long);
1793int journal_mark_freed(struct reiserfs_transaction_handle *,
1794 struct super_block *, b_blocknr_t blocknr);
1795int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);
a9dd3643
JM
1796int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,
1797 int bit_nr, int searchall, b_blocknr_t *next);
bd4c625c 1798int journal_begin(struct reiserfs_transaction_handle *,
a9dd3643 1799 struct super_block *sb, unsigned long);
bd4c625c 1800int journal_join_abort(struct reiserfs_transaction_handle *,
a9dd3643 1801 struct super_block *sb, unsigned long);
32e8b106 1802void reiserfs_abort_journal(struct super_block *sb, int errno);
bd4c625c
LT
1803void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);
1804int reiserfs_allocate_list_bitmaps(struct super_block *s,
3ee16670 1805 struct reiserfs_list_bitmap *, unsigned int);
bd4c625c
LT
1806
1807void add_save_link(struct reiserfs_transaction_handle *th,
1808 struct inode *inode, int truncate);
1809int remove_save_link(struct inode *inode, int truncate);
1da177e4
LT
1810
1811/* objectid.c */
bd4c625c
LT
1812__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);
1813void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,
1814 __u32 objectid_to_release);
1815int reiserfs_convert_objectid_map_v1(struct super_block *);
1da177e4
LT
1816
1817/* stree.c */
1818int B_IS_IN_TREE(const struct buffer_head *);
d68caa95
JM
1819extern void copy_item_head(struct item_head *to,
1820 const struct item_head *from);
1da177e4
LT
1821
1822// first key is in cpu form, second - le
bd4c625c
LT
1823extern int comp_short_keys(const struct reiserfs_key *le_key,
1824 const struct cpu_key *cpu_key);
1825extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);
1da177e4
LT
1826
1827// both are in le form
bd4c625c
LT
1828extern int comp_le_keys(const struct reiserfs_key *,
1829 const struct reiserfs_key *);
1830extern int comp_short_le_keys(const struct reiserfs_key *,
1831 const struct reiserfs_key *);
1da177e4
LT
1832
1833//
1834// get key version from on disk key - kludge
1835//
bd4c625c 1836static inline int le_key_version(const struct reiserfs_key *key)
1da177e4 1837{
bd4c625c 1838 int type;
1da177e4 1839
bd4c625c
LT
1840 type = offset_v2_k_type(&(key->u.k_offset_v2));
1841 if (type != TYPE_DIRECT && type != TYPE_INDIRECT
1842 && type != TYPE_DIRENTRY)
1843 return KEY_FORMAT_3_5;
1844
1845 return KEY_FORMAT_3_6;
1da177e4 1846
1da177e4
LT
1847}
1848
bd4c625c
LT
1849static inline void copy_key(struct reiserfs_key *to,
1850 const struct reiserfs_key *from)
1851{
1852 memcpy(to, from, KEY_SIZE);
1853}
1da177e4 1854
d68caa95
JM
1855int comp_items(const struct item_head *stored_ih, const struct treepath *path);
1856const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
a9dd3643 1857 const struct super_block *sb);
bd4c625c 1858int search_by_key(struct super_block *, const struct cpu_key *,
fec6d055 1859 struct treepath *, int);
1da177e4 1860#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)
a9dd3643 1861int search_for_position_by_key(struct super_block *sb,
d68caa95
JM
1862 const struct cpu_key *cpu_key,
1863 struct treepath *search_path);
ad31a4fc 1864extern void decrement_bcount(struct buffer_head *bh);
d68caa95
JM
1865void decrement_counters_in_path(struct treepath *search_path);
1866void pathrelse(struct treepath *search_path);
fec6d055 1867int reiserfs_check_path(struct treepath *p);
d68caa95 1868void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);
bd4c625c
LT
1869
1870int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
fec6d055 1871 struct treepath *path,
bd4c625c
LT
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,
fec6d055 1877 struct treepath *path,
bd4c625c
LT
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,
fec6d055 1883 struct treepath *path,
bd4c625c
LT
1884 struct cpu_key *key,
1885 struct inode *inode,
1886 struct page *page, loff_t new_file_size);
1887
1888int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
fec6d055 1889 struct treepath *path,
bd4c625c 1890 const struct cpu_key *key,
d68caa95 1891 struct inode *inode, struct buffer_head *un_bh);
bd4c625c
LT
1892
1893void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1894 struct inode *inode, struct reiserfs_key *key);
1895int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
995c762e 1896 struct inode *inode);
bd4c625c 1897int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
995c762e 1898 struct inode *inode, struct page *,
bd4c625c 1899 int update_timestamps);
1da177e4
LT
1900
1901#define i_block_size(inode) ((inode)->i_sb->s_blocksize)
1902#define file_size(inode) ((inode)->i_size)
1903#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))
1904
1905#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\
1906!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 )
1907
bd4c625c 1908void padd_item(char *item, int total_length, int length);
1da177e4
LT
1909
1910/* inode.c */
1911/* args for the create parameter of reiserfs_get_block */
bd4c625c
LT
1912#define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
1913#define GET_BLOCK_CREATE 1 /* add anything you need to find block */
1914#define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
1915#define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
1b1dcc1b 1916#define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */
bd4c625c
LT
1917#define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
1918
bd4c625c
LT
1919void reiserfs_read_locked_inode(struct inode *inode,
1920 struct reiserfs_iget_args *args);
1921int reiserfs_find_actor(struct inode *inode, void *p);
1922int reiserfs_init_locked_inode(struct inode *inode, void *p);
1923void reiserfs_delete_inode(struct inode *inode);
1924int reiserfs_write_inode(struct inode *inode, int);
1925int reiserfs_get_block(struct inode *inode, sector_t block,
1926 struct buffer_head *bh_result, int create);
be55caf1
CH
1927struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1928 int fh_len, int fh_type);
1929struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1930 int fh_len, int fh_type);
bd4c625c
LT
1931int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1932 int connectable);
1933
1934int reiserfs_truncate_file(struct inode *, int update_timestamps);
1935void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,
1936 int type, int key_length);
1937void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
1938 int version,
1939 loff_t offset, int type, int length, int entry_count);
1940struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);
1941
57fe60df 1942struct reiserfs_security_handle;
bd4c625c
LT
1943int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1944 struct inode *dir, int mode,
1945 const char *symname, loff_t i_size,
57fe60df
JM
1946 struct dentry *dentry, struct inode *inode,
1947 struct reiserfs_security_handle *security);
bd4c625c
LT
1948
1949void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1950 struct inode *inode, loff_t size);
1da177e4
LT
1951
1952static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,
bd4c625c 1953 struct inode *inode)
1da177e4 1954{
bd4c625c 1955 reiserfs_update_sd_size(th, inode, inode->i_size);
1da177e4
LT
1956}
1957
bd4c625c
LT
1958void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);
1959void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs);
1da177e4
LT
1960int reiserfs_setattr(struct dentry *dentry, struct iattr *attr);
1961
1962/* namei.c */
bd4c625c
LT
1963void set_de_name_and_namelen(struct reiserfs_dir_entry *de);
1964int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,
fec6d055 1965 struct treepath *path, struct reiserfs_dir_entry *de);
bd4c625c 1966struct dentry *reiserfs_get_parent(struct dentry *);
1da177e4
LT
1967/* procfs.c */
1968
1969#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )
1970#define REISERFS_PROC_INFO
1971#else
1972#undef REISERFS_PROC_INFO
1973#endif
1974
bd4c625c
LT
1975int reiserfs_proc_info_init(struct super_block *sb);
1976int reiserfs_proc_info_done(struct super_block *sb);
1977struct proc_dir_entry *reiserfs_proc_register_global(char *name,
1978 read_proc_t * func);
1979void reiserfs_proc_unregister_global(const char *name);
1980int reiserfs_proc_info_global_init(void);
1981int reiserfs_proc_info_global_done(void);
1982int reiserfs_global_version_in_proc(char *buffer, char **start, off_t offset,
1983 int count, int *eof, void *data);
1da177e4
LT
1984
1985#if defined( REISERFS_PROC_INFO )
1986
1987#define PROC_EXP( e ) e
1988
1989#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data
1990#define PROC_INFO_MAX( sb, field, value ) \
1991 __PINFO( sb ).field = \
1992 max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )
1993#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )
1994#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )
1995#define PROC_INFO_BH_STAT( sb, bh, level ) \
1996 PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \
1997 PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \
1998 PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )
1999#else
2000#define PROC_EXP( e )
2001#define VOID_V ( ( void ) 0 )
2002#define PROC_INFO_MAX( sb, field, value ) VOID_V
2003#define PROC_INFO_INC( sb, field ) VOID_V
2004#define PROC_INFO_ADD( sb, field, val ) VOID_V
ad31a4fc 2005#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V
1da177e4
LT
2006#endif
2007
2008/* dir.c */
c5ef1c42
AV
2009extern const struct inode_operations reiserfs_dir_inode_operations;
2010extern const struct inode_operations reiserfs_symlink_inode_operations;
2011extern const struct inode_operations reiserfs_special_inode_operations;
4b6f5d20 2012extern const struct file_operations reiserfs_dir_operations;
a41f1a47 2013int reiserfs_readdir_dentry(struct dentry *, void *, filldir_t, loff_t *);
1da177e4
LT
2014
2015/* tail_conversion.c */
bd4c625c 2016int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,
fec6d055 2017 struct treepath *, struct buffer_head *, loff_t);
bd4c625c 2018int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,
fec6d055 2019 struct page *, struct treepath *, const struct cpu_key *,
bd4c625c
LT
2020 loff_t, char *);
2021void reiserfs_unmap_buffer(struct buffer_head *);
1da177e4
LT
2022
2023/* file.c */
c5ef1c42 2024extern const struct inode_operations reiserfs_file_inode_operations;
4b6f5d20 2025extern const struct file_operations reiserfs_file_operations;
f5e54d6e 2026extern const struct address_space_operations reiserfs_address_space_operations;
1da177e4
LT
2027
2028/* fix_nodes.c */
1da177e4 2029
a063ae17 2030int fix_nodes(int n_op_mode, struct tree_balance *tb,
d68caa95 2031 struct item_head *ins_ih, const void *);
bd4c625c 2032void unfix_nodes(struct tree_balance *);
1da177e4
LT
2033
2034/* prints.c */
c3a9c210
JM
2035void __reiserfs_panic(struct super_block *s, const char *id,
2036 const char *function, const char *fmt, ...)
bd4c625c 2037 __attribute__ ((noreturn));
c3a9c210
JM
2038#define reiserfs_panic(s, id, fmt, args...) \
2039 __reiserfs_panic(s, id, __func__, fmt, ##args)
1e5e59d4
JM
2040void __reiserfs_error(struct super_block *s, const char *id,
2041 const char *function, const char *fmt, ...);
2042#define reiserfs_error(s, id, fmt, args...) \
2043 __reiserfs_error(s, id, __func__, fmt, ##args)
bd4c625c
LT
2044void reiserfs_info(struct super_block *s, const char *fmt, ...);
2045void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);
2046void print_indirect_item(struct buffer_head *bh, int item_num);
2047void store_print_tb(struct tree_balance *tb);
2048void print_cur_tb(char *mes);
2049void print_de(struct reiserfs_dir_entry *de);
2050void print_bi(struct buffer_info *bi, char *mes);
2051#define PRINT_LEAF_ITEMS 1 /* print all items */
2052#define PRINT_DIRECTORY_ITEMS 2 /* print directory items */
2053#define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */
2054void print_block(struct buffer_head *bh, ...);
2055void print_bmap(struct super_block *s, int silent);
2056void print_bmap_block(int i, char *data, int size, int silent);
1da177e4 2057/*void print_super_block (struct super_block * s, char * mes);*/
bd4c625c
LT
2058void print_objectid_map(struct super_block *s);
2059void print_block_head(struct buffer_head *bh, char *mes);
2060void check_leaf(struct buffer_head *bh);
2061void check_internal(struct buffer_head *bh);
2062void print_statistics(struct super_block *s);
2063char *reiserfs_hashname(int code);
1da177e4
LT
2064
2065/* lbalance.c */
bd4c625c
LT
2066int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,
2067 int mov_bytes, struct buffer_head *Snew);
2068int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);
2069int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);
2070void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,
2071 int del_num, int del_bytes);
2072void leaf_insert_into_buf(struct buffer_info *bi, int before,
2073 struct item_head *inserted_item_ih,
2074 const char *inserted_item_body, int zeros_number);
2075void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,
2076 int pos_in_item, int paste_size, const char *body,
2077 int zeros_number);
2078void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,
2079 int pos_in_item, int cut_size);
eba00305 2080void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,
bd4c625c
LT
2081 int new_entry_count, struct reiserfs_de_head *new_dehs,
2082 const char *records, int paste_size);
1da177e4 2083/* ibalance.c */
bd4c625c
LT
2084int balance_internal(struct tree_balance *, int, int, struct item_head *,
2085 struct buffer_head **);
1da177e4
LT
2086
2087/* do_balance.c */
bd4c625c
LT
2088void do_balance_mark_leaf_dirty(struct tree_balance *tb,
2089 struct buffer_head *bh, int flag);
1da177e4
LT
2090#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty
2091#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty
2092
bd4c625c
LT
2093void do_balance(struct tree_balance *tb, struct item_head *ih,
2094 const char *body, int flag);
2095void reiserfs_invalidate_buffer(struct tree_balance *tb,
2096 struct buffer_head *bh);
1da177e4 2097
bd4c625c
LT
2098int get_left_neighbor_position(struct tree_balance *tb, int h);
2099int get_right_neighbor_position(struct tree_balance *tb, int h);
2100void replace_key(struct tree_balance *tb, struct buffer_head *, int,
2101 struct buffer_head *, int);
2102void make_empty_node(struct buffer_info *);
2103struct buffer_head *get_FEB(struct tree_balance *);
1da177e4
LT
2104
2105/* bitmap.c */
2106
2107/* structure contains hints for block allocator, and it is a container for
2108 * arguments, such as node, search path, transaction_handle, etc. */
bd4c625c
LT
2109struct __reiserfs_blocknr_hint {
2110 struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */
3ee16670 2111 sector_t block; /* file offset, in blocks */
bd4c625c 2112 struct in_core_key key;
fec6d055 2113 struct treepath *path; /* search path, used by allocator to deternine search_start by
bd4c625c
LT
2114 * various ways */
2115 struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and
2116 * bitmap blocks changes */
2117 b_blocknr_t beg, end;
2118 b_blocknr_t search_start; /* a field used to transfer search start value (block number)
1da177e4
LT
2119 * between different block allocator procedures
2120 * (determine_search_start() and others) */
bd4c625c
LT
2121 int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed
2122 * function that do actual allocation */
1da177e4 2123
bd4c625c 2124 unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for
1da177e4 2125 * formatted/unformatted blocks with/without preallocation */
bd4c625c 2126 unsigned preallocate:1;
1da177e4
LT
2127};
2128
2129typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;
2130
bd4c625c
LT
2131int reiserfs_parse_alloc_options(struct super_block *, char *);
2132void reiserfs_init_alloc_options(struct super_block *s);
1da177e4
LT
2133
2134/*
2135 * given a directory, this will tell you what packing locality
2136 * to use for a new object underneat it. The locality is returned
2137 * in disk byte order (le).
2138 */
3e8962be 2139__le32 reiserfs_choose_packing(struct inode *dir);
1da177e4 2140
6f01046b
JM
2141int reiserfs_init_bitmap_cache(struct super_block *sb);
2142void reiserfs_free_bitmap_cache(struct super_block *sb);
2143void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);
2144struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);
bd4c625c
LT
2145int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);
2146void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,
2147 b_blocknr_t, int for_unformatted);
2148int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,
2149 int);
9adeb1b4 2150static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,
bd4c625c
LT
2151 b_blocknr_t * new_blocknrs,
2152 int amount_needed)
1da177e4 2153{
bd4c625c
LT
2154 reiserfs_blocknr_hint_t hint = {
2155 .th = tb->transaction_handle,
2156 .path = tb->tb_path,
2157 .inode = NULL,
2158 .key = tb->key,
2159 .block = 0,
2160 .formatted_node = 1
2161 };
2162 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,
2163 0);
1da177e4
LT
2164}
2165
9adeb1b4 2166static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle
bd4c625c
LT
2167 *th, struct inode *inode,
2168 b_blocknr_t * new_blocknrs,
3ee16670
JM
2169 struct treepath *path,
2170 sector_t block)
1da177e4 2171{
bd4c625c
LT
2172 reiserfs_blocknr_hint_t hint = {
2173 .th = th,
2174 .path = path,
2175 .inode = inode,
2176 .block = block,
2177 .formatted_node = 0,
2178 .preallocate = 0
2179 };
2180 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2181}
2182
2183#ifdef REISERFS_PREALLOCATE
9adeb1b4 2184static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle
bd4c625c
LT
2185 *th, struct inode *inode,
2186 b_blocknr_t * new_blocknrs,
3ee16670
JM
2187 struct treepath *path,
2188 sector_t block)
1da177e4 2189{
bd4c625c
LT
2190 reiserfs_blocknr_hint_t hint = {
2191 .th = th,
2192 .path = path,
2193 .inode = inode,
2194 .block = block,
2195 .formatted_node = 0,
2196 .preallocate = 1
2197 };
2198 return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);
1da177e4
LT
2199}
2200
bd4c625c
LT
2201void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
2202 struct inode *inode);
2203void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);
1da177e4 2204#endif
1da177e4
LT
2205
2206/* hashes.c */
bd4c625c
LT
2207__u32 keyed_hash(const signed char *msg, int len);
2208__u32 yura_hash(const signed char *msg, int len);
2209__u32 r5_hash(const signed char *msg, int len);
1da177e4
LT
2210
2211/* the ext2 bit routines adjust for big or little endian as
2212** appropriate for the arch, so in our laziness we use them rather
2213** than using the bit routines they call more directly. These
2214** routines must be used when changing on disk bitmaps. */
2215#define reiserfs_test_and_set_le_bit ext2_set_bit
2216#define reiserfs_test_and_clear_le_bit ext2_clear_bit
2217#define reiserfs_test_le_bit ext2_test_bit
2218#define reiserfs_find_next_zero_le_bit ext2_find_next_zero_bit
2219
2220/* sometimes reiserfs_truncate may require to allocate few new blocks
2221 to perform indirect2direct conversion. People probably used to
2222 think, that truncate should work without problems on a filesystem
2223 without free disk space. They may complain that they can not
2224 truncate due to lack of free disk space. This spare space allows us
2225 to not worry about it. 500 is probably too much, but it should be
2226 absolutely safe */
2227#define SPARE_SPACE 500
2228
1da177e4 2229/* prototypes from ioctl.c */
bd4c625c
LT
2230int reiserfs_ioctl(struct inode *inode, struct file *filp,
2231 unsigned int cmd, unsigned long arg);
52b499c4
DH
2232long reiserfs_compat_ioctl(struct file *filp,
2233 unsigned int cmd, unsigned long arg);
d5dee5c3 2234int reiserfs_unpack(struct inode *inode, struct file *filp);
bd4c625c 2235
11d9f653 2236#endif /* __KERNEL__ */
bd4c625c 2237
bd4c625c 2238#endif /* _LINUX_REISER_FS_H */
This page took 0.923955 seconds and 5 git commands to generate.